Papers of the Fifty-Third Algonquian Conference / Actes du cinquante-troisième Congrès des Algonquinistes (Papers of the Algonquian Conference) 1611864895, 9781611864892

Citation preview

Papers of the Fifty-Third Algonquian Conference Actes du cinquante-troisième Congrès des Algonquinistes

PA P E R S O F T H E A LG O N Q U I A N C O N F E R E N C E S ACTES DU CONGRÈS DES ALGONQUINISTES

William Cowan Founding Editor Inge Genee, Monica Macaulay, and Margaret Noodin Editors Laura Moquin and Macy Floyd Editorial Assistants Marie-Pierre Bousquet and Wesley Y. Leonard Associate Editors

53 Papers of the Fifty-Third Algonquian Conference

Actes du cinquante-troisième Congrès des Algonquinistes

edited by Inge Genee, Monica Macaulay, and Margaret Noodin

michigan state university press  |  east lansing

Copyright © 2024 by Michigan State University

i The paper used in this publication meets the minimum requirements of ANSI/NISO Z39.48-1992 (R 1997) (Permanence of Paper).

p Michigan State University Press East Lansing, Michigan 48823-5245 Paper ISBN:978-1-61186-489-2 PDF ISBN: 978-1-60917-759-1 Cover design by Erin Kirk Cover detail of a quill box is used courtesy of Julie L. Loehr

Visit Michigan State University Press at www.msupress.org

Contents

ix Preface Marie-Odile Junker and Christopher Cox



1 The Assumption of a Further Obviative Irene Appelbaum 17 Possessive Prefixes across Algonquian: Evidence for Two Separate Paradigms Ksenia Bogomolets

39 Stories of Empowerment: Indigenous Women’s Leadership Summer School Maggie Chittspattio, Nadia Ferrara, Ashley Guanish, Mary Jane Hannaburg, Denise Larocque, Amélie La Rose Gauthier, Anahi Morales Hudon, Anny Morissette, and Marie-Louise Perron 57 Animate Intransitive Consonant Stems in North East Cree Vincent Collette 75 sk and sp in Illinois David J. Costa 91 kwêyask kotahâskwâtam: The Effects of Altering Specificity in WordNet on the Accuracy of Computational Semantic Classifications of Plains Cree (nêhiyawêwin) Daniel Dacanay, Jolene Poulin, and Antti Arppe

107 The Meskwaki Past Tense Enclitic =iyo·we in Relation to an Affixal Past Tense Strategy Amy Dahlstrom 125 Ezhi-Aanikanootamaang Ogimaans: The Little Prince Translated into Ojibwemowin Jaymee Dhein, Angela Mesic, Margaret Noodin, and Susan Wade 141 Deverbal Noun Composition in Mille Lacs Ojibwe Hunter Johnson and Christopher Hammerly 161 Syllabicity of Dorsal Fricative [X] in Blackfoot: An Empirical Investigation Mizuki Miyashita 179 Patterns of Cree Preverb Usage in Early Child Language: A Longitudinal Case Study Burak Oney 197 Short Verbs in Algonquian Richard A. Rhodes 219 Reflections of a “Barbarous, Mentally Undeveloped” Linguist: On the Legacy and Obligations of Algonquian Studies and Linguistics Mskwaankwad Rice 235 Plains Cree Textual Analysis with PCA: Across the Bloomfield and Ahenakew-Wolfart Subcorpora Katherine Schmirler and Antti Arppe 253 Morphophonological Rule Development and Real-Time Rule Testing with XFST: A Model for Blackfoot Katherine Schmirler, Antti Arppe, and Inge Genee 269 On Diphthongs and Digraphs in Blackfoot Natalie Weber and Mizuki Miyashita

289 Rhythm and Intonation in Oji-Cree Matthew Windsor 311 Oblique Argument Licensing in Algonquian: Head Marking vs. Dependent Marking Yadong Xu 331 Contributors

Preface Marie-Odile Junker and Christopher Cox

T

he 53rd Algonquian Conference was organized and hosted by Carleton University faculty members Marie-Odile Junker and Christopher Cox. As with the preceding Algonquian Conference organized at the University of Wisconsin–Madison in 2020, uncertainties around the ongoing COVID-19 pandemic led us to plan the 2021 meeting as a fully online event. While this format facilitated broader access to the conference than a comparable in-person meeting may have in several respects, as we note below, it also required significant technical and logistical support from many staff members and offices at Carleton University to implement successfully. We wish to thank Mike Barker (School of Linguistics and Language Studies), Aaron Brown (Information Technology Services), Alex Pilkington (Carleton University Event Support), Steven Reid (University Communications), Andrew Riddles (Web Services), Delasie Torkornoo (Algonquian Dictionaries and Language Resources Project), Nick Ward (Faculty of Arts and Social Sciences—Communications), and Andrew Yuill (Faculty of Arts and Social Sciences Computing Support Unit) for their assistance with these aspects of the meeting. A dedicated team of Carleton University student volunteers—including Ashley Promislow, Donna Fenton, Lara Russo, and Meghana Akavoor Krishnannamboothiripad—also assisted

|

   ix

|

x   Preface

with moderating sessions, providing technical assistance, and editing conference video recordings. Special acknowledgment is due to Étienne Nemes for his months of dedicated work as the graduate student assistant responsible for all aspects of conference organization, and to Katie Martinuzzi for her help in editing conference video recordings. Financial support for the conference was provided by Carleton University’s Office of the Dean of the Faculty of Arts and Social Sciences and is gratefully acknowledged here. The conference was held from October 14–17, 2021, with over 120 registered participants taking part in conference presentations, special sessions, and social events using the digital teleconferencing platform Zoom. In addition to the thirty-four papers that were presented at the meeting, this conference also featured two workshops: one organized by Conor Quinn (University of Southern Maine) on the Minimal Course approach to teaching and learning Algonquian languages; and another organized by Anny Morissette (Université Saint-Paul) on the experiences of participants in the Quebec Native Women—Saint Paul University Summer School. A further special session on using Algonquian epistemologies to talk about Algonquian languages was organized by Wesley Leonard (University of California, Riverside) and Marie-Odile Junker (Carleton University), with workshops in both French and English dedicated to discussing the roles and relevance of Algonquian cultural norms in Algonquian language work. This session featured a dedicated question-and-answer session with Nicole Petiquay, coordinator of linguistic services at the Conseil de la Nation Atikamekw, entitled “Parler de ma langue dans ma langue” (Speaking about my language in my language). Video recordings of sessions throughout the conference are available on the 53rd Algonquian Conference website (https://algonquianconference.atlas-ling.ca/eng/conference/53rd-algonquian-conference/program/). While we would have been delighted to meet with conference participants in person in the Ottawa-Gatineau area this year, we recognized early on in the organizing process that the online-only format of the 53rd Algonquian Conference presented a special opportunity to work toward greater accessibility of the meeting, in general. As previous conference organizers had also reported, adopting a widely available video teleconferencing platform such as Zoom allowed us to open the conference to participants from many Algonquian Nations and countries around the world while keeping registration costs low. Moreover, with the assistance of the Paul Menton Centre for Students with Disabilities at Carleton University, the Sign Language Interpreting Associates Ottawa (SLIAO), and Louise Coté Interpretation,

|

Preface   xi

we were able to offer simultaneous interpretation in American Sign Language in nearly a dozen conference sessions, and simultaneous interpretation in English and/or French in two other workshop sessions—all with the aim of encouraging greater interaction between members of these language communities at the conference. We hope that similar efforts will take root as part of future Algonquian Conferences, whether virtual or in person, and that the meeting will continue to grow in its capacity to offer a welcoming, open, and inclusive meeting space for all those with an interest in Algonquian languages, cultures, and peoples.

Additional Conference Papers

In addition to the papers published in this volume, the following papers were presented at the conference: Antti Arppe (University of Alberta), Eddie Antonio Santos (University of Alberta), Andrew Neitsch (University of Alberta), Jolene Poulin (University of Alberta), Atticus Harrigan (University of Alberta), Katherine Schmirler (University of Alberta), Daniel Hieber (University of Alberta), Arok Wolvengrey (First Nations University): Towards a Morphologically Intelligent and User-Friendly Online Dictionary of Plains Cree. Ksenia Bogomolets (University of Auckland), Paula Fenger (Leipzig University), and Adrian Stegovec (University of Connecticut): The Blocking Effect of Negation on Initial Change: Rescue by Affix Deletion. Zoe Brown (University of Minnesota): “Aaniish gagwejimaasiwad . . .” : An Indirect Speech Act in Southwestern Ojibwe. Rachel Fedorchak, Vade Kamenitsa-Hale, Hunter Thompson Lockwood, and Monica Macaulay (University of Wisconsin–Madison): Strategies for Lexical Expansion in Algonquian Languages. Erik D. Gooding and Lily Lee Gooding (Minnesota State University Moorhead): Decolonizing Meshkwaki Patrilineality (Pre-1937/Post-1937): Coercion and Conflict in Regard to Meshkwaki Tribal Membership through Time and Space. Stephane Goyette (Université Dalhousie): Mens algonquiana in lingua gallica: l’obviatif en français métis. Atticus G. Harrigan and Antti Arppe (University of Alberta): Order as Alternation: The Status of Order as a Valid Category in Nêhiyawêwin. Corinne Kasper (University of Chicago): L2 Potawatomi Language Use.

|

xii   Preface

Robert E. Lewis Jr. (University of Winnipeg): Look! Directive Interjections across the Algonquian Language Family. Wendy Makoons Geniusz, Madalyn McCabe, and Aidan Sanfelippo (University of Wisconsin– Eau Claire): Language Revitalization Programs: What’s Working? What’s Not? Dustin Morrow (University of Minnesota): Consonant Elision in Southwestern Ojibwe Prefixation. Olivia Sammons (First Nations University): Applicative Constructions in Northern Michif. Katherine Schmirler (University of Alberta): Topicality, Overt Participants, and the Algonquian Person Hierarchy: A Plains Cree Corpus Study. Lucy Thomason (Smithsonian Institution): Horrified Realization: The Meskwaki Enclitic Phrase =wi·na=ne·hi. Natalie Weber, Evan Hochstein, Pinyu Hwang, Nico Kidd, Diana Kulmizev, and Hannah Morrison (Yale University): Blackfoot Words: Introducing a Database of Blackfoot lexical forms. Mitsuyoshi Yabe: Network Analysis Visualizations of the Abenakis in the Early 18th Century.

The Assumption of a Further Obviative Irene Appelbaum

A

ll Algonquian languages exhibit obviation.1 Typically, obviation is understood as a binary opposition. In relevant domains where there are at least two third persons, each third person is identified as either proximate or obviative, subject to the constraint that only one third person may be proximate. The referent of the proximate is typically thought to be most discourse prominent in that context. Discourse prominence, in turn, is said to encode a range of meanings including the “topic” (Bliss 2005:65; Bloomfield 1962:38; Oxford 2017:2; Wolvengrey 2005:424) or “focus” (Bliss 2005:65; Wolfart 1973:17) of the sentence, the character with whom the speaker has the most “sympathy” (Thomason 2003:98) or “empathy” (Dahlstrom 2017:52, n.d.:3-22), or whose “point of view” (Bloomfield 1962:38; Dahlstrom 2017:52; Thomason 2003:98) is being represented. Proximate and obviative NPs are additionally thought to participate in a referential hierarchy with proximate NPs ranking above obviative ones. In descriptions of several Algonquian languages, however—e.g., Potawatomi (Hockett 1939, 1948), Cree (Bloomfield 1946), Blackfoot (Frantz 1966; Uhlenbeck 1938), Meskwaki (Dahlstrom n.d.; Goddard and Dahlstrom 2022; Thomason 2003)—obviation is sometimes described as having three values: proximate,

|

   1

|

2    Irene Appelbaum

obviative, and an additional obviative. This additional obviative is variously referred to as “second obviative” (Goddard and Dahlstrom 2022; Hockett 1939, 1948), “secondary obviative” (Thomason 2003), “farther obviative” (Bloomfield 1946), “fifth person” (Frantz 2017; Uhlenbeck 1938), “sub-obviative” (Uhlenbeck 1938), and the term I use in this paper, “further obviative” (Dahlstrom n.d.; Goddard 1984; Oxford 2017). Further obviatives are thought to be grammatically realized and to rank below ordinary obviatives on referential hierarchies. The referent of a further obviative NP is thought to be less discourse prominent than that of an ordinary obviative within the same domain. Further obviative is not thought to be a general inflectional value such that any domain with at least three NPs will have a further obviative. Rather, it is only in two contexts that obviation has been described in tertiary terms: (i) possessed nouns with obviative possessors, and (ii) transitive clauses with two third-person obviative arguments. In the former context, it is the possessed noun that is said to be the further obviative; in the latter context, it is the object of a direct clause and the subject of an inverse clause that are each said to be further obviative. Claims for the existence of further obviatives have long been controversial. But much of the criticism has focused on further obviatives posited in the context of possessed nouns. (See, for example, Wolfart 1973 and 1978, as well as, more recently, Oxford 2017.) In this paper, I discuss claims for the existence of further obviatives in the second context: transitive clauses with two third-person obviative arguments (“double-obviative clauses”). I argue that we are not justified in positing further obviatives in this context either, because (i) there is no unambiguous morphological evidence for their existence, and (ii) the argument that they must exist is circular. Additionally, I argue that positing further obviatives is not theoretically innocuous because it serves to shield a deeply entrenched assumption—obviation determines direction—from empirical evaluation. My discussion is limited to Meskwaki because it is here that arguments for further obviatives in the context of transitive clauses with two obviative arguments are most explicit, but my remarks may be applicable to other languages that posit further obviatives in this context. This paper is organized as follows. In the first section, I examine the morphological data presented as evidence for further obviatives in Meskwaki and show that this data by itself does not allow us to infer that there are further obviatives in Meskwaki; instead, I argue, it is the assumption that there cannot be double-obviative transitive clauses that guarantees the conclusion that further obviatives exist. In the next section, I examine the argument that further obviatives must exist, and show that

|

The Assumption of a Further Obviative   3

it is wholly circular: the assumption that further obviatives must exist depends on the assumption that obviation status determines direction and this assumption, in turn, depends on the assumption that further obviatives exist. In the third section, I suggest that it is the commitment to the view that obviation determines direction, that ultimately fuels the postulation of further obviatives in Meskwaki. I end the paper with a brief summary and conclusion.

The Morphological Evidence for Further Obviatives

In Meskwaki, morphological data is presented for further obviatives in the context of transitive clauses. Dahlstrom n.d. and Thomason 2003 both assert that while there is no nominal inflection for further obviatives in Meskwaki, verbal inflection distinguishes obviatives from further obviatives: Meskwaki . . . has no special nominal inflection for the further obviative. Rather, the distinction between nearer and further obviative can be seen only in transitive verb inflection. (Dahlstrom n.d.:3-20) Verbal inflection . . . makes a distinction between primary and secondary, or nearer and farther, obviative. . . . Animate nominal inflection does not distinguish primary and secondary obviative. (Thomason 2003:8)

The reference to verbal inflection in the above quotations sets up the expectation that a further-obviative agreement morpheme will be identified. However, this is not the case. Looking at the morpheme-by-morpheme gloss of the double-obviative verb forms in (1) and (2) below, we see that there are three suffixes marking features of one of the arguments: -ni, -w, and -ani. In each example, the suffix -w indicates that this argument is third-person, the suffix -ni indicates that it is obviative, and the suffix -ani that it is (obviative) singular. None of these suffixes is a further-obviative morpheme.2 (1) wa·pam-e·-ni-w-ani look.at-dir-obv-3-obv.sg ‘He (obviative) looks at him/them (further obviative).’ (Dahlstrom n.d.:4-16, emphasis added)

|

4    Irene Appelbaum

(2) wa·pam-ekw-ni-w-ani look.at-inv-obv-3-obv.sg ‘He/They (further obviative) looks at him (obviative).’ (Dahlstrom n.d.:4-17, emphasis added)

Of course, the absence of an overtly realized further-obviative morpheme is compatible with there being unambiguous morphological evidence for further obviatives. It is sufficient for an absence of overt morphology to be uniquely associated with a further-obviative interpretation. That is, it is sufficient for other candidate interpretations to be ruled out. For example, the verbal template for Meskwaki transitive animate verbs is such that when a transitive clause has two third-person arguments—whether two obviatives, as in examples (1) and (2), or one-proximate, one-obviative, as in the more usual case—verbal inflection shows features for only one of the arguments. As Dahlstrom notes for a set of direct clauses with two third-person arguments, including the form in (1): “The suffixes following the theme sign encode person, number, and obviation features of the subject of the verbs. . . . Features of the object are not explicitly marked by the inflectional suffixes” (Dahlstrom n.d.:4-17). And for the inverse counterparts, including the form in (2), she states: “The suffixes following the theme sign encode number and obviation features of the object of the verb. The features of the subject are not marked explicitly” (Dahlstrom n.d.:4-17). Nevertheless, as Dahlstrom additionally notes (n.d.:4-17), the fact that the second argument is also third person can be unambiguously inferred. There are three candidate interpretations for this second argument: first person, second person, and third person. If it were either first or second person, there would be additional morphology indicating this. Since there is no such additional morphology, a third-person interpretation is the only possibility. My claim, however, is that the argument for further obviatives is not of this sort. Let’s take a closer look. In an ordinary transitive clause in Meskwaki, one argument will be proximate and the other obviative. In the direct clause in (3), the subject, ihkwe·wa ‘woman’, is proximate and the object, neniwani ‘man’, is obviative. In the inverse clause in (4), the subject, ihkwe·wani ‘woman’, is obviative and the object, neniwa ‘man’, is proximate.

|

The Assumption of a Further Obviative   5

(3) ehkwe·w-a kaka·čim-e·-w-a

neniw-ani

woman-sg joke.with-dir-3-sg man-obv ‘The woman (proximate) joked with the man (obviative).’ (Dahlstrom n.d.:3-13) (4) ehkwe·w-ani kaka·čim-ekw-w-a neniw-a woman-obv joke.with-inv-3-sg man-sg ‘The woman (obviative) joked with the man (proximate).’ (Dahlstrom n.d.:3-14)

But there are circumstances in which both arguments will be obviative. For example, if one of the arguments is third-person possessed and this possessor is proximate, both the subject and object will be obviative (Dahlstrom n.d.:3-12). Third-person-possessed NPs are obligatorily obviative, and since there can only be one proximate in a clause, the non-possessed argument is also obviative. Example (5) illustrates the case where it is the object, o·sani ‘his father’, that has a proximate possessor, and where the subject, ašahahi ‘the Siouxs’ [sic], along with ‘his father’ are therefore both obviative. In example (6), it is the possessor in the subject, owi·wani ‘his wife’, that is proximate, and ‘wife’ and the object, apenohehani ‘baby’, are therefore both obviative. (5) o-o·s-ani ašah-ahi

eh=nes-ekoniči

3poss-father-obv Sioux-obv.pl aor=kill-3”>3’.aor ‘The Siouxs [sic] (obviative) killed his (proximate) father (obviative).’ (Goddard and Dahlstrom 2022:3, emphasis added) (6) eh=anemi-nom-aniči

inini

apenoheh-ani ow-iw-ani.

aor=go.on-carry.on.back-3’>3”.aor that.obv baby-obv

3poss-wife-obv

‘His wife (obviative) went along with the baby (obviative) on her back.’ (Goddard and Dahlstrom 2022:14, emphasis added)

Double-obviative clauses can also arise if a proximate in a previous clause is still governing obviation in a later transitive clause, or if there are three non-coreferential NPs in a single clause. An example of the latter case is found in (7). There are two clauses in (7), but the last word, e·hwača·ha·či, is itself a clause with three arguments: ‘she’, ‘them’, and ‘it’, with ‘it’ referring to pešekesiwani ‘deer’. Since ‘she’ is proximate and there can only be one proximate, ‘them’ and ‘it’/‘deer’ are both obviative. An

|

6    Irene Appelbaum

example of the former case is found in (8). The passage begins with the clause in (8a), where kwi·yese·ha ‘a certain boy’ is introduced as proximate. Several clauses into the passage we find the clause in (8b). In this clause, the two arguments, ‘he’ (referring to neniwani ‘a certain man’) and mehte·hani ‘bow’, are both obviative even though there is no expressed proximate. This is because the latter clause is still considered within the scope of the earlier proximate kwi·yese·h-a ‘a certain boy’. (7) i·ni=ča·hi=’pi pešekesiw-ani e·škiki-ničini

e·h=wača·h-a·-t-i.

then=so=hrsy deer-obv fresh.creature-conj.ppl aor=cook-dir-3-mode ‘So then she (proximate) cooked fresh deer meat (further obviative) for them (obviative).’ (Thomason 2003:235) (8) a. nekoti=ye·toke našawe=kwi·yese·h-a. one=evid long.ago=boy-sg ‘There was a certain boy, long ago.’ (Thomason 2003:126) b. na·hka mehte·h-ani e·h=so·ken-a·-ni-t-i. also bow-obv

aor=hold-dir-obv-3-mode

‘And he (obviative) was holding a bow (further obviative).’ (Thomason 2003:126)

As noted above, when both arguments of a transitive animate clause are third person, the verb will be inflected for features of only one of the arguments. So, if the only person-marking inflection is for features of one third-person argument, we can be sure the other argument is third person as well. Whether the expressed person marking morphemes agree with the subject or the object of the clause depends on the theme sign (Dahlstrom n.d.:4-15). If the direct theme-sign is used, the features expressed with verbal suffixes agree with the subject of the clause; where the inverse sign is used, they agree with the object. This means that in transitive clauses with two third-person arguments, verbal inflection will always be for the higher-ranked argument, and in an ordinary transitive clause (either direct or inverse), this will be the proximate argument. The remaining—uninflected for—argument, in each case, can therefore unproblematically be inferred to be obviative. In the relatively unusual cases where both arguments of a transitive animate clause are obviative, the verbal inflection for the one expressed argument will always be for the higher-ranked obviative. If the theme sign is direct, this obviative argument will be the subject. This means the unexpressed inflection for the object

|

The Assumption of a Further Obviative   7

argument will be for a lower-ranked obviative and, hence, for a further obviative (Dahlstrom n.d.:4-17). If the theme sign is inverse, the obviative verbal inflection will agree with the object of the clause, and, in this case, the unexpressed inflection for the subject must be for a lower-ranked obviative. Hence, it, too, must be a further obviative (Dahlstrom n.d.:4-17). Thus, whenever the verb is inflected only for an obviative argument, the claim is that the other argument must be a further obviative. Although there is no individual further-obviative morpheme, so the argument goes, when the verbal inflection consists of only person features for one obviative argument, we know the second argument has to be further obviative. While the above reasoning seems to be parallel to that in which a second argument of a transitive clause is inferred to be (simply) third person, it is not. This is because the absence of morphology for features of the second argument cannot be unambiguously associated with a further-obviative interpretation. Specifically, this absence of inflectional morphology is compatible with either a further-obviative or a simple-obviative interpretation. A simple obviative reading would yield a direction-marked, transitive clause inflected for two obviative arguments. On this interpretation, the agreement morphology would in fact agree: there would be two obviative nominals and a verb inflected for two obviative arguments. By contrast, on a further-obviative interpretation, the agreement morphology would not agree because there would be two nominals inflected for obviative, while the verb would be inflected for one obviative and one further-obviative argument. The present point, however, is not that this second argument has to be a simple obviative, but rather that, because it might be, we are not justified in simply inferring that it is instead a further obviative. And the reason it might be, is that the verbal form here would look the same whether the unexpressed inflection for the second argument is interpreted as a further obviative or as a simple obviative. For example, although a second level of glossing in (9)3 shows one further obviative (3”) and one obviative (3’), this example could just as easily be glossed for two obviatives. That is, if the second-level gloss were instead “[kill 3’>3’.aor]”, the verb form would be identical. (9) o-o·s-ani

aša·h-ahi

e·h-nes-eko-ni-t-i

3poss-father-obv Sioux-obv.pl aor-kill-inv-obv-3-mode [kill 3”>3’.aor] ‘The Sioux (further obviative) killed his (proximate) father (obviative).’ (Dahlstrom n.d.:3-20)

|

8    Irene Appelbaum

Since there are two possibilities for the interpretation of the unexpressed inflection for the second argument, the (absence of) morphology does not justify the inference that it is for one of these, namely, for a further obviative. Now, proponents of further obviatives will here object that the theme sign rules out the possibility that the second argument is another simple obviative, and hence that the second argument can only be a further obviative. But it is not the theme sign per se that rules this out; it is rather the stipulation that there can be no direction-marked, double-obviative clauses. Contrary to this stipulation, though, a transitive verb morphologically marked for direction is perfectly compatible with both arguments being morphologically marked as simple obviatives. We know this is possible because such clauses exist: they are precisely the class of double-obviative clauses under discussion and seen, for example, in (5) and (6), above. To rule out, by fiat, the possibility of direction-marked double-obviative clauses is, at worst, to contradict the overt morphological facts and, at best, to beg the question about whether what seem to be double-obviative clauses, really are. But for present purposes, the point is just that what’s presented as unqualified morphological evidence for further obviatives, isn’t. There is no further-obviative morpheme, and the morphology that is present doesn’t allow us to unambiguously infer the existence of one. The verbal inflectional morphology is compatible with the second argument being either obviative or further obviative. Thus, it is the assumption that double-obviative direction-marked clauses are impossible, rather than the morphological data itself, that leads to the inference that Meskwaki has, as opposed to, might have, further obviatives.

The Logical Argument for Further Obviatives

If the above analysis is correct, it shows that there is no unambiguous morphological evidence for further obviatives in Meskwaki. But it does not show that the argument that they must exist, is wrong. Before evaluating this argument, let us reconstruct it explicitly, as follows: Meskwaki transitive clauses exhibit direction-marking. Direction presupposes clauses with unequally ranked arguments. Some transitive clauses have two obviative arguments and hence appear to be equally ranked. So, either these clauses are not marked for direction, or the arguments are unequally ranked, after all. Double-obviative clauses are marked for direction: They are found as inverse clauses as well as direct. Therefore, despite appearances, the two obviative

|

The Assumption of a Further Obviative   9

arguments in such clauses must be unequally ranked. Specifically, the object in a direct clause, and the subject in an inverse clause, must be further obviatives. The above argument is valid for the conclusion that the two arguments of a double-obviative clause must be unequally ranked. However, as an argument for the existence of further obviatives, it falls short. This is because the conclusion that the two obviative arguments need to be unequally ranked does not imply that they need to be ranked specifically in terms of obviation status. One might believe that obviation status determines direction-marking; indeed, it might. But this claim does not follow simply from the existence of direction-marking. It is an additional assumption and as such requires independent justification. Yet no justification whatsoever is given. It is simply taken for granted that to rank arguments is to rank them in terms of obviation status. In the passages below, for example, we see that simply citing the need to rank the arguments, immediately implies the presence of further obviatives. The object must rank below the subject on the person hierarchy. . . . Consequently where the subject is third person obviative, the object must be third person further obviative. (Dahlstrom n.d.:4-17, emphasis added) The subject must rank below the object on the person hierarchy. The subject is therefore . . . further obviative. (Dahlstrom n.d.:4-17, emphasis added)

Part of the difficulty here may be that the phrase “ranked obviatives” is ambiguous. It might mean that obviation as a grammatical category contains subcategories that have different degrees of obviative-ness. Alternatively, the phrase “ranked obviatives” might mean that there is only a single category of obviative, but obviative nouns in a sentence are nevertheless ranked in terms of some other factor(s). In this latter case, the obviatives are ranked, but they are not ranked in terms of obviation itself. The automatic association between ranking and ranking in terms of obviation status appears to be due to an implicit assumption that real-world distinctions need to be isomorphically mapped to morphological distinctions. That is, the assumption seems to be that if we are to rank three arguments of a clause with respect to discourse prominence instead of the more usual two, we must have corresponding morphology with three ranks. But this is surely a mistake. Consider an analogy. We might rank the English phrase five dogs above the phrase three dogs on the grounds

|

10    Irene Appelbaum

that the former refers to a greater number of dogs than the latter. But this ranking in no way implies that dogs in the former phrase is more plural than dogs in the latter phrase. The two NPs are ranked in terms of the number or size of their respective referents, not in terms of their degree of plurality: five dogs is more than three dogs, but the phrases five dogs and three dogs are equally plural. Similarly, we may rank participants in terms of their relative discourse prominence, but this does not by itself imply that they will each have a distinct obviative status. If there are two participants, in ordinary circumstances one will be proximate and the other obviative, but if there are three (or more) participants, the third one may be less discourse prominent than the second one, without having to be less obviative. As Cowell and Moss (2008:352) point out, in such contexts, “speakers still can judge an obviative person as more salient or topical in the discourse in relation to the other obviative person.”4 Nevertheless, the assumption that direction is determined by obviation status is widespread in Algonquian linguistics.5 Bliss (2005:62), for example, refers to “the implicit assumption made in the Algonquianist tradition that obviation and direct/ inverse work together as a concerted system.” Wolvengrey (2005:424), speaking of Cree and Algonquian generally, says: “Obviation . . . plays an especially important role in conjunction with Direct-Inverse morphology.” And Dahlstrom (2014 [1991]:93), speaking about Plains Cree, says explicitly: “The appearance of direct or inverse forms is governed by which argument is proximate and which obviative.” But the pervasiveness of this assumption doesn’t help close the gap between the claim that the arguments are ranked and the claim that they are ranked in terms of obviation status. And it is only the latter claim that leads to the conclusion that further obviatives must exist. As Zúñiga defines direction: “Direction denotes the grammatical distinction between expressions of transitive states of affairs expressed by syntactically bivalent predicates and informed by particular properties of the agentive and patientive arguments” (2014:335, emphasis added). It is part of the definition of direction that the arguments be ranked by properties of the arguments, but not that they be ranked in terms of obviation status. The claim that obviation determines direction does not follow simply from the existence of direction-marking; it is an additional assumption. Still, to identify “obviation determines direction” as an additional assumption is not to show that it is an unwarranted assumption. Whether the assumption is warranted depends on the empirical evidence—that

|

The Assumption of a Further Obviative   11

is, it depends on whether “obviation determines direction” is true. And, here, the existence of direction-marked, transitive clauses with two obviative arguments provides evidence that it is false. In such clauses, the morphology shows there are two equally obviative arguments, but the claim that obviation determines direction requires that the arguments have unequal obviation status. In linguistics, as in other scientific endeavors, when we have a conflict between theory and data, we have two basic options: adjust the theory or reject the data. To reject the data in this context is to deny that they provide even prima facie evidence against the theory. This is the tack taken by those who posit further obviatives: the data of double obviatives are redescribed to fit the theory (i.e., the assumption that obviation determines direction). On this approach, what looks like a simple obviative argument is nevertheless analyzed as a further obviative. In order to accomplish this reanalysis of the data, obviation status is decoupled from its morphological realization: When both arguments are obviative third person a ranking must be imposed upon the two obviatives, making one a first obviative (also called “nearer obviative”) and the other a second obviative (or “further obviative”). (Goddard and Dahlstrom 2022:5, emphasis added) If both subject and object of a transitive verb are obviative: one must be considered the “nearer obviative” and the other the “further obviative” for purposes of verb inflection. (Dahlstrom n.d.:3-20, emphasis added)

In this account, further obviatives are acknowledged to be “imposed” and are qualified as being “considered” further obviatives, suggesting a recognition that the morphology is otherwise and hence that further obviatives are to be taken more as logical than morphological entities. But any such acknowledgment, and any such qualification, is completely eclipsed by the fact that further obviatives (or “second obviatives”) are written in to the referential hierarchy and are thus treated on an absolute par with speech act participants (SAPs), proximates, and obviatives: The direct suffix . . . indicates that the subject outranks the object on the following hierarchy: . . . non-third person > third person proximate > third person (first) obviative > third person second obviative > inanimate (Goddard and Dahlstrom 2022:4, emphasis added)

|

12    Irene Appelbaum

On this hierarchy [of person] first and second persons are of equal rank and ranked highest, followed by proximate third person, then obviative third person, and finally the “further” obviative. (Dahlstrom n.d.:4-14, emphasis added)

Now, if one chooses to endorse the assumption that obviation determines direction, over the morphological evidence of double obviatives, it is necessary to postulate further obviatives. But it is not necessary to endorse this assumption. The alternative response to the conflict between the assumption that obviation determines direction and the existence of double-obviative clauses is to recognize that in these cases, at least, it cannot be obviation status that is doing the work of ranking the two arguments—their obviation status is equal. Once this is recognized, it is difficult not to take the next step and consider that, if double-obviative clauses can get by without obviation status doing the ranking work, perhaps the more usual direction-marked, transitive clauses with a single-proximate and a single-obviative argument can as well. In other words, taking this approach invites us to consider whether it was a mistake to assume that obviation status was determining direction in the first place.6 From this perspective, ordinary transitive clauses with one proximate and one obviative argument are ambiguous as to whether the arguments are ranked in terms of obviation status or merely correlate with it. These cases are ambiguous because whether obviation or discourse factors determine direction, the results will be the same: the more discourse-prominent argument will always be the proximate argument and proximate arguments will always outrank obviative arguments. From this perspective, double-obviative clauses are important precisely because they provide disambiguating evidence that arguments are not ranked in terms of obviation status. So, we have two types of response to the conflict between the assumption “obviation determines direction” and existence of morphologically marked, double-obviative clauses: posit further obviatives or re-evaluate the assumption “obviation determines direction.” The former tack is obviously more theoretically leveraged and less empirically grounded than the latter. But that is not the immediate problem. The immediate problem is that positing further obviatives to preserve the assumption “obviation determines direction” saves this assumption, but only by making the overall argument for the existence of further obviatives circular. For in order to conclude that further obviatives must exist, it must be true that obviation determines direction. As we’ve seen, this is an additional assumption. But

|

The Assumption of a Further Obviative   13

in order for this additional assumption to be true—given the existence of double obviatives—further obviatives must be assumed.

“Obviation Determines Direction”

So, just as the morphological evidence does not justify the conclusion that there are further obviatives, the logical argument does not justify the conclusion that there must be. Nevertheless, I’ll now suggest that it is neither empirical evidence nor logical argument that ultimately sustains the commitment to further obviatives. Instead, I think the lesson to be drawn from the above arguments is not, in the first instance, about that status of further obviatives themselves, but about the status of the assumption “obviation determines direction.” The heart of the matter, I believe, is whether we are allowed to bring the empirical data of direction-marked, double-obviative clauses to bear on this assumption. Double-obviative clauses are counterexamples to the assumption that obviation determines direction, but their status as such is not acknowledged by those who posit further obviatives. Or, rather, positing further obviatives is precisely a way to deny that such clauses are counterexamples. While “obviation determines direction” has the form of an empirical claim, it is not treated as empirically falsifiable. Another way to put this point is to ask: is there any empirical data that would count as showing this assumption to be false? For example, in cases where there are three morphologically marked obviatives within a single proximate span, as in (10), a Plains Cree clause discussed in Rhodes 2017, would this trigger a re-evaluation of the assumption that obviation determines direction, or would it, as Rhodes (2017:202) describes, simply lead to “positing of a further-further-obviative.” (10) ēh=kīh=nipah-iko-iyi-t

káhkákiwacen-ah

comp=pst=kill-inv-obv-3sbj Crow-obv ‘He [onâpêmiyiwa, ‘her obv husband obv’] had been killed by the Crow obv.’ (Rhodes 2017:201, emphasis added)

To the extent that the answer is the latter, the assumption of further (and further) obviatives functions to discredit the empirical evidence against the claim that obviation determines direction. To posit further obviatives, in other words, is to

|

14    Irene Appelbaum

treat the theory as inviolable and the data as corrigible. The assumption of further obviatives is thus not theoretically innocuous; it plays the tendentious role of insulating the assumption that obviative determines direction from empirical scrutiny.

Summary and Conclusion

In this paper, I have done two things. First, I have shown that positing further obviatives in the context of transitive clauses in Meskwaki is not justified. I have argued that the morphological evidence cited for them is equivocal and question-begging, and that the argument that they must be posited, depends on circular reasoning. Second, I have made it explicit that it is the assumption that obviation status determines direction that underwrites the commitment to further obviatives. I have shown that this assumption is so deeply entrenched that counterexamples—direction-marked, transitive clauses with two obviative arguments—are not permitted to surface as such. Instead, redescribing them as involving further obviatives functions to neutralize their status as counterexamples, and thereby to shield the assumption that obviation determines direction from re-evaluation. If this is correct, then the critical next step is to stop positing further obviatives and let the empirical chips fall where they may. Notes 1. Thanks to Amy Dahlstrom, Chris Hammerly, Will Oxford, and two anonymous reviewers for helpful feedback and suggestions; thanks to Amy Dahlstrom and Rich Rhodes for assistance with glosses. I am solely responsible for any errors. 2. The following abbreviations are used in interlinear glosses: 3 = third person, 3’ = third-person obviative, 3” = third-person further obviative, aor = aorist, comp = complementizer, conj = conjunct, dir = direct, evid = evidential, hrsy = hearsay evidential, inv = inverse, obv = obviative, pl = plural, poss = possessive, ppl = participle, pst = past, sbj = subject, sg = singular. 3. This example repeats the Meskwaki sentence in (5) from Goddard and Dahlstrom 2022, but here Dahlstrom identifies a further obviative in a second level of glossing only. 4. Unfortunately, in the same passage, Cowell and Moss (2008:352) identify the more discourse-prominent obviative as “functionally proximate,” effectively diluting the implication that direction need not be mediated by obviation status.

|

The Assumption of a Further Obviative   15

5. Rhodes 2017 and Rhodes 1990 are notable exceptions. 6. See Appelbaum 2019 for an analogous argument for Ktunaxa.

References Appelbaum, Irene. 2019. Double Obviatives and Direction-Marking in Kutenai. Proceedings of the 30th Western Conference on Linguistics, vol. 24, ed. by Trevor Driscoll, pp. 18–24. Fresno: Department of Linguistics, California State University, Fresno. Bliss, Heather. 2005. Topic, Focus, and Point of View in Blackfoot. Proceedings of the 24th West Coast Conference on Formal Linguistics, ed. by John Alderete, Chung-Hye Han, and Alexei Kochetov, pp. 61–69. Somerville, Massachusetts: Cascadilla Proceedings Project. Bloomfield, Leonard. 1946. Algonquian. Linguistic Structures of Native America, ed. by Cornelius Osgood and Harry Hoijer, pp. 85–129. New York: Viking Fund Publications in Anthropology. Bloomfield, Leonard. 1962. The Menomini Language, ed. by Charles F. Hockett. New Haven, CT: Yale University Press. Cowell, Andrew, and Alonzo Moss Sr. 2008. The Arapaho Language. Boulder: University Press of Colorado. Dahlstrom, Amy. 2014 [1991]. Plains Cree Morphosyntax. London: Routledge. Dahlstrom, Amy. 2017. Obviation and Information Structure in Meskwaki. Papers of the 46th Algonquian Conference, ed. by Monica Macaulay and Margaret Noodin, pp. 39–54. East Lansing: Michigan State University Press. Dahlstrom, Amy. [n.d.] Meskwaki Syntax. Manuscript. https://lucian.uchicago.edu/blogs/ adahlstrom/publications-2/selected-manuscripts/. Frantz, Donald G. 1966. Person Indexing in Blackfoot. International Journal of American Linguistics 32(1):50–58. Frantz, Donald G. 2017. Blackfoot Grammar. 3rd ed. Toronto: University of Toronto Press. Goddard, Ives. 1984. The Obviative in Fox Narrative Discourse. Papers of the 15th Algonquian Conference, ed. by William Cowan, pp. 273–286. Ottawa: Carleton University. Goddard, Ives and Amy Dahlstrom. 2022. Meskwaki (Algonquian) Evidence against Basic Word Order and Configurational Models of Argument Roles. Language Change and Endangered Languages: Studies in Honor of Lyle Campbell, ed. by Tiago C. Chacon, Nala H. Lee, and Wilson D. L. Silva. Edinburgh: Edinburgh University Press. Hockett, Charles F. 1939. Potawatomi Syntax. Language 15(4):235–248. Hockett, Charles F. 1948. Potawatomi I: Phonemics, Morphophonemics, and Morphological Survey. International Journal of American Linguistics 14(1):1–10.

|

16    Irene Appelbaum

Oxford, Will. 2017. Proximate DP, Obviative KP: Balancing the Morphosyntax and Pragmatics of Obviation. Proceedings of the 2017 Annual Conference of the Canadian Linguistic Association, ed. by Andrew Alexander Monti, pp. 1–15. https://cla-acl.artsci.utoronto.ca/ actes-2017-proceedings/. Rhodes, Richard A. 1990. Obviation, Inversion, and Topic Rank in Ojibwa. Berkeley Linguistic Society Parasession to the 16th Annual Meeting, ed. by David Costa, pp. 101–115. Berkeley: Berkeley Linguistics Society. Rhodes, Richard A. 2017. Obviation, Inversion, and the Notion of Topic in Algonquian. Papers of the 46th Algonquian Conference, ed. by Monica Macaulay and Margaret Noodin, pp. 197–211. East Lansing: Michigan State University Press. Thomason, Lucy. 2003. The Proximate and Obviative Contrast in Meskwaki. PhD thesis, University of Texas, Austin. Uhlenbeck, C. C. 1938. A Concise Blackfoot Grammar. Amsterdam: Noord-Hollandsche Uitgevers-maatschappij. Wolfart, H. Christoph. 1973. Plains Cree: A Grammatical Study. Transactions of the American Philosophical Society, n.s., vol. 63, part 5. Philadelphia. Wolfart, H. Christoph. 1978. How Many Obviatives: Sense and Reference in a Cree Verb Paradigm. Linguistic Structures of Native Canada, ed. by Eung-Do Cook and Jonathan D. Kaye, pp. 255–272. Vancouver: University of British Columbia Press. Wolvengrey, Arok. 2005. Inversion and the Absence of Grammatical Relations in Plains Cree. Morphosyntactic Expression in Functional Grammar, ed. by Casper de Groot and Kees Hengeveld, pp. 419–445. Berlin: Mouton De Gruyter. 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.

Possessive Prefixes across Algonquian Evidence for Two Separate Paradigms Ksenia Bogomolets

T

his paper investigates possessive prefixes across Algonquian.1 Possessive prefixes in Algonquian mark person features of the possessor on noun stems. Most independent nouns can be possessed, but possessor prefixes with independent nouns are nonobligatory—I will refer to these as alienable possessive prefixes (APPs). Algonquian languages also have a set of noun stems that are obligatorily possessed, i.e. they never occur without a possessive prefix. This set of dependent noun stems in Algonquian includes nouns referring to kinship terms, body parts, and some intimate possessions. This type of possessive construction is known as inalienable possession, and I will refer to these prefixes as inalienable possessive prefixes (IPPs).2 Traditionally, it has been assumed that Algonquian has a single set of possessive prefixes that is used with both independent and dependent noun stems, presumably because possessive prefixes in these environments can sometimes (subject to phonological conditions, see below) look identical, see (1):3 (1) a. Inalienable possession

b. Alienable possession

ni-kosis ni-ciimaan 1-son 1-boat

‘my son’

‘my boat’ (Woods Cree; Starks 1992:35, 64)

|

   17

|

18    Ksenia Bogomolets

In the Woods Cree examples in (1), the first-person-possessive prefix surfaces as ni- with a dependent noun stem (1a) and with an independent noun stem (1b). A well-known contrast between the APPs and IPPs, however, arises when they precede a vowel-initial stem. Consider (2), where before a vowel-initial stem, the first-person agreement prefix surfaces as n- with a dependent noun stem (2a), but as nit- with an independent noun stem (2b): (2) a. Inalienable possession

b. Alienable possession

n-atay nit-akohp 1-belly 1-blanket

‘my belly’

‘my blanket’ (Woods Cree; Greensmith 1985:65)

The contrast between IPPs and APPs illustrated in (2) is the main focus of this paper. This contrasting behavior of APPs vs. IPPs before a vowel-initial stem is most often described as an idiosyncratic phonological anomaly without any particular explanation (e.g., Fermino 2000:16–19 for Wampanoag; Frantz 1991:68–73 for Blackfoot; Starks 1992:22 for Woods Cree, among many others). Alternatively, a formal analysis of the contrast in (2) has been proposed in Newell and Piggott 2014 where the contrast is seen as evidence for distinct morphosyntactic positioning of the (same) possessive prefix when used with independent noun stems vs. dependent noun stems.4 The historical origin of these prefixes has also received conflicting analyses. Thus, Goddard (2007, 2015, and earlier works) reconstructs a single set of person agreement prefixes for Proto-Algonquian (PA); see Table 1. Under Goddard’s analysis, which appears to have been assumed explicitly or implicitly in most descriptions of possessive prefixes in the Algonquianist literature, the APPs and IPPs are underlyingly and historically identical. Proulx (1989), on the other hand, taking into account the contrasting behavior of APPs and IPPs illustrated in (2), reconstructs two separate sets of prefixes: one marking inalienable possession and another one marking alienable possession. The latter set, under Proulx’s reconstruction, is also used to mark agreement on verbs, as in Table 2. The question thus arises: How many paradigms of possessive prefixes are there in Algonquian languages, synchronically and diachronically? This paper argues that IPPs form a separate, underlyingly distinct paradigm from APPs.

|

Possessive Prefixes across Algonquian   19

TABLE 1. Reconstruction of a single set of PA agreement prefixes (based on Goddard

2007, 2015) Set 1

1ST

2ND

3RD

*ne-

*ke-

*we-

TABLE 2. Reconstruction of two sets of PA agreement prefixes (based on Proulx 1989) 1ST

Set 1: Inalienable Set 2a: Alienable & Verb Agr. before Vowel Set 2b: Alienable & Verb Agr. before Cons.

2ND

3RD

*n*k- *w*net- *ket- *wet*ne- *ke- *we-

The rest of this paper is structured as follows. In the next section, I present evidence for APPs and IPPs forming two separate paradigms. I argue that the APP paradigm has an additional underlying consonant (/t/ or /d/, depending on the language), which the IPP paradigm crucially does not have. A novel analysis in terms of liaison is proposed to account for the phonological behavior of APPs. I then offer a comparison of the current analysis and a formal account of the contrast between APPs and IPPs proposed in Newell and Piggott 2014. Finally, I briefly discuss the patterns found in verbal agreement prefixes across Algonquian and outline paths for future research.

APPs and IPPs Are Underlyingly Segmentally Different

In this section, I present the main argument for treating the APPs and the IPPs as two separate sets of morphemes. I will argue that APPs have an underlying, i.e., non-epenthetic, consonant—/t/ or /d/ (depending on the language)—which, crucially, the IPPs do not have as part of their underlying form. The following alternation is observed across Algonquian. alienable possessive prefixes end in a consonant—/t/ or /d/ if the following stem begins with a vowel, but they end in a vowel if the following stem begins with a consonant. Consider the alternation found in Arapaho in (3)– (4) as an illustration, with a representation of the general rule in (5) (recall also (1b–2b)):

|

20    Ksenia Bogomolets

(3) a. biixuut

b. ne-biixuut

shirt 1-shirt ‘a shirt’ (4) a. hotii

‘my shirt’

b. not-otii

car 1-car ‘a car’

‘my car’5

(5) ne-, e-, i- /__-C . . . net-, et-, it- / __-V . . .

(Cowell and Moss 2011:61)

In (3), an independent noun is prefixed with an alienable possessive morpheme ne- before a consonant-initial stem; in (4), an independent noun is prefixed with an alienable possessive morpheme /net-/ (not-) before a vowel-initial stem. Importantly, this alternation is never found with IPPs, which before a vowel-initial stem either remain the same as before a consonant-initial stem (cf. Arapaho examples in (6) below) or delete their vowel (cf. Woods Cree examples in (2)). (6) a. ne-biiˈʔoʔoo

b. ˈne-inoo

1-sweetheart 1-mother

‘my sweetheart’

‘my mother’

(Cowell and Moss 2011:266, 477)

The contrast between APPs and IPPs illustrated for Arapaho in (3)-(6) and in (1)(2) for Woods Cree is robust across Algonquian, i.e., we regularly find the /t, d/-∅ alternation in the possessive prefixes with independent nouns, but we never find /t, d/-final possessive prefixes with dependent noun stems (cf. Table 3).6 Previously, the /t, d/ consonants in APPs have been almost exclusively described as epenthetic to resolve vowel hiatus at the morpheme juncture (Cowell and Moss 2011; Goddard 2007, 2015; Newell and Piggott 2014; Oxford 2014, among many others; the only exception being Proulx’s reconstruction, cf. Table 2). In the remaining part of this section, I argue that the /t, d/ consonants in APPs are underlying rather than epenthetic. I propose a novel analysis of the /t, d/-∅ alternation in APPs in terms of a liaison pattern. The evidence for analyzing the /t, d/ consonants in APPs as underlying is threefold. First, the quality of the prefix-final consonant in APPs is constant across languages: /t/ or /d/, suggesting a common diachronic source for this consonant across Algonquian.7 Such uniformity is unexpected if the consonant

|

Possessive Prefixes across Algonquian   21

TABLE 3. Possessive prefixes across Algonquian ARAPAHO

Alienable Possession 1st ne(t)2nd e(t)3rd i(t)Inalienable Possession 1st ne2nd e3rd

i-

EAST CREE

PLAINS CREE

OJIBWE

PASS-MAL

WAMP.

ni(t)chi(t)u(t)-

ni(t)ki(t)o(t)-

ni(d)gi(d)o(d)-

n(t)k(t)w(t)-

nu(t)ku(t)wu(t)-

nichiu- (__C) w- (__V)

nikio- (__C) w- (__V)

nigi-

nk-

nuku-

o-

w-

wu-

is not part of the underlying form of the alienable paradigm. Second, the insertion of /t, d/ is not a regular strategy to resolve vowel hiatus: Algonquian languages have a number of productive hiatus resolution strategies all of which are cross-linguistically widely attested. These include deletion of one of the vowels, resyllabification of two short vowels at a morpheme juncture as a single nucleus, and epenthesis of glottal consonants (/ʔ/, /h/) or glides (/j/, /w/) (on the typologically unmarked status of laryngeals and glides as a resolution for vowel hiatus, see, for example, Blevins 2008; Żygis 2010). In fact, the /t, d/-∅ alternation is only attested in Algonquian with the person prefixes, while in all other cases vowel hiatus is resolved via one of the regular typologically unmarked strategies listed above.8 The final, and the most important, piece of evidence against analyzing the /t, d/ consonants in APP as epenthetic comes from the fact that these consonants surface in person prefixes when there is no vowel hiatus in a number of Algonquian languages, including Arapaho, Passamaquoddy-Maliseet, and Blackfoot. In the following sections, I first consider the Arapaho case in more detail, and I propose that the /t/-∅ alternation in APP can formally be analyzed as a case of liaison. I then turn to Passamaquoddy-Maliseet and Blackfoot and show how the liaison analysis can be extended to the patterns in these languages, which crucially cannot be captured with an epenthesis analysis.

|

22    Ksenia Bogomolets

The /t/-∅ Alternation in Arapaho: A Liaison Analysis

The /t/-∅ alternation in Arapaho is only ever attested with APPs, i.e., it is not found elsewhere in the language. Consonant epenthesis as a hiatus resolution strategy is only employed in Arapaho when a vowel sequence at the morpheme juncture cannot be resyllabified as a single nucleus, i.e., when a combination of two morphemes would result in an ‘extra-long’ trimoraic vowel sequence. In such cases, Arapaho employs epenthesis of typologically unmarked epenthetic consonants—glottal consonants or glides. Consider, for instance, (7) where a glottal fricative is epenthesized between a long vowel in the interrogative prefix koo- and the second-person possessive prefix e-: (7) koo-h-e-θooxe q-ep-2-glove ‘Is this your glove?’

(adapted from Cowell and Moss 2011:107)

Importantly, in all morphological contexts in the language, Arapaho allows for a combination of two short vowels at a morpheme juncture: consonant epenthesis never applies in such a phonological environment. Instead, two short vowels are always resyllabified as a single nucleus to avoid hiatus, cf. (8) below: (8) hisi-i



/hi.sii/

tick-sg.obv ‘a tick’

(Cowell and Moss 2011:59)

Crucially, the /t/-consonant in the APP surfaces before stems beginning with short vowels, i.e., in a phonological environment that does not warrant consonant epenthesis in Arapaho. Consider examples in (9); (9a) is repeated from (4); (9b) is adapted from Goddard (2015:362): (9) a. not-otii

b. net-eneecee-b

1-car 1-buffalo-poss

‘my car’

‘my buffalo’9

If the /t/-∅ alternation in APPs in Arapaho were to be analyzed as consonant epenthesis to resolve hiatus, it would thus be a unique and highly unexpected

|

Possessive Prefixes across Algonquian   23

process as (i) it would be a case of epenthesis of a consonant quality (i.e., /t/) that never occurs as epenthetic elsewhere in the language, and (ii) it would be a case of consonant epenthesis occurring in a phonological environment (i.e., between two short vowels) that never triggers epenthesis elsewhere in the language. For these reasons, I propose that /t/ in APPs is underlying, i.e., it is not epenthetic. I further propose that the /t/-∅ alternation observed in these prefixes in Arapaho should be analyzed as a case of liaison. The current account is formulated within the CVCV Phonology framework (Lowenstamm 1996; Scheer 2004). Within this theory, the timing tier consists of sequences of onsets and nuclei (CV). C and V positions on the timing tier may be phonologically null if they lack a link to a segment on the segmental tier. On the other hand, a segment may be floating if it is unassociated to a position on the timing (CV) tier.10 Floating elements have been proposed to account for multiple patterns cross-linguistically. One of the most well-known patterns is the liaison in French (Clements and Keyser 1983; Tranel 1986; Bennett 1991), as illustrated in (10).

In (10a), the final underlying consonant of the article is a floating element, and it remains silent as it is not associated with a position on the CV tier. In (10b), the following V-initial stem supplies an onset position for the floating final consonant of les, which results in it being pronounced. I propose that the /t/-∅ alternation in the alienable possessive prefixes in Arapaho (and more broadly, in Algonquian) should be analyzed in the same way. Thus, the prefix-final /t/ in Arapaho is a floating element, i.e., it is part of the

|

24    Ksenia Bogomolets

underlying form of the morpheme, but it is not underlyingly attached to a skeletal position on the CV tier (11).

The prefix-final /t/ then can only be pronounced if the following morpheme is V-initial and thus provides an empty onset slot for the /t/ to be associated with. Consider the schematic representation in (12)-(13):

In (12), the final underlying consonant of the APP is silent, as it is not associated with a position on the CV tier. In (13), however, the following V-initial stem supplies an onset position for the floating final consonant of the APP, and it can be pronounced. The representation of the APP in (11) not only provides a straightforward explanation for the /t/-∅ alternation in Arapaho, but it also makes the following prediction. Prefixes ending with a floating consonant are predicted to phonologically realize that consonant whenever they are followed by a vowel-initial morpheme

|

Possessive Prefixes across Algonquian   25

because the floating consonant will be able to link to an open C-position on the timing tier. This prediction is crucially different from the prediction made by an analysis treating the /t, d/ consonants in APPs as epenthetic to resolve vowel hiatus; compare the two in (14)-(15): (14) Prediction of the epenthetic /t, d/ account

/t, d/ will surface in a vowel hiatus environment:



∅]APP  t, d]APP / V__-V

(15) Prediction of the liaison account /t, d/ will surface if followed by a vowel-initial morpheme: ∅]APP t, d]APP / __-V

In the next section, I show that the prediction of the liaison analysis proposed here (15) is borne out in the Passamaquoddy-Maliseet patterns of the /t/-∅ alternation in person prefixes, which, crucially, cannot be explained as a hiatus resolution strategy of /t/-epenthesis (14). The /t/-∅ Alternation in Passamaquoddy-Maliseet: A Liaison Analysis

Passamaquoddy-Maliseet provides particularly clear evidence in favor of the liaison analysis of the /t/-∅ alternation in Algonquian. Similar to other languages within the family, the /t/-∅ alternation in Passamaquoddy-Maliseet is restricted to person prefixes. It is attested with APPs and with person-agreement prefixes with verbs (see discussion below), and it is not attested with IPPs.11 Consider the examples in (16): in (16a), an independent noun stem begins with a consonant, and the first-person-possessive prefix surfaces as n-; in (16b), an independent noun stem begins with a vowel, and the first-person-possessive prefix surfaces as nt-: (16) a. n-sakom-am b. nt-ipis-im 1-chief-poss 1-whip-poss

‘my chief’

‘my whip’ (Passamaquoddy-Maliseet; PMLP)

Observe that the vowel hiatus environment is not present in Passamaquoddy-Maliseet even when the noun stem begins with a vowel (16b). Contrast examples of the

|

26    Ksenia Bogomolets

alienable possession in (16) and the examples with inalienable possessive prefixes in (17) below. In (17), the first-person-possessive prefix is realized as n- before a consonant-initial dependent noun stem (17a) and before a vowel-initial dependent noun stem (17b): (17) a. n-pihtin b. n-ikuwoss 1-hand 1-mother

‘my hand’

‘my mother’

We thus observe that in Passamaquoddy-Maliseet, like in other Algonquian languages, the contrast between APP and IPP is neutralized before a consonant-initial stem ((16a), (17a)), but the difference in their phonology is evident before a vowel-initial stem ((16b), (17b)). Importantly, Passamaquoddy-Maliseet provides crucial evidence against treating /t, d/ in APP as epenthetic consonants to resolve vowel hiatus because person prefixes in Passamaquoddy-Maliseet have no vowel (cf. (16b), Table 3 above), and thus there is no vowel hiatus environment to trigger consonant epenthesis. Crucially, however, when the stem begins with a vowel, it provides a phonological environment for liaison. Thus, the prediction in (15) is borne out. Consider representations in (18)-(1E9).In (18), the /t/ consonant in the APP is a floating element and cannot be pronounced as it lacks an association with a slot on the CV-tier. In (19), on the other hand, when the following morpheme begins with a vowel, it supplies an empty C-slot on the timing tier, and the floating /t/ of the APP can get linked

|

Possessive Prefixes across Algonquian   27

TABLE 4. Alienable vs. inalienable possessive prefixes in Blackfoot (based on Proulx 1989) ALIENABLE POSSESSIVE PREFIXES

1st nit2nd kit3rd ot-

INALIENABLE POSSESSIVE PREFIXES

nko-

to it and be pronounced. The /t/-∅ alternation in APPs thus is straightforward in Passamaquoddy-Maliseet if analyzed as a liaison pattern, but it is surprising and idiosyncratic otherwise. The /t/ in APPs in Blackfoot

While Arapaho and Passamaquoddy-Maliseet provide strong evidence for the liaison analysis of the /t/-∅ alternation in APPs because in both languages the regular phonological environment for epenthesis is not met, Blackfoot provides clear evidence for treating /t, d/ as part of the underlying shape of the APPs. Possessive prefixes used with independent noun stems always have the final consonant /t/ in Blackfoot, i.e., there is no /t/-∅ alternation in this language. The set of APPs thus even on the surface looks quite different from the set of IPPs in Blackfoot; consider Table 4. The prefix-final /t/ in APPs cannot be analyzed as epenthetic in Blackfoot as there are no environments in which it would fail to surface as part of these morphemes. Consider the examples in (20) below. In (20a), the independent noun stem begins with a consonant, and an epenthetic vowel /i/ is inserted between the prefix and the stem;12 in (20b), the independent noun stem begins with a vowel: (20) a. nits-i-nʔixksinʔi b. nit-oʔpiimʔa 1-ep-song 1-rope ‘my song’

‘my rope’ (Blackfoot; adapted from Proulx 1989:56, 67)

The prefix-final /t/ in the APPs in Blackfoot must be analyzed as underlying rather than epenthetic. In contrast to the rest of Algonquian, the /t/ consonant in Blackfoot does not alternate with ∅. In the current analysis, this can be seen as a microparametric variation straightforwardly explained if /t/ in the Blackfoot prefixes is not a floating element; i.e., it is associated to a C-slot on the timing tier

|

28    Ksenia Bogomolets

in all cases. Consider the representations in (21)-(22). Blackfoot thus provides particularly clear evidence for analyzing /t, d/ as part of the underlying form of the APPs across Algonquian and against treating these consonants as epenthetic.

Discussion and Future Research

The proposals put forward in this paper have a number of theoretical consequences, some of which are considered briefly below. Contrasting Formal Accounts: Epenthesis vs. Liaison

The account of the contrast between APPs and IPPs proposed in this paper suggests that there is a segmental difference between two sets of prefixes that leads to their differing phonological behavior.13 This is markedly different from a previous formal account of the /t, d/ consonants in APPs, which derived the differences in the phonological behavior between APPs and IPPs from a morphosyntactic difference between alienable and inalienable possession (Newell and Piggott

|

Possessive Prefixes across Algonquian   29

2014, N&P henceforth).14 Newell and Piggott claim that the contrasting behavior of (what they consider to be a single set of) possessive prefixes before vowel-initial stems is due to the possessive prefixes being structurally ‘closer’ to the noun in the inalienable environment than in the alienable environment. While possessive prefixes are spelled out in the same phase (DP in Newell and Piggott’s analysis) with a dependent noun, possessive prefixes and a possessed independent noun are spelled out in different phases (DP and nP respectively). Focusing on Ojibwe, Newell and Piggott argue that vowel hiatus within a phase, as with the IPPs and vowel-initial noun stems, triggers vowel deletion to resolve hiatus, as in (23) below (the underlying forms of morphemes are presented as they are given in the source, but see note 15 below): (23) a. IPP with C-initial stem SF

b. IPP with V-initial stem

nika:d no:s

UF ni-ka:d

ni-o:s



1-leg

1-father



‘my leg’

‘my father’

(N&P:346, 349)

On the other hand, when vowel hiatus emerges between morphemes spelled out in different phases, as with the APPs and vowel-initial noun stems, hiatus should be tolerated as a general rule. Newell and Piggott themselves note that the appearance of the /d/ consonant in APPs before a vowel-initial stem is “difficult to explain” within their analysis (N&P:352). They further propose that it is not tolerated with APPs, contrary to the predictions of their analysis, because of an independent structural complication arising from a syntax-phonology mismatch. They stipulate that APPs are morphosyntactically external to the prosodic word (i.e., the stem they combine with), but for phonological reasons, these prefixes must be part of the same prosodic word as the stem. Thus, while APPs are spelled out in the morphology as a PWd adjunct, they then undergo an ‘adjustment process’ that incorporates the APPs into the PWd formed by the stem (24): (24) [APP [ . . .PWd]]  [APP [APP . . . PWd]]

(adapted from N&P:353)

|

30    Ksenia Bogomolets

Such a mismatch between the morphosyntactic position and the ‘surface’ phonological position of the APPs then triggers the /d/-epenthesis with this particular class of morphemes, as in (25): (25) a. APP with C-initial stem

b. APP with V-initial stem

SF

niʒi:ʃi:bim nidogima:m



ni-ʒi:ʃi:b-im

UF

ni-ogima:-m



1-duck-poss 1-leader-poss



‘my duck’

‘my chief, leader’ (N&P:346, 351)

Newell and Piggott then take the asymmetry in the presence/absence of the /d/-∅ alternation in APPs vs. IPPs as evidence for a difference in morphosyntactic positioning of possessive prefixes when used with dependent vs. independent nouns. Recall, it is proposed here that the /d/-∅ alternation in APPs and the contrast between APPs and IPPs can be explained in purely phonological terms (i.e., without the need to appeal to morphosyntactic or morpho-prosodic factors). Further evidence is thus required to support the structural difference between APPs and IPPs which Newell and Piggott argue for, namely that one is interpreted inside the same phase as the noun, and the other one is not. Such evidence in fact may come from phonological differences between the two sets of prefixes other than the presence/absence of the /t, d/-∅ alternation. I discuss these briefly below. Both APPs and IPPs show phonological interactions with the stem they attach to. The nature of these interactions, however, differs in a systematic way that may point not only to the two sets of morphemes being segmentally different (as argued in this paper), but also to their differing morphosyntactic locality with respect to the noun stem (as proposed in N&P). While phonological interactions between APPs and the noun tend to be fully regular and transparent, the interactions between IPPs and the stem may not be transparent synchronically and exhibit idiosyncratic patterns that could potentially be explained by their morphosyntactic ‘closeness’ to the stem. Fully regular phonological effects of the stem shape on APP are found across Algonquian languages. These include, for example, vowel harmony spreading from the stem to the APP (recall the Arapaho examples in (9a)) and palatalization processes affecting the underlying /t/ consonant (recall, for instance, the Blackfoot example in (20a)). IPPs on the other hand show phonological idiosyncrasies both

|

Possessive Prefixes across Algonquian   31

in their segmental shape and in their prosodic behavior across Algonquian. In Arapaho, for instance, in addition to the morphemes ne-, e-, i- (first, second, and third person respectively) exemplified above (in (6) and Table 3), IPPs can have the forms nei-, ei-, inii- or nee-, ee-, ii- for no synchronic phonological reasons. Consider (26) below. The obligatory possessive prefix with this noun has a long vowel in all forms and carries lexical falling tone (marked with the circumflex diacritic over the vowel): (26) a. nêe-ˈcet ‘my hand’ b. hêe-ˈcet ‘your hand’

(Arapaho; Goddard 2015:363)

The allomorphy of the IPP in forms like (26) is diachronically conditioned by the sensitivity of the possessive prefix to the phonological form of the particular noun root. Nouns like (26) historically had a root beginning with a cluster that became *ʔC in Proto-Arapaho-Gros Ventre and lengthened the vowel of the prefix when the glottal stop was lost, which in turn produced the idiosyncratic falling tone (Goddard 2015:363). The phonological shape of IPP in (27) below, similarly, cannot be explained by synchronic surface phonology, but is explained as a historical consequence of the phonological interaction between the particular noun root and the obligatory possessive prefix: (27) a. nei-hʔeh ‘my son’ b. hii-hʔeh ‘his/her son’

(Arapaho; Goddard 2015:364)

According to Goddard (2015), the vowel in the IPP in (27) arose as a regular consequence of the loss of PA root-initial *k. A formal analysis of such a “fusion” of IPP and noun stems is beyond the scope of this paper, but capturing the allomorphy patterns like (26) and (27) would likely require positing morphological rules referring to structurally local interactions between a particular root and the possessive morphemes. The precise formulation of the locality conditions involved in these interactions is currently under investigation (see also Newell et al. 2018). Possessive Prefixes and Person Agreement on Verbs

Another issue left for future research is the relation between possessive prefixes and person agreement prefixes found on verbs. Most Algonquian languages

|

32    Ksenia Bogomolets

provide clear evidence that verbal agreement prefixes and APPs historically formed a single paradigm, and synchronically show no or minimal difference between the two. Thus, person prefixes marking agreement on verbs in most Algonquian languages show the /d, t/-∅ alternation found in the APPs and exhibit phonologically transparent interactions with the stem they attach to. When the APP paradigm and the verbal paradigm do differ, interestingly, these differences are localized in the third person. For instance, the only difference between the APP paradigm and the verbal prefix paradigm in East Cree, Plains Cree, Ojibwe, Wampanoag, and Blackfoot is that they lack the third-person marker in the verbal paradigm, while the third-person morpheme is present in the APP paradigm (cf. Table 3). A small number of languages show more significant differences between the APP paradigm and the verbal agreement prefixes. For instance, while the APP prefixes are /t/-final in Arapaho, verbal prefixes are not; i.e., they never exhibit the /t/-∅ alternation. Another example of an interesting interaction between verbal and possessive prefixes comes from Woods Cree. The APP and the IPP paradigms in Woods Cree differ in way that is a familiar by now—while the APPs have an underlying /t/, the IPPs do not; this contrast is neutralized before a consonant-initial stem. Interestingly, the verbal paradigm makes use of both sets of the possessive prefixes; i.e., it fluctuates between using the APPs nit-, kit- and the IPPs n-, k- before a vowel-initial stem. Consider (28) below: (28) a. nit-apin b. n-aðahwa:w 1-sit.down 1-bury.ta.1>3

‘I am sitting down.’

‘I am burying him.’ (adapted from Greensmith 1985:64)

The ‘fluctuating’ Woods Cree pattern may suggest an ongoing language change and indicate that the two paradigms, IPPs and APPs, are undergoing a merger within the verbal agreement paradigm. A more precise description of the distribution of the alienable vs. inalienable paradigm to mark agreement on verbs in this language is left for future investigation.

|

Possessive Prefixes across Algonquian   33

Conclusion

In this paper, I have argued that Algonquian languages provide ample synchronic evidence in favor of analyzing the alienable possessive prefixes and the inalienable possessive prefixes as two separate paradigms. I have argued that APPs and IPPs are underlyingly segmentally different: while the APP paradigm contains an underlying consonant—/t, d/, the IPP paradigm does not. Crucially, I have argued that /t, d/ in APPs cannot be analyzed as epenthetic consonants to resolve vowel hiatus. Instead, I have proposed that the phonological behavior of these consonants represents a pattern of liaison. I have shown that such an analysis offers a better empirical coverage when compared to the earlier epenthesis accounts. Notes 1. I am grateful to Paula Fenger, Robert M. Leavitt, Adrian Stegovec, and Saurov Syed for their comments on earlier versions of this work. I also thank Peter Bakker, an anonymous reviewer, the editors of this volume, and the audience at the 53rd Algonquian Conference for their insightful suggestions. 2. The question whether the pre-stem possessive morphemes (as well as the pre-stem morphemes marking person agreement in verbs) are affixes or clitics is not relevant for the current paper, and for ease of exposition I will be referring to them as prefixes (see, however, Déchaine 1999; Halle and Marantz 1993). 3. The following abbreviations used in this paper are not included in the Leipzig Glossing Rules: ep = epenthetic, ic = initial change, obv = obviative, PWd = Prosodic Word, sf = surface form, ta = transitive animate, uf = underlying form. The orthography of the original is used in all examples except the IPA symbols for glottal stop /ʔ/ and voiceless dental fricative /θ/ are used here. The circumflex over a vowel marks lexical falling tone [ê]; vowel length is marked with colon [e:] or double vowels [ee], depending on the source; raised vertical line [ˈ] marks primary stress. 4. See section below for further discussion of Newell and Piggott 2014. 5. Note that /h/ is regularly inserted as the word onset for underlyingly vowel-initial stems in Arapaho as onsetless syllables are banned in the language (Bogomolets 2020:46–50). The vowel alternation /e/-/o/ in the prefixes is due to a productive vowel harmony process; see Cowell and Moss (2011:20–22). 6. It should be noted that different dialects of the languages listed in Table 3 might show slight differences in these inventories, but the relevant pattern always remains

|

34    Ksenia Bogomolets

the same where the APPs show the prefix-final /t, d/ before a vowel, and IPPs never do. The inventories of the possessive prefixes in Table 3 are based on the following sources: Cowell and Moss 2011 (Arapaho); Dahlstrom 1991 (Plains Cree); Fermino 2000 (Wampanoag); Junker 2000–2019 (East Cree); Leavitt 1996, Passamaquoddy-Maliseet Language Portal: PMLP, retrieved February 2022 (Passamaquoddy-Maliseet); Piggott 1980, Newell and Piggott 2014 (Ojibwe); Wolfart 1973 (Plains Cree). 7. I thank Peter Bakker who pointed out to me that these consonants appear to be an ancient element dating back to Proto-Algic as the same -t- element appears in the alienable possession construction in Wiyot (see Bakker 2006; Goddard 1975). 8. Newell and Piggott (2014) analyze /d/ as epenthetic in Ojibwe in another environment, namely when the future tense morpheme ga(d)- is followed by a vowel-initial stem. Consider the underlying form (UF) of this morpheme as analyzed in Newell and Piggott (2014:355): in (i), ga(d)- is followed by a C-initial morpheme while in (ii) it is followed by a V-initial morpheme:

(i) SF nigagi:we:se:



(ii) SF nigada:gamose:

UF ni-ga-gi:we:-ose:

UF ni-ga-a:gam-ose:

1-fut-go.home-walk 1-fut-snowshoe-walk

‘I will (probably) walk home.’

‘I will (probably) walk in snowshoes.’

The analysis of /d/ in this environment as an epenthetic consonant to resolve hiatus is, however, incorrect since /d/ in gad- is present underlyingly and can be traced back to its Proto-Algonquian source (/t/, iii) as well as to the cognates in other modern Algonquian languages (iv). Importantly, the final consonant of the future tense prefix does not syncopate in other Algonquian languages when the following morpheme is consonantinitial: (iii) *kataw- ‘want to, intend’  *kata future tense (iv)

PA *kata future tense > Ojibwe gad, Cree kata, Illinois kata, Arapaho het, Massachusett kat, Munsee kata (adapted from Pentland 2005:326)

A liaison analysis, parallel to the one proposed below for the /t, d/ consonants in person prefixes, can straightforwardly account for the /d/-Ø alternation in the future tense morpheme in Ojibwe (i)-(ii). Details are, however, beyond the scope of this paper. 9. A large number of independent nouns in Arapaho (and in Algonquian more generally), when possessed, are suffixed with the so-called ‘possessive theme marker’ -w/-b (9b).

|

Possessive Prefixes across Algonquian   35

These nouns refer to animate objects, body parts, or commonly possessed referents (see Cowell and Moss 2011:63; Goddard 2015:362). This possessive suffix is not discussed in the current paper. 10. On the significance of empty elements in the CVCV Phonology as well as on the principles behind Government and Licensing within the model, I refer the reader to Scheer (2004 and subsequent work). Structures given in this paper are simplified for the sake of space, and Government and Licensing relations are not marked in the structures. 11. I am grateful to Robert M. Leavitt for bringing to my attention that a small number of exceptional dependent vowel-initial noun stems take the /t/-final possessive prefixes in Passamaquoddy-Maliseet either regularly or as a variant in addition to the nonexceptional IPPs (which do not have the prefix-final /t/, cf. Table 3 above). The Online Passamaquoddy-Maliseet Dictionary (retrieved from http://www.pmportal.org) contains thirty-one such forms. All of these exceptional cases involve one of the following factors: (a) A noun has been misclassified by the authors of the dictionary as a dependent when in fact it is independent. (b) A noun is deverbal. Since verbs in Passamaquoddy-Maliseet use the Alienable paradigm prefixes for marking person agreement (see below), we find these used with dependent nouns, which are formed from verbs. (c) Both variants exist in the language—the regular Inalienable prefixation and the exceptional prefixation with the Alienable paradigm, suggesting, perhaps, language change in progress. (d) Relatedly, a number of dependent noun stems, which exceptionally take the /t/-final prefixes, also take the possessive theme marker -Vm, which is generally restricted to independent nouns across the family (see also note 9 above). This, according to Robert M. Leavitt (personal communication), appears to be a recent innovation. Such a ‘borrowing’ from the Alienable Possessive paradigm may be suggesting that these nouns are unstable in their dependent status as a result of a relatively recent language change. 12. Note that the epenthetic /i/ conditions assibilation of the prefix-final /t/ in (20a). The /i/-epenthesis in consonant clusters and the assibilation/spirantization exemplified in (20a) are regular phonological processes in Blackfoot (see, for example, Elfner 2004). 13. I have assumed that the /t, d/ consonants in the APP paradigm are part of the possessive prefixes. An anonymous reviewer mentions that, in addition to the evidence presented in this paper, speakers systematically classify possessive prefixes into two distinct paradigms, although I have not been able to confirm this. It should be noted, however, that an alternative analysis is possible where the /t, d/ consonants are not part of the possessive prefixes proper but are a separate morpheme by themselves. Postulating that the /t, d/ consonants are (or were historically) a morpheme separate from the possessive prefixes would mean that Algonquian possessive marking aligns more closely with a

|

36    Ksenia Bogomolets

cross-linguistically robust behavior of the alienable vs. inalienable marking. In languages that have both, alienable possession is virtually universally marked in a similar way to inalienable possession, but the marking for alienable is longer and more morphologically complex than that for inalienable possession. The additional morpheme in the alienable possessive paradigm has been described as an invariable ‘passive’, ‘possessive’, or ‘pertensive’ affix for a wide variety of languages (for a summary see Dixon 2009:277–297). Whether /t, d/ are analyzed as part of the APPs or as a separate possessive morpheme, the main claim of this paper remains the same, namely that these consonants must be underlying rather than epenthetic across Algonquian in the alienable possession environment. 14. I focus here specifically on the account proposed in Newell and Piggott (2014) as this is the only existing detailed formal account of the presence/absence of the /t, d/ consonants in APPs vs. IPPs.

