File loading please wait...
Citation preview
The Languages and Linguistics of Indigenous North America WOL 13.1
The World of Linguistics
Editor Hans Henrich Hock
Volume 13.1
The Languages and Linguistics of Indigenous North America A Comprehensive Guide Volume 1 Edited by Carmen Dagostino, Marianne Mithun, and Keren Rice
ISBN 978-3-11-059798-1 e-ISBN (PDF) 978-3-11-060092-6 e-ISBN (EPUB) 978-3-11-059869-8 Library of Congress Control Number: 2023932794 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.d-nb.de. © 2023 Walter de Gruyter GmbH, Berlin/Boston Cover image: jmatkins / iStock / Getty Images Plus Typesetting: Dörlemann Satz GmbH & Co. KG, Lemförde Printing and binding: CPI books GmbH, Leck www.degruyter.com
Preface
V
Preface This handbook is intended to provide broad coverage of topics of interest to linguists in general, and more specifically to community and academic scholars engaged in the study and revitalization of North American Indigenous languages. Particular attention has been given to new ideas and recent trends in research, to features of the languages that are typologically unusual or unusually well-developed in comparison with others outside of the area, and topics of special importance to communities. The general chapters include broad cross-linguistic coverage of each area and surveys of current work on the topic, as well as points that may be useful for language revitalization purposes. Many address topics that received less attention in earlier accounts, such as syntax, discourse, language change, and contact effects. Given the current blossoming of community-centered research (Bischoff and Jany 2018) and the formation of ever greater numbers of Indigenous linguists, the editors would like to ensure that this work is of value to the communities involved in language maintenance and revitalization. The volume is divided into two main parts, the first on general topics, and the second on revitalization and sketches of languages and families. Volume 1 describes different levels of structure: sounds and sound structures (acoustic phonetics, articulatory phonetics, tone, segmental phonology, prosodic phonology, word prosody, prosody beyond the word), words (identifying words, word classes), sentences (syntax within the clause, syntax beyond the clause, negation, questions and requests, information structure, relative clauses, subordination and complementation, switch reference and event cohesion), discourse (verbal art, conversation structure), and meaning (lexicalization and lexical meaning, lexicography, evidentiality, pluractionality and distributivity, mass versus count nouns, space, landscape, and orientation) and pragmatics. Following that are sections on language over time and space (how grammar emerges, language classification, language contact and linguistic areas, archival-based sociolinguistic variation, and community-based sociolinguistic variation). Volume 2 contains sections devoted to topics of importance in language acquisition and revitalization (outcomes of Mentor-Apprentice programs, child and child-directed speech, language pedagogies, digital tools for revitalization, the use of archival materials for language reclamation, and changing notions of fieldwork), followed by sketches of families and isolates. The geographic area over which the languages are spoken is extensive, extending from the Arctic in the north to the US border with Mexico in the south although some include Mexico in their definition of this area (Siddiqi, Barrie, Gillon, Haugen, and Mathieu 2020). This is the area traditionally covered in works on North American languages, such as Boas’ Handbook of North American Indian languages (1911, 1922), Voegelin and Voegelin’s Languages of the World: Native America fascicle one, Languages of the world: Native America fascicle two (1964, 1965), Campbell and Mithun’s The Languages of Native North America: A Historical and Comparative Assessment (1979), Campbell’s American Indian Languages: The Historical Linguistics of Native America (1997), Goddard’s Handbook of North American Indians 17: Languages (1996), and Mithun’s The https://doi.org/10.1515/9783110600926-201
VI
Preface
Languages of Native North America (1999), as well as the Routledge Handbook of North American Languages (2020) edited by Siddiqi et al. In general, we have opted for a greater number of shorter chapters rather than fewer longer ones, with the goal of covering as many relevant topics as possible while striving for user friendliness, though we recognize that the chapters necessarily vary somewhat in their accessibility and interest to different audiences. We were somewhat selective about languages and families, and if we did not include your favorite one, we hope that you will excuse us. It is an exciting time, with knowledge and ideas constantly evolving.
References Bischoff, Shannon T. & Carmen Jany (eds.). 2018. Insights from Practices in Community-Based Research. (Trends in Linguistics Studies and Monographs 319). Berlin/Boston: De Gruyter Mouton. Boas, Franz (ed.). 1911. Handbook of the American Indian languages, Part 1. (Bureau of American Ethnology Bulletin 40). Washington. Boas, Franz (ed.). 1922. Handbook of American Indian languages Part 2. (Bureau of American Ethnology 40). Washington. Campbell, Lyle. 1997. American Indian Languages: The Historical Linguistics of Native America. Oxford: Oxford University Press. Campbell, Lyle & Marianne Mithun (eds.). 1979. The Languages of Native America: A Historical and Comparative Assessment. Austin: University of Texas Press. Goddard, Ives (ed.). 1996. Handbook of North American Indians Volume 17: Languages, 137–157. Washington: Smithsonian Institution. Mithun, Marianne. 1999/2001. The Languages of Native North America. Cambridge: Cambridge University Press. Siddiqi, Daniel, Michael Barrie, Carrie Gillon, Jason Haugen & Eric Matthieu. 2020. Routledge Handbook of North American Languages. London and New York: Routledge. Voegelin, Carl F. & Florence M. Voegelin. 1964. Languages of the World: Native America fascicle one (Anthropological Linguistics). Vol. 6. Bloomington, IN: Anthropology Department, Indiana University. Voegelin, Carl F. & Florence M. Voegelin. 1965. Languages of the World: Native America fascicle two (Anthropological Linguistics). Vol. 7. Bloomington, IN: Anthropology Department, Indiana University.
Table of contents Volume 1 Preface V List of North American families, languages, and dialects Maps XLI
I Sounds and sound structure Sonya Bird, Rae Anne Claxton, and Tess Nolan 1 Acoustic phonetics 3 Heather Bliss, Sonya Bird, and Bryan Gick 2 Articulatory phonetics 39 Hiroto Uchihara 3 Tone 63 Colleen M. Fitzgerald and Matthew K. Gordon 4 Segmental phonology 89 Suzanne Urbanczyk 5 Prosodic morphology Matthew K. Gordon 6 Word prosody
109
135
Siri G. Tuttle 7 Prosody beyond the word
II Words Fernando Zúñiga 8 What is a word? Daniel W. Hieber 9 Word classes
205
183
155
XIII
VIII
Table of contents
III Sentences George Aaron Broadwell 10 Syntax within the clause
247
Elly van Gelderen 11 Negatives 267 Olga Lovick 12 Questions and requests in North American languages Anna Berge 13 Information structure
283
305
Tim Thornes 14 Clause-combining: Relative clauses
323
Amy Dahlstrom 15 Clause combining: Syntax of subordination and complementation Andrew McKenzie 16 Switch-reference and event cohesion
363
IV Discourse Anthony K. Webster 17 Verbal art 385 Olivia N. Sammons 18 Conversation structure
421
V Meaning Sally Rice 19 Lexicalization and lexical meaning Sally Rice 20 Lexicography
479
453
345
Tyler Peterson 21 Evidentiality
Table of contents
497
Robert Henderson 22 Pluractionality and distributivity Andrea Wilhelm 23 Mass and count nouns
511
527
Gary Holton and Andrea L. Berez-Kroeker 24 Sense of place: Space, landscape, and orientation Sihwei Chen and Lisa Matthewson 25 A sense of time and world
547
577
Elizabeth Bogal-Allbritten 26 Pragmatics 599
VI Languages over space and time Marianne Mithun 27 Languages as dynamic systems: How grammar can emerge Sarah Thomason 28 Language contact and linguistic areas Hannah J. Haynie 29 Language classification
647
669
Justin Spence 30 Archival-based sociolinguistic variation Kayla Palakurthy 31 Community-based sociolinguistic variation
689
701
619
IX
X
Table of contents
Volume 2 Preface
V
VII Language revitalization Onowa McIvor, Peter Jacobs, and Barbara Jenni 32 Reviving languages: Outcomes of a Mentor-Apprentice style learning study 719 Melvatha R. Chee and Ryan E. Henke 33 Child and child-directed speech in North American languages
741
Kari A. B. Chew, Wesley Y. Leonard, and Daisy Rosenblum 34 Decolonizing Indigenous language pedagogies: Additional language learning and teaching 767 Ashleigh Surma and Christina L. Truong 35 Digital tools for language revitalization
789
Megan Lukaniec 36 Using archival materials for language reclamation
807
Lorna Wanosts’a7 Williams and Ewa Czaykowska-Higgins 37 Changing notions of fieldwork 823
VIII Language families and isolates Richard Compton 38 Inuit-Yupik-Unangan: An overview of the language family Leslie Saxon 39 Dene – Athabaskan Will Oxford 40 Algonquian Nicole Rosen 41 Michif
951
931
875
843
Table of contents
Clarissa Forbes 42 Tsimshianic
985
T̕łat̕łaḵuł Patricia Rosborough and Daisy Rosenblum 43 Wakashan Languages 1013 Honoré Watanabe 44 Salish 1053 Philip T. Duncan, Valerie (Lamxayat) Switzler, and Henry B. Zenk 45 Chinookan family, with special reference to Kiksht and notes on Chinuk Wawa 1115 Joana Jansen 46 Sahaptian
1139
Andrew Garrett, Susan Gehr, Erik Hans Maier, Line Mikkelsen, Crystal Richardson, and Clare S. Sandy 47 Karuk 1169 M. Ryan Bochnak, Emily A. Hanink, and Alan Chi Lun Yu 48 Wáˑšiw 1201 Eugene Buckley 49 Pomoan 1223 Carmen Dagostino 50 California languages: Isolates and other languages Timothy P. Henry-Rodriguez 51 Chumashan 1275 Amy Miller 52 Yuman
1303
Eric Elliott and David Leedom Shaul 53 Uto-Aztecan 1333 Logan Sutton 54 Kiowa-Tanoan
1361
1247
XI
XII
Table of contents
Logan Sutton and Armik Mirzayan 55 Caddoan 1407 Armik Mirzayan 56 Sketch of the Siouan Language Family Daniel W. Hieber 57 Chitimacha
1519
Judith M. Maxwell and Patricia Anderson 58 Tunica 1545 Jack B. Martin 59 Muskogean
1577
Marianne Mithun and Ryan DeCaire 60 Iroquoian 1601 Raoul Zamponi 61 Unclassified languages
1627
1649 List of Authors Index of languages and varieties Index of names and subjects
1447
List of North American families, languages, and dialects The following is a list of families, languages, and dialects mentioned in the handbook. It is not an exhaustive list of all North American languages and dialects, but just those mentioned in the various chapters. Only North American languages are included here. The list includes the names as they appear in the chapters, in addition to any alternate names and spellings. While we do not intend to perpetuate misnomers and misspellings, we opted to include names that appear in major publications and those still in use so that relevant resources can be identified. We recognize that these names are likely going to change over time. For some languages and dialects, multiple names are widely used, and rather than choosing a single main entry, we decided to include each name as a main entry. Names that include cardinal directions, toponyms, etc. are listed once and cross-referenced (e.g., Apache, Plains -> See: Plains Apache). The third column reflects the genetic affiliation of particular languages and dialects. We chose to list the language family, as well as major branches for some of the larger families. If a name refers to a family, it is noted as such. Families, languages, dialects
Alternate names and spellings
Language family (branch)
‘Wuik̓ala
‘Uik̓ala, Ooweekyala, Oowekyala, Oweek’ala, Oweke(e)no
Wakashan (North Wakashan/ Kwakiutlan)
’Iipay
Ipai; Diegueño; Varieties: Mesa Grande, Barona, Santa Ysabel, Iñaja-Cosmit, San Pasqual
Yuman
’Nak̓wala
Dialect of Kwak’wala
Wakashan (North Wakashan/ Kwakiutlan)
’Uik̓ala
Wakashan (North Wakashan/ Kwakiutlan)
Abenaki, Eastern
See: Eastern Abenaki
Abenaki, Western
See: Western Abenaki
Achumawi
Achomawi
Palaihnihan
Acoma
Keresan
Adai
Isolate
Ahtna
Atna, Atnakenaege’, Copper River, Mednovskiy
Na-Dene (Dene, Northern)
Akokisa
Accokesaws, Arkokisa, Orcoquiza
Unclassified
https://doi.org/10.1515/9783110600926-202
XIV
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Alabama
Alibamu, Albaamaha, Albama, Alabamer, Muskogean Alabamu, Alaba, Albaamo, Aibamo, Aybamo, Halbama, Holbama, Alebamah, Alebamon, Albaamo innaałiilka, Limanu
Aleut
Anangax
Algic
Language family (branch)
Inuit-Yupik-Unangan Algic family
Algonquian
Algonkian
Branch of Algic
Algonquin
Anishinàbemiwin; Variety of Ojibwe
Algic (Algonquian)
Almosan
Hypothetical stock: Almosan (Algonquian-Wakashan)
Alsea
Isolate
Alutiiq
Alutiiq Alaskan Yupik; Pacific Gulf Yupik
Inuit-Yupik-Unangan
Amerind
Hypothetical superstock
Amotomanco
Unclassified
Anishininimowin
Oji-Cree
Apache, Jicarrilla
See: Jicarilla Apache
Apache, Lipan
See: Lipan Apache
Apache, Plains
See: Plains Apache
Apache, San Carlos
See: San Carlos Apache
Apache, Western
See: Western Apache
Apachean
Apache
Algic (Algonquian)
Subbranch of Na-Dene (Dene)
Apalachee
Muskogean
Apalachee-Spanish pidgin
Pidgin (Muskogean/ Indo-European)
Aranama
Anames, Jaranames, Juranames, Xaramenes
Unclassified
Arapaho
Hinónoʼeitíít
Algic (Algonquian)
Arikara Arizona Tewa
Caddoan Tewa, Hopi-Tewa
Asiatic Eskimo
Kiowa-Tanoan Inuit-Yupik-Unangan
Atakapa
Yukhiti
Isolate
Athabaskan
Athabascan, Athapascan, Athapaskan, Athabaskan, Dene
Branch of Na-Dene
Families, languages, dialects
List of North American families, languages, and dialects
Alternate names and spellings
Ati Piman
XV
Language family (branch) Uto-Aztecan (Southern, Tepiman)
Atikamekw
Cree, Attikamek, Tête de Boule, Attimewk, Atihkamekw, Atikamek, Attikamekw
Algic (Algonquian)
Atsugewi
Dialects: Hat Creek, Dixie Valley
Palaihnihan
Aztec-Tanoan
Hypothetical stock
Babine
Witsuwit’en, Nedut’en, Northern Carrier Na-Dene (Dene, Northern)
Baffin Inuktitut
Baffin Island Inuit, Qikiqtaaluup nigianimiutut
Bannock Barbareño Chumash
Inuit-Yupik-Unangan Uto-Aztecan (Numic)
šmuwič, Barbareño
Bay Miwok
Chumashan (Southern, Central) Utian (Miwokan)
Bayogoula
Bayagola, Bayagoula, Baya-Ogoula, Bayugla, Bayuk-okla
Unclassified
Bearlake
Sahtúot’i̜ ne yatǐ̜ , Dene, North Slavey
Na-Dene (Dene, Northern)
Bella Coola
nuxalk
Salishan
Bidai
Beadeye, Bedias, Bidey, Viday
Unclassified
Biloxi
Siouan-Catawban (Siouan, Ohio Valley/Southeastern)
Bitterroot Salish
Seliš, Flathead, Montana Salish
Salishan
Blackfoot
Pikanii, Siksiká, Siksika, Blackfeet, Niitsipowahsin, Pied Noir
Algic (Algonquian)
Caddo
Hasinay; Varieties: Hainai, Kadohadacho, Natchitoches, Yatasi
Caddoan
Caddoan
Caddoan family
Cahita
Uto-Aztecan (Yoemian/Cahitan)
Cahuilla
Ivilyuat; Dialects: Desert, Pass, Mountain Uto-Aztecan (Takic, Cupan)
Calusa
Calloosa, Caloosa, Caloose, Calos, Carlos Unclassified
Campo
See: Northeastern Kumeyaay
Carrier
Dakelh; varieties: Cheslatta, Lheidli, Saik'uz, Stuart Lake
Na-Dene (Dene, Northern)
Cascades
Dialect of Upper Chinook, Watlala
Chinookan
Catawba
Siouan-Catawban (Catawban)
XVI
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Cayuga
Gayogo̱hó꞉nǫʼ
Iroquoian (Northern)
Cayuse
Unclassified
Central Alaskan Yup’ik
Yugtun, Central Alaskan Yupik, Central Yupik, Yupik, Yupʼik, Yugtun, Yugcestun
Inuit-Yupik-Unangan
Central Pomo
Varieties: Hopland/Shanel, Yokaya/ Ukiah, Point Arena-Manchester/Boya, Anderson Valley
Pomoan
Central Siberian Yupik
St. Lawrence Island Yupik
Inuit-Yupik-Unangan
Chalon
Utian (Costanoan/Ohlone)
Chehalis, Lower
See: Lower Chehalis
Chehalis, Upper
See: Upper Chehalis
Chemakum
Chimakum, Chimacum
Chimakuan
Chemehuevi
Southern Paiute
Uto-Aztecan (Numic)
Cherokee
Jalagi, Tsalagi, ᏣᎳᎩ
Iroquoian (Southern)
Cherokee, Western
See: Western Cherokee
Cheyenne
Tsėhésenėstsestȯtse, Tsitsistas
Algic (Algonquian)
Chickasaw
Chikashshanompa’
Muskogean
Chico Maidu
Maiduan
Chikashshanompa’
Chickasaw
Muskogean
Chilcotin
Tsinlhqut’in, Tsilhqut’in, Tsilhqot’in, Tsilhqút’ín
Na-Dene (Dene, Northern)
Chimariko
Varieties: New River (Chimalakwe), Trinity River, South Fork
Isolate
Chinook
Chinookan; Languages: Lower Chinook/ Chinookan family Coastal Chinook, Kathlamet/Katlamat /Cathlamet, Upper Chinook/Kiksht/ Columbia Chinook
Chinook Jargon
Chinook Wawa, Chinuk Wawa, Chinook Jargon, Wawa, Jargon, Chinook
Pidgin
Chinook Proper
Chinook, Tsinuk, Shoalwater Chinook
Chinookan
Chinook, Lower
See: Lower Chinook
Chinook, Upper
See: Upper Chinook
Chinuk Wawa
Chinook Wawa, Chinook Jargon, Wawa
Pidgin
List of North American families, languages, and dialects
XVII
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Chipewyan
Dëne, Dene, Dené, Dene yatié, Dënesųłiné, Dënesųłiné Yatié, Dënësųłinë,́ Dëne Súline, Dene Sǫłiné, Denesųłiné, Denesuline, Dene Súline, Dene Sounlhine, Dene Soun'line, Dëne Dédlıné, Tetsǫ́t’ıné, Montagnais
Na-Dene (Dene, Northern)
Chiricahua Apache
Na-Dene (Dene, Apachean)
Chitimacha
Isolate
Chiwere
Baxoje-Jiwere (includes Ioway, Oto, and Missouria)
Siouan-Catawban (Siouan, Mississippi Valley)
Choctaw
Chahta, Chahta Anumpa
Muskogean
Chumashan
Chumashan family
Clackamas
Clackamas Chinook, Clawiwalla; dialect of Upper Chinook
Chinookan
Clallam
Klallam, Na’klallam, S’klallam, N@ xʷsň’ay’@múc@n, nəxʷsƛʼayʼəmʼúcən
Salishan
Classical Nahuatl
Aztec
Uto-Aztecan
Clatsop
Klatsop, Tlatsop, Chinook, Tsinuk
Chinookan
Coahuilteco
Isolate
Coast Miwok
Utian (Miwokan)
Coast Tsimshian
Tsmksian, Tsimshian, Coast Tsimshian, Sm’algya̲x; dialect of Lower Tsimshianic
Cocopa
See: Cocopah, Cucapá
Cocopah
Kwapá, Kwikapa, Cocopa, Arizona Cocopa
Yuman
Coeur d’Alene
Snchitsu’umshtsn (snčícuʔumšcn)
Salishan (Interior Salish, Southern)
Columbian
Moses-Columbia, nxaʔamxcín (Varieties: Salishan (Interior Salish, Columbia, Wenatchi) Southern)
Colville-Okanagan
Nsyilxcən; Varieties: Northern/Lakes, Southern/Colville
Salishan (Interior Salish, Southern)
Comanche
Nʉmʉ Tekwapʉ
Uto-Aztecan (Numic)
Comecrudo Comox
Tsimshianic
Comecrudan Comox-Sliammon; Varieties: Comox (Island); Sliammon-Homalco-Klahoose (Mainland)
Salishan (Coast Salish)
XVIII
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Comox-Sliammon
Comox, ʔayʔaǰuθəm; Ay-Ay-Ju- ThumEy7a7juuthem; Varieties: Comox (Island); Sliammon-Homalco-Klahoose (Mainland)
Salishan (Coast Salish)
Congaree Coosan
Unclassified Coos
Coosan family
Cora(n)
Uto-Aztecan (Southern, Coracholan-Nahuan)
Cotoname
Isolate
Cowlitz
ƛʼpúlmixq, sƛʼpúlmš
Cree
Salishan (Tsamosan) Algic (Algonquian)
Cree, Plains
See: Plains Cree
Cree, Swampy
See: Swampy Cree
Cree, Woods
See: Woods Cree
Creek
Muskogee, Muscogee, Mvskoke, Seminole
Muskogean
Crow
Apsáalooke, Apsaáloka
Siouan-Catawban (Siouan)
Cruzeño
Island Chumash
Chumashan (Southern)
Cucapá
Kuapá, Cucupá
Yuman
Cup’ik
Variety of Central Alaskan Yupik, Cup’ig, Yup’ikm Yup’ik, Yupic
Inuit-Yupik-Unangan
Cupan
Branch of Uto-Aztecan
Uto-Aztecan
Cupeño
Uto-Aztecan (Takic, Cupan)
Cusabo
Cusabe, Cusabee, Cusaboe, Coosaboys, Corsaboy
Unclassified
Dakelh
Varieties: Cheslatta, Lheidli, S̱aik'uẕ, Stuart Lake
Na-Dene (Dene, Northern)
Dakota
Dakhótiyapi, Dakȟótiyapi, Santee-Sisseton, Sioux, Dakhota, Santee
Siouan-Catawban (Siouan)
Dakota, Stoney
See: Stoney Dakota
Dakotan
Očhéthi Šakówiŋ - “Seven Council Fires” Siouan-Catawban (Siouan, (Dakota ~ Dakhota (Santee-Sisseton), Mississippi Valley) Bdewákhaŋthuŋwaŋ, Waȟpékhute, Waȟpéthuŋwaŋ, Sisíthuŋwaŋ, Dakota ~ Dakhota (Yankton-Yanktonai), Iháŋktȟuŋwaŋ, Iháŋktȟuwaŋna, Lakota ~ Lakhota ~ Lakȟóta), Nakota ~ Nakhota ~ Assiniboine, Nakoda ~ Stoney
List of North American families, languages, and dialects
XIX
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Dane-zaa
Beaver
Na-Dene (Dene, Northern)
Danezāgé’
Denek’éh, Kaska
Na-Dene (Dene, Northern)
Deg Xinag
Deg Hit’an
Na-Dene (Dene, Northern)
Delaware
Two languages: Munsee, Lenape
Algic (Algonquian)
Dena’ina
Tanaina
Na-Dene (Dene, Northern)
Dene
Term used as an alternative for Athabas Na-Dene (Dene, Northern) kan family; also used for language varieties spoken in Mackenzie Valley regions and elsewhere. Also: Slavey, Slave, Chipewyan
Dëne
Dene, Dené, Dene yatié, Dënesųłiné, Na-Dene (Dene, Northern) Dënesųłiné Yatié, Dënësųłinë,́ Dëne Súline, Dene Sǫłiné, Denesųłiné, Denesuline, Dene Súline, Dene Sounlhine, Dene Soun'line, Dëne Dédlıné, Tetsǫ́t’ıné, Montagnais, Chipewyan
Dëne Dédliné
Variety of Chipewyan, Dëne, Dene Sųłiné Na-Dene (Dene, Northern)
Dene Ké
Two languages: Slave or Kaska
Na-Dene (Dene, Northern)
Denek’éh
Kaska, Danezāgé’
Na-Dene (Dene, Northern)
Dene Sųłiné
Dëne, Dené, Dene yatié, Dënesųłiné, Na-Dene (Dene, Northern) Dënesųłiné Yatié, Dënësųłinë,́ Dëne Súline, Dene Sǫłiné, Denesųłiné, Denesuline, Dene Súline, Dene Sounlhine, Dene Soun'line, Dëne Dédlıné, Tetsǫ́t’ıné, Montagnais, Chipewyan
Dene-Dinjii
Term used for languages of the Macken- Na-Dene (Dene, Northern) zie Valley regions, including Gwich’in
Dene-Yeniseian
Hypothetical stock: Dene-Yeniseian
Dhegihan
Languages: Omaha – Ponca, Omaha ~ UmoNhOn, Osage, Kansa ~ Kaw, Quapaw
Sub-branch of Siouan-Catawban (Siouan, Mississippi Valley)
Diegueño
See: ’Iipay, Jamul Tiipay, La Huerta, Los Conejos, Northeastern Kumeyaay
diitiidʔaaʔtx̣
Ditidaht, Nitinaht, Nitinat, Ditidaht-Pacheedaht
Wakashan (South Wakashan, Nootkan)
Diné bizaad
Diné, Navajo, Navaho
Na-Dene (Dene, Apachean)
Ditidaht
Nitinaht
Wakashan (South Wakashan/ Nootkan)
XX
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Dogrib
Tłı̨ chǫ, Tłı̨ chǫ Yatıì, Tlicho, Tlinchon, Doné, Thlingcha-dinneh, Doné
Na-Dene (Dene, Northern)
East Cree
iiyiyuu ayimuun, iiyiyiuyimuwin, iinuuay- Algic (Algonquian) imuwin, iiyiyuuayimuwin
Eastern Abenaki
Penobscot, Caniba
Eastern Pomo
Kulanapo; Varieties: Habematolel/Upper Pomoan Lake, Big Valley, Robinson Rancheria
Algic (Algonquian)
Erie
Iroquoian (Northern)
Eskimo-Aleut
Inuit-Yupik-Unangan family
Esselen
Huelel
Isolate
Eudeve
Dohema
Uto-Aztecan (Opatan)
Eurasiatic
Hypothetical stock: Eurasiatic
Eyak
dAXunhyuuga’
Na-Dene
Fort Ware Sekani
Kwadacha Tsek’ene
Na-Dene (Dene, Northern)
Fox
Mesquakie (Meskwaki), Mesquakie-Sauk, Mesquakie-Sauk-Kickapoo, Sauk-Fox
Algic (Algonquian)
Gitksan
Gitxsan, Gitxsen, Gitxsanimaax, Giatikshan, Gityskyan, Gitxsanimx̲, Gitxsanimax̲, Gitxsenimx̲, Gitsenimx̲, Gyaanimx̲, Nass-Gitksan; dialect of Nass-Gitksan
Tsimshianic
Greenlandic
Greenlandic Kalaallisut; dialects: Kalaal- Inuit-Yupik-Unangan lisut/West Greenlandic; Tunumiisut/East Greenlandic; Inuktun; Thule
Greenlandic, West
See: West Greenlandic
Gros Ventre
Ananin, Ahahnelin, Ahe, A’ani, Atsina
Algic (Algonquian)
Guale
Unclassified
Guarijio
Uto-Aztecan (Southern, Tarahumaran)
Gwich’in
Kutchin, Loucheux, Takudh, Tukidh, Tukudh, Dinju Zhuh K’yuu
Na-Dene (Dene, Northern)
Haida
Haidah, Xaad Kil, X̱aat Kíl, X̱aadas Kíl, Isolate X̱aayda Kil, Xaaydaa Kil; Varieties: Skittagetan, Haidah, Tolewa Skittagits, Tahlewah Haidah, Queen Charlotte’s Island, Masset, Skidegate
Haida, Northern
See: Northern Haida
List of North American families, languages, and dialects
XXI
Families, languages, dialects
Alternate names and spellings
Haíɫzaqvḷa
Híɫzaqvḷa, Hailhzaqvla, Haílhzaqvḷa, Wakashan (North Wakashan/ Heiltsuk, Heiltsuq, Heiltsukv, Heiltsukvla, Kwakiutlan) Bella Bella
Halkomelem
Holkomelem, Hul’q’umi’num’, Halq̓eméylem, Hul̓q̓umín̓um̓, hən̓q̓əmin̓əm; Varieties: Chilliwack/ Upriver (halqʼəméyləm), Musqueam/Downriver (hənʼqʼəmínʼəmʼ), Cowichan/Island (həlʼqʼəminʼəmʼ)
Halkomelem, Upriver
See: Upriver Halkomelem
Hän
Han-Kutchin, Dawson, Moosehide
Na-Dene (Dene, Northern)
Hanis
Hanis Coos
Coosan
Hare
K’áshogot’ine xedə́, Dene, North Slavey
Na-Dene (Dene)
Havasupai
Havasu Baa Gwaawa
Yuman
Heiltsuk
Haíɫzaqv
Wakashan (North Wakashan/ Kwakiutlan)
Hidatsa
Hiraacá
Siouan-Catawban (Siouan, Missouri Valley)
Híɫzaqv
Heiltsuk, Hailhzaqvla, Bella Bella, Heiltsuk-Oweek’ala, Belbellah, Heiltsuk-Oowekyala
Wakashan (North Wakashan/ Kwakiutlan)
Hinono’eitiit
Arapaho
Algic (Algonquian)
ḥiškʷiiʔatḥ
Hesquiaht (dialect of nuučaan̓uł)
Wakashan (South Wakashan/ Nootkan)
Hitchiti Hocąk
Language family (branch)
Salishan (Coast Salish)
Muskogean Hocank, Hoocągra, Hochunk, Ho-Chunk, Siouan-Catawban (Mississippi Winnebago Valley Siouan)
Hokan
Hypothetical stock: Hokan
Hokan-Siouan
Hypothetical stock: HokanSiouan
Hopi
Hopílavayi
Uto-Aztecan
Hualapai
Walapai, Hwalbáy
Yuman
Huichol Hul’q’umi’num’
Uto-Aztecan (Coracholan- Nahuan) Hul̓q̓umín̓um̓; Halkomelem; Island Halkomelem
Salishan (Coast)
XXII
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Hupa
Hoopa, Na꞉tinixwe Mixine꞉wheʼ
Na-Dene (Dene, Pacific Coast)
Ichishkíin, Yakima
See: Yakima Ichishkíin
Innu
Montagnais, Innu-aimun
Algic (Algonquian)
Inuinnaqtun
Western Canadian Inuktun, Kangiryuarmiutun
Inuit-Yupik-Unangan
Inuit
Inuinnaqtun; Inuktitut; Inuktun; Inuttitut; Inuttut; Inuvialuktun
Inuit-Yupik-Unangan
Inuit-Yupik-Unangan
Eskimo-Aleut, Eskaleut
Inuit-Yupik-Unangan family
Inuktitut
Inuit, ᐃᓄᒃᑎᑐᑦ, Eastern Canadian Inuktitut, Inuktitut, Inuttitut, Inuttut
Inuit-Yupik-Unangan
Inuktitut, North Baffin
See: North Baffin Inukitut
Inuktitut, Tarramiut
See: Tarramiut Inukitut
Inupiaq
Iñupiaq, Iñupiatun, Inupik, Inupiak, Inupiat, North Alaskan Inuktitut, North Alaskan Inupiaq, North Alaskan Inupiat, North Alaskan, Inupiatun, Northwest Alaska Inupiat, Seward Peninsula Inupiaq
Inuit-Yupik-Unangan
Inuvialuktun
Western Canadian Inuktun, Inuktun
Inuit-Yupik-Unangan
Iroquoian
Iroquoian family
Jamul Tiipay
Jamul Diegueño, Tipai, Kumeyaay
Yuman
Ja’a
Tipey Aa, Tipei, Ha’a, Kumiay, Kumiai, Kumeyaay
Yuman
Jicarilla Apache
Na-Dene (Dene, Apachean)
Jova
Uto-Aztecan
Kalaallisut
West Greenlandic, Greenlandic
Kalapuyan
Inuit-Yupik-Unangan Kapaluyan family
Kalispel
Spokane
Salishan (Interior Salish, Southern)
Kanien’kéha’
Mohawk, Kanyen’kéha
Iroquoian (Northern)
Kansa
Kaw, Dhegiha Siouan
Siouan-Catawban (Siouan)
Kanyen’kéha
Mohawk, Kanien’kéha’
Iroquoian (Northern )
Karankawa
Isolate
Karkin
Utian (Costanoan/Ohlone)
Karuk
Karok
Isolate
List of North American families, languages, and dialects
XXIII
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Kashaya Pomo
Kashaya, Kashia, Southwestern Pomo (obsolete); Varieties: Stewart’s Point, Fort Ross, Haupt Ranch
Pomoan
K’áshogot’ine xedə́
Hare dialect of North Slavey, variety of Mackenzie Valley Dene
Na-Dene (Dene, Northern)
Kaska
Denek’éh, Dene K’é, Danezāgé’
Na-Dene (Dene, Northern)
Kathlamet
Katlamat, Cathlamet, Wakaikam,
Chinookan
Kato
Cahto
Na-Dene (Dene, Pacific Coast)
Kaw
Siouan
Kawaiisu
Uto-Aztecan (Numic)
Keres
Keresan
Isolate
Keresiouan
Hypothetical stock: Keresiouan
Kickapoo
Algic (Algonquian)
Kiksht
Upper Chinook
Chinookan
Kiliwa
Kiliwi, Koleeu, Ko’lew, Quiligua
Yuman
Kiowa
Cáuijògà/Cáuijò:gyà, Cáuígú, [Gáuigú, kɔ́ygú
Kiowa-Tanoan
Kiowa-Tanoan
Kiowa-Tanoan family
Kitanemuk
Uto-Aztecan (Takic/Serrano)
Kitsai
Kichai, Keechi
Caddoan
Klallam
nəxʷsƛʼáyʼəmʼucən, Clallam
Salishan (Coast Salish)
Klamath-Modoc
Varieties: Klamath, Modoc, Lutuamian
Isolate
Koasati
Coushatta, Koasáti
Muskogean (Eastern Muskogean)
Ko’alh
Ku’alh, Ko’al, Kw’aal
Yuman
Konkow
Maiduan
Konomihu
Shastan
Koyukon
Denaakk’e, Ten’a
Na-Dene (Dene, Northern)
Ktunaxa
Kutenai, Kootenay, Ksanka, Kootenai
Isolate
Kumeyaay
See: ’Iipay, Los Conejos, Northeastern Kumeyaay, Jamul Tiipay, La Huerta, Nejí, Ja’a, Peña Blanca, San Antonio Necua, San Jose de la Zorra, San José de Tecate
Kumeyaay, Northeastern
See: Northeastern Kumeyaay
XXIV
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Kutenai
Ksanka, Ktunaxa, Kootenai, Kootenay
Isolate
Kwadacha Tsek’ene
Fort Ware Sekani
Na-Dene (Dene, Northern)
Kw’ally
kʷʔaƚʸ, Kw’ally
Yuman
Kwak’wala
Kʷak’ʷala , Kwakwala, Kwakiutl, Wakashan (North Wakashan/ Kwagu’ł, Kwaguł, Kwagulth, Kwagu’tl, Kwakiutlan) Kwakwaka’wakw, Kwakwakawakw, Bakwumk’ala, Bak’wa̱mk’ala, Bak̓ʷəmk̓ala, Kwakwaka’wakw, Kwagiutl; (Varieties: ‘Nak̕wala, Gwat̕sala, G̱ut̕sala, Kwak̓wala, Liq’wala)
Kwatsáan
Quechan, Yuma, Kwtsaan
Kyuquot
Yuman Wakashan (South Wakashan/ Nootkan)
La Huerta
Tipay, Tipai, Cuchimí, Kumiay, Kumeyaay Yuman
Laguna Keres
Dialect of Keres
Lake Miwok
Isolate Utian (Miwokan)
Lakhota
Lakota, Lakhota, Lakȟótiyapi, Lakotiyapi, Siouan-Catawban (Siouan) Teton, Teton Sioux
Laurentian
St. Laurence Iroquois
Iroquoian (Northern)
Lekwungen
lək̓ʷəŋiʔnəŋ, Songhees; Songish; Northern Straits Salish
Salishan (Coast Salish)
Lenape
Lenape Delaware
Algic (Algonquian)
Lillooet
Sƛ̓áƛ̓imxəc, St̓át̓imcets; Varieties: Upper/ Salishan (Interior Salish, NorthFountain/Fraser River, Lower/Mount ern) Currie
Lingít
Tlingit
Lipan Apache
Na-Dene Na-Dene (Dene, Apachean)
Los Conejos
Kumeyaay, Diegueño, ’Iipay/Tiipay
Yuman
Loup
Nipmuck
Algic (Algonquian)
Lower Chehalis
ɬəwʼálʼməš (Varieties: Humptulips, Wynochee, Westport-Shoalwater)
Salishan (Tsamosan Salish)
Lower Chinook
Coastal Chinook; dialects: Shoalwater/ Chinook proper, Clatstop
Chinookan
Luiseño
Luiseño-Juaneño
Uto-Aztecan (Takic, Cupan)
List of North American families, languages, and dialects
XXV
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Lushootseed
xʷəlšucid, dxʷləšúcid, Puget Sound Salish, Skagit-Nisqually, Dxʷl@šucid, Northern Lushootseed, Northern Puget Sound Salish, dxʷləšúcid; Varieties: Northern (Skagit, Snohomish); Southern (Duwamish-Suquamish, Puyallup, Nisqually
Salishan (Coast Salish)
Macro-Algonquian
Hypothetical stock: Macro-Algonquian
Macro-Siouan
Hypothetical stock: Macro-Siouan
Mackenzie Valley Dene
Varieties of Dene spoken in the Na-Dene (Dene, Northern) Mackenzie Valley; Dene, Dene K’é, Slave, Slavey, North Slavey, South Slavey, K'áshogot'ine xedə́, Shúhtaot’ine yatí, Sahtú Dene, Deh Cho, Deh Gáh, MV Dene
Mahican
Algic (Algonquian)
Maidu
Northeastern Maidu, Mountain Maidu
Maidu, Mountain
See: Mountain Maidu
Maiduan
Maidun
Maiduan family
Mainland Comox
Comox-Sliammon
Salishan (Coast Salish)
Makah
Qʷiqʷidiččaq
Wakashan (South Wakashan/ Nootkan)
Maliseet
Malecite
Algic (Algonquian)
Mandan
Maiduan
Siouan-Catawban (Siouan)
Maricopa
Piipaash, Pee-Posh, Cocomaricopa; Xalychidom (Halchidhoma), Kavelychidom (Kavelchadom)
Yuman
Massachusett
Natick, Wampanoag, Wôpanâak
Algic (Algonquian)
Mattole
Mattole-Bear River
Na-Dene (Dene, Pacific Coast)
Mayo
Dialect of Cáhita
Uto-Aztecan (Tarahumaran)
Mednyj Aleut
Copper Island Aleut; bilingual mixed language [Aleut + Russian]
Mixed language
Menominee
Menomini, Oma͞ eqnomenew
Algic (Algonquian)
Mesa Grande ’Iipay
See: ’Iipay
XXVI
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Meskwaki
Mesquakie, Fox, Meskwaki-Sauk-Kickapoo
Algic (Algonquian)
Mi’kmaq
Mi’gmaq, Mi’kmaw, Micmac
Algic (Algonquian)
Miami-Illinois
Myaamia
Algic (Algonquian)
Michif
Métchif, Mitchif, Chippewa Cree, Michif Mixed language; generally clasCree, Heritage Michif, Southern Michif, sified as Algic (Algonquian) not-the-real Cree, Métif; bilingual mixed language [Cree + French], French Cree, Southern Michif
Mikasuki
Miccosukee, Hitchiti
Muskogean
Miluk
Miluk Coos
Coosan
Minto Lower Tanana
Na-Dene (Dene, Northern)
Miwok
Miwokan, Moquelumnan
Miwok, Plains
See: Plains Miwok
Miwok, Sierra
See: Sierra Miwok
Mobilian Jargon
Mobilian trade language, Mobilian Trade Jargon, Chickasaw-Choctaw trade language; Yamá, Moquelumnan
Pidgin
Mohave
Mojave, Hamakhav, Upriver Yuman
Yuman
Mohawk
Kanienʼkéha’, Kanyen’kéha
Iroquoian (Northern)
Mohegan-Pequot-Montauk
Utian (Miwokan)
Algic (Algonquian)
Mojave
See: Mohave
Molalla
Molala, Molele, Molalla-Sahaptian, Molalla
Molalla-Sahaptian
Mono
Owens Valley Paiute
Uto-Aztecan (Numic)
Mono, Western
See: Western Mono
Montagnais
Innu-aimun
Algic (Algonquian)
Montana Salish
Flathead, Kalispel–Pend d’Oreille, SelišQl’ispé, Salish-Pend d’Oreille, Séliš
Salishan (Southern Interior Salish)
Monyton
Moniton
Siouan-Catawban (Siouan, Ohio Valley/Southeastern)
Moose Cree
Algic (Algonquian)
Moses-Columbia
Nxaʔamxčín , Nxa’amxcin
Salishan
Mountain Dene
Shúhtaot’ine yatī; variety of North Slavey, Mackenzie Valley Dene
Na-Dene (Dene, Northern)
List of North American families, languages, and dialects
XXVII
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Mountain Maidu
Maidu
Maiduan
Multnomah
Dialect of Upper Chinook. Wakanasisi, Wacanassisi
Chinookan
Munsee
Munsee Delaware
Algic (Algonquian)
Muskogean
Muskhogean
Muskogean family
Muskogee
Muscogee, Creek, Mvskoke, Creek-Seminole, Muskokee, Mvskoke, Maskoke, Seminole dialect of Creek
Muskogean (Eastern)
Musqueam
Hən̓q̓əmin̓əm̓, variety of Halkomelem
Salishan
Mutsun
Costanoan, Ohlone (as in “Mutsun Ohlone”)
Utian (Costanoan)
Mutsun-Awaswas
Mutsun
Utian (Costanoan)
MV Dene
See: Mackenzie Valley Dene
Na-Dene (Dene)
Mvskoke
Muskogee, Creek, Muscogee, Creek-Seminole
Muskogean (Eastern)
Myaamia
myaamiaataweenki, Miami, Miami-Illinois, Illinois, Maumee, myaamia, Twightwee, Wea
Algic (Algonquian)
Na-Dene
Nadene, Athabaskan-Eyak-Tlingit
Na-Dene family
Nakota
Assiniboine
Siouan-Catawban (Siouan)
Nansemond
Unclassified
Nanticoke
Algic (Algonquian)
Narragansett
Algic (Algonquian)
Naskapi
Algic (Algonquian)
Natchez
Isolate
Naukan
Naukanski, Naukan(ski) Siberian Yupik
Inuit-Yupik-Unangan
Navajo
Dine, Diné, Dineh, Diné bizaad, Navaho, Naabeehó bizaad
Na-Dene (Dene, Apachean)
Nedut’en
Witsuwit’en, Northern Carrier, Babine
Na-Dene (Dene, Northern)
Nee’aanèegn’
Upper Tanana
Na-Dene (Dene, Northern)
nēhinawēwin
ᓀᐦᐃᓄᐍᐏᐣ, Swampy Cree, n-Dialect
Algic (Algonquian)
Nêhiyawêwin
ᓀᐦᐃᔭᐍᐏᐣ, Plains Cree, Y-Dialect
Algic (Algonquian)
Nejí
Tipey Aa, Tipei, Kumiay, Kumiai, Kumeyaay
Yuman
XXVIII
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Neutral Nevome
Iroquoian (Northern) Onabas Piman
New River Shasta Nez Perce
Uto-Aztecan (Tepiman) Shastan
Nez Percé, Nimiipuutímt, Niimiipuutímt, Sahaptian Nimipuutímt, Niimi’ipuutímt, Niimíipu, Nuumiipuutimt, Numí, Chopunnish, titoqatímt
Nez Perce Jargon Nimipuutímt
Language family (branch)
No affiliation See: Nez Perce
Nisenan
Maiduan
Nisg̱a’a
Dialect of Nass-Gitksan: Nishga, Niska’, Tsimshianic Nisk’a’, Nass, Nisgha, Nisk’a, Nishka, Niska, Nasqa’, Nisgha, Nisg̲a’amx̲, Lisims
Nishnaabemwin
Jibwemwin, Odawa, Ottawa, Chippewa, Eastern Ojibwa, Ojibway, Ojibwe
Algic (Algonquian)
Nitinaht
Nitinaht, Nitinat, Ditidaht, Southern Nootkan
Wakashan (South Wakashan/ Nootkan)
Nlaka’pamux
Thompson; Thompson River Salish; Nłeʔkepmxcín
Salishan (Northern Interior Salish)
Nomlaki
Wintu
Wintuan (Northern)
Nooksack
Lhéchalosem (ɬə́čələsəm)
Salishan (Coast Salish)
Nootka
Nuu-chah-nulth, Nuuchahnulth, nuučaan̓uɫ, Nutka
Wakashan (South Wakashan/ Nootkan)
North Baffin Inuktitut
Qikiqtaaluk uannangani
Inuit-Yupik-Unangan
North Slavey
Variety of Mackenzie Valley Dene. Dene, Slave, K'áshogot'ine xedə́, Shúhtaot’ine, Sahtú Dene
Na-Dene (Dene, Northern)
Northeastern Kumeyaay
Diegueño, Kumeyaay (proper), Tipai; Varieties: Campo, Imperial Valley, La Posta, Baron Long, Manzanita, Sycuan, ’Ewiiaapaayp (Cuyapaipe)
Yuman
Northeastern Pomo
Salt Pomo, Stony Creek, Stonyford
Pomoan
Northeastern Yavapai
Prescott Yavapai, Yavpé, Yavape
Yuman
Northern Carrier
Witsuwit’en, Nedut’en, Babine
Na-Dene (Dene, Northern)
Northern Haida
Dialect of Haida
Isolate
Northern Paiute
Paviotso, Numu, Northern Paiute- Bannock
Uto-Aztecan (Numic)
List of North American families, languages, and dialects
XXIX
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Northern Pomo
Redwood Cañon; Varieties: Pinoleville, Guidiville, Potter Valley, Coyote Valley, Sherwood Valley/Little Lake, Little Noyo River
Pomoan
Northern Straits
SENĆOŦEN, Saanich, Straits, NorthSalishan (Coast Salish) ern Straits Salish; Varieties: Sooke (T’Sou-ke), Saanich (SENĆOŦEN), Songhees/Songish (Lekwungen), Semiahmoo/ SEMYOME, Samish/ Xws7ámeshqen, Lummi (Xwlemi’chosen)
Northern Tepehuan
Uto-Aztecan (Southern, Tepiman)
Northern Tutchone
Dän kʼí
Northern Yuman
See: Hualapai, Havasupai, Yavapai
Nottoway
Na-Dene (Dene, Northern)
Iroquoian (Northern)
Nsyilxcen
n̓səl̓xcin̓, Okanagan, Colville-Okanagan, Nsyilxcən, Nsyilxcn, nsyílxcən, nsəlxcin
Nuu-wee-ya’
Three-dialect cluster including varieties Na-Dene (Dene, Pacific Coast) in s. Oregon and n. California. CoquilleRogue River-Tututni-Tolowa-Chetco-Galice-Applegate/Dakubeh
Nuuchahnulth
Nuu-chah-nulth; Nootka, Nuučaan̓uɫ, Nutka, Aht, West Coast, T’aat’aaqsapa; Varieties: Barkley, Central, Northern
Wakashan (South Wakashan/ Nootkan)
nuxalk
Bella Coola
Salishan
Nxa’amxcin
Moses Columbia Salish, Columbian Salish, Nxaʔamxčín
Salishan (Interior Salish, Southern)
O’odham
Papago, Pima-Papago
Uto-Aztecan (Tepiman)
Salishan (Interior Salish, Southern)
Obispeño
Chumashan (Northern)
Occaneechi
Siouan-Catawban (Siouan, Ohio Valley/Southeastern)
Ofo
Siouan-Catawban (Siouan, Ohio Valley/Southeastern)
Ohlone
Costanoan
Utian (Costanoan/Ohlone)
Oji-Cree
Variety of Ojibwe
Algic (Algonquian)
Ojibwe
Anishinaabemowin, Nishnaabemwin, Algic (Algonquian) Ojibway, Ojibwa, Otchipwe, Mississauga, Saulteaux, Chippeway, Southwestern Chippewa, Daawaamwin
XXX
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Okanagan
Nsyilxcən, Nsyilxcn, Nsəlxcin, Nsilxcín, Colville-Okanagan
Salishan (Interior Salish, Southern)
Okwanuchu
Shastan
Omaha
Omaha-Ponca
Siouan-Catawban (Siouan)
Oneida
Onʌyota’a:ka
Iroquoian (Northern)
Onondaga
Onoñdaʼgegáʼ nigaweñoʼdeñʼ
Iroquoian (Northern)
Oob No’ok
Uto-Aztecan (Tepiman)
Osage
Siouan-Catawban (Siouan)
Paipai
Paʔipa:y, Pa’ipai, Pa Ipai, Pai Pai, Akwa’ala
Paiute, Northern
See: Northern Paiute
Paiute, Southern
See: Southern Paiute
Pamlico
Carolina Algonquian
Pamunkey Passamaquoddy
Algic (Algonquian) Unclassified
Passamaquoddy-Maliseet
Patwin
Algic (Algonquian) Wintuan
Pawnee
Chaticks si chaticks, Cahriksicahriks, Cahiksicahiks; Varieties: Skidi, Skiri, South Band
Pee-Posh
See: Maricopa
Peña Blanca Pentlatch
Yuman
Caddoan
Yuman pənƛʼáč, pentl̓ach
Penutian
Salishan (Coast Salish) Hypothetical stock: Penutian
Peoria
Miami-Illinois
Algic (Algonquian)
Petun
Tionontati, Hkionontate
Iroquoian (Northern)
Picuris Northern Tiwa
Picuris Tiwa; Varieties: Tiwa, Picurís, Northern Tiwa
Kiowa-Tanoan
Pidgin Delaware
Pidgin Unami, Delaware Jargon, Trader’s Pidgin Jargon
Pidgin Inuit
There were two different ones.
Pidgin
Piipaash
See: Maricopa
Yuman
Pima
Pima Bajo; Mountain Pima; Lowland Pima; Nevome
Uto-Aztecan
Plains Apache
Na-Dene (Dene, Apachean)
List of North American families, languages, and dialects
XXXI
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Plains Cree
ᓀᐦᐃᔭᐍᐏᐣ nēhiyawēwin, Nêhiyawêwin
Algic (Algonquian)
Plains Indian Sign Language
Pidgin
Plains Miwok
Utian (Miwokan)
Pochutec
Uto-Aztecan (Southern, Coracholan-Nahuan)
Pomo
Pomoan
Pomo, Eastern
See: Eastern Pomo
Pomo, Kashaya
See: Kashaya Pomo
Pomo, Northeastern
See: Northeastern Pomo
Pomo, Northern
See: Northern Pomo
Pomo, Southeastern
See: Southeastern Pomo
Pomo, Southern
See: Southern Pomo
Pomoan
Pomo
Pomoan family
Potawatomi
Pottawatomie, Bodéwadmi, Bodéwadmimwen, Bodéwadmi Zheshmowen,
Algic (Algonquian)
Powhatan
Virginia Algonquian
Algic (Algonquian)
Purisimeño Ql'ispé
Pomoan family
Chumashan (Southern, Central) Upper Pend d’Oreille
Quapaw
Salishan (Southern Interior Salish) Siouan-Caddoan (Siouan)
Quechan
See: Kwatsáan
Quileute
Quillayute
Chimakuan
Quinault
kʷínayɬ (Varieties: Queets, Quinault)
Salishan (Tsamosan Salish)
qʷi·qʷi·diččaq
Makah
Wakashan (South Wakashan/ Nootkan)
Rio Grande Tewa
Varieties: Tewa, Nambé, Nanbé, Ohkay, Kiowa-Tanoan San Juan, Pojoaque, Santa Clara, Northern Tewa, San Ildefonso, San I [sæ̨n ˀaɪ], Tesuque
Ritwan
California Algic
Rumsen Saanich Sahaptian
Hypothetical branch of Algic Utian (Costanoan/Ohlone)
Northern Straits Salish, SENĆOŦEN
Salishan (Coast Salish) Sahaptian family
XXXII
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Sahaptin
Ichishkíin, Ichishkin, ’Ichishkin, ‘Ichish- Sahaptian kíin, Chíshkin, Shahaptin, Shawpatin; Variesties: Northeast dialects: Palouse, Palus, Peluuspem, Walla Walla, Walúula, Priest Rapids, Wanapam, Lower Snake, Wawyukma, Naxiyampam, Chamnapam; Northwest dialects: Klickitat, X̱ waɬx̱waypam, Klikitat, Kittitas, Pshwanwapam, Yakima,Yakama, Mamachatpam, Wanapum, Taytnapam, Upper Cowlitz, Taitnapam, Upper Nisqually, Mishálpam; Southern/Columbia River dialects: Umatilla, Imatalam, Rock Creek, Q’miɬama, John Day, Celilo, Wayamɬama, Tenino, Tinaynu, Tygh Valley, Warm Springs); Includes Tanino, Umatilla, Walla Walla, Yakama
Sahaptin, Yakima
See: Yakima Sahaptin
Sahtúot’i̜ ne yati̜
Sahtugot’ine yati, Sahtú, Bearlake, Dene; variety of Mackenzie Valley Dene
Na-Dene (Dene, Northern)
Salinan
Dialects: Antoniano, Migueleño
Isolate
Salish
Salishan
Salishan family
Salishan, Southern Interior
See: Southern Interior Salishan
Samala
Ineseño
San Antonio Necua San Carlos Apache
Language family (branch)
Chumashan (Southern, Central) Yuman
Variety of Western Apache
San Francisco Bay Costanoan
Na-Dene (Dene, Apachean) Utian (Costanoan)
San José de la Zorra
Tipey Aa, Kumiay, Kumiai, Kumeyaay
Yuman
San José de Tecate
Tipai, Kumiay, Kumiai, Kumeyaay
Yuman
Santee
Dakota
Siouan-Catawba (Siouan)
Saponi
Sappony
Siouan-Catawban (Siouan, Ohio Valley/Southeastern)
Sarcee
Tsúut’ínà, Tsuut'ina, Sarsi
Na-Dene (Dene, Northern)
Sauk
Meskwaki-Sauk-Kickapoo, Sac
Algic (Algonquian)
Sechelt
Shashishalhem (šášíšáɬəm), She shashishalhem
Salishan (Coast Salish)
Secwepemctsín
Shushwap, Shuswap, Secwepemc, At-nah, Shooswap, səxwəxcín, Secwepemctsía
Salish (Interior Salish, Northern)
List of North American families, languages, and dialects
XXXIII
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Sekani
Tse’khene, Tsek’ene
Na-Dene (Dene, Northern)
Sekani, Fort Ware
See: Fort Ware Sekani
Seliš
Séliš, Bitterroot Salish, Flathead, Montana Salish
Seminole
Salishan (Interior Salish, Southern) Muskogean
SENĆOŦEN
Sənčaθən, Saanich, Northern Straits Salish
Salish (Coast Salish)
Seneca
Onödowáʼga꞉
Iroquoian (Northern)
Seri
Comcaac, Comcáac, Comcáackg, ConIsolate caac, Concáac, Congcaac, Cuncaac, Komkak, Konkaak, Kunkaahac, Kunkaak, Könkáak
Serrano
Uto-Aztecan (Takic, Serran)
Sewee
Unclassified
Shasta
Shasta proper
Shasta, New River
See: New River Shasta
Shawnee She shashishalhem
Shastan
Algic (Algonquian) Sechelt
Shoccoree-Eno
Salishan (Coast Salish) Unclassified
Shoshone
Shoshoni, Shoshoni-Gosiute, Neme ta̲ikwappe, Sosoni’ ta̲ikwappe,
Uto-Aztecan (Numic)
Shúhtaot’ine yati
Mountain Dene; variety of North Slavey, Mackenzie Valley Dene
Na-Dene (Dene, Northern)
Shuswap
Secwepemctsín (səxʷepməxcín)
Salishan (Interior Salish, Northern)
Siberian Yupik
Bering Strait Yupik, Central Siberian Yupik, Saint Lawrence Island Yupik, Sivuqaghmiistun, Yoit, Yuit
Inuit-Yupik-Unangan
Sierra Miwok
Moquelumnan, Miwuk
Utian (Miwokan)
Sierra Miwok, Southern
See: Sourthern Sierra Miwok
Siletz
Salishan (Tillamook/hutyéyu)
Siletz Dee-ni
Tolowa
Na-Dene (Dene, Northern)
Siouan
Branch of Siouan-Catawban
Siouan-Catawban family
Siouan-Catawban Sirenik
Siouan-Catawban family Sirenikski
Inuit-Yupik-Unangan
XXXIV
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Siuslaw
Sayúskla, Siuslawan
Isolate
Sḵwx̱wu7mesh
Squamish, Sḵwx̱wu7mesh snichim
Salishan (Coast Salish)
Slave
Dene, Mackenzie Valley Dene, Slavey, Slavé, Slavi, North Slavey, South Slavey, Dené Tha, Dené, Deh Cho, Deh Gáh, Dene K’é, Got’ine, Mackenzian See also: North Slavey, South Slavey
Na-Dene (Dene, Northern)
Slavey Jargon
Slave Jargon
Pidgin
Sliammon
Comox-Sliammon, Éy7á7juuthem, ʔayʔaǰuθəm
Salishan (Coast Salish)
Sm’algya̱x
See: Coast Tsimshian
Snchitsu’umshtsn
Coeur d’Alene
Salishan (Interior Salish, Southern)
Solano
Unclassified
Sonoran O’otam
Uto-Aztecan (Tepiman)
Souriquois Jargon
Pidgin
South Slavey
Variety of Mackenzie Valley Dene. Dene, Dené, Dehcho Dene, Dené Tha, Deh Cho, Deh Gáh, Slave, Slavey
Na-Dene (Dene, Northern)
Southeastern Pomo
Varieties: Sulphur Bank/Elem, Lower Lake/Koi
Pomoan
Southeastern Yavapai
Kewevikopaya, Kwevkapaya, Kewevkopaya, Kewevkapaya
Yuman
Southern Interior Salishan
Branch of Salishan
Southern Paiute
Southern Paiute-Ute, Colorado River Uto-Aztecan (Numic) Numic, Ute-Chemehuevi; Dialects: Chemehuevi, Southern Paiute, Northern Ute
Southern Pomo
Salmonhole, West Creek, Dry Creek/ Mihilakhawna, Cloverdale, Lytton, Graton, Gallinoméro
Southern Sierra Miwok
Pomoan
Utian (Miwokan)
Southern Tiwa
Varieties: Tiwa, Isleta, Sandia, Ysleta del Sur, Ysleta D
Kiowa-Tanoan
Southern Tsimshian
Sgüüx̣s (dialect of Tsimshian
Tsimshianic
Southern Tutchone
Dän kʼè
Na-Dene (Dene)
Southern Ute
Uto-Aztecan (Numic)
Families, languages, dialects
List of North American families, languages, and dialects
Alternate names and spellings
Southwestern/Southeastern Tepehuan
XXXV
Language family (branch) Uto-Aztecan (Tepiman)
Spokane
Npoqínišcn
Salishan (Interior Salish, Southern)
Squamish
Sḵwx̱wú7mesh, sqʷx̣ʷúʔməš, sníchim
Salishan (Coast Salish)
St’át’imcets
Sƛ̓áƛ̓imxəc, St̓át̓imcets, Lillooet, Lillooet Salish, Lil’wat7úlmec (dialect)
Salishan (Interior Salish, Northern)
Stoney Dakota
Siouan-Catawban (Siouan)
Straits, Northern
See: Northern Straits
Susquehannock
Conestoga, Andaste
Swampy Cree
Iroquoian (Northern) Algic (Algonquian)
Tagish
Den k'é
Na-Dene (Dene, Northern)
Tahltan
Tāłtān
Na-Dene (Dene, Northern)
Tāłtān
Tahltan
Na-Dene (Dene, Northern)
Takelma
Takilma
Isolate
Tanacross
Neeʼaandegʼ
Na-Dene (Dene, Northern)
Tanana
Na-Dene (Dene, Northern)
Tanana, Minto Lower
See: Minto Lower Tanana
Tanana, Upper
See: Upper Tanana
Tanpachoa
Unclassified
Taos Northern Tiwa
Varieties: Tiwa, Taos, Northern Tiwa
Tarramiut Inuktitut
Canadian Inuit, Inuit, Inuktut, Nunavim- Inuit-Yupik-Unangan miutut, Nunavik dialect of Inuktitut
Tawasa Teguima
Kiowa-Tanoan
Timucuan Opata
Tepecano
Uto-Aztecan Opatan) Uto-Aztecan (Tepiman)
Tepehuan, Northern
See: Northern Tepehuan
Tepehuan, Southwestern/ Southeastern
See: Southwestern/Southeastern Tepehuan
Tetsǫ́t’ıné
Tetsǫ́t’ıné yatıé; variety of Dene Sųłiné/ Chipewyan
Na-Dene (Dene, Northern)
Tewa
Tano, dialect of Arizona Tewa
Kiowa-Tanoan
Tewa of Arizona
Arizona Tewa, Hopi-Tewa
Kiowa-Tanoan
Tewa, Rio Grande
See: Rio Grande Tewa
XXXVI
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Thompson
Nlaka’pamux, Nɫeʔkepmxcín, NtlakapSalishan (Interior Salish, muk, Thompson River Salish, Thompson Northern) Salish; Varieties: Lytton/Canyon, Nicola Valley; Varieties: Western, Eastern
Thompson River
Thompson, Thompson Salish
Salishan (Interior Salish, Northern)
Thompson Salish
Nlaka’pamux, Ntlakapmuk, Thompson, Thompson River Salish, Nɬeʔkepmxcín
Salishan (Interior Salish, Northern)
Tiipay
Sometimes used as a cover term for any and all Kumeyaay languages other than ’Iipay; see Kumeyaay. See also Jamul Tiipay
Tipai
See: Tiipay, La Huerta, Northeastern Kumeyaay, San José de Tecate
Tillamook
Hutyéyu
Timucua
Language family (branch)
Salishan Timucuan
Tiwa, Picuris Northern
See: Picuris Northern Tiwa
Tiwa, Southern
See: Southern Tiwa
Tiwa, Taos Northern
See: Taos Northern Tiwa
Tlingit
Tlingít, Lingít, Łingít, Lingit, Lingít yoo x̲’atángi
Na-Dene
Tłı̨ chǫ
Tłı̨ chǫ Yatıì, Dogrib, Thlingcha-dinneh
Na-Dene (Dene, Northern)
Tohono O’odham
O’odham; Papago-Pima, Papago
Uto-Aztecan (Piman)
Tolkapaya Yavapai
Tolkepaya, Upland Yuman, Western Yavapai
Yuman
Tongva
Gabrielino, Gabrieleño
Uto-Aztecan (Takic)
Tonkawa
Isolate
Towa
Varieties: Jemez, Pecos
Kiowa-Tanoan
Tsek’ene
Sekani; varieties include Kwadacha Tsek’ene, Fort Ware Sekani
Na-Dene (Dene, Northern)
Tse’khene
Tsek’ene, Sekani
Na-Dene (Dene, Northern)
Trader Navajo Tsilhq’otin
Pidgin Chilcotin, Tsinlhqut’in, Tsilhqut’in, Tsilhqút’ín
Tsimshian Tsimshian, Coast
Na-Dene (Dene, Northern) Tsimshianic family
See: Coast Tsimshian
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Tsimshian, Southern
See: Southern Tsimshian
Tsúut’ínà
Tsúūt’ínà, Tsuut’ina, Sarsi, Sarcee
XXXVII
Language family (branch)
Na-Dene (Dene, Northern)
Tubar
Uto-Aztecan (Southern, Tarahumaran)
Tübatulabal
Uto-Aztecan
Tümpisa
Timbisha, Timpisha, Tumbisha, Panamint
Uto-Aztecan (Numic)
Tumpisa Shoshone¨
Timbisha, Panamint, Koso
Uto-Aztecan (Numic)
Tunica
Tanico, Tonica, Tunihka, Tunixka, Yoroniku, Luhchi Yoroni, Yuron
Isolate
Tunumiisut
Inuit-Yupik-Unangan
Tuscarora
Ske:rù:ręˀ
Iroquoian (Northern)
Tutchone, Northern
See: Northern Tutchone
Tutchone, Southern
See: Southern Tutchone
Tutelo
Yesanechi
Siouan-Catawban (Siouan, Southeastern/Ohio Valley)
Tututni
Variety of Nuu-wee-ya’
Na-Dene (Dene, Pacific Coast)
Twana
sqʷuqʷúʔbəšq, Tuwaduq; Varieties: Quilcene, Skokomish
Salishan (Coast Salish)
Twulshootseed
Dxʷləšucid, Xʷəlšucid, Lushootseed, Puget Salish, Puget Sound Salish, Skagit-Nisqually
Salishan (Coast Salish)
Ucwalmícwts
Lil’wat7úl
Salishan
Unami
Delaware
Algic (Algonquian)
Unangam Tunuu
Aleut
Inuit-Yupik-Unangan
Upper Chehalis
Q̉ʷay̓áyiłq̉; Varieties: Satsop, Oakville Chehalis, Tenino Chehalis
Salishan (Tsamosan Salish)
Upper Chinook
Kiksht, Columbia Chinook; dialects: Multnomah, Clackamas, Cascades, White Salmon, Hood River, Wasco-Wishram
Chinookan
Upper Tanana
Nee’aaneegn’, Nee’aanèegn’, Nee’aanèègn’
Na-Dene (Dene, Northern)
Upriver Halkomelem
Halq̓eméylem, Holkomelem, Halkomelem
Salishan (Coast Salish)
Ute
Southern Paiute
Uto-Aztecan (Numic)
XXXVIII
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Ute, Southern
See: Southern Ute
Utian
Miwok-Costanoan
Utian family
Uto-Aztecan
Uto-Aztekan
Uto-Aztecan family
Uumarmiut
Inuit-Yupik-Unangan
Ventureño
Chumashan (Southern, Central)
Wáˑšiw
See: Washo
Waiilatpuan Wailaki
Hypothetical Cayuse-Molala family Eel River Athabaskan
Wakashan
Na-Dene (Dene, Pacific Coast) Wakashan family
Walapai
See: Hualapai
Wampanoag
Wôpanâak, Massachusett
Wappo
Algic (Algonquian) Yukian (Yuki-Wappo)
Wasco-Wishram
Dialects of Upper Chinook, Wasco, Wishram
Chinookan
Washo
Washoe; wá꞉šiw ʔítlu, Wašišiw, Wa:šiw, Wáˑšiw
Isolate
Wendat
Huron
Iroquoian (Northern)
Wenro
Neutral
Iroquoian (Northern)
West Greenlandic
Kalaallisut
Inuit-Yupik-Unangan
Western Abenaki
Algic (Algonquian)
Western Apache
San Carlos, San Carlos Apache, Cibecu, White River Apache, Ndee, Coyotero, Nnee biyáti’
Na-Dene (Dene, Apachean)
Western Cherokee
ᏣᎳᎩ ᎦᏬᏂᎯᏍᏗ, Tsalagi Gawonihisdi, dialect of Cherokee
Iroquoian (Southern)
Western Mono
Monache
Uto-Aztecan (Numic)
Western Yavapai
See: Tolkapaya Yavapai
Wichita
Kitikiti’sh, Kirikirish, Kirikir’is; Varieties: Tawakoni, Waco
Wikchamni Yokuts Winnebago
Yokutsan Ho-Chunk
Wintu Wintuan
Caddoan
Siouan-Catawban (Siouan) Wintuan (Northern)
Wintun
Wintuan family
List of North American families, languages, and dialects
XXXIX
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Wipuhk’a’bah
Wipukpaya, Verde Valley Yavapai
Yuman
Witsuwit’en
Nedut’en, Babine, Babine Carrier, North- Na-Dene (Dene, Northern) ern Carrier
Wiyot
Wishosk
Algic
Woccon
Siouan-Catawban (Catawban)
Woods Cree
Algic (Algonquian)
Wôpanâôt8âôk
Wampanoag, Wôpanâak
Wyandot
Algic (Algonquian) Iroquoian (Northern)
X̄a’islak̓ala (Kitimaat)-X̄enaksialak̓ala (Kitlope)
Haisla, X̣aʔislak’ala, Kitimat, Kitlope, Northern Kwakiutl
Wakashan (North Wakashan/ Kwakiutlan)
X̲aad Kil
Masset Haida, X̲aad kíl, dialect of Haida
Isolate
xʷməθkʷəy̓əm
Musqueam, Halkomelem
Salishan (Coast Salish)
Yakima Ichishkíin
Yakama, Yakima Sahaptin
Sahaptian
Yakima Sahaptin
Yakima Ichishkiin
Sahaptian
Yana
Varieties: Northern Yana, Central Yana, Southern Yana
Isolate
Yaqui
Hiaki, Yoeme, Yoem Noki
Uto-Aztecan
Yavapai
Varieties: Kwevikopaya (Southeastern Yavapai), Prescott (Northeastern Yavapai, Yavpé); Wipuhk’a’bah (Verde Valley Yavapai); Tolkapaya (Western
Yuman
Yawelmani
Yowlumne, Yauelmani
Yokutsan
Yoeme
Uto-Aztecan (Yoemian/Cahitan)
Yokuts
Yokutsan; Varieties: Poso Creek, General Yokutsan family Yokuts (Buena Vista and Nim-Yokuts)
Yokuts, Wikchamni
See: Wikchamni Yokuts
Yokutsan
Yokuts; Varieties: Poso Creek, General Yokuts (Buena Vista and Nim-Yokuts)
Yokutsan family
Yuchi
Euchee
Isolate
Yuki
Ukiah, Ukomno’m; Other varieties: Huchnom, Coast Yuki
Yukian/Yuki-Wappo)
Yuma
See: Kwatsáan
Yuman
Yuman Yuman, Northern
See: Northern Yuman
Yup’ik, Central Alaskan
See: Central Alaskan Yup’ik
XL
List of North American families, languages, and dialects
Families, languages, dialects
Alternate names and spellings
Language family (branch)
Yup’ik
Yupik, Yupiik, Yupiaq, Yupiat, Yugcestun, Inuit-Yupik-Unangan Yugtun, Central Alaskan, Central Alaskan Yup’ik
Yupik, Central Siberian
See: Central Siberian Yupik
Yupik, Siberian
See: Siberian Yupik
Yurok
Chillula, Mita, Pekwan, Rikwa, Sugon, Weitspek, Weitspekan
Algic
Zuni
Zuñi, Shiwi’ma, Ashiwi, Shiwi, Zuñi
Isolate
ʔayʔaǰuθəm
Comox-Sliammon, Comox, Island Comox, Mainland Comox, Tla A’min (Homalco/Xwemalhkwu–Klahoose– Sliammon/Tla A-min)
Salishan (Coast Salish)
Maps Introduction The following maps serve to illustrate the approximate territories of the languages and language families covered in this handbook. The maps stem from the Handbook of North American Indians, a series of handbooks published by the Smithsonian Institution over a thirty-year period in the late twentieth century.1 The maps are not intended as an authoritative representation of the territories, but rather as a general illustrative addition to the volume.2 Mapping in the Handbook of North American Indians began in 1978 and was based on the map by anthropologist Harold E. Driver with changes proposed by editors and chapter authors in the various volumes. The maps may cover different time periods (Ives Goddard, p.c.). While in the original source the maps were labeled “Key to Tribal Territories”3 and later “Key to Chapter Coverage”, the labels have been removed here for reprinting. The names of the languages and language groups shown on the maps have not been changed from the original source for reprinting in this volume, and they may not always represent current language names. The chapters covering language sketches and the list of languages included in this handbook can be consulted for more recently updated language names. The sequence of the maps included here follows the sequence of the Handbook of North American Indians series. It is arranged by culture and geographic area starting in the Arctic and moving south and then east. Maps from volumes 5-15 are included. In addition, an enlarged version of the whole continent map is included. It is a revised version of the map published in volume 17 of the Handbook of North American Indians and by the University of Nebraska Press. This map is included separately as a fold-out at the end of book and online here: https://www.degruyter.com/document/isbn/9783110597981/html
Acknowledgements We would like to extend our deepest gratitude to Laurie Burgess at the Department of Anthropology at the Smithsonian Institution for granting us permission to re-print the maps in this handbook. We are also eternally grateful to Ives Goddard for his helpful 1 Some maps that were originally produced decades ago have been redrafted, as noted by cartographer Dan Cole (p.c.). 2 The disclaimer published in the handbook volume 14 (page viii) can be applied more generally to all maps. 3 As noted by Ives Goddard (p.c.), the label “Key to Tribal Territories” was misleading. https://doi.org/10.1515/9783110600926-203
XLII
Maps
insights and guidance in obtaining permissions and for providing us with invaluable background information about the maps, which is included in the introduction and notes accompanying the maps. We would also like to thank the cartographer Dan Cole for graciously providing the digital files of the maps for re-printing.
References Native Languages and Language Families of North America. Compiled by Ives Goddard. Revised and enlarged edition, with additions and corrections, 1999. Department of Anthropology, National Museum of Natural History, Smithsonian Institution. Native Languages and Language Families of North America. Compiled by Ives Goddard. 1999. Studies in the Anthropology of North American Indians Series. Sheet map, folded. University of Nebraska Press. Sturtevant, William C. ed. 1971. Handbook of North American Indians: Arctic. Volume 5. David Damas, volume editor. Washington, DC: Smithsonian Institution. Sturtevant, William C. ed. 1982. Handbook of North American Indians: Subarctic. Volume 6. June Helm, volume editor. Washington, DC: Smithsonian Institution. Sturtevant, William C. ed. 1990. Handbook of North American Indians: Northwest Coast. Volume 7. Wayne Suttles, volume editor. Washington, DC: Smithsonian Institution. Sturtevant, William C. ed. 1978. Handbook of North American Indians: California. Volume 8. Robert F. Heizer, volume editor. Washington, DC: Smithsonian Institution. Sturtevant, William C. ed. 1979. Handbook of North American Indians: Southwest. Volume 9. Alfonso Ortiz, volume editor. Washington, DC: Smithsonian Institution. Sturtevant, William C. ed. 1983. Handbook of North American Indians: Southwest. Volume 10. Alfonso Ortiz, volume editor. Washington, DC: Smithsonian Institution. Sturtevant, William C. ed. 1986. Handbook of North American Indians: Great Basin. Volume 11. Warren L. D’Azevedo, volume editor. Washington, DC: Smithsonian Institution. Sturtevant, William C. ed. 1998. Handbook of North American Indians: Plateau. Volume 12. Jr. Deward E. Walker, volume editor. Washington, DC: Smithsonian Institution. Sturtevant, William C. ed. 2001. Handbook of North American Indians: Plains. Volume 13. Raymond J. DeMallie, volume editor. Washington, DC: Smithsonian Institution. Sturtevant, William C. ed. 2004. Handbook of North American Indians: Southeast. Volume 14. Raymond Fogelson, volume editor. Washington, DC: Smithsonian Institution. Sturtevant, William C. ed. 1979. Handbook of North American Indians: Northeast. Volume 15. Bruce G. Trigger, volume editor. Washington, DC: Smithsonian Institution. Sturtevant, William C. ed. 1996. Handbook of North American Indians: Languages. Volume 17. Ives Goddard, volume editor. Washington, DC: Smithsonian Institution.
Maps
XLIII
Map of the languages and major dialects in the Arctic culture area
Source: Department of Anthropology, National Museum of Natural History, Smithsonian Institution. Published in the Handbook of North American Indians: Arctic. Volume 5: ix. [Note: Some language names have been updated since originally published in the Handbook.]
XLIV
Maps
Map of the languages and dialects in the Subarctic culture area
Source: Department of Anthropology, National Museum of Natural History, Smithsonian Institution. Published in the Handbook of North American Indians: Subarctic. Volume 6: ix.
Maps
XLV
Map of the languages and language groups on the Northwest Coast To the left: North.
To the right: South.
Source: Department of Anthropology, National Museum of Natural History, Smithsonian Institution. Published in the Handbook of North American Indians: Northwest Coast. Volume 7: ix.
XLVI
Maps
Map of the languages and language groups in California
Source: Department of Anthropology, National Museum of Natural History, Smithsonian Institution. P ublished in the Handbook of North American Indians: California. Volume 8: ix.
Maps
XLVII
Map of the languages and language groups in the Southwest
Source: Department of Anthropology, National Museum of Natural History, Smithsonian Institution. Published in the Handbook of North American Indians: Southwest. Volumes 9 and 10: ix. [Note: There are some updates on the map since it was originally published in the handbook.]
XLVIII
Maps
Map of the languages in the Great Basin
Source: Department of Anthropology, National Museum of Natural History, Smithsonian Institution. Published in the Handbook of North American Indians: Great Basin. Volume 11: ix.
Maps
XLIX
Map of the languages and language groupings in the Plateau
Source: Department of Anthropology, National Museum of Natural History, Smithsonian Institution. Published in the Handbook of North American Indians: Plateau. Volume 12: ix.
L
Maps
Map of the languages, dialects, and language groupings in the Plains
Source: Department of Anthropology, National Museum of Natural History, Smithsonian Institution. Published in the Handbook of North American Indians: Plains. Volume 13.1 and 13.2: ix.
Maps
LI
Map of the languages, dialects, and regional groupings in the Southeast
Source: Department of Anthropology, National Museum of Natural History, Smithsonian Institution. Published in the Handbook of North American Indians: Southeast. Volume 14: ix.
LII
Maps
Map of the languages, dialects, and regional groupings in the Northeast
Source: Department of Anthropology, National Museum of Natural History, Smithsonian Institution. Published in the Handbook of North American Indians: Northeast. Volume 15: ix.
I Sounds and sound structure
Sonya Bird, Rae Anne Claxton, and Tess Nolan
1 Acoustic phonetics
Abstract: Acoustic phonetics is the subfield of linguistics dedicated to studying the details of speech sounds, by examining their sound waves. Although speech sounds have been studied for over 100 years, there are not many studies of North American Indigenous languages. In this chapter, we describe acoustic phonetics as it relates to Indigenous language revitalization and creating new speakers (Section 1.1). We discuss how to use existing and new recordings to document the acoustic features of speech (Section 1.2). We summarize acoustic phonetic work that exists on North American Indigenous languages, including studies on vowels, consonants, word-level prosody (the rhythm of individual words), and sentence-level intonation (how whole sentences sound) (Section 1.3). We end by discussing how people are coming together for community-based acoustic phonetic research on Indigenous languages, and how it is a key component of language documentation and revitalization (Section 1.4).
1.1 Introduction At first glance, languages around the world share many of the same sounds. For example, English (Germanic) and Hul’q’umi’num’ (Salish) both have a word written ten (‘mother’ in Hul’q’umi’num’). If we listen carefully though, we can notice that even though the spelling is the same, the sounds are slightly different: 1 is longer and more tense in Hul’q’umi’num’ than in English (spelled [e] vs. [ɛ] in the International Phonetic Alphabet, see Figure 2). Of course, languages also have sounds that are entirely different from those used in other languages, both in spelling and in pronunciation. For example, the sound spelled in the Hul’q’umi’num’ word ([t͡ɬʼam]; ‘to be enough’) does not exist in English. Unfamiliar sounds can be challenging for new speakers, especially if they do not know what exactly to pay attention to in trying to hear and pronounce them. Documenting a language’s sounds using acoustic phonetics and articulatory phonetics (see Bliss, Bird, and Gick, this volume) can help us understand the details of these sounds, and this can help us support new speakers in their language work. Phonetics is the subfield of linguistics that is focused on documenting and comparing the details of speech sounds, both within a language and across the languages of the world. Whereas articulatory phonetics (Bliss, Bird, and Gick, this volume) focuses on the actual articulation of speech, acoustic phonetics focuses on the sound waves that result from articulation. Both help us further our understanding of the human capacity for
1 Linguists use at least three sets of brackets: are used for spelling; /../ are used for sounds as they are encoded in our minds (see Chapter 5); [..] are used for sounds as they are actually pronounced. https://doi.org/10.1515/9783110600926-001
4
Sonya Bird, Rae Anne Claxton, and Tess Nolan
speech – what sounds we are able to make and how we use them to express ourselves. For Indigenous language documentation and revitalization, acoustic and articulatory phonetic research also creates a foundation for developing teaching materials that can help learners achieve what they think of as ‘authentic’ pronunciation. Many learners of Indigenous languages take very seriously their responsibility of speaking in a way that honors their elders (Bird and Kell 2017; Jenni et al. 2017; Bird and Miyashita 2019), but they are often lacking in resources to help them (McIvor 2015). Relatively few opportunities exist to interact with first language (L1, mother tongue) speakers and popular teaching approaches like Total Physical Response (Asher 1977) and Where Are Your Keys (Gardner 2019) tend to de-emphasize pronunciation. Analyzing the details of speech can be very technical work (see Sections 1.2.3 and 1.3) and can never replace working face-to-face with speakers. Nonetheless, it can play a valuable role in language learning, especially for learners at the intermediate to advanced level who are ready to fine-tune their speaking and listening skills. It allows us to “zoom in” on very specific aspects of speaking and listening, in the comfort of our own homes, in a space where we can really examine the details of speech without feeling like we’re putting ourselves and other speakers on the spot or making anyone repeat things over and over again. The examples provided throughout this chapter are transcribed using the International Phonetic Alphabet (IPA). The IPA was created by linguists to transcribe all the sounds in the world’s languages. Figures 1 and 2 summarize the symbols used for the most common consonants (Figure 1) and vowels (Figure 2) across languages. Both charts are organized based on articulation, and are described more fully in Bliss, Bird, and Gick (this volume). In the consonant chart, the columns correspond to where the sound is produced in the mouth/throat (place of articulation), from the lips (at the left) to the larynx (at the right). The rows correspond to how constricted the sound is (manner of articulation), from completely constricted stops (no air gets through, e. g., [p]) to unconstricted resonants (air flows freely, e. g., [l]). Within each cell, the two symbols represent sounds that differ only in what the vocal folds are doing; the sound on the left (e. g., [p]) is pronounced without vocal fold vibration (voiceless), while the sound on the right (e. g., [b]) is pronounced with vocal fold vibration (voiced). In the vowel chart, vowels are organized based on where the tongue is within the mouth, along two dimensions: frontback (towards the front of the mouth vs. towards the back of the mouth) and close-open (high to the palate vs. low on the mouth floor). Additional charts and lists of other IPA symbols and diacritics can be found on the website of the International Phonetic Association: (https://www.internationalphoneticassociation.org/IPAcharts/IPA_charts_2018. html). While the IPA is the standard alphabet used by the international community of linguists, in North America – at least in the Pacific Northwest – it is more common to use the North American Phonetic Alphabet (NAPA). The alphabets are very similar, with the exception of a few symbols (NAPA = IPA): x̣ = χ; ḥ = ħ; ƛʼ = t͡ɬʼ; s, c (or š, č) = ʃ, t͡ʃ; y = j; ṣ, c̣, ḷ, ị, ụ, ə̣, ạ = s̠ , c̠ , l̠ , i̠ , u̠ , ə̠, a̠ .
5
Acoustic phonetics
THE INTERNATIONAL PHONETIC ALPHABET (revised to 2015)
PHABET (revised to 2015)
CONSONANTS (PULMONIC)
Bilabial Labiodental
Plosive
Postalveolar Retroflex
Nasal
Palatal
Velar
Dental
Uvular
© 2015 IPA
Alveolar Postalveolar Retroflex
© 2015 IPA
Pharyngeal
Palatal
Velar
Uvular
Pharyngeal
Glottal
Glottal
Trill Tap or Flap Fricative Lateral fricative
Approximant Lateral approximant
Symbols to the right in a cell are voiced, to the left are voiceless. Shaded areas denote articulations judged impossible.
CONSONANTS (NON-PULMONIC)
VOWELS
Fig. 1: Basic (pulmonic) consonant chart (Source: https://www.internationalphoneticassociation.org/ Front Central Clicks Voiced implosives Ejectives content/ipa-chart, available under a Creative Commons Attribution-Sharealike 3.0 Unported License. Close Source: Copyright 2018 International Phonetic Bilabial Bilabial Examples: Association)
Back
left are voiceless. Shaded areas denote articulations judged impossible. Dental
(Post)alveolar VOWELS
s
Palatoalveolar Front
Dental/alveolar
Bilabial
Palatal
Dental/alveolar
Velar
Alveolar lateral Close
Central
Uvular
Close-mid
Back
Velar Alveolar fricative
OTHER SYMBOLS
eolar
icative
Voiceless labial-velar fricative Close-mid
Voiced alveolar lateral flap
Voiced labial-palatal approximant
Simultaneous
Open-mid Voiceless epiglottal fricative Voiced epiglottal fricative Epiglottal plosive
atives
ral flap
nd
ulations ymbols ary.
Open
Open
Alveolo-palatal fricatives
Voiced labial-velar approximant
Open-mid
SUPRASEGMENTALS
and
Primary stress
Affricates and double articulations can be represented by two symbols joined by a tie bar if necessary.
Secondary stress Long
DIACRITICS Some diacritics may be placed above a symbol with a descender, e.g. Voiceless
Where symbols appear in pairs, the one to the right represents a rounded vowel.
Where symbols appear in pairs, the one to the right represents Breathy voiced a rounded vowel. Dental
Voiced
Creaky voiced
Apical
Less rounded
Secondary stress Palatalized
Nasal release
Half-long Extra-short Minor (foot) group
Major (intonation) group SUPRASEGMENTALS Fig. 2: Vowel chart (Source: www.internationalphoneticassociation.org/content/ipa-chart, available under Aspirated Linguolabial Laminal Syllable a Creative Commons Attribution-Sharealike 2018break International Primary stress 3.0 Unported License. Source: Copyright More rounded Labialized Nasalized Linking (absence of a break) Phonetic Association) Advanced
Long Velarized
Lateral release
TONES AND WORD ACCENTS
Many Indigenous language communities in North America use LEVEL an Extra alphabetCONTOUR that is or or Rising Retracted Pharyngealized No audible release Half-long high adapted from either the Roman alphabet, the IPA, or the NAPA. Being able to read difol with a descender, e.g. High Falling Centralized Velarized or pharyngealized Extra-short ferent alphabets makes it much easier to access materials put togetherMid by differentHigh linDental rising Mid-centralized Raised ( = voiced alveolar fricative) Lowto Minor (foot)on group guists over time. In our work Salish languages, we have found that it is useful Low rising Syllabic Lowered ( = voiced bilabial approximant) Apical Extra Risingcreate conversion charts that includegroup IPA, NAPA, and the local writing system(s), so that Major (intonation) low falling Non-syllabic Advanced Tongue Root Downstep Global rise Laminal we have a handy reference to help decode various written materials. As an example, Syllable break Rhoticity Retracted Tongue Root Upstep Global fall Nasalized Table 1 provides a small section of a conversion chart that includes the alphabets for two Linking (absence of a break) neighboring Salish languages spoken on Southern Vancouver Island (Hul’q’umi’num’ Typefaces: Doulos SIL (metatext); Doulos SIL, IPA Kiel, IPA LS Uni (symbols) Nasal release TONES AND WORD ACCENTS (HUL) and SENĆOŦEN (DEA), NAPA, and IPA). The SENĆOŦEN orthography was created Lateral release LEVEL CONTOUR by the late Dave Elliott, W̱SÁNEĆ elder. DEA stands for Dave Elliott Alphabet; it uses all Extra or of four or Rising No audiblecapital release letters and a set high diacritics. High
= voiced alveolar fricative) = voiced bilabial approximant)
Mid Low
Extra low
Downstep
Falling
High rising Low rising Risingfalling
Global rise
6
Sonya Bird, Rae Anne Claxton, and Tess Nolan
Tab. 1: Sample conversion chart: Hul’q’umi’num’(HUL), SENĆOŦEN (DEA), NAPA, IPA HUL
DEA
NAPA
IPA
o u p’ lh ch kw xw
O E B Ƚ Ć Ȼ X̱
a ə p̓ ɫ č kʷ x̣ʷ
ɑ ə p’ ɬ t͡ʃ kʷ χʷ
The remainder of this chapter is structured as follows: we start by presenting methodological considerations in using and creating recordings that form the basis of acoustic analysis, how to make use of existing recordings, what to consider when making new recordings, and how to process and analyze these to learn about pronunciation (Section 1.2). We summarize acoustic phonetic work that has focused on North American Indigenous languages, focusing on some of the more unusual features of these languages (Section 1.3). We end the chapter with a brief discussion of new ways of conducting collaborative (university-community) acoustic phonetic research (Section 1.4).
1.2 Methods and materials produced 1.2.1 Using existing recordings and making new ones Our detailed understanding of speech sounds (reflected in the charts above) comes from analyzing speakers’ articulations (articulatory phonetics) and the resulting speech sound waves (acoustic phonetics), in addition to careful listening of course (auditory phonetics). Some sounds are easier to grasp referring directly to articulation, e. g., ones with complicated tongue positions like English [r] and [l]; others are easier to grasp referring to the acoustic signal (the sound waves), e. g., ones with a tightening in the throat like ejective [t’] (see Section 1.3.2.2). Combining auditory, acoustic, and articulatory phonetics gives us the best chance to document speech in a thorough, comprehensive way. In this chapter, we focus on acoustic phonetics (Bliss, Bird, and Gick, this volume, focuses on articulatory phonetics). Studying speech sound waves involves using specialized software (see below) to analyze audio recordings. These can either be existing, for example legacy recordings made with previous generations of speakers, or they can be made new. Existing recordings are immensely valuable for studying speech, especially in cases where there are currently no first language speakers. The earliest recordings of Indigenous
Acoustic phonetics
7
languages of North America were made in the late 1800s using wax cylinders (https:// www.historymuseum.ca/cmc/exhibitions/tresors/barbeau/mbf0300e.html). Since then, recording technology has evolved substantially, from magnetic tape recordings (reelto-reel and cassette) to digital media (DAT, minidisc, CD, DVD). Nowadays, recordings normally consist of digital audio files (WAV, MP3 and other formats), created and saved directly on hard drives and/or servers. The key to preserving and using recordings made on older media is digitization. Older media can easily become unusable, either because they deteriorate over time or because the equipment required to play back the recordings is no longer available. Many institutions have large collections of older recordings. For example, comprehensive collections of early wax cylinder recordings are housed at the Library of Congress in the United States and the Canadian Museum of Civilization in Canada. Other collections are housed more locally, e. g., the California Language Archive (http://cla.berkeley. edu/) contains a large collection of early recordings of Californian languages, thanks to a partnership with the IRENE project at the University of California, Berkeley (http:// irene.lbl.gov/). In addition, many universities, libraries and museums have the necessary equipment and capacity to digitize and archive recordings made using media that is now outdated. Partnerships between communities and these institutions can facilitate storing and accessing audio collections, and various funding opportunities exist to support such partnerships2. Existing recordings can provide us with wonderful breadth in terms of content, especially if they are of naturalistic speech (e. g., stories, conversations), allowing for fruitful exploration of all kinds of speech features. Working with these recordings can be relatively time-consuming (although thoroughly engaging!) in that it is often necessary to listen to (and transcribe) hours of recordings to hear the specific speech feature you are interested in. Where possible, working on existing recordings with fluent speakers (often elders) is very valuable; they can provide context and translations for what was recorded, and can also notice things that newer speakers may not notice about the speech styles of individual speakers, for example. In some cases, existing recordings may simply not contain enough examples to allow for an in-depth study of something you are interested in. This is where new recordings are useful: because they can be designed with a specific focus in mind, they allow for in-depth study (the trade-off being that we may miss important patterns by intentionally eliciting speech rather than letting it emerge naturally). Another advantage in making
2 Funding opportunities change over time, so research on the possibilities must be done at the time communities are ready to prepare an application. In 2021, some national-level possibilities include the Phillips Fund (American Philosophical Society), Jacobs Fund (Whatcom Museum), Social Sciences and Humanities Research Council of Canada (SSHRC; in Canada), and Dynamic Language Infrastructure/Documenting Endangered Languages (DLI/DEL) program (jointly funded by the National Science Foundation [NSF] and the National Endowment for the Humanities [NEH] in the US). Other funding opportunities may exist within specific institutions.
8
Sonya Bird, Rae Anne Claxton, and Tess Nolan
new recordings is that they provide the opportunity to work with fluent speakers, which can be a very rich experience. For example, in recording the details of pronunciation with elders, learners often have the opportunity to get valuable feedback on their own pronunciation. New recordings can take some time and effort to set up in terms of tailoring them to elicit specific aspects of speech but, once they are made, processing the recordings for analysis is relatively quick. In some contexts, making new recordings is not feasible (e. g., if there are currently no speakers of the language). Recording quality may also be a determining factor in choosing whether to work with existing or new recordings: if existing recordings were made on older media and/or were not made with acoustic phonetic research in mind, they may not be of high enough quality for acoustic analysis. Ultimately, the choice of whether to use existing or new recordings will differ depending on the research topic and on the feasibility of accessing or creating different types of recordings. In general, using a combination of the two types of recording can be very successful: existing recordings for breadth and new recordings for depth. One thing that we have learnt from our work together is that it is important not to underestimate how much time and effort it takes to review and process recordings (whatever the medium/media) and to get them in a format that is usable for different purposes. If possible, we recommend including someone in your team who is tech-savvy and can help to turn recordings into resources for community-based language learning, e. g., an app for listening to the language on your phone.
1.2.2 Methods for making new recordings 1.2.2.1 Recording procedures 1.2.2.1.1 Recording devices New recordings can be made using a range of devices. The Zoom H4N is currently (2020) a relatively inexpensive and high quality recorder (https://www.zoom-na.com/), and is very popular in linguistic documentation work. Recordings can also be made directly onto desktop or laptop computers, ideally using an external microphone (examples include the Rode NT-1 A, NT-USB, Sennheiser MKH 416 P48U3, or the Blue Yeti USB microphone) and free recording software like Audacity (https://www.audacityteam. org/). Nowadays, recordings can even be made on handheld devices. Paul De Decker and Jennifer Nycz (2011) compared the quality of sound recordings made on four devices: a professional Roland Edirol R-09 recorder, a first-generation iPhone, a Macbook Pro, and a Mino Video recorder, both original and uploaded to and then downloaded from YouTube. Compared to the baseline of the Edirol recorder, the iPhone and Macbook Pro were able to accurately record the lower frequencies (in vowels, the first two formants – see Section 1.3.2.1), but beyond that became unreliable. De Decker and Nycz found that the Mino Video recorder and the YouTube videos significantly altered and compressed
Acoustic phonetics
9
the audio of the recordings such that they were considerably warped and therefore unusable for acoustic analysis. In our opinion, even though microphones in non-dedicated consumer electronics have improved in recent years, it is not recommended to use them for significant, faithful, speech recording. In making recordings, the main factor to consider is what is called the signal to noise ratio (Ladefoged 2003): the difference in loudness between the sound you want to record (signal) and everything else (noise, including clocks, refrigerators, TVs, lawnmowers, etc.). Acoustic analysis requires audio recordings of relatively high quality, meaning a relatively loud signal compared to whatever background noise might be present. In a very quiet room and with talkers with relatively strong (loud) voices, it may be possible to get by without an external microphone, for example using the built-in microphone of the Zoom H4N. In a louder setting and/or with quieter speakers, or in cases where we want to record multiple speakers, one or more external microphones may be needed, and careful thought must go into choosing the appropriate one(s). 1.2.2.1.2 Microphones There are two main features to consider when it comes to microphones for recording speech: microphone type and polarity. Type refers to the way the microphone takes the physical property of sound (changes in air pressure) and converts it to electronic signals. There are a number of different types of microphones; the best type for speech is a condenser microphone. More specifically, small-diaphragm condenser microphones are better than large-diaphragm ones for their increased and more life-like responsiveness to speech sounds. These microphones (also called capacitor microphones) are good at capturing the more subtle acoustic properties of speech, but they do require an external power source. This power source (called “phantom power”3) will usually come from (a) the recorder (where it can be turned on in the settings, as is the case with Zoom recorders), (b) a powered pre-amp or computer (usually turned on with a switch), or (c) a battery pack included with the microphone itself. The second feature is microphone polarity. Polarity refers to the microphone’s sensitivity to sounds coming from different directions. Different polarities have different advantages and disadvantages, and different applications. For example, an omnidirectional microphone will record with equal loudness sounds from all the way around it (360 degrees). This can be useful for group conversations where there are not enough microphones for each individual but can be detrimental in noisy situations where just one person is being recorded. A cardioid polarity microphone is highly sensitive to sounds directly in front of the microphone, and very insensitive to sounds elsewhere. This makes it great for recording a single speaker while minimizing other sounds in the
3 On recorders and pre-amps phantom power can be labeled either “phantom power” or “48V” for the 48 volts that it provides.
10
Sonya Bird, Rae Anne Claxton, and Tess Nolan
environment (the signal to noise ratio mentioned above), but it is not so good if you are also interested in other speakers or ambient sound. If recording multiple speakers, it is best to have several cardioid microphones, one for each speaker (or group). Some microphones can be switched between different polarities, and this will often be indicated as a “multi-pattern” microphone. High quality microphones usually connect to a recorder or pre-amp with a special cable called an XLR cable. For example, the Zoom H4N mentioned above has two XLR ports for cables to connect to external microphones. Likewise, the Focusrite Scarlett 2i2 pre-amp has two ports for microphones. Some microphones (such as the Rode NT-USB or the Blue Yeti) connect directly to the computer via USB (which should be indicated in their title). Recorders and recording software allow various methods for recording audio from microphones into different channels. The most common are mono and stereo. In mono, all audio from all microphones is recorded to one channel; in stereo, the audio from at least two microphones is recorded into two channels. Some recorders, such as the Zoom H6N, allow multi-channel recording with up to four external microphones. Audio information from one microphone in one channel can be separated from audio information in another channel, allowing separate analysis if needed. The simplest recorders, with just one built-in microphone, allow (effectively) only mono recordings. The more microphones and microphone ports available on a recorder or a pre-amp, the greater the number of channels there are available for recording. This can often be reflected in the product name for a recorder, for example the Zoom H4N has (potentially) four microphones, two built-in and two ports for external microphones, while the Zoom H6N has (potentially) six microphones, two built-in and four ports for external microphones. These devices (or ones like them) are ideal for recording a group of elders in conversation, for example. Note that if you are recording to multiple devices rather than through multiple microphones on a single device, you will have to synchronize the recordings after the fact, for processing. This can be done by creating a loud sound, such as a clap, that appears at the beginning on all recordings. The recordings can then be synched later, by lining up the loud sound. 1.2.2.1.3 Audio file types Recorders often save recorded audio to a removable SD card or some other on-board storage, which can then be transferred to a computer. There is no quality difference between these storage types, though there can be quantity differences in terms of amount of storage available. The two most common audio file types found on portable recorders or accessible in recording software are uncompressed and compressed types. Uncompressed audio means that, within the given parameters for recording, all the acoustic information picked up by the microphone will be faithfully encoded into a digital file and be available for later analysis. No data or information is lost through the encoding of sound into digital information. This type commonly includes the WAV and AIFF file
Acoustic phonetics
11
types.4 Compressed audio includes both lossless and lossy file types. Lossless formats are compressed, that is, the data they hold have been manipulated from the original signal to create a smaller file size. They rely on algorithms to do this compression in a manner that preserves the full psycho-acoustic information of the original uncompressed format (meaning the human ear cannot normally tell that the file has been compressed). Examples of common lossless formats include FLAC and ALAC. Lossy formats involve the same kind of manipulation and transformation but result in an even smaller file sizes and greater information loss. Common types of lossy audio are MP3 or AAC formats. While compression saves on file size and therefore storage space, compression also entails loss of information (regardless of whether the listener can tell or not). We strongly recommend that original audio recordings be saved and stored in an uncompressed format (not MP3), unedited, and that this uncompressed format be used for analysis – see Dry (2009) for a wider discussion on the best practices of data and materials preservation. Whenever possible, decisions about recording set-ups should be made with guidance from a sound expert who is knowledgeable in the currently available technology. 1.2.2.1.4 Adding video recording Although the focus here has been on audio recordings, our experience is that we can learn a lot by being able to see as well as hear fluent speakers. Two kinds of video recordings are useful: (1) zoomed-in videos of speakers’ faces, to see things like lip rounding (see Bliss, Bird, and Gick,this volume, Section 2.3.1) and (2) zoomed-out videos of whole speakers, to see things like hand gestures and eyebrow movements, which are also part of how we communicate. In terms of pedagogical applications, a growing body of literature in language learning and teaching points to the benefits of multimodal speech perception (Bliss et al. 2018; Olson 2014). For any new recordings, it is worth considering simultaneous audio and video recording.
1.2.2.2 Elicitation materials and techniques If new recordings are being made, there are several options in terms of how to elicit speech. For basic information on the pronunciation of specific sounds, word list elicitations are useful. For example, a speaker might be asked to read a list of words, each one illustrating a different sound in the language. The Illustrations of the International Phonetic Association (IPA 1999; see Section 1.3.1) are largely based on word list elicitations. These short descriptions are designed to give the reader a fairly good sense of how the individual sounds of a language are pronounced.
4 Technically WAV refers to the file format that can hold many different types of metadata and audio at different levels of compression, but in speech recording “saving to WAV” usually means saving in an uncompressed format
12
Sonya Bird, Rae Anne Claxton, and Tess Nolan
To understand how a language sounds above the level of individual words, longer speech samples are needed. For example, to compare the intonation patterns of statements (e. g., ‘he went to the store.’) and questions (e. g., ‘he went to the store?’), whole sentences need to be elicited. Often this is done using props like pictures or storyboards (Figure 3) (Caldecott and Koch 2014). The website http://totemfieldstoryboards.org/ has an ever-growing collection of storyboards that can be used for research and teaching. Note that storyboards and similar visual props do not need to be highly professional-looking or digital; hand drawn pictures on notecards (Figure 3b) work very well. Speakers like them, they can be edited on the fly as needed, and they do not rely on technology (that sometimes crashes).5
(a) (b) Fig. 3: Examples of visual props to elicit conversational speech: (a) Picture with bunny holes that are movable and can be added/taken away to elicit ordinal numbers in a story where Coyote is looking for Rabbit (artwork created by co-author Claxton for a story elicitation session with elder Dr. Sti’tum’at Ruby Peter in September 2019); (b) Notecard from a story about borrowing a car (original source, following Creative Commons license CC-BYNC-SA 2.0: Caldecott and Koch [2014, Figure 2, p. 223]).
It is worth putting thought into designing elicitation sessions so that the recordings can be used for multiple research and pedagogical purposes. This is especially important if time with fluent speakers is limited. For example, stories and conversations can be used to study all kinds of things, from the tiniest details of speech (e. g., how the high front vowel /i/ transitions into the back consonant /q/; see Section 1.3.2.1) to the broadest patterns (e. g., sentence intonation; see Section 1.3.2.3). At the same time, they also provide a valuable cultural resource for communities, which can be used in any number of ways. Along these lines, as feasible and appropriate, it is always a good idea to record whole sessions with elders (rather than just a specific task). You just never know what might come up, and what unexpected things you might learn from these sessions!
5 Marion Caldecott, personal communication (May 5, 2020).
Acoustic phonetics
13
One thing that is important to remember about focused elicitation sessions is that speakers will not always sound the same in these recordings as they do in more natural settings, when they are speaking spontaneously. This means that how learners experience spoken language based on elicited speech may not translate to how they experience it in the ‘real world’: learners may find that they understand elicited speech better than they do more natural speech; similarly, they may find that they are expected to speak in community in a way that does not match how they speak in more careful sessions with their teachers and elders. This is partly because elicitation sessions tend not to capture the living nature of spoken language, and partly because speakers (often teachers/elders) participating in elicitation sessions are often particularly careful to create clear recordings, knowing that they will be used in part for pedagogical purposes. In any case, it is important to keep in mind that while doing focused pronunciation work is one component of language learning, it will never replace learning in community, and in a more holistic way. Finally, we must always remember that any recordings we make should be made available to the speakers and their community (as appropriate) in a timely manner. For example, when learners participate in recording sessions, they have very practical goals in mind; they expect the recordings to be processed quickly, so that they can get feedback sooner rather than later on their pronunciation. It is our responsibility to make this our priority as well.
1.2.2.3 Speakers Existing acoustic phonetic studies of North American Indigenous languages are by and large based on the speech of one or two first language (L1) speakers. It can be of great value though – in terms of linguistic inquiry and in terms of pedagogical applications – to compare pronunciation across different speakers (e. g., Fortier 2019). In the SENĆOŦEN-speaking community for example, young adults joke about sounding like 80 year old men, because the legacy recordings that have helped them learn their language were mostly recordings of elderly men (Bird and Kell 2017). Making recordings with speakers of all ages and across genders, in a range of tasks and activities, will ensure that we document the language as thoroughly as possible. Documenting the pronunciation of new speakers with varying fluency levels can be particularly useful, again for theoretical and practical reasons. There is an overwhelming focus on English in the second language (L2) pronunciation literature (Lee, Jang, and Plonsky 2015). Given how complex the sound systems are in many North American Indigenous languages, looking into pronunciation among learners of these languages will undoubtedly shed important light on L2 pronunciation as a whole. From the perspective of Indigenous language revitalization, documenting the speech of new speakers will help us determine what challenges they face, so that we can create tools and techniques to help them overcome these challenges.
14
Sonya Bird, Rae Anne Claxton, and Tess Nolan
No matter who is being recorded, it is important to follow appropriate protocols and best practices during the recording session. In working with elders, it is our responsibility to take care of them while they are working. This includes knowing what they like to eat and drink, being aware of dietary restrictions (e. g., diabetes or allergies) when we provide snacks, keeping an eye on their energy levels, being flexible when it comes to taking breaks vs. pushing through on a task, and making sure they have what they need to work (elicitation materials in large font sizes, spare batteries for hearing aids, notepads and pens, etc.). Many elders like to know ahead of time what they will be focusing on in a particular session, so getting materials to them in advance is important. In working with learners, we must always keep in mind that we are putting them on the spot by focusing on their pronunciation. Documenting their speech might feel uncomfortable and even intrusive, especially if they are sensitive about sounding different from their elders and/or peers. We must always proceed carefully, mindful of how learners are feeling, and keeping in mind that any form of feedback, no matter how constructive it is, can be discouraging. This section has focused on recording speech for future analysis (see Section 1.3.3). This kind of work is beneficial in two ways: First, it allows us – especially learners – to go off and figure out the details of speech offline, without the pressure of ‘getting it right’ in front of others. Second, it takes some burden off elders and teachers, in that they do not have to keep repeating the same things over and over again, as we try to make sense of what we are hearing. In an ideal situation, recordings are made by language learners, and contribute directly and immediately to their own learning process. With this in mind, we would like to end with a word of advice on doing pronunciation work with elders: remember that you are not necessarily going to see progress overnight. Your elders and teachers might offer you pronunciation tips and strategies that you are not immediately able to make use of. Take them away and work with them; give yourself time and patience, and practice, practice, practice. At the same time, try not to be shy when you are working with your teachers and elders; your time with them is precious, and you will get the most out of it if you are willing to take chances and make mistakes. In our experience, it works well to pick one thing to focus on, and to let your mentor(s) know what it is so they can help you. They will check your pronunciation, and will help you progress in a steady, step-by-step way. For example, if you are struggling with a particular sound, they may start you off with familiar words that you are comfortable saying, and help you work towards harder words. By balancing face to face work with work based on recordings, you will be able to make the most of different resources you have to work with.