References Bakker, Peter. 2006. Algonquian-Ritwan, (Kutenai) and Salish: Proving a Distant Genetic Relationship. University of British Columbia Working Papers in Linguistics 18:1–32. Bennett, William. 1991. Liaison in French. Word 42(1):57–88. Blevins, Juliette. 2008. Consonant Epenthesis: Natural and Unnatural Histories. Language Universals and Language Change, ed. by Jeff Good, pp. 79–109. Oxford: Oxford University Press. Bogomolets, Ksenia. 2020. Lexical Accent in Languages with Complex Morphology. PhD thesis, University of Connecticut, Storrs. Clements, George N. and Samuel Jay Keyser. 1983. CV Phonology. A Generative Theory of the Syllable. Linguistic Inquiry Monographs 9:1–191. Cambridge, MA. Cowell, Andrew and Alonzo Moss Sr. 2011. The Arapaho Language. University Press of Colorado. Dahlstrom, Amy. 1991. Plains Cree Morphosyntax. New York: Garland Publishing. Déchaine, Rose-Marie. 1999. What Algonquian Morphology is Really Like: Hockett Revisited. Papers from the Workshop on Structure and Constituency in Native American Languages, ed. by Leora Bar-el, Rose-Marie Déchaine, and Charlotte Reinholtz. Cambridge, MA: MIT Occasional Papers in Linguistics 17:25–72. Elfner, Emily. 2004. The Role of Sonority in Blackfoot Phonotactics. Honors thesis, University of Calgary. Fermino, Jessie Little Doe. 2000. An Introduction to Wampanoag Grammar. MA thesis,

|

Possessive Prefixes across Algonquian   37

Massachusetts Institute of Technology. Frantz, Donald. G. 1991. Blackfoot Grammar. Toronto: University of Toronto Press. Goddard Ives. 1975. Algonquin, Wiyot, and Yurok: Proving a Distant Genetic Relationship. Linguistics and Anthropology: In Honor of C. F. Voegelin, ed. by M. Dale Kinkade, Kenneth L. Hale, and Oswald Werner, pp. 249–262. Lisse: Peter de Ridder. Goddard, Ives. 2007. Reconstruction and History of the Independent Indicative. Papers of the 38th Algonquian Conference, ed. by H. Christoph Wolfart, pp. 207–271. Winnipeg: University of Manitoba. Goddard, Ives. 2015. Arapaho Historical Morphology. Anthropological Linguistics 57(4):345–411. Greensmith, Jennifer M. 1985. Future Markers in Woods Cree. Papers of the 16th Algonquian Conference, ed. by William Cowan, pp. 63–72. Ottawa: Carleton University. Halle, Morris and Alec Marantz. 1993. Distributed Morphology and the Pieces of Inflection. The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, ed. by Kenneth Hale and Samuel Keyser, pp. 111–176. Cambridge, MA: MIT Press. Junker, Marie-Odile (ed.). 2000–2019. The Interactive East Cree Reference Grammar. https:// www.eastcree.org/cree/en/grammar/. Leavitt, Robert M. 1996. Passamaquoddy-Maliseet. Munich and Newcastle: Lincom Europa. Lowenstamm, Jean. 1996. CV as the Only Syllable Type. Current Trends in Phonology: Models and Methods, ed. by Jacques Durand and Bernard Laks, pp. 419–443. Salford: European Studies Research Institute. Newell, Heather and Glynne Piggott. 2014. Interactions at the Syntax-Phonology Interface: Evidence from Ojibwe. Lingua 150:332–362. Newell, Heather, Glynne Piggott, and Lisa Travis. 2018. The Possessive Structure of Ojibwe: Support from Cupeño. Proceedings of the Workshop on the Structure and Constituency of Languages of the Americas, 21:169–184. Oxford, William Robert. 2014. Microparameters of Agreement: A Diachronic Perspective on Algonquian Verb Inflection. PhD thesis, University of Toronto. Passamaquoddy-Maliseet Language Portal (PMLP). Language Keepers and PassamaquoddyMaliseet Dictionary Project. http://www.pmportal.org. Pentland, David H. 2005. Preverbs and Particles in Algonquian. Papers of the 36th Algonquian Conference, ed. by H. Christoph Wolfart, pp. 323–338. Winnipeg: University of Manitoba. Piggott, Glynne. 1980. Aspects of Odawa Morphophonemics. New York: Garland Press. Proulx, Paul. 1989. A Sketch of Blackfoot Historical Phonology. International Journal of American Linguistics 55(1):43–82. Scheer, Tobias. 2004. A Lateral Theory of Phonology: What Is CVCV, and Why Should It Be? Berlin/New York: Mouton de Gruyter.

|

38    Ksenia Bogomolets

Starks, Donna Joy. 1992. Aspects of Woods Cree Syntax. PhD thesis, University of Manitoba. Tranel, Bernard. 1986. French Liaison and Extrasyllabicity. Studies in Romance Linguistics, ed. by Osvaldo Jaeggli and Carmen Silva-Corvalan, pp. 283–395. Dordrecht: Foris. Wolfart, H. Christoph. 1973. Plains Cree: A Grammatical Study. Transactions of the American Philosophical Society 63(5):1–90. Żygis, Marzena. 2010. Typology of Consonantal Insertions. ZAS Papers in Linguistics 52:111–140.

Stories of Empowerment Indigenous Women’s Leadership Summer School Maggie Chittspattio, Nadia Ferrara, Ashley Guanish, Mary Jane Hannaburg, Denise Larocque, Amélie La Rose Gauthier, Anahi Morales Hudon, Anny Morissette, and Marie-Louise Perron

I

n 2018, Quebec Native Women (QNW) and Saint Paul University (SPU) signed a Memorandum of Understanding to develop a summer school on Indigenous women’s leadership.1 The aims of this partnership were to offer Indigenous women leaders an opportunity to acquire tools and resources to support and strengthen their empowerment journeys as leaders, advocates, and agents of change in their community; and to offer building institutional capacities to develop pedagogical approaches inspired by Indigenous practices and knowledges through creating safe learning cultural spaces in which Indigenous women leaders can gather strength to continue to make a difference in their community. This paper is an opportunity for Indigenous women leaders, speakers, and organizers to reflect on their experience of the QNW-SPU Summer School. We explore the transformative experience and the journey of experiential growth lived by all who took part in this project. While there is a growing literature on the co-creation of university and Indigenous community programs, the journey of decolonization and indigenization of universities is rarely documented centering on first-person reflections of participants. We hope that documenting the QNW-SPU Summer School’s experience can contribute to the efforts to decolonize postsecondary education. The 2018 summer

|

   39

|

40    Chittspattio et al.

school truly embodied the concept of decolonial indigenization as described by Gaudry and Lorenz (2018:219): “Decolonial indigenization envisions the wholesale overhaul of the academy to fundamentally reorient knowledge production based on balancing power relations between Indigenous peoples and Canadians, transforming the academy into something dynamic and new.” By presenting this collection of first-person reflections from the participants of the program, we hope to reorient knowledge production, to build on and enrich a growing documentation of the voices of Indigenous and Algonquian women leaders beyond the official political sphere. Those voices and perspectives are needed for the advancement of Indigenous education programming and Algonquian studies as a whole. Indigenous women’s experiences need to be acknowledged and represented in their whole forms precisely due to the silencing of Indigenous women’s voices in Canadian society historically and contemporarily. Michi Saagiig Nishnaabeg scholar Leanne Betasamosake Simpson (2015) notes that “recognition for us is about presence, about profound listening and about recognizing and affirming the light in each other.” Decolonization is not only honoring the life of Indigenous women but also honoring their word. Here we provide some context on the program developed in the partnership, its content, and, most importantly, different narratives that document, each from a unique approach, the importance of restoration, empowerment, and connection for leadership training for Indigenous women. We hope these narratives will inspire others as we believe our experience demonstrates the potential of programs that embody Indigenous values of relational reciprocity where we can learn from one another.

Partnership History: Building Relationships

This journey began with an invitation to discuss a potential collaboration to develop a summer school in English for QNW in 2017. At that moment, QNW was planning a collaboration with the Université du Québec à Montréal for their first French summer school for Indigenous women leaders. We saw this project as a way to contribute to the efforts of decolonizing academia, to create spaces and projects that responded to Indigenous organizations and communities’ needs. In academia, little work experience exists in how to develop partnerships and to work in collaboration at every step of projects involving Indigenous peoples and organizations.

|

Stories of Empowerment   41

The first step in the right direction was to develop a Memorandum of Understanding to clearly establish the decisional process, the responsibilities, the resources available and the goals. We knew this project would involve important work from inside SPU to be respectful of the ways and goals of the partner. Generally, universities’ administrative staff and faculty are insufficiently prepared to concretely advance the efforts of decolonization. We mobilized allies within the university who understood the importance of taking the time to develop a partnership, people who knew we were going to face obstacles. It took months and considerable effort to educate and sensitize university employees, because even if the university leadership supports a project that innovates in its ways of doing, it takes constant vigilance to make sure consultation and validation are carried out properly at every step of the way. We hope that the summer school represents a contribution in this direction, principally, through what we have learned on how to build a meaningful relationship with our partner.

The Summer School

The 2018 QNW-SPU Summer School was a practical workshop that aimed to enable Indigenous women leaders to acquire and to share their knowledge, know-how, and skills as well as to strengthen their ability to take up their rightful place in the governance of their communities. The three-credit university course examined Indigenous community governance from an Indigenous woman’s perspective by exploring traditional and contemporary practices; changing roles and structures influenced by colonization. Using Indigenous pedagogical ways, the course offered participants a window into the unique knowledge of First Peoples’ political cultures, traditions of agency, leadership, and decision-making. Both partners, QNW and SPU, were involved in the content creation process. Co-creation involves prior discussions about the vision and needs of the Indigenous partner and the learners, back-and-forth drafts of the curriculum, suggestions, and approvals of activities and guest speakers. This process involved constant consultation. On several occasions, we had to rein in some of the actors involved at SPU, to explain the need to wait for approval from QNW because they had a voice in institutional decisions regarding this project. As an Indigenous scholar, author, and pioneer in the advancement of Indigenous education, Archibald (2008:90) rightly noted, “I believe that the Elders’ reminder to us to take time to talk in order

|

42    Chittspattio et al.

to ensure correct representation of their Indigenous knowledge is an example of engaging in both decolonization and transformative-action processes.” This is what we experienced. Elder Annie Kishkwanakwad St. Georges Smith taught us during the summer school: you have to take the time to do things right. We recognize that Canadian universities have a long way to go to properly and respectfully implement Indigenous education with and by First Peoples. As stressed by Laura McLennan and many others, “Indigenous peoples experiences with the mainstream education system have been negative as formal education has been a key component of colonization” (Crey and Fournier 1997 in McLennan 2014:5). The founders of Saint Paul University, the Oblates of Marie Immaculate, have participated in this colonization process as they were one of the congregations in charge of residential schools that aimed at acculturating Indigenous children with a Western education. Faced with the assimilation agenda, in the last three to four centuries, Indigenous knowledges, ways, and worldviews have been devalued (Haig-Brown 2000:4) and regrettably not recognized in formal education. It was clear to both partners that in order to provide a positive university experience, Indigenous knowledge had to be restored to its rightful place by recognizing its role in the summer school as part of formal education. The intended pedagogical direction was the holistic approach advocated within Indigenous education. By holistic approach we mean “educating the mental, emotional, physical, and spiritual being of the learner (Archibald 2008; Battiste and Barman 1995; Nee-Benham and Cooper 2000). All aspects of one’s being are equally important” (McLennan 2014:13). That is why we called upon a Kitigan Zibi Anishnabeikwe Elder to guide us. Elder Annie Kishkwanakwad Smith St. Georges has been involved in all levels of education and in several national groups for many years, sharing her traditional values, teachings and insights, and promoting harmony between one another and positive values of Indigenous peoples (Dion Stout 2009:28).2 As the land is pedagogy for First Peoples, it was of the utmost importance to recenter part of the summer school activities outside the university walls to establish this connection to the land. Since Indigenous education traditionally did not take place in a classroom setting, we wanted to provide a learning experience with various learning environments/places including social activities and events. For example, our afternoon at the Ottawa Inuuqatigiit Centre for Inuit Children, Youth and Families was an opportunity to discover this organization and to be introduced to Inuktitut and Inuit values and traditions. It was also an occasion to put into practice social leadership by sewing baby blankets for the

|

Stories of Empowerment   43

Inuit babies in Ottawa regional hospitals. The art therapy session with Dr. Nadia Ferrara was a highly emotional healing workshop for all those who participated. Jane Dickson, law professor; Gabrielle Fayant, founding member of Assembly of Seven Generations; and Carrington Christmas, Native Women’s Association of Canada Youth Advocate, among others, took part in the Speaker circles. It was important for us to be inclusive of Inuit, Métis, First Nation, LGBTQ2S+, and the voices of different generations throughout the summer school. The opening and closing ceremonies, teachings on Asinabka,3 and our graduation ceremony were noteworthy unifying events at the heart of our program. Elder Annie Kishkwanakwad Smith St. Georges brilliantly enacted what governance is all about by bringing us together as a group to learn from one another and build strength together. In traditional Indigenous education, the responsibility for passing knowledge does not rest solely on the shoulders of the Elders but on the whole community (Haig-Brown 2000:2). This demonstrates and embodies Indigenous values of relational reciprocity. With this in mind, we believe that the lessons to be learned from the summer school for both the university and the Indigenous partner organization must come from the women leaders themselves. Their full testimonies are a wealth of lessons to be learned and recommendations for the delivery of community-university programs of this kind.

Maggie Chittspattio: My Ideas and My Voice Were Heard

The summer school was a very nice experience for me. As an Indigenous woman, it has always been hard to feel comfortable outside my community. The summer school’s team made this all possible; they made it very comfortable and welcoming for us. My past experiences with formal education were never easy for me. I was very young when I dropped out of school. I was on my own and struggled and felt stressed. Education was not our way; it was pushed to us by Europeans. I always had that feeling of not receiving support and not wanting to be seen to achieve anything. The struggles I had as an Indigenous woman in a non-Indigenous community were not like what Saint Paul University offered us. The summer school felt like home, so much support. The Elder brought comfort. I felt secure, it was very welcoming, warm, at home. The most important feeling that I experienced was that my ideas and my voice were heard. This gave me self-confidence, the beginning of resilience

|

44    Chittspattio et al.

in me. I was a person who didn’t speak out, minded my own business, and zipped it. Today, I voice myself as much as I can. I got my confidence from taking this course. I voice myself more than I did. My experience helped me to better understand our ways as Indigenous people, taking back our ways, sharing our life experiences and how we all shared common fights wanting to get back our rights and our ways, relearning what has been taken away from us. The settlers brought their ways of education to my kind, taking away our ways. We were forced to learn the settlers’ way. We were taught our ways were not the way. Taking this course was eye-opening to me in how bad we were treated, the struggles First Nations had, and that there was a lot of good we had as First Nations. It was an opportunity to relearn how my ancestors taught us. We didn’t have schools or teachers. But we did have our lands and our Elders teaching us the importance of life and the land, of family and providing for the family. My ancestors were nomadic people, following the caribou as the seasons changed. Sadly, not anymore. We could never go back to our ways, another thing that was taken away from us. My bloodline carried strong women, starting from my great-grandmother Nisabat (Elizabeth). She was my teacher in our own ways. The stories she would tell us were always interesting. Her upbringing was so different from mine. My grandmother’s generation was introduced to alcohol and her two daughters were sent to boarding school where my aunt and mother faced a lot of discrimination! There have been strong ladies in my bloodline. They have faced so much pain and hurt and embarrassment throughout this change in their lives. Still, they all showed so much strength and power to me. If it weren’t for them, I wouldn’t be where I am today. I was very happy and honored to have shared my history with my super awesome classmates and professors about my family history of strong women. I felt so comfortable sharing, learning from them as well. I was glad I had the chance to attend this course with my mother, much needed for me.

Ashley Guanish: We Empowered One Another

Wachiya, my name is Ashley Guanish from the Naskapi Nation of Kawawachikamach. I would like to start off by sharing why I applied to the summer program. It was out of character because I get easily intimidated by the thought of higher

|

Stories of Empowerment   45

education, the language barriers. Like most arriving from a remote traditional territory, I was also intimidated by the vast diversity of urban living and stepping out of my comfort zone. It was a cultural shock, as they say, and it is intimidating. It took me a while to even put my name on the application form, to push my boundaries and be uncomfortable and be scared but to maybe inspire women, girls, and youth. It was an opportunity to bring something back and I thought of it that way as I filled out the form. My curiosity took over and quieted the doubt in my soul. Intrigued by the opportunity for personal growth and experience, and possibly to be sitting with kindred spirits that could contribute to my healing, I was grateful to be admitted into the program. Feeling vulnerable and with an open mind, I submerged myself into the course and did not anticipate what I found. We were very privileged to be gathered on Victoria Island with the Elder Annie St. Georges Smith. We were introduced to stories and the welcoming, the weather, and the food was perfect. It was the best way to start off and welcome the weary distant travelers. With formal education, we learn class material, readings, and competencies. Course content was introduced and delivered, although the difference with this summer school to other programs were the passions and emotions from the facilitators themselves, the openness of the content, dialogue, and the inviting environment that made it possible. Each of us students, Indigenous women leaders, were given the space necessary to reconnect as Nations, as women, as Matriarchs. All our sharing was felt in the stories, in the individual spoken words and in the hearts, the sufferings felt by our peoples everywhere, to the celebrations of life from time immemorial. Each session we would sit in a circle with open bodies, minds, and hearts. We connected with each other. I am honored to have sat in circles with these women as we empowered one another. Personally, I experienced identity revival, growth, and confidence as a Naskapi woman, mother, daughter, and friend with the help of my sisters, my classmates.

Mary Jane Hannaburg: Wari Goes to University

I am a Kanien’kehá:ka woman from Kanehsatake where I work as a mental health support worker. I have an educational and professional background in the field of support work. I completed an undergraduate degree with Concordia University. I am a registered addictions specialist, certified crisis worker, and a SafeTalk trainer

|

46    Chittspattio et al.

with Living Works. I also have a background in advocacy, as I have been actively involved in the Quebec Native Women movement for a number of years, and have completed three consecutive mandates as the vice president for the organization. The Indigenous women in the governance and leadership program truly benefited me in many ways. It was both intellectually stimulating as well as a healing experience for me. It has had a lasting impact on me and the relationships within both my personal and professional life. This experience brought me change and growth. I was reluctant to leave my comfort zone, my familiar environment, my work, and home life, where I feel I am needed. It was difficult for me to just pick up and leave my familiar setting. Participating in this program brought me opportunities to connect with the land, with my environment in ways that even as someone living in a rural community, I have been cut off from. It was very therapeutic for me to get out on the land. Had I not left, I may have not noticed the sounds of the water and the birds that I noticed while we were on the island. Noticing these things really had a strong impact on me while I was there. The experience is immersive and allowed me to connect in ways I would not have achieved had I remained at home, within my day-to-day patterns and routine. I felt this program allowed me to step out of the familiar and open up to new and stimulating experiences. We kept journals and wrote about our experience. In my first entry I depicted myself using a little doll-like figure “Wari goes to university.” She has eyes to take in information with, she sees. I wrote about teachings that were shared with me, things that really stuck with me. I wrote about the Elder who came to share messages about perseverance and how small changes have ripple effects and can lead to great change. I wrote about the story of an Indigenous woman who walked great lengths to deposit a letter at parliament to advocate for her community. How such a seemingly small gesture has such impact. It moved me. I wrote about my reflections on the natural world, our environment, and our relationships with the land. This course opened up my eyes, mind, and heart. This learning experience was markedly different from my previous learning experiences throughout my academic journey. My learning experiences have been more rigid in structure, strict, with a focus on writing and individual work. This course, through its format, content, and speakers, sparked my curiosity. I became curious about myself and my environment. It invited me to bring my emotions into my learning, not leave them aside. I learned differently.

|

Stories of Empowerment   47

Participating in this program supported the connection of my mind, body, and spirit as well as facilitating community connections by bringing Indigenous women together. The course could not have been what it was without the participation of everybody that was there. We are a group of women, each with our own individual gifts, brought together by this experience where we very organically shared our gifts with each other. I was reminded about that, about the gifts I have, and I built on and strengthened these as a result of my experiences within this program. There was a dynamic among our group that cannot be replicated or recreated. Overall, I felt it was a safe space for me to open up, be curious, learn, and connect and share. I carry with me the teachings and tools that I was gifted with during our exchanges. For example, I received teachings from Lee Maracle about the roots of trauma that really nourished my understanding, and this I continue to carry in my work within my community. We are facing long-term impacts that colonialism has had on us and continues to have on us as a result of our constant exposure to institutions based on and built around colonial ideologies. Decolonization means to me to restore everything that was taken: our languages, our practices, our values, our identities, our pride. I have been reflecting on this process of decolonization and how it is about change and how these changes start from the center; these changes start seemingly small and then grow as they move outwards, like ripples. Decolonization is a change that has to start with us. There have already been changes made in this way, and continue to be made, and we have a lot more work to do. I believe that programs like these are contributing to that ongoing work. I believe this course and initiatives like it are essential in empowering and supporting young Indigenous women, equipping them and nourishing them in ways that will have meaningful and significant impacts on their families, communities, and nations. By supporting young Indigenous women, we are supporting mothers, sisters, daughters, families, communities, nations.

Denise Larocque: From Students to Knowledge Keepers

My experience with the Indigenous Women’s Governance course was an experience that will forever be etched in my mind, a profoundly life-changing experience. I applied because of my love of learning, my thirst for knowledge, and a desire for

|

48    Chittspattio et al.

growth as a leader. I had been involved with the Quebec Native Women for over twenty-five years, and was, at the time, the District Representative for the Mi’gmaq Women of Quebec with a seat at the table. I enrolled with hopes to strengthen my leadership skills, which was important to me as it was my second time in this position. My love for learning was at times clouded by a past of poor grades in high school causing self-esteem issues, now showing itself as impostor syndrome. The summer school, however, felt different than any other learning environments I had become accustomed to. Our classroom felt like a safe place, devoid of judgment or competition. It allowed for our small group to share and connect from a place of vulnerability, wisdom, and life experiences, learning that many common threads connected us; Indigenous, women, strong, and determined. The classroom was not as expected, a glass box placed within a modern industrial communal space. The glass walls allowed us to watch the buzz of students either walking through or studying in the wide-open space with individually placed cubicles, plush couches, dining area, and shared tables. The exposed beams, pipes, and wires brought a warmth, paired with a cool gray cement floor, which made this such a unique space that I looked forward to returning to each day. An important part of the program was that the classroom was also the land, bringing importance to the use of ceremony and Elder, traditional medicines. Culture and knowledge transfer made this experience so moving and crucial; it created clarity on the importance of indigenizing education as not only beneficial but a form of study, as well. Out of the box and onto the land, we went from being students to knowledge keepers, having an Elder with us as we started with a gathering on Victoria Island. I was moved as I stood by the wigwam, looking across the water over to the Parliament Buildings; a feeling washed over me that the city grew up and around us, saving this little plot of land that was still somehow belonging to the First Nations People and we were still here! Smudging with the sage and abalone shell, I felt connected and rooted in culture, and learning could begin. We were five amazing women I grew to love and respect. Within the classroom, our glass box, I received a new way to see the work of our teacher Anny. How she worked tirelessly to gather this information to share with us: history of women in politics, the distinctions they have compared to their male counterparts, nurturers vs. providers, the ever-so-slightly changing system that still oppresses us. Sharing our impressions of our nightly readings, I was able to see how my lens is not the same

|

Stories of Empowerment   49

as others, yet my voice is equally important. Anny brought in speakers, authors, and youth who held wisdom, grace, and took their place in the world. The work they have done shows that each of us has a chance to leave a mark on this world. Through their shared knowledge, I found gratitude within those moments. The classroom was not only inside the industrial glass boxed room, or out on the land, but we also visited the Inuit Center where we helped make blankets for newborn babies and learned about the work the center does for its people and community. We learned of their culture, traditions, and the use of facial movements as a language, and also the language we share, laughter. Making the tiny quilts for babies with ulus, harpoons, knives, and other Inuit symbols on them allowed me to see the importance of connection to culture, the importance of seeing ourselves in the objects around us. A pivotal moment in the program was when Anny with loving, kind eyes thanked us. I stayed behind after class to ask why she was thanking us. In my eyes, she was the hero for the remarkable work she put in. I had imagined the number of hours spent researching, and putting together this knowledge in a comprehensive way to be passed on. The fact that colonialism has severed the way we once verbally passed down our history, traditions, and knowledge was a devastating loss, but I now was open to learning in this context that, even if not equal, still held importance. The link to storytelling and knowledge transfer was now being passed on to me through PowerPoints, articles, books, and with invited guest speakers who themselves wrote books. When she answered that she was thanking us because we were the ones out there doing the work, as cliché as it may sound, that was a light bulb moment for me! My mind at hyper-speed started to make the link that the very ones we were learning about—Mary Two-Axe Earley and her mission to advocate about gender discrimination in the Indian Act, the courage for change shown in the women who requested their Indian agent to ask the government for funds for sewing machines, which at the time may also have challenged the laws and norms of women gathering—were creating space for strength and voice to arise. Then there were Mi’gmaw women chiefs of Gaspé and Listuguj all leading the way for those who will come after, all women who in their own way, were out there doing the work. In that moment, I could see the common thread I shared with these women, that “I” being universal, which changes to “we,” always for the people, always for the next seven generations. Learning, in that moment, that our time and voice matter in making change in the world, I could see that volunteering my time to sit on Quebec Native Women’s

|

50    Chittspattio et al.

board or sacrificing some of my summer vacations to learn about women in leadership so that I could learn to lead better, was indeed a noble act. The course opened my eyes not only to my own worth and the important contribution I bring to this world; it also opened my eyes to the world around me, seeing importance in making the changes needed to be sure that our people have a say in the styles of learning, the acknowledgment that we learn differently, the connection through heart, culture, and traditions. Recognizing that land, medicines, and our relations are an important part of our learning and in who we are, validates our way of learning that differs from Western modalities. An example would be during the graduation ceremony when we were called up one by one to greet several people of the school along with the president of Quebec Native Women, and the Elder, who was furthest down the line, holding a handmade bag for us. All of us went to the Elder first; it was assumed that we wanted the bag over the certificate, but my response was: “In our culture our Elders are of the most importance, therefore we greet her first.” The positioning of her last in line shows the Western influence of the graduation. This is why it matters that we bring our opinions and voices to the table when creating programs for First Nations People, whether government programs, land and fisheries, housing, health, or education. In our counseling offices at my work hangs the sign Do nothing for us without us, and this message needs to trickle into all areas of social, political, and environmental decision-making agendas when it comes to First Nations People. To close out our program we were asked to prepare a personal project, which was not limited to a written paper. The project I selected was yoga, something I am passionate about, yet I have an untrained mind and often found it hard to get on the yoga mat. The process took searching for the perfect song, one with rhythm that would show the ebb and flow of each pose. Something with a heartbeat of a native drum that supported the calming movements of asana and breath. Getting on my mat every day to create and practice the routine was a struggle with the mind. Meticulously, I searched for the perfect dress. When it finally came time, I walked down the steps of the dimly lit auditorium, with one bright spotlight over the space where I would perform. A loud slapping sound filled the auditorium as the mat unrolled in the air and hit the floor, breaking the deafening silence of the room, where only my peers sat as on-lookers. Even after spending two weeks with them, I could feel my heart pounding as I walked bare-footed across the cold tile floor gathering my notes. I shared my struggle to get on the mat, under the warmth

|

Stories of Empowerment   51

of the spotlight, through a cracking voice trying to hold back the tears as I spoke of the space between self-love and self-sabotage and how it shows up in my failure to get on the mat and do something that I know will fill me up with immeasurable benefits, including the self-worth and growth that yoga gave me each day for those two weeks at school both on and off the mat. I shared my increase of strength of muscle and of mind, the clarity I received, the increase of joy in my ability to see my strength and stamina. When I finished, my classmates and teacher gave praise; some said it made them cry, others wanted to try yoga again after a failed attempt. Our ability to share our passions with each other that day showed pride in the form of a glow in our faces, possibly from the ability to move through the fear of presenting and a restored feeling, like we can take on the world. My gratitude for this experience will be shown in how I continue my journey through this life, how I move through this world; in the actions I take, the words I choose, and what I do with this life after my summer studies, which have continuously given me courage to take my place in this world, know my worth, and see more clearly the unique gifts each of us have. I know that our voices matter in and out of the class, self-compassion is important both on and off the mat and into the world, that by choosing to be enrolled in the course we were already being courageous . . . just by being there. I walked away with hope and the determination to want to inspire change.

Teachings from Participants: Restoration, Empowerment, and Connection

Unquestionably, for the participants, the summer school was a training experience different from formal education in its content, format, and dynamics. They appreciated the recognition and valorization of Indigenous cultures, Elders, territory, knowledge, ways of doing, and ancestral medicines. The ceremonies, talking circles, knowledge transfer activities, sharing of life experiences, and the importance of storytelling created a welcoming and safe learning environment. Being surrounded by Indigenous women, having ceremonies and Elders teaching in situ on the land, provided an environment conducive to cultural revival and identity revival for the participants. For the first time, Mary and Denise learned differently because introspection, emotions, and their whole being were part of the

|

52    Chittspattio et al.

teachings. A sense of being rooted in the cultural teachings of the program emerges from the testimonials. For Mary, restoring everything that was taken away is what decolonization is about. By validating their truths, encouraging them, emphasizing their personal abilities, the participants felt through the program that they have value, that their visions and opinions matter and make a difference. Promoting self-worth is a key component in empowering women. Denise noted, “I was able to see how my lens is not the same as others, yet my voice is equally important.” The focus on the participants’ unique voice and individual gifts was a cornerstone of the summer school. Hence, Mary “built on and strengthened these as a result of my experiences within this program.” Ashley and Mary both experienced change and growth during the program. A deep sense of connection or reconnection emerges from the women’s accounts of their experience. As Maggie noted, sharing their life experiences made the participants realize how much they had in common: their fights for their rights and their struggles to regain what they lost because of colonization, Western education, and discrimination. Attending the summer school with her mother gave Maggie the opportunity to reconnect with her. For Mary, the program with its immersive experience was an opportunity “to connect with the land,” to connect with her mind, body, and spirit. For Denise, the smudging performed by an Elder at the first morning gathering on Victoria Island allowed her to feel “connected and rooted in culture.” The teachings of the Elder on traditional sacred territory anchored the women leaders to Mother Earth, the source of their identity, and a source of empowerment. As mentioned earlier, the summer school education was not only intellectual, but it engaged the body, the emotions, and the mind. The knowledge passed down by the Elder also enabled the participants to have a balanced and respectful vision of their role as leaders. Participation in cultural and spiritual practices also strengthened the women’s solidarity because mutual aid, cooperation, and sharing allow women to remain strong and sometimes survive. With the help of the other participating women leaders, and because they all “were given the space necessary to reconnect as Nations, as women, as Matriarchs,” Ashley experienced “confidence as a Naskapi woman, mother, daughter and friend.” By facilitating community connections, the summer school was also an occasion to create networking. For Indigenous women, solidarity networks are of paramount importance because they strengthen their security, the defense of their interests, the improvement of their conditions, and their collective action as leaders.

|

Stories of Empowerment   53

For some of them, this feeling of connection or reconnection went beyond the link with culture, the land, with each other; it was finding oneself to make room for healing. Maggie experienced the beginning of resilience in herself as she felt that her ideas and voice were heard in an institutional non-Indigenous community where she was no longer invisible and felt comfortable sharing her family history of strong and powerful women. The land-based pedagogy, the holistic approach was also beneficial to Mary, as the course was therapeutic and opened her “eyes, mind, and heart.”

Conclusion

As Denise reminds us in her testimony, the adage “Do nothing for us without us” must be put into practice for the success of community-university programs like the QNW-SPU Summer School. The importance of involving Indigenous partners, through a Memorandum of Understanding or with the help of Elders acting as consultants and project leaders, must be the starting point. It is also necessary to create a safe learning environment where the truths and realities of each are respected and valued. This is how a relationship of trust can begin and the commitment of all those involved makes individual growth possible. Without a doubt, for Ashley the participants empowered one another. None of this would have been possible if the participants had not had the space to reveal their true selves or as Leanne Betasamosake Simpson (2015) put it “recognizing and affirming the light in each other.” Starting community-university programs with a ceremony on the territory is the way to go because, as Denise says, then “learning could begin.” There is no doubt that using the land, circles, ceremonies, and medicines as pedagogies are of prime importance so that the participants could feel at home. To achieve this sense of home, being surrounded by guest speakers who are also role models and inspirational leaders from various First Nations or generations is key. Thus, empowering each other becomes a driving force. The women leaders gave us teachings. The summer school team went into this experience with humility and a readiness to learn. We listened with our hearts, as there were tears, laughs, and moments of reflection. When the teacher becomes the student and when the settler is unsettled yet in a transformative fashion, it can leave you speechless and forever grateful for a learning experience that is not

|

54    Chittspattio et al.

temporary but lifelong. We were unsettled by the individuals’ resilience, strength, and generosity of spirit. The gift we received in being the learners during our time with these leaders is priceless. Notes 1. The authors are listed in alphabetical order. Author contributions, described using the CASRAI CRedIT typology (http://casrai.org/credit), are as follows: Conceptualization, A.M.; methodology, A.M., A.M.H.; writing—original draft preparation, A.M.; writing— review and editing, A.M., A.M.H., N.F., M.C., A.G., M.H., D.L., M.-L. P., A.L.R.G.; resources, A.M., A.L.R.G.; visualization, A.M.; supervision, A.M.; project administration, A.M. 2. https://nac-cna.ca/en/bio/annie-smith-st.-georges; https://www.algonquincollege.com/ kaleidoscope/past-conferences/annie-smith-st-georges-elder/, accessed February 22, 2022. 3. Located in the middle of the Ottawa River, Asinabka “the Place of the Glare Rock,” also known as Victoria Island, is a sacred place for the Anishnabeg (Adam 2013:26).

References Adam, Mohammed. 2013. NCC Treats Spence Camp as Case of Self-Expression. Ottawa Citizen. January 10, 2013. Archibald, Jo-Ann. 2008. Indigenous Storywork: Educating the Heart, Mind, Body, and Spirit. Vancouver: UBC Press. Battiste, Marie-Ann and Jean Barman (eds.). 1995. First Nations Education in Canada: The Circle Unfolds. Vancouver: UBC Press. Crey, Ernie and Suzanne Fournier. 1997. Stolen from Our Embrace: The Abduction of First Nations Children and the Restoration of Aboriginal Communities. Vancouver: Douglas & McIntyre. Dion Stout, Madeleine. 2009. Knowledge Exchange Workshop: Successful Approaches for the Prevention of Aboriginal Family, Final Report. Workshop in Gatineau, Quebec: Public Health Agency of Canada’s Family Violence Prevention Unit. http://www.awotaan.org/ downloads/Aboriginal_FV_Workshop_Final_Report.pdf. Gaudry, Adam and Danielle Lorenz. 2018. Indigenization as Inclusion, Reconciliation, and Decolonization: Navigating the Different Visions for Indigenizing the Canadian Academy. AlterNative: International Journal of Indigenous Peoples 14(3): 218–227. DOI: https://doi. org/10.1177/1177180118785382. Haig-Brown, Celia. 2000. Some Thoughts on Protocol in University/Community Partnerships. NALL

|

Stories of Empowerment   55

Working Paper (ERIC Document Reproduction Service No. ED460281). Ontario: OISE. McLennan, Laura. 2014. Indigenous Education: Empowerment Through Knowledge of Self. Unpublished manuscript, University of British Columbia, Vancouver, B.C. DOI: http:// dx.doi.org/10.14288/1.0055430. Nee-Benham, Maenette K. P. and Joanne E. Cooper. 2000. Indigenous Educational Models for Contemporary Practice: In Our Mother’s Voice. Vancouver: Lawrence Erlbaum. Simpson, Leanne Betasamosake. 2015, October 9. The Misery of Settler Colonialism: Roundtable on Glen Coulthard’s Red Skin, White Masks and Audra Simpson’s Mohawk Interruptus. Leanne Simpson (blog). https://www.leannesimpson.ca/writings/themiseryof-settler-colonialism-roundtable-n-gen-coulthards-red-skin-white-masksand-audrasimpsons-mohawk-interruptus.

Animate Intransitive Consonant Stems in North East Cree Vincent Collette

P

roto-Algonquian had a rule of allomorphy where a verb stem of the conjunct order ending in a vowel required *-t ‘third person singular (animate) [conjunct]’ while those ending in a consonant required *-k ‘third person singular (animate, inanimate) [conjunct]’ (Bloomfield 1946:101‒102).1 This rule is kept in many Algonquian languages including North East Cree as shown with animate intransitive (AI)2 verbs like /nîmi-/ > â=nîmi-t ‘as s/he dances’ (vowel stem) and / tikušin-/ > â=tikuših-k ‘as s/he arrives’ (consonant stem in n). However, with certain AI verb stems ending in t, an epenthetic i is inserted after the allomorph selection of -k as in /iht-/ > â=iht-i-k ‘as s/he fares, does.’ This rule ordering is opaque on the surface (since -k occurs after a vowel) and there is only a handful of AI verb stems ending in t. In this article I trace the comparative history of the archaic t-stem class and claim that in North East Cree and Western Innu, AI stems split not simply into vowel vs. consonant stem classes (as is commonly assumed) but into vowel stems, nasal consonant stems, and t-stem classes as shown by the distribution of the third person conjunct inflection -k. However, while the archaic t-stem class is kept in North East Cree and Western Innu, it has been, or is on the verge of being regularized in other dialects of Cree and Innu by reordering the i epenthesis and allomorphy selection,

|

   57

|

58    Vincent Collette

which leads to the reanalysis of the epenthetic vowel i as the common abstract final -i (e.g., /ihti-/ > e=ihti-t ‘as s/he fares, does’ in South East Cree inland subdialect).

East Cree Verb Structure and Inflection

East Cree is part of a series of connected dialects called the “Cree-Innu-Naskapi-Atikamekw” dialectal continuum. It includes (from West to East): Plains Cree, Woods Cree, Swampy Cree, Moose Cree, East Cree, Atikamekw, Innu, and Naskapi. East Cree is spoken in northwestern Quebec, on the East Coast, and in the hinterland of James Bay and splits into two major dialects. North East Cree (NEC) is spoken in Whapmagoostui, Chisasibi, Wemindji and part of Eastmain, while South East Cree (SEC) has two subdialects; the coastal dialect is spoken in Eastmain (but in some families only), Waskaganish, Washaw Sibi, and Nemaska, while the inland dialect is spoken in Mistissini, Waswanipi, and Ouje-Bougoumou. In this article I am concerned mainly with North East Cree, but I will also briefly discuss the distribution of -k in South East Cree and Western Innu (spoken in Pessamit) also known as Ilnu. The data presented in this article come from my fieldwork in Whapmagoostui and from written literature produced by speakers from Whapmagoostui, Chisasibi and Pessamit for Western Innu (taken from Drapeau and Machado Estevam 2019). The Algonquian verb has a complex internal structure that has been described by various authors (see Bloomfield 1946; Goddard 1990). Verbs are traditionally grouped into four morphosyntactic classes depending on the parameters of transitivity and animacy of the subject or the object. Some verbal lexemes can occur with up to four distinct stems that are distinguished in transitivity and in agreement with the notional absolutive in animacy, namely the subject of intransitive verbs―animate intransitive verbs (AI) and intransitive inanimate verbs (II)―or the object of transitive verbs―transitive animate verbs (TA) or transitive inanimate verbs (TI). In this article I will be concerned with AI verbs only. An Algonquian verb stem (i.e., a verb without its inflection) is internally complex, containing an initial, medial, and final (which can be concrete or abstract in the case of intransitive verbs). If a verb stem has only one element, it will be the initial (as shown in this article); if it has two elements, it will be an initial and one or two finals. Thus, nâsipâhyâw ‘s/he (on plane), it flies down to the water, to the coast’ (AI) comprises an initial nâsipâ- ‘toward the water, coast’, a concrete final -hy- ‘fly’, and an abstract final -â- followed by the inflection -w ‘third person singular

|

Animate Intransitive Consonant Stems   59

[independent order]’. Some verbs can also have a medial but it is always optional, and not required for the well-formedness of the verb stem. In the Algonquianist tradition AI verbs are known to end with an abstract final. On the phonological level, the North East Cree abstract finals comprise vowels (-â, -i, -î, -u), sequences that end in a vowel (-isi, -isî) (i.e., the vowel stem AI class), and those that end in a nasal like -in (i.e., the consonant stem class).3 The distribution of inflectional allomorphs is based on the value of the final segment of an AI stem. The inflections for the conjunct AI stems ending in a vowel (i.e., which stands as the abstract final) are shown in Table 1.4 Table 2 shows that verb stems ending in a nasal consonant like m or n (the latter being part of the abstract final -in) are aspirated into [h] on which is attached the third singular suffix -k. The plural is obtained by adding -wâw after -k but I have also recorded -č(č) ‘third person plural [conjunct]’.5 TI verbs also display this allomorphic rule as presented in Table 3. TABLE 1. AI stems ending in a vowel (abstract final -i-)

3sg -t

/â=nîm-i-t/ sub=dance-ai-3sg.cj

→ â=nîmit ‘as s/he dances’

3pl -č(č)

/â=nîm-i-č-ič/ sub=dance-ai-3-pl.cj

→ â=nîmič(č) ‘as they dance’*

* In NEC the corresponding third person plural conjunct form is -čč ~ -č due to the palatalization of k. More precisely, PA *-t-ik-i ‘third plural animate’ (with *-i ‘indicative mood’ which is lost in Cree) became -čik (as in Moose Cree), then -čič > -čč (often degeminates as -č) in North East Cree due to the palatalization of the second -k and the loss of the short/lax vowel i.

TABLE 2. AI stems ending in a nasal (abstract final -in-)

/â=pimis-in-k/ sub=lie.down-ai-3sg.cj /â=pimis-in-k-wâw/ 3pl -k-wâw sub=lie.down-ai-3-pl.cj 3sg -k

→ â=pimisihk ‘as s/he lies down’ → â=pimisihkwâw ‘as they lie down’ (~ â=pimisihč(č))

TABLE 3. TI stems + (theme sign im-)

3sg -k 3pl -k-wâw

/â=wâpiht-im-k/ sub=see.ti-th-3sg.cj /â=wâpiht-im-k-wâw/ sub=see.ti-th-3-pl.cj

→ â=wâpihtihk ‘as s/he sees it’ → â=wâpihtihkwâw ‘as they see it’ (~ â=wâpihtihč(č))

|

60    Vincent Collette

While this allomorphic rule for nasal stems exists in all dialects of Cree and Innu, it cannot explain all of the North East Cree data under study since -k can also occur with some AI verbs that show i before the inflection. For example, in (1) the independent verb iht-u ‘s/he does, fares thus’ (AI) inflects exactly as if the stem ended in a nasal consonant (as seen above), despite the presence of i. (1) Tân ây=iht-i-k? what sub=do.ai-?-3cj ‘What is he up to?’ (Whapmagoostui, 2012, woman, 64)

Besides a mention in Collette 2014:228–229 the occurrence of -k after a vowel is puzzling and has never been thoroughly documented, although it was likely widespread historically in Cree and Innu dialects. It is not mentioned in the online East Cree grammar (Junker et al. 2021), and Vaillancourt reports it for the Eastmain dialect (1978), but he did not analyze it and deliberately added an aspirate h―which is not present phonetically, phonemically, or historically―by analogy with other TI stems ending in a nasal consonant. Vaillancourt simply remarks that “quelques-uns prennent un “-hk” au-lieu du “-t” à la troisième personne du subjonctif. C’est ainsi qu’on dit tân âihtihk? et non tân âihtit” (1978:195). I show in this article that the presence of -k on these verbs is not a simple exception or an irregularity, but an archaic allomorphic feature of AI verb stems that goes back to Proto-Algonquian.

Proto-Algonquian Stem Classes and the Distribution of *-e-k ‘Third Person Singular [conjunct]’

As stated in the introduction, Proto-Algonquian had a rule of allomorphy where a verb stem of the conjunct order ending in a vowel required *-t ‘third singular [conjunct]’ while those ending in a consonant required *-k ‘third singular (animate, inanimate) [conjunct]’. But there is more to this rule. Comparative data shown in (2‒8) indicate that stems that required an epenthetic vowel before *-k ‘third singular [conjunct]’ ended in *p, *n, *t, and *θ as shown below (all reconstructed forms are from Hewson 1993, unless otherwise indicated; see also Goddard 1979:128 for examples in Unami and Munsee):

|

Animate Intransitive Consonant Stems   61

(2) *nep- ‘die’: Meskwaki êh=nep-e-k-i ‘he died’ (Bloomfield 1927:207); Menominee nep-ɛ-k ‘when he dies’ (Bloomfield 1962:178) (3) *men- ‘drink’: Miami mên-k-i ‘s/he is drinking’ (s.n. 2023); Shawnee men-wa ‘s/he drinks’, mên-e-k-a ‘s/he who drinks’ (Goddard 2015:405, fn. 89) (4) *nôn- ‘nurse’: Meskwaki êh=nôn-e-k-i (Goddard 2015:405, fn. 89) (5) *ent- ‘do so; fare so’: Old Algonquin in r ⟨ent-i-k⟩ ‘s’il fait ainsi’ (‘as he does so’) (Nicolas in Daviault 1994:175) (6) *kemôt- ‘steal’ (7) *meʔθekiθ- ‘be big’: Arapaho affirmative mood benê:séiθ-é-ʔ ‘s/he/it is big’ < PA*mêʔθekiθ-e-k-a ‘(one) who is big’ (Goddard 2015:370) (8) *kwâškwanθ- ‘jump’

Bloomfield (1927:206; see also Michelson 1939:78–79; Bloomfield 1962:178 for Menominee) describes three inflectional patterns for third person singular conjunct AI/TI verbs in his analysis of Meskwaki: -č after a vowel stem (i.e., an abstract final), -e-k-i after consonant stem, and -k-i after a nasal consonant. Pentland (1979:101–102) states the path of development from Proto-Algonquian to Cree as follows: “animate intransitive verbs ending in a consonant before the third person singular *-wa (except most AI verbs in *-nwa) have generalized the epenthetic *-e- which appeared in other environments, thus creating a unique set of words with stem-final *e in Cree.” Thus, comparative data show that in North East Cree the vowel i in ây-ihtik ‘as s/he fares, does so’ is not the abstract final i but an epenthetic i (from PA *e), hence the selection of -k. However, unlike Meskwaki, North East Cree and Western Innu reduced the PA distribution of the allomorph *-e-k ‘third singular [conjunct]’ to AI stems ending in ht and t (with t < PA*t or *θ). All other AI stems ending in a reflex of *p and the short monosyllabic stems ending in *n6 have been reanalyzed as vowel stems (e.g., North East Cree â=nipi-t ‘as s/he/it dies’) or as nasal stems that require an epenthetic i (e.g., Innu e=min-i-t ‘as s/he/it drinks’). Nevertheless, for descriptive

|

62    Vincent Collette

accuracy the verbal conjugation of North East Cree should be distributed into three different AI stem classes (vowel stem, t-stem, and nasal stems), as opposed to two (i.e., vowel vs. consonant stems) as is commonly done in grammatical analysis of Cree and Innu. There are other arguments in favor of this classification besides the presence of an epenthetic vowel i when followed by -k ‘third singular [conjunct]’. First, the epenthetic i does not trigger palatalization of a preceding t to č unlike the abstract final i (e.g., North East Cree /kustât-/ > â=kustâči-t ‘as s/he /it is afraid [conjunct]’). Second, the abstract final i can attach to AI or II stems while the epenthetic i can only attach to AI stems. Third, in the independent order (and likely in the dubitative conjunct) all of these verbs are inflected for the third person singular by adding -w ‘third person singular [independent]’ directly on the stem as in iht-u [ɪhtʊ] ‘s/he /it does, fares’ (but ni-t=iht-i-n ‘I do, fare thus’). In the next sections, I document the presence and extent of this small AI stem class in North East Cree and comment on its distribution in South East Cree (coastal dialect), as well as in Western Innu (spoken in Pessamit). Since in East Cree the AI t-stem class was greatly reduced from what it was historically, I offer a synchronic explanation in terms of rule ordering between allomorphy selection and epenthesis that accounts for the observed dialectal differences.

Allomorphy of AI Conjunct Verbs in North East Cree

As shown in the preceding section there are some remnants of PA conjunct allomorphy in North East Cree, as revealed by the presence of a reflex of PA *-e-k on a handful of AI verbs. Table 4 shows the third-person paradigms for the conjunct indicative (unmarked) and subjunctive (-â) moods (which does not require a conjunct subordinator) for three AI verbs. The presence of the subjunctive variant ihtičâ ‘if s/he fares’ is due to palatalization of k into č when followed by â. Since North East Cree conjunct allomorphy has been considerably modified from PA―as seen by the fact that the inflection -i-k now applies only to ai stems ending in (h)t―a synchronic explanation in terms of rule ordering seems promising. AI stems can be analyzed at an abstract level with final consonants (show simply as C below) in all East Cree dialects including Western Innu. The formation of the conjunct verb ây=ihtik ‘as s/he fares, does so’ is opaque because on the surface the morphophonemic rule of adding -k after consonant stems seems to have been overapplied to context where no consonant is present (and so *â=ihtit would seem

|

Animate Intransitive Consonant Stems   63

TABLE 4. Indicative and subjunctive mood inflections for conjunct AI verbs

3sg ind 3pl ind 3sg subj 3pl subj

VOWEL STEM

C-STEM (T)

C-STEM (N)

/nîm-i-/ ‘dance’ â=nîm-i-t ‘as s/he dances’ â=nîm-i-č(č) ‘as they dance’ nîm-i-č-â ‘if s/he dances’ nîm-i-t-wâw-â ‘if they dance’

/iht-/ ‘fare, do so’ â=iht-i-k ‘as s/he fares’ â=iht-i-k-wâw ‘as they fare’ iht-i-k-â (~ -č-â) ‘if s/he fares’ iht-i-k-wâw-â ‘if they fare’

/pimis-in-/ ‘lie down’ â=pimis-ih-k ‘as s/he lies down’ â=pimis-ih-k-wâw (~ -č(č)) ‘as they lie down’ pimis-ih-k-â ‘if s/he lies down’ pimis-ih-k-wâw-â ‘if they lie down’

TABLE 5. North East Cree: Opaque interaction UNDERLYING FORMS

/â=nîmi-C/

1. Allomorph selection â=nîmi-t 2. i epenthesis --3. Aspiration of n >[h] --[a:ni:mɪt] Surface forms ‘as s/he dances’

/â=iht-C/

/â=pimisin-C/

â=iht-k â=iht-i-k --[a:jhtɪk] ‘as s/he fares’

â=pimisin-k --â=pimisih-k [a:pɪmsɪhk] ‘as s/he lies down’

more natural). I claim that this rule opacity is provoked by the interaction between allomorph selection and i epenthesis between two obstruents like /t__k/. (Note that in the context /n__k/, aspiration of n > [h] occurs and epenthesis is not needed.) As shown in Table 5, in the phonological representation of North East Cree conjunct ai verbs, allomorph selection for the third-person suffix -k occurs before i epenthesis, creating an opaque interaction (see Kiparsky 1973:79). The differences between dialects (as shown below) are due to a different rule ordering between i epenthesis and allomorphy selection (for arguments that allomorphy selection can apply to different intermediate phonological representations and not only to surface forms, see Nevins 2011: section 3.1.1). North East Cree data shows that the distribution of -i-k occurs only with AI stems and one ambitransitive AI verb kičâmutu ‘s/he steals (it)’7 that end in t (< PA *t and *θ), with or without a preceding h. There are no more than thirty of these verbs, some of which are derivatives of ihtu.8 A list of NEC verbs belonging to this archaic AI t-stem class is presented in Table 6.

|

64    Vincent Collette

TABLE 6. North East Cree AI verb stems ending in /t/

mihčât-u ‘it is abundant’ misičit-u ‘s/he/it is big’ sâsâčit-u ‘s/he is barefoot’ tâwit-u ‘s/he opens his own mouth’ wîhčit-u ‘it is tasty’ yâhčit-u ‘s/he is light in weight’

čîht-u ‘it functions’ iht-u ‘s/he fares thus, is doing something’ kičâmut-u ‘s/he steals’ (ambitransitive AI)* kusikut-u ‘s/he is heavy’ kwâškuht-u ‘s/he jumps’ mâmâhčîht-u ‘s/he finds many things to do’

* This lexeme is expressed with simple and reduplicated stems in East Cree. The simple verb is čimatu ‘s/he steals’ while the reduplicated pattern for the second is CiCâ yielding kičâmutu.

Examples (9–12) are taken from written texts by North East Cree authors. I provide the original spelling in Roman orthography and a partial phonemic representation. (9) [. . .] asiniiyiu aa chii tikuskaatihk kiipwaa, aa chii saasaachistik. asinî-yiw

â=čî

tikuskâtih-k â=čî

sâsâčit-i-k.

stone.i-obv.sg sub=past step.on.ti-3cj sub=past barefoot.ai-ep-3cj ‘(. . .) he had stepped on a sharp rock while barefoot.’ (Webb [n.d.]; Chisasibi) (10) Aakw iyaakw kuyaashkuhtik aniyuu miichiwaahpiyuu [. . .] Âkw iyâkw kuyâškuht-i-k

an-iyiw

mîčiwâhp-iyiw (. . .)

then emp ic.jump.ai-ep-3cj ident.i-0’sg tipi.i-obv.sg ‘Then he jumped over the tipi (. . .)’ (Masty 2003:11; Whapmagoostui) (11) Naahaaw niniipuushtaan anit atihkw aa mihchaatik. (. . .) Naashch aa mihchaatich atihkuch. Nâhâw ni-nîpûštâ-n

anit

atihkw-Ø

â=mihčât-i-k. (. . .)

exactly 1-stand.in.relation.ai-sg.idp ident.loc caribou.a-sg sub=plenty.ai-ep-3cj Nâšč â=mihčât-i-č

atihku-č.

Really sub=plenty.ai-ep-3pl.cj caribou.a-pl ‘I will stand right there because the caribou is plenty. (. . .) There were lots of caribous.’ (Masty Sr. 2004:17; Whapmagoostui)

|

Animate Intransitive Consonant Stems   65

(12) Naashch aa chii wiichistik an ushikii aa muwaakiniwit. Nâšč â=čî

wîhčit-i-k

an

ušikiy-Ø

really sub=past tasty.ai-ep-3cj ident.a skin.a-sg â=muw-âkiniwi-t sub=eat.ta-passive.ai-3cj ‘Even the skin was good to eat.’ (Scipio [n.d.])

In order to verify the status of other verbs presented in Table 3, I paid special attention to naturally occurring discourse and I also conducted elicitation sessions with North East Cree speakers (in their fifties and sixties) from Whapmagoostui. I asked them to translate interrogative WH-questions and/or relative clauses (e.g., the one who . . . ) since the inflections of the conjunct order are required in this type of propositions. As shown below, most of the verbs in Table 4 require the inflection -i-k (13–14). Naturally occurring data (15–16) are also consistent with data obtained from written texts and elicitation. (13) an

awân-Ø

kâ=tâwit-i-k.

ident.a person.a-sg rel=open.mouth.ai-ep-3cj ‘that person with the open mouth (e.g., on a picture)’ (Elicitation, Whapmagoostui, man, 55) (14) Âwikun kâ=mâmâhčîht-i-k. focus rel=do.many.things.ai-ep-3cj ‘That’s how many things he finds to do.’ (Elicitation, Whapmagoostui, man, 55) (15) Tânith kâ=wâpim-it

an

kâ=misičit-i-k čisâyâkw-Ø?

Where sub.perf=see.ta-2/3.cj ident.a rel=big.ai-ep-3cj bear.a-sg ‘Where did you see that big bear?’ (2012, Whapmagoostui, man, 22) (16) Ni=čî=wâpihtim-wâ-n kây=iht-i-k. 1=past=see.ti-rl-sg.idp sub.perf=do.ai-3cj ‘I looked at what she was doing.’ (2012, Whapmagoostui, boy, 8)

Moreover, elicitation data also indicate that one ambitransitive AI verb kičâmutu ‘s/he steals (it)’ (17) (i.e., a morphologically intransitive verb with an optional direct object) behaves exactly as expected above.

|

66    Vincent Collette

(17) Čâkwâ-yiw wâhči kičâmut-i-k? why-obv ic.gen steal.ai.-ep-3cj ‘Why does s/he steal (stuff)?’ (Elicitation, Whapmagoostui, woman, 65)

I also found some textual examples of -k on t-stems in the coastal subdialect of South East Cree (coastal) spoken in Eastmain (18–19) and Waskaganish (20–21). (18) Meshuh wii kiipwaa chechiihtik chechiiunaapemit. Mišû

w-îya kiyipwâ če=čîht-i-k

if.in.the.event 3-pro emp

irr=work.ai-ep-3cj

če=čî unâpemi-t. irr=abl have.a.husband.ai-3cj ‘Was she the one who would be able (for whom it will work) to have a husband?’ (Mark-Stewart 1997:16) (19) Čâk kîpwâ âtimisichistik kâkiye an nâpesh. Čâk

kîyipwâ âti

misičit-i-k

kâkiye an

nâpeš-Ø.

Finally emp incep big.ai-ep-3cj emp ident.a boy.a-sg ‘Finally, that boy got bigger.’ (Mark-Stewart 1997:15) (20) Ekw u kâmisichistik iskwew chepichit. Ekw û so

kâ=misičit-i-k

iskwew-Ø

če=pîhče-t [. . .]

prox.a rel=big.ai-ep-3cj woman.a-sg irr=enter.ai-3cj

‘So a big woman will enter (. . .).’ (Blackned 1983:27) (21) [. . .] kiye nipihaat mwehch kaaitaakanut cheitik. Kiye nipih-ât

mwehč kây=it-âkanu-t

and kill.ta-3/3’cj just.like sub.perf=tell.ta-passive.ai-3cj če=iht-i-k irr=do.ai-ep-3cj ‘(. . .) and kill it, just as he had been told to do.’ (Blackned 1983:16)

In South East Cree (inland sub-dialect), the suffix -k ‘third singular [conjunct]’ is not used by speakers of the younger generation after AI t-stems and is considered archaic by some older speakers, as noted by Kevin Brousseau (personal

|

Animate Intransitive Consonant Stems   67

TABLE 7. South East Cree (inland): Transparent interaction UNDERLYING FORMS

/e=nîmi-C/

/e=iht-C/

/e=pimišin-C/

1. i epenthesis 2. Allomorph selection 3. Aspiration of n >[h]

--e=nîm-i-t --[ɛ:ni:mɪt] ‘as s/he dances’

e=iht-i-C e=iht-i-t --[ɛ:jhtɪt] ‘as s/he fares’

--e=pimišin-k e=pimiših-k [ɛ:pɪmʃɪhk] ‘as s/he lies down’

Surface forms

communication, 2017). This small class of AI has been regularized in the last century as seen in (22), an example taken from a text based on oral stories gathered in the early 1980 with speakers born at the turn of the twentieth century. (22) Ewiiche maak wehchi-ihtit kiya an iskweu. Ewîčče

mâk wehči ihti-t

kiya an

iskwew-Ø.

focus.dub coor ic.gen do.ai-3cj emp ident.a woman.a-sg ‘It is probably what that woman did.’ (Cooper 1981:33; Mistissini)

My analysis is that South East Cree (inland) has eliminated the opaque interaction seen in North East Cree and Western Innu in favor of a transparent one. As shown in Table 7 in the phonological representation of rule ordering, I epenthesis takes place before allomorphy selection and i is reanalyzed as part of the stem, and so -k is not selected for AI verbs belonging to the handful of archaic t-stems.

PA *-e-k in Western Innu

The inflection -i-k ‘third singular [conjunct]’ that is due to the same opaque interaction between epenthesis and allomorphy selection seen in NEC above also occurs in historical and modern Western Innu. Father Pierre Laure (1726) who resided in Chicoutimi (a trading post located on the south shore of the Saguenay river) in the early eighteenth century has ⟨ka chachatchitik⟩ kâ=sâsâčitik ‘déchausse, qui est nu-pieds’ (‘one who is barefoot’) (Laure 1726:292). The modern Innu form of this verb has been regularized as kâ=sâsâčitit—with the epenthetic i reanalyzed as the

|

68    Vincent Collette

abstract final i, exactly like e=ihtit ‘as s/he does, fares so’ in South East Cree. As a matter of fact, the allomorphy and inflections of AI/TI stems ending in a nasal are in a state of considerable fluctuation in Eastern Innu and to a certain extent also in South East Cree, since the aspiration of n into h is on the verge of being leveled out entirely in some dialects of East Cree and Innu. For instance, Drapeau (2014:166) indicates that in Innu AI and TI stems ending in n and m respectively are inflected in the conjunct with epenthetic i (apu takušin-i-t ‘as s/he did not arrive’), but not in Western Innu where the standard morphophonemic distribution of inflections -t and -k over vowel stems and nasal stems is kept (i.e., takušin > apu takushih-k ‘as s/he did not arrive’).9 Moreover, in Drapeau and Machado Estevam’s Pessamit Western Innu text collection (2019) one can find the inflection -i-k on AI stems like mišišt- ‘be big’ (23), it- ‘be thus’ (24), kwâškuht- ‘jump’ (25), and išpišt- ‘be a certain size’ (26). (Note that in Pessamit Western Innu the epenthetic vowel is ə because a and i merged into ə.) In the following examples, the first line displays the sentences in standard Innu spelling (which marks the -k inflection as ) while the second line displays a partly phonemic spelling used by Drapeau and Machado Estevam (2019): (23) Apu mishishtit, apishissishu ne auass. Apu məšišt-ə-k əpəšissəšə-w ne əwass-Ø. neg be.big.ai-ep-3cj be.small.ai-3idp med.a child.a-sg ‘He isn’t big, he is a small child.’ (Riverin 2019:S25) (24) Minuat tepishkanit, iapit e aitit [ne kukum..] ne tshishennu, iapit nikamu. Minwat tepəška-l-t again

yapət ekwən

sub.be.night.ii-obv-3cj still focus.i

ey=ayət-ə-k ne sub=redup.be.thus.ai-ep-3cj meda čəšelnu-Ø

yapət nəkəmŭ-w

old.man.a-sg still sing.ai-3idp ‘The next night, there again the old man goes on, he sings just the same.’ (Riverin 2019:S51) (25) Apu tshi pakau-kuashkutit. Apu či kəpe-kwâškŭht-ə-k neg pot disembark-jump.ai-ep-3cj

|

Animate Intransitive Consonant Stems   69

‘She cannot jump overboard.’ (St-Onge 2019:S83) (26) “nutim eshpishtit tshika ashiniu!” iteu. nuhtəm ešpəšt-ə-k čəkəy=əšəni-wə-w entirely ic.sub.be.certain.size.ai-ep-3cj 3irr=rock-dnm.ai-3idp it-e-w say.ta-dir.nl-3idp ‘“[. . .] will turn entirely into stone!” he says to him.’ (St-Onge 2019:S93)

I did not find examples in the Western dialects of Cree or in Naskapi, and it is likely that AI t-stems have been regularized in these dialects, but comparative data show that the archaic allomorphic rule can be reconstructed back to the immediate ancestral language. Unfortunately, the search and description of this archaic allomorphic pattern faces some challenges. First, since the distribution of -k is archaic, Innu translators often regularize it as -t per modern spelling conventions when they encounter it (see (23)–(26) above), so unless a phonemic spelling is used (as with Drapeau and Machado Estevam 2019) this important diachronic information will simply vanish from the records. Second, verb entries in some, if not most, dictionaries of Algonquian languages tend to have only the head word, some derivatives, and other inflectional details that cannot be guessed out. Crucially however, many dictionaries lack sentences altogether and so one cannot glean for usages of conjunct verbs.

Conclusion

This article argues that there are three distinct classes of AI verb stems in North East Cree and Western Innu (and likely Atikamekw as well): vowel stem, t-stem, and n-stem classes. This historical allomorphic distinction of AI stem classes is based on the distribution of the verbal inflection -i-k ‘third person singular [conjunct]’ (with epenthetic i from PA *e) on animate intransitive stems that end in ht or t (< PA *t and *θ). The epenthetic i does not palatalize a preceding consonant and only occurs with AI verbs stems, unlike the abstract final i (PA < *i) which does palatalize and attaches to AI or II verb stems. The AI t-stem class is numerically very small, so it is astonishing that such a remnant of Proto-Algonquian allomorphy was kept for so long. But while such archaism could seem non-economical and burdensome

|

70    Vincent Collette

at first sight, the fact that many of these verbs are highly frequent10 in everyday speech might explain why they have resisted inflectional regularization through rule reordering of allomorphy selection and i epenthesis, as seen in other dialects of Cree-Innu-Naskapi-Atikamekw. Notes 1. Acknowledgments: This paper was read at the 53rd Algonquian Conference (online). I have greatly benefited from discussion with Marie-Odile Junker, †David Pentland, Kevin Brousseau, and an anonymous reviewer who helped clarify the rule ordering of North East Cree. However, I have no reasons to think that they would necessarily endorse my points of view. I am responsible for all errors. 2. Abbreviations used: a = animate noun, abl = abilitative, ai = animate intransitive verb, C = consonant, cj = conjunct order, dir = direct marker, dnm = denominal verb, dub = dubitative mode, emp = emphatic, ep = epenthetic, gen =genitive, i = animate noun, ic = initial change, ident = identificator, idp = independent order, ii = intransitive inanimate, incep = inceptive, ind = indicative mode, irr = irrealis mode, loc = locative, m = mixed (a speech act participant and a third person), med = medial demonstrative, neg = negation, nl = non-local, third person participants only, obv = obviative, perf = perfective aspect, pl = plural, pot = potential, prox = proximal demonstrative, redup = reduplication, rel = relative clause, rl = relational verb, sg = singular, sub = subordinate clause, subj = subjunctive mode, ta = transitive animate verb, ti = transitive inanimate verb, V = vowel, 1 = first person, 3 = animate third person, 3’ = animate third person obviative, the equal sign = separates a proclitic from the stem; 〈 〉 = indicates a word as spelled in an historical source. In this article, I use a modified version of the East Cree Roman orthography where č and š stand for ch and sh respectively; long vowels are

written with a circumflex (e.g., â stands for aa). However, textual examples written in syllabics are presented in the standard Roman spelling. 3. Here is a list of all abstract finals along with a verb example: pimišk-â-w ‘s/he/it paddles’; nîm-i-w ‘s/he dances’; âhč-î-w ‘/s/he/it moves’; nikim-u-w ‘s/he/it sings’; wâp-isi-w ‘it is white’; pimât-isî-w ‘s/he/it lives’; tikus-in-Ø ‘s/he/it arrives’. The online East Cree grammar (Junker et al. 2021) states that there are also stems with the final -iwi- or -uwi(minitûš-iwi-w ‘s/he/it has worms)’ (AI), but this is not so: -iwi- is simply the denominal -iw- (minitûs ‘insect’ > minitûs-iwi ‘to have worms’) followed by the abstract final -i-. The present article also shows that there are AI verbs like iht-u ‘s/he/it fares, does’ that do not

|

Animate Intransitive Consonant Stems   71

have an abstract final. 4. Algonquian verbs have distinct inflectional paradigms known as verbal orders. Person, number, tense/aspect, and modality have different marking depending on the verbal order. The selection of a verbal order aligns with the types of propositions (but not always). Roughly put, the inflections of the independent order are mostly used in independent propositions, as well as in main clauses. Those of the conjunct order are used in dependent clauses and require initial change or a preverb of subordination, as well as in interrogatives sentences. Finally, the inflections of the imperative order are used to express wishes, orders, or commands. 5. The development from PA is as follows: PA *-k-ik-i (with *-i ‘indicative mood’, which is lost in Cree) yields -kik (as in Moose Cree), and -čik (as postulated for the palatalized dialects) becomes -č(č) in North East Cree (with palatalization of the second k and deletion of the short/lax vowel i). 6. Proto-Algonquian had two types of stems ending in *n: a) internally simple monosyllabic verbs of the type CVC required *e epenthesis, followed by *-k (these did not have an abstract final) (e.g., *nep-e-k-e ‘if he dies’); and b) internally complex verbs with an abstract final like -in- required *-k (e.g., *šênkihš-in-k-a ‘one who lies down’ (both PA forms are from Bloomfield 1946:102). This distribution based on phonotactics and internal complexity dispensed the need to memorize to which class an AI stem belong. 7. These verbs are also known as AI + Object verbs; these are morphologically intransitive verbs that can occur with an object or not, hence the label “ambitransitive” proposed by Drapeau (2014:143–144). 8. As noted by Kevin Brousseau (personal communication, 2021) the AI verb čîhtu ‘it works’ fuses the potential preverb čî and the verb stem iht- ‘do, fare’. 9. I gathered Tân eyâyihtik? ‘What is he up to?’ in Mashteuiatsh (a conservative dialect of Western Innu), but more work is needed to evaluate the extent of this AI class in this subdialect. The same rule of allomorphy occurs in Atikamekw ⟨tan e itik⟩ Tân eyittik? ‘What is up with him?’ (Chachai et al. 2021).

10. A few AI verbs seen in this article occur in the Leipzig-Jakarta list, which aims to

document semantic stability (see Tadmor et al. 2010). These are ‘do, fare thus’, ‘big’, ‘heavy’ (North East Cree), and ‘drink’ (Miami, Shawnee). The verb ‘die’ (Meskwaki, Menominee) occurs in the first version of the Swadesh list of a hundred items.

|

72    Vincent Collette

References Blackned, John. 1983. Etipâcimûc Chwân e-wîhtak tipâcimûwiniyu, ed. by Richard Preston. Chisasibi, QC: James Bay Cree Cultural Education Center. Bloomfield, Leonard. 1927. Notes on the Fox Language. International Journal of American Linguistics 4:181‒219. Bloomfield, Leonard. 1946. Algonquian. Linguistic Structures of Native America, ed. by Harry Hoijer, pp. 85‒129. New York: Viking Fund Publication in Anthropology. Bloomfield, Leonard. 1962. The Menomini Language, ed. by Charles Hockett. New Haven and London: Yale University Press. Chachai, Véronique, Marie-Odile Junker, and Nicole Petiquay (eds.) 2021. Dictionnaire Atikamekw. https://dictionnaire.langueatikamekw.ca/#/help. Collette, Vincent. 2014. Description de la morphologie grammaticale du cri de l’Est (dialecte du Nord, Whapmagoostui. PhD thesis, Université Laval. Cooper, Clara. 1981. Iskweuch kaaisi-nachiskahhko ewaapamaausuto. Mistissini, QC: Cree Publications. Daviault, Diane. 1994. L’algonquin au XVIIe siècle. Chicoutimi, QC: Presses de l’Université du Québec à Chicoutimi. Drapeau, Lynn. 2014. Grammaire de la langue innue. Québec: Presses de l’Université du Québec. Drapeau, Lynn, and Adriana Machado Estevam. 2019. General Introduction. International Journal of American Linguistics 85(S1):S1–S16. Goddard, Ives. 1979. Delaware Verbal Morphology. A Descriptive and Comparative Study. New York: Garland Publishing Inc. Goddard, Ives. 1990. Primary and Secondary Stem Derivation in Algonquian. International Journal of American Linguistic 56(4):449–483. Goddard, Ives. 2015. Arapaho Historical Morphology. Anthropological Linguistics 57(4):354– 411. Junker, Marie-Odile, Marguerite MacKenzie, Luci Bobbish-Salt, Alice Duff, Linda Visitor, Ruth Salt, Anna Blacksmith, Patricia Diamond, and Pearl Weistche (eds.). 2018. The Eastern James Bay Cree Dictionary on the Web: English-Cree and Cree-English, French-Cree and Cree-French (Northern and Southern dialects). https://dictionary.eastcree.org/. Hewson, John. 1993. A Computer-Generated Dictionary of Proto-Algonquian. Ottawa: Canadian Ethnology Service paper. Kiparsky, Paul. 1973. Abstractness, Opacity, and Linguistic Changes. Three Dimensions of Linguistic Theory, ed. by Osamu Fujimura, pp. 57‒86. Tokyo: Institute for Advanced Studies in Language.

|

Animate Intransitive Consonant Stems   73

Laure, Michel. 1726. Apparat français-montagnais (texte présenté et annoté par David Cooter). Québec: Presses de l’Université du Québec. Mark-Stewart, Florrie. 1997. Chichipayuu. The Nation, February 14. Masty, Mary Sr. 2004. Akaa kaa nipihtaat. The Nation, March 19 (vol. 11, no. 9). Masty, Sandy. 2003. Kaa Mitaaut. The Nation, May 2 (vol. 10, No. 12). Michelson, Truman. 1939. Contributions to Algonquian Linguistics. International Journal of American Linguistics 10(2/3):75–85. Nevins, Andrew. 2011. Phonologically-Conditioned Allomorph Selection. The Blackwell Companion to Phonology, ed. by Marc van Oostendorp, pp. 2357‒2382. Oxford: John Wiley & Sons, Ltd. Pentland, David H. 1979. Algonquin Historical Phonology. PhD thesis, University of Toronto. Riverin, Alexandre. 2019. Aiasheu. International Journal of American Linguistics 85(S1):S17– S62. Scipio, Maria. [n.d.] Miichim kaa chii kinuwaayihtaakiniwit. Originally published in The Nation. http://formersite.nationnewsarchives.ca St-Onge, Côme. 2019. Sheshiliss. International Journal of American Linguistics 85(S1):S63–S120. s.n. 2023. Miami Dictionary. https://mc.miamioh.edu. Tadmor, Uri, Martin Haspelmath, and Bradley Taylor. 2010. Borrowability and the Notion of Basic Vocabulary. Diachronica 27(2):226–246. Vaillancourt, Louis-Philippe. 1978‒1980. Cours de cri, dialecte québécois (2 volumes). Sillery, QC: Presses de l’Université du Québec. Webb, Elisa. [n.d.] Kaa paayikushit. Originally published in The Nation. http://formersite. nationnewsarchives.ca/ga-baygushit/

sk and sp in Illinois David J. Costa

T

he ancestor of the Algonquian languages, Proto-Algonquian, has long been reconstructed as having numerous word-medial consonant clusters. The consonant clusters of Proto-Algonquian were first comprehensively laid out by Bloomfield (1946:88–90), where he reconstructed thirty-one such clusters, and demonstrated how they appear in four of the daughter languages, i.e., Ojibwe, Meskwaki, Mesominee, and Plains Cree. Over the years only modest revisions to this scheme have been necessary, culminating in the thirty-three clusters recognized by Goddard (1982:25). In the last fifty years there has been continuous progress in determining the reflexes of these clusters across the Algonquian family in languages beyond Bloomfield’s (1946) original sampling, such as Arapaho, Massachusett, and Munsee (Goddard 1974, 1981, and 1982), Cree, Shawnee, Nawathinehena, Narragansett, and Nanticoke (Pentland 1979), and Miami-Illinois (Costa 2003:58–71, 79–87). For the most part the reflexes of the Proto-Algonquian clusters are the same throughout the documented history of the Miami-Illinois language, with the notable exception that certain instances of hk and hp in modern Miami-Illinois (ca. 1795–present) appear in the oldest Illinois records (ca. 1690–1725) as sk and sp. In Costa (2003:82–87) I was not able to systematically account for the appearance

|

   75

|

76    David J. Costa

of sk and sp in Illinois, due to the difficulty of searching for the relevant data in the Jesuit dictionaries. Since that time, the entirety of Largillier’s Illinois–French dictionary (n.d.) and LeBoullenger’s French–Illinois dictionary (n.d.), as well as a substantial portion of Pinet’s French–Illinois dictionary (n.d.), have been entered into a database (https://mc.miamioh.edu/ilda-myaamia/), thus rendering the vast amount of data in these manuscripts more searchable and easily usable than ever before. In this paper I will reconstruct the precise origins of the clusters sk and sp in Illinois, demonstrating which Proto-Algonquian clusters do or do not yield Illinois sk and sp, and discussing the philological difficulties that arise in interpreting the three-hundred-year-old Illinois manuscripts. I will also discuss how the consonant clusters of old Illinois compare with those of modern Miami-Illinois, and how these cluster reflexes compare with those of the other Core Central Algonquian languages. Finally, I will also discuss how the reconstructible clusters of the Core Central languages compare to those of the Algonquian family as a whole.

Miami-Illinois and the Illinois Language

“Miami-Illinois” is the customary name for one of the four branches of the Core Central subgroup of the Algonquian language family. Miami-Illinois consists of various closely related dialects, the best known being Miami, Wea, Peoria, and Kaskaskia. The term “Illinois” is used here to designate the form of the language found in the late seventeenth- and early eighteenth-century manuscripts from the Jesuit missions in what is now the state of Illinois. There is clear evidence that the manuscripts from the Illinois missions represent a mix of dialects, and thus the term Illinois is used here merely to designate a chronological time period of early attestation of the Miami-Illinois language, rather than a single form of speech.

Miami-Illinois and Core Central Algonquian

“Core Central” is the name for a subgroup of languages within the Algonquian family, consisting of Miami-Illinois, Shawnee, Ojibwe-Potawatomi, and Sauk-Meskwaki-Kickapoo. This grouping was first given this name by Goddard (1994:194–195), though he did not classify it as a genetic subgroup. Core Central was later put on firmer ground as a genetic grouping by Rhodes (2021), who further examined the

|

sk and sp in Illinois   77

morphological and phonological innovations characterizing the languages of this group. The Miami-Illinois reflexes of the Proto-Algonquian consonant clusters firmly place the language in the Core Central subgroup. Specifically, Miami-Illinois shares with the other Core Central languages the highly distinctive shift whereby the four Proto-Algonquian clusters *hr, *hθ, *ʔr, and *ʔθ all merge into the same reflex each language has for Proto-Algonquian *hs—in the case of Miami-Illinois, they all appear as hs (see Costa 2003:79–81; Goddard 1994:194–195; Rhodes 2021:306–307).1 Proto-Algonquian Consonant Clusters in Miami-Illinois

The reflexes of the Proto-Algonquian stop clusters in modern Miami-Illinois are very straightforward. The Proto-Algonquian nasal + stop clusters are retained in Miami-Illinois in all time periods, as in Ojibwe and Delaware (see Costa 2003:68). The remaining Proto-Algonquian stop, sibilant or affricate clusters, i.e., those with PA *p, *t, *k, *č, *s, or *š as second members, all become simple preaspirates in Miami-Illinois, i.e., hp, ht, hk, hc, hs, and hš. For example, uniquely in Algonquian, the six Proto-Algonquian clusters *hk, *šk, *θk, *sk (*xk), *čk, and *rk (*çk) all merge as simple hk in modern Miami-Illinois from the late eighteenth century onward (Costa 2003:61).

SC Clusters in Old Illinois

The realization of the Proto-Algonquian stop clusters in the oldest records of Miami-Illinois, namely the late seventeenth- and early eighteenth-century records of Illinois, is more complex than this, in that many of the hk-clusters of modern Miami-Illinois appear in Illinois as sk (variously written as either ‹sk› or ‹sc›), with numerous doublets of words transcribed with both sk and hk.2 For example, for ‘prairie’, LeBoullenger writes both ‹masc8te8i› and ‹mac8te8i› (phonemic maskoteewi and mahkoteewi; < PA *maškote·wi), though the reflex with sk is much more common. Indeed, it looks as though any word pronounceable with sk in Illinois could also be pronounced with hk as well. Likewise, a cluster sp is also found in Illinois,3 similarly alternating with hp; for example, for ‘it is tall’, the Largillier dictionary gives both ‹ispiki› and ‹ippiki› (phonemic iispiiki and iihpiiki; < PA stem

|

78    David J. Costa

*išpya·-). The obvious explanation for the variability in Illinois between sk and hk and between sp and hp is that in Illinois three hundred years ago, a sound change was underway, a process of debuccalization shifting sC-clusters to simple hC, the variability being perhaps due to dialect or the age of the speaker. At some point during the documentational gap of the Miami-Illinois language from the late 1720s to the 1790s, these s-clusters disappeared entirely: sk and sp are completely gone from all Miami-Illinois records by the 1790s, replaced by simple hk and hp.

Philological Problems

A major problem with determining the extent of sk in Illinois is that not uncommonly the French missionaries wrote ‹sk› or ‹sc› in cases when it is virtually certain that the cluster was actually hk. The most likely reason for this is simply that the missionaries spoke a language with no phoneme h, much less clusters of h + consonant, and so they consequently often either heard hk as [sk], or perhaps simply wrote ‹s› as the closest analog to preconsonantal h their language had. Confirming that this writing of h as ‹s› is no more than an auditory misperception is the fact that it is particularly common after front vowels, especially i. Moreover, the simple fact that preaspiration is not marked at all the great majority of the time in the Illinois records shows that hC sequences were a challenge for the missionaries to hear under the best of circumstances.4 Proto-Algonquian *hk in Illinois

The most persuasive examples of the missionaries writing ‹sk› or ‹sc› for what is in fact hk are found in words with clusters deriving from Proto-Algonquian *hk. Proto-Algonquian *hk does not appear as sk or šk in any language; it appears as hk in Miami-Illinois and all the other “Central Algonquian” languages, and as hk in most Eastern Algonquian languages save Delaware-Mahican, where it is further reduced to simple h. Nevertheless, twice Largillier writes reflexes of PA *ihkwa ‘louse’ (Goddard 2015:400) as though it had sk, in the forms ‹nitescama› ‘my louse’ and ‹atescame8a› ‘he has lice’ (phonemic nitehkama and atehkameewa, respectively). Given that otherwise PA *hk always gives hk in Illinois, the fact that ‘louse’ is written with plain ‹c› or ‹k› dozens of other times, and the fact that the cluster does not appear with a sibilant in any other language, it is extremely likely that the ‹s› in

|

sk and sp in Illinois   79

Largillier’s ‹nitescama› simply represents an attempt to render a preaspirate after a front vowel, perhaps something along the lines of a voiceless palatal fricative [ç].5 Either way, Proto-Algonquian *hk is highly unlikely to be one of the sources of Illinois sk.6 Proto-Algonquian *rk in Illinois

Similarly, Proto-Algonquian *rk (*çk) only ever appears as Illinois sk in a handful of forms deriving from PA *merkw- ‘blood, red’. The three examples I have found are: Largillier’s ‹amesc8ma› ‘one’s vein’ (ostensible *ameskoma), alongside more common hk-forms like nimehkoma ‘my vein’ (Gr/LB ‹nimec8ma›; see Costa 2019:56, 66); and LeBoullenger’s forms ‹misc8achiki kipicat8i› ‘metail rouge’ (ostensible *miskwaahšiki kiipihkatwi) and ‹nimisc8ng8e› ‘je me peins de rouge’ (ostensible *nimeskoonkwee), alongside far more common forms reflecting mehkw- or mihkw-. However, no other etyma from PA *rk ever show Illinois sk.7 Most likely, Illinois forms like ‹amesc8ma›, ‹misc8achiki› and ‹nimisc8ng8e› are the same as ‘louse’ above, examples of the missionaries writing hk with ‹sc› after front vowels due to the absence of h or hk in their own language.8 Proto-Algonquian *čk in Illinois

A Proto-Algonquian *k-cluster that never once gives Illinois sk is *čk. Among the etyma with Proto-Algonquian *čk reflected in Illinois as hk are nirakahkwa ‘my palate’ (Gr ‹niragac8a›, P ‹nirakag8a›, LB ‹niragac8o›; < PA *narakačkwi [Bloomfield 1946:89]); toohk- ‘move, touch’ as in toohkiiwa ‘he moves’ (LB ‹tohki8o›; < PA initial *to·čk- ‘wake, touch’ [Goddard 2020, personal communication]); and pahk- ‘remove a piece from’ as in nipahkamaa ‘I bite off a piece from him’ (Gr ‹nipaccama›; < PA *pačk-).9 Proto-Algonquian *θk in Illinois

Probably the most common Proto-Algonquian *k-cluster that does not yield Illinois sk is *θk. The main example I have found in the Illinois corpus of Proto-Algonquian *θk being written as though it were Illinois sk is mikihkwa ‘old woman’, with a final deriving from PA *-eθkw ‘woman’. Pinet variously writes this as ‹mikisc8a›, ‹mikisk8a› and ‹mikik8a›, Largillier writes it as ‹mikisc8a›, and LeBoullenger writes

|

80    David J. Costa

it as ‹mikisc8o›. However, no other Illinois derivatives of the Proto-Algonquian morpheme for ‘woman’ are ever transcribed with ‹sc› or ‹sk›, including words such as ahkweehsa ‘young unmarried woman’ (Gr/LB ‹ac8essa›, P ‹ak8essa›), ahkweehsimwa ‘female animal’ (Gr ‹ac8essem8a›, P/LB ‹ac8essim8a›), and niwiitikihkwa ‘my brother-in-law or sister-in-law, when married to two sisters or two brothers’ (Gr ‹ni8itikic8a›, LB ni8itikic8o›; < PA *ni·tekeθkwa ‘my sister (woman speaking)’). Given that ‘old woman’ is the only known Illinois word transcribed with sk from PA *θk, and that there are numerous examples of other Illinois words given with simple hk from PA *θk,10 it seems extremely likely that the ‹s› in the transcriptions of mikihkwa is spurious, and that the regular reflex of PA *θk in old Illinois was actually hk.

Proto-Algonquian Clusters Giving sk in Illinois

Of the six Proto-Algonquian *k-clusters that give hk in modern Miami-Illinois, I have argued that the four clusters discussed above, namely *hk, *čk, *rk, and *θk, all regularly give hk in Illinois, not sk. This leaves two Proto-Algonquian clusters that do in fact give Illinois sk, namely *sk (*xk) and *šk. Proto-Algonquian *sk Giving Illinois sk

I have so far found a total of eight Illinois etyma with Proto-Algonquian *sk (*xk) where this cluster appears as Illinois sk, more examples than seen with any of the other PA clusters discussed so far.11 Further confirming that these sk-clusters are genuine is the fact that about half of them occur after back vowels, an environment in which spurious sk-clusters never occur with *hk, *čk, *rk, or *θk. Table 1 contains the clear examples I have found of PA *sk giving Illinois sk.12

|

sk and sp in Illinois   81

TABLE 1. Illinois sk from Proto-Algonquian *sk

ILLINOIS WORD

GLOSS

TRANSCRIPTIONS

rooskanwi rooskimini

‘it is soft’ ‘flour’

rooskahkiwi

‘it is soft ground’

niskinši askinšimaawi

‘my forehead’ ‘one’s forehead’

aciskwi

‘stump’

aski mahtaski, loc. mahtaskinki

‘field’

LB ‹r8scan8i› ‘mol’ Gr ‹r8skimini› ‘farine’ LB ‹roscaki8i› ‘terre molle’ Gr ‹niskingi› Gr ‹askingima8i› LB ‹atchisc8i›, Gr ‹atchic8i› LB ‹aski› LB ‹mataski› ‘down’, R ‹mattaskinghi› Gr ‹rican8i›, LB ‹rescan8i› Gr ‹rabiscagane, LB ‹rapiscagane›, P ‹rapiskagane› LB ‹miroscami8i, ‹miroccami8i› Gr ‹makiski8i›, ‹makiki8i›, LB/P ‹makiski8i›

‘down’

riskanwi

‘it is dusk, almost night’

raapiskaakani

‘necklace’

mirooskamiwi

‘it is spring’

mahkiskiwi

‘herb, medicine’

PA FORM

*ro·sk- ‘soft’

*-skenθ- ‘forehead’ *očiskyi *askyi ‘land’ *reskanwi ‘evening falls’* *θa·pisk*mero·skami(·)*maškiskyiwi

* This reconstruction is courtesy of Ives Goddard (personal communication, 2006), and literally originally meant ‘things become invisible’; cf. Shawnee lehkaki ‘last night’ (Gatschet 1878–1879) and lhkanwi ‘it is moderating in the evening after a stormy day’ (Voegelin 1938–1940), Kickapoo neehkaki ‘yesterday’, Menominee nεhka·n ‘it is evening’ and Massachusett ‹nuhkan› ‘night’ (Mayhew 1709, Psalms 74:16). The *r in the PA initial *resk- is confirmed by Moose Cree liskwenam ‘his footprints grow faint’.

|

82    David J. Costa

Proto-Algonquian *šk Giving Illinois sk

The Proto-Algonquian cluster by far most often reflected as Illinois sk is PA *šk. Most Illinois words deriving from Proto-Algonquian etyma with this cluster appear at least optionally with sk, as seen in Table 2.13

TABLE 2. Illinois sk from Proto-Algonquian *šk PA OR COMPARATIVE FORM

ILLINOIS WORD

GLOSS

TRANSCRIPTIONS

apiiskwa ašiiskiwi

‘nighthawk’* ‘earth, land’

*pi·škwa *ašišk(y)iwi ‘mud, dirt’

(a)skimotayi

‘bag, pocket’

P ‹apisk8a›, LB ‹apisc8a› Gr/LB ‹achiski8i› Gr ‹askim8taï›, LB ‹skim8tae›, ‹askim8taie›, P ‹skim8tai›

ninteskonaan

‘I set it aside, leave part of it’

Gr/LB ‹nitesc8nan›

*iškw- ‘left over, remaining’

ninteskaraamwi

‘I drown’

eeskantia

‘chisel’

iskipakinwi, iiskipakinki

‘it is blue/green’

iiskihtaminki

‘watermelon’

askinwi, iiskinki

‘it is raw’

iskoteewi, skóteewi

‘fire’

iskwaanteemi

‘door’

ninkaaskahwaa kaaskihkioni ninkiiskišaan

‘I clean him (a hide), dress game’ ‘razor’ ‘I cut it, notch it’

Gr ‹nitescaram8i›, ‹nitescaram8i›, P ‹nite'skaram8i› Gr/LB ‹escantia›, P ‹eskantia› P ‹iskipakin8i›, LB ‹eskipakin8i›, ‹eskipakinghi› Gr/LB ‹iskitamenghi›, P ‹eskitamenghi› LB/P ‹askin8i›, Gr/P ‹iskinghi› Gr/P ‹isc8te8i›, LB ‹isc8te8i›, ‹sc8te8i›, R ‹isk8te8i› Gr/LB ‹esc8antemi›, ‹isc8antemi›, ‹ac8antemi› Gr ‹nicascah8a› Gr/LB ‹caskiki8ni› Gr ‹nikiskichan›

*maškimotayi

*iškaθa·mo-, < *išk- ‘tired’† *e·škan- ‘horn/ice chisel’‡ *aškipak- ‘blue/green’

*ašk- ‘raw’

*iškote·wi *iškwa·nte·mi *ka·šk- ‘shave, scrape’ *ki·šk- ‘cut, sever, break’

|

sk and sp in Illinois   83 PA OR COMPARATIVE FORM

ILLINOIS WORD

GLOSS

TRANSCRIPTIONS

ninkiiyaaskwee maskiikwi maskooteewi

‘I am crazy, foolish’ ‘swamp, marsh’ ‘prairie’

nimooskinatoo

‘I fill it’

Gr ‹nikiasc8e› Gr/LB/P ‹maskig8i› Gr/LB/P ‹masc8te8i› Gr ‹nim8skinat8›, P ‹nim8skinat8n›

waapiski-

‘white’ (initial and prenoun)

Gr ‹8abiski›, LB ‹8apiski›§ *wa·pešk-

iskaapeewa, iskaapia wiskirwa weeskiniikita wiiskapanwi, weeskapanki

‘chief’s attendant’

Gr ‹escapia›, ‹iscape8a›

‘he is young’ ‘she is young’

wiiskapeepiikwi

‘salt, sheep sorrel’

nintaatoskaan

taatoskaawi

‘I tear it open, burst it’ ‘he tears it open, bursts it’ ‘I tear him open, burst him’ ‘it bursts open, tears’

Gr ‹8iskir8a›, P ‹8skir8a› Gr/LB ‹8eskinikita› Gr ‹8iscaban8i›, ‹8escabanki› Gr ‹8iscapepic8i›, P ‹8iskapepik8i›, LB ‹8iscapepic8i› P ‹nitat8scan›

Gr/LB ‹tat8sca8i›

niraankiskaa

‘I run fast’

P/LB ‹niranghisca›

raankiskaata

‘he runs fast’

Gr ‹ranghiscata›

taatoskamwa nintaatoskawaa

‘it is sweet’

LB ‹tat8scam8› LB ‹nitat8sca8a›

Oj giiwaškwe ‘he is dizzy’ *maškye·kwi *maškote·wi *mo·šk- ‘full’

*ošk- ‘young’

*wi·škop-

*-eškam ‘by foot or body (TI)’ *-eškaw ‘by foot or body (TA)’ *-eška· (II) ‘by foot or body’ *-eška· (AI) ‘by foot or body’

* Illinois also has an alternate apeehkwa, as exemplified by Gr ‹apec8a› and the Pinet pl. ‹apek8aki›, which is also the only form attested in modern Miami-Illinois (cf. D ‹ă-pắk-wah›). Note also the possibly related apeeskoohsia ‘Turkey vulture’ (Gr ‹apesc8sia›, P ‹apesk8chia›, LB ‹apec8ssia›). † For the initial, see Goddard (2002:61); for the final, PA *-aθa·mo- ‘breathe’, I thank Ives Goddard (personal communication, 2020). ‡ See Rhodes (2021:323). § For examples of this, note waapiski-moohseewa ‘grub worm’ (Gr ‹8abiskim8sse8a›) and waapiskinihkiiwa ‘he has white hands’ (LB ‹8apiskineki8a›).

Proto-Algonquian *šp Giving Illinois sp

In addition to sk-clusters, old Illinois also had a cluster sp. Of the six *p-clusters of Proto-Algonquian (*hp, *šp,*čp, *sp,*θp, and *mp), *mp is unchanged, and three (*hp, *sp, and *θp) only ever give hp (no Illinois word deriving from PA *čp has been found).14 Only Proto-Algonquian *šp gives Illinois sp. However, this cluster is nowhere near as common in Illinois as sk: all the examples found so far derive from only two initials: Proto-Algonquian *išp- ‘high, up’ and *ki·špw- ‘sated, full (of food)’.

|

84    David J. Costa

TABLE 3. Illinois sp from Proto-Algonquian *šp ILLINOIS WORD

GLOSS

nikiispwi ‘I am satisfied, am full’ nikiispweerimaa ‘I have had enough of him’ ispeminki

‘up, above’

ispesiwa

‘he is tall’

ispiiwi

‘it is high, tall’

TRANSCRIPTIONS

PA FORM

Gr/LB ‹nikisp8i› Gr ‹nikisp8erima› Gr/LB ‹espeminghi›, ‹speminghi› Gr/P ‹ispessi8a›, P/LB ‹ispessi8a› Gr ‹ispi8i›, P ‹espi8i›

*ki·špw‘sated, full (of food)’ *išp‘high, up’

Both of these almost always appear in Illinois with sp, rarely with hp, as seen in Table 3. (Again, by the late 1700s, all these words appear with hp 100% of the time.)

Unexpected k-Clusters in Illinois

A few Illinois nouns with anomalous k-cluster reflexes remain unexplained by the above sound correspondences. A prominent example of this is Illinois ahkiskwa~ahkiskwi ‘drum, kettle’. 15 Given that this word reconstructs as PA *askehkwa, one might expect Illinois forms like *askihkwa~*askihkwi, which in fact never occur. Alongside several Illinois attestations of this word with simple ahkihkw- (Gr ‹akic8i›, P ‹akik8a›; also its modern form), it appears numerous times (especially in LeBoullenger’s dictionary) as ahkiskwa~ahkiskwi (LB ‹akisc8o›, ‹akisc8i›, P ‹akisc8a›). Given the overwhelming evidence that PA *hk does not give Illinois sk, the Illinois word seems to have flipped its clusters from an earlier (unattested) *askihkwa~*askihkwi. Thus these forms can perhaps be explained by positing a rule whereby word-medial sequences of older *skVhk flip this to hkVsk, by a process of metathesis. Another Illinois word that might well have undergone this same metathesis process is mahkiskiwi ‘medicine, herb’ (Gr ‹makiski8i›, ‹makiki8i›, LB/P ‹makiski8i›), which comes from a PA form *maškiskyiwi ‘grass’ (Goddard 2015:355). Given that PA *šk usually appears in Illinois as sk, while PA *sk gives Illinois sk much

|

sk and sp in Illinois   85

less consistently, it looks as though this word might have been earlier Illinois *maskihkiwi, which later metathesized to attested mahkiskiwi. As an alternate explanation, it is possible that mahkiskiwi in fact was earlier Illinois *maskiskiwi, and later underwent a dissimilation process whereby words with two syllable-adjacent *sk-clusters change the first cluster to hk. This latter scenario is supported by the fact that it is impossible to find any Illinois nouns with two sk-clusters separated by a single syllable. Another word with anomalous consonant clusters that could be explained this way is Illinois ahkiskiwi ‘land’ (Gr/P/R ‹akiski8i›); from a probable Proto-Algonquian form *aškiškyiwi,16 one might expect Illinois *askiskiwi, but in fact this word is never attested with sk for its first cluster. With this word and mahkiskiwi, dissimilation acting on words with two *sk-clusters seems indicated, with the s always being retained on the second k-cluster.

Conclusion: Core Central within Algonquian

At this point it is worth asking what the consonant reflexes of Illinois say about its place within Core Central Algonquian, and whether they shed any light on what these clusters were at the level of Proto-Core Central Algonquian. p-Clusters in Core Central Algonquian

Proto-Algonquian *p-clusters behave the same way in Illinois as in Ojibwe, in that they all merge into hp except for Proto-Algonquian *mp and *šp. *mp is still retained as such in both Ojibwe and Miami-Illinois, while *šp is preserved as šp in Ojibwe and as sp in Illinois. Every other *p-cluster appears as simple hp in all the other Core Central languages (fortis pp in Potawatomi and southern Ojibwe dialects). So one can posit that Proto-Core Central merged Proto-Algonquian *hp, *sp, *θp, and *čp, while *šp and *mp were still kept separate. k-Clusters in Core Central Algonquian

The reflexes of the PA *k-clusters in Core Central are considerably more complicated. There were six non-nasal + *k-clusters in Proto-Algonquian. Goddard (1994:195) has shown that relics of diminutive sound symbolism in Meskwaki

|

86    David J. Costa

prove that pre-Sauk/Meskwaki/Kickapoo once differentiated reflexes of *θk from reflexes of *hk, indicating that they were not yet merged at the Proto-Core Central level. Similarly, Ojibwe-Potawatomi uniquely keeps reflexes of Proto-Algonquian *rk distinct from all other clusters (as sk). Ojibwe also crucially shows that Proto-Algonquian *čk was separate at the Proto-Core Central level from *hk, in variously attesting *čk as kk, sk, and šk (Goddard 1997:31). If one does not include Illinois, all this suggests that at the Proto-Core Central level, of the seven relevant clusters, only Proto-Algonquian *hk and *sk had been merged. However, the fact that Illinois sometimes retains *sk as sk shows that PA *hk and *sk cannot have been merged at the Proto-Core Central level either. Taken together, this indicates that none of the PA *k-clusters were merged yet in Proto-Core Central, all seven of the PA *k-clusters being kept distinct. From the perspective of Algonquian as a whole, this is not too surprising, since modern-day Delaware has five distinct reflexes of the seven PA *k-clusters, merging PA *sk, *čk, and *rk into hk (see Goddard 1982:25). Consonant Clusters in Core Central Compared to the Rest of Algonquian

The preservation of all seven *k-clusters in Proto-Core Central stands in sharp contrast with the modern Core Central descendant languages, which are all characterized by an extensive merger of Proto-Algonquian clusters not matched elsewhere in Algonquian. In addition to the aforementioned merger of the four Proto-Algonquian glottal + liquid (*hr, *hθ, *ʔr, and *ʔθ) clusters into Proto-Core Central *hs, all the Core Central languages share a complete merger of Proto-Algonquian *hC and *ʔC clusters not shared by the nearest languages such as Menominee or Cree (Bloomfield 1946:88–90). Additionally, of the seven original Proto-Algonquian *k-clusters, modern Ojibwe preserves only a four-way contrast (nk, kk, sk, and šk), Potawatomi a three-way contrast (kk, sk, and šk), Illinois a three-way contrast (hk, sk, nk), modern Sauk-Meskwaki-Kickapoo and Shawnee only a two-way contrast (hk and šk), and modern Miami-Illinois a two-way contrast as well (hk and nk). The realization in modern Miami-Illinois of all the Proto-Algonquian stop and affricate clusters as either preaspirated or prenasalized represents a considerable simplification, since no other Algonquian language merges the six Proto-Algonquian *k-clusters that modern Miami-Illinois does, nor does any other Central or Eastern Algonquian language completely eliminate all sibilant + stop clusters. Nevertheless, the fact that the widespread collapsing of Proto-Algonquian *k-clusters found in

|

sk and sp in Illinois   87

the Core Central languages cannot be attributed to Proto-Core Central testifies to an extended period of mutual influence among the Core Central languages, continuing long after its breakup into its constituent languages and dialect groups. Notes 1. Proto-Algonquian reconstructions are given here in the system laid out in Goddard (1994). Thus, older PA *l is written as *r, *we is reconstructed as *o, and word-initial *i is written for older *e. Proto-Algonquian *sk and *rk replace Bloomfield’s (1946) *xk and *çk, respectively. Reconstructions from earlier sources have been converted to this transcription system. The following abbreviations are used in this paper: AI = animate intransitive, D = Jacob Dunn’s (n.d.) notes on Miami-Illinois, Gr = the Gravier (n.d.) Illinois-French dictionary, II = inanimate intransitive, LB = LeBoullenger’s (n.d.) French–Illinois dictionary, Mc = Truman Michelson’s (1928) Kickapoo notes, Oj = Ojibwe, P = Pinet’s (n.d.) French–Illinois dictionary, PA = Proto-Algonquian, R = Râle’s (1908) late-17th century Illinois Prayer Book, TA = transitive animate, TI = transitive inanimate, Tr = Trowbridge’s (n.d.) Kickapoo notes. 2. Moreover, sk is not especially rare in Illinois; a search of the ILDA database (https:// mc.miamioh.edu/ilda-myaamia/) for all examples of ‹sc› and ‹sk› in the LeBoullenger and Largillier dictionaries returns approximately 2,750 hits, at least half of them forms reflecting the finals PA *-eškaw (TA), *-eškam (TI) and *-eška· (AI and II) ‘action by body, foot’. 3. Illinois sp is much less common than sk; a search of the ILDA database for all examples of ‹sp› in the LeBoullenger and Largillier dictionaries returns 164 hits. This is due to the fact that p-clusters were much less common in Proto-Algonquian than k-clusters. 4. There are a handful of examples in the Illinois materials of the missionaries writing ‹s› for h before consonants other than k or p. For example, in one place LeBoullenger wrote ‹st› in ‹mistecapinte› for mihtekwaapinti ‘bowstring’ and similarly once wrote ‹mistêcamina› for mihtekamina ‘acorn’, both of which derive from PA *meʔtekw- ‘wood, tree’, with a PA *ʔt-cluster that gives ht in all stages of Miami-Illinois. In another place, LeBoullenger wrote ‘bowstring’ as ‹michtecopinte›, using ‹ch› to capture the preaspiration and voicelessness preceding the t. Elsewhere he wrote this word as ‹mitteg8pinte›, with the ‹tt› that was by far the most common means the missionaries had of explicitly writing ht.

|

88    David J. Costa

5. Among the many Proto-Algonquian etyma with *hk which are never written in Illinois with ‹sk› or ‹sc› are Illinois tahk- ‘cold’, as in tahkihsinwi ‘it is cold to the touch’ (< PA *tahk-); -iihkaan- ‘friend’ (< PA *-i·hka·n- [Ives Goddard, personal communication, 2020]); eemihkwaani ‘squash, pumpkin’ (< PA *e·mehkwa·na ‘spoon’); kiihkiihkwa ‘falcon’ (< PA *ki·hki·hkwa); -oohkom- ‘grandmother’ (< PA *-o·hkom-); and wiihkam- ‘summon him, call him’ (< PA *wi·hkom-). 6. The only other word where the missionaries write ostensible sk for *hk, Illinois ‹akisc8a› ‘drum, kettle’ (< PA *askehkwa), will be discussed below. 7. Among the Illinois forms always giving hk for PA *rk are noohkw- ‘lick’, as in Illinois noohkwaatam- ‘lick it’ (< PA *no·rkw-); ahk- ‘evaporate’, as in ahkiteewi ‘it evaporates, the water dries out, boils out’ (< PA *irk-); and ahkaawi ‘(his) tenderloin’ (< PA stem *-rka·w-). 8. It is also possible that these forms are influenced by the missionaries’ awareness of cognate Ojibwe forms such as miskwi ‘blood’. (Compare Preston’s 1796 modern Miami form ‹meque› ‘blood’, phonemic mihkwi.) 9. For this initial, see Pentland (1979:138, 153). Admittedly, *čk is one of the more elusive clusters in the Core Central languages, appearing as hk in Sauk-Meskwaki-Kickapoo and Shawnee, but variously as kk, sk and šk in Ojibwe (see Goddard 1997:31), to the point where Pentland (1979:384) suggests that the reflexes of *čk are so irregular across the different languages that it “probably [does] not represent a genuine set of correspondences”. 10. Some examples of Illinois words and morphemes always giving hk for PA *θk are ahkanimini ‘seed’ (< PA *waθkanimini); ahkici ‘on top of’ (< PA *waθkici); amehkwa ‘beaver’ (< PA *ameθkwa); arakiihkwi ‘tree bark’ (< PA *warake·θkwi); -hkan- ‘bone, leg’ (< PA *-θkan-); -hkon- ‘liver’ (< PA *-θkon-); -hkweekan- ‘neck’ (< PA *-θkwe·kankan-); -hpehkwan- ‘rib, back’ (< PA *-speθkwan-); mahkwa ‘bear’ (< PA *maθkwa); mehk- ‘find’ (< PA *meθk-); nahkom- ‘answer him’ (< PA *naθkom-); -nehk- ‘hand’ (< PA *-neθk-); niimihki ‘fort, palisade’ (metathesized < PA *mye·neθki); -pihkway- ‘bladder’ (< PA *-piθkway-); and -ehk- ‘dependent order marker for 2nd person singular subject with a 3rd person object’ (< PA *-eθk-). 11. Some examples of Illinois words and morphemes always giving hk for PA *sk are: aarahkwatwi ‘it is a cloud, cloudy’ (< PA *a·raskwatwi); anaahkii- ‘lay out a mat or rug’ (< PA *ana·skye·-); -aahkw ‘tree, wood’ (final)’ (< PA *-a·skw); mahkahkwi ‘barrel, cask’ (< PA *maskaskwi ‘nonearthen container’); mahkateeweewi ‘it is black’ (< PA *maskate·‘black’); mahkisini ‘shoe’ (< PA *maskeseni); mahsakahkwa ‘badger’ (< PA *meʔθakaskwa); miikaahkii- ‘fight’ (< PA *mi·ka·skye·-); nihkahkwani ‘my shin’ (< PA *naskaθkwani); pahkonam- ‘pull it out, extract it’ (< PA initial *paskw-); páhkia ‘ruffed grouse’ (< PA

|

sk and sp in Illinois   89

*(pas)paskiwa); pihkwaahkiwi ‘hill, range, knoll’ (< PA initial *peskw- ‘lump’); -hkaat‘foot’ (< PA *-ska·t-); -hkaš- ‘nail’ (< PA *-skanšy-); ahkici ‘bird’s tail’ (< PA *oskečy‘pipestem’/‘bird’s tail’); -hkiwan- ‘my nose’ (< PA *-skiwan-); and tipehkiiwi ‘it is night’ (cf. PA *tepeskwi ‘night’). LeBoullenger writes ‹sc› for expected hk once for the PA initial *kesk- ‘know, denote’ in the form ‹kiscaham8o› ‘he marks it’, but since this is the only example of this very common Illinois initial ever being transcribed with ostensible sk, the ‹s› in this form is presumably spurious. 12. Virtually all the forms in this and the following table can also appear with hk in place of sk, though to save space original transcriptions indicating hk are not given here. 13. There are not many words where Illinois sk never appears as the reflex of ProtoAlgonquian *šk; a few I have found are -iihkwan- ‘elbow’ (< PA *-i·škwan-); -hkiinšikw‘eye’ (< PA *-ški·nšekw-); šoohkwaham- ‘drag it’ (< PA initial *šo·škw- ‘smooth, slick’); and ahsahkwa ‘muskrat’ (< PA *we(ʔ)šaškwa). 14. Examples of stems from PA *hp, *sp and *θp giving only Illinois hp are seen in paahpi‘play, joke’ (< PA initial *pa·hp-); kahpakinwi ‘it is thick, coarse’ (< PA initial *kespak-); ahpena ‘potato’ (< PA *ospenya ‘tuber’); -hpikay- ‘rib’ (< PA *-spikay-); ahpwaa- ‘smoke a pipe’ (< PA *ospwa·-); and kihpir- ‘tie him’ (< PA *keθpiθ-). 15. Usually, the animate form of this noun is given as ‘drum’ and the inanimate as ‘kettle’, though there are instances in the Illinois sources where the animate form is glossed ‘kettle’. 16. Cf. Meskwaki aškiškiwi ‘mud’, Kickapoo askihkii, older askiskiwi ‘earth, ground, soil, land’ (Tr ‹oskiskeewēē› and Mc ‹a‘ckí‘ki› ‘dirt’), and Moose Cree aškiškiy ‘mud’. Some influence in the daughter languages from PA *ašišk(y)iwi ‘mud, dirt’ seems indicated as well.

References Bloomfield, Leonard. 1946. Algonquian Linguistic Structures of Native America, ed. by Harry Hoijer et al., pp. 85–129. New York: The Viking Fund. Costa, David J. 2003. The Miami-Illinois Language. Lincoln: University of Nebraska Press. Costa, David J. 2019. Rounding Dissimilation in Miami-Illinois. Papers of the 48th Algonquian Conference, ed. by Monica Macaulay and Margaret Noodin, pp. 53–68. East Lansing: Michigan State University Press. Dunn, Jacob P. [n.d.]. Various Notes on Miami. Manuscript, Indiana State Library, Indianapolis. Gatschet, Albert. 1878–1879. Notebook with Vocabularies, Texts, Notes (Shawnee and Potawatomi). Manuscript #68, National Anthropological Archives, Smithsonian Museum Support Center, Suitland, MD. Goddard, Ives. 1974. An Outline of the Historical Phonology of Arapaho and Atsina.

|

90    David J. Costa

International Journal of American Linguistics 40:102–116. Goddard, Ives. 1981. Massachusett Phonology: A Preliminary Look. Papers of the 12th Algonquian Conference, ed. by William Cowan, pp. 57–105. Ottawa: Carleton University. Goddard, Ives. 1982. The Historical Phonology of Munsee. International Journal of American Linguistics 48:16–48. Goddard, Ives. 1994. The West-to-East Cline in Algonquian Dialectology. Actes du 25e Congrès des Algonquinistes, ed. by William Cowan, pp. 187–211. Ottawa: Carleton University. Goddard, Ives. 1997. Addenda and Corrigenda for Leonard Bloomfield, “Algonquian” (1946). Algonquian and Iroquoian Linguistics 22(4):31–36. Goddard, Ives. 2002. Explaining the Double Reflexes of Word-Initial High Short Vowels in Fox. Diachronica 19(1):43–80. Goddard, Ives. 2015. Arapaho Historical Morphology. Anthropological Linguistics 57:345–411. Largillier, Jacques S. J. [n.d.; ca. 1700]. Illinois-French Dictionary. Manuscript, Watkinson Library, Trinity College, Hartford, CT. LeBoullenger, Antoine-Robert, S. J. [n.d.; ca. 1725]. French and Miami-Illinois Dictionary. Manuscript at the John Carter Brown Library, Brown University, Providence, RI. Mayhew, Experience. 1709. Massachuset Psalter. Boston: Nathaniel Green. Michelson, Truman. 1928. Notes on Kickapoo. Manuscript #3088, National Anthropological Archives, Smithsonian Museum Support Center, Suitland, MD. Pentland, David H. 1979. Algonquian Historical Phonology. PhD thesis, University of Toronto. Pinet, Pierre-François. [n.d.; ca. 1702]. French–Miami–Illinois dictionary. Manuscript, Archives des jésuites au Canada, Montréal, Québec. Preston, William. 1796. Miami Vocabulary. Manuscript #21, National Anthropological Archives, Smithsonian Museum Support Center, Suitland, MD. Râle, Sébastien. 1908. Facsimile of Pêre Marquette’s Illinois Prayer Book: It’s [sic] History by the Owner, Colonel J. L. Hubert Neilson, M.D. Quebec: Quebec Literary & Historical Society. [misattributed to Claude Allouez] Rhodes, Richard. 2021. The Case for Core Central Algonquian. Webs of Relationship and Words from Long Ago: A Festschrift Presented to Ives Goddard on the Occasion of his 80th Birthday, ed. by Lucy Thomason, David J. Costa, and Amy Dahlstrom, pp. 303–344. Petoskey, MI: Mundart Press. Trowbridge, Charles. [n.d.]. Notes on the Kickapoo. Manuscript at the Bentley Historical Library, University of Michigan, Ann Arbor, MI. Voegelin, Carl F. 1938–1940. Shawnee Stems and the Jacob P. Dunn Miami Dictionary. Indiana Historical Society Prehistory Research Series 1: 63–108, 135–167, 289–323, 345–406, 409–478 (1938–1940). Indianapolis.

kwêyask kotahâskwâtam The Effects of Altering Specificity in WordNet on the Accuracy of Computational Semantic Classifications of Plains Cree (nêhiyawêwin) Daniel Dacanay, Jolene Poulin, and Antti Arppe

T

he semantic classification of large-scale lexical resources such as dictionaries is an important preliminary step in the creation of various language resources.1 Tools such as semantically organized, or thematic, dictionaries (e.g., Visitor et al. 2013) can allow language learners (and by extension, instructors) to more effectively identify and fill lexical gaps by easing the retrieval of semantically relevant vocabulary. Additionally, the format of semantic groupings rather than alphabetical listing can serve as a more natural way of browsing vocabulary for languages with little orthographic standardization. Further, these “semantic dictionaries” permit linguists to more easily study patterns in lexicalization and lexical density, and can allow lexicographers to more easily identify potential gaps in a documented vocabulary. Finally, large-scale semantic classification is a fundamental step in the development of many modern digital language tools, including uses in machine translation, the improvement of predictive text and digital dictionary searches, the creation of intelligent spell-checkers, and the performance of textual sentiment analysis (Maas et al. 2011). As such, in pursuing the goal of providing useful linguistic resources for under-resourced languages, semantic classification of some kind inevitably becomes a necessity. However, the task of semantic classification has traditionally been a lengthy, repetitive, and costly affair, requiring months of dedicated manual work by linguists, often placing it out of the practical grasp of low-resource language

|

   91

|

92    Dacanay, Poulin, and Arppe

communities. In two previous papers (Dacanay et al. 2021a, 2021b), we have documented a method which appears capable of substantially reducing the workload necessary to produce a first-pass semantic classification of the entries in a bilingual dictionary. This approach leverages vector semantics with existing hierarchical semantic classification schemes to provide broadly usable classifications in as short as a matter of hours, but at the cost of these classifications tending toward inconsistent levels of accuracy compared to manual classifications (particularly relating to lexical specificity). Here, we will outline a refinement to this method, making use of the hierarchical structure of one of these existing semantic classification schemes, namely, the Princeton WordNet. This allows us to systematically adjust the degree of semantic generality in vector classifications by employing WordNet-indicated hypernyms, doing so with the intention of generating semantic domains with a more consistent and human-like degree of specificity. As in our previous works, this will be done in the context of a dictionary of Plains Cree/nêhiyawêwin (ISO:crk).

Background Plains Cree/nêhiyawêwin

Plains Cree (or nêhiyawêwin) is a language of the Cree-Montagnais-Naskapi branch of the Algonquian family, spoken throughout Alberta and Saskatchewan (Wolvengrey 2011). Although exact speaker population figures are contentious (ranging from 3,070 in Ethnologue (2015) to approximately 50,0252 in Statistics Canada’s 2016 Census of Population [Statistics Canada 2017]), Plains Cree’s speaker base in approximately the low ten-thousands makes it among the most widely spoken of Canada’s Indigenous languages. Despite this, most native speakers are elderly, and intergenerational transmission, though increasing, is still low. The largest contemporary bilingual dictionary of Plains Cree to English is nêhiyawêwin: itwêwina (Cree: Words; abbreviated CW), compiled throughout the 1990s and early 2000s by Arok Wolvengrey (2015). The dictionary, which is continually updated, is bilingual, being composed of roughly 21,000 part-of-speech-tagged Cree headwords, each given an English definition. For the purposes of our investigation, only Cree nouns and verbs (which together compose 88.4% of the dictionary’s content) were considered for semantic classification.

|

The Effects of Altering Specificity   93

Manual Semantic Classifications

As of 2021, CW has been fully semantically classified by manual annotators (Dacanay 2021a; Harrigan and Arppe 2021) using two separate hierarchical systems of semantic relationships (or ontologies), namely, the Princeton WordNet (Miller et al. 1990; abbreviated WN) and the SIL Rapid Word Collection Methodology (Moe 2003; abbreviated RW). Although semantic classification schemes designed specifically for Algonquian languages do exist (Visitor et al. 2013), we elected to make use of WN and RW on account of their potential for cross-linguistic comparison, with full classifications of a wide typological variety of languages existing already in both ontologies (Bosch and Griesel 2017; Boerger and Stutzman 2018). Although using “foreign” classification schemes, designed by Anglophones, rather than purpose-built Algonquian ones doubtless results in the loss of some degree of semantic nuance, the use of such ontologies remains the pragmatic choice from a standpoint of labor, particularly for obtaining first-pass classifications. Although designed predominantly by English speakers, both RW and WN are intended as relatively language-neutral lexico-semantic classification schemes, with groups of contextually synonymous or highly semantically proximate words organized into synonym sets, known as “synsets” in WordNet and “domains” in Rapid Words. These synonym sets are then arranged hierarchically via hypernymy and hyponymy into increasingly complex trees; for example, in WordNet, the English noun cod is contained in the synset (n) cod#2 alongside the synonymous term (n) codfish#1, with this synset having (n) saltwater_fish#1 as its immediate hypernym synset (with its hypernym tree ascending to (n) seafood#1, then (n) food#1, then (n) solid#1, and eventually up to (n) entity#1) and (n) salt_cod#1 as a hyponymic synset. In Rapid Words, the word cod would be a member of the semantic domain 1.6.1.5 Fish, which is subordinate to the semantic domains 1.6.1 Types of animals, then 1.6 Animals, and finally to 1 Universe, creation. Two of the chief differences between these ontologies concern size and part-of-speech classification; WordNet consists of 155,327 synsets, while Rapid Words contains 1,789 more general semantic domains, and while WordNet inherently organizes its synsets by part-of-speech (in the original English WN, nouns, verbs, adjectives, and adverbs), Rapid Words domains are defined based on semantic content alone, integrating vocabulary from multiple parts of speech. For the purposes of manual classification, CW entries were classified purely on the basis of lexical semantics, with entries being given correspondences to the synset(s) and domain(s) in WN and RW to which they were most closely

|

94    Dacanay, Poulin, and Arppe

semantically related. Part-of-speech was ignored entirely for WordNet synsets, and multiple classifications in either ontology were given for any Cree word that required multiple classifications to properly express the breadth of its meaning. This also aided in resolving cases in which the part-of-speech of the Cree entry differed from that of its closest correspondence in English (i.e., Plains Cree not having adjectives but rather expressing such attributes with verbs). For example, the Cree noun kinosêw ‘fish’ was given the WordNet classification (n) fish#1 and the Rapid Words classification 1.6.1.5 Fish, and the Cree verb kinosêwan ‘there are many fish’ was given the WordNet classifications (n) fish#1 and (adj.) abundant#1 and the Rapid Words classification 1.6.1.5 Fish. The details of this classification process are discussed in greater depth in Dacanay et al. (2021a). Our reasoning for manually classifying the full CW content was twofold; first, it ensured that the semantic classifications necessary for the various aforementioned functions would exist in a readily accessible and human-readable form, and second, it provided a “human” gold standard to serve as a benchmark for comparison with computational semantic classifications, namely those using vector semantics, of the same dictionary content. Vector Semantics

Vector semantics is the process of representing a lexical unit as “a point in a multidimensional semantic space” with the similarity of meaning between between two lexical units being “modeled as the distance between these points in the space” (Jurafsky and Martin 2021:5). More simply put, it is a means of computational semantic classification which relies on the Distributional Hypothesis (Firth 1957; Harris 1954), that is, the idea that words that occur in similar sentential contexts have similar meanings. Vector semantic models thus analyze the average distribution patterns of words across large corpora, enumerate aspects of these distributional contexts into sets of numbers (called “dimensions”), and use these numbers as a statistical representation of the meaning of the word as a whole. These sets of numbers, called “embeddings,” can be represented as vectors in a multidimensional space (corresponding to the aforementioned dimensions representing word meaning). These numbers may thus be compared by means such as cosine distance to determine their similarity; the more closely any two vectors are associated, that is, pointing toward the same direction in the multidimensional space, the more closely overlapping the average contexts of the compared words are, and thus, following

|

The Effects of Altering Specificity   95

the Distributional Hypothesis, the more similar these words are hypothesized to be semantically. As a means of determining distributional (and thus, theoretically, semantic) similarity between words, vector semantics appears a natural candidate for the computational generation of semantic classifications with a reduced (or nonexistent) need for manual intervention, having been used many times before for precisely this purpose (Jatnika et al. 2019; Al-Matham and Al-Khalifa 2021). To semantically classify the CW dictionary in this fashion, we generated vectors for each Cree entry by averaging the embeddings for all the individual English words in that entry’s definition as a “bag-of-words” (that is, the meaning of the definition as a whole was taken to be a simple aggregate of the meanings of all of its constituent words, weighting all words identically and ignoring word order). This was done largely as a matter of necessity, as existing Cree corpora are too small to provide a sufficient diversity of contexts to create Cree word vectors from purely Cree data (Harrigan and Arppe 2019); in addition, using the English definition words as a basis for the Cree word vectors also allowed us to leverage extremely large volumes of existing English training data (Mikolov et al. 2013). As a further by-product, this vector generation method is also theoretically applicable to any language which has a bilingual dictionary in English, regardless of the size of target language corpora. Finally, the use of English definition vectors to represent Cree words also permitted their comparison and classification with the English-based contents of WordNet and Rapid Words. This comparison is the basis of our classification method, as outlined in the next section. This same method was used to create vector representations for all of the WordNet synsets and Rapid Words domains. To create the vectors themselves, we used word2vec, a freely available word-vector generation tool drawing its contexts from the multibillion-word Google News Corpus (Mikolov et al. 2013). This process of vector generation is again discussed in more detail in previous publications (Harrigan and Arppe 2021; Dacanay et al. 2021a, 2021b). Classification Methods

For the sake of the human usability of the end product (as well as for the sake of comparison with human-created semantic classifications), when using vector semantics to semantically classify a large dictionary, it is necessary to classify the dictionary content into some organized system of semantic domains or groupings.

|

96    Dacanay, Poulin, and Arppe

To this end, we elected to make use of the two aforementioned ontologies: WordNet and Rapid Words. For every entry in the CW dictionary, the vectors of every WN synset and every RW domain were compared and ranked by relatedness (determined by cosine distance), creating a list of all WordNet synsets ranked by relatedness to the given Cree word, and a list of all Rapid Words domains ranked by relatedness to the given Cree word. For each of these two lists, we then determined where the top manual classification for the corresponding entry in the corresponding ontology was found, with the best (or most “human-like”) possible result being one in which the same synset or domain was selected as the top-ranking vector classification as was selected by a human annotator for the manual classification, indicating that both the word vectors and the human annotator deemed the same ontological category (synset or domain) as being the best possible fit for the given Cree word. Initial Results

In both WordNet and Rapid Words, significant differences in classification accuracy were observed between different Cree parts of speech. For WordNet, the median rank of the top manual classification among the vector classifications for Cree nouns was 15, whereas the median rank for Cree verbs was 333. In Rapid Words, the median ranks were 2 for Cree nouns and 15 for verbs. Instances in which the top vector classification and the top manual classification were an exact match, the theoretical ideal case for vector classifications, were uncommon overall. For the WordNet classifications, for example, this occurred only 315 times for the 13,669 Cree verbs in the dictionary, and only 726 times for the 5,212 nouns. As Table 1 shows, the overall results for the Rapid Words vector classifications were, by our metric, significantly more human-like than those with WordNet; that is, the top manual classifications were ranked higher among the vector classifications on average. However, although in absolute terms, Rapid Words produced more human-like results on average, proportional to the total number of possible classifications in each ontology (155,327 in WordNet vs. 1789 in Rapid Words), WordNet’s classifications were still superior, with Cree nouns having a median rank in the top 0.0127% of total classifications and verbs having a median in the top 0.283% in WordNet, compared to 0.203% for nouns and 1.53% for verbs in Rapid Words (Dacanay et al. 2021b). Qualitatively speaking, even in instances in which the exact manual classification was not ranked within the first, say, 100 vector classifications, the high-ranking

|

The Effects of Altering Specificity   97

TABLE 1. Percentile chart of median ranks of manual classifications among vector

classifications

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

VERBS (WN)

NOUNS (WN)

VERBS (RW)

NOUNS (RW)

1 5 18 52 137 333 762 1634 3554 9554 13,7352

1 1 2 4 7 15 28 59 164 864 121,883

1 1 1 3 6 15 38 80 161 327 983

1 1 1 1 1 2 4 10 24 69 976

This table shows the median vector-assigned ranks of the highest-ranked manual classifications among the vector classifications of Cree words, segmented by part of speech (PoS) and ontology. For example, 20% of the time, the top manual classification for a Cree verb in WordNet was within the top 18 vector classifications. The median ranks for each ontology-PoS pair are bolded.

vector classifications for most Cree words in both ontologies were still typically within the correct basic semantic domain. For example, for the Cree word apihkêsîs ‘spider’, while the human classification (n) spider#1 is only ranked 185th, the top ten WN classifications consist almost entirely of specific species and products of spiders: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

(n) spider_web#1 (Cosine Similarity: 0.79724957), (n) spider’s_web#1 (Cosine Similarity: 0.79724957), (n) spider_web#2 (Cosine Similarity: 0.79230372), (n) spider’s_web#2 (Cosine Similarity: 0.79230372), (n) genus_Ateles#1 (Cosine Similarity: 0.76677546), (n) orb_web#1 (Cosine Similarity: 0.75656461), (n) family_Majidae#1 (Cosine Similarity: 0.75422829), (n) funnel_web#1 (Cosine Similarity: 0.67104201), (n) spider_monkey#1 (Cosine Similarity: 0.66930821), (n) Ateles_geoffroyi#1 (Cosine Similarity: 0.66930821),

|

98    Dacanay, Poulin, and Arppe

Along these lines, one persistent issue that was encountered (particularly using WordNet) concerned the vector classifications’ degree of specificity. Initial observations revealed that high-ranking vector classifications for Cree words representing general semantic concepts (often generic animal terms, tools, and clothing) tended toward being either specific subtypes of, components of, or activities related to the meaning of the original Cree word, rather than the more appropriately general umbrella term. For example, the Cree words môhkomân ‘knife’ and mostos ‘cow, cattle, buffalo’ have top-ranking vector classifications of (n) knife_blade#1 and (n) cow_pasture#1, with their top human classifications ((n) knife#1 and (n) cattle#1) only being ranked in positions 18 and 53 respectively. This overspecificity problem was far more prevalent in the WordNet vector classifications than in those with Rapid Words, likely as a product of Rapid Words being a more general hierarchy of semantic classes, rather than being a hierarchy of individual words and terms like WordNet, and thus lacking dedicated classification categories for extremely specific lexicalized semantic concepts.

WordNet Structural Variations Potential Types of Structural Change

Based on both the quantitative results and general qualitative evaluations previously mentioned, it appears that ontological simplicity (which, in this context, largely denotes size and overall generality) does play at least some role in the accuracy of vector-based semantic classifications relative to those produced by humans. Given this generally observed trend among the RW vector classifications and the broad structural similarities between RW and WN that have already been noted, it would immediately appear that the most straightforward method to improve the quality of the current WN vector classifications would be to simply reduce the number of items from WN that are considered as classification categories evenly across all semantic domains, in line with the simpler structure of RW. However, there are a variety of potential strategies that can be applied to this end. As a first option, one could take the existing WordNet classifications and move them all up one or several levels in the hypernymy hierarchy. Alternatively, one could designate a specific level in the hypernym hierarchy as the maximum possible level of specificity and use only synsets at or above its level as classification labels. As a final option, one

|

The Effects of Altering Specificity   99

could, rather than simplifying or reducing the hierarchical structure of WordNet per se, take some number of existing top-level vector classifications for each Cree word (say, the top 100) and average their vectors, and then use this average to find a classification for the word as a whole. The first of these three methods is outlined in the coming section; the latter two are to be the subjects of future papers. Our rationale for beginning with the Hypernymy Method is a simple one; in short, this method would initially appear to be the most suited to resolve the fact that existent WordNet vector classifications, despite typically being in the correct basic semantic region, were too highly specific or imprecise to replace human annotation. Given that the Hypernymy Method would simply take the existing vector classifications for the CW content and move the classifications higher in their current branch of the hierarchy, this method, in theory, functions only to improve classification accuracy, as all previously correct classifications are retained and carried forward, along with the additional, newly correct classifications generated by moving upward in the hypernymy hierarchy (see Figure 1).

FIGURE 1. A demonstration of WordNet’s hypernymy hierarchy. If, for example, (n) meat_cleaver#1 (or any of its parallel synsets, such as (n) knife_blade#1 or (n) hunting_knife#1) is moved one level up, it becomes (n) knife#1, if it is moved up two levels, (n) edge_tool#1, and so on. Since this hierarchy is linear, once two classifications have converged on the same “central” line on the hierarchy, they will remain matched up to the highest level.

|

100    Dacanay, Poulin, and Arppe

TABLE 2. Median manual classification ranks with altered hypernym levels NOUNS

VERBS

VECTOR VECTOR 1 CONTROL LEVELS

VECTOR 2 VECTOR 3 VECTOR VECTOR 1 LEVELS LEVELS CONTROL LEVELS

VECTOR 2 VECTOR 3 LEVELS LEVELS

18

19

42.5

118

210

250

384

343

420

16

20

54

1205

93

94

102.5

MANUAL 2 LEVELS

1410.5

131

13

17.5

1475.5

149

49

44

MANUAL 3 LEVELS

2733

395

64

11

1624

156

62

34

MANUAL MANUAL 1 LEVELS

This table shows median ranks of manual classifications among vector classifications for Cree nouns and verbs after moving both classification types 1, 2, and 3 levels higher in the hypernymy hierarchy. Note the bolded, downward-diagonal patterns, showing the consistent increase in accuracy (or a decrease in median rank) as one moves up the hierarchy in equal steps for both classifications.

Results of the Hypernymy Method

The nature of these results reveals a much less straightforward pattern in vector classification overspecificity than was previously observed. While moving vector classifications higher in the hypernymy hierarchy does indeed improve the rank of the top manual classification among the vector classifications, the best result qualities were obtained when both vector classifications and the manual classifications to which they were being compared were moved higher in the hierarchy by the same number of levels. Indeed, the largest improvements for both Cree nouns and verbs are seen when both the manual and vector classifications are moved up three levels in the hypernymy hierarchy, with the median position of the manual classification decreasing from 15 to 11 in Cree nouns and 333 to 34 for Cree verbs. By contrast, when moving only vector classifications higher in the hypernymy hierarchy and retaining the position of the corresponding manual classifications (the combination initially hypothesized to be the most effective), the median position worsens across the board for Cree nouns and worsens when vector classifications are moved two levels up or higher for Cree verbs (indicated by the top row in Table 2). The worst results overall, however, were seen when vector classifications were retained in their control positions and the manual classifications were moved up in the hierarchy, with results worsening in quality across the board the higher in the hierarchy the manual classifications were moved, as is shown in columns 1 and 5 of Table 2. These results appear to contradict our initial observations that vector classifications tended to be domain-correct but overly specific. If this were the case, one

|

The Effects of Altering Specificity   101

would expect retaining the manual classifications in their initial positions and moving the vector classifications upward in the hypernym hierarchy to yield the most consistent improvement, as the overspecific vector classifications would converge with their more suitably general manual classifications when moved sufficiently higher in the hierarchy. However, our results indicate that, rather than being overly specific as a rule, WordNet vector classifications tend rather to be domain-accurate, but semantically imprecise, with high-ranking vector classifications tending toward being on more or less the correct level of specificity, but not necessarily the same precise branch of the hypernym tree as the manual classification. In such an instance, the manual and vector classifications would only converge if both were to be moved to a common ancestor node on the hypernym tree, rather than if only one of them were to be moved. For example, the top vector classification for the Cree verb pimohtêw ‘s/he walks, s/he walks along’ is (v) walk_in#1, a synset that, in WordNet’s structure, is parallel to, rather than a hyponym of, the synset that was chosen as the manual classification ((v) walk#1). As such, increasing the hypernym level of only the vector classification (or, indeed, only the manual classification) would function simply to further distance these existing classifications; rather, both must be moved to converge on a common hypernymic ancestor node. This phenomenon of semantic imprecision, rather than overspecificity, being at least partially responsible for the lesser quality of WordNet vector classification remains congruous with our previous assumption that Rapid Words exhibited superior classification quality due to its size. While this reduced size does indeed result in Rapid Words having fewer highly specific domains, it also results in Rapid Words having fewer domains on each level of specificity, reducing the possible degree of imprecision (that is, the number of incorrect choices) for a vector classification that is in the correct basic semantic area. Case Studies

As mentioned, increasing the hypernym level of both the manual and vector classifications seems to provide the most reliable improvements in the median position of the manual classification. For example, with the Cree noun wâposwayânakohp ‘rabbitskin robe, rabbitskin blanket’, the top manual classification ((n) robe#1) appeared in position 190 in the initial vector classifications, but when both vector and manual classifications were moved up by one level, the top vector classification ((n) sealskin#2) and the top manual classification both became (n) garment#1,

|

102    Dacanay, Poulin, and Arppe

providing an exact match for the highest-ranking classification. The general nature of this case (if not always its magnitude) is fairly typical of entries where an improvement was recorded after applying the Hypernymy Method. However, improvements of this nature are contingent on existing high-ranking vector classifications for a Cree word being on the same basic branch of the hypernymy hierarchy as at least one of the manual classifications; even in instances when top-ranking vector classifications are semantically relevant, this is not necessarily always the case. For instance, for the Cree noun kotikonikan ‘breech-loading gun’, the top-ranking vector classifications largely concern the breech itself (the top three, for example, are (n) rear_of_tube#1, (n) breech#1, and (n) rear_of_barrel#1). These synsets, though highly semantically related to the concept of breech-loading firearms, are hyponymic in WordNet to (n) opening#10, rather than to any synset overtly relating to firearms (such as (n) weapon#1). As such, short of receding to an extremely high-level, generic synset such as (n) artefact#1, there is no practical method in this case to make the top-ranking vector-selected synset converge with the manual classification. This being said, moving both all vector and manual classifications up in the hierarchy still improves the result quality for kotikonikan overall; with the default classifications, (n) breechloader#1 does not occur anywhere within the top 10,000 vector classifications, but when both classification types are moved up one level, their common synset (n) gun#1 occurs in position 25 (having previously been (n) cannon#4). There are, indeed, occasional instances in which moving the vector classifications while retaining the original manual classifications yields the best results, as our initial hypotheses predicted. For example, for the Cree noun môhkomân, the default position of the manual classification ((n) knife#1) is 52, but if one moves all (and only) vector classifications up by one level, (n) knife#1 appears in position 4 (previously having been (n) meat_cleaver#1), while by contrast the manual classification remains at 52 if one moves both the manual and vector classifications (with the former having been changed to (n) edge_tool#1). However, in addition to being few in number, cases such as this also tended to provide their best results when moved only a single level higher in the hierarchy, with the results declining in quality the more levels up one moves; this is in contrast to what occurs when manual and vector classifications are both moved up by the same degree, in which, once the two have converged, they remain a match indefinitely. For example, with môhkomân, although the manual classification’s median position is improved when the vector classification is moved up one level, if it is moved two levels, the

|

The Effects of Altering Specificity   103

previous matching vector classification at position 4 is lost, having been moved up to (n) edge_tool#1, no longer matching the original manual classification (n) knife#1.

Conclusion

Reducing the size of WordNet by exploiting its linear core hypernymy structure does indeed appear an effective and reliable method of improving vector semantic classification accuracy across part-of-speech boundaries. In effect, this “Hypernymy Method” functions to provide an intermediary stage between the thoroughness and semantic coverage of WordNet and the more simplistic semantic domains of Rapid Words, although it is still unable to match the median human-likeness of vector classifications made in the latter ontology. Despite this, it is clear that size reduction in WordNet has the potential to make the ontology more suitable to vector classification, and to this end, future research into alternative approaches to this size reduction (such as using only a preset hypernymy level for all classifications), as well as alternate means of selecting the ‘best’ vector classifications (such as using a weighted vote of existing classifications), should elucidate the precise means by which this reduction might be best approached. Notes 1. We would like to thank the Social Sciences and Humanities Research Council of Canada (SSHRC) for funding this research through their grant #895-2019-1012 (21st Century Tools for Indigenous Languages), as well as Dr. Arok Wolvengrey, for our continued use of his dictionary. 2. Although speaker counts for Plains Cree specifically are not mentioned, the figure of 50,025 is derived from the number of self-declared speakers of ‘Cree’ in Alberta and Saskatchewan

References Al-Matham, Rawan N., and Hend Al-Khalifa. 2021. SynoExtractor: A Novel Pipeline for Arabic Synonym Extraction using word2vec Word Embeddings. Complexity 3:1–13. doi:10.1155/2021/6627434. Arppe, Antti, Atticus Harrigan, Katherine Schmirler, and Arok Wolvengrey. 2018. A

|

104    Dacanay, Poulin, and Arppe

Morphologically Intelligent Online Dictionary for Plains Cree. Paper read at the 25th Stabilizing Indigenous Languages Symposium (SILS), University of Lethbridge, Lethbridge, Canada. Boerger, Brenda H., and Verna Stutzman. 2018. Single-Event Rapid Word Collection Workshops: Efficient, Effective, Empowering. Language Documentation and Conservation 12:147–193. Bosch, Sonja E., and Marissa Griesel. 2017. Strategies for Building WordNets for UnderResourced Languages: The Case of African Languages. Literator: Journal of Literary Criticism, Comparative Linguistics and Literary Studies 38(1):8. doi:10.4102/lit.v38i1.1351. Dacanay, Daniel, Atticus Harrigan, and Antti Arppe. 2021a. Computational Analysis versus Human Intuition: A Critical Comparison of Vector Semantics with Manual Semantic Classification in the Context of Plains Cree. Proceedings of the 4th Workshop on Computational Methods for Endangered Languages, ed. by Antti Arppe, Jeff Good, Atticus Harrigan, Mans Hulden, Jordan Lachler, Sarah Moeller, Alexis Palmer, Miikka Silfverberg, Lane Schwartz, 1:33–43. doi:10.33011/computel.v1i.971. Dacanay, Daniel, Atticus Harrigan, Arok Wolvengrey, and Antti Arppe. 2021b. The More Detail, the Better?—Investigating the Effects of Semantic Ontology Specificity on Vector Semantic Classification with a Plains Cree / nêhiyawêwin Dictionary. Proceedings of the First Workshop on Natural Language Processing for Indigenous Languages of the Americas (NAACL-HLT 2021), ed. by Manuel Mager, Arturo Oncevay, Annette Rios, Ivan Vladimir Meza Ruiz, Alexis Palmer, Graham Neubig, Katharina Kann, pp. 143–152. doi:10.18653/ v1/2021.americasnlp-1.15. Ethnologue. 2015. Plains Cree (crk). https://www.ethnologue.com/18/language/crk/. Firth, John R. 1957. A Synopsis of Linguistic Theory, 1930–1955. Selected Papers of J. R. Firth 1952–1959, ed. by John R. Firth, pp. 168–205. London: Longmans. Harrigan, Atticus, and Antti Arppe. 2019. Automatic Semantic Classification of Plains Cree Verbs. Paper read at the 51st Algonquian Conference in Montreal, Canada. Harrigan, Atticus, and Antti Arppe. 2021. Leveraging English Word Embeddings for SemiAutomatic Semantic Classification in Nêhiyawêwin (Plains Cree). Proceedings of the First Workshop on Natural Language Processing for Indigenous Languages of the Americas (NAACL-HLT 2021), pp. 113–121. doi:10.18653/v1/2021.americasnlp-1.12. Harris, Zellig S. 1954. Distributional structure. Word 10(2–3):146–62. Jatnika, Derry, Moch A. Bijaksana, and Arie A. Suryani. 2019. Word2Vec Model Analysis for Semantic Similarities in English Words. Procedia Computer Science 157:160–167. doi:10.1016/j.procs.2019.08.153.

|

The Effects of Altering Specificity   105

Jurafsky, Daniel, and James H. Martin. 2021. Vector Semantics and Embeddings. Speech and Language Processing, ed. by Daniel Jurafsky and James H. Martin, pp. 96–127. Hoboken, NJ: Prentice Hall Inc. Maas, Andrew L., Raymond E. Daly, Peter P. Pham, Dan Huang, Andrew Y. Ng, and Christopher Potts. 2011. Learning Word Vectors for Semantic Analysis. Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, ed. by Dekang Lin, Yuji Matsumoto, and Rada Mihalcea, pp. 142–150. Portland, Oregon. Mikolov, Tomas, Ilya Sutskever, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Distributed Representations of Words and Phrases and their Compositionality. Proceedings of Workshop at International Conference on Learning Representations (ICLR 2013), ed. by Yoshua Bengio and Yann Lecun, pp. 3111–3119. doi:10.5555/2999792.2999959. Miller, George. A., Richard Beckwith, Christine Fellbaum, Derek Gross, and Katherine J. Miller. 1990. Introduction to WordNet: An On-line Lexical Database. International Journal of Lexicography 3(4):235–244. doi:/10.1093/ijl/3.4.235. Moe, Ronald. 2003. Compiling Dictionaries using Semantic Domains. Lexikos 13:215–223. doi:10.5788/13-0-731. Statistics Canada. 2017. The Aboriginal Languages of First Nations People, Métis and Inuit. https://www12.statcan.gc.ca/census-recensement/2016/as-sa/98-200-x/2016022/98-200x2016022-eng.cfm. Visitor, Linda, Marie-Odile Junker, and Mimie Neacappo, eds. 2013. Eastern James Bay Cree Thematic Dictionary ([Northern/Southern] Dialect). Chisasibi: Cree School Board. Wolvengrey, Arok. 2011. Cree: Words. Regina: University of Regina Press.

The Meskwaki Past Tense Enclitic =iyo·we in Relation to an Affixal Past Tense Strategy Amy Dahlstrom

T

his paper describes the distribution and functions of a Meskwaki past tense enclitic, =iyo·we, and its relationship to a structurally distinct strategy for indicating past tense—the conjunct order mode suffix -ehe.1 The corpus for this study is drawn from narrative texts from the early twentieth century written by monolingual Meskwaki speakers.2 I propose that the core function of the enclitic =iyo·we is to express past tense in syntactic contexts where conjunct order verb inflection is not available, as described in the first section. I then briefly review the findings of Dahlstrom 2021, which describes the past tense and irrealis functions of the conjunct order mode suffix -ehe, as background for consideration of instances where the enclitic =iyo·we appears with verbs inflected in the conjunct order.3 The final section compares the Meskwaki phenomenon with Cree dialects in which there appears to be a competition between enclitic and affixal strategies of expressing evidentiality. I suggest that a similar process may be underway in the strategies of expressing past tense in Meskwaki, and that the Meskwaki enclitic may be gaining ground at the expense of the affixal strategy. Before turning to the specific examples, it should be noted that overt past tense marking is not obligatory in Meskwaki. A verb that does not bear a future prefix can be understood as referring to present time or to past time, depending on the context (e.g., the presence of an adverb such as ana·kowe ‘yesterday’). In (1), for example, a verb inflected in the independent indicative mode may have a reading

|

   107

|

108    Amy Dahlstrom

of either present tense or past tense; the same is true of the verb in (2), inflected in the aorist conjunct mode.4 (1) wača·howa wača·ho-wa cook-3.ind ‘She is cooking; she cooked.’ [independent indicative mode] (2) nekehke·nema·wa e·hwača·hoči ne-kehke·nem-a·wa e·h-wača·ho-či 1-know-1>3.ind

aor-cook-3.aor

‘I know that she is cooking; . . . that she cooked.’ [aorist conjunct mode]

Since overt past tense marking is not obligatory, it is of interest to see under what conditions the enclitic =iyo·we or the conjunct suffix -ehe is used.

Distribution of the Enclitic =iyo·we

The enclitic =iyo·we is one of twenty-nine Meskwaki enclitics discussed in detail in Goddard 2015. It is built upon a demonstrative stem (Proto-Algonquian *eyo· ‘this (inanimate)’), plus a suffix (Goddard 2015:113). In terms of its distribution, it may attach either to an NP, with scope over only that NP, or it may be a clausal enclitic, typically appearing in second position. If there is a string of more than one second-position enclitics, =iyo·we always appears last (Goddard 2015:82). Example (3) shows =iyo·we appearing in the final position of a clitic string: (3) pe·hkike·hwi·nameka·peheyo·we ketaniwose. pe·hki=ke·hi=wi·na=meko=a·pehe=iyo·we

ke-aniwose·-Ø

really=moreover=contr=emph=always=past 2-walk.a.lot-2.ind ‘But you used to walk very frequently.’ (W628)

|

The Meskwaki Past Tense Enclitic   109

Simple NPs

As mentioned above, =iyo·we may appear attached to an NP, with scope over only that NP. In such cases, the semantic effect of the enclitic is a reading of ‘former’, as in (4) and (5): 5 (4) e·hne·wa·či ona·pemaniyo·we e·h-ne·w-a·či

o-na·pem-ani=iyo·we

aor-see-3>3’.aor 3-husband-obv=past ‘She saw her former husband.’ (W932) (5) “. . . ” e·hineči tepowe·neni·hakiyo·we. “. . . ” e·h-in-eči

tepowe·neni·h-aki=iyo·we.

aor-say.thus.to-X>3.aor councilman-pl=past ‘ “. . . ,” the former councilmen were told.’ (W749)

In (4) the NP hosting the enclitic =iyo·we refers to an individual who was the subject’s husband in the past, but is no longer. Likewise, in (5) the NP functioning as host of =iyo·we refers to individuals who served as councilmen in the past. In both instances the gloss of ‘former’ for =iyo·we is appropiate. Clauses Containing an Independent Order Verb

Meskwaki, like most of the other members of the Algonquian family, displays a grammatical opposition between independent order and conjunct order verb inflection. In Meskwaki, independent order inflection appears in non-negated, non-modal, non-imperative, and non-narrative main clauses, while conjunct order inflection appears primarily in subordinate clauses, negated main clauses, and main clauses in narratives. If the syntactic or semantic context requires a verb to be inflected with independent order affixes, the conjunct order suffix -ehe (discussed in the following section) is not available to indicate past tense. Instead, the enclitic =iyo·we is used, as seen already in (3) and additionally in (6): (6) . . . , i·niniyoyo·we nekahkitawa·waki opešekesimwa·wani, . . . na·mepye·kiyo·we nesakapihtawa·waki, . . . i·nini=iyo=iyo·we ne-kahkitaw-a·waki

o-pešekesim-wa·w-ani

that.obv=for=past 1-hide.O2.from-1>3p.ind 3-deer-3p-obv

|

110    Amy Dahlstrom

na·mepye·ki=iyo·we ne-sakapihtaw-a·waki under.the.water=past 1-tie.O’s.O2.fast-1>3p.ind ‘. . . , for I had hidden their deer from them, . . . I had tied it up under the water, . . . ’ (W99)

This function of =iyo·we—marking past tense on independent order verbs—appears to be the core function of the enclitic. More than half of the examples in my corpus (79 out of 153) exhibit =iyo·we marking past tense in clauses with independent order verbs. Clauses Containing a Verb Inflected in the Potential Mode

Another verbal paradigm occurring in main clauses in Meskwaki is potential inflection, which may be glossed ‘would, could, should, might’. In (7) the speaker is a spirit addressing the culture hero Wisahkeha as a young boy; the spirit is describing how the human world of the future might have been if Wisahkeha had not committed a serious error. This contrary-to-fact state is expressed by the combination of potential inflection on the verb plus the past tense enclitic =iyo·we. (7) ayo·hiyo·we tako·mikatesa mešemeko·nahi ma·mi·čiya·ke. ayo·hi=iyo·we tako·mikat-esa meše=meko=i·nahi here=past exist-0.pot

anything [idiom]

ic-ma·-mi·či-ya·ke ic-redup-eat-1p>0.part.0 ‘Anything we eat would have existed here.’ (W54D) Verbless Constructions

Another environment in which the past tense enclitic =iyo·we appears is in constructions lacking a verb. In such constructions the alternate strategy, of suffixing -ehe to a verb inflected in the conjunct order, is naturally not available. In terms of the syntax of verbless constructions, they appear in main clause environments, similar to the use of verbs inflected in the independent order.

|

The Meskwaki Past Tense Enclitic   111

The most commonly encountered type of verbless construction is an equational sentence with a zero copula, as in (8): (8) manikohiyo·we we·či–pye·či–kakano·neti·hena·ni, . . . mani=kohi=iyo·we ic-oči–pye·či–kakano·neti·h-ena·ni this=certainly=past ic-from–come.to–converse.with-1>2.part.obl ‘This was why I came to talk with you, . . . ’ (W608K)

In (8) the lefthand element mani ‘this (inanimate)’ is equated to the headless relative clause on the right. Schematically, we may render (8) as “[this] = [the reason why I came to talk with you]”. Such equational sentences are extremely common in the textual corpus under investigation. In (8) the presence of the enclitic =iyo·we marks the equational sentence as being true in the past. Another verbless construction in Meskwaki is a predicate nominal used on its own: (9) pe·hkike·hiyo·we wa·waneška·haki. pe·hki=ke·hi=iyo·we

wa·waneška·h-aki

really=moreover=past rogue-pl ‘Moreover, in the past they used to be really bad boys.’ (W113G)

The predicate nominal construction resembles the equational construction seen in (8) because there is no verb present; it differs from the equational construction in that there is no overt lefthand element equated to the predicate nominal. As (9) demonstrates, the enclitic =iyo·we may appear attached to the first phonological word of a predicate nominal. Conjunct Order Verbs

The previous section demonstrated that the enclitic =iyo·we is used to indicate past tense in syntactic environments requiring a verb inflected in the independent order or potential mode, or in environments lacking a verb altogether. These syntactic environments, plus the use of =iyo·we with NPs, make up 75% of the instances of =iyo·we in my corpus (114 out of 153). Before turning to a consideration of the remainder of the =iyo·we tokens, I will briefly review the findings of Dahlstrom

|

112    Amy Dahlstrom

2021, which investigates the past tense and irrealis functions of the conjunct order mode suffix -ehe. Conjunct Order Verbs Suffixed with -ehe

The -ehe suffix is found on a subset of paradigms within the conjunct order, appearing in the slot for the mode suffix at the rightmost edge of the verb inflection.6 Filling the mode suffix slot with -ehe produces pairs of structurally related conjunct order modes, as seen in (10), in which the individual components of the inflectional suffixes are separately glossed:7 (10) a. a·kwi wača·ho-č-ini not cook-3-mode.suffix ‘She doesn’t/didn’t cook.’ [negative paradigm] b. a·kwi wača·ho-t-ehe not cook-3-past ‘She didn’t cook.’ [remote past] [negative preterite]

The paradigms in which the mode suffix slot is filled with -ehe are identified with the label ‘preterite’: negative preterite (as in (10b)), aorist preterite, preterite participle, changed preterite, and subjunctive preterite. The semantic effect of filling the mode suffix slot with -ehe varies. In some contexts it produces a clear case of a marked past tense, as in the relative clause in (11) where -ehe marks a remote or relative past tense, and in other contexts it produces a clear case of an irrealis function, as in the counterfactual conditional in (12). (11) ča·kimeko nemehkwe·neta e·nahina·čimohitehe ča·ki=meko ne-mehkwe·net-a ic-inah-ina·čimoh-itehe all=emph 1-remember-1>0.ind ic-redup-inform.so-3>1.pret.part.obl ‘I remembered everything she had ever told me.’ (Goddard 2006:117D) (12) pwa·wimata–ona·pe·miya·nehe wi·tamo·naka·wa·ha pwa·wi–=mata –ona·pe·mi-ya·nehe not–=rather

wi·tamaw-enaka·wa·ha

–have.husband-1.subjnct.pret tell-1>2p.potential

‘If I weren’t married I would tell you [plural].’ (W187)

|

The Meskwaki Past Tense Enclitic   113

A connection between past tense and irrealis constructions is not found only in Meskwaki, of course: similar patterns have been described for Cree-Innu-Naskapi (cf. James 1991; James et al. 2001; Drapeau 2014:179ff) and Miami-Illinois (Costa 2003:354), as well as for non-Algonquian languages (cf. Steele 1975; James 1982; Fleischman 1989; Larreya 2003; Hogeweg 2009, among others). In Dahlstrom 2021 I conjecture that for Meskwaki the past tense function of -ehe is the basic one, and that the key phenomenon providing a pathway from past tense to irrealis may be remote past uses such as the example in (13), with third-person inflection on the verb. More specifically, a verb inflected for both a remote past tense and a third-person subject entails speaker uncertainty about the truth of the statement. Once the -ehe suffix is associated with speaker uncertainty, it may then be extended to other irrealis functions. (13) a·kwiča·hye·toke nana·ši mešotehe i·niya nemešo·ha a·kwi=ča·hi=ye·toke nana·ši mešw-etehe not=so=it.seems i·niya

ever

hit.with.shot-X>3.neg.pret

ne-mešo·h-a

that.absent 1-grandfather-sg ‘My late grandfather was never hit in battle, it seems.’ (Michelson 1927:70.3)

In (13) the verb is inflected in the negative preterite, referring to a time when the speaker’s late grandfather was a young man taking part in battles. The speaker himself did not witness the event described. Dahlstrom 2021 also discusses two irrealis constructions formed by combining the future prefix wi·h- with the -ehe suffix. It is worth drawing attention to them here since we will see counterparts with =iyo·we below. The combination of future + past can be used to express unfulfilled intentions, as in (14), or a past modal reading, as in (15): (14) wi·hwi·če·nomaketeheča·hi wi·h-wi·če·nom-aketehe=ča·hi fut-play.with-1p>3.aor.pret=so ‘We (exclusive) were just going to play with him [but we didn’t].’ (W78)

|

114    Amy Dahlstrom

(15) na·hka ke·htenameko i·ni wi·hni·šwiha·tehe no·šisema oni·ča·nese·hahi na·hka ke·htena=meko i·ni wi·h-ni·šwih-a·tehe and surely=emph now fut-have.2.of.O-3>3′.aor.pret ne-o·šisem-a

o-ni·ča·nese·h-ahi

1-grandchild-sg 3-child.dim-obv.pl ‘And surely my grandson would have had two children now.’ (W928–929)

See Dahlstrom (2021:111–115) for more discussion of these constructions. Conjunct Order Verbs with the Enclitic =iyo·we

As mentioned above, the constructions discussed in the first section (simple NPs, independent order verbs, potential verbs, verbless constructions) account for about 75% of the tokens in my corpus of =iyo·we examples (114 out of 153). There are, however, instances of the enclitic appearing in clauses containing a verb inflected in the conjunct order, an environment in which the affixal strategy of using -ehe should theoretically be available. It is therefore of interest to look at some examples of the use of =iyo·we in conjunct order clauses: specifically, in clauses containing conjunct participles, negative inflection, aorist conjunct, and the irrealis constructions exhibiting both a future prefix and a past marker. Conjunct Participles

Participle forms are used in Meskwaki to express relative clauses. Such clauses may be used to modify a head noun; additionally, participles frequently appear on their own, forming an NP with an understood pronominal head noun. As discussed in Dahlstrom (2021:108–109), it is possible to indicate past tense on participles with an -ehe suffix in the mode slot, forming a preterite participle. An example of a preterite participle is in (16). (16) we·ča·hotehe ic-wača·ho-tehe ic-cook-3.pret.part.3 ‘the one who had cooked’

|

The Meskwaki Past Tense Enclitic   115

In the corpus of textual examples under investigation, there are several examples of the enclitic =iyo·we apparently taking the place of the expected conjunct order suffix -ehe to overtly mark a conjunct participle as past tense. Consider (17): (17) “. . . ” e·hinemeči me·šoška·čikiyo·we. “. . . ” e·h-in-emeči

ic-mešoška·-čiki=iyo·we.

aor-say.thus.to-X>3’.aor ic-be.smitten-3p.part.3p=past ‘ “. . . ,” the ones who had been smitten were told.’ (W1010)

In (17) the participle me·šoška·čiki ‘the ones who were smitten’ refers to two young men who had immediately become infatuated with two young women whom they happened to see naked. The enclitic =iyo·we overtly marks that the time when the young men were smitten was in the past. Presumably the past tense conjunct suffix -ehe could have been used here instead, but the enclitic strategy was chosen to express past tense. The effect of the combination me·šoška·čikiyo·we, then, is to indicate a relative past tense: the time of being smitten happened at an earlier time than the young men being told the quote in (17). Example (18) shows another instance of the past tense enclitic =iyo·we on a conjunct participle: (18) a·kwi owiye·hani wi·ha·hpečinanekowa·čini ki·ši–nesa·wa·činiyo·we. a·kwi owiye·h-ani wi·h-a·hpečinan-ekowa·čini not anyone-obv fut-kill.for.good-3’>3p.neg ic-ki·ši–nes-a·wa·čini=iyo·we. ic-perf–kill-3p>3’.part.3’=past ‘No one [obviative] whom they [proximate] had killed in the past will be killed for good by them [proximate].’ (W1047)

Example (18) is a rather complicated sentence to parse. It is taken from a passage of the long Wisahkeha text in which the culture hero is describing the afterlife, at which time murder victims will be given new life but those who had murdered them will not. The participle in (18) is ki·ši–nesa·wa·čini ‘the one(s) [obviative] whom they [proximate] had killed’, with the enclitic =iyo·we explicitly marking that the event of killing took place in the past; the participle modifies owiye·hani ‘anyone [obviative]’, forming a discontinuous NP. As in (17), the enclitic strategy in (18)

|

116    Amy Dahlstrom

seems to be taking the place of the suffixal past tense strategy, even though there is no morphological barrier to using the -ehe suffix on the conjunct participle in (18). The position of the enclitic =iyo·we is noteworthy. In (17) and (18) we see that it is attached directly to the conjunct participle, not in clause-second position. This position of the enclitic may be motivated in part by the fact that the conjunct participles may function as NPs. We have already seen that the enclitic =iyo·we may attach to a simple NP, such as ona·pemaniyo·we ‘her former husband’ in (4). But in (17) and (18) the interpretation of =iyo·we is that it marks past tense, not that it means ‘former’. In (17), for example, the young men are not “formerly smitten”: they are still smitten by the young women at this point in the text. An additional motivation for the appearance of =iyo·we immediately to the right of the verb is surely that doing so places the enclitic in a position similar to the suffix -ehe, which appears in the rightmost slot of verbal suffixes. With forms like the participle in (17), the enclitic =iyo·we is performing the function otherwise associated with the past tense suffix -ehe; it makes sense that the enclitic appears on the right edge of the verb, just as the past tense suffix would. Negative Verbs

A parallel phenomenon to the participle forms in (17) and (18) can be seen with verbs inflected in the negative mode. Such verbs occur in main clauses along with the main clause negative particle a·kwi. Morphologically, however, the negative verbs belong to the conjunct order, so a version containing the -ehe suffix is available, as seen earlier in (10b) and (13). Although it is morphologically possible for negative verbs to be marked for past tense using the suffix -ehe, some instances of negative verbs occur with the =iyo·we enclitic instead: (19) ni·nayo a·kwimeko ke·ko·hi nahi–inowe·ya·niniyo·we. ni·na=iyo a·kwi=meko ke·ko·hi nahi–inowe·-ya·nini=iyo·we. I=for not=emph anything ever–talk.so-1.neg=past ‘Me, I never said anything at all before.’ (W587) (20) a·kwike·hi wi·hni·mihena·ni ine·nemena·niniyo·we. a·kwi=ke·hi

wi·h-ni·mih-ena·ni

ine·nem-ena·nini=iyo·we.

not=moreover fut-cause.to.dance-1>2.aor think.thus.of-1>2.neg=past ‘And I wasn’t intending to have you dance.’ (W504)

|

The Meskwaki Past Tense Enclitic   117

In both (19) and (20) the enclitic appears not in the expected clause-second position but rather attached directly to the verb bearing negative inflection. Again, this positioning may be motivated by the fact that the enclitic is taking the place of the word-final past suffix -ehe. Aorist Conjunct Verbs

The aorist is the most frequently used mode of the conjunct order, appearing on complement clauses, adverbial clauses expressing reason, and main clauses in traditional narratives, among other functions. It is possible to indicate an overt past tense on an aorist conjunct verb by filling the mode suffix slot with the suffix -ehe (cf. Dahlstrom 2021:106–108), but it is also possible to mark past tense on an aorist conjunct verb by using the enclitic =iyo·we, as in (21): (21) e·hme·me·čike·netakiyoke·hmekoyo·we ma·wa·ka·neki e·hawiwa·či, e·h-me·me·čike·net-aki=iyo=ke·hi=meko=iyo·we

ma·wa·ka·n-eki

aor-be.sure.about-3>0.aor=for=moreover=emph=past village-loc e·h-awi-wa·či, aor-be.there-3p.aor ‘She had been sure that they were in the village.’ (W109)

The matrix verb in (21), ‘be sure about’, is inflected in the aorist conjunct because it appears in the narrative portion of a traditional story. The enclitic =iyo·we indicates that the time of the verb is in the past relative to the time of the story. =iyo·we and Irrealis Constructions

In the examples seen so far of aorists, participles, and negative verbs marked with the enclitic =iyo·we, the enclitic indicates past tense. But =iyo·we can also appear in irrealis constructions comparable to (14) and (15), in which a future prefix combines with a past tense morpheme on a conjunct order verb to express unfulfilled intentions or a past modal reading.8 In the following example of unfulfilled intentions, the enclitic =iyo·we appears to be used in place of the more frequently encountered past tense suffix -ehe:

|

118    Amy Dahlstrom

(22) wi·hokima·wihečikiyo·we ic-wi·h-okima·wih-ečiki=iyo·we ic-fut-make.O.chief-X>3p.part.3p=past ‘the ones who were going to be made chiefs’ (W652)

In (22) the enclitic =iyo·we attaches immediately to the right of the participle wi·hokima·wihečiki ‘the ones who will be made chiefs’, taking the place of the expected suffix -ehe. The combination of the past tense enclitic with the future prefix produces the reading of ‘the ones who were going to be made chiefs’. Example (23) shows another instance of the enclitic =iyo·we participating in the unfulfilled intentions construction. Here the enclitic attaches to the first phonological word of the subordinate clause: (23) ke·htenakoči ni·na ketene·nemene wi·hanemimekoyo·we–ne·ne·wona·ke. ke·htena=koči ni·na ke-ene·nem-ene truly=of.course I

2-think.thus.of-1>2.ind

wi·h-anemi–=meko=iyo·we –ne·-ne·w-ena·ke. fut-continue–=emph=past –redup-see-1p>2.aor ‘I really wanted for us to continue seeing you.’ W577F

Example (23) suggests that the past tense/irrealis function of the enclitic =iyo·we holds even if it is not in word-final position on the verb.