1.2.3 Analyzing audio recordings acoustically Acoustic phonetic data analysis can be quite technical, including detailed quantitative measurements of various acoustic parameters like vowel formants (Kharlamov and
Acoustic phonetics
15
Oberly 2019) and pitch contours (Martin and Johnson 2002) – see Section 1.3.2. This need not be the case though; qualitative visual inspection of the waveform and spectrogram associated with speech can be enough to clarify how speech sounds differ from one another, and this kind of work does not require a lot of technical expertise (Bird et al. 2019; Fish and Miyashita 2017). In terms of quantitative analysis, Ladefoged (2003) provides a useful summary of some of the acoustic features that are commonly measured, including pitch, loudness, and length (Ladefoged 2003: Chapter 4), vowel characteristics (Ladefoged 2003: Chapter 5), consonant characteristics (Ladefoged 2003: Chapter 6), and phonation types (Ladefoged 2003: Chapter 7). Studies illustrating some of these features and related measurements are discussed in Section 1.3.2 below. Note that, although technology-based analysis of speech can be very useful, it should not replace using our ears. Acoustic analysis provides an additional way of experiencing speech, one that can be incorporated into the process of understanding speech rather than one that replaces other approaches. In terms of acoustic data analysis tools, perhaps the most commonly used is Praat (http://www.fon.hum.uva.nl/praat/), a freely downloadable tool created by Paul Boersma and David Weenink at the University of Amsterdam (Boersma and Weenink 2022). Praat allows for visual inspection of waveforms, spectrograms, and pitch and intensity (loudness) contours of speech, as well as for taking quantitative measurements. Each (or all) of the visual displays can be paired with one or more “textgrids,” which can be used to segment and label the speech stream. Figure 4 below provides a sample waveform (upper panel), spectrogram (middle panel), and two textgrids6 (lower panel) of the word ‘acoustics’, as pronounced by Sonya Bird (co-author). Superimposed on the spectrogram are also the pitch contour (flat-ish lines) and intensity (loudness) contour (loopy lines). Such visuals can help us identify what the acoustic features of different sounds are. For example, we can see that the middle vowel ou [u] has the highest pitch and intensity (the pitch and intensity contours are higher up on the image), reflecting the fact that this vowel carries the primary, word-level stress (see 3.2.3). We can also see that the two k sounds in “acoustics” differ from one another: the first k ([kh]), which precedes the stressed vowel ou ([u]), has a lot of “aspiration” (visually: white noise) on it; this is the puff of air you can feel if you place your hand just in front of your lips and pronounce the word. The second k ([Ɂk]), which follows the unstressed vowel i ([ɪ]), looks quite different – it is not aspirated like the first one, and is preceded by spreading out and irregularity of the vertical lines – the pitch pulses – at the end of the vowel [ɪ] and before the airflow stops completely for [k]. This is called glottalization or laryngealization. Articulatorily, it corresponds to a tightening in the throat (a constriction at the glottis, more precisely), like the sound in the middle of uh-oh.7 6 Upper textgrid: English alphabet; lower textgrid: phonetic transcription using the International Phonetic Alphabet (IPA). 7 Each pitch pulse corresponds to one glottal pulse, or one open-close cycle of the vocal folds, as they vibrate to produce voiced speech sounds like vowels.
16
Sonya Bird, Rae Anne Claxton, and Tess Nolan
Fig. 4: Visual displays of the English word ‘acoustics’: top = waveform; middle = spectrogram (with superimposed pitch contour [flat-ish lines] and amplitude contour [big loopy]); bottom = textgrid for labeling sound files (upper: English orthography; lower: phonetic [IPA] transcription).
Praat also has a scripting function, which allows for automatically extracting and exporting measurements to an Excel spreadsheet for analysis. Another handy feature of Praat is that its textgrids are compatible with the annotation tiers in ELAN (ELAN 2019), a very popular tool used for annotating video and audio resources (https://tla. mpi.nl/tools/tla-tools/elan/). This means that it is possible to go between Praat and ELAN fairly seamlessly.
1.3 Acoustic phonetic descriptions of North American languages Based on many of the methods outlined in Section 1.2 above, different types of phonetic descriptions have been created for Indigenous languages of North America. In searching for such descriptions, we used three methods: (a) keyword searches (e. g., Salish + phonetics) through library and online databases, (b) searches within key journals focused on acoustic phonetics (e. g., Journal of Phonetics) and North American Indigenous languages (e. g., International Journal of American Linguistics), and (c) searches within reference sections of individual journal articles, including comprehensive summary articles like Gordon (2017) and McDonough and Whalen (2008). Based on our search, we created a database of existing acoustic studies, which allowed us to identify broad trends in acoustic phonetic research on Indigenous languages of North America, as well
Acoustic phonetics
17
as identify a number of studies that describe particularly unusual features of North American languages and/or provide useful models for future work. In the following two sections, we start by summarizing the types of acoustic phonetic resources that exist (Section 1.3.1); we then describe a number of acoustic phonetic studies of North American languages, as a sample of what exists, and as inspiration of what more could be done (Section 1.3.2).
1.3.1 Types of resources The majority of published acoustic phonetic studies on Indigenous languages of North America are relatively recent (since 2000), and not all language families or possible areas of study within acoustic phonetics are equally represented in the literature. The language families that have been most thoroughly studied are Salish (especially individual sounds, syllables, and prosody), Dene, and Muskogean (especially tone and pitch-accent). The majority of studies we found are on segmental phonetics: the acoustic features of consonants and vowels, and how they interact. These include descriptions of systems as a whole (common with studies on vowels) and ones about specific sounds or sets of sounds (common with studies on consonants). The remaining studies cover areas such as the acoustics of syllables and syllable rhythm, word-level prosody (stress, tone, and pitch-accent; see also Uchihara, this volume; Gordon, this volume), and sentence-level prosody and intonation (see also Tuttle, this volume). Perhaps the most readily available resources to community members are descriptions of sounds found in dictionaries and grammars of individual languages. For example, the SENĆOŦEN dictionary (Montler 2018) has a four-page description of the language’s consonants and vowels, including acoustic and articulatory features. In the past, such descriptions have often been somewhat opaque for non-specialists (very jargon-heavy), but this is changing, with linguists becoming increasingly aware of their responsibility to make their work accessible to the communities they work with (compare, for example, the descriptions in Montler’s 2018 dictionary with those in his (1986) Outline of the Morphology and Phonology of Saanich, North Straits Salish). In our experience, grammars rarely include thorough acoustic phonetic documentation of individual sounds, let alone of other pronunciation features, like word-level prosody or sentential intonation. For these, we must look to more specialized publications. The International Phonetic Association (https://www.internationalphoneticassoci ation.org) publishes thorough phonetic descriptions of individual languages in Illustrations of the International Phonetic Alphabet (IPA). First appearing in the Handbook of the IPA (1999), they are now published as short articles in regular volumes of the Journal of the International Phonetic Association. Audio files are also available for the Illustrations (https://richardbeare.github.io/marijatabain/ipa_illustrations_all.html). So far, only eight Illustrations have been written on Indigenous languages of North America: Spokane (Carlson and Esling 2001); Nuuchahnulth (Carlson, Esling, and Fraser 2001); Chickasaw
18
Sonya Bird, Rae Anne Claxton, and Tess Nolan
(Gordon, Munro, and Ladefoged 2001); Jicarilla Apache (Tuttle and Sandoval 2002); Mono (Olson 2004); Mono Lake Northern Paiute (Babel, Houser, and Toosarvandani 2012); Sahaptin (Hargus and Beavert 2014); and Gitksan (Brown et al. 2016). Similarly, broad phonetic descriptions of specific languages have been published in the Journal of Phonetics (Taff et al. (2001) on Aleut; Flemming, Ladefoged, and Thomason (2008) on Montana Salish), Anthropological Linguistics (Gordon, Munro, and Ladefoged (2000) on Chickasaw; Maddieson, Smith, and Bessell (2001) on Tlingit), and International Journal of American Linguistics (Gordon et al. (2001) on Western Apache; Gordon, Martin, and Langley (2015) on Koasati). A number of these were originally produced by researchers affiliated with the University of California Los Angeles’ Phonetics Laboratory (e. g., Gordon (1996) on Hupa), thanks to a National Science Foundation (NSF) grant Sounds of the World’s Languages awarded to Peter Ladefoged and Ian Maddieson, two of the founding fathers of modern phonetics (Gordon 2017). Aside from broad phonetic descriptions, journals like the Journal of Phonetics and the International Journal of American Linguistics have occasionally published scholarly papers on specific sounds of North American Indigenous languages, for example Bird et al. (2008) on glottalized resonants and Dart (1993) on Tohono O’odham stops and fricatives. The proceedings of the International Conference of Salish and Neighboring Languages have a number of acoustic studies on languages of the Pacific Northwest (e. g., Bessell (1997) on St’át’imcets; Jacobs (2007) on Skwxwu7mesh), and there are also studies on North American languages in the proceedings of the International Congress of Phonetic Sciences (e. g., Ingram and Rigsby (1987) on Gitksan). A handful of doctoral dissertations and master’s theses include acoustic phonetics of North American languages (e. g., Oberly (2008) on Southern Ute; Haynes (2010) on Numu; Neal (2018) on Kawaiisu). Finally, we have also come across two books that include acoustic phonetics, both on Dene languages: Athabaskan Prosody (Hargus and Rice 2005) and Navajo Sound System (McDonough 2003). In general, acoustic phonetic studies of North American languages are still fairly few and far between (McDonough and Whalen 2008), especially ones written in a way that is accessible to community members. However, there is an increasing call for phonetic resources that are accessible, for use in pronunciation learning and teaching (Fish and Miyashita 2017; Bird and Miyashita 2019). We are hopeful that the landscape is shifting in this respect.
1.3.2 Illustrative acoustic phonetic studies In this section we describe a small set of studies that reflect the types of acoustic phonetic research that has been done on North American Indigenous languages. These studies were chosen to cover a range of language families (Uto-Aztecan, Dene, Salish, Eskimo-Aleut, and Muskogean) and focus on a range of areas in acoustic phonetics, from vowels and consonants to word-level prosody and sentential intonation. Many of them also provide good models for future studies on similar features in other languages.
Acoustic phonetics
19
Within each subsection, we start with a small summary called What you need to know, for the benefit of readers who are not so interested in the technicalities of acoustic phonetic research but are looking for some information on the kinds of features to pay attention to in teaching and learning pronunciation.
1.3.2.1 Vowels What you need to know Two things are good to keep in mind when thinking about vowels: (1) Vowels can sound different in different languages, even if they are spelled the same way. Vowels can also sound different within a language, e. g., different elders and families might use different vowel variations. It is good to take the time to learn how different people you interact with pronounce their vowels, so that you can make conscious decisions about what variation(s) you want to use, and how to be consistent in your pronunciation. (2) Vowels can also sound different next to different consonants. Sometimes, when consonants are hard to tell apart, listening to the vowels can give you good clues to what the consonants are. This is the case with /k/ and /q/ (see 1.3.2.2), for example. Again, it is good to take the time to learn how the vowels sound in different contexts, so that you can incorporate this information into your speaking and listening practice. Many existing studies of vowels are acoustic descriptions of a language’s whole vowel system (Bessell 1997; Barthmaier 1998; Tuttle, Lovick, and Núñez-Ortiz 2011; Muehlbauer 2012; Nelson 2013; Nolan 2017; Fortier 2019; Kharlamov and Oberly 2019). These generally include vowel charts (see Figure 2), which show how vowels are distributed in an articulatory and acoustic space defined by the first two formants (F1 and F2) of each vowel (Ladefoged and Johnson 2015). Formants are bands of frequencies that are amplified by the shape of the vocal tract (see Figure 4, vowels [ə], [u] and [ɪ]). Broadly speaking, F1 is correlated with how close the tongue is to the palate (see Bliss, Bird, and Gick, this volume, Section 2.2.3): close vowels like [i] have a low F1; open vowels like [a] have a high F1. F2 is correlated with how far back the tongue is: front vowels like [i] have a high F2; back vowels like [u] (or [ə] and [a] in Figure 1) have a low F2. Vowels can be plotted with F1 on the y-axis and F2 on the x-axis, and with the origin point {0,0} at the top right (Figure 5a). With this lay-out, the acoustic distribution of vowels reflects where they are produced in the mouth, assuming a side view of the mouth with the lips at the left (Figure 5b). As an example, Figure 5 provides the vowel space of lək̓wəŋən (Central Salish) as spoken by a single speaker in the late 1960s (Nolan 2017). One thing that is interesting to notice is that lək̓wəŋən has a typologically unusual vowel space: it is asymmetrical, in that there is no mid back vowel [o] that corresponds to the mid front vowel [e], and [u] (not represented here) is uncommon, found primarily in borrowed words. While four-vowel systems are relatively common, they normally include at least one non-low back vowel (Gordon et al. 2001). Additionally, while the acoustic and artic-
20
Sonya Bird, Rae Anne Claxton, and Tess Nolan
ulatory spaces of [a] and [ə] overlap, they are distinctive in duration; [a] (as well as [i] and [e]) are on average ~180 milliseconds long, while [ə] is between ~100 and ~120 milliseconds on average (Nolan 2017).
(a) acoustic space
(b) articulatory space
Fig. 5: Acoustic (a) and corresponding articulatory (b) space of lək̓wəŋən vowels. Original source for 5(a) (following Creative Commons license CC-BY-NC-SA 2.0): Nolan (2017), Figure 4.9(d) “Lekwungen vowel space across three time points and a mean measure” (p. 74)
From the perspective of vowel typology, one of the most interesting aspects of acoustic descriptions is the detail often provided about how vowels vary as a function of adjacent consonants (Bessell 1997; Bird and Leonard 2009; Nolan 2017; Fortier 2019), position in the word (Gordon and Munro 2007), whether they are in a stressed position or not (Shahin and Blake 2004; Blake and Shahin 2008), etc. Kharlamov and Oberly (2019) documented the acoustic details of vowels in Southern Ute, a Uto-Aztecan language spoken in southeastern Colorado. They measured F1 and F2 for all the vowels and their systematic phonetic variations, or allophones.8 Among other things, they found that /ø/ had a significantly raised F1 (indicating a lowered tongue position) and lowered F2 (indicating a backed tongue position) before velars and uvulars (‘back of the mouth’ sounds), but not after them. Conversely, /a/ had a significantly raised F2 (indicating a fronted tongue position) after palatal (‘front of the mouth’) sounds /j/, /i/, or /ø/ but not before /j/, /i/, or /ø/. In other words, in Southern Ute, velar and uvular consonants tend to draw the previous vowel (/ø/ in particular) down and back, whereas palatal sounds tend to draw the following vowel (at least /a/) forward.9 Kharlamov and Oberly’s findings mirror those
8 See Fitzgerald, this volume, for a definition of ‘allophone’. 9 See Bliss, Bird, and Gick, this volume, for definitions of terms used to describe the place of articulation, like velar, uvular, and palatal.
Acoustic phonetics
21
of Nolan (2017), who found that in lək̓wəŋən (Central Salish), post-velar consonants also had the greatest influence on preceding vowels, whereas palatal consonants had the greatest influence on following vowels. Linguists use the term co-articulation to refer to the influence that adjacent sounds have on one another. Together, Kharlamov and Oberly’s (2019) and Nolan’s (2017) studies suggest that the observed asymmetry between palatal and velar/uvular consonants in their co-articulatory effects holds across languages and can be explained by physiological restrictions shared by speakers around the world. Not all co-articulatory effects are universal though. Bird and Leonard (2009) showed that, even within a single language family, co-articulatory patterns can vary. They considered existing descriptions of /iq/ and /qi/ sequences in Salish languages. These sequences are inherently challenging to pronounce because they require the tongue to move between opposite positions: front and high for /i/ and low and back for /q/ (see Gick and Wilson 2006). They found that co-articulatory effects differed across languages. In general, Central Salish languages were described as exhibiting asymmetrical co-articulatory effects: /qi/ sequences were pronounced with a lowered/retracted vowel, e. g., [qe], whereas /iq/ sequences were pronounced with /i/ intact, but with a transitional element between /i/ and /q/, e. g., [i-əq]. On the other hand, Interior Salish languages were described as exhibiting symmetrical co-articulatory effects: /qi/ and /iq/ were both either pronounced with a lowered/ retracted vowel ([qe] and [eq]) or with a transitional element ([q-ə-i] and [i-ə-q]). Figure 6 provides examples of two different versions of /qi/ within the Central Salish language SENĆOŦEN, pronounced by two different speakers. These show that real speech is in fact more variable than the generalizations above indicate. In 6a, we see that the vowel starts out as [i] (low F1 and high F2) but then transitions into a more central vowel [ə] before [q], with F1 raising and F2 lowering into [q]; in 6b, we see a stable, lowered and retracted [e] sound before [q] (raised F1 and lowered F2 compared to [i]). Knowing what strategies first language (L1) speakers use in their pronunciation can be very useful from a pedagogical perspective. Bird (2018) found that L1 speakers used a range of strategies to pronounce /iq/ and /qi/ sequences, all of which maintained the uvular place of /q/. In contrast, most second language (L2) speakers fronted the uvular /q/ to a velar [k] position, for example pronouncing /iq/ as [ik]. Having documented the differences between the pronunciations of L1 and L2 speakers, we can explain to L2 speakers what strategies L1 speakers use, so that they can choose among them rather than using a strategy that loses the /q/ sound (see also Bliss, Bird, and Gick, this volume, Section 2.3.2).
22
Sonya Bird, Rae Anne Claxton, and Tess Nolan
(a) Transitional vowel [iəq] in /st’iqəl/ (SDIKEL, ‘bog’)
(b) Stable, lowered and retracted [eq], in /st’iqəl/ (SDIKEL, ‘bog’) Fig. 6: Example strategies used for /iq/ articulation in SENĆOŦEN: [iəq] with a transitional vowel (a) vs. [eq] with a stable, lowered vowel (b). Adapted with permission (following Creative Commons license CC-BY-NC-SA 2.0) from Bird (2018), Figures 2 and 3.
Acoustic phonetics
23
1.3.2.2 Consonants What you need to know Especially for adults, it can be hard to learn how to hear and pronounce a lot of new consonants. Here are a few tips we have learnt in our own learning process: (1) Break down learning into steps: learn to hear a new sound first, then learn to pronounce it, and then learn where to use it (in what words). Celebrate the small victories along the way! (2) Languages of the Pacific Northwest in particular have a lot of consonants that might not be familiar to you. On top of this, they can occur together, in long strings (“clusters”). The elders we work with always tell us to slow down when we start speaking, to give each consonant its place. (3) Start with learning the sounds in short, simple words, and gradually build to more complex words and sentences. When you are learning to pronounce consonant clusters, try not to put vowels between them. (4) Once we have learned a sound that is unique to our language, we sometimes start to overuse it, to sound really authentic. This happens a lot with ejectives, for example (see below). When you are able to reliably pronounce a consonant, pay attention to what words have it – this is the next step in mindfully learning to hear and speak the language. In some languages, just a small change in a sound, e. g., glottalization (see below), can change the tense, or even the entire meaning of a word. This is the case for Hul’q’umi’num’ (Central Salish), and we have found that it is useful to learn and practice minimal pairs. Research on consonants tends to focus on specific sounds or sets of sounds that are not found in the dominant language of the area, usually English. Examples include stops, fricatives, and affricates (Dart 1993; Gordon et al. 2002; Holton 2001; McDonough and Wood 2008), post-velar consonants (Bessell 1993; McDowell 2004; Wilson 2007; Bird 2016), ejectives (Ingram and Rigsby 1987; Maddieson, Smith, and Bessell 2001; Wright, Hargus and Davis 2002; Ham 2008; Nelson 2010; Bird 2016), and glottalized resonants (Bird et al. 2008; Bird 2011). Many existing studies on consonants, including those listed above, have been done on languages of the Pacific Northwest, which have particularly rich consonant inventories. In this section we focus on the acoustic characteristics of sounds that involve some kind of constriction or tightening at the throat: ejective stops and glottalized resonants. The Americas are home to over half of the world’s languages that have ejective consonants, and they are found in almost every language of western North America from northern California to Alaska (Maddieson 2013). Ejective stops have been divided into two types, based on their articulatory and acoustic properties: strong/stiff/tense and weak/slack/lax (Lindau 1984). Strong ejectives have a strong ‘popping’ sound when they are released, while weak ejectives are lacking this pop, and sound more like creaky voiced (laryngealized) sounds. The exact pronunciation of ejectives varies across languages (Lindau 1984) as well as within languages (based on factors like word position; Percival 2019), and even within speakers (Wright, Hargus, and Davis 2002). A single North American language – Tlingit (Na Dene, spoken in South Eastern Alaska (US) and Western British Columbia and Yukon (Canada)) – has ejective fricatives (Maddieson,
24
Sonya Bird, Rae Anne Claxton, and Tess Nolan
Smith, and Bessell 2001; see also Bliss, Bird, and Gick, this volume, Section 2.3.5); these are extremely rare across languages of the world (Ladefoged and Maddieson 1996). Figure 7 provides an example of a strong (a) vs. weak (b) version of /t’/, in two repetitions of the SENĆOŦEN word DIL,EḴ /t’íl’əq/ (‘strawberry’) pronounced by a fluent speaking elder (SENĆOŦEN ejectives vary quite a bit in their realization). In the strong version, /t’/ is released with a strong burst (vertical band of noise), followed by a relatively long period of silence (no noise) before the start of the following vowel (this is also seen with the sudden drop in intensity between the burst and the vowel); the regularly-spaced pitch pulses (vertical lines) and relatively high pitch contour at the beginning of the vowel show that it begins with regular (modal) voicing. In the weak version, there is no silence between the /t’/ and the following vowel (there is only a slight dip in intensity),
(a) Strong /t’/ in SENĆOŦEN DIL,EḴ /t’íl’əq/ (‘strawberry’)
(b) Weak /t’/ in SENĆOŦEN DIL,EḴ /t’íl’əq/ (‘strawberry’) Fig. 7: Word-initial strong (a) vs. weak (b) /t’/ in two repetitions of the SENĆOŦEN word /t’íl’əq/ (DIL,EḴ ‘strawberry’) pronounced by a fluent speaking elder.
Acoustic phonetics
25
and creakiness can be observed at the beginning of the vowel, seen as spread out and irregular pitch pulses and a relatively low pitch contour (see also Figure 4 above). Wright, Hargus and Davis (2002) studied the acoustic features of /t’/ in a single language – Witsuwit’en, a Dene language spoken in the central interior of British Columbia, Canada. Their study provides an excellent model for studying ejectives, because they used a comprehensive set of measurements to characterize /t’/ (compared to plain /t/). Auditorily, they heard variation in production similar to that illustrated in Figure 7 for SENĆOŦEN: some /t’/ tokens were very strong and popping, others sounded more aspirated (breathy), and yet others were weak and sounded like voiceless stops followed by creaky voice. Ultimately, the acoustic measures they took placed Witsuwit’en /t’/ in the weak category, although the authors argued that a binary distinction could not fully capture the range of productions observed. Interestingly, pitch movement during the following vowel differed between men and women, illustrating how important gender can be in influencing the details of speech (see Johnson, Ladefoged, and McDonough (1993) on men and women’s pronunciation of Navajo vowels). Wright, Hargus, and Davis’ (2002) study alongside others that have described ejectives (Flemming, Ladefoged, and Thomason 2008; Ham 2008; Percival, Bird and Gerdts 2018) in Indigenous languages of North America shows us that ejectives do not necessarily fall into neat strong vs. weak categories, and that variation can be substantial both within and across speakers. One particularly noteworthy feature of ejectives in the context of language revitalization is that, at least in some languages, they seem to be strengthening over time: in cases where they have traditionally been described as weak, they are now produced as relatively strong (Bird 2020). In addition, learners are using them more than their elders do (Babel 2009; Bird, Gerdts, and Leonard 2016; Haynes 2010). This shift likely results from a combination of the pedagogical context in which the sounds are being transmitted (Saito and van Poeteren 2012; Uther, Knoll, Burnham 2006) and factors related to cultural identity (Gatbonton, Trofimovich, and Segalowitz 2011; Haynes 2010; Nance et al. 2016). Another set of sounds that involve a tightening of the throat is glottalized resonants (GRs). These are resonant consonants which include both an articulation in the upper vocal tract (e. g., at the lips for [m]) and constriction in the larynx (and sometimes the pharynx) (see Bliss, Bird, and Gick, this volume, Section 2.3.3). GRs are relatively rare sounds, found in only 20 of the 317 languages sampled by Maddieson (1984), but they are common in the languages of the Pacific Northwest. What is interesting about GRs is that glottalization varies both within and across languages, along two dimensions: (a) the timing of laryngeal constriction relative to the resonant articulation and (b) the phonetic realization of glottalization. Bird et al. (2008) document the timing between the laryngeal and resonant (oral) articulations in GRs in three different languages: St’át’imcets and NɬeɁkepmxcin (both Interior Salish), and Nuuchahnulth (Wakashan). They show that GRs are consistently pre-glottalized (e. g., [Ɂm]) in Nuuchahnulth, meaning glottalization occurs at the beginning of the resonant; they are consistently post-glottalized (e. g., [mɁ]) in NɬeɁkepmxcin, meaning glottalization occurs at the end of the resonant;
26
Sonya Bird, Rae Anne Claxton, and Tess Nolan
in St’át’imcets, their timing varies as a function of syllable position: pre-glottalized at the beginning of syllables (in onset position), post-glottalized at the end of syllables (in coda position), and mid-glottalized ([mɁm]) in intervocalic position (between two vowels). Figure 8 provides another example, from Hul’q’umi’num’ (Central Salish). In
(a) Pre-glottalized [Ɂl] in si’lu [siɁlə] (‘grandparent’)
(b) Post-glottalized [lɁ] in sisul’u [sisəlɁə] (‘grandparent-diminutive’)) Fig. 8: Pre-glottalized [Ɂl] in Hul’q’umi’num’ si’lu [siɁlə] (‘grandparent’) (a) vs. post-glottalized [lɁ] in sisul’u [sisəlɁə] (‘grandparent- DIM’) (b), as pronounced by co-author Rae Anne Claxton.
Acoustic phonetics
27
this language, full vowels (/a e i o u/) attract glottalization towards them; conversely, stressed /ə/ repels glottalization away from it. Thus, in Figure 8(a), /l’/ is pre-glottalized in si’lu [siɁlə] (‘grandparent’); in Figure 8(b), it is post-glottalized in the diminutive form sisul’u [sisəlɁə] (‘grandparent-diminutive’). Note the slight creakiness in [l] adjacent to the glottal stop [Ɂ] in both cases, also seen in Figures 4 and 7. Similar to ejectives, GRs also vary in terms of the phonetic realization of glottalization (Bird 2011). In some languages, glottalization seems to always be realized as a full glottal stop, e. g., Hul’q’umi’num’ (see Figure 8). In other languages, glottalization is more variable and often realized as mild creakiness, especially in unstressed environments, as for /l’/ in SENĆOŦEN DIL,EḴ /t’íl’əq/ ‘strawberry’ (Figure 7). The languages of the Pacific Northwest in particular have contributed immensely to our understanding of glottal articulations. As with other examples provided in Section 1.3, documenting the variation we see in ejectives and glottalized resonants is helpful not only from the perspective of phonetic typology, but also from the perspective of pronunciation teaching and learning: knowing how these sounds are pronounced by fluent speakers can help learners achieve what they think of as ‘authentic’ pronunciation. Because these sounds tend to be relatively easy to “see” on visual displays like Figures 7 and 8, they are good candidates for incorporating visual tools into learning (Olson 2014): learners can be taught to record themselves in Praat (see Section 1.2 above) and try to approximate an Elder’s speech visually as well as auditorily. From our experiences, this is an engaging way to learn the details of pronunciation, because learners can get a very concrete sense of what they are doing differently from their Elders, and how to approximate their speech more closely (Bird et al. 2019). Before moving on to prosody and intonation (Section 1.3.2.3), it is worth mentioning that many Indigenous languages of North America have length contrasts, either in their vowel systems (Apache (Gordon et al. 2001); Aleut (Taff et al. 2001); Chickasaw (Gordon 2000, Gordon et al. 2001); Plains Cree (Muehlbauer 2012); Upper Tanana Athabascan (Tuttle, Lovick, and Núñez-Ortiz 2011); Washo (Yu 2008)) or consonant systems (Washo, (Yu 2008)). In these languages, sounds include long and short versions of the same vowel or consonant. For example, Gordon and Munro (2007) studied the vowel length distinction in Chickasaw, across four different speakers and between short and long /a, aː/, /o, oː/, and /i, iː/. In Chickasaw, vowel length alone can distinguish between two different words, e. g., /iʃtalali/ (‘I bring it here’) vs. /iʃtalaːli/ (‘you set it upright’) differ in only the length of the vowel in the second-to-last syllable. In many languages with length contrasts, a short/long contrast is the whole story. In Chickasaw however, things are a bit more complex, in that short vowels can also be lengthened in prominent word positions (see Tuttle, this volume, on word prosody), leading to an intermediate vowel length. For example, the second short /a/ in /iʃtalali/ can be lengthened to [aˑ] because of its position in the word and the word-level prosody, leading to the pronunciation [iʃtalaˑli]. Gordon and Munro's results show short vowels are ~ 92 milliseconds long on average, the long vowels are ~155 milliseconds long on average, and the lengthened short vowels are ~140 milliseconds long on average.
28
Sonya Bird, Rae Anne Claxton, and Tess Nolan
1.3.2.3 Prosody and intonation What you need to know In language classes, we tend to focus listening and speaking work on individual sounds and words, but out in the “real world” (in community), intonation – the flow of whole sentences – is also very important. Learning the rhythm and flow of a language can be done somewhat separately from learning the consonants and vowels. A good way to learn is to replace all the syllables in a sentence with . For example, “the picture is beautiful” (with seven syllables) can be pronounced “ma MA ma ma MA ma ma” (capital letters mark prominent syllables). Doing this lets learners abstract away from challenging sounds, and focus on the flow of whole sentences – what syllables are more emphasized, and how emphasis is created. Once we have the flow down, we can start adding the consonants and vowels back in. Prosody and intonation refer to the rhythmic flow of words and sentences. Though less researched than segmental phonetics (vowels and consonants), prosody and intonation have also been the focus of a small number of studies. At the word level (see Tuttle, this volume, and Uchihara, this volume), the most commonly studied features are lexical tone and pitch accent (Martin and Johnson 2002; Gordon 2008; Uchihara 2013; Rice 2014; Hirata-Edds and Herrick 2017). For example, Martin and Johnson (2002) studied what they call “tonal accent” in Muskogee Creek, a Muskogean language spoken in eastern Oklahoma. Their goal was to acoustically confirm the auditory-based descriptions made by earlier researchers, in particular Haas (1977). Haas described three tones (level, falling, and rising) that could be borne by “key” (prominent) syllables in a word, creating grammatical (aspectual) contrasts. For example, the difference in tones creates the difference in aspect in three forms of the stem apoːk- ‘sit (of three or more)’: level tone [apóːkiːs] ‘we are (here),’ falling tone [apôːkiːs] ‘we have sat down, are in a sitting position,’ and rising tone [apǒːⁿkiːs] ‘we keep sitting and sitting’ (Martin and Johnson 2002, p. 29). Corroborating Haas’ descriptions, Martin and Johnson found that (1) level tones maintained a generally level pitch that slightly decreased over time; (2) falling tones varied by position: in initial position they had a rise and then a fall and, in medial and final positions, they gradually declined; (3) rising tones rose to a peak before slightly falling. Additionally, in words with multiple key syllables, tone was slightly lower on any key syllable following the initial one. Martin and Johnson also reported on a number of cases where Haas’ phonetic descriptions did not exactly match their acoustic findings. This raises two points to consider when doing acoustic work to corroborate earlier descriptive work: (1) the language may change over time or place (e. g., if speakers are from different communities), leading to observed discrepancies and (2) acoustic analysis does not always match up perfectly with auditory analysis (what we hear), for various reasons (Bird and Leonard 2009). In our experience, automatic pitch tracking in Praat is somewhat unreliable, and therefore should be confirmed with manual measurements and with our ears. Martin and Johnson’s
Acoustic phonetics
29
study is a good example of how the tools available now allow us to confirm and also go beyond previous work, giving us more insight into the fine phonetic details of speech, especially in ways that the ear may not pick up on unaided. In terms of prosody beyond the word (see Zúñiga, this volume), one common topic is question intonation, which refers to the flow of questions, including the rising and falling of pitch (Taff et al. 2001; Jacobs 2007; Oberg 2007; Koch 2008). It has been claimed that there is a universal tendency for languages to indicate yes/no questions with either a rising pitch (as in English ‘they are going to the store?’) or an overall higher pitch, and to indicate statements with a falling pitch (as in English ‘they are going to the store.’) or an overall lower pitch (Ladd 1981). Several North American languages show the latter pattern: an overall raised pitch in yes/no questions compared to statements. For example, Jacobs (2007) compared the intonation of declarative statements vs. yes/no questions in Skwxwu7mesh, a Central Salish language spoken in southwestern British Columbia, Canada. He measured (a) the overall change in pitch from the beginning to the end of the sentence and (b) the average pitches on the first two and last two syllables. He showed that both statements and questions had a fall in pitch from the start of the sentence to the end (Figure 9). In statements, the pitch fell consistently across the syllables, whereas in questions, the pitch initially rose through the question marker (first syllable) and then fell through the rest of the question, leading to an overall raised pitch compared to statements. Questions in Skwxwu7mesh are also marked by the presence of a question particle, making statements and questions syntactically distinct, and avoiding confusion between similar intonation patterns.
Fig. 9: Pitch contours of Skwxwu7mesh declaratives and yes/no questions. Original source (following Creative Commons license CC-BYNC-SA 2.0): Jacobs (2007), Table 9 “Comparison of pitch”, p. 247.
30
Sonya Bird, Rae Anne Claxton, and Tess Nolan
In their work on Aleut, an Eskimo-Aleut language spoken in southwestern Alaska,10 Taff et al. (2001) found that both statements and yes/no questions had falling intonation, and rises in pitch word-internally occurred in both types of sentences as well. A slight rise in the final content word was seen in questions for some speakers, but not for others, and this rise was also seen in some declaratives, so Taff et al. did not consider this feature to be distinctive of yes/no questions. They also compared the median pitches of statements and questions for each speaker and found that the pitch range of the questions for each speaker was within the range of their declarative sentences, neither higher nor lower. Taff et al.’s study shows that North American languages can “buck the trend” so to speak when it comes to assumed universals about intonation. The common theme emerging from the brief literature review presented in Section 1.3 is that Indigenous languages of North America have much to offer in terms of the richness of their phonetic structures. Research documenting these patterns has contributed and will continue to contribute greatly to our understanding of cross-linguistic as well as language-specific patterns of phonetic realization. In addition, it has provided a robust foundation for developing pedagogical resources for teaching and learning the details of pronunciation, which is necessarily language specific.
1.4 New wave of phonetic research: community-based and collaborative As mentioned in Section 1.3.1, much of the existing phonetic work on Indigenous languages of North America is highly technical, and as a result not terribly accessible to community members and language learners and teachers. Nonetheless, the studies cited in Section 1.3.2 have hopefully shown that conducting phonetic research can be valuable from the perspective of teaching and learning pronunciation in the context of language revitalization. In this final section, we offer our thoughts on how to conduct phonetic work in an accessible and collaborative way. Donna Gerdts (2010) talks about the linguist as one member of a language revitalization team. Indeed, in our experience, the most valuable kind of acoustic phonetic work combines the expertise of linguists, Elders and teachers, and learners (see also Bird and Miyashita 2019). Elders, teachers, and learners are in the best position to identify the challenges that learners face in speaking and hearing their language(s). These challenges most often correspond to features that differ from those of the majority language (English in our case) and that are particularly interesting from the perspective of phonetic typology. Community members also know best how to design research methods
10 Their study was in fact a very broad one, which covered consonant properties, vowel space, wordlevel stress, and sentence intonation.
Acoustic phonetics
31
for studying speech in culturally appropriate ways. Linguists (phoneticians) have the technical expertise to analyze the acoustic features of speech instrumentally. Once the phonetic details are understood, the team can work together to determine how best to make the research findings accessible and useful for teachers and learners. Naatosi Fish and Mizuki Miyashita (Fish and Miyashita 2017, Fish 2018)’s “pitch art” – stylized pitch contours of Blackfoot (Algonquian) tonal melodies (Figure 10) – is an excellent example of how teamwork can lead to engaging tools for teaching and learning the phonetic details of speech.
Fig. 10: Pitch Art of the Blackfoot word makóyi (wolf). Copied with permission from Fish (2018).
Collaborative phonetic work like Fish and Miyashita’s pitch art project has also been documented by Herrick et al. (2015) working on Cherokee and Bird et al. (2019) working on Hul’q’umi’num’. A recent Special Session at the International Conference of the Phonetic Sciences (Billington and Bird 2019) brought together linguists working with Indigenous and minority language communities around the world to share ideas about best practices for ensuring that we continue to work together in the most effective ways to support communities’ efforts in sound-related work, as part of larger language maintenance and revitalization efforts. We are enthusiastic about these continued partnerships and their contribution to our understanding of phonetic structures across the languages of the world and also to our ability to create resources for teaching and learning pronunciation. Acknowledgements: HÍSW̱ḴE SI,IÁM, huy tseep q’u sii’em’, and thank you to our community partners, who continue to shape the way we go about conducting acoustic phonetic work. Hul’q’umi’num’: late Dr. Sti’tum’at Ruby Peter, Swustanulwut Delores Louie, Dr. Donna Gerdts, Hul’q’umi’num’ Language & Culture Society, and students in the Hul’q’umi’num’ Language Academy. SENĆOŦEN: late Ivan Morris Sr. and late Ray Sam, Tye Swallow, teachers and students in the W̱,SENĆOŦEN IST program. Thank you also to the editors and reviewers for your valuable feedback.
32
Sonya Bird, Rae Anne Claxton, and Tess Nolan
References Cited in the chapter Asher, James J. 1977. Learning another language through actions: The complete teacher’s guidebook. Los Gatos, CA: Sky Oaks Productions. Babel, Molly. 2009. The phonetic and phonological effects of obsolescence in Northern Paiute. In James Stanford & Dennis R. Preston (eds.), Variation in indigenous minority languages, 23–46. Amsterdam: John Benjamins. Babel, Molly, Michael J. Houser & Maziar Toosarvandani. 2012. Illustrations of the IPA: Mono Lake Northern Paiute. Journal of the International Phonetic Association 42(2). 233–243. Barthmaier, Paul. 1998. Lushootseed vowels: A preliminary phonetic analysis. In Papers for the International Conference on Salish and Neighboring Languages (ICSNL) 33, 1–11. The University of British Columbia. Vancouver, BC. Bessell, Nicola. 1993. Towards a phonetic and phonological typology of post-velar articulation. Vancouver, BC: University of British Columbia dissertation. Bessell, Nicola. 1997. St’át’imcets vowels and consonants: A phonetic study. In Papers for the International Conference on Salish and Neighboring Languages (ICSNL) 32, 64–85. The University of British Columbia. Vancouver, BC. Bird, Sonya. 2011. The nature of laryngealization in St’át’imcets laryngealized resonants. International Journal of American Linguistics 77(2). 159–184. Bird, Sonya. 2016. Pronunciation change in SENĆOŦEN: A acoustic study of /k, kw, kw’, q, q’, qw qw’/ across generations of speakers. Paper presented at the annual meeting of the Acoustical Society of America. Salt Lake City, UT. Bird, Sonya. 2018. Phonetics in the context of language revitalization: Exploring pronunciation strategies. In Lisa Matthewson, Erin Guntly, & Michael Rochement (eds.), Wa7 xweysás i nqwal’utteníha i ucwalmícwa: He loves the people’s languages. Essays in honour of Henry Davis, 493–508. (University of British Columbia Occasional Papers in Linguistics 6). Vancouver, BC: University of British Columbia. Bird, Sonya. 2020. Pronunciation among Adult Indigenous Language Learners: the case of SENĆOŦEN /t’/. Journal of Second Language Pronunciation. 6(2). 148–179. Bird, Sonya, Marion Caldecott, Fiona Campbell, Bryan Gick & Patricia A. Shaw. 2008. Oral-laryngeal timing in glottalized resonants. Journal of Phonetics 36. 492–507. Bird, Sonya, Rae Anne Claxton, Luschiimtunaat Sonya Charlie, S-hwutstus Harvey George, Maida Percival & Sq’utxulenuhw George Seymour. 2019. Seeing speech: Teaching and learning Hul’q’umi’num’ pronunciation with Praat. Paper presented at the International Conference on Language Documentation and Conservation. Honolulu, HI. Bird, Sonya, Donna Gerdts & Janet Leonard. 2016. The realization of obstruents across speakers of Hul’q’umi’num’. Canadian Acoustics 44(3). 134–135. Bird, Sonya & Sarah Kell. 2017. The role of pronunciation in SENĆOŦEN language revitalization. Canadian Modern Language Review 73(4). 538–569. Bird, Sonya & Janet Leonard. 2009. The universality of articulatory conflict-resolution: Evidence from Salish languages. Northwest Journal of Linguistics 3(2). 1–29. Bird, Sonya & Mizuki Miyashita. 2019. Teaching phonetics in the context of Indigenous language revitalization. Proceedings of the 2018 International Symposium on Applied Phonetics (ISAPh), 39–44. https:// www.isca-speech.org/archive/ISAPh_2018/pdfs/07.pdf Billington, Rosey & Sonya Bird. 2019. Special Session at the 19th International Congress of Phonetic Sciences (ICPhS): Phoneticians in partnership with communities in language revitalization and maintenance. Melbourne, Australia, August 5–9.
Acoustic phonetics
33
Blake, Susan & Kimary Shahin. 2008. On the phonetics of schwa in Sliammon (M. Comox Salish): Implications for the representation of Salish vowels. Canadian Acoustics 36(4). 42–43. Bliss, Heather, Sonya Bird, Strang Burton, PEPAḴIYE Ashley Cooper & Bryan Gick. 2018. Seeing speech: Ultrasound-based multimedia resources for pronunciation learning in Indigenous languages. Language Documentation & Conservation 12. 315–338. Bliss, Heather, Sonya Bird & Bryan Gick. 2023 [this volume]. Articulatory phonetics. In Carmen Dagostino, Marianne Mithun & Keren Rice (eds.), The Languages and Linguistics of Indigenous North America: A Comprehensive Guide. Berlin & Boston: De Gruyter Mouton. Boersma, Paul & David Weenink. 2022. Praat: Doing phonetics by computer [Computer program]. Version 6.2.10, retrieved 17 March 2022 from http://www.praat.org Brown, Jason, Henry Davis, Michael Schwan & Barbara Sennott. 2016. Illustrations of the IPA: Gitksan. Journal of the International Phonetic Association 46(3). 367–378. Caldecott, Marion & Karsten Koch. 2014. Using mixed media tools for eliciting discourse in Indigenous languages. Language Documentation & Conservation 8. 209–240. https://scholarspace.manoa.hawaii. edu/bitstream/10125/24603/1/caldecott.pdf Carlson, Barry & John H. Esling. 2001. Illustrations of the IPA: Spokane. Journal of the International Phonetic Association 30(1–2). 97–102. Carlson, Barry, John H. Esling & Katie Fraser. 2001. Illustrations of the IPA: Nuuchahnulth. Journal of the International Phonetic Association 31(2). 275–279. Dart, Sarah N. 1993. Phonetic properties of O’odham stop and fricative contrasts. International Journal of American Linguistics 59(1). 16–37. De Decker, Paul & Jennifer Nycz. 2011. For the record: Which digital media can be used for sociophonetic analysis? University of Pennsylvania Working Papers in Linguistics 17(2). 51–59. https://repository.upenn. edu/pwpl/vol17/iss2/7 Dry, Helen Aristar. 2009. Preserving digital language materials: Some considerations for community initiatives. In Wayne Harbert, Sally McConnell-Ginet, Amanda Miller, & John Whitman (eds.), Language and Poverty, 202–222. Bristol: Multilingual Matters. ELAN (Version 5.6) [Computer software]. 2019, June 04. Nijmegen: Max Planck Institute for Psycholinguistics. https://tla.mpi.nl/tools/tla-tools/elan/ Fish, Naatosi. 2018. Pedagogy of pitch in L2 Blackfoot. University of Montana Undergraduate Theses and Professional Papers. 193. https://scholarworks.umt.edu/utpp/193/ Fish, Naatosi & Mizuki Miyashita. 2017. Guiding pronunciation of Blackfoot melody. In Jon Reyhner, Joseph Martin, Louise Lockard & Willard Sakiestewa Gilbert (eds.), Honoring Our Teachers, 203–210. Flagstaff, AZ: Northern Arizona University. Flemming, Edward, Peter Ladefoged & Sarah Thomason. 2008. Phonetic structures of Montana Salish. Journal of Phonetics 36. 465–491. Fortier, Kyra. 2019. An acoustic investigation of vowel variation in Gitksan. Victoria, BC: University of Victoria MA thesis. Gardner, Evan. 2019. Where Are Your Keys. https://whereareyourkeys.org/ Gatbonton, Elizabeth, Pavel Trofimovich & Norman Segalowitz. 2011. Ethnic group affiliation and patterns of development of a phonological variable. The Modern Language Journal 95(2). 188–204. Gerdts, Donna. 2010. Beyond expertise. The role of the linguist in language revitalization programs. In Lenore Grenoble & N. Louanna Furbee (eds.), Language documentation practice and values, 173–192. Amsterdam: John Benjamins. Gick, Brian & Ian Wilson. 2006. Excrescent schwa and vowel laxing: Cross-linguistic responses to conflicting articulatory targets. In Louis Goldstein, Douglas H. Whalen & Catherine T. Best (eds.), Laboratory Phonology VIII, 635–659. Berlin & New York: Mouton de Gruyter.
34
Sonya Bird, Rae Anne Claxton, and Tess Nolan
Gordon, Matthew. 1996. The phonetic structures of Hupa. In Ian Maddison (ed.), Fieldwork studies of targeted languages IV. (University of California Los Angeles Working Papers in Phonetics 93). 164–187. Los Angeles, CA: University of California. Gordon, Matthew. 2008. Pitch accent timing and scaling in Chickasaw. Journal of Phonetics 36. 521–535. Gordon, Matthew. 2017. Phonetic and phonological research on Native American languages: Past, present, and future. International Journal of American Linguistics 83(1). 79–110. Gordon, Matthew, Paul Barthmaier & Kathy Sands. 2002. A cross-linguistic acoustic study of voiceless fricatives. Journal of the International Phonetic Association 32(2). 141–174. Gordon, Matthew, Jack B. Martin & Linda Langley. 2015. Some phonetic structures of Koasati. International Journal of American Linguistics 81(1). 83–118. Gordon, Matthew & Pamela Munro. 2007. A phonetic study of final vowel lengthening in Chickasaw. International Journal of American Linguistics 73(3). 293–330. Gordon, Matthew, Pamela Munro & Peter Ladefoged. 2000. Some phonetic structures of Chickasaw. Anthropological Linguistics 42(3). 366–400. Gordon, Matthew, Pamela Munro & Peter Ladefoged. 2001. Illustrations of the IPA: Chickasaw. Journal of the International Phonetic Association 31(2). 287–290. Gordon, Matthew, Brian Potter, John Dawson, Willem de Reuse & Peter Ladefoged. 2001. Phonetic structures of Western Apache. International Journal of American Linguistics 67(4). 415–448. Haas, Mary R. 1977. Tonal accent in Creek. In Larry M. Hyman (ed.), Studies in stress and accent, 195–208. (Southern California Occasional Papers in Linguistics 4). Los Angeles, CA: University of Southern California. Ham, SooYoun. 2008. Tsilhqut’in ejectives: A descriptive phonetic study. Victoria, BC: University of Victoria MA thesis. Hargus, Sharon & Keren Rice (eds.). 2005. Athabaskan Prosody. (Current Issues in Linguistic Theory 269). Amsterdam & Philadelphia: John Benjamins. Hargus, Sharon & Virginia Beavert. 2014. Illustrations of the IPA: Northwest Sahaptin. Journal of the International Phonetic Association 44(3). 319–342. Haynes, Erin. 2010. Phonetic and phonological acquisition in endangered languages learned by adults: A case study of Numu (Oregon Northern Paiute). Berkeley, CA: University of California Berkeley dissertation. Holton, Gary. 2001. Fortis and lenis fricatives in Tanacross Athapaskan. International Journal of American Linguistics 67(4). 396–414. Ingram, John & Bruce Rigsby. 1987. Glottalic stops in Gitksan: An acoustic analysis. 11th International Congress of Phonetic Sciences (ICPhS XI), 134–137. Hirata-Edds, Tracy & Dylan Herrick. 2017. Building tone resources for second language learners from phonetic documentation: Cherokee examples. Language Documentation & Conservation 11. 289–304. Herrick, Dylan, Marcellino Berardo, Durbin Feeling, Tracy Hirata-Edds & Lizette Peter. 2015. Collaborative documentation and revitalization of Cherokee tone. Language Documentation & Conservation 9. 12–31. International Phonetic Association. 1999. Handbook of the International Phonetic Association: A guide to the use of the International Phonetic Alphabet. Cambridge: Cambridge University Press. Jacobs, Peter. 2007. Intonation of yes/no questions in Skwxwú7mesh. In Kristin M. Jóhannsdóttir & Martin A. Oberg (eds.), Papers for the International Conference on Salish and Neighboring Languages (ICSNL) 42, 236–255. (University of British Columbia Working Papers in Linguistics 20). Vancouver, BC: University of British Columbia. Jenni, Barbara, Adar Anisman, Onowa McIvor & Peter Jacobs. 2017. An exploration of the effects of mentor-apprentice programs on mentors’ and apprentices’ wellbeing. International Journal of Indigenous Health 12(2). 25–42. Johnson, Keith, Peter Ladefoged & Joyce McDonough. 1993. Do women listen to men’s voices with men’s mouths in mind? The Journal of the Acoustical Society 93. 2298. Doi: 10.1121/1.406484
Acoustic phonetics
35
Kharlamov, Viktor & Stacey Oberly. 2019. Phonetics of Southern Ute vowels. Journal of the International Phonetic Association 49(2). 1–19. Koch, Karsten. 2008. Intonation and focus in Nlhe7kepmxcin (Thompson River Salish). Vancouver, BC: University of British Columbia dissertation. Ladd, D. Robert. 1981. On intonational universals. In Terry Meyers, John Laver & John M. Anderson (eds.), The cognitive representation of speech, 389–397. (Advances in Psychology 7). Amsterdam & New York: North Holland Publishing Company. Ladefoged, Peter. 2003. Phonetic data analysis: An introduction to fieldwork and instrumental techniques. Malden, MA: Blackwell. Ladefoged, Peter & Keith Johnson. 2015. A Course in Phonetics. 7th edn. Stamford, CT: Cengage Learning. Ladefoged, Peter & Ian Maddieson. 1996. The Sounds of the World’s Languages. Hoboken, NJ: Wiley-Blackwell. Lee, Junkyu, Juhyun Jang & Luke Plonsky. 2015. The effectiveness of second language pronunciation instruction: A meta-analysis. Applied Linguistics 36(3). 345–366. Lindau, Mona. 1984. Phonetic differences in glottalic consonants. Journal of Phonetics 12. 147–55. Maddieson, Ian. 1984. Patterns of sound. Cambridge, MA: Cambridge University Press. Maddieson, Ian. 2013. Glottalized Consonants. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. https://wals.info/chapter/7 Maddieson, Ian, Caroline L. Smith & Nicola Bessell. 2001. Aspects of the phonetics of Tlingit. Anthropological Linguistics 43(2). 135–176. Martin, Jack B. & Keith Johnson. 2002. An acoustic study of “tonal accent” in Creek. International Journal of American Linguistics 68(1). 28–50. McDonough, Joyce. 2003. The Navajo sound system. (Studies in Natural Language & Linguistics 55). Dordrecht, Boston, London: Kluwer Academic Publishers. McDonough, Joyce, Douglas H. Whalen. 2008. Editorial: The phonetics of native North American languages. Journal of Phonetics 36. 423–426. McDonough, Joyce & Valerie Wood. 2008. The stop contrasts of the Athabaskan languages. Journal of Phonetics 36. 427–449. McDowell, Ramona. 2004. Retraction in Montana Salish lateral consonants. Vancouver, BC: University of British Columbia MA thesis. McIvor, Onowa. 2015. Adult Indigenous language learning in Western Canada: What is holding us back? In Kathryn A. Michel, Patrick D. Walton, Emma Bourassa & Jack Miller (eds.), Living our languages: Papers from the 19th Stabilizing Indigenous Languages Symposium, 37–49. Ronkonkoma, NY: Linus Learning. Montler, Timothy. 1986. An outline of the morphology and phonology of Saanich, North Straits Salish. (University of Montana Occasional Papers in Linguistics 4). Missoula, MT: University of Montana Linguistics Laboratory. http://saanich.montler.net/Outline/index.htm Montler, Timothy. 2018. SENĆOŦEN: A dictionary of the Saanich language. Seattle, WA: University of Washington Press. Muehlbauer, Jeffrey. 2012. Vowel spaces in Plains Cree. Journal of the International Phonetic Association 42(1). 91–105. Nance, Claire, Wilson McLeod, Bernadette O’Rourke & Stuart Dunmore. 2016. Identity, accent aim, and motivation in second language users: New Scottish Gaelic speakers’ use of phonetic variation. Journal of Sociolinguistics 20(2). 164–191. Neal, Thomas. 2018. A phonetic description of the Kawaiisu language. Tucson, AZ: University of Arizona dissertation. Nelson, Katherine. 2010. Ejectives in Nez Perce. Santa Barbara Papers in Linguistics 21. 1–19. Nelson, Katherine. 2013. The Nez Perce vowel system: A phonetic analysis. Proceedings of Meetings on Acoustics 19. 1–9.
36
Sonya Bird, Rae Anne Claxton, and Tess Nolan
Nolan, Tess. 2017. A phonetic investigation of vowel variation in Lekwungen. Victoria, BC: University of Victoria MA thesis. Oberg, Martin. 2007. Intonation contours in St’át’imcets. In Kristin M. Jóhannsdóttir & Martin A. Oberg (eds.), Papers for the International Conference on Salish and Neighboring Languages (ICSNL) 42, 357–370. (University of British Columbia Working Papers in Linguistics 20). Vancouver, BC. Oberly, Stacey. 2008. A phonetic analysis of Southern Ute with a discussion of Southern Ute language policies and revitalization. Tucson, AZ: University of Arizona dissertation. Olson, Daniel J. 2014. Benefits of visual feedback on segmental production in the L2 classroom. Language Learning and Technology 18(3). 173–192. Olson, Kenneth. 2004. Illustrations of the IPA: Mono. Journal of the International Phonetic Association 34(2). 233–238. Percival, Maida. 2019. Contextual variation in the acoustics of ejective stops. Proceedings of ICPhS 2019. Percival, Maida, Sonya Bird & Donna Gerdts. 2018. Laryngeal contrasts in first and second language speakers of Hul’q’umi’num’. Paper presented at the Annual Meeting of the Acoustical Society of America. November 5–9, Victoria, BC. Rice, Keren. 2014. On beginning the study of the tone system of a Dene (Athabaskan) language: Looking back. Language Documentation & Conservation 8. 690—706. Saito, Kazuya & Kim van Poeteren. 2012. Pronunciation-specific adjustment strategies for intelligibility in L2 teacher talk: results and implications of a questionnaire study. Language Awareness 21(4). 369–385. Shahin, Kimary & Susan Blake. 2004. A Phonetic Study of Schwa in St’át’imcets (Lillooet Salish). In Donna Gerdts & Lisa Matthewson (eds.), Studies in Salish Linguistics in honour of M. Dale Kinkade, 311–327. (University of Montana Occasional Papers in Linguistics 17). Missoula, MT: University of Montana Press. Taff, Alice, Lorna Rozelle, Taehong Cho, Peter Ladefoged, Moses Dirks & Jacob Wegelin. 2001. Phonetic structures of Aleut. Journal of Phonetics 29. 231–271. Tuttle, Siri G., Olga Lovick & Isabel Núñez-Ortiz. 2011. Vowels of Upper Tanana Athabascan. Journal of the International Phonetic Association 41(3). 283–312 Tuttle, Siri G. & Merton Sandoval. 2002. Illustrations of the IPA: Jicarilla Apache. Journal of the International Phonetic Association 32(1). 105–112. Uchihara, Hiroto. 2016. Tone and accent in Oklahoma Cherokee. Oxford: Oxford University Press. Uther, Maria, Monja Angelika Knoll & Denis Burnham. 2006. Do you speak E-NG-LI-SH? A comparison of foreigner- and infant-directed speech. Speech Communication 49. 2–7. Wilson, Ian. 2007. The effects of post-velar consonants on vowels in Nuu-chah-nulth: Auditory, acoustic, and articulatory evidence. Canadian Journal of Linguistics/Revue canadienne de linguistique 52(1). 43–70. Wright, Richard, Sharon Hargus & Katharine Davis. 2002. On the categorization of ejectives: Data from Witsuwit’en. Journal of the International Phonetic Association 32(1). 43–77. Yu, Alan C. L. 2008. The phonetics of quantity alternation in Washo. Journal of Phonetics 36. 508–520.
Other readings of interest Linguistics and Indigenous language revitalization Czaykowska-Higgins, Ewa. 2009. Research models, community engagement, and linguistic fieldwork: Reflections on working within Canadian Indigenous communities. Language Documentation & Conservation 3(1). 15–50. Fitzgerald, Colleen M. 2017. Motivating the documentation of the verbal arts: Arguments from theory and practice. Language Documentation & Conservation 11. 114–132.
Acoustic phonetics
37
Fitzgerald, Colleen M. 2017. The sounds of Indigenous language revitalization. Plenary address, 91st Annual General Meeting of the Linguistic Society of America. January 5–8, Austin, TX. Percival, Maida, Heather Bliss & Murray Schellenberg. 2017. Methodological trade-offs for dual-purpose phonetic fieldwork. Canadian Acoustics 45(3). 184–185. Rice, Keren. 2010. The linguist’s responsibilities to the community of speakers: Community-based research. In Lenore A. Grenoble & N. Louanna Furbee (eds.), Language documentation: Practice and values, 25–36. Amsterdam & Philadelphia: John Benjamins.
Methods Bird, Sonya. 2011. Phonetic fieldwork in the Pacific Northwest. International Congress of Phonetic Sciences (ICPhS) 17. Hong Kong.
Other broad phonetic descriptions Mithun, Marianne. 1999. The languages of Native North America. Cambridge: Cambridge University Press. Nelson, Katherine. 2013. An acoustic analysis and cross-linguistic study of the phonemic inventory of Nez Perce. Houston, TX: Rice University dissertation.
Consonants Bird, Sonya. 2004. Lheidli intervocalic consonants: Phonetic and morphological effects. Journal of the International Phonetic Association 34(1). 69–91. Kingston, John. 1985. The phonetics and phonology of the timing of oral and glottal events. Berkeley CA: University of California Berkeley dissertation. Kingston, John. 1990. Articulatory binding. In John Kingston & Mary E. Beckman (eds.), Between the grammar and physics of speech, 406–434. (Laboratory Phonology I). Cambridge: Cambridge University Press. Stonham, John & Eun-Sook Kim. 2008. Labialization in Nuuchahnulth. Journal of the International Phonetic Association 38(1). 25–50.
Vowels Herrick, Dylan. 2011. On Comanche’s central mid vowel. International Journal of American Linguistics 77(3). 373–396. Oberly, Stacey & Viktor Kharlamov. 2015. The phonetic realization of devoiced vowels in the Southern Ute language. Phonetica 72. 1–19. Russell, Kevin. 2008. Sandhi in Plains Cree. Journal of Phonetics 36. 450–464.
38
Sonya Bird, Rae Anne Claxton, and Tess Nolan
Syllables Bird, Sonya & Ewa Czaykowska-Higgins. 2016. Parsing Salish consonant clusters. In Martin J. Ball & Nicole Müller (eds.), Challenging Sonority: Cross-linguistic evidence, 159–199. Sheffield, UK: Equinox Publishing Home. Hargus, Sharon & Virginia Beavert. 2002. Yakima Sahaptin clusters and epenthetic [i]. Anthropological Linguistics 44(3). 231–277.
Word-level prosody/intonation Gordon, Matthew. 2004. A phonological and phonetic study of word-level stress in Chickasaw. International Journal of American Linguistics 70(1). 1–32. Gordon, Matthew & Edmundo Luna. 2004. An intergenerational study of Hupa stress. Annual Meeting of the Berkeley Linguistics Society (BLS) 30. 105–117.
Heather Bliss, Sonya Bird and Bryan Gick
2 Articulatory phonetics
Abstract: While the field of articulatory phonetics dates back millennia, throughout most of the field’s history, investigations have been largely impressionistic. With many relatively recent and continually evolving technological innovations, the tools and methods we can use to analyze speech articulation have opened up the space for new discoveries and approaches. Moreover, these technologies are increasingly affordable and portable, meaning that they can be used “in the field,” in community contexts with speakers of Indigenous languages in North America. As new technologies develop, our awareness is concurrently growing about the importance and necessity of conducting ethical and engaged research with Indigenous communities. In this chapter we survey tools and methods that have been used for articulatory research with Indigenous languages in North America, with a focus on how these tools and methods can enrich and enhance community-based research. Articulatory data can be an important part of both language documentation and language revitalization, as it can provide detailed information about the complexities and variation in linguistic sound systems, which can be used to assist learners in reclaiming their languages. As technologies and perspectives continue to evolve, the potential for articulatory phonetic research to inform language revitalization practices continues to grow.
2.1 Introduction Often, we talk about speech as sound, focusing on the acoustic signal associated with it (see Bird, Nolan, and Claxton, this volume). To understand sound though, we need to have an understanding of the movements responsible for creating it. Articulatory Phonetics is the subfield of linguistics focused on studying the articulations underlying speech sounds: movements of the tongue, the jaw, the larynx, and other structures used in speech production. Documenting the articulatory characteristics of speech tells us about the ways that humans are able to manipulate their vocal apparatus to create sounds unique to their languages. It also helps with teaching and learning new sounds not present in the language(s) we are familiar with. As detailed throughout this chapter, articulatory phonetic studies complement acoustic phonetic studies by providing detail on how we to produce speech sounds by manipulating our articulators. Unlike acoustic phonetics (see Bird, Nolan, and Claxton, this volume), articulatory phonetics requires technology that has historically been relatively inaccessible, particularly for use in Indigenous communities, because of cost, lack of portability, and the need for specialized expertise. However, as technology becomes increasingly affordable, portable, and user-friendly, new opportunities are emerging for exploring the articulatory properties of Indigenous languages and for developing resources for communities https://doi.org/10.1515/9783110600926-002
40
Heather Bliss, Sonya Bird and Bryan Gick
engaged in language documentation and revitalization, in particular ones focused on developing oral proficiency among new speakers. In this chapter, we start by introducing the field of Articulatory Phonetics, and how it can be a useful component of Indigenous language documentation (Section 2.2). We then describe some of the tools used for articulatory phonetic research, and how they can be applied in Indigenous language research (Section 2.3). We end with a brief discussion of areas of future research, in terms of articulatory features to explore as well as methods for doing this following collaborative, community-engaged models of research (Section 2.4).
2.2 What is articulatory phonetics and why is it important in Indigenous language research? 2.2.1 A brief history of the field of articulatory phonetics Articulatory phonetics has a long history, going back over two millennia. Many of the familiar terms and concepts we use today to describe speech sounds – including “articulator,” “place of articulation,” “stop,” “fricative,” “vowel,” and many others – were reintroduced by 19th-century British School phoneticians from the works of ancient Indian scholars such as Pāṇini and Patañjali (see Allen 1953). Notwithstanding its long history and technological advances, the ways we use articulatory phonetics to describe languages has not changed much in the course of its 2500-year history. Articulatory phonetic descriptions have long been a core part of the grammatical descriptions of Indigenous languages (e. g., Goddard 1907), though the phonetic portions of traditional grammars have often been very brief and impressionistic. Even recently, many descriptions of how speech sounds are produced seem more impressionistic than scientific. One survey of twenty descriptive grammars published between 1989 and 2000 revealed that the mean percentage of content devoted to sound systems is under 10 % and the large majority of this is phonological, not phonetic (Maddieson 2002). In recent years, however, technological advances have begun to allow us to “look inside” the moving vocal tract to gain a better understanding of the kinds of movements we make when we speak. As Rice (2006: 239) observes: “Many recent grammars…contain far more discussion of phonetics than do older grammars. …Technological advances have allowed phonetics to become a core part of linguistic training in a way that was difficult even fifteen years ago.” The particular technologies researchers have used have had an important influence on the shape of modern articulatory phonetics. Unlike the study of acoustic phonetics (see Bird, Nolan, and Claxton, this volume), in which large quantities of data can be easily recorded and analyzed, articulatory phonetics often relies on more involved, labor-intensive procedures for recording and analyzing data. As a result, articulatory
Articulatory phonetics
41
phoneticians are more likely to produce smaller case studies, or studies with a focus on individual variability. In some ways, this research tradition is well suited to Indigenous languages with small populations of speakers; unlike acoustic studies, there is less of a need to justify having small participant groups, as is true for all sciences with case-study design (see Bird 2011). We are better equipped today than ever before to tackle some of the big questions in the field of articulatory phonetics. Some of these include: What is the relationship between sounds and movements in speech? What kinds of basic physical processes or mechanisms do we use when we produce speech? Where do these come from and how do they develop in children? How do we coordinate the complex movements of speech sounds and transitions between speech sounds? What are the range and sources of variation in articulatory patterns? How can an understanding of the physical mechanisms of speech assist language learners (e. g., how can speech articulations be visualized to assist language teaching and learning)? To address any of these questions, it is essential to investigate a wide diversity of languages, particularly those that have not been traditionally included in typological studies. As an example, the very idea that “sound” is fundamental to speech is based on observations of widely spoken and widely studied mainstream languages. It was only recently that an instrumental study of two Indigenous languages in North America – Oneida (Iroquoian) and Blackfoot (Algonquian) – showed that entire syllables can be produced with no sound at all (see Gick et al. 2012). This and other studies will be described in more detail below.
2.2.2 Articulatory phonetics and Indigenous language documentation and revitalization Articulatory phonetic research can be expensive, laborious, and in some cases invasive for speakers. Why should we do this work then? From a theoretical standpoint, documenting the articulatory mechanisms used in languages across the world expands our understanding of the range of possibilities for human speech. Indigenous languages in North America provide numerous examples of the articulators in action through their use of ejectives, pharyngeals, soundless articulations, and so much more (McDonough and Whalen 2008; Gordon 2017). These languages are therefore immensely valuable in terms of what they can teach us about the human capacity for speech. From a documentation standpoint, articulatory recordings (e. g., video, ultrasound) provide an extra layer of detail beyond what is available in audio recordings, just as audio recordings provide an extra layer of detail beyond written records. From a language revitalization standpoint, articulatory phonetic research can also provide valuable tools for supporting teaching and learning pronunciation (see also Bird, Nolan, and Claxton, this volume). In his discussion of phonetic documentation, Maddison (2002: 420) states:
42
Heather Bliss, Sonya Bird and Bryan Gick
An ideal to have in mind is to provide enough phonetic information that a reader would be able to sound like a native speaker. A grammar writer might imagine he or she is describing the speech of the last generation of fluent speakers and aims to do it well enough that their grandchildren will be able to learn to sound like them and react like them to hearing the language spoken.
Currently in North America, many language revitalization efforts are focused on creating new speakers (see, e. g., McIvor and McCarty 2016). These speakers take very seriously their responsibility of pronouncing their language(s) in a way that honors their Elders (Bird and Kell 2017; Jenni et al. 2017), and documentation of the quality described by Maddison can provide valuable resources for them. Immersion-based approaches like Total Physical Response (TPR; Asher 1977) and Where Are Your Keys (WAYK; https:// whereareyourkeys.org) provide excellent opportunities for learners to hear and speak in a holistic way, but other approaches are necessary to tackle the details of pronunciation once learners are ready for them. One approach that is gaining increasing popularity, and that is directly based on articulatory phonetic research, is speech visualization, or “seeing speech,” which involves looking directly at what speakers are doing with their articulators (e. g., their tongues in ultrasound imaging) as they speak (see Bliss, Abel, and Gick 2018; Bliss et al. 2018). This approach complements similar work in acoustic phonetics, described in Bird, Nolan, and Claxton (this volume). We know that speech processing involves multiple modalities (e. g., Catford and Pisoni 1970; Navarra and Soto-Faraco 2007), and incorporating them into pedagogical techniques can be of great benefit to learners, especially when it comes to understanding articulations that are not easily accessible because they occur deep in the vocal tract, e. g., “back of the mouth” sounds like uvular and pharyngeal consonants.
2.2.3 Articulatory phonetics as the basis for phonetic descriptions and charts Linguists generally describe speech sounds based on how they are articulated. Consonants are normally described in reference to three primary dimensions: voicing, place, and manner. In its most basic version, voicing refers to whether the vocal folds are vibrating (voiced sounds) or not (voiceless sounds). Voicing also covers other, more complex articulations in the larynx, like those involved in making ejective stops and glottalized resonants (see Bird, Nolan, and Claxton, this volume, Section 1.3.2.2), or voice qualities like creaky voice (vocal fry) or breathy voice (see Figure 8). Place refers to where in the vocal tract the sound is being made. Figure 1 provides a schematic view of the vocal tract, labeled with the key landmarks in speech articulation, from the lips at the front/ top to the larynx at the back/bottom. Manner refers to how constricted the sound is, and consequently how freely air can flow through the vocal tract during the sound, e. g., stop consonants are sounds where the articulators are completely closed, stopping airflow altogether; resonant consonants are ones where the articulators remain relatively far apart, allowing air to flow freely through the vocal tract.
Articulatory phonetics
43
Fig. 1: Midline view of the human vocal tract showing some important anatomical landmarks for speech. (Source: Adapted from an image by Patrick J. Lynch, medical illustrator (CC BY 2.5 [http://creativecommons. org/licenses/by/2.5]), via Wikimedia Commons.)
As a shorthand for phonetic descriptions of sounds, linguists use the International Phonetic Alphabet (IPA). For example, the symbol [k] (also written in the Roman alphabet as ) is used for the sound we describe as a voiceless velar stop, e. g., in the English word kayak [kajak], borrowed from the Inuit word qayak [qajaq]. It is produced without vocal fold vibration (voicing = voiceless), by bringing the tongue body up against the back of the hard palate (place = velar) completely, thus completely stopping the airflow for a short time (manner = stop). Note that in the Inuit word, the first and last consonants are uvular [q] sounds, pronounced further back in the mouth than English [k]. This kind of sound adaptation is very common, and it happens naturally when a language borrows a word from another language that has an unfamiliar sound in it (Andersson, Sayeed, and Vaux 2017). Figure 2 summarizes the most common consonants found in the world’s languages, organized in an IPA chart. The columns correspond to places of articulation, from the front of the vocal tract on the left to the back of the vocal tract on the right. The rows correspond to manners of articulation, from the most constricted at the top to the least constricted at the bottom. Within each cell, the sound on the left is voiceless and the one on the right is voiced.
44
Heather Bliss, Sonya Bird and Bryan Gick
THE INTERNATIONAL PHONETIC ALPHABET (revised to 2015) CONSONANTS (PULMONIC)
© 2015 IPA
Bilabial Labiodental
Dental
Alveolar Postalveolar Retroflex
Palatal
Velar
Uvular
Pharyngeal
Glottal
Plosive Nasal Trill Tap or Flap Fricative Lateral fricative
Approximant Lateral approximant
Symbols to the right in a cell are voiced, to the left are voiceless. Shaded areas denote articulations judged impossible.
CONSONANTS (NON-PULMONIC)
VOWELS
Fig. 2: Basic (pulmonic) consonant chart (Source: www.internationalphoneticassociation.org/content/ipaBack Front Central Clicks Voiced implosives Ejectives chart, available under a Creative Commons Attribution-Sharealike 3.0 Unported License. Source: Copyright Close Bilabial Association)Examples: 2018Bilabial International Phonetic Dental
Dental/alveolar
Bilabial
(Post)alveolar
Palatal
Dental/alveolar
Close-mid
Palatoalveolar Velar Velar Vowels are also described articulatorily, usually in terms of tongue height and backOpen-mid Alveolar lateral or high Uvular Alveolar fricative 1 Close ness. vowels are ones with the tongue high up close to the palate; open or low vowels are ones with the tongue low against the floor of the mouth. Front vowels OTHER SYMBOLS Open have the tongue relativelyAlveolo-palatal far forward in the mouth; back vowels have the tongue Voiceless labial-velar fricative fricatives Where symbols appear in pairs, the one to the right represents a rounded vowel. or relatively far back. Figure 3 provides the vowels in a space that corresponds more Voiced labial-velar approximant Voiced alveolar lateral flap lessVoiced to these tongue-based distinctions. The vowel /i/, for example, is a close (or high) SUPRASEGMENTALS labial-palatal approximant Simultaneous and front vowel. A third dimension used for categorizing vowels is rounding, which refers Primary stress Voiceless epiglottal fricative Affricates and double articulations stress vowel; it to whether orfricative not the lips are rounded. For example, /y/ is also a Secondary close front Voiced epiglottal can be represented by two symbols joined by a tie bar if necessary. Long can also differ in differs from Epiglottal plosive /i/ in that it is rounded whereas /i/ is unrounded. Vowels Half-long terms of whether or not air can get out through the nose; in nazalized vowels, the velum DIACRITICS Some diacritics may be placed above a symbol with a descender, e.g. Extra-short is lowered, so that air escapes through the nose as well as through the mouth. Nasalized Voiceless Breathy voiced Dental Minor (foot) group vowels can be found inCreaky Dene languages, as Apical well as in languages like French. Voiced voiced Major (intonation) group The articulatory categories introduced here form the basis of phonological categoAspirated Linguolabial Laminal Syllable break ries,More discussed in Fitzgerald (this volume). Nasalized In our work with community members, we rounded Labialized Linking (absence of a break) have found that introducing the articulatory basis of sounds can help them grasp not Less rounded Palatalized Nasal release TONES AND WORD ACCENTS only how to pronounceVelarized these sounds, but also why they pattern the way they do in the Advanced Lateral release LEVEL CONTOUR Extra language. To this end, it can be useful to create schemas of the sound inventories that or or Rising Retracted Pharyngealized No audible release high areCentralized tailored to specificVelarized languages. Figure 4 provides an example from Hul’q’umi’num’, High Falling or pharyngealized High Mid Island and a Central Salish language spoken on the east side of Southern Vancouver on rising Mid-centralized Raised ( = voiced alveolar fricative) Low Low rising theSyllabic adjacent islands. Hul’q’umi’num’ 34bilabial consonants, 21 of which are not found in Lowered ( has = voiced approximant) Extra Risingfalling low English. Being able to Advanced summarize Non-syllabic Tongue Rootthem using visuals like this is particularly helpful, Downstep Global rise given all the possibilities. Rhoticity Retracted Tongue Root Upstep Global fall Typefaces: Doulos SIL (metatext); Doulos SIL, IPA Kiel, IPA LS Uni (symbols)
1 Note that describing vowels only in terms of what the tongue is doing (linguo-centric model) is a simplification of speech; in fact, jaw height and pharynx positioning also play important roles in vowel production and speech more generally (Esling 2005).
Articulatory phonetics
45
Fig. 3: Vowel chart within the vocal tract. (Source: Vocal tract image adapted from https://commons. wikimedia.org/wiki/File:Midsagittal_diagram_ unlabeled.svg. Public domain. Vowel chart image from www.internationalphoneticassociation.org/ content/ipa-chart, available under a Creative Commons Attribution-Sharealike 3.0 Unported License. Source: Copyright 2018 International Phonetic Association.)
Fig. 4: Hul’q’umi’num’ (Coast Salish) consonant chart showing anatomical landmarks corresponding to places of articulation. Red arrows indicate the points of contact (most often between tongue and palate) for different sounds. Consonants are written in the Hul’q’umi’num’ alphabet. Vocal tract image source as in Figure 3.
46
Heather Bliss, Sonya Bird and Bryan Gick
2.3 Tools for articulatory phonetics and their applicability in Indigenous language research and revitalization There is not enough articulatory data from North American languages to say anything broadly typological (but see Gordon 2017 for some generalizations based on articulatory and acoustic data). Rather than survey languages and/or classes of sounds, in this section, we survey tools and methods, with a focus on why they are useful and how they can be used effectively and ethically. For more general resources describing articulatory phonetic methodologies, we refer interested readers to Gick, Wilson, and Derrick (2013), Podesva and Zsiga (2013), and Ladefoged (2003).
2.3.1 Video Video is increasingly used as a basic recording method in language documentation to capture extralinguistic data such as gestures and facial expressions, as well as metadata variables surrounding the recording situation such as time of day, recording location, other people present, etc. (Ladefoged 2003; Margetts and Margetts 2012). However, video is easily overlooked as a tool for articulatory phonetic documentation, research, and pedagogy, despite being the most affordable and accessible tool described in this paper. Video can be used for measuring lip movements and lip aperture, which can inform investigations of vowel height differences (Gick et al. 2012; Bliss and Gick 2017), as well as distinctions between labial consonants and degrees of lip rounding (see Gordon, Martin, and Langley 2015 on Koasati). In addition to its utility as a research tool, video is useful as a pedagogical tool as it can be used to make data and images from ultrasound (and possibly other tools) more interpretable for untrained learners (see Bliss et al. 2018). Pedagogical applications are discussed in more detail in section 2.3.2 below. Here we focus on video as a research tool and briefly describe methods that have been used to measure lip aperture in two Indigenous languages in North America. Gick et al. (2012) report on two studies that measured lip aperture in Oneida and Blackfoot as a way to investigate “soundless” vowels – sounds that are produced by moving the lips and tongue but that do not make any audible sound, and so cannot be measured acoustically. For example, the correct way to say the Oneida word akohta’ (meaning ‘her shoes’) is to pronounce the first two syllables [agoht] and then to mouth the final vowel [a] while making no sound at all. If you were just listening (or, say, talking on the phone) to a speaker of Oneida, without looking at the speaker’s face to see the mouth opening for the final [a], you might well think the word ended with a final [t]. The usual way to write these “soundless” vowels (and sometimes consonants) in Oneida is with an underline, so that the word for ‘her shoes’ would be written akohta. One speaker
Articulatory phonetics
47
of Oneida (according to Gick et al. 2012) described these sequences as: “you say it, but you don’t say it.” In the Oneida and Blackfoot studies, video was a valuable tool for capturing “soundless” sequences. In both studies, a camcorder2 mounted on a tripod was focused on a speaker’s mouth as they produced words or phrases. Video editing software was later used to extract frames from the video (e. g., at the time corresponding to the midpoint of the vowel), and the lip opening was measured, either according to area (Oneida) or according to the vertical distance between reference points in the upper and lower lips (Blackfoot). Bliss and Gick (2017) used a similar method for investigating speaker variation in vowel articulations in Blackfoot, but they measure vertical and horizontal distances as a means to determine lip aperture (see Figure 5 below). The choice between these different measures depends on camera angle. In all three studies, pixels were converted to millimeters using known quantities in the frame; Maddieson (2002) suggests having a ruler in the frame for such purposes.
Fig. 5: Horizontal and vertical lines measuring lip aperture in Blackfoot
2 Margetts and Margetts (2012) note that even some professional-grade video cameras do not have professional audio inputs, and that most or all consumer cameras do not either. They advise using a mid-range “prosumer” camera with a secondary audio recording device to capture audio and visual streams in basic language documentation. For the purposes of measuring lip movements, a lower range consumer camera will likely suffice. As Smartphones and other readily accessible video-recording technologies become available, we expect the opportunities for using video in articulatory phonetic research will continue to grow.
48
Heather Bliss, Sonya Bird and Bryan Gick
2.3.2 Ultrasound There is a long history of ultrasound being used as a speech research tool, but only in the past 15–20 years has it been used in research and documentation of Indigenous North American languages. Historically, ultrasound was expensive and confined to a laboratory, but it has become increasingly portable and affordable over the past two decades (Gick 2002); there are now a number of handheld devices on the market that can connect to a laptop computer or smartphone. These developments have led to questions about how ultrasound could be used in the community-based settings effectively and ethically (see Gick, Bird, and Wilson 2005), and have opened the door for ultrasound to be used outside the laboratory and in Indigenous communities for articulatory phonetic documentation and research. With ultrasound, a transducer is held underneath a speaker’s chin, and a high-frequency sound is transmitted through the tongue and reflected back to the transducer, creating a 2-dimensional image of the tongue. Figure 6 provides two still images extracted from an ultrasound video, comparing the tongue position of the velar stop [k] (left) and the uvular stop [q] (right) (see Figures 2 and 4 above). The white line corresponds to the surface of the tongue, with the tongue tip at the right. The [k]~[q] contrast can be difficult to hear and to pinpoint acoustically (see Bird, Nolan, and Claxton, this volume, Section 1.3.2.2), but it is easy to see with ultrasound; the tongue body comes straight up for [k] but up and back (towards the uvula) for [q]. This contrast is very common in the languages of the Pacific Northwest (e. g., as in the Halq’eméylem (Coast Salish) words kálti [kalti] ‘candy’ and qá:l [qa:l] ‘to steal’), and we have found that “seeing” the articulation of the two sounds using ultrasound is beneficial for learners (Bird & Miyashita 2019).
Fig. 6: Ultrasound tongue contours of [k] (left) vs. [q] (right). Tongue tip is on the right. Red dots indicate where the tongue is highest in the mouth.
Ultrasound is particularly useful for imaging sounds that are differentiated along the mid-sagittal plane (side view of the head, as in Figure 1): vowel articulations, the place of articulation of coronal and dorsal consonants, and consonants that require coordination of different parts of the tongue such as lateral (“l-like”) and rhotic (“r-like”) sounds. For all of these sounds, it is easy to see the tongue contour and to get a good sense of what is hap-
Articulatory phonetics
49
pening where in the mouth. It is also possible to turn the transducer 90 degrees and image the tongue in the cross-sectional plane (front view of the head). In this view, it is possible to see what the sides and centre of the tongue are doing, which is especially relevant for differentiating lateral sounds from coronal sounds, as in Hul’q’umi’num’ (Coast Salish) [s] (e. g., in situn [sitǝn] ‘basket’) from [ɬ] (e. g., in lheel [ɬeːl] ‘to go ashore’) (see Figure 2). Ultrasound is non-invasive, and speakers who use ultrasound often report that they enjoy the process of seeing their tongues “in action” (see, e. g., Bliss et al. 2018 for discussion). The only drawback is that not all speakers image equally well. To enhance the ultrasound image, a thin layer of conductive gel is applied to the transducer. Because ultrasound imaging produces more or less clear images for different people, some speakers may also need to apply gentle pressure between the transducer and the neck, which may create mild discomfort for some speakers. To generate a clear image and minimize movements that can skew measurements, the speaker’s head and the transducer should ideally be stabilized, which can be achieved, for example, with a foam head rest affixed to a wall (see Gick et al. 2012) and an adjustable arm clamped to a chair to support the transducer (see Bliss et al. 2018). While there are numerous more invasive and more expensive stabilization tools and techniques that have been used by researchers, working in community settings – particularly with Elders – may require compromising on stabilization for the sake of the speakers’ comfort (see Percival, Bliss, and Schellenberg 2017). Moreover, as documented by Gick, Bird, and Wilson (2005), a simple stabilization set-up (such as that described above) can suffice to collect reliable data in the field. In Indigenous languages in North America, ultrasound has been used to document and describe a wide range of sounds and phenomena in a diverse range of languages. In combination with video (see 2.3.1), Gick et al. (2012) used ultrasound to study soundless vowels in Blackfoot and Oneida. As mentioned above, in these languages, vowels are articulated without any accompanying acoustic signal, meaning they can only be studied with articulatory methods. In combination with video recordings of speakers’ lips (see section 2.3.1), ultrasound recordings were used to demonstrate that different soundless vowels have distinct tongue shapes. Bliss and Gick (2017) expand on the 2012 study to include four additional speakers from two different Blackfoot dialects. Although the speakers exhibited certain differences in their tongue contours, all maintained a robust distinction between soundless /a/ and soundless /i/ vowels, as in the words ki’sómma [kiʔsumːa] ‘moon (proximate)’ and ki’sómmi [kiʔsumːi] ‘moon (obviative)’.3 Iskarous, McDonough, and Whalen (2012) also investigate variation – but amongst the productions of the velar fricative /x/ by a single speaker of Navajo (Athabaskan). They found that the tongue shape for this consonant varied considerably across tokens,
3 Proximate and obviative – the categories distinguished by soundless –a and –i in Blackfoot – are morphosyntactic categories observed in all Algonquian languages. See Chapter 59: Algonquian for details.
50
Heather Bliss, Sonya Bird and Bryan Gick
and was influenced by the following vowel (e. g., /xi/ sequences were further front than /xa/ sequences). Ultrasound has also been used to investigate VC and CV sequences in Salish languages, particularly in cases where the consonant has a co-articulatory effect on a neighboring vowel (see also Bird, Nolan, and Claxton, this volume, Section 1.3.2.1 on vowel-consonant co-articulation). Interior Salish languages are well-attested for having vowel retraction: uvular and pharyngeal consonants produced at the far back of the mouth trigger the preceding vowel to be retracted, i.e., produced further back in the mouth as well. Retraction is also attested as a secondary feature with some other consonants. McDowell (2004) was the first to conduct an articulatory study of this retraction phenomenon, focusing on lateral (“l-like”) sounds in Montana Salish. She finds that all four laterals exhibit retraction and co-articulatory effects, but they differ according to whether the retraction involves the tongue dorsum (back of the tongue) or the tongue root (see Figure 1 above). Namdaran (2006) and Hudu (2008) investigate retraction in another Interior Salish language, St’át’imcets, documenting articulatory properties of both inherently and secondarily retracted consonants and their co-articulated vowels. Bird, Leonard, and Moisik (2010) and Bird (2012) also look at VC and CV sequences in Salish, but in a different branch of the language family, and from a different perspective. They investigate the articulatory mechanisms that speakers of SENĆOŦEN (Coast Salish) use to produce /iq/ and /qi/ sequences, as in the words SḴITEW [sqitǝw] ‘mermaid’ and ȾIḴT [tθ’iqt] ‘ivory-billed woodpecker.’4 These sequences involve an “articulatory conflict,” because uvular /q/ is produced far at the back of the mouth, and high vowel /i/ is produced at the front. Bird (2012) finds that one speaker uses a tongue rolling strategy that allows him to pronounce the sequences without compromising the vowel or the uvular closure. In general, articulatory studies tend to be more involved than acoustic studies, both in terms of data collection and data processing. For this reason, ultrasound studies are sometimes done as focused follow-ups to broader acoustic studies, to add a deeper understanding of a particular aspect of pronunciation that is difficult to characterize from acoustic analysis. In the case of Bird (2012), ultrasound imaging explained why, in some cases, the speaker appeared to be pronouncing /qi/ and /iq/ without making any adjustments to ease his pronunciation (see Bird, Nolan, and Claxton, this volume, Section 1.3.2.1). It turned out he was making adjustments, but they were not perceptible in auditory or acoustic analysis. This kind of discovery is important not only for documentation and description, but also for language pedagogy, as it can provide learners with tips on how to overcome articulatory conflicts in their language. Towards the view of contributing to pedagogical materials for language revitalization, Bliss et al. (2018) describe their collaborations with three Salish communities
4 The SENĆOŦEN alphabet was created by the late Dave Elliott Sr. It uses all capital letters in combination with 4 diacritics (https://wsanecschoolboard.ca/sencoten-language/).
Articulatory phonetics
51
in British Columbia to develop ultrasound-enhanced videos that can help learners of SENĆOŦEN, Halq’eméylem, and Secwepemc to pronounce challenging sounds, particularly “back-of-the-mouth” sounds. These videos combine a profile view of a speaker’s head with an ultrasound image of their tongue, effectively making the tongue visible and salient for the learner. A screenshot of one of the videos is given in Figure 7; see also https://enunciate.arts.ubc for more ultrasound video resources, including an interactive IPA chart. Percival, Bliss, and Schellenberg (2017) adopt Bliss et al.’s method to produce ultrasound-enhanced videos of Hän (Athabaskan) speakers. This project focused not only on developing pedagogical resources but also documenting the articulatory properties of the complex consonant inventory, which includes 5–6 coronal contrasts and numerous affricates. With a more direct focus on pedagogy, Bird and Miyashita (2019) describe using ultrasound and other phonetic tools with Blackfoot (Algonquian) and Hul’q’umi’num’ (Coast Salish) language learners as a pronunciation aid. In our experiences working with Indigenous community members using ultrasound imaging, we found that it can be a useful tool for developing awareness of what is going on inside the mouth during speech production; learners are enthusiastic to use it and they often reflect on how helpful it is to see inside the mouth. Whether or not it actually benefits their pronunciation (specifically of the target sounds and/or more generally) is yet unclear, but studies of other languages (such as Cantonese, see Bliss et al. 2017) suggest that it does.
Fig. 7: Ultrasound-enhanced video screenshot for SENĆOŦEN.
2.3.3 Endoscopy In linguistic research, endoscopy – and, more precisely, laryngoscopy – is used to image the (lower) pharynx and larynx, to study different voice qualities (Esling et al. 2019), as well as sounds that are pronounced deep within the throat, with engagement of the pharyngeal and laryngeal structures (Esling, Fraser, and Harris 2005). There are two types of
52
Heather Bliss, Sonya Bird and Bryan Gick
endoscopes: rigid and flexible. Rigid endoscopes have a camera and light at the end of a rigid tube (approximately the size of a pencil). They are inserted into the very back of the mouth, such that the camera looks down at the pharynx and larynx. Flexible (fibre optic) endoscopes have much thinner, more flexible tubes. They are inserted through the nose and fed through the nasal passage and down through the pharynx as far as just above the larynx (or anywhere above this). Fibre optic endoscopes are more versatile in terms of their placement, and they interfere much less with articulation (Esling, Fraser, and Harris 2005; Podesva and Zsiga 2013). They are therefore generally more useful as a tool for characterizing speech articulations. Figure 8 provides an example of fibreoptic laryngoscopy data, showing the laryngeal articulator during a period of breathy voicing. Further images and explanations of laryngeal structures, states of the larynx, and phonation types may be found in Esling et al. (2019); companion materials, including videos, are available for free through the Cambridge University Press website.5
Fig. 8: Example of fibreoptic laryngoscopy data. Image shows a slight tightening of the laryngeal articulator during a period of breathy voicing. (Courtesy of John Esling, Department of Linguistics, University of Victoria, Canada).
Up until now, linguistic endoscopy research has generally been done in a laboratory setting. However, a quick Google search (conducted in spring 2019) shows that handheld endoscopy systems are now on the market and relatively inexpensive, potentially allowing for research in communities as well. Nonetheless, compared to other imaging techniques like video and ultrasound, endoscopy is somewhat invasive and, at least at some universities, ethics boards require that flexible endoscopes be inserted by a
5 From the following link, click on Resources > Resources > Companion files https://www.cambridge. org/ca/academic/subjects/languages-linguistics/phonetics-and-phonology/voice-quality-laryngealarticulator-model?format=HB.
Articulatory phonetics
53
medical professional. In short, serious thought should be put into deciding whether to incorporate endoscopy into articulatory phonetic research. The only endoscopic studies of Indigenous languages in North America that we know of have focused on Nlaka’pamux (Interior Salish) and Nuu-chah-nulth (Wakashan). These two language families feature pharyngeal and glottal sounds, which are produced very deep in the throat (similar to some of the Arabic sounds, for example), and which would be very difficult to accurately describe without directly looking at the articulatory mechanisms involved (Carlson, Esling, and Harris 2004; Esling 2003; Esling, Fraser, and Harris 2005). For these languages, endoscopy has been an extremely valuable component of language documentation. For example, Esling, Fraser, and Harris (2005) examines glottal stops, pre-glottalized resonants, and pharyngeal consonants in the Ahousaht dialect of Nuu-chah-nulth (Wakashan). The study is based on the pronunciation of two speakers, co-author Katie Fraser and her sister Łuuta Qamiina. The paper itself includes several sequences of still images (extracted from video) showing the articulatory features of Nuu-chah-nulth’s pharyngeal, glottalized, and glottal sounds, including which structures are involved (the arytenoid cartilages, the vocal folds, the ventricular folds (false vocal folds), the epilaryngeal tube, and the glottis) and their degree of constriction across sounds. Esling, Fraser, and Harris’ findings have made major contributions to our understanding of how laryngeal and pharyngeal structures contribute to speech, and of the vast range of configurations they can adopt – and consequently the vast range of sounds they can generate.
2.3.4 Palatography Palatography involves tracking points of contact between the tongue and the roof of the mouth, either at a single point in time (static palatography) or over time (dynamic palatography) (Anderson 2008). Static palatography has been used for many decades to study speech sounds, e. g., by Goddard (1907, 1912) in his work on Dene languages Hupa and Dene Sułiné. It involves painting a non-toxic, dark substance (often a mix of olive oil and powdered charcoal) on the surface of the tongue, and then using a mirror to observe where on the roof of the mouth the substance has left traces after articulating a specific sound (Anderson 2008; Ladefoged 2003).6 In our experience, the methodology is somewhat tricky: words have to be chosen carefully, so that only the sounds of interest create contact. The procedure is also fairly time-consuming since the tongue must be repainted after each utterance. One advantage of static palatography is that it is inex6 A related methodology is linguography, which involves painting the same substance on the roof of the mouth and inside surfaces of the upper teeth, instead of the tongue. Anderson (2008) notes that this method may be challenging, as the palate is less accessible to paint than the tongue (and may be ticklish for some speakers), but it may also be simpler than palatography as it does not require the use of a mirror to view the contact points.
54
Heather Bliss, Sonya Bird and Bryan Gick
pensive and portable. We have found that younger speakers and learners really engage in palatography sessions (in one session that co-author Bird was involved with, learners were streaming live on Facebook!). It is a good option for articulatory research done outside of laboratory settings and when the research budget is small. Figure 9 provides a close-up of a static palatography set-up, with the mirror showing the charcoal traces of tongue contact at the front and sides of the palate (for the English sound [θ], as produced by an 8-year old speaker).
Fig. 9: Static palatography image of charcoal traces left by the tongue on the palate during [θ] as pronounced by an 8-year old English speaker, seen through a mirror inserted into the mouth.