Conjunct Order Clauses with Both -ehe and =iyo·we

There are also examples in my corpus where both the -ehe suffix and the =iyo·we enclitic appear together in an irrealis construction. Example (24) contains a participle expressing unfulfilled intentions: (24) apinamekopi a·kwi kehke·netakini wi·hiši·hta·hiteheyo·we. apina=meko=ipi a·kwi kehke·net-akini even=emph=hrsy not know-3>0.neg

|

The Meskwaki Past Tense Enclitic   119

wi·h-iši·hta·hi-tehe=iyo·we. fut-work.so.dim-3.pret.part.obl=past ‘She even forgot what she was going to do, it’s said.’ (W847)

Example (25), on the other hand, shows a past modal construction formed on a participle with the future prefix, the past suffix -ehe, and the past enclitic =iyo·we: (25) “. . . ” wi·hinahinenakohakohiyo·we “. . . ” wi·h-inah-in-enakoha=kohi=iyo·we

fut-redup-say.thus.about-X>21.pret.part=certainly=past

‘ “. . . ” is what people would certainly have always said about us.’ (W514)

In (25) the two morphemes expressing past tense are separated by =kohi ‘certainly’. Not all clauses containing both =iyo·we and a conjunct verb inflected with -ehe exhibit the enclitic attached to the verb. For example, (26), with a negative preterite verb suffixed with -ehe, has the past tense enclitic attached to the first word of the clause: (26) mešeča·hiyo·we a·kwi ka·hkami a·čimoyanehe! meše=ča·hi=iyo·we a·kwi ka·hkami freely=so=past

a·čimo-yanehe!

not in.the.first.place narrate-2.neg.pret

‘But you didn’t say any of this in the first place!’ (W28)

Likewise, (27) contains a verb inflected in the changed preterite paradigm, which is associated with the particle keye·hapa ‘in fact’, often used to introduce surprising information (cf. Dahlstrom 2021:110–111). The past enclitic appears in clause-second position, not attached to the verb: (27) keye·hapake·hiyo·we mahkwa mi·kawiheti·yanehe keye·hapa=ke·hi=iyo·we mahkw-a ic-mi·kawiheti·-yanehe in.fact=moreover=past bear-sg ic-become.friends-2.ch.pret ‘In fact you struck up a friendship with the bear back then.’ (W574)

|

120    Amy Dahlstrom

Overview of the Distribution of the Two Strategies

Table 1 summarizes the distribution of the two possible ways of overtly indicating past tense in Meskwaki as reflected in the corpus under investigation here: using the enclitic =iyo·we alone; using the conjunct order suffix -ehe alone; using =iyo·we alone with a conjunct order verb; and using both =iyo·we and -ehe with a conjunct order verb. As can be seen in Table 1, simple NPs and the main clause constructions listed in the lefthand column permit the use of the enclitic =iyo·we. The other columns show how the various paradigms within the conjunct order may be explicitly marked for past tense. The second column, listing conjunct forms that may express past tense by suffixing -ehe, is the largest. The third column, listing conjunct forms that may express past tense with the enclitic =iyo·we and no suffix -ehe, contains four of the six constructions found in the second column. The last column lists conjunct forms that have been found in the corpus containing both the enclitic =iyo·we and the suffix -ehe. I conjecture that the past tense enclitic =iyo·we may be replacing the conjunct suffix -ehe as an overt marker of past tense. Examples such as (17–20), in which the enclitic attaches directly to the right edge of the verb instead of in the expected clause-second position, strengthen the case for considering the enclitic to be a replacement for a more archaic affixal strategy. Furthermore, I suggest that the forms containing both =iyo·we and -ehe may be an intermediate stage of this development.

TABLE 1. Distribution of =iyo·we and -ehe ONLY = IYO·WE

NPs independent verb potential verb Ø verb

CONJUNCT VERBS WITH -EHE

CONJUNCT WITH = IYO·WE

negative participle future + past aorist changed preterite subjunctive preterite

negative participle future + past aorist

CONJUNCT WITH BOTH

negative participle future + past changed preterite

|

The Meskwaki Past Tense Enclitic   121

Competition between Enclitics and Affixal Strategies in Cree Dialects

As stated above, most tokens of the enclitic =iyo·we in the corpus appear in environments in which the conjunct past tense suffix -ehe cannot be used: independent order verbs, potential verbs, verbless constructions, and adjoined to simple NPs. Examples (17–27), however, in which =iyo·we is used with conjunct order verbs, suggest that the enclitic’s distribution may be expanding, encroaching upon a more archaic affixal strategy of marking past tense. If the enclitic’s use is indeed expanding, this may provide a parallel case to a development seen in the Plains dialect of Cree. Blain and Déchaine (2007) find that evidential functions expressed by inflectional means in Innu, Naskapi, and East Cree are expressed by enclitics in Plains Cree. Table 2 reproduces their findings. Blain and Déchaine point out that Plains Cree speakers also have inflectional strategies available for expressing evidential notions, but the enclitics, such as dubitative êtokwê, seem to be preferred: “In Modern Plains Cree, the particle [êtokwê] occurs with greater frequency and its affixal counterpart is no longer widely used” (2007:278). Their observation is further supported by the following conjecture in Wolfart 1973, quoted in Blain and Déchaine: “But the most interesting question, which has to remain completely open at the present time, is whether there is an observable decline in the frequency of dubitative verb forms which would correlate with the increasing frequency of the non-verbal dubitative marker êtokwê” (Wolfart 1973:44). I suggest that a phenomenon parallel to the Plains Cree replacement of affixal strategies with enclitic particles may be occurring in Meskwaki, where the distribution of enclitic =iyo·we has begun expanding and taking the place of the suffix -ehe. TABLE 2. Indirect evidentials in Cree/Montagnais/Naskapi and Plains Cree (from Blain and

Déchaine (2007:283) PLAINS CREE MONTAGNAIS/ NASKAPI EAST CREE

êtokwê dubitative -tak present -tak present

êsa reportative -shipan nonpresent: past -shipan nonpresent: past

êska nonfactual (is)ka- . . . -ua nonpresent: subjective (is)ka- . . . -shapan-ua nonpresent: nonfactual

|

122    Amy Dahlstrom

Conclusion

This discussion has demonstrated that the primary function of the Meskwaki enclitic =iyo·we is to express past tense in parts of the grammar where the conjunct order suffix -ehe is not available. However, the use of =iyo·we with conjunct order verb forms suggests that the enclitic may be gaining ground against the inflectional suffix -ehe in conjunct environments, often appearing at the right edge of the verb in a position similar to that of the suffix -ehe. Parallel developments among the dialects of Cree suggest that this expansion of an enclitic’s distribution is a well-motivated morphosyntactic change for an Algonquian language. Notes 1. I thank the audience at the 53rd Algonquian Conference for helpful comments, and especially the two anonymous reviewers for their suggestions. 2. The primary source for the corpus is the 1,110-page manuscript by Alfred Kiyana (1913), supplemented with additional material by Kiyana and other Meskwaki authors (e.g., the anonymous author of the text in Goddard 2006). 3. The term ‘irrealis’ has been criticized by some (e.g., Bybee 1998) as not useful in studies of typology. Nevertheless, I employ it here as a term which covers both contrary-to-fact constructions and evidential notions such as speaker uncertainty. It is also widely used in other typological work (cf. Steele 1975; Larreya 2003). 4. 1p = first person exclusive plural, 21 = first person inclusive plural, 3′ = obviative, 0 = inanimate, absent = absentative demonstrative, anim = animate, aor = aorist prefix; aorist conjunct inflection, aor.pret = aorist preterite, ch.pret = changed preterite, cont = contrastive, dim = diminutive, emph = emphatic, fut = future, hrsy = hearsay, ic = initial change (ablaut rule), ind = independent indicative, loc = locative case, neg = negative inflection, neg.pret = negative preterite, O = (first) object, O2 = second object, obl = oblique head of relative clause, obv = obviative, part = participle, perf = perfective, pl = plural, pot = potential, pret.part = preterite participle, redup = reduplication, sg = singular, subjnct = subjunctive, subjnct.pret = subjunctive preterite, X = unspecified subject. Subject and object features in verb inflection are separated by > and are followed by identification of the verbal paradigm. The head of a relative clause is identified following the label part (participle). An en-dash (–) indicates a boundary between a preverb and the remainder of the verb. Vowel length is marked by a raised dot. Examples cited as W are from Kiyana 1913; the following number refers to the page

|

The Meskwaki Past Tense Enclitic   123

number of the Kiyana 1913 text. 5. Note that this construction is not typically used for reference to a deceased person, unlike in Cree, which uses the preterit suffix -pan for this purpose (Wolfart 1973:31). To refer to a deceased person in Meskwaki, an absentative demonstrative is employed (Goddard 2003:39); an example may be seen in (13). 6. The suffix -ehe has an allomorph -eha that appears after -(y)akw first person inclusive plural and -(y)e·kw second person plural; an example may be seen in (25). 7. Other examples in this paper do not have glosses of the individual components of the inflectional suffixes, because in many instances the subject and object values can only be identified by the combination of an inflectional prefix plus the inflectional suffixes, and the identification of a particular verbal paradigm may also depend on a combination of a prefix or initial change applying to the left edge of the verb plus the particular suffix filling the mode suffix slot. See Dahlstrom 2000 for discussion of such discontinuous morphemes in the Meskwaki inflectional system. 8. See also the irrealis reading of =iyo·we used with a main clause verb inflected in the potential paradigm in example (7).

References Blain, Eleanor M. and Rose-Marie Déchaine. 2007. Evidential Types: Evidence from Cree Dialects. International Journal of American Linguistics 73:257–291. Bybee, Joan L. 1998. “Irrealis” as a Grammatical Category. Anthropological Linguistics 40(2):257–271. Costa, David J. 2003. The Miami-Illinois Language. Lincoln: University of Nebraska Press. Dahlstrom, Amy. 2000. Morphosyntactic Mismatches in Algonquian: Affixal Predicates and Discontinuous Verbs. Proceedings from the Panels of the Chicago Linguistic Society’s Thirty-Sixth Meeting, ed. by Arika Okrent and John Boyle, pp. 63–87. Chicago: Chicago Linguistic Society. Dahlstrom, Amy. 2021. The Historical Semantics of Past Tense and Irrealis in Meskwaki. Webs of Relationship and Words from Long Ago: A Festschrift Presented to Ives Goddard on the Occasion of His 80th Birthday, ed. by Lucy Thomason, David J. Costa, and Amy Dahlstrom, pp. 99–117. Petoskey, MI: Mundart Press. Drapeau, Lynn. 2014. Grammaire de la langue innue. Québec: Presses de l’Université du Québec. Fleischman, Suzanne. 1989. Temporal Distance: A Basic Linguistic Metaphor. Studies in Language 13:1–50.

|

124    Amy Dahlstrom

Goddard, Ives. 2003. The Demonstrative Pronouns of Algonquian. Essays in Algonquian, Catawban, and Siouan Linguistics in Memory of Frank T. Siebert Jr., ed. by Blair A. Rudes and David J. Costa, pp. 37–102. Winnipeg: Algonquian and Iroquoian Linguistics. Goddard, Ives. 2006. The Autobiography of a Meskwaki Woman. Winnipeg: Algonquian and Iroquoian Linguistics, Memoir 18. Goddard, Ives. 2015. The Twenty-Nine Enclitics of Meskwaki. Papers of the 43rd Algonquian Conference, ed. by Monica Macaulay and J. Randolph Valentine, pp. 72–116. Albany, NY: SUNY Press. Hogeweg, Lotte. 2009. What’s So Unreal about the Past: Past Tense and Counterfactuals. Studies on English Modality: In Honour of Frank R. Palmer, ed. by Anastasios Tsangalidis and Roberta Facchinetti, pp. 181–208. Bern: Peter Lang. James, Deborah. 1982. Past Tense and the Hypothetical: A Cross-linguistic Study. Studies in Language 6:375–403. James, Deborah. 1991. Preterit Forms in Moose Cree as Markers of Tense, Aspect, and Evidentiality. International Journal of American Linguistics 57:281–297. James, Deborah, Sandra Clarke, and Marguerite MacKenzie. 2001. The Encoding of Information Source in Algonquian: Evidentials in Cree/Montagnais/Naskapi. International Journal of American Linguistics 67:229–263. Kiyana, Alfred. 1913. wisakea osani okyeni osimeani okomeseani. [Wisahkeha, his father, his mother, his younger brother, his grandmother.] Manuscript 2958-a in National Anthropological Archives, Smithsonian Institution, Washington, DC. Larreya, Paul. 2003. Irrealis, Past Time Reference and Modality. Modality in Contemporary English, ed. by Roberta Facchinetti, Manfred Krug, and Frank Palmer, pp. 21–45. Berlin: Mouton de Gruyter. Michelson, Truman. 1927. Contributions to Fox Ethnology. Bureau of American Ethnology Bulletin 85. Washington, DC: Government Printing Office. Steele, Susan. 1975. Past and Irrealis: Just What Does it All Mean? International Journal of American Linguistics 41:200–217. Wolfart, H. Christoph. 1973. Plains Cree: A Grammatical Study. Transactions of the American Philosophical Society, n.s., vol.63, part 5. Philadelphia.

Ezhi-Aanikanootamaang Ogimaans The Little Prince Translated into Ojibwemowin Jaymee Dhein, Angela Mesic, Margaret Noodin, and Susan Wade

T

he little prince, in Antoine de Saint-Exupéry’s philosophical novel for all ages, tells readers: “On ne voit bien qu’avec le cœur. L’essentiel est invisible pour les yeux.”1 In English this line is commonly translated as: “It is only with the heart that one can see rightly; what is essential is invisible to the eye.” Translating this line into Ojibwemowin moves it away from the belief that a heart represents emotion and our eyes lead us to logic. To express the way the unseen has value requires concepts to become verbs: giwii-nisidawendaan apii boonigidetaazoyan bwaa-waabanjigaadeg maamawigichi-inendaagwag or, more literally, “you understand when you let go of everything, that what is unable to be seen is most important.” Translation requires letting go of the comfort and perceived precision experienced by using only one language, and looking instead for what is not easily seen in comparisons and cultural connotations. As Moyes mentions, any action of translation between colonial and Indigenous languages forces the translators to “recognize the imbrication of translation, power, and the production of knowledge” (2018:73). Le Petit Prince was chosen for this project not as an act of resistance but an act of inclusion because it contains elements of oral narratives, place-based identity, and complex relationships between animals, humans, and plants.

|

   125

|

126    Dhein et al.

Three versions of the same story are discussed in this paper, and we use three different titles to distinguish them: the original Le Petit Prince written in French (de Saint-Exupéry 1986), The Little Prince translation in English (de Saint-Exupéry 2000), and the Ojibwemowin translation, Ogimaans (de Saint-Exupéry 2021). As translators of the Ojibwemowin version, the authors of this paper trace the process of translating an enduring novella from one of the world’s dominant, colonizing languages into a minority Indigenous language currently being revitalized in North America. While theoretical frameworks are worthy of discussion, this paper primarily serves as a documentation of pedagogical practice. Lessons learned by the translators in working on this project can assist future translators doing similar work, and contribute to the study of translation and literature. Mii omaa mazina’iganing inga-dibaadaamin ezhi-aanikanootamaang mii aadizookeyaang: This paper is the story of making a new version of an old story.

History of the Text

Ogimaans, the ‘little king, little chief, or little prince’, is the first translation of Saint-Exupéry’s classic parable into an Indigenous language of northern North America, the region now dominated by the United States and Canada. Le Petit Prince has been translated into 445 languages and dialects and several more translations are currently underway. Together the translations offer a kaleidoscopic view of how one being connects with others and recognizes their place in the universe. This particular modern novella was selected for translation into Ojibwemowin because it is an aadizookaan, a teaching story, intended to help the reader along their own journey. As more students gain proficiency in Ojibwemowin, there is an increased need for monolingual texts. Numerous monolingual books in Ojibwemowin have been produced by elders and teachers today (DeBungie 2014; Treuer and Sullivan 2021). Many of them focus, as they should, on characters, settings, and plots that are culturally and geographically familiar to Ojibwe readers. The translation of Le Petit Prince is intended to add further dimension to the corpus of monolingual Ojibwemowin texts. It was originally a narrative composed for postmodern readers shaken by the implications of the First and Second World Wars. However, for Indigenous readers living with the continual impact of postcolonial cultural dissonance, and humans watching world wars continue, it is a text worthy of scrutiny.

|

The Little Prince Translated into Ojibwemowin   127

Any translation should be approached within a historical framework. It is crucial that the translator knows the background of the author and the period in which it was created. Antoine de Saint-Exupéry wrote Le Petit Prince while living in New York in 1942, partly as a love letter to his wife, partly as a cautionary tale for humanity, and partly as a fable set during the innocence of his own childhood. He filled his tale about a young space traveler with memories of his own adventures as a pilot who flew over deserts and his boyhood experiences growing up in SaintMaurice-de-Rémens in the early 1900s. As the story unfolds, he exposes his own vulnerabilities through the characters and questions the meaning of existence. Saint-Exupéry was a pioneer of aviation. As a commercial mail pilot, he worked for Aeropostale and had mail routes in Europe, Africa, and South America. He also flew air races attempting to break speed records. By 1927, Saint-Exupéry was a manager of an airfield in Cape Juby, now Tarfaya, in western Sahara. This was a Spanish colonial territory, which bordered colonial French Sahara. In the Sahara he was introduced to the fennec, or desert fox, who was no doubt an inspiration for the fox in The Little Prince. In 1935 he crashed, along with his navigator, in the Libyan Desert. They, like Saint-Exupéry’s character, the narrator in Le Petit Prince, had very little food or fluids. Walking for four days, both men started to hallucinate and were finally rescued by a Bedouin. Flying many times during his career over unpopulated areas, Saint-Exupéry had the opportunity to see the stars clearly. Many national astronomical societies existed by 1900, and in 1930 Pluto was discovered causing a renewed global interest in the earth’s position in the universe. Saint-Exupéry wrote to his mother that he pondered the human experience, not only from the perspective of the vast heavens he flew in, or the oceans he flew over, but also from his childhood room where he was often alone with his thoughts (Webster 1993:243). World War II began in September 1939 when Nazi Germany invaded Poland. However, the precursors to World War II are found in the treaties that ended World War I, and in European politics in the years between the two world wars. Countries such as England, France, Spain, and Germany expanded their colonial territories around the globe, including North and South America, and the African continent. World War II was a continuation of the assertion of a few European languages, cultures, and economies over many other nations, including many Indigenous people and lands. In May 1940 when German forces invaded France, Saint-Exupéry went into exile and moved to New York, often giving talks trying to convince the

|

128    Dhein et al.

United States to become more active in preventing the dominance of the Nazi regime. He spent a little over two years in New York and then in April 1943 he was approved to return to battle and assigned to fly with the Free French Air Force on the side of the Allies. A year later, he took off on an unarmed P-38 Lightning plane, on a reconnaissance mission from an airbase in Corsica but never returned. His plane was finally recovered near Marseille, France, in 2000. Saint-Exupéry wrote Le Petit Prince the year before his death, and it was published posthumously. Drawings of the little figure who became the main character can clearly be seen in letters and manuscripts to a friend who suggested he write a children’s book. Written in French, Le Petit Prince was published in 1943 in both French and English, with Katherine Woods creating the first English translation. Although not without its own translation errors, this version was the standard for over fifty years until 1995 when three new English versions were published. Le Petit Prince was banned by the Vichy Regime because Saint-Exupéry had flown for the Free French Airforce. However, after the liberation of France it was republished in 1946 and has since become a revered classic.

Translation into Ojibwemowin

The translation of Le Petit Prince into Ojibwemowin began in 2008 when Dr. Walter Sauer from Edition Tintenfass reached out to Margaret Noodin. Edition Tintenfass, based in Germany, and the Jean-Marc Probst Foundation for the Little Prince, based in Switzerland, have worked for many years to support linguistic diversity and global human relations through hundreds of translations of Le Petit Prince. There were many barriers to completing the task of translating Le Petit Prince into the southwestern dialect of Ojibwemowin, an Anishinaabe language in the Algonquian linguistic family. There were no examples of similar modern novellas being translated. Not many students and elders were interested in the art of literary translation. Beginning language students did not have enough background to help with the monumental task of translating this complex literary piece. Finally, in 2020, through the Electa Quinney Institute for American Indian Education (EQI) at the University of Wisconsin–Milwaukee (UWM), Isabel Bader provided the funding to finish the project. The goal was to produce a monolingual text for the second-year students on the UWM campus and throughout Anishinaabewakiing. During the decade that passed since the project began, speaking and writing abilities in

|

The Little Prince Translated into Ojibwemowin   129

Ojibwemowin had increased at UWM, and several current and former teachers were hired to assist with the project. The translation team was led by Margaret Noodin and support was provided by Angela Mesic, a first-year Ojibwemowin teacher at UWM. Additional support was provided by Susan Wade, a former first-year Ojibwemowin teacher at Michigan State University (MSU), and Michael Zimmerman Jr., a former instructor of Ojibwemowin at the Indian Community School and current Potawatomi-language instructor at UWM. Jayme Dhein read the final product as a student of Ojibwemowin and Translation Studies at UWM. The project was an experimental literary translation and together the team allowed the story to be reshaped by the principles of Ojibwe style and epistemology. After nine months, the team finished the translation, which became the 134th edition of Le Petit Prince published by Edition Tintenfass for the Jean-Marc Probst Foundation. To date, there are 6,233 editions in 496 different languages and dialects. The foundation provided the translation team with one thousand free copies to give away as a teaching tool. MSU Press has the North American rights to the translation and maintains the book for sale as a title in their catalog, providing readers with a unique intersection of language revitalization and education. While recognizing that the use of Ojibwemowin within the Anishinaabe diaspora spans 143 communities across the Midwest, the translation uses the Roman alphabet and the Fiero Double Vowel spelling system for Ojibwemowin, which is the most common writing system in the southern Great Lakes region. The Fiero Double Vowel system utilizes Roman letters and letter combinations to represent Ojibwe sounds (Meuers 2012). It was not intended to be a prescriptive orthographic choice but was merely the most practical for the teachers who served as translators for this edition and their primary audience of readers. It may be the catalyst to establishing one or more formalized, written variants of this Indigenous language, as has been the case with other languages of the world—Norwegian’s Bokmål ‘book tongue’ and Arabic’s Literary or Classical Arabic come to mind. However, it is important to note that multiple variants would be supported by Edition Tintenfass, which has a proclivity for producing a wide array of dialect-based translations of Le Petit Prince. To complete the translation the team referenced both the English and French first edition texts and often consulted the original French edition to confirm translation choices related to morphology, rhythm, and style. Each member of the team had a different history of encountering the text. Noodin first read the French version in 1980 and also studied it during a university semester in Nevers, France. Wade read the English version in 1987 as an English speaker in Montreal, Canada,

|

130    Dhein et al.

where she also studied French. Zimmerman and Mesic first read the book in English as part of their work on this project. Dhein read a German version in 2010 prior to learning Ojibwemowin. Knowing the text would be read by teachers, students, and linguists in the future, the team worked on multiple documents in a shared drive, which allowed comments, questions, and updates to be preserved. The two main documents in use were the translation and style guide. The translation of the twenty-seven chapters was formatted as a three-line document with English as the base language because the majority of Ojibwemowin students and teachers speak English as their first language. It is also the language used by all the translators, serving as the means of collaboration. The second line and third lines are, respectively, the Ojibwemowin translation and an English gloss of the Ojibwemowin to serve as a reference tool for morphology and grammar. The gloss reveals the ways in which ideas changed and words were added, or left out, as the text moved from English and French into Ojibwemowin. To ensure multiple edits for all passages, each translator was assigned a set of chapters to draft and a set of chapters to edit. Then, when the text was complete, knowing the publisher would not have copyeditors for the Ojibwemowin, Zimmerman and Noodin read all chapters one more time to catch any final errors. The translation style guide2 was created to ensure consistency and archive editorial decisions, such as the use of punctuation, grammar, and overall formatting. The intent was not to set standards, but to document the questions and resolutions as they arose. One key concept that was clarified in the style guide was the pronoun perspective and use of obviation. Ogimaans, the little prince himself, was considered the third person, which meant all other persons not first or second required obviation markers. This is clearly the example set by first speakers of Ojibwemowin and the task of carrying it throughout a novel was another one of the challenges faced by the team. In the end it provides future readers and students with an extended narrative example of obviative endings on nouns as well as indicative, dubitative, and preterit verbs in both independent and conjunct forms. A third important shared document, which was created during the translation process, was a record of the lessons learned during this particular translation. This document provided more in-depth analysis of issues that arose while translating from a subject-centered language to one that is verb centered. The lessons document includes more detailed examples of Ojibwemowin aesthetics than the style guide; for example, when and how often to repeat pronouns, specific patterns of discourse markers used so that the chapters do not have jarring inconsistencies by

|

The Little Prince Translated into Ojibwemowin   131

different translators, and instances of replacing Indo-European adverbs used for description and emphasis with preverbs or emphatic particles, which are more natural in Ojibwemowin. The style guide document also includes a short glossary of neologisms and the logic behind them. When there were no precedents for the expression of an action, idea, or item, the practice of creating a lexical item through description of core actions or features was used, for example, wayaazakonenjiged (‘lamplighter’, literally ‘someone who lights the streetlamp’) and netaaminikweshkijin (‘tippler’, literally ‘someone who has become skilled in drinking regularly’). The translation style guide was a living document and evolved as each team member made discoveries in the translation process and added their knowledge to the record. The style guide and glossary of words created for the translation are published on Ojibwe.net where a page dedicated to the text allows readers to learn more about the history of the book, interesting vocabulary included in the translation, famous quotes, and grammar choices made by the translators (see https://ojibwe. net/ogimaans-the-little-prince-book/). Also published on Ojibwe.net are audio segments so readers can hear several passages, including the most iconic quotes from the book, and consider the ways the Ojibwemowin translation differs from the original French and English, as can be seen in examples (1)–(4) below.3 (1) From the King: “Awashime zanagad

wii-dibaakonidizoyan apii dibaakonadwaa.

much more it is difficult fut-you judge yourself when you judge them Giishpin gidaa-dibaakonidiz if

gwayak, geget igo

you may-you judge yourself correctly surely emph

nibwaakaayan aapiji.” you are wise very “C’est le plus difficile. Il est bien plus difficile de se juger soi-même que de juger autrui. Si tu réussis à bien te juger, c’est que tu es un véritable sage.” “It is much more difficult to judge oneself than to judge others. If you succeed in judging yourself rightly, then you are indeed a man of true wisdom.”

|

132    Dhein et al.

(2) From the Aviator: “Gakina netaawigijig gii-oniijaaniziwag all

aanawi aanind

grown ones past-they were children although some

minjimendamowaad geyaabi.” remember it

still

“Toutes les grandes personnes ont d’abord été des enfants. (Mais peu d’entre elles s’en souviennent.)” “All grown-ups were once children—although few of them remember it.” (3) From the Little Prince: “Noongom giwii-wiindamoon gaadooyaan, now

wenipanag:

you will-I tell you what I keep secret simply

Mii eta dash giwii-nisidawendaan

apii boonigidetaazoyan

So only then you will-you understand.it when you.forgive/release bwaa-waabanjigaadeg maamawigichi-inendaagwag.” not able-without.visibility what.is.altogether-most-important “Voici mon secret. Il est très simple : on ne voit bien qu’avec le coeur. L’essentiel est invisible pour les yeux.” “And now here is my secret, a very simple secret: It is only with the heart that one can see rightly; what is essential is invisible to the eye.” (4) From the Rose: “Ahaw, indaa-zhiibendaan bi-ayaawaad ok

niizh gemaa niswi

I must-I endure it here-they exist two or

three

wemakwaayaanenhyag giishpin wii-ani-nakweshkawagwaa memengwaag.” small caterpillars

if

fut-on the way-I meet others butterflies

“Il faut bien que je supporte deux ou trois chenilles si je veux connaître les papillons.” “Well, I must endure the presence of two or three caterpillars if I wish to become acquainted with the butterflies.”

Although the spirit of the story remains the same, as The Little Prince becomes Ogimaans, numerous small changes demonstrate the way the greatest teaching stories are aadizookanag, animate narratives that do much more than simply tell the facts. From changes in phrasing, description, and vocabulary, to changes in hyphenation, conjugation, and length, the translated story becomes a descendant

|

The Little Prince Translated into Ojibwemowin   133

of the original, maintaining markers of the first voice, but including traces of other voices and linguistic frameworks.

Lessons Learned

When undertaking a translation into an endangered Indigenous language, it is important to remember and consider the potential long-lasting impacts the translated text may have. The impact of the Ogimaans translation from dominant Indo-European languages into a minority Indigenous language is threefold: the expansion of the corpus available for curriculum, the creation of new lexical items, and the contribution to global translation studies. When initially deciding to translate Le Petit Prince into Ojibwemowin, one of the motivating factors was the opportunity to expand the availability of language learning materials, primarily for adults. While there are child-oriented materials (in style and complexity of linguistic elements) already available, ensuring adults as well as children have access to content suitable to their linguistic abilities is crucial for promoting continued and lasting language acquisition. By creating the linguistically complex, long-form content, advanced learners will be able to recognize and reinforce what they have heard in conversations and learned in formal classes. Ogimaans is useful for taking theoretical concepts and structured lessons from the classroom and giving them practical applications. Beyond the scope of adult learners, Ogimaans will become an option for young learners who age out of first readers and picture books. For adult students of Anishinaabe literature, the text offers an example of Anishinaabe literary aesthetics appearing in a story with global appeal. This opportunity to provide a text that could be used to hypothesize comparative literary theories about Indigenous culture and philosophy presented as a narrative was another reason the translators were initially interested in the task. The collective sense upon completion was one of awe and inspiration regarding the value of detailed interlinguistic analysis of language and creative expression. Rarely have Indigenous languages been included in the study of world literature other than through the filter of translation. A second impact of this translation is the production of new lexical items for modern and culturally unfamiliar terms. Because the novel was written after the attempted linguicide of Ojibwemowin, and many other Indigenous languages, there

|

134    Dhein et al.

were no texts from the middle of the twentieth century to use as resources for capitalist, postindustrial, political, and philosophical terminology. One might imagine that a number of code-talkers, including the Anishinaabe veterans of World War I and World War II, may have created these terms but they were not preserved. The translators were also aware that the Anishinaabe approach to neologisms is highly individualized and multiple terms for the same idea or item are frequently accepted as long as there is morphemic fidelity. When the translators encountered a word or concept for which there existed no established Ojibwemowin equivalent, they had to come to an agreement on how to handle these words. Whenever possible, endonyms, or their equivalent, were used. For example, the little prince and the pilot talk about the Sahara Desert in Chapter 25 (Mesic et al. 2020:140). Recognizing “Sahara” as itself an Indigenous word from Arabic, the translators decided to preserve its foreignness, italicizing the term to indicate it as such. In other cases, an entirely new Ojibwemowin word was created to supplant the French and English. Of particular interest are the words invented to name the characters the little prince meets on other planets. For example, in Chapter 12, he meets the buveur or ‘tippler’, terms that both have connotations of excess without complete incapacitation. To reflect the essence of the character, the translators built netaaminikweshkijid, a word that expresses the idea of a person who has practice in habitual drinking. In the center of the word is minikweshki ‘to drink habitually’, which is itself a variation of the animate intransitive verb minikwe ‘to drink’, with the pejorative suffix -ish. Added to this is the preverb nitaa-, which is used to indicate an action someone has done many times, even becoming skilled in it. Use of the word as a proper noun dictates an initial vowel change from /i/ to /e/. The nominalization is signaled by the addition of -d. A loose back-translation may read something like ‘someone who has become skilled in drinking regularly’. Using this pattern, all the other characters who were known by their occupation were given similar names in Ojibwemowin. Not only does the creation of these terms present learners with the lexical material needed to talk about a wide array of everyday ideas and topics, but it also invites discourse and provokes thought about the efficacy of these terms, perhaps encouraging others to suggest other options. Third, the translation of Ogimaans can lead to numerous hypotheses in the study of translation. One theory, often discussed in early stages in the formal education of budding translators, warrants close examination. As a theorist, Friedrich Schleiermacher developed his doctrines of translation and philosophy in Germany during the late eighteenth and early nineteenth centuries. Schleiermacher was a

|

The Little Prince Translated into Ojibwemowin   135

proponent of a particular method in translation, advocating for “[leaving] the writer in peace as much as possible and [moving] the reader toward him” (2012:49). This is known as foreignization, a strategy of translation that seeks to leave intact that which linguistically and culturally signifies to the reader the “otherness” of the original text while preserving its readability and intelligibility. Schleiermacher also identified an opposing strategy, known as domestication, which seeks the opposite, instead “[moving] the writer toward [the reader]” (2012:49), thereby masking the author’s foreignness. As stated, Schleiermacher is strongly in favor of foreignization, touting it as the “true . . . goal of translation as a whole” (62) and claiming that “any attempt to combine [the two strategies] being certain to produce a highly unreliable result [in which] writer and reader might miss each other completely” (49). Ogimaans demonstrates that there is room for both strategies in a single translation. References to known worldly places and things (e.g., the Sahara Desert) are left unaltered and foreign. However, the translators of Ogimaans endeavored to “domesticate” the story and its lessons into an Anishinaabe worldview, adjusting the framing of the lessons. Terms and expressions were adapted to better suit the translating culture and language. Additionally, as previously discussed, the translators made efforts to incorporate new words into the lexicon, adopting words where possible and inventing new terms where necessary. In these ways, they have followed and supported Schleiermacher’s ideas that language “can most vigorously flourish and develop its own strength . . . through extensive contact with the foreign” (2012:62), while simultaneously challenging his notions of foreignization as an ideal framework. Although Schleiermacher’s writings are arguably imperialistic and nationalistic in his framing of Germans and the German language to be well suited to “unite all the jewels of foreign science and art together” (62), he says something else of note. Later, he writes not about the benefits of translating into German for the world, but of the benefits of translating for the language itself. He recognizes that what is “beautiful and strong” in his language was “developed, or restored from oblivion, only through translation” (62; emphasis added). Perhaps Schleiermacher is underscoring foreignization as a tool for revitalization. Even earlier in the text, he writes, “languages are not invented . . . rather they are discovered [through] science and art” (51). We can see the same thing at play in Ogimaans with Ojibwemowin. Given the examples of using familiar words and phrases to mirror unfamiliar metaphors, and creating neologisms as needed, we see how translation can introduce

|

136    Dhein et al.

the translating language to new terms and ideas. When this happens, the translator must decide how to proceed, choosing either domestication, presenting the new concept in a culturally relevant way, or foreignization, preserving that which is novel in favor of expanding the translating language’s capacity for expression. Through Schleiermacher’s own writings, we can see foreignization recognized as a versatile and valuable tool in language revitalization endeavors. However, the translators have also strayed from this method, where possible, to insert elements from the receiving culture into the story, to make it more relevant to Anishinaabe readers. The translators of Ogimaans are not the first to challenge colonial ideals established long ago. In fact, their work aligns with that of several postcolonial theorists. Translator David Bellos writes about the differences between “translating UP” and “translating DOWN,” describing them as translating into “a language of greater prestige” and translating into a language “with a smaller audience . . . with less cultural, economic, or religious prestige” (2011:168). However, this is not said to disparage either practice, but to recognize the differences in impact and implication the act of translation has on the receiving culture. When “translating DOWN,” to use Bellos’s term, as the Ogimaans translators have done, special care must be taken when encountering concepts, words, or references to physical items that do not exist at all or as such in the receiving culture. Languages serve their speakers. They have vocabulary to express that which is useful, common, and meaningful to those who use them. As we have discussed, the Ogimaans translators, where appropriate, took the opportunity to propose solutions to various translational challenges, such as the nonexistence of a term or concept in Ojibwemowin. Here, the translators invented words, which align with the conventions of Anishinaabe grammar, or swapped in more culturally relevant terms. The only things kept intact from the source are the names of global beings and locations, such as the Sahara Desert. In doing so, the translators have done what others have done before them. Some translations of the Bible, for example, have changed a fig tree to a banana tree, because the receiving culture has no concept of a fig tree (Bellos 2011:174). The importance of the symbolism is retained without clouding the message through a lengthy explanation. While many Anishinaabe readers of Ogimaans may readily identify with a Western way of expressing certain ideas or life lesson, by adapting the messages within Le Petit Prince, the translators seek to (re)introduce readers to an Anishinaabe way of seeing universal truths. The lessons presented in the story of Le Petit Prince, and which are so intrinsic to its cultural and literary value, lend themselves to being reimagined through an

|

The Little Prince Translated into Ojibwemowin   137

Anishinaabe lens using Ojibwemowin as the language. Schleiermacher himself argues the aim of translation is to “[preserve] the unity of the impression made by the work” (2012:48). When translating into a dominant or colonizing language, retaining all the peculiarities and novel cultural elements of the original may be seen as ideal or even virtuous. When undertaking the opposite endeavor, there may be an even stronger imperative to domesticate these features. In terms of language revitalization and cultural reclamation, such decisions may be crucial to the success of these efforts. Indigenous peoples still face the challenges of “ongoing legacies of colonization, ethnocide, and linguicide” (McCarty and Lee 2014:103). As such, in the “fight for cultural and linguistic survival” (McCarty and Lee 2014:103), the modification of certain literary elements to be more culturally relevant should be understandable and encouraged. The Ogimaans translators demonstrate this by completely altering a metaphor in the same scene with the fox from Chapter 21, with which this paper began (see example (3) above). In English, the line reads, “It is only with the heart that one can see rightly; what is essential is invisible to the eye” (Mesic et al. 2020:126). What the fox is telling the little prince and the reader is that material things are not important. The greatest things in life are those we cannot touch but instead can only feel. However, this idea expressed as such does not fit into Ojibwe notions of discussing emotional understanding: the metaphors are entirely changed in Ogimaans. There is no comparison between seeing with one’s heart or with one’s eyes. There is instead a reminder that by letting go of what is material one can best understand emotions and ideas. Translations of this nature, by and for Indigenous people, incorporating their own cultural perspectives and teachings, may be key in furthering the reclamation of these “traditional cultural values” (Reyhner 2017:6). This is an example of what is known as Indigenous education sovereignty, an element of general tribal sovereignty (McCarty and Lee 2014:101, 103). As McCarty and Lee point out, “language is vital to cultural continuity and community sustainability because it embodies both everyday and sacred knowledge” (2014:109). With long-form literary translations such as this, readers are introduced to a wide range of vocabulary. Topics are present that may not be discussed in a classroom or in the home, like how to prevent a baobab infestation. However, they also teach readers relevant vocabulary for useful life lessons, like the importance of discipline. By taking these lessons and applying an Indigenous lens to them, the translators have taken a step toward reclaiming Anishinaabe narrative notions. By including them in this book, which is so very well suited for adults and children alike, the translators have ensured these messages

|

138    Dhein et al.

and the Indigenous way of understanding and communicating are preserved for generations to come.

Conclusion

Ogimaans, and all his relations, are part of the vast network of linguistic systems that have encoded human knowledge since humans began using sound to represent thought. This paper has offered a glimpse of its popularity, which was never witnessed by the author. It has also documented the process of translating a lengthy, monolingual work of literature into an endangered Indigenous language and attempted to explain the value of such work. The translation team recently met with another teacher working to translate Le Petit Prince into an Algonquian language. Yvette Mollen, a professor at Université de Montréal, is working on an Innu version of the classic tale. She commented that some of the methodology used for the Anishinaabemowin version, such as the three-line translation and the creation of a style guide, was useful to her project. And then the group wandered into a discussion of la forêt vierge, “the primeval forest,” and how these gestures toward pure untouched virginity, something from the primus (first) aevum (age), are best described in an Algonquian language. In Le Petit Prince, the narrator remembers: Lorsque j’avais six ans j’ai vu, une fois, une magnifique image, dans un livre sur la Forêt Vierge qui s’appelait Histoires Vécues.

In English, this became: Once when I was six years old I saw a magnificent picture in a book, called True Stories from Nature, about the primeval forest.

In Ojibwemowin, the boy recalled a fancy picture in a book about the truth that is found in a community of elder trees. Ingoding apii gaa-ingodwaaso-biboonigiziyaan chitwaa mazinichigan gaa-waabandamaan mazina’iganing, gaa-izhinikaadeg Debwemigad Gaa-Ezhiwebag Chi-Gete-Megwaayakong.

|

The Little Prince Translated into Ojibwemowin   139

In Innu, Mollen noted, the phrase leaned toward a “une forêt qui n’a pas été détruit,” ‘a forest that has not yet been destroyed’, because as a translator she was reminded, “les forêts d’aujourd’hui sont habitées, déboisées, détruites par les mines . . . une forêt vierge est tout le contraire” (‘today’s forests are inhabited, deforested, destroyed by mines’ . . . a virgin forest is the opposite of that’). Kutuasht ka tatupipuneshian nuapamati, peikuau, meshta-minushishit akunikanit/ kaiakunakanit anite mashinaikanit uauitakanu neme minashkuau nasht eka ka pikunikanit assi, ishinikatepan Tshitshue tipatshimuna ka ishpaniti.

This attention to shifts in perspective illustrates how multifaceted the truth can be. Nature is pure and feminine, the woods are where life began ages ago, trees are elders who hold the truth, and we must attend to the destruction of woods capable of sustaining life on the planet. As this paper began, it ends with a reminder to look for that which cannot be seen with the eyes, and that which cannot be understood without all the global metaphors for life and learning. Notes 1. Authors are listed alphabetically. 2. Available at https://ojibwe.net/wp-content/uploads/2021/02/Ogimaans-Style-Sheet.pdf. 3. Abbreviations used: emph = emphasis, fut = future.

References Bellos, David. 2011. Is That a Fish in Your Ear? New York: Faber and Faber. de Saint-Exupéry, Antoine. 1943. The Little Prince. Translated by Katherine Woods. New York: Harcourt Brace & Co. de Saint-Exupéry, Antoine. 1986. Le Petit Prince. Stuttgart: Klett. de Saint-Exupéry, Antoine. 2000. The Little Prince. Translated by Richard Howard. New York: Mariner Books. de Saint-Exupéry, Antoine. 2021. Ogimaans. Translated by Angela Mesic, Margaret Noodin, Susan Wade, and Michael Zimmerman Jr. Neckarsteinach, Germany: Edition Tintenfass. DeBungie, Rosemarie, Anna Gibbs, Nancy Jones, Gordon Jourdain, Marlene Stately, Eugene Stillday, Rose Tainter, Wesley Ballinger, Lucia Bonacci, Dustin Burnette, Lisa LaRonge, John Nichols, Keller Paap, Michael Sullivan, and Anton Treuer. 2014. Wiijikiiwending.

|

140    Dhein et al.

Minneapolis: Wiigwaas Press. McCarty, Teresa L., and Tiffany S. Lee. 2014. Critical Culturally Sustaining/Revitalizing Pedagogy and Indigenous Education Sovereignty. Harvard Educational Review 84(1):101– 124. doi:10.17763/haer.84.1.q83746nl5pj34216. Mesic, Angela, Margaret Noodin, Susan Wade, and Michael Zimmerman Jr. 2020. Ogimaans— Three Line Version. Unpublished manuscript. Mesic, Angela, Margaret Noodin, Susan Wade, and Michael Zimmerman Jr. 2020. Style Guide for Ogimaans (The Little Prince). https://ojibwe.net/ogimaans-the-little-prince-book/. Meuers, Michael. 2012. Ojibwemowin Writing Systems. Red Lake Nation News, February 20, 2012. https://www.redlakenationnews.com/story/2012/02/20/news/ojibwemowinwriting-systems/022020120835545495863.html. Moyes, Lianne. 2018. From One Colonial Language to Another: Translating Natasha Kanapé Fontainte’s Mes lame de tannage. TranscUlturalAl 10(1):64–82. Ogimaans (The Little Prince) Book. 2021. August 30, 2021. https://ojibwe.net/ogimaans-thelittle-prince-book/. Reyhner, Jon. 2017. Affirming Identity: The Role of Language and Culture in American Indian Education. Cogent Education 4(1). doi:10.1080/2331186X.2017.1340081. Schleiermacher, Friedrich. 2012. On the Different Methods of Translating. Translated by Susan Bernofsky. The Translation Studies Reader, 3rd edition, ed. by Lawrence Venuti, pp. 43–63. New York: Routledge. Sherman, Beatrice. 1943. A Prince of Lonely Space. New York Times, April 11, 1943, 9. Treuer, Anton, and Michael Sullivan Sr. 2021. Ge-ni-aabadak Giniigaaniiminaang. St. Paul: Minnesota Historical Society Press. Webster, Paul. 1993. Antoine de Saint-Exupéry: The Life and Death of The Little Prince. London: Pan Macmillan.

Deverbal Noun Composition in Mille Lacs Ojibwe Hunter Johnson and Christopher Hammerly

W

hile Algonquian languages are typically noted for their verb-centricity, there are various nominal elements that play an important role in the grammar.1 The goal of this paper is to provide a description and preliminary analysis of the formation of a particular type of noun known as deverbal nouns—nouns that are formed from verbs through the addition of “nominalizing” morphology—through a case study of the Mille Lacs variety of Southwestern Ojibwe (henceforth simply “Ojibwe” unless disambiguation is needed). We look at both broader patterns of deverbal noun formation through the creation and analysis of a small corpus sourced from The Ojibwe People’s Dictionary (OPD, Livesay and Nichols 2022), then test these patterns by working with a first speaker of Ojibwe from Mille Lacs. To understand where deverbal nouns fit into the grammar, we can consider the different types of nominal elements in the language. Some nouns in Ojibwe are simply nouns and are not formed by applying a derivational process to a word of a different category: for example, akik ‘pail’ and ikwe ‘woman’. Languages like Ojibwe also have productive processes of forming nouns or nominal-like elements from other parts of speech such as participial relative clauses (1) and compounds (2).2

|

   141

|

142    Hunter Johnson and Christopher Hammerly

(1) Participle formation (from Sullivan 2016) a. wayaabishkiiwejig

ic.waabishkiiwe-d-ig



ic.is.white-3-pl



‘white people’ (lit. ‘Those who are white’)

b. zhayaazhiibaabagizojig

ic.zhaazhiibaabagizo-d-ig



ic.hoop.dances-3-pl



‘hoop dancers’ (lit. ‘Those who hoop dance’)

(2) Nouns via compounding (from OPD) a. giiyose-w-inini hunt.vai-comp-man ‘hunter’ b. mazina’ige-mazina’igaans gets.on.credit.vai-card

‘credit card’

The focus of our research is on deverbal nouns: nouns formed by attaching a nominalizing morpheme to a verb stem. This process is demonstrated in (3). (3) Deverbal nouns (from OPD and Joe Nayquonabe) a. odaminow-aagan play.vai-nmlz ‘doll’ b. bakwezh-igan cut.vti-nmlz ‘bread’ c. anishinaabemo-win speak.Ojibwe.vai-nmlz ‘Ojibwe language’ d. ziinikiigomaa-n blow.nose.vai-nmlz ‘snot’

|

Deverbal Noun Composition   143

All these deverbal nouns are formed by adding a nominalizer (bolded, glossed as nmlz) to an existing verb stem like odamino ‘s/he plays’, bakwezh ‘s/he cuts a piece off something animate’, anishinaabemo ‘s/he speaks Ojibwe’, or ziinikiigomaa ‘s/he blows h/ nose’. Ojibwe speakers use this strategy to productively form nouns following the rules of the grammar, and this project sets out to ask the question: what restrictions, if any, exist in the deverbal noun-making process in Ojibwe?

Deverbal Noun Formation in Ojibwe

This project analyzed four productive nominalizers that can be used to form deverbal nouns in Ojibwe: -win, -aagan, -gan, and -n (for a related study on Oji-Cree, see Hoffman and Oxford 2021; for Plains Cree, see Giesbrecht and Lachler 2021). Since these combine with a verb to form a noun, a bit of background on verb types in Ojibwe is necessary. As summarized below, there are four classes of verb stems in Ojibwe based on transitivity and the grammatical animacy of their arguments. (4) a. VAI: Verb Animate Intransitive (single animate subject in a sentence) b. VTA: Verb Transitive Animate (an animate object is being acted on) c. VII: Verb Inanimate Intransitive (single inanimate subject in a sentence) d. VTI: Verb Transitive Inanimate (an inanimate object is being acted on)

Each of the four nominalizers show different patterns in terms of what types of verb stems they combine with. Below we present what previous work has shown in each case. Nouns Formed with -win

The first type of deverbal nouns are formed with the nominalizer -win (Nichols 1980:81; Bloomfield 1958, §11.52). Giesbrecht and Lachler (2021) investigated nouns formed by a similar nominalizer in Plains Cree and found that -win rarely forms deverbal nouns from inanimate verb stems such as VTIs (roughly 17% of all -win nouns) or VIIs (roughly 1%). Instead, -win most frequently combined with VAIs (56%) and VTAs (24%). They found that -win nominals in Plains Cree are never semantic agents: the “doers” of the clause as in play-er, the one who plays. Instead,

|

144    Hunter Johnson and Christopher Hammerly

-win nominals are made up of other semantic types, such as predicates, patients, instruments, and locations. Some examples of Ojibwe -win nouns are given in (5). (5) -win nouns a. abi-win be.at.home.vai-nmlz ‘room’ b. agaji-win be.ashamed.vai-nmlz ‘shame’ c. baapi’idiwag-win laugh.at.each.other.vai-nmlz ‘humor’ d. bimaadizi-win live.vai-nmlz ‘life’

The examples in (5) demonstrate that -win can be used to form nouns such as abiwin ‘room’, agajiwin ‘shame’, baapi’idiwin ‘humor’, and bimaadiziwin ‘life’. The majority of the nouns formed by -win take an abstract as opposed to concrete meaning, as in (5b–d)—that is, they do not refer to physical objects. The particular abstract meaning of the deverbal noun often reflects the essence of the meaning of the verb stem: the -win form of the verb bimaadizi ‘s/he lives’ is ‘life’, the -win form of the verb agaji ‘s/he is ashamed’ is ‘shame’. This pattern holds across most of the -win nouns with some exceptions, including (5a) abi-win ‘a room’. Nouns Formed with -n

The second type of noun that this study investigates is formed with the nominalizer -n (Nichols 1980:78; Bloomfield 1958, §11.25). Similar to -win nouns, -n combines exclusively with intransitive verb stems, primarily VAIs (Valentine 2001:502). Consider (6).

|

Deverbal Noun Composition   145

(6) -n nouns a. abwaa-n roast.vai-nmlz ‘a roast’ b. apishimo-n lie.vai-nmlz ‘mattress’ c. bimikawaa-n leave.tracks.vai-nmlz ‘footprints’ d. waabikwe-n grey.hair.vai-nmlz ‘grey hair’

When compared to -win nouns in (5), the meaning of nouns formed from -n shown in (6) are generally more concrete. By ‘concrete’ we mean that -n nouns tend to be instruments used to perform actions like apishimon ‘mattress’ or some physical result or product of performing an action as in abwaan ‘roast’, bimikawaan ‘footprints, tracks’. Nouns Formed with -gan

The third type of noun is formed with -gan (Nichols 1980:77–78; Bloomfield 1958, §11.27). Such nouns are almost exclusively made from transitive verb stems, most commonly VTIs. The glosses of the verbs are thus different and reflect transitive verbs. Below are some examples of -gan deverbal nouns. (7) -gan nouns a. aabaabika’-igan unlock.vti-nmlz ‘key’ b. baasaabikiz-igan blast.vti-nmlz ‘dynamite, explosives’

|

146    Hunter Johnson and Christopher Hammerly

c. dibaabiishkooj-igan weigh.vti-nmlz ‘scale’ d. agwaakwa’-igan nail.vti-nmlz ‘poster’

Nouns formed with -gan are often associated with instruments or tools, such as aabaabika’igan ‘key’ (instrument of locking), baasaabikiz-igan ‘dynamite, explosive’ (instrument of explosives), or dibaabiishkooj-igan ‘scale’ (instrument of weighing). However, some nouns simply do not fit as cleanly in these semantic categories. As shown in (7d), for example, the verb stem agwakaa’ means ‘hang it up on the wall with a nail’, and the noun formed from adding -gan is agwaakwa’-igan ‘poster’. We faced this challenge with all the nominalizers. Instead of grouping them into semantic categories, we instead grouped them by abstract or concrete—a much clearer task. Nouns Formed with -aagan

The final type of deverbal nouns in this study are those formed with -aagan (Nichols 1980:80; Bloomfield 1958, §11.28). There are fewer examples of these nouns, and like nouns formed from -gan, they are primarily used in creating deverbal nouns from transitive stems. However, unlike -gan, which primarily combines with VTIs, -aagan combines with VTAs. (8) -aagan nouns a. dasoon-aagan trap.vta-nmlz ‘a trap’ b. gikinoo’amaw-aagan teach.vta-nmlz ‘student’ c. wiijiiw-aagan accompany.vta-nmlz ‘partner, companion’

|

Deverbal Noun Composition   147

d. nagadenim-aagan be.familiar.vta-nmlz ‘a friend’

Like -igan nouns, -aagan nouns are generally concrete. For example, an instrument noun like dasoonaagan ‘trap’ is something used to capture an animal, while other nouns formed with -aagan can be thought of as patients—the ones being acted upon by the verb stem. Patient nominalizations include gikinoo’amawaagan ‘student’ (concrete, the patient of the teaching), wiijiiwaagan ‘partner, companion’ (concrete, the patient of accompany), or nagadenimaagan ‘friend’ (the patient of being familiar with someone).

Methods

We created a spreadsheet containing 420 Ojibwe nouns by consulting the OPD. The OPD contains partially complete lists of words formed from each of the nominalizers that we are examining, and these nouns made up the bulk of our sample. To ensure that we found all possible examples, once these lists were exhausted, we used the advanced search function to look for words that ended with each nominalizer. The spreadsheet entries for the nouns included detailed information about the noun, the translation, the nominalizer used, the animacy, the verb stem, the verb stem type, the verb stem composition, whether the noun was abstract or concrete, and the OPD entry link. A summary and example of the extracted information is given in Table 1. In all cases, except the abstract/concrete distinction, which is not indicated in the dictionary, we went by what the dictionary reported. Having completed this stage, we rescanned the dictionary to verify that we had not missed any entries and to ensure that all the information was complete and accurate. We further set out to decompose some of the verb stems that appeared decomposable to us based on recognizable morphology. The data from the spreadsheet was then analyzed using the R statistical environment (R Core Team 2021) in RStudio (RStudio Team 2021). For each of the nominalizers, we calculated the total number of nouns sampled from the dictionary, the percentage of nouns that were animate versus inanimate, the percentage of nouns that were abstract versus concrete, and the percentage of nouns formed from

|

148    Hunter Johnson and Christopher Hammerly

TABLE 1. Sample of data compiled from the dictionary to form each entry of the corpus

NOUN TRANSLATION NOMINALIZER ANIMACY STEM STEM MEANING STEM VERB TYPE ABSTRACT/ CONCRETE

-GAN

-WIN

-AAGAN

-N

bakwezhigan ‘bread’ -gan Animate bakwezh ‘cuts a piece off’ VTA

gikaawin ‘old age’ -win Inanimate gikaa ‘old’ VAI

odaminowaagan ‘doll’ -aagan Animate odamino ‘plays’ VAI

ziinikiigomaan ‘snot’ -n Animate ziinikiigome ‘blows nose’ VAI

concrete

abstract

concrete

concrete

each of the four stem types (VTA, VTI, VAI, and VII). These findings are reported in Table 2 and discussed in detail in the following section. Having consulted the dictionary and formed our generalizations, we worked with our native speaker collaborator Mr. Nayquonabe to corroborate the findings. While working with Mr. Nayquonabe, we followed the generalizations we gathered from the dictionary and generated examples to test these generalizations against native speaker collaborator data and judgments. The fieldwork was conducted over Zoom between January and June 2021. We recorded audio and video for each session and transcribed the data elicited with Mr. Nayquonabe. For each noun, we elicited the singular form, plural form, and noun with a demonstrative, since the animacy of a noun is apparent from the form of the plural marker and demonstrative.

Results

In this section we present the results of our dictionary survey, summarized in Table 2. Of the 420 nouns collected, 136 examples were formed with -gan, 94 were formed with -n, 170 were formed with -win, and 20 were formed with -aagan. The majority (90%, n=378) of the nouns were inanimate and the majority (64.5%, n=271) were formed from VAI verb stems. The relationship between the nominalizer and the output noun’s animacy is most clear with -win, where 97.8% (n=166) of -win nominals are inanimate. The other nominalizers showed more variation in the animacy of the output noun. For example, 9.6% of -gan nouns are animate (n=13),

|

Deverbal Noun Composition   149

TABLE 2. Dictionary data summary NOMINALIZER

-gan -n -aagan -win TOTAL

# OF NOUNS

% VAI

% VII

% VTA

% VTI

% ANIMATE

% ABSTRACT

136 94 20 170 420

1.5 100 30 99.4 64.5

0 0 0 0.6 0.2

0.7 0 60 0 3.1

97.8 0 10 0 32.1

9.6 16 60 1.2 10

2.2 2.1 5 75.3 31.9

and 16% of -n nouns were animate (n=15). The only nominalizer that had majority animate nouns was -aagan, where 60% of nouns were animate (n=12). The relationship between nominalizer and the transitivity of the verb stem was more clear cut. Both -gan and -aagan primarily paired with transitive verb stems (VTA,VTI). More precisely, -gan combined with transitive VTAs and VTIs 98.5% of the time (n=134) and VAIs only 1.5% of the time (n=2). Meanwhile, -n combined with VAIs 100% of the time (n=94). -win was combined with VAIs 99.4% of the time (n=169) and only 0.6% with VIIs (n=1). While most deverbal nouns found in the dictionary are concrete (68%, n=285), -win largely formed abstract nouns with 75% (n=128) of -win nouns being abstract. Across the board, we see that the most common verb stem for deverbal nouns in Ojibwe is VAI (64.5%, n=271), with VTI the second most common (32.1%, n=134). VTA and VII appeared with much smaller frequency, with only 3.1% (n=13) of deverbal nouns being formed from VTAs and 0.2% (n=1) with VIIs. From the patterns observed in this data set, we extract three major generalizations about deverbal noun formation: (i) Ojibwe disallows VII stem nominalizations, (ii) deverbal nouns in Ojibwe formed with these nominalizers cannot denote agents, and (iii) the deverbal noun animacy is not always directly tied to which nominalizer was used. Below, we support these generalizations with native speaker judgments elicited with Mr. Nayquonabe. Inanimate Verb Stems (VIIs)

In line with Giesbrecht and Lachler’s (2021) survey of -win nominalizations in Plains Cree, our data reveal that Ojibwe has a strong dispreference against deverbal nouns formed from VII stems. The authors’ research found that only 1% (n=6) of

|

150    Hunter Johnson and Christopher Hammerly

-win nouns in Plains Cree were formed from VII stems. Of the 420 deverbal nouns surveyed here, only one noun in the entire data set was formed from a VII verb stem: gizhaatewin ‘heat, hot weather’. We consider this to be a genuine exception to an otherwise robust generalization that Ojibwe speakers do not nominalize VII stems with any of the nominalizers presented here—expanding the conclusion of Giesbrecht and Lachler’s (2021) to include -gan, -aagan, and -n in addition to -win. To corroborate this generalization, we created a list of VII verb stems and combined them with -win. We then consulted with our native speaker collaborator, Mr. Nayquonabe, for his acceptability judgments. For each example we first confirmed that Mr. Nayquonabe was familiar with the meaning of the VII stem. We then presented him with the VII+win form to ask (i) whether it was a meaningful word in Ojibwe and (ii) what exact meaning it conveys. To summarize the results, Mr. Nayquonabe struggled to attribute any meaning to these words. While he could squeeze out the same meaning as the underlying verb, these VII+-win “nouns” are not valid words in Ojibwe and are certainly not nouns. Taking a step back to unpack this finding, overall, deverbal nouns formed from VAIs with -win adopt a variety of meanings (Giesbrecht and Lachler 2021), but they generally reflect the essence of the verb stem. For example, consider the three VAI deverbal nouns in (9). (9) VAI + -win: a. gimoodi ‘s/he steals’ → gimoodi-win ‘theft’ b. ikwewi ‘s/he is a woman’ → ikwewi-win ‘womanhood’ c. debwe ‘s/he tells the truth’ → debwe-win ‘truth’

The deverbal noun meanings in (9) reflect the essence of the original verb stem. By nominalizing gimoodi ‘s/he steals’ it becomes gimoodi-win ‘theft’, the concept or the act of stealing. By nominalizing ikwewi ‘s/he is a woman’ it becomes ikwewi-win ‘womanhood’, the essence of being a woman. By nominalizing debwe ‘s/he tells the truth’, it becomes debwe-win ‘truth’, the essence of telling the truth. If the VII verbs pattern like the VAI verbs when nominalized with -win, their expected meaning should be similar. VII verbs are used to describe inanimate things: miskwaa ‘it is red’, agaasaa ‘it is small’, or abawaa ‘it is warm (weather)’. Consider the following four VII stems and their anticipated nominalized meanings in (10).

|

Deverbal Noun Composition   151

(10) VII + -win (anticipated meanings): a. miskwaa ‘it is red’ → miskwaa-win ‘redness/red’ b. agasaa ‘it is small’ → agasaa-win ‘smallness’ c. abawaa ‘it is warm (weather)’ → abawaa-win ‘warm (weather)’

When we presented Mr. Nayquonabe with VII+-win forms, he indicated that there is little to no difference in meaning between the VII stem alone and the invented deverbal form. He repeatedly described it as ‘describing the thing’, which is what the bare verb already does. Consider the example miskwaa ‘it is red’ in (11), which might be expected to mean ‘redness’ when -win is added, as in (12).3 (11) mɪskwɑː red.vii ‘it is red.’ (12) mɪskwɑː-wɪn red.vii-nmlz Forced meaning: ‘it is red.’

Instead, Mr. Nayquonabe indicated that this form carries the same meaning with or without the added -win: it is describing something that is red, not necessarily the property of being red. As an example, he provided: “an Ojibwe speaker might say miskwaa-win when describing somebody’s house but it does not clearly mean ‘redness’ or the essence of being red.” Other examples come from verbs like agaasaa ‘it is small’ (13) whose nominalized form in (14) we might expect to mean ‘smallness’, or the property of being small. (13) ɑgɑsɑː small.vii-nmlz ‘it is small.’ (14) ɑgɑsɑː-wɪn small.vii-nmlz Forced meaning: ‘it is small.’

|

152    Hunter Johnson and Christopher Hammerly

Instead, Mr. Nayquonabe said that the -win does not change the meaning, it still carries the meaning of the VII stem. Mr. Nayquonabe mentioned that he had heard (14) in the past but it is not something he would say. These data add support to the generalization abstracted from the dictionary survey that Ojibwe VIIs are not viable candidates for deverbal noun formation. Meaning and Nominalization

In Ojibwe, just as in Plains Cree (Giesbrecht and Lachler, 2021) and other languages, nouns are classified into different semantic categories depending on what entity or property they pick out in the natural world. A noun such as akik ‘pail’ falls into the category of instruments, bimikawaan ‘footprints, tracks’ might be considered a result deverbal noun, and gikinoo’amaagewinini ‘teacher’ falls into the category of agents. Sometimes, these categories are associated with particular nominalizing morphology as in the productive agentive nominalizer -er in English: kick+-er → kicker. Cross-linguistically these agentive nominalizers are somewhat common, i.e. the French -euse/-eur as in nettoyer ‘to clean’ → nettoyeuse ‘cleaner (fem.)’, nettoyeur ‘cleaner (masc.)’ the Spanish -ador(a) as in trabajar ‘to work’ → trabajadora ‘worker (fem.)’ or trabajador ‘worker (masc.)’, etc. Semantic agents are the “doers” of a sentence or verb and cross-linguistically, agentive nominalizations are rather common. Based on the dictionary data and our fieldwork, it appears that semantic agents cannot be formed by the four nominalizers in Ojibwe. These findings align with those in Plains Cree outlined in Giesbrecht and Lachler (2021), who found that deverbal nouns formed with -win in Plains Cree cannot denote agents.4 For example, in Ojibwe, when the verb odamino ‘s/he plays’ is nominalized the resulting noun is ‘doll’: the thing played with. Meanwhile, one of the deverbal nouns of the verb play in English is play-er ‘the one playing’. None of the following deverbal noun data from Ojibwe create agent nominalizations (English counterparts given for contrast). (15) a. akwaandawe ‘they (sg,anim) climb’ → akwaandaw-aagan ‘ladder’: instrument

b. ‘climb’ → ‘climb-er’: agent

(16) a. adaawaage ‘they (sg,anim) sell’ → adaawaaga-n ‘something for sale, merchandise’: product

b. ‘s/he sells’ → ‘seller’: agent

|

Deverbal Noun Composition   153

(17) a. biindaagibagizo ‘they (sg,anim) do a hoop dance’ biindaagibagizo-win ‘a hoop dance’: result

b. ‘s/he does a hoop dance’ → ‘a hoop dancer’: agent

(18) a. baasaabikiz ‘they (sg,anim) blast h/’ → baasaabikiz-igan ‘an explosive, dynamite’: instrument

b. ‘blast’ → ‘blaster’: agent (or Instrument)

We further confirmed this pattern through our elicitation sessions with Mr. Nayquonabe: none of the nouns elicited with him are semantic agents. While there are no agentive nominals formed with these nominalizers, Ojibwe has other productive processes to form agents through (i) participle formation as seen above in (1) and (ii) noun compounding as seen above in (2). Consider another example of noun compounding used in the word gikænɑːʔɑmɑge-w-enene ‘teacher’ (agent, the one who teaches). Instead of nominalizing the verb ‘to teach’, with one of the four nominalizers, Mr. Nayquonabe adds w-inini ‘man’. (19) gikænɑːʔɑmɑge-w-enene teach.vai-comp-man ‘teacher (masc.)’

In the process of identifying the lack of semantic agents, we noticed an interesting pattern related to the semantics of deverbal nouns: certain nominalizers appear to be related to forming either abstract or concrete nouns. We found that -win is largely associated with abstract nouns and -n, -igan, and -aagan are associated with concrete nouns. In fact, less than 2.5% of -gan (n=3), 2.5% of -n (n=2), and 5% (n=1) of -aagan nouns were abstract but a majority, 75%, of -win (n=128) nouns were abstract. Deverbal Noun Animacy

In our analysis we found that most of the nouns (90%, n=378) were inanimate, while 10% (n=42) were animate. The breakdown of animacy within each type of nominalizer shows that -win had the strongest relationship with a particular animacy as only 1.2% (n=2) of the deverbal nouns were animate. Compare that to -aagan,

|

154    Hunter Johnson and Christopher Hammerly

which patterns most strongly with animates as 60% (n=12) are animates.5 The -igan and -n nominals share similar animacy distributions: 9.6% of nouns formed with -igan are animate, whereas 16% of -n nouns are animate. While working with Mr. Nayquonabe, we consistently elicited both the plural and demonstrative forms of each of the nouns to determine the animacy. For the most part, his judgments pattern like the dictionary in the sense that the majority of the nouns are inanimate.

Discussion

Having presented in detail the results from the dictionary sample and native speaker judgments, we turn now to consider some of the extensions and complications with the presented analysis, particularly with respect to the syntactic structures that underlie deverbal noun formation. We consider the possibility that the nominalizers are internally complex and the consequences of the patterns of animacy for current theories of the relationship between noun class assignment and nominalization. Decomposition and the Structure of Nominalization

So far we have treated the four nominalizers -gan, -win, -aagan, and -n as single morphemes—this follows the classification scheme used in the OPD. In this section, we show that these nominalizers have internal complexity. Following Valentine (2001:502) for Nishnaabemwin and Nichols (1980:78) for Mille Lacs Ojibwe, we adopt the view that the nominalizer -gan is in fact built from the combination of a detransitivizing morpheme -ge and the nominalizer -n.6 While not discussed by either Nichols or Valentine, we further extend this proposal to -aagan, which differs only in -gan in the presence of the -aa augment morpheme.7 The proposed breakdowns are outlined in (20). (20) a. -gan = -ge + -n

b. -aagan = -aa + -ge + -n

With this, we can move toward a more precise characterization of nominalization and an explanation of why -gan and -aagan appear to overwhelmingly nominalize transitive verbs, while -n (and -win) only nominalizes intransitives: in fact, only intransitive verbs, and more particularly animate intransitive verbs

|

Deverbal Noun Composition   155

(i.e., VAIs), can be the stem for a deverbal noun. Any apparent cases of deverbal noun formation with VTI and VTA stems are better described as first undergoing a process of detransitivization with -ge, which decreases the valency of the verb by eliminating the internal argument or patient, creating a VAI. This is followed by nominalization proper with the addition of -n. This two-step process is outlined in (21) for the deverbal noun bakite’igan ‘hammer’. (21)

a. VTI Stem: bakite’ ‘They (sg,anim) strike/hit it (inan)’



b. VAI formation via -ge → bakite’ige ‘They (sg,anim) strike/hit’



c. Nominalization via -n → bakite’igan ‘hammer (inan)’

This breakdown gives rise to a critical question of why 30% (n=5) of deverbal nouns formed from -aagan should have a VAI stem—these stems are already VAIs and should not have to undergo detransitivization in order to be nominalized. In other words, it should be possible to directly form deverbal nouns by the addition of -n and the absence of -ge. While our analysis remains tentative, we note the conspicuous fact that one of the five stems (mimigoshkam ‘They (SG, ANIM) threshes something’) is labeled in the OPD as a VAI+O—a VAI stem that can take objects and inflect with transitive morphology—and the other four, shown in (22), all appear to have meanings that include implicit objects.8 (22) a. odamino ‘They (sg,anim) play’; odaminwaagan ‘doll (anim)’

b. ziko ‘They (sg,anim) spit’; zikwaagan ‘spitoon (inan)’



c. agoodoo ‘They (sg,anim) hang a snare’; agoodwaagan ‘snare (inan)’



d. akandoo ‘They (sg,anim) lie in wait for game’; akandoowaagan ‘hunting stand (inan)’

While we have followed the OPD in our classification of each of these stems, we note the possibility that odaminwaagan “doll (anim)” is in fact formed by the addition of the nominalizer -n to the VAI+O stem odaminwaage ‘They (sg,anim) play with something as a toy’, which is formed from the VAI odamino by the addition of the VAI+O final -aage.9 Indeed, despite the fact that Goddard (1990:473–474) asserts that -aagan may be a single morpheme, and therefore non-complex, we acknowledge the possibility that there are intermediate VAI+O derivations for all of these forms, as seen with odamino. Turning now to -win, the temptation may arise to decompose it into two parts: our familiar nominalizer -n and a morpheme -wi. At present, it is not at all clear

|

156    Hunter Johnson and Christopher Hammerly

that such an analysis is warranted, in accordance with Bloomfield (1946:106), Bloomfield (1958:68), and Goddard (1990:472), who all stipulate that there is no synchronic morpheme boundary. While -wi is indeed a common morpheme in Ojibwe, its general function is denominal verb formation—in particular, taking a noun and turning it into a verb of being (Valentine 2001:363). This is inconsistent with the current context of deverbal noun formation. To our knowledge, there is no other plausible analysis of -wi within this context, so we treat -win as a nominalizer in its own right, with no direct derivational relationship to -n (and by extension -aagan and -gan). This split between deverbal nouns formed from -n versus -win is further supported by broad semantic differences that arise between the deverbal nouns in each case: the majority of deverbal nouns formed by -win denote abstract concepts, while the majority of nouns formed by -n denote concrete objects or beings. Animacy and the Structure of Nominalization

We situate our analysis of the structure of deverbal noun formation within the framework of Distributed Morphology (DM; e.g., Halle and Marantz 1993, 1994; Marantz 1997, 2001). One of the major tenets of the theory is the idea that the familiar categories noun and verb are derived—there are no elements that are directly stored in the lexicon as nouns or verbs. Nouns and verbs are formed when an abstract, categoryless root (√) combines with a nominalizer n (“little n”) or a verbalizer v (“little v”). These category-forming heads from DM can be linked to the verb- and noun-forming final morphemes (Brittain 2003; Mathieu 2014). Homing in on the nominalizing heads, previous work within DM has argued that n is responsible not only for creating nouns from roots or verb stems, but also for introducing noun class features (Kihm 2005; Kramer 2014, 2015; Hammerly 2019). Following this proposal, we might assume there should be a close relationship between the use of a particular nominalizer (i.e., -win or -n) and the animacy or grammatical gender of the deverbal noun. In other words, each type of n head might be expected to be exclusively associated with either the animate or inanimate class. The results of the current study reveal a mixed result here. With -win only 1.2% (n=2) of the deverbal nouns were grammatically animate—the overwhelming majority were classified as inanimate. This result is consistent with the observations of Valentine (2001) for Nishnaabemwin and Giesbrecht and Lachler (2021) for

|

Deverbal Noun Composition   157

Plains Cree, and supports a strong (though not absolute) relationship between noun class and nominalizing morphology. In contrast, the nominalizations formed from -n, -gan, and -aagan were more mixed. All told, 16% (n=40) of the nouns in this group where inanimate—a minority to be sure, but difficult to classify as purely exceptional. This raises the question, left to future work, of how to reconcile the present findings with the theory that n introduces not only a noun-forming function but also a particular noun class. There are two observations that further hem in this eventual account of noun class assignment in Ojibwe. First, with limited exceptions, the best predictor of noun class in Ojibwe is whether the referent of the noun is living or non-living (see Dahlstrom 2022 and Goddard 2002). In other words, it appears that noun class in Ojibwe is almost entirely semantic in the sense of Corbett (1991). Second, there is no relationship between the “animacy” of the underlying verb stem (i.e., whether it is a VAI/VTA versus VII/VTI stem) and the deverbal noun animacy. Again, this is in line with previous observations by Valentine (2001) and Giesbrecht and Lachler (2021).

Conclusion

This paper set out to identify restrictions in the deverbal noun formation process in Ojibwe. Through an extensive dictionary survey, we identified three major restrictions: (i) VII verb stems cannot be nominalized, (ii) there are no semantic agents formed from these nominalizers, and (iii) the grammatical gender of the noun is not directly tied to which nominalizer was used. Native speaker judgments elicited through fieldwork were used to verify and support the accuracy and strength of these restrictions. This investigation broadens our knowledge of how nouns are formed in Ojibwe and ultimately in languages as a whole. Based on these three generalizations and our findings presented here, there are countless avenues for further research. We highlight two particularly puzzling questions: (i) why are VIIs immune to nominalization (or, why do only VAIs serve as the stems for deverbal nouns)?; and (ii) how exactly is grammatical gender determined in Ojibwe deverbal nouns? Other avenues of future research may involve a deep analysis of deverbal nouns in other dialects of Ojibwe or related languages in the Algonquian family to see whether these patterns hold.

|

158    Hunter Johnson and Christopher Hammerly

Notes 1. Chi-miigwech to our native speaker collaborator Joe Nayquonabe. Mr. Nayquonabe was a joy to work with on this project and we cannot thank him enough for dedicating his time and effort. Also, a big thank you to professor Jean-Philippe Marcotte for his contributions to the conceptualization of the project, and to the participants of the 53rd Algonquian Conference for their very helpful feedback—special thanks to Richard Rhodes for his comments. This research was funded by a UROP grant to Johnson & Hammerly, and the President’s Postdoctoral Fellowship to Hammerly, both from the University of Minnesota. All errors are the responsibility of the authors. 2. We follow Leipzig glossing conventions with the addition of the following abbreviations: anim = animate, comp = compounding morpheme in Algonquian, h/ = his or her, inan = inanimate, nmlz = nominalizer, vai = Verb Animate Intransitive, vii = Verb Inanimate Intransitive, vta = Verb Transitive Animate, vti = Verb Transitive Inanimate. 3. Examples including data elicited with Mr. Nayquonabe are represented with a broad transcription using the International Phonetic Alphabet. Examples including dictionary data are represented using the double vowel orthography as reported in the OPD. 4. As mentioned above, the focus of this study and that of Giesbrecht and Lachler (2021) is nouns formed with particular nominalizers. As with Ojibwe, speakers of Plains Cree have their own way of forming deverbal agent nouns. Wolfart (1973:69) shows that the -w morpheme may be used for creating agent nouns and Goddard (1990:473) shows that this is a general pattern diachronically, but not synchronically, in Algonquian. While both Ojibwe and Plains Cree have ways of forming deverbal agent nouns, these nominalizers cannot be used to do so. 5. This is likely because -aagan forms deverbal nouns from VTA verbs that implicitly have an animate patient. 6. For additional discussion, see Mathieu (2014), Bloomfield (1946:106), Bloomfield (1958:66), and Goddard (1990) for specific analyses of the composition of -igan. 7. However, Bloomfield (1958:67) proposes that -a:kan consists of -kan with prefinal a:-. In addition, see Goddard (1990:473) for a diachronic analysis of the composition of -aagan; however, as Goddard (1990:474) admits, this is likely not the synchronic analysis. 8. As a helpful reviewer pointed out, there is only explicit evidence that mimigoshkam “They (sg,anim) threshes something” is a VAI+O verb stem; this may be an easy analogy for speakers to make to other VAI+O verb stems. 9. https://ojibwe.lib.umn.edu/word-part/aage-final.

|

Deverbal Noun Composition   159

References Bloomfield, Leonard. 1946. Algonquian. Linguistic Structures of Native America, ed. by Harry Hoijer, pp. 85–129. New York: Viking Fund Publications in Anthropology. Bloomfield, Leonard. 1958. Eastern Ojibwa: Grammatical Sketch Texts, and Word List. Ann Arbor: University of Michigan Press. Brittain, Julie. 2003. A Distributed Morphology Account of the Syntax of the Algonquian Verb. Proceedings of the 2003 Annual Conference of the Canadian Linguistic Association, ed. by Sophie Burelle and Stanca Somesfalean, pp. 26–41. Toronto: Canadian Linguistic Association. Corbett, Greville G. 1991. Gender. Cambridge: Cambridge University Press. Dahlstrom, Amy. 2022. Meskwaki Syntax. Unpublished manuscript, University of Chicago. https://lucian.uchicago.edu/blogs/adahlstrom/publications-2/selected-manuscripts/ meskwaki-syntax-book/. Giesbrecht, Lex, and Jordan Lachler. 2021. Nominalization Strategies in Plains Cree: An Analysis of the -win Suffix. Papers of the 50th Algonquian Conference, ed. by Monica Macaulay and Margaret Noodin. East Lansing: Michigan State University Press. Goddard, Ives. 1990. Primary and Secondary Stem Derivation in Algonquian. International Journal of American Linguistics 56:449–483. Goddard, Ives. 2002. Grammatical Gender in Algonquian. Papers of the 33rd Algonquian Conference, ed. by H. Christoph Wolfart, pp. 195–231. Winnipeg: University of Manitoba. Halle, Morris, and Alec Marantz. 1993. Distributed Morphology and the Pieces of Inflection. The View from Building 20, ed. by Kenneth Hale and Samuel Jay Keyser, pp. 111–176. Cambridge, MA: MIT Press. Halle, Morris, and Alec Marantz. 1994. Some Key Features of Distributed Morphology. MIT Working Papers in Linguistics 21, ed. by Andrew Carnie and Heidi Harley, with Tony Bures, pp. 275–288. Cambridge, MA: MIT Press. Hammerly, Christopher. 2019. Limiting Gender. Gender and Noun Classification, ed. by Éric Mathieu, Myriam Dali, and Gita Zareikar, pp. 93–118. Oxford: Oxford University Press. Hoffman, Sarah, and Will Oxford. 2021. Derivational Functions of Theme Signs in Oji-Cree. Papers of the 50th Algonquian Conference, ed. by Monica Macaulay and Margaret Noodin, pp. 135–153. East Lansing: Michigan State University Press. Kihm, Alain. 2005. Noun Class, Gender, and the Lexicon-Syntax-Morphology Interfaces: A Comparative Study of Niger-Congo and Romance Languages. The Oxford Handbook of Comparative Syntax, ed. by Guglielmo Cinque and Richard S. Kayne, pp. 459–512. Oxford: Oxford University Press.

|

160    Hunter Johnson and Christopher Hammerly

Kramer, Ruth. 2014. Gender in Amharic: A Morphosyntactic Approach to Natural and Grammatical Gender. Language Sciences 43:102–115. Kramer, Ruth. 2015. The Morphosyntax of Gender: Evidence from Amharic. Oxford: Oxford University Press. Livesay, Nora, and John Nichols. 2022. The Ojibwe People’s Dictionary. https://ojibwe.lib.umn. edu. Marantz, Alec. 1997. No Escape from Syntax: Don’t Try Morphological Analysis in the Privacy of Your Own Lexicon. University of Pennsylvania Working Papers in Linguistics 4.2, ed. by Alexis Dimitriadis, Laura Siegel, Clarissa Surek-Clark, and Alexander Williams, pp. 201–225. Philadelphia: University of Pennsylvania. Marantz, Alec. 2001. Words and things. Handout of a talk, MIT, Cambridge, MA. Mathieu, Éric. 2014. Nominalizations in Ojibwe. In Cross-linguistic Investigations of Nominalization Patterns, ed. by Ileana Paul, pp. 3–24. Amsterdam: John Benjamins Publishing Company. Nichols, John. 1980. Ojibwa Morphology. PhD thesis, Harvard University. R Core Team. 2021. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. https://R-project.org/. RStudio Team. 2021. RStudio: Integrated Development Environment for R. RStudio, PBC, Boston, MA. http://rstudio.com/. Sullivan, Michael D. 2016. Relativization in Ojibwe. PhD thesis, University of Minnesota. Valentine, J. Randolph. 2001. Nishnaabemwin Reference Grammar. Toronto: University of Toronto Press. Wolfart, H. Christoph. 1973. Plains Cree: A Grammatical Study. Transactions of the American Philosophical Society 63:1–90.

Syllabicity of Dorsal Fricative [X] in Blackfoot An Empirical Investigation Mizuki Miyashita

S

yllable is a term used commonly in both linguistic and meta-linguistic senses.1 In the literature on Blackfoot phonology, the notion of syllables is also used as a theoretical tool. For example, Frantz (2017) as well as Taylor (1969) refer to the syllable when describing pitch accent. Despite the use of the notion of the syllable in the literature, I find it challenging when others ask me what the syllable structure of Blackfoot is like. In this context, a linguist typically expects an answer involving coda allowance, coda condition, onset complexity, syllable contact, and syllable weight. And in Blackfoot, words such as imitáá [i.mi.táː] ‘dog’ or nínna [nɪn.na] ‘my father’ have syllable structures that are easily determined because of the sequence of alternating consonants and vowels.2 But not all Blackfoot words are like these. For example, ómahkokata [ómʔxkokata] ‘gopher’, ímahkihkinaa [ímʔxkçkinaː] ‘sheep’, and isttohkihkiitaan [isttxwkçkiːtaːn] ‘pancake’ are puzzling cases with respect to syllabification because of the inclusion of the dorsal fricatives [x], [ç], and [xw], which are phonetic variations derived from /ax/, /ix/, and /ox/, respectively. (The three phonetic variants [x], [ç], and [xw] are hereafter represented with a capital [X], underspecifying their place, because the segments’ place features do not matter for the current topic.)

|

   161

|

162    Mizuki Miyashita

This paper addresses the syllabification challenge of the dorsal fricative [X] in Blackfoot by investigating its prosodic status, which was previously assumed to be syllabic but was understudied. To contribute to linguistic description, theoretical analysis, and pedagogical application, the study reported on here consulted native speakers of Blackfoot and examined whether the speakers would tap on the sound in question while pronouncing a word that includes the sound. Based on the results, I conclude that the dorsal fricative [X] is, unlike a vowel, not inherently syllabic, but it is licensed to be syllabic in certain environments. I also explore the possible theoretical treatment of the segment’s prosodic status and suggest that the sound might be moraic without being parsed to a syllable.

The Blackfoot Dorsal Fricative

Orthographically, the dorsal fricative is represented with a digraph: a vowel plus h. (The h alone represents a velar fricative in Blackfoot.) It never occurs word-initially or prevocalically; in other words, it is never an onset. Frantz (2017) states that the sound represented with two orthographic symbols (a vowel and h) is pronounced as one sound, and in fact this sound has been analyzed as a vowel-consonant coalescence (Miyashita 2018): The three phonetic variants [x], [ç], and [xw] are derived from /ax/, /ix/, and /ox/, respectively, and their place features are determined based on the quality of the underlying vowels: /a/, /i/, or /o/. When a long vowel precedes the dorsal fricative underlyingly, the surface form occurs as a sequence of a short vowel and [X]. The examples in (1) show orthographic representations of the dorsal fricatives in italics, their underlying forms in IPA in slashes, and their surface forms, also in IPA, in square brackets. (1) Orthographic, underlying, and surface forms of [X] Underlying short vowel

Underlying long vowel

ah /ax/  [x]

aah /aax/  [ax]

ih /ix/  [ç]

iih /iix/  [iç]

oh /ox/  [xw]

ooh /oox/  [oxw]

As for the role of [X] in a syllable, when it follows a vowel, as in innóóhksisii [in:óXksisiː] ‘elephant’, the sound [X] surfaces as a coda: [in.nóX.ksi.si]. However, its prosodic status presents a problem when there is no preceding vowel at the

|

Syllabicity of Dorsal Fricative [X] in Blackfoot   163

surface, such as the case of ómahkokata ‘gopher’ shown in (2). The example shows two possible syllabifications: one when [X] is assumed to be syllabic (left) and the other when it is not (right). (2) ómahkokata ‘gopher’3 /ómaxkokata/  [óm.X.ko.ka.ta] or [ómX.ko.ka.ta]

Existing articles on Blackfoot phonological analysis tend to assume that [X] is syllabic (e.g., Frantz 2017; Weber 2016). On the other hand, I have anecdotal support for this sound being non-syllabic from my experience assisting a Blackfoot language instructor at the Piegan Institute in 2008 and 2009. The instructor, who is a native speaker of the language, had created a word list for her students, and each word was written with spaces between the “syllables” to help students with pronunciation. Table 1 shows selected forms from the teaching material followed by a gloss. Orthographic and IPA transcriptions based on recordings of the instructor are also shown, as well as an indication of the source lesson, list, and item number. I found it intriguing that [X] was sometimes written as part of a syllable containing a vowel (a–b) and sometimes as part of a syllable consisting only of an onset consonant and [X] (c–d). Example (e) includes both types. This dorsal fricative has been a puzzling factor in Blackfoot’s syllable structure. There may be something that makes speakers indecisive about syllabification when a word includes the sequence of C[X]. Understanding the sound in terms of its syllabicity should contribute not only to the field of linguistics but also to instructors’ pedagogical efforts and teaching strategy.

TABLE 1. Some Blackfoot words from language teaching material

a. b. c. d. e.

PEDAGOGICAL FORM

GLOSS

ORTHOGRAPHY

IPA

ID

Omah-ksi-ki-mi A-patoh-sohts Nii-tah-taa Oo-tah-koi-nat-tsi Mis-tak-omah-kih-kina

‘lake’ ‘north’ ‘river’ ‘yellow’ ‘mountain goat’

ómahksikimi apátohsoohts niítahtaa ótahkoinaattsii míístakómahkihkinaa

[ómʔxksikimi] [apátxwsoxwts] [niítxtaː] [ʊ́ txkoinatːsi] [míːstakómʔxkçkinaː]

(4.6.8) (2.2.1) (4.6.7) (1.6.8) (4.3.8)

|

164    Mizuki Miyashita

Descriptions of Blackfoot Syllables

There are some descriptions and discussions of Blackfoot syllables in the literature. Taylor (1969) briefly talks about syllable canon and states that a Blackfoot syllable can be complex, allowing up to five consonants in the onset and three consonants in the coda, as shown in (3). (3) Taylor’s Blackfoot syllable canon (C)(C)(C)(C)(C)V(V)(C)(C)(C)

Taylor uses the word ksstsiists ‘snail shells’ as an example showing the maximal use of the syllable canon (4a). But Kaneko (1999, 2000) counters this claim by pointing out that the sequences ks and ts should be considered affricates: [ks] and [ts] respectively.4 This analysis simplifies Taylor’s syllable canon, as shown in (4b).5 (4) Syllabification of ksstsiists ‘snail shells’ a. ksstsiists CCCCCVVCCC b. ksstsiists CCCVVCC

Frantz (2017) uses the notion of the syllable, as mentioned earlier, to identify pitch accent location. However, [X] is never a pitch-bearing unit, so there is no pitch associated with the segment at the surface. And [X] is similarly outside the discussion by Elfner (2006), whose theory-driven analysis investigates the syllable-internal structure of Blackfoot words that include geminates. Since the dorsal fricative never occurs as a geminate, Elfner’s analysis does not apply to it. Thus, the dorsal fricative is not a center of discussion in these studies.

Syllabic Consonants in Blackfoot

Currently, there is no study that focuses on the Blackfoot dorsal fricative’s prosodic status. However, Derrick (2007) studied the syllabicity of another fricative consonant, [s], and claimed that inter-consonantal long [sː] is syllabic in Blackfoot. Derrick (2006) also provides instrumental support for a long [sː] being syllabic, showing that it has a duration that resembles a vowel. This hypothesis has been adopted and utilized for further theoretical work (e.g., Denzer-King 2009; Goad and Shimada 2014).

|

Syllabicity of Dorsal Fricative [X] in Blackfoot   165

TABLE 2. Comparison between distributions of [s] and [X]

a. Phonetic variations b. Long vowel precedes c. Followed by a vowel d. Long consonant

[s]

[X]

   

   

While the duration of [sː] alone is not a solid determination factor for its syllabicity, if one assumes that the hypothesis is correct and inter-consonantal long [sː] is syllabic, the next question might be whether the same analysis can be applied to [X]. The dorsal fricative [X] and alveolar fricative [s] are the only fricatives in the Blackfoot consonant inventory, and both are often observed in complex consonant clusters. Thus, one may assume that there are some similarities between them, suggesting that [X] may also be a syllabic consonant. However, these two sounds must not be generalized too far, since they are fundamentally distinct in several ways. As shown in Table 2, the alveolar fricative has no phonetic variants, while the dorsal fricative has three phonetic variants (a). A long vowel can precede [s], but can never occur before [X] at surface level (b). A vowel can follow [s], but not [X] (c). And there is a long counterpart of [s] but not [X] (d). The last characteristic is particularly significant, since [s] is syllabic only when it is long, while there is no long [X] and, thus, it is impossible to find [X] in the same environment where [s] can be syllabic. So the syllabicity of [sː] cannot be used as an argument for the syllabicity of the dorsal fricative, and the current study aims to examine and reveal the syllabicity of [X].

Theoretical Interest

Investigation of the dorsal fricative’s prosodic status is also theoretically motivated. Weber (2016) analyzes accent and prosody in Blackfoot verbs, including words with this sound. In Weber’s analysis, the sound is treated as a syllabic voiceless vowel and pre-aspiration of the following consonant rather than a coalescence. Frantz also describes the sound as a “voiceless syllable” (2017:6) and states that “the vowel and h are pronounced simultaneously” (2017:21); however, it must be remembered

|

166    Mizuki Miyashita

that Frantz’s account is impressionistic, without an experimental investigation, and that, as mentioned earlier, he used the notion of the syllable as a tool to describe accent locations on full vowels.

Though it is not at the syllable level, the analysis in Elfner 2006 hypothesizes that [X] is moraic. Adopting this view of [X], the example shown in (2) of ómahkokata ‘gopher’ can be addressed again with the representations given in (5), in which possible syllable structures for the first three phonetic segments [omX] are illustrated. If the sound [X] is syllabic, the internal syllable structure of the first part of the word may be illustrated with two syllable nodes, as shown on the left in (5). If it is not syllabic, the structure may be illustrated with one syllable node, with [X] analyzed as a coda, as shown on the right in (5).6 The issue is that, to my knowledge, there is no evidence or empirical support formally reported for either structure. This study addresses this issue with an empirical investigation.

Speaker Intuition Consideration

The syllabification tools that are made available based on commonly studied languages like English are of limited use for Blackfoot prosodic analysis, because such languages do not include sound sequences like those in Blackfoot. However, meta-linguistically, a syllable is often understood as a pronunciation unit or a chunk of pronounceable sounds that native speakers can intuitively identify or segment. Therefore, speaker intuition is a significant part of studying syllables. Still, researchers need to be careful about using the term syllable, especially with regard to a language whose prosody is understudied. Since the concept is indeed an abstract one, depending on the prosodic characteristic of a language, how speakers of the language treat this unit could vary from language to language or even from speaker to speaker. It may well vary even among linguists, for that matter. It is also important to note that syllable analysis of uncommonly researched languages should rely on data that consists of words syllabified by native speakers.

|

Syllabicity of Dorsal Fricative [X] in Blackfoot   167

For example, the famous case of Berber’s syllabification of vowelless syllables (e.g., Dell and Elmedlaoui 1985, 1988; Ridouane 2008) is not what was first predicted with theoretical analysis. The syllable boundaries were given by the native speaker author for analysis of the sonority scale. The notions of ambisyllabicity and syllabic consonants in English have also been presented by researchers who are native speakers. In the case of the Blackfoot dorsal fricative, native speaker consultation is especially valuable because there are multiple possibilities for it to surface in terms of prosody: syllabic, not syllabic, or both. However, there is no documentation or formal study reported on Blackfoot syllabification that is based on native speaker intuition. Since the author is not a native speaker of Blackfoot, the study reported here involved consultation with native speakers.

Tapping Sessions

In pursuing prosodic consultation, I adopted the tapping method that is often used to find stressed syllables (Fitzgerald 1997). While this study is not about finding stress, the first step of this method is to find stress-bearing units, which are typically syllables, and therefore this step was adopted. The term syllable was not used for speaker consultation; beat was used instead, for the following reasons. The term syllable and its interpretation seem to work well with those languages in which the concept is already familiar to most speakers, typically via language education and especially language arts (e.g., poetry). Also, while the syllable tends to serve as a counting unit in many languages, it is nonetheless not an absolute universal. For example, the counting unit of a haiku in Japanese is the mora, as opposed to the syllable in English haiku (Cole and Miyashita 2006; Lehiste 1997). So, for an understudied language whose counting unit is yet unknown, it is safer to use a term other than syllable during tapping sessions to avoid invoking preconceptions based on speakers’ English knowledge. By doing so, influence from English knowledge is reduced. This decision was especially important in this study because the speaker consultants are bilinguals who have gone through education in English and have no formal education in Blackfoot literacy. Two native speakers were consulted: a female from Kainai (recorded in 2016) and a male from Siksiká (recorded in 2017).7 Both were in their sixties at the time of recording. They were instructed to “tap the beats” while pronouncing a list of words, provided in the orthography in isolation.8 Prior to the recording, a short

|

168    Mizuki Miyashita

training session was conducted using words with CV sequences, such as imitáá ‘dog’ or ponoka ‘elk’, whose syllable structure is obvious. For the actual tapping session, a list of forty-three words for animals was used, of which thirteen included the sound [X].9 For this study, recordings of twelve words were used; one word was excluded as it was not pronounced with the sound in question.10 The recordings were then processed using Praat (Boersma and Weenink 1992–2021). An example of a processed and annotated sound file is shown in (6). The first two rows show the wave form and spectrogram. Below the spectrogram, the first tier shows the IPA transcription, the second tier indicates the location of the taps, and the third tier is the entire word’s orthographic transcription.11 Next, a transcription of the utterance and tapping was created, as shown in Table 3.12 An asterisk indicates a tap, and it is placed immediately to the left of the segment where tapping occurred. For example, the word in (6) was transcribed as [*kaːn*ɪssk*iː*naː]. As shown in (6), a tap is not necessarily aligned with the onset of each vowel—the first tap, for example, occurred slightly before the onset of the vowel utterance, after [k]. Only Speaker B demonstrated this misalignment, and only infrequently.

Results and Analysis

Based on the results from the tapping session, the tapped beats are considered to be syllable nuclei for the sake of presentation simplicity. In other words, this study currently assumes that the tapped segment is a syllable nucleus. This of course may be altered, as it is possible that the tapped elements are some other prosodic unit (e.g., a mora, a head of foot, or a mix of prosodic units). As shown in Table 3, five words (a–e) included the dorsal fricative preceded by a vowel (i.e., CV[X], as

|

Syllabicity of Dorsal Fricative [X] in Blackfoot   169

TABLE 3. Transcription of Blackfoot words with tapping

a. b. c. d. e. f. g. h. i. j. k. l.

SPEAKER A

SPEAKER B

ORTHOGRAPHY

GLOSS

*inn*oxwks*is*iː k*ɛjssk*axp*i *oːn*ist*ax*ss *ɪmm*oj*oxwk*ok*in*iː *ɛːs*aj*oxwk*om*iː *omʔxk*ap*iʔs*i *omʔxk*sstoːkiː *omʔxk*ok*at*aː *omʔxk*at*aːj*o s*ikxwkj*aːj*o *omʔx(*)kçk*in*aː m*iːst*aks*omʔx*kçk*in*aː

*inn*oxwks*is*iː k*ajssk*axp *oːn*ist*axss *ɛmm*oj*oʔk*in*i *ɛːs*aj*oxwk*om*iː *omʔxk* ap*iʔs*i *omʔxk*sst*oːk*iː *omʔxk*ok*at*aː *omʔxk*at*aːj*o s*ikxwkj*aːj*o *iːmʔx*kçk*in*aː *miːst*ak*iomʔxk*çk*in*aː

innóóhksisii kaaysskááhp onistaahs áímmoyoohkini áísaayoohkomi ómahkapi’si omahksstooki ómahkokata omahkatayo sikohkiaayo íímahkihkinaa miistáksoomahkihkinaa

‘elephant’ ‘porcupine’ ‘calf’ ‘lion’ ‘bull’ ‘wolf’ ‘donkey’ ‘gopher’ ‘cougar’ ‘black bear’ ‘sheep’ ‘bighorn sheep’

in [innóxwksisi:] ‘elephant’), five (f–j) included the dorsal fricative not preceded by a vowel (i.e., C[X], as in [ómʔxkokata] ‘gopher’), and two (k–l) included two dorsal fricatives intervened by a consonant and not immediately preceded by a vowel (i.e., C[X]C[X], as in [ímʔxkçkina:] ‘sheep’). Both speakers located taps on the same places except for one word: In (c), for ‘calf’, Speaker A tapped on the final long [s:] while Speaker B did not. See below for further discussion. These three [X] environment types, labeled as Types A, B, and C, and the results of the tapping sessions are described in the following sections. For each type, one representative word is used to describe the results. Type A: CV[X]

The first case is when the dorsal fricative is preceded by a vowel. The example word is innóóhksisii ‘elephant’. The IPA transcription of this word is shown on the left in (7). The tapped segments are indicated with an asterisk immediately above the sound. As shown here, the four taps are aligned with the vowels only; no tapping on the dorsal fricative [X] occurred when it was preceded by a vowel. This can be analyzed as indicating that the dorsal fricative is acting as a coda of the syllable whose nucleus is the preceding vowel. Thus, a representation of the word’s syllable structure would look like the illustration on the right in (7).

|

170    Mizuki Miyashita

Type B: C[X]

The word ómahkokata ‘gopher’ is used as an example for the second type: words in which the dorsal fricative is not preceded by a vowel. This is an example of vowel-consonant fusion (Miyashita 2018). If it is tapped, [X] may be syllabic and represented as shown on the left in (5), projecting its own syllable node. On the other hand, if not tapped, [X] is not syllabic, and perhaps serves as a coda attached to the preceding syllable, as shown on the right in (5). The result from the tapping session is shown on the left in (8): The speakers tapped only along with the vowels, not on [X]. The fact that the speakers did not tap on [X] indicates that it is not syllabic, and the representation on the right in (5) is preferred. The schematic illustration of the word’s syllable structure is shown on the right in (8).

Type C: C[X]C[X]

The word ímahkihkinaa ‘sheep’ is an example of the final case: a word that includes two dorsal fricatives with no vowel between them. Both speakers felt challenged by the tapping task with words of this type. Speaker A tapped this word twice with two different results, once omitting the tap on [X] (shown as (*) in Table 3). Speaker B tried the tapping exercise several times with this word before reaching an acceptable level of confidence. There are three vowels in this word, all of which were tapped as expected. As for the two dorsal fricatives, the first fricative was never tapped, as in Type B words (in which [X] also follows a consonant). Interestingly, however, the second [X] was tapped, as shown on the left in (9) (twice by Speaker B and once by Speaker A). Thus, the first [X] serves as a coda of the initial syllable, as shown on the right in (9). And, under the assumption that a tapped segment is a syllable nucleus, the second [X] projects a syllable node under which are the dorsal

|

Syllabicity of Dorsal Fricative [X] in Blackfoot   171

fricative and the preceding consonant (i.e., [k]), also shown on the right in (9). It is worth repeating that both speakers found it difficult to conduct the tapping with these Type C words with two dorsal fricatives in a row, while they did not have this problem with other word types. I will come back to this point in the next section.

Summary and Discussion

In summary, the prosodic characteristics of the Blackfoot dorsal fricative based on previous studies and the current experiment are: [X] is moraic (Elfner 2006), [X] is not inherently syllabic (Types A and B), and [X] may be syllabic when following a syllable with another dorsal fricative (Type C). There are some theoretical considerations of these outcomes. One is that the Foot Binarity constraint may be in effect, since it seems that three moras within a syllable are avoided (10; left). Instead, in a Type C word like word ímahkihkinaa ‘sheep’, [kç] (or [kX]) is projected as a vowelless syllable at the risk of Peak violation (10; right).13

This analysis of a second dorsal fricative projecting its own syllable raises the following issues. First, as mentioned earlier, both speakers found it difficult to conduct tapping with words containing two dorsal fricatives and attempted tapping several times before reaching a satisfactory tapping production. So, the projecting of [kç] (=[kX]) as a syllable is based on the result of a forced decision rather than confident speaker intuition. This fact suggests that considering the sound [X] as a syllable may also be unnatural. Second, the unit in question is intrinsically different

|

172    Mizuki Miyashita

from a typical syllable, which serves as a tone-bearing unit, because [X] lacks a vowel and therefore has no association with pitch or amplitude. Taking these problematic issues into account, I propose that the second [X] is better analyzed as an unparsed mora, at the risk of Parse-μ violation, as illustrated in (11). The dotted oval indicates no association status line connected to a syllable node.

Under this analysis, a dorsal fricative [X] is not necessarily a syllabic segment, but it does function as a type of rhythmic unit. Therefore, the mora association is the priority, and not the syllable. Arguably, this treatment creates an odd form without a syllable node. However, an even more extreme case is reported in Hyman 2011, in which the word structure in Gokana can be analyzed only with mora, without syllables at all. Prosody in each language should be examined in its own right, and it is not surprising to find data that cannot be explained with theoretical tools that are based on languages like English.

If a Blackfoot [X] is an unparsed mora, it is also possible to posit that [X] is always moraic and non-syllabic, as illustrated in (12). The fact that the speakers did tap on [X] in certain contexts suggests that the tapped [X]s are syllabic. This could be interpreted as indicating that a syllabic node was projected when necessary, meaning when the speakers decided to tap on them. Such treatment invites many theoretical questions. For example, it would be interesting to investigate how an unparsed mora can be treated in metrical stress theory (Hayes 1995). Hyman’s view—“Claimed syllable universals are often highly theory-dependent. . . . What’s so great about being universal?” (2010:114)—is a good reminder for researchers who

|

Syllabicity of Dorsal Fricative [X] in Blackfoot   173

work on the prosody of a language that is not their own language. What elements are considered as tapping units or syllabic must involve a carefully designed consultation.

Conclusion: Implications and Further Issues

I have presented an empirical investigation of the Blackfoot dorsal fricative [X] with regard to its syllabicity. The speakers consulted only tapped on [X] when it was the second of two [X]s separated by a single consonant. Thus, it is a segment that is tapped and also not tapped. Under the assumption that a tapped segment is a syllable nucleus, [X] is not inherently syllabic, as it is an unparsed mora, but can be licensed to be syllabic in certain contexts, such as when the speakers in this study tapped on it. There are several theoretical implications of this claim. First, previous literature has only hypothesized that the sound [X] is moraic (Elfner 2006), and this study supports that analysis. The next step may be to examine whether its moraicity is inherent. Second, the analysis of [X] as an unparsed mora is also an interesting question to discuss further. Some scholars understand unparsed moras as a case of mora deletion (e.g., Jensen 1995). Other scholars consider unparsed moras as existing moras that are not parsed into syllables, parallel to the idea of unparsed syllables that are not deleted but also not parsed into feet. The latter aligns with the case of Blackfoot unparsed mora proposed in this study. While the current study suggests that the Blackfoot dorsal fricative may be unparsed, it definitely contributes to the rhythm of the language. This means that not being parsed into syllables is not the same as deletion or reduction. Thus, the study will assist further research on prosodic analysis in Blackfoot and beyond (e.g., Weber 2013, 2016). Since Blackfoot presents a unique case, further exploration of [X] will shed light in the literature on prosodic typology. There also are social implications. When community members talk about the uniqueness and linguistic identity of their language, many refer to its polysyntheticity and typically say that their language is “descriptive.” This study adds another unique feature that they can include as part of Blackfoot’s linguistic identifiers. In addition, because studying syllables helps us understand the language’s rhythmic structure, it also provides valuable information to language teachers and learners

|

174    Mizuki Miyashita

with respect to Blackfoot pronunciation. For example, the method used in this study can be modified and used as a tool to teach pronunciation. It can also be combined with a pronunciation teaching tool for word melody that was developed by a community linguist (Fish and Miyashita 2017). In conclusion, Blackfoot [X] is a unique sound that contributes to the language’s unusual prosodic structure. This study presented an empirical investigation of Blackfoot [X], consulting native speakers for their intuition about the rhythmic component of the segment. Since only two speakers performed the tapping sessions for this study, and the Type C environment that presents a critical view toward the analysis of [X] as an unparsed mora that can become syllabic has only two example words, a repeated study with a larger word sample with more speakers may confirm or modify the current analysis and will help reveal the syllabicity of [X] in Blackfoot. Notes 1. I would like to thank the people I consulted for my study: Rod Scout and Bernadine Tallman—native speakers of the Blackfoot language; linguist mentors the late Don Frantz and Inge Genee; and my former research assistants, Naatosi Fish and Dannii Yarbrough. I would also like to thank many community members from the Blackfoot speaking bands, including the late Darrell Kipp, the cofounder of Piegan Institute; Rosella Many Bears, a former instructor of Blackfoot; and others who have encouraged and supported my research in Blackfoot. This work is partially funded by NSF DLI-DEL (formerly DEL) grant (BCS-1251684) and the Phillips Fund of the American Philosophical Society (2013–2014). 2. While the IPA symbol [ː] is commonly (but not always) used to represent Blackfoot long segments (vowels and consonants), in this paper, long consonants are shown with double letters due to the focus on syllabification. This representation enables the labeling of such consonants as the coda of the preceding vowel and the onset of the following vowel when appropriate. Low vowels are transcribed broadly with [a] throughout the paper. 3. There is a glottal stop insertion between [m] and [X], which is also transcribed in Derrick (2006, 2007) and Miyashita (2018). In this paper, it is considered as a type of coarticulation occurring with a preceding nasal consonant [m] and [n] and not included in the transcriptions. 4. Kaneko (2000) also suggests that [ps] should be considered an affricate [ps], but this paper does not include this idea because I believe /ps/ does not behave like the affricates

|

Syllabicity of Dorsal Fricative [X] in Blackfoot   175

[ks] and [ts]. For example, a word never begins with /ps/, while words do begin with /ks/, as in ksíkksinááttsii ‘it is white’, or /ts/, as in tsimá ‘where’. 5. In this example, the long vowel is shown with double letters, following Taylor. 6. Theoretically, [m] could be associated with the second mora. However, in this paper it is assumed to be part of the first mora because this example includes a glottal stop coarticulation. Derrick (2006, 2007) has a slightly different arrangement for [X] that has the coarticulation of [ʔ], making [X] the nucleus and the glottal stop the onset. 7. Blackfoot is spoken by the members of four geographic and cultural bands: Siksiká, Aapátohsipikani, and Kainai in Alberta, and Aamsskáápipikani in Montana. 8. Both speakers consulted were comfortable with the orthography, although many speakers are not. 9. Note that the list used was developed for the purpose of teaching the Blackfoot language by a language educator, in the spirit of conducting research while supporting language revitalization efforts. 10. The excluded word is sikihtsisoo ‘moose’. Both speakers pronounced this word as [siksstsisʊ:], without [X], which is a common pronunciation. The orthography used here is taken from the dictionary (Frantz and Russell 2017). 11. This step is intended to find segments that are tapped when pronounced, which indicates the tapping or counting unit. It is not intended to reveal syllable boundaries. The example uses a word with no [X]. 12. This word appears as áíi.mmoyo’kiniomita in the dictionary. The consultants produced a short version without the -omita ending, and Speaker A produced it with /x/ instead of a glottal stop. The list was prepared by Speaker A, hence the word is included although Speaker B produced it with a glottal stop. Regarding the discussion in the Syllabic Consonants in Blackfoot section about whether the duration of long [s:] is evidence for its syllabicity: The data here include the word omahksstooki ‘donkey’ with inter-consonantal long [s], which was tapped by both speakers. Though this study is not centered on long [s], this may serve as a small piece of evidence for the syllabicity of [s] as Derrick (2006, 2007) proposed. The orthographic representation miistáksoomahkihkinaa for ‘bighorn sheep’ is based on the utterance produced by Speaker A and the dictionary (Frantz and Russell 2017). Speaker B’s slightly different utterance would be rendered as miistákiomahkihkinaa. The same lexical item is shown in Table 1 with the orthographic representation míístakomahkihkinaa based on the pronunciation of the language teacher who created the materials referenced in the table. Note that the teacher provided ‘mountain goat’ for its English equivalent. Although

|

176    Mizuki Miyashita

the two speakers in this study and the language teacher pronounced this word slightly differently, all three produce the same sequence involving [X]: {mxkç}. 13. I adopt the definitions of Foot Binarity (“Feet are binary under moraic or syllabic analysis”) as described for Ft-Bin in Kager (1999) and Peak (“Syllables have one vowel”) as in Archangeli (1997). The understanding of Parse-μ in the literature seems to vary. For example, Blevins and Harrison define it as follows: “All moras must be parsed by feet” (1999:221), Roca and Al-Ageli says moras are “parsed into syllables” (1999:129), and Broselow et al. state that “Parse Mora . . . requires moras to be incorporated into higher prosodic structure” (1995:125). The current discussion assumes the interpretation described by Roca and Al-Ageli (1999).

References Archangeli, Diana. 1997. Optimality Theory: An Introduction to Linguistics in the 1990s. Optimality Theory: An Overview, ed. by Diana Archangeli and Terry Langendoen, pp. 1–32. Malden, MA: Blackwell. Blevins, Juliette, and Sheldon P. Harrison. 1999. Trimoraic Feet in Gilbertese. Oceanic Linguistics 38(2):203–230. Boersma, Paul, and David Weenink. 1992–2021. Praat: Doing Phonetics by Computer [Computer program]. Version 6.1.50. https://www.praat.org. Broselow, Ellen, Marie Huffman, Sui-I Chen, and Ruohmei Hsieh. 1995. The Timing Structure of CVVC Syllables. Perspectives on Arabic Linguistics VII: Amsterdam Studies in the Theory and History of Linguistic Science. Series IV, Current Issues in Linguistic Theory, ed. by E. F. Konrad Koerner, pp. 119–138. Philadelphia: John Benjamins. Cole, Deborah, and Mizuki Miyashita. 2006. The Function of Pauses in Metrical Studies: Acoustic Evidence from Japanese Verse. Formal Approaches to Poetry, ed. by B. Elan Dresher and Nila Friedberg, pp. 173–192. New York: Mouton de Gruyter. Dell, François, and Mohamed Elmedlaoui. 1985. Syllabic Consonants and Syllabification in Imdlawn Tashlhiyt Berber. Journal of African Languages and Linguistics 7:105–130. Dell, François, and Mohamed Elmedlaoui. 1988. Syllabic Consonants in Berber: Some New Evidence. Journal of African Languages and Linguistics 10:1–17. Denzer-King, Ryan. 2009. The Distribution of /s/ in Blackfoot: An Optimality Theory Account. MA thesis, University of Montana. Derrick, Donald. 2006. Duration of Blackfoot /s/: A Comparison of Assibilant, Affricate, Singleton, Geminate and Syllabic /s/ in Blackfoot. Proceedings of WSCLA XI, ed. by

|

Syllabicity of Dorsal Fricative [X] in Blackfoot   177

Atsushi Fujimori and Maria Amélia Reis Silva, pp. 53–60. University of British Columbia Working Papers in Linguistics. Vancouver. Derrick, Donald. 2007. Syllabification and Blackfoot /s/. Proceedings of NWLC 22, pp. 62–76. Simon Fraser University Working Papers in Linguistics. Burnaby. Elfner, Emily. 2006. Contrastive Syllabification in Blackfoot. Proceedings of WCCFL 25, ed. by Donald Baumer, David Montero, and Michael Scanlon, pp. 141–149. Somerville, MA: Cascadilla Press. Fish, Naatosi and Mizuki Miyashita. 1997. Guiding Pronunciation of Blackfoot Melody. Honoring Our Teachers, ed. by Jon Reyhner, Joseph Martin, Louise Lockard, and Willard Sakiestewa Gilbert, pp. 203–210. Flagstaff: Northern Arizona University. Fitzgerald, Colleen. 1997. O’odham Rhythm. PhD thesis, University of Arizona. Frantz, Donald, G. 2017. Blackfoot Grammar. 3rd edition. Toronto: University of Toronto Press. Frantz, Donald G., and Norma Jean Russell. 2017. The Blackfoot Dictionary of Stems, Roots, and Affixes. 3rd edition. Toronto: University of Toronto Press. Goad, Heather, and Akiko Shimada. 2014. In Some Languages, /s/ Is a Vowel. Supplemental Proceedings of the 2013 Annual Meeting on Phonology, ed. by John Kingston, Claire MooreCantwell, Joe Pater, Robert Staubs. DOI: https://doi.org/10.3765/amp.v1i1.50. Hyman, Larry M. 2010. Does Gokana Really Have No Syllables? Or: What’s So Great About Being Universal? Phonology 28(1):55–85. Jensen, John. 1995. Constraints and Opaque Interactions. Cahiers linguistiques d’Ottawa 23:1–9. Kager, René. 1999. Optimality Theory. Cambridge University Press. Kaneko, Ikuyo. 1999. A Metrical Analysis of Blackfoot Nominal Accent in Optimality Theory. MA thesis, University of British Columbia. Kaneko, Ikuyo. 2000. Velar Spirantization and Velar Phonemes in Blackfoot. Algonquian Papers-Archive 31:200–210. Lehiste, Ilse. 1997. The Phonetic Realization of the Haiku Form in Estonian Poetry, Compared to Japanese. Speech Production and Language: In Honor of Osamu Fujimura, ed. by Shigeru Kiritani, Hajime Hirose, and Hiroya Fujisaki, pp. 241–250. Berlin: Mouton de Gruyter. Miyashita, Mizuki. 2018. Vowel-Consonant Coalescence in Blackfoot. Papers of the 47th Algonquian Conference, ed. by Monica Macaulay and Margaret Noodin, pp. 217–235. Ridouane, Rachid. 2008. Syllables without Vowels: Phonetic and Phonological Evidence from Tashlhiyt Berber. Phonology 25:321–359. Roca, Iggy, and H. M. Al-Ageli. 1999. Optimal Metrics. Issues in Phonological Structure: Papers from an International Workshop, ed. by S. J. Hannahs and Mike Davenport, pp. 127–148.

|

178    Mizuki Miyashita

John Benjamins Publishing. Taylor, Allan. 1969. A Grammar of Blackfoot. PhD thesis, University of California, Berkeley. Weber, Natalie. 2013. Accent and Pro-DPS in Blackfoot. Proceedings of the 2013 Annual Conference of the Canadian Linguistic Association. Weber, Natalie. 2016. Accent and Prosody in Blackfoot Verbs. Papers of the 44th Algonquian Conference, ed. by Monica Macaulay, Margaret Noodin, and J. Randolph Valentine, pp. 348–369. Albany, NY: SUNY Press.

Patterns of Cree Preverb Usage in Early Child Language A Longitudinal Case Study Burak Oney

T

he question of what drives first language acquisition has been one of the most fundamental questions in acquisitional studies.1 There have been a number of approaches proposed under certain theoretical stances (see Ambridge and Lieven 2011, for further discussion), and among them, Generativist and Usage-Based accounts are the ones that are widely proposed and investigated. While the former argues for a developmental pattern shaped by ‘grammatical complexity’, also referred to as ‘markedness’, within the latter approach, the emergence and development of a certain part of the grammar (or lexical items) are correlated with its frequency in the input. Although these two accounts have been subject to a relatively large amount of research, much of the data analyzed are coming from only a relatively small number of languages, many of which tend toward the isolating end of the typological scale, mostly Indo-European languages (Kelly et al. 2014). Given the lack of research on polysynthetic languages on this matter, languages with large and complex verbal paradigms such as Cree offer unique puzzles and insights to our understanding of language acquisition. In order to contribute to this under-represented body of research, this paper investigates the acquisition of a class of morpheme called “preverb” for one

|

   179

|

180    Burak Oney

child code-named “Daisy” (3;08.10–5;11.25), who is learning Northern East (NE) Cree (Central Algonquian, ISO 639-3 code crl) as her first language. Previously, the preverb development of a younger child code-named Ani (2;01 to 4;03) was investigated by Brittain and Rose (2021) (henceforth referred to as Brittain and Rose), where they claim that grammatical complexity appears to be a driving factor in determining the child’s development pattern. This study builds on their study and includes one session of child-directed speech (CDS) to see whether Daisy’s linguistic behavior is a feature of acquisition (shows a developmental pattern) or resembles more an adult-like behavior that reflects the frequency distribution of preverbs in the language.

Cree Preverbs and Their Acquisition

Preverbs in NE Cree are a small set of morphemes that precede the verb stem and function to express grammatical (tense, aspect, mood), lexical, and deictic (directional) information.2 All Algonquian languages have preverbs (Bloomfield 1946), and they have been the subject of a relatively large amount of research to date (James 1991; Clarke et al. 1993; Jancewicz and MacKenzie 1998; Pentland 2005, among others). Although preverbs in Cree are phonologically independent words (Goddard 1990), they are part of the verb complex—indeed, traditional descriptive practice is to identify them as elements of the verb stem, despite their status as independent words. The only research to date undertaken on the acquisition of preverbs is Brittain and Rose (2021). In their study, they examine Ani’s corpus (2;01 to 4;03) to investigate the pattern of emergence of the preverb system. According to Brittain and Rose, Ani’s pattern of development suggests a learning path that is driven by grammatical complexity: she starts using preverbs with the potentially less complex independent (2;04:10) seven months before she produces preverbs with the conjunct (2;11:16); she does not employ the rules of initial change to any preverb; and forty-seven of them are a type of preverb that can only occupy one position in the verb stem. Based on this pattern of Ani’s preverb acquisition, Brittain and Rose identify five potential challenges that a Cree learning child may encounter while acquiring the preverb system: (1) the usage of the conjunct in general, and specifically in combination with preverbs; (2) initial change and phonological instability; (3) positional instability; (4) preverb combination; and (5) unique preverb-inflection combinations. In this

|

Patterns of Cree Preverb Usage   181

study, I investigate grammatical complexity regarding preverbs focusing on these areas of grammar of NE Cree. Potential Complexities of NE Cree Preverbs

The first possible difficulty with which a Cree learner may be presented concerns the use of the conjunct inflectional paradigm. First, the conjunct is restricted to certain environments: subordinate and relative clauses, clauses that contain interrogative pronouns (wh-expressions), and some main clauses such as focus constructions. Independent inflection, on the other hand, appears to be employed elsewhere (e.g., main clauses). Another way in which the independent may be less challenging for the learner is the nature of the inflection. The conjunct inflection in NE Cree is fusional while the independent order is more agglutinative.3 Given the fact that agglutinative morphology is argued to be easier to acquire than fusional (Slobin 1982), the independent, again, is potentially easier to acquire for the learner compared to the conjunct. The last potential difficulty of the conjunct with which the Cree learner is presented is initial change. Preverbs may undergo initial change, and changed forms are referred to as “phonologically unstable” by Brittain and Rose as they show an alteration in form, depending on the inflection they are paired with and the position where they appear.4 Additionally, a number of preverbs have no changed forms because they can only combine with the independent order, or they do not appear in initial position in the conjunct: these are the forms referred to by Brittain and Rose as “phonologically stable.” As mentioned earlier, Ani does not produce any changed form of preverbs (Brittain and Rose 2021); all of her preverb productions involve either lexicalized conjunct preverbs or the citation forms of preverbs that are able to undergo initial change. Grammatical preverbs in NE Cree, to the best of our knowledge, can only appear in preverb position; therefore, they are referred to as positionally stable preverbs by Brittain and Rose. Lexical and directional preverbs, on the other hand, have a more complex distribution than grammatical preverbs in that many can be found in either preverb position or initial (root) position. This is a phenomenon referred to as preverb bumping (Goddard 1990:479), the supposed leftward displacement of a morpheme from initial position to the preverb position. In Ani’s speech, forty-seven out of forty-eight preverbs are from the grammatical set (positionally stable). Another potentially complex area of NE Cree grammar is preverb combination. Preverbs in Algonquian languages very commonly appear in combinations of two

|

182    Burak Oney

or three. Ani does not combine preverbs in her production, suggesting that the use of single preverbs precedes in acquisition of the ability to combine. Last, certain preverbs in NE Cree can only combine with one of three inflectional sets, while the others can combine with more than one type of inflection. In Ani’s corpus, for example, there are five preverbs (pâchi (toward), chîh (past), wîh (want to), chîh (able to), and uhchi (past/negation)) that can combine with more than one type of inflection. However, Ani always uses preverbs with only one inflectional order, even in cases where a given preverb combines with more than one order.

Methodology

Chisasibi Child Language Acquisition Study (CCLAS) is the first in-depth longitudinal naturalistic first language (L1) acquisition study of an Algonquian language for which quality data are available.5 Filming occurred between November 2004 and April 2007 with three participants, and approximately fifty-five hours of child speech and CDS have been obtained through around eighty video recordings in total. All CCLAS recordings are stored and processed in Phon (Rose et al. 2006). The present study focuses on a subset of recording sessions of Daisy’s corpus. Daisy was recorded for twenty-seven months in thirty-two sessions, eleven of which form the data set on which this research is based (covering age 3;08–5;07). The eleven sessions under scrutiny are evenly spaced, and the study thus covers a span of approximately twenty-three months. Additionally, in one session the adult’s preverbs were analyzed in order to see a sample of preverb usage in the CDS. For reasons of feasibility, this session is not one of the sessions included in the study of Daisy’s preverb usage. We selected a session for which the adult speech was largely already parsed; because the adult speech is more complex than the child speech, adult speech is not fully parsed in the database but is done on a need-to-do basis. Table 1 details the sessions selected for examining the child and the adult’s speech.

Results

A total of 2,527 verb tokens appear in Daisy’s inventory, 1,434 of which Daisy combines with at least one preverb: 1,264 verbs have one preverb, 168 verbs have

|

Patterns of Cree Preverb Usage   183

TABLE 1. Video recordings selected for study, for Daisy and Adult CHILD’S AGE

DATE OF RECORDING

LENGTH OF THE SESSION (MINS:SECS)

1. 03;08.10 2. 03;09.23 3. 03;11.11 4. 04;02.16 5. 04;04.04 6. 04;06.02 7. 04;07.28 8. 04;10.03 9. 05;00.20 10. 05;04.12 11. 05;07.03 12. 05;05.03

2004-12-14 2005-01-27 2005-03-15 2005-06-10 2005-08-08 2005-10-06 2005-12-02 2006-02-07 2006-04-24 2006-08-16 2006-11-07 2006-09-07

35:48 37:16 58:32 50:25 40:04 41:58 43:28 43:25 35:17 33:31 32:26 31.37

CDS is examined only in the last session (12), while the rest were used for Daisy’s speech.

two preverbs, and two verbs have three preverbs. In total, she produces 1,603 preverb tokens, representing fourteen different morphemes: twelve grammatical, two directional, and two lexical. As Table 2 shows, Daisy uses preverbs from the earliest session (age 03;08.10). There is no observable pattern of increase in the number of any of the preverb types, although the findings show that there are preverbs she uses frequently right from the first sessions (03;08.10), and preverbs that are fairly uncommon in her inventory. The most preverb diversity is observed at age 03;11.11 and 4;10.03 with sixteen and seventeen different types respectively, and the least variety is found at age 03;08.10, the earliest session in this study, as well as at ages 05;04.12 and 05;07.03, with eleven and twelve different types respectively.6 On average, Daisy makes use of 13.5 different types across the sessions. Table 3 tabulates the number of tokens of each preverb type, and their proportions relative to her entire preverb production. The vast majority of Daisy’s preverb production consists of grammatical preverbs, accounting for 92.07% (1,476/1,603). Directional and lexical preverbs, on the other hand, constitute only 7.91% (127/1,603) of Daisy’s inventory. The conjunct preverb kâ is the most frequent preverb in Daisy’s inventory, at 23.58% (378/1,603) of the total number of preverbs. It is present in every recording session and, correctly, appears only with the conjunct order. When examining the usage of

|

184    Burak Oney

TABLE 2. Preverb production in Daisy’s speech AGE

NUMBER OF PREVERBS

LENGTH OF THE SESSION

03;08.10 03;09.23 03;11.11 04;02.16 04;04.04 04;06.02 04;07.28 04;10.03 05;00.20 05;04.12 05;07.03 Total

79 167 308 148 171 96 159 151 135 42 147 1603

35:48 37:16 58:32 50:25 40:04 41:58 43:28 43:25 35:17 33:31 32:26

kâ over time, numbers fluctuate in the sessions in between so that no meaningful pattern of increase is observed. Although the distribution and functions of the conjunct preverb kâ are not yet fully described for adult grammar, we can observe that it is used in at least the following environments: in main clauses (in past-tense wh-questions and focus constructions), in dependent clauses (in complement clauses and adverbial clauses), and in relative clauses (among others, see Brittain 2001 for western Naskapi). Daisy uses the preverb kâ in each of these contexts with a variety of verb stems. Below I provide two instances of kâ in main clause focus construction (1) and in past-tense wh-construction (2).7 (1) Mâutih wîyi this

kâ

api-t

James tâtâuch anitâh tâtâuch

pron.3 lic.pvb sit.ai-cin.3sg name middle here middle

‘Here is where James sat, he sat in the middle.’ (2) Awân kâ

(Daisy, 03;11.11)

wîchâw-ik

wh lic.pvb accompany.ta-cin.s:1sg/o:3sg ‘Who did I go with?’

(Daisy, 04;02.16)

The next most frequent preverb type in Daisy’s inventory is chîh (past tense), produced in 269 utterances (16.78%), and it is always found denoting past tense

|

Patterns of Cree Preverb Usage   185

TABLE 3. Daisy’s preverb inventory, by preverb type PREVERB TYPE

PREVERB

Conjunct neutral preverb âh Conjunct preverb

kâ

Tense

chîh uhchi châ châ

Modality

Aspect

ki chiki kiti chîh wîh pih chipih chîh ati chîsh

Directional Lexical Total

pâchi uhchi ishi wîchi

FUNCTION/MEANING

TOKENS

(% OF TOTAL N)

complementizer/relativizer complementizer/relativizer, past tense in some contexts, focus constructions past tense past tense, negated clause future future, functioning as imperative (second-person form only) independent non-third future independent third future future ‘able to’ ‘want to’ non-third ‘should, could’ third ‘should,’ ‘could’ ‘should have’ inchoative completive ‘completed doing something’ ‘toward’ speaker ‘away from’ speaker ‘in a certain manner’ ‘together’

218

13.59%

378

23.58%

269 28 109

16.78 1.74% 6.79%

48

2.99%

199 50 3 38 101 2 6 3 9

12.41% 3.11% 0.18% 2.37% 6.3% 0.12% 0.37% 0.18% 0.56%

15

0.93%

95 22 5 5 1603

5.92% 1.37% 0.31% 0.31% 100%

on a great number of different verb stems. It is also one of Ani’s earliest preverbs, appearing first at the age of 2;07.06. Daisy mostly produces chîh (past) in isolation, using it singly in 217 utterances. I show one occurrence of chîh (past) with the independent verb below in (3). (3) Chîh chishwâwâpiyiu pvb.past make.loud.noise.ai.iin.3sg ‘She was very loud.’

(Daisy, 03;08.10)

|

186    Burak Oney

The conjunct preverb âh is produced by Daisy at the rate of 13.59% (218/1,603). Similar to kâ, the understanding of the distribution and functions of âh is not complete; however, âh seems to occur in certain types of subordinate clauses such as purpose clauses, when-clauses, relative clauses that have a habitual interpretation, and gerund and infinitive structures (Brittain and Rose 2021). In Daisy’s speech, this preverb appears in all these contexts. Below in (4), she uses âh in a relative clause construction. (4) Awân utih âh

âpitisî-t

nâpâu

wh here lic.pvb work.ai-cin.3sg man ‘A man that works here.’

(Daisy, 03;09.23)

All the other conjunct and tense preverbs are present in all the sessions, but her use of châ in imperative construction is sporadic and does not begin until age 03;11.11. Daisy’s first token of châ in imperative is given in (5). (5) Châ kiniwâyim-it lic.pvb.fut care.for.ta-cin.s:2sg/o:3sg ‘Look after her.’

(Daisy, 03;11.11)

Among the other grammatical preverbs, certain preverbs appear less than ten times in total, and these preverbs are kiti (independent third allomorph), pih (should, could, non3), chipih (should, could, 3), chîh (should have), and ati (inchoative).8 While directional preverbs are used productively on a variety of verb stems, the usage of lexical preverbs is quite limited: she uses wîchi (together) and ishi (relative root) in five utterances, and both are produced in only one session, though the former is always attached to the English root “school” while the latter always appears attached to another preverb. The CDS, on the other hand, consists of a total of 293 utterances. In these utterances, 164 tokens of verbs (with or without a preverb) and 109 tokens of preverbs are found. The adult produces fourteen different types of preverbs from all three sets (i.e., grammatical, lexical, and directional). The majority of preverb production comes from the grammatical set (92.66%), and within this set, tense and conjunct preverbs are the most frequent types, comprising 87.14% (47.70% and 39.44% respectively) of the preverb production. Directional and lexical preverbs, on the other hand, constitute only 7.34% of the preverb production altogether

|

Patterns of Cree Preverb Usage   187

(8/109). The preverb kâ is the most frequently used type in CDS with thirty-three tokens overall, appearing in main clauses (13/33), relative clauses (8/33), past-tense wh-clauses (7/33), past-tense subordinate clauses (5/33), and the other conjunct preverb châ occurs in the adult corpus seventeen times. Tense preverbs chîh (past) and ki (future non3) follow the conjunct preverbs in frequency, appearing in fourteen and thirteen utterances respectively. The adult uses the preverb âh in subordinate clauses such as gerunds (3/10), infinitives (2/10), when-clauses (1/10), relative clauses (1/10), and wh-clauses (1/10). Other types of preverbs occur in CDS less frequently, forming 12.84% of all preverb production. Inflectional Orders

Daisy uses preverbs frequently with conjunct and independent verbs, but relatively little with imperative verbs; since grammatical preverbs are incompatible with imperative verbs (Jancewicz and MacKenzie 1998), and most of her preverbs are grammatical, this is expected. Table 4 shows the total number of inflectional orders for each session, for verbs that are combined with a preverb. As shown in Table 4, even though both paradigms are quite abundant, Daisy pairs preverbs with the conjunct more than she does so with independent verbs, and in eight sessions, the conjunct is used more than the independent. However, note that the majority of Daisy’s preverb-conjunct productions (646/694, 93.08%) consist of only the three conjunct preverb: âh, kâ, and châ. She combines eight other preverb types with the conjunct, but these constitute only 6.91% (111/222) of the forms where she uses a preverb with a conjunct verb. Table 5 tabulates the number of tokens of each preverb. Similar to the conjunct, she uses eleven different preverb types with the independent. However, more diversity is observed in the usage of the independent: five types of preverbs make up 94.02% of her preverb production with the independent, while 5.97% come from the other six types. Table 6 shows the number of each preverb with the independent for each session. Daisy uses preverbs with the imperative in just six sessions, in fourteen utterances, using just three types of preverbs. Except for directional pâchi (toward), no preverb appears in more than one session. In CDS, all three orders are present and similar to Daisy; the conjunct and independent are quite prominent. The most frequent order is the conjunct, employed in sixty-five tokens of verbs, accounting for 63.72% of her verb production that includes a preverb or preverbs. Following is the

|

188    Burak Oney

TABLE 4. Daisy’s use of single preverbs with different inflectional orders per session AGE

CONJUNCT INDEPENDENT IMPERATIVE TOTAL

03;08.10 03;09.23 03;11.11 04;02.16 04;04.04 04;06.02 04;07.28 04;10.03 05;00.20 05;04.12 05;07.03 Total

29 83 116 43 84 44 79 57 51 16 92 694

25 59 109 77 43 37 68 55 38 20 26 557

1 4 2 2 2 2 13

55 146 227 120 129 83 149 112 89 36 118 1264

In addition, she produces 108 conjunct and 62 independent verbs with multiple preverbs.

TABLE 5. Daisy’s preverbs combined with a conjunct verb PREVERB ÂH TOTAL

178

PREVERB

UHCHI PAST. NEG

TOTAL

1

KÂ

CHÂ FUT

CHÂ IMP

CHÎH PAST

CHÎH SHOULD HAVE

326

98

44

2

ATI INCH

PÂCHI TOWARD

1

CHÎSH COMP

3

3

5

WÎH WANT TO

13

UHCHI FROM

20

TABLE 6. Preverbs combined with an independent verb PREVERB

CHÎH PAST

KI FUTURE NON3

CHIKI FUT 3

KITI FUT 3

CHÎH SHOULD HAVE

UHCHI PAST NEG

TOTAL

221

169

50

1

1

24

PREVERB

CHÎH ABLE

TOTAL

8

WÎH WANT

CHÎSH CMP

55

15

PÂCHI TOWARD

9

WÎCHI TOGETHER

4

independent, occurring in thirty-two tokens of utterances with a preverb, and two tokens with multiple preverbs, forming 33.33% of the adult’s preverb production. Last, imperative order appears in only three utterances.

|

Patterns of Cree Preverb Usage   189

Initial Change and Preverbs

In total, Daisy produces 237 tokens of verb complexes affected by the synchronic process of initial change, including thirty-four preverbs.9 Details of how she uses initial change with verb stems remain to be investigated by future work, but these figures show that she is using initial change in both locations on the verb complex: affecting the stem and to lesser extent preverbs. The thirty-four changed form preverbs involve four preverb types: wâh (IC.wîh), wâhchi (IC.uhchi (from)), âti IC.ati (inchoative), and piyâchi (IC.pâchi (toward)). The changed form wâh is found in nine utterances in five sessions, and the variety of the verbs stems suggests a productive usage of this changed form. The example in (6) shows her first utterances with wâh. This appears in a main clause, which is likely to be a focus construction because of the first word being the demonstrative mâu; Brittain and Rose (2021) discuss examples of Ani using main clause conjunct verbs in the same kinds of structures. (6) Mâu wâh

utin-imân

this ic.pvb.want take.ti-cin.s:1sg/o:0sg ‘This is what I want to take.’

(Daisy, 03;08.10)

The preverb wâhchi is the changed form of uhchi (from), which often appears in wh-questions, forming part of the interrogative phrase. This form appears first at the age of 03;11.11, it is the most frequently changed form in Daisy’s speech, accounting for 61.76% (21/34) of her changed form preverbs. It is mostly employed in questions that includes the wh-pronoun châkwân~châkwâ ‘what’ (châkwân wâhchi = ‘why’), as shown in (7). (7) Châkwâ-yiu wâhchi wh-obv

îsi-nihkâsu-t

anitâh

ic.pvb.from rr-name.ai-cin.3sg that

‘Why is he called (like) that?’

(Daisy, 04;10.03)

The preverb wâhchi is the only changed form in CDS, produced four times with four different verb stems in the session under consideration. The remaining two changed forms, âti and piyâchi, appear less frequently, occurring in three and one utterances consecutively.

|

190    Burak Oney

Preverb Combination

Daisy quite frequently uses constructions accommodating multiple preverbs: thirty-seven different combinations of two preverbs are found in 176 preverb tokens in Daisy’s speech, though twenty types of 2-preverb sequences only appear in one or two contexts. However, here again, there is no pattern suggesting any kind of development. The most diversity is found at the age of 03;11.11 (15 different types of combination) and the least is at the age of 05;04.12 (three different types). The most common preverb combination in Daisy’s speech is kâ+pâchi (toward), with twenty-nine tokens in eight sessions. I provide one instance of this combination in (8). (8) Nîyi kiyâh kâ 1

pâchi

nitiwâpim-it

ni-wîchâwâkin

and lic.pvb pvb.toward go.see.ta-cin.s:3sg/o:1sg 1-friend (Daisy, 03;11.11)

‘My friend came to see me too.’

The sequence âh+chîh (past) is the second most frequent type in her corpus, produced in twenty-two utterances in seven sessions productively. It is mostly employed in subordinate clauses with the past meaning, as in (9). (9) Wîyi mâk âh

chîh

nihchinâkunîsu-t nimâ

pron.3 and lic.pvb pvb.past dress.ai-cin.3sg right ‘She was the one who dressed herself, right?’

(Daisy, 04;10.03)

Additionally, Daisy is found combining three preverbs in her speech, though it’s quite limited: only one type, ki (future non3)+chîh (able to)+pâchi (toward), occurs, and she produces only two tokens of this combination. Preverb combination is also present in the adult input: seven tokens appear with six different types of combination.

Discussion

Daisy produces a great number of preverbs with great variety from the earliest session, and all supposedly more challenging forms that are argued by Brittain and Rose are present from the earliest session; therefore, no evidence of development is found in Daisy’s speech. From the beginning, Daisy is adept with preverbs,

|

Patterns of Cree Preverb Usage   191

and her use of preverbs seems very much like the adult. Of twenty preverb types present in Daisy’s inventory, twelve of them are absent in Ani’s corpus, and six do not appear in the CDS. Of these six preverbs that are not present in the CDS, five types (chîshi (completive), chipih (should, could, 3), kiti (future 3), chîh (should have), pih (should, could, non3)) are also among the least frequent preverbs that we do not have evidence of productivity for Daisy, which suggests an adult-like behavior for Daisy. In terms of inflectional paradigms, even though both orders are quite abundant in Daisy’s speech, she pairs preverbs with the conjunct (901/1,579) more than she does so with independent verbs (662/1,579). A similar pattern is also found in CDS, in which the adult uses preverbs with conjunct verbs (65/102) more than independent verbs (34/102). As mentioned previously, Ani produces preverbs more with the independent (27/48) than conjunct (20/48), and her early production of preverbs is, except for one token with imperative, paired with independent forms (Brittain and Rose 2021). In Daisy’s speech, as mentioned earlier, thirty-four tokens of changed forms are found with four preverb types (wîh (want to), uhchi (from), ati (inchoative), pâchi (toward)). Among them, uhchi (from) is the most frequent form (21/34) and is the only changed form in the CDS, with four tokens. Conjunct and tense preverbs seem to be the most frequent preverb types in NE Cree, though it remains to be established by future research what the distribution is in the adult language. Crucially, these frequent preverbs are not subject to the synchronic process of initial change, therefore, it is not surprising that a few changed forms appear in Daisy’s speech. When the verb stems are examined, 237 tokens of changed forms have been found in Daisy’s corpus. Therefore, it is clear that she knows how to apply initial change, but given that synchronic changed preverbs are not common, she only produces a handful of them. The majority of Daisy’s preverbs are of the grammatical set, forming 92.07% (1,476/1,603) of her preverb production. Directional and lexical preverbs, on the other hand, make up only 7.29% (117/1,603) and 0.62% (10/1,603) of Daisy’s inventory. These findings for Daisy are consistent with findings for Ani, for whom 47/48 of preverbs in her inventory were of the grammatical type. Brittain and Rose (2021) speculate that the reason for Ani’s early use of grammatical over lexical preverbs could be that they are phonologically and positionally stable, and thus easier for the child to learn. However, Ani’s and Daisy’s use of mostly grammatical preverbs may reflect the distribution of these types of preverb in the language in general (or,

|

192    Burak Oney

at the very least, the distribution of preverbs in the CDS).10 This is, indeed, what is found in one session of the CDS. Quite similar to Daisy, the majority of preverbs are of the grammatical set (92.66%), while directional and lexical preverbs form only 7.34% of the adult’s preverb production altogether (8/109). This suggests that if we take the CDS to be reflective of the language in general, here again, Daisy’s developmental path is parallel to the probable distribution of preverbs in similar genres (casual conversation) in the language. As for preverb combination, even though Daisy’s corpus clearly starts after the onset of her ability to use preverbs, we would expect to see some evidence of her being able to use this area of the grammar in increasingly more sophisticated ways over time. However, here again, there is no pattern suggesting any kind of development: unlike Ani, Daisy quite frequently uses constructions accommodating multiple preverbs. Preverb combination is also present in the adult input: seven tokens appear with six different types of combination. Therefore, in terms of preverb combination, Daisy and Ani behave quite differently, and again, Daisy’s use of preverbs is very much adult-like in this area of the grammar, as multiple preverbs are quite common in NE Cree (Brittain and Rose 2021). Another interesting finding is that in Daisy’s corpus, contrary to Ani who never combines preverbs with more than one order, five preverb types (chîh (past), chîh (should have), uhchi (past neg.), wîh (want to), chîsh (completive)) appear with both conjunct and independent verbs, one type (ati (inchoative)) appears with the conjunct and imperative, and one type (pâchi (toward)) occurs with all three inflectional orders. Even though Daisy is able to pair preverbs with multiple orders, she does not do this very often. I attribute this to the fact that most frequent preverbs are restricted to occurring with only one inflection order; kâ, âh, and châ only occur with the conjunct, and ki (future non3) and chiki (future 3) only occur with the independent. This is also reflected in the CDS, in which only one token appears with two orders: uhchi (from) with one token in a conjunct verb, and five tokens in an independent verb.

Conclusion

In summary, Daisy is quite adept with preverbs, using them greatly in a good number of contexts from the earliest session. Given the level of the analysis presented here, and the volume of data reviewed, no clear pattern of development over time is

|

Patterns of Cree Preverb Usage   193

evident in Daisy’s speech like we have in Ani’s speech; the forms that were proposed to be potentially more complex by Brittain and Rose (appearing late in Ani’s inventory) are used by Daisy from the beginning. In general, Daisy’s production is more like what is found in the single session of the CDS examined: the most frequent preverbs in Daisy’s speech are also the most common preverbs in the adult’s speech, and preverbs that are not present in CDS are the ones that are least frequent in Daisy’s speech; preverbs are used more with the conjunct, which follows from the frequency of the conjunct preverbs; changed forms (excluding lexicalized conjunct preverbs) are relatively infrequent; and grammatical preverbs make up the majority of preverb production for Daisy, the adult, as well as Ani. Additionally, given that preverb combination is common in NE Cree, her ability to combine preverbs with such great diversity is, again, reflective of the language. Last, given that most of the frequent preverbs can only be paired with one particular order, and that very few preverbs appear in combination with more than one inflectional order in Ani, Daisy, and the adult’s speech, I speculate that the appearance of a preverb with multiple inflectional orders is, indeed, relatively infrequent in the language. Notes 1. Sincere thanks to Julie Brittain, Phil Branigan, and Will Oxford for their insightful feedback, comments, and questions, and many thanks to the language consultant Alice Duff. This research has received financial support from funding awarded to me from Dr. Brittain’s Social Sciences and Humanities Research Council of Canada Insight Grant (#435-2013-1297), and from a Graduate Fellowship from the School of Graduate Studies, Memorial University. Any errors are my own. 2. See Junker et al. (2015) for a detailed list of preverbs with their usages and examples. 3. As one of the reviewers pointed out, there are certain aspects of independent order that may pose a challenge for the learner. For example, independent pronominal inflections (many of them, at least) are circumprefixal, which is typologically rare and arguably more complex, while the conjunct pronominal inflections are not. However, as I follow Brittain and Rose’s diagnostics for complexity in this study, I leave this issue for further investigation. 4. The conjunct preverbs kâ, âh, and châ are considered lexicalized changed forms (Brittain and Rose 2021) as they have no unchanged (citation) for counterparts that can be derived transparently via the synchronous process of initial change. 5. See Brittain et al. (2007) for more information on CCLAS and the data-processing

|

194    Burak Oney

methodology. 6. Of these 12 grammatical preverbs, two are subject to allomorphy: first, the independent future forms are ki~chiki dependent on person features, with an alternation of chiki~kiti based on dialect variation, and second, modal ‘should/could’ has two forms pih~chipih also dependent on person features. 7. The following abbreviations are used in data glossing: 0 = inanimate proximate singular, 1 = first person, 2 = second person, 3 = third person animate, AI = animate intransitive, CIN = conjunct indicative neutral, FUT = future tense, IIN = independent indicative neutral, IC = initial change, LIC = lexicalized initial change, O = object, OBV = obviative, PRON = pronoun, PVB = preverb, RR = relative root, S = subject, SG = singular, TA = transitive animate, TI = transitive inanimate, WH = wh word. 8. As mentioned earlier, the modal preverb has two allomorphs, pih~chipih, the former occurs with first- and second-person subjects while the latter with the third-person subject. 9. In this number, I do not count the lexicalized changed form conjunct preverbs âh, kâ, châ. 10. Note that no study has been made on Cree adult-to-adult interactions to see how it may differ in nature from Cree CDS. Strictly speaking, we do not know whether the CDS captured in CCLAS is accommodating to the children, differing from the general distribution.

References Ambridge, Ben, and Elena Lieven. 2011. Child Language Acquisition: Contrasting Theoretical Approaches. Cambridge: Cambridge University Press. Bloomfield, Leonard. 1946. Algonquian. Linguistic Structures of Native America, ed. by Henry Hoijer and Cornelius Osgood, pp. 85–129. New York: Viking Fund Publications in Anthropology 6. Brittain, Julie. 2001. The Morphosyntax of the Algonquian Conjunct Verb: A Minimalist Approach. New York: Routledge. Brittain, Julie, Carrie Dyck, Yvan Rose, and Marguerite MacKenzie. 2007. The Chisasibi Child Language Acquisition Study (CCLAS): A Progress Report. Papers of the 38th Algonquian Conference, ed. by H. C. Wolfart, pp. 1–17. Winnipeg: University of Manitoba. Brittain, Julie, and Yvan Rose. 2021. The Development of Preverbs in Northern East Cree: A Longitudinal Case Study. First Language 41(4):376–405. Clarke, Sandra, Marguerite MacKenzie, and Deborah James. 1993. Preverb Usage in Cree/

|

Patterns of Cree Preverb Usage   195

Montagnais/Naskapi. Papers of the 24th Algonquian Conference, ed. by William Cowan, pp. 32–45. Ottawa: Carleton University. Goddard, Ives. 1990. Primary and Secondary Stem Derivation in Algonquian. International Journal of American Linguistics 56(4):449–483. James, Deborah. 1991. Preverbs and the Function of Clauses in Moose Cree. The 1990 Belcourt Lecture. Winnipeg: Voices of Rupert’s Land. Jancewicz, Bill, and Marguerite MacKenzie. 1998. Preverbs in Naskapi: Function and Distribution. Papers of the 29th Algonquian Conference, ed. by David H. Pentland, pp. 150–168. Winnipeg: University of Manitoba. Junker, Marie-Odile, Luci Salt, and Marguerite MacKenzie. 2015. East Cree Verbs (Northern Dialect). [Revised and expanded from 2006 original edition]. The Interactive East Cree Reference Grammar. https://www.eastcree.org/cree/en/grammar/northern dialect/verbs/ preverbs/ Kelly, Barbara, Gillian Wigglesworth, Rachel Nordlinger, and Joe Blythe. 2014. The Acquisition of Polysynthetic Languages. Language and Linguistics Compass 8(2): 51–64. Pentland, David H. 2005. Preverbs and Particles in Algonquian. Papers of the 36th Algonquian Conference, ed. by H. C. Wolfart, pp. 323–338. Winnipeg: University of Manitoba. Rose, Yvan, Brian MacWhinney, Rodrigue Byrne, Gregory Hedlund, Keith Maddocks, Philip O’Brien, and Todd Wareham. 2006. Introducing Phon: A Software Solution for the Study of Phonological Acquisition. Proceedings of the 30th Annual Boston University Conference on Language Development, ed. by David Bamman, Tatiana Magnitskaia, and Colleen Zaller, pp. 489–500. Boston: Cascadilla Press. Slobin, Dan I. 1982. Universal and Particular in the Acquisition of Language. Language Acquisition: The State of the Art, ed. by Eric Wanner and Lila R. Gleitman, pp. 128–170. Cambridge: Cambridge University Press.