Another advantage of palatography is that it records contact points between the tongue and the hard structures (teeth and palate) above it, something that ultrasound, for example, is unable to do. In some cases, being able to image the points of contact on the tongue and on the palate simultaneously is very useful. For example, in Coast Salish languages, /θ/ (often spelled
) is articulated differently than in English, even though the sounds have generally been transcribed the same way in the IPA, and the contrast between /θ/ and /s/ is quite difficult to perceive for people unfamiliar with these sounds (see Mellesmoen 2018 on Comox-Sliammon). To teach learners to pronounce these sounds, it would be very handy to know what part of the tongue makes contact with what part of the upper teeth/palate, something that palatography (as opposed to ultrasound) can help with. Dart (1991) used static palatography to describe Tohono O’odham (Uto-Aztecan) sounds that involve raising the tongue front, as in the words a’ada [ˈaʔad̪a] ‘a kind of ritual’ and aḍawi [aˈdawi̥ ] ‘buffalo gourd.’ Her findings confirmed, for example, that the “dental” (front) /d̪/ and the “alveolar” (back) /d/ differed in the following ways: (a) which part of the tongue made contact with the roof of the mouth – tongue tip and blade for /dental d̪/ and tongue tip only for /alveolar d̪/, (b) where contact was made – at the upper teeth for /d̪/ and behind the upper teeth for /d/, and (c) how large the area of contact was – greater (longer) for /dental d̪/ than for /alveolar d̪/. Avelino and Kim (2003) also use static palatography to investigate similar contrasts in Pima, a related language. The detailed descriptions generated by Dart’s (1991) and Avelino and Kim’s Articulatory phonetics 55 (2003) research could be very useful for explaining to new speakers how to articulate these sounds, which are important to distinguish in Tohono O’odham, but correspond to a single /d/ sound in English and Spanish (the colonial languages spoken in Tohono O’odham territory). Dynamic palatography, or electropalatography (EPG), involves creating customized false palates for individual speakers that have electrodes embedded in them in a gridlike fashion (see Figure 10). These palates are inserted in the speakers’ mouths, and track contact over time within each cell of the grid (Ladefoged 2003; Gick, Wilson, and Derrick 2013). Having to create custom palates is often not feasible, for various reasons such as cost and time, and so EPG has not often been used in-community. No studies exist that have used EPG to study North American languages, but it has been used to investigate Aboriginal languages in Australia (Tabain, Fletcher, and Butcher 2011). Many Australian languages have extensive sets of sounds articulated with the front of the tongue against the roof of the mouth, which differ in precisely which part of the tongue makes contact, and where along the roof of the mouth. EPG has provided valuable information on the details of these sounds; again, this information can be used in pedagogical applications as well as to broaden our understanding of phonetic typology. Fig. 10: A custom palate (left) and tongue-palate contact patterns (right) using electropalatography. The visualizations of contact patterns, from the Kay Palatometer Database CSL 4333, indicate (using small blue squares) where the tongue is touching the palate for a male English speaker during (middle image) and after (right image) the release of an alveolar stop consonant [t]. 2.3.5 Aerodynamic methods Aerodynamic methods track the airflow and/or air pressure inside a speaker’s mouth during the production of speech sounds. This is particularly helpful for determining airstream mechanism, e.g, whether a given sound is ejective (produced with an explosive release of pressure created by a glottalic airstream) or not (pulmonic airstream). Maddieson (2002) notes that the term “glottalized” is used in many phonetic descriptions but is imprecise in that it has been used to refer to both ejective consonants as well as non-ejective consonants preceded or followed by constriction of the glottis. Understand- 56 Heather Bliss, Sonya Bird and Bryan Gick ing what is meant by glottalization in phonetic descriptions can be investigated using aerodynamic methods that measure airflow and air pressure.7 Aerodynamic methods are not new; Goddard (1907, 1912) investigated air pressure in Hupa and Dene Sułiné using a now-archaic tool known as a kymograph. The kymograph (literally ‘wave drawing’) included flexible tubing with a rubber diaphragm on one end and a reed pen on the other, which would make tracings on a rotating drum in response to air pressure changes created by holding the diaphragm against the mouth (oral airflow), larynx (laryngeal activity), or up the nose (nasal airflow). Today, an oral or nasal mask with a plastic tube is still used, but instead of a kymograph recording pressure changes on a rotating drum, these same changes are picked up by a transducer and sent to a computer with specialized software that can digitize them. This kind of system for measuring air pressure/flow may be suitable for community-based research as it is highly portable, but it is costly (see http://www.sciconrd.com/price.aspx) and may require specialized training to interpret the results. Its operation is fairly simple; speakers produce words or phrases with the plastic tube between their lips, the opening of which rests lightly at the roof of the mouth. While not particularly invasive, some researchers report that it can be difficult to position the tube so it is not blocked by the tongue or obstructed with saliva. The nasal mask may be uncomfortable for some speakers, but it can be removed between recordings. In studies of North American languages, aerodynamic methods have been largely used to investigate the properties of consonants that have been described as ejectives (see also Bird, Nolan, and Claxton, this volume, Section 1.3.2.2). For instance, Maddieson, Smith, and Bessell (2001) measure the intra-oral pressure of a class of consonants described as ejective fricatives in words such as x’áax’ [x’aːx’] ‘crabapple’ in Tlingit (Dene) and compare it with that of non-ejective (pulmonic) fricatives in words such as xaak [xaːk] ‘empty seashell.’ They note that ejective fricatives are typologically rare and are not attested in any other North American language. Moreover, there is reason to question whether these sounds are truly ejective; Bessell (1996) notes that they are longer in duration than would be expected for a true ejective fricative, given that the volume of air would be relatively small for a sound produced with glottal closure and the release time would in turn be relatively short. However, the results of Maddieson, Smith, and Bessell’s (2001) study indicate that Tlingit does have ejective fricatives that are distinguishable from regular (non-ejective) fricatives by having a higher intra-oral pressure and a differ- 7 Maddieson (2002) also uses aerodynamic methods to determine that a consonant contrast previously described as “fortis” versus “lenis” in Leggbo (a Niger-Congo language of Africa) is in fact a distinction between singleton and geminate consonants. We know of no similar studies for North American languages, but we note that this methodology may be useful for enhancing phonetic descriptions of some North American languages. Articulatory phonetics 57 ent pattern of release.8 Flemming, Ladefoged, and Thomason (2008) use similar methods to compare ejective and non-ejective stops and affricates in Montana Salish, and they also report higher pressure for the ejectives than non-ejectives. Finally, McDonough and Tucker (2012) use contemporary methods to replicate Goddard’s (1912) aerodynamic study of Dene Sułiné consonants, including the lateral affricate ejective [tɬ’] as in the word [tɬ’iːze] ‘fly.’ Their results are consistent with Goddard’s initial findings; all speakers consistently produced canonical ejectives, supporting the efficacy of Goddard’s kymographic methods and Dene Sułiné’s resistance to sound change, even after 100 years. 2.3.6 Other tools There are many other tools of modern articulatory phonetics that have not yet been brought to bear on studies of the Indigenous languages in North America. For the most part, this is because these require either highly invasive or highly involved laboratorybased procedures. The most common of these procedures involve either point-tracking or imaging techniques. Point-tracking techniques, such as magnetometry, x-ray microbeam, or optical tracking, involve attaching small sensors to different parts of the vocal tract (e. g., tongue, lips, jaw or face) and tracking their movement over time. Depending on the type of tracking device, sensors can be tracked using either a magnetic field (as in magnetometry), x-ray (as in x-ray microbeam), or reflected light (as in optical tracking). These methods all involve expensive, highly specialized equipment that is normally only available at highend research centers, and none are portable. Some of the more intensive imaging techniques used in laboratory-based articulatory phonetics research include x-ray, computed tomography (CT), and structural magnetic resonance imaging (structural MRI). These techniques can produce very detailed images, and some (CT and MRI) can even image in 3 dimensions (“volumetrics”), but they have some serious drawbacks when it comes to investigating Indigenous languages. For example, both x-ray and CT present significant health concerns, while for CT and MRI, speakers need to sit or lie still for an extended period of time. All three methods are generally available only in a clinical hospital or research university setting, and all are expensive and require a team of professionals to operate. 8 An anonymous reviewer questions whether aerodynamic methods can distinguish between true ejectives versus consonant-glottal stop sequences. While Maddieson, Smith, and Bessel (2001) do not discuss this directly, they do note that Tlingit has both types of sounds and that the difference between them is in regards to the timing of the glottal release (see also Howe and Pulleyblank 2001 on Nuu-chah-nulth). Given that aerodynamic methods can track release times, we speculate that they could be used to distinguish between ejectives and consonant-glottal stop sequences. 58 Heather Bliss, Sonya Bird and Bryan Gick 2.4 Conclusions and future directions This chapter has surveyed articulatory phonetic tools and methods that can and have been used in the description, documentation, and revitalization of Indigenous languages in North America. While not a new field, articulatory phonetics is continually evolving, not only in terms of the research technologies employed but also in terms of researchers’ perspectives on community engagement. Regarding the former, as new innovations lead to improved tools and technologies, there is increased potential to conduct articulatory phonetic research in the field and with communities in ethical and effective ways (see Gick, Bird, and Wilson 2005). Regarding the latter, there is a growing awareness of the importance and necessity for community-based articulatory research (e. g., Czaykowska-Higgins 2009; Rice 2010; Fitzgerald 2017; Bird and Miyashita 2019). As technologies become increasingly portable, affordable and user-friendly, Indigenous communities can take the lead in conducting articulatory research on questions of direct relevance to their own linguistic needs and interests. While detailed phonetic documentation is a crucial ingredient in Indigenous language revitalization, so is the creation of new speakers, who can benefit from pedagogical applications of articulatory research to improve their pronunciation and carry their languages forward in an authentic way. As more Indigenous communities engage in articulatory phonetic research and pedagogy, the field will continue to shift and evolve. Acknowledgments: Nitsiko’tahsi’taki, HÍ,SW̱ḴE SI,IÁM, huy tseep q’a’ sii’em’, and thank you to our community and affiliated partners. Blackfoot: Ikino’motstaan Noreen Breaker, late Tootsinam Beatrice Bullshields, Tony Black Water, Natalie Creighton, and Piohkomiaaki Rachel Ermineskin. Halq’eméylem: Strang Burton and Siyamiyateliyot Elizabeth Phillips. Hul’q’umi’num’: Dr. Sti’tu’mat Ruby Peter, Delores Louie, Dr. Donna Gerdts, Hul’qumi’num’ Language & Culture Society, and students in the Hul’q’umi’num’ Language Academy. SENĆOŦEN: late Ivan Morris Sr. and late Ray Sam, Tye Swallow, and teachers and students in the W̱,SENĆOŦEN IST program. Thank you also to the editors and reviewers for their helpful feedback. References Allen, W. Sidney. 1953. Phonetics in ancient India. (London Oriental Series 1). London: Oxford University Press. Anderson, Victoria B. 2008. Static palatography for language fieldwork. Language Documentation & Conservation 2(1). 1–27. Andersson, Samuel, Oliver Sayeed & Bert Vaux. 2017. The phonology of language contact. Oxford Handbooks Online. Asher, James J. 1977. Learning another language through actions: The complete teacher’s guidebook. Los Gatos, CA: Sky Oaks Productions. Articulatory phonetics 59 Avelino, Heriberto & Sahyang Kim. 2003. Variability and constancy in the articulation and acoustics of Pima coronals. In Pawel Nowak, Corey Yoquelet & David Mortensen (eds.), Proceedings of the Twenty-ninth annual meeting of the Berkeley Linguistics Society, 43–54. Ann Arbor: Sheridan Books. Bessell, Nicola J. 1996. Tlingit F’: Ejective or glottalized? Paper presented at the annual meeting of the Society for the Study of Indigenous Languages of Americas, San Diego, January 4–7. Bird, Sonya. 2011. Phonetic fieldwork in the Pacific Northwest. Proceedings of the International Conference of Phonetic Sciences XVII. 76–79. Bird, Sonya. 2012. Cool thing about ultrasound #17: Now I can pronounce /hiqət/! In Joel Dunham, John Lyon & Natalie Weber (eds.), Proceedings of the 47th International Conference on Salish and Neighbouring Languages (UBC Working Papers in Linguistics), 1–12. Vancouver, BC: University of British Columbia. Bird, Sonya & Sarah Kell. 2017. The role of pronunciation in SENĆOŦEN language revitalization. Canadian Modern Language Review 73(4). 538–569. Bird, Sonya, Janet Leonard & Scott Moisik. 2010. A motion vector analysis of tongue motion in SENĆOŦEN / qV/ and /Vq/ sequences. Paper presented at Ultrafest V, Haskins Laboratories, March 19–21. Bird, Sonya & Mizuki Miyashita. 2019. Teaching phonetics in the context of Indigenous language revitalization. Proceedings of the 2018 International Symposium on Applied Phonetics (ISAPh), 39–44. https://www.isca-speech.org/archive/ISAPh_2018/pdfs/07.pdf Bliss, Heather, Jennifer Abel & Bryan Gick. 2018. Computer-assisted visual articulation feedback in L2 pronunciation instruction: A review. Journal of Second Language Pronunciation 4(1). 129–153. Bliss, Heather, Sonya Bird, PEPAKIYE Ashley Cooper, Strang Burton & Bryan Gick. 2018. Seeing speech: Ultrasound-based multimedia resources for pronunciation learning in Indigenous languages. Language Documentation & Conservation 12. 315–338. Bliss, Heather, Lauretta Cheng, Murray Schellenberg, Zoe Lam, Raymond Pai & Bryan Gick. 2017. Ultrasound Technology and its Role in Cantonese Pronunciation Teaching and Learning. In Mary O’Brien & John Levis (eds.), Proceedings of the 8th Annual Conference on Pronunciation in Second Language Learning and Teaching (PSLLT), 33–46. Bliss, Heather & Bryan Gick. 2017. Blackfoot final vowels: What variation and its absence can tell us about communicative goals. In Betsy Sneller (ed.), Select proceedings of New Ways of Analyzing Variation (NWAV) 45. (Penn Working Papers in Linguistics 23[2]). Philadelphia: University of Pennsylvania. https:// repository.upenn.edu/pwpl/vol23/iss2/6/ (accessed 09 August 2019). Carlson, Barry F., John H. Esling & Jimmy G. Harris. 2004. A laryngoscopic phonetic study of Nlaka’pamux (Thompson) Salish glottal stop, glottalized resonants, and pharyngeals. In Donna Gerdts & Lisa Matthewson (eds.), Studies in Salish Linguistics in Honor of M.D. Kinkade, 58–71. Missoula, MT: University of Montana. Catford, J. C. & David B. Pisoni. 1970. Auditory versus articulatory training in exotic sounds. The Modern Language Journal 54(7). 477–481. Czaykowska-Higgins, Ewa. 2009. Research models, community engagement, and linguistic fieldwork: Reflections on working within Canadian indigenous communities. Language Documentation & Conservation 3(1). 15–50. Dart, Sarah. 1991. Articulatory and acoustic properties of apical and laminal articulations. Los Angeles: University of California Los Angeles dissertation. Esling, John H. 2003. Glottal and epiglottal stop in Wakashan, Salish and Semitic. Proceedings of the 15th International Congress of Phonetic Sciences 2. 1707–1710. Esling, John. 2005. There Are No Back Vowels: The Laryngeal Articulator Model. Canadian Journal of Linguistics 50(1/2/3/4). 13–44. Esling, John H., Katie Fraser & Jimmy G. Harris. 2005. Glottal stop, glottalized resonants, and pharyngeals: A reinterpretation with evidence from a laryngoscopic study of Nuuchahnulth (Nootka). Journal of Phonetics 33(4). 383–410. 60 Heather Bliss, Sonya Bird and Bryan Gick Esling, John, Scott Moisik, Allison Benner & Lise Crevier-Buchman. 2019. Voice quality: The laryngeal articulator model. Cambridge: Cambridge University Press. Fitzgerald, Colleen. 2017. The sounds of Indigenous language revitalization. Plenary address presented at the 91st annual meeting of the Linguistic Society of America, Austin, January 5–8. Flemming, Edward, Peter Ladefoged & Sarah Thomason. 2008. Phonetic structures of Montana Salish. Journal of Phonetics 36. 465–491. Gick, Bryan. 2002. The use of ultrasound for linguistic phonetic fieldwork. Journal of the International Phonetic Association 32(2). 113–122. Gick, Bryan, Sonya Bird & Ian Wilson. 2005. Techniques for field application of lingual ultrasound imaging. Clinical Linguistics & Phonetics 19(6–7). 503–514. Gick, Bryan, Heather Bliss, Karin Michelson & Bosko Radanov. 2012. Articulation without acoustics: “Soundless” vowels in Oneida and Blackfoot. Journal of Phonetics 40. 46–53. Gick, Bryan, Ian Wilson & Donald Derrick. 2013. Articulatory phonetics. Maldon, MA and Oxford: Wiley-Blackwell. Goddard, Pliny Earle. 1907. The phonology of the Hupa language. Part I: the individual sounds. Berkeley: University of California Press. Goddard, Pliny Earle. 1912. Chipewyan texts and analysis of the Cold Lake dialect, Chipewyan. (Anthropological Papers of the American Museum of Natural History 10). New York: Published by order of the Trustees of the American Museum of Natural History. Gordon, Matthew. 2017. Phonetic and phonological research on native American languages: Past, present, and future. International Journal of American Linguistics 83(1). 79–110. Gordon, Matthew, Jack Martin & Linda Langley. 2015. Some phonetic structures of Koasati. International Journal of American Linguistics 81(1). 83–118. Howe, Darin & Doug Pulleyblank. 2001. Patterns and timing of glottalisation. Phonology 18. 45–80. Hudu, Fusheini. 2008. The low vowel and retraction in St’at’imcets: An ultrasound investigation. SKY Journal of Linguistics 21. 67–81. Iskarous, Khalil, Joyce McDonough & D. H. Whalen. 2012. A gestural account of the velar fricative in Navajo. Laboratory Phonology 3(1). 195–210. Jenni, Barbara, Adar Anisman, Onowa McIvor & Peter Jacobs. 2017. An exploration of the effects of mentor-apprentice programs on mentors’ and apprentices’ wellbeing. International Journal of Indigenous Health 12(2). 25–42. Ladefoged, Peter. 2003. Phonetic data analysis: An introduction to fieldwork and instrumental techniques. Malden, MA: Blackwell Publishing. Maddieson, Ian. 1998. Collapsing vowel harmony and double-articulated fricatives: Two myths about Avatime phonology. In Ian Maddieson & Thomas J. Hinnebusch (eds.), Language history and linguistic description in Africa, 155–166. Trenton NJ: Africa World Press. Maddieson, Ian. 2002. Phonetics in the field. In Julie Larson & Mary Paster (eds.), Proceedings of the twentyeighth annual meeting of the Berkeley Linguistics Society, 411–429. Ann Arbor: Sheridan Books. Maddieson, Ian, Caroline L. Smith & Nicola Bessell. 2001. Aspects of the phonetics of Tlingit. Anthropological Linguistics 43. 135–176. Margetts, Anna & Andrew Margetts. 2012. Audio and video recording techniques for linguistic research. In Nicholas Thieberger (ed.), The Oxford handbook of linguistic fieldwork, 13–53. Oxford: Oxford University Press. McDonough, Joyce M. & Benjamin Tucker. 2012. Replicating P. E. Goddard: A contemporary airflow and EGG study of Dene Sułiné. University of Rochester Working Papers in the Language Sciences 4(1). 45–56. McDonough, Joyce M. & D. H. Whalen. 2008. Editorial: The phonetics of native North American languages. Journal of Phonetics 36. 423–426. McDowell, Ramona. 2004. Retraction in Montana Salish lateral consonants: An ultrasonic study. Vancouver: University of British Columbia MA thesis. Articulatory phonetics 61 McIvor, Onowa & Teresa L. McCarty. 2016. Indigenous bilingual and revitalization-immersion education in Canada and the USA. In Ofelia García & Angel Lin Cham (eds.), Encyclopedia of language and education, vol. 5: Bilingual and multilingual education, 1–17. Basel, Switzerland: Springer International. Mellesmoen, Gloria. 2018. An acoustic comparison of /θ/ and /s/ in Comox-Sliammon. Journal of the Acoustical Society of America 144(3). 1939–1939. Namdaran, Nahal. 2006. Retraction in St’at’imcets: An ultrasonic investigation. Vancouver: University of British Columbia MA thesis. Navarra, Jordi & Salvador Soto-Faraco. 2007. Hearing lips in a second language: visual articulatory information enables the perception of second language sounds. Psychological Research 71. 4–12. Podesva, Robert J. & Elizabeth Zsiga. 2013. Sound recordings: Acoustic and articulatory data. In Robert J. Podesva & Devyani Sharma (eds.), Research methods in linguistics, 169–194. Cambridge: Cambridge University Press. Percival, Maida, Heather Bliss & Murray Schellenberg. 2017. Methodological trade-offs for dual-purpose phonetic fieldwork. Canadian Acoustics 45(3). 184–185. Rice, Keren. 2006. Let the language tell its story? The role of linguistic theory in writing grammars. In Felix K. Ameka, Alan Dench & Nicholas Evans (eds.), Catching language: The standing challenge of grammar writing, 235–268. Berlin: Mouton de Gruyter. Rice, Keren. 2010. The linguist’s responsibilities to the community of speakers: Community-based research. In Lenore A. Grenoble & N. Louanna Furbee (eds.), Language documentation: Practice and values, 25–36. Philadelphia: John Benjamins. Tabain, Marija, Janet Fletcher & Andrew Butcher. 2011. An EPG study of palatal consonants in two Australian languages. Language & Speech 54(2). 265–282. Hiroto Uchihara 3 Tone Abstract: Tone is an integral part of some indigenous languages spoken in North America, even if it may not be a common feature. Tone can present challenges due to the difficulty of learning to hear tones, of establishing tonal contrasts, and the complexity of tonal phonology. However, tone can be an integral part of the language, serving to convey both lexical and grammatical functions and thus cannot be ignored. With adequate training and basic knowledge of tones, tone can be transcribed and represented in collaboration with the speakers. This chapter describes such a methodology and how tones can be represented in the orthography. In addition, this chapter surveys the functions of tones, tone processes, and interactions of tones with other parts of the grammar, as well as the diachrony of tones in the indigenous languages of North America. 3.1 Introduction Tone is an integral part of some indigenous languages spoken in North America,1 even if it may not be a common feature. Tone can present challenges for community members as well as for linguists working on documentation and revitalization, due to the difficulty of learning to hear tones, of establishing tonal contrasts, and the complexity of tonal phonology. Tone is a linguistic use of pitch; speakers of all languages use pitch to encode linguistic information, but its function can be fundamentally different depending on the language. In a tonal language pitch can distinguish different words or different forms of the same word, while in English or French speakers use pitch for intonation to convey pragmatic information. If a speaker of Oklahoma Cherokee says ga̋:du with a rising pitch on the first syllable it means ‘bread’, while if he says gadú with level low pitch on the first syllable it means ‘on top of’; in this way, pitch contributes to the difference in meaning. On the other hand, an English speaker can pronounce the word bread with any pitch, without changing its meaning, but they can differ in the pragmatics; if a speaker pronounces the word bread with a falling pitch contour, he may be answering a question, ‘Do you want bread or rice?’, while if a speaker pronounces bread with a rising pitch contour, a speaker may be asking a question. More specifically, the definition of a canonical tonal language is where each syllable is specified for a tone (Hyman 2006); that is, speakers need to know the tone of each syllable. Typical examples of tonal languages are Mandarin and Vietnamese. Among North 1 In this chapter, as in other chapters of the book, “North America” refers to the regions that currently fall within the territories of the United States and Canada; Mexico is excluded from the discussion, despite its geographical and linguistic contiguity. https://doi.org/10.1515/9783110600926-003 64 Hiroto Uchihara American languages, many Dene/Athabaskan languages qualify as true tonal languages in this sense, since each syllable is specified for a tone. However, there are also less canonical tonal languages in North America. First, in some languages tonal contrasts are more restricted, such as only on accented (or stressed) syllables: this is the case in some Northern Iroquoian languages (Chafe 1977; Michelson 1988), Takelma (Sapir 1912), and some Uto-Aztecan languages such as Hopi (Jeanne 1982) and Yaqui (Demers et al. 1999). Secondly, there are so-called pitch-accent languages (Hyman 2009; Gordon 2014), which employ pitch to encode lexical and grammatical information as in tonal languages, but where each syllable is not specified for a tone but rather one syllable is specified for accent and the pitch on the other syllables is predictable (a typical pitch-accent language is Tokyo Japanese). This is the case with some Algonquian languages such as Arapaho and Cheyenne, Siouan languages such as Crow and Dhegiha, Muskogean languages and possibly Kiowa-Tanoan languages. Tonal languages in North America include the following: Kiowa-Tanoan languages (Sutton, this volume), Lingít/Tlingit and some Dene/Athabaskan languages (Saxon, this volume), some Algonquian such as Montagnais and Kickapoo (Oxford, this volume), Cherokee (Iroquoian; Mithun and DeCaire, this volume), Halq̓eméylem/Upriver Halkomelem (Salishan; Watanabe, this volume), Heiltsuk (Wakashan; Rosenblum and Rosborough, this volume), Haida, and other more marginally tonal languages in the senses described above. Tonal inventories of these languages are typically simple: they mostly contrast two pitch levels, low and high (and possibly the combinations of these two tones, rising and falling). Some languages are reported to have three pitch levels, low, mid and high and their combinations, including Tsuut’ina/Sarcee (Dene/Athabaskan; Sapir 1925; Cook 1978; McDonough et al. 2013) and Haida (Hori 1996). Other languages may have more complex inventories, such as Oklahoma Cherokee (Uchihara 2016). The organization of the rest of this chapter is as follows. First, I will discuss methodologies for establishing and identifying tonal contrasts, which will be crucial in documentation projects (§ 3.2). I will further discuss issues in the representations of tones, both orthographic and linguistic in § 3.3. As was already mentioned, tone can serve to distinguish different words or different forms of the same word. § 3.4 discusses such functions of tones. Tone can further be complicated by various tonal processes (§ 3.5), and interactions with non-tonal phonology and morphology (§ 3.6). Many tonal languages in North America developed tones recently. § 3.7 discusses how tones can develop from various sources. 3.2 Methodologies for establishing and identifying tones For most of the languages spoken in North America basic descriptions of tones already exist. However, sometimes we would like to check validity of the previous studies, or we Tone 65 might suspect that a new system with new tonal patterns is emerging, and thus we feel the necessity for establishing tonal contrasts. In addition, even when a basic description of tones exists, often we do not always know what tone each word has. In this section, we will explore various methodologies for establishing tonal contrasts and identifying tones of each word. There is a fallacy that speakers easily ‘hear’ tones, and otherwise only those with absolute pitch can hear tones. This is not the case; it is very rarely the case that speakers can hear tones without any training (even in languages with fairly simple two-level pitch accent languages such as Tokyo Japanese). On the other hand, most people, whether speaker or non-speaker, or linguist or non-linguist, can learn to hear tones employing certain methodologies (Snider 2018: 34), as we will see in this section. Another misconception is that one can establish tonal contrasts and identify tones of each word by using acoustic tools such as Praat (to be discussed later). Tonal contrasts are categorical representations in speakers’ minds, which acoustic tools cannot access (Cruz & Woodbury 2014: 515; Snider 2018: 34). Linguists have been wary of overreliance on acoustic tools since they became available; for instance, Kenneth Pike remarks: instrumental records… do not contribute greatly to their analysis since (1) it is the relative pitch of tonemes which is significant, rather than their absolute pitch, and (2) tonemes change under various conditions, so that the intervals do not remain fixed in such a way that they may be mechanically discovered; instruments merely record gross fluctuation, rather than analyzing it in terms of deviations of units within a system. (Pike 1948: 21). How, then, can one establish tonal contrasts and identify tones of each word? Pike (1948), Cruz & Woodbury (2014) and Snider (2018) detail methodologies for establishing tonal contrasts. The languages they are interested in and familiar with are Otomanguean languages spoken in Mexico and Bantu languages in Africa, and their typological tonal characteristics are distinct from those in North American languages. However, we can apply their methodologies to these languages as well. Their methodology is developed so that we can know how speakers categorize each tone pattern; acoustically, speakers physically produce various pitch patterns with different frequencies, but what we are interested here is how speakers categorize which pitch patterns as the ‘same’ and which pitch patterns as ‘different’. This is the same with the segments; both Spanish speakers and English (or Japanese) speakers physically produce [s] and [z], but Spanish speakers categorize these sounds to belong to the same category /s/ and produce [z] when an adjacent consonant is voiced, while English or Japanese speaker as two distinct categories, /s/ and /z/. By looking at acoustic realizations alone, we would not be able to determine how speakers categorizes these sounds. In (1) I summarize the steps that they propose to follow in establishing and identifying tones: 66 (1) Hiroto Uchihara Steps for establishing tonal contrasts I. Find tonal minimal pairs, that is pairs of words which are segmentally identical, differentiated only by tones. This would confirm at least how many and which tone patterns are contrastive. II. Group words according to the phonological characteristics which can affect the perception of tones. This includes the following: a. The number of syllables: monosyllabic (one syllable), disyllabic (two syllables), trisyllabic (three syllables), etc. b. Vowel length: short or long c. Glottalization: whether the vowel is followed by a glottal stop or not. d. Syllable structure: open syllable (no final consonant in the coda, such as me) vs. closed syllable (syllable ends in a consonant, such as meat); if there is coda, resonant (consonants such as n, m, l, r) vs. obstruent coda (consonants such as p, b, t, d, k, g, etc.). III. Within the groups established in (II), classify each word according to the tone patterns. This can be achieved by comparing two words in sequence, asking speakers whether the two words have the same tonal patterns, or asking them to ‘hum’ or ‘whistle’ these words to see if they do have the same tonal patterns. As an illustration, we can try to establish tonal patterns in Oklahoma Cherokee. First, consider Oklahoma Cherokee minimal pairs or near minimal pairs which contrast only by tones. Here, we are still not concerned with the specifics of the tones, especially their labels such as ‘high’ or ‘low’, or how they are written in the orthography, but rather we concentrate on how these words are (mostly) differentiated solely by tones; see § 3.3 on how to represent tones. (2) (3) (4) (5) (6) a. a. a. a. a. ama ‘water’ higa ‘Eat it!’ ǐ:ga ‘noon’ ga̋:du ‘bread o:hni ‘back’ b. b. b. b. b. á:ma ‘salt’ híga ‘You just ate it’ i:ga ‘day’ gadú ‘on top of’ ò:hna ‘lawn’ Next, limiting ourselves to disyllabic words (words with two syllables) with a short vowel in the first syllable, we find two groups of words, most of which belong to the first group. Here we disregard whether the first syllable is glottalized (marked with the symbol ʔ) or not: (7) A ada ‘wood’, ama ‘water’, daksi ‘turtle’, joʔi ‘three’, kwana ‘peach’, sali ‘persimmon’, sihgwa ‘pig’, sohi ‘hickory nut’, svkta ‘apple’, tehga ‘toad’, tsgili ‘ghost’, wahya ‘wolf’, yansa ‘buffalo’ (8) B áʔni ‘strawberry’, táʔli ‘two’ Tone 67 When we ask speakers to pronounce the words in A in a sequence, they should all have the same ‘melody’. On the other hand, if we mix words from class B, their melodies should sound somewhat different. Often speakers can recognize whether the words have the same ‘melody’ or not. Another thing to note here is that we are still not labeling each tonal pattern as terms like low, mid, high, falling, but rather abstract labels such as A or B. Sometimes, speakers and linguists are quick to label each tonal category with concrete tonal labels such as ‘low’, ‘high’, etc., but labels are not important at this point; what is important is whether the tonal melody of one word is the same as the tonal melody of the model word. In fact, Ken Pike and Thom Smith-Stark warn us of the danger of labelling: up to this point there has been no essential need for tonal transcription. It is the grouping as such which has been important… Serious efforts to make an early “correct” transcription only tend to confuse us. (Pike 1948: 53) Whenever anyone said “these ones are mid-high to low, these ones are high to mid-low,” Thom smiled and said, “Never mind about that; as long as the tone sounds the same in each pile!” (Cruz & Woodbury 2014: 501 – 502) The following are words with a long vowel in the first syllable; here, we find five melodies: (9) A a:dla ‘rubber’, do:sa ‘mosquito’, gi:hli ‘dog’, i:ga ‘day, i:ya ‘pumpkin’, ji:sdu ‘rabbit’, lo:lo ‘locust’, nv:da ‘moon, sun’, su:li ‘buzzard’, yv:gi ‘fork’ (10) B á:ma ‘salt’, gv́:na ‘turkey’, jí:sgwa ‘bird’, ó:si ‘stove’ (11) C kǎ:hwi ‘coffee’, nǔ:na ‘potato’, sě:lu ‘corn’, tǔ:ya ‘bean’, wě:sa ‘cat’, yǒ:na ‘bear (12) D dlà:yhga ‘bluejay’, nv̀:ya ‘rock’, sv̀:gi ‘onion’ (13) E a̋:ta ‘young woman’, ga̋:du ‘bread’, gő:la ‘winter’, i̋:ga ‘noon’ When we encounter new disyllabic words, we can identify their tonal patterns by comparing their tones with the model words listed above. This methodology is easier to apply for languages where the shape of the root is constant, such as CV or CVCV, such as Otomanguean languages spoken in Mexico. On the other hand, tonal languages spoken in North America tend to be polysynthetic (that is, one ‘word’ can contain information that would be conveyed in a phrase or a sentence in English) and thus polysyllabic; the longer the word is, the harder it is to find other words which have exactly the same number of syllables and other phonological characteris- 68 Hiroto Uchihara tics to compare with. However, unlike languages in Mesoamerica, Asia or Africa, tonal systems of North American languages generally involve only a few tonal patterns (§ 3.1); this makes it easier to apply the methodology for establishing tonal contrasts. After establishing tonal contrasts following the methodology described above, one can confirm the results with acoustic software such as Praat; such acoustic tools are especially useful for labelling each tonal pattern established employing the methodology above with concrete tonal labels such as low, high, high-falling, etc. Praat can be downloaded at http://www.fon.hum.uva.nl/praat/. If you open a recording with Praat and click on ‘view and edit’, you will see a screen like the following. This is word [a:˨˩giʔ˦gv:˧˦ʔi˥˩] ‘I ate it’ in Oklahoma Cherokee; we can see the low fall tone on the first syllable, a relatively high tone on the second, and the drastic pitch fall on the last syllable. Tone, or pitch, is drawn as a blue line. Fig. 1: Pitch trace, spectrogram and spectrum of [a:˨˩giʔ˦gv:˧˦ʔi˥˩] ‘I ate it’ in Oklahoma Cherokee (Junior Scraper, speaker) 3.3 Representation of tones As we have seen, tone is a linguistic use of pitch. This section discusses how tones can be represented in orthography, if community members decide to represent them. In developing orthographic systems, speakers may decide whether or not they would like to represent tones. Note that not all the tonal languages with orthographic conventions represent tones in their orthography. For instance, various dialects of Japanese have pitch accent contrasts, but they are never represented in orthography. Representing tones in orthography can have advantages and disadvantages. If tone is contrastive, that is if tone distinguishes different words and forms of words (§ 3.4), Tone 69 ambiguity can be avoided if all the tones are represented. Moreover, representing all the tonal contrasts will be helpful for second-language learners and linguists. On the other hand, having to represent tones in all syllables all the time may discourage some speakers from writing their language, since as we have seen above, identifying tones is not easy for all the speakers. See Bird (1998), Bernard et al. (2002) and Roberts et al. (2019), among others, on whether marking tone helps comprehension in tonal languages spoken in Africa. If we decide to represent tone in orthography, how do we represent it? Snider (2018: Ch.1) classifies various systems of tone representation into the following major systems: the Chao system (or the IPA system), which includes the numeric version of the Chao system; the Pike system; and the diacritic system.2 First, the Chao system is shown in (14) and illustrated in (15) with an example from Oklahoma Cherokee: (14) Chao system ˥ Extra high ˦ High ˧ Mid ˨ Low ˩ Extra Low (15) Oklahoma Cherokee a˧dv:˧˩ne:˧˦li:˦˥sgi ‘actor’ ˩˥ rising ˥˩ falling ˦˥ high rising ˩˧ low rising ˧˥˩ rising-falling The Chao system has a numeric version; here, the higher numbers correspond to higher pitch, illustrated in (17) with a Cherokee example: (16) Numeric Chao system 55 Extra high 15 rising 44 High 51 falling 33 Mid 45 high rising 22 Low 23 low rising 11 Extra Low 343 rising-falling (17) Oklahoma Cherokee a3dv:31ne:34li:45sgi ‘actor’ The Pike system, commonly used among Mesoamericanists (mainly in the previous century) also employs numbers to represent tones, but the numbers are reversed, so that the higher numbers represent lower pitch; (18) is the same Cherokee word with this representation: 2 Snider (2018) further introduces the bar system, but here I omit the discussion of this system since this system is seldom employed for North American languages. 70 (18) Hiroto Uchihara Cherokee a3dv:35ne:32li:21sgi ‘actor’ Lastly, the diacritic system employs accent diacritics above vowels, illustrated in (20): (19) Diacritic system a̋ Extra high ǎ rising á High â falling ā Mid à Low ȁ Extra Low (20) Oklahoma Cherokee ādv̀:ně:li̋ :sgi ‘actor’ Each system has its own advantages and disadvantages. The Chao (IPA) system and numeric systems might be precise, but they are harder to type and might be unintuitive for community members. On the other hand, the diacritic system might be easier to type but it can be difficult to represent all the tonal contrasts sometimes, such as various types of falling (high-falling, mid-falling) or rising tones (mid-rising, mid-falling). In this chapter, I employ the diacritic system to represent tones. This is first because it is conventional for the majority of the tonal languages spoken in North America, and second because North American languages rarely have complex tone inventories (with the possible exception of Oklahoma Cherokee), with usually two tone levels and contours combining them and thus accent diacritics can represent all of the tonal contrasts in these languages fairly easily. In general, I follow the conventions used in the original sources. Such sources usually use the diacritic system presented in (19), except that with systems with more than two levels (such as Tsuut’ina/Sarcee or Oklahoma Cherokee), the intermediate tone is left unmarked, rather than representing it with a macron (ā). Even within the diacritic system, there can be great variation. For instance, for Oklahoma Cherokee, at least seven orthographic systems have been proposed and employed in the linguistic literature; the first ‘Pulte & Feeling (1975) system’ corresponds to the numeric Chao system, while the remaining six systems are all diacritic systems, with differences in how to represent vowel length or which tones to represent or not (in some systems the low tone is left unmarked, since this tone is the least marked tone phonologically; cf. § 3.5.3). Tone 71 Tab. 1: Comparison of Oklahoma Cherokee tonal orthography systems V length (i) Pulte & Feeling (1975) (ii) Scancarelli (1987) Munro (1996) (iii) Feeling et al. (2003), Uchihara (2009) MontgomeryAnderson (2008) (iv) Community orthography Modified community orthography short LF (short) L H SH (short) ạ1 ạ2 ạ3 ạ4 ȁ à á a̋ – à á – à a á a̋ à a á – à ā á à a á long LF LL HH LH HL SH a1 = a21 a2 a3 a23 a32 a4 ȁ: à: á: ǎ: â: a̋: àa àà áá àá áà áa àà/àa aa áá aá áa a̋a̋ aà aa áa aá áà áá à: ā: á: á̄: á̄: a̋: à: a: á: ǎ: â: a̋: A word ‘actor’, for example, would be represented as follows according to each system. It might be important for community members and linguists to agree on how tones can be represented, to avoid future confusion. (21) ‘actor’ in Oklahoma Cherokee a. Pulte & Feeling (1975) b. Scancarelli (1987) c. Munro (1996) d. Feeling et al. (2003) e. Montgomery-Anderson (2008) f. Community orthography g. Modified community orthography ạ2dv1ne23li4sgi àtv̏:ně:li̋ :ski àdv̀vnèélíiski atv̀vn ̀ eéli̋ i̋ ski atvv̀neélííski ādv̀:nēˊ:li̋ :sgi adv̀:ně:li̋ :sgi 3.4 Functions of tones In a tonal language, tones can be used to distinguish lexical items (§ 3.4.1), or to encode grammatical information (§ 3.4.2). 3.4.1 Lexical contrast Tones can be used to distinguish different lexical items, as segments do. For instance, see the following minimal pairs and triplets which are distinguished solely by tones. 72 Hiroto Uchihara First in (22) from Oklahoma Cherokee, the second syllable has a low tone in (a) and a high tone in (b), which contribute to the difference in meaning. Similarly, in (23) from K’ashógot’i̜ ne Xədə́/Hare (variety of Sahtú Dene/North Slavey, Dene/Athabaskan), the low tone in (a) and the high tone in (b) is the only difference between the two lexical items. Finally, in (24) from Kiowa (Kiowa-Tanoan), the high tone in (a), the low tone in (b), and the falling tone in (c) all contributes to the difference in meaning: (22) Cherokee ((Uchihara 2016: 3) a. gaʔdvsga b. ‘I am growing’ (23) K’ashógot’ine Xədə́ (Rice 2014: 692) a. sa b. sá ‘bear’ ‘beaver’ (24) Kiowa (Miller 2018: 43) a. kˀɔ́: ‘cold’ b. gaʔdv́sga ‘I am hanging it up’ kˀɔ ̀: ‘to lay there’ c. kˀɔ̂: ‘knife, cut’ 3.4.2 Grammatical use of tones Tones can also be used to encode grammatical information, such as tense, aspect or person. In some languages, the person category can be encoded solely by the differences in tones. For instance, in Oklahoma Cherokee (25), for some verbs the only difference between the 1sg and the 3sg forms is the tone, namely the lowfall tone and the low tone in the first syllable. In some Dene/Athabaskan languages, such as Diné bizaad/Navajo (26) or Tsuut’ina/Sarcee (27), the 2sg and 3sg forms are differentiated by tones on the prefix ni-: (25) Cherokee (Feeling 1975) a. gò:hlv:sga ‘I make it’ (26) Dene bizaad/Navajo (Goosen 1995: 4) a. nílį́ b. nilį́ ‘you are’ ‘he is’ (27) Tsuut’ina/Sarcee (Cook 1978: 171) a. níɣá b. niɣá ‘you will grow up’ ‘he will grow up’ b. go:hlv:sga ‘he makes it’ Tone can also encode the tense/aspect/mode categories in North American languages. For instance, the punctual and imperative forms in Oklahoma Cherokee are distin- Tone 73 guished by a tone difference in the penultimate syllable (28); the non-iterative and iterative aspect in Oklahoma Cherokee are encoded by a tone difference in the first syllable, in addition to vowel length (29); finally, the imperfective and perfective aspects in Dene/ Athabaskan languages such as Tsuut’ina/Sarcee (30) and Diné bizaad/Navajo (31) can be differentiated by tones of the stem or the prefix: (28) Cherokee (Uchihara 2016: 3) a. ho:hwe:lv̂:ga (Uchihara 2016: 3) ‘you just wrote it’ (29) Oklahoma Cherokee (Pulte & Feeling 1975: 254) a. ù:go:hé:ʔi (Feeling 1975: 252) b. úgo:hé:ʔi ‘he reportedly saw it’ ‘He reportedly saw it again’ (30) Tsuut’ina/Sarcee (Cook 1978: 169) a. -kˀós ‘to throw (imperfective)’ (31) Dine bizaad/Navajo (McDonough 1999: 509) a. yicha b. ‘he is crying’ b. b. ho:hwe:lv̀:ga ‘Write it!’ -kˀòs ‘to throw (perfective)’ yícha ‘he cried’ Finally, tone can mark nominal number in Montagnais Algonquian (Cowan 1983): in (32), the singular form has a high tone on the last syllable (a) while in plural it has a low tone (b). (32) Montagnais (Cowan 1983: 68) a. ustá:xw b. ustà:xw ‘ax’ ‘axes’ In summary, tone can serve to convey both lexical and grammatical information. Thus, in a way, ‘tone can do everything segments and non-tonal prosodies can do’ (Hyman 2011: 214). 3.5 Tonal processes and constraints In § 3.2, we took a look at how tone melodies of each word can be identified, and in § 3.3 we saw how tones can be represented in the orthography. However, when such words are placed in contexts or are inflected, their tones can undergo changes. For instance, in Diné bizaad/Navajo, “I am walking along” is yìšá:ɬ, with a high tone on the last syllable, but its negated form, “I am not walking along”, is dò: yìšâ:ɬ=dà (Leer 2001: 82); here, the syllable šâ:ɬ now has a high-low falling tone. This is due to a tonal process that changes the lexical tones, which we will take a look in § 3.5.1. Some such tonal processes can be motivated by cross-linguistically common tonal constraints, that is certain tonal 74 Hiroto Uchihara sequences are not preferred in general (§ 3.5.2). Lastly, some tonal processes can be better understood by employing the concept of tonal markedness (§ 3.5.3). 3.5.1 Tonal processes For North American languages, the most common tonal process is tone spreading, where a tone of one tone-bearing unit (such as syllable) spreads to another tone-bearing unit. The example above from Diné bizaad/Navajo illustrates spreading of a low tone, where the low tone of the enclitic =dà spreads to the preceding long vowel with a high tone, forming a falling tone (Leer 2001). Here, the first line represents how speakers pronounce, and the second line shows a more abstract level of analysis which shows how each word is pronounced in isolation. Thus, the syllable šâ:ɬ has a falling tone in the first line, as pronounced by speakers, while it has šá:ɬ, a high tone, in the second line, reflecting how it is pronounced without the following =dà. Between these two lines we see a line connecting the syllable dà in the first line and the symbol L which represents a low tone, as well as a dashed line which connects this L with the syllable šâ:ɬ. This means that the original low tone on dà spreads to the preceding syllable šá:ɬ, which originally has a high tone. Due to this low tone that spreads from the following syllable dà, the syllable šâ:ɬ now forms a high-low falling tone.3 (33) Diné bizaad/Navajo dò: yìšâ:ɬ=dà (Leer 2001: 82) L dò: yìšá:ɬ=dà neg I.am.walking.along=neg ‘I am not walking (along)’ The conventions for representing tones as in (33) may look complex at first glance, but such a representation is employed in many linguistics works on tones; such a representation is called autosegmental representation (Goldsmith 1976). It can be useful for learners in understanding why the tones are not always as expected. (33) above shows that the low tone spreads leftward in Diné bizaad. A low tone in Diné bizaad can also spread rightward from any high tone verb prefix to a short prefix syllable that does not have a high tone (Leer 2001: 81). Thus, the low tone on xà= ‘up/ out’ in (34) spreads for two syllables. Again, the solid line represents where the tone is lexically associated to, and the broken lines indicate where they spread to: 3 In the examples, the first line shows the surface forms, while the second line (if any) shows the segmented form with lexical representations. The third line provides the gloss. The abbreviations used in this chapter are as follows: a: set A prefix (agentive-series); an: animate; cmp: compact; du: dual; ex: exclusive; foc: focus; ind: indicative; neg: negative; obj: object; pct: punctual; prs: present; sg: singular; subj: subject. Tone (34) 75 Diné bizaad/Navajo (Leer 2001: 81) xàxònìštšà:d L xà=xo-ni-š-l-tšà:d up/out=areal.obj.-thematic.prefix-1sg.subj.-thematic.prefix-fluff.up ‘I fluff up the soil’ Such tone spreading processes are reported in other Dene/Athabaskan languages (Hoijer 1943; Cook 1989; Rice 1989; Leer 2001, Holton 2005; Rice & Hargus 2005), Oklahoma Cherokee (Lindsey 1985; Uchihara 2016: Ch.6), Kiowa (Watkins 1984; Harbour 2003; Miller 2018: 96 ff.), Heiltsuk (Wilson 1987) and Haida (Hori 1996). When tone shifts from one tone-bearing unit (such as syllable) to another, this results in tone displacement. For instance, in Dakelh/Carrier (Dene; Athabaskan), the negative enclitic =íloh has a high tone on its first syllable (indicated with H in the second line), as shown in (35a), but this high tone is shifted to the preceding stem with certain nouns (Story 1989: 105). Below, the unlinked line is indicated with =, meaning that this syllable no longer has a high tone. Tone displacement is also reported in Fort Nelson Slavey (Dene/Athabaskan; Rice 1989) and Oklahoma Cherokee (Uchihara 2016: Ch.6). (35) Dakelh/Carrier (Story 1989: 105) a. xoh=íloh H xoh-íloh goose-neg ‘not a goose’ b. lhéz=iloh = H lhez-íloh dust-neg ‘not dust’ In some languages, some morphemes are associated with a floating tone. A floating tone is not realized on the morpheme it is associated to but rather is realized on the adjacent morpheme or a word (which is commonly known as tone sandhi) when they are combined. For instance, in Oklahoma Cherokee, some stems are associated with a floating high tone which docks onto the preceding morpheme, most commonly a pronominal prefix. As can be seen in (36b), the pronominal prefix ci- 1sg.a is lexically low-toned; however, when followed by verb stems such as -H:tlo:- ‘strap’, which is associated with a floating high tone (indicated with a raised H), the pronominal prefix receives a high tone as in (36a).4 4 In the examples from Oklahoma Cherokee, the first line has letters such as d, j, g, while the second line has t, c or k corresponding to them. This is because the first line reflects the community-based orthography, while the second line reflects the conventions employed by linguists. 76 (36) Hiroto Uchihara Oklahoma Cherokee (Feeling 1975: 92, 97) a. jí:dlo:híha b. H ci- :tlo:-híh-a 1sg.a-strap-prs-ind ‘I am strapping it.’ ji:jagalí:ʔa ci-:cakal-íh-a 1sg.a-rip-prs-ind ‘I am ripping it.’ When grammatical information is expressed by tones (see § 3 4.2 above), this could be also due to a morpheme that consists only of a floating tone. This is the case with the iterative prefix in Cherokee that we saw in (29b), or the possessive suffix in Dënesu̜ ɬiné/ Chipewyan and Tłı̨ chǫ/Dogrib (Saxon & Wilhelm 2016). Finally, in some languages a boundary tone is assigned to the last tone bearing unit (such as syllable) of a (phonological) word, to mark the end of a word or a phrase. For instance, in Oklahoma Cherokee, the word-final vowels are not pronounced in general, but when one is pronounced, this syllable usually has a high tone (Lindsey 1985: 125, 168; Haag 2001: 414; Johnson 2005: 17), even if lexically this vowel does not have a high tone, as shown in (37). Here, the final syllable hli is pronounced with a high tone; that this high tone is only phonetic (and thus not represented in the orthography) is indicated by putting these forms in the square brackets []. That the lexical tone of this syllable is low is justified by the form in (37b), where the word ‘dog’ is followed by a focus enclitic =dv́: and thus the syllable hli is no longer final in the word. A word-final boundary tone is also found in some Dene/Athabaskan languages (Rice 1989). (37) Oklahoma Cherokee (DF 1972) a. [gi:ɬí] gi:hli dog ‘dog’ b. [gi:ɬidʌ́ :] gi:hli=dv́: dog=foc ‘(it is the) dog’ 3.5.2 Tonal constraints Some of the tonal processes discussed in § 3.5.1 are motivated by cross-linguistically common tonal constraints; that is, certain tonal sequences are not generally preferred. First, in general a sequence of the same tones in the adjacent syllables that are not due to spreading is avoided (the Obligatory Contour Principle; Leben 1973; Myers 1997). For instance, in Oklahoma Cherokee, when two morphemes with a high tone come next to each other, one of them has to be deleted (Uchihara 2016: 110). Thus, in (38), the forms in (a) and (b) share the verb root -stoóʔ- ‘crush’, with the lexical high tone as can be observed in (a), but when preceded by another high tone as in (b), this high tone is deleted. Here, the deletion of tone is indicated by = across the line which links H, a high tone, to the syllable sdo: Tone (38) Oklahoma Cherokee (Feeling 1975: 48, 17) a. à:sdó:ʔa b. H a-stoóʔ-a 3sg.a-crush:prs-ind ‘He is crushing it.’ 77 à:gî:sdoʔa = H H a-kíi(ʔ)+stoóʔ-a 3sg.a-eat+crush:prs-ind ‘He is chewing it.’ Another cross-linguistic tendency is to avoid a tonal dip (*Trough; Yip 2002: 137; Hyman 2009: 229). For instance, again in Oklahoma Cherokee, a sequence of high-low-high tone is avoided (Uchihara 2016: 123). 3.5.3 Tonal markedness In a tonal language, not all the tones have an equal status. Some tones are more common than others; some tones may trigger tonal processes more than others. For instance, in Oklahoma Cherokee, the low tone is the most frequent tone in the language and never affects tones of other syllables. Reflecting this fact, in some orthographic representation of tones in Oklahoma Cherokee, the low tone is not marked with any diacritic, as we saw in § 3.3. In such a case, we say that the low tone in Oklahoma Cherokee is unmarked, underspecified or that syllables that carry a low tone are toneless (Hyman 2001). Unmarked tones are those which are phonologically not active (for instance, they do not spread to other syllables), and/or occur more frequently than other tones (Maddieson 1976: 30; Hyman 2001). The status of markedness can change over time or vary between varieties: Leer (2001) presents evidence that Lingít/Tlingit and Southern Dene/Athabaskan languages were originally low-marked, but the marked tone has shifted to the high tone (Leer 2001). Tones which are superficially identical can differ at a more abstract level. Such is the case of abstract tonal contrasts. Such analysis has been proposed for Dakelh/Carrier (Story 1989: 104 ff.), Neeʼaandegʼ/Tanacross (Holton 2005: 260 ff.) and Oklahoma Cherokee (Cornelius 2018). For instance, in Oklahoma Cherokee, a high tone spreads to the preceding mora (roughly speaking, a short vowel has one mora and a long vowel has two moras) if complex phonological and morphological conditions are met (cf. § 3.5.1) as in (39a), but in some cases this high tone fails to spread to the preceding mora even though the conditions are met (39b). Cornelius (2018) attributes this difference in behavior to the abstract contrast between the underlying low tone and the underlying underspecified tone, both of which are realized as a low tone on surface. That is, in (39a), the syllable hwee is underspecified (or carries no tone) and thus no letter below hwee in the second line, while in (39b) the syllable nvv carries a low tone, represented with the letter L in the second line.5 5 The asterisk (*) means that the forms are not pronounced this way. 78 Hiroto Uchihara (39) Oklahoma Cherokee (Feeling 1975) a. goohweélíʔa6 b. nvvwóoti (*nvv́wóoti) L H ‘medicine’ H ‘he is writing’ 3.6 Interactions with other parts of the grammar Tone is not independent from the rest of the phonology and grammar, and it can i nteract with other parts of the grammar: phonology, such as syllable structure or accent (§ 3.6.1); morphology, or the internal structure of words (§ 3.6.2). 3.6.1 Interactions with phonology Tones can interact with syllable types. For instance, contour tones (rising and falling tones) can only occur on long vowels, as in Cherokee (Lindsey 1985) or Tsuut’ina/Sarcee (Cook 1978). Tones can also interact with the accent (or stress). Thus, in some languages tone is contrastive only on accented syllables, as in Northern Iroquoian (Chafe 1977; Michelson 1988) and Yaqui (Uto-Aztecan; Demers et al. 1999). In the following examples from Kanienʼkéha/Mohawk, a Northern Iroquoian language, the accent is assigned to the penultimate syllable, where the high tone in (40) contrasts with a falling tone in (41): (40) Kanien’kéha/Mohawk (Michelson 1988: 54) í:weʔs ‘she, it is walking around’ (41) Kanien’kéha/Mohawk (Michelson 1988: 54) ì:reʔs ‘he goes’ Tone can also attract accent or vice versa. Thus, in the K’ashógot’i̜ ne Xədə́/Hare variety of Sahtú Dene/North Slavey, a high tone is attracted to the accented syllable (Rice 1990); on the other hand, in Neeʼaandegʼ/Tanacross (Holton 2000), and Tse’khene/Sekani (Hargus 2005) high tone attracts accent. The interaction of tone and accent is also reported for Crow (Siouan; Kaschube 1954) and Heiltsuk (Kortlandt 1975). 6 In these examples the long vowels are represented with doubling the vowels instead of a colon, so that the tonal process can easily be seen. Tone 79 3.6.2 Interactions with morphology Some tonal processes discussed in § 3.5 can be morphologically conditioned; that is, the application of such processes can be limited to certain units (or domains), such as morphemes, words, or phrases. For instance, high tone spreading is restricted to the stem domain in Oklahoma Cherokee, which excludes pronominal prefixes (Uchihara 2016: Ch.7). In (42a), the high tone associated with the syllable gí spreads leftward by one mora due to the high tone spreading discussed in § 3.5.1 above, forming a lowhigh rising tone. On the other hand, in (42b) the high tone on the syllable gí cannot spread to the preceding syllable because this syllable belongs to the pronominal prefix oostii- 1du.ex.a, which is outside of the domain of high tone spreading in Oklahoma Cherokee. The domain boundary is indicated by a square bracket ([). (42) a. Oklahoma Cherokee (DF, July 2013) Spreading à:sdǐ:gíʔa H aa-[stiik-íʔ-a 3sg.a-eat.lg-prs-ind ‘He is eating it (something long).’ b. OK Cherokee (Feeling 1975:47) No spreading ò:sdi:gíʔa H oostii-[k-íʔ-a 1du.ex.a-eat-prs-ind ‘He and I are eating it.’ Similarly, in Kiowa, the domain of high tone spreading excludes pronominal prefixes (Harbour 2003: 548; Miller 2018). Thus, the high tone of the last syllable of the incorporated kíísɔ́ ‘afternoon’ spreads to the first syllable of the verb stem dęįkˀɔ́ɔ́ ‘lie asleep’ in (43a), while the high tone of the pronominal prefix á- fails to spread (43b): (43) Kiowa (Harbour 2003: 548) a. Spreading kíísɔ́dę́įḱ ˀɔ́ɔ́ H [kíísɔ́+dęįkˀɔ́ɔ́ afternoon+lie.asleep ‘?’ b. No Spreading ádęįkˀɔ́ɔ́ (*ádę́įḱ ˀɔ́ɔ́) H á-[dęįkˀɔ́ɔ́ they.an-lie.asleep ‘They lie asleep’ Tone spreading in Dene/Athabaskan languages are also sensitive to the morphological structures (Story 1989; Rice 1989). For instance, in Fort Nelson Slavey, high tone spreading is only applied within the word and thus applies in compounds which form one word (44a), but it fails to apply in phrasal compounds which form two words (44b) (Rice 1989: 240). A similar morphological conditioning of the tone spreading is also reported in Crow (Siouan; Matthews 1959). 80 (44) Hiroto Uchihara Fort Nelson Slavey (Athabaskan) (Rice 1989: 240) a. Spreading b. No Spreading [sáhdhéh] [dechį][jíh] (*dechį́jíh) H H sah+dhéh dechį jíh bear+skin wood hook ‘bearskin’ ‘wood hook’ 3.7 Diachrony of tones Not all tonal languages have always been tonal. Most tonal languages spoken in North America have recently developed tones from some segments (§ 3.7.1) or syllable structure through accent (§ 3.7.2). Their original source is traceable in many cases through internal reconstruction or the comparative method. This development of tones from other sources is called tonogenesis. 3.7.1 Tonogenesis from segments In the languages of North America, the most common segmental source of tones is the glottal stop. It has been well known that glottal stop has tonal effects on the preceding vowel (Hombert et al. 1978; Kingston 2011). A glottal stop itself can induce a higher tone; this is the case with some Dene/Athabaskan languages such as Slavey or Dënesu̜ ɬiné/ Chipewyan (Krauss 2005) and Montagnais Algonquian (Cowan 1983). A glottal stop can also induce a lower tone, via creakiness; this is the case with some other Dene/Athabaskan languages such as Tłı̨ chǫ/Dogrib or Diné bizaad/Navajo (Krauss 2005); Northern Iroquoian (Michelson 1988); Keres (isolate; Miller 1964: 17–18); Hopi (Manaster Ramer 1986); Takelma (Sapir 1912: 20); Quileute (Chimakuan; Hoard 1993); Heiltsuk (Rath 1986; Wilson 1987); Coast Tsimshian (Sasama 1997); and Sanya-Henya Lingít/Tlingit (Leer 1991: 12–18). Note that some Dene/Athabaskan languages have developed a low tone and others a high tone from a glottal stop (so-called ‘tonal flip-flop’). Even dialects of the same language have developed opposite tones, as in Dän kʼè/Northern Tutchone and Dän kʼí/ Southern Tutchone (Krauss 2005). For instance, the word for ‘arrow’ in Dän kʼè/Northern Tutchone is k’éʔ with a high tone, while in Dän kʼí/Southern Tutchone it has a low tone, k’àʔ (Krauss 2005: 94). More extreme cases are Oklahoma Cherokee (Uchihara 2009; 2016) or Halq̓eméylem/Upriver Halkomelem (Brown 2004), which has developed both higher and lower tone from a glottal stop depending on the morphological and phonological contexts. For instance, in Oklahoma Cherokee, one and the same morpheme Tone 81 can alternate between a high tone and a lowfall tone, depending on the morphological context. Thus, in (45), both forms share the stem which can be internally reconstructed as *-yooʔst- ‘break (compact)’, but has an allomorph (another form of the same morpheme in another context) with the lowfall tone in the imperative form in (a) while in the punctual form it has a high tone (b). (45) a. Oklahoma Cherokee (Feeling 1975) hiyò:sda hi-yòò(ʔ)st-Ø-a 2sg.a-break.cmp-pct-ind ‘Break it (compact)!’ b. Oklahoma Cherokee (EJ 2011) hiyó:sda hi-yóo(ʔ)st-Ø-a 2sg.a-break.cmp-pct-ind ‘You just broke it (compact)’ Another common segmental source of tones is fricatives. A glottal fricative h has induced a lower tone in Montagnais (Algonquian; Cowan 1983), Northern Iroquoian (Michelson 1988), and Hopi (Uto-Aztecan; Manaster Ramer 1986); recall that Montagnais and Northern Iroquoian have also developed tones from a glottal stop. In Kickapoo (Algonquian; Gathercole 1983), fricatives θ, s, h induced a lower tone on the preceding vowel, and thus now speakers can employ whistle speech to communicate with each other, where segmental information is absent but the information is carried by the tones (Voorhis 1971).7 3.7.2 Tonogenesis from syllable structure and accent Tones can also develop differently depending on syllable structure or accent, as in Haida (Isolate; Hori 1996; Enrico 1998) or Heiltsuk (Wakashan; Wilson 1987). In Heiltsuk, the original stressed syllables acquired a high tone, while other syllables now have a low tone (Wilson 1987). In Haida, a high tone is found when the syllable has a long vowel, as in (46a) or (46b), or a sequence of a vowel + resonant, that is w, y, l, ɬ, m, n or ŋ as in (46c). A low tone is found on a word-initial syllable with a short vowel without any consonant in the final position of the syllable, as in (46d), or when this short vowel is followed by an obstruent, that is b, d, g, ɢ, p, t, k, q, t’, k’, q’, ʔ, sɬ, x, χ, or h as in (46e). A short syllable which follows a low-pitched syllable has a low tone, due to low-tone spreading (see § 3.5.1 of this chapter for tone spreading), as shown in (46 f ). Otherwise, a mid tone is found; that is, with a short syllable not followed by any consonant or a short syllable followed by an obstruent, which is not word-initial or preceded by a low-toned syllable. This is exemplified by the second syllable of the form in (46g): 7 This is practiced by courting adolescents in the Nacimiento community in Mexico. The whistle is produced by “cupping the hands and placing them together to form a chamber into which they blow with the lips placed against the knuckles of the thumbs. The pitch is controlled by lifting the fingers from the back of the chamber” (Voorhis 1971: 238). 82 (46) Hiroto Uchihara Haida a. /ɬguu/ [ɬgú:] ‘heron’ b. /kiis/ [kí:s] ‘pus’ c. /qayd/ [qáyt] ‘tree’ d. /qu/ [qò:] ‘sea otter’ e. /mad/ [màt] ‘mountain goat’ f. /q’aχada/ [q’à:χàdà] ‘dogfish’ g. /ɬtalga/ [ɬtál̩ ga] ‘nest’ The generalization appears to be that in an accented (mostly initial) syllable, a high tone is found when the rhyme (the vowel and the following final consonant if there is one within the syllable) is heavy (a long vowel or a sequence of a vowel + a resonant), while a low tone is found when the rhyme is light (a short vowel or a sequence of a short vowel + an obstruent); in Haida, coda obstruent is not counted for the syllable weight, which is typologically common (Gordon 2006). In a non-accented syllable a mid tone is assigned, which could be the unmarked tone (cf. § 3.5.3), unless a low tone spreads from the preceding syllable. 3.8 Conclusion As we have seen in this chapter, tone can be challenging both for community members and for linguists working on such languages. However, tone can be an integral part of the language, serving to convey both lexical and grammatical functions (§ 3.4) and thus cannot be ignored. With adequate training and basic knowledge of tones, tone can be transcribed and represented in collaboration with the speakers (§ 3.2, § 3.3). Tonal contrasts can emerge from other sources such as segments or accent (§ 3.7). Tone in North American languages is still underdocumented and understudied. These languages can have typologically and theoretically interesting characteristics which are not reported in tonal languages of other parts of the world, and thus documentation and analysis of tonal systems of such languages is urgent for both language revitalization and linguistics. Acknowledgements: I am thankful to Karin Michelson, two anonymous reviewers and the editors of the volume, especially Keren Rice, for their feedback on this chapter. Tone 83 3.9 References 3.9.1 General Beckman, Mary. 1986. Stress and Non-Stress Accent. Dordrecht: Foris. Bernard, Russell., George Mbeh & Penn Handwerker. 2002. Does marking tone make tone languages easier to read? Human Organization 61(4). 339–349. Bird, Steven 1999. When Marking Tone Reduces Fluency: An Orthography Experiment in Cameroon. Language and Speech 42. 83–115. Buckley, Eugene. 2019. Stress, tone and accent. In Ch. 3 of Daniel Siddiqui, Michael Barrie, Jessica Coon, Carrie Gillon, Jason Haugen & Eric Mathieu (eds.), Routledge Handbook of North American Languages, 68–90. New York, NY and Oxford: Routledge. Caballero, Gabriela & Matthew Gordon. 2021. Prosody in North American Indian languages. In Carlos Gussenhoven & Aoju Chen (eds.), The Oxford Handbook of Language Prosody, 395–407. Oxford: Oxford University Press. Goldsmith, John A. 1976. Autosegmental Phonology. Cambridge, MA: Massachussetts Institute of Technology dissertation. Gordon, Matthew. 2001. A typology of contour tones. Studies in Language 25(3). 423–462 Gordon, Matthew. 2006. Syllable weight: phonetics, phonology, typology. New York, NY and Oxford: Routledge. Gordon, Matthew. 2014. Disentangling stress and pitch accent: Toward a typology of prominence at different prosodic levels. In Harry van der Hulst (ed.), Word stress: Theoretical and typological issues, 83–118. Cambridge, UK: Cambridge University Press. Gordon, Matthew. 2016. Consonant-Tone Interactions: A Phonetic Study of Four Indigenous Languages of the Americas. In Avelino, Heriberto, Matt Coler & Leo Wetzels (eds.), The Phonetics and Phonology of Laryngeal Features in Native American Languages, 129–156. Leiden: Brill. Gordon, Matthew. 2017. Phonetic and phonological research on Native American Languages: past, present, and future. International Journal of American Linguistics 83(1). 79–101. Hyman, Larry. 2001. Private tone in Bantu. In Shigeki Kaji (ed.), Cross-linguistic studies of tonal phenomena, 237–257. Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa. Hyman, Larry. 2006. Word-prosodic typology. Phonology 23. 225–257. Hyman, Larry. 2009. The representation of tone. In UC Berkeley Phonology Lab Annual Report (2009). 108–132. Hyman, Larry. 2009. How (not) to do phonological typology: a case of pitch-accent. Language Sciences 31. 213–238. Hyman, Larry. 2011. Tone: is it different? In John Goldsmith, Jason Riggle & Alan Yu (eds.), The handbook of phonological theory, 197–239. 2nd edn. Oxford: Blackwell Maddieson, Ian. 1976. Universals of tone. In Joseph Greenberg (ed.), Universals of human language, vol. 2: Phonology, 335–365. Stanford, California: Stanford University Press. Roberts, David, Dana Basnight-Brown & Valentin Vydrin. 2019. Marking tone with punctuation: an orthography experiment in Eastern Dan. In Yannis Haralambous (ed.), Proceedings of the conference Grafematik: Graphemics in the 21st century from graphemes to knowledge, 293–327. Brest, France: Fluxus. Snider, Keith. 2018. Tone Analysis for Field Linguists. Summer Institute of Linguistics. Yip, Moira. 2002. Tone. Cambridge: Cambridge University Press. 84 Hiroto Uchihara 3.9.2 Na-Dene Cook, Eung-Do. 1971. Vowels and tones in Sarcee. Language 47. 164–179. Cook, Eung-Do. 1989. Chilcotin tone and verb paradigms. In Eung-Do Cook & Keren Rice (eds.), Athabaskan linguistics,145–98. Berlin: Mouton de Gruyter. Goossen, Irvy. 1995. Dine Bizaad: Speak, Read, Write Navajo. Salina bookshelf. Hargus, Sharon. 1985. Lexical Phonology of Sekani. Los Angeles, CA: University of California, Los Angeles dissertation. Hargus, Sharon. 2005. Prosody in two Athabaskan languages of northern British Columbia. In Sharon Hargus & Keren Rice (eds.), Athabaskan Prosody, 393–423. Amsterdam: John Benjamins. Hargus, Sharon & Keren Rice (eds.). 2005. Athabaskan Prosody. Amsterdam: John Benjamins. Holton, Gary. 2000. The Phonology and Morphology of the Tanacross Athabaskan Language. Santa Barbara, CA: University of California, Santa Barbara dissertation. Holton, Gary. 2005. Pitch, tone and intonation in Tanacross. In Sharon Hargus & Keren Rice (eds.), Athabaskan Prosody, 249–275. Amsterdam: John Benjamins. Hoijer, Harry. 1943. Pitch accent in the Athapaskan languages. Language 19. 38–41. Hung, Beverly. 1959. On the phonemic status of Navaho stress. Anthropological Linguistics 19. 20–3. Kingston, John. 2005. The phonetics of Athabaskan tonogenesis. In Sharon Hargus & Keren Rice (eds.), Athabaskan Prosody, 136–184. Amsterdam: John Benjamins. Krauss, Michael E. 2005. Athabaskan Tone. In Sharon Hargus & Keren Rice (eds.), Athabaskan Prosody, 55–136. Amsterdam: John Benjamins. Landar, Herbert. 1980. On stress in Apachean languages. International Journal of American Linguistics 46. 227–30. Leer, Jeff. 1991. The schetic categories of the Tlingit verb. Chicago, IL: University of Chicago dissertation. Leer, Jeff. 1999. Tonogenesis in Athabaskan. In Shigeki Kaji (ed.), Cross-linguistic studies of Tonal Phenomena, Tonogenesis, Typology, and Related Topics, vol.1, 37–66. Tokyo: Institute of the Study of the Languages and the Cultures of Asia and Africa. Leer, Jeff. 2001. Shift of tone markedness in Northern Tlingit and Southern Athabaskan. In Shigeki Kaji (ed.), Proceedings of Symposium: Cross-Linguistic Studies of Tonal Phenomena, Tonogenesis, Japanese Accentology, and Other Topics, 61–86. Tokyo: Institute of the Study of the Languages and the Cultures of Asia and Africa. McDonough, Joyce. 1999. Tone in Navajo. Anthropological Linguistics 41(4). 503–39. McDonough, Joyce. 2003. Navajo sound system. Dordrecht: Kluwer Academic. McDonough, Joyce, Jared O’Loughlin & Christopher Cox. 2013. An investigation of the three tone system in Tsuut’ina (Dene). The Journal of the Acoustical Society of America 133. 3571. https://doi. org/10.1121/1.4806542 Pike, Eunice. 1986. Tone contrast in Central Carrier (Athapaskan). International Journal of American Linguistics 52. 411–18. Rice, Keren. 1976. Hare phonology. Toronto, ON: University of Toronto dissertation. Rice, Keren. 1989. The Phonology of Fort Nelson Slave Tone: Syntactic Implications. In Eung-Do Cook & Keren Rice (eds.), Athabaskan linguistics, 229–264. Berlin: Mouton de Gruyter. Rice, Keren. 1990. Prosodic constituency in Hare (Athapaskan): Evidence for the foot. Lingua 82. 201–245. Rice, Keren. 2014. On beginning the study of the tone system of a Dene (Athabaskan) language: Looking back. Language Documentation & Conservation 8. 890–706. Rice, Keren & Sharon Hargus. 2005. Introduction. In Sharon Hargus & Keren Rice (eds.), Athabaskan Prosody, 1–45. Amsterdam: John Benjamins. Sapir, Edward. 1925. Pitch accent in Sarcee, an Athapaskan language. Journal de la Société des Americanistes de Paris n.s. 17. 185–205. Tone 85 Saxon, Leslie & Andrea Wilhelm. 2016. The “possessed noun suffix” and possession in two Northern Dene (Athabaskan) languages. International Journal of American Linguistics 82(1). 35–70 Story, Gillian. 1989. A Report on the Nature of Carrier Pitch Phenomena: With Special Reference to the Verb Prefix Tonomechanics. In Eung-Do Cook & Keren Rice (eds.), 1989, Athabaskan linguistics, 99–144. Berlin: Mouton de Gruyter. Tuttle, Siri. 2003. Archival Phonetics: Stress and Tone in Tanana Athabaskan. Anthropological Linguistics 45(3). 316–336 3.9.3 Kiowa-Tanoan Harbour, Daniel. 2003. The Kiowa case for feature insertion. Natural Language and Linguistic Theories 21(3). 543–578. Miller, Taylor. 2018. The Phonology-Syntax Interface and Polysynthesis: A Study of Kiowa and Saluteaux Ojibwe. Newark, DE: University of Delaware dissertation. Sivertsen, Eva. 1956. Pitch problems in Kiowa. International Journal of American Linguistics 22. 117–130. Watkins, Laurel. 1980. A grammar of Kiowa. Lawrence, KS: University of Kansas dissertation. 3.9.4 Iroquoian Chafe, Wallace. 1977. Accent and related phenomena in the Five Nations Iroquois languages. In Larry Hyman (ed.), Studies in stress and accent, 169–81. (Southern California Occasional Papers in Linguistics 4). Los Angeles, CA: University of Southern California. Cornelius, Samantha. 2018. Prosodic Phonology in Oklahoma Cherokee. Arlington, TX: University of Texas at Arlington dissertation. Haag, Marcia. 2002. Cherokee tone associations with overt morphology. Chicago Linguistic Society 37(2). 413–423. Herrick, Dylan, Marcellino Berardo, Durbin Feeling, Tracy Hirata-Edds & Lizette Peter. Collaborative Documentation and Revitalization of Cherokee Tone. Language Documentation and Conservation 9. 12–31. Hirata-Edds, Tracy & Dylan Herrick. 2017. Building Tone Resources for Second Language Learning from Phonetic Documentation: Cherokee Examples. Language Documentation & Conservation 11. 289–304. Johnson, Keith. 2005. Tone and Pitch Accent in Cherokee Nouns. UC Berkeley Phonology Lab Annual Report (2005). http://www.linguistics.berkeley.edu/phonlab/annual_report/2005/Johnson1-48. pdf#search='Cherokee Tone’ Lindsey, Geoffrey. 1985. Intonation and interrogation: tonal structure and the expression of a pragmatic function in English and other languages. Los Angeles, CA: University of California, Los Angeles dissertation. Lindsey, Geoffrey. 1987. Cherokee Pitch Phonology. Ms. London: University College at London. ms. Michelson, Karin. 1988. A comparative study of Lake-Iroquoian accent. Dordrecht: Kluwer. Uchihara, Hiroto. 2009. High Tone in Oklahoma Cherokee. International Journal of American Linguistics 75(3). 317–36. Uchihara, Hiroto. 2016. Tone & Accent in Oklahoma Cherokee. Oxford Studies of Endangered Languages. Oxford: Oxford University Press. Wright, Richard. 1996. Tone and Accent in Oklahoma Cherokee. In Pamela Munro (ed.), Cherokee papers from UCLA, 11–22. (University of California Occasional Papers in Linguistics 16). Los Angeles: University of California, Los Angeles. 86 Hiroto Uchihara 3.9.5 Algonquian Bogomolets, Ksenia. 2023. Accent and tone in Plains Algonquian languages. In Ksenia Bogomolets & Harry van der Hulst (eds.), Word prominence in languages with complex morphology, 219–247. Oxford: Oxford University Press Cowan, William. 1983. The development of suprasegmental inflections in Montagnais. International Journal of American Linguistics 49. 64–71. Franz, Donald. 1972b. The origin of Cheyenne pitch accent. International Journal of American Linguistics 38. 223–5. Gathercole, Geoffrey. 1983. Tonogenesis and the Kickapoo tonal system. International Journal of American Linguistics 49. 72–6. Leman, Wayne. 1981. Cheyenne pitch rules. International Journal of American Linguistics 47. 283–309. Martin, Pierre. 1980. Des tons en montagnais? Cahier de Linguistique de l’Université du Québec 10. 175–94. Voorhis, Paul. 1971b. Notes on Kickapoo whistle speech. International Journal of American Linguistics 37. 238–43. 3.9.6 Siouan Gordon, Raymond, Jr. 1972. Pitch accent in Crow. International Journal of American Linguistics 38. 191–200. Kaschube, Dorothea. 1954. Examples of tone in Crow. International Journal of American Linguistics 20. 34–6. Kim, Michael. 1996. The tonal system of accentual languages. Chicago, IL: University of Chicago dissertation. Matthews, George. 1959. On tone in Crow. International Journal of American Linguistics 25. 135–6. 3.9.7 Uto-Aztecan Demers, Richard, Fernando Escalante & Eloise Jelinek. 1999. Prominence in Yaqui words. International Journal of American Linguistics 65. 40–50. Jeanne, LaVerne Masayesva. 1982. Some phonological rules of Hopi. International Journal of American Linguistics 48(3). 245–70. Manaster Ramer, Alexis. 1986. Genesis of Hopi tones. International Journal of American Linguistics 52. 154–60. 3.9.8 Pacific Northwest Brown, Jason C. 2004. Some tonogenetic properties of Upriver Halkomelem. In Lea Harper & Carmen Jany (eds.), Proceedings from the Seventh Workshop on American Indigenous Languages, 40–48. Santa Barbara: University of California, Santa Barbara Enrico, John. 1990. Masset Haida phonology. Berkeley, CA: University of California, Berkeley dissertation. Enrico, John. 1991. The lexical phonology of Masset Haida. (Alaska Native Language Center-Research Paper 8). Alaska Native Language Center, University of Alaska Fairbanks: Fairbanks, Alaska. Hoard, James. 1993. Quileute tones. In Mattina Montler & Timothy Montler (eds.), American Indian Linguistics and Ethnography in Honor of Laurence C. Thompson, 417–28. Missoula, MT: University of Montana. Hori, Hirofumi. 1994. Pitch in Skidegate Haida. In Osahito Miyaoka (ed.), Languages of the North Pacific Rim, 197–215. Sapporo: Hokkaido University. Hori, Hirofumi. 1996. Pitch assignment rules in Skidegate Haida. Gengo Kenyku 110. 28–51. Tone 87 Kortlandt, Federik H. H. 1975. Tones in Wakashan. Linguistics 146. 31–34. Rath, John. 1985. Predictable and unpredictable tones in Heiltsuk Wakashan. Papers for the International Conference on Salish (and Neighbo(u)ring) Languages 20. 313–24. Sasama, Fumiko. 1997. A report on Coast Tsimshian ‘interrupted vowels’. Languages of the North Pacific Rim 2. 47–60. Sapporo: Hokkaido University. Wilson, Stephen. 1987. The development of tones in Heiltsuk. Berkeley Linguistics Society 13. 321–9. 3.9.9 Takelma Sapir, Edward. 1912. The Takelma Language of Southwestern Oregon. Handbook of American Indian Languages. Washington: Government Printing Office. Shipley, William. 1969. Proto-Takelman. International Journal of American Linguistics 35. 226–230. Colleen M. Fitzgerald and Matthew K. Gordon 4 Segmental phonology Abstract: This chapter explores some of the special characteristics of sounds (or segments) in North American languages. There are numerous ways in which the sounds of North American languages are special; for example, certain types of consonants that are common crosslinguistically do not occur in some languages in North America, e. g. nasals, bilabials, while there are other sounds that are relatively rare in languages of the world but that are more widespread in North American languages, e. g. ejectives, creaky voiced sonorants. These consonants combine in many languages to form potentially elaborate syllables. Because North American languages are known for their complexity in word structure, there are also many alternations in the realization of segments that depend on position in a word and neighboring sounds. Language revitalization efforts benefit from a better understanding of these special characteristics. Indigenous communities often focus on learners sounding “right”, achieving an accent that closely approximates fluent first language speech and minimizes “accentedness” (Munro and Derwing 2015, Bird and Kell 2017). An essential part of that is understanding the distinctiveness of the pronunciation of ancestral languages. Those distinctive aspects of sound systems are the focus of this chapter. 4.1 Introduction The sound systems of North American indigenous languages exhibit some relatively unique patterns of consonants and vowels, termed “segments”, in comparison with other languages of the world. These patterns have been understudied and possess characteristics that are dramatically different from English, Spanish, and French, the most widely spoken languages in North America. Increasing knowledge of the sound systems found in North American languages is beneficial for communities seeking to reclaim their languages. The next section (section 4.2) highlights some of the special and rare patterns observed in the inventory of segments found in particular North American languages, language families, and geographic regions. Section 4.3 delves into how sounds can affect each other and trigger changes of different kinds. Section 4.4 explores how segments may be combined to form larger units, known as syllables. The chapter concludes by synthesizing implications of the study of segments for language revitalization. https://doi.org/10.1515/9783110600926-004 90 Colleen M. Fitzgerald and Matthew K. Gordon 4.2 Consonants and vowels Considerable diversity exists in the inventory of segments found in languages of the world, both in the number and type of segments observed. For example, North American English has about 36 segments (24 consonants and 12 vowels), where the exact number varies between dialects and speakers, while Latin American Spanish varieties have approximately 22 segments (17 consonants and 5 vowels). North American indigenous languages vary considerably in their segment inventories, with some having fewer than English and Spanish and others having considerably more. For example, the Iroquoian language Seneca has 16 segments (9 consonants and 7 vowels), while the Na-Dene language Dëne Sųłinë́ (Chipewyan) has 63 segments (39 consonants and 24 vowels). The wide variation between languages in their number of segments is matched by comparable diversity in the type of segments that occur. Because of their diversity, the study of North American indigenous languages makes a significant contribution to the field of typology, which is concerned with identifying which properties are common, uncommon, and absent across languages and explaining the reasons for these distributions. Sections 4.2.1–4.2.3 explore these properties further for segment inventories in North America. 4.2.1 Segment inventories While a large portion of this chapter will focus on consonants and their properties, it is notable that one typological rarity fairly well-attested in North American languages, and more generally in the languages of the Americas, is related to their vowel systems. Native American languages often display relatively small-sized vowel inventories (Maddieson 2013g). For example, Central Alaskan Yup’ik (an Eskimo-Aleut language of Alaska), and the two languages of the southeastern United States, Chickasaw (a Muskogean language) and Caddo (a Caddoan language), all have vowel systems with only three different vowel qualities, which are distinguished through tongue position (height and backness) and the presence vs. absence of lip rounding. Chickasaw has the low central unrounded vowel /a/, the high front vowel unrounded /i/ and the mid back rounded vowel /o/ (Munro and Willmond 1994: ix), while the vowel systems of Central Alaskan Yup’ik (Jacobson 1984) and Caddo (Chafe 2018) consist of /a/ and /i/, but instead of /o/, they have the high back rounded vowel /u/. All three languages, however, have variants of these three basic vowel qualities differing in other dimensions. One of these dimensions is length, where short vowels contrast with long vowels that are produced by stretching out the short vowel without changing its quality (i. e. tongue and lip position). For example, the Central Alaskan Yup’ik word /juk/ (spelled yuk in the Yup’ik orthography) ‘one person’ has a short /u/ while the word /juːk/ ‘two people’ (spelled yuuk) has a long /uː/ (Jacobson 1984:8), which is represented with the International Phonetic Alphabet (IPA) length markː. Chickasaw Segmental phonology 91 and Caddo (but not Central Alaskan Yup’ik) also distinguish vowels based on whether they are oral or nasal(ized). Oral vowels are the default case cross-linguistically, while nasal vowels are produced by venting air through the nose, as for the vowels in French words like lent ‘slow’, pronounced [lã], or long ‘long’, pronounced [lõ]. In Chickasaw, the middle vowel in /takõlo/ ‘peach’ and the first vowel in /ĩnakfiʔ/ brother (of a woman)’ are nasalized. The combination of length and nasality is also responsible for the large number of vowels in Dëne Sųłinë́ mentioned above, as there are six vowel qualities all of which can occur in short and long variants and oral and nasal form, i. e. there are both short and long oral vowels and both short and long nasal vowels. Turning to consonants, several languages with unusually small inventories of consonants are found in North America, e. g. Iroquoian languages. For example, Lukaniec (2018) analyzes Wendat, an Iroquoian language, as having nine consonants, which is a small consonant inventory in comparison with the vast majority of the world’s languages (Maddieson 2013h). The consonants of Wendat appear in (1), where each consonant is represented using a sound of the International Phonetic Alphabet (IPA), a universal set of symbols that encodes any sound found in any language of the world. Most IPA symbols correspond to those used in writing systems based on the Latin alphabet, such as those used for English and most North American indigenous languages, but there are some differences; a couple of common ones are mentioned below. The columns in (1) reflect place of articulation, i. e. where in the mouth a sound is produced, such that sounds at the left have more forward articulations than those to their right. Labial sounds, those produced at the lips, thus appear on the far left, while sounds produced at the larynx (or glottis), which contains the vocal folds, fall on the right side. In between the lips and larynx are different places of articulation, including the alveolar ridge (the bony ridge behind the upper teeth), which is contacted by the tongue blade in the production of sounds like /t/, /s/, /n/, and /r/, the hard palate, the point of contact for the tongue for the sound /j/, and the velum, which serves as the contact point for /k/ and /w/ and is the gateway to the nasal cavity—when raised, the velum blocks off the nasal chamber; when lowered, the result is a nasal sound like the alveolar /n/ and nasalized vowels, as in Chickasaw (see above). Rows in the chart in (1) encode manner of articulation, or the degree of constriction in the vocal tract. Plosives have a complete closure in the oral cavity and a closed velum, such that no air escapes through the nose, whereas nasals also have a complete oral occlusion but a lowered velum that allows venting of air through the nose. (Note that plosives are also often termed stops, a category that technically includes nasal consonants as well since there is a closure of the mouth for both plosives and nasals.) Finally, approximants have a relatively open vocal tract without air passing through the nose; acoustically, they are characterized by regular (termed “periodic”) vocal fold vibration (see chapter 1 on acoustics). Note that some sources draw a further division among approximants between the two glides /j, w/ and the liquids, which include lateral sounds, such as /l/, and rhotic sounds such as /ɹ/ (the r-type sound of English) and the Spanish trilled /r/. 92 (1) Colleen M. Fitzgerald and Matthew K. Gordon Wendat consonant inventory (Lukaniec 2018: 43) Labial Alveolar Palatal Velar Plosive t k Fricative s Nasal n Approximant ɹ j w Glottal ʔ h The majority of the consonants in Wendat roughly correspond to sounds in English with the exception of /ʔ/, the symbol for glottal stop, a sound that is produced by closing the vocal folds, as in the middle of the English word uh-oh where it is pronounced but not spelled. It may also be noted that there are two sounds in (1) that are found in English but spelled differently in the Wendat orthography. The palatal approximant (glide) /j/ is written as ‘y’ and is pronounced like English ‘y’ (e. g. in the word yam), a convention also seen in the Central Alaskan Yup’ik example presented earlier. The alveolar approximant (rhotic) /ɹ/ is written ‘r’ and pronounced like English ‘r’ (e. g. in the word rat). Wendat is also of interest because it lacks labial plosives. Of the 567 languages categorized by Maddieson (2013e), fewer than 1 % lack a labial consonant—and all are from North America, including another Iroquoian language, Oneida, and Wichita (Kirikirʔi:s), a Caddoan language related to Caddo. Another uncommon feature of Wichita (besides its lack of labial consonants) that contributes to its small inventory of 10 consonants, is its lack of nasal consonants (Rood 1975), a feature shared with its fellow Caddoan language, Pawnee (Parks and Pratt 2008). An absence of nasal consonants is a more general areal feature (see chapter 28 on language contact and linguistic areas) of many Pacific Northwest languages (which neither Wichita nor Pawnee are), including the Salish languages Twana and Lushootseed, the Chimakuan language Quileute, and the Wakashan languages Makah and Nitinat (Thompson and Thompson 1972). Historically, these languages are argued to have had nasals, but from that earlier stage, the sounds developed into voiced stops, leading to this inventory gap. At the other end of the inventory size continuum are the large consonant inventories (34 or more consonants per Maddieson 2013h), spread out in the Pacific Northwest across several unrelated language families. The Na-Dene language Dëne Sųłinë́ mentioned earlier has 39 consonants. The Coast Tsimshian language Sm’algy̠a̠ x has a similarly large consonant inventory with 38 distinct consonants. We will examine this language a bit further below in section 4.2.2, since Sm’algy̠a̠ x also has some consonant types found infrequently in other languages. 4.2.2 Cross-linguistically underattested segments Maddieson (2013f) identifies several kinds of “uncommon” segment types: pharyngeals, labial-velars, clicks, and “th” sounds (his categorization of dental/alveolar non-stri- 93 Segmental phonology dent fricatives as in the English word think). The first two of these categories occur in Native American languages and will be discussed in this section, alongside several other cross-linguistically less common segment types that are found in many languages of North America: uvulars, glottalized nasals and approximants, ejectives, and laterals. Pharyngeals are produced by retracting the very back of the tongue, the tongue root, toward the rear wall of the pharynx (or back of the throat). Pharyngeal sounds occur in only 4.1 % of the languages surveyed in Maddieson (2013f). The Interior Salish language, Snchitsu’umshtsn (Coeur d’Alene) includes the pharyngeal /ʕ/, described in teaching materials by the fluent speaker Lawrence Nicodemus (1975: xvi) as having “no equivalent” in English, “best learned by listening to native speakers”. Observations in Doak and Montler (2000) suggest that that pharyngeal consonants are not commonly produced nor perceived in the second language pronunciation (and the writing) of Snchitsu’umshtsn. Another type of consonant produced far back in the mouth are uvulars, which involve a constriction, either a closure as for a plosive or a narrowing as for a fricative, at the uvula, the dangling appendage visible at the back of the throat. Maddieson’s survey (2013f) indicates that uvulars are found roughly twice as frequently as pharyngeals (which still makes them relatively rare), and he notes they often occur in the indigenous languages of the western part of the United States and Canada. The Wakashan language Nuuchahnulth, spoken in British Columbia, Canada, has two pharyngeal consonants, /ʕ ħ/, which occur in fairly robust distribution in the language and are derived historically from uvular fricatives and glottalized uvular stops (Carlson, Esling, and Fraser 2001). The consonant inventory for Sm’álgyax (Coast Tsimshian) is presented in (2) and features uvulars among its 38 distinct consonants, one of the largest consonant inventories found across the world’s languages. (2) Sm’álgyax consonant inventory (adapted from Mulder 1994: 20) Labial Alveolar Palataliz- Velar Labialized velar ed velar Plosive: Ejective pʼ tʼ k jʼ kʼ kwʼ j Voiceless p t k k kw Voiced b d gj g gw Affricate: Ejective tsʼ Voiceless tsʼ Voiced dzʼ Fricative s ɬ Nasal: Plain Voiced m n Glottalized ʔm ʔn Uvular Glottal qʼ q ɢ ʔ χ h 94 Colleen M. Fitzgerald and Matthew K. Gordon Approximant: Plain Voiced Glottalized l ʔl j ʔj ɰ ʔɰ w ʔw The labels for the charts in (1) and (2) encode categories like place and manner of articulation (both of which were discussed earlier), as well as other dimensions involving the configuration of the vocal folds and the source of the air for the consonants. We discuss here those additional dimensions applicable to plosives and return in section 4.2.3 to those affecting nasals and approximants. Voiced sounds are produced with vibration of the vocal folds, whereas voiceless sounds have no vocal fold vibrations. Ejectives involve a more complex orchestration of timing relations between the vocal folds and the oral constriction. They are produced by closing the vocal folds, as for glottal stop, at about the same time as the oral closure for the consonant is produced and then raising the entire larynx upward to compress the air between the closed vocal folds and the oral closure for the consonant. Finally, the oral closure and then shortly afterwards the vocal fold closure are released, allowing air to rush out of the mouth to produce a distinctive “popping” sound that makes ejectives one of the most auditorily salient type of segments. In ejectives, the source of the air responsible for the sound, termed the airstream mechanism, is not the lungs as in most other sounds, but the cavity between the vocal folds and the consonant’s constriction. Sm’álgyax has a majority of its consonants articulated in the back of the mouth, at the velum or further back. Sm’álgyax also has a variety of complex segments. Complex segments have two articulations, articulations that could represent a sequence of two distinct segments, except that these two articulations pattern as a single unit in the language. Complex segments include affricates, which start with a stop that releases into a fricative, often at the same place of articulation. English has two affricates, /ʧ/, sometimes phonetically represented as /č/ and heard at the beginning of chat, and /dʒ/, sometimes phonetically represented as /ǰ/ and heard at the beginning and end of judge. Unlike the English affricates, which are produced between the palatal and alveolar regions, Sm’álgyax’s affricates are alveolars and include an ejective. Another type of complex segment in Sm’álgyax are the glottalized nasals and approximants / ʔm ʔn ʔl ʔj ʔw ʔɰ /. Glottalized nasals and approximants are particularly common in the northwestern portion of North America (Maddieson 2013c). Finally, complex segments are also reflected in the velar segments that occur in (2) with secondary articulations, a j-like palatalization, represented by a superscripted /j/, e. g. /kj/, and a w-like labialization, represented by a superscripted labial-velar approximant /w/, e. g. /kw/. It may also be noted that Sm’álgyax has a voiceless lateral fricative, which is produced by lowering one or both sounds of the tongue just slightly so the air passing over the tongue creates noise, similar to the /l/ if you produce the English word clay in slow motion. Near the far end of the large and typologically intriguing consonant inventories is Tlingit, a Na-Dene language with 42 consonants shown in (3). (3) Segmental phonology 95 Tlingit consonant inventory (adapted from Maddieson, Smith and Bessell 2001: 139) Alveo- Post- Pala- Late- Velar Lablzd Uvular Lablzd Glottal velar uvular lar alveo- tal ral lar Plosive: Ejective tʼ kʼ kwʼ qʼ qwʼ ʔ w w Unaspirated t k k q q Aspirated th kh kwh qʰ qwh Fricative: Plain s š ɬ x xw χ χw h w w Ejective sʼ ɬʼ xʼ x ʼ χʼ χ ʼ Affricate: Ejective tsʼ ʧʼ tɬʼ Unaspirated ts ʧ tɬ h h Aspirated ts ʧ tɬh Nasal n Approximant j w None of the Tlingit consonants are labials, but complex segments, uvulars and laterals are abundant. The 10 uvulars include plain and labialized uvular plosives and fricatives. In addition to featuring ejectives, the Tlingit consonant inventory includes an aspiration contrast for the plosives; the unaspirated plosives are simple voiceless ones whereas the aspirated ones are associated with a puff of air when they are released, much like the English initial plosives in pan, tan, and can. Tlingit is also typologically remarkable in possessing ejective fricatives, which are found in only 2.2 % of languages according to the UCLA Phonological Segment Inventory Database (UPSID) (http://web.phonetik.unifrankfurt.de/upsid_info.html). Ejective fricatives present an articulatory challenge since they require simultaneously a tight enough constriction to allow for pressure to build up for the salient ejective release while also allowing for enough air to escape through the constriction to produce the turbulent airflow required for a fricative. One of the ejective fricatives of Tlingit is /ɬʼ/, even though Tlingit lacks the far more common lateral approximant /l/. Tlingit is noteworthy in having five laterals, none of which are approximants. Among North American indigenous languages, Ahtna (a Na-Dene language of Alaska), Nuuchahnulth, Kutenai (a language isolate spoken in Idaho and Montana), and Kiowa (a Tanoan language of Oklahoma) also lack a sonorant /l/, but have laterals produced using other manners of articulation. 96 Colleen M. Fitzgerald and Matthew K. Gordon 4.2.3 Phonation and voicing In this section, we focus on properties of sounds that involve the vocal folds. Many languages employ a contrast between voiceless and voiced sounds. A voicing contrast is particularly common for plosives as in Sm’álgyax. Less common are voiced vs. voiceless distinctions for other types of sounds. For example, many languages have only voiceless fricatives and not voiced ones as in Tlingit. (This differs from English, which does contrast voiceless and voiced fricatives, e. g. the pair of words sip vs. zip.) There is an aerodynamic reason for the bias in favor of voiceless fricatives. Vocal fold vibration is dependent on supraglottal air pressure, above the vocal folds, being lower than subglottal pressure below the vocal folds since air flows from areas of higher pressure to areas of lower pressure. Fricatives, however, require sufficient supraglottal pressure behind the oral constriction to create the turbulent air flow requisite for a fricative. What is particularly rare is for a language to make a voicing distinction for consonants other than plosives but not for plosives. Mesa Grande ’Iipa (Diegueño), a Yuman language spoken in southern California, is thus unusual in having a voicing contrast only for laterals; this contrast occurs at both the alveolar (4a) and palatal (4b) places of articulation. It may be noted that the voiceless laterals are fricated in Mesa Grande ’Iipa as is typical across languages. (4) Mesa Grande ’Iipa voicing contrast (Langdon 1966: 32–33) Voiced laterals Voiceless laterals a. xəlul ‘he plays the flute’ wəl̥ itʃ ‘he is bad’ b. wəʎak ‘he lies down’ xəʎ̥aː ‘moon’ Another feature relevant to the vocal folds is aspiration, which we saw earlier in the discussion of Tlingit. Aspiration is produced when the vocal folds are open (abducted) widely enough apart to allow substantial airflow from the lungs into the oral cavity but still close enough together to create turbulence as the air passes through the vocal folds. Consonants can differ in whether the aspiration precedes or follows the consonant with which it is associated. In Tlingit (and also English), there are post-aspirated plosives, meaning that the vocal fold abduction gesture extends after the release of the oral closure for the plosive. In some languages, however, plosives can be preaspirated, meaning that the vocal fold opening gesture precedes the oral constriction. Preaspiration is considerably rarer than postaspiration across languages. Tohono O’odham (Papago), a Uto-Aztecan language of southern Arizona and Sonora, Mexico, is a language with preaspiration; it occurs with voiceless plosives preceding a vowel (5a) but does not affect voiced stops (5b). The lack of preaspiration with voiced stops is expected since the vocal fold abduction gestures needed for aspiration is incompatible with the vocal fold adduction gesture necessary for the vocal fold vibration associated with voicing. (5) Segmental phonology 97 Preaspiration in Tohono O’odham stops (Fitzgerald 1996) a. máhkai ‘doctor’ máːgina ‘car (Spanish loan)’ b. wáːhpag ‘hole’ (plural) čaːbo ‘short-legged one’ There is significant variation across languages in the specific aspects of preaspiration, such as how much frication occurs and what degree of voicing occurs. Tohono O’odham’s Uto-Aztecan relatives like Southern Paiute, Chemehuevi, Ute (Miller, Elzinga, and McLaughlin 2005), Hopi (Whorf 1946) and Rarámuri (Tarahumara; Valiñas 2001) also have preaspiration, as do the Central Algonquian languages Menominee, Fox, Cree and Ojibwa (Bloomfield 1935). While it is most common for stops and fricatives to be voiceless, for vowels, it is the opposite—voiceless vowels are almost always limited to specific contexts, in particular adjacent to voiceless consonants. A number of indigenous languages, particularly in the western United States, devoice vowels, especially Uto-Aztecan languages. Southern Ute speakers regularly devoice unstressed vowels (see chapter 6 on stress) before a voiceless consonant (Oberly and Kharlamov 2015); this devoicing process is a feature that has been a challenge in language revitalization for learners (Oberly 2013). The examples in (6) illustrate vowel devoicing from words written in both the IPA (on the left) and the tribal orthography (on the right). Voiceless vowels are indicated with an open circle under the vowel in the IPA and with underlining in the orthography. Note that the IPA symbol ˈ indicates that the following syllable is stressed (see chapter 6 on stress). (6) Devoicing in Southern Ute (Oberly and Kharlamov 2015: 3) a. ʧi̥ ˈʧiɣɛ chi̱ chige ‘to be hard’ b. tɯ̥ˈkwa tü̱kwa ‘to be deep’ c. ko̥ˈpokitiː ko̱pokitii ‘to cause to break’ d. ku ̥ˈpɛnapɯ ku︬ penapü ‘bat, drum stick’ Voicing may also be subject to modifications based on the configuration of the vocal folds. These differences in voicing are often termed phonation types and include a common kind found in languages of North America, creaky voicing. In creaky phonation, the tension of the vocal folds is increased, resulting in vibratory cycles that are irregular and lower in fundamental frequency or pitch (see chapter 1 on pitch). The Wakashan language, Kwakw’ala (Kwakiutl) contrasts regular or modal voiced nasals (7a), with creaky voiced nasals, indicated by a tilde under the nasal (7b). It may be noted that consonants produced with glottal constriction, including ejectives (see above), are often termed glottalized and may be represented with a superscripted glottal stop. For example, the creaky voiced /m̰ n̰/ in Kwakw’ala may alternatively be represented as /ˀm ˀn/. (7) Creaky nasals in Kwakw’ala (Gordon and Ladefoged 2001: 387) a. nəm ‘one’ naka ‘drinking’ b. n̰an̰əm̰a ‘nine’ n̰ala ‘day’ 98 Colleen M. Fitzgerald and Matthew K. Gordon Contrasts involving creaky voicing are underattested cross-linguistically, but Gordon and Ladefoged (2001) observe they occur in a number of indigenous languages in the northwest, including Montana Salish, Hupa, and Kashaya Pomo (see Howe and Pulleyblank 2001 for more on glottalization in languages of North America). 4.3 Phonological processes In discussing the voiceless vowels of Southern Paiute, it was noted that voiceless vowels are limited to certain contexts, in particular before a voiceless consonant. Another condition on voiceless vowels in Southern Paiute is that they are unstressed. In other words, vowels that are either stressed or not adjacent to a voiceless consonant do not become voiceless. In fact, the same vowel may be voiceless in one form of a word and voiced in another form if the context changes between the two forms. This situation often arises when a morpheme, a basic element of meaning, is added to a word and creates alternations in the realization of sounds. This can be seen in the Southern Ute forms in (8), in which the base forms in (8a) and (8c) end in a voiceless vowel but their counterparts with a suffix, the plural -u in (8b) and the locative -vwan ‘on’ in (8d), have a voiced vowel in place of the original voiceless vowel. (8) Voicing alternation in Southern Ute vowels (Oberly and Kharlamov 2015: 4) a. mamaʧi̥ mamachi ‘woman’ b. mamaʧiu mamachiu ‘women’ c. tɯkaɁnapɯ̥ tüka’napü̲ ‘table’ d. tɯkaɁnapɯvwan tüka’napüvwan ‘on the table’ This alternation, or phonological process, illustrates how sound patterns change when morphological processes lead to new or different phonological environments by creating different combinations of sounds in different places. In (8), the word-final environment is removed when morphemes are added. These phonological processes can produce differences that fluent speakers may not be conscious of, but their absence is noticeable and perceivable as sounding “different”. Returning again to Nuuchahnulth, the glottalized nasals /ˀm ˀn/ (along with their glottalized glide counterparts /ˀj ˀw/) may trigger creakiness in the vowels occurring on both sides of the glottalized resonant (Carlson, Esling, and Fraser 2001). This a natural process resulting from the overlap of the constriction gesture for the glottalized nasals with adjacent vowels. A similar type of process triggered by gestural overlap, but of a different type of gesture, is observed in English when the lowered velum position for a nasal consonant spills over to an adjacent vowel creating a nasalized vowel. For example, the vowel in the word tan is nasalized but this nasalization is predictable from the environment, namely being adjacent to a nasal consonant. This differs from the nasalization in Chickasaw or French, which can distinguish words with different Segmental phonology 99 meanings, e. g. the French pair beau ‘beautiful’ (phonetically [bo]) vs. bon ‘good’ (phonetically [bõ]). Consonants can also be affected by context. Klamath, a language of California and Oregon, allows plosives to contrast three ways: voiceless unaspirated, voiceless aspirated, and ejective. However, voiceless aspirated and ejective plosives surface as voiceless unaspirated plosives before another obstruent (9c, f). The aspiration or ejective property is retained when the plosive occurs before vowels (9a, d) or sonorants (9b, e). (9) Ejectives and Aspirated Plosives in Klamath (Blevins 1993: 246) Stem mphet’ ‘floats’ a. C’V mphet’iːqi ‘floats up’ b. C’S mphet’wa ‘floats in water’ c. C’O mphetplanʧ’a ‘floats downstream’ Stem pheʧh ‘foot’ d. CʰV pheʧʰiːqi ‘puts a foot over’ e. CʰS pheʧʰwa ‘puts a foot into the water’ f. CʰO pheʧk’wa ‘puts a foot across’ The examples of alternations thus far have involved a segment being affected by an adjacent one. Alternations can sometimes operate at a distance. One such example of a long-distance alternation is provided by vowel harmony, a process involving spreading of certain properties of a vowel, e. g. frontness or rounding, from one vowel to another across adjacent consonants. In languages with vowel harmony, vowels can be divided into two (or potentially more depending on how many properties are involved) groups that are mutually exclusive in a word, i. e. all vowels in a word belong to one set or the other set but may not be mixed between the two groups. Vowel harmony is observed in Nimipuutímt (Nez Perce), a Sahaptin language of the western U. S. on the Columbia Plateau. The language has five vowels, /i e a o u/ that can be separated into two sets on the basis of their patterning: if one set of vowels (called the dominant series) is present, the other vowels (called the recessive set of vowels) do not appear in the word, with a few exceptions (Aoki 1966). The examples in (10) show how /a o/ belong to the dominant set and trigger a shift of the vowel in the prefix neʔ- (10a vs. 10b) and the suffix -eʔ (10c vs. 10d). (10) Vowel Harmony in Nimipuutímt (Aoki 1966: 759–60) a. neʔ-ˈmeχ ‘my paternal uncle’ b. naʔ-ˈtoːt ‘my father’ c. ˈmeq-eʔ ‘paternal uncle!’ d. ˈtoːt-aʔ ‘father! ’ Vowel harmony occurs in a number of Native North American languages, including Salish languages like Snchitsu’umshtsn (Doak 1992), and the Pomoan language, Kashaya (Oswalt 1961, Walker 2013). 100 Colleen M. Fitzgerald and Matthew K. Gordon Consonants may also be involved in long distance harmony processes. Tahltan, a Na-Dene language, displays harmony among coronal consonants (i. e., articulations using the blade or tip of the tongue). In the Tahltan consonant inventory in (11) all the consonants from dental to postalveolar (delimited by bold lines) are coronal consonants, and the subset of those in shaded columns are the ones that participate in consonant harmony. (11) Tahltan consonant inventory (adapted from Shaw 1991: 144) Labial Dental Lateral Inter- Alveo- Postalve- Velar Labializdental lar olar ed velar b d dl dð dz dz g gw t tɬ tθ ts ʧ k kw tʼ tɬʼ tθʼ tsʼ ʧʼ kʼ k wʼ ɬ l m n, ˀn θ ð s z ʃ ʒ x ɣ j w xwʼ ɣw Uvular Glottal ɢ q qʼ ʁ χ ʔ h While the process is not limited to the first person singular subject marker ‘I’ -s-, that morpheme shows how the different sets of coronals act in this system. In (12), the first person morpheme is in bold, showing that it surfaces as -θ- in (12a–b), as -ʃ- in (12c–d), and as -s- in (12e–f). The determining factor for which variant appears is in whether there are any coronals from one of the shaded sets in (11). If there is an interdental or postalveolar consonant to the right of the first person -s- marker, then the -s- matches the place of articulation of that consonant but otherwise remains -s- if there are no interdental or postalveolar consonants to its right. It may be noted that the triggering sound may be either immediately adjacent to the first person morpheme, as in (12c–d) or may be separated from it, as in (12a–b). (12) Coronal Harmony in Tahltan (morpheme /s/, “I”/1st sg. subject; Shaw 1991: 144) a. dɛθkʷʊθ ‘I cough’ b. mɛθɛθɛθ ‘I’m wearing (on feet)’ c. hudiʃʧa ‘I love them’ d. nɛʃjɛɬ ‘I love them’ e. ɛsdan ‘I love them’ f. sɛsxɛɬ ‘I’m going to kill it’ The Na-Dene language family has various patterns of coronal consonant harmony, as do other languages also have consonant harmony, such as Chumashan languages (Shaw 1991, Mithun 1997). Segmental phonology 101 4.4 Syllable structure The previous sections examined consonants and vowels in two general ways: the patterns of the inventories of each in a given language (and across languages), and the influences that segments wield on each other’s features as word-building processes generate new segment sequences. This section describes how consonants and vowels are organized into larger units, known as syllables. Vowels are typically the core elements (often termed the nucleus) of syllables, potentially preceded by consonants (where they are termed the syllable onset) and/or followed by consonants (where they are termed the syllable coda). In Karuk (see chapter 47 on Karuk), syllables consist of “[a]ny consonant plus an immediately following vowel, plus any immediately following consonant that is not immediately followed by a vowel” (Bright 1957: 11). The syllabification patterns in Karuk are shown for one syllable words in (13a–b) and two syllable words in (13c–g). The period symbol . is used to indicate boundaries between the different syllables, while C stands for any consonant, V for any short vowel, and Vː for any long vowel. Note that the acute accent mark in certain forms marks tone, which is not relevant to the discussion of syllable structure (see chapter 3 on tone). (13) Syllables in Karuk (Bright 1957) Syllable Shape One Syllable a. tas CVC b. xuːn CVːC c. d. e. f. g. Two Syllables sára púraf ʔamkir pusjaːh túːjʃip CV.CV CV.CVC CVC.CVC CVC.CVːC CVːC.CVC Gloss ‘fence’ ‘acorn mush’ ‘bread’ ‘deer’ ‘table’ ‘toyon berry’ ‘mountain’ The maximal syllable in Karuk is CVC and syllables without an onset consonant are prohibited. This kind of syllable structure is complex, but not overly so, and in the survey of languages in Maddieson (2013j), roughly half of the languages display moderate complexity in syllables. In contrast, a language with maximally simple syllable structure starts with a single onset consonant, but does not have any coda consonants. Syllable position plays an important role in the sound patterns of many languages of the world, since many restrict certain consonants to either onsets or coda position. In the Na-Dene language Deg Xinag, there is a three-way contrast between voiceless unaspirated, voiceless aspirated, and ejective plosives in onset position but a two-way contrast between voiceless and voiced plosives in coda position. Conversely, there is a tripartite distinction between plain voiced, glottalized, and voiceless 102 Colleen M. Fitzgerald and Matthew K. Gordon nasals in coda position but only plain voiced nasals in the syllable onset (Hargus 2012: 165–6). Returning to Karuk, the syllable complexity is strictly regulated. When word-building processes lead to sequences of segments that are not permissible as syllables in Karuk, phonological processes alter those sequences to conform to the syllable canon of CV(C) (where the parentheses around the C indicate optionality of the coda consonant). In (14), vowel-initial stems in Karuk contrast with consonant-initial stems. There are two prefixes, nani- ‘my’ and mu- ‘his’, which end in a vowel. While this does not create any challenges for stems that begin with a consonant, vowel-initial stems show a syllable repair strategy for many speakers, who say mú-psi:h instead of the expected form mu-apsi:h and naní-psiːh instead of expected naní-apsiːh (though some speakers do say mú-ʔapsi:h and naní-ʔapsiːh). (14) Vowel Deletion in Karuk (Bright 1957) V-initial stem Gloss C-initial stem a. ápsiːh ‘leg’ típah b. naní-psiːh ‘my leg’ nani-típah c. mú-psiːh ‘his leg’ mu-típah Gloss ‘brother, male cousin’ ‘my brother, male cousin’ ‘his brother, male cousin’ What is the problem, in terms of syllable structure, with a form like *mu-apsiːh? Karuk does not allow onsetless syllables and repairs them by eliminating the first vowel of the stem in mú-psiːh. It may be noted that that is not the only available strategy for avoiding onsetless syllables in cases of vowel sequences (a situation that is called hiatus) arising at morpheme boundaries—another option, which many languages opt for, would be to insert a consonant between the vowels. In fact, the careful reader may have noticed that the form ápsiːh ‘leg’ begins with a vowel, which would appear to violate the requirement that syllables have an onset consonant. As it turns out, speakers add a glottal stop to words that would otherwise begin with a vowel. Thus, ápsiːh is realized as ʔápsiːh. This process of adding a sound is called epenthesis. The Yakima variety of Sahaptin, a Sahaptian language of Washington and Oregon, displays more complex syllable structure, with certain kinds of consonant clusters permitted in the syllable onset and coda, e. g. skɨtks ‘fringe’, and sɨt’χws ‘Brodiaea hyacinthine’ (Hargus and Beavert 2002). However, these consonant clusters display an internal organization such that sonorants (also termed resonants), which are a class of consonants with periodic (i. e. non-noisy) energy (see chapter 1) that includes approximants and nasals, must be closer to the vowel nucleus than obstruents, a set of consonants that is characterized by noisy energy and includes fricatives and plosives. An obstruent followed by a sonorant is thus possible in the onset (15a–c) and a sonorant followed by an obstruent is viable as a coda cluster (15d–e), but the opposite order is not permitted in either position. Onset and coda clusters, including those consisting of two obstruents (15f–k), are allowed subject to these sonority sequencing principles. Following Hargus and Beavert’s practice, the examples in (15) are schematized with O for obstruent and S for sonorant. (15) Segmental phonology 103 Consonant Clusters in Yakima Sahaptin (Hargus and Beavert 2002: Appendix) Cluster type Gloss a. χmɨɬ OSɨO ‘ridged’ b. ʧʼmɨt OSɨO ‘red elderberry, Sambucus racemosa’ c. ʃnɨm OSɨS ‘Crataegus douglassi’ d. ʃɨlk OɨSO ‘cricket’ e. kwɨnʧ OɨSO ‘tree moss’ f. ɬɨɬχ OɨOO ‘dirt’ g. ʦʼɨps OɨOO ‘released, popped off’ h. χɨʧʼk OɨOO ‘removing quickly’ i. kʼpɨs OOɨO ‘cold’ j. wɨtk SɨOO ‘half, piece’ k. qw’ʃɨm OOɨS ‘mischievous’ The patterns of syllable structure in Yakima Sahaptin illustrate a trend identified by Maddieson (2013j) of a positive correlation between syllable complexity and inventory size such that languages with more segments tend to allow greater syllable complexity. English shares with Yakima Sahaptin a high threshold for syllable complexity in terms of number of consonants, e. g. in words like strengths (phonetically /stɹɛŋθs/) and squelched (/skwɛlʧt/). However, English allows far fewer sonority plateaus or reversals involving a higher sonority consonant closer to the nucleus. Consequently, the combinations of consonants occurring in many North American indigenous languages present challenges to learners acquiring both a new place of articulation, airstream mechanism, or phonation type while simultaneously deploying that new articulatory dimension in a complex syllable. 4.5 Conclusions The incredible segmental diversity of the indigenous languages of North America plays an essential role in the cross-linguistic understanding of what is possible in phonology, not only for these languages, but abstractly for language more generally. Especially noteworthy are the consonant systems, from the size of the inventory, the number of contrasting places of articulation, or the permissible consonant clusters, among other features. Some of these aspects are also vulnerable to change or loss as languages have fewer fluent first language speakers. The contrast between glottalized and non-glottalized consonants is one example. While first language speakers of Dëne Sųɬiné (Chipewyan) contrast three sets of stops, /d t tʼ/, Cook (2005) observes that adults whose first language is English, but who have learned Dëne Sųɬiné deglottalize the ejectives (so producing /tɬ/ rather than /tɬʼ/). Recent research has developed a better understanding for effective teaching of indigenous languages (i. e., Hermes 2007, Johnson 2014, Czaykowska-Higgins, Burton, 104 Colleen M. Fitzgerald and Matthew K. Gordon Marinakis, McIvor 2017, McIvor, Jacobs, and Jenni 2018; see also chapter 34). That literature is growing, but only a small proportion attends to the pronunciation and phonological implications for indigenous language revitalization. One challenge is that more documentation and analysis of the phonetics and phonology of indigenous languages is needed (Whalen, Di Canio, and Shaw 2011, Fitzgerald 2017b). Most of the world’s languages have been studied in far less detail than English or other major world languages so there is certainly a need for more phonetic and phonological documentation and analysis of the sounds systems of indigenous languages. Another challenge, as demonstrated in this chapter, is that the sound systems of North American indigenous languages differ anywhere from slightly to substantially from those of English, Spanish or French (Whalen, Di Canio, and Shaw 2011; Bird and Miyashita 2018). Accordingly, achieving a pronunciation that sounds like the elders is challenging because it means being able to perceive and produce potentially a large number of novel contrasts, e. g. the difference between velars and uvulars, between plain and ejective stops, modal voiced and creaky voiced nasals, etc., and being able to articulate all of these differences when such consonants occur consecutively (Bird and Kell 2017). Despite these challenges, communities focused on language revitalization can draw from a number of important principles grounded in the research on second language pronunciation for other languages. Evidence suggests that pronunciation should be explicitly taught, even at the beginning of instruction (Thomson and Derwing 2015, Bird and Kell 2017). Language instruction benefits from an understanding of the linguistic aspects of sound structure like the segmental inventory or topics covered in this and the other phonetic and phonological chapters of this volume. Exposure to authentic language use, especially steady streams of spoken language from elders in conversation or even stories plays a role in helping learners perceive the different sounds and where word and phrasal breaks occur (Derwing and Munro 2014: 50–51, as well as studies cited there). Using culturally grounded materials like those that are part of the verbal arts tradition of that language, such as genres of storytelling, song, or poetry (Fitzgerald and Hinson 2015; Fitzgerald 2017a) is one way to provide that authentic input. Finally, resources exist for communities and linguists, such as the Teaching Pronunciation in Indigenous Languages course (Teaching Pronunciation for Indigenous Languages 2018) that has been taught several times at the Institute on Collaborative Language Research or through collaborative projects in teaching and research that focus on phonetics, phonology, or pronunciation cited here. Segmental phonology 105 References Aoki, Haruo. 1966. Nez Perce vowel harmony and proto-Sahaptian vowels. Language 42. 759–767. Bird, Sonya & Sarah Kell. 2017. The role of pronunciation in SENĆOŦEN language revitalization. The Canadian Modern Language Review/La Revue Canadienne des language vivantes. 73(4). 538–569. Bird, Sonya & Mizuki Miyashita. 2018. Teaching phonetics in the context of Indigenous language revitalization. Proceedings of the International Symposium on Applied Phonetics (ISAP). 39–44. doi 10.21437/ ISAPh.2018-7 Blevins, Juliette. 1993. Klamath Laryngeal Phonology. International Journal of American Linguistics 59. 237–280 Bloomfield, Leonard. 1935. On the sound-system of Central Algonquian. Language 1. 130–156. Bright, William. 1957. The Karok language. Volume 13. Berkeley, CA: University of California publications in linguistics. Carlson, Barry F., John H. Esling & Katie Fraser. 2001. Illustrations of the IPA: Nuuchahnulth. Journal of the International Phonetic Association 31(2). 275–279. Chafe, Wallace. 2018. The Caddo Language: A grammar, texts, and dictionary based on materials collected by the author in Oklahoma between 1960 and 1970. Petoskey, MI: Mundart Press. Cook, Eung-Do. 2006. The patterns of consonantal acquisition and change in Chipewyan (Dëne Sųłiné). International journal of American linguistics 72(2). 236–263. Czaykowska-Higgins, Ewa, Strang Burton, Onowa McIvor & Aliki Marinakis. 2017. Supporting Indigenous language revitalisation through collaborative post-secondary proficiency-building curriculum. In Wesley Y. Leonard & Haley De Korne (eds.), Language Documentation and Description. Vol 14, 136–159. London: EL Publishing. Derwing, Tracey & Murray J. Munro. 2014. Once you have been speaking a second language for years, it’s too late to change your pronunciation. In Linda Grant & Donna M. Brinton (eds.), Pronunciation myths: applying second language research to classroom teaching, 34–55. Ann Arbor, MI: University of Michigan Press. Doak, Ivy. 1992. Another Look at Coeur d’Alene Harmony. International Journal of American Linguistics 58(1). 1–35. Doak, Ivy & Timothy Montler. 2000. Orthography, lexicography, and language change. In Nicholas Ostler & Blair Rudes (eds.), Endangered Languages and Literacy. Proceedings of the Fourth FEL Conference, 132–138. Bath, England: FEL. Fitzgerald, Colleen M. 1996. Aspiration in Tohono O’odham. paper presented at the LSA Annual Meeting, San Diego, CA, January 4–7, 1996. Fitzgerald, Colleen M. & Joshua D Hinson. 2015. Using listening workshops to integrate phonology into language revitalization: Learner training in Chickasaw pronunciation. Paper presented at the 4th International Conference on Language Documentation and Conservation, University of Hawai’i, Honolulu, Hawai’i, February 26–March 1, 2015. Fitzgerald, Colleen M. 2017a. Motivating the documentation of the verbal arts: Arguments from theory and practice. Language Documentation and Conservation 11. 114–132. Online http://hdl.handle. net/10125/24728 (accessed 10 December 2019). Fitzgerald, Colleen M. 2017b. The Sounds of Indigenous Language Revitalization. Invited plenary address at the Annual Meeting of the Linguistic Society of America. Austin, TX, January 5–8, 2017. www.youtube. com/watch?v=wrPe_6KdoOo&list=PLc4TBef_CiuokIawF2lrXL2q6vyP4IIs7 (accessed 3 February 2017). Gordon, Matthew K. & Peter Ladefoged. 2001. Phonation types: a cross-linguistic overview. Journal of phonetics 29(4). 383–406. Haas, Mary R. 1940. Tunica. (Handbook of American Indian Languages, vol. 4, no. 40, part 4a). New York: JJ Augustin. 106 Colleen M. Fitzgerald and Matthew K. Gordon Hargus, Sharon. 2012. Deg Xinag rounding assimilation: a case study in phonologization. Laboratory Phonology 3.1. 163–193. Hargus, Sharon & Virginia Beavert. 2002. Predictable versus underlying vocalism in Yakima Sahaptin. International Journal of American Linguistics 68(3). 316–340. Hermes, Mary. 2007. Moving toward the language: Reflections on teaching in an Indigenous immersion school. Journal of American Indian Education 46(3). 54–71. Howe, Darin & Douglas Pulleyblank. .2001. Patterns and timing of glottalisation. Phonology 18. 45–80. Doi: 10.1017/S0952675701004018. Jacobson, Steven. 1984. Yup’ik Eskimo Dictionary. Fairbanks: University of Alaska, Alaska Native Language Center. Johnson, M. K. (Sʔímlaʔxw). (2014). yaʕ̓tmín cqw əlqwilt nixw, uł nixw, ul nixw, I need to speak more, and more, and more: Okanagan-Colville (Interior Salish) Indigenous second-language learners share our filmed narratives. Language Documentation and Conservation 8. 136–167. Kaisse, Ellen M. & Patricia A. Shaw. 1985. On the theory of Lexical Phonology. Phonology 2(1). 1–30. Langdon, Margaret H. 1966. A Grammar of Diegueño: The Mesa Grande Dialect. Berkeley,CA: University of California dissertation. https://escholarship.org/uc/item/08t8n3db (accessed 31 December 2019). Lukaniec Megan. The elaboration of verbal structure: Wendat (Huron) verb morphology. Santa Barbara, CA: University of California, Santa Barbara dissertation. Maddieson, Ian. 2013a. Voicing and Gaps in Plosive Systems. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/5 (accessed 17 February 2019). Maddieson, Ian. 2013b. Uvular Consonants. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/6 (accessed 17 February 2019). Maddieson, Ian. 2013c. Glottalized Consonants. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/7 (accessed 17 February 2019). Maddieson, Ian. 2013d. Lateral Consonants. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/8 (accessed 17 February 2019). Maddieson, Ian. 2013e. Absence of Common Consonants. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/18 (accessed 17 February 2019). Maddieson, Ian. 2013f. Presence of Uncommon Consonants. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/19 (accessed 17 February 2019). Maddieson, Ian. 2013g. Vowel Quality Inventories. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/2 (accessed 20 October 2019). Maddieson, Ian. 2013h. Consonant Inventories. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/1 (accessed 20 October 2019). Maddieson, Ian. 2013i. Voicing in Plosives and Fricatives. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/4 (accessed 2 November 2019). Maddieson, Ian. 2013j. Syllable Structure. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. https://wals.info/chapter/12 (accessed 17 February 2019). Segmental phonology 107 Maddieson, Ian, Caroline L. Smith, & Nicola Bessell. 2001. Aspects of the Phonetics of Tlingit. Anthropological Linguistics 43(2). 135–176. http://www.jstor.org/stable/30028779 (accessed 9 November 2019). McIvor, Onowa, Peter Jacobs & Barbara Jenni. 2018. Adult Indigenous contributions to reviving languages in British Columbia through Mentor-Apprentice style learning: Research Report. Victoria, British Columbia, Canada: University of Victoria. Retrieved from www.netolnew.ca/resources. Mithun, Marianne. 1999. The Languages of Native North America. New York: Cambridge University Press. Miller, Wick R., Dirk Elzinga & John E. McLaughlin. 2005. Preaspiration and gemination in Central Numic. International journal of American linguistics 71(4). 413–444. Mulder, Jean G. 1994. Ergativity in Coast Tsimshian (Sm’algyax). Volume 124. Berkeley,CA: University of California Press. Munro, Murray J. & Tracy M. Derwing. 2015. A prospectus for pronunciation research in the 21st century: A point of view. Journal of Second Language Pronunciation 13. 11–42. Munro, Pamela & Catherine Willmond. 1994. Chickasaw: An Analytical Dictionary. Norman, OK: University of Oklahoma Press. Nicodemus, Lawrence. 1975. Snchitsu’umshtsn: The Coeur d’Alene Language: a Modern Course, Albuquerque, NM Southwest Research Associates. Available online at https://plateauportal.libraries.wsu.edu/digitalheritage/snchitsuumshtsn-coeur-dalene-language-page-xvi-xvii (accessed 8 November 2019). Oberly, Stacey. 2013. The phonetic correlates of Southern Ute stress. In Shannon Bischoff, Amy V. Fountain, Deborah Cole & Mizuki Miyashita (eds.), The persistence of language: constructing and confronting the past and present in the voices of Jane H. Hill, 85–106. Amsterdam: John Benjamins. Oberly, Stacey & Viktor Kharlamov. The phonetic realization of devoiced vowels in the Southern Ute language. Phonetica 72(1). 1–19. Oswalt, Robert L. 1961. A Kashaya grammar (Southwestern Pomo). Berkeley, CA: University of California dissertation. Parks, Douglas R. & Lula Nora Pratt. 2008. A Dictionary of Skiri Pawnee. Lincoln, NE: University of Nebraska in Cooperation with the American Indian Studies Research Institute, Indiana University. Rood, David S. 1975. The implications of Wichita phonology. Language 51(2). 315–337. Shaw, Patricia A. 1991. Consonant harmony systems: The special status of coronal harmony. In Carole Paradis & Jean-François Prunet (eds.), Phonetics and phonology, vol. 2: The special status of coronals: internal and external evidence, 125–57. San Diego, CA: Academic Press. Thomson, Ron I. & Tracy M. Derwing. 2015. The Effectiveness of L2 Pronunciation Instruction: A Narrative Review. Applied Linguistics 36(3). 326–344. Thompson, Laurence C. & M. Terry Thompson. 1972. Language universals, nasals, and the Northwest coast. In M. Estellie Smith (ed.), Studies in Linguistics in Honor of George L. Trager, 441–456. The Hague: Mouton. Teaching Pronunciation for Indigenous Languages. 2018. Workshop offered at the Institute on Collaborative Research (CoLang). Instructors, Ewa Czaykowska-Higgins & Colleen Fitzgerald, instructors. Gainesville: University of Florida. https://colang.lin.ufl.edu/home/colang-2018-workshops/ (accessed 17 February 2019). Valiñas, Leopoldo. 2001. Lengua, dialectos e identidad étnica en la Sierra tarahumara. In Claudia Molinari & Eugeni Porras (eds.), Identidad y cultura en la sierra Tarahumara, 105–125. Mexico: Instituto Nacional de Antropología e Historia. Walker, Neil A. 2013. A Grammar of Southern Pomo. An Indigenous Language of California. Santa Barbara, CA: University of California, Santa Barbara dissertation. Whalen, D. H., Christian T. DiCanio & Patricia A. Shaw. 2011. Phonetics of endangered languages. Acoustics Today 7. 35–42. Whorf, Benjamin. 1946. The Hopi Language, Toreva Dialect. In Harry Hoijer et al (ed.), Linguistic Structures of Native America, 158–183. (Viking Fund Publications in Anthropology 6). New York: Cornelius Osgood. Suzanne Urbanczyk 5 Prosodic morphology Abstract: Prosodic morphology refers to a wide range of processes to make new words by reference to the sound patterns in a word, such as adding syllables or deleting or changing the order of segments. These processes are less common than simply adding segments to indicate a new meaning, but are found in many Indigenous languages of North America. This chapter provides an overview of the kinds of prosodic morphology processes found as well as a brief outline of how understanding these patterns can aid in teaching and learning new words. 5.1 Introduction Words are made of sounds and have meanings. Some words are simple, in terms of having just a few sounds and one meaning. Other words can be more complex, with several identifiable meanings and sound groupings. In many languages of North America, new words can be created by a variety of ways that do not involve simply adding an affix such as a prefix like re- or a suffix like -s to the edge of a base, as in the word re-write-s. A base is the part of the word to which some word formation processes apply. For example, in Hiaki (Uto-Aztecan) the way to express habitual meaning involves copying the first syllable of the base, as illustrated in (1) below. This chapter provides an overview of a range of prosodic morphology patterns, as found in the languages of North America, starting with a general introduction to the field. The bulk of the chapter illustrates the many ways that prosodic structure is relevant to understanding word patterns. It wraps up by discussing some ways that concepts like syllable structure can be helpful in teaching and learning word patterns like this. 5.1.1 Word patterns The sets of words in (1–6) below show a range of ways that a new meaning can be expressed by reference to units like syllables. Relevant portions of the word are underlined and syllable boundaries are indicated with a period, where it could be helpful. The linguistic terms used to refer to the patterns are indicated above each example. For example, in (1) the first syllable is copied to express that the action is ‘habitual’, and this way to make new words, by having a copy of part of a word is knowns as reduplication. The underlined portion of the words below indicate what is added. As can be seen, it looks like an exact copy of the first syllable. https://doi.org/10.1515/9783110600926-005 110 (1) Suzanne Urbanczyk Reduplication – ‘habitual’ Hiaki (Haugen 2014: 511) vu.sa vu.vu.sa ‘awaken’ či.ke či.či.ke ‘comb one’s hair’ In Southern Sierra Miwok (Miwok), the syllable structure of the base changes, either by making a consonant or a vowel longer or shorter (long consonants and vowels are represented with a colon following the symbol for the long segment). The base (or stem) has a specific target or shape it needs to be, depending on the suffix that is attached. For example, notice that in Stem 3 in (2), the second consonant is always long, and in Stem 4, the second consonant is always followed by another consonant. (2) Templatic morphology – Southern Sierra Miwok (Broadbent 1964: 38) Stem 1 Stem 2 Stem 3 Stem 4 meaning mus:alo:tkowta- musahlotkowat- mus:aʔlot:uʔkow:at- musʔalotʔukowta- ‘to be ashamed’ ‘to catch’ ‘to bump into’ In Dakota (Siouan), an affix meaning ‘I’ (indicated with underlining) is added inside the root, going after the first syllable of the word as in (3). (3) Infixation – Dakota, 1st person (Boas & Deloria, 1939) na.pcu ‘swallow’ na-wa-pca ‘I swallow it’ la.k’o.ta ‘Lakota’ la-ma-k’o.ta ‘I am a Lakota’ In Koasati (Muskogean), plural meaning is expressed by removing part of a syllable. The part of the syllable removed is underlined in the first column below in (4). (4) Subtractive morphology – Koasati ‘plural’ rhyme deletion (Kimball, 1991) singular plural meaning pi.táf-fi-n a.ko.co.fót-li-n ti.wáp-li-n si.mát-li-n pít-li-n a.ko.cóf-fi-n tíw-wi-n sím-mi-n ‘to slice up the middle’ ‘to jump down’ ‘to open something’ ‘to cut up tanned skin’ Lushootseed (Salish) shows that a vowel change can occur when copying a syllable to indicate the meaning of ‘diminutive’, as illustrated in (5). The accent indicates that the vowel is stressed. (5) Ablaut – Lushootseed diminutive reduplication (Urbanczyk, 2006) ǰə́səd ‘foot’ ǰí-ǰəsəd ‘little foot’ tədzíl ‘lie in bed’ tí-tədzil ‘lie in bed for a little while’ Prosodic morphology 111 And finally, in Saanich (Salish) the syllable structure changes in yet another way, by reordering the consonant and vowel, to indicate an action is occurring now as in (6). (6) Metathesis – Saanich ‘actual’ (Montler, 1986) θk̓ʷə́t ‘straighten it out’ θə́k̓ʷt ‘straightening it out’ ƛ̓pə́x̣ scatter’ ƛ̓ə́px̣ ‘scattering’ The shape, location and conditions of these changes are frequently described in terms of syllables and rhythmic patterns, or what is often referred to as prosodic structure. The next section discusses some key concepts related to prosody and prosodic morphology. 5.1.2 Prosodic morphology One way to see how prosodic structure is important to understanding word structure comes from taking a closer look at the Hiaki word for ‘awaken’ as represented in (7) below. The base has two syllables indicated by the symbol σ above the segments. The segments that are grouped into syllables are linked to the syllables by lines. Next to vusa ‘awaken’, one can see that the ‘habitual’ is formed by copying the first syllable. The syllable that is affixed is indicated with shading. Notice that it is an exact copy of the syllable that comes after it. We can describe this pattern as prefixing a syllable. (7) σ σ vusa ‘awaken’ → σ σ σ vuvusa ‘habitual’ When looking at the word patterns (morphology) above, the changes that are associated with a new meaning can be understood in terms of prosodic structure, and fall under the domain of prosodic morphology. As one can see, the concept of prosodic unit, such as a syllable, is central to describing the patterns and a brief explanation of what these prosodic units are is useful to understanding the field. In essence, prosodic structure is relevant to understanding how segments are grouped into larger units, such as syllables. Syllables are then grouped into larger prosodic units, called feet, in which one syllable is more prominent than the other, and which frequently reflects the rhythmic pattern of the language. Feet are essential units in the prosodic word. Just as there can be more than one syllable in a word, so too, there can be more than one foot. The following prosodic hierarchy illustrates the various prosodic units and what the larger category is above it. The most elemental prosodic unit is a mora, which can be referred to as a timing or a weight unit. Notice that a mora (µ) is below the syllable; only some segments in a syllable are linked to moras, and more about this will be discussed below. The symbols and abbreviations used to represent these prosodic units are indicated at the far left. 112 (8) Suzanne Urbanczyk Prosodic Hierarchy (Selkirk 1980) PWd prosodic word Ft foot σ syllable µ mora McCarthy and Prince (1986, 1990) proposed that meaningful units (morphemes) can be prosodic units; they developed several principles to account for some of the processes illustrated in (1–6). The basic idea is that rather than affixes and bases being composed solely in terms of segments, affixes and bases can also be prosodic units, which lack segments: they are empty prosodic units (or templates), that need to be filled by segments. A template is a meaningful prosodic unit, like syllable or foot, but does not have any segments in it. The cornerstone of this approach to understanding words like this is the Prosodic Morphology Hypothesis (PMH) and is defined as follows. (9) Prosodic Morphology Hypothesis (McCarthy and Prince 1990: 209) Templates are defined in terms of the authentic units of prosody: mora (µ), syllable (σ), foot (Ft), prosodic word (PWd) and so on. Other principles refer to how the prosodic units are filled with segments. There are also ways to identify the base, by referencing prosodic structure, as for example with the infixing pattern in (3). The patterns found in the languages of North America illustrate a wide range of ways in which prosodic units can be added to bases (as affixes) or can be targets to which processes apply (as bases) and we will look at patterns for each of these in sections 2 and 3 below. The process of reduplication involves doubling segments to indicate meaning, and there are some interesting patterns that arise when more than one reduplicative or prosodic morpheme is added to a word, which is the focus of section 4. 5.2 Prosodic units as affixes One of the key ways that prosodic morphemes are used to express meaning is when they are affixes. The basic insight here is that any type of prosodic unit can be associated with meaning and then affixed. Empty prosodic units (those that have no segments) gain segmental content by copying segments from the base to which they are attached. This section outlines patterns found to support each level of the prosodic hierarchy, starting with the mora, and building up to the syllable, then the foot and prosodic word. Prosodic morphology 113 5.2.1 Mora affixation A mora can be described as a timing unit, representing vowel and consonant length; it is also a unit of weight, which is relevant to locating stress (Hyman, 1985; Chapter 7, by Gordon). Let’s start by first discussing the relevant parts of a syllable. There are three main parts to a syllable. The onset is the first part of the syllable, coming before the vowel, and is always at least one consonant. The nucleus is the centre of the syllable and is usually a vowel, and the coda refers to any consonants at the end of a syllable. The nucleus and the coda are sometimes collectively called the rhyme. Interestingly, onsets do not play a role in determining where stress falls and so are not ever moraic. The following illustrates how length is encoded via moraic structure, with simple syllables. Following convention, C is used to represent any consonant and V is used to represent any vowel. The difference between a short and long vowel is in the number of moras: one for a short vowel and two for a long vowel, as in (10.a). (10) a. b. Moras as timing units Vowel length Short Long σ σ µ C V Consonant length Short σ σ µ µ C V C V µµ C V Long σ σ µ µ µ C V C V A short consonant is linked to just one position in the syllable, as in (10.b). Notice that it is only linked to the onset of the syllable – there are no moras for short consonants. On the other hand, long, or geminate consonants are represented in (10.b) and are linked to the mora and coda of the syllable before it, as well as the onset of the second syllable. The double linking represents double length. A range of patterns have been analyzed as affixation of a mora: vowel lengthening, consonant gemination, reduplication, epenthesis, subtractive morphology, and metathesis (Saba Kirchner, 2010; Zimmermann, 2017). The following examples from Alabama (Muskogean) straightforwardly show how adding a mora either results in a lengthened vowel (11.a) or a lengthened consonant (11.b), depending on the syllable structure and size of the base. Recall that length is represented by having a colon after the symbol that is long. For instance, in the forms in (11.a), the second to last vowel of the base is long in 114 Suzanne Urbanczyk the imperfect, and in (11.b) the consonant that begins the second to last syllable is long in the imperfect. The affected segments are indicated with shading. (11) Alabama imperfect (Hardy & Montler, 1988) base imperfect gloss a. b. Vowel lengthening hofna hó:fna ‘smell’ isko í:sko ‘drink’ isi í:si ‘take, catch’ coba có:ba ‘big’ ibakpila ibakpí:la ‘turn upside down’ Consonant lengthening (gemination) bala:ka bál:a:ka ‘lie down’ coko:li cók:o:li ‘sit down’ ilkowatli ilków:atli ‘move’ ataka:li atak:a:li ‘hang one object’ afinapli afin:apli ‘lock up’ The following diagrams illustrate how adding a mora lengthens the vowel (12.a) as well as the consonant (12.b). The affixed mora is indicated with shading under the column labeled ‘imperfect’. Notice that the shaded mora is affixed inside the first syllable – it is the rightmost mora in the first syllable. In (12.a) the mora affix is linked to a vowel – so the vowel becomes long – it has two moras. In (12.b), the mora affix is linked to a consonant – so is realized as a long consonant (a geminate). The length of the vowel or consonant is represented by how many moras it is linked to. (12) mora affixation a. b. Vowel lengthening base imperfect σ σ σ σ µ µ µµ µ c o b a c o b a Consonant gemination base imperfect σ σ σ σ σ σ µ µµ µ µµ µµ µ c o k o l i c ó k o l i ‘big’ ‘sit down’ Prosodic morphology 115 This way of representing the affix as a mora helps to explain the two different changes of lengthening as coming from the same input – a segmentally empty prosodic unit that needs to be filled. Because there are multiple ways to fill a mora as above, affixing a mora can also account for different forms of a morpheme (allomorphs), depending on what the segmental content of the base is. Other ways to fill a mora can be seen in Kwak’wala (Wakashan) in which the mora is realized as either vowel lengthening, when the initial syllable is one mora as in (13.a) when the first syllable of the base is a light syllable, or as reduplication, if the initial syllable is two moras – or heavy – as (13.b). Notice that nasal consonants n and m make the first syllable heavy, equivalent to a long vowel. There is independent evidence from stress that plain sonorant consonants are moraic, but obstruents and glottalized resonants are not. Kwak’wala shows that languages vary as to which coda consonants can be moraic. (13) a. b. Kwak’wala (Wakashan, Saba Kirchener 2013: 240) Vowel lengthening c̓əx → c̓a:x.m̓u:t ‘singe’ ̓kəxʷ ̓ ̓ → ka:xʷ.mu:t ‘suck with whole mouth’ t̓əs → t̓a:s.m̓u:t ‘crack barnacles’ Reduplication wən → wən.wə.mu:t ‘drill with auger’ ma: → ma:.mə.mu:t ‘(serpent) crawl’ c̓a:s → c̓a:.c̓əs.mu:t ‘eel grass’ ̓kəmƛ → k̓əm.k̓əɫ.m̓u:t ‘adze’ Another example of how mora affixation is associated with a range of allomorphs comes from Saanich (Salish): reduplication (14.a), metathesis (14.b), and glottal stop insertion/ infixing (14.c) have all be analyzed as prefixing a mora (Stonham, 1994). (14) Straits (SENĆOŦEN) actual allomorphy (Montler, 1986; Stonham, 1994) a. reduplication t̓ᶿéʔ ‘be on top’ t̓ᶿét̓ᶿəʔ ‘riding (a horse)’ s-kʷúl ‘school’ s-kʷúkʷəl ‘going to school’ ɫík̓ʷsən ‘trip’ ɫíɫək̓ʷsən̓ ‘tripping’ b. metathesis sx̣ə́t ‘push it’ sə́xṭ ‘pushing it’ θk̓ʷə́t ‘straighten it out’ θə́k̓ʷt ‘straightening it out’ ƛ̓pə́x̣ scatter’ ƛ̓ə́px̣ ‘scattering’ c. glottal stop infix (epenthesis) čáq̓ʷəŋ ‘sweat’ čáʔqʷəŋ̓ ‘sweating’ wéqʷəs ‘yawn’ wéʔqəs ‘yawning’ xʷítəŋ ‘jump’ xʷíʔtəŋ̓ ‘jumping’ 116 Suzanne Urbanczyk How the mora is filled depends on the segments in the base. Affixing a mora provides a way to see what these different patterns have in common. While they look like completely different processes, the range of patterns can be explained as filling an affixed empty mora. 5.2.2 Syllable reduplication A frequent shape of reduplicative morpheme is a syllable. As was seen above, there are light (one mora) and heavy (two moras) syllables. Each of these is attested in reduplication patterns. Examples of both light and heavy syllable reduplication are found in Plains Cree (Algonquian). Ahenekew & Wolfhart (1983, p. 370) note that the long vowel of the reduplicant is “typically followed by devoicing”, indicated by /h/ below. The reduplicant (copied portion) is indicated with underlining below. (15) a. b. c. Plains Cree (Ahenekew & Wolfart, 1983, p. 371) Base pimohte:‘walk along’ Light syllable ni-pa-pimohta:n ‘I am walking along’ Heavy syllable ni-pa:h-pimohta:n ‘I walk off and on’ In the examples above, only the first consonant is copied, as the vowel quality is [a]. The different shapes arise because a monomoraic syllable is affixed in (15.b), while a bimoraic syllable is affixed in (15.c). The difference in length is then achieved because the segmental quality of the vowel links to the empty moras in the syllable to fill the two moras, and creates a long vowel in (15.c) (Marantz, 1982; McCarthy & Prince, 1986). Hiaki (Uto-Aztecan) illustrates an instance of true syllable copying, where the first syllable of the base is copied (Haugen, 2014). If the stem begins with an open, codaless syllable, the whole syllable is copied (16.a). If the base stem begins with a closed syllable, as in (16.b), then that syllable including the coda is copied. (16) a. b. Syllable copying reduplication: Hiaki habitual (Haugen, 2014: 511) CV.CV-initial stems vu.sa vu.vu.sa ‘awaken’ či.ke či.či.ke ‘comb one’s hair’ ču.pa ču.ču.pa ‘grow (transitive verb)’ he.wi.te he.he.wi.te ‘agree’ CVC.CV-initial stems vam.se vam.vam.se ‘hurry’ čep.ta čep.čep.ta ‘jump over’ čuk.ta čuk.čuk.ta ‘cut with a knife or saw’ hit.ta hit.hit.ta ‘make a fire’ Hiaki syllable reduplication is rarer than other types of syllable reduplication: the shape of syllable doesn’t matter – it is copied exactly as it is in the base. So, it is proposed that Prosodic morphology 117 the affix is a syllable – it gets segments by copying the first syllable. No specification of weight is needed, as the weight varies. The example of ‘awaken’ was presented with full syllable structure in (7). 5.2.3 Foot/prosodic word reduplication When the reduplicant is two syllables in size, this often corresponds with being a foot in size. There are two basic types of prosodic feet, identifiable in terms of which syllable is more prominent (Hayes, 1995). When the initial syllable is emphasized (the strong syllable or head) it is a trochaic foot. When the final syllable is stronger (the head), it is an iambic foot. Each of these foot types can have various properties, including being monosyllabic. The basic classification and representation of feet as iambic and trochaic is provided below. Trochaic feet are binary, where the two elements are either having two syllables (for a syllabic trochee, as in 17.a) or having two moras (for a moraic trochee, as in 17.b). For languages with the system in (17.a), it doesn’t matter what kind of syllables they are, only that two syllables are grouped into a foot. Iambic feet are always quantity sensitive, which means that the number of moras is important, not the number of syllables. There is a preference for a light (one mora)-heavy (two moras) pattern, as seen in the initial foot type in (17.c). The stressed or strong syllable is indicated with an S below it. (17) a. b. c. Basic inventory of prosodic feet (Hayes, 1995) Syllabic trochee Ft σ σ S Moraic trochee Ft Ft σ σ σ µ µ µ µ S S Iamb Ft Ft Ft σ σ σ σ σ µ µ µ S µ µ S µ µ S 118 Suzanne Urbanczyk The following examples of Yurok (Algic) ‘repetitive’ reduplication illustrate what appears to be a moraic trochee. All the different types of moraic trochee are found in the language. The copied portion can be two light syllables (18.a) or two different types of heavy syllable: CVC- (18.b) and CV:- (18.c). The choice of reduplicant form depends on segmental properties of the base (Garrett, 2001). (18) a. b. c. Yurok ‘repetitive’ reduplication (Garrett, 2001) Bisyllabic meno:ʔ ‘pull anything’ meno-meno:tek̓ ‘I repeatedly pull’ yekʷoh(s-) ‘to fold’ yekʷo-yekʷoh ‘to fold several things’ CVCkep̓eɫ ‘housepit’ kep̓-kep̓eɫ ‘there are several housepits’ ket̓ey ‘to park, moor’ ket̓-ket̓ey- ‘to lie (boats)’ CV:maʔepet‘to tie up’ ma:-maʔepet‘to tie right up’ moʔohkeloyt ‘to make into a ball’ mo:-moʔohkeloyt ‘make several balls’ In all the forms above, CVCV, CVC and CV: are different ways to fill a trochaic foot affix. So, while it might seem that there are three distinct patterns, they can all be understood as a trochaic foot. The reduplicated words below from Fox (Algonquian) illustrate that bisyllabic size can correspond to a prosodic word. Recall from the introduction that there is a hierarchy of how prosodic units are built, where syllables can be grouped into feet and that feet are grouped into prosodic words. Teasing apart what is a foot vs. a prosodic word involves looking at what happens at the ends of words and seeing if that happens at the end of the reduplicant. If end of the reduplicant has the same properties as the end of a word, it is likely a prosodic word. Dahlstrom (1997, p. 218) observes that the second syllable of the reduplicant has the same patterning as syllables at the ends of words: both end with a short vowel, regardless of what the second syllable of the base is. Notice that even if the vowel in the second syllable of the base is long, as in (19.a) or is long and followed by a consonant, as in (19.b), the second syllable of the reduplicant always has the shape CV. (19) Fox prosodic word reduplication (Dahlstrom, 1997) a. či:.pi-či:.pi:.kʷe:-wa ‘he winks’ b. ko.kʷa-ko.kʷa:š.ke:-wa ‘he is jerked’ Having looked at ways that prosody is relevant to adding meaningful content to a base, we now turn to patterns in which segments are deleted and how that can be described in terms of syllable structure as well. Prosodic morphology 119 5.2.4 Subtractive morphology & truncation There are two general ways that material is deleted to indicate a new meaning in a word. In one, the segments associated with a prosodic unit are deleted. This is generally referred to as subtractive morphology: the meaning is indicated by subtracting a prosodic unit. In the other, the base is reduced in size to a prosodic unit; the segments not associated with a larger prosodic unit are deleted, which is generally referred to as truncation. Muskogean languages have subtractive morphology, with deletion of a specific prosodic unit, such as a rhyme or coda (Fitzgerald, 2016), as seen in the Koasati words indicating ‘plural’ meaning below. The deleted part of the word from the ‘singular’ is indicated with shading and underlining. Notice that this is the end of the syllable: either the rhyme (20.a) or coda (20.b). (20) Koasati ‘plural’ (Kimball, 1991) a. Rhyme-Deletion singular plural b. pitáf-fi-n akocofót-li-n tiwáp-li-n simát-li-n ataká:-li-n albití:-li-n apoɬó:-ka-n Coda-Deletion singular pít-li-n akocóf-fi-n tíw-wi-n sím-mi-n aták-li-n albít-li-n apóɬ-ka-n ‘to slice up the middle’ ‘to jump down’ ‘to open something’ ‘to cut up tanned skin’ ‘to hang something’ ‘to place on top of’ ‘to sleep with someone’ plural gloss asikóp-li-n atóf-ka-n kacáɬ-ɬi-n akapós-kan asiko:-li-n ató:-ka-n kacá:-li-n akapó:-kan ‘to breathe’ ‘to melt’ ‘to bite something’ ‘to be pinched’ gloss Another way that segments are deleted to indicate meaning is called truncation. In this case, rather than stating what is deleted in terms of syllable parts (like codas), the process involves deletion of everything so that the new word has a specific prosodic shape. An example of truncation from Central Alaskan Yup’ik (Eskimo-Aleut) illustrates the pattern, in which the target shape is a foot or prosodic word. These are names and so that is why they are all spelled with a capital letter. 120 Suzanne Urbanczyk (21) Central Alaskan Yup’ik (Woodbury, 1985) Full noun Proximal Vocative Aŋukaɣnaq Nupiɣak Kalixtuq Aɣnaɣayaq Aŋ ~ Aŋuk Nup ~ Nupix/Nupik Kal ~ Kalik Aɣən The variation in the number of syllables, as seen in the first form is because feet can have more than one shape, as noted above. A couple of points are worth noting here. First, these forms illustrate how the prosodic hierarchy works, from the prosodic word to the segments. Each form is a prosodic word (indicated as PWd below), which is exactly one foot. Second, the structure of the foot is different, but both forms are well-formed iambic feet. Recall from (17.c) that iambic feet can be one or two syllables. (22) PWd PWd Ft Ft σ σ σ µ µ µ µ µ a ŋ a ŋ u k 5.2.5 Prosodic repairs In some cases, there are changes or alternations to a base that appear to be motivated by improving the prosodic structure: these are referred to as prosodic repairs. For example, copying a single consonant has been analyzed as providing an onset to create wellformed syllables. There is a well-known preference for syllables to begin with consonants rather than vowels, and so copying a consonant helps to make it a better syllable. The following example from Spokane (Salish) illustrates that the repetitive morpheme /-e-/ can be an infix, if the stem begins with a cluster (23.a). If the stem begins with a single consonant, that consonant is copied to provide an onset for the repetitive /-e-/, as in (23.b). The copied segment is indicated with shading. (23) Spokane (Bates & Carlson 1992) a. Infix /-e-, šl̓-n̓-t-ən̓/ š-e-l̓n̓tén̓ rep, chop-ctr-tr-1sgTrS ‘I chopped it up repeatedly’ /-e-, lč̓-n̓-t-ən̓/ l̓-e-č̓n̓tén̓ rep, tie-ctr-tr-1sgTrS ‘I tied it over and over’ Prosodic morphology b. Consonant reduplication /-e-, šl̓/ rep, chop /-e-, nič̓-n̓-t-əxʷ/ rep, cut-ctr-tr-2sgTrS 121 še-šíl̓ ‘chopped repeatedly’ n̓e-n̓íč̓n̓txʷ ‘you kept cutting’ A second type of prosodic repair is vowel strengthening or ablaut and it has been motivated to avoid having schwas be said with stress or more emphasis when they are reduplicated (Urbanczyk, 2001). The following examples from Lushootseed (Salish) illustrate that the change of schwa to /i/ is motivated by stress, rather than being a segment associated with diminutive CV- reduplication. The stressed syllable is indicated with an accent mark above the vowel. Crucially, the vowel is only changed when stress falls on the reduplicant vowel (which is indicated with shading), as in (24.a). If stress falls on a schwa that is not reduplicated, it stays as schwa. If stress occurs later in the word, as in (24.b), the schwa is retained. (24) Lushootseed diminutive reduplication (Bates, Hess & Hilbert, 1994; Urbanczyk, 2006) a. Stressed reduplicant ǰə́səd ‘foot’ ǰí-ǰəsəd ‘little foot’ z z təd íl ‘lie in bed’ tí-təd il ‘lie in bed for a little while’ b. Unstressed reduplicant ƛ̓əládiʔ ‘sound, noise’ ƛ̓ə-ƛ̓əládiʔ ‘little noise’ táləɫ ‘nephew/neice’ tə-táləɫ ‘little nephew/neice’ As one can see, when the stress mark is on the shaded vowel, it is [i], but when the shaded vowel is not stressed, the vowel is [ə]. There are some good reasons proposed or why schwa avoids being stressed, mostly related to it being a neutral vowel. 5.2.6 Summary This section illustrated a range of patterns related to how the affixation of segmentally empty prosodic units can account for reduplication, metathesis, and deletion. This not only illustrates an important insight about prosodic morphology, but it also maintains a concatenative approach to morphology, especially in the face of what might otherwise be seen as evidence against that approach. We have also seen how having well-formed prosodic structure can account for when consonants are copied and also when vowel quality changes with the addition of a new meaningful unit to a base. The hypothesis that segmentally empty prosodic units are affixes, differs from other research, which aims to derive the prosodic shapes from constraints on morpho-prosodic well-formedness (Downing, 2005; Kennedy 2008; McCarthy & Prince 1999; Urbanczyk, 2006). In this view, the prosodic shapes of morphemes are emergent, rather 122 Suzanne Urbanczyk than directly listed in the input to the creation of a word. We turn next to looking at how prosodic structure is relevant in having well-formed bases to which morphological operations may apply. 5.3 Prosodic units as bases Two ways that prosodic units are relevant to morphological bases include the location of infixes and requiring certain prosodic shapes of bases. Each of these will be examined in turn. 5.3.1 Infixing location Yu (2007) defines an infix as an affix that separates a morpheme into two parts that would not otherwise occur independently. This is necessary to distinguish infixing from other types of internal modification, like the ablaut pattern discussed above. Yu’s typological survey of the location of infixes revealed that they are located in noticeable, or salient positions, which he refers to as pivots. There are two classes of pivots: edge pivots and prominence pivots. The use of the term ‘edge’ is meant to refer to the first and last parts of words, and prominence is related to stress. Yu provides the following list of potential pivot locations. The parentheses identify positions which have been observed by others, which Yu did not find unequivocal evidence to support. (25) Potential pivot locations (Yu, 2007, p. 67) Edge pivots Prominence pivots First consonant First vowel (First syllable) Last syllable Last vowel (Last consonant) Stressed foot Stressed syllable Stressed vowel An example of an infix found with an edge pivot is the first-person subject marker ‘I’ in Dakota (Siouan): it comes after the first vowel, which also happens to be the first syllable in the words below. (26) Dakota 1st person (Yu, 2007: 102, citing Boas & Deloria 1941, Moravscik 1977) ʔi.kto.mi ‘Iktomi’ ʔi-ma-kto.mi ‘I am Iktomi’ na.pcu ‘swallow’ na-wa-pca ‘I swallow it’ la.k’o.ta ‘Lakota’ la-ma-k’o.ta ‘I am a Lakota’ na.wi.zi ‘jealous’ na-wa-wi.zi ‘I am jealous’ Prosodic morphology 123 An example of a prominence pivot can be found with infixing reduplication. The language isolate Washo illustrates infixing to the stressed syllable, to indicate plural meaning. A mora is attached and receives its segmental content by copying segments from the base, similarly to Kwak’wala reduplication. Notice that stress (marked with an accent over the stressed vowel) in all the words below is on the second to last syllable, regardless of whether or not there is reduplication. The words in (27.a) illustrate that the segments copied are adjacent, and the words in (27.b) show that a coda consonant can be skipped in order to copy the onset and nucleus of the stressed syllable. (27) Washo plural infixing reduplication (Jacobsen, 1974; Winter, 1970) singular plural gloss a. b. ʔélel gúšuʔ dámal bókoŋ ʔéw.šiʔ nén.t̓uš mók.go ʔe-lé-lel gu-šú-šuʔ da-má-mal bo-kó-koŋ ʔe.-ší-w.šiʔ ne.-t̓ú-n.tuš-u mo.-gó-k.go ‘mother’s father’ ‘pet’ ‘to hear’ ‘to snore’ ‘father’s brothers’ ‘old women’ (u=nominalizing suffix) ‘shoe’ 5.3.2 Stem modifications based on prosodic units In addition to prosodic units determining what can be affixed to a word and where an affix is placed, prosodic units can also serve to require a particular shape and size of the base. The patterns either relate to expansion of the stem to some target size or to a requirement that affixes be attached to a particular type of base (where the base adjusts its syllable structure for a particular affix). In terms of stem expansion, Cupeño (Uto-Aztecan) habilitative meaning is expressed by identifying the stressed foot (indicated in square brackets on the left) and expanding the form, so the stressed syllable is followed by two syllables. The expansion occurs by copying the vowel and inserting a glottal stop [ʔ] onset. The copied portion is indicated by underlining on the right, and the two syllables after the stressed syllable are shaded. (28) Cupeño habilitative (Hill, 1970; McCarthy & Prince, 1990; Crowhurst, 1994) Verb Stem Habilitative a. [čál] čá-ʔaʔa-l ‘husk’ [tə́w] tə́-ʔəʔə-w ‘see’ hə[lyə́p] həlyə́-ʔəʔə-p ‘hiccup’ kə[láw] kəlá-ʔaʔa-w ‘gather wood’ b. [páčik] páči-ʔi-k ‘leach acorns’ [čáŋnəw] čáŋnə-ʔə-w ‘be angry’ 124 Suzanne Urbanczyk c. d. [pínəʔwəx] [xáləyəw] čí hú ʔáyu pínəʔwəx xə́ləyəw číʔ húʔ ʔáyu ‘sing enemy songs’ ‘fall’ ‘gather’ ‘fart’ ‘want’ As one can see, there is no affix per se, but rather the ‘habilitative’ meaning is expressed by the expansion of the base to meet a target size. How many vowels are copied depends on the number of syllables in the base and where stress falls, but the output is the same: the stressed syllable is followed by two unstressed syllables, as indicated by shading. In (28.a) two syllables are added, because the base has final stress. In (28.b), only one syllable is added, because stress is on the second to last syllable in the verb stem. In (28.c), the stressed syllable is already followed by two syllables, so no copying is needed. And interestingly, in (28.d), the verb stems all end in vowels, so a different pattern is used. Languages in which affixes require the base to have a specific shape include Zuni (isolate), Sierra Miwok (Utian), and Yowlumne (Yokuts). In Sierra Miwok, there are several stem shapes which change, depending on the suffix, and the root’s composition of consonants and vowels. The following illustrates four different stem shapes found. These examples show that different bases can have different shapes, depending on the class, changing the length of the vowel or consonant or adding extra material. (29) Templatic morphology – Southern Sierra Miwok (Broadbent 1964, p. 38) Stem 1 Stem 2 Stem 3 Stem 4 meaning mus:alo:tkowta- musahlotkowat- mus:aʔlot:uʔkow:at- musʔalotʔukowta- ‘to be ashamed’ ̓to catch’ ‘to bump into’ As you can see above, each class has a set pattern regarding the length of consonants, such as geminating the second consonant in the third class. 5.3.3 Summary This survey has illustrated a range of ways in which prosodic structure is relevant in determining the base of some morphological operation: either by identifying the location of infixes, or by adjusting the shape of the base for a particular affix. Some patterns such as truncation also show how bases can be adjusted to have a specific shape to express a particular meaning, an example of that was given from Yup’ik in (21). When learning new words that use prosodic morphology, it may appear that the processes are not familiar or that there are many different forms to keep track of. However, once one can see that prosodic structure is relevant, one can make equivalences between forms, which could be useful thus provide a guide for how to create new words. Prosodic morphology 125 5.4 Multiple prosodic morphemes and multiple exponence Much of the work in prosodic morphology has focused on how a single morpheme or meaning is realized. A growing body of research has started to explore what happens when more than one prosodic morpheme is added or there are multiple parts to the prosodic morphemes. This section outlines some of the patterns that occur when more than one prosodic morpheme is added to a word. 5.4.1 Double reduplications Zimmermann (2021) outlines a typology of patterns found when there is more than one reduplicative morpheme used to form a new word, referring to these as multiple reduplication. She observes three patterns: all reduplicants surface as expected (faithful multiple reduplication), only a single reduplicant surfaces (avoidance of multiple reduplication), or all reduplicants surface, but one is smaller than expected (subtracting multiple reduplication). She identifies multiple reduplication occurring in Algonquian, Salish and Wakashan languages, as well as languages spoken outside of North America. Examples of faithful multiple reduplication are found in Thompson (Salish) and Fox (Algonquian). Interestingly, the outermost reduplicant appears to copy the adjacent segments, as can be seen in the Fox example below. The base word is presented in (30.a). Notice that in (30.b) the prosodic morpheme only copies the consonant; instead of copying the vowel, a fixed long vowel /a:/ occurs, as indicated with shading. The form in (30.c) illustrates that the second reduplicative morpheme is a bisyllabic foot, faithfully copying the vowels in the adjacent base. In (30.d) we see the pattern with both reduplicative morphemes. (30) Faithful multiple reduplication in Fox (Dahlstrom, 1997, 206–207) a. wi:tamaw-e:wa ‘he tells him’ b. wa:-wi:tamaw-e:wa ‘he tells him over and over’ c. wi:ta-wi:tamaw-e:wa ‘he keeps telling him’ d. wa:wi-wa:-wi:tamaw-e:wa ‘he keeps telling him over and over’ When the bisyllabic reduplicant is located before the /C1a:-/ reduplicant in (30.d), it copies the long vowel of the reduplicant, not the root vowel. This shows that the outermost reduplication occurs after the first one is added; otherwise it would be puzzling as to how both reduplicants can have the fixed /a:/ vowel. This discussion of attaching the prosodic morphemes in a specific order will also be relevant below. Other examples of faithful multiple reduplication can be found in Salish languages, such as Thompson (Broselow, 1983). The basic stem is given in (31.a). A CVC- reduplicant indicates ‘distributive’ meaning in (31.b). Diminutive is CV- (31.c), and both are as 126 Suzanne Urbanczyk expected when combined in (31.d): the vowels and consonants are as expected, and the shapes are also as expected. Notice that it looks like the ‘distributive’ skips over the ‘diminutive’ in order to copy the second consonant of the root. (31) Faithful multiple reduplication in Thompson (Broselow 1983, Thompson & Thompson 1991) a. sil ‘calico’ b. sil-síl ‘calico, distributive’ c. sí-sil’ ‘calico, diminutive’ d. sil-sí-sil’ ‘calico, distributive-diminutive’ Broselow (1983) notes that this skipping effect is because the ‘diminutive’ is actually an infix. It is attached to the stressed syllable, as can be seen with longer words with non-initial stress, shown in the examples below. (32) Thompson infixing reduplication a. qʷintwáxʷ ‘talk to each other’ b. qʷintwáw’xʷ ‘talk to each other, diminutive’ The ‘distributive’ is a true prefix, so must attach to the root. Subtracting multiple reduplication is defined as: “One of the reduplicants is smaller than its form in isolation” (Zimmermann, 2021, p. 542). An example of this pattern can be found in Lushootseed (Salish), with ‘distributive’ reduplication: this morpheme is shaded in the examples below to show how it changes depending on where it occurs in the word. The ‘distributive’ usually has the shape CVC as in (33.b). When it comes after ‘diminutive’ Ci- reduplication in (33.c), it has its usual CVC shape, as illustrated in (33.d). However, when the ‘distributive’ precedes ‘diminutive’, it is reduced in size to CV-, as in (33.e). (33) Subtracting multiple reduplication in Lushootseed (Bates, Hess & Hilbert 1994) a. bədaʔ ‘child, offspring’ b. bəd-bədaʔ ‘children’ ‘distributive’ c. bíʔ-bədaʔ ‘young child’ ‘diminutive’ d. bí-bəd-bədaʔ ‘litter (of animals); dolls’ ‘diminutive-distributive’ e. bí-bi-bədaʔ ‘young children’ ‘distributive-diminutive’ The examples below verify that this is in fact reduplication, rather than affixation, as was suggested by a reviewer. The only difference in the patterns is that there is deletion of the base vowel with ‘diminutive’ reduplication. (34) Subtracting multiple reduplication in Lushootseed (Bates, Hess & Hilbert 1994) pástəd ‘white person, Caucasian’ pas-pastəd ‘many white folks’ ‘distributive’ pa-pstəd ‘white child, white friend’ ‘diminutive’ pa-pa-pstəd ‘many white children’ ‘distributive-diminutive’ Prosodic morphology 127 The reduction in size of the ‘distributive’ to CV- from CVC- has been argued by Urbanczyk (2001) to be due to purely phonological processes: it can be attributed to avoiding doubled consonants or geminates. Lushootseed doesn’t have any geminates, and the ill-formed *bíb-bi-bədaʔ would be independently ruled out because a sequence of two segments like bb is never found in the language (contra Broselow 1983 and Zimmermann 2021). The following section looks at cases in which two reduplicative morphemes are avoided. 5.4.2 Coalescence of prosodic morphemes Turning to avoidance of multiple reduplication, Zimmermann (2021) refers to this as coalescence and discusses examples from Wakashan languages. In Nuu-chah-nulth (Wakashan), when more than one morpheme is present that is associated with reduplication, only one reduplicant arises: the two prosodic units are fused by coalescence. Nuu-chah-nulth, like other Wakashan languages, has a number of stem changes associated with various suffixes. These stem changes include triggering reduplication, lengthening or shortening the vowel in the base or the reduplicant, and also some segmental changes (not indicated here). The changes that the suffix triggers in the base are indicated in the second line below, where R indicates reduplication, and L indicates that the base or reduplicant vowel is lengthened. Multiple reduplications occur, but only in very restricted contexts, as illustrated in (35). Stonham (2007) presents evidence that when multiple reduplication does happen, it is when the reduplicative morphemes are associated with different stages of word formation. As in the example below, the word for ‘sea otter’ is created first, with reduplication being triggered by the suffix meaning ‘look for’. Then, once this is created, the word is fully formed and then it can be pluralized by having CV- reduplication. (35) Multiple reduplication in Nuu-chah-nulth (Stonham, 2007) k̓ʷa-k̓ʷa-k̓ʷaƛ̓iiḥ RED-RED-k̓ʷaƛ-iiḥ PL-RED-sea.otter-look.for[R] ‘sea otter hunters’ So, the first reduplication occurs when the stem is created, the second reduplication occurs after this, when the final word is created. However, if two suffixes are added that trigger reduplication at the same level, only one reduplication is found. This can be seen in the word in (36) below. Notice that it has two suffixes that trigger reduplication, but only one reduplicant (underlined) occurs in the word. The reduplicant has a long vowel as resulting from coalescing the two prosodic units associated with the two suffixes ‘at.leg’ and ‘really’. 128 (36) Suzanne Urbanczyk Avoidance of multiple reduplication in Nuu-chah-nulth1 (Kim, 2003) ƛ̓uu-ƛ̓uukʷan̓łap RED-ƛuk-aan̓uł[R+L]-apa[RL+L] RED-broad-at.leg-really ‘His legs are really big’ An interesting feature of this avoidance pattern is that when reduplication co-occurs with a suffix it can be omitted if needed, as it is not the sole marker of meaning. Plural reduplication, as in (35), is one of the only reduplication patterns that does not co-occur with a suffix. This raises the question of whether the avoidance of multiple reduplication occurs when reduplication co-occurs with an affix, or whether the effect is related to avoiding multiple reduplications altogether. Some examples from Halkomelem (Salish), discussed in § 5.4.3, seem to suggest that multiple reduplication is avoided when it is the sole indicator of meaning. Another pattern of coalescence can be found in Halkomelem (Salish). The prosodic morphemes for ‘plural’ and ‘imperfective’ are coalesced when they co-occur with some bases. The base form of the verb is provided in (37.a). The imperfective is CV- reduplication in (37.b), with plural being a CəC- prefix, as in (37.c). (37) Halkomelem (Salish) coalescence of prosodic morphemes (Hukari, 1978) a. t̓íləm ‘sing’ b. t̓í-t̓ələm̓ ‘singing’ imperfective c. t̓əl-t̓íləm ‘they sing’ plural d. t̓íl-t̓ələm̓ ‘they are singing’ plural-imperfective When both morphemes are combined, the result is CVC- reduplication with initial stress as in (37.d), indicated with shading. It should be pointed out that the ‘imperfective’ for the most part is associated with initial stress, so this looks like the prosodic unit associated with imperfective coalesces with the syllable associated with plural reduplication to result in CVC- initially stressed word. 5.4.3 Avoiding reduplication with allomorph selection A fourth pattern of how meanings are expressed with multiple prosodic morphemes has been identified (Mellesmoen and Urbanczyk, 2021). Halkomelem (Salish) illustrates a pattern in which reduplication is avoided if another allomorph can be chosen to express the meaning. The ‘imperfective’ has a range of allomorphs, including reduplication 1 An anonymous reviewer has asked for examples of intermediate words. A search of available sources including Rose (1980), the source for many of Stonham’s (2007) examples, has not revealed any words that would be relevant, where they have the same root or base and the relevant suffixes to trigger reduplication. Prosodic morphology 129 (38a), metathesis (38b), h- prefixation (38c), ablaut (38d) and glottal stop insertion/infixing (38e; Hukari, 1978), as illustrated below. The relevant portion of the word that shows the allomorph is indicated with shading. (38) Halkomelem imperfective allomorphy (Hukari & Peter, 1995) a. reduplication t̓íləm ‘sing’ t̓i-t̓ələm̓ ‘singing’ b. metathesis pqʷát ‘break it’ páqʷt ‘breaking it’ c. h- prefixation mə́q̓ət ‘swallow it’ hə́m̓q̓ət ‘swallowing it’ d. ablaut č̓ə́kʷx̣ ‘fry’ č̓ékʷx̣ ‘frying’ e. infix hésəm ‘sneeze’ héʔsəm ‘sneezing’ Likewise, the ‘plural’ has several different allomorphs. Two of these allomorphs involve reduplication, and one does not. In (39a), the plural is C1əC2- reduplication and in (39b) it is a -C1i- infix. As can be seen in (39c), plural meaning is expressed with a /-l̓-/ infix. (39) Halkomelem plural allomorphy (Hukari, 1978) a. C1əC2- reduplication lémət ‘look at him/her/it’ ləm-lémət t̓íləm ‘sing’ t̓əl-t̓íləm b. -C1i- infixing reduplication kʷəmləxʷ ‘root’ kʷə-kʷí-mləxʷ šə́yəɫ ‘older sibling’ šə-ší-yəɫ c. -l- infix técəl ‘arrive, reach’ tél̓əcəl méčəs ‘match’ mél̓əčəs ‘look at them’ ‘they sing’ ‘roots’ ‘older siblings’ ‘they arrive’ ‘matches’ When these prosodic morphemes are combined with ‘diminutive’ reduplication on verbs (which is always reduplicative), the range of allomorphs found with the plural is reduced to just the /-l̓-/ infix, avoiding the reduplicative forms. The plural is indicated with shading to illustrate where it occurs in the word (the location of the epenthetic schwa varies). (40) Halkomelem multiple morphemes (Hukari, 1978) diminutive-imperfective diminutive-imperfective-plural a. ɬiʔɬəɬə́nəm̓ ‘weaving’ ɬəliʔɬəɬə́nəm̓ ‘weaving’ b. piʔpaqʷt ‘breaking it’ pəliʔpaqʷt ‘breaking them’ While allomorphy is related to choosing one of several forms to express a meaning, there are also patterns in with several forms are required to express a single meaning. We turn to this next. 130 Suzanne Urbanczyk 5.4.4 Multiple exponence: Affix-triggered reduplication Multiple exponence refers to when a single meaning is expressed by multiple realizations within a word (Harris, 2017, p. 9). There is very little research on multiple exponence and prosodic morphology, so the following discussion represents a preliminary overview, which should be subject to verification by a balanced typological survey. In some languages, reduplication co-occurs with the addition of a segmentally specified affix. This affix-triggered reduplication can be found in the Wakashan languages, as illustrated in the Nuu-chah-nulth examples in (35 – 36) above. For example, in (35), the suffix meaning ‘look for’ triggers copying the first consonant and vowel of the base. The striking feature about affix-triggered reduplication is that it seems to be found only with languages which have independently motivated reduplication as well as a very large inventory of affixes. For example, in Nuu-chah-nulth, reduplication on its own indicates plurality, but there are approximately 80 suffixes which trigger reduplication (as well as other alternations to the base). Nuu-chah-nulth has over 400 affixes, and approximately 20 % of the suffixes are also associated with reduplication. A preliminary investigation of this pattern found that most of the suffixes that trigger reduplication have homophonous or near homophonous counterparts, so reduplication may serve to disambiguate the suffixes from each other (Lee and Urbanczyk, 2006). The following is a partial list of some of the homophonous affixes in Nuu-chah-nulth, with the kind of change indicated after the meaning. (41) Nuu-chah-nulth homophonous affixes (Sapir and Swadesh, 1939; Lee and Urbanczyk, 2006) form meaning modification to base a. b. c. -aɬc̓a -aɬc̓a -im -im -paɬ -paɬ -paɬ ‘at fault’ ‘at an upright surface’ ‘through an aperture’ ‘… thing’ ‘on each side’ ‘in the same group with’ ‘season of’ reduplication, vowel lengthening no change reduplication, vowel lengthening no change reduplication vowel lengthening no change While this pattern is interesting, much further work is needed to document the specific details of the words as well as to verify if the observation holds that languages with affix-triggered reduplication also have homophonous affixes, and this is not an accidental correlation. To summarize, this section has reviewed some examples of multiple prosodic morphemes. A general typology of multiple reduplication as found by Zimmerman (2021) was presented, along with other evidence that reduplication is avoided when possible. Prosodic morphology 131 5.5 Relevance to language revitalization A key issue that arises in the context of language revitalization, is to identify the best way to teach and learn these patterns. Some of the patterns seem quite complex when first looking at them, and they often reflect important and common meanings that are acquired or learned early, like plurality or that an action is on-going. For example, the Halkomelem ‘imperfective’ forms in (38) express a very common meaning in the language. Viewing the range of changes as adding a syllable or a mora helps to see what gets added to the word, when it is not simply a suffix with consonants and vowels in it, like English -ing. There is very little published research on effective ways to teach and learn new words that have prosodic morphemes, even for languages outside North America that have reduplication and which have many speakers and learners (for example Chinese). Some recent work on this is outlined in the hopes of inspiring others to share their approaches more widely. Tohono O’odham (Uto-Aztecan) makes use of several patterns of reduplication to express plurality. Beer, Cruz, Hirrel and Kerfoot (2014) describe several patterns found that depend on the syllable shape, in particular the number of moras in the base, vs. the reduplication pattern. They outline a path to follow to teach the pattern, where students first learn to identify consonants and vowels and what counts as moraic in Tohono O’odham, without using formal linguistic terms. They then provide a series of exercises to help learners identify the patterns, asking learners to predict the correct form. This provides a way for learners to be able to find the patterns and apply them to new forms. Related to learning patterns in this way, I recently had the opportunity to co-develop some fun ways to present Halkomelem imperfective patterns with a cohort of language learners. One example uses flash cards for Halkomelem, in which the base form is written on one side of a card, while the ‘imperfective’ form is written on the other side (Claxton, Urbanczyk, & Hul’q’umi’num’ Language Academy, 2019). Once we did this and developed a game, learners expressed that it was easier to internalize the pattern by playing a game, trying to match the form with the meanings, rather than just going through all the rules. Other ideas one can use are to make schemas of word patterns, using symbols like C and V, and filling the blanks with the correct letter, to create new words. It also might be helpful to link some of the patterns of repetition in reduplication with other culturally significant patterns of repetition, such as with knitting or weaving (Urbanczyk, 2021). This is an important and understudied area of language revitalization, and it is hoped that by presenting the patterns with representations of the prosodic morphemes, people will start to share their ideas and successes with teaching and learning these patterns. 132 Suzanne Urbanczyk 5.6 Conclusions The proposal that morphemes can be prosodic units has been a key development in understanding a range of word formation processes which are not strictly affixational. Some of the patterns found can be described as affixing prosodic units such as moras, syllables, or feet. Doing so allows one to see that there is just one affix, but it is pronounced differently, depending on what the sounds of the base are. Prosodic structure and patterns are also relevant to identifying bases for some affixes. Multiple reduplication patterns can be found in several language families (Algonquian, Salish, Wakashan), and some general patterns regarding how these are realized was discussed as well. A few final thoughts about how these patterns may be taught was also shared. References Ahenekew, Freda & Hans Christoph Wolfart. 1983. Productive reduplication in Plains Cree. In William Cowan (ed.), Proceedings of the Algonquian Conference, 369 – 378. Ottawa, ON: Carleton University Press. Bates, Dawn & Barry F. Carlson. 1992. Simple syllables in Spokane Salish. Linguistic Inquiry 23. 653–659. Bates, Dawn, Thomas. M. Hess & Vi Hilbert. 1994. Lushootseed dictionary. Seattle: University of Washington Press. Beers, Keiko, Robert Cruz, Laura Hirrel & Iphigenia Kerfoot. 2014. Describing reduplication patterns in Tohono O’odham with language learners in mind. Proceedings of the High Desert Linguistics Society Conference, 43–55. University of New Mexico. Boas, Franz & Ella C. Deloria. 1941. Dakota grammar. (Memoirs of the National Academy of Sciences 23. part 2). Washington, DC: US Government Publishing Office. Broadbent, Sylvia M. 1964. The Southern Sierra Miwok language. Berkeley, CA: University of California Press. Broselow, Ellen. 1983. Subjacency in morphology: Salish double reduplication. Natural Language and Linguistic Theory 1. 317 – 346. Claxton, Rae Anne, Suzanne Urbanczyk & Hul’q’umi’num’ Language Academy. 2019. Hul’q’umi’num’ verb classes: Phonological word schemas. Presentation at 54th International Conference on Salish and Neighbouring Languages, Victoria, BC. Crowhurst, Megan. 1994. Foot extrametricality and template mapping in Cupeño. Natural Language and Linguistic Theory 12. 177–201. Dahlstrom, Amy. 1997. Fox (Mesquakie) reduplication. International Journal of American Linguistics 63(2). 205 – 226. Downing, Laura. 2006. Canonical forms in prosodic morphology. Oxford UK: Oxford University Press. Fitzgerald, Colleen. 2016. Morphology in the Muskogean languages. Language and Linguistic Compass 10. 681 – 700. Garrett, Andrew. 2001. Reduplication and infixation in Yurok: morphology, semantics and diachrony. International Journal of American Linguistics 67. 264–312. Hardy, Heather K. & Timothy R. Montler. 1988. Imperfective gemination in Alabama. International Journal of American Linguistics 54. 399 – 415. Harris, Alice. 2017. Multiple exponence. Oxford, UK: Oxford University Press. Haugen, Jason. 2014. Hiaki (Yaqui) reduplication redux: morphological and prosodic targets in syllabic reduplication. International Journal of American Linguistics 80(4). 507 – 532. Hayes, Bruce. 1995. Metrical stress theory: principles and case studies. Chicago: University of Chicago Press. Prosodic morphology 133 Hill, Jane H. 1970. A peeking rule in Cupeño. Linguistic Inquiry 1. 534–539. Hukari, Thomas. 1978. Halkomelem nonsegmental morphology. Proceedings of the 13th International Conference on Salish and Neighbouring Languages, 157 – 209. https://lingpapers.sites.olt.ubc.ca/ icsnl-volumes/. Hukari, Thomas & Ruby Peter. 1995. The Cowichan dictionary of the Hul’q’umi’num’ dialect of the Coast Salish people. Duncan BC: Cowichan Tribes. Hyman, Larry. 1985. A theory of phonological weight. Dordrecht: Foris Publisher. Kennedy, Robert. 2008. Evidence for morphoprosodic alignment in reduplication. Linguistic Inquiry 39. 589–614. Kim, Eun-Sook. 2003. Theoretical issues in Nuu-chah-nulth phonology and morphology. Ph.D. thesis, University of British Columbia. Kimball, Geoffrey D. 1991. Koasati grammar. Lincoln, NE: University of Nebraska Press. Lee, Sunghwa & Suzanne Urbanczyk. 2006. Reduplication and vowel lengthening/shortening in Nuu-chah-nulth. Presentation at 2nd International Conference on Wakashan Languages, Victoria, BC: University of Victoria. Marantz, Alec. 1982. Re reduplication. Linguistic Inquiry 13. 483–545. McCarthy, John & Alan Prince. 1986. Prosodic morphology. Manuscript. University of Massachusetts, and Brandeis University. McCarthy, John & Alan Prince. 1990. Foot and word in prosodic morphology: The Arabic broken plural. Natural Language and Linguistic Theory 8. 209 – 283. McCarthy, John & Alan Prince. 1999. Faithfulness and identity in prosodic morphology. In R. Kager, H. van der Hulst & W. Zonneveld (eds.), The prosody morphology interface, 218–309. Cambridge, UK: Cambridge University Press. Mellesmoen, Gloria & Suzanne Urbanczyk. 2021. Avoiding multiple reduplication without Integrity. In D. K. E. Reisinger & M. Huijsmans (eds.), Proceedings of 37th West Coast Conference on Formal Linguistics, 169–178. Boston MA: Cascadilla Press. Montler, Timothy R. 1986. Morphology and phonology of Saanich. Missoula, MT: University of Montana Press. Moravscik, Edith. 1977. On rules of infixing. Bloomington, IN: Indiana University Linguistics Club. Saba Kirchner, Jesse. 2010. Minimal reduplication. Ph.D. thesis, UC Santa Cruz. Saba Kirchner, Jesse. 2013. Minimal reduplication and reduplicative exponence. Morphology 23. 227–243. Sapir, Edward & Morris Swadesh. 1939. Nootka texts: tales and ethnological narratives, with grammatical notes and lexical materials. Philadelphia, PA: Linguistic Society of America, University of Pennsylvania. Selkirk, Elisabeth O. 1980. The role of prosodic categories in English word stress. Linguistic Inquiry 11. 563 – 605. Stonham, John. 1994. Combinatorial morphology. (Current Issues in Linguistic Theory). Amsterdam: John Benjamins. Stonham, John. 2007. Nuu-chah-nulth double reduplication and Stratal Optimality Theory. Canadian Journal of Linguistics 52. 105 – 130. Thompson, Lawrence. C. & M. Terry Thompson. 1991. The Thompson language. Missoula, MT: University of Montana Occasional Papers in Linguistics. Urbanczyk, Suzanne. 2001. Patterns of reduplication in Lushootseed. New York, NY: Garland. Urbanczyk, Suzanne. 2006. Reduplicative form and the root-affix asymmetry. Natural Language and Linguistic Theory 24. 179–240. Urbanczyk, Suzanne. 2021. A note on how weaving and knitting can enhance learning Salish reduplication patterns. In M. Desmarais & J. Wu (eds.), Working Papers of the Linguistics Circle of the University of Victoria, Vol. 31(1), 148–156. Victoria, BC: University of Victoria. Woodbury, Anthony. 1985. Meaningful phonological processes: A consideration of Central Alaskan Yupik Eskimo prosody. Unpublished manuscript. University of Texas, Austin. Yu, Alan. 2007. A natural history of infixation. Oxford, UK: Oxford University Press. 134 Suzanne Urbanczyk Zimmermann, Eva. 2017. Morphological length and prosodically defective morphemes. Oxford UK: Oxford University Press. Zimmermann, Eva. 2021. Two is too much…in the phonology!: A phonological account of unfaithful multiple reduplication” The Linguistic Review 38(3). 537–572. https://doi.org/10.1515/tlr-2021-2075 Matthew K. Gordon 6 Word prosody Abstract: Many phonological properties operate over the domain of the word. For example, stress is typically bounded by the word, as are other phenomena sensitive to word edges, including tone, prosodic minimality conditions, harmony processes, and segmental alternations. North American Indian languages have provided a fertile ground for studying word level prosody due to their characteristic high degree of morphological synthesis, which gives rise to long words that supply a rich backdrop for word-bounded phonological properties. This chapter examines phonetic and phonological aspects of word prosody in North American Indian languages from both a synchronic and diachronic perspective, with a particular focus on stress and metrical structure, areas of research in which North American languages have greatly informed typological knowledge. 6.1 Introduction Word prosody refers to features of a language’s sound system that are sensitive to the domain of the word. For example, stress is typically bounded by the word, as are other phenomena sensitive to word edges, including tone, word minimality conditions, harmony processes, and segmental alternations. North American Indigenous languages have provided a fertile ground for studying word level prosody due to their characteristic high degree of morphological synthesis, which gives rise to long words that supply a rich backdrop for word-bounded phonological properties. Word prosody is also an important area of study from the point of view of language acquisition since accurate production of prosodic features is crucial for intelligibility and comprehensibility. This chapter examines phonetic and phonological aspects of word prosody in North American Indian language from both a synchronic and diachronic perspective, with a particular focus on stress and metrical structure, areas of research in which North American languages have greatly informed typological knowledge. 6.2 Prosodic structure of the word A prosodic word is composed of one or more syllables, which in turn consist typically of a vowel nucleus (syllable peak) and one or more consonants on either side of the nucleus (see Fitzgerald & Gordon this volume, on syllable structure). A language may also provide evidence for a unit smaller than the word comprised of a group of two adjacent syllables. This unit, the foot, is diagnosed primarily through stress, which is https://doi.org/10.1515/9783110600926-006 https://doi.org/10.1515/9783110600926-000 136 Matthew K. Gordon increased prominence associated with certain syllables in a word. For example, in the English noun INsight, the first syllable is stressed whereas the verb inCITE minimally differs in its pronunciation in having stress on the second syllable. Longer words often have a secondary stress that is less prominent than the primary stress but stronger than other syllables in a word. For example, the word HEliCOPter has two stresses, a primary stress on the first syllable and a secondary stress rhythmically placed two syllables away on the third syllable. Speakers of some languages, including English, have strong intuitions about stress that are often accessible through the tapping test in which the speaker taps once while saying a word. Typically, a single tap will align with the stressed syllable of a word and a second tap, if introduced, will fall on a secondary stressed syllable. In other languages, however, speakers’ intuitions about stress may be less salient and other phonetic diagnostics of stress may be necessary to consider. It is common to characterize rhythmic stress patterns in terms of disyllabic feet, where the foot corresponds to the groupings commonly used to describe many poetic traditions, e. g., iambic pentameter in Shakespearean verse and dactylic hexameter in Greek and Latin poetry. For example, the word HEliCOPter can be divided into two feet indicated by parentheses, i. e., (HEli)(COPter). In this word, the two feet are termed trochaic feet since they consist of a stressed syllable followed by an unstressed syllable. Feet are discussed further in section 6. A schematic diagram of the prosodic structure of the word (HEli)(COPter) appears in Figure 1 (see Urbanczyk 2023 [this volume] for more on the prosodic structure of the word). Word Ft Ft Syll Syll Syll Syll HE li COP ter Fig. 1: The prosodic structure of the word ‘helicopter’ 6.3 Phonetic correlates of stress There are many physical correlates of stress. In some languages, consonants and vowels have noticeably different realizations depending on stress. Stressed syllables are typically associated with stronger or hyperarticulated sounds and unstressed syllables are characterized by weaker or hypoarticulated sounds. For example, in English, voiceless stops (sounds produced without vibration of the vocal folds) are aspirated (i. e., produced with a puff of air after release) before stressed vowels (though not after /s/) and most unstressed vowels centralize to schwa. Thus, in the word apPENdix, the /p/ (in Word prosody 137 bold) is aspirated in the onset of the stressed second syllable but the same /p/ is not aspirated in APpenDECtomy because it occurs before an unstressed vowel. The vowel in the second syllable also alternates in quality between the two words as a function of stress; it is realized with a more front quality [ɛ] when stressed in appendix but as a central, schwa-like, vowel [ə] when unstressed in appendectomy. Stress is typically realized phonetically through one or more acoustic and articulatory properties, including greater duration, increased intensity (loudness), higher fundamental frequency (pitch), and greater gestural displacement and/or velocity (see Bird, Nolan, and Claxton, this volume, and Bliss, Bird, and Gick, this volume, for more on acoustic and articulatory features, respectively, of North American Indian languages). Pitch (Hz) 250 100 0 1.212 Time (s) ˈs a p a̰ b u ˈb u Fig. 2: Spectrograms (frequency range from 0–8kHz), pitch track (in blue) and intensity trace (in red) for the Lakota words /ˈsapa/ ‘to know,’ with initial stress, and /buˈbu/ ‘husky,’ with final stress1 The acoustic exponents of stress can be visualized using displays like the one in Figure 2, which shows a spectrogram, pitch track (in blue), and intensity trace (in red) for the pair of Lakota (Siouan; North and South Dakota) words /ˈsapa/ ‘to know’ with initial stress 1 Figure is based on the following recording from the UCLA Phonetics Archive: http://archive.phonetics. ucla.edu/Language/LKT/lkt_word-list_1983_01.wav 138 Matthew K. Gordon and /buˈbu/ ‘husky’ with stress on the second syllable. A spectrogram provides information about the distribution of energy across different frequencies, where darker areas indicate greater energy. Time is on the x-axis and frequency on the y-axis in a spectrogram. The pitch track (in blue) is useful for looking at the fundamental frequency of the voice, the property which, in English, distinguishes the end of yes/no questions (ending in a pitch rise) and statements (ending in a pitch fall). Finally, the intensity trace (in red) shows the volume (or loudness) of the voice. The mark above and to the left of the stressed syllable indicates stress in the International Phonetic Alphabet (IPA). In the word with initial stress on the left, the vowel in the first syllable has higher pitch (acoustically equivalent to fundamental frequency) and greater intensity than the second vowel. The second vowel is also realized with creaky phonation (indicated with a subscripted tilde), evident from the increased distance between the vertical lines in the spectrogram. Creaky phonation is commonly associated with unstressed syllables, particularly in final position. The second word, on the other hand, has higher pitch and greater intensity on the second vowel, which is also longer and lacks creakiness. Acoustic correlates of stress vary from language to language (see Gordon and Roettger 2017 for a survey) and may not all converge on stressed syllables. For example, stressed vowels in Witsuwit’en (Na Dene; British Columbia) are longer, have greater intensity and higher fundamental frequency (pitch) than their unstressed counterparts (Hargus 2005), whereas duration and intensity but not fundamental frequency are used to signal stress in Witsuwit’en’s close linguistic and geographic relative Sekani (Hargus 2005). 6.4 Typology of stress systems in North America An interesting and productive area of prosodic research involves the typology of stress systems. The investigation of stress on a broad cross-linguistic basis sheds light on the diversity of stress systems along multiple dimensions: its function (e. g., to distinguish words with different meanings or to mark the edges of words to facilitate the comprehension of connected speech), its location (e. g., near the left or right edge of a word), its relationship to other prosodic features (e. g., tone), and its rhythmic properties. This section examines how a relatively small number of parameters can be used to describe the superficially vast array of stress systems found throughout the world, including in North America. 6.4.1 Phonemic and morphological stress Languages differ not only in how stress is realized but also in the principles governing the location of stress within a word. A fundamental division can be drawn between lan- Word prosody 139 guages in which stress is phonemic and used to distinguish words or morphemes with different meanings and those in which stress is predictable based on prosodic properties. Phonemic stress is observed in some North American languages, e. g., Yuchi (isolate; Oklahoma), in which the words ˈʃaja ‘squirrel’ and ʃaˈja ‘weeds’ (Wagner 1933: 309) differ only in their stress patterns. The location of stress for each word in this pair must be learned along with the sounds. Words that may differ not only in terms of stress but also in terms of sounds can also diagnose phonemic stress. For example, the Lakota pair of words in figure 2 both contain the same string of consonant-vowel-consonant-vowel (CVCV) but differ in stress position. The parallel word structure but divergent stress pattern suggests that stress is phonemic and must be learned for each Lakota word. In Lakota, there are thus many pairs of words with the same syllable structures but different stress positions (1). (1) Lakota stress distinctions (Rood and Taylor 1996) a. Initial stress:ˈmaza ‘iron’, ˈmila ‘knife’, ˈtʰoka ‘enemy’ b. Final stress: paˈha ‘hill, mountain’, tʰeˈzi ‘belly’, hoˈhu ‘bone’ Lakota also exemplifies another function of stress: to signal a grammatical contrast, as in the distinction between the non-active verb xopˈxopa ‘to be good-looking’ and its active counterpart ˈxopxopa ‘to pose, try to appear one’s best’ (Boas and Deloria 1941: 38). In many languages, stress is sensitive to morphological structure. In some languages, particular morphemes may either attract stress or reject stress, whereas in others, the distinction between root morphemes and affixes (prefixes, suffixes) is important in predicting stress. For example, many Interior Salish languages display intricate relationships between morphology and stress (see Czaykowska-Higgins 1998 for an overview of stress in Salish languages). Nxa’amxcin (Salish; Washington) exemplifies the ways morphology can influence stress (Czaykowksa-Higgins 1993). Prefixes in Nxa’amxcin thus reject stress while roots preferentially attract stress over certain suffixes (recessive suffixes). However, certain suffixes (dominant suffixes) pull stress off the root. Certain suffixes and roots also have the ability to force stress rightward off of an immediately following suffix. 6.4.2 Prosodically predictable stress In many languages, stress is neither a property of individual words (as in Lakota) nor morphemes (as in Nxa’amxcin) but rather is predictable from the prosodic structure of a word. In these languages with predictable stress, there are two fundamental dimensions to which a language’s stress system may be sensitive: the distance of a syllable from a word edge and the role of the internal structure, or weight, of a syllable in positioning stress. 140 Matthew K. Gordon 6.4.2.1 Fixed stress Considering the first of these dimensions, syllable location, many languages position stress a fixed and predictable distance from a word edge. For example, stress in Mohawk (Iroquoian; Ontario, Quebec, New York) falls on the penultimate (second-to-last) syllable of most words, as shown in (2) (Chafe 1977; Michelson 1988; Mithun in press). (2) Penultimate stress in Mohawk (Mithun in press) (syllable boundaries marked by.) a. ak.ˈhsot.ha ‘my grandmother’ b. a.kon.ˈhahw.ha ‘my son-in-law’ c. wa.ke.niaʔ.ˈtat.hens ‘I am thirsty.’ To take an example of stress oriented toward the left edge of a word, stress is fixed on the initial syllable in Southeastern Pomo, as shown in (3) (Pomoan; California) (Moshinsky 1974; Buckley 2013). (3) Initial stress in Southeastern Pomo (Buckley 2013) a. ˈxela ‘friend’ b. ˈxelataj ‘friends’ c. ˈhaliqmattat ‘(two) discuss, plan’ 6.4.2.2 Weight-sensitive stress On the other hand, there are many languages in which stress is not fixed on the same syllable in all words but is sensitive to syllable structure, or syllable weight. Aspects of syllable structure that can play a role in predicting the location of stress include whether the syllable contains a long vowel or not and whether it ends in a consonant or not. Syllables ending in a consonant (a coda consonant), are considered closed syllables (CVC), while those ending in a vowel are open syllables. Most languages in which syllable weight impacts stress fall into one of two types: either only syllables containing a long vowel (CVV) are stress- attracting (i. e., heavy) or both syllables containing a long vowel or a coda consonant are heavy. Aleut (Eskimo-Aleut; Alaska, Russia) is an example of the former type of stress system in which syllables with long vowels are heavy. Stress may fall on either the penultimate or the final syllable of a word depending on the structure of the final two syllables (Taff et al. 2001). In most cases, stress is on the penult (4a). However, if the final syllable contains a long vowel and the penult does not, the final syllable, the heavier one, attracts stress (4b). (4) Syllable weight in Aleut (Taff et al. 2001: 91–92) (ː indicates that the vowel is long) a. Penult stress: ˈsi.ʧiŋ ‘four’, a.ˈmax.six ‘spend the night’, a.ˈðaː.ðaː ‘their father’ b. Final stress: si.ˈʧiːŋ ‘nine’ Word prosody 141 Hopi (Uto-Aztecan; Arizona) (Jeanne 1982) also has a weight-sensitive stress system, but unlike Aleut it treats not only CVV but also CVC as heavy (5). Stress falls on either the first or the second syllable: the first syllable if it is closed by a consonant or contains a long vowel (5a), otherwise the second syllable (5b). (5) Syllable weight in Hopi (Jeanne 1982: 253–54) a. Initial stress: ˈnap.na ‘shirt’, ˈnaː.tɨ.ho.ta ‘to hurt oneself’. b. Second syllable stress: ko.ˈjo.ŋo ‘turkey’, qø.ˈtø.som.pi ‘headband’ The form ˈnap.na ‘shirt’ demonstrates that CVC is heavy in Hopi since it attracts stress in initial position. This may be compared with the Aleut form in (4) ˈsi.ʧiŋ ‘four’, in which the final CVC fails to attract stress. Another property of syllables that can also influence stress is vowel quality. Most commonly, this sensitivity to vowel quality manifests itself as a distinction between schwa and peripheral vowels (vowels other than schwa), where schwa is more resistant to stress. For example, in Lushootseed (Salishan; Washington), stress falls on the leftmost peripheral vowel in a word (6a) and on the first syllable in words consisting only of schwa (6b) (Bianco 1995). Note that morphology also plays a role in Lushootseed stress, as in many other Salish languages (see section 5.1). (6) Syllable weight in Lushootseed (Bianco 1995: 128) a. jə.ˈla.ʧiʔ ‘both hands’, ˈda.da.tut ‘morning’ b. ˈp’ə.ʧ’əb ‘bobcat’, ˈχə.gʷəd ‘blackberry’ As the form jə.ˈla.ʧiʔ shows, the attraction of stress to the left edge of the word is superseded by a stronger prohibition against stress on schwa. Only in words where there is no peripheral vowel does schwa attract stress, as in ˈp’ə.ʧ’əb. Weight distinctions are often encoded by linguists as differences in the number of timing units, or moras, associated with syllables (Hayes 1989). Syllables that are heavy (i. e., stress-attracting) have more moras than those that are light. Most short vowels have one mora, while long vowels have two moras, i. e., CVµVµ. Languages differ in whether they assign a mora to a syllable-final (coda) consonant or not. In Aleut, which treats only long vowels as heavy, coda consonants do not receive a mora, whereas in Hopi, which treats syllables with either a long vowel or a coda consonant as heavy, coda consonants carry a mora, e. g., Aleut CVµC vs. Hopi CVµCµ. Heavy syllables under this approach are those with two moras, whereas light syllables have fewer than two. Moraic theory can be extended to account for stress in languages like Lushootseed if one assumes that schwa is different from other vowels in not having any mora associated with it. Heavy syllables thus have at least one mora in Lushootseed whereas light syllables have no moras (under the assumption that coda consonants are also non-moraic in Lushootseed). A key feature of weight-sensitive stress is that it is virtually never sensitive to the structure of the syllable onset, i. e., onset consonants do not have a mora. A syllable is thus equally heavy regardless of whether it has two or more onset consonants. 142 Matthew K. Gordon 6.4.3 The role of final consonants in stress systems An interesting twist on weight-sensitive stress is observed in languages in which syllables require extra weight to count as heavy in final position. For example, in Stoney Dakota (Siouan; Alberta), the final syllable is stressed only if it is closed by two consonants; otherwise, the penultimate syllable is stressed even if the final syllable is closed by a single consonant (Shaw 1985). Stress is thus final in a.ˌki.da.ˈbinʧ ‘they looked at it’, but falls on the penult in a.ˌki.da.ˈbik.taʧ (Shaw 1985: 189). Note that secondary stress (marked with ˌ) falls predictably on the second syllable, an inheritance from the dominant pattern of primary stress found in closely related Lakota (see section 5.1). The requirement that the final syllable have two coda consonants to attract stress is typically treated as an instance of extrametricality (Hayes 1989), the application of a more stringent type of weight criterion at the periphery of a word. Extrametricality is largely limited to final syllables, though it has been claimed to also apply at the left edge of the word in Kashaya Pomo (Pomoan; California) (Buckley 1994). 6.4.4 Tone-driven stress A rarer type of property that predicts stress is tone in certain languages that possess both lexical tone (see Uchihara, this volume, on tone) and stress. Tone corresponds to pitch but is used to describe languages in which words are distinguished through pitch differences. Cross-linguistically, in languages with tone-driven stress, high tones and tones involving a high starting or ending point (i. e., falling or rising tones) preferentially attract stress over lower tones (De Lacy 2004), a pattern that also holds in North American languages, e. g., in Xaad Kil (isolate; Alaska, British Columbia) (Enrico 1991) and Fort Ware Sekani (Na-Dene; British Columbia) (Hargus 2005). For example, in Fort Ware Sekani, the low-toned final syllable rejects stress in ˈxə́nə ̀s ‘raft’, whereas the hightoned final syllable is stressed in tɬ’ə́ˈnə́s ‘snake’ (Hargus 2005: 412). 6.4.5 Hybrid stress systems Stress systems in many languages do not fit neatly into a single category but rather reflect a hybrid system employing elements of multiple types. For example, even Lakota, which was introduced earlier as an example of a language with phonemic and morphological stress, has stress on the second syllable in the majority of words, vestiges of which are also observed in the secondary stress patterns of related Dakota (see section 6.4.3). Similarly, Chimariko (isolate; California) (Jany 2009) positions stress on the penultimate syllable of the root but also stresses long vowels not in the penult, a system that reflects a combination of fixed stress, morphological stress, and weight-sensitive stress. Fort Ware Sekani (Hargus 2005) is sensitive to a complex array of factors, including Word prosody 143 morphology (stems attract stress over affixes), vowel quality (more peripheral vowels attract stress over schwa), tone (high tone attracts stress over low tone), and position (stress is pulled to the left if other factors are controlled for). 6.4.6 Rhythmic stress Another dimension along which stress systems differ concerns the number of stresses per word. In most of the languages discussed thus far, there is a single stress per word. In many languages, however, words, especially when they have many syllables, may contain multiple stresses occurring at rhythmic intervals. For example, the seven syllable Central Alaskan Yupik (Eskimo-Aleut, Alaska) word qayangecigsugnarquq, phonetically [qa.ˈjaː.ŋə.ˈʧiq.suʁ.ˈnaχ.quq], ‘he will probably get a kayak’ (Miyaoka 1985: 61) displays a rhythmic prominence pattern involving stress on alternating even-numbered syllables. Rhythmic stress systems can be classified according to two parameters: the edge of the word at which the rhythmic stress pattern originates and whether the first syllable at that edge is stressed or not. Combining these two dimensions yields four basic types of stress patterns: left-to-right rhythm starting with a stressed syllable, left-to-right rhythm starting with an unstressed syllable, right-to-left rhythm commencing with a stressed syllable, and right-to-left rhythm initiating with an unstressed syllable. All four of these patterns are attested in North America and are distinguished by considering a range of longer words with an odd and even number of syllables. Osage (Siouan; Oklahoma) instantiates the second system (left-to-right beginning with an unstressed syllable) (7a) (Altshuler 2009). Kutenai/Ktunaxa (Isolate; Montana, Idaho, British Columbia) reflects the mirror image of the Osage pattern (7b) (Garvin 1948). Tübatulabal (Uto-Aztecan; California) is similar to Kutenai in parsing the word from right-to-left but differs in starting with a stressed syllable (7c) (Voegelin 1935). Finally, Cahuilla (Uto-Aztecan; California) employs a left-to-right parse beginning with a stressed syllable (7d) (Seiler 1965). Note that the IPA symbol ˌ represents a secondary, or less prominent, stress in a word. (7) Rhythmic stress patterns a. Osage: xoː.ˈʦo.ðiːb.ˌrɑ̃ ‘smoke cedar’, ɑ̃ː.ˈwɑ̃.lɑː.ˌxy.ɣe ‘I crunch up my own (e. g., prey) with teeth’ b. Kutenai/Ktunaxa: ˌkqa.qa.ˈnaɬk.qaːʦ ‘automobile’, ˌhu.qaɬ.ˌwiyn.ʔoːɬʔ.ˈxup.xa ‘I want to know’ c. Tübatulabal: pɨ.ˌtɨt.pɨ.ˌtɨː.di.ˈnat ‘he is turning it over repeatedly’, ˌɨm.bɨŋ.ˌwi. ba.ˈʔat ‘he is wanting to roll string on his thigh’ d. Cahuilla: ˈtax.mu.ˌʔat ‘song’, ˈtax.mu.ˌʔaʔ.tih ‘song (objective case)’ As the examples above illustrate, the primary or main stressed syllable is almost always the stressed syllable closest to the edge at which the rhythmic pattern originates. Thus, in left-to-right systems, the leftmost stress is the primary one, whereas in right-to-left systems, the rightmost stress is the main one. 144 Matthew K. Gordon 6.5 Foot structure The foot was introduced in section 6.2 in the context of the English word (HEli)(COPter), which consists of a trochaic foot, meaning the stressed (or head) syllable precedes the unstressed syllable. There is another common type of foot that has the opposite structure: the iamb, which consists of an unstressed syllable followed by a stressed syllable. Left-to-right stress systems beginning with an unstressed syllable, as in Osage, are thus iambic, whereas those beginning with a stressed syllable, e. g., Cahuilla, are trochaic. Right-to-left systems starting with an unstressed syllable, e. g., Kutenai, are also trochaic, whereas those beginning with a stressed syllable, e. g., Tübatulabal, have iambic feet. 6.5.1 Iambic/trochaic asymmetries Comparison of iambic and trochaic stress, particularly those spoken in North America where iambic stress is disproportionately common relative to other areas of the world, suggests an interesting rhythmic asymmetry in the two types of systems. Iambic systems typically lengthen stressed syllables as in the Central Alaskan Yupik word (maˈɬuː)(suˈtuː) (ɬiˈniː)(luˈni) ‘he apparently always hunted for beached sea mammals’ (Woodbury 1987: 696), which displays lengthening of vowels in non-final stressed syllables, or the second syllable in each iambic foot (feet surrounded by parentheses). Iambic lengthening in Yupik (as in many other languages) is suppressed in final syllables and in syllables closed by a consonant, which are already heavy without vowel lengthening by virtue of the syllable-final consonant, e. g., (uˈtəχ)(tənˈʁil)(ŋuʁˈni) ‘(ones) who do not come home (loc.)’ (Woodbury 1987: 696). Iambic lengthening can alternatively affect the consonant immediately following stressed vowels. For example, in Lenape (Algonquian; United States) (Goddard 1982), a voiceless consonant other than /h/ is lengthened (i. e., geminated) after a stressed short vowel. Thus, the /t/ after the stressed vowel in the word (nəˈmət)(təˈmeː) ‘I follow a trail’ surfaces as long unlike its counterpart before a stressed vowel (məˈtə)(meː) ‘he follows a trail’ (Goddard 1979: xiii). The post-stress consonant lengthening can be interpreted as a strategy for increasing the duration of the stressed syllable if one assumes the standard analysis in which a long (geminate) consonant is shared between two syllables. On the other hand, in (mə.ˈtə.)(meː), unlike (nə.ˈmət.)(tə.ˈmeː), the short /t/ belongs only to the second syllable. Note that the /m/ after the stressed vowel in the first word does not lengthen because it is a voiced consonant. The lengthening process observed in the iambic stress languages of Central Alaskan Yupik and Lenape offers support for the hypothesis that the optimal iamb consists of a light unstressed syllable followed by a heavy stressed syllable (Prince 1990). Unlike iambic stress systems, trochaic systems are less consistent in the pervasiveness and magnitude of the lengthening effect. Trochaic stress languages thus limit stress-induced lengthening to the primary stressed syllable in a word and the length- Word prosody 145 ening effect, where it occurs, is characteristically smaller than that observed in iambic languages. This durational asymmetry between iambic and trochaic stress systems is often termed the Iambic-Trochaic Law (see Hayes 1995 and Hyde 2011 for overviews), which has analogs in musical grouping preferences and has been amply documented in psycholinguistic experiments. When presented with sequences of nonce syllables that are durationally equivalent but alternate between more intense and less intense, e. g., [DA da DA da DA] (where greater intensity is indicated via capital letters), listeners prefer to group them into binary units in which the first syllable is perceived to be stressed, i. e., (DA da) (DA da) DA, On the other hand, when presented with alternating sequences that are durationally imbalanced but equivalent in intensity, e. g., [daa da daa da daa], the preference is for groupings in which the second member is the long syllable, i. e., daa (da daa) (da daa). The relevance of the grouping asymmetry in different genres suggests a deeper basis for the Iambic-Trochaic Law that transcends the modality of language. The relationship between stress and length is bidirectional in many languages with rhythmic stress. Not only do syllables lengthen when stressed but syllables that are heavy also attract stress. In such languages with rhythmic weight-sensitive stress, the alternating stress pattern is interrupted by heavy syllables that attract stress even when adjacent to another stress. For example, in Chickasaw (Muskogean; Oklahoma) (Munro and Willmond 1994, 2005; Gordon 2004), which employs a left-to-right iambic stress pattern, all closed syllables and syllables containing a phonemic long vowel attract stress and the left-to-right alternating stress pattern resumes after the heavy syllable. As in Central Alaskan Yupik, non-final stressed vowels in open syllables lengthen in Chickasaw. Thus, in the word, (ˌtok)(saˌliː)(liˈtok) ‘I worked’, the first syllable is stressed by virtue of its weight whereas the third vowel is stressed by the rhythmic stress pattern (and lengthened because it is stressed). The final syllable is both heavy and in a rhythmically prominent position. The attraction of stress by heavy syllables creates the potential for multiple adjacent stresses in a Chickasaw word, e. g., (ˌok)(ˌʧaː)(ˌlin)(ˈʧiʔ) ‘savior’, which has four consecutive stressed syllables. 6.5.2 Metrical structure beyond stress In all of the examples discussed thus far, the evidence for metrical feet has come from stress and the rhythmic grouping preferences comprising the Iambic-Trochaic Law. In certain languages, phenomena other than stress provide the most compelling support for foot structure. One such language is Muskogee (Muskogean; Florida), in which the occurrence of high tone is predictable from an iambic metrical parse oriented toward the left edge of a word. In Muskogee words (lacking a lexically assigned high tone), a predictable high tone spans the domain between the leftmost and the rightmost foot head in a word (Haas 1977; Martin 2011). For example, in the words (nokó)(sótʃí) ‘bear cub’, (awá)(nájí)ta ‘to tie to’ (Martin 2011: 75), a high tone plateau extends from the 146 Matthew K. Gordon second syllable to the rightmost strong syllable: the final syllable in (nokó)(sótʃí) and the penultimate syllable in (awá)(nájí)ta. In the word (í:)(káná) ‘land’, the high tone span begins with the first syllable, which is the leftmost strong syllable due to its heavy status. Vowel deletion (syncope) and shortening patterns can also provide evidence for metrical structure. Tonkawa (isolate; Texas, Oklahoma, New Mexico) displays complementary processes of vowel shortening and deletion that can be accounted for with reference to metrical feet (Gouskova 2003). Deletion targets short vowels in even-numbered non-final syllables, while shortening makes phonemic long vowels short in the same context (Hoijer 1933, 1946). For example, the second (long) vowel is shortened and the fourth (short) vowel is lost in the word /ke-taː-notoso-oʔs/ ‘he stands with me’, which surfaces as (ˈketa)(ˈnot)( ˈsoʔs). Gouskova (2003) suggests that both processes are driven by foot well-formedness conditions stemming from the trochaic metrical parse of Tonkawa. Long vowels shorten to avoid a trochaic foot in which the unstressed syllable is heavier than the stressed syllable, e. g., (ˈketa) instead of *(ˈketaː), reflecting a more general strong cross-linguistic bias against unstressed heavy syllables. Short vowels delete to create monosyllabic feet in which the stressed syllable is heavy, e. g., (ˈnot) instead of *(noto), in keeping with the cross-linguistic preference for stressed syllables to be heavy. Poetry and songs can also provide evidence for metrical structure. In her analysis of Tohono O’odham (Uto-Aztecan; Arizona, Mexico) songs, Fitzgerald (1998) shows that recurring characteristics of song lines can be captured by assuming that the song meter consists of trochaic feet. First, neither the second nor the last syllable of a line is stressed and, second, adjacent stresses are prohibited. Both constraints follow if one assumes that lines adhere to a trochaic parse and that there is a requirement that song lines begin and end with a trochaic foot. The trochaic parse is enforced through a process of vacuous reduplication (copying of sounds; see Urbanczyk 2023 [this volume] for discussion of reduplication) whose only function is to ensure that the song is metrically well-formed. The role of reduplication in the song’s verse is apparent through comparison of words appearing in a song with the spoken equivalent lacking reduplication. For example, the word ˈwaw ‘rock’ surfaces as ˈwa-wai before a word beginning with a stressed syllable in a song (Fitzgerald 1998: 24). 6.6 Prominence at different levels One of the challenges intrinsic to the study of stress is the intertwined relationship between word level stress and prominence associated with larger prosodic units. Although stress is typically conceived as a word level property, many descriptions of stress are based on words uttered in isolation, a context in which the word is equivalent to the entire utterance. Prosodic units larger than the word have their own prominence characteristics, which often are realized through the same acoustic cues as word stress. Word prosody 147 For example, intonation is a phrase and utterance level feature that is conveyed by pitch, which is the perceptual correlate of fundamental frequency, the same acoustic property used to signal stress in many languages. The use of pitch as a cue to both word stress and phrasal intonation can make it difficult to determine whether a prominence on a syllable is attributed to stress or to intonation. For example, the question Is it a mammal? has a pitch rise on the second syllable of mammal even though the first syllable is stressed. This terminal pitch rise is a general property of yes/no questions for many speakers of American English. The stress on the first syllable of mammal is more easily perceived when mammal is placed in non-final position of an utterance where there is no terminal pitch rise, e. g., A mammal couldn’t live in that climate. The stress on the first syllable of mammal is even more apparent when the word is focused, as in the utterance A fox is a mammal not a reptile. As these examples of the word mammal in different contexts illustrate, the same word can have very different prosodic realizations depending on properties related to the utterance. Stress patterns may accordingly differ depending on the position of a word in an utterance. For example, word-final syllables in Central Alaskan Yupik are unstressed when followed by a large prosodic boundary even if they are predicted to be stressed by the weight-sensitive iambic stress system that treats CVV as heavy. The effect of a word’s position can be observed by comparing the first word of the phrase in (8a), which has a large boundary separating the words, with the first word of the phrase in (8b), which has a small boundary between the words. Conversely, word-final syllables followed by another word in the same phrase are stressed in Yupik even if not predicted to be stressed by the typical stress rules treating (non-initial) CVC as light (8c). (8) The effect of phrasal position on stress in Central Alaskan Yupik (Miyaoka 1985) a. nuˈnaːkaː taˈmaːna his land that (extended one) ‘that (extended one) is his (emphasis) land’ b. nuˈnaːˌkaː taˈmaːna his land that (extended one) ‘that (extended one) is his land’ c. qaˈjaːˌmun teˈkiːtuq to the kayak he came ‘he came to the kayak’ (cf. phrase-final *qaˈjaːmun) In summary, it is important to consider the prosody of a word in varied contexts to determine which syllable or syllables are consistently prominent and can be reliably treated as stressed (see Tuttle, this volume for more on the relationship between word prosody and prosody associated with larger units). 148 Matthew K. Gordon 6.7 Prosodic evolution Like other linguistic properties, stress and metrical structure can evolve over time due to either internal changes or due to contact with other languages. The evolution of stress systems in languages of North America has been the subject of relatively little study but there is enough comparative research on certain language families to suggest various mechanisms of prosodic development. In a study of Yupik varieties, Leer (1985) examines the development from a relatively simple left-to-right weight-sensitive iambic stress system, still observed in Central Siberian Yupik, to more complex versions in which the core rhythmic and weight elements of the inherited system are preserved but have been supplemented with additional features. For example, Alaskan Yupik varieties have introduced a modification to the weight system whereby initial closed syllables containing a short vowel (CVC) attract stress. Central Alaskan Yupik varieties have extended this attraction of stress by closed syllables to positions to the right of the initial syllable, where a CVC syllable preceding a non-final CV syllable attracts stress away the following CV. The progression in the weight of CVC can be observed in comparing the stress of the word for ‘my big boat’ in three Yupik varieties: aŋˈjaχpaka in Central Siberian Yupik, ˈaŋjaχˈpaːka in the Norton Sound dialect of Alaskan Yupik, and ˈaŋˈjaχpaka in Central Alaskan Yupik (Jacobson 1985: 21), where the lack of stress on the final syllable reflects the pre-boundary realization (see section 6.6). Northern Iroquoian languages display a different type of prosodic development based on a shift in directionality of stress assignment. The ancestor language to modern Northern Iroquoian languages, Proto-Northern Iroquoian, originally positioned stress on the penultimate (second-to-last) syllable (Chafe 1977; Foster 1982; Michelson 1988). This system is essentially preserved in Mohawk (Chafe 1977; Mithun in press) and Tuscarora (Williams 1976). Cayuga and Seneca, however, have introduced a left-to-right iambic system (see section 6.5.1 on iambs) that places stress on the syllable immediately preceding the penult (the antepenult) if it is an even-numbered one counting from the beginning of the word. The difference in the location of stress between Mohawk and Tuscarora, on the one hand, and Cayuga and Seneca, on the other hand, can be seen in the cognate set for ‘mind, spirit’ in (9). Note that the circumflex ˆ in the Mohawk form indicates falling tone, (see Uchihara 2023 [this volume] for more on tone) a reflex of the following /h/ that is still preserved in the Tuscarora form. (9) Stress differences in Northern Iroquoian languages (Julian 2010: 558–9) a. Penultimate stress in Mohawk and Tuscarora Mohawk: oɁniˈkũ̂ːɹaɁ Tuscarora: oɁtiˈkə̃hrɛh b. Iambic stress in Cayuga and Seneca Cayuga: kɁaˈnikõhaɁ Seneca: hoɁˈnikɔɛ̃ Ɂ ̃ Word prosody 149 Comparison of stress in Northern Iroquoian languages demonstrates how related languages may follow different evolutionary paths in the development of their prosodic systems. In this case, a relatively simple shift from counting syllables from right-to-left to left-to-right by Cayuga and Seneca speakers produced a divergence in the location of stress from their linguistic relatives speaking Mohawk and Tuscarora. 6.8 Prosody and language contact The relative malleability of stress systems often makes it difficult in practice to distinguish between interlanguage convergence attributed to internal sources and convergence due to language contact (see Rice 2010 for discussion of this issue in the context of stress in North American languages). The potential for language contact to influence stress is exemplified, however, by Michif, a mixed language arising through contact between Plains Cree (Algonquian; Saskatchewan, Alberta, Manitoba, Montana) and French speakers in Manitoba during the 19th century (Rhodes 1977, 1986; Rosen 2006, 2007). Michif’s stress system shows elements of both the French and the Cree systems of stress. The basic pattern is for stress to fall on odd-numbered syllables in a word counting from the right edge. The choice of which stress is the primary one in words with multiple stresses varies as a function of word length. In words with three syllables, the shortest word shape that can support more than one stress, the primary stress is on the final syllable (10a), a pattern that mirrors French. In words of at least four syllables, the second to the last stress, the one on the antepenult (third-to-last syllable in a word), is the primary one, a pattern that corresponds to Cree stress (10b). (10) Michif stress (Rosen 2007: 219–220) a. Three-syllables: ˌmiʧʊˈwak ‘they are eating’, ˌʃɔkɔˈla ‘brown’ b. Four-syllable and longer: nɪˈmiʧʊˌnɑn ‘we are eating’, ˌkɪmiˈʧʊnɑˌwɑw ‘you (pl.) are eating’ Rosen (2007) suggests that the Michif divergence in the location of primary stress as a function of word length is attributed to a difference in the characteristic length of words with a French source vs. those of Cree origin. Because words sourced from French, predominantly nouns, were typically no more than three syllables, the French system of primary stress on the final syllable was adopted in trisyllabic words. On the other hand, the polysyllabic words, predominantly verbs that came from Cree, followed the Cree system in which primary stress falls on the antepenultimate syllable. This difference in primary stress placement only emerges in words with at least three syllables, as both French and Cree words have only a single (primary) stress on the final syllable in words shorter than three syllables. As Rosen points out, a key feature of the Michif stress system is its internal coherence: primary stress is predictable from word length and not directly from the language from which a word originates. Plains Cree words with three 150 Matthew K. Gordon or fewer syllables thus have primary stress on the final syllable, e. g., ˌmiʧʊˈwak ‘they are eating’ (Rosen 2007: 219), while French words with more than three syllables position primary stress on the antepenult, e. g., oˈtomɔˌbil (Rosen 2007: 220). As the consistency of stress across different strata of the lexicon demonstrates, the influence of the source language’s stress system exerts its impact indirectly through the probabilistic correlation between word length and its source. 