Short Verbs in Algonquian Richard A. Rhodes

T

he conventional wisdom about Algonquian languages is that traits shared between the various languages are a product of diffusion (see, e.g., Proulx 1984). But with Goddard’s (1980) compelling morphological argument for Eastern Algonquian as a genetic subgroup, followed by his (1994) analysis of phonological innovations across the family, the outlines for a genetic classification of the family on phonological and morphological grounds have become clear. Rhodes (2021) extends Goddard’s lines of argument to establish Core Central Algonquian, consisting of Ojibwean, Miami-Illinois, Meskwakian, and Shawnee, as a genetic subgroup next to Eastern Algonquian and cites reasons to believe that both arose from a single dialect with common innovations. Both Goddard and Rhodes also cite lexical evidence consistent with their respective analyses. But looking across the whole vocabulary, the lexical picture is complex due to borrowing. There is substantial evidence for borrowing between languages and dialects in significant amounts. Bloomfield (1962:21–22) cites Ojibwe influences and borrowings in Menominee. Rhodes (1989) discusses the unexpectedly large number of lexical similarities between Cree dialects and Ojibwe dialects. Pentland (2003),

|

   197

|

198    Richard A. Rhodes

cited in Rhodes (2020b), demonstrates that borrowings between varieties of Cree and Ojibwe run in both directions, continuing into historic times. Rhodes and Costa (2003) show how number terminology spread from Ojibwe as far as Cheyenne. Costa (2013) shows that there was extensive borrowing among the southern Great Lakes languages, Meskwaki-Sauk-Kickapoo, Miami-Illinois, Shawnee, and Potawatomi. One of the unnoticed facts about this line of research is the relative lack of verb examples. This is not without reason. Rhodes (1989, 2017) attempted comparisons of Algonquian vocabulary based on the Swadesh two hundred–word list (Swadesh 1971). Rhodes pointed out significant problems arising from both stem agreement and the bipartite structure of Algonquian verbs for the Swadesh list methodology. Biedny et al. (2022), using the Leipzig-Jakarta list (Tadmor 2009) attempted a statistical analysis of lexical similarity across a small set of languages, simply omitting verbs for similar reasons. But no real picture of Algonquian lexical relations can be developed absent some meaningful way to compare verbs. Rhodes (2020a) is a step in that direction. It is an in-depth study of a small number of verb initials showing that initials are independently borrowable. They behave as if they were parts of phrases rather than bound morphemes. This paper takes up another part of the problem in comparing Algonquian verbs. The core question here is: what are the most basic verbal meanings across the family? To the extent that basic vocabulary is classified as such because it occurs at a high frequency, a significant portion of basic vocabulary might be expected to differ across language families because of culture, especially where those cultures have distinct subsistence regimens. Farmers talk about different things than hunter-gatherers. That means that a universalist approach to basic meanings, like the Swadesh list or the Leipzig-Jakarta list, is too blunt an instrument. I propose, instead, to use length as the independent measure to arrive at a list of the most basic verb meanings in Algonquian. The logic is simple. Zipf’s law of abbreviation holds that word length inversely correlates with frequency (Zipf 1949). Zipf included three unrelated North American languages as test cases: Nootka, Plains Cree, and Dakota (actually Lakhota). All three show a clear statistical correlation between length and frequency (1949:78–86). Of course, Zipf recognized that inflection was a potential distorting factor, so he measured the length of stems rather than words. (He developed a notion of “holophrase,”

|

Short Verbs in Algonquian   199

the maximal inflectable form of a lexeme, or what linguists generally refer to as a stem.) Based on Zipf’s work, we can make the following equation: the shortest is the most frequent is the most basic. Length in Algonquian stems, however, can be measured in one of two ways: either by the number of phonemes or by the number of morphemes, and these measures are independent. For example, the stem of Ojibwe onzo ‘it (an.) is boiling’ is four phonemes long and consists of three morphemes, as shown in (1a), while the stem of Ojibwe nookizi ‘it (an.) is soft’ is six phonemes long but consists of only two morphemes, as shown in (1b). (1) Ojibwe examples of stem length by morpheme a. onzo ‘it (an.) is boiling’

[on-z-o]-w



[boil-by.heat-an]-3



[stem]-inflection

b. nookizi ‘it (an.) is soft’ [nook-izi]-w

[be.soft-an]-3



[stem]-inflection

Given the morphological and semantic complexity of Algonquian verbs, counting morphemes yields better results. Therefore we will call a verb stem that contains a single lexical morpheme a short verb, regardless of the number of phonemes it contains. Thus, Ojibwe nookizi ‘it (an.) is soft’, with one lexical morpheme nook- ‘soft’ is short and Ojibwe onzo ‘it (an.) is boiling’, with two lexical morphemes, on- ‘boil’ and -z- ‘by heat’, is not. A list of basic verb meanings for Algonquian can be developed comparatively. By collecting the short verbs in each language and selecting the most widely attested meanings, we arrive at a master list of basic meanings for the family. This method is available for Algonquian languages because the vast majority of Algonquian verb stems have at least two lexical parts. Examples from a range of languages are given in (2) and (3).1 The forms in (2) are stems with one lexical morpheme. Those in (3) have two. This is not the same distinction as initial and final. All initials are lexical. But only concrete finals are lexical. Abstract finals are not, as I discuss below.2

|

200    Richard A. Rhodes

(2) Stems with a single lexical morpheme a. Munsee

b. Meskwaki

pákeew ‘it is flat’

matanêwa ‘s/he overtakes h/’

[pak-ee]-w [mata-n]-ê-wa flat-inan-3 c. Plains Cree

overtake-an-an.obj-3.an

d. Cheyenne

akohcin ‘s/he is in the water’

é-noōhta ‘s/he left it’

[akoht-in]-w é-[noo-ht]-á

be.in.water-an-33 3-leave-inan-inan

(3) Stems with two lexical morphemes a. Miami

šoowiteehiaani ‘I worry’



[šoow-iteeh-ee]-aani



sore-heart-an-1

b. Oji-Cree

otihkopinaan ‘s/he pulls h/ up onto something higher’

ot-[ihko-pi-n]-aa-an 3-at.a.higher.place-pull-an-an.obj-obv c. Shawnee

halemahkamikatwi ‘it begins’



[halem-ahkamik-at]-wi

starting-be.an.event-inan-3.inan d. Menominee

keesesam ‘s/he cooks it’



[kees-es]-am-w



finished-use.heat.inan-inan-3

It is important to note that taking verb stems as object of study entails the need to attend to each of the four possible stem forms for any given meaning: inanimate intransitive (II), animate intransitive (AI), transitive inanimate (TI), and transitive animate (TA). The last morpheme in each of the stems in (2) and (3) marks animacy agreement. Where that morpheme is not realized as a portmanteau with a lexical morpheme, it is called an abstract final. In practice, transitivity turns out to be more important than animacy, because almost all meanings can be predicated of both animates and inanimates. But for

|

Short Verbs in Algonquian   201

any given meaning, there is an inherent, or notional, transitivity. Simple motion verbs are inherently intransitive, predicating only a theme, the entity moving, whereas a notion like ‘hold’ is inherently transitive. Holding is a relation between the holder and the thing held. We use the term notional transitivity to reflect the number of participants in the semantic frame. Notional transitivity is distinct from grammatical transitivity. For example, Ojibwe wiisini ‘he eats’ is grammatically intransitive but refers to a notionally transitive event. It is important to keep notional transitivity in mind, because certain transitive meanings are expressed in a given Algonquian language with a morphologically intransitive verb form. Bloomfield (1946:95) called such verbs pseudo-transitives, now generally known as AI+Os, animate intransitives with objects. An example from our emergent set of basic verbs is notionally transitive ‘roast’. In Plains Cree and Munsee the verb stems expressing ‘roast’ are AI+O, but in Cheyenne, Menominee, Ojibwe, Meskwaki, and Miami-Illinois the verb stems are morphologically transitive. The data are given in (4). (4) Terms meaning ‘roast s.t.’ a. Expressed as AI+O

i. Plains Cree

apwēw ‘s/he roasts (it)’ [apw-ē]-w heat-an-3

ii. Munsee

wtapoosiin ‘s/he roasts it’ wŭt-[apw-usi]-n

3-roast-an-obj

b. Expressed with grammatical transitive

i. Cheyenne

é-honoto ‘s/he roasts it (an)’ é-honóhta ‘s/he roasts it’ é-[honó-t]-ó

é-[honó-ht]-á

3-roast-an-3.obv

3-roast-inan-inan



ii. Menominee

apuanεεw ‘s/he roasts it (an)’ apuatam ‘s/he roasts it’

[apwaa-n]-εε-w

[apwaa-t]-am-w

roast-an-an.obj-3 roast-inan-inan-3

|

202    Richard A. Rhodes

iii. Ojibwe

odabwânaan ‘s/he roasts it (an)’ odabwaadaan ‘s/he roasts it’



od-[abwaa-n]-aa-an



3.erg-roast-an-an.obj-obv 3.erg-roast-inan-inan-obj

od- [abwaa-d]-am-n

iv. Meskwaki

apwânêwa ‘s/he roasts it (an)’ ’apwâtamwa ‘s/he roasts it’



[apwâ-n]-ê-wa

[apwâ-t]-am-wa



roast-an-an.obj-3.an

roast-inan-inan-3.an

3

v. Illinois (arch.)

nintapwaalaa ‘I roast it (an)’ nintapwaataan ‘I roast it’



nint-[apwaal]-aa

nint-[apwaat]-am -n



1-roast.an -an.obj

1-roast.inan -inan.obj-obj

As is obvious from the forms in (4), the root for ‘roast’ across these languages is broadly cognate, including the Cheyenne stem forms.4 Only Miami has a non-cognate root. It is also worth noting that differences in morphosyntax like those in (4) also exist between dialects. Different varieties of Ojibwean choose one or the other strategy (or both). See Table 1 below. At this point, the notion of a short verb needs to be refined. The basic idea is that the stem may only contain one lexical morpheme. But there are also secondarily derived verb stems that contain only one lexical morpheme (Goddard 1990). Following the principle that short verbs are also semantically basic, it needs to be further specified that a secondarily derived stem does not count as a short stem. Even if your theory holds that animacy agreement is derivational rather than inflectional, it is primary derivation (Goddard 1990). In fact, the animacy agreement marker, when required, is the morph that completes a stem. TABLE 1. Morphosyntax of the ‘roast’ root in varieties of Ojibwe

Northwest Odawa Nipissing Oji-Cree Eastern Southwest

AI+O

TA/ TI

+ + + + – –

– – – + + +

|

Short Verbs in Algonquian   203

The condition that a short stem is not secondarily derived helps to resolve some difficult cases. One potential candidate for being a short stem is the reflex of PA *a·čime·wa ‘s/he tells about h/’. Reflexes of PA *a·čime·wa are widely attested across the family. (5) PA *a·čime·wa ‘s/he tells about h/’ PC ācimēw, Ch. é-hósemóho, Men. aacemεw, M-I aacimeewa, Mesk. âčimêwa

However, this form is ambiguous between having the structure root + abstract final and having a bipartite structure. The two possible interpretations for the structure of the Plains Cree reflex are given in (6). (6) Possible interpretations of Plains Cree âcimêw ‘s/he tells about h/’ a. As short verb âcimêw ‘s/he tells about h/’ [ât-im]-ê-w tell.about-an-an.obj-3 cf. akohcimêw ‘s/he puts h/ into water’ [akoht-im]-ê-w be.in.water-an-an.obj-3 b. As bipartite verb âcimêw ‘s/he tells about h/’ [ât-im]-ê-w tell.about-by.speech.an-an.obj-3 cf. kisîmêw ‘s/he makes h/ angry by what she says’ [kisî-im]-ê-w be.angry-by.speech.an-an.obj-3

The analysis is disambiguated in favor of the short verb analysis by the structure of the TI form, shown in (7b). It shows that the im in the TA is an abstract final as proposed in (6a) and not a concrete final as in (6b). (7) Plains Cree âcimêw as short stem a. âcimêw ‘s/he tells about h/’ (TA) [ât-im]-ê-w tell.about-an-an.obj-3

|

204    Richard A. Rhodes

b. âtotam ‘s/he tells about it’ (TI) [ât-ot]-am-w tell.about-inan-inan-3

Somewhat trickier are the words for ‘hear’ and ‘listen to’ in (8) and (9), both candidates for short word status, and attractive as basic for intuitive reasons. (8) PA *no·ntawe·wa ‘s/he hears h/’ Ch. é-néstovóho, Men. noohtawεw, Ojibwe noondawaad, M-I noontaweewa (9) PA *pesentawe·wa ‘s/he listens to h/’ Ch. é-áahtovóho, Ojibwe bizindawaad, M-I pisentaweewa, Mesk. pesetawêwa

The potential problem is that the final for ‘hear’ is as in (10). (10) ‘hear’ PA *-ehtaw- TA, *-eht(am)-

So the question arises whether the forms in (8) and (9) are bipartite, containing an allomorph of ‘hear’. In the end it is a judgment call. We include ‘hear’ and ‘listen’ as basic concepts because the concepts are intuitively basic and the structure allows for a short verb interpretation. This study is focused on seven languages: Plains Cree, Cheyenne, Menominee, Southwest Ojibwe, Miami-Illinois, Meskwaki, and Munsee Delaware. These languages encompass the whole of the family, with representation from every segment of the 1994 Goddard classification. The particular dialects are chosen because they have the advantage of having good lexical resources available. It is important to have an extensive lexicon to work from because accidental gaps are inevitable and an inadequate word list could distort the results.5 When all the short verbs are collected from these languages, it becomes immediately obvious that the task actually consists of two distinct parts. In most languages there are many more short intransitive verbs than short transitive verbs, by a factor of about three. There is a further complication in that the roots of many short intransitive verbs function as productive initials in bipartite verbs, as in (11). In this paper there is only space enough to address the transitive stems.

|

Short Verbs in Algonquian   205

(11) Intransitive roots as initials

a. Plains Cree

wâkisiw (AI), wâkâw (II) ‘be bent, curved’ wâkinam ‘s/he bends it’

wâkipitam ‘s/he bends it (down) by pulling on it’

wâkikâpawiw ‘s/he stands bent over’ wâkikâtêw ‘s/he has crooked legs’

b. Munsee

níisksuw (AI), níiskeew (II) ‘be dirty’ niiskeelúndam ‘s/he thinks it is dirty’ niiskalóhkeew ‘s/he does a messy job (of it)’ niiskeekhíikeew ‘s/he writes messily’ niiskcháaleew ‘s/he has a dirty nose’

When one compares the lists of short transitives between languages, two things stand out. First, the lists are all of different lengths. Meskwaki and Menominee have more short verbs than Cheyenne or Munsee, both in absolute terms and as a proportion of the total number of transitive verbs in their respective lexicons. Second, many short verbs are unique to a single language, i.e., verbs whose meaning is not expressed by a short verb in any other language, as in (12). (12) Men. kohkaanεεw ‘s/he catches h/ [fish] with a hook, on hook and line’, εεsiiw ‘s/he grinds h/’, Ojibwe ozha’aad ‘scare h/ [game] off’, M-I nipwaameewa ‘teach h/’, Mesk. anawinêwa ‘s/he sneaks up on h/’

A number of these unique forms are intransitive roots with a transitive abstract final added to them, as in (13). (13) Men. kohkaanεεw ‘s/he catches h/ [fish] with a hook, on hook and line’ < kohkεεw ‘s/ he angles, fishes with a hook and line’.

Since the aim of this exercise is to try to arrive at a list of basic concepts, the lists were assembled into a single table, sorted by meaning, and any meaning not expressed by a short verb in at least four of the seven languages was eliminated. The number four was not arrived at in a principled way; rather, it is a Goldilocks number. Any lower and the list burgeons, and too much higher (i.e., 6) and the list shrinks in half. Choosing four rather than five gives a slightly more generous list. The result is a list of fifty-two basic meanings given in a condensed form of just TAs in the appendix. The rest of the paper is devoted to some of the basic terms of particular linguistic interest, either morphologically or semantically.

|

206    Richard A. Rhodes

The meanings that have to do with dogs and bodily fluids are perhaps the biggest surprise. The meaning ‘bark at’ is attested in all languages, and is reconstructable, as in (14a). Munsee has a short non-cognate, given in (14b). The forms in all the languages are built with the same logic: add the abstract final reflex of *ł /*t to the AI to form the TA and TI respectively. (14) Terms meaning ‘bark at’ a. PA *mekiłe·wa TA, *mekitamwa TI (< AI *mekiwa). PC mikitêw, mikitam; Ch. é-maetóho, é-maēsta; Men. mekeenεεw, mekeetam; SWO miginaad, migidang; M-I makileewa, makitamwa; Mesk. mekinêwa, mekitamwa b. Mun. wugáaleew, wungáatam < wúngeew ‘s/he barks’, cf. Blackfoot áakohkatsiiwa ‘s/he will bark at h/’ < áakohkiwa ‘s/he will bark’

‘Spit on’, ‘defecate on’, and ‘urinate on’ are attested in all languages except for Munsee, which does not have ‘urinate on’ listed, and ‘defecate on’ is not found in Miami-Illinois (Costa, p.c.). The idea that these forms are basic gains strong support from the fact that they display reconstructable irregularities. The data sets are given in examples (15), (16), and (17). (15) Terms meaning ‘spit on’ PA *sehkwa·łe·wa TA, *sehkwa·tamwa TI (but AI *sehkwiwa) PC sihkwâtêw, sihkwâtam (AI sihkow), Men. sεhkuanεεw, sεhkuatam (AI sεhkoow), Ill. Nisehkawaaraa ‘I spit on him’, nisehkawaataan ‘I spit on it’ (AI nisehkarwi ‘I spit, slobber’), SWO zikwaanaad, zikwaadang (AI ziko), Mesk. sehkwânêwa, sehkwâtamwa (AI sehkwiwa); Mun. sŭkhwáaleew, [TI unattested], (AI súhkwiiw)

As the forms cited in (15) show, ‘spit on’ has an extended root allomorph that differs from the simple AI root and that extended root is reconstructable. Illinois further elaborates that allomorphy by epenthesizing an [a] before the /w/ and reshapes the AI, as well. 6 Like ‘spit on’, the transitive forms for ‘defecate on’ are unpredictable from the AI, but the allomorphy is reconstructable. The forms are given in (16). Miami-Illinois reports no forms in this meaning.7

|

Short Verbs in Algonquian   207

(16) Terms meanings ‘defecate on’ PA *mi·čiłe·wa TA, *mi·čitamwa TI (but AI *mi·si·wa). PC mîcitêw, mîcitam (AI mîsîw); Ch. é-mésetóho, [é-mēmésta TI, é-mée’kahe AI]; Men. meecenεεw, meecetam (AI meeseew); SWO miijinaad, miijidang (AI miizii); Mesk. mîčinêwa, mîčitamwa (AI mîsîwa); Munsee TI maskchítam (AI másktuw)

As can be seen in (16), the Cheyenne TA is cognate (PA *č > Ch. s), but the AI and TI are reformed.8 The Munsee forms are not cognate. The forms for ‘urinate on’ are given in (17). Munsee reports no forms in this meaning. (17) Terms meaning ‘urinate on’ PA *šekiłe·wa TA, *šekitamwa TI (< AI *šekiwa). PC sikitêw, sikitam (sikiw AI); Ch. é-xaetóho, é-xaésta, (é-xaa’e AI); Men. sekeenεεw, sekeetam (sekeew AI); SWO zhiginaad, zhigidang (zhigi AI); M-I šikileewa (only TA attested); Mesk. shekinêwa, shekitamwa (shekiwa AI)

The transitive forms of ‘urinate on’ in all the languages are built with the same morphological logic: add the abstract final reflex of *ł /*t to the AI stem to form the TA and TI respectively, the same morphological operation as with ‘bark at’. Two opposite meanings appear in the list in (14): ‘carry away’ and ‘leave behind’. There is clearer agreement across the languages in the shape of the forms for ‘leave behind’, given in (18). (18) Terms meaning ‘leave behind/abandon’ PA *nakałe·wa TA, *nakatamwa TI. PC nakatêw, nakatam; Ch. é-nootóho, é-noōhta; Men. nakaanεεw, nakaatam; SWO naganaad, nagadang; M-I nakaleewa, nakatamwa; Mesk. nakanêwa, nakatamwa; Mun. ngáleew, ngátum

The forms for ‘carry away’ are less consistent semantically and morphologically. There is no word in Cheyenne with exactly that meaning. The Ojibwe and Munsee meanings imply that the theme is heavy and so is glossed ‘haul’, and neither have any implied direction. Cree, Ojibwe, and Munsee have restructured the initial syllable. The forms are given in (19).

|

208    Richard A. Rhodes

(19) Terms meaning ‘carry away’ PA *awałe·wa TA, *awato·wa TI. PC āwatêw, āwatōw; Men. awaanεεw, awaatam; SWO aawanaad, aawadang ‘haul’; Miami-Illinois TA awal- ‘lead him, take him’, TI awatoo‘carry it away, take it’; Mesk. awanêwa, awatamwa; Mun. (TA unattested), éewatoow ‘haul a load’

I conclude this paper by addressing the class of meanings expressed by short verbs in all the languages where the forms are not all cognate. The most notable of these is ‘have’. The forms are listed in (20). Note that Menominee has two verbs glossed as ‘have’. (20) Terms meaning ‘have’ a. from a PA root *ay-: PC ayâwêw, ayâw; SWO ayaawaad, ayaang; Mun. ayúweew, ayúm b. from PA *aʔłe·wa, *aʔto·wa ‘put, place’: Ch. é-ho’hóho, é-hó’tse; Men. aqnεw, aqtaw; M-I ahseewa, ahtoowa c. from PA *awi- ‘be a N’: Mesk. awiwêwa, awiwa (TI3) d. from PA *tał- ‘at’: Men. tanεεw, tanaam

The underlying logic for all the forms in (20) is that they have an implicit or explicit locative meaning, in their respective languages. Meskwaki AI awiwa means both predicative ‘be’ and ‘be at’. The corresponding AI forms in Cree and Ojibwe mean ‘be at’. This is a different take on the notion ‘have’ than in English where ‘have’ is understood in terms of control rather than collocation. Worth noting in this connection is that the reflexes of PA ‘put’ still mean ‘put/place’ in all the languages, including those that also use their reflex to mean ‘have’. The meaning ‘put/place’ is also basic. The forms are given in (21). (21) Terms meaning ‘put/place’ a. PA *aʔłe·wa, *aʔto·wa: PC ahêw, astâw; Men. aqnεw, aqtaw; SWO asaad, atood; M-I ahseewa, ahtoowa, Mesk. asêwa, ahtôwa, Mun. áhleew, áhtoow b. from PA *pen- ‘down’: Ch. é-anȯhanóho, é-anȯhāna c. from PA *po:n- ‘put down’: M-I pooneewa, poonamwa

Cheyenne is the sole exception to this meaning being expressed as morphologically basic. Cheyenne has a bipartite form based on PA *pen- ‘down’ for ‘put’, following the logic that most of the time the act of ‘putting’ involves putting

|

Short Verbs in Algonquian   209

something down. Miami also has a synonym for ‘put’ using the root poon-, which had the historical meaning of ‘put down’. The reflexes of PA *po:n- are found elsewhere in the family, meaning variously to ‘put something (like fuel) in’ or ‘something (like food) on the fire’. In Ojibwe the root is used to mean ‘lay eggs’. But in Shawnee the poon- root is the regular form for ‘have’, again showing the Algonquian connection between location and having. There are many other interesting observations to be had from the forms realizing the meanings listed in the appendix. For example, the details of what ‘employ’ means are significantly different from English. Also notable is that the non-cognate forms for ‘overtake’ lack clear PA sources. This is different from, e.g., the non-cognate forms for ‘have’, or the non-cognate forms for ‘see’, each of which has a PA source. In conclusion, two points are worth mentioning. First, I started working on this problem with an eye to building a better tool for future statistical analysis, but the dataset that arises here is small enough to explore directly, observing which concepts are stable family-wide and which are common innovations or borrowings. The main results are that Menominee is by far the most conservative language with only one innovative form out of the set of fifty meanings, aside from the problematic ‘have’, and is missing only one. Cheyenne is the most innovative, with nineteen innovated forms and two missing. Munsee has the second most lexical innovations, with eighteen and two missing. Plains Cree has eleven; notably, six of those are shared with Ojibwe. In the Core Central languages, Ojibwe has eight innovations, Meskwaki has nine, and Miami-Illinois has six but is missing six forms. How many of the missing forms in Miami-Illinois are accidental gaps may never be known. Working with these forms and coming to understand their meanings in the various languages is an excellent exercise in decolonization. Once we stop trying to fit Algonquian words to putatively universal concepts like the Swadesh list or the Jakarta-Leipzig list, we can start to see the world through Algonquian eyes.

|

210    Richard A. Rhodes

Appendix: Basic TA Verbs

The TA forms of the fifty identified basic Algonquian meanings are given here. The languages are abbreviated as follows: PCr, Plains Cree; Ch, Cheyenne; Men, Menominee; SWO, Southwestern Ojibwe/Od, Odawa; M-I, Miami-Illinois; Mesk, Meskwaki; and Mun, Munsee (Delaware). When the TI is given because the TA is non-cognate, it is marked with †. When the form is AI+O rather than TA, it is marked with ‡. ‘Eat’ is so important and irregular that both TA and TI are listed. Forms that are not reflexes of the proto-form, i.e., that are innovations, are given in italics.

āwatēw

matēw

akimēw

(i)tōtawēw

minihkwēw ‡ é-mane ‡

mīcitēw

mowēw

mīciw †

kitamwēw

kostēw

asamēw

nātēw

nōtinēw

*awał-

*mał-

*akim-

*to·taw-

*man- ‡

*mi·čin-

*mow-

*mi·č-†

*kitamw-

*koʔł-

*ahšam-

*na·ł-

*mi·ka·ł-

*mełkaw- miskawēw

pētāw †

*pye·łmanaad

tootawεw

ahsaamεεw

koqnεw

ketaamiiw

miicwah †

miiw

miičineewa

menuah ‡

é-mé’ovóho

é-méoto

mεhkaawεεw mikawaad

meekaanεεw miigaanaad

naanaad

ashamaad

gosaad

gidamwaad

miijid †

amwaad

miijinaad

minikwed ‡

doodawaad

aagimaad

aawanaad

biinaad

maamanεw/ iiwεεw akeemεεw

MIAMI-ILL.

menwa ‡

(i)tôtawêwa

akimêwa

manêwa

awanêwa

pyênêwa

pôtânêwa

MESKWAKI

mihkaweewa

miikaaleewa

naaleewa

ahšameewa

kohseewa

‘fetch’

‘feed’

‘fear’

‘eat all’

‘eat it’

‘eat it (an.)’

‘put feces on’

‘drink’

‘do to’

‘count’

‘copulate with’

‘carry away’

‘bring’

‘blow on’

GLOSS

‘find’

matahkháaleew fight’

náaleew

xámeew

kxwéew

ktámweew

míichuw †

mhwéew

maskchíhtum

mŭnéew ‡

líhkaweew

akíimeew

wihpéemeew

éewatoow †

péetoow †

pootáaleew

MUNSEE

mehkawêwa móxkaweew

mîkâtîhêwa

nânêwa

ašamêwa

kosêwa

câkamêwa

kitamweewa/ ceekameewa

amwêwa mîčîwa †

miiciwa †

amweewa

mooyahweewa mîčinêwa

minwa ‡

ihpenaleewa

akimeewa

maleewa

awaleewa

piileewa

boodaanaad pootaaleewa

SW OJIBWE

awaanεεw

piinεεw

—

MENOMINEE

é-nó’oohé’tovȧtse naanεεw

é-hoxomóho

é-e’hóho

é-mȧhaeto

é-mese †

é-mevo

é-mésetóho

é-tsėhetanóho

é-hoemóho

é-hestse’eme

—

é-ho’eohtseho

é-éstovóhta †

pōtātēw

*po·ta·ł-

CHEYENNE

PLAINS CREE

PROTOALG

PROTO-ALGONQUIAN SHORT TRANSITIVE ANIMATE VERBS (PART 1)

Short Verbs in Algonquian   211

|

PLAINS CREE

miyēw

minahēw

akotēw

ayāwēw

pēhtawēw

kātēw

pakamahwēw

miswēw

nipahēw

ocēmēw

nakatēw

natohtēw

kanawāpamēw

sākihēw

wīhēw

atotēw

atimēw

PA

*mi·l-

*menah-

*akoł-

—

*no·ntwa-

*kya·l-

*pakam-

*mešw-

*neʔł-

*oči·m-

*nakał-

*pesentaw-

*wa·pam-

*tapa·l-

*wi·nł-

*alo·ł-

*atem-

é-hohtamóho

é-hotse’oto

é-veho

é-méhoto

é-vé’hoomóho

é-áahtovóho

é-nootóho

é-vȯsémóho

é-na’hóho

é-nȧha’emȧsóho

é-oomóho

é-nó’oetóho

é-néstovóho

é-ho’hóho

é-hoésemóho

é-manȯhóho

é-méto

CHEYENNE

SW OJIBWE

atεεmεεw

anoonεεw

weehnεw

tapaanεεw

waapamεw

pεhtaawεεw

nekaanεεw

ociimεεw

nεqnεw

mesiiw

pakaamεεw

kianεεw

MIAMI-ILL.

adimaad

anoonaad

wiinaad

zaagi’aad

ganawaabamaad

bizindawaad

naganaad

ojiimaad

nisaad

mizhwaad

ahtemeewa

alooleewa

wiinteewa

tapaaleewa

waapameewa

pisentaweewa

nakaleewa

soopalohpweewa

ankiheewa

mišweewa

pakameewa

kyaaleewa

bakite’aad

noontaweewa

gaanaad/ gkinaad (Od)

ahseewa

akoleewa

menaheewa

miileewa

noondawaad

ayaawaad

tanεεw/ aqnεw noohtawεw

agoonaad

mina’aad

miinaad

akoonεεw

menaahεεw

meenεεw

MENOMINEE

PROTO-ALGONQUIAN SHORT TRANSITIVE ANIMATE VERBS (PART 2) MESKWAKI

matanêwa

anohkânêwa

wîswihêwa

tepânêwa

wâpamêwa

pesetawêwa

nakanêwa

pôtetonêhpwêwa

nesêwa

mešwêwa

pakamêwa

kahkinêwa

nôtawêwa

awiwêwa

akônêwa

menahêwa

mînêwa

MUNSEE

matáleew

alóoleew

wíhleew

àhkwáaleew

pŭnáweew

kŭlústaweew

ngáleew

waangóomeew

níhleew

—

pakámeew

káaleew

púndaweew

ayúweew

wéhlaleew

mŭnáheew

míileew

GLOSS

‘overtake’

‘employ’

‘name’

‘love’

‘look at’

‘listen to’

‘leave behind’

‘kiss’

‘kill’

‘hit shooting’

‘hit’

‘hide’

‘hear’

‘have’

‘hang’

‘give to drink’

‘give’

212    Richard A. Rhodes

|

é-hetóho

é-vóomóho

nisitawinawēw

apwēw ‡

maskamēw

itēw

*nenaw

*apwe·-

*maxkam-

*eł-

é-nénoto

ācimēw

sikitēw

âpacihêw

pēhēw

*a·čim-

*šekił-

*aw-

*pi·h-

nōhēw

sihkwâtêw

kimotamawêw

*sihka·ł-

*kemootem-

kohcipaýihêw

kitotēw

*kałoł-

*no·nł-

é-oese’hanoto

é-nomáhtsėhóho

akohcimēw

*akwi·n(č)-

*kom-/konł-

é-éestsėstovóho

miýāmēw

*mela·m-

é-tonoomóho

é-ho’ōtse †

é-xaetóho

é-hósemóho

—

é-hoésemóho

é-matomo

é-amóho

wāpamēw

pimwew

*ne·w-

*pemw-

é-šéenóho

é-honoto

é-nanovóho

é-ho’a’ó’tsenoto

ahēw

*ahł-

CHEYENNE

PLAINS CREE

PA

piihεεw

awεεw

sekeenεεw

aacemεw

koohnεw

noohnεw

kemootemεw

sεhkuanεεw

keeketotawεw

akiihcemεw

menaamεεw

pemiiw

nεεwεεw

enεεw

mahkaamεεw

apuanεεw

nenaawεεw

aqnεw

MENOMINEE

PROTO-ALGONQUIAN SHORT TRANSITIVE ANIMATE VERBS (PART 3)

ahkawaaheewa

aweewa

aabiji’aad/ ayood † (Od) bii’aad

šikileewa

aacimeewa

konteewa

noonteewa

kimotemeewa

sihkawaaleewa

kaloleewa

akwincimeewa

milaaweewa

pimweewa

neeweewa

ileewa

mahkameewa

apwaaleewa

sehkwânêwa

apwîhêwa

awêwa

šekinêwa

âčimêwa

komêwa

nôtêwa

kemôtemêwa

péeheew

awéeheew ‡

—

kchíhlaleew

kwíhkeew

nóhleew

kŭmóotŭmeew

sŭkhwáaleew

kihkŭlóoleew

kamúkhweew kanônêwa

mŭláaweew

menâmêwa

payáxkheew

néeweew

léew

shihkwihtáasuw

apóosuw ‡

náweew

áhleew

MUNSEE

akwîčimêwa/ akwînêwa

pemwêwa

nêwêwa

inêwa

manihêwa

apwânêwa

nenawêwa

asêwa

ahseewa/ pooneewa ninaweewa

MESKWAKI

MIAMI-ILL.

zhiginaad

dibaajimaad

gonaad

nooni’aad

gimoodimaad

zigwaanaad

ganoonaad

agwanjimaad

biijimaamaad

bimwaad

waabamaad

inaad

makamaad

abwaanaad

nisidonawaad

asaad

SW OJIBWE

‘wait for’

‘use’

‘urinate on’

‘tell about’

‘swallow’

‘nurse’

‘steal from’

‘spit on’

‘talk with’

‘soak’

‘smell’

‘shoot’

‘see’

‘say to’

‘rob’

‘roast’

‘recognize (by sight)’

‘place’

GLOSS

Short Verbs in Algonquian   213

|

|

214    Richard A. Rhodes

Notes 1. The data in this paper (including glosses) are drawn from standard lexicographic and documentary sources listed in the References, and are cited using the orthographies found in those sources. Plains Cree: Wolvengrey (2001). For Ojibwean: Potawatomi: Welcher (n.d.); Odawa: Rhodes (1985); Eastern Ojibwe: Wilson (1874); Nipissing: Cuoq (1886), Lemoine (1909), and McGregor (1994); Southwestern Ojibwe: Baraga (1853) and Nichols (2012–2021); Northwest/Berens River Ojibwe: Fiero (2013); Oji-Cree: Nichols (2014). For Munsee: O’Meara (1996); For Cheyenne: Fisher et al. (2017). For Menominee: Bloomfield (1975). For Meskwakian: Goddard and Thomason (2014). For Miami: Baldwin and Costa (2005). For Illinois, Costa (p.c. 2022). For Shawnee: Longhorn et al. (1991), Pearson (1991), and Schaefer (2019a, 2019b). Proto-Algonquian forms are taken from Aubin (1975) when they are listed there, with reinterpretation of *θ as *ł. 2. Glossing follows Leipzig rules, with the added abbreviations: AN = animate, h/ = him/her, INAN = inanimate, and OBV = obviative. 3. Costa (p.c.) indicates that the TI stem is only attested in Illinois into the 1800s. After that, apwee- is used as an AI+O. 4. Cheyenne phonological development is complex. The relevant changes here are *ł > > t, *w > > n, *p > Ø, [+long] > [hi tone], *a > o: PA *apwa:ł- > apwa:t- > apna:t > ana:t- > anát> hanát- > honót-. See Goddard 1988 for an account of the phonological development of Cheyenne. 5. It turns out that there may be a problem of accidental gaps with Miami-Illinois. See the appendix. 6. Costa (p.c.) points out that there is an Illinois TA stem sehkwaal- with the expected shape root, which survives in several attestations but has the meaning ‘rejoice at his misfortune’. The forms of ‘spit on’ are only found in Illinois sources. 7. Costa (p.c.) notes the Illinois forms nimiisilaa ‘I shit him’ (TA) and nimiisitaan ‘I shit it’ (TI) are regularly derived from the AI stem miisiiwa ‘he shits’. 8. The Cheyenne TI é-mēmésta is reduplicated but is otherwise cognate. PA *(o)mi·čite:ni yields Ch. *-mésėsta by sound law, which in turn contracts to *-mésta. Reduplication yields the attested form -mēmésta.

References Aubin, George F. 1975. A Proto-Algonquian Dictionary (No. 29). Ottawa: National Museums of Canada. Baldwin, Daryl, and David J. Costa. 2005. Myaamia neehi peewaalia kaloosioni mahsinaakani: A

|

Short Verbs in Algonquian   215

Miami-Peoria Dictionary. Miami, OH: Myaamia Project at Miami University. Baraga, Frederic. 1853. A Dictionary of the Otchipwe Language, Explained in English. Cincinnati: Joseph A. Hemann. Biedny, Jerome, Andrea Cudworth, Sarah Holmstrom, Monica Macaulay, Gabrielle Mistretta, Joseph Salmons, Charlotte Vanhecke, and Bo Zhan. 2022. Lexical Relationships in Central Algonquian. Papers of the 51st Algonquian Conference, ed. by Monica Macaulay and Margaret Noodin, pp. 1–18. East Lansing: Michigan State University Press. Bloomfield, Leonard. 1946. Algonquian. Linguistic Structures of Native America, ed. by Cornelius Osgood and Harry Hoijer, pp. 85–129. Viking Fund Publications in Anthropology 6. New York: Viking Fund. Bloomfield, Leonard. 1962. The Menomini Language, ed. by Charles F. Hockett. New Haven, CT: Yale University Press. Bloomfield, Leonard. 1975. Menomini Lexicon. Milwaukee: Milwaukee Public Museum. Costa, David J. 2013. Borrowing in Southern Great Lakes Algonquian and the History of Potawatomi. Anthropological Linguistics 55(3):195–233. Cuoq, Jean-André. 1886. Lexique de la langue Algonquine. Montréal: J Chapleau et fils. Fiero, Charles E. 2013. An Ojibwe Word List: Vocabulary from Communities in Northwestern Ontario and Eastern Manitoba. Unpublished manuscript. Fisher, Louise, Wayne Leman, Leroy Pine Sr., and Marie Sanchez. 2017. Cheyenne Dictionary. Lame Deer, MT: Chief Dull Knife College. Goddard, Ives. 1980. Eastern Algonquian as a Genetic Subgrouping. Papers of the 11th Algonquian Conference, ed. by William Cowan, pp. 143–158. Ottawa: Carleton University. Goddard, Ives. 1988. Pre-Cheyenne y. In Honor of Mary Haas, ed. by William Shipley, pp. 345–360. Mouton de Gruyter. Goddard, Ives. 1990. Primary and Secondary Stem Derivation in Algonquian. International Journal of American Linguistics 56(4):449–483. Goddard, Ives. 1994. The West-To-East Cline in Algonquian Dialectology. Actes du VingtCinquième Congrès des Algonquinistes, ed. by William Cowan, pp. 187–211. Ottawa: Carleton University. Goddard, Ives, and Lucy Thomason. 2014. A Meskwaki-English and English-Meskwaki Dictionary. Petosky, MI: Mundart Press. Lemoine, Georges. 1909. Dictionnaire français-algonquin. Chicoutimi: G. Delisle. Longhorn, Leonard, Meredith Wapekeche, Jeanette Bayless, and Helen Ramirez. 1991. Shawnee Language Dictionary. Shawnee, OK: Absentee Shawnee Tribe of Oklahoma. McGregor, Ernest. 1994. Algonquin Lexicon. Maniwaki: Kitigan Zibi Education Council. Nichols, John D. 2014. Anihshininiimowin Oji-Cree Dictionary (Severn River and Winisk River).

|

216    Richard A. Rhodes

Sioux Lookout: Kwayaciiwin Education Resource Centre. Nichols, John D. (ed.). 2012–2021. The Ojibwe People’s Dictionary. https://ojibwe.lib.umn.edu/. O’Meara, John. 1996. Delaware-English, English-Delaware Dictionary. Toronto: University of Toronto Press. Pearson, Bruce L. 1991. Shawnee Language Dictionary. Shawnee, OK: Absentee Shawnee Tribe of Oklahoma. Pentland, David. 2003. Interaction between Cree and Ojibwa: A Historical Perspective. Paper presented at the Workshop on Structure and Constituency in the Languages of the Americas, March 8. Proulx, Paul. 1984. Two Models of Algonquian Linguistic Prehistory: Diffusion versus Genetic Subgrouping. Anthropological Linguistics 26(4):393–434. Rhodes, Richard. 1989. The Cree Connection. Paper read at the 28th Conference on American Indian Languages, Washington, DC. Rhodes, Richard A. 1985. Eastern Ojibwa-Chippewa-Ottawa Dictionary. Documentation No. 5. Berlin: Mouton. Rhodes, Richard A. 2017. Towards a Semantic Dictionary of Algonquian. Papers of the 45th Algonquian Conference, ed. by Monica Macaulay, Margaret Noodin, and J. Randolph Valentine, pp. 177–190. East Lansing: Michigan State University Press. Rhodes, Richard A. 2020a. The Nature of Algonquian Bipartite Verbs and Implications for Borrowing. Paper presented at the 52nd Algonquian Conference. Rhodes, Richard A. 2020b. Language Shift in the Subarctic and Central Plains. The Language of Hunter-Gatherers, ed. by Tom Güldemann, Pat McCovell, and Richard A. Rhodes, pp. 552–576. Cambridge: Cambridge University Press. Rhodes, Richard A. 2021. The Case for Core Central Algonquian. Webs of Relationships and Words from Long Ago, ed. by Lucy Thomason, David Costa, and Amy Dahlstrom, pp. 303–344. Petoskey, MI: Mundart Press. Rhodes, Richard A., and David Costa. 2003. The History of Algonquian Number Words. Essays in Algonquian, Catawban and Siouan Linguistics in Memory of Frank T. Siebert, Jr., ed. by Blair A. Rudes and David J. Costa, Memoir 16 Algonquian and Iroquoian Linguistics, pp. 181–215. Winnipeg: Algonquian and Iroquian Linguistics. Schaefer, Carl. 2019a. Lexicon of the Alford Gospels (Shawnee). Unpublished manuscript. Schaefer, Carl. 2019b. Lexicon of the Shawnee Laws. Unpublished manuscript. Swadesh, Morris. 1971. The Origin and Diversification of Language, ed. by Joel Sherzer. Chicago: Aldine. Tadmor, Uri. 2009. Loanwords in the World’s Languages: Findings and Results. Loanwords in the World’s Languages: a Comparative Handbook, ed. by Martin Haspelmath and Uri

|

Short Verbs in Algonquian   217

Tadmor, pp. 73–74. Berlin: de Gruyter. Welcher, Laura. [n.d.] Potawatomi Dictionary. Unpublished manuscript. Wilson, Edward F. 1874. The Ojebway Language: A Manual for Missionaries and others Employed among the Ojebway Indians. Venerable Society for Promoting of Christian Knowledge. Toronto: Roswell and Hutchison. Wolvengrey, Arok. 2001. Nêhiyawêwin: Itwêwina, Cree: Words (Volumes 1 and 2). Regina: University of Saskatchewan Press. Zipf, G. K. 1949. Human Behavior and the Principle of Least Effort. Boston: Addison-Wesley Press.

Reflections of a “Barbarous, Mentally Undeveloped” Linguist On the Legacy and Obligations of Algonquian Studies and Linguistics Mskwaankwad Rice

A careful study of the Chippewa language has brought the writer to the conviction that the Indians belonging to the Algonquin family of nations must have attained to a high degree of civilization at a remote period of time and that their subsequent lapse into barbarism was due to incessant wars and migrations. Our opinion is based on the following reason: A nation’s language is a true and reliable index of the mental capacity and intellectual status of its people. A barbarous, mentally undeveloped race cannot originate a systematic, philosophically regular, grammatical language. I think this needs no proof. Now, any one who has a grammatical knowledge of the Chippewa language, will concede that it is wonderfully systematic, regular, euphonic, plastic, and expressive. It must, therefore, have originated with a people mentally well developed. —Chrysostom Verwyst, Life and Labors of Rt. Rev. Frederic Baraga, First Bishop of Marquette, Mich.

The Western scientific tradition has a long, exploitative, and damaging history with Indigenous peoples worldwide. While linguists never kidnapped people or robbed graves, the discipline has been employed as a tool in justifying acts

|

   219

|

220    Mskwaankwad Rice

of genocide and has exploited people for its own gain; all must reckon with the history of the discipline and their roles in perpetuating these practices. The earliest linguistic writings about the Ojibwe language were made by missionaries, motivated by the evangelical goal of assimilating Indigenous peoples based on a white supremacist belief in the inferiority of Indigenous peoples and cultures. As it is not customary for linguists to address their positionality or the history of their sources and the discipline in their research, the legacy of these early works as tools of assimilation has generally been ignored in linguistic literature—to the extent that when the sources are commented on at all, it is often in a positive manner. Since linguistics is the study of human language, it is inherently tied to the dynamics of interaction of language, peoples, and cultures. The discipline must therefore acknowledge its history and role in the erasure of Indigenous peoples. In this paper I suggest a blueprint for how linguists and Algonquianists specifically can address both their own positionality and that of their colonial sources.

Positionality Statement

My name is Mskwaankwad Rice and I am from Waasaaksing First Nation.1 I am a learner/new speaker of Nishnaabemwin, and at the time of writing I am in my second year of the linguistics PhD program at the University of Minnesota, completing preliminary requirements. I study linguistics for its utility in understanding the intricacies of how languages function, and I seek to apply this knowledge to Ojibwe language reclamation efforts. I am a generally Indigenous-coded, cisgender male of Nishnaabe/Anishinaabe and Eurocanadian heritage, which results in a range of social interpretation of my ethnicity, identity, and experience. I am cognizant of the implications of this as a descendant of both a colonized population and of settlers benefiting from Indigenous displacement.2 My own positionality is multifaceted and shifting.3 I am an Indigenous person, technically also a linguist as I have a degree in linguistics and undertake linguistic work, and also (somewhat begrudgingly because the academic study of peoples in such a manner is problematic) an Algonquianist. While this paper is written for the Papers of the Algonquian Conference (PAC), I hope that it will have a wider reach, and for that reason I have attempted to make it accessible outside of the context of linguistics or Algonquian studies.

|

Legacy and Obligations of Algonquian Studies   221

Missionary Linguists and the Ojibwe

What is known of Algonquian languages from a linguistic standpoint is rooted in missionary linguistics, and it is not a habit of documentary or theoretical linguists to reflect critically upon this legacy. Missionary linguist Frederic Baraga is cited in at least sixty-eight volumes of PAC, almost always as a neutral source, without discussion of the impacts of his missionary work among the Anishinaabeg. In fact, in the niche field of Algonquian linguistics, writers even celebrate missionary linguists for providing valuable historical sources, praised as “well-resourced,” “timeless,” and of “enormous value.”4 This paper highlights the works of Baraga, Rev. Edward Wilson, and Chrysostom Verwyst, though numerous others missionized to the Anishinaabeg, a program that continues to this day. Baraga’s grammar and dictionary were built upon by linguistic successors, many of whom were missionaries themselves, and today those original works are commonly cited in linguistic papers written about Ojibwe and linguistically related languages. Very few PAC papers discuss the problematic history of missionizing among our peoples, and those mentioning the issue have a historical or anthropological bent. Podruchny (1995) notes that there is debate over the general issue of missionizing among Indigenous North Americans, and others such as Long (1985), Krieger (1989), Brown (2012), and Westman (2012) discuss missionary work—albeit in a noncritical, descriptive manner. I argue in this paper that a linguist engaging in an ethical research practice must note the context of legacy materials such as Baraga’s. The Slovenian Roman Catholic Bishop Frederic Baraga missioned in Anishinaabe country in the nineteenth century, producing the first major linguistic works examining Ojibwe. Baraga’s (and other missionary linguists’) explicit goal was the conversion of Native people to Christianity, a point made in the preface of his 1878 grammar: My principal intention in publishing this Grammar is, to assist the Missionaries in the acquirement of the Otchipwe language and its kindred dialects, as I know by experience how useful it is the Missionary to know the language of the people whom he is endeavoring to convert to God. (Baraga 1878:vii)

This motivation to convert Indians is rooted in the complementary assumptions of a) the inferiority of Indigenous spirituality and b) the superiority of European

|

222    Mskwaankwad Rice

religion. These notions of white supremacy are made more explicit elsewhere.5 In footnotes in his grammar, Baraga notes that the “pluperfect, and the imperfect tense, are not so sharply distinguished in Otchipwe, as they are in English, or in other civilized languages,” implying that Ojibwe is uncivilized, and adding that the tenses are used “promiscuously” (98–99). A good linguist knows not to attempt to place Eurocentric linguistic conventions of pluperfect and imperfect upon a language unrelated to English (though this fact took the discipline some time to realize). It is also known to linguists today that language speakers do not haphazardly select one verb form over another, and the linguist’s goal is to determine the semantic and pragmatic nuance carried in this selection, not relegating the apparent opacity to a deficiency of the people whose language is under study. The idea of Ojibwe language and people (and by extension Indians generally) being inferior was also expressed by Rev. Edward Wilson in his 1874 Ojibwe manual for missionaries. He notes that the language, “in common with those of other Indian tribes[,] is not a written one, and, though by some considered musical, is very deficient in phonetic elements” (1874:iv). Again, the modern linguist knows that no language can have a “deficient” phonology. Wilson too extends the notion of deficiency of the language to a deficiency of the people: It seems a marvellous thing, indeed, that these poor ignorant Indians, with no knowledge of literature, or the general principles upon which languages are based, should have handed down so complex a dialect as the one before us, with all its multitudinous inflections, affixes, and prefixes, from one generation to another. (31)

The white supremacist notion of the inferiority of the non-European was so ingrained in this man of God that when faced with evidence that by his own logic proved otherwise, he could only marvel at its inexplicability rather than use it to reconfigure his racist ideals. The epigraph to this chapter, citing Verwyst (1900), is a striking example of this cognitive dissonance, and when modern linguists and Algonquianists uncritically cite (let alone revere) the works of these missionaries, they uphold the legacy that espouses the notions cited above.

|

Legacy and Obligations of Algonquian Studies   223

Why Does This Matter?

One cannot simply “separate the art from the artist,” and take the works of missionary linguists on face value as precious historical records. The white supremacist notions subscribed and adhered to by the missionary linguists had tangible, detrimental effects upon the Native people they sought to convert. Harvey (2015) examines the interrelation of the concepts of language and race in colonial North America, and how an assumed link led to the justification of assimilationist and genocidal efforts of colonial governments. He notes that “U.S. officials based their efforts to extinguish Native languages on a century of increasing philological knowledge that blended notions of descent and intellectual difference” (2015:15). Works by early linguists such as those cited above provided “evidence” of the inferiority of the Indian mind, as exemplified by so-called deficiencies of their “uncivilized” languages. As seen in the previous section, in the face of contradictory evidence, the authors marvel at such anomalies or explain away the matter in feats of mental gymnastics. These early sources are thus not neutral documents, as the racist ideals they are built upon were instrumental in genocidal colonial projects carried out against Indigenous peoples. We therefore cannot dismiss their racism as simply a byproduct of their times. The summary of the final report of the Truth and Reconciliation Commission of Canada (2015; TRC henceforth) describes how the presence of the missionary helped to justify the European colonial program in both expansion of empire and in the destruction of those refusing to capitulate (2015:48). Modern apologists might argue that missionary work and residential schools were well intentioned, but regardless of intention, these powers dismantled the traditional lifeways of Indigenous peoples. The TRC notes that “although missionaries often attempted to soften the impact of imperialism, they were also committed to making the greatest changes in the culture and psychology of the colonized,” and this was realized by disrupting “relationships to the land, language, religion, family relations, educational practices, morality, and social custom” (48). The detrimental effects of the traumas inflicted upon Native peoples—from the era of forced removals and the reservation system, to the establishment of residential/boarding schools, to the continued exercise of state-sponsored oppression and

|

224    Mskwaankwad Rice

violence today witnessed in the over-policing of our communities and continued apprehension of Indigenous children from their homes—cannot be understated. The field of Algonquian studies cannot ignore the fact that its forefathers held racist beliefs about the peoples whose languages they studied, nor that their intention for those people was religious conversion, which entailed sociocultural destruction.6 Algonquianists must acknowledge these facts when conducting research that inherently takes Indigenous peoples (specifically, those the field have delineated as ‘Algonquian’) as research subjects.7 There is a current and pertinent discussion regarding linguistics and race, and this includes the discipline’s relationship and history with Indigenous peoples. Charity Hudley et al. (2018) brought up the fact that linguistics has not addressed the issue of race, leading to, among other things, a Linguistic Society of America (LSA) Statement on Race (2019).8 Part of this movement has included an effort to remodel the typically white executive committee of the LSA, and saw the creation of the Natives4Linguistics special interest group, which seeks to increase Indigenous participation in linguistics and ameliorate the discipline’s hegemonic position taken in relation to Indigenous peoples and languages. In their responses to Charity Hudley et al. (2020), Gaby and Woods (2020) and Leonard (2020) discuss the issue as it relates to Indigenous peoples, highlighting such hegemony and harms inflicted upon Indigenous peoples as a result. A manifestation of this within the discipline itself is the use of esoteric language and gatekeeping of what is deemed valid knowledge, discussed in the following section.

Why Should We Do Something about It?

Chickasaw scholar Jenny L. Davis (2017) explains the importance of rhetoric in discussing Indigenous languages, noting the concept of the erasure of colonial agency, a practice that “minimises the historical and ongoing causes of language endangerment and dormancy, sometimes to the extent of misattributing agency for such realities onto Indigenous communities themselves” (37). It is thus important that one not whitewash the legacy of canonical sources by citing them without qualification. This is a discipline-wide issue not specific to Algonquianist linguistics, and it is noted throughout the discussion spurred by Charity Hudley et al. (2018). Linguistics maintain a position of power over Indigenous peoples; there are many cases in which individuals and bodies become self-made authorities on a

|

Legacy and Obligations of Algonquian Studies   225

language. The case of Frank Siebert and the Penobscot language is one example (Gregory 2021). Siebert effectively came to “own” the language via US copyright law by having written it down, and upon his death, he willed the materials to the American Philosophical Society. With no remaining first-language speakers, the language of the Penobscot Nation is now difficult to access physically, as well as on account of Siebert’s esoteric orthography. Huisken (2013) examines how nineteenth-century missionaries among the Dakȟóta similarly employed knowledge of Dakȟótiyapi to serve their purposes of Christianization and assimilation, while also benefiting the field of ethnography to further their own academic careers. The tradition of assuming authority over Native peoples and languages is rooted in notions of superiority and has allowed linguists and the field itself to commoditize and benefit from them to the detriment of the peoples and languages. The phenomenon of linguistics treating Indigenous languages as objects is central to the current issue and is discussed by Miami scholar Leonard (2018) and Errington (2008), among many others. Such practices are immoral for their one-sided extractivism and disregard for Indigenous research protocols (Snow et al. 2016; Holton et al. 2022), a demonstration of how Western science by extension devalues the peoples themselves. That peoplehood and language are inherently connected is a tenet of many Indigenous ontologies and not one of Western science (see Clarke 1996; Harvey 2015; and Leonard 2021b), and Western academia is not self-reflexive of Western hegemony affecting Indigenous peoples negatively. Kubota (2020) and Mellow (2015) discuss how Western science perpetuates this hegemony and note ways to counteract or decolonize it. In linguistics, the point of contact between researcher and research subject is found in fieldwork activity. Rice (2006) provides a background on ethics in linguistic fieldwork that demonstrates linguistics’ hegemonic position even from the beginning of its discussion of ethics, and makes suggestions for improving relationships in this regard. Zuni linguist Adrienne Tsikewa (2021) demonstrates that a linguist-centered, extractive approach to fieldwork continues today, as contemporary fieldwork training ignores Indigenous research methodologies and avoids discussion of ethics generally. Among the many instances of culture shock I experienced in returning to an academic space over a decade after completing my undergraduate studies was being struck by glaringly problematic instances of how the linguistics field views Indigenous peoples. A sidebar in a phonology textbook noted an instance of

|

226    Mskwaankwad Rice

linguists finding that the Yowlumne language previously thought “extinct” still had a number of “surviving” speakers and, exalting the language solely for its import to linguistics and phonology, exclaimed that “we nearly missed it!” (Zsiga 2013:242; emphasis added).9 Having recently lost my grandmother, who was the second-to-last first-language speaker of Nishnaabemwin in my family, this reduction of a people’s value to solely their importance to the field of linguistics struck me. In addition to the implications of terms such as “language death,” these terms themselves sidestep or sanitize the fact that the “death” of a language comes only with the death of human beings who themselves are/were living, real people who those of us excluded from the “we” of linguistics knew and loved and yet mourn. As I explored historical sources examining Ojibwe, I additionally came across examples that claim that my people are uncivilized, calling us “ignorant,” “barbarous,” and “mentally undeveloped.” If Algonquianists continue to uncritically center their problematic sources and moreover ignore the fact that linguistics assumes Indigenous nonpresence, they will continue to subject Indigenous students and Indigenous linguists to the abasement of their peoples, creating a hostile and exclusionary environment, in addition to adding yet another barrier for Indigenous peoples navigating the already exclusionary space of academia. A large portion of legacy materials in linguistics and elsewhere is problematic in the same manner as the sources noted above. Adams-Campbell et al. (2015) discuss archives with respect to their relation to Indigenous peoples and their role in legitimizing and upholding colonial state power. In linguistics as elsewhere, sources and archives are misconceived as being neutral and objective, a point also made by Errington (2008) and Leonard (2021b). The latter, along with Gaby and Woods (2020), describe how this perspective is used in gatekeeping knowledge to deny Indigenous access and presence in academia. The concept of a neutral and objective archive empowers sources to erase Indigenous presence, and a parallel is made with settler colonialism’s need for the “non-encounter” to deny Indigenous presence and thus also deny the settler violence upon Indigenous “others” that predicates the settler state (Adams-Campbell et al. 2015:110). Reminders of my presumed status as an object of study, rather than as a linguist, are pervasive in this field. On my first day as a student of linguistics, I picked up an LSA pamphlet on endangered languages (Woodbury n.d.). Among other content I found striking was the heading of one section reading, “What does language extinction mean for a community—and for the rest of us?,” a statement implying

|

Legacy and Obligations of Algonquian Studies   227

that I myself, a member of a community facing “language extinction,” am not included in the “us” of the community of linguists. This relates to the significance of language and terminology; “extinction” is relevant to discussions by Davis (2017) and Bobaljik (1998) noted above in erasing colonial agency, in addition to the use of an exclusionary “us,” whereas the LSA Statement on Race cites the importance of inclusion in the discipline. Outside of explicitly exclusionary language—whether by overt racism, as in the historical Algonquianist sources, or by implication of Indigenous nonpresence, as in the LSA pamphlet on endangered languages—the esoteric language of linguistics serves to create another barrier to people whom the discipline takes as subjects yet inherently excludes. Gaby and Woods (2020) note that linguistic metalanguage renders the outputs of language documentation and description inaccessible to community members. As an undergraduate student in the early 2000s, I spent many hours in university libraries often exploring sources and subjects outside of what I was meant to be studying at the time, and in one such instance I came upon the PAC. I was thrilled to discover this long-standing community of practice committed to the study of my heritage language, which I had not yet begun to pick up in earnest. I wondered how I’d never heard of this seemingly well-established Algonquian Conference before, myself with a budding interest in language revitalization. When I read the papers, however, I was utterly confused by the content and language used within and ill equipped to learn anything from them. This deflated my initial feeling of hopefulness that the presence of this group would surely mitigate to a degree the threat of language loss. Yet in the nearly two decades since I became aware of the Algonquian Conference, the reality on the ground is that the number of Anishinaabemowin speakers in my community and across Ojibwe country has seen a steady decline. The fact that Algonquian is a term and concept in itself is an example of the hegemony of the white institution imposing foreign concepts upon Indigenous peoples and languages. Valentine (1994) and Anishinaabe linguist Migizi (Michael) Sullivan (2020) discuss the inadequacies and inaccuracies of exonyms imposed upon the languages/dialects and communities of Anishinaabeg, and the subsumption of numerous cultural identities under a linguistically based delineation is an academic invention that ignores the Native reality. While my people recognize and maintain relationships with our neighbors and relatives of similar languages and cultures, our concept of relationality differs from that implied by “Algonquian.” Harvey (2018) recognizes that Native conceptions of identity and relation differ from

|

228    Mskwaankwad Rice

European norms, and relationality is not necessarily dependent upon nor restricted to linguistic affiliation. The delineation of what is now called “Algonquian” is the result of the framing of my people as merely objects of study without consideration of the views of the peoples themselves. They do not fit the European framework imposed, just as the concepts of imperfect and pluperfect do not suit Ojibwe grammar. This speaks to the fact that there is an inherent disconnect between Western and Indigenous ways of knowing, and works such as Anishinaabe scholar Sharla Mskokii Peltier’s “Anishinaabewin: An Epistemological and Research Method Framework” (2021), among many others, illustrate this.

What Can We Do?

A number of specific actions can be taken to address the legacy of Algonquian studies and linguistics generally by both the individual linguist and by larger entities such as journals as editing bodies and institutions as sites of education. As individuals we can explicitly position ourselves in our work, published and otherwise. This is common in discussions of social issues but practically nonexistent elsewhere, and linguistics as a whole does not address such issues. Individuals have different backgrounds and motivations for doing linguistics that color our work— and it is good practice to acknowledge these things for purposes of accountability. We can also explicitly position the sources in our writings. Baraga and others are not neutral sources, as illustrated above, and we must engage critically with them. For example, works citing Baraga could include a footnote like this: This paper draws heavily on Baraga (1878). As a missionary linguist, Bishop Baraga’s explicit motivation in his work was the religious conversion (p. vii) of the Anishinaabe people among whom he worked. The report of the Truth and Reconciliation Commission (TRC) of Canada (2015) recognizes the damage inflicted upon the sociocultural wellbeing of Indigenous peoples by missionary efforts. Moreover, racist notions of Indigenous inferiority supported and espoused by Baraga and others were used in colonial governmental policy to justify acts of genocide such as the residential/boarding school system (see Harvey 2015).

Tsikewa (2021) notes that reflexive positionality is not often discussed in linguistics, and this is certainly also the case in Algonquian studies. Reiterating the fact that

|

Legacy and Obligations of Algonquian Studies   229

missionizing continues among Anishinaabeg today, Algonquianists have stated their Christian faith in their publications, or have noted collaboration with SIL International, an evangelical Christian organization that has a problematic history with Indigenous peoples worldwide.10 With such issues in mind, ethical research involving Indigenous peoples requires a positioning of the researcher in addition to the sources and resources utilized. Methodologies of the discipline must also be critically evaluated. The 53rd Algonquian Conference, at which this paper was presented, held a special session titled “Drawing from Algonquian epistemologies to talk about Algonquian languages.” Aside from the Western conceptualization of “Algonquian” here, such a discussion is a good step, but reckoning with the field’s legacy must happen alongside any project of decolonization, since the racism that is built into the discipline will continue to harm Indigenous communities and Indigenous linguists unless it is critically examined. Editors and institutions can likewise adopt best practices for addressing the history and sources, adopting a style guide for acknowledging them, as in the example footnote above. The PAC style guide (2022) is purely formal in nature and notes nothing about the contextualization of one’s work or sources. Eurocentric conventions should be reassessed, as in, for example, the National Science Foundation’s (NSF) adherence to ISO codes in funding applications, which see the use of inaccurate exonyms such as Chippewa for the Ojibwe language.11 Younging’s Elements of Indigenous Style (2018) is a comprehensive resource offering best practices for citational praxis while explaining their necessity. In addition to simple and straightforward actions addressing positionality, bodies such as the Algonquian Conference can make statements acknowledging their histories and the legacies of their founders; in fact, discussion of this is presently underway. This is an area in which the Algonquian Conference can use its power to affect transformation in how research is conducted in the discipline. As of 2019, the Society for the Study of Indigenous Languages of the Americas (SSILA) requires that abstracts include a statement on the social outcomes/impacts/ implications of work presented, in recognition of the fact that such work has implications for and impacts upon Indigenous peoples.12 I suggest that a similar practice be adopted by the Algonquian Conference. The Australian Journal of Rural Health, Canadian Journal of Rural Medicine, and Rural and Remote Health collaborated in 2022 to create a statement on research and reconciliation with Indigenous peoples. In addition to committing to signifying which contributing authors are

|

230    Mskwaankwad Rice

Indigenous, they will reject papers “that concern Indigenous communities but do not acknowledge an Indigenous author or provide evidence of a participatory process of Indigenous community engagement” (Lock et al. 2022:1). Tuck and Yang 2012 (among many others) discuss how true decolonization is a difficult and unsettling process, and the position taken by the above journals is a signpost for decolonial action. The prospect of a radical self-examination of Algonquian studies may be disconcerting—since it is unclear how the discipline would hold up under that level of scrutiny—and that means that we are looking in the right direction. Notes 1. I gratefully acknowledge the reviewers who have made this paper much better than I could have on my own, and any errors are mine. In addition to anonymous reviewers, two friends I acknowledge for their kindness and help are Ayah Helmy and Geraldine King. I am also forever grateful to my advisor Claire Halpert for her ongoing guidance, help, and support, and especially for her familiarity with and understanding of issues brought up in this paper. 2. Numerous terms are used interchangeably in this paper including Anishinaabe(g), Nishnaabe(g), Ojibwe, and even exonyms like Indian and Native, themselves terms used by my people in an exercise of reappropriation of a modicum of social selfdetermination. Words denoting the language itself such as Ojibwe, Anishinaabemowin, and Nishnaabemwin are also used interchangeably. 3. Leonard (2021a) speaks to particular challenges faced by Indigenous linguists. 4. I have chosen to not cite the authors of these quotes as my intent is only to highlight how Algonquianists describe the sources under discussion, and not to single out individuals who are in fact respected colleagues, friends, and mentors. I myself have uncritically utilized Baraga as a source. 5. While this term conjures images of neo-Nazis and the KKK in the popular perception, it is accurate in this context referring to the belief in the superiority of white/European people and culture over others, a view clearly expressed by sources cited herein. This is itself a simplified definition that does not capture the nuance of a theoretical definition, which includes structural and societal considerations. 6. The masculine-gendered term ‘forefathers’ is used purposely in reference to the tradition of colonial white, male dominance of this field. A look at the list of contributing authors to PAC in its early days will illustrate this, though recent years have seen greater inclusion of women and even Indigenous contributors.

|

Legacy and Obligations of Algonquian Studies   231

7. There are nonetheless many Indigenous linguists of ‘Algonquian’ nations doing great work for/with their languages such as Yolanda Pushetonequa (Meskwaki), Bernard Perley (Maliseet), Migizi Sullivan (Ojibwe), and Brendan Kishketon (Ojibwe/Kickapoo). Ethnographers such as Mary Ann Corbiere (Odawa) and Alan Corbiere (Odawa) contribute meaningfully to the Ojibwe corpus, in addition to William Jones (Meskwaki) in the late nineteenth and early twentieth centuries. 8. Parallel organizations such as the Canadian Linguistic Association and the Canadian Association of Applied Linguistics exist in Canada that themselves could undertake self-reflection discussed in this paper. I discuss the LSA as it is relevant to my personal experience as a graduate student in the USA. 9. Bobaljik (1998) cautions against terms such as “language death” for their implication that language loss is a natural phenomenon, while in the case of Indigenous languages under the assault of colonization, it is not. 10. SIL was formerly known as the Summer Institute of Linguistics but now use the abbreviation as a name (https://www.sil.org/). Dobrin and Good (2009) discuss the issue of linguists accepting and thus legitimizing SIL missionary presence in Indigenous communities, where the former often utilize frameworks and resources established by the latter, and the authors note the necessity of reevaluating linguistics’ relationship with SIL. 11. The classification of languages has been standardized under International Organization for Standardization (ISO) conventions. Note also that a great deal of this work has been carried out by SIL International. 12. See https://www.ssila.org/social-impact-and-outcomes for information and examples.

References Adams-Campbell, Melissa, Ashley Glassburn Falzetti, and Courtney Rivard. 2015. Introduction: Indigeneity and the Work of Settler Archives. Settler Colonial Studies 5(2):109−116. Baraga, Frederic. 1878. A Theoretical and Practical Grammar of the Otchipwe Language for the Use of Missionaries and Other Persons Living Among the Indians, second edition. Montreal: Beauchemin & Valois. Bobaljik, Jonathan D. 1998. Researcher-Activist-Indigenous Collaborations in Indigenous Language Maintenance. Bicultural Education in the North: Ways of Preserving and Enhancing Indigenous Peoples’ Languages and Traditional Knowledge, ed. by Erich Kasten, pp. 13−28. Münster: Waxmann Velag. Brown, Jennifer S. H. 2012. As for Me and My House: Zhaawanaash and Methodism at Berens

|

232    Mskwaankwad Rice

River, 1874–1883. Papers of the 40th Algonquian Conference, ed. by Karl S. Hele and J. Randolph Valentine, pp. 79–96. Albany, NY: SUNY Press. Charity Hudley, Anne H., Christine Mallinson, Mary Bucholtz, Nelson Flores, Nicole Holliday, Elaine Chun, and Arthur Spears. 2018. Linguistics and Race: An Interdisciplinary Approach towards an LSA Statement on Race. Proceedings of the Linguistic Society of America 3(8), ed. by Patrick Farrell, pp. 1–14. DOI: https://doi.org/10.3765/plsa.v3i1.4303. Charity Hudley, Anne. H., Christine Mallinson, and Mary Bucholtz. 2020. Toward Racial Justice in Linguistics: Interdisciplinary Insights into Theorizing Race in the Discipline and Diversifying the Profession. Language 96(4):e200−e235. Clarke, Damon. 1996. What My Hualapai Language Means to Me. Stabilizing Indigenous Languages, ed. by Gina Cantoni, pp. 82–85. Flagstaff, AZ: Center for Excellence in Education. Davis, Jenny L. 2017. Resisting Rhetorics of Language Endangerment: Reclamation through Indigenous Language Survivance. Language Documentation and Description 14:37–58. Dobrin, Lise M. and Jeff Good. 2009. Practical Language Development: Whose Mission? Language 85(3):619–629. Errington, J. Joseph. 2008. Linguistics in a Colonial World. Oxford: Blackwell. Gaby, Alice and Leslie Woods. 2020. Toward Linguistic Justice for Indigenous People: A Response to Charity Hudley, Mallinson, and Bucholtz. Language 96(4):e268–e280. Gregory, Alice. 2021. How Did a Self-Taught Linguist Come to Own an Indigenous Language? New Yorker. https://www.newyorker.com/magazine/2021/04/19/how-did-a-self-taughtlinguist-come-to-own-an-indigenous-language. Harvey, Sean P. 2015. Native Tongues. Cambridge, MA: Harvard University Press. Harvey, Sean P. 2018. Native Views of Native Languages: Communication and Kinship in Eastern North America, ca. 1800–1830. The William and Mary Quarterly 75(4):651–684. Holton, Gary, Wesley Y. Leonard, and Peter L. Pulsifer. Indigenous Peoples, Ethics, and Linguistic Data. 2022. The Open Handbook of Linguistic Data Management, ed. by Andrea L. Berez-Kroeker, Bradley McDonnell, Eve Koller, and Lauren B. Collister, pp. 49–60. Cambridge, MA: MIT Press. Huisken, Dylan Fredric. 2013. “We Have Found the Native Tongue Indispensable”: Missionary Interactions with the Dakota Language, 1834–1893. PhD thesis, University of Montana. Krieger, Carlo J. 1989. Ethnogenesis or Cultural Interference? Catholic Missionaries and the Micmac. Actes du 20e Congrès des Algonquinistes, ed. by William Cowan, pp. 193–200. Ottawa: Carleton University Press. Kubota, Ryuko. 2020. Confronting Epistemological Racism, Decolonizing Scholarly Knowledge: Race and Gender in Applied Linguistics. Applied Linguistics 41(5):712–732.

|

Legacy and Obligations of Algonquian Studies   233

Leonard, Wesley Y. 2018. Reflections on (De)colonialism in Language Documentation. Reflections on Language Documentation 20 Years After Himmelmann 1998, ed. by Bradley McDonnell, Andrea L. Berez-Kroeker, and Gary Holton, pp. 55–65. Honolulu: University of Hawai‘i Press. http://hdl.handle.net/10125/24808. Leonard, Wesley Y. 2020. Insights from Native American Studies for Theorizing Race and Racism in Linguistics (Response to Charity Hudley, Mallinson, and Bucholtz). Language, 96(4):e281–e291. Leonard, Wesley Y. 2021a. Centering Indigenous Ways of Knowing in Collaborative Language Work. Sustaining Indigenous Languages: Connecting Communities, Teachers, and Scholars, ed. by Lisa Crowshoe, Inge Genee, Mahaliah Peddle, Joslin Smith, and Conor Snoek, pp. 21–33. Tucson: Northern Arizona University Press. Leonard, Wesley Y. 2021b. Toward an Anti‐Racist Linguistic Anthropology: An Indigenous Response to White Supremacy. Journal of Linguistic Anthropology 31(2):218–237. Linguistic Society of America. 2019. LSA Statement on Race. https://www.linguisticsociety.org/ content/lsa-statement-race. Lock, Mark, Faye McMillan, Bindi Bennett, Jodie Lea Martire, Donald Warne, Jacquie Kidd, Naomi Williams, Paul Worley, Peter Hutten-Czapski, and Russell Roberts. 2022. Position Statement: Research and Reconciliation with Indigenous Peoples in Rural Health Journals. Australian Journal of Rural Health 27(1):1–2. Long, John S. 1985. Rev. Edwin Watkins: Missionary to the Cree, 1852–1857. Papers of the 16th Algonquian Conference, ed. by William Cowan, pp. 91–117. Ottawa: Carleton University Press. Mellow, J. Dean. 2015. Decolonizing Western Science, Research, and Education: Valuing Linguistic Diversity. Honoring Our Elders: Culturally Appropriate Approaches for Teaching Indigenous Studies, ed. by Jon Reyhner, Joseph Martin, Louise Lockard, and Willard Sakiestewa Gilbert, pp. 45–60. Flagstaff, AZ: Northern Arizona University. Papers of the Algonquian Conference. 2022. Style Guide. https://algonquianconference.atlasling.ca/wp-content/uploads/2015/10/PAC-Style-Sheet-04-18-17.pdf. Peltier, Sharla. 2021. Anishinaabewin: An Epistemological and Research Method Framework. Sustaining Indigenous Languages: Connecting Communities, Teachers, and Scholars, ed. by Lisa Crowshoe, Inge Genee, Mahaliah Peddle, Joslin Smith, and Conor Snoek, pp. 49–62. Tucson: Northern Arizona University Press. Podruchny, Carolyn. 1995. “I Have Embraced the White Man’s Religion”: The Relations between the Peguis Band and the Church Missionary Society, 1820–1838. Papers of the 26th Algonquian Conference, ed. by David H. Pentland, pp. 350–378. Winnipeg: University of Manitoba Press.

|

234    Mskwaankwad Rice

Rice, Keren. 2006. Ethical Issues in Linguistic Fieldwork: An Overview. Journal of Academic Ethics 4:123–155. Snow, Kevin. C., Danica G. Hays, Guia Caliwagan, David J. Ford Jr., Davide Mariotti, Joy Maweu Mwendwa, and Wendy E. Scott. 2016. Guiding Principles for Indigenous Research Practices. Action Research 14(4):357–375. Sullivan, Michael D. Sr. 2020. Relativization in Ojibwe. Lincoln: University of Nebraska Press. Truth and Reconciliation Commission of Canada. 2015. Honouring the Truth, Reconciling for the Future: Summary of the Final Report of the Truth and Reconciliation Commission of Canada. Winnipeg: Truth and Reconciliation Commission of Canada. Tsikewa, Adrienne. 2021. Reimagining the Current Praxis of Field Linguistics Training: Decolonial Considerations. Language 97(4):e293–e319. Tuck, Eve, and K. Wayne Yang. 2012. Decolonization Is Not a Metaphor. Decolonization: Indigeneity, Education & Society 1(1):1–40. Valentine, J. Randolph. 1994. Ojibwe Dialect Relationships. PhD thesis, University of Texas at Austin. Verwyst, Chrysostom. 1900. Life and Labors of Rt. Rev. Frederic Baraga, First Bishop of Marquette, Mich. Milwaukee, WI: M.H. Wiltzius. Westman, Clinton N. 2012. Cree Pentecostalism and Its Others. Papers of the 40th Algonquian Conference, ed. by Karl S. Hele and J. Randolph Valentine, pp. 395–418. Albany, NY: SUNY Press. Wilson, Rev. Edward F. 1874. The Ojebway Language: A Manual for Missionaries and Others Employed among the Ojebway Indians. Toronto: Rowsell and Hutchison. Woodbury, Anthony C. n.d. Betty Birner (ed). What Is an Endangered Language? Linguistic Society of America Pamphlet. https://www.linguisticsociety.org/resource/faq-whatendangered-language. Younging, G. 2018. Elements of Indigenous Style: A Guide for Writing by and about Indigenous Peoples. Edmonton, AB: Brush Education Inc. Zsiga, Elizabeth C. 2013. The Sounds of Language: An Introduction to Phonetics and Phonology. Malden, MA: Wiley-Blackwell.