6.9 Conclusions: The importance of word prosody Although prosodic properties play a very important role in language, they often receive less attention in language instruction than segmental properties such as consonants and vowels. As we have seen, in many languages, differences in the location of stress can differentiate words that are otherwise identical, as in the Yuchi minimal pair ˈʃaja ‘squirrel’ vs. ʃaˈja ‘weeds’. Even if a language does not distinguish any pairs of words solely on the basis of stress, stress still plays an important role in facilitating effective communication, which is a critical component in language learning (see Derwing and Munro 2009 for an overview of issues related to communication for second language learners). Numerous studies have shown that accuracy in the production of stress patterns among second language learners is a better predictor of both intelligibility, the ability to be understood by others, and comprehensibility, the perceived naturalness or nativelikeness of speech, than other properties such as individual sounds and syllables (Anderson-Hsieh, Johnson, and Koehler 1992; Hahn 2004; Isaacs and Trofimovich 2012). Furthermore, because stress typically occurs at or near a word edge, it can be used as a marker of word boundaries aiding in the chunking of the string of speech into individual words. The importance of stress placement can easily be observed if one tries to interpret an English sentence in which stress is inaccurately placed, e. g., eLEphants are mamMALS with stress on the second rather than the first syllable of elephants and on the second rather than the first syllable of mammals. Stress can also impact the realization of consonants and vowels in significant ways that impact meaning and intelligibility. For example, the rhythmic lengthening of stressed vowels in Yupik and Chickasaw discussed in section 6.5.1 creates vowels that are identical in length to (or virtually identical to) phonemic long vowels. Thus, the second rhythmically lengthened vowel in the Yupik word qaˈjani ‘his own kayak’ is equivalent in length to the phonemic long vowel in the word qaˈjaːni ‘in his own kayak’ (Jacobson 1985: 30) potentially leading to homophony or near homophony between words. On a deeper level, prosody is a crucial component of a language’s identity. The overall sound of a language depends heavily upon the stress, timing, and rhythm of individual words that form the building blocks of phrases, sentences, and larger discourses (see Buckley 2020 for more on word prosody in North American languages). Word prosody 151 References Altshuler, Daniel. 2009. Osage fills the gap: The quantity insensitive iamb and the typology of feet. International Journal of American Linguistics 75. 365–398. Anderson-Hsieh, Janet, Ruth Johnson & Kenneth Koehler. 1992. The relationship between native speaker judgments of nonnative pronunciation and deviance in segmental, prosody, and syllable structure. Language Learning 42. 529–555. Bianco, Violet. 1995. Stress in Northern Lushootseed – a preliminary analysis. In Papers for the 30th (XXX) International Conference on Salish and Neighbouring Languages, 127–36. Victoria: University of Victoria. http://lingpapers.sites.olt.ubc.ca/files/2018/03/1995_Bianco.pdf (accessed 5 July 2020). Bird, Sonya, Tess Nolan & Rae Anne Claxton. [this volume]. Acoustic phonetics. In Carmen Dagostino, Marianne Mithun & Keren Rice (eds.), The languages and linguistics of Indigenous North America: A comprehensive guide. Berlin & Boston: De Gruyter Mouton. Bliss, Heather, Sonya Bird & Bryan Gick. [this volume]. Articulatory phonetics. In Carmen Dagostino, Marianne Mithun & Keren Rice (eds.), The languages and linguistics of Indigenous North America: A comprehensive guide. Berlin & Boston: De Gruyter Mouton. Boas, Franz & Ella Deloria. 1941. Dakota grammar. (Memoirs of the National Academy of Sciences 23). Washington: U.S. Government Printing Office. Buckley, Eugene. 1994. Persistent and cumulative extrametricality in Kashaya. Natural Language and Linguistic Theory 12. 423–464. Buckley, Eugene. 2013. Prosodic structure in Southeastern Pomo stress. Paper presented at the annual meeting of the Society for the Study of the Indigenous Languages of the Americas. https://www.ling. upenn.edu/~gene/papers/Buckley2013_pse_stress.pdf (accessed 5 July 2020). Buckley, Eugene. 2020. Stress, tone, and pitch accent. In Daniel Siddiqi, Michael Barrie, Carrie Gillon, Jason Haugen & Éric Mathieu (eds.), The Routledge handbook of North American languages, 68–90. New York: Routledge. Chafe, Wallace. 1977. Accent and related phenomena in the Five Nations Iroquois languages. In Larry Hyman (ed.), Studies in stress and accent, 169–81. (Southern California Occasional Papers in Linguistics 4). Los Angeles: University of Southern California. https://dornsife.usc.edu/assets/sites/56/docs/ SCOPIL4-studies-in-stress-and-accent.pdf (accessed 5 July 2020). Chafe, Wallace. 2014. A grammar of the Seneca language. UC Berkeley: University of California Press. Czaykowksa-Higgins, Ewa. 1993. Cyclicity and stress in Moses Columbia Salish (Nxa’amxcin). Natural Language and Linguistic Theory 11. 197–278. De Lacy, Paul. 2004. Markedness conflation in Optimality Theory. Phonology 21. 1–55. Derwing, Tracey M. & Murray J. Munro. 2009. Putting accent in its place: Rethinking obstacles to communication. Language Teaching 42. 476–490. Enrico, John. 1991. The lexical phonology of Masset Haida. ANLC-RP 8. Fitzgerald, Colleen. 1998. The meter of Tohono O’odham songs. International Journal of American Linguistics 64. 1–36. Foster, Michael. 1982. Alternating weak and strong syllables in Cayuga words. International Journal of American Linguistics 48. 59–72. Garvin, Paul. 1948. Kutenai 1: Phonemics. International Journal of American Linguistics 14. 37–43. Goddard, Ives. 1979. Delaware verbal morphology. New York: Garland Goddard, Ives. 1982. The historical phonology of Munsee. International Journal of American Linguistics 48. 16–48. Gordon, Matthew. 2004. A phonological and phonetic study of word-level stress in Chickasaw. International Journal of American Linguistics 70. 1–32. Gordon, Matthew & Timo Roettger. 2017. Acoustic correlates of word stress: A cross-linguistic survey. Linguistics Vanguard 3.1. https://doi.org/10.1515/lingvan-2017-0007. 152 Matthew K. Gordon Gouskova, Maria. 2003. Deriving economy: Syncope in Optimality Theory. Amherst, MA: University of Massachusetts, Amherst dissertation. Available online on the Rutgers Optimality Archive, ROA 610, http://roa.rutgers.edu. Haas, Mary. 1977. Tonal accent in Creek. In Larry Hyman (ed.), Studies in stress and accent, 195–208. (Southern California Occasional Papers in Linguistics 4). Los Angeles, CA: University of Southern California. Hahn, Laura. 2004. Primary stress and intelligibility: Research to motivate the teaching of suprasegmentals. TESOL Quarterly 38(2). 201–223. Hargus, Sharon. 2005. Prosody in two Athabaskan languages of northern Brittish Columbia. In Sharon Hargus & Keren Rice (eds.), Athabaskan prosody, 393–423. Amsterdam: John Benjamins. Hayes, Bruce. 1995. Metrical stress theory: principles and case studies. Chicago: University of Chicago Press. Hayes, Bruce. 1989. Compensatory lengthening in moraic phonology. Linguistic Inquiry 20. 253–306. Hoijer, Harry. 1933. Tonkawa: An Indian language of Texas. In Franz Boas & Harry Hoijer (eds.), Handbook of American Indian languages. Glückstadt; New York: J.J.: Columbia University Press. Hoijer, Harry. 1946. Tonkawa. In Cornelius Osgood (ed.), Linguistic structures of Native America, 289–311. New York: Viking Fund Publications in Anthropology. Hyde, Brett. 2011. The iambic-trochaic law. In Marc van Oostendorp, Colin Ewen, Elizabeth Hume & Keren Rice (eds.), The Blackwell companion to phonology, vol. II, 1052–1077. West Sussex, UK: Wiley-Blackwell. Isaacs, Talia & Pavel Trofimovich. 2012. Deconstructing comprehensibility: Identifying the linguistic influences on listeners’ L2 comprehensibility ratings. Studies in Second Language Acquisition 34. 475–505. Jacobson, Steven. 1985. Siberian Yupik and Central Yupik prosody. In Michael Krauss (ed.), Yupik Eskimo prosodic systems: Descriptive and comparative studies, 25–46. Fairbanks, Alaska: Alaska Native Language Center. Jany, Carmen. 2009. Chimariko grammar: Areal and typological perspective. Berkeley, CA: University of California Press. Jeanne, La Verne. 1982. Some phonological rules of Hopi. International Journal of American Linguistics 48. 245–270. Julian, Charles. 2010. A history of the Iroquoian languages. Winnipeg: University of Manitoba dissertation. Leer, Jeff. 1985. Prosody in Alutiiq (the Koniag and Chugach dialects of Alaskan Yupik). In Michael Krauss (ed.), Yupik Eskimo prosodic systems: Descriptive and comparative studies, 77–134. Fairbanks, Alaska: Alaska Native Language Center. Martin, Jack B. 2011. A grammar of Creek (Muskogee). University of Nebraska Press. Michelson, Karin. 1988. A comparative study of Lake-Iroquoian accent. Dordrecht: Kluwer. Mithun, Marianne. in press. A grammar of Mohawk. Miyaoka, Osahito. 1985. Accentuation in Central Alaskan Yupik. In Michael Krauss (ed.), Yupik Eskimo prosodic systems: Descriptive and comparative studies, 51–76. Fairbanks: Alaska Native Language Center. Moshinsky, Julius. 1974. A grammar of Southeastern Pomo. Berkeley: University of California Press. Munro, Pamela & Catherine Willmond. 1994. Chickasaw: An analytical dictionary. Norman: University of Oklahoma Press. Munro, Pamela & Catherine Willmond. 2005. Chikashshanompa’ kilanompoli’. Los Angeles: UCLA Academic Publishing. Prince, Alan. 1990. Quantitative consequences of rhythmic organization. Parasession on the Syllable in Phonetics and Phonology Chicago Linguistic Society 26. 355–398. Rhodes, Richard. 1977. French Cree: A case of borrowing. Actes du huitième congrès des algonquinistes, 6–26. Ottawa: Carleton University Press. Rhodes, Richard. 1986. Métchif: A second look. Proceedings of the 16th Congress of Algonquianists, 287–296. Ottawa: Carleton University Press. Word prosody 153 Rice, Keren. 2010. Accent in the native languages of North America. In Harry van der Hulst, Rob Goedemans & Ellen van Zanten (eds.), A survey of word accentual patterns in the languages of the world, 155–248. Berlin & Boston: De Gruyter Mouton. Rood, David S. & Allan R. Taylor. 1996. Sketch of Lakhota, a Siouan Language. Handbook of North American Indians, vol. 17: Languages, 440–482. Washington DC: Smithsonian Institution. Rosen, Nicole. 2006. Language contact and Michif stress assignment. STUF 59(2). 170–190. Rosen, Nicole. 2007. Domains in Michif phonology. Toronto: University of Toronto dissertation. Seiler, Hansjakob. 1965. Accent and morphophonemics in Cahuilla and Uto-Aztecan. International Journal of American Linguistics 31. 50–9. Shaw, Patricia. 1985. Coexistent and competing stress rules in Stoney (Dakota). International Journal of American Linguistics 51. 1–18. Taff, Alice, Lorna Rozelle, Taehong Cho, Peter Ladefoged, Moses Dirks, & Jacob Wegelin. 2001. Phonetic structures of Aleut. Journal of Phonetics 29. 231–271. Uchihara, Hiroto. [this volume]. Tone in North American Languages. In Carmen Dagostino, Marianne Mithun & Keren Rice (eds.), The languages and linguistics of Indigenous North America: A comprehensive guide. Berlin & Boston: De Gruyter Mouton. Urbanczyk, Suzanne. 2023 [this volume]. Prosodic morphology. In Carmen Dagostino, Marianne Mithun & Keren Rice (eds.), The languages and linguistics of Indigenous North America: A comprehensive guide. Berlin & Boston: De Gruyter Mouton. Voegelin, Charles Frederick. 1935. Tübatulabal grammar. University of California Publications in American Archaeology and Ethnology 34. 55–190. Wagner, Günter. 1933–1938. Yuchi. In Franz Boas (ed.), Handbook of American Indian languages, vol. 3, 292–384. New York: Columbia University Press. Williams, Marianne Mithun. 1976. A grammar of Tuscarora. New York: Garland Press. Woodbury, Anthony. 1987. Meaningful phonological processes: a consideration of Central Alaskan Yupik Eskimo prosody. Language 63. 685–740. Siri G. Tuttle 7 Prosody beyond the word Abstract: This chapter will discuss the study of prosodic effects that extend beyond single words in connected speech. These patterns are realized in contrasts in pitch, duration, and loudness. They represent both extensions of and interactions between different prosodic systems such as word stress, lexical tone, and phrase and utterance intonation. While it can be difficult to pull together information of this type from written sources on North American Indigenous languages, there is valuable information to be gathered even from the oldest grammatical descriptions. More recently, linguists increasingly address the prosody of connected speech in their descriptions. 7.1 It’s all about how you say it When we listen to a traditional story told by an elder, we hear more than the sounds of consonants and vowels. We hear how they use their voices to let us know when a new sentence is beginning; when a piece of the speech is important; and even where we are in the story. Consider this example from a Lower Tanana Dene story told by Moses Charlie in the early 1960s (Charlie, Krauss and Kari 1991). Fig. 1: From “Raven and Bear” by Moses Charlie (Charlie, Krauss and Kari 1991) The pitch track above the printed words in this example represents some aspects of the performance of this piece of a story. The wavy lines in the middle box of the graphic show how the storyteller lengthened and shortened his words and raised or lowered https://doi.org/10.1515/9783110600926-007 156 Siri G. Tuttle his pitch. The pitch track is generated using a phonetic analysis program called Praat (Boersma and Weenink 2018; see Bird, Nolan, and Claxton, this volume) so it can be analyzed quantitatively. However, it can also be understood intuitively. Longer lines mean a longer time taken to say a word; higher marks mean a syllable or a word was pronounced with higher pitch in this recording. Notice in this example how the first tsoni ‘brown bear’ takes longer to say than the next one, and there is a pause after it. The storyteller repeats the word. The grizzly is the main antagonist in this story, a very important character. The introduction of this character gets special treatment, both in words – the repetition – and in the prosodic treatment of the words. However, the storyteller also remembers to stress the first syllable in the word, not the second, using effects of length and pitch, so that word prosody is respected as well. Prosody is letting us know about all these things at once. This example also shows the stretchability of prosodic structure – in this case, intonational boundary marking. Notice again the first pronunciation of tsoni ends with an intonational low tone before the pause – but when it is included in a longer utterance (the second pronunciation), the final low pitch is assigned to the word at the end of that utterance (nokhuniłtsinh). Thus, some of the prosodic effects that we hear in connected speech relate directly to linguistic information: they let us know when words and utterances are beginning and ending. However, they may also encode information that is called paralinguistic – letting us know more about the speaker’s attitude, intention, and feelings. When we use recorded stories as linguistic data, we have to sort out the linguistic from the paralinguistic in order to make generalizations. In the case of prosody, we also have to sort out effects that relate to syllables and words from effects driven by the composition of longer utterances. It can be difficult to decide where to start, especially if the prosodic systems of the language we are studying have not been fully described. This chapter therefore connects to all the other chapters in this book that deal with prosody. In words, prosody organizes stress and tone patterns to provide cues to word identity, and also cues the listener to word boundaries (see Gordon, this volume, on word prosody and Uchihara, this volume, on tone for descriptions and examples from numerous languages). Prosody also organizes speech in domains that are, or can be, much larger than a prosodic word. Some prosodic patterns play out over shorter or longer stretches of speech, from a single syllable to a long utterance. Intonation is one of these systems. However, it is impossible to figure out what intonation is doing in a language unless we know what prosodic phenomena exist at the word level as well. For this reason, the examples in this chapter will involve interaction among prosodic systems: word stress, tone, and intonation. Despite the importance of phrase and utterance-level prosody to the understanding and use of language, this field of study is still developing in Native American linguistics. Prosody beyond the word 157 In this chapter, many of the examples come from the author’s work with Dene (ath) languages, which provide a wealth of opportunities to study the interactions between prosodic systems. 7.2 Prosody beyond the word What we observe most often when we study prosody is contrast and variation in three major phenomena of spoken language: the relative pitches at which we hear words spoken, the relative length of syllables, and the relative loudness of different syllables. Pitch is a given in any speech, because each of our voices have a basic pitch, reflecting its fundamental frequency. All syllables take time to utter, because each must be uttered in turn. And loudness, reflecting the acoustic amplitude of our utterances, has to be there or we cannot hear each other. Prosody is the study of how we use these basic properties of human speech in communication. What is fascinating about it is that despite the universality of the cues of prosody, the ways it is used are very particular to individual languages. Beckman and Venditti (2011: 486) refer to prosody as a “multidimensional taxonomy of phonetic form in relationship to linguistic function.” As such, its study requires us to be aware of all the possible functions related to each measurable cue. Among these are boundary creation (helping the listener find the edges of words or larger units), realization of prominence (marking the most important syllable in a string of language) and lexical identification (serving as a required element for distinguishing one morpheme from another). Considering only these three functions, three possible cues, and several possible types of prosodic systems, we can see some of the many dimensions to which Beckman and Venditti refer (Table 1). Tab. 1: Functions and Cues of Prosodic Systems F0 (Pitch) Duration (Length) Intensity (Loudness) Boundary Creation Boundary tones (phrases) Edge-aligned stress (words) Intonational slowing (phrases) Intonational fading (phrases) Word identification Tone, stress, accent Stress Increased loudness: Stress in words Prominence Stress, emphasis Stress, emphasis Stress, emphasis The use of the same auditory cues to signal different prosodic systems, as shown in Table 1, can result in a high level of complexity. It is problematic for learners and describers of languages that prosodic patterns of this type are not always fully described, and their interaction even less often. There are several reasons for this. It can be very 158 Siri G. Tuttle challenging to provide this description because it requires knowledge of each of the prosodic systems that might be present, and tools to tease them apart. But there may also be a bias toward more intensive study of the linguistic elements that are encoded in orthography – that is, segmental elements that can be spelled using commonly available symbols. Such bias may have consequences for the creation of reference works that are used to build pedagogical materials. This in turn may exacerbate the issues that can arise from too great dependence on written language in teaching (see Chew, Leonard, and Rosenblum, this volume). Prosodic patterns operate within domains of application: stress patterns usually play out within words, and tone systems also have word-level patterns. However, when words are combined into phrases, both stress and tone systems may apply patterns in the larger domain as well, and these patterns may be similar to or different from those that apply within words. When words are joined into phrases, and phrases into longer utterances, prosodic phenomena that operate within different domains may also come into competition. This happens when systems that employ the same phonetic means of expression – pitch, duration, or loudness – need to use the same tone-bearing unit (a vowel, say, or a syllable, where the pitch can be heard) for their expression. In some cases, conflicting values for a phonetic cue may all be expressed, a phenomenon called co-articulation; in other cases, one system’s expression may obscure the expression of another system. A high lexical tone at the end of an intonational phrase may be affected by intonational final lowering, for example. Word or phrase stress may use pitch as part of its expression, and the pitch required may obscure lexical tone. Other prosodic patterns include vowel harmony systems, in which vowels must agree for some feature within a particular domain. They may also include nasalization patterns, and rhotacization patterns. These patterns work much the same way as those that use pitch, loudness and timing as their auditory cues, in that their domains are not defined by the boundaries of single linguistic units. However, these systems are realized segmentally, rather than in pitch and timing. Robins (1958, 1967) describes a vowel harmony process in the Algic language Yurok (Algic; yur), which makes any non-high vowel in a prefix be pronounced as [ɚ], a rhotic or r-like vowel, if it precedes the rhotic vowel in the first syllable of the stem word. In this example from Robins’ 1958 grammar, the initial syllable of ‘my saliva’ is pronounced with the rhotic vowel written as [ɹ], harmonizing with the following rhotic in the first stem syllable. (1) Yurok vowel harmony (Robins 1958:26) (Robins’ orthography) (ʔ)nelɹhpɹyeł ~ (ʔ)nɹlɹhpɹyeł ‘my saliva’ Harmony processes are considered prosodic because they do not just affect vowels and consonants that are adjacent to each other but work across syllable boundaries (as in 1) and even across word boundaries. Prosody beyond the word 159 Prosody beyond the word, even more than word-level prosody, is often not notated in writing systems. Conventional devices like punctuation can be used to suggest boundaries within which prosodies may operate. Usually, the choice is made to keep wordlevel spelling consistent, rather than to transcribe prosody using the writing system. More general guidelines for reading, or phonological rule statements, may be provided to help learners. For this reason, information about prosody beyond the word is sometimes found buried in the caveats presented by researchers when they are describing word-level systems or attempting to tease one prosodic system apart from another. For example, Krauss (2005: 105) wrote of the Dene languages Gwich’in (gwi) and Hän (haa), that “both…show phrase-final lowering of unmarked pitch, apparently neutralizing the contrast between marked and unmarked…” and of Lower Tanana (taa), another language with sparse low tone (p. 119) that “verb stem tone is (apparently) neutralized in the usual final position”, though he did not specify intonational final lowering as the reason for this neutralization. Thus, while it is not always possible to find an article that explicitly references prosody beyond the word, it is often possible to find clues in other linguistic descriptions to systems that need investigation, as they emerge as confounds in studies of other systems. Another example from Moses Charlie’s Lower Tanana Raven and Bear story, recorded by Krauss himself, shows the source of his observation. In Figure 2, a low-toned syllable ends the utterance, and in Figure 3, a non-tonal syllable is in the same position. Measurement in Praat shows that these two syllables are both pronounced at about the same pitch – by the same speaker, in the same narrative. The numbers in Hz (Hertz) represent measurement of the acoustic property fundamental frequency, which we hear as pitch. Fig. 2: Low toned syllable in utterance-final position in Lower Tanana (Charlie, Krauss and Kari 1991) 160 Siri G. Tuttle Fig. 3: Non-low syllable in utterance-final position in Lower Tanana (Charlie, Krauss and Kari 1991) The relevant information lies in the pitch track over these two utterances. Both end with a low pitch on the last syllable. However, the final syllable in Figure 2 is lexically low-toned, meaning that it can be expected to sound lower than the last one in Figure 3. However, in this intonational final position, the two are almost identical in pitch. (There is a difference in the voice quality, which is historically related to Lower Tanana’s low tone – but the difference does not show up as a difference in pitch.) This example shows how important it can be to learn about all the prosodic systems in a language we are studying. Because prosodic systems with different domains may use the same acoustic cues, it is possible to draw false conclusions if we are thinking only about lexical tone patterns, or only about stress, or only about intonation. 7.3 Methods in prosodic research There are numerous methods and tools available for the investigation of prosodic systems. This applies to the selection or collection of data, its analysis, and the implications of research for practical applications. Xu Yi (2011) describes the development of international prosodic research as a passage through a number of different methods, including introspection, transcription, hypothesis-testing, and more sophisticated quantitative modeling, on the way to predictive knowledge – which is what learners of a language hope to gain, in order to speak correctly. Researchers in Native American prosodic research have employed the same methods with the same goal. However, language endangerment and language shift make the status of prosodic data an especially crucial element in this research. Prosody beyond the word 161 As in other aspects of linguistic research, direct elicitation, structured narration, and the recording of spontaneous speech are all useful tools in data collection. Direct elicitation of words rarely gives enough information for a description of prosodic effects over phrases or utterances. Elicitation can interfere with the collection of valid data. There is, for example, the lab-speech effect, caused by lack of context and speaker discomfort, which can cause a speaker’s prosody to come out distorted in any number of ways. However, prosodic interaction itself can interfere with the gathering of word-level information. Studying spontaneous or semi-spontaneous narration offers both benefits and risks. The benefits include access to authentic and spontaneous connected speech. For some situations, where few first-language speakers remain, narration recorded and archived in earlier decades may provide the only opportunity to hear fluent speech. The risks include the possibility of confusing constraints on performance, such as markers of register or genre, with constraints on ordinary language. Because fluent texts may have been recorded by relatively few narrators in some endangered language situations, there is also the possibility of one person’s performative style being understood to represent the state of prosody for the language’s population – or to be taken as evidence for significant language change. For this reason, whenever possible we want to review narrative audio (and when possible, video) with the original narrator in order to correct possible confusions. When the original narrator is not available, family members or others who had a chance to hear the narrator in person can be extremely helpful in interpretation, even when they themselves do not feel they could tell a traditional story in the language with confidence. Some studies of phrasal prosody (Mirzayan 2010, for example) use both elicited data and spontaneous narration. This study of Lakhota prosody is based on archival narrative recordings from the early 1970s, as well as recordings by the author done between 2006 and 2009. In the author’s work, recordings include spontaneous speech, but also scripted phrases and conversation, as well as elicited words and phrases (Mirzayan 2010: 32). An example from Mirzayan’s study (Figure 4) shows how multilayered prosodic analysis needs to be. The graphic shows the waveform generated by the program Praat (Boersma and Weenink 2018). Below the waveform, a pitch track shows the relative height of pitches in the recorded utterance. The next tier down is marked using the ToBI prosodic annotation system (Beckman and Elam 1997), which marks “pitch accents” that reflect intonational patterns. These high-pitched phrase-level accents line up with the word-level stresses. The tier below the ToBI annotation shows the segments of the utterance, divided by syllables; below that we see the words of the utterance. Since high and low pitches can be contributed to an utterance by both word-level and utterance-level requirements, this layered annotation style can be highly informative, though it is, of course, also very time-consuming to create. As a dissertation, Mirzayan’s work has room for a few very complete graphics demonstrating the analysis. Many other studies, including those that modern work depends on, contain no such intuitively readable display. 162 Siri G. Tuttle 300 F0 (Hz) 255 210 165 120 (L) 75 ma H* štíŋ maštíŋča 0 L ča H* phu té L ^H* L lí =phuté la líla Time (s) !H* L% ó tah óta 2.705 Fig. 4: Mirzayan (2010: 82): a Lakhota sentence about buffalo berries. Despite the brevity of some older descriptions of phrasal prosody, no description should be ignored. Historically, prosodic description has been reported based on accumulated intuitive observation, including qualitative study of recorded narration. Prosodic facts may be presented very briefly as part of much longer and more complex descriptions of segmental phonology and morphology. Still, even these descriptions can encapsulate crucial facts about pronunciation. These generalizations are not based on first impressions, but often on decades of work by a linguist with a language community. For example, Pitkin’s (1984) grammar of the Wintu language of Northern California (Wintuan, wnw) contains only three pages that describe prosody above word level, but they include information about smaller and larger domains, effects of pitch, duration and pause, and phrasal accent. It is no accident that the grammar is based on nearly thirty years of experience with the language. Sivertsen (1956) considers multiple issues that make the phonemic status of suprasegmentals in Kiowa (kio) hard to define. Her work was based on data collection at the 1954 Summer Institute of Linguistics at the University of Oklahoma, with a number of Kiowa participants. She approaches the description of phrase-level prosody as a problem of sorting confounds, but this does not prevent the inclusion of several generalizations, including that of final lowering and downdrift (p. 122), raising of pitch on high-toned syllables “under special emphasis” (p. 123) and patterns of laryngealization on open syllables, involving interaction with the tonal system (p. 124). This work is qualitative but includes instrumental measurement. Quantitative and qualitative methods are both useful, in the study of prosody beyond the word. These represent two different, and equally important, means of arriving at generalizations. Quantitative studies measure and count observations, and then filter the measurements through statistical models to reduce the possibility of observer Prosody beyond the word 163 bias or results that are due to chance. Qualitative studies demonstrate and describe phenomena learned through uncounted, but valid observations. These two styles of working with prosodic data complement each other. Usually, the results are consistent across methodologies, when they can be compared (Tuttle 2003). Both quantitative and qualitative methods can use instrumental measurement. Acoustic graphics such as those shown in Figures 1–4 can represent typical or exemplary tokens of a phenomenon, where the linguist has learned through experimental measurement and statistics; they may also represent good examples of phenomena learned about primarily through auditory study. In both cases, the linguist is showing an example that they believe correctly represents their generalization. Digital tools for data collection, analysis, description, and demonstration of prosodic patterns have become increasingly accessible. The pitch tracks shown in Figures 1–4, for example, were derived from the program Praat (Boersma and Weenink 2018; see Bird, Nolan, and Claxton, this volume). It takes only a few minutes to make these measurements and create a graphic display to demonstrate them using this free program. Affordable access to programs like this has only become available in the last twenty years, and this has changed the study of sound in language in important ways. For one thing, it means that when we make a claim about pitch, length, quietness, or some other phonetic effect, we can back it up, sometimes with quantitative data and often with intuitively communicative images. For another, it means that when we work with prosodic systems, we really need to provide this phonetic evidence – because we can. The end result can become a body of descriptive work that is both scientifically reproducible, and adaptable to applications like language learning and teaching. Researchers working with Dene languages have used quantitative methods in a number of studies: Holton (2005) for Tanacross (tcb), Lovick and Tuttle (2012) for Dena’ina (tfn), Tuttle (1998) for Lower and Middle Tanana, Bird (2004) for Lheidli (crx), Berez (2011) for Ahtna (aht), Alderete and Bob (2005) for Tahltan (tht), Tuttle (2005) for Apache (apw), and Palakurthy (2019) for Diné Bizaad (Navajo, nav) have used such methods. Researchers of Salish languages have also employed quantitative methods to explore prosodic patterns: Caldecott (2016; St’át’imcets, lil) follows on a burst of research on prosody for this language family, including works by Barthmaier (2004; Okanagan, oka), Beck (1991; Lushootseed, lut) Caldecott and Czaykowska-Higgins (2012; Nxaʔamčín/ Moses-Columbian Salish, sal), among others. Kinkade and Mattina (1996) offer several useful caveats for those wishing to embark on prosodic description using published and archived written narratives from the languages of North America. They point out (p. 253) that “Non-speaking specialists in a language may be able to make some evaluation of texts based on comparisons between texts narrated by different speakers, observation of certain stylistic features or complexities, or vocabulary selection, but even then expectations based on Euro-American literary traditions may color their judgments.” They also note that published texts are 164 Siri G. Tuttle often “severely edited” to make them easier to read. Those who wish to understand the prosody of the performances on which these texts are based should, if at all possible, obtain the recordings from which they were transcribed and compare the two versions of the recorded language. Sometimes, an original narration is corrected or adapted by the narrator in further work on the story. There can be multiple layers of information in recorded field sessions discussing the story, as well as in the original performance recording. When available, all versions should be compared. Recorded narration is not the only form of verbal art that can be used to study prosody (see Webster, this volume). The study of verse has been used as a method for learning about word stress in many languages, because of the tendency for poetic forms to match word prominence to metrical prominence. Musical rhythms can be thought of as analogous to poetic meter in some cases, making text setting another potential area of study in the understanding of prominence in language. Lower Tanana song lyrics show that word prominence can be linked to intentionally created structure as well as the naturally occurring rhythms of utterances. Tuttle (2012), a study of text-setting in songs from the village of Minto (Lower Tanana), finds that syllables that would carry main stress in speech – mostly syllables that correspond with the roots of words – are the syllables most likely to be linked to a moving pitch, a note held over more than one beat, or a higher pitch than a neighboring syllable. In Figure 5, two lines of a short song from a Lower Tanana story are analyzed for their rhythm and pitch movement in relation to word stress. The lines are: sech’etthila’ selotl’ogh no’i’oyh, yozronh eya, yozronh ‘Give me my ax back! Clear sky, clear sky.’ In this song, which follows the rhythm of speech, word roots are stressed and are also associated with higher pitches and longer-held notes (Tuttle 2012). In the table in Figure 5, syllables in the song lyric are lined up with other aspects of the song: whether the pitch changes on a syllable, whether the pitch in the melody goes up or down, how long a syllable is held for, whether the syllable is placed on a downbeat in the rhythm. By carefully listening to multiple recordings and paying attention to how singers place words in melodies, it is possible not only to learn about song making, but also to uncover facts about the language that did not show up in conversational registers. In this case, it was the importance of word roots to the stress pattern in the language. Prosody beyond the word 165 Yozronh, line 1 Pitch movement D/U Beats Linguistic stress Segmental Gloss high D 1 X se my lower U 1 . ch'e its flat D 1 X tthi chop flat U 1 . la' poss flat D 1 X se my flat U 1 . lo band flat DU 2 X tl'ogh to flat DU 2 X e voc lower DU 2 X ya voc Yozronh, lines 2- 3 Pitch movement D/U Beats Linguistic stress Segmental Gloss higher lower D U 1 1 X X no ’i back you flat DU 2 X ’oyh give flat flat D UDU 1 2-3 X X yo zronh sky clear higher D 1 X yo sky flat UDU 2-3 X zronh clear Fig. 5: Two lines from a Lower Tanana song (Tuttle 2012), annotated 7.4 Intonational boundaries Intonation organizes several different kinds of functions in a language. Intonation groups parts of an utterance and signals beginnings and endings; it may also be used for emphasis and contrast. In some languages, it can encode mood (declarative, interrogative, imperative). Intonational patterns can also distinguish speech registers and genres of expression. Boundaries occur at the edges of units. Prosodic units have been suggested to exist in a hierarchy – reaching from the level of the syllable to multi-sentence groupings, something like paragraphs in written language (Nespor and Vogel 1986). Intonation does not always play a role in differentiating smaller from larger units, but some studies have found intonational differences between larger utterances and phrases within those utterances. Beginnings and ends of intonational units – utterances, and sometimes prosodic phrases within utterances – are often marked by an intonational pattern involving a change in pitch, slowing of speech rate, or a decrease in volume. A common finding is “final lowering.” Declarative utterances in many languages, including some Native American languages, have this pattern, in which the final syllable of an utterance carries a lower tone than the rest of the syllables. Caldecott and Czaykowska-Higgins (2012), for example, report final lowering as well as low initial boundary tones in Salish languages. Gordon, Martin, and Langley (2015) describe the prosodic system of Koasati (cku), a Muskogean language spoken in Louisiana and Tennessee, using a database of words 166 Siri G. Tuttle spoken in citation form, find that the one-word utterances in their data set show evidence of phrase-final tones – high or low, depending on the speaker. The data in this study consists of one-word utterances. Figure 6, from Gordon, Martin, and Langley, shows the sharp fall in pitch at the end of a Koasati utterance. They interpret this pitch pattern as intonational. Fig. 6: Koasati elicitation, showing final lowering Final lowering is not uncommon in Dene languages. Lovick and Tuttle (2012: 310) found final lowering in the Alaskan Dene language Dena’ina; Tuttle (1998, 2003) discusses utterance-final lowering in Lower Tanana, in an instrumental study using both elicited and text data. Final lowering may be found in Dene languages outside Alaska, as well. De Reuse (2006: 58) notes a final lowering effect in San Carlos Apache, which seems to co-articulate with high tones, creating a falling tone where a high toned syllable occurs at the end of an utterance. If an utterance ends with a low pitch, and the speaker begins a new utterance with a higher pitch, both beginning and end of utterance can be said to be intonationally marked. Berez (2011) finds that high or “reset” pitch marks the beginning of utterances in Ahtna narration. Final lengthening, sometimes referred to as “rhyme lengthening” because both consonants and vowels may be extended, is reported for Dena’ina by Lovick and Tuttle 2012, in intermediate intonational units (abbreviated “IU” in the graphic, and smaller than their “story units, ” abbreviated as “SU”). Palakurthy (2019) finds final lengthening at the ends of intonational units in a set of Navajo recorded narratives, as well as reset at the beginnings of units. This paper also reports significant pause at the ends of syntactic, as opposed to intonational units, a finding that harks back to the work of Landar Prosody beyond the word 167 Fig. 7: Rhyme length in prosodic domains in Dena’ina (Lovick and Tuttle 2011) (1963), a work that sought to organize Navajo grammar based on intonational group ings. Figure 7 shows a graphic representing statistical findings for the Dena’ina study (Lovick and Tuttle 2012: 315) on rhyme lengthening in intermediate intonational units: In this study, as in many others, statistics are used to reduce the probability that listener bias will affect generalizations about acoustic measurements. The boxplot in Figure 7 shows the median length (the line in the middle of the boxes) of syllable rhymes in milliseconds (thousandths of a second) as well as the variability of the findings (the height of the boxes). The position of the median line in the 4-IU-Fin position (position 4, Intonational-Unit-final) demonstrates the difference in rhyme length for this comparison. Statistical diagrams like this one provide information in graphic shorthand. They do not substitute for evaluation of the statistical studies themselves but serve to illustrate patterns that emerge with quantitative analysis. Boundary tones can be used to indicate other distinctions as well as to delimit utterances. In some cases, boundary markings may differ for different utterance types – as in English, where certain types of questions may be marked differently from declaratives. Lovick (2020) considers intonational and other accentual patterns in Upper Tanana Dene (tau) in the process of describing lexical tone effects. She finds (p. 153) that the low lexical tone in the Northway dialect of this language can be obscured by high-rising intonation in yes-no questions. Declaratives in this language show declination, or a gradual fall in pitch. Lovick (this volume) provides numerous other examples of intonational marking of different types of questions in various languages of North America. Final lengthening – lengthening of certain segments or slowing of speech rate at the ends of utterances – has also been described for Chickasaw (cic) in Gordon and 168 Siri G. Tuttle Munro (2007). Final vowels in Chickasaw words are lengthened whether the words are included in a phrase, end a phrase, or end a larger intonational unit. These vowels also become breathy, and the duration of the breathiness increases with the size of the domain ended by the vowel. To summarize: intonation structures utterances using relative pitch, loudness, and length. In the Native American languages discussed in this section, pitch and length have been the cues that define intonational boundaries. 7.5 Lexical tone beyond the word Lexical tone, in languages that have it, distinguishes words, different forms of words, and morphemes from one another (Uchihara, this volume). As seen in Table 1 (repeated here), tone is one of the linguistic systems that uses voice pitch as a cue. When words that have lexical tones are combined, tonal systems also play out in larger domains, such as phrases and utterances. Because lexical tone shares pitch as a cue with intonation, and sometimes with word stress, interactions between these different pitch-using systems can also change tonal contours. Tab. 1: (repeated). Functions and Cues of Prosodic Systems F0 (Pitch) Duration (Length) Intensity (Loudness) Boundary Creation Boundary tones (phrases) Edge-aligned stress (words) Intonational slowing (phrases) Intonational fading (phrases) Word identification Tone, stress, accent Stress Increased loudness: Stress in words Prominence Stress, emphasis Stress, emphasis Stress, emphasis I follow van der Hulst (2011) and Welmers (1973) in defining a language with tone as one in which “both pitch phonemes and segmental phonemes enter into the composition of at least some morphemes.” but also including languages that allow affixes that consist solely of tones. This means that the word “tone” in this chapter refers to pitch effects that are associated with particular morphemes or with particular morpheme-level meanings. These definitions of tone all have to do with words, not with domains outside the word. However, tonal phenomena do extend to the phrasal level, and also often interact with the other prosodic systems that can use pitch as a cue: stress and intonation. Though Welmers (1973) was writing for researchers of African languages, his observations on tone and its relationship with other prosodic systems are valid for North Prosody beyond the word 169 American languages as well. He points out that languages that use lexical tone may have many syllables that get their pitch from another prosodic system, and that tone is not always lexically significant (p. 79). Some of the tonal phenomena encountered outside the domain of the word include tonal spreading, downstep, downdrift, and interaction with intonational use of pitch, as discussed below. Tonal spreading, when a tone that exists on one syllable changes the tonal value of syllables near it, can occur both within and between words. Holton (2005: 252) describes word-internal spreading of tones in Tanacross, an Alaska Dene language with high lexical tone. This phenomenon results in a co-articulation of tones, resulting in rising or falling tone in certain morphologically complex contexts. Spreading across word boundaries has also been documented in the Dene languages of Alaska; Ritter (1990), for example, points out rightward spreading of Hän low tone from lexical items immediately preceding non-tonal verbal prefixes. Downstep is a phrasal tonal phenomenon in which high tones following certain other tones are pronounced with relatively lower pitch. Thompson and Thompson (1966) describe downstep in Tillamook (til); while a pitch pattern is referred to as stress, and particular accents followed by downstep or de-accentuation are described under intonation, the patterns are all shown over sequences without spaces, which indicates a judgment that they are “words.” Miner (1979) describes downstep in Winnebago (Ho-Chunk; Siouan, win) as a terracing effect over utterances, causing the pitches of word accents to occur at successively lower levels. In both the Ho-Chunk and the Tillamook case, the downstep effect is described as affecting word accent or stress, rather than lexical tone. However, the effect is the same, as pitches that are assigned at the level of the word are realized differently depending on their position in a phrase or utterance. Martin and Johnson (2002) describe tonal downstep in Creek (Muskogean, mus) as being triggered between words by a stress-assigned low pitch at the beginnings of words. Thus, when there are two words with high tones, the second high tone will be relatively lower than the first due to interaction with the stress system. These authors also describe downdrift: this is a gradual declination in the pitch of tonal syllables, towards a low at the end of an utterance. Tuttle (1999) found downdrift in the low tones of San Carlos Apache – but no clear downdrift or downstep in the high tones of the same data set, in a study intended to uncover intonational patterns. Figure 8 shows average pitch for high and low tones in a set of recordings of the sentences used in the experiment. The high tones cluster within a range of ten Hertz – between 120 and 130 Hertz in the male speaker’s voice. The low tones, however, fall from about 105 to 80 Hertz over the course of the utterances, suggesting that high and low tones seem to be treated differently in this language. Without studying tones in the context of longer utterances, this kind of pattern is much harder to observe. In example (2) below, a tone-intonation interaction is shown that demonstrates a competitive relationship between two prosodic systems: low tone, which in this lan- 170 Siri G. Tuttle Fig. 8: Average high and low tones in different utterance positions in a Western Apache experimental set guage is found sprinkled sparsely through utterances, and an intonational pitch that can be used to signal utterance type. A rising pitch at the end of yes-no questions in Lower Tanana is often used by bilingual speakers of the language in place of the low-toned question marker –(h)i, and this intonation can override low tone marking.’ The result can be a question contour that is very familiar to English speakers (M, H and L here refer to pitch levels Mid, High and Low): (2) Lexical tone vs. question intonation – Lower Tanana (Session notes November 17, 2005) a. Contour when lexical tone is realized: M H L L L Khuzrunhts’e ghestà’ Well PFV:1s.sbj-sleep:pfv ‘I slept well’ b. Contour with question marker realized: M H L M H L M M L Nenh chukhw, khuzrunhts’e ghintà’ì’ You also well PFV:2sSUB-sleep:PFV-Question ‘And you, did you sleep well?’ c. Contour with question intonation on verb stem: M H L M H L M H Nenh chukhw, khuzrunhts’e ghintà’ì’ You also well PFV:2sSUB-sleep:PFV:Question And you, did you sleep well?’ Prosody beyond the word 171 The variation among these possibilities pits an intonational contour (which may or may not have been borrowed from English and favors a high pitch at the end of the question) against a tonal melody (which realizes the low tone on the low-toned question marker). Sometimes, the lexical-tonal version of the question wins out, and version (b) is heard; sometimes, the intonational pattern in (c) is realized instead. Both patterns are understood to convey the same meaning. Holton (2005: 273), a qualitative study, reports interaction between lexical tone and intonation in Tanacross, which he shows to have distinctive high and low tones. He finds that final lowering interacts with the high lexical tone in this language, with final lowering obscuring high tone in utterance-final syllables. Fig. 9: Pitch track for declarative with high tone stem in Tanacross (Holton 2005). Figure 9 illustrates Holton's (2005:269) analysis of the interaction of high tone and final lowering in Tanacross. While the high tone here is expressed, its expression is seen in the early part of the utterance-final syllable, while the utterance-final low is expressed at the end of the syllable, creating a falling tone. Similar phenomena are found in neighboring Lower Tanana. In Figures (10) and (11), two “versions” of a word are shown. The word is shath, which means ‘wart’. The speaker first gives the word as a translation, in citation form, in Figure 10. Notice that because it happens at the end of the utterance, it is pronounced with a low pitch (155 Hz). In Figure 11, recorded just after (10), she pronounces the word in context, in the middle of a sentence. Now, the word is pronounced at 183 Hz, which is quite a bit higher. If you were testing the word to find out if it bore low lexical tone, what would you decide based on Figure 10 or Figure 11? This is an example of how we need to examine words in different contexts to understand their patterns. 172 Siri G. Tuttle Fig. 10: Lower Tanana shath ‘wart’ in citation Fig. 11: Lower Tanana shath ‘wart’ in sentence context Summary: distinctive tones assigned to words participate in prosodic interactions in larger domains as well. Sometimes, this interaction results in loss of information, as when an intonational pitch overrides a lexical tone. Assimilation or spreading, downstep, downdrift, and interaction with intonational contours are all documented in Native American languages. Prosody beyond the word 173 7.6 Stress beyond the word Stress is the term used for patterns of prominence within words (see Gordon, this volume). When stress systems are predictable, their regular patterns may be assigned on the basis of syllable position, syllable quantity, rhythmic pattern, or some combination of these. However, stress may also be assigned to particular morpheme types (such as roots), or to a syllable in a certain position in a word. When words are placed in larger context, stress assignment can change due to constraints on stress class or lapse. However, phrases and utterances are also domains for prominence. In longer stretches of speech, certain word stresses are often promoted to the level of the “phrase accent,” a term originated by Gösta Bruce in his study of Swedish accentual patterns. Mirzayan’s (2010) study of Lakhota (lkt) documents phrasal accents in this Siouan language. Mirzayan’s work also shows how transcriptional techniques developed for very different languages can be adapted for the languages of North America. The ToBI (Tones and Break Indices) system was developed for English by Mary Beckman and others (see Beckman and Elam (1997), and it has also been adapted by Sun-Ah Jun (1993) for Korean and other languages. In fluent communication, speech rate is much quicker than in elicitation or teacher-pronunciation. This can mean that in a somewhat longer sequence, one word stress will be more prominent than others. In Lower Tanana, each word is stressed on its root, as well as on any full vowel and closed syllable. The most prominent word stress falls on the last stress in the word – usually a word root. In a multiple-word utterance, as shown in the Lower Tanana sentence in (3), the most prominent of these word stresses is the last one, and the others may be reduced in prominence. The ticks before stressed syllables indicate primary phrasal stress when high, and secondary stress when low. (3) Lower Tanana (Session Notes, November 17, 2005) Khedoch’ekhwdel’ikhde ghesdò’ kheˌdoˌch’ekhwˌdelˈ’ikhde ˌghesˈdò’ SchoolatPFV:1sSUB-sit:PFV ‘I sat in school’ Word stress in this language is cued by length of the vowel; a stressed vowel is nearly twice as long as an unstressed vowel (Tuttle 1998). The stem syllable -dò’ will have the strongest stress in the word. Because it comes at the end of the utterance as well, this syllable will also be the most prominent – and longest – in the utterance. Studies of stress often include consideration of both citation, or slow-speech form, and of connected speech in the form of spontaneous text or elicited reading. The results of studying material from these two sources may differ because of phrase-level patterns, so it is important to pay attention to the source of data and methods used for analysis. By studying a list of words recorded in isolation, boundary markings of declarative or 174 Siri G. Tuttle list-like utterances can be learned. By examining the same words in connected speech, both word-level and utterance-level prosody can be clarified. Descriptions of intonational phenomena such as phrase accent may refer to fastspeech phenomena rather than connected speech. For example, Mesquakie (sac) (Voorhis 1971) is reported by Voorhis to have both deliberate and casual forms of speech, with the deliberate forms known at that time only by older speakers. There are numerous rhythmic adjustments described in this article as effects of rapid or casual speech, which might also be considered effects of word-level prosody. This article is somewhat unusual in that that fast-speech forms are described. Many descriptions of North American languages idealize away from these effects, a practice that has made the task of language learners far more difficult as they try to reconcile the speech they hear with idealized language as recorded in grammars. Caldecott (2016) demonstrates a phrase or “nuclear” accent in St’át’imcets (Salishan, lil), which occurs on the last full vowel of an utterance and makes clear that the nuclear accent is not used to mark focus, and not associated with the particle used to mark focus. Phrase accents are not always realized with higher pitch. In (3) above, we saw how word-level prominence is promoted to utterance-level prominence in connected speech in Lower Tanana. However, this prominence does not coincide with high pitch, due to the complex relationships among the language’s prosodic systems. Final lowering provides intonational boundary marking, so a final syllable will be pronounced at lower pitch as well as being lengthened, no matter its lexical tone, and still carry primary stress prominence. In (4), two low tones, one in denigi and one on the mode morphology of the verb, draw tones downward, as does the final intonational low tone. Syllables preceding and following low tones are high in comparison. The stress and pitch patterns are shown in (4). The L, M and H stand for low, mid and high pitch. (4) Lower Tanana (Session Notes, November 17, 2005) L L H H M L L L Denìgi k’osenàthdeyo Deˈnìgi ˌk’oseˌnàthdeˈyo Moose around-3S-1sOBJ-PFV-follow ‘The moose followed me around.’ The pattern in (4) shows how high pitches associate with a syllables that are stressed, but may not bear primary stress; both tonal and intonational patterns provide more of the pitch contour than stress does. There is much more that can be discovered about stress patterns in phrases. Because the acoustic correlate, or audible cue, for stress in languages can differ widely, different effects may be expected and should be explored. As in all areas of prosody in connected speech, it is clear here that isolating stress-related effects from those associated with lexical tone or intonation can be tricky. Patience and an open mind are very important to discovering patterns. Prosody beyond the word 175 7.7 Information structure and intonation The study of information structure intersects with intonation when categories important to the tracking of information in discourse are signaled with prosodic contrasts – changes in pitch, duration or intensity (see Berge on information structure, this volume). Higher or changed pitch are often associated with focused categories in the languages that have been studied, but there are also other strategies for calling the listener’s attention to a segment of speech. Intonational marking of different kinds of focus is found in some North American languages. As Berge (this volume) shows, this term can be used to describe elements in an utterance that provide new (as opposed to given) information, or elements that contrast with alternative elements already present in the conversation. The contrasted-not-contrasted distinction interacts with other important informational distinctions, particularly Given-New. Givenness is a property of linguistic items that are already present in the discourse, while Newness is a property of items that are introduced; after introduction, they also become Givens (Krifka 2008). Katz and Selkirk (2011), looking at English intonation, also found differences in the expression of prominence for items that were New versus those that appeared in Contrast. Applying the techniques used for languages like English to the Native languages of North America requires attention to information structure strategies that do not involve prosodic marking. Fery, Skopeteas and Hoernig’s 2010 study of a variety of worldwide languages found that strategies for marking Newness and Contrast varied among syntactic, morphological, and prosodic expression. It is significant that the unity these authors find among such focusing strategies is that of moving focus toward the end of an utterance and keeping given objects toward the beginning. This pattern has been reported for languages of the Americas as well: Harbour et al (2012), for example, investigates information structure in Kiowa entirely in relation to the linear presentation of noun phrases, without consideration of prosodic marking, and finds a relationship between postverbal placement of noun phrases and transition points between phases of narrative. The authors suggest that further research into clause prosody would be a useful development. Methods for studying information structure in the field have been advanced by innovative empirical research, as discussed in Bohnemeyer (2015). Structured narrative, structured elicitation through the use of games, and other strategies for eliciting rarely used constructions assist in discovering how information structure is conveyed. Because both new information and contrasting information can be marked morphologically or lexically, languages do not always use intonation to indicate the presence of focus. However, some do. Mirzayan (2010) found that in addition to a lexical marker for focus in Lakhota, an intonational pattern including a high accent was frequently, but not invariably present. Dene languages employ multiple strategies for expressing newness and contrast, including manipulation of word order, morphological marking, and prosodic marking. 176 Siri G. Tuttle Lovick and Tuttle (2013) compared texts in three Alaska Dene languages and found morphological, syntactic, and prosodic strategies used for marking newness and contrast. When prosodic prominence is used, higher pitch can be involved, but interactions with low tone (present in both Upper and Lower Tanana) can affect the expression of information structure. However, the use of lexical markers that encode information structure is common in all three languages. Clearly the availability of this strategy for tracking information structure makes prosody redundant in many cases. 7.8 Additional considerations The study of prosody in connected speech can be complicated by the previous experience of researchers. Because prosodic patterns are not usually written, and are governed by subconscious knowledge, it can be particularly difficult to catch ourselves when we (again subconsciously) apply assumptions about prosodic patterns that would be correct for our first language, but not for the language we are studying. It is very easy for our first language prosodic systems to interfere in our understanding of the functions of the prosody we are listening to. For English speakers, it is easy to confuse tone or accent for stress or prominence, since we do not use pitch distinctively in English. It is also easy for us to misinterpret pitch or voice quality effects for markers of speaker attitude, when they may be doing an entirely different job in the language. Understanding of social expectations in the language community is essential to working with prosodic questions. In the present context of language endangerment, methodologies are emerging that can bring together these different areas of knowledge. Prosodic research especially benefits from research directed by language community members, who have a great deal of social and pragmatic understanding even when they have not had the opportunity to become fluent speakers of their community’s language. Non-community members with complementary skills can learn to adjust their questions and analytic approach to follow the lead of these researchers. 7.9 Conclusion In this chapter several aspects of phrase-level prosody have been addressed: those expressing some aspect of a lexical tone system; those involved with stress; and those involved in intonation. Examples of interaction between word-level and phrase or utterance level systems have also been discussed. Examples of the study of prosody beyond the word include many descriptions that start with the goal of describing one prosodic system, but where it is found that some other system, competing for the means of expression, must be described in order to get Prosody beyond the word 177 to the system that is the intended focus of the study. However, numerous descriptions of prosodic phenomena focus on larger domains for functions involving pitch, rhythm, and variation in intensity. The resolution of competing prosodic systems from languages in contact is a fruitful area for research that has not been fully explored for most North American languages. The reason for this may lie in respect for conservative versions of the languages, and for some learners and researchers, in a desire to master or understand the language as spoken by those least affected by language contact. So many aspects of language find expression in prosody that in order to learn about the form or function of a system, it is very important to keep track of context. For this reason, despite potentially difficult conditions where languages are endangered, it is important to continue the study of prosody, especially if members of the speech community can design and carry out research. This does not mean, of course, that we should stop doing our homework before we make new recordings, or that we should ignore the analyses of previous researchers. We should also be making use of archived narrative. Bringing old narration back to speakers and learners for commentary, re-narration and discussion can create a rich environment for learning about the function and form of prosody, as well as many other topics of interest. It can also stimulate the production of new narration that can add to archived material for study. For many languages of North America, prosodic study is incomplete. There are many opportunities not only to describe prosodic systems, but to define the questions learners, teachers and linguists will be asking about these systems for a long time. Naturally, we worry about describing prosody that may have been introduced through language contact, or that may show differences from patterns described long ago. The possibility of linguistic change should not lesson our curiosity but intensify it. If we are lucky, earlier observations are available for comparison. But if we are the first ones to ask, that is a delightful privilege as well. References Barthmaier, Paul. 2004. Intonation units in Okanagan. In Donna Gerdts & Lisa Matthewson (eds.), Studies in Salish linguistics in honor of M. Dale Kinkade, 30–34. (University of Montana Occasional Papers in Linguistics 17). Missoula: University of Montana. Beck, David. 1999. Words and prosodic phrasing in Lushootseed. In T. Allan Hall & Ursula Kleinhenz (eds.), Studies on the phonological word, 23–46. Amsterdam: Benjamins. Beckman, Mary E. & Gayle Ayers Elam. 1997. Guidelines for ToBI Labeling. Columbus, Ohio: Ohio State University Research Foundation. Beckman, Mary E. & Jennifer J. Venditti. 2011. Intonation. In John Goldsmith, Jason Riggle & Alan C. L. Yu (eds.), Handbook of Phonological Theory, 485–532. Malden, MA: Wiley-Blackwell. Berez, Andrea. 2011. Prosody as a genre-distinguishing feature in Ahtna: A quantitative approach. Functions of Language 18(2). 210–236. 178 Siri G. Tuttle Bird, Sonya. 2004. Lheidli intervocalic consonants: Phonetic and morphological effects. Journal of the International Phonetic Association 34(1). 69–91. Boersma, Paul & David Weenink. 2018. Praat: doing phonetics by computer [computer program]. Version 6.0.37, retrieved 14 March 2018 from http://www.praat.org/. Bohnemeyer, Jürgen. 2015. A Practical Epistemology for Semantic Elicitation in the Field and Elsewhere. In M. Ryan Bochnak & Lisa Matthewson (eds.), Methodologies in Semantic Fieldwork, 13–46. Oxford: Oxford University Press. Bruce, Gösta. 1977. Swedish Word Accent in Sentence Perspective. Lund: Gleerup. Caldecott, Marion. 2016. St’át’imcets intonation contours: a preliminary study. Canadian Journal of Linguistics/Revue canadienne de linguistique 61(2). 119–155. doi: 10.1017/cnj.2016.11 Caldecott, Marion, & Ewa Czaykowska-Higgins. 2012. Prosodic phrasing in Nxaʔamxčín (Salish) declarative clauses. Canadian Acoustics 40(3). 16–17. Chafe, W. 1976. Givenness, contrastiveness, definiteness, subjects, topics and points of view. In Charles N. Li (ed.), Subject and topic, 27–55. New York: Academic Press. Charlie, Moses, Michael E. Krauss & James Kari. 1991. The Moses Charlie Collection. Manuscript. Fairbanks: Alaska Native Language Archive: TNMN981K1991d. DeReuse, Willem. 2006. A Practical Grammar of San Carlos Apache. München: LINCOM Europa. Gordon, Matthew & Pamela Munro. 2007. A phonetic study of final vowel lengthening in Chickasaw. International Journal of American Linguistics 73. 293–330. Gordon, Matthew, Jack B. Martin & Linda Langley. 2015. Some Phonetic Structures of Koasati. International Journal of American Linguistics 81(1). 83–118. Harbour, Daniel, Laurel J. Watkins & David Adger. 2012. Information structure, Discourse Structure, and Noun Phrase Position in Kiowa. International Journal of American Linguistics 78(1). 97–126. Holton, Gary. 2005. Pitch, Tone and Intonation in Tanacross. In Sharon Hargus & Keren Rice (eds.), Athabaskan Prosody, 249–274. Amsterdam: John Benjamins. Jun, Sun-Ah. 1993. The Phonetics and Phonology of Korean Prosody. PhD dissertation. Ohio State University. Kinkade, M. Dale & Anthony Mattina. 1996. Discourse. In William C. Sturtevant (ed.), Handbook of North American Languages. Vol. 17: Languages, 244–274. Washington: Smithsonian Institution. Landar, Herbert. 1963. Navaho Syntax. Ph.D. dissertation, Yale University. Lovick, Olga. 2020. A Grammar of Upper Tanana. Volume 1. Lincoln: University of Nebraska Press. Lovick, Olga & Siri Tuttle. 2012. Prosody of Dena’ina Narrative Discourse. International Journal of American Linguistics 78(3). 293–334. Martin, Jack & Keith Johnson. 2002. An Acoustic Study of “Tonal Accent” in Creek. International Journal of American Linguistics 68(1). 28–50. Miner, Kenneth L. 1979. Dorsey’s Law in Winnebago-Chiwere and Winnebago Accent. International Journal of American Linguistics 45(1). 25–33. Mirzayan, Armik. 2010. Lakota Intonation and Prosody. Linguistics Graduate Theses & Dissertations. Paper 7. Ph.D. dissertation. University of Colorado. Nespor, Marina & Irene Vogel. 1986. Prosodic Phonology. Dordrecht: Foris Publications. Palakurthy, Kayla. 2019. Prosody in Diné Bizaad Narratives: A Quantitative Investigation of Acoustic Correlates. International Journal of American Linguistics 85(4). 49–531. Pitkin, Harvey. 1984. Wintu Grammar. (University of California Publications in Linguistics 94). Berkeley: University of California Press. Ritter, John. 1989. Han Literacy Session, Whitehorse, Yukon, December, 1989. Manuscript. Fairbanks: Alaska Native Language Archive. HN976R1989. Sivertsen, Eva. 1956. Pitch Problems in Kiowa. International Journal of American Linguistics 22(2). 117–130. Thompson, Laurence & M. Terry Thompson. 1966. A Fresh Look at Tillamook Phonology. International Journal of American Linguistics 32(4). 313–319. Prosody beyond the word 179 Tuttle, Siri G. 2012. Language and music in the songs of Minto, Alaska. In Niclas Burenhult, Arthur Holmer, Anastasia Karlsson, Håkan Lundström & Jan-Olof Svantesson (eds.), Language Documentation and Description, vol 10: Special Issue on Humanities of the lesser-known: New directions in the description, documentation and typology of endangered languages and musics, 82–112. London: SOAS. Tuttle, Siri G. 2008. Phonetics and word definition in Ahtna Athabaskan. Linguistics 46(2). 439–470. Tuttle, Siri G. 1999. Preliminary findings on intonation in San Carlos Apache. Paper presented at the 1999 Athabaskan Languages Conference, Albuquerque, New Mexico. Tuttle, Siri G. 1998. Metrical and tonal structures in Tanana Athabaskan. Ph.D dissertation, Washington: University of Washington. Tuttle, Siri & Olga Lovick. 2013. Prosodic marking of information structure in Alaskan Athabascan text. Workshop on Phonetic/phonological typology and fieldwork, Max Planck Institute for Evolutionary Anthropology, Leipzig. August 14, 2013. Voorhis, Paul H. 1971. New Notes on the Mesquakie (Fox) Language, International Journal of American Linguistics 37(1). 63–75. Xu, Yi. 2011. Speech Prosody: a methodological review. Journal of Speech Sciences 1(1). 85–115. II Words Fernando Zúñiga 8 What is a word? Abstract: The recent literature on wordhood has shown the importance of disentangling patterns of sound (the phonological domain) from patterns of grammar (the morphosyntactic domain) for defining units that can be meaningfully called “words.” The present paper outlines the main issues involved when identifying, analyzing, and classifying these phonological words and grammatical words from the perspective of Indigenous North American languages. It summarizes what we already know about how diverse and elusive they can be, as well as the areas where we need further research to learn more about complex and heterogeneous phenomena related to word(-like) units. 8.1 Introduction Both language-internally and cross-linguistically, there are different kinds of units that linguists, teachers, and learners like to call words, and it is important (i) to tell them apart terminologically and analytically, and (ii) to see that they show significant, and perhaps also systematic, variation. The present paper outlines the main issues raised by such structural units in the context of Indigenous North American languages. Section 2 presents the two main wordhood domains, namely sound structure and syntagmatic relations (i. e., how units are chained together to build phrases, clauses, and sentences), while Section 3 addresses some possible relationships between units identified in each domain. Section 4 deals with the subtypes of grammatical words found along two particular dimensions that have received much attention in the literature, namely the morphological make-up of grammatical words and the kinds of grammatical words distinguished by the lexicon and the grammar of a language (i. e., “word classes” or “parts of speech”). Section 5 closes the paper with some summarizing remarks and suggestions for further research. 8.2 Wordhood domains Different units of linguistic structure can be distinguished in two domains: phonology and morphosyntax. Intermediate-level units called “words” are normally expected to show not only higher autonomy than smaller units but also stronger cohesion than larger units, in either domain. Before discussing these domains, however, some comments on orthographic “words” are in order. The definitions of kinds of “words” introduced in this section are summarized in Table 1 below for ease of reference. https://doi.org/10.1515/9783110600926-008 184 Fernando Zúñiga Tab. 1: Selected kinds of units Orthographic word In some writing systems, uninterrupted string of graphic symbols preceded and followed by some boundary sign (e. g., space, punctuation mark, etc.) Phonological word Unit of phonology, smaller than an intonational phrase and larger than a segment, defined by phonological rules and features Grammatical word Unit of grammar (= morphology and syntax), smaller than a phrase and larger than a morpheme, defined by grammatical behavior Morpheme Minimal form-meaning pair in language; it can be either autonomous or bound in grammar; it may or may not correspond to a simple phonological unit Clitic Morpheme with a mixed autonomy status (e. g., an autonomous grammatical word that is part of a phonological word) 8.2.1 Orthographic words Separating orthographic words (e. g., through spaces) is only about one thousand years old (Saenger 1997). Most importantly, even though the use of spaces is connected to some analysis of linguistic structures, orthographic words are conventional and related to other kinds of words rather idiosyncratically. The treatment given to compound words and to functional elements varies significantly. For instance, Spanish object clitics are written separately before verbs but as parts of the verb when they follow, as in e. g., me lo da ‘s/he gives it to me’ vs. dámelo ‘give it to me!’, but this does not reflect any difference in the linguistic status of those elements. Compare also English bedroom with living-room and pairs like instead of and in spite of. Arbitrary though orthographic words generally are, normal language users, language learners and teachers, and linguists are accustomed to them. The influence of the orthographic conventions of English, French, and Spanish in North America is considerable to this very day, however—which poses several problems in the present context. First, those conventions are not perfectly designed or implemented to meet the complex needs of their own writing communities; especially the conventions of English and French are quite idiosyncratic and difficult to master. Second, many Indigenous languages show structures, both in their phonology and their grammar, that are so different from those of the European languages of reference, that devising and implementing suitable (European-inspired) writing systems and orthographies for them is a challenging task. Moreover, norms are subject to competing motivations that are not always easy to harmonize; striking a satisfactory balance between a convention that is relatively easy to learn (and therefore probably close in many concrete and abstract respects to the convention of whatever European language is dominant in primary school) and a convention that does justice to the Indigenous language regarding not only its structures but also its cultural and political context (which may depart in numerous respects from What is a word? 185 some of the familiar European conventions) is usually a formidable job for interdisciplinary teams of people intimately familiar with the language and its speakers. Without going into specifics here, it seems that consistency is the single most important requisite for a useful and successful writing convention. Considering structural issues in particular, and irrespective of whether the script is alphabetic or syllabic, boundary marks like spaces should ideally separate units of grammar, which is the implicit default option in most conventions. Spaces are not normally used to separate exclusively units of sound; they are only variably and irregularly used with this function. Trying to use spaces for both purposes in any kind of principled way would probably lead to numerous inconsistencies and difficulties that would most likely incentivize speakers not to write at all (see Section 3). 8.2.2 Phonological words At their simplest, phonological words consist of smaller units of the stream of speech called segments (consonants and vowels). Sequences of segments smaller than phonological words are morae, syllables, and feet (see Urbanczyk 2023 [this volume]); they make up what the specialized literature calls the “metrical hierarchy.” For instance, the English phonological word segments /s/, /e/, /g/, /m/, /e/, /n/, and /t/ can be grouped in two syllables /seg/ and /ment/, which constitutes a foot /seg.ment/, which is also a phonological word. Sequences of phonological words constitute larger units like clitic groups, phonological phrases, and intonational phrases; this is the “prosodic hierarchy” (see Gordon [this volume]; Tuttle [this volume]). They can be defined language-specifically by segmental and prosodic features/rules. Segmental features include structural and phonotactic constraints. For instance, verbal phonological words are minimally bisyllabic in Mohawk (Iroquoian), i. e., they consist of at least two syllables; smaller units are parts of verbal phonological words (Adler 2016). Example (1) shows how an initial i and a medial e appear on a verb form that would otherwise have no vowels (*kks); these vowels do not contribute to the unit’s meaning but merely make it a formally admissible phonological word: (1) Mohawk (Iroquoian; Marianne Mithun, p.c.) Í:keks. i-k-e-k-s proth-1sg.A-ep-eat-hab ‘I eat.’ Moreover, phonological-unit boundaries are frequently restricted. For instance, most consonant clusters can occur only medially in a phonological word in Cayuga (Iroquoian); initially and finally in a phonological word, they are severely constrained, e. g., kr, sn, and tsy can occur only initially and medially, and nr, and kth can occur only medially (Dyck 1999). Lastly, segments are often realized differently at phonological word 186 Fernando Zúñiga boundaries. In Cup’ik (Eskimoan), for instance, stem-final uvular and velar sounds are realized as continuants word-medially (or before a sonorant-initial enclitic), as plosives word-finally and utterance-finally (or before non-sonorant-initial enclitic), or as either continuant or plosive word-finally when followed by a sonorant-initial phonological word (Woodbury 2002). Prosodic features include the sensitivity of tone and stress assignment to phonological-unit boundaries. For instance, morphemes in Arapaho (Algonquian) bear different pitches that are redistributed at the phonological word level (Cowell and Moss 2008: 22–40). In (2), the verb stem ciinén- ‘put (something) down’ has high pitch on its short second syllable; this surfaces normally if the following syllable is low pitched, as in (a) (iinén-o becomes iinéno), but the pitch pattern is altered—in order to avoid inadmissible sequences of high-pitch syllables—if the following syllable is long and high pitched, as in (b) (iinén-óú becomes íínenóú). Such redistribution occurs within phonological words, not in larger units. (2) Arapaho (Algonquian; Cowell and Moss 2008: 29, my glosses) a. Ceniinénowoo. ciinén-owoo put.down-1sg ‘I am putting it down.’ b. Ceníínenóú’u. ciinén-óú’u put.down-3pl ‘They are putting it down.’ Fixed stress is not widespread in Indigenous North American languages, but it can be found on the first syllable (e. g., in Comanche; Charney 1993: 40–44) or the penultimate syllable (e. g., in Sm’álgyax; Dunn 1995: 5) (see also Michelson 1988 for Lake Iroquoian). Other relevant phenomena include segmental harmony, i. e., where phonological segments agree in a particular phonological feature; vowel harmony is especially frequent (e. g., in Sahaptian, Interior Salishan, Washo, and Yokutsan; Mithun 1999: 478–479, 494, 557, 568), but consonant harmony is also found (e. g., coronal harmony in Athabaskan; Hansson 2007). For an example of vowel harmony, see in (3) how the verb stem wé·yik‘go across’ and the 1st-person-singular affix -se (a) both adapt their “recessive” e vowel to the “dominant” a vowel of the recent-past suffix in (b). This process is found within phonological words, but not in larger phonological units. (3) Nez Perce (Sahaptian; Aoki 1966: 760, my glosses) a. Wé·yikse. wé·yik-se go.across-1sg ‘I am going down.’ What is a word? b. 187 Wá·yiksaqa. wé·yik-se-qa go.across-3pl-rec.pst ‘I went across recently.’ The usual kinds of segmental and prosodic rules used to define phonological words are summarized in Table 2 below. Tab. 2: Summary of potentially relevant phonological word regularities Structural constraints Do some units show a minimal number of morae, syllables, or feet? → If so, these units may be phonological words. Phonotactic constraints Can/must some segments (not) appear at the edge of some units? Are some segments regularly realized differently at the edge of some units? → If so, these units may be phonological words. Tone/stress Within what kind of unit does tone and/or stress become (re)assigned? → These units may be phonological words. Harmony phenomena Within what kind of unit do consonants, vowels, or features (e. g., nasality, rhoticity) harmonize? → These units may be phonological words. At least two factors complicate this simple picture. First, since the prosodic hierarchy does not necessarily include the metrical hierarchy, having the phonological word as the place where both meet cannot be taken for granted. See Selkirk (1986), Inkelas (1989, 1993) and Dyck (2009) for arguments in favor of treating these hierarchies as independent from each other. Second, and more importantly, phonological words need be neither present nor unambiguous everywhere; languages may have one kind of phonological word, several, or none (Schiering, Bickel, and Hildebrandt 2010). For instance, the string in (4) from Cup’ik as analyzed by Woodbury (2002) can be phonologically grouped in two different ways (according to different sets of rules). Small phonological words exclude, while large phonological words include, enclitics like the reportative =gguq: (4) Cup’ik (Woodbury 2002: 91–93) small phonological word added unit large phonological word mallu-ssu-tu-llini-luni gguq beached.whale-hunt-always-apparently-3sg.app rep Mallussutulliniluni-gguq. ‘S/he is apparently always hunting beached whales, it is said.’ 188 Fernando Zúñiga 8.2.3 Grammatical words Grammatical words consist of smaller units called morphemes and appear in larger units (e. g., phrases and clauses). They can be defined language-specifically by criteria based on structural cohesion and behavior. Before presenting these criteria, however, some introductory remarks on morphemes are in order. The morpheme is the smallest meaningful unit of a language.1 For instance, the English expression the ducklings consists of four morphemes: the, duck, -ling, and -s. Morphemes are typically classified as roots (which carry specific, “lexical,” semantic content; here: duck) and affixes (which carry general, “grammatical,” semantic content; here: -ling and -s), sometimes further distinguishing clitics (which, as discussed in Section 3.1, are diverse; here: possibly the). In many languages, roots can be expanded to form bases/stems via compounding and/or affixation; in languages that allow simpler units, roots, stems, and grammatical words may exactly correspond to each other (like English duck). Regarding structural cohesion, languages show morphemic strings that show integrity along several parameters, namely insertion potential, displacement potential, fixed ordering, and cross-slot dependencies (Bickel and Zúñiga 2017). Insertion potential stipulates the conditions under which an element can interrupt the string. Restrictive grammatical words thus defined would ideally be uninterruptible, and noun incorporation—a possible instance of “free forms (=words) interrupting (other) words”—has been sometimes seen as potentially posing a problem (see, e. g., the brief, and incomplete, discussion in Julien 2002: 34–35 of some Pawnee data). Nevertheless, it is extremely rare to find true “free forms” as nominal incorporates in Indigenous North American languages (Marianne Mithun, p.c.); incorporates are normally stems, rather than fullfledged grammatical words.2 Displacement potential refers to constitutive elements of words being moved or copied by a general operation independently of each other, and fixed ordering refers to a particular order not being determined arbitrarily (e. g., rather than by semantics or pragmatics). An instance of the latter can be seen in (5) from Blackfoot; the verb root ssi ‘wipe’ can appear in the simple stems ssi-i ‘wipe (something inanimate)’ and ssi ‘wipe (something animate)’, but also in the complex stem ssi-ika-asi ‘wipe one’s feet’, with the so-called medial -ika ‘foot’. The order ssi-ika is fixed; *ika-ssi is unacceptable. 1 Many studies use the term morph instead of morpheme, while others use them interchangeably, but there is no consensus as to how to best employ this pair—which is, at least in principle, roughly parallel to phone (or segment) and phoneme. I use the more familiar term morpheme here. 2 Incorporated stems that are routinely also grammatical words are much less rare in Indigenous South American languages. (5) What is a word? 189 Blackfoot (Algonquian; Dunham 2009: 5, glosses adapted) Anná Conrad áíssiikawatsii. ann-wa Conrad á-ssi-ika-atsi dem-prox.sg C. ipfv-wipe-foot-ai ‘Conrad is washing his feet.’ Lastly, cross-slot dependencies refer to cases where neither constitutive element can be deleted without simultaneously deleting another element. For instance, in (6), the negative prefix Imáát- requires the concomitant appearance of the non-affirmative suffix for singular 1st or 2nd person -hpa: (6) Blackfoot (Algonquian; Frantz 2009: 18, 82, glosses adapted) a. Nitáóoyi. nit-á-ooyi 1-ipfv-eat.ti ‘I am eating.’ b. Nimáátáóoyihpa. n-Imáát-á-ooyi-hpa 1-neg-ipfv-eat.ti-naff.1sg/2sg ‘I am not eating.’ Constraints on displacement, re-ordering, and deletion are widespread in North American language families showing highly complex verb morphology, like Algonquian, Athabaskan, and Iroquoian. Future studies might show specific tests for grammatical words to be useful in the North American context, even though they do not yield cross-linguistically consistent results. First, the impossibility of eliding units smaller than grammatical words in coordination (also called cross-slot dependency diagnostic) works well in many languages. The second test is a special case of the displacement potential and concerns extractability (i. e., the possibility of displacing a noun from its original location in a larger nominal unit), which can distinguish Noun-Noun compounds from Noun Phrases (as in Italian; Bisetto and Scalise 1999: 38–39).3 Nevertheless, extractability is probably best seen as a property of (referential) phrases and can be restricted in some languages (Haspelmath 2011: 52). The third test concerns anaphoric islandhood, i. e., expressions whose parts cannot be anaphorically related to other parts of the sentence. Outbound anaphora (i. e., an external pronoun referring to an element of a grammatical word) is primarily determined by semantic and pragmatic factors (Ward, Sproat, and McKoon 1991). For instance, in English weapons of mass destruction are controversial because its effects are 3 In the compound nave ospedale ‘hospital boat’, ospedale is unextractable: *ospedale, hanno costruito una nave _ ‘hospital, they have built a _ boat’ is ungrammatical. In the noun phrase trasporto dei passeggeri ‘passenger transport’, by contrast, dei passeggeri is extractable: dei passeggeri, è efficiente il trasporto _ ‘of the passengers, the transportation is efficient’ is grammatical. 190 Fernando Zúñiga terrible, the anaphoric pronoun its cannot refer to destruction; in cases like McCartyites are now puzzled by him, the pronoun him can refer to McCarthy, because the latter is a name and therefore referential (Haspelmath 2011: 51). By contrast, inbound anaphora (i. e., an internal, referential, pronoun within a grammatical word) is impossible in some languages (e. g., English: *McCarthy was happy that many him-ites were at the meeting is ungrammatical) but possible in others (e. g., Georgian: Ševardnaʒe icnobs? K’i, imis-iania ‘Does she know Shevardnaze? Yes, she is a him-ite’; Harris 2006: 199). Another criterion concerns morphophonological idiosyncrasies, which are allegedly common in stem-affix combinations but excluded in host-clitic combinations (Zwicky and Pullum 1983: 505). This cross-linguistic tendency can be relevant language-specifically; in Blackfoot, for instance, some noun stems show a slightly different form after prefixes, like ponoká- ‘elk’ appearing as -innoka- after sik- ‘black’ in siksínnoka- ‘black elk’ (Frantz 2009: 78–80). See also § 8.3.1 for more on cliticization. Other cohesion-related criteria used to identify grammatical words appear to be less useful. One criterion is similar to an old “semantic word” notion but is nowadays usually framed in terms closer to Bloomfield’s (1933: 178) “Minimal Free Forms”: can a given unit occur on its own as a well-formed, complete utterance? In principle, this test helps to distinguish larger units that can occur on their own (whether grammatical words or phrases) from smaller units that cannot (e. g., affixes), but its implementation is often difficult. In English, for instance, autonomous units exclude not only functional elements like articles and prepositions but also some transitive verbs, which are not free forms but are traditionally regarded as “words.” Moreover, the test cannot adequately distinguish between compounds and phrases (e. g., English compounds like firewater or blood(-)red consist of free forms that would qualify as phrases). Two further criteria are even more problematic. The first concerns deviations from one-to-one correspondences between meanings and forms (zero marking, multiple exponence, cumulative exponence, and morphomic patterns), which would apply to affixes.4 Nevertheless, since all four deviation types exist in syntax as well (Spencer 2001), the test does not distinguish affixes from grammatical words (Haspelmath 2011: 54–59). The second criterion is related to the interruptibility criterion mentioned above and is based on the possibility of splitting up units larger than, or identical to, grammat4 Zero marking consists in one meaning being expressed by the absence of form; consider, for instance, the opposition between the Blackfoot stems sina-aki- ‘draw, write’ (ai), sina-i- ‘draw/take a picture of (inanimate)’ (ti), and sina- ‘draw/take a picture of (animate)’ (ta), where the latter stem has a zero socalled “final.” In other words, the absence of form carries a meaning analogous to the one carried by the non-zero forms -aki and -i in the other two stems (Frantz & Russell 2017: 252). Multiple exponence consists in several forms expressing one meaning, e. g., Latin rēg-s-ī (spelled rexi) ‘I directed’, where the perfect is encoded three times, viz. by stem vowel length (ē), by a temporal-aspectual suffix (-s), and by the person-number suffix (-ī) (Haspelmath 2011: 46). Cumulative exponence consists in one form expressing several meanings, e. g., Blackfoot -wa, which usually encodes both ‘singular’ and ‘proximate’. Lastly, morphomic patterns consists in one form expressing no meaning, e. g., Blackfoot m and n at the end of some nominal stems in some forms but not in others (pokón-wa ‘ball’ vs. pokó-íksi ‘balls’) (Frantz 2009: 11). What is a word? 191 ical words via pauses (Hockett 1958: 166, Langacker 1972: 41). It is unclear, however, whether this is better understood as a phonological word criterion (Dixon and Aikenvald 2002: 24). Most probably, pausing is not closely related to language structure per se but derives from production and processing conventions (Haspelmath 2011: 37). Finally, a grammatical word criterion related to behavior rather than cohesion concerns the selectivity of dependent elements. Affixes tend to be highly selective as to their possible hosts (e. g., only nominal or verbal stems), whereas function words (e. g., prepositions) are frequently less so. This criterion is used often, but non-selective nonwords (“edge inflection,” Zwicky 1987; “promiscuous inflection,” Stump 2001: 126) pose a potential problem. There are non-selective tonal markers, for instance, which cannot be words because they are not segmental units.5 Similarly, English ’s is non-selective but triggers haplology (i. e., the deletion of identical consecutive strings, as in English the kids’ ideas), which is typical of morphological markers. Simply accepting non-selective markers as parts of words—as many studies do—renders the selectivity test useless (Haspelmath 2011: 47). The most useful criteria for defining grammatical words presented above are summarized in Table 3 below. Tab. 3: Summary of potentially useful grammatical word diagnostics Insertion potential Can strings of adjacent elements be interrupted by autonomous elements? → If insertion potential is high, those adjacent elements probably belong to different grammatical words. Displacement potential Can either morpheme in a string be moved/copied independently of the other? → If they can, those adjacent elements probably belong to different grammatical words. Fixed ordering Is a particularstring order arbitrarily fixed, or does it allow the converse order with a different meaning? → If the order is fixed, the adjacent elements in the string may belong to the same grammatical word. Cross-slot dependencies Can either morpheme in a string be deleted without simultaneously deleting another morpheme? → If they cannot, those elements may belong to the same grammatical word. 5 For instance, in Òko, a Benue-Congo language of Nigeria, the locative is expressed as a high tone on the first syllable, cf. ùgbègbèn ‘mirror’ vs. úgbègbèn ‘in the mirror’ (Atoyebi 2009: 94). 192 Fernando Zúñiga 8.3 Relationships between phonological words and grammatical words Even though phonological words and grammatical words that have the same form can be found in natural languages, such matches cannot be simply assumed to hold everywhere. Mismatches seem to be rather frequent, but they have not been systematically studied from a typological perspective. Preliminary exploration of phonological and “long” grammatical words suggests that mismatches between phonological words and grammatical words probably show considerable variation (Bickel and Zúñiga 2017). I limit myself here to outlining some important kinds of mismatch in the North American context, starting with elements that show grammatical autonomy but limited phonological autonomy (§ 8.3.1) and then proceeding to other mismatches (§ 8.3.2). 8.3.1 Grammatical autonomy with limited phonological autonomy One possible mismatch between phonological and morphosyntactic structures consists in elements showing high grammatical autonomy but limited phonological autonomy. Special cases thereof are known as clitics. The term ‘clitic’ is used in fairly disparate ways in the linguistic literature. Older descriptive studies use it to refer to units that somehow fall between words and affixes— typically without being technical about the definition of words. Recent theoretical and/ or comparative studies use the term more technically,6 to refer to prosodically deficient grammatical words (e. g., Klavans 1985).7 Some studies in this vein see clitics as having the distribution of (functional) grammatical words and the form and prosodic properties of affixes. More precisely, Spencer and Luís (2012) define their canonical clitic as follows. Distributionally, words and canonical clitics occur non-selectively at the edge of phrases and show wide scope in coordination, while affixes have limited host choice and narrow scope in coordination. Formally, affixes and canonical clitics are usually prosodically unspecified monomoraic consonant-vowel syllables that are dependent, while words are only minimally monomoraic consonant-vowel syllables that are subject to phonotactic constraints. According to this view, in English the canonical clitic auxil6 Recent studies in the Chomskyan tradition also have a technical definition, distinguishing as they do between clitics and affixes rather sharply. They typically employ a bundle of parameters shown by well-behaved instances of either category, including formal boundedness, obligatoriness, position, and tam-conditioned allomorphy (see, e. g., Nevins 2011 and Woolford 2016). 7 Occasionally, studies distinguish between kinds of bound morphemes based on morphosyntactic cohesion only, without employing phonological criteria but using the term clitic nonetheless. For instance, Watanabe defines “clitics” in Mainland Comox (Salishan) as showing higher insertion and displacement potential than “affixes,” and as not showing morphophonological idiosyncrasies (2003: 38–47). What is a word? 193 iary ’s in he’s come differs both from the suffix s in books and from the strings book and books, which are both grammatical words. Some scholars resort to additional criteria to justify treating particular elements as something different from both words and affixes, frequently without providing detailed prosodic and grammatical characterizations for such elements. Most notably—and potentially problematically, see the end of § 8.2.3—, these criteria include a grammatical autonomy property, namely selectivity: non-selective grammatical elements are routinely classified as clitics (see, e. g., Mithun 1999: 39). For instance, the Plains Cree (Algonquian) elements ni and ki are typically treated as clitics because they can occur on both verbs (e. g., nimâcîn ‘I hunt’ and kimâcîn ‘you (sg) hunt’) and nouns (e. g., nitem ‘my horse/dog’ and kitem ‘your (sg) horse/dog’). Other common criteria are whether the elements in question can be assigned a part of speech, and whether they have a dedicated position in the clause; residual “particles” on the one hand, and second-position or clause-final elements on the other, are prone to clitic treatment (e. g., the Lakota interrogative marker he in (7) below). Older studies also consider whether the elements in question have “full” counterparts that occur under complementary conditions; those that do not are less frequently regarded as clitics. Sometimes cliticization is defined more widely. For instance, while some clitics appear to be canonical clitics and form part of a phonological word defined by phonological features and rules, others, which is presumably more common, are “simply added—as an extra, unstressed syllable—to a fully articulated phonological word after all processes and rules have applied” (Dixon and Aikhenvald 2002: 25). I call the latter annexed clitics here. Different kinds of elements frequently receive clitic treatment across languages. For instance, so-called enclitics are closer to canonical clitics in Lillooet (Salishan) but to annexed clitics in Creek (Muskogean). The former (e. g., specifiers like ti= and =a in ti=n-citxw=a [art=my-house=reinf] ‘my house’, or the question marker =ha in nuk’wʔ-ancíh-as=ha? [help-tr-you-he=q] ‘did he help you?’; Van Eijk 1997: 46–47) tend to remain unstressed, but they are within the domain of stress of phonological words and can bear stress under appropriate circumstances (Van Eijk 1997: 17). By contrast, some Creek case markers like subjective =(i)t in sokha-há:tka=t ‘opossum (sbj)’ and nonsubjective =(i)n in sokha-há:tka=n ‘opossum (nsbj)’ are outside the domain of stress (Martin 2011: 80). These elements also appear depending on several grammatical, semantic, and pragmatic factors and occur at the edge of NPs rather than nouns (Martin 2011: 336–343). Perhaps even more frequently, different kinds of elements receive underspecified clitic treatment in the same language. For instance, recent grammatical descriptions of Lakota (Siouan) treat as enclitics a number of post-verbal elements that express notions related to polarity, tense, aspect, modality, and evidentiality (e. g., Rood and Taylor 1996: 473 f and Ullrich 2011: 820 f ). These elements, however, are phonologically heterogeneous: some are integrated into the verb and closer to canonical clitics (e. g., unreal/ potential ktA and negative šni) while others are autonomous and perhaps best seen as annexed clitics or even nonclitic particles (e. g., evidential kéyA and male assertive 194 Fernando Zúñiga yeló). The examples in (7) illustrate two such elements. The unreal marker ktA in (a) is clearly part of the preceding phonological word ú ‘come’ (regarding stress, nasalization, and other phonological processes/rules). By contrast, the hearsay evidential kéyA in (b) bears its own stress and shows other signs of autonomy; it originated in a Proto-Siouan verb *ke·he ‘say (the preceding)’, and it can also be found in a plural form kéyapi, which suggests it has grammaticalized only incompletely (Rankin et al. 2002: 197): (7) Lakota (Siouan; Ullrich 2011: 821–822, my glosses) a. Ú kta he? ú=ktA=he come=unreal=q ‘Will he come?’ b. Tokhéčela máni kéye. tokhečéla máni kéyA barely walk evid ‘It is said he was barely walking.’ 8.3.2 Other mismatches between phonology and morphosyntax One possible case of mismatch consists in elements showing high phonological autonomy but limited grammatical autonomy. I have called such mirror images of clitics as found in Mapudungun (unclassified, South America) anti-clitics (Zúñiga 2014). Simple instances thereof would be long, cohesive grammatical words that consist of several phonological words, but it is still rather unclear whether complex grammatical words in Eskimoan, Wakashan, Salishan, Algonquian, or Athabaskan can actually behave in this way. To my knowledge, there are no comparative studies on such phenomena. The verbal grammatical words of Blackfoot may be such a case of structurally cohesive units that are phonologically looser. Consider the verbal strings given in (8): (8) Blackfoot (Algonquian) a. Nitsikákomimmotsspinnaan(a). nit-iikakomimm-oti-hpinnaan 1-love.ta-inv-1pl ‘We (excl) are loved.’ (Frantz 2009: 61) b. Nóko’siksi áyo’kaayaawa. n-óko’s-iksi á-yo’kaa-yi-aawa 1-kid-anim.pl ipfv-sleep.ai-3pl-3pl ‘My kids are sleeping.’ (Fox and Frantz 1979: 152; glosses are adapted and tones added) c. Áánistsisa ikkámáakaaistoosi. waanIt-is ikkám-yáak-waaistoo-si say.ta-2sg→3.imp if-fut-come.ai-3.sbjv ‘Ask him if he will come.’ (Frantz 2009: 138) 195 What is a word? Many studies on Algonquian languages regard elements like nit- (8a) and n- (8b) as proclitics, mainly based on some interpretation of the selectivity criterion. Some elements following the verb stem, like -oti and -hpinnaan in (8a), -yi in (8b), and -is in (8c), are routinely considered affixes. Others, like -aawa ‘3pl’ in (8b) and -si ‘3.sbjv’ in (8c), are often considered enclitics, the former on syntactic grounds—it appears only under certain syntactic (and pragmatic) conditions—and the latter on phonological grounds—it is less tightly integrated into the preceding unit—(see Stacy 2004 and references therein). As in Lakota (see § 8.3.1), such “enclitics” are apparently heterogeneous regarding their phonological autonomy; upon closer inspection, -si and other comparable elements may well turn out to be anti-clitics. Regarding their grammatical autonomy and leaving aside the selectivity criterion, Blackfoot verb complexes (including personal markers before and after the stem) seem to be rather tight units. Another case where a systematic kind of mismatch is not analyzable as run-of-themill cliticization comes from Kwak’wala (as already noted in Anderson 1984). Consider (9); the elements da ‘det’ and laχ ənoχ ‘prep 1pl’ are phonologically hosted by the units preceding them but morphosyntactically hosted by the units following them: (9) Kwak’wala (Wakashan; Janzen 2011: 102) phonological unit phonological unit grammatical unit phonological unit grammatical unit ləmis gaχuʔi da ʔixpoʔom a then come det fruit obj Ləmis gaχu’ida ʔixpo’omalaχənoχ gukw. ‘Then the fruit was brought up to our house.’ phonological unit grammatical unit laχ prep ənoχ 1pl gukw house The comparative literature calls such elements ditropic clitics (Embick and Noyer 1999, Cysouw 2005). They seem to occur in several languages of the Northwest Coast, but much more work is needed on them. To the extent that descriptive studies address them, they use terms like “ambivalent affixes” for them (see, e. g., Suttles 2004: 28–29 for a short mention of such (optionally) encliticized elements in Musqueam [Salishan]). 8.4 Selected parameters of variation of grammatical words This section presents two important parameters of variation of grammatical words, namely (i) their internal make-up and (ii) the overall word classes into which they fall. The former is one of the central topics of morphological typology, has been debated in detail since the mid-1990s, and is the object of § 8.4.1. The latter has become highly 196 Fernando Zúñiga controversial in the 20th century —particularly since the 1970s—, and would merit a comprehensive treatment, but I limit myself to sketching its essentials in § 8.4.2. 8.4.1 Morphological make-up Regarding the internal structure of grammatical words, Indigenous North American languages show variation (Sapir 1921: Chh. V–VI; Mithun 1999: Ch.2; Zúñiga 2017, 2019). While grammatical words that potentially consist of a high number of morphemes are frequent, this is not universal in the region. Moreover, those that do have such “polysynthetic” grammatical words show different subtypes thereof. First, morphemic arrangement can be templatic (e. g., in Athabaskan and Iroquoian) or layered (e. g., in Eskimoan). In the former, roots and affixes occur in specific positions and in an invariant order with respect to the other morphemes; in the latter, at least some affixes can occur in different positions, showing semantic and grammatical scope over the material to their left. The examples in (10) (slightly adapted from Mithun 1999: 42–43) illustrate these two subtypes. The templatic arrangement in (a) has both the verb root (-e ‘go’) and the affixes around it occupying specific slots in a fixed order. The layered arrangement in (b) has the verb root (ayag- ‘go’) occupy the leftmost position, but the modal suffix -yugnarqe ‘probably’ can immediately follow either the verbal root and its future marker (ayag-ciq- ‘will go’) or the verbal suffix and its past tense marker, (-ni-llru ‘claimed’), with a difference in meaning: (10) a. b. Tuscarora (Iroquoian) Yaʔnə̨́:tsyə̨:t. y-aʔ-ʔn-ə̨ts-ye-e-t trans-fct-du-rep-indf.a-go-pfv ‘They two went back there.’ Yup’ik (Eskimoan) Ayagciqsugnarqnillruuq. ayag-ciq-yugnarqe-ni-llru-u-q go-fut-probably-claim-pst-ind-3sg ‘He said he would probably go.’ Ayagciqnillruyugnarquq. ayag-ciq-ni-llru-yugnarqe-u-q go-fut-claim-pst-probably-ind-3sg ‘He probably said he would go.’ Second, the morphemes combined into grammatical words may belong to one clause (the default case) or to separate clauses (e.g, in Wakashan and Eskimoan). The examples in (11) from Pittman (2009: 141) illustrate cases where a lexical suffix even allows the complex construction to accommodate a new argument. In the Nuuchahnulth example (a), the intransitive verbal root waʔič- ‘sleep’ combines with the transitive verbal suffix -ʕiƛ ‘come upon’ to form a single grammatical word; note that it is the lexical suffix that introduces the new 1st-person-singular subject. Similarly, in the Inuktitut example (b), the transitive verbal root niri- ‘eat’ licenses two arguments (Miali ‘Mary’ and tuktu ‘(the) caribou’) and the transitive verbal suffix -nira ‘say’ introduces the new subject Jaani What is a word? 197 ‘John’. In both cases, the verbal roots and the verbal suffixes head their own (subordinate and superordinate) clauses, which have been conjoined: (11) a. b. Nuuchahnulth (Wakashan) Waʔičʕiƛitsiš Ken. waʔič-ʕiƛ-mit-siiš Ken sleep-come.upon-pst-1sg.ind K. ‘I found Ken sleeping.’ North Baffin Inuktitut (Eskimoan) Jaani-up niri-qqau-nira-lauq-taa tuktu J.-erg eat-rec.pst-say-rem.pst-3sg→3sg caribou[abs] ‘(A while ago) John said that Mary was eating the caribou.’ Miali-mu. M.-all This parameter of variation has received comparatively little attention and needs to be explored in greater detail (see Zúñiga 2019 and references therein). Lastly, morphemes of different kinds can be combined into grammatical words: roots may combine with grammatical affixes (as in many North American and Indo-European languages), lexical affixes (e. g., in Eskimoan, Wakashan, Salishan, Chimakuan, and Tsimshianic), and/or other roots (“incorporation,” e. g., in Iroquoian, Caddoan, Siouan, Kiowa-Tanoan, Uto-Aztecan, Athabaskan, Tsimshianic, and several unclassified languages). For instance, the nominal root nayir ‘seal’ combines with the lexical affix -cur ‘hunt’ in the Yup’ik grammatical word in (12a); the Onondaga (Iroquoian) grammatical word in (12b) is different: nominal -ʔseht- ‘vehicle’ is a stem (it consists of ʔse- ‘drag’ plus the nominalizer -ht; Marianne Mithun, p.c.) and verbal -hninu ‘buy’ is a root, and either can occur as the sole lexical element of its own grammatical word. (12) a. b. Yup’ik (Mithun 1999: 49) Nayircurtuq. nayir-cur-tuq seal-hunt-3sg.ind ‘He is seal-hunting.’ Onondaga (Iroquoian; Woodbury 1975: 17) Tom waʔhaʔsehtahní·nuʔ. Tom waʔ-ha-ʔseht-a-hninu-ʔ T. pst-3sg.m.A-vehicle-ep-buy-tam ‘Tom bought a vehicle.’ There is much to say about lexical affixation in North American languages, but I limit myself to outlining the basics here (see Mattissen 2006 and Mithun 1997, 1999: 48–56, 2017 for more details). First, note that the mirror image of (12a) is also found; in (13), the verbal root k’wəs- ‘burn’ combines with the affix -cəs ‘hand’: 198 (13) Fernando Zúñiga Musqueam (Salishan; Gerdts 1998: 95) Ni cən k’wəscəs. ni cən k’wəs-cəs aux 1sg.sbj burn-hand ‘I burned my hand.’ Thus, Indigenous North American languages differ regarding the formal status of elements that are semantically comparable to roots, words, or even phrases in many other languages: their grammatical autonomy may vary considerably, and even if they are bound, they can range from root-like to affix-like. Some of the latter are noun-, adjective-, or adverb-like (e. g., Musqueam -cəs), can be found in Northwest-Coast languages and in Algonquian, and express notions like location, direction, instruments, manner, degree, quantification, chronology, and body parts. Others are verb-like (e. g., Yup’ik -cur-), can be found in Eskimoan, Wakashan, and Algonquian, and show semantics that ranges from general (e. g., ‘exist’, ‘have’, ‘make’) to specific (e. g., ‘hunt’, ‘eat’, ‘smell like’). In any given language, lexical suffixes may have non-affixal counterparts (which may be etymologically related to them or not), and the former are often more general (e. g., metaphorically and/or metonymically extended) than the latter. Like incorporated roots/ stems, lexical affixes are both vocabulary-extending and discourse-structuring elements, backgrounding as they do ancillary or established information. Most present-day scholars think lexical affixes probably originated in compounds. For instance, either the verbal or the nominal element in erstwhile Verb+Noun compounds retained the capability of being used as an independent stem while the companion element became affixal; so-called bipartite stems (Jacobsen 1980) arose when both elements of erstwhile Verb+Verb compounds became dependent. 8.4.2 Parts of speech There is no universally accepted answer to the twofold question of what parts of speech (notions like nouns, verbs, etc.) are and how they are identified. Some scholars—often those working in the Chomskyan tradition (e. g., Baker 2003)—see parts of speech as universal notions. Others—often those working in the functional-typological tradition (e. g., Croft 2000)—regard them as language-specific but showing strong cross-linguistic tendencies. Approaches to parts of speech also differ as to the role played by prototypicality and/or gradience in their definition (e. g., Aarts 2007). See Polinsky (2012) for a recent illustration of the main approaches. Languages evidently differ regarding the morphological possibilities and/or necessities of their grammatical words: the formal marking of case, number, gender, person, tense, aspect, etc. shows significant cross-linguistic variation, often even among closely related languages. Morphologically, most languages of Native North America clearly distinguish not only between function and content words but also between different What is a word? 199 subtypes of the latter; these language-specific subtypes can be termed, in the spirit of the model outlined in § 8.2, morphological nouns, morphological verbs, etc. Syntactically, languages often distinguish between elements that can head noun phrases from those that head clauses, which could in turn be termed syntactic nouns, syntactic verbs, etc. As with phonological words and grammatical words, matches between morphological units and syntactic units do exist, but they cannot be assumed to hold everywhere: mismatches exist as well. Several language groups of Indigenous North America have figured prominently in the debates about parts of speech—most notably: Eskimoan, Iroquoian, Salishan, Wakashan, and Chimakuan (see Mithun 1999: 56–67). The issue has often been framed in terms of, and even reduced to, the nature and the degree of distinction between nouns and verbs. The relevant descriptive questions asked in the literature are: (i) can most content grammatical words (including proper names) be used referentially and predicatively, (ii) do such grammatical words need special morphology to perform each task, and (iii) is such morphology available to most roots/stems or just to subsets thereof? Irrespective of the theoretical use to which the answers are put, there cannot be any doubt on the variation encountered in North American languages regarding this issue. Athabaskan and Algonquian languages, for instance, do not normally do what Northwest Coast languages famously can; even though the determiner tit crucially differentiates between the predicative string preceding it and the referential one following it in Upper Chehalis, the structural symmetry shown by pairs like (14) is striking nonetheless: (14) Upper Chehalis (Salishan; Kinkade 1983: 30) a. ʔAc-táw-ɬ tit ʔac-lə́pxw-ɬ. stat-big-intr det stat-hole-intr ‘The hole is big.’ b. ʔAc-lə́pxw-ɬ tit ʔac-táw-ɬ. stat-hole-intr det stat-big-intr ‘The big one is a hole.’ 8.5 Conclusions The previous sections have taken studies from the 2000s onwards on the relevant phenomena seriously: when using the term “word” to address linguistic units, one needs to distinguish between different kinds of words, both within and across languages. The first parameters of variation of words are domain-related (i. e., phonology and morphosyntax), and we saw that phonological and grammatical words can, but need not, match. Several subtypes of grammatical words can be distinguished, based on which kinds of morphemes are combined into words, and how, as well as how many clauses are involved. 200 Fernando Zúñiga Native North American languages show variation regarding all these parameters, and the emerging picture is a complex one. Long polymorphemic grammatical words are not hard to find, and comparative work seems to confirm the traditional impression that many languages of the region share this general feature (see, e. g., Bickel and Nichols 2013 for a worldwide survey of inflectional synthesis on verbs). Incorporation and lexical affixation, by contrast, are hallmarks of the languages of some areas and families only; other languages of the region show neither. Both templatic and layered grammatical words are robustly represented in the region; multiclausal polysynthesis and a tenuous correlation between different morphological grammatical words and syntactic grammatical words (i. e., “weak noun-verb distinction”), under-described though both phenomena are typologically, seem to occur only in few linguistic groups. Comparative work on different kinds of grammatical words and their distribution is still in its infancy, and such analysis of phonological words has not been conducted yet. Further research on these topics will help determine patterns, formulate generalizations, and possibly also identify more cases of mismatches between different kinds of words. We have also much to learn from future work on the origins and the development of different kinds of grammatical words and phonological words. Regarding the issue of parts of speech, rather than arguing in favor or against the adequacy of the received pigeonholes, descriptive and comparative work should focus on the regularities and the oddities found in the lexicon and grammar of the languages under scrutiny. Detailed inventories of subtypes of nouns, verbs, adverbs, and particles, as well as their formal properties, are particularly interesting and useful. Acknowledgements: I am grateful to the editors of the present volume and two anonymous reviewers for their comments on earlier versions of this article. Abbreviations ABS AI ALL ANIM APP ART AUX CL DET DU EP ERG FCT FUT HAB absolutive animate intransitive allative animate appositional article auxiliary classifier determiner dual epenthesis ergative factual future habitual IC IMP IND INDF.A INTR INV IPFV NAFF NEG NSBJ OBJ PFV PL PREP PROTH initial change imperative indicative indefinite agent intransitive inverse imperfective nonaffirmative negative nonsubject objective perfective plural preposition prothesis PST Q REC REINF REM REP SBJ SBJV SG What is a word? past question recent reinforcing remote repetitive subject subjunctive singular SO STAT TA TAM TI TR TRANS UNREAL 201 sticklike object stative transitive animate tense-aspect-modality transitive inanimate transitive translocative unreal References Aarts, Bas. 2007. Syntactic gradience: The nature of grammatical indeterminacy. Oxford: Oxford University Press. Adler, Jeffrey. 2016. The nature of conspiracy: Implications for parallel versus serial derivation. Santa Cruz: University of California at Santa Cruz MA thesis. Aikhenvald, Alexandra. 2002. Typological parameters for the study of clitics, with special reference to Tariana. In R. M. W. Dixon & Alexandra Aikhenvald (eds.), Word: A cross-linguistic typology, 42–78. Cambridge: Cambridge University Press. Anderson, Stephen. 1984. Kwakwala syntax and the Government-Binding theory. In Eung-Do Cook & Donna Gerdts (eds.), The syntax of Native American languages, 21–75. New York: Academic Press. Aoki, Haruo. 1966. Nez Perce vowel harmony and Proto-Sahaptian vowels. Language 42(4). 759–767. Atoyebi, Joseph Dele. 2009. A reference grammar of Oko. Leipzig: University of Leipzig dissertation. Baker, Mark. 2003. Lexical categories: Verbs, nouns, and adjectives. Cambridge: Cambridge University Press. Bickel, Balthasar & Johanna Nichols. 2013. Inflectional synthesis of the verb. In Matthew Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/22 Bickel, Balthasar & Fernando Zúñiga. 2017. The ‘word’ in polysynthetic languages: Phonological and syntactic challenges. In Michael Fortescue, Marianne Mithun, & Nicholas Evans (eds.), The Oxford handbook of polysynthesis, 158–185. Oxford: Oxford University Press. Bisetto, Antonietta & Sergio Scalise. 1999. Compounding: morphology and/or syntax? In Lunella Mereu (ed.), Boundaries of morphology and syntax, 31–48. Amsterdam: John Benjamins. Bloomfield, Leonard. 1926. A set of postulates for the science of language. Language 2(3). 153–164. Bloomfield, Leonard. 1933. Language. New York: Holt, Rinehart & Winston Charney, Jean Ormsbee. 1993. A grammar of Comanche. Lincoln: University of Nebraska Press. Cowell, Andrew & Alonzo Moss, Sr. 2008. The Arapaho language. Boulder: University Press of Colorado. Croft, William. 2000. Parts of speech as language universals and as language-particular categories. In Petra Vogel & Bernard Comrie (eds.), Approaches to the typology of word classes, 65–102. Berlin: Mouton de Gruyter. Cysouw, Michael. 2005. Morphology in the wrong place: A survey of preposed enclitics. In Wolfgang Dressler, Dieter Kastovsky Oskar Pfeiffer, & Franz Rainer (eds.), Morphology and its demarcations, 17–37. Amsterdam: John Benjamins. Dixon, R. M. W. & Alexandra Aikhenvald. 2002. Word: A typological framework. In R. M. W. Dixon & Alexandra Aikhenvald (eds.), Word: A cross-linguistic typology, 1–41. Cambridge: Cambridge University Press. Dunham, Joel. 2009. Noun incorporation in Blackfoot. Manuscript. University of British Columbia. 202 Fernando Zúñiga Dunn, John Asher. 1995. Sm’algyax: A reference dictionary and grammar for the Coast Tsimshian language. Seattle: University of Washington Press/Sealaska Heritage Foundation. Dyck, Carrie. 1999. Cayuga syllable structure. Linguistica Atlantica 21. 69–105. Dyck, Carrie. 2009. Defining the word in Cayuga (Iroquoian). International Journal of American Linguistics 75(4). 571–605. Embick, David & Rolf Noyer. 1999. Locality in post-syntactic operations. In David Embick & Rolf Noyer (eds.), Papers in morphology and syntax, Cycle two, 265–317. Cambridge, MA: MIT Press. Fortescue, Michael. 2017. What are the limits of polysynthesis? In Michael Fortescue, Marianne Mithun & Nicholas Evans (eds.), The Oxford handbook of polysynthesis, 115–134. Oxford: Oxford University Press. Fox, Jacinta & Donald Frantz. 1979. Blackfoot clitic pronouns. In William Cowan (ed.), Papers of the Tenth Algonquian Conference, 152–166. Ottawa: Carleton University. Frantz, Donald. 2009. Blackfoot grammar. 2nd ed. Toronto: Toronto University Press. Frantz, Donald & Norma Jean Russell. 2017. Blackfoot dictionary of stems, roots, and affixes. 3rd ed. Toronto: University of Toronto Press. Gerdts, Donna. 1998. Incorporation. In Andrew Spencer & Arnold Zwicky (eds.), The handbook of morphology, 84–100. Oxford: Blackwell. Hansson, Gunnar Ólafur. 2007. On the evolution of consonant harmony: The case of secondary articulation agreement. Phonology 24(1). 77–120. Harris, Alice. 2006. Revisiting anaphoric islands. Language 82(1). 114–130. Haspelmath, Martin. 2011. The indeterminacy of word segmentation and the nature of morphology and syntax. Folia Linguistica 45(1). 31–80. Hockett, Charles. 1958. A course in modern linguistics. New York: MacMillan. Inkelas, Sharon. 1989. Prosodic constituency in the lexicon. San Francisco: Stanford University dissertation. Inkelas, Sharon. 1993. Deriving cyclicity. In Sharon Hargus & Ellen Kaisse (eds.), Phonetics and phonology, Vol. 4: Studies in lexical phonology, 75–110. San Diego: Academic Press. Jacobsen, William. 1980. Washo bipartite verb stems. In Kathryn Klar, Margaret Langdon & Shirley Silver (eds.), American Indian and Indoeuropean studies, 85–99. The Hague: Mouton. Janzen, Jonathan. 2011. Phrase- and word-level prosody in Kwak’wala. University of British Columbia Working Papers in Linguistics 30. 101–107. Julien, Marit. 2002. Syntactic heads and word formation. New York: Oxford University Press. Kempf, Luise. 2010. In erober: vnd plünderung der Statt: Wie Ellipse von Wortteilen entstand. Beiträge zur Geschichte der deutschen Sprache und Literatur 132(3). 343–365. Kinkade, M. Dale. 1983. Salish evidence against the universality of ‘noun’ and ‘verb’. Lingua 60. 25–40. Klavans, Judith. 1985. The independence of syntax and phonology in cliticization. Language 61(1). 95–120. Langacker, Ronald. 1972. Fundamentals of linguistic analysis. New York: Harcourt Brace Jovanovich. Lewis, Geoffrey. 1967. Turkish grammar. Oxford: Oxford University Press. Martin, Jack B. 2011. A Grammar of Creek (Muskogee). Lincoln: University of Nebraska Press. Mattissen, Johanna. 2006. The ontology and diachrony of polysynthesis. In Dieter Wunderlich (ed.), Advances in the theory of the lexicon, 287–353. Berlin: Mouton de Gruyter. https://doi. org/10.1515/9783110197815.287 Mattissen, Johanna. 2017. Sub-types of polysynthesis. In Michael Fortescue, Marianne Mithun, & Nicholas Evans (eds.), The Oxford handbook of polysynthesis, 70–98. Oxford: Oxford University Press. Michelson, Karin. 1988. A comparative study of Lake-Iroquoian accent. Dordrecht: Kluwer. Mithun, Marianne. 1997. Lexical affixes and morphological typology. In Joan Bybee, John Haiman, & Sandra Thompson (eds.), Essays on language function and language type, 357–371. Amsterdam: John Benjamins. Mithun, Marianne. 1999. The languages of Native North America. Cambridge: Cambridge University Press. Mithun, Marianne. 2017. Polysynthesis in North America. In Michael Fortescue, Marianne Mithun, & Nicholas Evans (eds.), The Oxford handbook of polysynthesis, 235–259. Oxford: Oxford University Press. What is a word? 203 Nevins, Andrew. 2011. Multiple agree with clitics: Person complementarity vs. omnivorous number. Natural Language and Linguistic Theory 29(4). 939–971. Piggott, Glyne L. 1995. Epenthesis and syllable weight. Natural Language and Linguistic Theory 13(2). 283–326. Pittman, Christine. 2009. Complex verb formation revisited: Restructuring in Inuktitut and Nuu-chah-nulth. In Marc-Antoine Mahieu & Nicole Tersis (eds.), Variations on polysynthesis: The Eskaleut languages, 135–147. Amsterdam: John Benjamins. Polinsky, Maria (ed.). 2012. Are lexical categories universal? The view from Chamorro. Special issue of Theoretical Linguistics 38.2. Rankin, Robert, John Boyle, Randolph Graczyk & John Koontz. 2002. A synchronic and diachronic perspective on ‘word’ in Siouan. In R. M. W. Dixon & Alexandra Aikhenvald (eds.), Word: A cross-linguistic typology, 180–204. Cambridge: Cambridge University Press. Rood, David & Allan Taylor. 1996. Lakhota. In Ives Goddard (ed.), Handbook of North American Indians, vol. 17: Languages, 440–482. Washington: Smithsonian Institution. Sadock, Jerrold. 1986. Some notes on noun incorporation. Language 62(1). 19–31. Saenger, Paul. 1997. Space between words: The origins of silent reading. Stanford: Stanford University Press. Sapir, Edward. 1921. Language: An introduction to the study of speech. New York: Harcourt Brace. Schiering, René, Balthasar Bickel & Kristine Hildebrandt. 2010. The prosodic word is not universal, but emergent. Journal of Linguistics 46(3). 657–709. Selkirk, Elisabeth. 1986. On derived domains in sentence phonology. Phonology Yearbook 3. 371–405. Spencer, Andrew. 2001. The paradigm-based model of morphosyntax. Transactions of the Philological Society 99(2). 279–314. Spencer, Andrew & Ana Luís. 2012. The canonical clitic. In Dunstan Brown, Marina Chumakina, & Greville Corbett (eds.), Canonical morphology and syntax, 123–150. Oxford: Oxford University Press. Stacy, Elizabeth. 2004. Phonological aspects of Blackfoot prominence. Calgary: University of Calgary MA thesis. Stump, Gregory. 2001. Inflectional morphology: A theory of paradigm structure. Cambridge: Cambridge University Press. Suttles, Wayne. 2004. Musqueam reference grammar. Vancouver: University of British Columbia Press. Ullrich, Jan. 2011. New Lakota dictionary: Lakȟótiyapi-English, English-Lakȟótiyapi & incorporating the Dakota dialects of Yankton-Yanktonai, Santee-Sisseton. 2nd ed. Bloomington: Lakota Language Consortium. Urbanczyk, Su. 2023 [this volume]. Prosodic morphology. In Carmen Dagostino, Marianne Mithun & Keren Rice (eds.), The languages and linguistics of Indigenous North America: A comprehensive guide. Berlin & Boston: De Gruyter Mouton. Van Eijk, Jan. 1997. The Lillooet Language: Phonology, Morphology, Syntax. Vancouver: University of British Columbia Press. Ward, Gregory; Richard Sproat & Gail McKoon. 1991. A pragmatic analysis of so-called anaphoric islands. Language 67(3). 439–474. Watanabe, Honoré. 2003. A morphological description of Sliammon, Mainland Comox Salish, with a sketch of syntax. Osaka: Osaka Gakuin University, Faculty of Informatics, Endangered Languages of the Pacific Rim. Woodbury, Anthony. 1975. Onondaga noun incorporation: Some notes on the interdependence of syntax and semantics. International Journal of American Linguistics 41(1). 10–20. Woodbury, Anthony. 2002. The word in Cup’ik. In R. M. W Dixon & Alexandra Aikhenvald (eds.), Word: A cross-linguistic typology, 79–99. Cambridge: Cambridge University Press. Woolford, Ellen. 2016. Two types of portmanteau agreement: Syntactic and morphological. In Géraldine Legendre, Michael Putnam, Henriëtte de Swart & Erin Zaroukian (eds.), Optimality-Theoretic syntax, semantics, and pragmatics: From uni- to bidirectional optimization, 111–135. Oxford: Oxford University Press. 204 Fernando Zúñiga Zúñiga, Fernando. 2014. (Anti-)cliticization in Mapudungun. Morphology 24(3). 161–175. Zúñiga, Fernando. 2017. On the morphosyntax of indigenous languages of the Americas. International Journal of American Linguistics 83(1). 111–139. Zúñiga, Fernando. 2019. Polysynthesis: A review. Language and Linguistics Compass 13.4. Zwicky, Arnold. 1987. Suppressing the Zs. Journal of Linguistics 23(1). 133–148. Zwicky, Arnold & Geoffrey Pullum. 1983. Cliticization vs. inflection: English n’t. Language 59(3). 502–513. Daniel W. Hieber 9 Word classes Abstract: This chapter is an introduction to word classes (parts of speech) in indigenous North American languages. It explains theoretical approaches to the study of word classes (language-particular vs. typological) as well as how word classes are classified (lexical vs. functional classes and open vs. closed classes). The core of the chapter is a survey of the word classes commonly found in North American languages. It covers the lexical categories of noun, verb, adjective, and adverb, and the functional categories of adposition, article, auxiliary, particle, and pronoun. There are however many outstanding questions in word-class research. Two of the most prominent ones – locus of categoriality and the noun-verb distinction – are discussed in detail in the latter half of this chapter. The diversity of North American languages continues to challenge our understanding of the nature of word classes. 9.1 Introduction Word classes (traditionally called parts of speech) are groups of words in a language that fill similar slots in an utterance (Croft 2001: 11) and share some linguistic properties, whether those properties are semantic, syntactic, or morphological (Anward, Moravcsik, and Stassen 1997: 171–172; Anward 2001: 726; Rijkhoff 2007: 709; Schachter and Shopen 2007: 1–2; van der Auwera and Gast 2011: 166).1 For example, the class of words that can fill the slot in the utterance the big _____ are typically called “nouns” in English. Noun, verb, and adjective are the best-known classes, but linguists argue for the existence of many others. Languages vary in the number of word classes they have, the characteristics that define those classes, and the number of words that fall into each class (Schachter and Shopen 2007: 1; Velupillai 2012: 122; Smith 2015). Native North American languages have a unique part to play in research on word classes. These languages challenge traditional conceptions of word classes because they do not cleanly map onto the categories of Greek and Latin, which were thought to be universal (Anward, Moravcsik, and Stassen 1997: 167; Vogel and Comrie 2000: xiii). As a result, early Americanist linguists sought to analyze languages on language-internal 1 This definition is intentionally broad, because linguists disagree – often fundamentally – on what word classes are, and how to define them in particular languages (see § 2). Bernard Comrie (p.c.) points out that the present definition could include inflectional classes or valence classes, which are not traditionally considered distinct parts of speech. The tradition in linguistics is that the term word class refers to categories like noun, verb, pronoun, etc. (Haspelmath 2001: 16538). However, some linguists, particularly those that adopt the perspective of construction grammar (see especially Croft 2001), happily accept different inflectional classes or valence classes as types of word classes. See § 2 for more detail. https://doi.org/10.1515/9783110600926-009 206 Daniel W. Hieber evidence alone, rather than impose grammatical models from other languages and traditions (Sapir 1921: 125). The subsequent quest to accurately describe Native American languages on their own terms motivated – and continues to motivate – a large portion of the research into the nature of linguistic categories. An understanding of word classes is useful to speakers and language learners because knowing the category of a word provides speakers with information about how that word is used. The part of speech of a word can indicate which affixes that word is allowed to take, how that word combines with other words or affixes to create new words, and what roles that word can play in a sentence, among other information. This chapter has two primary goals: 1) to introduce the study of word classes with a focus on current approaches 2) to highlight the unique place and contribution of native North American languages in this research Section 2 presents two competing theories of word classes. Section 3 explains the main types of word classes: lexical vs. functional and open vs. closed. Section 4 is a brief survey of some common word classes. Section 5 summarizes two central issues in word class research in North American languages specifically. Section 6 concludes by summarizing the distinct contribution of North American languages to the study of word classes. 9.2 Theories of word classes Today, there are two diametrically opposed perspectives on the nature of word classes (Croft 2001: 63). The first, more traditional approach, argues that individual languages have large, cohesive word classes such as noun, verb, and adjective, but that these categories vary considerably across languages, with perhaps some languages lacking certain categories entirely. Researchers that adopt this perspective differ as to whether they view word classes as clearly defined or fuzzy and prototypical, but they agree that it is possible to define and describe major categories for every language. This is the particularist (that is, language-particular) approach to word classes. The second approach argues that the behaviors of individual words in a language are so diverse that it is impossible to formulate broad definitions for word classes. In this approach, languages do not have major word classes like noun, verb, and adjective. Instead they have a proliferation of tiny categories or constructions. The major word classes are emergent and epiphenomenal (Croft 1991: 30; Croft 2005: 436), arising from the human propensity to view the world through the cognitive prototypes of objects, actions, and properties, and the fact that discourse is fundamentally a sequence of referents and predicates (Sapir 1921: 119; Croft 1991: 124). This cognitive propensity is reflected in various subtle ways in the grammars of all languages. This is the typological (that is, crosslinguistic), constructional, or functional prototype approach to word classes (Croft 2001: 102–104). Word classes 207 It is impossible to discuss word classes without at least implicitly committing to one of these two perspectives. Nearly all the studies referenced in this chapter adopt the particularist approach to word classes. The typological approach to word classes is still fairly recent, and little research has looked at North American languages from a constructional perspective (though see Hieber [2018] and Vigus [2018]). However, since this chapter is a crosslinguistic survey, I adopt the typological approach here. When I use terms like noun or verb in describing a language, I am referring to crosslinguistic prototypes or comparative concepts (Haspelmath 2010), rather than making a claim about the existence or nonexistence of that particular part of speech in that particular language. 9.3 Types of word classes Word classes are typically described along two dimensions: they may contain lexical (“content”) words or grammatical (“function”) words, and they may be open to new members or closed to new members. This section describes each of these types. 9.3.1 Content words vs. function words One way to describe word classes is in terms of the meanings of their words, dividing them into lexical categories or functional categories (Haspelmath 2001; Rijkhoff 2007). Lexical categories contain “content words” which prototypically refer to things, events, or properties in the world. Below are some lexical words in Arapaho (Algonquian). Section § 4.1 discusses lexical categories in more depth. (1) Arapaho (Algonquian) (Cowell and Moss 2008: 56, 61, 74–75) hébes ‘beaver’ hébesii ‘beavers’ wóxhoox ‘horse’ woxhóóxebii ‘horses’ ho’óeet ‘clay’ ho’óeetno ‘(clay-based) ceremonial paints’ bes ‘wood’ béxo ‘sticks’ biixúút ‘shirt’ nebiixúút ‘my shirt’ nííhooyóúʼu ‘they (inanimate) are yellow’ nííhoonéíh(i)t ‘s/he (animate) is yellow) nonóóhowó’ ‘I see him/her’ neihoownoohówoo ‘I don’t see him/her’ 208 Daniel W. Hieber In contrast, functional categories contain words which indicate grammatical relationships or specify features about content words. These are called “function words”, and they typically have abstract meanings. Below are some function words in Creek (Muskogean). (2) Creek (Muskogean) (Martin 2011: 304, 331–332, 357–359, 360–362) leykauxiliary verb, ‘be (while sitting)’ hoyɬauxiliary verb, ‘be (while standing)’ wa:kk- auxiliary verb, ‘be (while lying)’ =ta:t(i) focus of attention =a:t(i) referential (definite/emphatic) The first three words in (2) are auxiliary verbs – words which provide additional information about a main verb (see § 4.2.3). In Creek, auxiliary verbs express aspect, possibility, or strength of assertion (Martin 2011: 298). The last two words in (2) are enclitics – morphemes which behave phonologically like suffixes, but syntactically like independent words, with scope over the entire phrase. While =ta:t(i) attaches to noun phrases and indicates that the noun is the focus of attention or topic of the discourse, =a:t(i) attaches to verb phrases to indicate definiteness or identifiability (Martin 2011: 357, 360). North American languages have a great diversity of functional categories like these. Section § 4.2 describes several common ones in more detail. The terms lexical category and functional category are not used the same way by all researchers. Both lexical category and functional category are sometimes used to refer to word classes as defined here (e. g., Payne 1997: 32). Sometimes word class is used to refer to lexical categories (Rijkhoff 2007: 710). It is also common to use the term grammatical categories for word classes, although this term more typically refers to formally marked features of a word such as person, tense, or number (Crystal 2008: 68–69; 186–187; Trask 1993: 122). Another related term is syntactic categories; this is sometimes used in the equivalent sense of lexical categories, sometimes in the broader sense of word classes (see Rauh [2010] for an extended discussion). It is helpful to be aware of these terminological differences when reading linguistic research. The distinction between lexical and functional categories is not always a clear one. Adpositions (prepositions and postpositions) often have both lexical and functional uses (Haspelmath 2001: 16539; Smith 2015: 178). Consider the two uses of the word by in English in (3). (3) English (Indo-European) (Corpus of Contemporary American English; Davies 2020) a. Remember the last time you passed by your favorite park b. If your life was destroyed by the money that paid for this thing In example (a), by is lexical, meaning ‘next to’ or ‘in proximity to’. In (b), by is functional, marking the agent of a passive clause. Adpositions in Chitimacha (isolate) also have both lexical and functional uses. In (4), the postposition hix may mean ‘with; by means of’ (its lexical sense, in (a)) or mark the agent of a transitive verb (its functional sense, in (b)). (4) Word classes 209 Chitimacha (isolate) (Swadesh 1939a: A10k.2, A11c.10) a. hus mahci kuh hix qapx nehpapuyna hus mahci kuh hix qapx neh-pa-puy-na 3sg tail feather with refl cover-caus-hab-nf.pl ‘they adorn themselves with his tail feathers’ b. qix hix hi koomicukix qix hix hi kow-ma-cu(y)-ki-x 1sg erg and call- plact-irr-1sg.a-cond ‘if I call them’ The reason for this gradation between lexical and functional uses of the same word is that function words derive historically from lexical words, a process known as grammaticalization (Hopper and Traugott 2003). A language will often retain the older, lexical meaning alongside the newer, functional meaning. Another example of the cline between lexical and functional uses of a word is the use of words meaning ‘sit’, ‘stand’, and ‘lie’ as auxiliary verbs indicating progressive or continuative aspect in Siouan (Mithun 1999: 115–116), some Muskogean languages (Munro 1984; Broadwell 2006: 209–211), and Chitimacha (Hieber 2019: 350–352), among others. Example (5) shows lexical and functional uses of ‘sit’, ‘stand’, and ‘lie’ in Mandan (Siouan). (5) Mandan (Siouan) (Kennard 1936: 31) a. wɛ́rɛx nakóc wɛ́rɛx nak-oc pot sit-prs ‘A pot was there (sitting).’ b. mah ísɛkanakeròmakoc ‘he was (sitting) making an arrow’ c. múixtɛ ̀na tɛ́romakoc múi-xtɛ-na tɛ-romakoc village-big-emph stand-narr.pst ‘there was a big village’ d. ptáhakekaʼ ‘he was running around (upright)’ e. máːta makómakoc máːta mak-omakoc river lie-narr.pst ‘the river was there’ f. miníxamakɛkaʼ ‘he was playing (prone)’ 210 Daniel W. Hieber 9.3.2 Open classes vs. closed classes Another way to describe word classes is by how open they are to new members. Open classes are typically large and have new words added to them frequently, whereas closed classes are typically small, limited to a fixed set of words, and add new members only slowly and infrequently (often through grammaticalization) (Robins 2014: 214–215; Schachter and Shopen 2007: 3; Velupillai 2012: 115). English articles, for example, are a closed class of just two words (the and a/an), while English nouns are in principle unlimited, with more words added all the time. There is gradation here as well: English prepositions are generally considered a closed class even though they constitute a large group of words (greater than 100 members), because new prepositions are not created easily. Nonetheless, new prepositions do occasionally arise. For example, prepositional uses of the word absent (as in the utterance absent those ropes, we’d float to a new and faraway place [COCA]) arose only in the 1940s (Harper 2020). In North American languages, one somewhat common closed class of words is the preverb category, words which form a semantic unit with their verb, and often indicate things like direction or aspect (Los et al. 2012: Ch. 1).2 Chitimacha has a closed set of 10 preverbs, shown below in (6) (Hieber 2018). By contrast, Menominee (Algic) has a large open class of preverbs (Bloomfield 1962: 214). (6) Chitimacha (isolate) (Hieber 2018: 19) hi ‘to, there’ his ‘back to, back there’ kap ‘up, beginning, becoming’ kaabs ‘back up’ ka ‘across’ kas ‘back across, apart, reverse’ ni ‘down’, definite qap ‘here, coming’ qapx ‘back here, coming back’, reflexive, reciprocal While open classes tend to be lexical ones and closed classes tend to be functional ones, this is just a tendency (Velupillai 2012: 115). Some Australian languages (Dixon 1976: 615–768; Dixon 1980: 280–281) and Papuan languages (Foley 1986: 113–118) have small, closed classes of verbs (Anward 2001: 728). However, I know of no North American language which has a closed class of verbs like this. Closed adjective classes are likewise less common in North America. In a balanced sample of 27 languages in Mexico and north2 In some language families, the term preverb is used for certain types of verbal prefixes with lexical meanings, rather than for syntactically distinct words. This is the case in many Dene languages, for example (Rice & de Reuse 2017: Sec. 23.2.2). Interestingly, the functions and meanings of these affixal preverbs are similar to those of syntactically free preverb classes in other languages, suggesting that preverbs are a coherent typological class whose morphological boundedness is a cline. Word classes 211 ward, Velupillai (2012: 127–128) finds that 7 have a closed adjective class. Velupillai analyzes most languages in the sample as lacking an adjective class entirely (17 languages), and the few languages with an open adjective class are constrained to Mesoamerica (3 languages). Cupeño (Uto-Aztecan) has fewer than 100 adjectives (though Hill [2005: 202] notes that “the classes of adjectives and adverbs are not closed by structural principle but simply have relatively few members”). In Wichita (Caddoan), property concepts are expressed through verbs; however, a handful of words behave like inflected noun stems rather than inflected verb stems. The only words in this category are Riwa·c ‘big’, Rikic ‘little’, riya·s ‘old’, and colors such as khac ‘white’ and kʷah·c ‘red’ (Rood 1996: 594–595). 9.4 Word classes in North American languages This section describes the major lexical categories (§ 4.1) and a sample of functional categories (§ 4.2) in North American languages from a crosslinguistic perspective, in keeping with the functional-typological approach presented in § 2. 9.4.1 Lexical classes The four most widely-discussed lexical classes are nouns, verbs, adjectives, and adverbs. This section briefly defines each in turn. 9.4.1.1 Nouns Nouns are words whose prototypical function is to refer to things (Croft 1991: 51–52; 2001: 66, 89). The best exemplars are “time-stable” entities such as people, places, and things (Givón 2001: 51), but nouns frequently refer to non-prototypical concepts as well, such as abstract ideas and feelings. Distributionally, the prototypical function of nouns is to serve as a participant in a clause, or as the head of a noun phrase that does so. Typologically, nouns regularly have special forms or markers for the grammatical categories of number, possession, definiteness/specificity, noun class (“gender”), or case/ grammatical relations (Croft 1991: 79; Haspelmath 2001: 16541; Dixon 2010: 54–55; Velupillai 2012: 125). However, every one of these features may be marked on verbs as well, meaning that the presence of these grammatical categories is not a failproof diagnostic for distinguishing nouns from verbs. I demonstrate a few such cross-cutting examples in the remainder of this section. Number: On nouns, number marking indicates plurality of the referent; analogously, some languages have a kind of number marking on verbs called pluractionality (also event number or verbal number). Pluractionality indicates that the event happened 212 Daniel W. Hieber multiple times, or that the action was distributed among multiple participants (Mithun 1988: 215–218; Mattiola 2020). Most North American languages surveyed by Mattiola (2020) have pluractional morphology. Possession: While many languages indicate a possessive relationship between two nouns by marking either the possessor noun or the possessed noun, Nuuchahnulth (a.k.a. Nootka; Wakashan) also allows possessive marking on verbs. When the possessive suffix -ʔa·k appears on nouns, it indicates that the noun is possessed by the subject of the clause. When the suffix appears on verbs, it indicates that the subject of the verb is the possessor of the noun phrase. The two examples in (7) illustrate this contrast. Possession is not an exclusively nominal category. (7) Nuuchahnulth (Wakashan) (Nakayama 2001: 128) a. ʔaapḥiiʔiš ɬuucmaakqs ʔaːp-ḥi·-ʔi·š ɬuːcma-ʔa·k-qs kind-dur-ind.3 wife-poss-sbj.1sg ‘My wife is kind.’ b. ʔaapḥiiʔaks ɬuucma ʔaːp-ḥi·-ʔa·k-s luːcma kind-dur-poss-1sg wife ‘My wife is kind.’ Definiteness: Verbs may have special morphology indicating that the speaker is referring to a definite (identifiable) action, or a definite/indefinite participant involved in the action. Chitimacha, for example, has a preverb ni which marks the verb as definite. The verb gast- ‘plant’ is transitive, usually taking an object, but with the preverb ni it becomes intransitive and means ‘plant it’, indicating that there is some definite thing being planted, whose identity can be understood from context. Definiteness is therefore also not criterial of a noun category. Noun Class: Nouns in many languages take affixes that signify some inherent property about the item, such as its animacy, spatial orientation, or shape (Mithun 1999: 95–103), separating nouns into classes. For example, the Iroquoian and Algonquian languages make a morphological distinction between animate and inanimate nouns, with different affixes for each. In Yurok (Algic), however, adjectives also make this distinction (see § 4.1.3; Robins 1958: 93–95). Dene languages have an entire set of classificatory verbs whose stems change based on the countability, number, animacy, and shape/consistency of their absolutive argument (Jaker, Welch, and Rice 2019: 497). Classification is therefore not a category unique to nouns. Word classes 213 9.4.1.2 Verbs Verbs are words whose prototypical function is to predicate – that is, to state something about a referent (Searle 1969: 23–24; Croft 1991: 109–111; Croft 2001: 66, 89). The best exemplars of verbs are actions, events, and processes (Givón 2001[1984]: 52), but verbs frequently convey static meanings as well, such as location or knowledge. Distributionally, the prototypical function of verbs is to serve as the head of a clause. Typologically, it is common for verbs to mark the grammatical categories of tense, aspect, mood, polarity (negative/positive), evidentiality (source of knowledge), epistemic modality (attitude towards the statement), event number, verb class, or grammatical relations (information about the participants in the clause and their relations to one another) (Croft 1991: 79; Haspelmath 2001: 16541; Dixon 2010: 52–54). In § 4.1.1, I noted that nouns also indicate grammatical relations; it is quite common for North American languages to indicate grammatical relations on both nouns and verbs. An important difference is that markers of grammatical relations on nouns indicate their own role in the clause, while markers of grammatical relations on verbs indicate the role of its arguments in the clause. The grammatical categories most commonly associated with verbs may be found on other categories as well. Since the noun-verb distinction is treated at length in § 5.2, I will mention just two cross-cutting cases here: Although tense is the most canonical grammatical category associated with verbs, Makah (Wakashan) nouns may also appear with tense markers. Compare the predicative and referential uses of the tense marker in (8). (8) Makah (Wakashan) (Jacobsen 1979: 113) a. baʔasʔu house-pst-ind.3 ‘It was a house.’ b. baʔasʔuq house-pst-art ‘the former house’ Similarly, although aspect is also canonically associated with verbs, Nuuchahnulth (also Wakashan) allows aspect markers on nouns: (9) Nuuchahnulth (Wakashan) (Nakayama 2001: 48) a. ḥaaḥuupač̓ak ḥaːḥuːp-(y)a-č̓ak instructing-cont-instrument ‘This is a teaching.’ b. ḥaaw̓iɬaƛ̓aƛ qaaḥma ḥaːw̓iɬaƛ-ʼaƛ qaaḥma young.man-tel name ‘Qaahma was a young man.’ 214 Daniel W. Hieber When nouns are used in this non-prototypical way, they are inherently durative, and may only form existential, classifying, or identifying expressions (Nakayama 2001: 48). 9.4.1.3 Adjectives Adjectives are words whose prototypical function is to modify – that is, to specify additional features, qualities, or attributes of a referent (Searle 1969: 23–24; Croft 1991: 109–111; Croft 2001: 66, 89; Schachter and Shopen 2007: 13). Adjectives always modify nouns; words which modify other kinds of items are classified as adverbs (see § 4.1.4). The best exemplars of adjectives are words which attribute properties having to do with value, dimension, age, speed, physical property, and color (Dixon 1977), but adjectives may convey a vast diversity of concepts depending on the language. Adjectives may have distinct forms for comparatives (taller), superlatives (tallest), and equatives (as tall as). Adjectives are not the only way that languages can convey information about attributes and properties. They may have verbs meaning ‘have quality X’, or nouns meaning ‘thing with quality X’. Consequently, adjectives crosslinguistically tend to associate with either nouns or verbs (Croft 1991: 131; Wetzer 1996: 19; Croft 2000: 94). This is especially true for North American languages – there are few if any morphosyntactic devices dedicated to modification. However, words for property concepts usually exhibit behaviors which are different from other words in their class, often justifying the recognition of a subclass of verbs or nouns. There are North American languages with a large, open class of adjectives such as Chitimacha (see discussion in § 5.1) or Central Pomo (Pomoan) (Mithun 1999: 474), but this is rare. Slightly more common are languages with a small, closed class of adjectives. In a sample including 23 North American languages, Velupillai (Velupillai 2012: 128) finds only 6 of those languages have a distinct but closed class of adjectives. Southern Paiute (Uto-Aztecan) has only about a dozen adjectives, for the concepts large, small, long, short, new, old, good, high, strong, hard, and cold (Sapir 1930: 77–79). We have already seen the small class of adjectives in Wichita (Rood 1996: 594–595). Tłı̨ cho Yatıì (a.k.a. Dogrib, Na-Dene) likewise has a closed set of 20 adjectives which are distinguished by their lack of inflectional morphology (Welch 2016). Most North American languages arguably lack an adjective class, such that property concepts are a subcategory of noun or verb or divided between both. Only a few North American languages encode property concepts as nouns; most languages code property concepts as a subclass of verbs. Rincón Luiseño (Uto-Aztecan) is one language which codes some property concepts like nouns. While most modifiers in Luiseño are derived from verbs, the most prototypical property concepts take noun endings, e. g., yoot ‘large’ and kiháat ‘small’ (Kroeber and Grace 1959: 59). The following examples illustrate the morphological similarity between nouns and adjectives. (Note that the Word classes 215 “absolutive” suffix3 in these examples has various realizations – here -ch, -l, or -sh – and that the plural of ‘girl’ is irregular/suppletive.) (10) Rincón Luiseño (Uto-Aztecan) (Elliott 1999: 27–28) a. nawítma-l yawáywi-sh girl-abs pretty-abs ‘pretty girl’ b. nánatma-l-um yawáywi-ch-um girl-abs-pl pretty-abs-pl ‘pretty girls’ c. Yaʼá-sh tóow-q nawítma-l-i yawáywi-ch-i. man-abs see-prs.sg girl-abs-obj pretty-abs-obj ‘The man sees a pretty girl.’ d. péshli-chal yawáywi-chal dish-ins pretty-ins ‘with the pretty dish’ The Maidu (Maidun) and Cherokee (Iroquoian) languages are like the Chitimacha language mentioned above in that they contain an open class of adjectives, all of which are formed from verbs (Dixon [1911: 716–717] for Maidu; Lindsey and Scancarelli [1985], Chafe [2012], and Barrie and Uchihara [2019] for Cherokee). Unlike Chitimacha, however, these languages use nominal rather than adjectival affixes for modifiers. Adjectives in these languages are therefore a subclass of nouns which are all derived from verbs. By far the most common way to encode property concepts in North American languages is as a subclass of verbs. The following examples illustrate the use of such property concepts in a selection of languages, comparing them to action verbs in the same language. (11) Navajo (Na-Dene) (Young and Morgan 1980: 216, 290) a. yi-sh-cha ipfv-1sg-cry ‘I am crying’ b. ni-s-neez n.ipfv-1sg-tall ‘I am tall’ 3 In the North American tradition, the term absolutive sometimes refers to the default or unmarked form of a noun rather than the single argument of an intransitive verb (as most linguists use the term today). Grammatical descriptions of Luiseño often use this former, more traditional sense of the term. I retain that usage in the examples here. 216 Daniel W. Hieber (12) Quileute (Chimakuan) (Andrade 1933: 267, 257) a. čaːč-a-ø fly-dur-3abs ‘it is/was flying’ b. tsiʔda-ʔa-ø handsome-dur-3abs ‘he is handsome’ (13) West Greenlandic (Eskaleut) (Fortescue 1984: 120, 121) a. isir-puq ingil-luni-lu come_in-3sg.ind sit_down-4sg.ctmp-and ‘she came in and sat down’ b. illu-at kusanar-puq house-3pl.poss pretty-3sg.ind ‘their house is pretty and warm’ Many North American languages make a distinction in their verbal person marking between agents – the argument in the clause that performs, effects, instigates, or controls the event – and patients – the argument in the clause which lacks one or more of these properties. In languages with property verbs that exhibit agent-patient marking, property verbs often use patient person markers, although this is just a typological tendency. Examples include Alabama (Muskogean; Wetzer [1996: 216–217]), Kiowa (Kiowa-Tanoan; Wetzer [1996: 187]), Lakota (Siouan; Pustet [2002: 388]), and Mojave (Yuman; Wetzer [1996: 187]), among many others. Central Pomo is one North American exception to this tendency: basic adjectives appear with agent markers unless they are inchoative (‘becoming’) (Mithun 1991: 521). As mentioned, property verbs often exhibit slightly different behavior from prototypical event verbs within a language. The most common difference is that property verbs are limited in the range of inflectional possibilities they can take (what Croft [1991; 2000; 2001; 2003; 2010] calls their behavioral potential). They may be limited to the stative or durative aspects, for example. Another common difference is that property verbs may modify nouns directly, but event verbs require some type of additional nominalizing or relativizing morphology to modify nouns. For instance, in a thorough review of evidence for adjective classes in several Siouan languages, Helmbrecht (2006; in progress) reports that in Hocank (a.k.a. Winnebago; Siouan) nouns may be modified using relative clauses. Relative clauses in Hocank are structurally nearly identical to noun phrases in the language. They typically require a determiner and person inflection, and they may take tense and aspect marking. However, when a property word is used to modify a noun, it does not require a determiner, is never inflected for person, and never takes tense or aspect marking. In Choctaw (Muskogean), property words are morphological verbs, but there are clear semantic regularities which set them apart from other verbs. In Muskogean languages, verb stems undergo certain phonological changes such as nasalization, h-inser- Word classes 217 tion, lengthening, etc. to indicate their aspect. These sets of phonological changes are called grades in the Muskogean literature. When applied to property words, however, these grades have the semantic effect of intensifiers or comparatives rather than aspect (Haag 1995, 1997). Secondly, when these property words appear after nouns, they optionally show nominal morphology, with a penultimate pitch accent and final glottal stop (Broadwell 2006: 223). In the case of Nuuchahnulth, Nakayama (2001) argues that the adjective class is a discourse tendency rather than a clearly-defined set of properties that pick out a mutually exclusive set of words. He reports that adjectivals are words which do not take objects, and which may be combined with nominals to form a phrase (Nakayama 2001: 50). They may however sometimes also serve directly as arguments. A more unusual pattern of behavioral differences for property concepts occurs in Yurok, where numbers and about eleven roots expressing property concepts vary the form of their stem (that is, their stems are suppletive) based on the category of the noun they modify. The categories include human, animal, plant (non-tree), tree, stringlike, flat, and others. Each one of these stems in Yurok potentially has a different form of the stem for each one of these categories. Example (14) shows the different stems for ‘big’ and ‘small’. (14) Yurok (Algic) (Robins 1958: 93–95) Category ‘big’ ‘small’ human beings animals and birds round things body parts, utensils, clothes stringlike things flat things houses boats peloy plɹʔɹy ploy(keloy) plep plep ploks pleʔloy pleyteloy cey(kel) cɹykɹʔɹy ceykoh cey(kel) cey(kel) cey(kel) ceykoh cey(kel) In other languages there seems to be no substantive behavioral differences between property words and event verbs. In Seneca (Iroquoian), words expressing property concepts do belong to a subclass of verbs that are limited to stative aspect, but there are numerous other, non-property verbs which also belong to this class (Chafe 2012). Example (15) lists representative sets of property words and event words in Seneca, both of which are restricted to the stative aspect. 218 (15) Daniel W. Hieber Seneca (Iroquoian) (Chafe 2012: 13–14) a. osde’ ‘it’s heavy’ otgi’ ‘it’s dirty’ odö:sgwi:h ‘it’s wrinkled’ o:ni:yöh ‘it’s hard, tough’ ojiwagëh ‘it’s sour, bitter’ hohsë:h ‘he’s fat’ hodí’gyö’ ‘he’s shy’ b. otga:h ‘it’s making a noise’ owë h́ de’ ‘it has something added to it’ hotö:de’ ‘he hears it’ honö́hdö’ ‘he knows it’ hóío’de’ ‘he’s working’ hohse’ ‘he’s riding on its back’ ho:wísdagá’de’ ‘he has a lot of money’ ho’áshägéhde’ ‘he’s carrying a basket on his back’ Chafe (2012) considers seven possible criteria that might identify a class of adjectives in Seneca (and by extension all of Northern Iroquoian) and finds that all the criteria are subject to the same problem in that they include non-property concepts as well. Finally, some languages distribute property concepts across multiple word classes. In Chinook (Chinookan), words expressing speed, color, and a few words for human propensity are particles, while words expressing age are verbs, and words expressing dimension, value, and other human propensity concepts are nouns (Dixon 1977: 53–54). In sum, the encoding of property concepts in North American languages shows tremendous diversity. Some languages have a large, open class of distinct adjectives, others have a small closed class, but in most North American languages property words are a subset of either nouns or verbs. And in a few cases, even the existence of such a subclass is difficult to motivate. 9.4.1.4 Adverbs Adverbs, like adjectives, are words whose prototypical function is modify; however, they differ from adjectives in that adjectives only modify nouns, while adverbs may modify essentially anything else (Haspelmath 2001: 16543; Velupillai 2012: 130), including verbs (run quickly), adjectives (quite happy), other adverbs (very quickly), prepositions (right out), noun phrases (quite the party), entire utterances (unfortunately), but not individual nouns (*dog quickly) (Velupillai 2012: 130). Semantically, adverbs prototypically convey meanings such as manner (quickly), degree (extremely), time (now), location (there), or evidential/epistemic attitude (probably, frankly) (Quirk et al. 1985; Velupillai 2012: 130). Like adjectives, adverbs may have distinct forms for comparatives (faster), superlatives (fastest), and equatives (as fast as). Word classes 219 Adverbial constructions are not strongly grammaticalized in North American languages. In a recent handbook of North American languages, discussions of adverbs appear only sparingly, and the term “adverb” does even not appear in the index (Siddiqi et al. 2020).4 Languages use a variety of other strategies for conveying prototypical manner concepts instead. In Dëne Sųłıné (Na-Dene), locative nouns may function as adverbs (Cook 2004: 303), and adverbs in most Dene languages are derived from relative clauses (Jaker, Welch, and Rice 2019: 497). Chitimacha has a set of suffixes expressing manner, including -di ‘doing horizontally’, -duwa ‘doing suddenly’, -kint ‘by pushing’, and -ti ‘by handling’. In Nuuchahnulth adverbial concepts like ‘also’, ‘for two days’, and ‘still’ are encoded with intransitive predicates (Nakayama 2001: 51–53). Nuuchahnulth also has a number of lexical suffixes expressing location, such as -ʽis ‘being on the beach’, -ʼas ‘being on the ground’, -ʼa· ‘being on the rock’, or -ʽiɬ ‘being in the house’. Otherwise, adverbial concepts are expressed through verb serialization. The examples below show serial verbs expressing manner, time, and location, respectively. (16) Nuuchahnulth (Wakashan) (Nakayama 2001: 100) a. ƛawaʔiiʔaƛquuč kʷaačiƛ ƛawa-ʔiː-ʼaƛ-quː-č kʷaː-či(ƛ) near-reaching-tel-cond.3-infer move.backwards-mom he.would.go.near move.backwards ‘[While he was dancing] he would go near [him] moving backwards.’ b. qiis waɬyuu qiː-s waɬ-yu· for.long-1sg go.home-done I.for.long at.home ‘For a long time I stayed at home.’ c. yacaaqtuu ʔucačiƛ ʔuuƛ̓aqči yac-a·qtu· ʔu-ca-či(ƛ) ʔuuƛ̓aqči step-going.over it-going.to-mom name walked.over.the.hill went.there Odlaqutla ‘They went over [the high land] to Odlaqutla.’ 9.4.2 Functional classes As mentioned in § 3.1, North American languages exhibit a large variety of function words. This section covers just a few functional classes for which North American languages exhibit unique or interesting behaviors – adpositions, articles, auxiliaries, particles, and pronouns. 4 This point is a comment on the structure of North American languages rather than a criticism of the volume’s coverage. 220 Daniel W. Hieber 9.4.2.1 Adpositions Adpositions are words that govern a noun phrase and signal a relationship between the noun phrase and another word in the clause (Hagège 2010: 1; Kurzon and Adler 2008: 2). It can be difficult to distinguish adpositions from case markers since they both signal relationships between elements of a clause, and there is often a diachronic and synchronic cline between them resulting from grammaticalization (Hagège 2010: Sec. 2.2). Adpositions may also be clitics (Hagège 2010: 18). There are several types of adpositions: prepositions precede the noun phrase, postpositions follow the noun phrase, and circumpositions consist of two elements, one which precedes the noun phrase and one which follows it. Shoshone (Uto-Aztecan) has another type called an inposition which occurs inside the noun phrase (Dryer 2013c). Many second-position clitics could also be considered a type of inposition. Perhaps because North American languages signal many of the relationships among participants using affixes (especially verbal affixes), adpositions are not a robust word class in most North American language families. Even when present in a language, they are sometimes a marginal class, only lightly grammaticalized. Creek (Muskogean), for example, has just a small set of nouns such as naɬkapá ‘middle’ which have grammaticalized into postpositions indicating location (Martin 2011: 147–148). Not all languages have adpositions however (Croft 1991: 144). In Nuuchahnulth, relationships between referents are always communicated by predicates, as shown below. (17) Nuuchahnulth (Wakashan) (Nakayama 2001: 53) šišaa ʔuuʔatup kʷakuucuk šiš-(y)a· ʔu-ʽatup kʷakuːc-uk clean-cont it-doing.for grandchild-poss cleaning doing.for.them her.grandchildren ‘She would peel them for her grandchildren.’ 9.4.2.2 Articles As mentioned in § 4.1.1, one of the categories that can be indicated on a noun is definiteness. One way languages do this is by using definite/indefinite articles (such as the, a/ an in English). Lakota has a set of two definite articles and three indefinite articles with slightly different uses, exemplified in (18). Articles in Lakota follow their noun phrase rather than precede it. (18) Lakota (Siouan) (Van Valin 1977: 63) a. ki ‘the’ definite b. kʔũ ‘the aforementioned’ definite c. wã ‘a (specific)’ indefinite d. wãẑi ‘one (nonspecific)’ indefinite e. cha ‘a (contrastive)’ indefinite Word classes 221 Quileute (Chimakuan) only has an indefinite article (Andrade 1933: 246). Chitimacha uses a demonstrative adjective as a definite article, a common strategy in North America (Dryer 2013b). 9.4.2.3 Auxiliaries Auxiliary verbs are verbs which provide grammatical information about the main verb they accompany (Velupillai 2012: 146). In example (19) from Southern Pomo (Pomoan), the future tense marker appears on the auxiliary verb yo- rather than the predicate kac:i ‘cold’. (19) Southern Pomo (Pomoan) (Walker 2013: 356) kac:i yo-kʰ:e cold aux-fut ‘it will be cold’ While the grammatical features that auxiliary verbs carry typically include tense, aspect, person, number, etc., a prevalent feature of North American languages is that auxiliary verbs may also provide information regarding the spatial orientation of their subjects – usually sitting, standing, or lying position. This is especially common in the Siouan languages (Mithun 1999: 115–116) and the languages of the U.S. Southeast (Campbell 1997: 342). See § 3.1 above for an illustration of positional auxiliary verbs in use in Siouan. 9.4.2.4 Particles Language descriptions often include a word class called particles, but particles are not a coherent typological class. The term “particle” is a morphological term, typically referring to words which are invariable and/or do not have inflectional morphology (Crystal 2008: 352). However, the functions of uninflectable words are not consistent across languages. Particles in Algonquian cover a wide array of functions such as quantifiers, numerals, adjectives, adverbs, prepositions, and conjunctions (Oxford 2007; 2019: 511). Particles in Chitimacha, on the other hand, are used for preverbs, postpositions, negation, topic marking, discourse markers, and interjections. There is no typological prototypical core to particles as a word class. 9.4.2.5 Pronouns Pronouns are words that either refer to discourse participants (I, you, s/he), refer anaphorically to referents that are activated in the discourse (Kibrik 2011: 73), or otherwise stand in for nouns. Pronouns referring to discourse participants are called personal 222 Daniel W. Hieber pronouns, while the others are sometimes called proforms (Bhat 2004: 5). Personal pronouns may be syntactically free words (free pronouns), affixes on the verb (bound pronouns), or clitics (clitic pronouns). The following example from Mohawk (Iroquoian) includes both the free pronoun í:se’ ‘you’ and the bound pronoun sa- ‘you’. (20) Mohawk (Iroquoian) (Mithun 2013: 292) Í:seʼ tókaʼ wà:kehreʼ tókaʼ thé:nenʼ íseʼ tokaʼ waʼ-k-ehr-eʼ tokaʼ thenenʼ you maybe fact-1sg.a-think-pfv maybe something you maybe I thought maybe something sarì:waienʼ ne ahsheriʼwanón:tonhseʼ. sa-rihw-a-ien-ʼ ne a-hshe-riʼwanonton-hs-eʼ 2sg.p-matter-have-stat the irr-2sg/fi-ask-ben.appl-pfv you issue have the you would ask her ‘I thought that you might have some questions to ask her.’ Southern Pomo has clitic pronouns rather than affixes: (21) Southern Pomo (Pomoan) (Walker 2013: 229) mihyanakʰ:eʔwamt̯ aʔa mihyana-kʰ:e=ʔwa=mt̯ a=ʔa kill-fut=cop.evid=2sg.p=1sg.a ‘I’m going to kill you.’ All languages have free pronouns, irrespective of whether they also have bound or clitic pronouns. In North American languages, discourse participants are predominantly expressed using bound pronouns on the verb (Dryer 2013a). In these languages, the functions of the pronouns are divided between the bound and free forms. The bound pronouns are used to refer to and track referents in the discourse, while the free pronouns accomplish the various other functions, such as focus/emphasis, cleft constructions, topicalization, antitopicalization, etc. (Mithun 2003; 2013). There are many types of proforms, including demonstrative (22), indefinite (23), interrogative (24), possessive (25), and relative (26), among others. (22) Demonstrative: Potawotami (Algic) (Lockwood 2017: 58) Apte ode gminen. apte ode gminen half dem.near I.give.you ‘I’ll give you half of this.’ (23) Indefinite: Seneca (Iroquoian) (Chafe 2015: 118) Ëké:owi’ sö:ga:’. I’ll tell them someone ‘I’ll tell someone.’ Word classes (24) Interrogative: Southern Pomo (Pomoan) (Walker 2013: 231) čaʔ:aʔkam:u ʔaṭʰ:a ʔahsoduy čaʔ:a=ʔka=m:u ʔaṭʰ:a ʔahso-duy-∅ who=inter=3sg gravel throw.many.small-dir-pfv ‘who threw the gravel?’ (25) Possessive: Creek (Muskogean) (Martin 2011: 144) ca-nâːki-t ôː-s 1sg.p-thing-t be.fgr-ind ‘It’s mine.’ (26) Relative: Tuscarora (Iroquoian) (Mithun 2012: 270) Thwé:ʼn waʼkayęʼnaʼnitʼúthahs haʼ káhneʼ kayęʼnęʼnę́nhyahr. all he put them to sleep the who they are guarding him ‘He put to sleep all those who were guarding him.’ 223 9.5 Issues in word-class research This section describes the most prominent themes in research on word classes in North America. The difficulties in determining word classes in North American languages are decidedly different from those presented by languages in other areas of the world. For North American languages, there are three recurring questions in the study of word classes and lexical categories in particular: 1) at what level a word is categorized (root, stem, or entire inflected word; § 5.1), 2) whether a given language distinguishes noun and verb (§ 5.2), and 3) whether a given language has an adjective category (which has already been discussed in § 4.1.3). The widespread (but not ubiquitous) presence of (poly)synthesis in North American languages (Mithun 2017b: 235; Rice [this volume]) means that a morphological distinction between nouns, verbs, and, when present, adjectives, is often quite clear. Words tend to have multiple affixes indicating their word class. In the following example from Nez Perce (Sahaptian), there are a tense marker and a perfective aspect marker – both categories typically associated with verbs. (27) Nez Perce (Sahaptian) (Deal 2010: 57) hi-pe-nees-ex-n-e 3.sbj-plSBJ-plobj-see-pfv-rem.pst ‘they saw us/you (pl.)/them’ Similarly, nouns in Nez Perce are marked for case (their role in the sentence) (Deal 2010: 32), a feature which is typically associated with nouns. Given these clear morphological distinctions, it may seem surprising that there could be any ambiguity regarding word classes in North American languages. Nonethe- 224 Daniel W. Hieber less, the potential for ambiguity in lexical categories can occur at the root, stem, or even whole word level, and words may be categorized differently at different levels (Jacobsen 1979: 100; Mithun 1999: 56; Haag 2006: 143; Lois et al. 2017: 102; Mithun 2017a: 155; Clemens 2019: 372). Section 5.1 shows how this ambiguity surfaces at these different levels in the languages of North America, and how categorization depends on the level of analysis (root, stem, or word). 9.5.1 Locus of categoriality In morphologically complex languages, words have an internal structure, so that some morphemes are more central to the core meaning of the word than others. The morpheme that provides the core sense of a word is called the root. For example, in Chitimacha the root ni- ‘water’ is used as the base for a number of different words, including nen- ‘go out of water’, nicwa- ‘approach water’, nitgext- ‘dump into water’, niduwa- ‘fall into water’, and others (Swadesh 1939b: 44). Each of the forms just listed are called stems, defined as the part of the word which serves as the basis for all its inflected forms. The stem nicwa-, for example, serves as the base for the inflected forms nicwi ‘s/he approaches water’, nicwicuki ‘I will approach water’, nicwipuyna ‘they used to approach water’, etc. Each of these inflectional possibilities is called an inflected word. Words may be categorized differently depending on whether one is analyzing the root, stem, or inflected word. In the West Greenlandic (Eskaleut) language, the lexical category of a word is typically obvious at all three levels. In example (28) the nominal root aamaruti- ‘coal’ takes various suffixes which create new stems, changing the word at different points from a noun to a verb and back again. Affixes which change the class of a word are called derivational affixes. At each step of derivation in West Greenlandic, the category of the word is clear. (28) West Greenlandic (Eskaleut) (Fortescue 1984: 315) aamaruti-ssar-siur-vi-tua-a-suq coal-fut-look.for-place-only-be-intr.ptcp n>n>v>n>n>v>n ‘which is the only place for getting coal’ In other North American languages, roots do not seem to be categorized for word class. In these languages, stems can be categorized but roots cannot. Haag (2006) argues that Cherokee is one such language. Cherokee has many words which are composed of multiple roots compounded together; however, it is impossible to determine what the category of the resulting compound will be based on the roots. The roots are simply put together in a way that makes sense for their meanings, and then a suffix is added that clarifies the lexical category (Haag 2006: 138). Example (29) shows two roots in Cherokee. (29) Word classes 225 Cherokee (Iroquoian) (Haag 2006: 137) a. -jaʔt- ‘attach asymmetrically at an indentation’ b. -húú- ‘stoma, opening’ (allomorph -ʔúú) Example (30) shows four compounds that can be formed using these roots. (30) Cherokee (Iroquoian) (Haag 2006: 137–138) a. tii-húú-jaʔt-î pl.obj-opening-attached-thing ‘lunchbox (with two handles)’ b. c. d. a-húú-jaʔt-î sg-opening-attached-thing ‘pitcher’ jii-ʔúú-jaʔt-v́vkâ 1sg-opening-attached-imm.pst ‘I just now attached a handle to something (e. g., a bucket)’ ‘I just now caught something by the mouth with a hook or attachment.’ tee-jíí-ʔúú-jaʔt-v́vkâ pl.obj-1sg-opening-attached-imm.pst ‘I just now attached something with more than one handle to something.’ Though in each case the stem is formed from the same combination of roots, in (a) and (b) the result is a noun, and in (c) and (d) the result is a verb. Haag takes this and other evidence to suggest that lexical categorization is not relevant to Cherokee roots, only stems. A similar situation occurs in Algonquian languages, in which lexical stems are formed of a combination of up to three components, called initial, medial, and final in the Algonquian literature (Goddard 1990; Macaulay and Salmons 2017; Lockwood 2017: 63–64; Oxford [this volume]). The initial is generally considered the root of the word, but it is the final component which determines the lexical category of the stem. Roots in Algonquian languages are therefore unspecified for lexical category. Examples (31) and (32) demonstrate how the same initial (shown in boldface) can be used to form either a noun or verb stem in Ojibwe and Menominee (both Algonquian languages). (31) Ojibwe (Algic) (Nichols 2020) a. miskozi miskw-izi red-3sg.ind ‘it is red’ b. miskobag miskw-bagw red-leaf ‘red leaf’ 226 (32) Daniel W. Hieber Menominee (Algic) (Monica Macaulay, p.c.) a. maehkuakom maehkw-akom red-skin/hide/covering/garment ‘red blanket’ b. maehkīhotaw maehkw-hot-a‑w red-paint‑theme-3sg ‘s/he paints it red’ In the Ojibwe example in (31), the same initial miskw- ‘red’ is used to form both a noun ‘red leaf’ and a verb ‘it is red’, while in the Menominee example in (32) the initial maehkw- ‘red’ is likewise used to form both the noun ‘red blanket’ and the verb ‘s/he paints it red’. Thus, in Algonquian it is only stems which are categorized for lexical category, not the root components. In some languages, even the stem can be neutral or ambiguous with respect to lexical category. Frachtenberg (1922: 318) claims that any stem in Coos (Coosan) may be used either nominally or verbally as appropriate. This is illustrated in (33). (33) Coos (Coosan) (Frachtenberg 1922: 329, 330, 328) a. poːʷkw-is slave-nmlz ‘slave’ b. ŋ-poːʷkw-its 1sg-enslave-tr ‘I enslaved him’ c. huːʷmis ‘woman’ d. n̥-huːʷmis-its 1sg-marry-tr ‘I marry (her)’ e. tsoːweˣtɬ ‘grease’ f. n̥-tsoːʷˣtɬ-ts 1sg-grease-tr ‘I greased it’ g. tɬʼkwiː ‘blanket’ h. tɬʼkwi-t cover-tr ‘she covered (them) with blankets’ Word classes 227 For the Tonkawa (isolate) language, Hoijer (1933: 23–24) famously claimed, “To apply the classificatory notion of “parts of speech” to Tonkawa would do extreme violence to the spirit of the language.” He provides the following example as evidence of his claim: (34) Tonkawa (isolate) (Hoijer 1946: 297) a. notox-ʔaː-la hoe-def-nom.sg ‘the hoe’ b. notx-o-ʔ hoe-decl-3.prs ‘he hoes it’ Andrade (1933: 179) likewise analyzes Quileute as a language where stems may be used as either noun or verb, assuming their function in context. In other languages, such as Hopi (Uto-Aztecan), most stems are specified for category, but a subset are ambivalent and may be used as either noun or verb (Whorf 1946: 163). Even fully-inflected wordforms with clear morphological marking of their class may nonetheless blur the distinction between noun and verb. In many North American languages, fully-inflected morphological verbs may be used as nominals without any special affixes or other modification, as the following examples illustrate. (35) Chitimacha (isolate) (Swadesh 1939b: 56; Swanton 1920: 17) a. dzampuyna dza-m-puy-na thrust-plact-hab-nf.pl ‘they usually thrust/spear (with it)’ ‘spear’ b. pamtuyna pa-m-tuy-na ford-plact-hab-nf.pl ‘they usually cross (it)’ ‘bridge’ (36) Cayuga (Iroquoian) (Mithun 2000: 200) a. o̜tekho̜nyáʔthaʔ ye-ate-khw-o̜ni-aʔt-haʔ indf.a-refl-meal-make-ins-ipfv ‘one makes a meal with it’ ‘restaurant’ b. kao̜tanéhkwi ka-ro̜t-a-nehkwi neut.a-log-ep-haul.ipfv ‘it hauls logs’ ‘horse’ 228 (37) Daniel W. Hieber Navajo (Na-Dene) (Young 1989: 316) a. tsinaaʼeeɬ tsi(n)-naaʼeeɬ wood-it.moves.about.floating ‘ship, boat’ b. chahaɬheeɬ it.is.dark ‘darkness’ For Cayuga (and other Iroquoian languages), some morphological verbs have been so fully lexicalized as nouns that they may no longer be used with their verbal meanings. The default meaning of kao̜tanéhkwi for Cayuga speakers is ‘horse’, not ‘it hauls logs’. Other verbs may retain both uses, while others lack any nominal meaning at all. Morphological verbs in Iroquoian therefore each sit on a cline from fully verbal to fully nominal, with many cases in between (Mithun 2000). In other languages, fully-inflected nouns and verbs can appear superficially similar, taking affixes of the exact same form, but nonetheless belong to clearly distinct word classes. In Central Alaskan Yup’ik, for example, the forms of noun inflections are a subset of the forms of verb inflections (Sadock 1999: 386). That is, noun endings all look like verb endings (but not vice versa), and even have similar functions, as the following examples illustrate: (38) Central Alaskan Yup’ik (Eskaleut) (Mithun 2017a: 161) a. qaya-q ‘kayak’ sg kaigtu-q ‘he/she/it is hungry’ sg b. qaya-k ‘two kayaks’ du kaigtu-k ‘they two are hungry’ du c. qaya-t ‘three or more kayaks’ pl kaigtu-t ‘they all are hungry’ pl Possessive suffixes on nouns likewise share their forms with transitive person suffixes on verbs: (39) Central Alaskan Yup’ik (Eskaleut) (Mithun 2017a: 161) a. angya-qa ‘my boat’ 1sg/3sg ner’a-qa ‘I am eating it’ 1sg/3sg b. angya-gka ‘my two boats’ 1sg/3du ner’a-gka ‘I am eating both of them’ 1sg/3du c. angya-nka ‘my boats’ 1sg/3pl ner’a-nka ‘I am eating them’ 1sg/3pl d. angya-a ‘his/her boat’ 3sg/3sg nera-a ‘he/she/it is eating it’ 3sg/3sg e. angya-k ‘his/her two boats’ 3sg/3du ner’a-k ‘he/she/it is eating both of them’ 3sg/3du Word classes f. angya-i nera-i ‘his/her boats’ ‘he/she/it is eating them’ 229 3sg/3pl 3sg/3pl However, any transitive verb whose object is not a third person has suffixes which never appear in nominal inflections, such as the examples in (40). (40) Central Alaskan Yup’ik (Eskaleut) (Sadock 1999: 386) takua-anga ‘s/he sees me’ 3sg/1sg takua-atigut ‘s/he sees us’ 3sg/1pl takua-akkit ‘I see you (sg.)’ 1sg/2sg takua-rma ‘you (sg.) see me’ 2sg/1sg The reason for these similarities is that many verbal inflections arose historically from nominalizations (Jacobson 1982; Woodbury 1985; Mithun 2008; Berge 2016). This is an example of a process known as insubordination, where subordinate clauses or noun phrases are reanalyzed as main clauses (Evans 2007; Mithun 2008; Evans and Watanabe 2016). Despite having a common origin as noun suffixes, verbal and nominal endings in Yup’ik are now nonetheless two distinct sets of affixes belonging to different word classes. Another case of superficial similarity between nouns and verbs comes from Menominee: (41) Menominee (Algic) (Monica Macaulay p.c.) a. askēhnen askēhnen-w be.fresh-3sg ‘it is fresh/raw’ b. askēhnen askēhnen-w be.fresh-nmlz ‘raw thing’ (Monica Macaulay p.c.) While the words in (41) have the same surface and underlying forms,5 this is merely a historical accident; the third person -w suffix and the nominalizing -w suffix are unrelated. Not only the category label, but the size of the category can vary depending on the level of analysis. Lindsey and Scancarelli (1985), for example, argue that Cherokee has a large, open class of adjectives when considering the level of the inflected word, but a small, closed class of adjectives when considering the level of the root. More drastically, Chitimacha lacks adjective stems entirely, but nonetheless has an open class of adjectives at the word level. All adjectives in Chitimacha are formed by adding an adjectivizing suffix to a verb stem, as shown in the examples in (42). 5 Note that the final /w/ in both examples is lost due to a synchronic process of final consonant cluster reduction. 230 (42) Daniel W. Hieber Chitimacha (isolate) (Swadesh 1939b) bixtigi ‘industrious’ affix. An independent word (also referred to as particle) may become cliticized and affixed, after which the negative is renewed. That presumably happened in (4), where te- represents the older negative that is now attached to the verb and tąʔą is the renewal. Grammaticalization followed by renewal can be seen as a cycle and, in the case of negatives, it is called the Negative Cycle (as in Jespersen 1917). Mithun (2016; 2018b) provides examples of both grammaticalization and renewal in a number of NA languages and I too will use this cycle to describe the variation among closely related languages. Another example of variation due to grammaticalization and renewal is found in Yuman. Taking a snip from the map in Figure 1, we see Jamul Tiipay listed as having a Negatives 271 particle (dark blue), Hualapai and Diegueño with a negative auxiliary (light blue), and Maricopa as having double negation (yellow). Hualapai Maricopa Diegueño/Kumeyaay Jamul Tiipay Fig. 2: Variation in Yuman negatives This variety shows the result of a Yuman Negative Cycle. The Jamul Tiipay from is an original negative verb with a switch reference marker, i. e., maaw; the Hualapai and Diegueño/Kumeyaay negatives are still inflected verbs; and the Maricopa original negative verb aw is now a suffix –ma with a clause marker (w)aly- as its renewal. This reconstruction is from Munro (1974a) and will be discussed more in section 4. 11.3 Negatives and word order In this section, I discuss the position of the negative in the sentence as early (in the clause) or as late. For the most part, the position of the negative depends on the position of the verb. Negation typically precedes the verb (Dryer 2013b) and therefore will be sentence-initial in verb-initial languages and more towards the end in verb-final languages. The languages of North America are no exception to this pattern. All 11 VSO languages listed in WALS are negative-initial, namely Kalispel, Sahaptin, Quileute, Makah, Halkomelem, Musqueam, Squamish, Kyuquot, Heiltsuk, Bella Coola, and Nisgha. An example from one of these languages, Squamish, is shown in (9), and an example from Nuuchahnulth, a verb-initial language not included in the WALS sample, is shown in (10). (9) Squamish (Salish) (Kuipers 1967: 194, cited after Kroeber 1999: 156) háw q-ʔan-c’ic’áp’ ti-scíʔs neg irr-1S-work art-today ‘I don’t work today.’ 272 (10) Elly van Gelderen Nuuchahnulth (Wakashan) (Nakayama 2001: 119) wikiimits wik -i:p -it -s neg obtaining pst 1S ‘I didn’t catch anything’ The sole VOS language listed in WALS, Chinook, is negative-initial as well. This pattern of verb-initial with the negative in pre-verbal position is also known from Austronesian and Celtic languages. The negative appears in initial position with other word orders as well. WALS (map 144P) gives a few examples, e. g., Northern Paiute, Pima Bajo, Mono, Wichita, Cupeño, Kiowa, and Comanche, as (11) shows, where the verb is in final position. (11) Comanche (Uto-Aztecan) (Charney 1993: 220, gloss simplified) ke nɨɨ toHtɨn-kahtu miʔaRƗ neg 1S [name]-toward go ‘I didn’t go to Lawton.’ Languages that place verbs in final position often have a later negative, either preceding or following the verb. The 24 North American languages in the WALS atlas (Dryer 2013b) that have the negation placed in final position are all OV. This pattern of verb-last and late negation is also known from other languages, e. g., from those in the Indo-European family. The North American OV-languages where negation is last include Hare (6), Chipewyan (7), Osage, Lakhota, and Diegueño (12); on the other hand, Hopi, as in (13), Tlingit, and Cahuilla have a negative that precedes the final verb. (12) Diegueño/Kumeyaay (Yuman) (Munro 1976a, cited after Langdon 1970) ʔ-a:m-x ʔ-ma:w 1-go-irr 1-neg ‘I don’t/didn’t go.’ (13) Hopi (Uto-Aztecan) (Kalectaca 1978: 41) I’ nagahu qa wehe. this medicine neg spill ‘This medicine didn’t spill.’ The negative in (12) develops from a negative verb as Munro (1974a: 37) argues. Note the inflection which makes it an auxiliary verb (light blue in Figure 2), and not a particle (dark blue). In some languages, as shown in Figure 1, there are two negative markers. In Athabascan, the sentence initial negative (e. g., doo in (2)) is the more recent renewal and the final negative (e. g., da in (2)) is the older one. Similar renewals can be found in Hare (6) and Apache (14), where doo is optional. Negatives (14) 273 (Western) Apache (Bray and The White Mountain Apache Tribe 1998: 109) (doo) nchad da neg 2S.cry neg ‘Don’t you cry.’ This development fits the Negative Cycle, mentioned around (6). Again, Mithun (2016; 2018b) presents other such cases. Concluding, we can say that the negative is typically close to the verb: in initial position when the verb is initial and final when the verb is final. Renewed negation results in the double marking of Figure 1/Table1. 11.4 Negative verbs and auxiliaries Payne (1985: 222) argues that the origin of a negative marker is often the negative of the verb ‘to be’ and Givón (1978) mentions a negative verb approximating ‘to lack’ as a source. Mithun (2018b: 315–320) provides examples of negatives deriving from verbs meaning ‘to die/not exist’ and I will add others. NA languages show many negatives that are still verbs, e. g., Diegueño (12), or that occur in structures that consist of two clauses, e. g., Salish and Algonquian, discussed below. The latter is evidence of the verbal nature of the negative. In this section, I first sketch the reconstruction in Yuman mentioned before and in Uto-Aztecan and Tanoan that show the negative is verbal in origin. I then look at Salish and Algonquian for evidence of more than one clause, with the negative originating in a cleft construction. I finish with Athabascan, Muskogean and geographically related isolates, and EskimoAleut. 11.4.1 Yuman, Uto-Aztecan, and Tanoan Compare the data from Maricopa in (1) to those in Diegueño in (12) and Mohave in (15). This shows the negative affixes -ma and -mo in Maricopa and Mohave, respectively, but an independent verb in Diegueño. (15) Mohave (Yuman) (Munro 1976a: 106 with gloss adapted) ʔnyeč ʔ-iyem-mo-t-m I 1-go-neg-emph-real ‘I didn’t go’ Munro (1976a) reconstructs (16), where aw is a verb that has a clause to its left. The –m in the embedded clause marks that the subject is different from that of the main clause (also known as a switch reference marker). 274 (16) Elly van Gelderen (S O V]-m aw neg-Verb ‘It is not that (I went)’ In Diegueño (and Tolapaya Yavapai), the negative main verb is still inflected but, in Maricopa and Mohave, it is a suffix. Munro (1976: 65–8) shows that Mohave has the complex verbal negative suffix -m-ot- because, when the realis marker p-č is used in (17), the negative is split. The -m is therefore an earlier Different Subject marker. (17) Mohave (Yuman) (Munro 1976a: 106 with gloss adapted) ʔnyeč ʔ-iyem-m-p-o-t-č I 1-go-neg-real-neg-emph-real ‘I didn’t go’ More evidence that what remains in –o/-a:w/-a is originally a verb *–aw, with an embedded clause to its left is that the realis marker of the entire sentence, which is lexically determined by the main verb, is determined by the negative, i. e., it has to be –m, as in (15), or –p-č, as in (17); it cannot be –k, even if the lower clause requires that. Uto-Aztecan uses particles for which Langacker (1977: 33) suggests two reconstructed forms, *ka and *ka-y. This is clear from comparing negation in some of the languages, e. g., either káa in Yaqui (Dedrick and Casad 19992), ki in River Warihío (Armendáriz 2007: 1063), qa in Hopi (12) (Kalectaca 1978: 39), and ke in Comanche (Charney 1993: 220) or k…y, e. g., qay in Luiseño (Hyde 1971: 13), kai in Shoshone (Crum and Dayley 1993: 11), Cupeño qay (Jacobs 1972: 71), Cahuilla kilye (Hioki 1971: 24), and Northern Paiute kai (Nichols 1974: 228). Miller (1967: 49) also reconstructs *ka and *kai. According to Langacker, the latter is a negative and the verb ‘to be’, forming a negative cleft, as in ‘It is not the case that …’.4 Kroskrity (1984) shows that, in Arizona Tewa, a Tanoan language, there are two negative markers, the prefix we- and the suffix –dí, as in (18). The suffix –dí is also used for subordinate clauses in (19). This dual use points to a bi-clausal origin of the negative where we in (18) was the main negative and mun the subordinate verb, perhaps cleftlike, as in ‘It was not the case that the man saw the woman’. The original subordinator was then reanalyzed as part of the negative complex. 2 Although Yaqui káa can have enclitic subject pronouns when it appears clause-initially. 3 River Warihío procliticizes on the lexical verb and cliticizes to ‘te in existential and copular constructions. 4 Langacker (1977: 33) cites an unpublished paper by Munro (1974b) with another possible bi-clausal reconstruction where *-y is a nominalizer with an optional case suffix *ta for forms such as Tubatulabal haainda. Negatives (18) Arizona Tewa (Kiowa-Tanoan) (Kroskrity 1984: 95) sen kwiyó we-mán-mun-dí man woman neg-3/3-see-neg ‘The man didn’t see the woman.’ (19) Arizona Tewa (Kiowa-Tanoan) (Kroskrity 1984: 95) he’i sen na-mɛn- dí ‘o-yohk’ó that man 3-go-sub 1-be.asleep ‘When that man went, I was asleep.’ 275 In fact, all Tanoan languages (except Jemez) have a prefix and suffix where the latter is identical with a subordinator. The data for Picuris Tiwa are given in (20). (20) Picuris Tiwa (Kiowa-Tanoan) (Zaharlick 1977: 238, cited after Kroskrity 1984: 96) ‘u-wa-wəle-mᶒ 3P-neg-go-sub.pres ‘They are not going.’ 11.4.2 Salish and Algonquian Salishan languages have many constructions where a negative is followed by a subordinate structure, which is indicative of the negative’s earlier verbhood (see Kroeber 1999: 99; 108; 138; 155–9; 204; 207; 213; 230–2). For instance, in Kalispel (21), the verb following the negative has a possessive inflection and nominalizer, which means the verb looks like an earlier embedded structure. The i- possessive and s- nominalizer only occur with a negative particle, which currently no longer looks like a verb. (21) Kalispel (Salish) (Vogt 1940: 75, cited after Kroeber 1999: 108) tá i-s-ən-té neg 1s.pos-nom-in-think ‘I don’t think so,’ Davis (2005: 50) argues that Proto-Salish had two patterns that “probably involved a biclausal structure” with a negative verb selecting either a nominalized complement, as in (21), or a subordinate clausal one, marked by CJ (conjunctive) in (22). (22) Comox (Salish) (Kroeber 1999: 156) xwuxwaʔ=č huǰ-an neg=1s finish-1s.cj ‘I am not finished yet.’ 276 Elly van Gelderen Algonquian languages are well-known for having some negators that choose the conjunct verb form (typical for embedded clauses) rather than the independent one. MacKenzie (1992) summarizes the negatives for the Cree-Montagnais-Naskapi continuum. A variant of nama is used with independent verbs and eka with conjunct verbs in Plains Cree, Moose Cree, and Swampy Cree. Montagnais has the innovative apu, typically with a conjunct verb, instead of nama. Variation has been reported as to independent and conjunct forms with the negatives. 11.4.3 Athabascan, Eskimo-Aleut, Muskogean and neighboring isolates As mentioned in connection with Athabascan in section 3, a number of scholars reconstruct the negative in (6) and (7) (either particle or suffix) as a verb. So, yíle in (6) and hilɛ in (7) are seen as erstwhile negative auxiliaries or copula verbs. For reasons of space, I will not go into this more here, but see Leer (2000: 123), Rice (1989: 1108, n. 1), and van Gelderen (2011: 325–6). This verbal origin is also the reason that they vary for tense, mood, and aspect (TMA), as will be shown in section 5. The renewal mentioned around (14) is not of verbal origin. According to Fortescue (1999), Eskimo-Aleut also uses a negative enclitic based on a copular stem, the u- in –ulax in (23). Other tenses/aspects use different markers. (23) Aleut (Eskimo-Aleut) (Fortescue 1999, fn 5) awa-za-ĝ-ulax work-hab-3s.neg ‘He never works.’ In Alabama and Koasati, Muskogean languages, the negative is complex because of the relationship between agreement and the negative form. This connection is to be expected if the negative is originally a verb or auxiliary. Haas (1977) argues for such a connection on the basis of the negative verbs present in Chitimacha, an isolate in the same ‘Gulf’ region. Montler and Hardy (1991: 16) do not agree and argue that the origin as an auxiliary is not so much “by weight of evidence” but they do not “offer a better reconstruction”. From a grammaticalization point of view, Haas’ reconstruction makes sense. Chitimacha has a clearly inflected negative auxiliary, as in (24), as well as an uninflected variant in (25). The lexical verb kaakwi in (24) is clearly marked as non-finite (through GER(und)) whereas the lexical verb in (25) is inflected for person and number and the negative is a particle. (24) Chitimachi (isolate) (Hieber 2015: 3, unpublished material from Swadesh) we qaxinjadi cuntk hi waytm kaakwi gay-ik. old man about more know.ger neg-1sg ‘I do not know any more about the old man.’ Negatives (25) 277 Chitimachi (isolate) (Hieber 2015: 3, unpublished material from Swadesh) waqax qam haanaa ne kaakw-iki gan. others what they.happen even know-1sg neg ‘I do not know what happened to the others.’ In this section, several North American families/languages have been shown to have a negation pattern focused on the negative verb. Although the Uralic languages, some Nilo-Saharan languages, and a few others also show auxiliary negation on the WALS map, this pattern is not that wide-spread. Austronesian negation is ambiguous between particle and verb as well. Related to the verbal status of the negative is the specialization of the negative for TMA which I will turn to next. 11.5 Negatives specialized for tense, mood, and aspect (TMA) Languages of North America specialize as to whether their negative modifies a particular tense or aspect or mood. In main and subordinate clauses, the negative may be different because the mood of the main clause is independent and that of the subordinate is usually dependent. Cross-linguistically, a very frequent special negative is the prohibitive mood (a negative imperative), and I will start with the prohibitive in Zuni. I then examine Athabascan and Central Pomo negatives that are marked for aspect. Languages can have the regular negative in the imperative or a special one. Van der Auwera and Lejeune’s (2013) map shows that NA languages are varied. Maricopa, Diegueño, Pomo, Klamath, Lakhota, Wichita and others use the same negative for declaratives and imperatives but Zuni, Acoma, Oneida, Northern Paiute, Mandan, and others use a different form. For instance, Zuni, an isolate, distinguishes between imperative/permissive verbs on the one hand and verbs inflected for person on the other (Newman 1965: 73), which are declarative. The former have the negative particle ʔełł, as in (26), whereas the latter have kwaʔ, as in (27), both in clause-initial position. In addition, there is a form of naʔma that attaches to the verb in both (26) and (27). (26) Zuni (isolate) (Newman 1965: 74) ʔełł tešlan-naʔma-ø neg afraid-neg-imp ‘Don’t be afraid.’ (27) Zuni (isolate) (Newman 1965: 74) kwaʔ tešlan-nam-ka neg afraid-neg-pst ‘He wasn’t afraid.’ 278 Elly van Gelderen Bunzel (1935: 513–5) adds more on the variation of naʔma when used with present tense and certain aspects. It is optional when the negative subjunctive is marked by a suffix –cukwa in addition to the initial kwaʔ, as in (28). (28) Zuni (isolate) (Bunzel 1935: 514; Bunzel 1933: 57, line 97) k waʔ ał-cukwa neg sleep-neg.subj ‘She will not sleep (if the baby cries).’ As for the connection between TMA and the negative, we have already seen some evidence for that link in the previous section. For example, Picuris Tiwa (20), Aleut, and Muskogean make such connections. Here I discuss Athabascan and Pomoan. Leer (2000: 103) finds a connection between negation and verbal aspect in Athabascan. In many languages, e. g., Ahtna (29), the negative replaces conjugation markers, such as the ø-imperfective, with a negative s- prefix, as shown in the non-negative imperfective (30). (29) Ahtna (Athabascan) (Kari 1992: 123) ‘ele’ k’e-s-t’aaz-e neg indef-neg-cut-neg ‘He isn’t cutting something.’ (30) Ahtna (Athabascan) (Kari 1992: 123) k’e-t’aas indef-cut.impf ‘He’s cutting it.’ In addition to the –s, there is a negative suffix –e, probably a remnant of a copula, and a renewed form ‘ele’. Mithun (1998) presents data from Central Pomo (Pomoan) where two negatives (generally5) specialize for imperfective or perfective aspect. The čhów negative in (31) marks the perfective whereas the t̪ hín in (32) is used for everything else. (31) Central Pomo (Pomoan) (Mithun 1998: 78) ʔaˑ čá-ˑn-ka-w čhów. 1.agt run-impf-caus-prf neg ‘I didn’t drive.’ (32) Central Pomo (Pomoan) (Mithun 1998: 78) t̪ áwhal yhé-t̪ -aʔ t̪ hín =ka. work do-multiple-impf.pl neg=inferential ‘They must not be working.’ 5 As Mithun (1998) points out, there is fluidity depending on pragmatic choices. Negatives 279 This section has provided examples of languages where the negative depends on the mood or aspect of the clause. This is to be expected if the negative derives from a verb or auxiliary, which are marked for TMA. Next, I turn to negative pronouns which can turn a sentence into a negative as well. 11.6 Negative indefinite pronouns In this section, I show that the negative indefinite in NA languages is often based on an indefinite pronoun and may occur with sentential negation. The WALS map (Haspelmath 2013) shows that all NA languages listed use an additional sentential negation but that is hard to argue in languages like Navajo, as I will show. Haspelmath’s (1997) overview of how indefinite pronouns are marked in a sample of 140 languages includes eight NA languages. His NA data show that all, except Slave, have interrogative-based positive indefinites. For instance, in O’odham (Uto-Aztecan), heḍai means ‘who’ but, when combined with the negative pi, it means ‘nobody.’ The same is true in Hopi (33). This word has to occur at the beginning of the clause, unlike the sentential negative qa in (34), which occurs before the verb. Yaqui and Tümpisa Shoshone follow a similar pattern, according to Haspelmath. (33) Hopi (Uto-Aztecan) (Malotki 1979: 123, Haspelmath 1997: 251) Qa-háqaqw kwii-kwitsi. neg-where rdp-smoke ‘Smoke comes out nowhere.’ (34) Hopi (Uto-Aztecan) (Kalectaca 1978: 41) I’ nagahu qa wehe. this medicine neg spill ‘This medicine didn’t spill’. In Mohave (35), there is a special kind of negative that occurs with indefinites, so again no special negative indefinites appear. (35) Mohave (Yuman) (Munro 1976b: 71) kuč ʔ-ičo:-poʔa:və something 1-make-neg ‘I am not making anything.’ The indefinites are really positive on their own. That is true in Navajo as well where the negative indefinite háída also translates as ‘someone’ so the sentential negation is responsible for the negative indefinite meaning, as in (36). 280 (36) Elly van Gelderen Navajo (Athabascan) (Hale and Platero n.d.) doo háída níyáa da neg someone arrived neg ‘Nobody has arrived.’ Indefinites go through cycles, from indefinites with sentential negation, as in (35), to negative indefinites, as in (33), to adding another negative (not shown). Cf. Willis (2010) for a cycle involving these pronouns. In short, cross-linguistically in the NA languages, there are no special negative indefinites. The data show that a negative is either affixed to an indefinite pronoun, as in Hopi, or that, as in Mohave, the indefinite occurs in a negative clause. 11.7 Conclusion All languages have negative markers, and North American languages are no exception. The negatives can be affixes or particles and are often placed close to the verb. In many of the languages we have discussed, negatives show evidence of (still) being verbs or auxiliaries or having (once) been verbal. The negative form also differs depending on the tense, mood, and aspect (TMA) of the clause. Finally, negative indefinites are often formed by combining the negative with an indefinite or interrogative pronoun. References Armendáriz, Rolando. 2007. A Grammar of River Warihío. Munich: Lincom. Bray, Dorothy & the White Mountain Apache Tribe. 1998. Western Apache-English Dictionary. Tempe: Bilingual press. Bunzel, Ruth. 1933. Zuni Texts. New York: G.E Stechert. Bunzel, Ruth. 1935. Zuni. New York: G.E Stechert. Charney, Jean Ormsbee. 1993. A Grammar of Comanche. Lincoln: University of Nebraska Press. Cook, Eung-Do. 2004. A Grammar of Dëne Sųłiné (Chipewyan). Winnipeg: Algonquian and Iroquoian Linguistics. Crum, Beverly & Jon Philip Dayley. 1993. Western Shoshoni Grammar. Boise State University. Dahl, Östen. 1979. Typology of sentence negation. Linguistics 17. 79–106. Davis, Henry. 2005. On the syntax and semantics of negation in Salish. International Journal of American Linguistics 71(1). 1–55. Dedrick, John & Eugene Casad. 1999. Sonora Yaqui Language Structures. Tucson: University of Arizona Press. Dryer, Matthew. 2013a. Order of Negative Morpheme and Verb. In Matthew Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online, Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://wals.info/chapter/143, Accessed on 2018-05-28.) Dryer, Matthew. 2013b. Position of Negative Morpheme with Respect to Subject, Object, and Verb. In Matthew Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Negatives 281 Max Planck Institute for Evolutionary Anthropology. (Available online at http://wals.info/chapter/144, Accessed on 2018-05-28.) Dryer, Matthew. 2013c. Negative Morphemes. In Matthew Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://wals.info/chapter/112, Accessed on 2018-06-13.) Fortescue, Michael. 1999. The rise and fall of polysynthesis in the Eskimo-Aleut family. In Nicholas Evans & Hans-Jürgen Sasse (eds.), Problems of polysynthesis, 257–75. Berlin: Akademie Verlag. Gelderen, Elly van. 2008a. The Negative Cycle. Linguistic Typology 12(2). 195–243. Gelderen, Elly van. 2008b. Cycles of Negation in Athabaskan. Working Papers in Athabaskan Languages 7. 49–64. Gelderen, Elly van. 2011. The Linguistic Cycle. Oxford: Oxford University Press. Givón, Talmy. 1978. Negation in Language. In Peter Cole (ed.), Syntax & Semantics 9, 69–112. New York: Academic Press. Gordon, Lynn. 1986. Maricopa Morphology and Syntax. Berkeley: University of California Press. Haas, Mary. 1977. From auxiliary verb phrase to inflectional suffix. In Charles Li (ed)., Mechanisms of Syntactic Change, 525–38. Austin: University of Texas Press. Hale, Ken & Paul Platero. n.d. Negative Polarity Expressions in Navajo. Unpublished manuscript. http:// lingphil.mit.edu/papers/hale/papers/hale014.pdf Haspelmath, Martin. 1997. Indefinite Pronouns. Oxford: Oxford University Press. Haspelmath, Martin. 2013. Negative Indefinite Pronouns and Predicate Negation. In Matthew Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online, Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://wals.info/chapter/115, Accessed on 2018-08-25.) Hieber, Daniel. 2015. To be not or to not be. Unpublished manuscript. (https://files.danielhieber.com/ publications/Chitimacha-negation-behavioral-profile/pdf.pdf) Hioki, Kojiro. 1971. Zur Beschreibung des Systems der Klitika im Cahuilla. Unpublished manuscript. Hyde, Villiana. 1971. Introduction to the Luiseno Language. Banning: Malki Museum Press. Jacobs, Roderick. 1972. Syntactic Change: a Cupan Case Study. UCSD Dissertation. Jespersen, Otto. 1917 [1966]. Negation in English and other Languages. Copenhagen: A.F. Høst. Kalectaca, Milo. 1978. Lessons in Hopi. Tucson: University of Arizona Press. Kari, James. 1992. Some Concepts in Ahtna Athabaskan word formation. In Mark Aronoff (ed.), Morphology Now, 107–131. Albany: SUNY Press. Kopris, Craig Alexander. 2001. A Grammar and Dictionary of Wyandot. PhD SUNY Buffalo. Kroeber, Paul. 1999. The Salish Language Family. Lincoln: The University of Nebraska Press. Kroskrity, Paul. 1984. Negation and subordination in Arizona Tewa. International Journal of American Linguistics 50. 94–104. Kuipers, Aert. 1967. The Squamish Language. The Hague: Mouton. Langacker, Ronald. 1977. Studies in Uto-Aztecan Grammar, I. Arlington: SIL. Langdon, Margaret. 1970. A Grammar of Diegueño: The Mesa Grande Dialect. Berkeley: University of California. Leer, Jeff. 2000. The Negative/Irrealis category in Athabaskan-Eyak-Tlingit. In Ted Fernald & Paul Platero (eds.), The Athabaskan Languages, 101–138. Oxford: Oxford University Press. Li, Fang Kuei 1967. Chipeywan. In C. Osgood & H. Hoijer (Eds.), Linguistic Structures of Native America. Viking Fund Publications 6: 398-423. [reprinted from 1946] MacKenzie, Marguerite 1992. Negative Markers in East Cree and Montagnais. Papers of the 23rd Algonquian Conference, Carleton Ottawa. 274–284. Malotki, Ekkehart. 1979. Hopi-Raum: Eine sprachwissenschaftliche Analyse der Raumvorstellungen in der Hopi-Sprache. Tübingen: Gunter Narr Verlag. Miestamo, Matti. 2003. Clausal Negation: A Typological Study. Helsinki: University of Helsinki Doctoral Dissertation. 282 Elly van Gelderen Miller, Amy. 2001. Grammar of Jamul Tiipay. Berlin: Mouton de Gruyter. Miller, Wick. 1967. Uto-Aztecan Cognate Sets. Berkeley: University of California. Mithun, Marianne. 1995. Affixation and Morphological Longevity. In Geert Booij & Jaap van Marle (eds.), Yearbook of Morphology 1994, 73–97. Dordrecht: Springer. Mithun, Marianne. 1998. Fluid Aspects of Negation in Central Pomo. In Leanne Hinton & Pamela Munro (eds.), Studies in American Indian Languages, 77–86. Berkeley: University of California Press. Mithun, Marianne. 2016. What cycles when and why? In Elly van Gelderen (ed.), Cyclical Change Continued, 19–45. Amsterdam: John Benjamins. Mithun, Marianne. 2018a. Negation against a polysynthetic backdrop: Mohawk. Presentation at Syntax of the World’s Languages 8, Paris, 3–5 September. Mithun, Marianne. 2018b. Shaping typology through grammaticalization: North America. In Heiko Narrog & Bernd Heine (eds.), Grammaticalization from a typological perspective, 309–336. Oxford: Oxford University Press. Montler, Timothy & Heather Hardy. 1991. The phonology of negation in Alabama. IJAL 57(1). 1–23. Munro, Pamela. 1974a. Topics in Mohave Syntax. La Jolla, CA: UCSD Dissertation. Munro, Pamela. 1974b. On the Morphology of Shoshonean Negatives. Unpublished manuscript. Munro, Pamela 1976a. Subject Copying, Auxiliarization, and Predicate Raising: The Mojave Evidence. International Journal of American Linguistics 42.2: 99–112. Munro, Pamela. 1976b. Mojave Syntax. New York: Garland. Nakayama, Toshihide. 2001. Nuuchanulth (Nootka) Morphosyntax. Berkeley: University of California. Newman, Stanley. 1965. Zuni Grammar. Albuquerque: UNM Press. Nichols, Michael. 1974. Northern Paiute Historical Grammar. University of California Berkeley Dissertation. Payne, Thomas. 1985. Negation. In Timothy Shopen (ed.), Language typology and syntactic description, vol. I, 197–242. Cambridge: Cambridge University Press. Rice, Keren. 1989. A Grammar of Slave. Berlin: Mouton de Gruyter. van der Auwera, Johan. 2010. On the diachrony of negation. In Larry Horn (ed.), The Expresssion of negation, 73–101. Berlin: De Gruyter Mouton. van der Auwera, Johan & Ludo Lejeune (with Valentin Goussev). 2013. The Prohibitive. In Matthew Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://wals.info/chapter/71, Accessed on 2018-10-09.) Vogt, Hans. 1940. The Kalispel Language. Oslo: Det Norske Videnskaps Akademie. Willis, David. 2010. Negative polarity and the Quantifier Cycle. http://rhssl1.uni-regensburg.de/SlavKo/ conferences/gces/abstracts/Willis.pdf Young, Robert W. & William Morgan. 1987. The Navajo Language: A Grammar and Colloquial Dictionary. Albuquerque: University of New Mexico Press. Zaharlick, Ann Marie. 1977. Picuris syntax. Washington D.C.: American University Dissertation. Olga Lovick 12 Questions and requests in North American languages Abstract: Questions and requests are speech acts that call upon the hearer to do something. Questions typically ask for a verbal response, while requests usually elicit a non-verbal response. Both are important in language teaching and often among the first expressions that a learner acquires. Questions fall into two groups: polar questions, which can be answered with “yes” or “no”; and content questions, which require a more elaborate answer. Depending on the language, there are different types of requests, depending on factors such as identity of the addressee, politeness, etc. This paper provides an overview over the formation of different types of questions and requests in Indigenous languages of North America. Common strategies include intonation, (verbal) inflection or special markers. The paper closes with a brief discussion of how questions and requests can be used instead of one another. 12.1 Introduction Questions and requests are related types of speech acts that call upon the hearer to do something (Searle 1976: 11). The first type, questions, calls on the hearer to provide a verbal response of some sort—to answer the question. The second type, requests1, calls on the hearer to perform (or to refrain from performing) a certain activity. The difference between these speech acts, then, lies in the type of action that the speaker wants the hearer to perform, as illustrated in (1) and (2), where a question and a request are paired with possible responses. (1) Question A: Do you like chocolate? B: Yes./No./Only milk chocolate./… (2) Request A: Pass the salt, please. B: [Passes the salt to A] 1 This type of speech act is often called “command” or “imperative” in the literature. I use “request” instead of “command” due to the hierarchical connotations of the latter term. The term “imperative” is used here for the linguistic (default) form used to express a request. https://doi.org/10.1515/9783110600926-012 284 Olga Lovick Of course, it is possible to answer questions non-verbally (e. g., by nodding or shaking one’s head) and requests verbally (e. g., by saying “Sorry, I don’t see the salt anywhere.”), but (1) and (2) are prototypical examples. Questions and requests are very common in everyday interactions and thus important for learners working towards fluency. They also are used prominently in several approaches often used in Indigenous language teaching: Hinton, Florey, Gessner and Manatowa-Bailey (2018: 128) point to the importance of questions and requests in the Master-Apprentice approach; and the Total Physical Response approach by Asher (1969) has the explicitly stated goal of improving language students’ comprehension of spoken requests. In this chapter, I review basic types of questions (§ 2) and requests (§ 3) and describe how they can be formed in different Indigenous languages. In § 4, I look at how questions can be used to form requests and vice versa. 12.2 Questions Questions call on the hearer to provide a verbal response; they are a request for information. Depending on the response being sought, two types of questions can be distinguished. Polar questions are used when a speaker seeks to confirm or disconfirm a certain state of affairs, as in (1) above. Because they typically allow a limited set of answers such as “yes” and “no”, they are also known as yes/no questions. Content questions are used when the speaker is attempting to obtain a missing piece of information. If I notice that cookies have disappeared from the cupboard, I might ask a question such as (3). (3) Who ate the cookies? Content questions always contain a question word, such as who, what, when, where, why, or how. In English, this type of question is also known as wh-question. 12.2.1 Polar questions Polar questions can be formed in several different ways. In a few languages, they differ from a declarative by intonation only. This is illustrated using a sentence pair from Upper Tanana (Dene). The declarative clause in (4a) has gently declining pitch across the utterance, while the question in (4b) has a sharp rise on the last syllable. In both clauses, the word order is the same and no special question marker is used in (4b).2 2 The following abbreviations are used in this paper: 1/2–first person inclusive, AG–agent, ALL–allative, ANIM–animate, ASRT–assertion, CONCESS–concessive, CONT–continuative, DAT–dative, DEL–de- (4) Questions and requests in North American languages 285 Upper Tanana (Dene) (Roy H. David, Sr. p.c. May 28, 2013, Avis Sam p.c. August 12, 2009) a. Ay chih dindeey heldeeł. and also people 3pl.sbj:eat:ipfv ‘And they eat people.’ b. Shoh aldeeł? bear 2pl.sbj:eat:ipfv ‘Do you (addressing plural) eat bear?’ ↘ ↗ This strategy is used in almost a fifth of all languages worldwide but seems to be quite rare in North America according to Dryer (2013a). Much more common, both typologically and in North America, is the use of question particles, e. g., in Nakota (Siouan). No special intonation is used in this language, even though the question particle he is often omitted, and the word order is identical to a declarative (5). (5) Nakota (Siouan) (Cumberland 2005: 377) Tʰaspą yúta he? apple 3ag:eat q ‘Did s/he eat an apple?’ In Lakhota (Siouan), the form of the question particle depends on the gender of the speaker: male speakers use hųwo, female ones hųwé (Trechter 1995: 48–49).3 Some languages have a special interrogative mood paradigm. Lupardus (1982: 194–195) shows that in Alabama (Muskogean), the interrogative mood causes nasalization of the stem-final vowel. This is accompanied by a prosodic pattern “consist[ing] of a drop followed by a slight rise, all on the final syllable” (p. 195), as illustrated in (6). While only about a quarter of the world’s languages use special interrogative inflection, Dryer (2013a) suggests that this is not at all uncommon in the languages of North America. layed, DEM–demonstrative, DEON–deontic, DES–desire, DIR–direction, DUR–durative, EMPH–emphatic, FOC–focus, FUT–future, GEN–genitive, IMM–immediate, IMP–imperative, INAM–inanimate, INCEP– inceptive, INST–instrumental, INTENS–intensifier, IPFV–imperfective, IRR–irrealis, ITER–iterative, MOD–modal, NEG–negation, NMLZR–nominalizer, NONREF–non-referential, OBJ–object, OBL–oblique, OPT–optative, PAST–past, PAT–patient, PL—plural, POL–polite, POSS–possessor, POT–potential, PROH– prohibitive, PROX–proximate, PUNC–punctual, Q–question, QUOT–quotative, REF–referential, SBJ– subject, SEQ–sequential, SG–singular, SPEC–specific, SR–switch reference, STV–stative, TR–transitive, TRNS–translocative, UNR–unrealized. Most grammatical glosses are identical to those in the source; only a few were standardized. 3 Trechter (1995:48) also notes that the female question particle is largely obsolete. 286 (6) Olga Lovick Alabama (Muskogean) (Lupardus 1982: 195) a. ispalõ is-(i)pa-lo-~ 2sg.ag-eat-fut-q ‘Will you eat it?’ b. ákpolo ák-(i)pa-o-lo neg.1sg.ag-eat-neg-fut ‘I will not eat it’ The expected (or preferred) answer may have an influence on the form of a question. Rice (1989) notes for example that questions formed with the question particle in Slave (Dene) are neutral regarding the possible answers (p. 1123 f.), while intonation questions in that language are used when the speaker expects an affirmative answer (p. 1128). 12.2.2 Content questions All content questions contain an interrogative word. As alread noted by Sadock and Zwicky (1985: 184), languages differ enormously in the number of interrogative words and the distinctions they make. Cook (2004: 99) lists seven interrogative stems in Dëne Sųłiné, translating to ‘who, what, when, where, why, how’, and ‘how many’. Cumberland (2005: 381) lists twelve interrogative words (plus additional derived forms) in Nakota; this language distinguishes, for example, between realized versus unrealized when asking about the time of an event (7). This distinction is also attested in some Dene languages (see, e. g., Lovick 2020: 280 for Upper Tanana). (7) Nakota (Cumberland 2005: 383) a. Tóhą hí he? when:past 3ag:arrive q ‘When did he arrive?’ b. Tohą́ n hí-pi-kta when:fut 3ag:arrive-pl-pot ‘When will they arrive?’ he? q More unusual is the interrogative inventory of Ute (Uto-Aztecan). This language distinguishes twelve interrogative pronouns depending on subjecthood, animacy, number, and referentiality (Givón 2011: 316). (8) illustrates the number contrast for the animate non-referential subject and object pronouns ‘what kind of (animate)’. As Givón (2001: 316) points out, the expected answer to this type of question identifies the “type of individual, not its specific reference”. (8) Questions and requests in North American languages 287 Ute (Uto-Aztecan) (Givón 2001: 316) a. ’u-vwa-tʉ ’íni ’uni’ni(-y)? there-at-dir q:nonref:anim:subj do/be-imm ‘What kind (of a person/animate) is there? b. ’u-vwa-tʉ ’ini-u ’uni’ni-kya-y? there-at-dir q:nonref:anim:subj-pl do/be-pl-imm ‘What kind (of persons/animates) are there? The non-subject stems participate in a number of morphological processes: they can be suffixed with one of several case markers or can join with a postposition or a verb, yielding complex forms such as those in (9). (9) Ute (Uto-Aztecan) (Givón 2001: 321) a. ’agha-naagha karʉ-i? q:inan:obj-in sit-imm ‘Where-in is s/he sitting?’ b. ’agha-vaa-tukhwa paghay’wa-y? q:inan:obj-at-go walk-imm ‘Where-to is (s/he) walking? Thus, languages differ not only in the number of interrogative stems, but also in the number of forms these stems can occur in. One important distinction typologically concerns the position of the interrogative word in the sentence. In many languages the interrogative word occurs obligatorily in clause-initial position. This is illustrated in (10) for Onondaga (Iroquoian). While the word order in that language is generally rather free (Woodbury 2017: 348–349), the interrogative word always occurs in initial position (p. 321). (10) Onondaga (Iroquoian) (Woodbury 2017: 322) Nwa·hóʔdęʔ naʔ snaʔjyęhá·wiʔ? what asrt 2sg.ag:pail:carry:stv ‘What is that pail you are carrying?’ Clause-initial interrogative words seem to be the dominant type in North America; typologically, this word order appears to occur in about one third of cases (see also Dryer 2013b). In languages such as Nakota (Siouan), the interrogative word occurs in situ, i. e., in the same place where the phrase it replaces would ordinarily occur. When asking about the patient, tuwé ‘who’ occurs in its regular position between the agent noun phrase and the verb (11). The question marker he (discussed already in 2.1) is optional. (11) Nakota (Siouan) (Cumberland 2004: 379) John tuwé į-yų́ ǧa he? John who stv-3ag>3pat:ask q ‘Whom did John ask?’ 288 Olga Lovick In some languages, such as Slave (Dene; Rice 1989: 1158–1160) or Osage (Siouan, Quintero 2004: 471), interrogative words may either be fronted or occur in situ with no apparent difference in meaning. Such mixed systems are quite rare in the world’s languages (Dryer 2013b). In many languages, the question word is followed by a focus marker. In Nishnaabemwin (Algonquian), the “emphatic focussing particle” dash is so common that it often fuses with the question word (Valentine 2001: 979). The question word wenen ‘who’, for example, coalesces with dash to wenesh (12). (12) Nishnaabemwin (Algonquian) (Valentine 2001: 980) Wenesh maanda debendang zhibiihgaans? who:foc this 3sg.prox>3sg:own pencil ‘Whose pencil is this?’ Another strategy that is common in Indigenous languages is the use of interrogative verbs in content questions. This is described in detail by Munro (2012) for the Takic subgroup of Uto-Aztecan (13), where this appears to be a highly productive process. (13) Mountain Cahuilla (Uto-Aztecan) (Munro 2012: 278) Hem-hí-yax-we? 3pl.sbj-ind-say-dur ‘What are they saying?’ Munro (2012) also notes the existence of interrogative verbs in Muskogean and Yuman languages. Interrogative verbs also attested in several Salish languages (14). (14) St’át’imcets (Salish) (van Eijk 1997: 134) kaxw kánəm 2sg.sbj be.what.state ‘How are you doing?’ In Musqueam (Salish), a substantial inventory of interrogative verbs is derived from a few basic stems (Suttles 2003: 395). Interrogative verbs can also be formed by incorporating a question word into a form of the verb ‘to be’ as done in Dëne Sųłiné (Dene). The verb in (15) is inflected for second person singular subject and imperfective mode. (15) Dëne Sųłiné (Dene) (Cook 2004: 99) ʔedlǫ́ lyeʔá ʔedlá-ǫlye ʔá what-2sg.sbj:be.named:ipfv q ‘What is your name?’ Questions and requests in North American languages 289 12.3 Requests Requests can be categorized along several different dimensions (Aikhenvald 2010), such as the identity of the addressee of the request (the person intended to carry out the action; in this sense, the term “addressee” is not necessarily identical to the term “hearer”), § 3.1, or polarity (whether the request is to do something or to abstain from doing something), § 3.2. Many languages make additional distinctions, § 3.3. It is important to note here that these dimensions may intersect, and that languages can have many different ways of phrasing a request depending on the context. Finally, some languages have a dedicated set of non-verbal expressions that are used in requests, § 3.4. In this section, I do not offer comparisons on how frequent a particular feature is typologically or even within North America. This is because the World Atlas of Language Structures draws different distinctions than I do in this chapter. Additionally, while question formation is discussed in somewhat comparable fashion in reference grammars, the same cannot be said for requests; information on this topic is often spotty. 12.3.1 Identity of the addressee A prototypical request involves getting the hearer to do something; Aikhenvald (2010: 18) calls such second-person requests canonical. The most basic, or common, form of a canonical request is often called an imperative; I follow this practice here. Requests with a first person (usually plural, and usually inclusive, i. e., including the second person) addressee are often called hortatives or exhortatives. Jussives is a common label for requests with a third person addressee. 12.3.1.1 Canonical imperatives Canonical imperatives are in many languages formally unmarked and thus the shortest possible verb form. Aikhenvald (2010: 19) comments that this lack of formal marking possibly reflects their functionally unmarked status as the most common type of request. Unmarked canonical requests are attested in many North American languages. In Northern Paiute (Uto-Aztecan, Thornes 2017: 151), canonical requests consist of the bare stem. Subject number may be expressed by suppletive (number-marked) stems (16) or by the use of overt pronouns (p. 153). (16) Northern Paiute (Uto-Aztecan) (Thornes 2017: 151) Yadua! talk.sg ‘Speak!’ (addressing singular) 290 Olga Lovick Another common strategy is to simply use a second person form. In Slave (Dene), the second person imperfective functions as canonical imperative (17). (17) Slave (Dene) (Rice 1989: 1109) Kágodahwhi! 2pl.sbj:go.out:ipfv ‘Go outside!’ (addressing plural) Other languages employ a special imperative marker. The Aleut (Eskaleut) suffix -da, -ada -‘imp’ in (18) never co-occurs with an overt subject but is interpreted as second person singular or plural depending on context (Golovko 2001: 301). (18) Aleut (Eskaleut) (Golovko 2001: 301) Qangu-da! come.in-imp ‘Come in!’ (addressing singular or plural) Aleut additionally employs a special intonation pattern involving lowered pitch and a lengthened final syllable (p. 302). In Onondaga (Iroquoian), an imperative suffix combines with an agent marker (Woodbury 2017: 101). The second person singular agent marker has a special imperative form (19). (19) Onondaga (Iroquoian) (Woodbury 2017: 101) Sęni·hęh́! s-ęni·hę-h’ 2sg.imp-stop-imp ‘Stop!’ (addressing singular) Another group of languages has a special imperative verb paradigm or set of paradigms. In the Algonquian language family, this is known as the ‘imperative order’, which contrasts structurally and functionally with the independent and conjunct orders as illustrated in Table 1 for Innu. Tab. 1: Innu (Algonquian) paradigms of ‘work’, indicative mood, present tense (Baraby 2017: 56) independent conjunct imperative 1SG 2SG 3SG 4SG 1PL 1/2PL 2PL 3PL nit-atusse-n tshit-atusse-n atusse-u atusse-nua nit-atusse-nan tshit-atusse-nan tshit-atusse-nau atusse-uat atusse-ian atusse-in atusse-t atusse-niti atusse-iat atusse-iaku atusse-ieku atusse-ht – atusse – – – atusse-tau atusse-ku – Questions and requests in North American languages 291 Outside the Algonquian language family, imperative paradigms are also reported for Klamath (Isolate; Stegnij 2001: 81) and several Eskaleut languages. 12.3.1.2 (Ex)hortatives Hortatives (or exhortatives) have a first-person addressee. In many languages, including Innu (Algonquian) in Table 1, they are limited to the first-person plural including the second person, e. g., atussetau ‘let’s (you and I) work’. In this language, the hortative belongs to the same imperative paradigm as second-person requests. In languages such as Ute (Uto-Aztecan), hortative and imperative formations differ. Second person requests use the bare stem plus an optional emphatic affix, while hortatives employ an irrealis suffix and a first person inclusive subject marker (20). (20) Ute (Uto-Aztecan) (Givón 2011: 305, 307) a. Yugwi! pl:sit ‘Sit!’ (addressing several) b. wʉ́ ʉka-qha-paa-rawi! work-pl-irr-1/2pl ‘Let’s (you and we) work!’ Some languages do not have a dedicated strategy for the formation of hortatives. In Moses-Columbia Salish, imperatives contain the enclitic =(t)aʔ and no subject inflection (21a). Hortatives can be formed in several ways, e. g., by adding a modal marker (21b) or by adding an ‘unrealized’ verbal prefix (21c). (21) Moses-Columbia (Salish) (Mattina 1999: 2, 13) a. ʔác’x̌-nt=(t)aʔ look.at-tr=imp ‘Look at it!’ b. saʔk ʔác’x̌-nt-m deon look.at-tr-1pl.sbj ‘Let’s look at it!’ c. t’íl’ ka-s-ʔác’x̌-nt-m emph unr-s-look.at-tr-1pl.sbj ‘Let’s look at it!’ 292 Olga Lovick Some languages limit hortatives to non-singular subjects (including minimally the speaker and the hearer), but not all of them do. Koasati, for instance, has a dedicated first person singular hortative suffix illustrated in (22).4 (22) Koasati (Muskogean) (Kimball 1991: 263) /má:mik cimmą́ hkaták/ má:mik cim-mán,h,ka-ták then 2sg.dat-tell,h:grade-1sg.sbj:imp ‘Then let me tell you about this!’ Singular hortatives can be difficult to translate into English. In his discussion of Uummarmiut hortatives, Lowe (1985: 155) notes that “[i]t is as if the speaker, talking to himself and trying to make up his mind whether he should do something or not, was urging himself to finally do it”, illustrating it with (23): (23) Uummarmiut (Eskaleut) (Lowe 1985: 155) niuqqar-la-nga have.tea-opt-1sg.sbj ‘I might as well have tea!’ Hortatives are not attested in all languages; e. g., Cumberland (2005) does not report their existence in Nakota (Siouan). 12.3.1.3 Jussives Jussives are requests directed at a third person, i. e., neither the speaker nor the hearer. They may belong to the same paradigm as canonical requests and hortatives as for example in Klamath. In this language, imperative markers for all three persons belong to what Stegnij (2001: 81) terms “class 23” suffixes, an obligatory class of modal and tense markers. A hortative form is shown in (24a), a jussive form in (24b). A canonical imperative is given in (26) below. (24) Klamath (Isolate) (Stegnij 2001: 86, 85) a. hak hay sle?-i:k emph excl see-imp:1sg ‘let me see it!’ a. q’ay mis sa sle:-tgi neg 2sg.o 3pl.s see-imp.3sg ‘do not let them see you!’ 4 “h-grade” in the gloss means that this form uses a particular stem form instead of the simple root. This language distinguishes between root imperatives and h-grade imperatives; Kimball (1991: 269) notes that h-grade imperatives “are used to request an action rather than to order it”. Questions and requests in North American languages 293 Jussives may also belong to a different paradigm. In Slave (Dene), the basic imperative uses the imperfective mode (17), while hortatives and jussives both employ the optative (25). (25) Slave (Dene) (Rice 1989: 1111) dezǫa kare nágoguyeh! children outside 3pl.sbj:play:opt ‘Let the children play outside!’ Like hortatives, jussives do not exist in all languages; Innu for example does not have them (see Table 1) nor does Northern Paiute (Thornes 2017: 161). 