Plains Cree Textual Analysis with PCA Across the Bloomfield and Ahenakew-Wolfart Subcorpora Katherine Schmirler and Antti Arppe

T

his paper presents an analysis of Plains Cree text types using a morphosyntactically tagged corpus and Principal Component Analysis (PCA).1 The corpus used herein, containing 152,405 words of Plains Cree, offers opportunities to explore the frequency of morphosyntactic features in the language, with many directions for research questions to take. As this corpus of Plains Cree contains texts from many speakers in various communities, of different ages in different times, one must consider to what extent internal differences are obscured when the corpus is taken as a whole, and in what ways the corpus can be analyzed for a fuller picture of Plains Cree spoken text. However, the sheer number of features can quickly overwhelm when narrowing down a research question, and so a dimension reduction technique, such as PCA, becomes invaluable for exploring overall patterns and devising research questions. In this paper, two main distinctions in text types are highlighted: 1) a distinction between narrative and counseling texts; and 2) within narratives, a distinction in how a story is presented, with or without considerable direction quotation. These and other distinctions are discussed throughout, as well as their overlap with each other and with traditionally identified text types.

The Corpus

The Plains Cree corpus consists of two subcorpora, the Bloomfield subcorpus (abbreviated BT, “Bloomfield texts”) collected and edited by Leonard Bloomfield

|

   235

|

236    Katherine Schmirler and Antti Arppe

in the 1920s (Bloomfield 1930, 1934), and the Ahenakew-Wolfart (A-W) subcorpus collected and edited by Freda Ahenakew and H. C. Wolfart in the latter decades of the twentieth century (Ahenakew 2000; Bear et al. 1992; Kâ-Nîpitêhtêw 1998; Masuskapoe 2010; Minde 1997; Vandall and Douquette 1987; Whitecalf 1993). Together, these volumes contain 152,405 orthographic words of Plains Cree (241,152 tokens including non-Cree words, numerals, and punctuation). The corpus includes 1) manually verified morphological analyses and 2) syntactic tags provided by a Constraint Grammar-based automatic parser (Arppe et al. 2020; Harrigan et al. 2017; Schmirler et al. 2018, 2023; Snoek et al. 2014). The corpus is divided by chapter; these are referred to as “texts” throughout. The text types are drawn from the descriptions of the texts by the speakers or editors. A commonly cited distinction between text types in Plains Cree is between âtayôhkêwina ‘sacred stories’, events that occurred before the world was “in its present, definitive state” (Bloomfield 1930:6), and âcimowina, everything else, including discourses and narratives. Freda Ahenakew comments on subcategories of âcimowina in Vandall and Douquette (1987:xii–xiii): kayâs-âcimowina ‘old-time stories’ (the distant past, but not sacred), kakêskihkêmowina ‘counseling texts’ (comments on differences between traditional Cree life and modern life), wawiyatâcimowina ‘funny stories’, and âcimisowina ‘stories about oneself’.

Method

In this section, we briefly introduce register analysis and Principal Component Analysis, two methods used in the analyses of this paper, as well as the data preparation process. Register Analysis

Register analysis is an approach to textual analysis that focuses on linguistic features in combination with the context in which the text is produced and how these features function for the communicative purpose of the text. Register analysis, as laid out by Biber and colleagues (e.g., Biber 1991; Biber et al. 1998, 2002; Biber and Conrad 2019), involves three main steps. The first step describes the situational context: Who is speaking to whom? What is their relationship? What is the purpose of the interaction? The second step considers the relative frequencies of various

|

Plains Cree Textual Analysis with PCA   237

linguistic features in each type of text: certain features are more frequent in certain text types. The third step identifies functional relationships between the context and the linguistic features: Why do those features occur in those text types? This process is cyclical: after the initial analysis, one returns to the situational context, re-evaluating based on linguistic features. For example, a cursory exploration of texts in Plains Cree found that personal stories include more first person exclusive, while counseling texts include more first person inclusive. The former includes speakers talking about their own lives or families, which do not include the listeners, while the latter involves shared Cree identity with the audience (Schmirler and Arppe 2020). A simplified version of this method is employed herein for the sake of brevity. Principal Component Analysis

Principal Component Analysis (PCA) is one of many statistical techniques that reduce the dimensions of large, sparse datasets—an apt description for natural language data used in register analysis (e.g., Biber and Conrad 2019). Unsurprisingly, datasets drawn from corpora can be sizable, with many linguistic features, many words in which they occur in combination, and still other features of the texts themselves to consider. Reducing the dimensions of such datasets is invaluable, as the occurrence and co-occurrence of hundreds of features are not easily interpreted. In PCA, the simplification of large datasets allows for the most important independent variables to be extracted by creating new variables from the original, calling these principal components (PCs). PCs are created as combinations of all independent variables in the data, so as many PCs are generated as original variables, and the correlation between PCs is zero. PCs are then ranked by how well they account for the variance in data, with PC1 accounting for the largest proportion of variance, PC2 the next largest proportion, and so forth. Though PCs are notoriously less interpretable than the original variables, the PCs help to visualize the relationships and identify features and interactions to explore in future, more precise analyses. We focus on the first and second PCs in the following analyses. Each PC consists of “positive features” that characterize texts on the positive end of each axis in a plot and “negative features” that characterize texts on the negative; the texts are plotted with PC1 on the x-axis and PC2 on the y-axis, giving a two-dimensional visualization of similarities and differences among the texts. The features and plots can be used together to see how texts differ, and how they group with respect to similar features,

|

238    Katherine Schmirler and Antti Arppe

and to link groups to existing text type labels in Plains Cree. Where groups cannot be linked to existing labels, new categories can be explored. The Data

Before the analysis, non-Cree tokens were removed from the morphosyntactically tagged corpus (English and French words, punctuation, etc.); this choice is motivated by previously noted differences between the two subcorpora, namely far more non-Cree words and punctuation in A-W than in BT. The point of interest for the present investigation is how Plains Cree features differ between texts and text types, rather than how English use is reflected in the analysis.2 After these forms were removed, the tags were converted into a table where each row represents one chapter, identified by speaker initials, chapter number, and in the case of BT as speakers overlapped, the volume, and each column represents a morphological feature or syntactic function tag. Each cell contains the number of times a particular feature tag occurs in a text. This method does not consider how morphological features interact (e.g., how many VAIs are 1Sg, how many actors are nouns vs. pronouns). Further tags (rather than whole words and all their features) were then trimmed from the table; these were primarily metalinguistic tags, identifying orthographic variation or an uncertain analysis. Additionally, only actor and goal tags are retained for the purposes of this investigation, removing tags for negation, prepositional phrases, quantifiers, and temporal and locative particles. Initial modeling found related features that ranked similarly within the same PC: future would group with its subtypes, definite and intentional, indicating that future tense overall, rather than a particular type, characterized certain texts, so the subtypes were trimmed. Similarly, immediate and delayed imperatives co-occurred with the overall imperative tag and so were trimmed, leaving only the imperative. PCA is undertaken herein using the prcomp function in R; the frequencies are normed, dividing the number of occurrences of each feature in a chapter by the total number of words, the variables are rotated, centered, and scaled.3 The results are presented in three stages: the first includes the full corpus, the second just BT, and the third just A-W. This approach demonstrates similarities and differences between and within the subcorpora; the by-chapter division demonstrates similarities and differences between texts produced by the same speaker.

|

Plains Cree Textual Analysis with PCA   239

Results

The results are divided into three sections. First we present the results of the full corpus analysis and then those for the Bloomfield and Ahenakew-Wolfart subcorpora respectively. Full Corpus

The full Plains Cree corpus consists of 140 chapters from twenty speakers, drawn from nine volumes. Before preprocessing the morphosyntactically tagged corpus for PCA, the full Plains Cree corpus contains 241,152 tokens (37,941 types), including non-Cree words, punctuation, and metadata. After preprocessing, these are reduced to 152,405 Plains Cree word tokens (31,616 types). The dataset constructed from this preprocessed data includes 571 feature tags, which are trimmed to eighty-eight for the purposes of PCA. Of the 483 tags removed, 378 (78.26%) were preverb or prenoun tags. The first two PCs account for 17.72% and 7.65% of variance respectively, totaling 25.37%. These PCs are visualized in Figure 1. This plot shows the texts of the full corpus, divided by subcorpus with their distributions shown by ellipses.

FIGURE 1. Full corpus PCA: Visualizing PC1 and PC2

|

240    Katherine Schmirler and Antti Arppe

PC1 occurs along the x-axis, so features on the positive side of PC1 characterize A-W and features on the negative side characterize BT. For PC2 along the y-axis, both subcorpora contain features from the positive and negative ranges. Even without reference to features, we can observe that the subcorpora, while different in some ways, are similar in others. The features characteristic of the positive and negative ends of each PC are given in Table 1. Not all features are discussed here for the sake of brevity; instead, similar or mutually exclusive features that occur in opposition are highlighted, alongside those that reinforce patterns. In PC1, there is a contrast between more particles and pronouns on the positive side (i.e., A-W) and more verbs on the negative (i.e., BT). Within word classes, especially verbs, we see the occurrence of past tense and conjunct forms increasing positive scores along PC1, and the occurrence of present, future, independent, and imperative verb forms contributing to a negative score along PC1; mixed VTAs on the positive and nonlocal, direct, and local VTAs on the negative; first persons (both singular and plural) on the positive side, and third person (both proximate

TABLE 1. Full corpus PCA: Features of PC1 and PC2 POSITIVE

PC1

PC2

NEGATIVE

Verbs Present, future tense Independent, imperative verbs 3′, 3′O, Px3Sg, 3Sg, Px3′, 0′Sg, Px2Sg, 2Sg Nonlocal, direct, local VTAs Obviative, dependent, animate, vocative nominals @, @GOAL> Particles, pronouns Past tense Conjunct verbs Excl, 0Sg, 1Sg, Px21Pl, X, Incl, Px1Pl Mixed VTAs Inanimate, plural nominals

|

Plains Cree Textual Analysis with PCA   241

and obviative) and second person singular on the negative. There are inanimate and plural nominals on the positive side, and obviative, dependent, animate, and vocative nominals on the negative. Many of these features can be interpreted in groups on each side of the PC as well, especially the negative: we see the occurrence of nonlocal verbs alongside third person proximate and obviative tags, as well as more nominals with animate features, such as obviatives, dependent nouns, and vocatives. These features co-occur with the prevalence of overt actor tags. The negative side also shows a prevalence of imperative verbs and second person tags, which likely occur together, and, despite the relative lack of speech act participants, local VTAs as well—perhaps due to the overall prevalence of verbs here. Groups are less clear on the positive side, though a prevalence of first persons may be connected to the prevalence of mixed VTAs, and the prevalence of inanimate person tags to the prevalence of inanimate nominals. While PC1 highlights differences between the subcorpora, PC2 highlights similarities in their internal variation. In PC2, there is a contrast between VTAs, quotative verbs, and personal pronouns on the positive side (the top half of the plot), with VAIs and particles on the negative. Future tense contrasts with past, as for PC1, but while imperative and future conditional verbs characterize the positive side, no verbal order characterizes the negative. There is a contrast in persons as well, though now more strongly divided between SAPs and non-SAPs, though 3SgO does occur alongside the SAPs. Again, the person contrasts align with VTA feature contrasts: where there are SAPs, mixed and local VTAs also occur, in contrast to third persons and nonlocal VTAs. Obviative and animate nominals again characterize the negative side, alongside third persons and overt actors and goals. However, dependent nouns now occur alongside the first and second persons: likely the possessors of these nouns. We identify two main patterns in these PCA results for the full Plains Cree corpus. First, we bring to attention the similarities between the features identified in PC1 and PC2: many of the features that divide the subcorpora also explain their internal variation. Second, focusing on PC2, the features that characterize each end of this range point toward not necessarily different text types themselves, but different ways of presenting narrative, which occur in both subcorpora. In addition to the grouped features discussed for PC2 above, the positive range is characterized by quotative verbs and thus more quotation. It is likely that the first and second person features, interrogative particles, interjections, and imperatives occur within the quotation. The less frequent use of quotation and greater use

|

242    Katherine Schmirler and Antti Arppe

of third persons and other particle classes group together on the negative side of PC2. Though further research is needed to confirm the relationships between these features, their co-occurrences may be attributed to how dialogue is presented: with or without direct quotation. BT Subcorpus

The BT subcorpus consists of eighty-two chapters from eight speakers, drawn from two volumes. Before preprocessing the morphosyntactically tagged corpus for PCA, BT contains 102,962 tokens (15,287 types), including non-Cree words, punctuation, and metadata. After preprocessing, these are reduced to 72,475 Plains Cree word tokens (15,267 types). The dataset constructed from this preprocessed data includes 354 feature tags, which are trimmed to eighty-four for the purposes of PCA. Of the 270 tags removed, 187 (69.26%) were preverb or prenoun tags. For BT, the first two PCs account for 11.92% and 8.84% of variance respectively, totaling 20.76%. These PCs are visualized in Figure 2. This plot shows the texts of the BT subcorpus, divided by volume with their distributions shown by ellipses. The volumes are used here as a rough approximation of text type, as Sacred Stories of the Sweetgrass Cree (SSSC, Bloomfield 1930) is entirely sacred stories and Plains Cree Texts (PCT, Bloomfield, 1934) is mostly nonsacred stories with a small section of sacred stories. Unlike for the full corpus, the plot does not present any immediate division between the volumes. However, the distribution of SSSC is almost entirely within that of PCT, suggesting that sacred stories behave as a subtype of narrative. Individual coding of chapters for text type will refine this picture in future research. The features characteristic of the positive and negative ends of each PC are given in Table 2. Again, not all features are discussed. Many of the patterns discussed for the full corpus are also evident here for BT. In PC1, there is a distinction between particles and verbs, though other verb classes also distinguish each end of the spectrum; past tense contrasts with future; conjunct verbs contrast with other orders. Similar patterns occur in PC2, with particles and some verb classes contrasting with other verb classes, as well as tense and order contrasts—in both the full corpus and BT analyses, conjunct verbs are characteristic of one side of the PC while other orders group in opposition. The patterns of person and narrative features are, once again, the most salient. PC1 is a little different from the others seen thus far; rather than a strong distinction between SAPs and nonSAPs, we see 3Pl and obviative contrasting with other persons. However, the other

|

Plains Cree Textual Analysis with PCA   243

FIGURE 2. BT subcorpus PCA: Visualizing PC1 and PC2

TABLE 2. BT subcorpus PCA: Features of PC1 and PC2 POSITIVE

NEGATIVE

Particles, VIIs Past tense Conjunct verbs

Verbs, VTAs, quotative verbs Future tense Future conditional, imperative, independent verbs Px1Sg, 2SgO, 1SgO, 3SgO, 1Sg, 3Sg, 2Sg, Px3Sg, Px2Sg, Px1Pl

Px3Pl, 0′Sg, Px3′ PC1

PC2

Plural, inanimate nominals, proper nouns Quantifiers, numerals, locative, temporal, negative particles, indeclinable nominals @ACTOR> Particles, VIIs, quotative verbs, VTIs, VAIs Future, past tense Independent, future conditional verbs 1Sg, 0Sg, 0′Sg, 1SgO, Px1Sg, Px21Pl, 2SgO, Excl Local VTAs Inanimate nominals

Dependent, singular, vocative nominals Interrogative particles, interjections

VTAs, pronouns Present tense Conjunct verbs 3′O, Px3Sg, 3′′O, 3′, Px2Sg Nonlocal, direct VTAs Obviative, animate, dependent, plural nominals @, @ACTOR>, @, @ACTOR>

Future tense Conjunct, future conditional verbs Px21Pl, Incl, 0Sg, Px2Sg, 2SgO, 2Sg Inanimate, singular nominals Indeclinable nominals, negative particles @, @ACTOR>, @ ] ;

The rules can then proceed: for the most part, they are implemented very much like the rewrite rules with only minor changes in the syntax. At other times, more complex workarounds are needed. The implementation of the gemination rule is given in (3). This is a case where the rule can be implemented almost exactly like the rewrite rule, though to simplify the necessary code, the rule is reproduced for every potential geminate stop. (3) Modeling the Gemination Rule (C1 → C2 / _+C2) a. define Ck2kkRule [ Cshort -> k || _ %> k ] ;

b. define Cp2ppRule [ Cshort -> p || _ %> p ] ; c. define Ct2ttRule [ Cshort -> t || _ %> t ] ;

d. define CC2GemRule [ Ck2kkRule .o. Cp2ppRule .o.

Ct2ttRule ] ;

(4) Gemination Rule: Rewrite to Regular Expression with C1

→

C2

/

_

+

C2

Cshort

->

k

||

_

%>

k

We begin by walking through the relationship between the rewrite rule given by Frantz (2017:176) and the model rule in (3). The rule also begins with a definition,

|

Morphophonological Rule Development   257

though now the name of the rule is given instead of a label for a set of characters. In (3a), the rule name is Ck2kkRule: the name tells us this is a rule, where a sequence of C + k goes to (“2”) k + k. The rule itself is then given; though the formalism uses some different characters, this has almost a one-to-one relationship to the rewrite rule, as in (4). C1 corresponds to Cshort, C2 to , the slash to the double pipe , and the suffix boundary to %>; the arrows and underscore are essentially identical. This rule can thus be read as “a short consonant becomes when it occurs before a suffix boundary and another ”. The rule is then repeated for each possible geminate stop, before the group is composed together using the operator .o. in (3d). The rule adaptation proceeds in this manner down the list in Frantz (2017), with testing examples given after each rule (see the rule-testing examples in (6) and (7)) for how these are used). At the end of the rule list, a final definition is created that composes all the rules together; this is the point at which rule ordering is undertaken. Some rule ordering changes were needed, though more likely on the basis of how they were formulated rather than because of errors in Frantz’s analysis. Though Frantz (2017) lists rules numbered from 1 to 26, these do not align perfectly with the composed rule list, as will be seen in the testing examples below when compared to (1) above.

Testing the Rules

This section begins with a more detailed explanation of the script, for those more interested in the finer computational details. The following subsection, which gives examples of the rule testing, begins with a less computational description of the process, for those who are so inclined. The Details of Scripting

A crucial feature in the development of computational models of (morpho)phonology is the ability to test the functioning and impact of every individual rule as part of the entire rule set specifying the overall phonological processes. In the computational modeling of morphology and phonology, the first dominant paradigm was the two-level model (twolc, Koskenniemi 1983, a variant of FSTs, e.g., Beesley

|

258    Schmirler, Arppe, and Genee

and Karttunen 2003), which in the 1980s and 1990s (and even later) was used for full-scale models of the inflection, derivation, and morphophonology of the entire core vocabulary of languages. In this paradigm, the individual phonological rules were understood to apply in parallel, which was desirable at the time, as the computational compilation of these rules required significantly less memory than the alternative, a cascade of rewrite rules that achieve the same changes (Roark and Sproat 2007:102ff). Importantly, compilers for two-level rules, such as XFST and HFST (Helsinki FST, Lindén et al. 2011) included a functionality for scrutinizing the application of all the individual rules constituting the entire rule set (e.g., hfst-pair-test in the HFST suite). However, comprehensive sets of two-level rules are notoriously difficult to troubleshoot and debug, in particular when trying to resolve the interdependencies of these rules (as experienced, e.g., by Harrigan et al. 2017 in their modeling of Cree verbs). Additionally, the significant increases in available computer memory and processing capacity have made the computational compilation of ordered cascades of phonological rewrite rules more practical and feasible since the late 1990s—an approach that an increasing number of computational linguists have since adopted (M. Hulden, pers. comm., June 2014). Dunham (2014:175) describes his testing process for Blackfoot rewrite rules: test cases are included in the rule file, and then the database has built-in tools to determine whether the rules overall give the desired output from the input. However, as far as we are aware, in contrast to twolc, there has not existed any practical implementation for the examination of the effect of individual rewrite rules, rather than the effects of the full rule list, within an ordered cascade of such rules, a functionality that we found would be necessary in the incremental development and testing of a phonological rule set for any language. To address this need, a pair of shell scripts were written that operate on the source code of the phonological rewrite rules written using the xfscript formalism, where, as seen above for gemination, each rule is first individually defined and assigned a name (in effect defining a mini FST), and then the ordered sequence of the application of all the rules is specified as a regular expression composing together (and thus ordering) each individually defined rule (i.e., these aforementioned mini FSTs) at the very end of the source code file (introduced by the regex function, see (5)). The first of the shell scripts, compile-rewrite-rules.sh,5 takes as its primary argument the path to the phonology.xfscript file, which contains the source code for the rewrite rules and their master composition, and creates, using the foma compiler (by default, with HFST as the alternative FST format), the

|

Morphophonological Rule Development   259

full set of mini FSTs corresponding to each individual rewrite rule in their final order (stored in a subdirectory named foma/, or hfst/). The second of the shell scripts, test-rewrite-rules.sh,6 again takes as its first argument the path to the phonology.xfscript file with the source code, and takes one or more pairings of an underlying word form and the corresponding expected surface word form through the ordered sequence of the mini FSTs created by the first script. In effect, the output of each preceding rule FST becomes the input to the following rule FST. The second argument determines whether the effects of each individual rule are shown (long), or if only the rules that have been triggered for some particular intermediate input are shown (diff), or whether only the aggregate result of the sequence of all rules is presented and compared with the expected output (default). (5) Composing the Rules in Order read regex [ sConnRuleB .o. sConnRuleA .o. CC2GemRule .o. sInsRule .o. h2ssRule .o. oReplaceRule .o.

wi2oRuleA1 .o. wi2oRuleA2 .o. wi2oRuleB .o. k2ksRule .o. i2NeutRule .o. Vowel2GlideRule .o. VV2VRule .o.

iLossRule .o. tAffricRule .o. iAbsorbRule .o. ihLossRule .o. PresibRule .o. GlideReductRule .o. GlideLossRule .o. iyiReductRule .o. PostsibRule .o. GlotMetaRule

.o. GlotLossRule .o. GlotAssRule .o. GlotReductRule

.o. VowelEpenRule .o. sss2ssRule .o. AccentRule .o. deep2surfRule ] ; Examples of Rule Testing

For those less interested in the minutiae of scripting, we put it more plainly: in order to test the rules, a script was used that takes an underlying word form (e.g., a string of morphemes with no sound changes applied), applies the rules in order, and compares the resulting form to an expected surface form. These testing pairs of underlying and surface forms can be provided separately or included in the rule file. In the case of Blackfoot, we have done the latter and marked out these to-betested pairs by a special character sequence at the beginning of the line;7 typically, the testing pairs are given in conjunction with the rule they are most relevant to. The latter of the scripts described above can be used in two ways: we can test an individual rule or small set of rules by including just one test pair as the input to the

|

260    Schmirler, Arppe, and Genee

script, or we can use the entire file with a search for the special characters so that all the test pairs are included in the testing. When testing one pair, the long form of the script’s output is used: this format presents every rule in the list, indicates whether or not it applies to the underlying form in question, and after all rules have been applied, the resulting surface form is compared to the expected surface form and marked as successful or not. Thus, the developers can see whether a rule that should apply is not applying or whether a rule that should not apply is applying, and adjustments can be made. This is the crucial capability of the scripts that differs from the testing process in previous work such as Dunham (2014), where the impact of individual rules is not shown, and thus the developer must find for themselves which rules are under- or over-applying—thus the scripts are key to making the development process more transparent and efficient. The example pair given for gemination in (1) is given for the testing example here in (6). This pair also demonstrates two other sound changes, as well as a final rule that removes morpheme boundaries. Sections where no rules apply are elided for the sake of brevity; this is a reporting option provided by the script (diff). This testing output shows various stages of the rule application, starting with stage 0 of the underlying form of the word, with standing in for Frantz’s , and two suffix boundaries. The testing then proceeds down the ordered list of rules: a hyphen following the number shows that the rule does not apply, and the plus sign shows that a change occurs. As each rule applies, its output becomes the input for the next, thus rule 3 here changes to and this change continues throughout. Rule 11, neutralization, then applies, changing to , followed by glide loss in rule 20, and the morpheme boundaries are deleted by rule 30. The final lines give the underlying and surface forms, where the equal sign indicates that the surface form output by the model matches the expected surface form in the test pair, and then a summary that is more useful in the full file testing, but otherwise reiterates that the model successfully changed the underlying form to the surface form. (6) Testing Gemination, Neutralization, and Semivowel Loss 0: LEXC

nitáni2t>k>wa

3: CC2GemRule

+ nitáni2k>k>wa

20: GlideLossRule

+ nitánik>k>a

---------------------|--------------11: i2NeutRule

+ nitánik>k>wa

|

Morphophonological Rule Development   261

30: deep2surfRule

+ nitánikka

---------------------|---------------

1: 1-30: nitáni2t>k>wa -> nitánikka (=)

STATS: Correct: 1/1 - Wrong: 0/1 - Missed: 0/1 SUMMARY - SUCCESS: 1/1 - FAIL: 0/1 - PARTIAL: 0/1 - OTHER: 0/1

To examine the overall efficacy of the rules during development, the overall test looks at all the pairs in the file, and instead of examining the output of each rule for each pair, the short output shows each underlying and expected surface pair and indicates whether the given surface form matches the expected surface form. The final summary then gives the overall success and failure rates, indicating how well the rules work, at least for the limited testing set included in the model file. Example (7) shows a shortened version of the output, marking with an asterisk the failure that results from incorrect interactions between rules, to be explored in further development. The issue arises between glottal metathesis and glottal deletion; the metathesis rule does not accurately retain morpheme boundaries, which then blocks the deletion rule from applying. Where an error occurs, the form in parentheses gives the actual output, and the form following the -/-> is the expected output, that is, what the model is ideally trying to achieve. From this output, we can then focus on the individual example cases where errors arise and use the long output to see which rules are (mis)applying and correct them. (7) Full File Test Results with Failure in Word 32







1: 1-30: nitáni2t>k>wa -> nitánikka (=)



STATS: Correct: 1/1 - Wrong: 0/1 - Missed: 0/1

2: 1-30: nitáni2t>awa -> nitánistawa (=)

STATS: Correct: 1/1 - Wrong: 0/1 - Missed: 0/1



3: 1-30: ann>yi2hka -> annisska (=)







STATS: Correct: 1/1 - Wrong: 0/1 - Missed: 0/1 . . .

* 32: 1-30: á7 áí′sttohkohpiy′ssi

( áí′sttohkohpiy′ssi) |



-/-> áísttohkohpiy′ssi

STATS: Correct: 0/1 - Wrong: 1/1 - Missed: 1/1

|

262    Schmirler, Arppe, and Genee





. . .



37: 1-30: atsikí>istsi -> atsikíístsi (=)



SUMMARY - SUCCESS: 36/37 - FAIL: 1/37 - PARTIAL: 0/37





STATS: Correct: 1/1 - Wrong: 0/1 - Missed: 0/1 - OTHER: 0/37

Thus, we might say that, on the basis of a very limited development set of 37 words, our phonological model is 97.3% accurate. This is of course a very small set, which moreover consists of relatively “clean” examples selected by Frantz (2009, 2017) to demonstrate his rules, though it is nevertheless a promising start of the development of a phonological model. To further examine the efficacy of the rules, we compiled a longer list of words drawn randomly from other sections of the grammar (Frantz 2017). The model performed much less effectively against these additional 169 words, with only 60 (35%) of them being output correctly. Of the 109 failed words, 70 (64%) failed due to the rules for accented vowels, 10 (9%) were due to rules for deletion of a vowel at the beginning or end of a word, and 16 (15%) involved both issues. The remaining 13 (12%) failed pairs largely involved interactions between vowels at morpheme boundaries. These three phenomena have not undergone any detailed modeling and thus examples were not included in the initial development set. The failed words are an excellent indication of the rules that were not fully specified in Frantz’s appendix and thus were not initially included, or other rules that were not implemented properly and require further work. Now that the rule list has undergone initial modeling and has demonstrated both the application of the scripts and the usefulness of existing grammatical description in development, we can look to previous work and further Blackfoot examples in testing to continue the modeling process in future research.

Comparing Approaches

In this section, we highlight some of the main differences between the rules in the Blackfoot model presented here and those developed by Dunham (2014). We intend for this section to add another layer to the tutorial aspect of this paper, as there is no one correct way to present rules in the XFST syntax, and different

|

Morphophonological Rule Development   263

linguists or developers will prefer different approaches in the creation of rule sets for any language. To compare our approach to the model developed by Dunham (2014), we must first address key differences in the stages of model development. For our purposes, we wanted to create a first iteration of a model where the rules were applied as directly as possible from Frantz (2009, 2017), tested against the examples given.8 Dunham (2014) instead detailed a fuller model aiming to capture the phonology of Blackfoot as accurately and precisely as possible while still allowing for phonological and orthographic variation in a larger database. As a result of this, our rules are much more general: for example, the gemination rule in examples (3) and (4) is given very generally by Frantz, but in actuality only applies to consonants before (i.e., this is the only context that ever occurs to trigger the rule), and thus the gemination rule in Dunham (2014) contains only two sub-rules: the first to geminate after and a second to deal with the sequence , which, within a collection of related forms, surfaces as , as in (8). Thus, the two models are not directly comparable at this level, and instead we look to the formulation and ordering of the rules.9 (8) Gemination (Dunham 2014:257) define gemination [





[ a n i s t -> a n i k | | _ "-" k ] .o.

[ t -> k || _ "-"k ] ] ;

Differences in our approach and that of Dunham (2014) can be seen immediately in the character and set definitions. Dunham (2014:255) defines groups for all the phonemes, then vowels, accented vowels, consonants, obstruents, stops, plosives, and glides. Our definitions (see (2)) include long and short vowels and consonants, each then grouped respectively into vowels and consonants, then separate definitions for accented vowels and for glides. Despite each model defining eight groups, the groups certainly differ. Different groups necessarily result in different types of rules, as we reference different categories. As our model is only in the initial stages of development, the sets are far from finalized and we will likely take further inspiration from Dunham; long and short vowels and consonants are referenced rarely in our rules and may be removed in future development. However, we do not foresee using groups for obstruents, stops, and plosives. While these groups are useful in phonological descriptions, overuse of features in rule development

|

264    Schmirler, Arppe, and Genee

can lead to less readability in the rules, which can be detrimental for debugging, especially in projects that span years and involve many researchers who need to understand and edit the rules over time. Considerations of readability and clarity for the development and debugging process also influenced our rule formatting, which differs in some respects from Dunham (2014). Recall again the gemination rule, where three separate sub-rules are defined and then composed together into a single gemination rule: this approach is used at several points throughout our rules. Dunham (2014) simplifies this into a single step, composing his two rules together under one heading, and uses this approach in all rules with different sub-rules, such as in (8) (cf. (3) above). While elegant, this can again lead to less readability and it becomes harder to pinpoint individual issues. In our testing process, we have the option of applying the composed rule, or we could have each sub-rule apply, and thus more easily find which part of a multi-part rule causes the issue in question. However, as the scripts we use for testing are new developments, this was perhaps less of a concern for Dunham. With the scripts, we would recommend the use of clearer, more detailed multi-part rules. The treatment of morpheme boundaries in the rules also differs between our approach and Dunham’s. First, Dunham (2014) uses the boundaries as the basis for epenthesis: when a new character is inserted, a rule treats it as replacement of the morpheme boundary, as in (9a), where the morpheme boundary becomes when it occurs between a stop and . In the same rules in our model, we instead have a sequence of characters including the morpheme boundary, which changes to another sequence to retain the boundary, as in (9b), where a single changes to a sequence of two, with the boundary (Bx) as part of the context, retaining it in the output.10 By retaining the morpheme boundaries throughout as much as possible, with a final rule included that deletes them, the last rule can be excluded from the model in certain cases; for example, an online dictionary can create paradigms where individual morphemes are indicated by color or formatting (e.g., https://southern.verbs.eastcree.org/; https://itwewina. altlab.app/). Second, Dunham marks all morpheme boundaries with the same character, a hyphen, whereas we prefer to distinguish between prefixes (). This distinction allows for different rules to easily be marked to apply at specific boundaries when necessary, and for prefixes and suffixes to be marked differently, as in dictionary paradigms.

|

Morphophonological Rule Development   265

(9) Morpheme Boundary Rule Comparison a. Dunham (2014:259)

define sConnection [

[ "-" -> s || stops _ s ] .o. . . . ] ;

b. Current model

define sConnRuleA [ s -> s s || Consonant Bx _ ,, s -> s s || Consonant Bx _ s ] ;

Finally, there are differences in rule ordering between our rules and Dunham’s, as a result of the different approaches to modeling the rules. Frantz (2009, 2017) offers comments on rule ordering at the end of the rule appendix, which are for the most part borne out in the rules devised for our model and by Dunham (2014). However, some differences do arise, particularly where our rules apply too broadly or too narrowly, and thus interact in ways not foreseen by Frantz or that do not appear in the example test cases. The first case is exemplified by the gemination rule: because it is too broad, it interacts with the s-connection rules, which thus needed to be ordered first in our rule list, though the correction of the gemination rule to only include stops should remove this interaction. The second case is the interaction between Frantz’s rules 9 and 12, desyllabification ({ i → y, o → w} / V+_V) and i-absorption (i → Ø / s_{a,o}) respectively (Frantz 2017:177), which occur as rules 13 and 17 in (5) above. Frantz (2017:179) notes that i-absorption should be ordered before desyllabification, though because the example cases did not include these particular contexts and Frantz’s rule ordering was not closely attended to during initial modeling, the incorrect order was not seen in the testing process and not yet corrected.

Conclusion

The first stages of a Blackfoot morphological model have demonstrated 1) the viability of grammatical descriptions as a basis for computational modeling: where a list of rules exists, they form an excellent starting point for modeling, and 2) the use of scripts to streamline the testing and development of these rules. This rule development can be undertaken by a linguist who is not intimately familiar with the phonology of the language in question, as long as they have some familiarity with the xfscript formalism. Furthermore, the testing scripts offer immediate evaluation

|

266    Schmirler, Arppe, and Genee

of rules throughout the development, speeding development and allowing for individual errors or incorrectly ordered rules to be quickly identified. The same scripts can be used to assess the overall effectiveness of the rules, either on a small test set as demonstrated here, or on a larger set once the data is available. The phonological rules here are ultimately intended to be linked with the morphotactic side of a model, which will also allow for further development and refinement. We end with three main observations: 1) efficient testing, such as that provided by the scripts described here, is key to speeding development; 2) one does not have to start with the morphotax when constructing a morphological model when clearly laid out phonological rules are readily available;11 and 3) one need not base a new model on an old one: while another’s work can be excellent for comparison, starting from scratch can create a model that works better for the current team’s needs, rather than trying to parse and restructure previous rules. Notes 1. Frantz uses a capital (as in (1b)) to indicate what he labels a “breaking i”, which triggers particular morphophonological changes that a “non-breaking i” does not (2017:34). 2. The source code for this Blackfoot phonological model can be found at https://github. com/giellalt/lang-bla/blob/main/src/fst/bla-phonology.xfscript (bbea741) (Schmirler and Arppe under development). 3. The font Courier New is used throughout to indicate code elements or script filenames, (e.g., Vshort refers to the list of short vowels in the code).

4. Abbreviations: AccentRule = accent spread rule, Bx = any morpheme boundary, CC2GemRule = consonant plus consonant goes to geminate rule, Ck2kkRule =

consonant plus k goes to k plus k rule, Clong = long vowel, Cp2ppRule = consonant plus p goes to p plus p rule, Cshort = short vowel, Ct2ttRule = consonant plus t

goes to t plus t rule, deep2surfRule = underlying to surface form rule, FST = FiniteState Transducer, GlideLossRule = semivowel loss rule, GlideReductRule = semivowel reduction rule, GlotAssRule = glottal assimilation rule,

GlotLossRule = glottal loss rule, GlotMetaRule = glottal metathesis rule,

GlotReductRule = glottal reduction rule, Gx = any glide, h2ssRule = x-sibitation rule, HFST = Helsinki Finite-State Transducer, i2NeutRule = I-neutralization rule, iAbsorbRule = i-absorption rule, ihLossRule = ih loss rule, iLossRule

= i-loss rule, iyiReductRule = y-reduction rule, k2ksRule = breaking k rule,

|

Morphophonological Rule Development   267

oReplaceRule = o-replacement rule, PostsibRule = post-sibitation rule,

PresibRule = presibilation rule, regex = regular expression, sConnRuleA = s-connection rule part A, sConnRuleB = s-connection rule part B, sInsRule =

s insertion rule, sss2ssRule = sss-shortening rule, tAffricRule = t-affrication

rule, twolc = two-level compiler, Vacc = accented vowel, Vlong = long vowel,

Vowel2GlideRule = desyllabification rule, VowelEpenRule = vowel epenthesis

rule, Vshort = short vowel, VV2VRule = vowel shortening rule, wi2oRuleA =

coalescence rule A, wi2oRuleB = coalescence rule B, XFST = Xerox-style Finite-State Transducer.

5. https://github.com/giellalt/giella-core/blob/main/devtools/ruletest/compile-rewriterules.sh. 6. https://github.com/giellalt/giella-core/blob/main/devtools/ruletest/test-rewrite-rules.sh. 7. The special character convention that we adopted from the twolc formalism is a pair of lines in the format !!€ underlying-form and !!€ surface-form. We have now also added an auxiliary script, extract-test-cases.sh, to retrieve these

pairs from the phonology source file: https://github.com/giellalt/giella-core/blob/main/ devtools/ruletest/extract-rule-test-cases.sh, which can then be piped for evaluation with test-rewrite-rules.sh

8. Our model forms only the initial stages of the phonological component of a full morphological model for Blackfoot, with much more testing and development to be done. 9. Note that Dunham (2014) makes reference to an earlier edition of the Blackfoot grammar, Frantz (1991). There are minor differences in the phonological rule appendix, though these do not greatly affect the respective phonological models. 10. Our rule is also more detailed, explicitly indicating cases where both or occur in the surface form. The double comma indicates a list of rules; in such a simple case this is deemed fine for readability considerations. Further testing may show that this level of detail is not necessary. 11. We use the term ‘morphotax’ for the rules that determine how morphemes can be combined in sequence to form words.

References Arppe, Antti, Atticus Harrigan, Katherine Schmirler, Lene Antonsen, Trond Trosterud, Sjur Moshagen, Arok Wolvengrey, Jean Okimâsis, Dorothy Thunder, Jordan Lachler, and Conor Snoek. Under development. Finite-State Transducer-Based Computational Model of

|

268    Schmirler, Arppe, and Genee

Plains Cree Morphology. https://github.com/giellalt/lang-crk/tree/main/src/fst. Beesley, Kenneth R., and Lauri Karttunen. 2003. Finite-State Morphology: Xerox Tools and Techniques. Palo Alto: CSLI Publications. Dunham, Joel. 2014. The Online Linguistic Database: Software for Linguistic Fieldwork. PhD thesis, University of British Columbia. Frantz, Donald G. 1991. Blackfoot Grammar. Toronto: University of Toronto Press. Frantz, Donald G. 2009. Blackfoot Grammar (2nd ed.). Toronto: University of Toronto Press. Frantz, Donald G. 2017. Blackfoot Grammar (3rd ed.). Toronto: University of Toronto Press. Harrigan, Atticus G., Katherine Schmirler, Antti Arppe, Lene Antonsen, Trond Trosterud, and Arok Wolvengrey. 2017. Learning from the Computational Modelling of Plains Cree Verbs. Morphology 27(4):565–598. Hulden, Mans. 2009. Foma: A Finite-State Compiler and Library. Proceedings of the EACL 2009 Demonstrations Session, pp. 29–32. https://aclanthology.org/E09-2008.pdf. Koskenniemi, Kimmo. 1983. Two-Level Morphology: A General Computational Model for WordForm Recognition and Production, Publication No. 11. Department of General Linguistics, University of Helsinki. Lindén, Krister, Erik Axelson, Sam Hardwick, Tommi A. Pirinen, and Miikka Silfverberg. 2011. HFST—Framework for Compiling and Applying Morphologies. Systems and Frameworks for Computational Morphology (second international workshop), ed. by Cerstin Mahlow and Michael Piotrowski, pp. 67–85. Berlin: Springer. Roark, Brian, and Richard W. Sproat. 2007. Computational Approaches to Morphology and Syntax. Oxford: Oxford University Press. Schmirler, Katherine, and Antti Arppe. Under development. Finite-State Transducer-Based Computational Model of Blackfoot Morphology. https://github.com/giellalt/lang-bla/ tree/main/src/fst.

On Diphthongs and Digraphs in Blackfoot Natalie Weber and Mizuki Miyashita

T

his paper is a preliminary study of the realization of Blackfoot digraphs in different phonological contexts, across multiple speakers and dialects.1 We survey descriptions of digraphs—referred to as “diphthongs” in the literature (Frantz 2017)—and provide results of an acoustic study of how four Blackfoot speakers pronounce various digraphs. Both aspects of the paper reveal a large amount of variation in how digraphs are pronounced within and across speakers. We argue that interchanging the terms digraphs and diphthongs is misleading, because digraphs can be realized phonetically either as monophthongs (mid vowels) or diphthongs. Throughout the paper, we use the following definitions for digraphs and diphthongs. A digraph is an orthographic entity, where two symbols are used to represent a single sound. For example, English booth includes two digraphs: for [u], for [θ].2 In Blackfoot, , , and are digraphs. A diphthong is a phonetic entity, which starts with one vowel quality and ends in a different vowel quality. Some English words with diphthongs are [baj] , [bej] , [baw] , and [bɔj] . Phonetic entities that do not change vowel qualities are monophthongs, such as English [bæt] and [bɛt] . The remainder of the paper is laid out as follows. First, we present an overview of the previous research on Blackfoot digraphs and diphthongs, focusing on the phonetic and phonological facts as well as highlighting how these terms came to

|

   269

|

270    Natalie Weber and Mizuki Miyashita

be used interchangeably in Blackfoot linguistics. Second, we show how these terms have led to false claims about Blackfoot vowels in the literature. Finally, we give our preliminary results from the acoustic study, followed by some implications.

Background

We conducted a survey of the previous literature in Blackfoot to determine whether researchers explicitly discussed diphthongs, digraphs, or both. What follows is a brief overview of our findings. In particular, we found that researchers often used the term diphthongs while discussing digraphs. This usage occasionally misleads other researchers, a point that we later return to. On Diphthongs

In Blackfoot descriptions and phonology papers, diphthongs are always treated as a non-contrastive phonetic variant of monophthongs (Frantz 1978, 2017:2–3; Kinsella 1972; Lowery 1979; Taylor 1969; Uhlenbeck 1938). Diphthongs have also been mentioned as part of phonological analyses. For example, Elfner (2006) talks about diphthongization of /o+i/ to [oj] as one vowel hiatus resolution strategy among others. Diphthongs are discussed only sporadically in the literature, and there are no acoustic or phonological studies that focus on diphthongs. This may be due in part because researchers often conflate diphthongs with digraphs, and may not realize that diphthongs could be studied in their own right. On Digraphs

In contrast to Blackfoot diphthongs, which are only rarely studied, researchers frequently refer to orthographic representations (digraphs). We surveyed the previous literature to determine which properties of digraphs the authors focused on. We organized the results into three themes, as shown in (1). (1) Digraphs represent: a. A family of non-contrastive sounds b. Underlying vowel sequences (across morpheme boundaries) c. Underlying phonemic contrasts (morpheme-internally)

|

On Diphthongs and Digraphs in Blackfoot   271

Digraphs: A Family of Non-Contrastive Sounds

Researchers frequently note that a single digraph can be pronounced in many ways. For example, Frantz (2017:2–3, 183) discusses how there is a family of non-contrastive sounds in Blackfoot that are all represented by the same digraphs.3 In some cases, these sounds are positional variants (allophones) that only occur next to certain sounds. In other cases, these sounds are dialectal variants. As shown in (2), before long consonants, is pronounced as [ɛ] (like English said),4 as [ɔ] (like English dawn, for speakers who pronounce this differently than the name Don), and like [i] but with rounded lips. The pronunciation of here is a high, front round vowel, which is commonly written in linguistic sources as [ü], and which we represent in (2) as the IPA symbol [y]. The transcriptions in double square brackets are created by the authors, and are based on Frantz’s (1978, 2017) description of how the digraph is pronounced in these positions, as well as his explanation of how the orthography maps to pronunciation. These transcriptions also match what we have heard from other speakers. (2) Digraph pronunciations before long consonants (Frantz 2017:2) Orthography

Phonetic (IPA)

Translation

a. áínnisiwa

⟦ɛ́nːisiwḁ⟧

‘He descends.’

b. áóttakiwa

⟦ɔ́tːakiwḁ⟧ ‘bartender’

c. nitáakotoissikópii ⟦nitâːkotysːikopi⟧ ‘I will go to rest.’

Before a glottal stop, is pronounced as [ej] (like English paid) in the Káínai dialect, or as [aj] (like English bite) in the Siksiká dialect. The diphthong is pronounced as [aw] (like English out) in all dialects in this position. (3) Digraph pronunciations before glottal stop (Frantz 2017:2–3, 183) Orthography Káínai (IPA)

Siksiká (IPA)

Translation

a. áí’poyiwa

⟦ájʔpojiwḁ⟧

‘He speaks.’

⟦éjʔpojiwḁ⟧

b. ákao’toowa ⟦ákawʔtoːwḁ⟧ = ⟦ákawʔtoːwḁ⟧

‘He has arrived.’

In all other positions, is pronounced as [æː] (like English plaid) in the Káínai dialect, or as [ej] (like English paid) in the Siksiká dialect. The diphthong is pronounced as long [ɔː] (like English dawn) and as [oj] (similar to English coin) in all dialects.

|

272    Natalie Weber and Mizuki Miyashita

(4) Digraph pronunciations in all other environments (Frantz 2017:2–3) Orthography Káínai (IPA)

Siksiká (IPA)

Translation

a. áípotaawa ⟦ǽːpotaːwḁ⟧ ⟦éjpotaːwḁ⟧ ‘airplane’ b. áókska’siwa ⟦ɔ́ ːkskaʔsiwḁ⟧ = c. nohkóíksi

⟦noxʷkójksi̥⟧ =

⟦ɔ́ ːkskaʔsiwḁ⟧ ‘He’s running.’ ⟦noxʷkójksi̥⟧ ‘my sons’

To summarize, digraphs represent multiple pronunciations that vary based on phonological context and dialect. These pronunciations include phonetic monophthongs and diphthongs. The above discussion uses the descriptions of allophones and dialectal variants from Frantz (2017) for heuristic reasons, while other researchers report many other possible pronunciations. The digraph in particular exhibits a lot of variation. For example, Frantz (1978:312) lists [eˑ], [ɛ], and [æˑ] as variants of , and [ɔˑ] and [ao] as variants of . Lowery (1979) notes that “/ay/ varies from [əy] to [ey] to [ɛ], depending on the speaker” (Lowery 1979:75). Taylor (1969:86–88) describes the pronunciation of as [ai] in open syllables and [ɛi] in closed syllables, and as [ɛɛ] unless subsequently shortened. He also describes different phonetic values than Frantz (2017, 1978) for : [au] (if unstressed before /x/), as [ɔ] (before ), or as [ao] otherwise. There is also intra-speaker variation in pronunciation, even when the phonological context is held constant. This can be seen in research that includes narrow phonetic transcriptions, such as Kaneko (1999) and Peter (2014). It is clear that researchers do not agree on the number of surface allophones, nor on their distribution. Digraphs: Underlying Vowel Sequences (Across Morpheme Boundaries)

Digraphs represent underlying vowel sequences across morpheme boundaries (Frantz 1978; Elfner 2006; Kinsella 1972:20; Frantz 2017:186; Taylor 1969:3, 86–93; Weber 2020). Frantz (2017:186) observes that any time that the sequences /a+i/ or /a+o/ were expected based on morphology, [æ] or [ɔ] occur instead.5 He interprets those two vowel sounds as sequences of underlying vowels and he spells them as such, using letters that are independently needed. For example, the transitive animate (TA) final -istoto ‘caus’ begins in a vowel [i], which is apparent when this final occurs after a consonant-final initial, such as sap- ‘correct’, (5a). When this final occurs after a vowel-final initial like sata‘offended’, (5b), the orthography represents the underlying vowel sequence. The

|

On Diphthongs and Digraphs in Blackfoot   273

actual pronunciation of this vowel sequence depends on dialect and phonological context, as discussed in the preceding section; we have represented the digraph as [ǽː] in (5b), matching Frantz’s (2017) description of the Káínai pronunciation.6 (5) Underlying /a+i/ vowel sequences (Weber 2020:241)7 a. ⟦sapístotóːsḁ⟧ b. ⟦satǽːstotoːsḁ⟧

sapístotóósa satáístotoosa

sap-istoto-:s-Ø sata-istoto-:s-Ø

correct-caus.ta-2sg:3.imp-cmd offended-caus.ta-2sg:3.imp-cmd ‘Reach an agreement with him!’ ‘Purposely make her angry!’

The digraph in (5b) reflects the underlying vowel sequences, before vowel coalescence. The same is true for the vowel sequences /a+o/ and /o+i/. (See Weber 2020 for examples.) This creates morphological transparency in the orthography, but masks the true pronunciation. Many researchers claim that these vowels only occur across morpheme boundaries. Taylor notes that “the vowels /ɛ/, /ɔ/, and /ő/ [e.g., our [y]—NW&MM], in non-borrowed words, are always the result of vowel contractions of one kind or another” (Taylor 1969:3). Kinsella notes that “/ɛ(ɛ)/ and /ɔ(ɔ)/ are always the result of contractions of //a(a) + I(I)// or //a(a) + i(i)//; and //a(a) + o(o)// respectively” (Kinsella 1972:20). Next, we show that this is not always true. Digraphs: Underlying Phonemic Contrasts (Morpheme-Internally)

As argued in Weber (2020), digraphs may represent underlying phonemic contrasts. Although most instances of [æː] and [ɔː] occur across a morpheme boundary, some instances of [æː] and [ɔː] are tautomorphemic (morpheme-internal).8 These vowels are not predictable elements of the word, and represent abstract units of contrast. There is even a minimal pair for [ɔː] vs. [aː], (6). (6) Underlying tautomorphemic in contrast with (Weber 2020:42) a. [ɔːníːt]

b. [aːníːt]

aoníít aaníít ao-nii-t-Ø aan-ii-t-Ø hole-by.needle.ti-2sg.imp-cmd say-ai- 2sg.imp-cmd ‘Pierce it!’ (BB) ‘Say (s.t.)!’ (BB)

|

274    Natalie Weber and Mizuki Miyashita

Although there are no minimal pairs for tautomorphemic [æː] and other long vowels, there are examples that show that [æː] occurs in overlapping environments with other long vowels. For example, [æː], [aː], and [oː] can all occur in the environment [#s__ta . . . ], as shown in (7). We give a four-line gloss for each example so that it is clear that the [æː] is tautomorphemic. (7) Tautomorphemic [ɛː], [aː], and [oː] in same environment (Weber 2020:42) a. %⟦sæːtamit⟧

saitamit

sait-ami-t-Ø

breathe-be.ai-2sg.imp-cmd ‘Breathe!’ (Piikáni dialect) b. ⟦saːtápiksit⟧

saatápiksit

saat-apiksi-t-Ø across-throw.ti-2sg.imp-cmd ‘Throw it out!’ c. ⟦sóːtamanistaːwḁ⟧ sóótamanistaawa

sootam-an-ist-aa-Ø-wa despite-tell-ta-3obj-ind-3 ‘I told her in spite of all else!’

The examples in (6) and (7) show that [æː] and [ɔː] contrast with other long vowels in Blackfoot. Since [æː] and [ɔː] each exist in a family of non-contrastive sounds, it is reasonable to ask what the abstract phonemic representation of these sounds could be. The back vowel is simpler. Recall that Frantz (2017) states that is pronounced like [ɔ] before long consonants, like [aw] before a glottal stop, and [ɔː] elsewhere. Since [ɔː] is the allophone that occurs in the widest set of environments, we take /ɔː/ to be the phonemic representation of this sound, with positional allophones [ɔ] and [aw]. The front vowel has dialectal variants, summarized from Frantz (2017:2–3) in (8).

|

On Diphthongs and Digraphs in Blackfoot   275

(8) Variants of

Káínai

Siksiká

Before long consonants [ɛ]

[ɛ]

Before glottal stop

[ej]

[aj]

Elsewhere

[æː] [ej]

Here, if the underlying phoneme matched the elsewhere allophone it would be /æː/ in Káínai and /ej/ in Siksiká. However, this would mean that /ej/ is the only phonemic diphthong in the language, and that it only exists in one dialect. For this reason, we assume that the underlying phoneme is a monophthong, which we represent as /ɛː/ to create a symmetrical vowel system. The symbol [ɛ(ː)] is also one of the most frequently used symbols for this vowel by other researchers (Denzer-King 2009; Derrick 2007; Elfner 2006; Frantz 1978; Kaneko 1999; Kinsella 1972; Peter 2014; Stacy 2004; Taylor 1969; Uhlenbeck 1938; Van Der Mark 2003; Weber 2020; Windsor 2017). It also is consistent with the Káínai pronunciation. In the corpus of texts we used for our study below, the pronunciation of ranges from [ɛː] to [æː] for the Káínai speaker. This changes the vowel inventory, which must have five long vowels, /iː ɛː aː ɔː oː/ and three short vowels /i a o/, as shown in Table 1. To summarize, the correspondence of digraphs to phonetic and phonological representations is complex. A digraph can represent a phonetic monophthong or diphthong. Digraphs are also used to represent (underived, contrastive) tautomorphemic long mid vowels, as well as (derived, predictable) underlying vowel sequences. These relations are summarized in (9). (9) Orthography

Phonetic pronunciation



Abstract representation Tautomorphemic

Derived

a.

[ɛ] ~ [æː] ~ [ej] ~ [aj]

/ɛː/

/a+i/

b.

[ɔ] ~ [ɔː] ~ [aw]

/ɔː/

/a+o/

c.

[y] ~ [oj]

TABLE 1. Blackfoot vowel phoneme inventory FRONT HIGH MID LOW

i iː ɛː

CENTRAL

a aː

BACK

o oː ɔː

/o+i/

|

276    Natalie Weber and Mizuki Miyashita

As shown from this discussion, digraphs do not correspond with a unique phonetic realization, nor do they necessarily correspond to a precise phonemic representation. Unfortunately, when researchers mention digraphs, they do not always clearly distinguish the three properties in (1). This can lead to confusion in the literature, as we discuss next. On the Conflation of Terms

While we distinguished digraphs from diphthongs in the above discussion, many researchers use “diphthong” instead of “digraphs,” following the terminology used in the Blackfoot Grammar (Frantz 2017). The first few chapters of Frantz (2017) were written for those who have no linguistics training (Miyashita and Many Bears 2018). The sound inventory is introduced by referring to orthographic symbols rather than the actual sounds. The section in which , , and are described with their variations is called Diphthongs. It is clear that the section describes digraphs rather than diphthongs, because some pronunciations are described as monophthongs and others as diphthongs, as we discussed in a preceding section. Two earlier grammars, Tims (1889:1) and Uhlenbeck (1938:1–4), also refer to digraphs as diphthongs, even when the description of these sounds includes a mix of monophthongs and diphthongs. This leads to a situation where “Blackfoot literature refers to both true and orthographically represented diphthongs” (Stacy 2004:9). Conflating these two terms has two effects. First, the descriptions in reference materials have misled researchers in several ways. We describe these in the next section. The implication is that one must read previous research carefully to determine which of the above properties are being discussed. Second, we suspect that the conflation of “diphthongs” and “digraphs” leads many to suspect that diphthongs are rare, or only occur as positional allophones in well-defined contexts. Specifically, the variation in how digraphs are pronounced has not been studied acoustically. We address this gap with our study below.

Confusion in the Literature

As the most recent reference grammar, Frantz (2017) is often the first resource that linguists access to learn more about some particular aspect of Blackfoot

|

On Diphthongs and Digraphs in Blackfoot   277

grammar. However, the Diphthongs section title may mislead both learners and researchers in several ways. Although some of these studies include phonetic analyses of the language (e.g., Van Der Mark 2003 on pitch accent, and Derrick 2007 on assibilants), none focus on diphthongs. Even in these acoustic studies, the assumptions about the Blackfoot vowel inventory rely on the descriptive literature. In this section we discuss these by type, along with quotes illustrating the confusion. Digraphs may be confused with phonetic diphthongs. This can happen whenever the researcher takes the section title at face value. n

Frantz (1997) observes three diphthongs in Blackfoot: [ai], [ao] and [oi]” (Peterson 2004:4).

Digraphs may be confused with phonemes. Because Frantz’s (2017) section is embedded in a chapter that describes the contrastive sounds of Blackfoot, it is easy to mistake these digraphs as phonemic diphthongs. Indeed, digraphs can represent phonemic contrasts when they are tautomorphemic, although the abstract phoneme is better represented as a monophthong. However, none of Frantz’s examples in this section include tautomorphemic digraphs. n

n

n

n

“The phones [eˑ], [ɛ], and [æˑ] are variants of the diphthong /ai/”, and long [ɔˑ], short [ɔ], and [ao] are variants of /ao/ (Frantz 1978:312). “/ay/ varies from [əy] to [ey] to [ɛ], depending on the speaker” (Lowery 1979:75). The phonological inventory in Kaneko (1999:13–14) includes three phonemic short vowels, three phonemic long vowels, and three phonemic diphthongs, and she states that “three kinds of diphthongs /ai/, /ao/, and /oi/ are also found in the language” (Kaneko 1999:15). Van Der Mark (2003) claims that the Siksiká dialect has five contrastive vowel qualities, with long and short counterparts: /i iː ɛ ɛː a aː ɔ ɔː u uː/, and that “Blackfoot also has the diphthong, /oj/” (Van Der Mark 2003:8). She notes that the mid vowels are spelled as and (Van Der Mark 2003), showing that she takes the three digraphs to be contrastive segments.

|

278    Natalie Weber and Mizuki Miyashita

n

n

n

“Blackfoot’s only true diphthong is /ɔj/” (Stacy 2004:9–10). “Blackfoot also has long diphthongs, and the mid front vowel /ɛ/ is a reduction of the long diphthong /ai/” (Derrick 2007:1). Denzer-King (2009:17) includes a phonemic inventory with three monophthongal vowel qualities and three phonemic diphthongs, /ai/, /ao/, and /oi/.

Researchers may underestimate the extent of phonetic variation for digraphs. This is because the orthography has the effect of “masking” allophonic and free variation. (See Gambarage 2017 for a similar discussion of the effects of Swahili orthography.) The following authors do not mention a diphthongal allophone of and at all. n

n

n

“Orthographic diphthongs include ai, and ao realized as long monophthongs, /ɛ:/ or /e:/, and /ɔ:/, respectively” (Stacy 2004:9–10). “The vowels [eː], [ɛː] and [æː], which are always long, are the surface outputs of underlying sequences of /a+i/, while [ɛ] is the realisation of this sequence when it is followed by a coda consonant. Similarly, [ɔː] is the surface realisation of /ao/, which is shortened to [ɔ] in closed syllables” (Elfner 2006:14). “Blackfoot also has long diphthongs, and the mid front vowel /ɛ/ is a reduction of the long diphthong /ai/” (Derrick 2007:1).

In the next section we describe a preliminary acoustic study of digraph pronunciations.

Preliminary Corpus Study

Our research goal is to study the variation between monophthongs and diphthongs within and across speakers of Blackfoot. To address these questions, we used connected speech in recorded narratives as our corpus of speech data for the

|

On Diphthongs and Digraphs in Blackfoot   279

following reasons. First, diphthongs often occur at morpheme boundaries and connected speech would provide many examples of polymorphemic words. Second, connected speech is more likely to be produced fluently and naturally, in contrast to a controlled but less natural wordlist task. Third, traveling was not an option because of the COVID-19 pandemic, and online recording can include extralinguistic artifacts (Sanker et al. 2021). Thus, we used recordings that were available to us as the data source for this study. Corpus

The four recorded narratives used in the study are summarized in Table 2. The first two (a, b) were from the Glenbow Museum’s website (Brown 2002; Krmpotich 2005). Both are told by Mr. Leo Fox (Káínai dialect). The other two (c, d) are stories recorded by the second author (Miyashita). The speaker of recording (c) is Ms. Shirlee Crow Shoe (Aapátohsipiikani dialect; recorded in 2009). The speaker of the recording (d) is the late Chief Earl Old Person (Aamsskáápipiikani dialect; recorded in 2012). The recordings (c, d) were stories told by the speakers while being shown scenes from a picture-only book, “One Frog Too Many” (Mayer and Mayer 1975). The Siksiká dialect is not represented in this corpus, which means that three of the four major dialects are represented. Hypotheses

For the current study, we developed two hypotheses, shown in (10), based on the description and claims found in Frantz (2017). (10) Specific Hypotheses 1. Digraphs before glottal stops are pronounced as diphthongs (Frantz 2017:2–3). 2. Digraphs before long consonants are pronounced as short-lax vowels (Frantz 2017:2–3; 1978:312; Taylor 1969:86–88). Methods

Vowels can be defined acoustically by their first, second, and third formant values (F1, F2, and F3). These are bands of resonating frequencies in the sound wave, and each vowel quality has a unique set of formants. Over the course of a monophthong,

|

280    Natalie Weber and Mizuki Miyashita

TABLE 2. Recordings of connected speech (narrative) (age 50s–80s) SPEAKERS

GENDER DIALECT

STORIES

a. Leo Fox

M

Káínai

b. Leo Fox

M

Káínai

Blood Clot Naapi ki siikokiinis

F

Aapátohsipikani

M

Aamsskáápipikani

c. Shirlee Crow Shoe d. Earl Old Person

DURATION RECORDING (MM:SS)

2:27

Glenbow M.

1:26

Glenbow M.

Friends

8:42

Rec. 2009

One Frog Too Many

7:52

Rec. 2012

formant values remain relatively stable. Over the course of a diphthong, formant values will change from the frequencies associated with the first vowel quality to the frequencies associated with the second vowel quality. We focused on the orthographic digraphs , , and . We hand-segmented the vocoids corresponding to each digraph in a textgrid in Praat (Boersma and Weenink 2019). We then used a Praat script to extract measurements of formants F1, F2, and F3 at 20% and 80% of the vocoid’s duration. This allowed us to approximate the vocoids formant trajectories over time while avoiding transitions out of and into the surrounding consonants. The 20% and 80% marks are frequently used in acoustic studies of diphthongs (e.g., Elvin et al. 2016). Praat uses the Burg method of analysis for formant estimation, and we used the default settings, except that we used a time step of 0.005s for all speakers. For Earl Old Person we searched for six formants with a maximum frequency of 5500 Hz, and for Leo Fox and Shirlee Crow Shoe we searched for five formants with a maximum frequency of 5000 Hz. These settings did reasonably well at tracking F1 and F2. We spot checked any tokens that had an unexpectedly small or large change in F1 or F2, and manually corrected five tokens where the formants had not been correctly tracked. Praat did not always reliably track F3, so we focus here on F1 and F2. Diphthongs involve a marked articulatory and acoustic transition (Gay 1970). To approximate the magnitude of acoustic transition for each vowel token, we used Euclidean distance. The formula is given in (11). This formula takes the square root of the change in F1 squared added to the change in F2 squared over the vowel’s duration. Therefore, a larger number means a longer trajectory and is more likely to correspond to a diphthong. A smaller number means that the formant value did not change much and is more likely to correspond to a monophthong.

|

On Diphthongs and Digraphs in Blackfoot   281

(11) √((F2b – F2a)2+(F1b – F1a)2), where F1a and F2a are the formant measurements at 20% of the vowel duration, and F1b and F2b are the formant measurements at 80% of the vowel duration

To test our two hypotheses, we annotated each token with the immediately following context. Finally, we categorized each token as either a monophthong or a diphthong in the following way. We found that for each digraph for each speaker, the Euclidean distances clustered around a range of frequencies, and a few outliers had very large Euclidean distances. We marked the outliers as diphthongs and the rest as monophthongs. We checked a few tokens on either side of this divide to confirm that they sounded like either monophthongs or diphthongs, and that the formants in the spectrograms either did not or did have perceptible slope. The exact cutoff point differed by speaker and digraph. Although crude, this method reliably found the division between monophthongs and diphthongs. Results

There were a total of 149 tokens, which are shown by speaker and digraph in Table 3. The digraph is most frequent and the digraph is very infrequent. Table 4 shows the number of diphthongal and monophthongal tokens for each digraph, along with their variance in terms of Euclidean distance. Because the number of tokens is so small, variance is represented simply with the mean and standard deviation. There are several things to note. First, the Euclidean distance for diphthongs is much greater than that for monophthongs. Second, speakers use different mixes of strategies for each token. For example, Leo Fox only has diphthongal for two of thirty-one tokens, but has diphthongal for nine of eighteen tokens. Third, no speakers in this corpus had a monophthongal realization of /o+i/. We now turn to our hypotheses. The first hypothesis was that digraphs before glottal stops are pronounced as diphthongs. This is false. All speakers use a mix of diphthongs and monophthongs before glottal stops (Table 5). Results are pooled across all three digraphs. The second hypothesis was that digraphs before long consonants are pronounced as short lax vowels. This is tentatively true for but not for , which was always diphthongal (Table 6). However, there were not many tokens of digraphs before long consonants in the corpus at all.

|

282    Natalie Weber and Mizuki Miyashita

TABLE 3. Total number of total by speaker and digraph SPEAKER





TOTAL

Leo Fox Shirlee Crow Shoe Earl Old Person

31

18

5

54

39

13

2

54

29 107

11 42

1 8

41 149

TOTAL

TABLE 4. Variance in Euclidean distance for diphthongs and monophthongs DIPHTHONGS

MONOPHTHONGS

SPEAKER

DIGRAPH

COUNT

MEAN (HZ)

SD (HZ)

COUNT

MEAN (HZ)

Leo Fox

2

574.095

192.761

29

160.017 101.728

Shirlee Crow Shoe

7

498.608

132.946

32

187.872 76.107

Earl Old Person

2

622.028

165.379

27

187.464 98.310

Leo Fox

9

464.306

171.471

9

105.923 73.314

Shirlee Crow Shoe

2

292.485

47.778

11

129.773 52.457

Earl Old Person

7

308.673

43.163

4

127.511 69.109

Leo Fox

5

729.628

308.014

0

---

---

Shirlee Crow Shoe

2

1197.663

5.068

0

---

---

Earl Old Person

1

566.015

n/a

0

---

---

TABLE 5. Realization of digraphs before glottal stops

Leo Fox Shirlee Crow Shoe Earl Old Person

TOTAL

DIPHTHONG

MONOPHTHONG

17 16 7

5 6 4

12 10 3

SD (HZ)

|

On Diphthongs and Digraphs in Blackfoot   283

TABLE 6. Realization of digraphs before long consonants TOTAL DIPHTHONG

Leo Fox Shirlee Crow Shoe Earl Old Person

5 7 0

MONOPHTHONG

4 (all ) 1 () 0 7 (all ) -----

Outside of those two contexts, we found that all three speakers use a mix of diphthongs and monophthongs, contra Frantz’s (2017) description. If these are positional variants, then they are not conditioned by the following context. As shown in Table 4, the choice between monophthongs and diphthongs is partly determined by the choice of digraph: is mostly monophthongs, is mixed, and is always a diphthong.9 Our point is that the distribution of monophthongs and diphthongs is more complicated than what is described in the literature. Additionally, we noticed that phonetic diphthongs also correspond to many other orthographic sequences besides digraphs. Example (12) is from Naapi ki siikokiinis by Leo Fox, and (13) is from Friends by Shirlee Crow Shoe. This suggests that phonetic diphthongs are more widespread in Blackfoot than previously thought. Future work must take this into account. (12) [inɔ́wʔksêjʔsːkə́ mːiçkḁ]

(13) [s:kaiʔsawaxsim:i:waj]

iinao’k-saw-sskaa-mm-iihk sska’-isaw-yaahsimm-ii-w=ayi however-neg-break.ai-ind-report shock-neg-feel.pleased.ta-dir-3=3obv ‘but it did not break’

‘he is not very pleased’

Summary

To summarize, Blackfoot diphthongs and digraphs are distinct concepts, but are often confused in the Blackfoot literature. Disconnecting the terms diphthongs and digraphs may lead to new avenues of research, such as studying diphthongs in their own right. It may also aid language teaching and learning. It is clear that the orthography might not tell a learner all the information they need about how words are pronounced. Each digraph can be pronounced in a number of ways, depending

|

284    Natalie Weber and Mizuki Miyashita

on dialect and phonological context. A larger-scale acoustic study may lead to generalizations which language teachers can develop into pronunciation teaching materials when introducing words containing diphthongs or written with digraphs. Our preliminary corpus study results show that (a) all three speakers in the corpus use a mix of phonetic monophthongs and diphthongs, (b) phonological context affects pronunciation but not exactly as described in Frantz (2017), and (c) some variation remains unexplained. Some phonetic diphthongs occurred for other orthographic sequences that are not digraphs, suggesting that diphthongs are more widespread in Blackfoot than previously believed. Future acoustic studies should focus on all phonetic diphthongs, no matter how they are spelled. Notes 1. We would like to acknowledge the Blackfoot speakers whose stories were used in the corpus study: Leo Fox (Káínai; Blood) and the Glenbow Museum, Shirlee Crow Shoe (Aapátohsipikani; Northern Piegan), the late Earl Old Person (Aamsskáápipikani; Blackfeet or Southern Peigan), and the late Beatrice Bullshields (Káínai; Blood). We also thank Rosella Many Bears (Káínai; Blood) who helped to transcribe and translate some of these stories. This project was partially supported by grants awarded to Mizuki Miyashita by the American Philosophical Society and Humanities Montana. Some data processing was completed by Samantha Prins (spring 2019) and Kathryn Boyle (spring 2021), student research assistants at the University of Montana. Finally, this paper is completed in memoriam of Donald G. Frantz and Earl Old Person, both of whom passed away shortly before the paper was presented at the 53rd Algonquian Conference in fall 2021. 2. The following notations are used in this paper: square brackets, [ ], for phonetic transcription; slashes, / /, for phonemic transcription; angled brackets, < >, for orthographic transcription; and doubled square brackets, ⟦ ⟧, for phonetic transcriptions created from orthography. The symbol % marks a dialectal variant.

3. Frantz (2017) refers to these as diphthongs rather than digraphs. Throughout this section, we use the terms diphthong and digraph as defined in our introduction, even when this mismatches the terminology used in original sources. 4. In this section the parenthetical notes are Frantz’s description. The IPA symbols are our interpretation. 5. Vowel hiatus may be resolved in different ways at different points within the verbal complex. Windsor (2017) notes that coalescence is obligatory, optional, or prohibited at different morphological boundaries. Diphthongization is taken as phonological evidence

|

On Diphthongs and Digraphs in Blackfoot   285

by Bliss (2013:150–152) that the demonstrative is a separate prosodic constituent (a “word”) from the noun in Blackfoot. In a detailed case study, Peter (2014) looks at the hiatus resolution patterns of seven prefixes in Blackfoot. Six of these prefixes end in a glide, while the seventh ends in a vowel, and the prefixes were elicted before morphemes beginning in the vowels a, ai, i, o, or ao. Peter was trying to confirm the description of semivowel alternation in Frantz (2017:88), but instead he found a wide variety of hiatus resolution patterns. 6. Abbreviations include: 2 = second person, 3 = third person, ai = animate intransitive, caus = causative, cmd = command, dir = direct, imp = imperative, ind = indicative, neg = negation, obj = object, obv = obviative, report = reportative, sg = singular, ta = transitive animate, ti = transitive inanimate. Examples spoken by Beatrice Bullshields are marked with “(BB)”. 7. We have made slight changes in the morpheme gloss line for examples cited from Weber (2020) throughout this paper. 8. The sequence [oi] always occurs across a morpheme boundary. That is, it never contrasts with other vowels. 9. We do not expect these patterns to hold over a larger corpus. The digraph is probably usually a monophthong because of the dialects sampled here. According to Frantz (2017), the Siksiká dialect frequently pronounces as a diphthong. The digraph does not occur before long consonants in our corpus, which is the only position where it is described as a monophthong. In addition, also had the smallest sample size and may have much more variation in pronunciation than what is represented in our corpus.

References Bliss, Heather. 2013. The Blackfoot Configurationality Conspiracy: Parallels and Differences in Clausal and Nominal Structures. PhD thesis, University of British Columbia. Boersma, Paul, and David Weenink. 2019. Praat: Doing Phonetics by Computer [Computer program]. Version 6.0.50. http://www.praat.org/. Brown, Alison. 2002. Nitsitapiisinni: Our Way of Life: The Blackfoot Gallery, Glenbow Museum Calgary, Alberta, Canada. Museum Anthropology 25(2):69–75. Denzer-King, Ryan Edward. 2009. The Distribution of /s/ in Blackfoot: An Optimality Theory Account. MA thesis, University of Montana. Derrick, Donald. 2007. Syllabification and Blackfoot /s/. Proceedings of the Northwest Linguistics Conference (NWLC) 22 Burnaby, BC, Canada, February 18–19, ed. by Nicole

|

286    Natalie Weber and Mizuki Miyashita

Carter et al., pp. 62–76 (Simon Fraser University Working Papers in Linguistics 1). Vancouver, BC: Department of Linguistics, Simon Fraser University. Elfner, Emily. 2006. The Mora in Blackfoot. MA thesis, University of Calgary. Elvin, Jaydene, Daniel Williams, and Paola Escudero. 2016. Dynamic Acoustic Properties of Monophthongs and Diphthongs in Western Sydney Australian English. The Journal of the Acoustical Society of America 140(1):576–581. Frantz, Donald G. 1971. Toward a Generative Grammar of Blackfoot (with Particular Attention to Selected Stem Formation Processes). Norman: SIL, University of Oklahoma. Frantz, Donald G. 1997. Blackfoot Grammar. 1st ed., reprinted. Toronto: University of Toronto Press. Frantz, Donald G. 2017. Blackfoot Grammar. 3rd ed. Toronto: University of Toronto Press. Gambarage, Joash J. 2017. Unmasking the Bantu Orthographic Vowels. Africa’s Endangered Languages: Documentary and Theoretical Approaches, ed. by Jason Kandybowicz and Harold Torrence, pp. 449–484. Oxford: Oxford University Press. Gay, Thomas. 1970. A Perceptual Study of American English Diphthongs. Language and Speech 13(2):65–88. Kaneko, Ikuyo. 1999. A Metrical Analysis of Blackfoot Nominal Accent in Optimality Theory. MA thesis, University of British Columbia. Kinsella, Norman George. 1972. The Vowel System of Blackfoot. MA thesis, University of Calgary. Krmpotich, Cara, and David Anderson. 2005. Collaborative Exhibitions and Visitor Reactions: The Case of Nitsitapiisinni: Our Way of Life. Curator: The Museum Journal 48(4):377–405. Lowery, Brenda. 1979. The Phonological System of Blackfoot. Contributions to Canadian Linguistics, pp. 67–91. (National Museum of Man, Mercury Series, Canadian Ethnology Service Paper no. 50.) Ottawa: University of Ottawa Press. Mayer, Mercer, and Marianna Mayer. 1975. One Frog Too Many. New York: Dial Books. Miyashita, Mizuki, and Rosella Many Bears. 2018. Blackfoot Grammar. International Journal of American Linguistics 84(4):583–587. Peter, Naoki. 2014. Hiatus Resolution in Blackfoot. Masterarbeit, Universität Bern. Peterson, Tyler. 2004. Theoretical Issues in the Representation of the Glottal Stop in Blackfoot. Proceedings from the 7th Workshop on American Indigenous Languages [WAIL 7], ed. by Lea Harper and Carmen Jany, pp. 106–121. (Santa Barbara Papers in Linguistics vol. 15). Santa Barbara: Department of Linguistics, University of California Santa Barbara. Sanker, Chelsea, Sarah Babinski, Roslyn Burns, Marisha Evans, Jeremy Johns, Juhyae Kim, Slater Smith, Natalie Weber, and Claire Bowern. 2021. (Don’t) Try This at Home! The Effects of Recording Devices and Software on Phonetic Analysis. Language

|

On Diphthongs and Digraphs in Blackfoot   287

97(4):e360–e382. Stacy, Elizabeth. 2004. Phonological Aspects of Blackfoot Prominence. MA thesis, University of Calgary. Taylor, Allan. 1969. A Grammar of Blackfoot. Berkeley: PhD thesis, UC Berkeley. Tims, John W. 1889. Grammar and Dictionary of the Blackfoot Language. London: Society for Promoting Christian Knowledge. Uhlenbeck, Christianus Cornelius. 1938. A Concise Blackfoot Grammar Based on Material from the Southern Peigans. Amsterdam: Noord-Hollandsche Uitgevers-Maatschappij. Van Der Mark, Sheena C. 2003. The Phonetics of Blackfoot Pitch Accent. MA thesis, University of Calgary. Weber, Natalie. 2020. Syntax, Prosody, and Metrical Structure in Blackfoot. PhD thesis, University of British Columbia. Windsor, Joseph W. 2017. Predicting Prosodic Structure by Morphosyntactic Category: A Case Study of Blackfoot. Glossa: A Journal of General Linguistics 2(1):Art. 10. 1–17.