12.3.2 Polarity All requests surveyed so far were directives that called upon someone to do something. There are also negative directives that call upon someone to not do something. These are often called prohibitives. Prohibitives can be compositional or non-compositional. Compositional prohibitives consist of a regular imperative, hortative, or jussive negated by the same negator used in standard negation (negation of a declarative clause). Klamath uses compositional prohibitives characterized by the cooccurrence of the second-person imperative suffix, which occurs in affirmative and negative requests alike, and the standard negator q’ay (26). (26) Klamath (Stegnij 2001: 81, 88) a. Gida ?is Gis-lG-i! here 2sg>1sg step-down-2sg:imp ‘Here you step down on me!’ b. Q’ay del-ditgo:l-i! neg look-out.from.under-2sg:imp ‘Don’t look out!’ (addressing one) Typologically more common are non-compositional prohibitives. Van der Auwera, Lejeune, and Goussev (2013) note that these may differ from standard-negated imperatives with respect to the negator, the verb form, or both. Upper Tanana (Dene) illustrates non-compositional prohibitives. Standard negation is achieved by negative inflection plus the negative particle k’a(t’eey), while prohibitives use affirmative inflection and the prohibitive particle sǫ’ (27). While imperatives are usually in the imperfective, prohibitives are almost always in the optative (Lovick and Tuttle 2019: 131). 294 (27) Olga Lovick Upper Tanana (Dene) (Lovick and Tuttle 2019: 131) a. Ha’áát tsät įįtthèèł ch’a. out wood 2sg.sbj:chop:ipfv foc ‘Chop wood outside!’ (addressing one) b. Sǫ’ tsät ǫǫtthèèl! proh wood 2sg.sbj:chop:opt ‘Don’t chop wood!’ (addressing singular) c. K’àt’eey tsät į ̀į ̀tthéél de’… neg wood 2sg.sbj:chop:pfv:neg if ‘If you (sg.) had not chopped wood…’ The distinctions noted for affirmative requests may be available in negative ones. Onondaga, for example, allows negative jussives (as well as negative hortatives), (28). (28) Onondaga (Iroquoian) (Woodbury 2017:106) a. ahgwi tho hęhséh! ahgwi tho h-ę-hs-e-h proh loc trns-fut-2sg.a-go-imp ‘don’t go there!’ (addressing singular) b. ahgwí ęhatdó·gah! ahgwi ę-h-atdog-ah proh fut-3m.sg.a-notice-imp ‘he shouldn’t notice!’ Other distinctions, such as that between immediate and delayed requests in Algonquian (see § 3.3.2) are only available in affirmative requests, not in negative ones. 12.3.3 Other distinctions Many languages make additional distinctions within the functional domain of requesting. Different forms may be used depending on politeness, urgency, the relationship of speaker and addressee and so forth (§ 3.3.1). Some languages draw distinctions that have to do with the timing of a request (§ 3.3.2). More rarely, the form of a request is influenced by properties of the speaker or addressee (§ 3.3.3). 12.3.3.1 Softening and strengthening a request The forms described in § 3.1 and 3.2 are often perceived as somewhat rude, and many languages have requesting strategies that may be more appropriate in certain circumstances. Thornes (2017: 154) finds that in Northern Paiute, a request involving a verb form inflected for tense accompanied by a second person pronoun plus a deontic modal Questions and requests in North American languages 295 enclitic (28) is perceived as “softer” than one using the uninflected form illustrated in (15) above. (29) Northern Paiute (Uto-Aztecan) (Thornes 2017: 154) ɨɨ=sakwa ka=i=mia-si i=puh-nagi! you=mod obl=1=go-seq 1=inst/eye-chase ‘You should keep watching me where I go!’ Another strategy identified by Thornes (2017: 155) involves the bare stem imperative preceded by an intensifier, “indicat[ing] a kind of emphatic permissive command that lends an immediacy to the expected action” (29): (30) Northern Paiute (Uto-Aztecan) (Thornes 2017: 155) ɨnɨ mia! nɨ mai yongo-kwi,… intens go I dem evening-fut ‘Go ahead and go! I’ll stay here for the night,…’ Lovick (2016) notes that in Upper Tanana, requests in the imperfective mode signal that the speaker expects the request to be complied with, while those in the optative do so to a much lesser degree. The imperfective request in (31a) has the force of an order, while the optative request in (31b) has the force of a suggestion. (31) Upper Tanana (Dene) (Lovick 2016: 285, 286) a. An staanįįdaay! away:all 2sg:sg.go.away:iter:ipfv:imp ‘Go away! (addressing one) b. Shyiign t’axoh natonshya’! down:all finally 2sg:sg.go:iter:opt ‘It is time for you to go back down!’ (addressing one) In Aleut, on the other hand, optative requests are stronger than plain imperatives, although Golovko (2001: 307) additionally notes that intonation can also strengthen or soften a request. Crow (Siouan) has three strategies (Graczyk 2007: 151–153). The “simple imperative” in (32a) contains the imperative suffix -h; some verbs additionally contain a second person agent prefix.5 The “mild or polite imperative” additionally contains the suffix -kawe (32b). The “emphatic imperative” contains a suffix -wa and “adds a note of insistence: ‘do it or else!’” (32c). (This kind of marker is sometimes called “apprehensive”, cf. Aikhenvald (2010: 226)). 5 The presence of the second person agent marker seems to depend on the initial syllable of the verb stem. It is absent in (31a, b) but present in the form in (31c), where the verb begins with an unstressed i-vowel. 296 (32) Olga Lovick a. b. c. Crow (Siouan) (Graczyk 2007: 151, 152, 153) Baláxi-h! sing-imp ‘Sing!’ (addressing one) Disshi-káwe-h! dance-pol.imp-imp ‘Please dance!’ (addressing one) d-iháwi-wa-h 2sg.ag-sleep-emph.imp-imp ‘Sleep (or else)!’ (addressing one) In a similar vein Givón (2011: 308) describes a “strong obligative mode” for Ute, composed of the irrealis nominalizing suffixes. This mode may be used in affirmative and negative forms (33). (33) Ute (Uto-Aztecan) (Givón 2011: 308) a. wʉ́ ʉka-vaa-na-mʉ work-irr-nmlzr-2sg ‘you must work’ (addressing one) b. ’ʉmʉy-aqh-’uru ka-’ini-vaa-’wa-na 2sg/gen-it-that neg-do-irr-neg-nmlzr ‘you must not do this’ (addressing one) 12.3.3.2 Time of execution Some languages distinguish requests based on when they should be carried out. The imperative order in Algonquian languages allows the speaker to distinguish between immediate requests, to be carried out immediately, and delayed ones, to be carried out at a later time (34).6 (34) Nishnaabemwin (Algonquian) (Valentine 2001: 991, 993) a. Zhngishni-n doopwining. lie.prone-2sg:imp:imm on.table ‘Lie down on the table.’ (addressing one) b. Baamaa waabang bi+zhaa-kan… later tomorrow come-2sg:imp:del ‘Come tomorrow…’ (addressing one) 6 In the Algonquianist tradition, the hyphen is usually used to separate preverbs, while I use it in this paper to indicate morpheme boundaries. (33b) contains a preverb that I separated from the rest of the form using +. Questions and requests in North American languages 297 Baraby (2017: 67–74) suggests that in Innu (also Algonquian), forms similar to (33b) should be more appropriately analyzed as evidential imperatives, arguing that the action is to be carried out in the absence of the speaker. In Asiatic Eskimo7, imperatives can take the continuative marker, which contributes the meaning “urging to continue/not to terminate an action in progress” (Vaxtin 2001: 133): (35) Asiatic Eskimo (Eskaleut) (Vaxtin 2001: 133) a. negh-i eat-imp:2sg ‘Eat!’ (addressing one) b. negh-aq-i eat-cont-imp:2sg ‘Go on eating!’ (addressing one) A similar construction is reported by Quintero (2004: 297) for Osage (Siouan). Rice (1989: 1109–1110) reports that Hare (a variety of Slave; Dene) has a timing distinction within negative directives. Those in the optative mode warn against an activity not yet begun, while those in the imperfective direct an individual to stop the activity (36). (36) Hare (Dene) (Rice 1989: 1110) a. ʔelá dewǫtl’éle! boat 2sg.sbj:paint:opt:neg ‘Don’t paint the boat!’ (addressing one, painting has not yet begun) b. ʔelá dįtl’éle! boat 2sg.sbj:paint:ipfv:neg ‘Stop painting the boat!’ (addressing one) Koasati (Muskogean) seems to be particularly rich in timing distinctions within the domain of requesting. Kimball (1991, chap. 6) reports a distinction between first delayed (37b), second delayed (37c), and polite continuing imperatives (37d), all of which differ in turn from the bare root canonical imperative (37a). (37) Koasati (Muskogean) (Kimball 1991: 264, 267, 268) a. /ip!/ i:pa-Ø-DEL eat-2sg.imp-phr-term ‘eat!’ (addressing singular) 7 This is the language name used by Vaxtin; I suspect that this is the language known these days as “Siberian Yup’ik”. 298 Olga Lovick b. c. d. /amawí:cįh/ am-awí:ci-Ø-˛Vh 1sg.dat-help-2sg.imp-delay ‘help me later!’ (addressing singular) /taɬą́ :hah/ taɬ-:˛V́hah weave(imp)-sec.delay ‘weave it a lot later!’ /ɬopótcitik/ ɬopót-ci-tik pass.through-2sg.sbj-pol ‘please continue on passing through!’ 12.3.3.3 Indexing speaker or addressee properties In some languages the form of a request depends on certain properties of the speaker or addressee. Lakhota, for example, uses different request particles depending on the speaker’s gender. This intersects with distinctions regarding the relationship between speaker and addressee. Table 2 replicates a summary table provided by Trechter (1995: 57). Her table also includes gender-marked enclitics that are not used in requests; these have been omitted here. Tab. 2: Gender-marked request enclitics in Lakhota (Trechter 1995: 57) Male Female Command Familiar command Entreaty yo yetho ye ye nitho na Eastern Pomo (Pomoan) employs a special form when making a request of an in-law to display respect. The semantically unmarked form of a request with a singular addressee employs the imperative suffix -im (38a). If the addressee is an in-law, the plural imperative suffix -me is used with a singular stem (38b). For plural (non-in-law) addressees, the plural imperative suffix combines with a plural stem and/or a plural agent suffix (38c). (38) Eastern Pomo (Pomoan) (McLendon 1996: 530) kaˑki-m! sit.down:sg-sg.imp ‘Sit down!’ (addressing one) Questions and requests in North American languages b. c. 299 kaˑki-me! sit.down:sg-pl.imp ‘Sit down!’ (addressing daughter in-law) naˑphó-k-aki-me! several.to.be.located-punc-pl.ag-pl.imp ‘Sit down!’ (addressing several) Eastern Pomo in fact distinguishes a very large number of different types of requests. In addition to the basic form in (38a, c) and the in-law form in (38b), McLendon (1996) mentions a “polite imperative” which involves the concessive clitic (39a), “instructions” (which seem to function similar to delayed requests, 39b), “requests” (which seem to be more polite, 39c), and “supplications” (used in prayers, 39d). (This language also has “exhortations”, i. e., hortatives in the terminology used in this paper.) (39) Eastern Pomo (Pomoan) (McLendon 1996: 530, 531) a. xaˑká=bi yóx-k’i-l-im flint=concess become-do/make-dur-sg.imp ‘Please make an arrowpoint!’ (addressing one) b. xódaˑmal máˑ-l ʔáˑm čiki-pʰila after.awhile you(pl)-pat thing happen-if(sr) kuˑnúˑla-bùˑc’ike-heʔmìˑpal maˑ xaˑdí-yaki-baʔè Coyote-Old.Man-spec(pat) you(pl.ag) go.get-pl.ag-subjunc ‘After awhile, if something (bad) happens to you, you should go get Old Man Coyote.’ (addressing several) c. má=ti diˑt’áq-aʔèˑle you(sg.ag)=subjunc think-cond ‘would you think (about) it?’ d. yowx, maˑʔáy=ti moˑkoˑš prayer.opening food=subjunc become-des ‘May it turn into food!’ It is unclear how common languages with such rich inventories in request forms are. Reference grammars tend to list the more common forms, but, as noted by Aikhenvald (2010: 2), “[t]he possibilities for [the expression of] commands are immense, and openended”. 12.3.4 Imperative-only lexemes Many languages in North America have special imperative-only lexemes, non-verbal expressions used exclusively in requests (Aikhenvald 2010: 317). They exist in Musqueam (Salish; Suttles 2003: 470–474), Ute (Uto-Aztec; Givón 2011: 429), Koasati (Muskogean, Kimball 1991: 504–507), Innu (Algonquian, Baraby 2017: 63–64), Upper Tanana (Dene, 300 Olga Lovick Lovick 2020: 302–305), and probably many other languages. The semantic range of these lexemes is considerable, as comparison of the inventories in Koasati (40) and Innu (41) shows. (40) Koasati (Muskogean) (Kimball 1991: 504–507) a. hé:! ‘go on!’ b. himá:kǫ! ‘wait a minute!’ c. kappá:l! ‘shut up!’ d. máh! ‘listen!’ (41) Innu (Algonquian) (Baraby 2017: 63) a. matshi! ‘go ahead, go away!’ b. ashtam! ~ashtamite! ~mua! ‘come here!’ c. akua! ‘watch out! be careful!’ d. eshku! ‘wait a minute! hang on! stop a second!’ e. kata! ‘wait!’ f. eka pitama! ‘wait, just a moment!’ g. ushte! ‘move further away! get away from here!’ h. eku! ‘let’s go!’ 12.4 Questions and requests: using one as the other The preceding sections provide an overview over both the formal range of question and request formation in North American languages and over the range of distinctions made by individual languages. This overview is by necessity too brief and cursory in many places. Before closing, however, I want to draw attention to one interesting way in which questions and requests interact in many languages. At the outset, I noted that questions and requests have in common that they direct the actions of another person. In light of this, it is not surprising that oftentimes the functions of questions and requests overlap, allowing one to be used to express the other. Givón (2011: 314–315) observes that in Ute, negative polar questions often serve as polite requests. (42) Ute (Uto-Aztecan) (Givón 2011: 315) … Mʉ́ ni-áa nʉ́ nay kach ’u-vwaa-tʉ 2pl.s-q 1sg.o neg there-at-dir ‘… Won’t you please carry me there?…’ nɵ́ ɵ’wa-way-ura?… carry-neg-be McLendon (1996: 531) observes the very same strategy for Eastern Pomo, and Watahomigie, Bender, and Yamamoto (1982: 130) for Hualapai (Yuman). In Hualapai, there is additionally a systematic ambiguity between a certain type of negative request and a type of negative question—a conventionalized way of forming an affirmative request. Questions and requests in North American languages 301 The two sentences in (43) consist of the same morphosyntactic material and differ only in intonation: the prohibitive is characterized by falling intonation, while the negative question/affirmative request has rising intonation (Watahomigie, Bender, Watahomigie, Yamamoto, Mapatis, Powskey, and Steele 2001: 130). (43) Hualapai (Yuman) (Watahomigie et al. (2001: 131) a. Miya:m-a mđé! go-imp neg:asrt ‘Don’t go!’ (addressing one) b. Miya:m-a mđé! go-imp neg:asrt ‘Why don’t you go? (i. e., Go!)’ (addressing one) ↘ ↗ Lovick and Tuttle (2019) report that in Upper Tanana and Koyukon (both Dene), one conventionalized strategy to avoid uttering a prohibitive is to question the addressee’s behavior (44). (44) Upper Tanana (Dene) (Lovick and Tuttle 2019: 144) Dii xah ch’a utüh tidhįį’ia tl’aan why foc 3sg:over 2sg:step:incep:pfv and ‘Why did you step over it when you got there?’ ni’įįhaał? 2sg:sg.arrive:ipfv:prog In Inuktitut, we find the opposite. Changing the intonation of a canonical imperative to typical question intonation with rising pitch makes it into a “polite question” (45). (45) Inuktitut (Eskaleut) (Mallon 1991: 21) tiituriit? drink.tea-imp:2sg ‘Drink tea?’ This tendency towards functional overlap between questions and requests is attested in many languages across the globe. 12.5 Summary In this chapter, strategies for the formation of different types of questions and requests in the Indigenous languages of North America were reviewed. Questions and requests are common speech acts in routine interactions and are central to several approaches used in Indigenous language teaching. Two types of questions can be distinguished. Polar questions are questions that typically elicit a “yes” or “no” answer. In some languages, polar questions are marked only by a special intonation pattern. More common is the use of question particles or special interrogative paradigms. 302 Olga Lovick Content questions query a missing piece of information; they always contain a question word. The languages surveyed here differ considerably with respect to the number of question words. Another difference concerns the placement of the question word. In many languages, the question word always occurs in clause-initial position; in other languages, it occurs in the position of the phrase it replaces. In a few languages, either order is attested with no apparent change in meaning. Several Indigenous languages have interrogative verbs. In the domain of requesting, many types of distinctions can be drawn. One is the addressee: the addressee of a canonical request is the hearer (second person). The most common form of a canonical request is often called the imperative. In some languages, imperatives are bare stems; in other languages, special imperative morphemes or even paradigms exist. In yet another group, second-person subject forms are co-opted as imperatives. Requests directed at the first person (hortatives) and at the third person (jussives) are also common. In some languages, they belong to the same paradigms as imperatives, but this is not always the case. One can also distinguish requests to do something from those not to do something; the latter type is called prohibitives here. Prohibitives sometimes are simply imperatives combined with the same negative marker(s) as used in declarative sentences, but more frequently, they use special forms. In many languages (Indigenous and other) there are ways to soften or strengthen requests. Some languages distinguish immediate from delayed or general requests or use different forms depending on the identity of speaker or addressee. A number of languages have special imperative-only lexemes that can only be used in requests. Since requests and questions both direct the addressee to provide a (nonverbal or verbal) response, it is not surprising that in many languages, there is functional overlap between these types of speech acts. Thus, learning about question and request formation in any language (Indigenous or not!) does not merely involve learning to construct the forms: one also needs to learn when and how to use them in an appropriate fashion, which will be different for every language. References Aikhenvald, Alexandra Y. 2010. Imperatives and commands. Oxford: Oxford University Press. Asher, James J. 1969. The total physical response approach to second language learning. The Modern Language Journal 53(1). 3–17. Baraby, Anne-Marie. 2017. Imperatives and evidentiality in Innu. In Daniël Van Olmen & Simone Heinold (eds.), Imperative and directive strategies, 53–77. Amsterdam: John Benjamins. Cook, Eung-Do. 2004. A grammar of Dëne Sųłiné (Chipewyan). Algonquian and Iroquoian Linguistics – Special Athabaskan Number, Memoir 17. Winnipeg: Algonquian/Iroquoian Linguistics. Questions and requests in North American languages 303 Cumberland, Linda A. 2005. A grammar of Assiniboine: A Siouan language of the Northern Plains. PhD diss., Indiana University. Dryer, Matthew S. 2013a. Polar questions. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. https:// wals.info/chapter/116. Dryer, Matthew S. 2013b. Position of interrogative phrases in content questions. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. https://wals.info/chapter/93. Givón, Talmy. 2011. Ute reference grammar. Amsterdam/Philadelphia: John Benjamins Publishing Company. Golovko, Evgenij V. 2001. Imperative in Aleut. In Viktor S. Xrakovskij (ed.), Typology of imperative constructions, 300–314. Munich: LINCOM. Graczyk, Randolph. 2007. A grammar of Crow: Apsáaloke Aliláau. Lincoln: University of Nebraska Press. Hinton, Leanne, Margaret Florey, Suzanne Gessner & Jacob Manatowa-Bailey. 2018. The Master-Apprentice language learning program. In Leanne Hinton, Leena Marjatta Huss & Gerald Roche (eds.), The Routledge Handbook of Language Revitalization, 127–136. New York: Routledge. Kimball, Geoffrey D. 1991. Koasati grammar. Lincoln: University of Nebraska Press. Lovick, Olga. 2016. Įįjih and request formation in Upper Tanana Athabascan: Evidence from narrative text. Anthropological Linguistics 58(3). 248–298. Lovick, Olga. 2020. A grammar of Upper Tanana, Volume 1: Phonology, lexical classes, morphology. Lincoln: University of Nebraska Press. Lovick, Olga & Siri G. Tuttle. 2019. Video elicitation of negative directives in Alaskan Dene languages: Reflections on methodology. In Aimée Lahaussois & Marine Vuillermet (eds.), Methodological tools for linguistic description and typology. Language Documentation & Conservation, Special Publication No. 16, 125–154. Lowe, Ronald. 1985. Uummarmiut uqalungiha ilihaurr̂ û tikrâ ngit: Basic Uummarmiut Eskimo Grammar. Inuvik: Committee for Original Peoples Entitlement. Lupardus, Karen Jaque. 1982. The language of the Alabama Indians. PhD diss., University of Kansas. Mallon, Mick. 1991. Introductory Inuktitut reference grammar. Montreal: Arctic College – McGill University Inuktitut Text Project. Mattina, Nancy. 1999. Moses-Columbia Imperatives and Interior Salish. Anthropological Linguistics 41(1). 1–27. McLendon, Sally. 1996. Sketch of Eastern Pomo, a Pomoan language. In Ives Goddard (ed.), Handbook of North American Indians. Volume 17: Languages, 507–550. Washington: Smithsonian. Munro, Pamela. 2012. Interrogative verbs in Takic. UCLA Working Papers in Linguistics, Theories of Everything 17(32). 274–284. Quintero, Carolyn. 2004. Osage Grammar. Lincoln: University of Nebraska Press. Rice, Keren. 1989. A Grammar of Slave. Berlin & New York: Mouton de Gruyter. Sadock, Jerrold M. & Arnold M. Zwicky. 1985. Speech act distinctions in syntax. In Timothy Shopen (ed.), Language typology and syntactic description. Volume I: Clause structure, 155–196. Cambridge: Cambridge University Press. Searle, John R. 1976. A classification of illocutionary acts. Language in Society 5(1). 1–23. Stegnij, Viktor A. 2001. Imperative constructions in Klamath. In Viktor S. Xrakovskij (ed.), Typology of imperative constructions, 78–97. Munich: LINCOM. Suttles, Wayne. 2003. Musqueam reference grammar. Vancouver: UBC Press. Thornes, Tim. 2017. On the heterogeneity of Northern Paiute directives. In Alexandra Y. Aikhenvald & R. M. W. Dixon (eds.), Commands: A cross-linguistic typology, 146–168. Oxford: Oxford University Press. Trechter, Sara. 1995. The pragmatic functions of gender deixis in Lakhota. PhD diss., University of Kansas. Valentine, J. Randolph. 2001. Nishnaabemwin reference grammar. Toronto: University of Toronto Press. 304 Olga Lovick Van der Auwera, Johan, Ludo Lejeune & Valentin Goussev. 2013. The prohibitive. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. https://wals.info/chapter/71. Van Eijk, Jan. 1997. The Lillooet language: Phonology, morphology, syntax. Vancouver: UBC Press. Vaxtin, Nikolaj B. 2001. Imperative sentences in Asiatic Eskimo. In Viktor S. Xrakovskij (ed.), Typology of imperative constructions, 129–144. Munich: LINCOM. Watahomigie, Lucille J., Jorigine Bender, Philbert Watahomigie, Sr., Akira Y. Yamamoto, Elnor Mapatis, Malinda Powskey & Josie Steele. 2001. Hualapai reference grammar. Expanded and revised edition. Osaka: Endangered Languages of the Pacific Rim. Watahomigie, Lucille J., Jorigine Bender & Akira Y. Yamamoto. 1982. Hualapai reference grammar. Los Angeles: American Indian Studies Center, UCLA. Woodbury, Hanni. 2017. A reference grammar of the Onondaga language. Toronto: University of Toronto Press. Anna Berge 13 Information structure Abstract: The way people decide to present what they want to say reflects the choices they make about what it is they want to focus on, or what they think other people in a conversation know. The study of how people package the information they want to convey in a sentence is called information structure. Some important concepts in the study of information structure include the notions of focus (what a speaker considers particularly important information), topic (what a speaker is talking about), and word order (what information a speaker decides to present first). Indigenous languages, being so different syntactically to widely spoken languages like English, have much to offer to studies of information structure. For example, how are notions such as focus and topic indicated when a sentence may consist of a single word? In this chapter, I present the basic notions of and approaches to information structure and the challenges that Indigenous languages have posed to studies of information structure, then conclude with a brief look at why an understanding of information structure is helpful for language revitalization efforts. 13.1 Introduction Most discussions of syntactic structure in a language focus on how words are combined into phrases, clauses, and sentences. Another way to look at a sentence or clause, however, is how information is packaged within it: how new vs. old information is indicated, how important or newsworthy information is differentiated from backgrounded information, how the use of nominal forms of verbs or verbal forms of nouns affects the interpretation of the message, and so forth. This approach to understanding language is called Information Structure. The packaging of information happens at all levels; thus, we can ask whether or not a language has a single word for a concept, such as caused events. In English, we have separate words for ‘to die’ and ‘to cause to die,’ i. e. ‘to kill.’ In Kalaallisut (Eskaleut), the same concepts are formed from the same lexeme: toqu- ‘to die,’ toquC- ‘to kill’ (ultimately from toqu-t- ‘die-caus’). Information structure, however, specifically refers to information packaging within sentences. In (1a–1c) from Kalaallisut (Eskaleut), the same information (tamuatsivaaq ‘seal skin and fat chewing gum’) is packaged in four different ways in three consecutive clauses: as an incorporated noun in (1a), as a part of the verb and the verb ending in (1b), as an independent pronoun taanna ‘that one’ also in (1b), and as an independent noun in (1c). (1) Kalaallisut (Eskaleut) (Berge 2011:61–62) a. tamuatsivaar-tor-luta tassa tamuatsivaaq-eat-1pl.ct that.is https://doi.org/10.1515/9783110600926-013 306 Anna Berge b. c. tassa taanna tamua-rujoor-tar-parput that.is that.one.abs chew-casually-hab-1pl.subj/3sg.obj.ind ‘…we used to chew it continually’ tamuatsivaa-mik taasar-parput uagut taqqava-ni tamuatsivaaq-inst call-1pl.subj/3sg.obj.ind 1.pl.abs south-loc ‘…we in the south used to call it “tamuatsivaaq”’ Why do this? Varying the way we speak makes a story more interesting. All languages have ways of manipulating the way that information is packaged. The study of information structure is relatively new in terms of the field of linguistics, and major ideas and advances in our understanding thereof are still being developed, but several components of information structure are generally accepted as fundamental: word order, or the order in which we present information; topic, or what it is we are talking about, and focus, or the new information we want to communicate about the topic. In section 2, I will develop these concepts in more detail. The theory of information structure was initially developed for languages very different from the Indigenous languages of North America. Over the past several decades, theories of information structure have been greatly enriched and informed by the studies of Indigenous languages. Several characteristics shared by many of these languages have especially contributed to the discussion, including different word orders, nominal marking systems, and different concepts of word and sentence boundaries. In section 3, I will address the contributions that Indigenous languages have made to these aspects of information structure. In the process, I hope to show how important an understanding of information structure is to the study of the Indigenous languages, and why this is vital to revitalization efforts. 13.2 What is information structure? The concept of a formal level of grammatical structure based on the organization of information rather than words has a long tradition in fields such as philosophy or logic. Within linguistics, this concept was first developed and found useful for characterizing the relatively flexible word order of Slavic languages (cf. Firbas 1964), as in (2a) and (2b), which differ in the order of the subject (ya ‘I’) and object (knigi ‘books’): (2) Russian (Firbas 1964) a. Ya knig-i 1nom.sg book-acc.pl ‘I love books’ b. knig-i ya book-acc.pl 1nom.sg ‘books, I love (them)’ lyubly-u love-1sg.pres lyubly-u love-1sg.pres Information structure 307 Changing the expected word order of a sentence without changing grammatical marking allows a speaker to highlight something; what comes first in a sentence is typically what the speaker wants to talk about, or the theme, or topic, and what comes after is what the speaker wants to say about it, or the rheme, or comment. As the importance of information structure was recognized, it was taken up in studies of other language types, including English. Word order is less free in English, and variations tend to incur a change in structure, as in the addition of the pronoun ‘them’ in the translation of (2b). These early discussions focused on languages with typically separate verbs and arguments, or in other words, languages with structures often quite different from those of the Indigenous languages of America. How does one talk about information structure in languages with sentences that frequently consist of a single word, as in Unangam Tunuu (Eskaleut) aniqduĝikuqing ‘I have a child’? Nevertheless, it is widely accepted that we all make choices about how we structure the presentation of information, or how we package the information, in order to convey how important one piece of information is with respect to another, regardless of the type of language we speak.1 Different languages impose different strategies for expressing what is important on the speakers, but all languages have a variety of strategies for doing so. In (3a) and (3b) from Unangam Tunuu, information structure strategy relies not on word order but on decisions about word building: (3) Unangam Tunuu (Berge 2016:144–145) a. aniqdu-ĝi-ku-qing child-have-ind-1sg ‘I have a child.’ b. aniqdu-x̂ mata-ku-qing child-abs.sg have-ind-1sg ‘I have a child.’ Whenever there are two different ways of communicating what appears to be the same information, the information is in fact not the same (Mithun 1983). There are subtle differences between these two sentences, best explained as a function of speaker and hearer presuppositions about why the speaker is giving this information, what he/she expects the hearer to understand, etc. (Berge 2016). The analytic structure in (3b) is found in a wider variety of contexts today, whereas the incorporated structure in (3a) 1 This is evident in the long history of studies of related sentences, encompassing many syntactic and semantic theories, e. g. Chomskyan transformational and generative grammar, Fillmore and Kay’s studies of semantic roles in different sentence types, Foley and Van Valin’s functionalist approaches to syntax, and speech act theories, e. g. regarding the effects of different performative verbs such as ‘command’ vs. ‘request’ on the meaning of a sentence. Information Structure is one approach to understanding the way people package information in a clause; cross-linguistic studies such as those by Vallduví and Engdahl (1996) highlight the different structural options available to speakers for packaging the same information. 308 Anna Berge often has either special connotations or restrictions in use. Aniqduĝikuqing can mean ‘I have a child’ or ‘I am with child’ (i. e. pregnant); and although in principle, it can mean ‘I have children,’ the analytic structure is preferred for plurals: aniqdun matakuqing ‘I have children.’ Information structure, therefore, requires us as linguists to not just understand the sentence structure, but also the factors involved in the choice of sentence structure. These include understanding presuppositions and knowledge shared between the speaker and hearer (pragmatics), the flow of information negotiated during a conversation (conversational analysis), or the choice of words (lexical analysis) or perspective (semantics) that best convey the speaker’s meaning. Key works on information structure in which these concepts are elaborated include Chafe (1976), Lambrecht (1994), Vallduví (1991, 1995), Gundel (1988), and Krifka (2006) among others. Although information structure is an important component of linguistic knowledge, there is no agreement on either the key elements of information structure or their definitions. This, unfortunately, makes it difficult to compare information structure between languages in general, and even more so between the many less-widely studied languages. Nevertheless, there are concepts commonly taken up in most studies, including especially word order, focus, and topic. 13.2.1 Word order One of the main premises of information packaging is that we limit the amount of information in any given utterance to something manageable to the hearer; one pervasive idea is that an utterance preferably includes no more than one new piece of information. For example, saying ‘John likes Mary’ pragmatically suggests that both ‘John’ and ‘Mary’ are new participants, and in Cayuga (Iroquoian), a sentence with both ‘John’ and ‘Mary’ overtly expressed would seem odd. In normal speech, speakers are more likely to have introduced one of the two previously, and therefore to say either (4a), (4b), or (4c) (4) Cayuga (Iroquoian) (Mithun, 1992:18) a. Mary shakó-n̥ǫ̥hwe’-s Mary he/her-like-hab ‘He likes Mary’ b. John shakó-n̥ǫ̥hwe’-s John he/her-like-hab ‘John likes her’ c. Shakó-n̥ǫ̥hwe’-s he/her-like-hab ‘He likes her’ This necessarily results in an utterance that contrasts the new information with the already known, or given information. This is the basis of much terminology in the literature, e. g., the following pairs of terms, which are not necessarily synonymous and which Information structure 309 may overlap in some ways: theme/rheme; given/new; backgrounded/foregrounded information; and topic/comment, or topic/focus. The contrast between new and given information is expressed in many ways, but one of the most obvious is through word order. Because we can only say one word at a time, what gets expressed first sets the tone for what follows. In a neutral statement (with no particular emphasis or unexpected use of language), the order of given vs. new information is assumed to be constant, and a change in the expected word order is a signal to pay special attention. In many early studies, the claim was that given information precedes new information; thus, in the statement ‘Mary is making some cookies,’ ‘Mary’ is given information (using the name ‘Mary’ with no other context assumes that she is known to the hearer), and ‘is making some cookies’ is new (hence the use of the indefinite ‘some’). This order has been found in Yaqui (Uto-Aztecan) (Guerrero and Bellero 2010). However, new and/or important information precedes given and/or backgrounded information in languages as different as Ojibwa (Algonquin), Ute (Numic), Tuscarora (Iroquoian), Coos (Coosan) and Unangam Tunuu (Eskaleut) (Mithun 1992:17, Mithun 1983, Berge 2015). In Unangam Tunuu, one or more of the verbal arguments is likely to be pronominally marked on the verb; this may result in a sentence consisting only of the verb, as in (3a), or in a sentence with the new information first, as in (5b): (5) Unangam Tunuu (Edna Floyd speaker, Berge field notes 2013) a. Kiin ing’ suxuraada-n agu-na-n? Who these cookie-pl make-part-3pl ‘Who made these cookies?’ b. Mariiya-m agu-laana-ngin. Mary-erg make-recent.past-3pl.an ‘Mary made them’ Word order studies have grown in sophistication and complexity, in part as a result of data from Indigenous languages. The order given first/new last may not be appropriate for all languages, or in all circumstances; and neither may a description that relies on overt or expressed noun phrases to the exclusion of information packaging within the verb. Nevertheless, normal discourse typically involves more than one-word sentences, and the distinction between given/new, important/background, etc. is valid and useful. Even in extremely polysynthetic languages like Kalaallisut, while each clause may consist of a verb only, it more frequently includes particles, adverbs, and postpositional phrases, as in (1a)-(1c). Furthermore, word order variations have important effects in any language: moving a word to the front of a clause may signal what the sentence will be about (i. e. the topic) or it may signal new information (i. e. the focus), whereas moving a word to the very end may signal an antitopic, a different kind of emphasis, a reminder of given information that is only expressed inflectionally on the verb, etc. In Koyukon (Na-Dene), Tlingit (Na-Dene), and Unangam Tunuu, all verb-final languages, moving a noun phrase after the verb signals that it is old but important information, whereas at least in the latter two languages, moving a noun phrase before the clause 310 Anna Berge signals that it is new and important information (cf. Thompson 2000, Crippen 2014, and Berge 2016 for the respective languages). Word order thus interacts with the notions of focus and topic. 13.2.2 Focus Focus and topic are often discussed with reference to and in opposition to each other, but they are separate concepts. Perhaps one of the most widely cited definitions of focus is that of Lambrecht (1994), for whom focus is new information about what is being talked about, or topic. He distinguishes three types of focus. The location of focus in a sentence can be an argument of the verb, such as a subject or object; in (5a) and (5b), the focus is on the subject (kiin ‘who’ and Mariyam ‘Mary’, respectively). Focus on an argument is often contrastive, as in ‘Is John or Mary making some cookies?’ ‘Mary’. When the predicate of the sentence is in focus, we tend to have a typical topic-comment structure, where the subject is the topic, or given information, and the rest of the sentence is the new information, as in ‘making some cookies’ in ‘Mary is making some cookies’. Finally, the sentence as a whole may describe new information, as in the answer to ‘What’s going on?’ ‘Mary’s making some cookies’. Another influential understanding of focus emphasizes the role of contrast rather than newness. For Krifka (2006), focus is “the presence of alternatives that are relevant for the interpretation of linguistic expressions”; in ‘Who’s making some?’ there are a set of relevant people who could be making cookies, while in ‘What’s going on?’ there is a set of relevant activities that could be happening. Whether focus is about newness or contrast, the linguistic indication of focus may include the use of stress, a particle, a cleft structure (using a complex sentence to focus one element, as in ‘It’s Mary that’s making the cookies’), or moving an argument, either to the right or to the left of the clause. These two understandings of focus are not mutually exclusive. In (6a), from Meskwaki (Algonquian), the focus (in bold) is the new information, or comment; in (6b), the focus is contrastive (it is also a cleft structure): (6) Meskwaki (Algonquian) (Dahlstrom, 2003:152) a. ni∙na a∙kwi wi∙h-na∙kwa∙-ya∙nini I not fut-leave-1 ‘As for me, I’m not leaving’ b. a∙kwi ni∙na wi∙h-na∙kwa∙-ya∙nini not I fut-leave-1 ‘I’M not leaving; it’s not me who’s leaving’ Information structure 311 13.2.3 Topic The term topic has also been applied to a variety of phenomena. It is commonly defined as the given part of a sentence, and thus it is often correlated with a) the first part of a sentence, as discussed above; b) the subject, since most world languages have a word order in which the first element is the subject; c) a primary (preferably animate) participant, since that is what is likely to be the subject of the discourse and therefore already given; d) backgrounded information, since what is given or known tends not to be highlighted; e) pronouns, since given information need not be repeated in full. All of these are illustrated with ‘she’ in the following: ‘Mary is in the kitchen. She is baking a cake.’ The topic in (6a) is ni∙na ‘I’; it is also the first element, the subject, and the given information. A topic can be focused (because it is being contrasted, or it is a new topic, or it is made more prominent by some form of movement, such as fronting, cf. Wilbur 2012:464; Gundel 1988:296); this is the case in (6b). On the other hand, we can think of topic as what the sentence is about; this is often correlated with the notion of contextual frames: by stating the topic directly, the speaker signals that he or she is assuming that the hearer shares a set of expectations generally associated with the topic. For example, in talking about baking, we make assumptions about the types of food, equipment, people and/or personality types, etc. involved in baking. Some languages, such as Lakota (Siouan) and Haida, are known as topic-prominent languages; they mark this kind of topic grammatically, with a topic particle or displacement of the topic (often to the left of the sentence). In Haida (7), the topic chiin iiwaanda ‘big fish’ is the first element in the sentence and is indicated by the following particle uu: (7) Haida (Eastman and Edwards 1983:58) chiin iiwaandaa uu l guulaagan fish big top she likes ‘Big fish are what she likes.’ 13.2.4 Identifying focus and topic Focus and topic are identified in a variety of ways at all levels of grammatical structure. Thus, focus can be indicated phonologically by stress, as in the English translation of (6b). Morphologically, a topic may simply be indicated by a pronominal inflection on the verb, as in the indication of the subject ‘I’ in the inflectional ending -qing in (3a-b) or by a particular case marker, determiner, or enclitic; contrastive focus may be indicated by the decision to use a separate expression, rather than a fused or incorporated structure (3a-b). Syntactically, a topic or focused element may be in front of a sentence, clefted, or expressed with some other non-neutral structure (6b). Lexically, either focus or topic or both may be indicated by a particle. In Western Apache (Na-Dene), both focus and topic are distinctively marked: =(h)í is an enclitic that can mark topics (8a-b), and =go an enclitic that can mark focus (8d); (8c) is a neutral statement: 312 (8) Anna Berge Western Apache (Na-Dene) (de Reuse 2001:14) a. Gósé=hí ápos yíyą̄ą̄ dog=top apple it.ate.it ‘Speaking about the dog, it ate the apple.’ b. Gósé ápos=í yíyą̄ą̄ dog apple=top it.ate.it ‘Speaking about the apple, the dog ate it.’ c. Shiye’ hish’į̄ į̄ ni’ my.son I.see.him past ‘I saw my son.’ d. Shiye’=go hish’į̄ į̄ ni’ my.son=foc I.see.him past ‘It is my son that I saw (and not someone else).’ Such clear marking of focus and topic is uncommon, however, and in many languages, a particle is optional. In Kalaallisut, for example, tassa ‘that is’ indicates that a neighboring phrase is focused (1a-b), but a phrase may be focused without the use of this particle. It can often be confusing to distinguish between topic and focus because the two are not necessarily mutually exclusive, and many strategies used for one may also be used for the other. Frequently, multiple strategies are used, and since these strategies overlap, it is unclear what indicates topic and what indicates focus, if these can be isolated. For example, both fronting and the use of a particle may indicate topic, but they are also common strategies for focus. This is further compounded by different uses of the terms topic and focus in the literature, as well as by different understandings of their characteristics over time. In describing topic in Haida, for example, Eastman and Edwards (1983) suggest that all topics are contrastive, and that sentence-initial elements not marked by the topic marker uu are focused, but not contrastive, as we see by comparing (7) with (9): (9) Haida (Eastman and Edwards 1983:58) chiin iiwaandaa l guulaagan fish big she likes Focus Subj Verb ‘She likes big fish.’ However, contrast is often reserved for discussions of focus. Distinguishing the two terms is especially confusing when comparing strategies cross-linguistically: where one language uses one particular strategy for signaling topics, another might use the same strategy for focused elements. Finally, the needs of information structure frequently clash with those of a different level of grammatical structure, as in when both sentential subjects and topics are generally first in a sentence, but the topic is something other than the subject. Information structure 313 13.3 Indigenous language studies and information structure The Indigenous languages of America have challenged earlier understandings of information structure in several ways, as a result of which our understanding of information structure has been enormously enriched. Although there are still relatively few studies of information structure in Indigenous languages, compared to others, information structure is a major topic of interest in studies of semantics in these languages (Matthewson 2017:6). As a result of studies of these languages, theories of information structure have had to address issues such as 1) verbal inflection, its effect on word expression and word order, and the very notion of basic word order, 2) the different nominal and head marking systems, and 3) different levels of morphological synthesis, the notion of a unit of meaning within a sentence, and indeed the notion of a sentence. 13.3.1 Word order We typically speak of word order with respect to subject, object, and verb: SOV, SVO, VOS, etc. Almost all possible word orders are found in the North America; we speak of Ojibwa or Haida as having VOS order, for example, and Na-Dene and Eskaleut languages as having SOV word order. Discussing word order involves assumptions that there is a basic word order, and that subject and object are overtly expressed. Mithun (1992) argues that some languages, such as Coos and Cayuga, may not have a basic word order, and that preferred word order in any give sentence is related to the pragmatic or communicative needs at the time of utterance. On the other hand, some languages are claimed to have relatively flexible word order, but they have been found to have very definite preferences; Yaqui, for example, strongly disprefers word orders in which the object precedes the subject (Guerrero and Belloro 2010:123). Basic word order is in opposition to non-neutral word order, which implies fronting, postposing, or extraction of a subject or object from the sentence. A language with basic SOV word order might allow non-neutral orders such as OSV (fronting of O), OVS (postposing of the subject), or S, OV (extraction of the subject, signaled through intonation, indexing,…). In Unangam Tunuu, a language that has generally been characterized as having fixed SOV word order, constructions with an extracted topic (10) may actually be more natural in casual conversation. (The object is said to be extracted here both because of its position before the subject (O, SV) and its inflectional marking): (10) Unangam Tunuu (Bergsland 1997:143) ayx̂aasi-x̂ anu-m aga-ti-ku-u boat-abs current-rel take-caus-ind-3sg/3sg.an ‘the boat, the current took it’ = ‘the boat was taken by the current’ 314 Anna Berge In many languages, the subject and sometimes the object are indexed on the verb with pronominal inflection. A language with pronominal indexing on the verb may allow the subject or object to be unexpressed if it is understood from context (4c). In this case, a clause may have no overt noun phrases; this tends to be the case in mid discourse, when the participants have already been mentioned, and the focus is on an activity described by the verb. A clause may, however, have one, two, or more rarely three overt noun phrase arguments. Time and again, it has been reported that a sentence with two or more overt noun phrases is rare in discourse (Mithun 1992:20; Berge 2009, 2011), and even disallowed (Beck 2010:42) in part because too much information is being introduced simultaneously, and in part because speakers prefer not to repeat previously introduced information. Far more frequently, a sentence will have at most one overt argument, which may be either a topic or a focused element. Thus, in a language with a verb final or verb initial basic word order, neutral word order by itself may not be a good indicator of topic or focus. In consecutive clauses (11a-c) from Unangam Tunuu, the topic is x̂ulustaakam uluu ‘seal meat,’ but both the topic and some new information riisax̂ ‘rice’ are focused by being overtly expressed and sentence initial: (11) Unangam Tunuu (Susie Merculief speaker, Berge field notes 2007) a. …x̂ulustaaka-m ulu-u kakliitka-x̂si-ĝi-i bachelor.fur.seal-rel meat-3sg.posm meatball-make-pas.part-3sg.posm ‘having made seal meatballs’ b. …x̂ulustaaka-m ulu-u su-x̂ta-lix bachelor.fur.seal-rel meat-3sg.posm take-cont-conj ‘using seal meat’ c. riisa-m una-ĝi-i su-x̂ta-lix… rice-rel cook-pas.part-3sg.posm take-cont-conj ‘using cooked rice’ In combination with other factors, however, such as whether or not an argument is overt, a subject or object, in neutral or non-neutral order with respect to the verb (cf. (10)), accompanied by a particle (cf. (8a-d), or (7) vs. (9)), and so forth, word order can be an indicator of topic or focus status. Many languages in North America show a very strong tendency, perhaps even requirement, for overt focus and topics, whether newly introduced, contrastive, or continued, to be toward the beginning of a sentence. In Kiowa (Tanoan), a verb-final language with relatively fluid word order, focus and topic must precede the verb (12a-b); they cannot follow the verb (12c); and they must precede a verbal particle (e. g. an) if one is present (12d), as opposed to non-topics, which would be found after a verbal particle: (12) Kiowa (Tanoan) (Harbour et al, 2012:104) a. Focus: Hâatêl gɔ–góp? who.q 3s:2s–hit.pf ‘Who hit you?’ Information structure b. c. d. Focus: Focus: Topic: 315 Carl ę́ –góp. Carl 3s:1s–hit.pf ‘Carl hit me.’ *Ę́góp Carl 3s:1s–hit.pf Carl ‘Carl hit me.’ Kɔ́ígú an ém–thɔ̨ɔ̨-téttɔ hágyá Kiowa.i hab 3a:refl–story-tell-impf sometime ɔnk!îi gya–dɔ́ɔ́mêi nɔ past 3p–be.evid and.df ‘The Kiowas tell a story about a past time when . . .’ In Unangam Tunuu too, both focus and topic tend to be sentence-initial, but there are interesting differences between them arising from factors such as argument and overtness. For example, an overt object is almost always new, or focused, whereas an overt subject may be equally new or given. These tendencies correlate with the overwhelming preference for objects to come before the verb, and the frequent occurrence (up to 30 % in conversations) of subjects after the verb, or postposition of subjects. Postposed elements, on the other hand, tend to function differently, although there is a great deal of variation in how they are interpreted. In many cases, they are arguments that in other circumstances might have been unexpressed. Harbour et al. (2012:113) see them as having a role in discourse structure, beyond the sentence; Mithun (2018) calls them antitopics, noting that in Cayuga, they are often intonationally separate from the sentence, and Berge (2011) finds the same for Kalaallisut. In Unangam Tunuu, however, they are almost always pronominal subjects and so much a part of the sentence that they are becoming cliticized to the preceding verb. In Yaqui, too, they are definite, identifiable, and previously introduced in the discourse. In Koyukon (13), they are almost always definite, human, and associated with important participants. (13) Koyukon (Thompson 2000:231) Huyeɬ gheel beyeenhooldlet go tsook’aale and.then 3s.became.angry dem old.woman ‘And then the old woman became angry.’ Tersis and Carter-Thomas (2005:494) find the same for Tunumiisut (Eskaleut), but they also show that postposing can be a focusing strategy that results in topic-comment word order. North American Indigenous languages show that the formulation of topic-comment or given-new characterization of information structure is simplistic. Czipionka (2007) finds a correlation between word order and focus, with a greater tendency to have focus first in SOV languages. However, North American languages are not uniform (Yaqui, an SOV language, has topic-comment order). Furthermore, they have shown that languages with flexible word order do, in fact, have word order requirements, and freedom is constrained by information structure and pragmatic factors. 316 Anna Berge 13.3.2 Case marking Nouns are often marked to differentiate them from other nouns. One common type of differentiation is case, which is the inflectional marking on nouns that allows a speaker to distinguish the various roles that nouns play within sentences, such as subject and direct object. Some languages, such as Na-Dene languages, make little to no use of case marking, while in other languages, case marking is much more extensive. There are a number of case marking systems. Nominative-accusative systems, in which sentential subjects are inflected in one way (with nominative case) and direct objects in another (with accusative case) are well-known from studies of European languages, including the Slavic languages described by early studies of information structure and a relatively small number of North American languages such as Miwok (Utian), Maidu (Maiduan), Yaqui, or the Yuman languages. The latter two groups are unusual in having a special marker for the nominative case, whereas the vast majority of nominative-accusative languages have marking on the accusative case. In many of the former type of nominative-accusative languages, nominative has been found to coincide with topic. Unfortunately, there is little discussion of information structure or discourse structure in North American marked-nominative languages. Handschuh (2014) notes that a noun being topicalized or focused in Maidu may have an emphatic marker and sentence-initial position (14), but this is not enough to conclude that subjects are topics: (14) Maidu (Handschuh 2014:126, quoting Dixon 1911:711) sü-m has nik do’kan dog-nom emph 1sg.acc bite ‘The dog bit me.’ Another system, known as ergative-absolutive case marking, is found in languages along the west coast, such as Eskaleut languages, Tsimshian, and Chinook. In ergative-absolutive systems, the subject of an intransitive verb and the object of a transitive verb are marked in the same way (with absolutive case), and the subject of a transitive verb is marked differently (with ergative case); in other words, case does not match up with the syntactic categories of subject and object, but it does reflect information structure. This pattern of case marking seems to reflect the introduction of new information in discourse (Du Bois 1987). In extended discourse, however, absolutive case occurs far more frequently than ergative case does, and in most cases, it does not coincide with new information, but rather with topic. This has been argued for Inuit (Tersis and Carter-Thomas 2005, Berge 2011, Johns and Kucerova 2017). Ergative subjects in Kalaallisut are infrequently expressed, but when they are, they are focused, and not topical within the clause; in Unangam Tunuu, they are specifically non-topics (Berge 2009). A very different type of nominal marking involves differentiating the roles of 3rd person participants in discourse (i. e. participants other than the speaker and hearer). In some languages, gender is an important distinction, although gender does not always relate to information structure as discussed here. In others, the distinction is between Information structure 317 more and less important or prominent participants, with prominence often correlated with animacy (how participants are ranked on a hierarchy from animate to inanimate). The most prominent 3rd person argument in a sentence is designated as proximate, while all other 3rd person arguments are obviative. This type of distinction is found in some Salishan languages, Kutenai, and Algonquian languages; Navajo and some other Na-Dene languages have a similar distinction, without marking on the nouns. There is a consistently close association between information structure and obviative-proximate marking, with proximate often associated with topic (Junker 2004; Zúñiga 2014); obviative has been described as a non-topic. Proximate and obviative are distinct from subject and object; proximate in neutral clauses is the subject (15a), but the obviative can be a subject (15b), in which case a special indication is made on the verb, as illustrated in East Cree (Algonquian), a VOS language: (15) East Cree (Algonquian) (Junker 2004: 349–350) a. miyeyim-e-u uyuuh atim-h uu awaash like.ta-dir-3 this.obv dog-obv this.prox child.prox ‘a child likes dogs’ b. miyeyim-iku-u uyuuh awaash-ah uu atim like.ta-inv-3 this.obv child-obv this.prox dog.prox ‘this child likes this dog’ Almost all word orders are found in natural discourse and either or both subjects and objects can precede the verb. Fronted noun phrases are said to be focused (16a). However, when both precede the verb, the obviative, or non-topic, cannot be first (16b): (16) East Cree (Algonquian) (Junker 2004: 349) a. uu awaash uyuuh atim-h this.prox child.prox this.obv dog-obv ‘It is this child that likes this dog’ b. * uyuuh atim-h uu awaash this.obv dog-obv this.prox child.prox ‘It is this child that likes this dog’ miyeyim-e-u like.ta-dir-3 miyeyim-e-u like.ta-dir-3 In Plains Cree (Algonquian), this restriction is even more pronounced: even if only one noun phrase is in non-neutral position, the obviative cannot precede the proximate (17a-b): (17) Plains Cree (Algonquian) (Bliss 2005:61) a. Am-a nínaa-wa íístini-m an-i í’ksisako-yi dem-prox man-prox cut-dir dem-obv meat-obv ‘The man cut the meat’ b. *An-i í’ksisako-yi íístini-m am-a nínaa-wa dem-obv meat-obv cut-dir dem-prox man-prox ‘The man cut the meat/The meat was cut by the man’ 318 Anna Berge Thus, nominal marking in many North American languages very clearly has an important role to play in information structure, distinct from word order yet obviously related. Each nominal marking system makes different distinctions which appear to correlate the default nominal marking (case or person) with topic; but a satisfactory and comprehensive cross-linguistic comparison of topic and focus remains to be done. 13.3.3 Polysynthesis Languages differ in how much information they pack into individual words, and in the types and methods of packaging that information. If words can be analyzed as having component parts, they can be thought of as synthetic, which means ‘put together’. Some languages allow or require quite a few component parts to be put together to form words, in which case they are known as polysynthetic. (18) illustrates polysynthesis in Wichita: (18) Wichita (Caddoan) (Mattissen 2004:204, citing Rood 1976:75) kiya:kí-riwa:c-ʔáras-a-ri-kita-ʔa-hí:rik-s evid.aor:3-big-meat-coll.u-tr-top-come-iter-impf ‘by making many trips, he carried the large (quantity of) meat up into it [the tree]’ Although polysynthesis is found elsewhere, it is particularly well represented in the Americas. There are different understandings of what constitutes polysynthesis (Mithun 2009, 2017), and there are different types of polysynthesis, but key to the discussion here is that many ideas are packaged into a single word, and this must have an implication for word order, sentence structure, and so forth. If only one unit of new information is preferentially introduced per sentence, how does this work in polysynthetic languages, and what is a unit of information? Is a word a clause, and is a sentence (one or more clauses) the same across languages? In some ways, polysynthesis may not be as problematic to information structure as it seems. In terms of the units of meaning in a clause, even in a highly polysynthetic structure, many of the elements of meaning give flavor to the clause, but the actual new content may not exceed what we otherwise find in non-polysynthetic clauses. Although (18) is decontextualized, the glosses suggest that most of the information is grammatical, leaving ‘big meat’ and ‘come’ as the places to look for new information. As for word order, a one-word clause may include a topic indexed on the verb and proposition-level focus; and many clauses in fact have multiple words and phrases that indicate relative levels of focus or backgrounding. In other ways, polysynthesis presents some interesting challenges to theories of information structure. For example, sentences in polysynthetic languages frequently consist of chains of clauses, and therefore tend to resemble paragraphs, rather than the short, decontextualized examples found in this article. They may reflect topic continuity in discourse (Berge 2011), but information structure is specifically supposed to be a Information structure 319 feature of the syntax. Another characteristic of polysynthesis is the ability to incorporate independent words into a verbal structure, as in the independent noun aniqdux̂ ‘child’ as in (3a), or ʔáras ‘meat’ in (18). Incorporation therefore affects the flow of information. It may be a mechanism for introducing a new topic (Berge 2011; Mithun 2009); but not all incorporated nouns are new, nor are they all topical (Mithun 1983; Berge 2011). Incorporation is one of many options for managing information. Both the effects of clause chaining and incorporation on information structure need more investigation. 13.4 Conclusions Information structure is an important, if still imperfectly understood part of linguistic knowledge. Time and again, concepts such as neutral and non-neutral word orders, topic tracking, strategies to focus noteworthy information, and choices in the packaging of information within words and sentences have been shown to have explanatory value in understanding the grammar of Indigenous languages, often capturing insights that do not seem to be answered by appeal to syntax or morphology alone. There appear to be some interesting correlations or trends that need further investigation: for example, there is a preference for a verbal argument to the right of the sentence to be given information, or backgrounded, rather than new; nominal marking, regardless of the marking system, seems to mark topic in the languages illustrated here; and even in highly polysynthetic languages, patterns such as limiting the amount of new information in a clause are evident. Information packaging is complex, and as important as cross-linguistic studies are in improving our understanding of information structure, we also need to expect and study variation in information structure within a given language. For example, because of the linear nature of speech, all languages have strategies to manipulate word order for effect. We can therefore expect there to be neutral word orders and unexpected word orders that will catch the attention of the listener in some way. A good speaker will play with his or her options to do so, and we can therefore expect differences in information structure between speakers of a language, with more eloquent or creative speakers using a wider variety of strategies, or with some speakers preferring some strategies over others. Therefore, true knowledge of a language must include an understanding of that language’s information packaging strategies. In addition to cross-linguistic and language internal differences in information structure, we can expect differences in information structure across time within a language, as a result of either changes in notions of aesthetics, or perhaps of changes in other parts of the grammar, either through language internal developments or as a result of language contact. Change is inevitable, it is a healthy part of language adaptation through time. But information structure may also be affected by language loss and replacement, with the structure of a dominant or prestigious language affecting that of 320 Anna Berge the affected language. As such, further studies of information structure in Indigenous languages are desperately needed, especially for language maintenance and revitalization. information structure has rarely been prominent in discussions of language revitalization and the teaching of heritage languages, and efforts have generally focused on learning vocabulary and the generation of simple conversations. But learning a language also requires learning its unique approach to packaging concepts, or the way speakers manipulate the language for effect. References Beck, David. 2010. Communicative Structure in Lushootseed Syntax: Thematicity and Focalization. In José Camacho, Rodrigo Gutiérrez-Bravo & Liliana Sánchez (eds.), Information Structure in Indigenous Languages of the Americas: Syntactic Approaches, 39–64. Berlin & Boston: De Gruyter Mouton. Berge, Anna. 2009. Tracking topics: A comparison of “topic” in Aleut and Greenlandic discourse. In Marc-Antoine Mahieu & Nicole Tersis (eds.), Variations in Polysynthesis: The Eskaleut Languages, 185–200. (Typological Studies in Language 86.) Amsterdam: John Benjamins. Berge, Anna. 2011. Topic and Discourse Structure in West Greenlandic Agreement Constructions. Lincoln: University of Nebraska Press. Berge, Anna. 2013. Object Reduction in Aleut. Transitivity and Its Related Phenomena. Asian and African Languages and Linguistics 7. 5–23. Berge, Anna. 2014. A Comparison of Information Structure in Greenlandic and Aleut. Workshop on Information Structure in Head Marking Languages, Nijmegen, March 28–29. Berge, Anna. 2016. Pribilof Anĝaĝigan Tunungin/The Way We Talk in the Pribilofs. Fairbanks: Alaska Native Language Center. Berge, Anna. 2016. Polysynthesis in Aleut (Unangam Tunuu). In Tokusu Kurebito (ed.), Linguistic Typology of the North 3, 1–38. Tokyo: ILCAA. Bergsland, Knud. 1997. Aleut Grammar. Fairbanks: Alaska Native Language Center. Bliss, Heather. 2005. Topic, Focus, and Point of View in Blackfoot. In John Alderete, Chung-hye Han & Alexei Kochetov (eds.), Proceedings of the 24th West Coast Conference on Formal Linguistics, 61–69. Somerville, MA: Cascadilla Proceedings Project. Carol M. Eastman & Elizabeth A. Edwards. 1983. Pragmatic factors and Haida syntax, Word 34(2). 57–65. DOI: 10.1080/00437956.1983.11435737. Dahlstrom, Amy. 2003. Focus Constructions in Meskwakwi (Fox). In Miriam Butt & Tracy Holloway King (eds.), Proceedings of LFG03 ConferenceUniversity at Albany, State University of New York, 145–163. Stanford, CA: CSLI Publications. De Reuse, Willem. 2001. Topic, focus, definiteness, specificity, and referential prominence in Western Apache noun phrases and relative clauses. In Jeanie Castillo (ed.), Santa Barbara Papers in Linguistics 11, Proceedings from the fourth Workshop on American Indigenous Languages July 6–8, 2001, 13–19. Santa Barbara: UCSB Department of Linguistics. Du Bois, John W. 1987. The Discourse Basis of Ergativity. Language 63/4. 805–855. Eastman, Carol M. & Elizabeth A. Edwards. 1983. Pragmatic factors and Haida syntax. Word 34(2). 57–65. DOI: 10.1080/00437956.1983.11435737. Firbas, Jan. 1964. On Defining the Theme in Functional Sentence Analysis. In Josef Vachek (ed.), Travaux Linguistiques de Prague 1: L’École de Prague d’Aujourd’hui, 267–280. Prague: Editions de l’Académie Tchécoslovaque des Sciences. Givón, Talmy. 1983. Topic Continuity in Discourse. Amsterdam: Benjamins. Information structure 321 Gundel, Jeanette K. 1988. Universals of topic-comment structure. In Michael Hammond, Edith A. Moravcsik & Jessica R. Wirth (eds.), Studies in Syntactic Typology, 209–240. Amsterdam: John Benjamins. Guerrero, Lilián & Valeria A. Belloro. 2010. On word order and information structure in Yaqui. In José Camacho, Rodrigo Gutiérrez-Bravo & Liliana Sánchez (eds.), Information Structure in Indigenous Languages of the Americas: Syntactic Approaches, 117–139. Berlin & Boston: De Gruyter Mouton. Handschuh, Corinna. 2014. A Typology of Marked-S Languages. (Studies in Diversity Linguistics 1). Berlin: Language Science Press. http://langsci-press.org/catalog/book/18. Harbour, Daniel, Laurel J. Watkins & David Adger. 2012. Information Structure, Discourse Structure, and Noun Phrase Position in Kiowa. International Journal of American Linguistics 78(1). 97–126. Johns, Alana & Ivona Kučerova. 2017. On the Morphosyntactic Reflexes of Information Structure in the Ergative Patterning of Inuit Language. In Jessica Coon, Diane Massam, Lisa Demena Travis (eds.), The Oxford Handbook of Ergativity, 397–418. Oxford: Oxford University Press. DOI 10.1093/ oxfordhb/9780198739371.013.17. Junker, Marie-Odile. 2004. Focus, Obviation, and Word Order in East Cree. Linga 114. 345–365. Krifka, Manfred. 2006. Basic Notions of Information Structure. In Caroline Féry, Gisbert Fanselow & Manfred Krifka (eds.), Interdisciplinary Studies on Information Structure 06, 13–56. (ISIS/Working Papers of the SFB 632). Potsdam: Universtitätsverlag Potsdam. Lambrecht, Knud. 1994. Information Structure and Sentence Form: Topic, Focus, and the Mental Representation of Discourse Referents. Cambridge: Cambridge University Press. Matthewson, Lisa. 2017. Semantics in Indigenous American Languages: 1917–2017 and Beyond. International Journal of American Linguistics 83(1). 141–172. Mattissen, Johanna. 2004. A Structural Typology of Polysynthesis. WORD 55(2). 189–216. Mithun, Marianne. 1983. The Genius of Polysynthesis. In James S. Thayer (ed.), North American Indians: Humanistic Perspectives, 221–242. (University of Oklahoma Papers in Anthropology 24.2). Norman, OK: University of Oklahoma Press. Mithun, Marianne. 1992. Is basic word order universal? In Doris L. Payne (ed.), Pragmatics of Word Order Flexibility, 15–61. (Typological Studies in Language 22). Amsterdam: John Benjamins. Mithun, Marianne. 2009. Polysynthesis in the Arctic. In Marc-Antoine Mahieu & Nicole Tersis (eds.), Variations on Polysynthesis: The Eskimo-Aleut Languages, 3–18. (Typological Studies in Language 86). Amsterdam: John Benjamins. Mithun, Marianne. 2018. Factors behind variation in marking information structure: Contributions from Central Pomo. In Evangelia Adamou, Katharina Haude & Martine Vanhove (eds.), Information Structure in Lesser-Described Languages: Studies in Prosody and Syntax, 119–156. Amsterdam: John Benjamins. https://doi.org/10.1075/slcs.199.05mit. Tersis, Nicole & Shirley Carter-Thomas. 2005. Integrating syntax and pragmatics: word order and transitivity variations in Tunumiisut. International Journal of American Linguistics 71(4). 445–472. Thompson, Chad L. 2000. Iconicity and Word Order in Koyukon Athabaskan. In Theodore B. Fernald & Paul R. Platero (eds.), The Athabaskan Languages: Perspectives on a Native American Language Language Family, 228–251. (Oxford Studies in Anthropological Linguistics). Oxford: Oxford University Press. Vallduví, Enric & Elisabet Engdahl. 1996. The Linguistic Realization of Information Packaging. Linguistics 34. 454–519. Zúñiga, Fernando. 2014. Inversion, Obviation, and Animacy in Native Languages of the Americas: Elements for a Cross-Linguistic Survey. Anthropological Linguistics 56. 3/4. 334–355. 322 Anna Berge Further Reading Dalrymple, Mary & Irina Nikolaeva. 2011. Objects and Information Structure. Cambridge: Cambridge University Press. Foley, William. 2007. A Typology of Information Packaging in the Clause. In Timothy Schopen (ed.), Language Typology and Syntactic Description, 362–446. 2nd edn. Cambridge: Cambridge University Press. Tim Thornes 14 Clause-combining: Relative clauses Abstract: This chapter presents a survey of constructions in the languages of Native North America that could properly be considered relative clauses. Relative clauses are subordinate clauses or clause-like structures that describe nouns and either fully occupy the position of a noun phrase in a sentence or appear as dependents of a noun phrase, either modifying or delimiting the head noun or associated statement. From the perspective of language acquisition, one may consider the primary function of a restrictive relative clause as assisting the speaker in directing the attention of the listener to a particular entity or type—an identifying function. 14.1 Introduction: What is a relative clause? A language requires some means for speakers to identify and describe entities in a discourse. For example, we can imagine two people in conversation at a busy dog park, and one person says, “That dog looks thirsty!” The other person asks, “Oh! Which one?” “The black lab that’s standing near the gate over there,” is the response. The questioner’s request for help in narrowing down the list of possible referents of “that dog” has led to a response that supplies more specifics. These specifics include breed (it’s a lab) and color (black), but also, as a busy dog park no doubt has more than one black lab in it, a statement of the posture and location (it’s standing near the gate) of the referent “dog.” By specifying something about the dog—in this case, where it is and what it is doing— attention has been directed to a particular dog. The dog is identified in this case not only by color, breed, etc., but by relating it to a proposition. This proposition is what is known as a relative clause, indicated between [square brackets] in example (1a): (1a) {The dog [that’s standing near the gate over there]} looks thirsty. In this example, we consider the noun “dog” to be the head noun (in boldface), since it is the entity (or referent) that is being described by the relative clause. Head nouns have two roles in a complex sentence containing a relative clause, 1) the role of the head noun inside the relative clause and 2) its role in the main clause. The head noun plus relative clause together form a {noun phrase}, surrounded by curly brackets.1 It is important to note that sentence (1a) consists of essentially two statements expressed as two clauses. These include a relative clause (a type of subordinate or dependent clause) whose function is to help the listener identify a specific dog, and a main clause whose function is to convey the primary information (to assert something) about the dog so identified (i. e. that it looks thirsty). 1 We will treat the determiner “the” as part of the noun phrase {NP} as well. https://doi.org/10.1515/9783110600926-014 324 Tim Thornes Notice that the entire noun phrase, including the relative clause, serves in the role of subject of the main clause, which expresses the main statement and includes a verb, “looks,” and a predicate adjective, “thirsty.” In most cases, the role of the complex noun phrase (the one containing the relative clause) in the main clause does not make a difference in the grammatical form of the relative clause itself. Note: (1b) Do you see {the dog [that’s standing near the gate over there]}? The noun phrase in curly brackets is identical in form to the one in example (1a), despite the fact that it is serving in the role of object of the main clause verb “see” in example (1b). In both examples, one imagines that the speaker felt that this additional information regarding the dog was necessary in order to both describe and identify one particular dog to the listener. Consider, by comparison, the following example from the Siouan language, Hidatsa: (2) Hidatsa (Missouri River; Siouan) (Park 2012:517) { masúga [agu–mii–náhci–s] } adáàsi–hgua dog rel–1b–bite–def outside–loc ‘The dog that bit me is outside.’ háhgu–c be.around–decl In (2), masúga ‘dog,’ is the head noun, followed by the complex verb form aguwiiráhcis operating as a relative clause meaning ‘(one) that bit me.’ Perhaps the context involves my friend wondering why I am nervous about leaving his house to go home, and so I offer him (2) by way of explanation. At any rate, the relative clause narrows the referential range of all possible dogs to a particular one. We can consider masúga as playing a role in both the relative clause (the biter) and the main clause (the one outside). It is the role in the former that is important here, as that is the role that determines what type of relative clause it is (see section 14.3). In some cases, we find relative clauses that do not serve an identifying function by narrowing reference to a specific entity, but instead simply provide additional information about an entity whose identity is already known. Consider: (3) Felix, [who works at the capitol], has been lobbying for the latest farm bill. In this example, the proper noun, Felix, does not require further description to assist in identifying him, since we know his name. The relative clause in square brackets in example (3) is what is known as a non-restrictive relative clause type, since the relative clause construction here does not serve the purpose of restricting the range of possibilities to some entity. That is, reference to Felix’ work is not required in order for us to know who has been lobbying for the farm bill (although it may explain why he has). We do not put Felix in bold, since it is not, properly speaking, the head of a noun phrase that includes the non-restrictive relative clause.2 By comparison, the relative clause type 2 Also, being set off by commas, it is not part of the same intonational phrase as the would-be head noun. Clause-combining: Relative clauses 325 exemplified in (1a-b) and (2) is called a restrictive relative clause. It is this latter type that we will be most concerned with in this chapter. As a brief note of caution, I think it important to acknowledge the influences of methodology on relative clause structures as they are represented and described in the languages of Native North America. The practice of direct elicitation, which usually involves asking speakers for the translational equivalents of engineered sentences in English, say, into the target language, increases the possibility of calques from English. That is to say, when one presents for translation an English example of a relative clause, it increases the likelihood that the translation equivalent will share grammatical properties in common with English. On the other hand, in searching a corpus of spontaneous speech for examples of relative clause structures, many of those studying Native North American languages find them to be either rare or absent. 14.2 Parameters of variation This chapter presents a survey of grammatical constructions in the languages of Native North America that are or could be properly considered relative clauses or, alternatively, serve the typical functions associated with relative clauses. Although we find a great deal of variation across the languages of the continent, this variation occurs within a limited set of parameters. We have already seen a couple of these parameters of variation at play in the discussion of examples in the introduction. Here are a few of the ways in which relative clauses may vary grammatically from language to language: 1) the position of the head noun vis a vis the relative clause, 2) the presence or absence of a special relativizer or relative pronoun, 3) the use of a specialized verb form or verbal morphology particular to relative clauses, 4) how the grammatical role of the head noun within the relative clause is identified (“recovered”), and 5) whether there are restrictions on tense or other distinctions that can be coded in the relative clause, in other words, the degree to which relative clauses resemble main clauses in finiteness (section 14.4). We will draw some connections between these parameters, since they interact with each other within the ecology of the various languages. Some theoretical issues will also be touched upon, but the goal of the present chapter is to lay a groundwork of facts, not a settlement of theory-specific concerns. For the student and teacher of Native North American languages, I hope that this survey provides for a deeper understanding of this complex area of grammar while also advancing possibilities for teaching and learning these interesting languages. 326 Tim Thornes 14.2.1 Presence and position of the head noun 14.2.1.1 External heads One parameter of variation is the position of the head noun with respect to the relative clause. In our English examples above, the head noun always comes before the relative clause (i. e. to its left), and the relative clause may therefore be called a left-headed relative clause.3 The examples below also demonstrate this feature: (4) Northern Paiute (Uto-Aztecan) (Thornes 2003:464) { umɨ [kai u= pidzabi–dɨ] } ɨmɨ–nɔ tuʔi they not 3sg:acc= like–nmlz them–with try ‘Those that didn’t like it would try to fight with them.’ (5) Creek (Muskogean) (Martin 2011:392) { axéy ifá [ a:–hóyL–a:ti–t ] } lopéyc–i:–t that dog dir–stand.fgr–ref–t nice–dur–t ‘That dog standing over there is friendly.’ (6) Okahoma Cherokee (Iroquoian) (Montgomery-Anderson 2008:523) { a–skaya [ ji–jii–ali–hnohehtiisk–vv’i ] } a–ahnika 3a–man rel–1a.an–mdl–talk.with:inc–exp/sub 3A–leave:imm ‘The man that I was talking to left.’ na–koiwɨnai–ʔyakwi mm–fight.against–hab ó:–s be.fgr–ind The example in (4) has the pronoun umɨ ‘they’ acting as the noun phrase head whose identity is restricted to some subset of those present (in this case, those who were attending an Indian boarding school). In (5) the subject noun phrase of the main clause (in curly brackets { } ) headed by ifá ‘dog’ has a modifying relative clause (in square brackets [ ] ) a:hóyLa:tit ‘standing over there.’ Head nouns can be understood to play two roles in these complex sentences: 1) their grammatical role in the main clause and 2) their role in the relative clause. In both (4) and (5), the head nouns in bold are playing the role of subject in both the main clause and their respective relative clauses. This is not a necessary circumstance, however. In (6), the head noun askaya ‘man’ is followed by the restrictive relative clause jijiialihnohehtiiskvv’i ‘that I was talking to’ in square brackets. The head noun of the complex noun phrase ‘the man that I was talking to’ is playing the role of indirect object in the relative clause, while the entire noun phrase headed by askaya is serving the role of subject of the main clause centered on the verb of the main clause aahnika ‘he left.’ 3 Many authors prefer to call these post-nominal relative clauses, based on the position of the relative clause with respect to the head noun. Since much of our focus is on the presence or position of the head noun, our preference will be to label relative clause types with respect to the property, or parameter, of “headedness.” Clause-combining: Relative clauses 327 Although our examples thus far are left-headed, it turns out that headedness is a flexible concept in Cherokee, which allows for right-headed relative clauses as well, as in (7): (7) Okahoma Cherokee (Iroquoian) (Montgomery-Anderson 2008:523) { [ ji–kinii–atuuliisk–oi ] altiithla } rel–1b.dl–want:cmp–hab\sub car kaayuula khilo uu–hwas–ei already someone 3b–buy:cmp–nxp ‘The car we want has already been bought.’ lit. ‘The car that we want, someone already bought it.’ The right-headed relative clause type is found much less frequently in the languages of Native North America. Most examples I have found that illustrate this type demonstrate it as an alternative to the left-headed variety, as we have seen with Cherokee (and is reported as rare, but possible, in Creek, cf. Martin 2011). Ktunaxa, widely known as Kutenai,4 a language isolate5 of the northern Idaho panhandle and neighboring British Columbia, favors right-headed relative clauses as in (8): (8) Ktunaxa (isolate) (Dryer 2007:202) wiɬqaʔ–ni {[niʔ k=u big–indic def subord=1.subj ‘the woman that I saw was tall’ wukat] see paɬkiy} woman Notice here that the head noun paɬkiy ‘woman’ follows to the right of the relative clause. Unlike Cherokee, flexibility of head position appears not to be a characteristic of relative clauses in Ktunaxa. Tlingit (Na-Dene) represents another case whereby right-headedness has been analyzed as strict (Crippen 2012)6. To be clear, a strong preference for right-headed relative clauses is not at all uncommon in the world’s languages. Rather, it represents something of an anomaly in the indigenous languages of North America. 14.2.1.2 Internal heads In most cases of left- versus right-headedness, there is an assumption that the head noun is outside of the relative clause (external), as we have seen, especially where there is flexibility with respect to the order of the head and the relative clause, as in Cherokee. There are, however, a significant number of languages of Native North America that 4 The preferred name for the language is Ktunaxa (Matthew Dryer, in personal communication). 5 A language isolate is one for which a relationship to other languages in a language family has not been established or is in question. 6 I thank Ashlyn Nutting for alerting me to the Tlingit data and analysis in Crippen (2012). 328 Tim Thornes offer examples of head nouns appearing inside the relative clause (internal). The following examples from Osage (Siouan), Navajo (Athabaskan), and Jamul Tiipay (Yuman) demonstrate the internally-headed relative clause type. (9) Osage (Siouan) (Quintero 2004:467) {[sitó̩i̩ níhkaši wi̩ a–ø–chí–api–ðe]} yesterday man a prev–a3s–arrive.here–pl–decl John–a akxaíi–ø–ø–ðe–api–ðe John–syl subj prev–p3s–a3s–see–pl–decl ‘John saw the man that came yesterday’ (10) Navajo (Athabaskan) (Platero 1974:204) {[tl’éédá̩á̩’ ashkii aLhá̩á̩’–á̩á̩]} last.night boy 3sg.impfv.snore–rel.past ‘The boy who was snoring last night will speak.’ (11) Jamul Tiipay (Yuman) (Miller 2001:210) {[Manweel kaamiis txiil]}–pu nyaach Manuel shirt wear–dem i+sj ‘The shirt Manuel is wearing, I sewed it.’ yádooLtih fut.3sg.speak shuukwil sew In (9), the head noun níhkaši ‘man’ appears surrounded by other elements of the relative clause. This is also the case with ashkii ‘boy’ and kaamiis ‘shirt,’ in (10) and (11), respectively. One characteristic of relative clauses with internal heads is that, in most cases, the head nouns appear in the same position inside the relative clause as they would in an ordinary main clause, e. g. before or after the predicate. Chimariko, a language isolate from northern California, also has internally-headed relative clauses, as is most clearly in evidence in example (12), although the source analyzes (13) as an example as well. (12) Chimariko (isolate) (Jany 2013:433) {[mo’a phuncar h–uwa–tku–rop ]} yesterday woman 3–go–dir–dep ‘That woman who came yesterday told me.’ (13) Chimariko (isolate) (Jany 2013:433) {[h–iman–tamo–rop map’un ]} h–i’am–ta 3–fall–dir–dep that.one 3–beat–der ‘Those fellows that went down got beaten.’ pha’yi–nip thus.say–pst These examples also illustrate flexibility in the placement of the head noun within the relative clause as preceding the relative clause verb huwatkurop in (12) while following the verb himantamorop in (13). In both cases, the head noun bears the intransitive subject relation to the relative clause. According to Jany (2013:434), elements within a noun phrase in Chimariko show significant flexibility, while still remaining together as a Clause-combining: Relative clauses 329 syntactic unit, or constituent. The dependent marking suffix –rop consistently indicates the status of the verb as that of a relative clause. Flexibility is also noted in the case of Jamul Tiipay, which allows “pre-posing” (movement to the beginning of the clause) of head nouns, as in the following: (14) Jamul Tiipay (Yuman) (Miller 2001:210) { [vineen nyaap wich]–pe–ch } m–aaxway–x–s poison me+abs have–dem–sj 3/2–kill–irr–emp ‘The poison that I have will kill you.’ By comparing the position of the object here to that in (11) above, we see that the object “poison” precedes the subject in (14), as opposed to the object “shirt” following it in (11). Languages with flexible word order—and there are many such languages in Native North America—typically use variable word order for pragmatic, rather than strictly grammatical, purposes (see Broadwell (this volume) and Berge (this volume)). Highlighting the relative importance of certain characters in the discourse, for example, is one common function of flexibility in word order. 14.2.1.3 Headlessness Finally, we will consider relative clause structures that appear without an overt head noun—what are commonly called either headless relative clauses or, alternatively, “free” relatives. In these cases, the relative clause itself is sufficient for narrowing down the range of possible referents, as in the following examples from English and Northern Paiute: (15) I’ll have { [what she’s having.] } (16) Northern Paiute (Uto-Aztecan) (Thornes 2003:502) { [ tɨ=tɨ–da–kwɨhɨ–na ] } pɨnau owi–tu hani–ʔyakwi logo= apass–IP/feet–get–ptcp again there–to do–hab ‘. . what it grabs (with its claws), (it) brings back to there.’ In example (15) the reference to “what she’s having” narrows the range of choices (e. g. at a restaurant) to another person’s order. In (16), from a traditional narrative about a flying monster that carries people away, the reference is narrowed to what or whoever is captured. Most of the languages surveyed allow for headless relative clause structures. Examples (17)-(21) illustrate the diverse range of languages that share this property. (17) Blackfoot (Algonquian) (Franz 1991:129) n–Imaat–ssksino–a:–yi {[om–iksi k–omoht–yIIstap–oo–at–a:–iksi]} 1–neg–know–dir–3pl that–3pl 2–from–away–go–fin–dir–3pl ‘I don’t know those you went away from.’ 330 Tim Thornes (18) Saanich (Salishan) (Montler 1993:260) kwən–nəxw = sən {[kwsə tɬ’itɬ’əw]} see–trans = 1sg:subj dem escaping ‘I saw the one who was getting away.’ (19) Wappo (isolate; proposed Yukian) (Thompson, et.al. 2006:122) {[maʔa mi thal mes–ta]} ah paʔe–siʔ just 2sg what make–pst:dep 1sg:nom eat–fut ‘I’ll eat whatever you made.’ (20) Creek (Muskogean) (Martin 2011:393) {[iLki–acól–i léyk–a:t]} L–óLho:y–ati:–s 3pat.father–old–I sit.sg.fgr–ref dir–reach.du.lgr–past5–ind ‘They got to where their elderly father lived.’ (21) Ktunaxa (isolate) (Dryer 2007:202) wiɬqaʔ–ni {[niʔ k=u wukat]} big–indic def subord=1.subj see ‘the person that I saw was tall’ or ‘the thing that I saw was big’ For some languages, like Blackfoot and Northern Paiute (according to Franz 1991 and Thornes 2012, respectively), it is reported that clearly headed relative clauses appear only rarely in natural discourse.7 We turn now to another feature of variation across relative clause structures, namely, the use of a specialized relativizer or relative pronoun, usually to introduce a relative clause but also correlating with other features of grammatical expression. 14.2.2 Presence of a relativizer or another boundary identifier The presence of a special relativizer or relative pronoun often serves to identify the boundary of a relative clause and, in some cases, to determine the role of the target entity within the relative clause. Looking back at the English examples above, “that” serves as relativizer in (1a) and (1b), while “who” is a relative pronoun in example (3). The main difference between the two is that a relativizer is generally fixed in form, typically serving only to indicate the presence or boundary of a relative clause. A relative pronoun, on the other hand, is sensitive to other features, like animacy, humanness, or number, that are assigned to the head noun. Note, for example, that while you can say “the man that” or “the man who” 7 It has been suggested that there may be a correlation between a high frequency in the use of headless relative clauses and the nominalization strategy discussed in section 14.4.2 (cf. Comrie and Estrada-Fernandez 2012). Clause-combining: Relative clauses 331 in English, one cannot say “*the house who,” since a house is not animate. The invariant relativizer “that” can be used with any head noun type, but the relative pronoun changes form, depending upon these and other features. What counts as animate, or human, is a judgment that speakers of English make all the time. For some, “the dog who” is okay (dogs being closely associated with humans), but “the starfish who” is strange (despite being technically animate), unless one is referring to an animated character in a movie with human-like characteristics. Yakima Ichishkíin, a Sahaptian language of the Columbia Plateau region, has subordinate clauses, including relative clauses, which appear to be identical to an ordinary main clause (cf. Jansen, 2010:396–397). The only distinction from main clauses is the presence of a subordinating prefix (or clitic) ana– (what we could consider a relativizer REL), as in the following: (22) Yakima Ichishkíin (Sahaptian) (Jansen 2010:400) ku awkú pa–tamápaysh–ta {[ana–kuunak húuy and then 3sg.s –report.on–fut rel–obj in.vain pá–talax̱itk–sha]} inv–discipline–impv ‘And then they will report on that one who is disciplining them in vain.’ This boundary phenomenon appears to be the minimal requirement for identifying relative clauses as well as other subordinate clause types. Note that ana– in (22) attaches to the object pronominal kuunak to form a kind of relative pronoun translating roughly as ‘the one that.’ We see it elsewhere attached to interrogative pronouns like ‘who’ or ‘where’ as in the following: (23) Yakima Ichishkíin (Sahaptian) (Jansen 2010:399) {[ana–shín i–wiláalakw–ta pɨnmikínk tmíyu–t–ki]} rel–who 3sg.s–outrace–fut 3sg.pn.inst plan–nzr–inst awkú kushk i–wá–ta then winner 3sg.s–cop–fut Whoever comes up with the best plan (to create a night) will be the winner.’ In keeping with the English translation, ana– could be interpreted as ‘–ever’ as with ana–shín ‘whoever,’ ana–mún ‘whenever,’ and ana–miɬ ‘however many,’ among others. Notice, too, that both (22) and (23) represent examples of headless relative clauses. In Crow (Siouan), there are two “basic relativizers,” according to Graczyk (2007:254), which also appear as prefixes to the verb of the relative clause. Their forms are sensitive to whether or not the subject versus some other notional role is being played by the head noun in the relative clause. For subject relative clauses, the relativizer ak– is used, as in the following: 332 (24) Tim Thornes Crow (Siouan) (Graczyk 2007:253) {hinne bacheé–m [ak–óopiia–sh]} is–bilé this man–det rel–smoke–det 3:poss–fire ‘this man who was smoking’s fire was burning down’ awá–ss–dee–m earth–goal–go–ds Here, we can identify bacheém ‘the man’ as the one doing the smoking, the subject of the relative clause. It is isbilé ‘his fire’ that is the subject of the main clause verb awássdeem ‘going (burning) down.’ Recall that it is the notional role of the head noun in the relative clause that typically affects various aspects of the grammar of relative clauses. As another example of the use of ak–, consider the following headless relative clause: (25) Crow (Siouan) (Graczyk 2007:253) {[Baáhpuuo ak–kukaa–húua–sh]} Pryor rel–source–come–det ‘the one who came from Pryor…’ Despite the absence of a true head noun in (25), we still understand the relative clause to be identifying a target entity playing the role of subject with respect to the verb ‘come.’ Not all relative clause types require a relativizer in Crow, however. In many cases, like the following object relative clause, it is sufficient to use a determiner suffix to mark the right boundary of a relative clause: (26) Crow (Siouan) (Graczyk 2007:262) {shikáaka–m [xapíi–o–sh]} kuú–k boy–det lost–caus.pl–det come.back–decl ‘the boy that they lost has come back’ In (26), the definite determiner (det) –sh marks the boundary of the noun phrase whose head noun is shikáakam ‘the boy’ playing the role of object to the relative clause verb ‘lost.’ Note that the complex noun phrase shikáakam xapíiosh ‘the boy they lost’ is subject of the main clause verb kuúk ‘come back.’ Slave, a Northern Athabaskan language, uses separate words that Rice (1989) refers to as complementizers both to signal the final boundary of a relative clause and to indicate whether or not the head noun is definite. By definite, we mean that something is known to both the speaker and the hearer. Compare the following examples: (27) Slave (Athabaskan) (Rice 1989:1320) a. {ɬue [ɬan˜ihdé síi]} seghán˜ilã fish 3.killed.3pl:obj comp 3.gave.1sg>3pl:obj ‘he gave me the fish he killed’ b. {ɬue [ɬan˜ihdé líi]} seghán˜ilã fish 3.killed.3pl:obj comp 3.gave.1sg>3pl:obj ‘he gave me whatever fish he killed’ Clause-combining: Relative clauses 333 The complementizer síi in (27a) signals that the head noun is definite (referring to some particular, known bunch of fish), while líi in (27b) indicates that it is indefinite (e. g. an unknown quantity or a variety of fish that is involved). Montgomery-Anderson (2008) reports that Oklahoma Cherokee represents a somewhat unusual case of boundary marking with regard to relative clauses. Cherokee relative clauses appear with both a relativizer (REL) (formally referred to as a “prepronominal prefix,” ji–) and a distinctive high-fall tone on the rightmost long vowel of the relative clause verb. Although tone is not an unusual property in the indigenous languages of North America (see Uchihara, this volume), this particular grammatical use of tone is otherwise unattested, as far as I am aware, for these languages.8 14.3 Relative clause types We began this chapter by looking at language-particular features in section 14.2, where we explored some basic ways in which relative clause formation can differ from language to language. Now we will look at how these features interact with one another to help determine what role the key referent (either in the form of a head noun or the understood target entity of a headless relative clause) is playing within the relative clause itself. Several times in our discussion, we have made a point of identifying the grammatical role of the head noun within the relative clause. This role is distinct from the role of the complex noun phrase in the main clause. This duality of roles is important to track, as it is most often the case that the role of the head noun (or target entity in the case of headlessness) is essential for describing the form and function of relative clauses. To clarify, consider the following English examples: (28) a. b. {The dog [that chased the cat]} bit me. {The cat [(that) the dog chased]} got away. Subject relative clause Object relative clause These represent two distinct relative clause types, a subject relative clause (the head noun “dog” plays the role of subject of “chased”) in (28a) and an object relative clause (the head noun “cat” plays the role of object of “chased”) in (28b). In the subject relative clause of (28a), the relativizer “that” does the important work of marking the initial boundary of a relative clause “that chased the cat.” Note that without it, the sentence reads rather oddly, as one is led to believe that “the dog chased the cat” is the main clause. With the object relative clause in (28b), on the other hand, the relativizer is optional (and so appears in parentheses), and without it, we still recognize the role of the head noun in the relative clause, since “the cat” has been moved from its expected 8 The use of tone as an identifier of relative clauses is found elsewhere in the world, however, as in the Ebembe (Bantu; Niger Congo) language of eastern Congo and adjacent Tanzania (Iorio 2014:150 ff), where a verb’s final vowel in a relative clause carries high tone. 334 Tim Thornes object position following the verb “chased.” The change in the position of the object noun leaves what linguists often refer to as a gap in the relative clause-level grammar. This gap is one strategy English has for identifying the role of the head noun in the relative clause. The relative clause verb “chase” requires an object to follow it, which, in the case of (28b), precedes it as the head of the relative clause. In the languages of Native North America, several factors operate to minimize the use of the gap strategy. The prevalence of internally-headed and headless relative clauses entails either that the head noun itself occupies the syntactic position where one may find a gap in other languages, or that there is no head noun available for “gapping.” In the case of external heads, the preponderance are left-headed, which essentially leaves the head of a subject relative clause in its usual position in the main clause. Finally, the extensive cross-indexing of participants on the verbs of both relative and main clauses in many Native North American languages leaves open the question as to whether one ought to consider syntactic gaps as such since the capacity these indexing patterns have for avoiding ambiguity is evident in the morphology. Since the gap strategy is largely unavailable on both practical and theoretical grounds, most Native North American languages use different means for indicating the role of a head noun in the relative clause. Considering Yakima Ichishkíin examples (22) and (23) above, we see that the relativizing prefix ana– can be found attached to an interrogative pronoun that is itself sensitive to which participant is being relativized. In the earlier examples, the target entity for relativization was left underspecified, since the relative clauses were of the headless variety. As another example, this time of an internally-headed relative clause, consider: (29) Yakima Ichishkíin (Sahaptian (Jansen 2010:399) ku kuuk awkú á–sɨ́ nwi–ya and now then 3o–say–pst {[ana–túun sapsikwʼa–t i–náktux̱–ɨnm–a]} rel–what.obj teach–nmlz 3sg.s–carry.back–cisl–pst ‘that’s when he spoke the teaching he brought back’ Here, the nominalized verb sapsikw’at ‘teaching’ serves as the object of the relative clause. Again, we see the importance of classifying relative clauses for an understanding of many aspects of relative clause grammar as compared to basic, or main clause grammar. Studies of relative clauses, especially studies designed to make broad comparisons between and among languages from all over the world, have taken pains to see what sorts of limitations there may be on particular relative clause types (like subject, object, indirect object, etc.). Some languages only allow for subject relative clauses,9 9 This has been most famously described for Malagasy, an Austronesian language of Madagascar (Keenan and Comrie 1977). Clause-combining: Relative clauses 335 for example, while others, like English and Crow, allow head nouns to play nearly any grammatical role within the relative clause. Given the introductory nature of this survey and space restrictions, a more detailed comparison of how or whether such restrictions play out in the languages of Native North America will be set aside for another study. More important, for our purposes, is to explore the different ways various relative clause types may be expressed in the grammars of the individual languages. 14.4 Finiteness and nominalization Typical English relative clauses, like those in examples (1), (3), and (28), do not use special verb forms or harbor restrictions on indicating things like tense or aspect—what would otherwise distinguish them from main clauses. Instead, tense can be indicated within relative clauses, as in “the dog that’s standing” (present) versus “the dog that stood” (past) or “the dog that will stand” (future). Relative clauses that replicate many of the features, including tense marking, of main clauses are said to be finite, while those that show more restrictions in form are referred to either as less finite or even fully non-finite. Many linguists consider finiteness to be a scalar concept, meaning that there can be degrees of finiteness. As such, different subordinate clause types or relative clauses in different languages (related or not) may show variation in which or how many finite properties they share with independent clauses. 14.4.1 Finite properties in relative clauses By way of grasping the notion of finiteness further, consider that in English the form of the verb in a direct command, like ‘Go!’ or ‘Leave that alone!’ is almost completely non-finite. One cannot say ‘*Went!’ (past) or ‘*Will leave that alone!’ (future). Clausal complements in English illustrate variation in finiteness (see Dahlstrom, this volume, on subordinate clauses and complementation), as in ‘I told her to go’ or ‘I hope she will leave.’ The complement-taking verbs “tell” and “hope” have different finiteness requirements of their complements. The complement of “tell” requires an infinitive form of the verb. The complement of “hope,” on the other hand, cannot appear in the infinitive, but maintains the finite properties of the independent main clause, ‘she will leave.’ The forms of the pronouns that follow the complement-taking verbs are different as well. Following “tell” is “her,” the object form, behaving as the object of “tell.” Following “hope,” however, we have “she,” the subject form, in its role as subject of the complement clause. We could, on the basis of these two simple examples, conclude that the complement of “tell” is less finite than the complement of “hope.” That is, a hope-complement shares more properties with a fully finite main clause. 336 Tim Thornes Let us now consider finiteness with regard to relative clauses in Native North American languages. As examples of relative clauses that are highly finite, note the following from the Salishan language Saanich. (30) Saanich (Salishan) (Montler 1993:256) ʔəw’ Xchi–t = sən { kwsə asp know–trans = 1sg:subj dem ‘I know the man who hit you.’ swəy’qaʔ male [t’əm’–ə–sə ]} hit–trans–2obj (31) Saanich (Salishan) (Montler 1993:258) ʔəw’ Xchi–t = sən { kwsə asp know–trans = 1sg:subj dem ‘I know the man who you hit.’ swəy’qaʔ male [t’əm’–ət–əxw ]} hit–trans–2subj There appears to be little to distinguish relative clauses from main clauses in Saanich. Indeed, Montler (1993) prefers the more general term attributive construction in describing these construction types based on the function of these structures to attribute particular features to some entity. The difference between the subject relative clause in (30) and the object relative clause in (31) can only be determined by the verb agreement with a second person object in the former, leaving swəy’qaʔ as the subject, and a second person subject in the latter with swəy’qaʔ the object.10 Other Salishan language specialists assert that relative clauses are simply not present in some languages of the family, either because one does not find a unique syntactic structure that can be referred to as such—one that is distinct from other structures that perform attributing functions or from general clausal subordination—or that there is little, if any, structural difference between a relative clause within a main clause and a sequence of two main clauses. This is due in large part to the highly finite nature of subordinate clauses generally in Salishan languages. Thompson River Salish arguably has relative clauses of both the headed (32) and headless (33) varieties, according to Kroeber (1997). Consider: (32) Thompson River (Salishan) (Kroeber 1997:385) y’e–mín–ne { he seʔlís t ə [qwəz–t–exw ] } good–trz–1s.ts art knife att art use–trz–2s.ts ‘I like the knife that you use/used.’ (33) Thompson River (Salishan) (Kroeber 1997:385) pze–ne {ɬ [xwuy’ kn–ci–s] } meet–1s.ts art fut help–trz+2s.obj–3.ts ‘I met the one who is going to help you.’ 10 This can lead to potential ambiguities in the case of only third person participants as third person categories are typically unmarked. Clause-combining: Relative clauses 337 It is interesting that in headed relative clauses, both Saanich and Thompson River Salish appear to place the verbal predicate after other elements of the clause, while in main clauses, the preference appears to be verb initial. Whether this sort of pre-posing is true for relative clauses more generally in Salishan merits exploration.11 Athabaskan languages are also noted for having strongly finite subordinate, including relative, clauses. In Bearlake, a Northern Athabaskan language, object marking on the main clause verb is missing when the object position is filled by an object relative clause. (34) Bearlake (Athabaskan) (Rushforth and Gorbet 1989:457) ’Ehkee { tLi̩ [nadéhtLa] } ghái̩ i̩ dá boy dog 3leave.perf 3.see.perf ‘The boy saw the dog that left (to return).’ Without object-marking on ‘see,’ the complex noun phrase containing the relative clause must be interpreted as the syntactic object of the main clause. Navajo, another Athabaskan language, carries a relativizing suffix –á̩á̩ that both marks the final boundary of the relative clause and indicates past tense, as we saw in example 10). As a counterpart to –á̩á̩, the relativizing suffix –ígíí indicates that the time frame for the relative clause is the present, as in the headless relative clause in 35): (35) Navajo (Athabaskan) (Platero 1974:204) { [KinLání–góó deeyáh–ígíí] } bééhonisin Flagstaff–to 3sg.go–rel.nonpast 3sg(obj).imperf.1sg(subj).know ‘I know (the person) who is going to Flagstaff.’ Morphologically, the only difference between relative clauses in Navajo and main clauses appears to be the use of these tense-sensitive relativizing suffixes. Platero (1974:203) notes that the non-past form –ígíí also functions as a general nominalizer, perhaps placing Navajo relative clauses on a scale of slightly reduced finiteness. At the non-finite end of the scale, we can take into account those subordinate clauses that bear more resemblance to noun phrases than finite clauses. We turn now to nominalization proper, a key strategy for forming relative clauses in several languages and language families of Native North America. 11 A general preference for left-headedness may be due, in part at least, to the connection between relative clause and like structures and a discourse function like topic or focus (a la Schachter 1973). The fronting of constituents is a common strategy for indicating discourse prominence. 338 Tim Thornes 14.4.2 Nominalization as strategy Relative clause structures that are the least finite share numerous features one expects of a noun phrase, rather than a clause. These features are well-known from numerous typological studies (cf. Keenan and Comrie 1977, Givón 1990, 2012, and Andrews 2007). Noun- or noun phrase-like structures that derive from verb- or clause-like structures are referred to as nominalizations. A simple nominalization that changes the word class (part of speech) of a verb into a noun can be illustrated with the verb “teach” and the noun “teacher.” Adding the suffix –er to the verb results in a noun meaning “one who teaches.” Northern Paiute, an indigenous Uto-Aztecan language of the northern Great Basin region of the western U.S. also uses a suffix –dɨ for the same task, turning the verb tɨničui ‘teach; tell stories’ into the noun tɨničuidɨ ‘teacher; story-teller.’ Northern Paiute and other Native North American languages, however, extend the use of the nominalization process toward turning an entire clause into a complex noun phrase. The examples in (36a) and (36b) illustrate this process: (36) Northern Paiute (Uto-Aztecan) (Thornes 2003:428) a. su=tɨhɨča oʔo wɨnnɨ nom=deer there stand.sg ‘The deer is standing out there.’ b. nɨ { ka=tɨhɨča [oʔo wɨnnɨ–dɨ] } punni I acc=deer there stand.sg–nmlz see:dur ‘I see the deer that’s standing out there.’ In example (36a), we have a fully finite main clause. The simple form of the verb wɨnnɨ ‘stand.sg’ indicates that the act of standing is presently ongoing. In (36b), with the addition of the nominalizing suffix (nmlz), we have a structure that best translates as a relative clause in English. The head noun, tɨhɨča ‘deer,’ is being modified and referentially restricted by the relative clause in square brackets, [oʔo wɨnnɨdɨ]. We could translate this relative clause as ‘the one standing out there’ just as we could interpret tɨničuidɨ as ‘the one who teaches.’12 Looking back at examples (24) and (25) from Crow, it is not a coincidence that both ak– and ala– function as nominalizers as well as relativizers. In both Crow and Northern Paiute, the form of the nominalizer is a reflection of the relative clause type. Just as –dɨ in Northern Paiute appears on verbs in subject relative clauses, so, too, does ak– in Crow. In order to form a non-subject relative clause in Crow, the prefix ala– is used, while in Northern Paiute, the nominalizing participle suffix –na appears, as in: 12 It doesn’t work too well, however, to interpret the bracketed noun phrase in 38) to mean ‘over there stander.’ (37) Clause-combining: Relative clauses 339 Northern Paiute (Uto-Aztecan) (Thornes 2003:428) su={ miidɨ [ ɨ= kuhani—na ]} kai toki kamma nom= meat your= cook–ptcp neg correct taste ‘The meat you cooked doesn’t taste right.’ Aside from the use of a nominalizing suffix, a major feature of nominalized clauses in Northern Paiute and many other languages is that the subject of the relative clause appears in its possessive form. In (37) the ‘your’ possessor form ɨ= is the would-be subject of cook. Possession in noun and pronoun forms is a feature of noun phrases, not of clauses. In Mohave and other Yuman languages, the prefix kw– that appears on the verb of a subject relative clause does double duty as subject/agent nominalizer as well. In the Yuman case, however, instead of having a counterpart for non-subject relative clauses, the subject nominalizer kw– contrasts with its absence. Compare the following examples from Mesa Grande Diegueño: (38) Mesa Grande Diegueño (Yuman) (Dryer 2007:200) a. {[’ehatt gaat kw–akewii]}= ve= ch nye–chuukuw dog cat rel.subj–chase]=def=subj 1obj –bite ‘the dog that chased the cat bit me’ b. {[’ehatt gaat akewii]}= ve= ch chepam dog cat chase]=def=subj get.away ‘the cat that the dog chased got away’ The order of subject, object, and verb remains the same in both examples, a reflection of the SOV (subject-object-verb) word order in main clauses. The only difference in form between the two relative clauses is in the presence of the kw– prefix in (38a), signaling that subject of the relative clause ’ehatt ‘dog’ is the head noun of the complex noun phrase in curly brackets. The absence of this prefix in (38b) indicates that it is the object gaat ‘cat’ that is the head. So it is the combination of subject relative marking and word order that allows us to identify the head noun and its role in the relative clause. The Uto-Aztecan language Cupeño also bears key features of nominalization in forming relative clauses, including restrictions on verb finiteness. As Hill (2003) points out, Cupeño “verb constructions in complement and relative clauses are marked only for mood, not tense (116, emphasis added).” Note the following: (39) Cupeño (Uto-Aztecan) (Hill 2003:298) Aya { ataxa–m [ pe–m kwaw–in–t–am ] } then person–pl det–pl call–in–npn–pl ramaada–’i paas mekwel–pe’–men–wen. ramada–o thrice go.around–3pl–in.pl–pipl ‘And the people who were invited went three times around the ramada.’ 340 Tim Thornes Nominalized relative clauses typically have one or more of three key characteristics of non-finiteness: 1) restricted finite verb coding (e. g. lack of tense; mood marking only in Cupeño), 2) the presence of a determiner to mark the boundary and definiteness of the entity identified by the relative clause, and 3) possessive/genitive case forms on participants (especially the subject role). Wappo, a California language isolate, expresses relative clause-like properties of modification and narrowing reference with internally-headed, dependent clauses. Subject relative clause: (40) Wappo (isolate) (Thompson, Park, and Li 2006:115–116) { [ cek’ew olol ] –i } i peh–khiʔ dem man dance:dep –nom 1sg look:at–stat ‘The man who’s dancing is looking at me.’ Object relative clause: (41) Wappo (isolate) (Thompson, Park, and Li 2006:115–116) { [ i chuya t’um –t ] –i } shoy’i–khiʔ 1sg house buy –pst:dep –nom burn–stat ‘The house that I bought burned down.’ It is not readily apparent that these examples are in fact internally-headed, since in both cases, either a pronoun or a demonstrative precedes the head noun (perhaps as proclitic). In (40), the demonstrative appears to be functioning as a determiner, as does, arguably, the first-person singular pronoun in (41). This pronoun functions elsewhere as an inalienable possessor form. By this analysis, it may be possible to propose that what the authors call simply a “suffix for verbs in dependent clauses (xvii)” is actually a suffix indicating a nominalization. Chimariko, another isolate of California, illustrates many of the core features of nominalization in the formation of relative clauses as well. In Chimariko relative clauses, there is a complete lack of tense, aspect, or mood marking (the latter two of which are required on main clauses)—taking finite restrictions a step further than Cupeño, which still can be marked for mood, as we have seen. The dependent marker –rop, as we have seen above in examples (12) and (13) and as discussed in Jany (2013), could thereby be interpreted as a nominalizer. Clause-combining: Relative clauses 341 14.5 Summary and questions What I have attempted in this survey is to explore a few broad issues with respect to relativization phenomena in this large and diverse linguistic landscape. First among these is the presence and position of a head noun whose modification or identification is the main function of the relative clause. Second has been the identification of distinct relative clause types and how the role of the head noun or, barring one, the target entity, is understood from the grammatical cues present in the language. Finally, we have explored the presence, in full or in part, of properties of nominalization as a strategy for forming structures one could reasonably claim to be relative clauses as distinct from other grammatical constructions. As a related issue, finiteness can serve to mask structural differences between relative and main clauses. Relative clauses have held great interest among typologists and syntactic theorists, with parameters for comparison and variation established for Standard Average European (SAE) and a few other well-known languages, usually with extensive literary traditions. Some such parameters either 1) are ambiguous when applied to Native North American languages or 2) do not apply to some of these languages, particularly when grammatical relations and alignment patterns are not of the SAE variety. Grammatical studies of Native North American languages have suffered, until relatively recently, from a narrow focus on morphological phenomena and a reliance upon direct elicitation as a barometer for the presence of certain readily identified, and identifiable, syntactic structures. The assumption not only of the presence of such structures in the target language under study, but of a similar, if not identical, set of associated functions, has often remained unquestioned. Thornes (2012) and Álvarez González (2016) focus on the referential function of (mostly headless) relative clauses in Uto-Aztecan languages over the modifying function, in keeping with what Álvarez González refers to as “the prototypical function of nominalization (135).” Indeed, Bickel (2011) asserts most simply that “relative constructions turn a propositional expression into a referential one (428).” It turns out that the distribution of the so-called “nominalization strategy” for forming relative clauses is scattered across the languages surveyed, but also appears to be a family trait, for example, in the Yuman, Uto-Aztecan, and Siouan families. Input from a greater range of languages is necessary both for establishing characteristics of relative clauses among related languages, and to look for possible areal influences from long term contact between languages. As relative clause structures appear to be particularly prone to syntactic calquing and other grammatical developments, like the use of interrogative pronouns as relative pronouns (Mithun 2012), one might expect to find specific features to be present across well-known linguistic areas like the Pacific Northwest, northern California, or the North American Southeast. 342 Tim Thornes References Álvarez González, Albert. 2016. ‘The evolution of gramatical nominalizations in Cahita languages.’ In Claudine Chamoreau & Zarina Estrada-Fernandez (eds.), Finiteness and Nominalization, 107–139. Amsterdam: John Benjamins. Andrews, Avery D. 2007. Relative clauses. In Timothy Shopen (ed.), Language Typology and Syntactic Description. Vol. 2: Complex Constructions, 206–236. Cambridge: Cambridge University Press. Bickel, Balthasar. 2011. Grammatical Relations. In Jae Jung Song (ed.), Oxford Handbook of Linguistic Typology, 399–444. Oxford: Oxford University Press. Chamoreau, Claudine & Zarina Estrada-Fernandez. 2016. Finiteness and Nominalization. Amsterdam: John Benjamins. Comrie, Bernard & Zarina Estrada-Fernández (eds.). 2012. Relative Clauses in the Americas. Amsterdam: John Benjamins. Comrie, Bernard & Tania Kuteva. 2005. Relativization strategies. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://wals.info/chapter/96, Accessed on 2019-01-09.) Crippen, James A. 2012. Exploring Tlingit relative clauses: Morphology and syntax. University of British Columbia. https://tlingitlanguage.com/wp-content/uploads/2015/01/Crippen-2012-relativeclauses.pdf Dryer, Matthew S. 2013. Relationship between the Order of Object and Verb and the Order of Relative Clause and Noun. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://wals.info/chapter/96, Accessed on 2019-01-09.) Dryer, Matthew S. 2007. Noun phrase structure. In Timothy Shopen (ed.), Language Typology and Syntactic Description. Vol. 2: Complex Constructions, 151–205. Cambridge: Cambridge University Press. Frantz, Donald G. 1991. Blackfoot Grammar. Toronto: University of Toronto Press. Givón, Talmy. 2012. Toward a diachronic typology of relative clauses. In Bernard Comrie & Zarina EstradaFernández (eds.), Relative Clauses in the Americas, 3–25. Amsterdam: John Benamins. Givón, Talmy. 1990. Syntax: A Functional-Typological Introduction. Vol. 2. Amsterdam: John Benjamins. Graczyk, Randolph. 2007. A Grammar of Crow. Lincoln: University of Nebraska Press. Hill, Jane. A Grammar of Cupeño. Berkeley: University of California Press. Iorio, David E. 2014. Subject and Object Marking in Bembe. Ph. D. dissertation, University of Newcastle upon Tyne. Jansen, Joana Worth. 2010. A Grammar of Yakima Ichishkíin/Sahaptin. Ph. D. dissertation, University of Oregon, Eugene. Jany, Carmen. 2011. Clausal Nominalization as Relativization Strategy in Chimariko. International Journal of American Linguistics 77. 429–443. Keenan, Edward & Bernard Comrie. 1977. Noun phrase accessibility and universal grammar. Linguistic Inquiry 8. 63–99. Kroeber, Paul D. 1997. Relativization in Thompson River Salish. Anthropological Linguistics 39(3). 376–422. Langacker, Ronald 1977. Studies in Uto-Aztecan Grammar, I. Arlington: SIL. Martin, Jack B. 2011. A Grammar of Creek (Muskogee). Lincoln: University of Nebraska Press. Miller, Amy. 2001. Grammar of Jamul Tiipay. Berlin: Mouton de Gruyter. Mithun, Marianne. 2012. Questionable relatives. In Bernard Comrie & Zarina Estrada-Fernández (eds.), Relative Clauses in the Americas, 269–300. Amsterdam: John Benamins. Mithun, Marianne 1999. The Languages of Native North America. Cambridge: Cambridge University Press. Montgomery-Anderson, Brad. 2008. A Reference Grammar of Oklahoma Cherokee. Ph.D. dissertation, University of Kansas, Lawrence. Clause-combining: Relative clauses 343 Montler, Timothy. 1993. Relative clauses and other attributive constructions in Saanich. In Anthony Mattina (eds.), American Indian Linguistics and Ethnography in honor of Laurence C. Thompson, 241–63. (Occasional papers in linguistics). Missoula, MT: University of Montana. Munro, Pam. 1976. Mojave Syntax. New York: Garland. Nakayama, Toshihide. 2001. Nuuchahnulth (Nootka) Morphosyntax. Berkeley: University of California Press. Park, Indrek. 2012. A Grammar of Hidatsa. Ph. D. dissertation. Indiana University, Bloomington. Platero, Paul. 1974. The Navajo relative clause. International Journal of American Linguistics. 40. 202–46. Quintero, Carolyn. 2004. Osage Grammar. Lincoln: University of Nebraska Press. Rushforth, Scott & Larry Gorbet. 1989. Notes on Bearlake Athapaskan relative clauses. International Journal of American Linguistics. 55. 455–67. Rice, Keren 1989. A Grammar of Slave. Berlin: Mouton de Gruyter. Schachter, Paul. 1973. Focus and relativization. Language. 49. 19–46. Starks, Donna. 1995. Subordinate clauses in Woods Cree. International Journal of American Linguistics. 61. 312–27 Thompson, Sandra A., Joseph Sung-Yul Park & Charles N. Li. 2006. A Reference Grammar of Wappo. (University of California Publications in Linguistics 138). Berkeley: University of California Press. Thornes, Tim. 2012. Functional underpinnings of diachrony in relative clause formation: the nominalization-relativization connection in Northern Paiute. In Bernard Comrie & Zarina Estrada-Fernández (eds.), Relative Clauses in the Americas, 147–170. Amsterdam: John Benamins. Thornes, Tim. 2003. A Northern Paiute Grammar with Texts. Ph. D. dissertation. University of Oregon, Eugene. Whaley, Lindsay. 2011. Syntactic Typology. In Jae Jung Song (ed.), Oxford Handbook of Linguistic Typology, 465–486. Oxford: Oxford University Press. Amy Dahlstrom 15 Clause combining: Syntax of subordination and complementation Abstract: This chapter examines the phenomenon of subordination and how it is realized in the indigenous languages of North America. We will start by defining terms and looking at examples of different types of subordinate clauses, then consider ways in which various North American languages indicate that a clause is subordinate. One way in which languages can differ is whether the subordinate material is expressed in a full clause or one that is reduced in some way. Moreover, languages may exhibit different patterns of case marking, word order, or negation in subordinate clauses compared to the patterns found in main clauses. The final section considers some tricky cases in which it may be hard to tell whether a specific clause is a main clause or a subordinate clause. 15.1 Some definitions Let us start by defining some terms. A clause consists of a predicate (usually a verb), the arguments required by the predicate (for example, a subject and an object), plus optional modifiers providing information about the time of the event (for example “yesterday”) or the reason for the action, etc. Subordinate clauses are clauses in which the whole clause functions either as an argument or as a modifier in another clause, the main clause. In the Cherokee example in (1), the underlined portion is a subordinate clause expressing the time of the main clause:1 1 Abbreviations in the examples: 0 inanimate, 1p 1person exclusive, 3r third person reflexive, a Set A pronominal prefixes (Cherokee), a actor (Lakota), a A case (Northern Pomo), a transitive subject (Yupik), abl ablative, abs absolutive, Acomp adverbial complementizer, adv adverb, agt agent, ai Animate Intransitive verb, all allative, and andative, aor aorist, app applicative, area areal pronoun, art article, aux auxiliary, b Set B pronominal prefixes, C complementizer, c1 changed conjunct 1, cl clitic, cmp completive, cn conjunction, cnsq consequential, conj conjunct, ctl control, cvb converb, decl declarative, def definite, det determiner, di ditransitive suffix, ds different subject, dst distributive, e ergative, erg ergative, evid evidential, f feminine, fut future, ger gerund, hab.pres habitual present, hsy hearsay, I intransitive (Cherokee), I agreement class I (Choctaw), ic initial change, impv imperfective, inc incompletive, ind indicative, inf infinitive, infin infinitive, int.part interrogative participle, ipf imperfective, ipfv imperfective, irr irrealis, loc locative, lv linking vowel, m masculine, nc non-determinative connective, ncbr non-clause-bounded reflexive, neg negative, nom nominative, ns nonsingular, nxp nonexperienced past, nz nominalizer, o direct object, ob object, obj object, obl oblique, obv obviative, opt optative, ord ordinal, osp oblique-subject participle, part participle, pasv passive, pat patient, perf perfective, pf perfective, pft perfective, pl plural, po possessive, pp postposition, pr present, prc present continuous, pst, pt past, q yes/no question marker, refl reflexive voice, rel relative, rroot relative root, s singular https://doi.org/10.1515/9783110600926-015 346 (1) Amy Dahlstrom Cherokee (Montgomery-Anderson 2015: 338) sanaale yijayééja jalagi hadahntesgéesdi sanaale yi-ja-yéej-a jalagi hi-adanvhtesg-éesdi morning irr-2b-wake(i):cmp-cvb Cherokee 2a-think:inc-pft ‘In the morning when you wake up, think Cherokee!’ The main clause is ‘think Cherokee!’ and could be used on its own. The underlined part is a clause (it has a verb ‘wake up’, a subject ‘you’, and a modifier ‘in the morning’) and that whole clause functions as a modifer in the main clause, identifying the time when you should ‘think Cherokee!’. Notice that the underlined part (the subordinate clause) could not be said in isolation, except as a fragment answering a question like “When should I think Cherokee?” In the next section we will see a number of ways in which languages can mark clauses to indicate that they are subordinate: in (1) the subordinate clause is marked by a “converb” suffix -a on the subordinate clause verb, as well as by a change in tone on the verb. The opposite of subordination is coordination, where two (or more) simple clauses are combined into a single sentence and each half of the sentence is of equal importance. (2) is a Cherokee example of coordination: (2) Cherokee (Montgomery-Anderson 2015: 316) aniisgay aàníina aniichúújahno anii-sgaya anii-na anii-chúúja=hno 3a.ns-man 3a.ns-sit(ns):prc 3a.ns-boy=cn ‘The men are sitting and the boys are standing.’ aàniidóòna anii-dóòna 3a.ns-stand(ns):prc (2) contains two main clauses: that is, each clause in (2) could be used on its own as a simple clause; moreover, the activity described by each clause in (2) is given equal weight. The part of (2) that means ‘and’ is the enclitic conjunction =hno, which attaches to the first word of the second clause. (Enclitics and proclitics are discussed in section 2.3 below.) 15.1.1 Types of subordinate clause: complement vs. adjunct Subordinate clauses like the one in (1) are called adjunct or adverbial clauses: they perform functions similar to simple adverbs identifying the time or reason or other circumstance related to the main clause. Another type of subordinate clause is known (Choctaw, Northern Pomo, Southern Paiute) s subject (Nishga), s. someone/something, s3 subsidiary 3rd person, sap speech act participant, sbj subject, sg singular, sm subordinator marker, sr switch reference, ss same subject, su subject, sub subordinator, \sub subordinate tone, temp temporal, top topic, tr transitive, transloc translocative, u undergoer, vai verb, Animate Intransitive, vis visible, vti verb, Transitive Inanimate, X unspecified subject, wh question word. Clause combining: Syntax of subordination and complementation 347 as a complement clause, a clause which expresses the subject or object of the verb of the main clause. For example, consider (3), also from Cherokee: (3) Cherokee (Montgomery-Anderson 2015: 319) uùnaduulis jalagi uuniiwooniíhisdi uunii-aduuli=s jalagi uunii-wooniíhisdi 3b.ns-want:prc=q Cherokee 3b.ns-speak:inf ‘Do they want to speak Cherokee?’ The underlined part of (3) is a complement clause, here functioning as the object of ‘want’. Like the adjunct clause in (1), the complement clause in (3) cannot be used on its own as an independent clause. Complement clauses are different from adjunct clauses, however, in that adjunct clauses are optional: in (1) the speaker has chosen to give extra information about the time when the main clause takes place. Complement clauses are not optional – if the underlined portion of (3) were omitted the remaining portion would not be a complete sentence. 15.1.2 Types of complements 15.1.2.1 Embedded statements Let us look at additional examples of complement clauses. Complement clauses often express statements, as in the Kutenai example in (4): (4) Kutenai (Morgan 1991: 445) Qa ʔupxni mi¢’qaqas qa ‿ ʔupx ‿ ni mi¢’qaqas neg ‿ see/know ‿ ind chickadee niʔs ksakiɬ hakiɬwi¢kiɬiɬ niʔ-s k ‿ sak-iʔɬ‿ hakiɬ-wi¢ki-ɬ-iɬ the-s3 sm ‿ still-adv ‿ keep-watch-di-pasv ‘The chickadees don’t know (that) they are being watched.’ The statement ‘they are being watched’ in (4) is what the chickadees don’t know. 15.1.2.2 Embedded questions It is also possible to use a complement clause to report a question that a subject asked or wondered about as in the Washo example in (5): 348 (5) Amy Dahlstrom Washo (The Washo Project Online Dictionary) béverli gó:beʔ hálaŋa hé:š yák’aš-i Beverly coffee still Q warm-ipfv gó:beʔ métuʔ-mámaʔ-i coffee cold-finish-ipfv ‘Beverly asked if the coffee was still warm, but the coffee was cold.’ ʔ-í:d-i-š 3.sbj-say-ipfv-sr The question ‘is the coffee still warm?’ is what Beverly asked. (The second line of (5) is a coordinate clause, conjoined with ‘Beverly asked…’.) 15.1.2.3 Future or hypothetical events Complements of verbs like ‘want to’ or ‘try to’ express future events or hypothetical events. We already saw a Cherokee example of this type in (3); another, from Northern Pomo, is in (6): (6) Northern Pomo (O’Connor 1992: 36) wayʔe duhu man natka early leave 3sf.A try ‘She tried to leave early.’ In (6), ‘leave early’ is what the subject of (6) tried to do. 15.1.3 Types of adjuncts Examples (3)-(6) illustrated types of complement clauses; now let us look at different types of adjunct clauses that a language might employ.2 2 Not discussed in this chapter are relative clauses, which are subordinate adjunct clauses which modify a head noun, as opposed to the subordinate adjunct clauses discussed here, which modify the main clause. For example, in the following Sahaptin sentence the underlined portion meaning ‘who rode across’ is a relative clause modifying ‘that man’, providing more information about which man is referred to. (i) Sahaptin (Rigsby and Rude 1996: 688) ín=aš á=qʼinu-šan-a kwaaná I=1sg 3abs-see-impv-pst that.obj ‘I saw that man who rode across.’ ɨwínš-na man-obj.sg Relative clauses are discussed in Thornes, this volume. ana-pɨ́n rel-3sg i-qásu-yayč-a 3nom-on.horse-cross-pst Clause combining: Syntax of subordination and complementation 349 15.1.3.2 Temporal adjuncts As we saw earlier in the Cherokee example in (1), adjunct clauses often provide information about the time of the event reported in the main clause relative to the time of another event. A similar example is seen in the Tonkawa sentence in (7): (7) Tonkawa (Hoijer & Wier 2018: 50) ʔawas-wa·-ka ya·c-ayco-na-l-ʔok meat-obv-nom.pl look-up-abl-3-when he-ylap-an-cʔel-ʔa·-yʔik yele·la-k-laknoʔo. refl-stand-ger-top-def-all sit-part-evid ‘When the buffalo looked up, he was sitting on top of the tree.’ Other temporal adjuncts may express temporal relations such as ‘while…’, ‘before…’, and ‘after…’. 15.1.3.2 Reason clauses Another semantic type of adjunct clause is one that identifies the reason for the event in the main clause occurring, as in the Northern Pomo example in (8): (8) Northern Pomo (O’Connor 1992: 257) tiʔ xama diṭhal-kan mo:w ncbr.obl foot hurt-Acomp 3sm.A ‘He’s not dancing because his foot hurts.’ khemane-nha dance-neg According to O’Connor (1992: 39), the suffix -kan, glossed Acomp for adverbial complementizer, indicates that “[a]ction in suffixed clause precedes action in main clause, and main clause event is seen as resulting from event in suffixed clause.” (The abbreviation ncbr, non-clause-bounded reflexive, indicates that the possessor of the foot is the same as the person who is not dancing.) 15.1.3.3 Purpose clauses Other adjunct clauses make the goal or purpose of the action of the main clause explicit, as in the Haida example in (9): (9) Haida (Enrico 2003: 1045) dang-ga tla.ad-ee-ran hl you-pp help-infin-for I ‘I’ve come in to help you.’ qats’a-ang come.in-pr In (9), the speaker asserts that he or she has come in in order to help the addressee. 350 Amy Dahlstrom 15.1.3.4 Concessive clauses Another type of adjunct clause is often translated in English with ‘although’ or ‘even though’, expressing a state of affairs that contrasts with what is expressed in the main clause. (10) Nishga (Tarpent 1987: 418) c’ə n’í[t]=ɬ qásq’an-ə-n=ɬ ɬáq’askw although that’s=nc dislike.food-ctl-2s=nc seaweed ʔi: məq’ap ḳíp-t and 2e must.eat.s.-3 ‘Even though you dislike seaweed, you have to eat it.’ In the Nishga example in (10) the speaker knows that the addressee does not like seaweed, but asserts that the addressee must eat it nevertheless. 15.1.3.5 Conditional clauses The final type of adjunct clause exemplified here is the conditional clause, or ‘if clause’. A conditional adjunct clause identifies a hypothetical state of affairs; the main clause expresses the consequence of that hypothetical condition: (11) San Carlos Apache (de Reuse 2006: 337, glosses added from de Reuse’s word list) Ńch’ii=yúgo, doo dadányu nahikai da it.is.windy=if neg outside we.(pl.).are.(around) neg ‘If it is windy we don’t go outside.’ In the Apache example in (11), the hypothetical condition is ‘if it is windy’ and the consequence of that condition is expressed in the main clause: ‘we don’t go outside’. 15.2 How are clauses identified as subordinate? Now that we have a sense of the range of functions subordinate clauses play in a sentence, let us investigate how languages distinguish subordinate clauses from main clauses and how different types of subordinate clauses might be identified. 15.2.1 Complementizer or particle Some languages indicate that a clause is subordinate by using a separate small word, often called a particle or a complementizer. (English uses this strategy with comple- Clause combining: Syntax of subordination and complementation 351 mentizers like that or whether introducing complement clauses and words like when or if introducing various types of adjunct clauses.) An example of this strategy was seen above in the Nishga concessive clause in (10), which is introduced by a separate particle glossed ‘although’. More examples of particles can be seen in the Nez Perce examples in (12): (12) Nez Perce (Deal 2015: 412, 410) a. ke kaa Angel-nim C then Angel-erg ‘when Angel calls them’ b. ke-x kaa Angel-nim C-1 then Angel-erg ‘when Angel calls me’ hi-nees-cewcew-téetu 3subj-O.pl-call-hab.pres hi-cewcew-téetu 3subj-call-hab.pres In (12) the combination of ke and kaa indicate that the clause is a temporal adjunct clause, glossed ‘when’. See Aoki (1970: 126–127) for the wide range of particles found in Nez Perce subordinate clauses. As Aoki points out, some of the particles can be the host for subject and/or object inflection. In (12b), for example, the particle ke is followed by -x, which indicates the first person singular object of the verb ‘call’. 15.2.2 Affix Many of the languages of North America exhibit complex morphology on the verb, with agreement for both subject and object, incorporated objects, incorporated adverbial material, etc. (see Zúñiga, this volume, and Broadwell, this volume) It is therefore not surprising that some of the languages of North America indicate that a clause is subordinate by adding an affix to the verb of the subordinate clause. We have already seen several instances of this strategy in the examples above: for example, the Cherokee example in (1) has a “converb” suffix -a attached to the adjunct clause, and the Tonkawa example in (7) includes a suffix -ʔok glossed ‘when’. Languages with switch reference (see McKenzie, this volume), where the switch reference system extends to subordinate clauses, often have paired suffixes marking specific types of subordinate clauses. In such systems, one member of the pair of suffixes indicates that the subject of the subordinate clause is the same as the subject of the main clause and the other member of the pair indicates that the subject of the subordinate clause is different from the subject of the main clause. Consider the pair of Choctaw sentences below: (13) Choctaw (Broadwell 2006: 263) a. Kaah sa-nna-haatokoosh, iskali’ ittahobli-li-tok car 1sI-want-because:ss money save-1sI-pt ‘Because I wanted a car, I saved money.’ 352 Amy Dahlstrom b. Kaah banna-haatoko, iskali’ ittahobli-li-tok car want-because:ds money save-1sI-pt ‘Because he wanted a car, I saved money.’ Both sentences of (13) exhibit adjunct clauses of the reason type, and both adjunct clauses are identified by a suffix on the verb of the subordinate clause. In (13a) the suffix expressing ‘because’ is -haatokoosh while in (13b) the suffix glossed ‘because’ is -haatako. The suffix in (13a) also indicates ‘same subject’ – that is, the subject of ‘want’ and the subject of ‘save’ are the same person. In (13b) the suffix indicates ‘different subject’, because the subject of ‘want’ is not the same as the subject of ‘save’. 15.2.3 Clitic As discussed in Zúñiga (this volume), clitics resemble affixes in being phonologically dependent on a host word to be pronounced, but in other respects have properties of separate words. Some languages of North America identify subordinate clauses by using a clitic, rather than an affix or a separate particle. A clitic which precedes the host it attaches to is called a proclitic; one which follows the host is called an enclitic. An example of an enclitic was seen in (2), with the Cherokee conjunction =hno ‘and’. For an example of a proclitic consider the following Caddo example, in which a proclitic nat appears in a temporal adjunct clause: (14) Caddo (Melnar 2004: 94–95) nappáwdihšiyah … nahašʔnáwwá·yáh nat#wa-wid(i)-ih-šiyah nak ašnáw-wa-yáh temp.sub#pl-arrive-and-transloc.perf transloc.ind#meal-pl-eat ‘When they arrived there, … they would eat something there.’ dikaʔháy dikaʔháy something The symbol # after nat indicates that the usual word-internal phonological processes of Caddo do not apply at that boundary, reflecting a difference between clitics and ordinary affixes. 15.2.4 Special paradigms of subject/object agreement In other languages, the difference between main and subordinate clauses may be indicated by the use of a separate paradigm, that is, the set of affixes agreeing with the subject (and object, in some languages). For example, in the Algonquian language Meskwaki third person singular is expressed by the suffixes -w-a in the independent indicative paradigm, which is only used for verbs in main clauses. See (15a): (15) Clause combining: Syntax of subordination and complementation 353 Meskwaki (Dahlstrom 2000: 76–78) mahkate·wi·wa mahkate·wi·-w-a fast-3-sg ‘he/she fasts; he/she fasted’ [main clauses only] b. mahkate·wi·tehe mahkate·wi·-t-ehe fast-3-mode.suffix ‘if he/she had fasted, …’ c. e·hmahkate·wi·či e·h-mahkate·wi·-t-i aorist-fast-3-mode.suffix that he/she fasts; that he/she fasted.’ d. me·hkate·wi·či ic-mahkate·wi·-t-i ic-fast-3-mode.suffix ‘when he/she fasted, …’ Other paradigms, the conjunct forms, are used mostly in subordinate clauses.3 A sampling of conjunct paradigms is shown in (15b-d): in all three forms third person is expressed by the suffix -t, which is palatalized to [-č] if it is followed by [-i], as in (15c-d). Other parts of the conjunct verbs indicate what role the subordinate clause plays. For example, (15b) is a contrary-to-fact ‘if’ clause, (15c) is a complement clause which could be used with a main clause such as ‘I know…’, and (15d) is an adjunct clause expressing ‘when’ in the past. These functions are identified with a combination of the final suffix (glossed mode suffix in (15b-d) above) and what appears on the left edge of the verb. On the left edge of the verb there are three possibilities: nothing at all (15b), a prefix e·h-, glossed aorist (15c), or a process known as initial change (ic) in the Algonquianist literature which changes the length and quality of the vowel of the first syllable of the verb. In (15d) initial change has changed the short [a] of the stem to a long [e·]. Note that for the purpose of glossing examples, initial change is represented as if it were a prefix on the left edge of the verb stem, both in the morphological breakdown in line 2 and in the glosses in line 3. In each form in (15b-d), neither the mode suffix at the right edge of the verb nor what appears on the left edge of the verb (if anything) can be identified as marking the verb’s function on its own: each of the mode suffixes appears in other conjunct paradigms, as does the aorist prefix and initial change. It is the combination of the material on the left and right edges of the verb that indicates the function of the verb’s clause. See Dahlstrom (2000: 76–78) for more discussion. 3 See section 5 for discussion of conjunct forms used in main clauses. 354 Amy Dahlstrom 15.2.5 Nominalization In some languages of North America, forming a subordinate clause involves turning the clause into a noun.4 Many Salish languages employ this strategy (Kroeber 1999: 100): (16) Lushootseed (Kroeber 1999: 102) ʔi ‿cən ʔa·-t kwθə ‿dáktə aux ‿ 1s.su.cl call.tr art ‿ doctor ‘I called a doctor to look me over’ kwə s‿ -ƛ’èm-ə-θ-ám’š-s art ‿ nz-look-lv-tr-1s.ob-3.po In (16) the final word of the sentence expresses a purpose clause ‘[for him] to look me over’. The purpose clause is formed by turning the whole clause into a noun, indicated by the prefix s- (in boldface) on the verb ‘look’; s- is glossed nz for nominalizer. As a result of the nominalization, the subject of ‘look’ is expressed as a possessor (the boldfaced suffix -s at the right edge of the nominalized clause). The purpose clause can thus be translated literally as “his looking me over”. Another consequence of turning the purpose clause into a noun is that the subordinated material is introduced with an article kwə, appearing on the left of the nominalized clause. 15.2.6 Tonal contour The tonal language Cherokee uses tone to identify subordinate clauses. All subordinate verbs in Cherokee bear a “highfall” tone; for some subordinate clauses this is the only indication that they are subordinate. (17) Cherokee (Montgomery-Anderson 2015: 342–343) duùhwahtvvhe taliine aániihlinaʔééʔi dee-uu-hwahtvvh-e tali-iine anii-hlinaʔ-éeʔi dst-3b-find:cmp-nxp two-ord 3a.ns-sleep(ns):inc-nxp\sub ‘He found them asleep again.’ The gloss \sub indicates highfall tone, here on the nonexperienced past suffix. According to Montgomery-Anderson (2015: 467), the highfall tone rises higher than regular high tone, and falls a little at the end. 4 Nominalization as a strategy in relative clause formation is also discussed in some detail in Thornes, this volume. Clause combining: Syntax of subordination and complementation 355 15.2.7 No marking There sometimes is no special marking identifying a clause as a subordinate clause. Watkins (1984: 235) states that some complement clauses in Kiowa are “simply juxtaposed to the main clause.” (18) Kiowa (Watkins 1984: 235) nɔ́·-p’ì· ę́-tét my-sister [(2,3sg/agt):1sg/pat:∅/obj]-tell/pf á-k’ì·-dè Carnegie-kù ∅-bá·n-ê· her-husband-poss Carnegie-to [3sg]-go-ipf/hsy ‘My sister told me that her husband was going to Carnegie.’ 15.2.8 Embedded questions The special type of complement clause expressing a question often contains a question particle, if the embedded question is a yes-no question, or a question word, if the embedded question is of the question word type. Here are examples of an embedded yes-no question and an embedded question word question from Slave: (19) Slave (Rice 1989: 1175) John [ ʔeyi t’eere sú húhshu ] kodįhshǫ́ le that girl Q 3 opt.marry 3 know.neg ‘John doesn’t know if that girl is getting married.’ (20) Slave (Rice 1989: 1181) [ ʔamíi ʔat’į ] keodįhsǫ́ le who 3 is 3 knows area.neg ‘She doesn’t know who it was.’ See also the Washo example of an embedded yes-no question in (5). Meskwaki, on the other hand, employs its rich system of verbal paradigms (discussed earlier in 2.4) to indicate that a subordinate clause is an embedded question. In (21) there is no independent question word corresponding to ‘who’. Instead, the subordinate verb is identified as an embedded question by the combination of initial change on the first syllable of the compound verb plus the suffixes on the verb ‘eat’. Furthermore, the final suffix -a on the verb ‘eat’ indicates that the element being questioned is the subject of the verb ‘eat’.5 5 See Dahlstrom 2019 for a detailed discussion of these Meskwaki forms, known as interrogative participles in the Algonquianist literature. 356 (21) Amy Dahlstrom Meskwaki (Dahlstrom 2019: 77) e·hwe·pi–nana·tohtawi·nameki . . . e·h-we·pi–nana·tohtaw-i·nameki . . . aor-begin–ask-X>1p/aor e·škike·hi–mi·čikwe·na ic-aški–=ke·hi –mi·či-kwe·n-a ic-first.time–=and –eat-3>0/int.part-3 ‘They (unspecified) began to ask us . . . and who ate it first.’ 15.3 Full clause or reduced? Besides the differences illustrated in the preceding section regarding the strategies used to indicate that a given clause is subordinate, there are also differences seen among the languages of North America in terms of whether the subordinate clause expresses the full range of information that would be found in a main clause. Some languages, such as the Algonquian language Blackfoot and the Salish language Halkomelem, have no infinitive forms (Ritter and Wiltschko 2004). In other words, every subordinate clause in those languages expresses the full range of information found in main clauses. Other languages exhibit various restrictions on what grammatical information is expressed in a subordinate clause. For example, Rood (1996: 590) reports that Wichita marks fewer tense/aspect distinctions in subordinate clauses than are found in main clauses. 15.3.1 Infinitives Some North American languages are described as having infinitive forms of the verb in some subordinate clauses. Lakota is an example: (22) Lakota (Ullrich 2018: 16) Inúŋwaŋ iblútȟe. inúŋwAŋ i-bl-(y)útȟA swim try-1sg.a-stem ‘I tried to swim.’ In (22) the main clause verb ‘try’ is infixed with a first singular agent morpheme but the verb of the subordinate clause has no marking for subject. Nevertheless, the subject of ‘try’ is understood to also be the subject of ‘swim’ in the Lakota example, just as it is in the English translation which uses an infinitive form of ‘swim’. Similar examples were seen above in (6) for Northern Pomo and (9) for Haida. Clause combining: Syntax of subordination and complementation 357 A different sense of the term “infinitive” is used in Montgomery-Anderson’s (2015) description of Cherokee. Cherokee infinitives have prefixes identifying the subject of the verb of the subordinate clause but do not indicate the tense/aspect of the verb.6 (23) Cherokee (Montgomery-Anderson 2015: 75) aànehldi uudagesv uuyoohuúsehdííʔi a-anehldi uudagesv uu-yoohuúsehdííʔi 3a-try:prc weight 3b-lose:inf ‘He is trying to lose weight.’ Another example of a Cherokee infinitive was seen above in (3). 15.3.2 Both main and subordinate predicates in a single verb The most extreme example of reducing the subordinate “clause” is found in languages which permit incorporation of complement predicates into the main clause verb. For example, in the Yupik example in (24) the main clause predicate is ‘say’ (in boldface) and the predicate of the complement of ‘say’ is ‘wait for’, which is expressed as part of the same verb: (24) Central Alaskan Yupik (Woodbury 2017: 555) atanqe-ciq-ni-llru-ateng ama-ni wait.for-future-say-past-cnsq.3sga+3r.plO there-loc ‘Because he said that (he) will wait for them there’ Notice that Yupik allows separate tense markers for the two predicates in (24): ‘wait for’ is future tense and ‘say’ is past tense. Another example of this type can be seen in the Meskwaki example in (21), where ‘begin’ and ‘ask’ are compounded into a single verb. 15.4 Interactions with case-marking, word order, and negation The difference in syntactic contexts between main clauses and subordinate clauses can have an effect on other parts of the grammatical system of a language. For example, the Uto-Aztecan language Southern Paiute exhibits different case-marking patterns depending on whether the clause is main or subordinate. Subjects in main clauses take nominative case, while subjects in subordinate clauses take oblique case. 6 Examples like the Cherokee forms in (3) and (23) suggest that finiteness is perhaps best thought of as a matter of degree, rather than a clear-cut opposition between finite and non-finite. 358 (25) Amy Dahlstrom Southern Paiute (adapted from Bunte 1986: 283) nʉ’ pʉsuchuxwai-yʉ-ak Johni-ung kiritsi-ang I.nom know-pr-3vis John.obl-art cat.obl-art arungwa-ngkʉ-kai-naya-anga-n uy-app-perf-osp.obl-3svis-1s I know (it) that John bought the cat for me.’ In (25) the first-person subject of the main clause is in nominative case, while ‘John’, the subject of the complement clause, takes an oblique case marker. The basic word order pattern of a language may also differ depending on whether the clause is a main clause or a subordinate clause. For Quileute, Andrade (1933: 278) reports, “In the main clause the normal order is (1) verb, (2) subject, (3) object. In the subordinate clause the subject precedes the verb.” (26) Quileute (Andrade 1933: 280, 285) toqò·l dâ·kil yik hadós·t’ot’ ki’ k’ade´’ya’a´k replied then the elder.sister the her.younger.sister “hé.sekłli tca’à· ha’ tcè·k u t’łotóloo’t tas há.kutax̣a‘.” I.prefer yonder that large star 3rd.person.conditional come ‘Then the elder girl said to her younger sister, “I should prefer that big star yonder would come.”’ In (26) the first clause is a main clause, with the subject (in boldface) following the verb. In the second line there is a complement clause in which the subject ‘that big star yonder’ (in boldface) precedes the verb of the complement clause, ‘come’. Another example of how the difference between main and subordinate clauses can affect other parts of the language is in the expression of negation. In Potawatomi, two different strategies are used to negate a verb, depending on whether the verb is in a main clause or in a subordinate clause. The following examples are from the Forest County, Wisconsin, dialect of Potawatomi: (27) Potawatomi (Lockwood 2017: 115) a. Jo wi nin nwi-byasi jo=wi nin n-wi-bya-si neg I 1-fut-come.vai-neg ‘I’m not coming tomorrow.’ b. ga-bwa-wje-bontawat ic.gi-pwa-wje-bonet-awat pst-neg-rroot-quit.vti-3.pl.conj ‘why they did not quit it’ wabek wabek tomorrow (27a) is a main clause; negation here is expressed with a negative particle jo plus a suffix -si on the verb (both in boldface). In (27b), however, the subordinate clause is negated with a preverb pwa compounded with the verb and there is no negative morpheme suffixed to the verb. Clause combining: Syntax of subordination and complementation 359 15.5 Some tricky cases Most of the time it is easy to decide whether a particular clause in a language is a main clause or a subordinate clause. There are, however, a few tricky cases to be aware of, which will be covered in this section. 15.5.1 “Cosubordination” Occasionally languages have constructions which seem to exhibit features of both coordination and subordination; the term which has been coined for such constructions is cosubordination. An example can be seen in the Siouan language Crow. Crow has a switch reference system, but the system operates only on coordinate clauses. In the Crow construction which is described as cosubordination, chains of clauses are connected by switch reference markers, with only the final clause of the chain bearing a marker indicating the speech act type: (28) Crow (Graczyk 2007: 402) alápasshi-ss-basaa-(a)k dáakbachee-sh hii-ák kukaaaaxp-ák direction-goal-run-ss his.son-det reach-ss embrace-ss óhchikaap-ak iispáschi-k huu-k greet-ss kiss-decl say.pl-decl ‘he ran toward him, he reached his son, he hugged him, he greeted him, he kissed him’ (Lk 15:28) The same-subject markers on the non-final clauses are in boldface, as is the declarative speech act marker -k on the final verb of the chain, ‘kiss’. (The final word in (28) is a reportative evidential which also has the declarative speech act marker.) The effect of the declarative marker extends over the entire chain, and the non-final clauses cannot be used on their own — features which suggest subordination. However, the Crow switch reference system does not otherwise appear on subordinate clauses, only on coordinate clauses, making constructions like (28) difficult to classify as involving either subordination or coordination. For a further example of a language analyzed as exhibiting cosubordination, see Jacobsen (1992), a lengthy discussion of various subordinate and cosubordinate constructions in the Wakashan language Nootka. 15.5.2 Formally subordinate constructions used in main clauses In some languages, clauses bearing morphology which usually indicates that the clause is subordinate can used as main clauses in certain contexts. For example, the Algonquian language Plains Cree exhibits a similar distinction between independent and 360 Amy Dahlstrom conjunct paradigms as the one discussed above in 2.4: in Cree, second person singular is indicated by a prefix ki- plus a suffix -n in the independent paradigm, used only in main clauses (29a), while conjunct verbs require a suffix -yan to express second person singular (29b). (29b) shows that a verb bearing conjunct inflection can be used in a main clause: (29) Plains Cree (Cook 2014: 140) a. kinôhtêhkatân cî ki-nôhtêhkatê-n cî 2-hungry.vai-sap q ‘Are you hungry?’ [independent inflection] b. ê- nôhtêhkatêyan cî ê- nôhtêhkatê-yan cî c1-hungry.vai-2 q ‘Are you hungry?’ [conjunct inflection] According to Cook (2014: 140), the form in (29a), with the independent inflection normally used for main clauses, is used to ask someone “out of the blue” if they are hungry, such as when someone has come to visit. The form in (29b), using conjunct morphology otherwise found on subordinate clauses, is used when the context creates a presupposition relevant to the utterance, perhaps if the addressee is rummaging in the fridge looking for food. 15.5.3 Historical change/reanalysis In the section above we saw that verbs bearing subordinate clause marking may be used in a main clause in certain contexts. For some languages, such variation between subordinate clause functions and main clause functions eventually results in certain constructions being reanalyzed as being main clauses only.7 As a result, the verbal morphology which originally marked the clause as being subordinate still appears on the newly reanalyzed clause, even though the clause is no longer subordinate. An example of this can be seen in the Algonquian language Menominee, where question-word questions (as opposed to yes-no questions) require conjunct inflection on the verb. (30) Menominee (Johnson and Macaulay 2015: 344) Tāq kēs–mēcek? wh pst–eat.ai.3conj ‘What did he eat?’ 7 See van Gelderen, this volume for this type of historical change involving negative particles. Clause combining: Syntax of subordination and complementation 361 As in other Algonquian languages, the primary function of conjunct morphology in Menominee is to indicate that the clause is subordinate. Given that, we might expect the translation of (30) to be something like “What is the thing which he ate?”, with ‘eat’ appearing in a subordinate clause. Johnson and Macaulay (2015), however, present evidence which indicates that the verb in questions like (30) is not part of a subordinate clause but is instead the main clause verb of the question. Menominee questions are therefore an instance where the syntax of the construction has changed over time, reanalyzing an originally subordinate clause to be a main clause, but where the morphology of the construction still reflects the syntax of the older construction.8 15.6 Conclusion This chapter has defined various types of subordination, surveyed ways in which subordination may be indicated, shown how the syntactic context of subordination can affect other parts of a language’s linguistic system, and ended by pointing out a few pitfalls that analysts of a language may need to watch out for in classifying clauses as subordinate. References Andrade, Manuel J. 1933. Quileute. In Franz Boas (ed.), Handbook of American Indian languages 3, 151–292. New York: Columbia University Press Aoki, Haruo. 1970. Nez Perce grammar. University of California Publications in Linguistics. Berkeley: University of California Press. Broadwell, George Aaron. 2006. A Choctaw reference grammar. Lincoln: University of Nebraska Press. Bunte, Pamela A. 1986. Subordinate clauses in Southern Paiute. International Journal of American Linguistics 52. 275–300. Cook, Clare. 2014. The clause-typing system of Plains Cree: Indexicality, anaphoricity, and contrast. Oxford: Oxford University Press. Dahlstrom, Amy. 2000. Morphosyntactic mismatches in Algonquian: Affixal predicates and discontinuous verbs. In Arika Okrent & John Boyle (eds.), Proceedings from the panels of the Chicago Linguistic Society’s thirty-sixth meeting, 63–87. Chicago: Chicago Linguistic Society. Dahlstrom, Amy. 2019. Embedded questions in Meskwaki: Syntax and information structure. In Monica Macaulay & Margaret Noodin (eds.), Papers of the 48th Algonquian Conference 69–85. East Lansing: Michigan State University Press. 8 Another example of subordinate clauses being reanalyzed as main clauses can be seen in auxiliary verb constructions in Yuman languages, where the main verb and the auxiliary are separated by the suffix which otherwise marks ‘same-subject’ in a switch reference system. In such constructions the verb which is now the main verb must have originated as the verb of a subordinate clause, subordinated to the verb which now functions merely as an auxiliary (McKenzie 2015: 436). 362 Amy Dahlstrom Deal, Amy Rose. 2015. A note on Nez Perce verb agreement, with sample paradigms. In Natalie Weber, Erin Guntly, Zoe Lam & Sihwei Chen (eds.), Proceedings of the International Conference on Salish and Neighbouring Languages 50, 389–413. Vancouver: University of British Columbia Working Papers in Linguistics. Enrico, John. 2003. Haida syntax. Lincoln: University of Nebraska Press. Graczyk, Randolph. 2007. A grammar of Crow. Lincoln: University of Nebraska Press. Hoijer, Harry & Thomas R. Wier. 2018. Tonkawa texts: A new linguistic edition. Norman: University of Oklahoma Press. Jacobsen, William H., Jr. 1992. Subordination and cosubordination in Nootka: Clause combining in a polysynthetic verb-initial language. In Robert D. Van Valin (ed.), Advances in Role and Reference Grammar, 235–274. Amsterdam: John Benjamins. Kroeber, Paul D. 1999. The Salish language family: Reconstructing syntax. Lincoln: University of Nebraska Press. Johnson, Meredith, & Monica Macaulay. 2015. A monoclausal analysis of Menominee wh-questions. International Journal of American Linguistics 81. 337–377. Lockwood, Hunter Thompson. 2017. How the Potawatomi language lives: A grammar of Potawatomi. Madison: University of Wisconsin dissertation. McKenzie, Andrew. 2015. A survey of switch reference in North America. International Journal of American Linguistics 81. 409–448. Melnar, Lynette R. 2004. Caddo verb morphology. Lincoln: University of Nebraska Press. Montgomery-Anderson, Brad. 2015. Cherokee reference grammar. Norman: University of Oklahoma Press. Morgan, Lawrence. 1991. A description of the Kutenai language. Berkeley: University of California dissertation. O’Connor, Mary Catherine. 1992. Topics in Northern Pomo grammar. New York: Garland. de Reuse, Willem J. & Philip Goode. 2006. A practical grammar of the San Carlos Apache language. (LINCOM studies in Native American linguistics 51). Munich: Lincom Europa. Rice, Keren. 1989. A grammar of Slave. Berlin: Mouton de Gruyter. Rigsby, Bruce & Noel Rude. 1996. Sketch of Sahaptin, a Sahaptian language. In Ives Goddard (ed.), Handbook of North American Indians, Volume 17, 666–692. Washington, DC: Government Printing Office. Ritter, Elizabeth, & Martina Wiltschko. 2004. The lack of tense as a syntactic category: Evidence from Blackfoot and Halkomelem. In J. C. Brown & Tyler Peterson (eds.), Proceedings of the 39th International Conference on Salish and Neighbouring Languages 14, 341–370. Vancouver: University of British Columbia Working Papers in Linguistics. Rood, David. 1996. Sketch of Wichita, a Caddoan language. In Ives Goddard (ed.), Handbook of North American Indians, Volume 17, 580–608. Washington, DC: Government Printing Office. Tarpent, Marie-Lucie. 1987. A grammar of the Nisgha language. Victoria: University of Victoria dissertation. Ullrich, Jan. 2018. Modification, secondary predication, and multi-verb constructions in Lakota. Düsseldorf: Heinrich Heine University dissertation. The Washo Project. Online dictionary. Chicago: University of Chicago. http://washo.uchicago.edu/dictionary/ dictionary.php (accessed 27 May 2019). Watkins, Laurel J. 1984. A grammar of Kiowa. Lincoln: University of Nebraska Press. Woodbury, Anthony. 2017. Polysynthesis in Central Alaskan Yup’ik. In Michael Fortescue, Marianne Mithun & Nicholas Evans (eds.), The Oxford handbook of polysynthesis, 536–559. Oxford: Oxford University Press. Andrew McKenzie 16 Switch-reference and event cohesion Abstract: This chapter considers some of the ways that grammars encode how speakers build discourse structures larger than sentences, focusing on switch-reference systems. Canonically these indicate that the clauses share subjects, but findings in many North American languages show that switch-reference indicates various kinds of event cohesion as well. The chapter also discusses different clause types that switch-reference can be found with and how those types affect the kinds of cohesion available. Finally, it compares switch-reference to a number of phenomena that have been linked to it over the years. 16.1 Introduction Most grammatical study emphasizes how clauses and sentences are built, but people do not generally utter sentences in isolation. Rather, we situate sentences as part of a larger conversation (or discourse) between speech participants. In the course of a conversation, we will link certain sentences together as involving the same theme, event, or topic— that is, how much cohesion they have. We also set other sentences as distinct from one another. Adverbials like then or while fulfill this purpose, but in many of the languages of the world, specific forms in the grammar provide speakers with the means to indicate cohesion across clauses. In the languages of North America, one very common form that marks cohesion is called switch-reference, which usually indicates that the two clauses have the same grammatical subject or different subjects. For instance, in Kiowa (Kiowa-Tanoan), the connective meaning ‘when’ has forms that indicate switch-reference (SR). The form in (1a) indicates that the clauses share the ‘same subject (SS)’, and the form in (1b) indicates they each have a ‘different subject (DS)’.1 (1) Kiowa (Christina Simmons, p.c.) a. tôy à=héːbà=tsę ː̀ èm=bǫ́ ː house.in 1sg=enter.pfv=when.SS 1sg>2sg=see.pfv ’When I went into the house, I saw you.’ 1 Glossing conventions follow Leipzig rules, with the following additions and clarifications: !: surprise, A: agent, D: dative/applicative argument, decl: declarative mood, DS: different subject/situation, dur: durative, evid: indirect evidentiality, exp: personal experience evidential, h: h-grade aspect, INC: Incomplete aspect, indic: indicative mood, irr: irrealis mood, mv: middle voice, obl: oblique, OR: open reference, Pma: perfective, male addressee, prev: previously mentioned, ptcp: participle, S: subject, seq, sequential in time, sim: simultaneous, spec: specific, SS: same subject/situation, TEM: temporal pivot. Some glosses from various sources have been adjusted for clarity or to standardize abbreviations. https://doi.org/10.1515/9783110600926-016 364 Andrew McKenzie b. tôy à=héːbà=ę ː̀ ę́=bǫ́ ː house.in 1sg=enter.pfv=when.DS 2sg>1sg=see.pfv ’When I went into the house, you saw me.’ SR markers are not usually translated, since English and other European languages lack them. However, their use is crucial for fluent conversation in the languages they appear in, so care must be taken to document and understand them. 16.2 Cohesion via role identity When thinking about cohesion, we want to keep in mind exactly what is being held as cohesive. It is typically held that a marker of cohesion actually indicates some kind of semantic identity across clauses. That is, the same particular role in each clause is filled by the same real-world object.2 A marker of non-cohesion actually indicates some kind of semantic non-identity across clauses; the same role is fulfilled by different real-world objects. Thus, in (2), the -ku suffix in Tümpisa Shoshone (Numic) indicates that the two clauses’ time arguments are filled by the same interval in time (sim), while -ka/ha indicates two distinct intervals (3): (2) Tümpisa Shoshone (Dayley 1989: 347–8) a. ümaku tammü kahni kuppa weekikkwantu’ih rain-sim.DS we.incl house in enter.pl-going to ‘When it rains, we’re going to go in the house.’ b. the time of it raining = the time of us going into the house (3) Tümpisa Shoshone (Dayley 1989: 347–8) a. Nümmü [tatsa naakkiha] nümmü supe toya mantu mi’a we.excl we.incl get-seq.DS in there mountain to go When [=after] it gets summer, we go there to the mountains.’ b. the time of it getting summer ≠ the time of us going to the mountains When we consider cohesion across clauses in terms of identity of a particular role-filling item, we can try to organize these markers based on the roles, and we observe that usually, different kinds of semantic objects fill different roles. 2 If one is talking about unreal situations or hypotheticals, these ‘real-world’ objects might also be hypothetical. Switch-reference and event cohesion 365 16.2.1 Subject identity A common form of cohesion across clauses involves subject identity. For instance, in some subordinate clauses in Chickasaw (Muskogean) the −t suffix indicates that the subject of the main clause is the same as the subject of the subordinate clause. The nasalvowel suffix −V indicates that the subjects are distinct. This bivalent marker is called switch-reference, and in (4) it is the only overt difference between the two sentences. (4) Chickasaw (Munro 2017: 122) a. Amposhi’ achifa-kmat kashooch-a’ni dish wash-if:SS dry-can ‘if she1 washes the dishes she1 can dry them’ b. Amposhi’ achifa-kma kashooch-a’ni dish wash-if:DS dry-can ‘if she1 washes the dishes she2/*1 can dry them’ Switch-reference was first named by Jacobsen (1967) when describing the Washo (Isolate) and Tonkawa (Isolate) languages, and it was swiftly found in dozens of other languages in North America and then around the world (Austin 1981; Haiman & Munro 1983; Roberts 1997; McKenzie 2015; van Gijn 2016). Many American languages allow arguments to be unexpressed; sentences might just consist of a verb. Because of this, switch-reference can tell listeners who is doing what when there are two people involved. For instance, the Crow (Siouan) sentence uá dappeék can mean that someone killed his wife, or that his wife killed someone. In a sentence with SR, this important distinction becomes clear. (5) Crow (Graczyk 2007: 417) a. bachée-sh húu-laa uá ∅-∅-dappeé-k man-det come-SS his:wife 3O-3A-kill-decl ‘The man came and killed his wife.’ b. bachée-sh húu-m uá ∅-∅-dappeé-k man-det come-DS his:wife 3O-3A-kill-decl ‘The man came and his wife killed him.’ This effect can be very helpful in a narration. In the following excerpt from a Kiowa story, seven consecutive sentences (b-g) start with an SR-marker. Each marker helps listeners understand the action, since the subject is not expressed verbally in any of them, and the overt nouns are not case-marked. (6) Kiowa (P. McKenzie n.d. “Poolant’s Killing”) a. pólą́ ːtè tsę̂ ː á=p’ɔ́y-hyèl … t’ɔ́khɔ́y+k’íː mɔ́n Poolant horse 3sS›3sD=be lost-evid White+man probably á=pɔ ̀ːdôː 3sA›3sO›3sD=care for ‘Poolant had lost a horse … a White man must have been watching it for him.’ 366 Andrew McKenzie b. c. d. e. f. g. nɔ ̀ hɔ́ndéʔ+tʰàːlyìː ∅=pɔ ̀ː+kʰɔ̂ː+tòl-hèl and:DS some+boy 3sA›3sO=bring+get+send.pfv-evid ‘So he sent a boy to bring it back.’ (DS → Poolant) nɔ ̀ mɔ́n ∅=ɔ́ːgyá+tɔ ̀ː and:DS probably 3sS=refuse to relinquish+act.pfv ‘But he refused to give the horse up.’ (DS → the White man) nɔ ̀ hègɔ́ ɔ̨́ːg̣ɔ ̀ gyà=ây-hyèl and:DS then self 3sA›3pO=start.off-evid ‘So then he went [to get it] himself.’ (DS → Poolant) gɔ ̀ mɔ́n háyá t’ɔ́khɔ́y+k’íː mɔ́ː and:SS probably somehow White+man somewhat ∅=éppátéː+dɔ ̀ːpèː 3sA›3sO=with force+ask.pfv ‘He must have somehow tried to coerce the White man a bit. (SS → Poolant) nɔ ̀ hègɔ́ mɔ́n ∅=sɔ̨́ɔ̨ ːdè ̀ and:DS then probably 3sS=become angry.pfv ‘and then he got angry’ (DS → the White man) gìgɔ́ ∅=thɔ́ttè-hèl gɔ ̀ ∅=hól-hèl and:SS|then 3sA›3sO=shoot.pfv-evid and:SS 3sA›3sO=kill.pfv-evid ‘and shot him and killed him’ (SS → the White man) Switch-reference is very common in North America and will form the focus of the rest of this chapter. We will see how it interacts with other expressions of cohesion, and how it often ignores subjects in favor of vaguer forms of cohesion. 16.2.2 Events and times Switch-reference in many languages can ignore subjects altogether, to mark a continuity between the events of the clause rather than the subject. This is sometimes called ‘non-canonical switch-reference.’ Watkins (1993) finds that in Kiowa, SS marking can be used with different subjects when the two events are linked. (7) Kiowa (Watkins 1993: 149) a. Kathryn gyà=gút gɔ ̀ Esther=àl K. 3sA›3pO=write.pfv and:SS E.=also ‘Kathryn wrote a letter and Esther wrote one, too.’ b. Kathryn gyà=gút nɔ ̀ Esther=àl K. 3sA›3pO=write.pfv and:DS E.=also ‘Kathryn wrote a letter and Esther wrote one, too.’ gyà=gút 3sA›3pO=write.pfv gyà=gút 3sA›3pO=write.pfv McKenzie (2012) investigates this link and reports that speakers like (7a) if for instance the letters are being talked about as parts of a campaign under discussion, like writing Switch-reference and event cohesion 367 the governor to pardon a particular person. If there is no such campaign, or even if there is, but it is not significant to the discussion, then DS marking is chosen (7b). DS marking can also indicate that the events being described are distinct, even if the clauses’ subjects are identical. This is the case in Kiowa and also in Lakota. (8) Lakota (Siouan) (Dahlstrom 1982: 73) mazophiyeta wa’i yũkhã čhuwe wãblake store-to 1.go and:DS sister 1.see ‘I went to the store and I saw my sister.’ In the terms we use here, we can say that these cases involve ordinary switch-reference, but with an event or situation as the locus of cohesion instead of the subject. Events and situations are real-world objects that we can describe or refer to with anaphora. Events can be verb arguments, and they can be related to each other as identical or not. SS marking in cases like these indicates that the event (or situation) that the first clause describes is the same event as the event that the second clause describes. DS marking indicates that the clauses describe different events or situations that are not being joined in this sentence. In (7a), the speaker chooses the same event and applies the two sentences to it; the letter-writings are parts of that event. In (7b), the speaker chooses each letter-writing event separately. This choice is schematized in Figure 1. e₃ = e₁ + e₂ + … e₃ : the letter-writing campaign e₁ : Kathryn writes a letter e₂ : Esther writes a letter e₃ – Kathryn wrote a letter and.SS e₃ – Esther wrote one, too e₁ – Kathryn wrote a letter and.DS e₂ – Esther wrote one, too Fig. 1: Speaker’s choice of event continuity and its effect on SR marking. It can be hard for investigators to ‘see’ what the difference is between these choices, but native speakers clearly distinguish them. SS marking is used to track events in several ways in a number of other languages. One use is to ‘zoom in’ on a scene. In the Crow sentence in (9), a boy looks over at the campsite, and the events he sees are linked by SS. (9) Crow (Graczyk 2007: 415) chiláakshe shikáakee-sh asall-ák kuss-íkaa-lee-m ashé ah-ak morning boy-det go out.SS goal-look-!-DS lodge many-SS bilaxpáake chiwakálaa-(a)k dahkú-m people go back and forth-SS continue-DS ‘in the morning the boy went out, he looked in the direction of [the old campsite], and to his surprise, there were lots of lodges, and people going back and forth.’ 368 Andrew McKenzie Another event-oriented use of SR is described by Mithun (1993), who notes that the Central Pomo (Pomoan) sentence (10), uttered as part of a legal discussion, describes two clauses “packaged as a single event explaining the lack of documents, even though their subjects were different.” (10) Central Pomo (Mithun 1993: 132) mu:l ʔe khe pápil=ʔel s-ts’á-ba that copula 1.obl paper=the with.liquid-destroyed-and:SS ʔa: qów=mča-w=ʔkhe č’o-č=ya 1A out=throw-pfv=fut happen-semelfactive=exp ‘My papers got wet and I just had to throw them away’ čalél just Mithun proposes that SR only marks event cohesion, and it just so happens that subject sharing is a key component of two events being cohesive enough for SS marking. In narratives, DS marking can also be used with event orientation, indicating the boundary between two episodes in a story, as exemplified in (11), from the Mandan (Siouan) language. DS links two clauses with the same subject but marks the story’s boundary between the landing and the traveling. Once the second part begins, SS marking links Coyote’s actions. (11) Mandan (Mixco 1997:248) kipxeak kirątɛrį kasi:wįowąkoʔš ki-pxe-ak ki-rątE-rį ka-si:-wį-o:wąk-oʔš mv-land-DS mv-get up-SS inceptive-travel-prog-npst-Pma ‘Coyote landed, got up, and started traveling.’ kirųwąʔkšis ki-rųwąʔk-ši-s ?-man-good-def It can be difficult to determine what drives speakers to choose to describe one event or two when linking two clauses. Pustet (2013) quantitatively analyzes the use of SR in Lakota narratives, finding that subject identity, identity of the event’s time, contrast, and probability all play a role in choosing SS or DS. Mithun (2020) discusses similar factors playing roles in Pomoan languages, along with possible routes of the markers’ origins. 16.3 Additional morphology of switch-reference This section discusses some of the morphology of switch-reference in more detail, focusing on the general traits one can expect to find if it is present in a language. 16.3.1 Diagnosing switch-reference Switch-reference (SR) markers can be diagnosed as such if they have all of the following features, based on McKenzie (2015: 418). Switch-reference and event cohesion 369 1. They occur at or near a clause juncture, including coordination or subordination. 2. They have two values in a complementary pair, such that the forms have identical meanings except for their SR value. 3. The SR value depends solely on the reference of the subject or prominent argument of each joined clause.3 The complementary forms are known as ‘same subject’ and ‘different subject’ forms, or SS and DS. Some of these pairs are shown in Table 1. Notably, the SS/DS abbreviations are used even when the prominent argument is not the subject. Tab. 1: Complementary pairs of SR markers Language SS DS Choctaw (Muskogean | cho) Mojave (Yuman | mov) Crow Tohono O’odham (Piman | ood) -t -k -laa c -V̨ (nasal vowel) -m -m ku Not all SR morphemes have all three properties on this list. For instance, a handful of languages only have a DS form, like Washo -š. No language has been observed to have only SS forms. Some languages have what researchers call “open reference” (OR), where the SS form can only occur with same subjects, while an ‘open’ form can be used with same or different subjects. One such language is Northern Pomo (Pomoan), where –(e)n is SS (12a), but -da is open reference (12b-c). One might suspect this is actually a form of non-canonical or event-related SR; McKenzie’s (2015) survey lists these forms as ‘non-canonical’, requiring more investigation. (12) a. b. Northern Pomo (O’Connor 1993: 231, 232) ša-nam maʔa-n man mo:wal baʔol-e fish-spec eat-SS 3s.fem 3s.masc call-pres ‘While she1 ate the fish she1/*2 called him.’ ša-nam maʔa-da man mo:wal baʔol-e fish-spec eat-DS 3s.fem 3s.masc call-pres ‘She1 vomited because she2/*1 ate the fish.’ 3 The term ‘prominent argument’ is a bit vague because the mechanism of prominence varies across languages. In some languages the prominent argument is very clearly the grammatical subject of the verb. In others, it can be a topic (i. e., an argument of a discourse), while in some (like Choctaw), a topical “broad subject” can be the focus of SR rather than the actual grammatical subject (Broadwell 2006). In languages where the event or situation argument is the focus of SR, that is the prominent argument. 370 Andrew McKenzie c. tiyi ša-nam maʔa-kan logophor fish-det eat-because.OR ‘She1 vomited because she1 ate the fish.’ maːdal 3s.fem.pat yat-ye vomit-past Most SR morphemes do have all the features on this list. For cases where not all of them are met, if a morpheme has two out of three of the features, it is likely switch-reference. McKenzie (2015) discusses how some systems described with the term ‘switch-reference’ do not meet these structural criteria of a distinct morpheme, even though they do involve a mandatory switch in reference. For instance, Enrico uses ‘switch-reference’ to describe a Haida phenomenon where coreference depends on the presence or absence of a pronoun. Woodbury (1983) describes a “fourth-person” marking in Central Yup’ik that appears in instances of a third-person that is not co-referent to an anaphoric third person. Similar systems are found in other languages of the Eskimo-Aleut family (Pittman 2005, Berge 2011). However, they do not meet criterion 2, and their restriction to third person suggests rather that they are in the family of pronominal anaphors. They do feature in marking cohesion across clauses, however, and are discussed in Section 6. Many languages have multiple forms of SR markers, such as Kiowa, whose SR markers are fused with different connectives (Table 2). Each set of SR markers forms a complementary pair. The first two are sentential conjunctions, and the second two are adverbial subordinators. Tab. 2: Kiowa SR forms (Watkins 1984: 236) SS DS ‘and’ ‘and’ (contrary to expectation) ‘when’ ‘if, as’ gɔ ̀ k’ɔ ̀t =tsę ː̀ =gɔ ̀ nɔ ̀ ɔ ̀t =ę ː̀ =nɔ ̀ 16.3.2 Fusion with other morphemes SR markers in many languages fuse with another morpheme, creating a portmanteau. The SR markers thus seem to indicate two or more things at once. We saw this with Kiowa in the previous section, where the SR marker also indicates some sentential connection. As linguists we often will separate the SR meaning from the other meaning in our analysis, the way we often do with fused tense, aspect, or mood markers, because phenomena like temporal connectives or conjunctions operate independently from indications of reference. In Chemehuevi (Numic), the SR markers also indicate simultaneity or sequentiality. Between those two values, and the SS/DS pair, the language has four distinct markers of SR (Table 3). Switch-reference and event cohesion 371 Tab. 3: Chemeheuvi SR markers (Press 1979:108) SS DS same event time (sim) different event time (seq) -ga(i) -c(i) -g(u) -k(a) The Chemehuevi case is also interesting because it demonstrates how tricky it can be to rely on linguistic descriptions without further investigation. Linguists over the years have used the same terms in different ways, and that difference can cause confusion to a reader in our time. In the case of Chemehuevi, Press (1979) does not use the term ‘switch-reference’, but simply states that some adverbial clause markers depend on whether the subject is the same as that of the main clause. We also see that the marker appears at a clause boundary, occurs in a complementary pair, and indicates subject identity or non-identity. These are all signs of switch-reference, so we can consider it as SR, even if the author did not. When we consider the description of the time relation in the SR markers, we also must take care. Press (1979: 108) describes the same time (sim) SR markers as contemporaneous due to being durative, and different time (seq) markers as momentaneous (which leads to a sequential reading). However, Press then states that the ‘momentaneous’ markers “refer to antecedent actions”— things that had previously occurred. Press claims that these adverbial markers also indicate whether or not the clauses’ ‘tense’ matches, but this cannot be the case, as Press points out elsewhere that Chemehuevi subordinating clauses are not marked at all for tense and are interpreted with the tense of the main clause (13). The clauses therefore always have the same tense, so “different tense” marker could not be used here. We can see from the description that the marker indicates that the clauses’ event times are distinct. They are non-contemporaneous, as one occurs after the other is complete. (13) Chemehuevi (Press 1979: 108) Ann ijapaka-c tɨrawaiʔi-kwai-vɨ Ann be scared-SS.seq dash-away-past ‘Ann got scared and ran away.’ This discussion is not intended to cast doubt or aspersions on previous work, but to warn readers about encountering terminology. Modern usage of the term tense usually does not apply to the event time (when the action actually happens) but rather to the topic or reference time (the time frame the sentence is about). However, this usage did not become widespread until after much of the descriptive literature was already published, including the description of Chemehuevi. When examining a language’s reference materials, a reader should keep this kind of usage difference in mind. 372 Andrew McKenzie 16.3.3 Where switch-reference is found Switch-reference is found in languages in many parts of North America. McKenzie (2015) finds SR in nearly 70 languages and dialects in indigenous languages of the US and northern Mexico. Many language groups lack it altogether, notably the Algonquian, Iroquoian, and the Athapaskan groups. In many language groups, virtually every language has it, including the Numic, Yuman, and Muskogean groups. However, one of SR’s interesting features worldwide is that it spreads through contact from one language to its neighbors (Austin 1981). In North America, this kind of diffusion explains why many language families have one or a few members that use SR, while the rest do not. Among the Siouan languages, for instance, languages whose speakers were in contact with speakers of SR languages often came to have SR, including Crow in the North (near Numic communities) and Biloxi in the south (near Muskogean communities). However, those that were not in such contact never adopted it, like Osage or Kansa. Kiowa is an even more striking case, for it is the only Kiowa-Tanoan language with SR, having changed from its speakers’ close contact with the Crow and the Comanche, who both speak SR languages. Another consequence of areal diffusion is that we cannot assume that a language has SR forms or does not simply based on the languages related to it. On the other hand, the presence of SR markers can reveal interesting pieces of a language’s history (Mithun 2020). 16.4 Switch-reference with subordination Switch-reference frequently occurs with subordinate clauses. The SR marker appears at the end of the subordinate clause, usually attached to or fused with the subordinating connective. It connects the embedded clause to the main clause and relates its subject to the main clause subject: SS if the subjects are identical and DS if they are not. In this section’s examples, subordinate clauses are enclosed within [square brackets]. (14) Tümpisa Shoshone (Dayley 1989: 348) Puhakantün puuhawinna [ung kamma-ku] shaman cure you.acc be sick-when.DS ‘The shaman cures you when you’re sick’ If there are multiple subordinate clauses in a single sentence, each clause’s SR marker compares its subject to the main clause subject, not to that of any adjacent subordinate clause. (15) Switch-reference and event cohesion Maricopa (Yuman) (Gordon 1983:93) [’iipaa-ny-sh nya-vaa-k]A [’ayuu ’rav-m]B man-det-subject when-come-SS something 1-rav-DS ‘When the man came, he helped me because I was sick.’ 373 ny-wik-k 3>1-help-aspect 16.4.1 Adjunct clauses Switch-reference is commonly found with adverbial clauses, subordinate clauses that modify the event of the main verb. These include when-clauses, if-clauses, and asclauses. (16) Mojave (Munro 1976: 44–5) nya-avač-ku:ʔe:-kum ahwer-k when-arrive.pl-poor.pl-DS fence in-SS ‘When the [poor] parents had a chance to come in, they [the Whites] had fenced the place off’ (17) Choctaw (Broadwell 2006:292) Tiballichi-li-kmã am-anooli-h err-1s-if:DS 1sD-tell-tense ‘If I make a mistake, tell me.’ Choctaw (18) Tolkapaya Yavapai (Yuman) (Hardy & Gordon 1980:185) qwaloyawa ’-nwirk-k ’-wu-m [ssah-a ’um-t-m] chicken 1>3-cook-SS 1-do-inc spoil-irr neg-tem-DS ‘I cooked the chicken before it spoiled.’ 16.4.2 Complement clauses Complement clauses are arguments to the verb, usually a verb expressing a mental state like think or know, or a verbal action like say or tell. SR can occur with complement clauses in some languages, though not very many. (19) Choctaw (Broadwell 2006: 271) Lynn-at ik-ikháan-o-h [iy-aachĩ-ka-t] Lynn-nom agr-know-neg-tense go-irr-comp-SS ‘Lynn does not know that she will go’ 374 Andrew McKenzie 16.4.3 Relative clauses It is uncommon for relative clauses to bear SR marking, but it does occur. Choctaw and Chickasaw permit SR marking to be used in place of the case-marking that the relative clause normally bears. (20) Choctaw (Broadwell 2006:299) [Ofi’ ipiita-li-k-aash-mã] dog feed-1s-tense-prev-dem.DS ‘The dog that I fed ran away’ (21) Chickasaw (Gordon & Munro 2017:13) [Ofi’ ipita-li-kaash-oot] isso-li-tok dog feed-1s-prev-focus.SS hit-1s-past ‘I hit the dog I fed’ balii-t run-ptcp kaniiya-h go away-tense Washo employs DS marking with internally-headed relative clauses. (22) Washo (Peachey 2006:7) k’ák’aʔ dá: ∅-gé:gel-i-š-ge ∅-yá:m-aʔ heron there 3-sit-ipfv-DS-3O 3-speak to-aorist ‘She spoke to a heron who was sitting there.’ It is not yet known why SR appears so much more frequently with adverbial clauses than with complement or relative clauses, or why any particular language employs SR on one type but not another. 16.5 Switch-reference with other clause types Switch-reference is not restricted to the juncture of subordinate clauses to main clauses. It occurs in many languages in coordinating constructions, and its properties in those cases differ slightly from those of SR with subordination. It also occurs with clausechains, where its properties again differ slightly. 16.5.1 Sentential coordination A good number of languages use SR with sentential coordination, where two complete sentences are joined together. The SR marker is fused with the conjunction in such cases, like in Washo. (23) Switch-reference and event cohesion 375 Washo (Jacobsen 1967) píteliʔ yát’umuwaʔ-aʔ ʔ-i-š-da géwe geʔišúwam-i-da lizard go down in-aorist X-ipfv-ds-loc Coyote pursue him-ipfv-ds ‘The lizard went in and then Coyote pursued him and. . .’ Coordination contrasts with subordination in several ways. With coordination, the SR markers appear at the beginning of the clause (clause-initial) attached to the conjunction, while with subordination they appear at the end of the clause (clause-final) near the subordinating morpheme. The subject of a coordinated SR clause is compared with the previous clause that it is coordinated with, and there is no skipping of main clauses (see (6)). Also, coordination allows for non-canonical SR in many languages (7a), but subordinating SR is strictly canonical. 16.5.2 Clause-chaining Many languages use switch-reference with clause chains. Clause-chaining is a type of clause linking where a series of incomplete ‘medial’ clauses are chained one after another and capped off by a main ‘final’ clause. (24) Central Pomo (Mithun 1993:121) ʔa: čáw=yó-ba máti ʔ-čhá:-č-ba maʔá 1A in=go-SS down by gravity-sit-inch-SS food ‘I came into the house, (I) sat down, and (I) started to eat.’ qa:-yúʔč’i-w biting-begin-pfv Clause-chains skirt the line between subordination and coordination. Clause-chaining behaves like subordination in many ways. A chained clause’s tense and aspect are determined by the main clause. In a sentence with a long chain of clauses, all of their tense and aspect values depend on those of the main clause. However, chained clauses also behave like coordination in many ways. For instance, anaphora proceeds in a linear fashion, rather than from main clause to subordinate. Since their clause structure is distinct from coordination and subordination, chained structures are translated into English as coordination, subordination, or other structures altogether, as we can see in the sequences in (25), in the unrelated languages Jamul Tiipay (Yuman) and Creek (Muskogean). (25) a. b. Jamul Tiipay (Yuman) (Miller 2001:240) servees me-si-x-pu m-aa-chm uuyaaw beer 2-drink-irr-dem 2-go-DS know ‘I know you went there to drink beer’ w-amp-ch ’al’al-ch w-aam-s 3-walk-SS wobble-SS 3-go away-emphatic ‘He staggered away’ 376 Andrew McKenzie c. d. chaw-k uuyaaw-x fix-irr.SS know-irr ‘I will find out how to fix it (or, I will know how and I will fix it)’ Creek (Muskogean) (Martin 2011:315) ísti hamk-ín faccí:ca-n háhy-i:-t person one-DS judge-acc make.h-agr-SS ‘We’ll make one person the judge…’ With SR, clause-chains skirt the line as well. An SR marker appears at the end of the chained clause, as it would in a subordinate clause. However, SR’s meaning behaves likes it does with coordination, in that it compares the medial clause to the next one, whether it is another chained clause or the main clause. Also, SR can be non-canonical with clause-chains, like it can with coordination. 16.5.3 Pro-verbs Another structure that carries SR is the pro-verb. A pro-verb is a verb linking two sentences together, and which stands in place of the main verb of the first of the two sentences. The Tonkawa example in (26) demonstrates this effect. The two sentences are linked by the verb translated as do so, which bears the SR marker. (26) Tonkawa (Hoijer 1949: 43) hostaxso:n xilipa:nanoklaknoʔo ha:ʔako:nwa:ʔa:la. in the morning he always went out hunting.evid that man ʔe:lʔila ʔaweykak kwa:lowkak ya:lo:nanklaknoʔo he so doing.SS many deer big ones he always killed them.evid ‘That man went out hunting every morning, it is said. So doing, he always killed many big deer, it is said.’ Muskogean languages routinely use pro-verbs in discourse. For instance, the Creek verb mom ‘be so’ is used to link two sentences, often in narratives. (27) a. b. c. Creek (Martin 2011: 354–355) isti hámk-it inókk-i:-t wâ:kk-ati:-s person one-nom sick-dur-SS lie.sg-histpst-ind ‘A man lay sick.’ mo:m-it i-heywa ó:c-i:-t ô:m-ati:-s be so-SS one-nom exist-dur-SS be.sg-histpst-ind ‘And he had a wife.’ mo:m-it hatâm ifá-n sólk-i:-n ó:c-i:-t be so-SS again dog-acc many-dur-DS exist-dur-SS ô:m-ati:-s be.sg-histpst-ind ‘And he also had many dogs’ Switch-reference and event cohesion 377 Pro-verbs are used in languages that do not have SR on coordinating conjunctions. SR on coordinating conjunctions occurs at the start of its sentence, allowing a link to previous sentences. SR on subordinate or chained clauses is clause-final, so that link is not really feasible. It is thus likely that the use of pro-verbs offers a strategy that allows a switch-reference effect between main clauses in languages whose grammar only allows it between a main and subordinate or chained clause. 16.6 Other structures This section discusses a few other methods of indicating coherence across sentences. These methods are sometimes linked to switch-reference, and sometimes described as switch-reference, but they do not meet the diagnostic criteria for it. 16.6.1 Obviation Obviation is a distinct system of grammar that is widely found in some regions and language families, marking arguments based on their relative discourse prominence. Obviation has been linked to switch-reference on multiple occasions (Hale 1992; Wichmann 2007; Muehlbauer 2012), because it indicates a kind of cohesion throughout a discourse. However, obviation is a distinct system with several differences from SR, with its own mechanisms. These differences are summarized in Table 4. More obviously, there exist languages with both obviation and switch-reference, like Hopi (Uto-Aztecan) (Hale 1992). Tab. 4: Comparison of switch-reference and obviation property SR obviation 1. at clause juncture 2. occurs in complementary pair 3. depends on reference or cohesion marked on nouns restricted to 3rd person interacts with agreement depends on discourse prominence marked on conjunctions or subordinating connectives yes yes yes no no no no yes no yes no yes yes yes yes no While obviation and switch-reference can co-occur, such co-occurrence might be a rare accident. The language groups most widely known for obviation are Algonquian (see Oxford, this volume) and Athapaskan (see Saxon, this volume), and they all lack 378 Andrew McKenzie switch-reference, with the likely exception of Arapaho (Algonquian) (Cowell & Moss 2008), whose speakers spent many years interacting with speakers of SR-languages on the Northern Plains. 16.6.2 Special kinds of reflexives In the Eskimo-Aleut languages, some person suffixes on embedded verbs and possessors vary depending on whether they co-refer with the main clause subject. (28) Central Yup’ik (Woodbury 1983: 296) wangkugneng-tawaam tangvakuneng aavurciiqut us.du-however if they (3.refl) see they will be amused ‘If they see the two of us, they will be amused’ (29) West Greenlandic (Fortescue 1991: 53) nulia-ni takuaa wife-3.refl see-3sg.agt/3s.obj.indic ‘hei saw hisi wife’ This marking is sometimes described as “fourth person” or as “switch-reference” (Woodbury 1983, Fortescue 1991; Pittman 2005; Berge 2011), although it differs in crucial ways from switch-reference morphology (Table 5). Tab. 5: Comparison of switch-reference with Central Yup’ik special reflexives property SR special reflexive 1. at clause juncture 2. occurs in complementary pair 3. depends on reference or cohesion marked on nouns restricted to 3rd person interacts with agreement marked on conjunctions or subordinating connectives yes yes yes no no no yes no no yes yes yes yes no This form also appears on objects and can be considered a “special reflexive,” more like long-distance anaphors (cf. Bok-Bennema 1991) than switch-reference. Berge (2011) shows that these anaphors can refer back to a discourse topic rather than the main subject (30–31), and that is also a feature of long-distance anaphors cross-linguistically (Kitagawa 1981, Xu 1994). In (30–31), the previously established topic is cigarettes, and that is what the co-referent marker (3coref) points back to, rather than the 1st singular subject of the previous clause. Switch-reference and event cohesion 379 (30) West Greenlandic (Berge 2011: 202–3) 1987-arsimi tassanngaannarsuaq pujortarunnaarama 1987-ars-mi tassannga-innaq-suaq pujortaq-junnaaq-gama 1987-years-loc from.then.on-only-big smoke-no.more-1sg.causative ‘in 1987 from then on I stopped smoking [I no longer smoked]’ (31) West Greenlandic (Berge 2011: 203) uanga cigaritsip ajuleramimga uanga cigaritsi-p ajor-leq-gaminga 1sg cigarette-relative be.bad-begin-3coref.sg.subj/1sg.obj.causative ‘cigarettes didn’t like me anymore [began to be bad for me]’ At a deep level of the grammar, it is likely these special reflexives emerge from the same sorts of syntactic phenomena that switch-reference emerges from (Baker & Camargo Souza 2019). However, in a descriptive sense this is a distinct phenomenon of clause cohesion. 16.7. Conclusion This chapter has introduced the concept of switch-reference and discussed how it expresses different kinds of cohesion between clauses. The most common way is marking whether two clauses’ subjects are identical, but it can also mark identity of events or situations from one sentence to another. The chapter also showed the variety of ways that switch-reference can appear in different North American languages. Switch-reference’s meaning, function, and frequency all contribute to its importance in discourses like conversations and narratives. This importance makes SR vital to understand for communities interested in maintaining and revitalizing their languages. SR is a key component of putting sentences together, and thus a key component of teaching learners how to build conversations. It is difficult to translate, though, and even including its meaning in a translation can be very awkward. Due to this difficulty, teachers cannot rely on comparison to the structures of more familiar languages when discussing SR. It has to be described on its own terms by people familiar with how it works, and it is hoped that this chapter can help people gain that familiarity. References Austin, Peter. 1981. Switch-Reference in Australia. Language 57(2). 309–334. Baker, Mark C. & Livia Camargo Souza. 2019. Switch-Reference in American Languages: A Synthetic Overview. In Daniel Siddiqi, Michael Barrie, Carrie Gillon, Jason Haugen & Eric Mathieu (ed.), The Routledge Handbook of North American Languages, 210–232. London: Taylor and Francis. 380 Andrew McKenzie Berge, Anna. 2011. Topic and Discourse Structures in West Greenlandic Agreement Constructions. Lincoln, NE: University of Nebraska Press. Bok-Bennema, Reineke. 1991. Case and Agreement in Inuit. Berlin: Foris Publications. Broadwell, George Aaron. 2006. A Choctaw Reference Grammar. Lincoln, NE: University of Nebraska Press. Cowell, Andrew & Alonzo Moss, Sr. 2008. The Arapaho Language. Boulder, CO: University Press of Colorado. Dahlstrom, Amy. 1982. A functional analysis of switch-reference in Lakhota. In Kevin Tuite, Robinson Schneider & Robert A. Chametzky (eds.), Chicago Linguistics Society (CLS) 18. Chicago: University of Chicago Press. Dayley, Jon. 1989. Tümpisa (Panamint) Shoshone grammar. University of California Publications in Linguistics Vol. 115. Berkeley, California: University of California Press. Fortescue, Michael. 1991. Switch reference anomalies and ‘topic’ in West Greenlandic: A case of pragmatics over syntax. In Jef Verschueren (ed.), Levels of Linguistic Adaptation, 53–80. Amsterdam: John Benjamins. Gijn, Rijk van. 2016. Switch-reference in South America. In Rik van Gijn (ed.), Switch-Reference 2.0. (Typological Studies in Language 114). Amsterdam/Philadelphia: John Benjamins. Gordon, Lynn. 1983. Switch-reference, clause order, and interclausal relationships in Maricopa. In John Haiman & Pamela Munro (eds.), Switch-reference in Universal Grammar, 83–104. Philadelphia: John Benjamins. Gordon, Lynn & Pamela Munro. 2017. Relative clauses in Western Muskogean languages. Glossa: A Journal of General Linguistics 2(1). 30. Graczyk, Randolph. 2007. A Grammar of Crow. Lincoln, NE: U of Nebraska Press. Haiman, John & Pamela Munro (eds.). 1983. Switch-reference and Universal Grammar. Amsterdam: John Benjamins. Hale, Ken. 1992. Subject obviation, switch-reference, and control. In Richard K. Larson, Sabine Iatridou, Uptal Lahiri & James Higginbotham (eds.), Control and Grammar. Studies in Linguistics and Philosophy, 51–77. Dordrecht: Kluwer. Hardy, Heather K. & Lynn Gordon. 1980. Types of adverbial and modal constructions in Tolkapaya. International Journal of American Linguistics 46(3). 183–196. Hoijer, Harry. 1949. Tonkawa syntactic suffixes and anaphoric particles. Southwestern Journal of Anthropology 5(1). 37–55. Jacobsen, William H., Jr. 1967. Switch-Reference in Hokan-Coahuiltecan. In Dell Hymes & William Bittle (eds.), Studies in Southwestern Linguistics, 238–263. The Hague: Mouton. Kitagawa, Chisato. 1981. Anaphora in Japanese: Kare and Zibun. In Ann Farmer & Chisato Kitagawa (eds.), Coyote Papers 2: Proceedings of the Arizona Conference on Japanese, 61–75. Tucson, AZ: University of Arizona Linguistics Circle. Martin, Jack. 2011. A Grammar of Creek (Muskogee). Lincoln, NE: University of Nebraska Press. McKenzie, Andrew. 2012. The role of contextual restriction in reference-tracking. Amherst, MA.: University of Massachusetts Amherst dissertation. McKenzie, Andrew. 2015. A survey of switch-reference in North America. International Journal of American Linguistics 81(3). 409–448. McKenzie, Parker. n.d. Account of Poolant’s killing. Recorded by Laurel Watkins. Oklahoma Historical Society Research Center. Miller, Amy. 2001. A Grammar of Jamul Tiipay. The Hague: Mouton de Gruyter. Mithun, Marianne. 1993. Switch-reference: Clause combining in Central Pomo. International Journal of American Linguistics 59(2). 119–136. Mithun, Marianne. 2020. Inside contact-stimulated grammatical development. In Patience Epps, Danny Law & Na’ama Pat-El (eds.), Historical Linguistics and Endangered Languages: Exploring Diversity in Language Change. Routledge Series in Historical Linguistics. New York: Routledge. Switch-reference and event cohesion 381 Mixco, Maurício. 1997. Mandan Switch Reference: A Preliminary View. Anthropological Linguistics 39(2). 220–298. Muehlbauer, Jeffrey. 2012. The relation of switch-reference, animacy, and obviation in Plains Cree. International Journal of American Linguistics 78(2). 203–238. Munro, Pamela. 1976. Two Stories by Nellie Brown (Mojave). In Yuman Texts, Margaret Langdon ed. Supplement to the International Journal of American Linguistics 1:3. University of Chicago Press. p. 43–50. Munro, Pamela. 2017. Chickasaw quantifiers. In Denis Paperno & Edward L. Keenan (eds.), Handbook of Quantifiers in Natural Language. Vol. II, 113–201. Amsterdam: Springer Netherlands. O’Connor, Mary Catherine. 1993. Disjoint reference and pragmatic inference: Anaphora and switch reference in Northern Pomo. In William A. Foley (ed.), The role of theory in language description, 215–242. (Trends in Linguistics, Studies and Monographs 69). Berlin: Mouton de Gruyter. Peachey, Robert M. 2006. On switch-reference and the Internally-Headed Relative Clause construction in Washo. Unpublished manuscript. Press, Margaret L. 1979. Chemehuevi: A Grammar and Lexicon. Berkeley, CA: University of California Press. Pustet, Regina. 2013. Switch-reference or coordination? A quantitative approach to clause linkage in Lakota. International Journal of American Linguistics 79(2). 153–188. Roberts, John. 1997. Switch-Reference in Papua New Guinea. In Andrew Pauley (ed.), Papers in Papuan Linguistics No. 3, 101–241. Canberra, ACT, Australia: Australian National University. van Gijn, Rik. see Gijn, Rijk van. Watkins, Laurel. 1984. A Grammar of Kiowa. Lincoln, NE: University of Nebraska Press. Watkins, Laurel. 1993. The Discourse Function of Kiowa Switch-Reference. International Journal of American Linguistics 59(2). 137–164. Wichmann, Søren. 2007. The reference-tracking system of Tlapanec: Between obviation and switchreference. Studies in Language 31(1). 801–827. doi: 10.1075/sl.31.4.04wic. Woodbury, Anthony C. 1983. Switch reference, syntactic organization, and rhetorical structure in Central Yup’ik Eskimo. In John Haiman & Pamela Munro (eds.), Switch-reference in Universal Grammar, 291–315. Philadelphia: John Benjamins. Xu, Liejiong. 1994. The antecedent of ziji. Journal of Chinese Linguistics 22(1). 115–137. IV Discourse Anthony K. Webster 17 Verbal art Abstract: This chapter outlines some of the important linguistic and cultural issues concerning Native American verbal art. While not attempting to be exhaustive, this chapter does attempt to highlight a variety of verbal artistic traditions. After providing a definition of verbal art and discussing some of the ethical issues concerning documenting verbal art, the chapter turns to an overview of various topics related to verbal art. Topics include the importance of performance for thinking about verbal art (1), the question of locally understood genres (2), poetic structuring (3), the ways speech play provides key poetic devices (4), and the evaluative criteria by which verbal art is interpreted (5). Particular focus is placed on concerns with punning, ideophony, poetry, song, narratives, placenames, and parallelism (of various kinds). At the end of the chapter is an extended bibliography for further readings. Introduction Native American verbal art covers a diverse range of discursive practices. This chapter outlines a number of key aspects of such practices. It does not pretend to be inclusive of the full range of that diversity. As a useful heuristic, Sherzer and Woodbury (1987: 8) define verbal art this way: “Verbal art” is a community’s own conception of what in language use is aesthetically or rhetorically pleasing, the forms and processes that members of the community label or otherwise demonstrate they consider to be verbally artistic. To add to this and to sharpen our focus a bit, it is well to keep in mind that verbal art as a cover term is a bit misleading. First, verbal art is often multimodal, including gesture (Wiget 1987) or sign languages such as Plains Indian Sign Language (Farnell 2002) or various inscriptive practices such as O’odham calendar sticks (see Darling and Lewis 2007; Schermerhorn 2019). Second, verbal art should not be an a priori assumption. Rather, what is and is not considered verbal art according to local aesthetic standards needs to be investigated and not imputed. Third, any Western derived distinction between “art” and other modes of action—ritual or otherwise—needs to be dispensed with. Native American verbal art can do multiple things simultaneously and local theories of language functions need to be attended to. It is simply not the case for many Native American aesthetic traditions to talk of “art for art’s sake”—rather verbal art is often used to restore the world, cure the sick, or otherwise affect changes in the world. Such a recognition leads, then, to a realization that there may be limits on the kinds of verbal art that can be documented or shared. Among the Zuni (ISO zun), for example, https://doi.org/10.1515/9783110600926-017 386 Anthony K. Webster there is a distinction (B. Tedlock 1995) between tso’ya and attsani (the beautiful and the dangerous), where beautiful things—including certain genres of verbal art—can be shared and dangerous things should not be shared. Some genres are public, other genres private. Likewise, certain genres may be off limits to certain people based on gender, as is the case among the Mvskoke (ISO mus) (Innes 2010) or based on community status (whether or not one is initiated into certain ritual societies), as described for the Tewa of Arizona (ISO tew) (Kroskrity 2012). Even genres, like Navajo Coyote stories, that have been much documented, may need to be returned and destroyed after the death of the narrator (Toelken 1998). Such considerations should remind us of the social nature of verbal art and, as well, of our ethical obligations towards such verbal art and the people who create such verbally artistic discourse. Much of this research has been done under the name of ethnopoetics (Hymes 1981), for purposes here we can think of the ethno- as highlighting the need for ethnography (that is, talking with people about their verbal art) and the -poetics as indicating some attention to the literary devices used in verbal art (Webster 2020). In what follows I will discuss the issue of performance (1), then the question of locally understood genres (2), then I will discuss poetic structuring (3), I turn after that to a discussion of various poetic devices (4), and I conclude by discussing the evaluative criteria by which verbal art is interpreted (5). 17.1 Performance Performance is a key concept in thinking about Native American verbal art because it challenges us to think beyond a text-centered view of verbal art. Rather, it forces researchers to confront the situated nature of the doing of verbal art (Bauman 1984). Such performances are emergent as well, sensitive to and creative of context (Bauman 1984). Narrators may adjust the narrative according to who is present or how the audience responds. Verbal art becomes, then, less about the product, but rather the process, the doing-things with languages and gestures (and whatever else may be used) in the sphere of meaningful behavior. Textualized versions, that is versions written down, of verbal art often obscure the performance features of loudness, prosody, pacing, accompanying gestures, and the like (see Section 3.2). Verbal art is enacted. In many of the early texts collected by the Boasians, the audience was often the narrator and the anthropologist or linguist, sometimes as well an interpreter.1 Contrary to what might be 1 Briefly, the “Boasians” refers to anthropologists and linguists trained by the anthropologist Franz Boas and later his students in the early twentieth century. Documenting Indigenous languages was a cornerstone of the early tradition of Boasian anthropology. Such work was done before the wide-spread use of sound recording devices and so the forms of verbal art were often taken down through the slow process of elicitation. Here the anthropologist and/or linguist (the distinction was not so clear at that Verbal art 387 assumed, such texts often reveal the ways that performers used such interactions for complex, often locally meaningful, forms of social action. Verbal art is here in the service of reflexive commentary on the interactional moment—that is, it is a way for the narrator to comment on the participants involved (including the anthropologist). Darnell (1989) provides a useful analysis of the various contextual issues that informed—and constrained in certain respects—the performance of a Cree narrative (ISO cre). Wiget (1987) provides a detailed analysis of a video-recorded Hopi Coyote narrative told by Helen Sekaquaptewa (ISO hop). Having said this, it is also the case that some genres (discussed below) are more or less open to creative adaptations. Some genres, ideally, are to be performed verbatim—that is, they are assumed to be verbatim reproductions of prior performances and are evaluated accordingly. 17.2 Genres It is sometimes assumed that verbal art must be long and elaborate. But the question of what constitutes verbal art in a particular community is, in fact, an empirical question. It is a question of genre. Some genres are rather lengthy, taking hours to complete. Other genres, the skillful deployment of a place-name for example, may be rather fleeting. So too the use of puns—a form of speech play that can be mobilized in the service of verbal art—can be fleeting. Some genres are meant to be repeated verbatim. Certain songs, for example, tend to be relatively stable over time, even if the ordering of songs may differ from performance to performance (Bahr 1986; Bahr et al 1997, Loether 1993). Other genres are understood to be more ephemeral. An assumption inherited from the Boasian tradition saw verbal art as residing in some static whole text, the complete narrative. But verbal art need not be a single “complete” narrative. Scollon (2009: 261) reports that, “within the Athabaskan storytelling tradition one doesn’t waste words or insult one’s listener by telling somebody something he or she already knows. A truly knowledgeable person really only requires an allusion to the story.” Such skillful allusions, then, constitute a site of verbal artistry. Likewise, Kendall (1979: 146) notes that for Yavapai and Hualapai (ISO yuf), Coyote tales—a particular genre that involves the trickster figure Coyote—were meant to be elliptical—the idea was that people knew the stories and could fill in various parts. Such expectations about the shape and form of various narratives need to be investigated. time) wrote the narrative in a phonetic system (later a phonemic system) while the Indigenous narrator told the narrative (often slowly and often repeatedly). Given the emphasis of the Boasians on using the texts to create a grammar and dictionary, careful attention to the verbal artistry of the narratives was often not done (see Epps, Webster, Woodbury 2017 for a history of this tradition). The narratives were almost always treated as prose. Songs were treated as poetry. Ethnopoetic work, to be discussed below, has suggested that such narratives are better represented as a series of lines and verses and not as prose. 388 Anthony K. Webster Some genres may have recognizable names—that is to say, they are locally categorized as distinct genres. Various Native communities differ in the elaboration of such named genres. For example, Shaul (2002: 6) gives the following list of named genres among the Hopi (see also Sekaquaptewa, Hill, and Washburn 2015): (1) Hopi (Shaul 2002: 6) tuuwutsi (stories) (a subcategory istutuwutsi ‘coyote stories’) tsa’alawu (announcements) wimtsa’alawu (ritual announcements) qa wimtsa’alawu (secular announcements) unangwvàasi (prayers) wimlavayi (ritual speech) lavay’oyi (speech) taawi (song) puwati/puwvitstawi/titaptawi (lullabies) tsakotawi (children’s game songs) tuwutstawi (story songs) sosokwtawi (gambling songs) ngumantawi (grinding songs) yewatawi (songpoems) katsintawi (kachina songs) Among Zuni, narratives fall into two broad categories: chimiky’ana’kowa (origin stories and considered true) and telepnaawe (tales and considered fiction) (Tedlock 1983: 159– 160). Cup’ik (ISO esu) narrative traditions are also divided into two broad categories: Qanemcit ‘narratives’ and Qulirat ‘tales’. Qanemcit are understood to have originated with a particular person, often based on real people; whereas Qulirat are said to have originated with remote ancestors (this seems their defining feature) (Woodbury 1984: 13). Genres may often be demarcated by formal styles, a co-occurrence, that is, among various linguistic features (formality being a local categorization). Consider, for example, the Seneca (ISO see) ways of speaking described by Chafe (1993). Chafe highlights the ways Seneca speaking styles diverge along four dimensions (prosody, formulaicness, sentence structuring, and source of knowledge), as seen in Table 1: Tab. 1: Dimensions of Seneca Styles of Speaking (adapted from Chafe 1993: 86) Conversation Good Message Thanksgiving Speech Prosody Formulaicity Sentences Epistemology Free Low Fragmented Uncertain Somewhat stylized Moderate Somewhat integrated Unexpressed Highly stylized High Highly integrated Certain Verbal art 389 In Seneca, then, the Thanksgiving Speech is marked by way of a particular formal style that includes highly stylized prosody, formulaic utterances, a highly integrated sentence structure, and explicit use of epistemic particles such as wai(h) (‘indeed’, ‘for sure’) and verbs such as tkaye:i’ (‘it is a fact’) (Chafe 1993: 86). Genres are often recognized by way of opening and closing formulas (Zuni, Eastern Pomo). In Kumeyaay, spoken in southern California and Baja California (ISO dih), for example, trickster tales begin with the following formulaic opening: (2) Kumeyaay (Field and Cuero 2012: 325) Ke’nápa nyuuchs It’s an old story. Nyuuch yúsa. It’s old. Nyuu, It’s old, nyuu yus ’i mat. it’s old, I say. Ke’nápa nyuuch nyáasa: It’s an old story I am telling you. In Zuni telapnanne, they often begin with the formulaic opening so’nahchi said loudly and concludes with the formulaic closing lee semkonikya (Tedlock 1983: 160–161). In both cases, neither word is meant to be understood referentially or semantically; rather it indexes that a telapnanne is about to begin or to end. It creates, then, a narrative frame. Inside the frame, one is to understand what is being said as a part of the genre telapnanne. Genres can also be recognized by the use of genre signature devices (particular linguistic forms that must be present often throughout a particular genre) (Shaul 2002: 7). In both Hopi (a Uto-Aztecan language) and Tewa of Arizona (a Kiowa-Tonoan language), traditional narratives employ an evidential particle (respectively yaw and ba) which occurs in almost every clause of a narrative outside of quoted speech. Here is an example of the use of ba ‘so they say’. I follow Kroskrity’s (2012: 162–163) formatting with the one exception that I have bolded ba when it occurs: (3) Tewa of Arizona (Kiowa-Tanoan) (Kroskrity 2012: 162–163) Kídí di-da-kelen ‘haedi ba And then, after they got quite strong ‘a:khon-ge-pe’e ba over to the plain, so ‘óóbé-khwóóli-n-di im-bi yiyá-‘in-di. they were flown by their mothers. When we think of verbal art as performance we realize, as well, that non-linguistic features can also signal genres. For example, some narrative traditions could only be told at night and in the winter—such was the case for Coyote stories among the Navajo (Toelken and Scott 1981). Recalling the two broad genres among the Zuni, chimiky’ana’kowa can be told at all times of the year, while telepnaawe should be told only at night and during the winter (Tedlock 1983: 160). 390 Anthony K. Webster 17.2.1 Names In some Native communities, place-names and/or personal names can be considered forms of verbal art. Among the Western Apache (ISO apw), there is a genre of verbal art called yałti’ bee’ízhí (speaking with names) (Basso 1996: 80). When this genre of verbal art occurs, place-names are “fixed up” (náyidlé). That is to say, they occur both in their long form and with an emphatic enclitic (=né ‘at this very place’). Thus, T’iis Tl’áh ‘Olį í ̨ ’́ (Water Flows Inward Under a Cottonwood Tree) becomes, in this genre, T’iis Bitł’ah Tú ‘Olį ń é (Basso 1996: 90). Here is an example of “speaking with names” from Basso (1996: 79). While place-names are considered aesthetically pleasing ways of speaking, the genre is also used for interpersonal and moral instruction, and revolves around the uttering of the fixed-up place-names. In this case, two Apache women (Lola and Emily) are attempting to comfort another Apache woman (Louise) (I have followed the formatting conventions of Basso 1996: 79). Notice that in the examples below, the enclitic =né is variously realized as =yé (in Lola’s first turn and Emily’s reply) and then as =né (in Lola’s second turn) (I have bolded the enclitic). (4) Apache (Basso 1996: 79) Lola: Tséé Hadigaiyé yú ‘ágodzaa. (It happened at Line of White Rocks Extends Up And Out, at this very place!) [Pause: 30–45 seconds] Emily: Ha’aa. Túzhį’ Yaahigaiyé yú ‘ágodzaa. (Yes. It happened at Whiteness Spreads Out Descending To Water, at this very place!) [Pause: 30–45 seconds] Lola: Da’aníí. K’is Deeschii’ Naaditiné yú ‘ágodzaa. (Truly. It happened at Trail Extends Across A Red Ridge With Alder Trees, at this very place!) While in “Speaking with Names,” a single place name may be uttered, its force comes from being associated with ‘ágodzaahí (historical tales) (Basso 1996: 49). Indeed, one key genre feature of Western Apache narratives is that they begin at named places (Basso 1996). Such uses of place-names to anchor narratives is quite common and can also be found among the Eastern Pomo (ISO peb) (McClendon 1977) and the Zuni (Tedlock 1983), for example. In each case, the use of the place-names grounds the narrative in a particular named and knowable world. Personal names too can be forms of verbal art. Whiteley (1998) describes the way Hopi names are forms of literature. They are, according to Whiteley (1998: 107), “individually authored poetic compositions that comprise a literary genre.” Northern Yuman (ISO yuf) personal names can also be understood as compressed texts, in need of interpretation and exegesis (Kendall 1980). Verbal art 391 17.2.2 Puns While puns are not often considered in the Western tradition as forms of verbal art, in many Native communities, puns are a highly valued aesthetic practice. They can either stand alone, as fleeting puns in conversations, or they can be used in other genres (from narratives to emceeing at Powwows to rituals). This is yet another place where speech play and verbal art come together. Puns can be of two sorts. The first sort are what we can call intralingual puns (these are puns within a language): (5) Seneca (Chafe 1998: 187) agwásgane:s ‘we wish’ agwás gane:s ‘very long penis’ Note the way two words can be punned with one single word. The second sort of puns we can call interlingual puns (these are puns that work because they cross languages). The following example is a rather old pun on the Navajo Nation. (6) Navajo (ISO nav) (Webster 2010: 291) television ‘television’ télii alizhgo ‘urinating donkey’ Gelo (1999) and Palmer (2003) report the Kiowa (ISO kio) and English interlingual pun of bót ‘cow innards’ with ‘boat’ leading to such puns as love bót (after the television show The Love Boat). Puns can both stand alone as forms of verbal art or can occur in a variety of verbally artistic genres (see below). 17.2.3 Songs Among the more studied aspects of Native American verbal art are the song traditions. Loether (1993) notes that, for Western Mono (ISO mnr), songs are one place where Indigenous languages may continue after their use in day-to-day interactions has ceased. Important linguistic analysis of specific song traditions can be found, for example, in Hinton (1984) on Havasupai (ISO yuf) songs and in Enrico and Stuart (1996) on Northern Haida (ISO hdn) songs. Hinton (1980) and Frisbie (1980) have discussed the role of vocables—semantically meaningless syllables—as structuring devices in the song traditions of Havasupai and Navajo, respectively (see also Hymes 1981). The early ethnomusicological work on Navajo song traditions by McAllester (1954, 1980a) highlighted a crucial aesthetic sensibility: namely, that for many Navajos, “beauty is that which does something” (McAllester 1954: 72). McAllester (1980a) outlines a number of the poetic devices found in Navajo songs, including repetition, interruption, alternation, ambiguity, imitation, identification, and continuation. Both Hinton (1990) for Havasupai and Fitzgerald (1998) for Tohono O’odham (ISO ood) have shown that meter can be found in Native North American verbally artistic traditions. 392 Anthony K. Webster A number of song traditions use a special song language (see Bahr 1984). Such song languages may manipulate everyday words by adding syllables or they may reduce grammatical details (deleting grammatical affixes or particles, for example). Shoshoni (ISO shh) song traditions often omit the subject, leaving the line ambiguous, and will use obscure or obsolete words as well (Crum, Crum, and Dayley 2001: 19–20). To give a sense of the changes involved in song languages, below are the first lines of the First Snake Song as sung by Ray Winnie in 1957. The song language is first, then everyday Navajo, and then an English translation (following McAllester’s 1980 presentation). Note the addition of syllables like -ye, -yoya, -yeye, and the reduction of long vowels to short vowels. Note as well the use of vocables at the end of the song lyrics for the first verse. (7) Navajo (McAllester’s 1980: 3) K’a dine dighini-ye, k’a K’ad dinééh dighiní k’ad Now man holy now Tł’iyitso Tł’iistsoh Snake big yołtełigo yołtéełgo holding while dine dinééh man nohani yo-he nohaanííyá to us has come dighini-ye dighiní holy nohani-yoya-yeye nohaanííyá to us has come yaha ‘eye neya ŋeye ŋa yaha ‘eye neya ŋeye ŋa Here we have seen how ordinary Navajo can be changed into song language (through the addition of syllables and the reduction of vowel length). Thus song language is a manipulated form of everyday language—it is a form of speech play, in that respect in the service of verbal art. Bahr (1983) and Bahr, Paul, and Joseph (1997) point out, in discussing Tohono O’odham songs, that it is necessary to produce four versions: a transcript of the song language, an ordinary language version, an ordinary English language version and, importantly, an English song language version. Here is an example of an O’odham Ocean song (Bahr 1991:544). For the sake of brevity, I have not included the ways Bahr indicates various rhythmic features of the song. Note as well, as with the Navajo example, the song language includes more syllables than the ordinary language version. (8) O’odham (Bahr 1991:544) Song Language Ka ci me su na ni Ka ci me su na ni ku kugi me Toi ta ge da mai Ma ma to ne wu pa him Ordinary Language and Literal Translation Ka:cim su:dagi Staying water Ka:cim su:dagi kukughim Staying water terminating T-oidag da:m Our-field on-top Mamtod wurpahim. Seaweed tossing. Verbal art 393 Free Translation STAYING WATER, STAYING WATER EXTENDS, OUR FIELDS’ TOPS TOSSED WITH SEAWEED Bahr translates the song language into ordinary O’odham. He then translates that ordinary O’odham into a literal English translation (represented directly below the ordinary O’odham). Finally, the literal English translation is translated into a free English translation (the All caps indicating this is song language). Much recent work on Native North American song traditions has been published. See, for example, Browner 2009 and Levine and Robinson 2019 for useful recent edited volumes. 17.2.4 Stories As we saw above concerning the genres of Hopi verbal art, narratives too are often divided into locally relevant genres. Such genres may be recognizable, as discussed earlier, based on form (genre signatures, for example) or content or function—or in combination (genres need not always be clearly distinguishable). Basso (1996: 48–51) provides a useful breakdown of Western Apache narrative genres. (9) Western Apache narrative genres (Basso 1996: 48–51) nagoldi’é ‘narrative story’ Major categories godiyįhgo nagoldi’é (myth) ‘ágodzaahí (historical tales) nłt’éégo nagoldi’é (saga) ch’idii (gossip) Minor categories ma’ highaałyú nagoldi’é (Coyote stories) binííbaa’ nagoldi’é (seduction tales) The major categories can be organized based on when the narrative is set, and on the function or purpose of the narrative, as illustrated in Table 2: 394 Anthony K. Webster Tab. 2: Genre Distinctions for Major Categories of Western Apache narratives (after Basso 1996: 50) Narrative Category Temporal Locus of Events Purpose godiyįhgo nagoldi’é (myth) godiyąąná’ (in the beginning) doo’ánííná (long ago) dííj̢í̢ígo (modern times) k’ad (now) to enlighten, to instruct ‘ágodzaahí (historical tales) nłt’éégo nagoldi’é (saga) ch’idii (gossip) to criticize, to warn, to “shoot” to entertain, to engross to inform, to malign 17.2.5 Poetry Verbal art does not remain static. New genres may emerge, inspired and informed by other genres. Societies, cultures, and verbally artistic traditions are not hermetically sealed. For example, a number of Native Americans write poetry. This poetry is influenced by both larger traditions of poetry (Western or Eastern), and by locally relevant verbally artistic traditions. Contemporary Native American poets write and perform in English, local varieties of English, their Indigenous languages, or in some combination of them. Among the Tohono O’odham poems are called ha-cegĭtodag (‘thoughts’) (Zepeda 2019: 8). Some Navajos classify contemporary poetry as a form of hane’ (‘narrative’, ‘story’) (Webster 2009). The use of punning and ambiguity more generally are important features in Navajo verbal art and life, and can also be found in the poetry of a number of Navajo poets. Many of the poems in Rex Lee Jim’s (1995) saad are predicated on puns. So, for example, a poem may play with the ambiguity of na’astsǫǫsí (‘mouse’, ‘the one that goes about sucking’), which can be heard as náá’ásts’ǫǫs (‘to perform a sucking ceremony again’) (see Webster 2018: 32). 17.3 Poetic structuring Following the lead of Melville Jacobs (1959), a number of scholars have sought to rethink Native American narrative traditions in ways other than prose. One particularly fruitful school of thought has been ethnopoetics. This work has sought, through a variety of presentational formatting principles, to highlight various poetic structuring devices, often presenting such narratives in a format resembling Western poetry (thus narratives were presented in lines). Early work, especially work focused on narrative traditions taken down in dictation, tended to focus on recurrent forms such as particles and form-content parallelism (Hymes 1981, 2003). Lines were then defined by such recurrent forms. Other work, often focusing on performances that were recorded with audio-recordings, highlighted prosody and pause structure (Tedlock 1983). Lines here are defined by pause structure. Much work has sought to combine the insights of both “particle” and Verbal art 395 “pause” (see Bright 1979; Woodbury 1985, 1987; Kroskrity 1985, 1993; McClendon 1982; V. Hymes 1987; Doak 1991; Watahomigie and Yamamoto 1998.). I turn now to the use of particles, both initials and finals, as line creating devices. It is important to stress that such uses need to be investigated and not imputed. Likewise, while I have divided the discussion into initials and finals, this does not mean that the two are mutually exclusive. There is often an interplay between, for example, initial adverbial particles and a clause final quotative particle (Hymes 2003). The two work in conjunction to create lines. Likewise, Kimball (1993: 5) notes that lines in Koasati (ISO cku) narratives are created by way of the use of a switch-reference suffix -n and/ or initial conjunctions. 17.3.1 Particles In this section, I highlight the use of particles and enclitics in the creation of meaningful discursive units, following the lead of Hymes (1981). In general, Hymes’s method sought to discern various structuring principles of narratives. The narratives could then be represented as a series of lines, verses, stanzas, acts, and scenes. 17.3.1.1 Initials Initials are words or particles that appear at the beginning of a clause. They are often referred to as discourse particles. Initials, as we will see below, often have an adverbial quality to them. They are one key site for the poetic structuring of discourse. Hymes’s (1981; 2003) early work on Chinookan narrative structuring focused on the use of initial particles in the creating of verses (a grouping of lines). For example, in Wasco-Wishram (ISO wac), the initial particles aga kwapt (‘now then’), nawit (‘straightaway’), agawit’a (‘now again’), kwadau (‘and’, ‘before’), and qidau (‘that is the way’) were shown to be crucial in the structuring of narratives (Hymes 2003: 217). In Upper Chehalis (Salish) (ISO cjh), speaker Silas Heck uses the initial particle húy (‘and so’) as a major structural device (Kinkade 1987: 283). In Ojibwe (Algonquian) (ISO oji) narratives, mii (‘so’) and mii dash (‘and then’) play a major structural role and act as a key rhetorical device (Ghezzi 1993; Valentine 1995; Spielmann 1998). Spielmann (1998: 201), following the work of Valentine (1995), suggests that “mii marks the advancement of the storyline.” Its use doesn’t just signal particular discourse units (lines, stanzas or verses), but also keys the listener into important moments of plot advancement. Ghezzi (1993) qualifies this by suggesting that the use of mii dash may be a part of a particular narrator’s style. We will see this again when we look at reduplication in Southern Paiute (ISO ute) narratives, where reduplication is a poetic option, actualized by individual narrators as a part of their style (Bunte 2002). It is, of course, only possible to attend to individual style when it can be held up in relief against the options other narrators have taken. 396 Anthony K. Webster 17.3.1.2 Finals Finals are forms that occur at the end of clauses. In the discussion below, they are either enclitics or quotatives that are clause final. In Chiricahua Apache (Athabaskan) (ISO apm) Coyote narratives, the narrative enclitic =ná’a ‘so they say’ is often used at the end of every clause outside of quoted speech. It serves at least three discursive functions: 1) it places the narrative within the voice of authority; 2) it acts as a genre signature; and 3) it creates meaningful discursive units that—following Hymes (1981)—we can call lines (Webster 1999). In Maidu (ISO nmu), the sentence final-quotative ac’ójʔam ‘it is said’ occurs throughout Maidu narratives (Nevins 2017: 30). Here again, it functions both as a genre signature device and as a structuring principle, creating lines through its rhythmic repetition. Such interplay between grammatical functions and poetic structuring are examples of what Sherzer (1990: 18) calls the poeticization of grammar. 17.3.1.3 A Karuk example Here, then, is an example of the representation of a narrative based on the Hymesian principles of measured verse. The example comes from Bright’s (1984) |