Rhythm and Intonation in Oji-Cree Matthew Windsor

E

ven to the casual listener, Oji-Cree sounds quite distinct from other Ojibwe languages.1 This is in part due to well-known segmental differences; for example, the fortis consonants of Ojibwe correspond to h + consonant clusters in Oji-Cree. But the distinct character of Oji-Cree is also owed to differences in its most salient prosodic features: word rhythm and phrasal intonation. The first goal of this paper is to sketch the essential patterns of Oji-Cree word rhythm (non-primary stress) and briefly relate them to the overall grammatical system as well as to other documented varieties of Ojibwe. The second goal is to characterize the basic intonation patterns of Oji-Cree and their relation to word rhythm. One type of pattern places pitch prominence on a rhythmically strong syllable, while another type operates independently of word rhythm. The resulting prosodic system displays several typologically interesting features including a lack of cumulative word stress and intonation patterns that operate unanchored to word-level rhythm (Himmelmann and Ladd 2008).

Background

Oji-Cree is spoken in thirteen communities of northwestern Ontario, at the northern extremity of the Ojibwe language complex. Varieties of Ojibwe can be divided into

|

   289

|

290    Matthew Windsor

three dialectal tiers: northern, central, and southern (Valentine 1994). Oji-Cree is the most divergent variety of the northern tier, having developed significant innovations through relative isolation from other dialects as well as through prolonged contact with Swampy Cree. The influence of Swampy Cree (and to a lesser extent Plains Cree) is observable in the phonology, morphosyntax, and vocabulary of the language. Oji-Cree can be itself divided into three dialects, which differ more superficially: a western dialect within the Severn River basin, a southern dialect located at the headwaters of the Severn River, and an eastern dialect located along the Winisk River system (Valentine 1994; Nichols 2014). This study was undertaken in Kingfisher Lake, of the Winisk River dialect. Most speakers in the community are bilingual in English with the youngest active speakers in their twenties. A small group of elders remain monolingual speakers of Oji-Cree. Rhythm across Algonquian and Ojibwe

Most Central Algonquian languages have been described as having primarily iambic rhythm, including Menominee (Bloomfield 1962:19–20; Milligan 2005), Potawatomi (Lockwood 2017:25), Ojibwe (Valentine 1994), Miami-Illinois (Costa 2003:98) and East Cree (Brittain 2000:189). Most of these also display some subset of idiosyncratic vocabulary. Eastern Algonquian languages are split, with Passamaquoddy (LeSourd 1993) and Mi’gmaq (Olson 2012) described with a primarily trochaic pattern, and other languages such as Munsee Delaware (Goddard 1982:19) as iambic or ambiguous (see also Milligan 2005 and Oxford 2020 for summaries of rhythm types within the family). On the western periphery of Algonquian, the Plains languages Arapaho, Cheyenne, and Blackfoot have been noted for their complex prosodic systems, which feature the independent operation of lexical stress and tone, the use of vowel length and stress to distinguish phonemes, and a lack of culminative accent (Bogomolets in press). Within Ojibwe, a useful point of departure is Valentine’s (1994, 2001) summary of stress patterns in the Eastern and Odawa dialects, with reference to metrical stress theory as developed in Hayes (1984, 1995). In Eastern Ojibwe, syllables are organized into iambic (weak-strong) feet starting from the beginning of the phonological word, including person prefixes but not preverbs. The system is quantity-sensitive: syllables containing a long vowel (aa, ii, oo, e) are “heavy” (intrinsically strong).

|

Rhythm and Intonation in Oji-Cree   291

The final syllable of a word is also always intrinsically strong (“extrametrical” in Hayes’s terms). A few examples from Eastern Ojibwe are given in (1), with parentheses indicating foot structure, and underlining indicating metrical stress placement. Final strong syllables appear in angle brackets. (1) Eastern Ojibwe metrical structure (Valentine 2001:52–58) a. (dago) ‘s/he arrives’ b. (gida)go ‘you arrive’ c. (maki) ‘shoe’ d. (nima)ki ‘my shoe’ e. (maki)zi ‘shoes’ f. (e)(siba) ‘raccoons’ g. (niwaa)(bami)go ‘s/he sees us’ h. bi-(niga) ‘come sing!’

The system allows for two consecutive strong syllables (1a,f), termed a rhythmic clash. Person prefixes (ni-, gi-) are included in the domain of rhythm (1b,d), though preverbs form a separate domain and do not affect the metrical structure of the stem (1h bi- ‘come’). With Eastern Ojibwe for reference, Valentine (1994) then provides a sketch of rhythmic variation across all documented varieties of Ojibwe based on previous literature and his own fieldwork. One outlier includes Odawa, which deletes certain weak vowels, creating a system of vowel syncopation. Northwestern Ojibwe (of the central dialect tier) also stands out in that rhythmic prominence is associated with pitch, and in the avoidance of rhythmic clashes. The resulting pattern of alternating pitch applies across strings of preverbs, which cohere to form a single phonological word with noun and verb stems. Last, Swierzbin (2003) represents the only attempt to measure acoustic correlates of stress in any dialect of Ojibwe. She found vowel duration to be the primary indicator of rhythmic stress, paralleling my own observations for Oji-Cree. Rhythm in Oji-Cree

Valentine (1994) further notes some divergent aspects of rhythm in Oji-Cree, observing that many words composed entirely of light syllables begin with initial

|

292    Matthew Windsor

stress on the first syllable, rather than on the second as expected. I repeat a few of his examples in (2) using the northern roman and syllabic orthographies standard to Oji-Cree:2 (2) Tri-syllabic Oji-Cree words with rhythmic “shift” (adapted from Valentine 1994:155) a. ᒪᐦᑭᓯᐣ

mahkisin ‘shoe’

b. ᒥᓂᑯᐠ

minikohk ‘enough’

c. ᒥᑭᓀ

mikine

d. ᒥᐣᑭᓯ

minkisi ‘eagle’

e. ᒪᓂᒎᓑ

manicoohsh ‘insect, spider’

f. ᒥᓂᑎᐠ

‘s/he barks’

minitik ‘island’

g. ᒪᑐᑳᐧᐣ

matokwaan ‘traditional tipi’

h. ᐱᒧᐦᓭ

pimohse

‘s/he walks along’

i. ᐱᒥᐸᐦᑑ pimipahtoo ‘he runs along’

Examples with no “stress shift” in his list include nikamo ‘s/he sings’, kahkina ‘all’, and awiya ‘someone’.3 Valentine also notes a shift in strings of light syllables within some longer words such as inahkwashiwe ‘he paddles there’, and in words composed of two light syllables. It seems the shift is particularly visible in how it affects the deletion of weak vowels between homorganic consonants. For example, Algonquin (an Ojibwe variety of the northern tier) and Oji-Cree delete vowels in different locations within cognate words. This is presumably because the weak vowels in Oji-Cree occupy different positions. While the correspondences are not all straightforward, (3) shows a tidy example for illustration. Underscores indicate the location of deleted /i/. (3) Short vowel deletion in Algonquin vs. Oji-Cree ‘s/he runs along’ a. Algonquin: m_mibatoo (with assimilation of initial b>m; p. 152) b. Oji-Cree: pim_pahtoo (c.f. p. 154)

Valentine provisionally suggests that these rhythmic shifts in Oji-Cree may be caused when stress is attracted to initial syllables with labial onsets, with a deficient foot being created at the beginning of the word (pi)(mipa). In what follows I will illustrate a different analysis using a grid structure. Although labial onsets may be associated with stress to some degree, the stress-initial pattern also applies to a significant set of words lacking labial onsets:

|

Rhythm and Intonation in Oji-Cree   293

(4) Initial prominence in Oji-Cree words lacking labial onsets a. ᐅᐣᑎᓯᐃᐧᐣ ontisiwin ‘payment’ b. ᑯᓯᑲᐧᐣ c. ᑕᑯᑯᓯ

d. ᑕᑯᐦᔑᐣ

kosikwan

‘it is heavy’

takokosi

‘s/he arrives (at a new camp/residence)’

takohshin ‘s/he arrives’

I also regard this rhythmic shift as sufficiently pervasive to describe Oji-Cree as having primarily trochaic rather than iambic foot structure. Primary Stress

All discussion thus far has concerned non-primary stress, which I will continue to refer to as rhythmic stress to better distinguish it as a separate phenomenon (Van der Hulst 2014a, 2014b). Most Algonquian languages have received descriptions of primary word stress (see Milligan 2005 for a summary), though in several cases there is a lack of consensus. Plains Cree, for example, has received multiple analyses of primary stress which disagree on its nature and placement (see Muehlbauer 2006 for discussion), though none to my knowledge take up Wolfart’s (1996:431) call for the need to explore stress patterns in relation to sentence-level intonation. In her study of Menominee vowels, Cudworth (2019:252) similarly notes that stress analysis for that language “would benefit from a closer analysis of the vowel or syllable position within a phrase or connected speech” (252). Voorhis’s (1974) description of Kickapoo describes the domain of primary stress as the “sentence,” with examples illustrating single-word versus multi-word phrases (14–16). This distinction between word-level stress and phrase-level prominence is increasingly recognized as important in the field of prosodic typology (Gordon 2014; Van der Hulst 2014b). Turning to Ojibwe, Bloomfield’s (1958) classic description does not specify primary stress assignment. Valentine (1994:156–157) expresses doubt concerning previous studies that provided a neat analysis of primary word stress for Eastern Ojibwe, though in his grammar he does supply Kaye’s (1973) analysis as a rule of thumb: main stress falls on the leftmost strong syllable from the end of the word within a window of three feet. Regarding Oji-Cree, Valentine describes primary stress as most often falling on the pre-final foot, though sometimes on a final heavy syllable, or even on the third-final foot. He attributes some of this variation to intrinsic properties of vowels

|

294    Matthew Windsor

such as /aa/, which are naturally more sonorous and may attract primary stress. Across Ojibwe then, the realization of primary stress seems less straightforward than the more salient organization of rhythmic structure. The final part of this paper will describe intonational pitch prominences, which I argue have been previously taken for word-level stress.

Method and Terminology

For this study I began by examining a small corpus of recorded Oji-Cree words and phrases, then testing apparent patterns with my colleague Jessie Atlookan, a first-language Oji-Cree speaker. We recorded an additional word-list initially based on the appendix of Valentine (1996) and later expanded to include an additional 150 words. I subsequently combed through a corpus of natural texts recorded from elders, focusing on common words with relevant syllable patterns and paying attention to prosodic and pragmatic context.4 Throughout the process, I noted rhythmic patterns as heard, then verified them using software for acoustic analysis, PRAAT (Boersma and Weenink 2021). This step was particularly helpful for identifying the exact site of boundary tones (a High-Low contour that falls over the pre-final foot). In the remainder of this paper I will use terminology following the work of Van der Hulst (2012, 2014b) and Himmelmann and Ladd (2008), whose approaches are grounded in typological perspective and are conservative in terms of theoretical mechanisms. The term prominence refers to the relative perceptual salience of a syllable (or some other unit). Prominence is understood to be made up of the combined effects of accent, rhythm, boundary effects, word stress, and intonation-level prosody, with languages differing in how they make use of each of these. Accent refers to the phonologically or lexically specified location of prominence in a particular word, while stress is the phonetic realization of accent. Rhythm refers to the organization of non-primary stresses (or rhythmic beats) within a word—the alternation of strong and weak syllables (Van der Hulst 2012). Intonational pitch prominence refers to the prominence of a syllable within an intonation unit cued by pitch, including the boundary tones that mark their edges (Himmelmann and Ladd 2008).

|

Rhythm and Intonation in Oji-Cree   295

Word Rhythm in Oji-Cree

As in other Ojibwe languages, rhythmic beats in Oji-Cree fall on heavy syllables (those containing long vowels) and word-final syllables. Otherwise, beats fall on alternating light syllables. At the beginning of a word, alternating beats typically begin on the initial syllable, representing a trend away from the iambic pattern of other Ojibwe languages. As in Northwestern Ojibwe, rhythmic lapses are preferred over clashes in OjiCree. This is represented in Figure 1 using a grid structure, with each rhythmic beat marked with an ‘x’ and the intrinsically strong (heavy and final) syllables in bold.5 In Figure 1 rhythmic beats occur on alternating syllables beginning at the start of the word. The beat that would fall on hsi is skipped to avoid a clash with the following heavy syllable naa, instead creating a lapse—a series of two unstressed syllables. Lapses are created wherever a series of light syllables runs into the final syllable (b in Figure 2) or a heavy syllable (c,d). Similarly, rhythmic clashes (d) only occur between intrinsically strong (heavy and final) syllables. The beat-initial “trochaic” pattern applies equally to nouns (5):

FIGURE 1. Rhythmic lapse in Oji-Cree

FIGURE 2. Basic Oji-Cree lapse and clash patterns

|

296    Matthew Windsor

(5) a. ᒪᐦᑭᓯᐣ

mahkisin ‘shoe’

b. ᒪᐦᑭᐃᓯᓇᐣ mahkisinan

‘shoes’

c. ᐸᐦᑫᑲᒥᐠ pahkekamik

‘room’

d. ᐸᐦᑫᑲᒥᑯᐣᐠ pahkekamikonk ‘in a room’

A subset of vocabulary included in the study displayed alternating beats beginning on the second syllable (the iambic pattern).6 In some cases the sequence ah appears to carry an inherent level of prominence, as in sakahkinike ‘s/he’s tidying’. In other cases, there was no measurable difference in prominence between two adjacent light syllables, rendering the rhythmic pattern indeterminate. Acoustic Correlates of Rhythm: Length and Reduction Processes

In Oji-Cree, rhythmic stress is most closely associated with vowel duration. As in Menominee (Milligan 2005), the longest syllables contain phonemically long vowels; the next longest are rhythmically stressed syllables containing short vowels; the shortest are weak syllables (containing short vowels). Weak syllables are recognizable by their level of reduction: their vowels are often shorter in duration, devoiced, deleted, or subject to processes like unrounding.

FIGURE 3. Duration and voicing across syllables in takohshin

|

Rhythm and Intonation in Oji-Cree   297

Figure 3 shows a waveform and pitch contour for an elicited pronunciation of ᑕᑯᐦᔑᐣ takohshin ‘s/he arrives’, showing longer duration on strong syllables and devoicing across the weak syllable. For this isolated utterance, pitch falls over the duration of the word, which is also the typical pattern in natural speech. The pitch contour of individual tokens varied according to their position within a phrase and did not correlate with word rhythm. More significant pitch movements are associated with multi-word phrases, discussed in the second part of this paper. Aspiration Spreading

One reduction process that deserves additional attention is aspiration spreading, whereby the glottis retains a spread position across a sequence of weak syllables adjacent to an h + consonant cluster. For example, underlying ᓲᑭᐦᑅᐣ sookihpwan ‘it’s snowing’ is frequently pronounced soohki̥hpwan [soːʰkɪ̥ʰpwʌn], with the underlying aspiration of [hp] spreading across the preceding weak syllable to create a voiceless vowel and pre-aspirated [ʰk]. This phenomenon is also observable in the tendency of speakers trained in transcription to locate the underlying /h/ segment in different positions within a word. Rhythmically strong syllables preceding /hC/ are also partially devoiced or otherwise glottalized, but unlike weak syllables they are frequently pronounced with significant intensity, as if to maintain their prominence through compensation for lost voicing. Recorded examples include kihtimowin ‘laziness’ and pahkinaake ‘he wins’. Aspiration spreading in Oji-Cree bears resemblance to aspiration floating in Tariana (Aikhenvald 2003) as well as floating glottal features proposed for Klamath (Blevins 1993), Mixtec (Macaulay and Salmons 1995), and Cuzco Quechua (Parker and Weber 1996). Similar processes of vowel devoicing preceding hC clusters also occur in several other Algonquian languages, including but not limited to East Cree (Knee 2018), Miami-Illinois (Costa 2003:103) and Cheyenne (Leman and Rhodes 1978). Primary Stress

As described in the literature, the most perceptually salient syllable in Oji-Cree is most often the strong syllable of the pre-final foot, though the relevant domain appears to be the phrase or intonation unit, rather than the individual word.

|

298    Matthew Windsor

Notably, many longer words such as siikahantawaawahsinaaniwan from Figure 1 do not seem to carry a single culminative stress, the long vowels all being about equal in duration, and pitch movements varying across tokens. Oji-Cree therefore appears to lack culminative accent, meaning that words may contain multiple rhythmic prominences, but do not necessarily carry a single most prominent stress (Himmelmann and Ladd 2008). The lack of culminative accent is a property said to be associated with morphologically complex languages, where significant prominences are more likely to be conditioned by phrasal intonation (Van der Hulst 1997). This appears to be the case for Oji-Cree. Person Prefixes

Unlike in Eastern Ojibwe, the addition of person prefixes in Oji-Cree does not affect the rhythm of noun and verb stems. In Table 1, the stem rhythm of the verb takohshin (row a) is unaffected by the addition of person prefix ki- (row b). The same pattern applies to possessed nouns (rows c-d).7 It is unclear from the existing literature how widespread this pattern may be across varieties of Ojibwe. Preverbs and Prenouns

Oji-Cree preverbs and prenouns do not affect the rhythm of the stem, as in most other varieties of Ojibwe. What is notable is that longer strings of preverbs group together to form rhythmic units that roughly correspond to morphological slots. Table 2 shows a simplified morphological ordering of preverbs in conjunct order, with rhythmic domains differentiated with bold borders.8 Table 3 provides a few illustrative forms with the root takohshin ‘s/he arrives’. Rhythmic clashes involving

TABLE 1. Person prefixes and rhythm in Eastern Ojibwe and Oji-Cree

EASTERN OJIBWE

Oji-Cree

b.

dagoshin gi-dagoshin

takohshin ki-takohshin

c.

makizin

mahkisin

ᒪᐦᑭᓯᐣ

d.

ni-makizin

ni-mahkisin

ᓂᒪᐦᑭᓯᐣ

a.

GLOSS

ᑕᑯᐦᔑᐣ ᑭᑕᑯᐦᔑᐣ

‘he arrives’ ‘you arrive’ ‘shoe’ ‘my shoe’

|

Rhythm and Intonation in Oji-Cree   299

TABLE 2. Simplified morphological ordering of preverbs in conjunct order

SUBORDINATION

TENSE/ MODALITY

ASPECT/DIRECTION

kii- pst pi- toward speaker, ci-kii- ability continuous aspect

e- sub

RELATIVE PREVERBS

(SEMI-) LEXICAL PREVERBS

ishi- ‘in such a way/place’

kakwe‘try/make an effort to’

ASPECT/ DIRECTION

ani- gradual onset

TABLE 3. Preverbs and rhythmic domains

syllabic

Roman

Gloss

a.

ᐁᑮ ᑕᑯᐦᔑᐣᐠ

e- kii- [takohshink

‘[when] he arrived’

b.

ᒋᑮ ᐊᓂᐱᑕᑯᐦᔑᐣᐠ

ci-kii- [ani-pi- [takohshink

‘he can start to arrive (here)’

c.

ᒋᑮᐃᔑ ᑲᑫᐧ ᐊᓂᑕᑯᐦᔑᐣᐠ

ci-kii- _shi- [kakwe- [anitakohshink

‘he can try to start to arrive (here, in such a way)’

d.

ᐁᑮ ᓂᐦᓲ ᑕᑯᐦᔑᓂᐋᐧᐨ

e-kii-nihsoo[takohshiniwaac

‘[when] they three arrived’

light syllables demonstrate the boundaries between rhythmic domains. Rhythm can therefore function to disambiguate preverbs from the beginnings of lexical stems. The contrasting phrases in (6) differ only by rhythm; otherwise they contain the same string of segments. In (6b), a rhythmic clash serves to separate the preverb kii- from the beginning of the verb stem shiki ‘pee’. Otherwise, the string kii- is interpreted as the beginning of the verb kiishiki ‘be full grown’ (6a). (6) a. ᑳᐏᐣ ᒋᑮᔑᑭᔮᐣ. Kaawin ci-kiishiki-yaan. neg sub:irr-be.full.grown-cj.1sg ‘I’m not grown up/fully grown.’ b. ᑳᐏᐣ ᒋᑮ ᔑᑭᔮᐣ.

Kaawin ci-kii- [shiki-yaan. neg sub:irr-able-pee-cj.1sg ‘I’m unable to pee.’

|

300    Matthew Windsor

Further evidence that the units in Table 3 form a psychologically real unit comes from orthographic practices. Literate speakers in Kingfisher Lake optionally insert orthographic spaces after preverbs at one or more of these rhythmic boundaries. In row c, for example, two spaces are inserted: one following the relative preverb ᐃᔑ ishi- ‘in such a way, place’ and one following the semi-lexical preverb ᑲᑫᐧ kakwe- ‘try/ make an effort’. Along with pause placement, rhythm is one phonological criterion that serves to identify preverbs as a phonological unit within a larger grammatical word (cf. Wolfart 1973:75–76 for discussion of preverb “compounding” in Plains Cree). Interim Summary

In comparison to other varieties of Ojibwe, the main distinctive features of Oji-Cree word rhythm therefore include: 1. a

preference for rhythmic stress beginning on the first syllable, creating a trochaic rather than iambic pattern; 2. a preference for rhythmic lapses (a series of two consecutive weak syllables) over rhythmic clashes (two consecutive strong vowels); 3. person prefixes that do not affect the rhythm of noun and verb stems; 4. preverbs that group together into rhythmic units, corresponding to the organization of templatic morphological slots. Functional Advantages to the Oji-Cree System

A functional advantage to the organization of Oji-Cree word rhythm is that it maintains the initial rhythm of noun and verb stems, making it easier to recognize the same lexical word in different grammatical contexts. At the other end of the Ojibwe complex, Odawa has developed a very different strategy to achieve the same goal. In this dialect, where person prefixes do affect word rhythm, a new generation of speakers has innovated new prefix forms (underlying nidoo-, gidoo-, odoo-). Because the new prefixes contain a long vowel, they no longer affect the initial (iambic) rhythm at the beginning of word stems. The Odawa forms in Table 4 are adapted from Valentine (2001), with deleted weak vowels reinstated for ease of comparison. In this way Odawa and Oji-Cree have developed different strategies for maintaining the transparency of noun and verb stems.

|

Rhythm and Intonation in Oji-Cree   301

TABLE 4. Odawa original and innovative person prefixes in parallel to Oji-Cree

SINGULAR ‘SHOE’

POSSESSED ‘MY SHOE’

STEM RHYTHM AFFECTED

Oji-Cree

ᒪᐦᑭᓯᐣ mahkisin

ᓂᒪᐦᑭᓯᐣ ni-mahkisin

no

Odawa

makizin

Original: ni-makizin (58)

yes

Innovative: nidoo-makizin (199)

no

Also note the rhythmic correspondence of prefixed forms between dialects in Table 4. The Oji-Cree system may have come about through the reanalysis of such forms, with the original rhythm of prefixed stems (ni-mahkisin) being extended to non-prefixed stems (mahkisin). A comparable process appears to have created nouns with lexicalized rhythm in Menominee, for example (Bloomfield 1962:93).

Intonation and Word Rhythm

The other most significant prosodic feature of Oji-Cree is intonational pitch prominence. Speakers of all languages chunk their speech into segments often called intonation units (IUs), which typically function to present one salient new idea at a time (Chafe 1994). The nature of Algonquian grammar is such that in natural speech IUs often consist of a single word. Oji-Cree IUs are characterized by a cohesive pitch contour that is relatively flat, typically with an initial pitch reset and a falling final contour. The speech of some elders is characterized by final devoicing at the end of each IU, and significant pauses between IUs are also standard in extended monologue. The following sections will provide an overview of a few of the essential intonation patterns, focusing on the relationship between intonation and rhythm. Intonation of Statements

Statements are typically characterized by a pitch prominence on the pre-final foot of the IU. In monologues, this contour occurs most prominently at the end of

|

302    Matthew Windsor

FIGURE 4. Basic statement intonation

thematic groupings of clauses, and a preliminary look at recordings of conversation indicates they may be associated with the organization of turn-taking.9 Figure 4 shows a very long IU with fairly flat pitch until the last word, where a High-Low contour falls over the pre-final foot. Another common intonation contour associated with emphasis or surprise involves a high pitch contour in the same location. In (7) this occurs on the final noun omahkisiniwaa ‘their shoes’ just before the speaker begins to laugh. She is telling about a period when all the women of her generation wanted groundhog fur on their shoes, so they were continually going out to set snares for groundhogs. The high boundary tone is marked with an acute accent. (7) ᒦᑕᔑᐦᐃ ᐁᐃᐧᓈᐸᒋᐦᐋᐋᐧᐨ ᐁᐃᐧᒦᓴᐁᐧᑕᐦᐃᑳᑫᐋᐧᐨ ᐅᐦᐅᒫ ᐅᐦᒪᑭᓯᓂᐋᐧ. mii tahsh ihi e-wii-inaapacihaawaac e-wii-miihsawecahikaakewaac ohomaa o(mahki)(síni). ‘They wanted to use it to sew the fur onto their shoes!’ (followed by laughter)

In these cases, pitch prominence occurs on a strong syllable as determined by word-level rhythm, sometimes bleeding into a following light syllable. Intermediate phrases are also sometimes identifiable within an IU, characterized by a much smaller pitch prominence (again on the pre-final foot). Intonation in Questions

Oji-Cree question intonation shows a different type of pattern in which boundary tones operate independently of word-level rhythm. Content questions carry a pronounced High-Low contour beginning on the second last syllable of the IU (the prominent high tone is indicated with an acute accent).

|

Rhythm and Intonation in Oji-Cree   303

(8) ᐋᐣ ᐁᔑᓈᑯᓯᐨ ᑰᐣ? aan eshinaakosí-c koon? how/what ic\look.so.ai-cj.3sg snow ‘What does the snow look like?’

Yes/no questions share a very similar intonation pattern and are typically also marked with the question particle ᓇ =na. Na is a second-position clitic that occurs after the first grammatical word and behaves like a suffix for the purpose of word rhythm. In (9) the main verb kikii-metawe ‘you played’ occupies a different position in each question, yet the high tone always remains on the second-to-last syllable of each IU. (9) a. ᑭᑮᒣᑕᐁᐧ ᓇ ᑫᑮᐣ? Ki-kii-metawe=na kékiin? 2-pst-play.ai=q you.too ‘Did you play too?’ b. ᑭᑮᒣᑕᐁᐧ ᓇ?

Ki-kii-metawé=na? 2-pst-play.ai=q ‘Did you play?’

c. ᐋᔕ ᓇ ᑭᑮᒣᑕᐁᐧ? Aasha=na ki-kii-metáwe? yet=q 2-pst-play.ai ‘Did you play yet?’

What is notable about the last example (9c) is that the pitch peaks on a rhythmically weak syllable. Figure 5 shows the pitch contour for (9c), with an initial pitch prominence on the initial questioned element aasha ‘yet’ and a high tone on the weak penultimate syllable of metawe ‘play’, which nevertheless retains its short duration. The pitch contour of individual words can similarly vary according to their position within statements. Observe the contrast between two instances of the verb pakitinank ‘s/he offers (it) to him/her’, uttered in non-final position (10a) and in the final, prominent position at the end of an IU (10b).

|

304    Matthew Windsor

FIGURE 5. Waveform and pitch contour for aasha na ki-kii-metawe

(10) a. ᑳᐃᐧᐣ ᒋᐅᐣᒋ ᐸᑭᑎᓇᐣᐠ ᐅᑭᒫᐃᐧᐣ ᐊᓄᐦᑮᐃᐧᐣ ᒥᔕᒥᐦᑯᐄᐧᐦᔑᐣᐠ. Kaawin ci-onci-pakitinank okimaawin anohkiiwin Mihshamihkowiihshink. ‘The government didn’t offer any jobs in Mishamikoweesh.’ [SW01.32] b. ᑳᐃᐧᐣ ᒋᐅᐣᒋ ᐸᑭᑎᓇᐣᐠ.

Kaawin ci-onci-pakitinank. [pʌkɪtń̩ʌnk] ‘He didn’t offer it.’ (elicited)

In (10a), pakitinank is pronounced within a very flat pitch contour in the non-prominent part of the IU. In (10b), the rhythm of pakitinank is shown clearly by the loss of rhythmically weak /i/ between homorganic consonants, yet the pitch peaks on the remaining syllabic /n/. Cases like this illustrate high pitch prominence occurring on a rhythmically weak (and phonetically reduced) syllable. In most languages, lexical stress serves as an intonational anchor for pitch prominence so that intonational pitch accents must always occur on a stressed syllable (Himmelmann and Ladd 2008). In the above examples, intonational prominence is instead assigned in a top-down fashion, independently of word rhythm. A handful of languages with similar systems are mentioned by Himmelmann and Ladd (2008) and Gordon (2003, 2014). The prosodic system of Wolof,

|

Rhythm and Intonation in Oji-Cree   305

for example, presents interesting similarities, having relatively flat pitch contours and intonational pitch prominences that occur on unstressed syllables (Rialland and Stéphane 2001). This pattern underscores the importance of studying word rhythm in different phrasal positions rather than solely through the elicitation of isolated words, which speakers pronounce with varying intonation patterns.

Conclusion

This paper has sketched some of the distinctive features of Oji-Cree rhythm and intonation and demonstrated how they interact. Regarding rhythm, Oji-Cree words primarily stress odd-numbered syllables and avoid rhythmic clashes. The rhythm of grammatical word stems is always maintained regardless of prefixation, contributing to Oji-Cree’s relative morphological transparency (Valentine 1994:48). Regarding intonation, statements are characterized by a pitch prominence on the strong syllable of the pre-final foot, sometimes bleeding into a following light syllable. Questions contain a high-low boundary tone beginning on the secondto-last syllable regardless of word-level rhythm. This last pattern is typologically rare in that pitch prominence is assigned in a top-down fashion, independently of word-level prominence (Gordon 2014). Within Kingfisher Lake, this study contributes to a foundation of linguistic analysis undergirding wider efforts to create resources in and about the Oji-Cree language. Future work on prosody should identify the ways fluent speakers use prosody to express specific pragmatic constructions, manage turns in conversation, and create rhetorical effect in narrative. For those seeking to better understand the prosody of Algonquian languages, the case of Oji-Cree points to the need to carefully distinguish word-level prosody from intonational phenomena. Notes 1. I am continually grateful to my colleagues at Mishamikoweesh corp. for their collaboration in all things Oji-Cree. My thanks goes especially to Jessie Atlookan for her help with pronunciation and in the recording of additional phrases for this study. I am also grateful to the participants of the 53rd Algonquian Conference, two reviewers who gave feedback during the revision of this paper, and Monica Macaulay. A non-technical

|

306    Matthew Windsor

summary for community members is available at www.kingfisherculture.org/grammar. 2. In the northern roman orthography, “lenis” consonants are represented with voiceless characters and “fortis” consonants with a preceding . In this example I use underlining to represent “main stress” as marked by Valentine. 3. In the current study many recorded tokens of kahkina and awiya were pronounced with rhythmic stress (longer duration) on the first and third syllables. 4. Media recordings are informally viewable at www.kingfisherculture.org; one subcollection is available in print as Mamakwa and Windsor (2021). 5. The use of grid structures here follows Van der Hulst (2014a), whose approach has its roots in the original grid structures proposed in (Prince 1983). 6. Words with rhythm beginning on the second syllable included: i. ᓇᐸᑲᐣᑕᐦᐊᐣ

ii. ᓇᒪᑕᐱ

iii. ᓂᑲᒧ

napakantahan

‘flatten it!’

namatapi

‘s/he sits on the ground’

nikamo

‘s/he sings’

iv. ᓂᐱᒥᓂᐦᔕᐦᐋᐧ ni-piminihshahwaa ‘I’m following him/her’ v. ᑭᐦᑐᐦᒋᑲᐣ

vi. ᐅᐣᒋᑯᓯ

kihtohcikan

‘guitar’

oncikosi

‘s/he moves from there’

7. Abbreviations: 1 = first person, 2 = second person, 3 = third person, ai = Animate Intransitive verb, cj = conjunct, irr = irrealis, neg = negative, pst = past, q = yes/no question particle, sg = singular, sub = subordinator. 8. A few speaker-oriented preverbs such as frustrative aana- and indirect evidential mateoccur directly after the tense/modality slot. Volitional wii- occurs in a slot following relative preverbs, and the “semi-lexical” slot can be divided into preverbs involving a degree of control (maaci- ‘start’) followed by those expressing manner, quality, or number. Ability is actually signaled by the special combination of a future or irrealis prefix (not only ci-) plus kii-. 9. See Couper-Kuhn and Selting (2018:69–72) for an overview of how speakers use prosodic cues to open “transition relevant space” between conversational turns.

References Aikhenvald, Alexandra. 2003. A Grammar of Tariana, from Northwest Amazonia. Cambridge: Cambridge University Press. Blevins, Juliette. 1993. Klamath Laryngeal Phonology. International Journal of American

|

Rhythm and Intonation in Oji-Cree   307

Linguistics 59(3): 237–279. Bloomfield, Leonard. 1958. Eastern Ojibwa: Grammatical Sketch, Texts, and Word List. Ann Arbor: University of Michigan Press. Bloomfield, Leonard. 1962. The Menomini Language. New Haven, CT: Yale University Press. Boersma, Paul, and David Weenink. 2021. PRAAT (Version 6.1.53) [software]. Bogomolets, Ksenia. In press. Accent and Tone in Plains Algonquian Languages. Word Prominence in Polysynthetic Languages, ed. by Ksenia Bogomolets and Harry van der Hulst. Oxford: Oxford University Press. Brittain, Julie. 2000. A Metrical Analysis of Primary Stress Placement in Southern East Cree. International Journal of American Linguistics 66(2):181–217. Chafe, Wallace. 1994. Discourse, Consciousness, and Time. Chicago: University of Chicago Press. Costa, David J. 2003. The Miami-Illinois Language. Studies in the Native Languages of the Americas. Lincoln: University of Nebraska Press. Couper-Kuhn, Elizabeth, and Margret Selting. 2018. Interactional Linguistics: Studying Language in Social Interaction. Cambridge: Cambridge University Press. Cudworth, Andrea. 2019. Menominee Vowels: Quality, Quantity, and a Hierarchical Model of Representation. PhD thesis, University of Wisconsin. Goddard, Ives. 1982. The Historical Phonology of Munsee. International Journal of American Linguistics 49(1):16–48. Gordon, Matthew. 2003. The Phonology of Pitch Accents in Chickasaw. Phonology 20:173–218. Gordon, Matthew. 2014. Disentangling Stress and Pitch-Accent: A Typology of Prominence at Different Prosodic Levels. Word Stress: Theoretical and Typological Issues, ed. by Harry Van der Hulst, pp. 83–118. Cambridge: Cambridge University Press. Hayes, Bruce. 1984. A Metrical Theory of Stress Rules. New York: Garland. Hayes, Bruce. 1995. Metrical Stress Theory: Principles and Case Studies. Chicago: University of Chicago Press. Himmelmann, Nikolaus, and Robert Ladd. 2008. Prosodic Description: An Introduction for Fieldworkers. Language Documentation and Conservation 2(2):244–274. Kaye, Jonathan. 1973. Odawa Stress and Related Phenomena. Odawa Language Project, Second Report, ed. by Glyne Piggot and Jonathan Kaye, pp. 42–50. Toronto: University of Toronto Press. Knee, Sarah. 2018. Vowel Devocalization in Northern East Cree. Canadian Journal of Linguistics 59(3):303–338. Leman, Wayne, and Richard Rhodes. 1978. Cheyenne Vowel Devoicing. Papers of the 9th Algonquian Conference, ed. by William Cowan, pp. 3–24. Ottawa: Carleton University Press.

|

308    Matthew Windsor

LeSourd, Philip S. 1993. Accent and Syllable Structure in Passamaquoddy. New York: Garland. Lockwood, Hunter Thompson. 2017. How the Potawatomi Language Lives: A Grammar of Potawatomi. PhD thesis, University of Wisconsin-Madison. Macaulay, Monica, and Joseph Salmons. 1995. The Phonology of Glottalization in Mixtec. International Journal of American Linguistics 61(1):38–61.

Mamakwa, Lydia, and Matthew Windsor (eds.). 2021. ᑳᐱᒫᒌᑯᓯᔮᐣᑭᐸᐣ | When We Moved: Memories of Mishamikoweesh. Kingfisher Lake: Lulu Press. DOI: 10.5281/zenodo.4571185. Milligan, Marianne. 2005. Menominee Prosodic Structure. PhD thesis, University of Wisconsin-Madison. Muehlbauer, Jeff. 2006. Pitch as Accent in Plains Cree Nominals. Actes du Trente-Deuxieme Congres des Algonquinistes, ed. by H. C. Wolfart, pp. 299–268. Winnipeg: University of Manitoba. Nichols, John D. (ed.). 2014. Anihshininiimowin Oji-Cree Dictionary (Severn River and Winisk River). Sioux Lookout: Kwayaciiwin Education Resource Centre. Olson, Erin. 2012. Describing the Stress System of Listuguj Mi’gmaq. Papers of the 44th Algonquian Conference, ed. by Monica Macaulay, Margaret Noodin, and Randolph J. Valentine, pp. 268–283. Albany, NY: SUNY Press. Oxford, Will. 2020. Algonquian. Routledge Handbook of North American Languages, ed. by Daniel Siddiqi, Michael Barrie, Carrie Gillon, Jason Haugen and Eric Mathieu, pp. 504–523. London: Routledge. Parker, Steve, and David Weber. 1996. Glottalized and Aspirated Stops in Cuzco Quechua. International Journal of American Linguistics 62(1):70–85. Prince, Alan. 1983. Relating to the Grid. Linguistic Inquiry 14(1):19–100. Rialland, Annie, and Robert, Stéphane. 2001. The Intonational System of Wolof. Linguistics 39(5):893–939. Swierzbin, Bonnie. 2003. Stress in Border Lakes Ojibwe. Papers of the 34th Algonquian Conference, ed. by H. C. Wolfart, pp. 341–370. Winnipeg: University of Winnipeg. Valentine, J. Randolph. 1994. Ojibwe Dialect Relationships. PhD thesis, University of Texas at Austin. Valentine, J. Randolph. 1996. Phonological Parameters of Ojibwe Dialect Variation. Papers of the 27th Algonquian Conference, ed. by David Pentland, pp. 287–323. Winnipeg: University of Manitoba. Valentine, J. Randolph. 2001. Nishnaabemwin Reference Grammar. Toronto: University of Toronto Press. Van der Hulst, Harry. 1997. Primary Accent Is Non-Metrical. Rivista di Linguistica 9(1):99–127. Van der Hulst, Harry. 2012. Deconstructing Stress. Lingua 122(13):1494–1521.

|

Rhythm and Intonation in Oji-Cree   309

Van der Hulst, Harry. 2014a. Representing Rhythm. Word Stress, ed. by Harry van der Hulst, pp. 325–365. Cambridge: Cambridge University Press. Van der Hulst, Harry. 2014b. The Study of Word Accent and Stress: Past, Present, And Future. Word Stress, ed. by Harry van der Hulst, pp. 3–55. Cambridge: Cambridge University Press. Voorhis, Paul. 1974. Introduction to the Kickapoo Language. Bloomington: Indiana University. Wolfart, H. Christoph. 1996. Sketch of Cree, an Algonquian Language. Handbook of North American Indians, vol. 17 Languages, ed. by Ives Goddard and William C. Sturtevant, pp. 390–439. Washington: Smithsonian Institution. Wolfart, H. Christoph. 1973. Plains Cree: A Grammatical Study. Transactions of the American Philosophical Society, n.s., vol. 63, part 5. Philadelphia.

Oblique Argument Licensing in Algonquian Head Marking vs. Dependent Marking Yadong Xu

G

oddard (2020) shows that the definite oblique argument in Unami Delaware triggers agreement on the verb, characterized by the appearance of the set of central suffixes known as the n-endings (Goddard 1979:103–106) as well as the peripheral endings (Goddard 1979:38).1 In this paper, I explore whether oblique arguments trigger the same agreement inflection in the other two languages: Ojibwe (Central Algonquian) and Maliseet-Passamaquoddy (Eastern Algonquian), based on data from Valentine 2001 and Rhodes 2010 for Ojibwe; and Sherwood 1983, Bruening 2001, and LeSourd forthcoming for Maliseet-Passamaquoddy. Ojibwe and Maliseet-Passamaquoddy are investigated because their secondary object agreement patterning has been observed by Xu (forthcoming) for bordering with Unami Delaware in a cline. Secondary objects refer to the object of an AI+O verb or the theme (direct object) of a TA+O verb. AI+O verbs are transitive verbs taking two arguments (a subject and an object) and TA+O verbs are ditransitive verbs requiring three arguments (a subject, an indirect object, and a direct object). Specifically, in Ojibwe (the Nishnaabemwin variety) the secondary object of AI+O verbs (Goddard 1979) can be indexed by the combination of n-endings and peripheral endings, but the secondary object TA+O verbs (Goddard 1979) cannot be indexed by this collocation. In the Eastern languages, the secondary object of AI+O and TA+O both can be indexed by the set of n-endings and peripheral endings, but their agreement differs in degrees of restrictiveness: in Unami Delaware the secondary

|

   311

|

312    Yadong Xu

object needs to be definite in order to trigger the agreement markers while the Maliseet-Passamaquoddy secondary object is not subject to such restriction.2 After a cross-linguistic comparison, I will show that the two Eastern languages pattern together for allowing the n-endings and peripheral endings to index oblique arguments, whereas Ojibwe patterns differently for disallowing these agreement suffixes to index oblique arguments. Additionally, even within the Eastern languages that permit agreement with oblique arguments, an interesting divergence is found: as long as the oblique NP bears the case marker *-enki (Dahlstrom 2022), it is no longer available for agreement. The difference arising from case marking therefore leads us to the implication that the variability of the oblique arguments for verb agreement also depends on how they are morphologically marked in the grammar. This paper proceeds as follows. First, I provide the background on grammatical functions and explain the collocation of the n-endings and peripheral agreement with various arguments in Unami Delaware. In what follows, I distinguish oblique arguments into four types, compare each of their agreement pattern in Unami Delaware, Ojibwe, and Maliseet-Passamaquoddy, and show that the boundary of obliques available for agreement rests on case marking. Then, I focus on two types of oblique arguments, pointing out the implication that their different agreement behaviors reflect distinct morphological markings and Case assignments. Finally, I conclude the paper.

Grammatical Relations and Peripheral Agreement

In typological research, grammatical relations can be divided into two general categories: core arguments and oblique arguments (Comrie 1989:66, 179; Dixon 1994:122–123). Core arguments are noun phrases that have close associations with semantic roles determined by predicates (for a much more detailed discussion, see Andrews 2007:152, 157). Subject, direct object, and indirect object are instances of a core argument. In contrast, oblique arguments are noun phrases that provide additional circumstances such as location, time, and manner. Most importantly, unlike core arguments, the form or the meaning of oblique arguments usually is not determined by predicates. In English, the core arguments often are bare NPs or pronouns, whereas the oblique arguments usually are nominals preceded by a preposition, such as the instrumental in (1a) and the locative in (1b). Furthermore, both oblique arguments (underlined) in (1) behave like adjuncts because they can be dropped.

|

Oblique Argument Licensing in Algonquian   313

(1) a. John chased the bear with a stick. b. John ran to the store.

Sometimes the situation is not always as this simple, in that the oblique argument may be obligatory for the predicate rather than being optional like adjuncts.3 The English verb put requires three arguments: a subject, a direct object, and a locative oblique argument. As exemplified below, the example in (2a) containing all three arguments is permissible. In contrast, the example in (2b) lacking the PP phrase and the example in (2c) missing the NP object are impermissible. (2) a. Alice put the key in the box. b. ✗Alice put the key.

c. ✗Alice put in the box.

In Algonquian languages, verbs have an elaborate morphological distinction subcategorizing the requirement on the arguments. A derivational morpheme termed final (Bloomfield 1946:104–111; Goddard 1990) specifies the transitivity of the verb as well as the gender of the argument. Four basic verb classes are thus distinguished by finals. As shown in Table 1, intransitive verbs (i.e., AI and II) are categorized by the animacy of their subjects, and monotransitive verbs (i.e., TA and TI) are categorized by the animacy of their objects. Beyond the basic four-way distinctions, two additional verb classes, AI+O verbs and TA+O verbs, may select an object. As explained earlier in this paper, the former is a transitive verb taking two core arguments (a subject and an object)

TABLE 1. Verb classes and stem shapes (examples from Unami Delaware, Goddard

2021:48–49) CLASS

STEM (FINAL BOLDED)

MEANING: maxk- ‘RED’, pak- ‘HIT’

AI (animate intransitive)

maxkəsi·-

‘something.an is red’

II (inanimate intransitive)

maxke·-

‘something.in is red’

TA (transitive animate)

pakam-

‘to hit someone/something.an’

TI (transitive inanimate)

pakant-

‘to hit something.in’

|

314    Yadong Xu

and the latter is a ditransitive verb requiring three core arguments (a subject, an indirect object, and a direct object). However, these two verb classes do not employ special derivational morphology but use the same set of finals of the AI or the TA. In addition, the object argument selected by these two classes are not specified for gender and therefore can be animate or inanimate. Focusing on the object arguments, they are further distinguished by two kinds: primary object and secondary object (Goddard 1979; Rhodes 1990). The former refers to the object of a TA/TI verb or the goal of a TA+O verb while the latter refers to the object of an AI+O verb or the theme of a TA+O verb. Regarding peripheral agreement, it is a suffix that occurs at the right edge of the main clause-type verb and expresses the nominal features of a third-person argument (gender, number, and obviation). In Algonquian ditransitives, the goal patterns with the object of the monotransitive verb (i.e., primary object), while the theme (i.e., secondary object) patterns differently. The Unami Delaware examples provide evidence that all three types of core arguments can be indexed by peripheral agreement (bolded): subject as in (3a), primary object as in (3b), and secondary object as in (3c). Note, despite the similarity in their phonological forms, the w-suffix with the umlaut diacritic in (3a) is categorized as one of the m-endings by Goddard (1979:103) and is distinct with the w-suffix in (3b). Two most noticeable differences can tell apart these two w-suffixes. First, the m-ending -w (written as -ẅ in the interlinear gloss, a notation introduced by Nilsen 2017) never co-occurs with the prefix, while the w-ending one always does.4 Second, the m-ending gives rise to umlaut to the stem (the diacritic indicates its umlauting effect), while the w-ending does not trigger umlaut to the stem. (3) a. Subject kəntəké·yɔk kəntka·-ẅ-ak dance-3-3pl ‘They (an) dance.’ (Unami Delaware, Goddard 1979:167) b. Primary object nəmi·lá·ɔk nə-mi·l-a·-w-ak 1-give-3.obj-1sg-0pl ‘I give to them (an).’ (Unami Delaware, Goddard 1979:171)

|

Oblique Argument Licensing in Algonquian   315

c. Secondary object nəmi·lá·na nə-mi·l-a·-n-a 1-give-3.obj-1sg-0pl ‘I gave them (in) to him/her.’ (Unami Delaware, Goddard 2020:104)

Looking more closely at the last two examples in (3), peripheral agreement is preceded by the suffix called central endings (Goddard 1979:38), which consists of a formative element and a number morpheme (unmarked for singular participants, -əna·n for 1pl participant, and -wa·w for 2pl or 3pl participants). The central suffix may inflect for one of the three sets given the formative element it contains: m-endings, w-endings, and n-endings (Goddard 1979:103–106). The selection of the formative element actually correlates with the behavior of peripheral agreement (Goddard 2007:264). As exemplified in (3b), the w-ending is used when peripheral agreement indexes a lower-ranked animate primary argument. However, the use of n-endings is not conditioned by gender. As put forth by Goddard (2020), the n-endings appear in a much broader contexts in that they are not only used in the inflection involving an inanimate argument (the TI inflection and the TA inanimate subject inflection) but also in cases where peripheral agreement indexes the secondary object of either gender, or other nominal elements (oblique arguments, relative root complements, and adjuncts). Following Goddard (2007), I analyze the central suffix -w in (3b) and that of -n in (3c) as a singular suffix, specifying the number of the primary participant (its person feature is expressed by the prefix, here nə- 1). Returning to the discussion of peripheral agreement in terms of its ability to index various grammatical functions, as discussed in Goddard (2020), it is capable of indexing oblique arguments in Unami Delaware. The example with an instrumental oblique is provided in (4). Most importantly, the oblique agreement is indicated by the peripheral suffix (morphologically zero for the 0sg category) accompanied by the n-ending central suffix. The other Eastern language under investigation, Maliseet-Passamaquoddy, patterns with Unami Delaware and its examples will be provided in the following section.

|

316    Yadong Xu

(4) Instrumental pɔ́ k·ama·n ə-pakam-a·-n-Ø 3-hit-3.obj-3sg-0sg ‘He struck him/her with it.’ (Unami Delaware, Goddard 2020:106)

In contrast to (4), the instrumental oblique in Nishnaabemwin (Ojibwe), instantiated in (5), does not give rise to the same agreement suffixes. Instead, the verb inflection uses the intransitive AI morphology, with peripheral agreement indexing the subject, in parallel with the Unami Delaware AI example given in (3a). (5) Instrumental wezhho gye gkizhe wezhho-Ø-Ø

gye gkizhe.

adorn.refl-3-3sg also charcoal ‘He paints himself with charcoal.’ (Nishnaabemwin, Valentine 2001:659)

Above all, peripheral agreement in Unami Delaware is versatile for indexing various kinds of grammatical functions: subject, primary object, secondary object, and oblique argument. When it comes to oblique agreement, a language-based division is observed: the Eastern languages permit oblique agreement while Ojibwe forbids it. When the verb allows agreement with secondary objects and oblique arguments, the same inflection is used (the n-endings and peripheral agreement).

Cross-Linguistic Patterns of Oblique Agreement

This section describes four morphological strategies to code an oblique argument in Algonquian languages. The nominal element regarded as the oblique argument may be unmarked, appearing as a bare NP, or it could be morphologically indicated by three means: via case marking (Dahlstrom 2022), a relative root (Bloomfield 1946:120; Rhodes 1990), or both at the same time.5 In this section, I start with Ojibwe examples for the purpose of illustrating each strategy. We will see that agreement is not available with oblique arguments of any kind in Ojibwe. Then, I present the corresponding examples from Unami Delaware and Maliseet-Passamaquoddy, arguing that obliques introduced by the former two strategies are available for

|

Oblique Argument Licensing in Algonquian   317

agreement, however, those introduced by the latter two strategies are unavailable for agreement. Patterns in Ojibwe

The Ojibwe examples are from the Eastern varieties of Nishnaabemwin and Odawa, which are virtually identical in their oblique agreement patterning. In the first strategy, the oblique argument appears as a bare NP, such as the temporal oblique ngo-bboon ‘for one winter’ illustrated in (6). I will refer to this kind of obliques as the bare type. Zooming in the verb inflection, the prefix ni- and the zero central suffix -Ø agree with the 1sg subject. Due to lacking peripheral agreement, there is no agreement with the bare type obliques. (6) Bare type ngo-bboon ngii-yaa widi ningo-biboon ni-gii-ayaa-Ø iwidi one-winter 1-past-be-1sg there ‘I went there for one year.’ (Odawa, Rhodes 2010:308)

The example in (7) shows an instance of a locative oblique. Concentrating on jiimaaning ‘in the canoe’, the locative oblique is constructed by adding the locative suffix -ing to the noun. I follow Dahlstrom (2013, chapter 3) to regard -ing as a case marker. Henceforth, I will refer to the NPs bearing the case suffix as the case type. The verb again shows the intransitive AI inflection. Therefore, no agreement with the case type oblique occurs in Nishnaabemwin. (7) Case type jiimaaning ngii-boozmi miinwaa bezhig kwe jiimaan-ing n-gii-boozi-min

miinawaa bezhig kwe

canoe-loc 1-past-embark-1pl and

one

woman

‘She and I and another woman got into a canoe.’ (Odawa, Rhodes 2010:314)

The case suffix *-enki not only can express various types of location: goal (e.g., ‘for’), source (e.g., ‘from’), or path of motion (e.g., ‘to, toward’), or stationary location (e.g., ‘in, on, at’), but also can express non-locative meanings that are associated with manner or appearance (Dahlstrom 2022:3–29). The Nishnaabemwin noun

|

318    Yadong Xu

mhiing ‘wolf’ in (8) bears the case suffix -ing, illustrating the non-locative function, and accordingly is translated as ‘like wolves’. (8) mhiinging zhinaaziwag gonad nimookajiinyig mhiingn-ing zhinaazi-ẅ-ag gonad

nimmokaajiiny-ig

wolf-loc look.so-3-3pl these.3pl dog-3pl ‘These dogs look like wolves.’ (Nishnaabemwin, Valentine 2001:664)

With regard to the third strategy, the oblique NP may be introduced by relative roots. I will refer to this type as the RR type. Relative roots (glossed as ‘rr’) are preverbal elements that obligatorily make a reference to a nominal element in the clause (Bloomfield 1946:120; Wolfart 1973:66; Rhodes 1998, 2010). Relative roots can be understood as an adposition-like morpheme that is morphologically part of the verb but takes a freestanding phrase as its complement. The example in (9) exemplifies an oblique introduced by the relative root, dso- ‘in a certain number, so many’. The sentence initial NP mdaaswi-shi-naanan ‘fifteen’ is selected as the complement of the relative root, contributing to the semantics of manner. The verb inflection again shows intransitive agreement morphology, cf. (5), lacking oblique agreement.6 (9) RR type mdaaswi-shi-naanan dso-bboon’-gizi midaaswi-shi-naanan dso-biboon-gizid-Ø-Ø ten-plus-five

so.many.rr-year-have-3-3sg

‘He is fifteen years old.’ (Nishnaabemwin, Valentine 2001:664)

Digressing on relative root complements, they can be overt NPs like midaaswi-shi-naanan in (9), deictics, or pronouns. It is also possible for them to be covert elements. Such as in (10), the relative root complement refers to the spatial deictic ‘there’ (noted as Ø in boldface), even though it is not morphologically present in the clause. (10) Covert baamaa go ga-zhaami baamaa go Ø

gi-ga=iN-yaa-min

afterwards ptcl there 2-fut=to.rr-go-1pl ‘Afterwards we (inclusive) will go there.’ (Odawa, Rhodes 2010:314)

|

Oblique Argument Licensing in Algonquian   319

TABLE 2. Ojibwe: Agreement by types of oblique arguments

OBLIQUES

BARE TYPE

RR TYPE

CASE TYPE

MIXED TYPE

Nish









Last, the oblique NP may be introduced using two means at the same time, via case marker and a relative root. As exemplified in (11): the relative root is iNexpressing manner, and the corresponding relative root complement is aniniwing ‘in the form of a man’ which bears the case suffix. In what follows, I will refer to this kind as the mixed type. Focusing on the verb inflection, the TA inflection is shown: the prefix o- and the central suffix -Ø index the proximate subject, and peripheral agreement -an indexes the obviative object. Again, the co-occurrence of the n-endings and peripheral agreement is not found to support for oblique agreement in Nishnaabemwin. (11) Mixed type niniing-sh go naa wgii-naabmaan niwi mnidoon aniniw-ing=sh go naa o-gii=iN-waabam-aa-Ø-an

aniwi manidoo-an

man-loc=ptcl ptcl ptcl 3-past=in.rr-see-3.obj-3sg-3′ this.3′ spirit-3′ ‘He saw the spirit in the form of a man.’ (Odawa, Rhodes 2010:308)

Table 2 sums up the patterns of oblique agreement in Nishnaabemwin. In a word, oblique agreement does not take place regardless of which type of strategies being used. Patterns in Eastern Algonquian

As mentioned earlier, oblique arguments are accessible for peripheral agreement in Unami Delaware and Maliseet-Passamaquoddy. In what follows, I further argue that the four types do not behave the same in terms of agreement. Specifically, when the oblique argument involves case marking, agreement with oblique arguments is prohibited. First, the Unami Delaware example showing the bare type is repeated as (12). Without repeating the previous depictions, the instrumental oblique triggers

|

320    Yadong Xu

agreement, manifested by the combination of the central suffix -n and peripheral agreement (in boldface). To save space, the Maliseet-Passamaquoddy bare type example is omitted here but will be shown in (15a). (12) Bare type pɔ́ k·ama·n wə-pakam-a·-n-Ø 3-hit-3.obj-3sg-obl ‘He struck him/her with it.’ (Unami Delaware, Goddard 2020:106)

Moving on to the RR type, the Maliseet-Passamaquoddy example in (13a) involves a comitative oblique. The relative root wit- ‘with’ selects nuhsimísok ‘my younger siblings’ as its complement. The Unami Delaware example in (13b) illustrates the use of the relative root əli·-, which expresses manner, ‘in a certain way’, and therefore is glossed as ‘so’. The relative root complement is the emphatic demonstrative ná=nə. Apparently, in these two Eastern Algonquian languages, the RR obliques trigger the n-endings as well as peripheral agreement. (13) RR type a. nìl n-wit-ayyanénuuk nuhsimísok nìl n-wit-ayya-nénənw-ak n-uhsimís-ok I 1-with.rr-play-1pl-3pl 1-young.sibling-3pl ‘I play with my younger siblings.’ (Maliseet-Passamaquoddy, LeSourd 2020) b. ná=nə ntə́ li-kí·spwi·n ná=nə nət-əli·-ki·spwi·-n-Ø foc=that 1-so.rr-be.full-1sg-0pl ‘I got full on it.’ [lit. ‘I got full that way.’] (Unami Delaware, Goddard 2020:105)

It is worth mentioning that the relative root əli·- in (13b) corresponds to PA *eθ- ‘thither, thus’ (Bloomfield 1946:120), which is quite common across different daughter languages.7 Other than indicating manner, it also can function as the allative oblique, expressing path of motion or goal, and thus is translated as ‘in a certain direction, to (somewhere)’ (Rhodes 1998:10; Valentine 2001:421; Goddard 2021:45). The previous Nishnaabemwin iN- in (10) also illustrates the allative function, so does the Unami Delaware əl- in (17a).

|

Oblique Argument Licensing in Algonquian   321

As opposed to the previous two types, the third type—the case-marked oblique argument—is no longer available for peripheral agreement in Eastern Algonquian. Using Maliseet-Passamaquoddy for instance, the NP katkohk ‘pot’ in (14) bears the case suffix -ək. In this example, epahka- is an AI verb. It is important to point out that the combination of the central suffix -n and the peripheral suffix -əl demonstrates the AI+O inflection because it agrees with the plural secondary object pəcetesəl ‘potatoes’. Therefore, no agreement occurs with the locative oblique. (14) Case type ntepahkanəl pəcetesəl katkohkək nt-epahka-n-əl pəcetes-əl katkohk-ək 1-throw-1sg-0pl potato-0pl pot-loc ‘I throw potatoes (in) in the pot.’ (Maliseet-Passamaquoddy, Sherwood 1983:121)

The convincing evidence clarifying the unavailability for peripheral agreement with case-marked oblique arguments is given in Sherwood (1983:122–123). The contrast shown in (15) lends support that the key to agreement with obliques lies in whether the NP is marked by the locative suffix -ək. The bare NP wik ‘house’ in (15a) triggers the n-suffix and peripheral agreement. However, the NP wikək ‘in the house’ in (15b) takes the locative case; the same agreement morphemes are disallowed. (15) a. Bare type məwin kwəssəyotan məcəyehsəwəl wik məwin w-kəssəyota-n-Ø məcəyehsəw-əl wik bear 3-move.in-3sg-0sg Partridge-3′

house.0sg

‘Bear moved in into Partridge’s house.’ b. Case type *məwin kwəssəyotan məcəyehsəwəl wikək məwin w-kəssəyota-n-Ø məcəyehsəw-əl wik-ək bear 3-move.in-3sg-0sg Partridge-3′

house-loc

Intended: ‘Bear moved in into Partridge’s house.’ (Maliseet-Passamaquoddy, Sherwood 1983:122)

Alternatively, if the case-marked wikək must be maintained, as given in (16), the grammatical form must use the AI inflection. Namely, the central suffix -ẅ as

|

322    Yadong Xu

well as the zero peripheral suffix together agrees with the subject məwin ‘bear’. As a result, oblique agreement is disallowed. (16) Case type məwin ksəyote məcəyehsəwəl wikək məwin kəssəyota-ẅ-Ø məcəyehsəw-əl wik-ək bear move.in-3-3sg Partridge-3′

house-loc

‘Bear moved in into Partridge’s house.’ (Maliseet-Passamaquoddy, Sherwood 1983:123)

Last, the mixed type is also attested in the Eastern languages. As given in (17a) for Unami Delaware, the case-marked NP mux·ó·link ‘in/to the boat’ is linked to the relative root əli- ‘to’. In the Maliseet-Passamaquoddy example in (17b), the locative NP qospemok ‘in/to the lake’ is the complement for the relative root oloq- ‘to, that way’. Crucially, both oblique arguments again cannot be indexed by peripheral agreement. In Unami Delaware, the verb in (17a) shows the AI inflection agreeing with the subject. The Maliseet-Passamaquoddy verb in (17b) use the TA inflection: the proximate subject is indexed by central agreement and the obviative object is indexed by peripheral agreement -l. Again, agreement is not found for the mixed type obliques. (17) Mixed type a. mux·ó·link lí-pó·s·o·p mux·ó·l-ink əli-po·si·-ẅ-əp-Ø boat- loc to.rr-embark-3-pret-3sg ‘He went aboard a boat.’ (Unami Delaware, Goddard 2021:45) b. ’toloqaphal oloqiw qospemok ’t-oloq-aph-a-Ø-l

oloqiw qospem-ok

3-that.way.rr-track-3.obj-3sg-3′ that.way lake-loc ‘She tracked him (obv) toward the lake.’ (Passamaquoddy, Bruening 2001:169)

Taken together, an interesting split regarding oblique agreement is found within the two Eastern Algonquian languages. As summarized in Table 3, obliques identified as the bare type and the RR type can be indexed by the collocation of the n-endings and peripheral agreement, whereas obliques identified as the case type and the mixed type are not available for the same combination of agreement suffixes.

|

Oblique Argument Licensing in Algonquian   323

Tying together with Ojibwe, Table 4 synthesizes the accessibility of peripheral agreement with each type of oblique argument. In Ojibwe, oblique arguments are inaccessible for peripheral agreement in spite of the strategy that is employed. In contrast, in the two Eastern languages, even though peripheral agreement is able to index oblique arguments, the arguments that involve case marking fail to trigger agreement.

Morphological Markings and Case Licensing

At this point, the question arises out of the split pattern in the Eastern languages: why do case-marked obliques block agreement? In this section, I focus on the case type and the relative root (RR) type, proposing that their morphological markings shed light on the divergence in participation of agreement. When it comes to indicating grammatical relations, Nichols (1986) identifies that the argument is either marked on a dependent constituent as case markers or on the head as agreement affixes, and accordingly, the former is known as dependent marking and the latter is known as head marking. The Chechen example

TABLE 3. Eastern Algonquian: Agreement types of oblique arguments OBLIQUES

BARE TYPE

RR TYPE

CASE TYPE

MIXED TYPE

Unami









Mal-Pas









TABLE 4. Summary: Agreement types of oblique arguments OBLIQUES

OJIBWE

MALPAS

UNAMI

Bare type







RR type







Case type







Mixed type







|

324    Yadong Xu

shown in (18) exemplifies dependent marking because the core arguments, the subject, the direct object, the indirect object, bear morphological cases (erg, dat, nom), but the head, the verb, does not use agreement affixes. (18) da:-s

woʕa-na ur-Ø

tü:xira

father-erg son-dat knife-nom struck ‘The father stabbed the son.’ [lit. ‘father struck son with knife’] (Chechen, Nichols 1986:61)

On the contrary, the Abkhaz example in (19) embodies head marking strategy given that the grammatical relations (subject, indirect object, and direct object) are indexed by agreement affixes on the verb. (19) a-xàc’a a-pħoə̀ s

a-šoqo’ə̀ Ø-lə̀ -y-te-yt’

the-man the-woman the-book it-to.her-he-gave-finite ‘The man gave the woman the book.’ (Abhkaz, Nichols 1986:61)

In the literature on Case theory (Chomsky 1981; Vinokurova 2005; Legate 2008; among others), arguments must be licensed, thematically by the assignment of theta-roles, and syntactically by abstract Case (Lochbihler 2012:13). There are two ways for nominals to be assigned Case, either via lexical Case by appearing in “Cased positions” (Polinsky and Preminger 2014:9), such as English PPs, or via structural Case by moving to a position where they can receive Case, such as the EPP (Extended Projection Principle) requirement demanding the overt nominative subject in English. In what follows, I argue that the case type oblique reflects the dependent marking and thus is licensed by means of lexical Case, whereas the RR type oblique reflects the head marking and therefore is licensed by virtue of structural Case. Returning to Algonquian oblique arguments, the previous Maliseet-Passamaquoddy example is repeated as (20). The case type exactly follows the dependent marking morphology because the locative argument is indicated by the morphological case -ək on the noun phrase wik ‘house’. Furthermore, no agreement with the locative argument appears on the head.

|

Oblique Argument Licensing in Algonquian   325

(20) Case type məwin ksəyote məcəyehsəwəl wikək məwin kəssəyota-ẅ-Ø məcəyehsəw-əl wik-ək bear

move.in-3-3sg Partridge-3′

house-loc

‘Bear moved in into Partridge’s house.’ (Maliseet-Passamaquoddy, Sherwood 1983:123)

As for the RR type, relative roots can be understood as the adposition-like morpheme that is marked on the head, and its complement occurs as a freestanding phrase. The previous Maliseet-Passamaquoddy comitative example is repeated as (21). Similar to the way that the benefactive argument ‘the woman’ being licensed by the prefix lə̀- ‘to her’ in (19), the comitative NP appears in the caseless form, and crucially, the corresponding relative root is precisely a prefix appearing on the head and licenses the argument ‘my younger siblings’. (21) RR type nìl n-wit-ayyanénuuk nuhsimísok nìl n-wit-ayya-nénənw-ak n-uhsimís-ok I 1-with.rr-play-1pl-3pl 1-young.sibling-3pl ‘I play with my younger siblings.’ (Maliseet-Passamaquoddy, LeSourd 2020)

After recognizing that the two types of obliques result from distinct morphological markings, I further suggest that these two types can be formally captured by different Case assignments. It is worth noting that both types of obliques enter the derivation inside the VP and below the core arguments in that, cross-linguistically, oblique arguments are c-commanded by other core arguments (Marantz 1993).8 As schematized in (22a), the dependently marked oblique arguments appear in the Cased position as the complement of PP. Consequently, the head P assigns the locative case to its complement. Therefore, Algonquian locatives bearing the suffix *-enki receive Case lexically. However, I propose that the head marked oblique receives abstract Case structurally from V. Following Bruening (2001:170), I assume a relative root complement is introduced by the relative root phrase (RRP) as schematized in (22b). Significantly, the relative root oblique nuhsimísok ‘my younger siblings’ is merged at the position in which structural Case is assigned. Therefore, the relative root complements are assigned Case structurally. Due to the scope of this paper, I leave the complete

|

326    Yadong Xu

(22a)

(22b)

clausal structure as well as the explanation about the linear sequence of the fronted relative root for future research. The importance of attributing these two distinct morphological markings to different Case assignments is threefold. First, it captures the profound consequence of distinct Case assignments on the reflection of Algonquian morphology. Namely, the lexical Case is phonologically pronounced as case markers in Algonquian languages, whereas the structural Case is not phonologically realized by morphological cases (i.e., core arguments and the RR obliques). Second, it connects with the theoretical literature on variability of the nominals for participating in syntactic operations. In generative syntax, PP is considered as a phase (Abels 2012:202–220; Bošković 2014) and prevents its internal argument from taking part in agreement. In a similar vein, the inability of case-marked obliques triggering the agreement suffixes exactly results from being blocked by the PP-shell. In contrast, the RR obliques are accessible for agreement due to not being deactivated by the PP-shell. Third, it corroborates the cross-linguistic behaviors of peripheral agreement with secondary objects.9 The secondary object of TA+O verbs and relative root obliques share the same patterning in the three languages examined in this paper: both kinds of arguments can be indexed by peripheral agreement in the Eastern languages but cannot be indexed in Ojibwe. Why we get the same pattern is because the theme and the relative root complement receive Case from the verb and originate low in the syntax.

|

Oblique Argument Licensing in Algonquian   327

Conclusion

This paper explores the patterns of oblique agreement, indicated by the n-endings and peripheral agreement, in three Algonquian languages. The first observation is that none of the examined oblique arguments give rise to agreement in Ojibwe. The second observation is, even though peripheral agreement in Eastern Algonquian languages can index the oblique arguments, an important restriction about where the limit lies is revealed. Concretely, case marking blocks NPs from being targeted by peripheral agreement. In accounting for the divergence regarding oblique agreement within the Eastern languages, the case type and the RR type are considered to reflect distinct morphological markings, which fundamentally differ in receiving Case in the syntax. This paper has contributed to extending the observations on the cline of peripheral agreement described in Xu (forthcoming) from core arguments to oblique arguments. In addition, this paper offers a novel analysis in connecting differential availability of peripheral agreement with oblique arguments more profoundly with the way they are derived in the grammar. Notes 1. Morpheme glosses follow the Leipzig Glossing Rules, with these additions: 3 = third proximate person, 0 = inanimate, 3´/obv = obviative, ai = intransitive animate, ai+o = transitivized intransitive animate, an = animate, ii = intransitive inanimate, in = inanimate, pret = preterite, ptcl = particle, rr = relative root, ta = transitive animate, ta+o = ditransitive with animate goal, ti = verb transitive inanimate. 2. Definiteness plays a role only in Unami Delaware because it preserves the ProtoAlgonquian parallel inflection paradigms where the definite object is specified by peripheral agreement while the indefinite is not. Such definiteness-driven paradigms are lost in the languages of Maliseet-Passamaquoddy and Ojibwe therefore their examples no longer show sensitivity to definiteness. 3. The research on subcategorization (Chomsky 1965; Kaplan and Bresnan 1982) and valency theory (Tesnière 1953) contribute to explaining the number as well as the type of arguments required by the predicate. 4. The umlauting -ẅ corresponds to Goddard’s (2007) |-w|m or Pentland’s (1999) -w’. 5. A class of derivational suffixes called instrumental finals (Valentine 2001:660) can express the meaning associated with instruments. However, I do not consider this type of affix as a comparable oblique argument-introducing strategy. First, they are

|

328    Yadong Xu

derivational morphology, contributing to the formation of the verb stem. Second, they do not introduce arguments and so do not participate in agreement (theme sign, central agreement, or peripheral agreement). 6. Rhodes (1990, 2010) shows that the animate relative root complements can be advanced on a par with the object, triggering the animate object agreement marker (theme sign). However, the examples he provides are in the conjunct inflection. It is unclear whether the same kind of gender-based advancement could take place in the independent inflection in which the n-endings and peripheral agreement occur. 7. The reflexes of *eθ- ‘thither, thus’ are: iN- in Nishnaabemwin (Valentine 2001:421), əl- or əš- in Unami Delaware (Goddard 2021:45), and oli- in Passamaquoddy (Bruening 2001:169). In Blackfoot, *eθ- is developed as itap- denoting the general allative meaning (Russell et al. 2012:65). 8. In Passamaquoddy, subjects at least c-command into locative phrases that are added by a relative root (Bruening 2001:170). 9. These two Eastern languages allow the collocation of n-endings and peripheral agreement to index secondary objects of both AI+O and TA+O verbs. However, Nishnaabemwin can index the secondary object of AI+O verbs but not that of TA+O verbs.

References Abels, Klaus. 2012. Phases: An Essay on Cyclicity in Syntax. Berlin: Walter de Gruyter. Andrews, Avery D. 2007. The Major Functions of the Noun Phrase. Language Typology and Syntactic Description, ed. by Timothy Shopen, pp. 132–223. Cambridge: Cambridge University Press. Bošković, Željko. 2014. Now I’m a Phase, Now I’m Not a Phase: On the Variability of Phases with Extraction and Ellipsis. Linguistic Inquiry 45(1):27–89. Bloomfield, Leonard. 1946. Algonquian. Linguistic Structures of Native America, ed. by Harry Hoijer, pp. 85–129. New York: Viking Fund Publications in Anthropology. Bruening, Benjamin. 2001. Syntax at the Edge: Cross-Clausal Phenomena and the Syntax of Passamaquoddy. PhD thesis, Massachusetts Institute of Technology. Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Comrie, Bernard. 1989. Language Universals and Linguistic Typology, 2nd edition. Chicago: University of Chicago. Dahlstrom, Amy. 2022. Meskwaki Syntax Book. Unpublished manuscript. https:// lucian.uchicago.edu/blogs/adahlstrom/publications-2/selected-manuscripts/

|

Oblique Argument Licensing in Algonquian   329

meskwaki-syntax-book/. Dixon, R. M. W. 1994. Ergativity. Cambridge: Cambridge University Press. Goddard, Ives. 1979. Delaware Verbal Morphology: A Descriptive and Comparative Study. New York: Garland. Goddard, Ives. 1990. Primary and Secondary Stem Derivation in Algonquian. International Journal of American Linguistics 56:449–483. Goddard, Ives. 2007. Reconstruction and History of the Independent Indicative. In Papers of the 38th Algonquian Conference, ed. by H. C. Wolfart, pp. 207–271. Winnipeg: University of Manitoba. Goddard, Ives. 2020. The Paradox of Partial Similarity. Unami Secondary-Object Marking and Other Clones of the N-endings. Papers of the 49th Algonquian Conference, ed. by Monica Macaulay and Margaret Noodin, pp. 99–113. East Lansing: Michigan State University Press. Goddard, Ives. 2021. A Grammar of Southern Unami Delaware (Lenape). Potoskey, MI: Mundart Press. Kaplan, Ronald, and Joan Bresnan. 1982. Lexical-Functional Grammar: A Formal System for Grammatical Representation. In The Mental Representation of Grammatical Relations, ed. by Joan Bresnan, pp. 173–281. Cambridge, MA: MIT Press. Legate, Julie Anne. 2008. Morphological and Abstract Case. Linguistic Inquiry 39(1):55–101. LeSourd, Philip S. Forthcoming. Explaining an Unusual Agreement Pattern in MaliseetPassamaquoddy. Papers of the 52nd Algonquian Conference, ed. by Monica Macaulay and Margaret Noodin. East Lansing: Michigan State University Press. Lochbihler, Bethany. 2012. Aspects of Argument Licensing. PhD thesis, McGill University. Marantz, Alec. 1993. Implications of Asymmetries in Double Object Constructions. Theoretical Aspects of Bantu Grammar, ed. by Sam A. Mchombo, pp. 113–150. Stanford: CSLI. Nichols, Johanna. 1986. Head-Marking and Dependent-Marking Grammar. Language 62:56–119. Nilsen, Campbell. 2017. The Historical Development of the Menominee Independent Indicative. Undergraduate senior thesis, University of Oklahoma. Pentland, David H. 1999. The Morphology of the Algonquian Independent Order. Papers of the 30th Algonquian Conference, ed. by David H. Pentland, pp. 222–266. Winnipeg: University of Manitoba. Polinsky, Maria, and Omer Preminger. 2014. Case and Grammatical Relations. Routledge Handbook of Syntax, ed. by Andrew Carnie, Dan Siddiqi, and Yosuke Sato, pp. 150–166. New York: Routledge. Rhodes, Richard A. 1990. Ojibwa Secondary Objects. Grammatical Relations: A Cross

|

330    Yadong Xu

Theoretical Perspective, ed. by Katarzyna Dziwirek, Patrick Farrell, and Errapel MejíasBikandi, pp. 401–414. Stanford: CSLI. Rhodes, Richard A. 1998. Clause Structures, Core Arguments and the Algonquian Relative Root Constructions. Winnipeg: Voices of Rupert’s Land. Rhodes, Richard A. 2010. Relative Root Complement: A Unique Grammatical Relation in Algonquian Syntax. Rara and Rarissima, ed. by Jan Wohlgemuth and Michael Cysouw, pp. 305–324. Berlin: De Gruyter Mouton. Russell, Lena, Inge Genee, Eva van Lier, and Fernando Zúñiga. 2012. Referential Hierarchies in Three-Participant Constructions in Blackfoot: The Effects of Animacy, Person, and Specificity. Linguistic Discovery 10:55–79. Sherwood, David Fairchild. 1983. Maliseet-Passamaquoddy Verb Morphology. PhD thesis, Yale University. Tesnière, Lucien. 1953. Esquisse d’une syntaxe structurale. Paris: Klincksieck. Valentine, J. Randolph. 2001. Nishnaabemwin Reference Grammar. Toronto: University of Toronto Press. Wolfart, H. Christoph. 1973. Plains Cree: A Grammatical Study. Transactions of the American Philosophical Society, new series, 63(5):1–90. Philadelphia: American Philosophical Society. Vinokurova, Nadya. 2005. Lexical Categories and Argument Structure: A Study with Reference to Sakha. PhD thesis, Utrecht: LOT Dissertation Series: UiL-OTS. Xu, Yadong. Forthcoming. A Survey of the Patterning of Peripheral Agreement. Papers of the 52nd Algonquian Conference, ed. by Monica Macaulay and Margaret Noodin. East Lansing: Michigan State University Press.

Contributors

Irene Appelbaum received a PhD in philosophy and an MA in linguistics from the University of Chicago. She is professor of linguistics at the University of Montana in Missoula. Her research focuses on Ktunaxa language documentation and linguistics. Antti Arppe is an associate professor of quantitative linguistics at the University of Alberta, as well as the founder of the Alberta Language Technology Lab and the partnership director of the SSHRC-funded partnership “21st Century Tools for Indigenous Languages.” Ksenia Bogomolets (PhD 2020, University of Connecticut) specializes in phonology and morphology of polysynthetic languages. She is particularly interested in theoretical issues pertaining to morpho-phonology of Algonquian languages. On the empirical side, she has a keen interest in documentation and revitalization of understudied and endangered languages, and she has conducted fieldwork with speakers of Coorgi (Dravidian), Kyrgyz (Turkic), and Arapaho (Plains Algonquian). She currently holds an honorary research fellow position at the linguistics

|

   331

|

332   Contributors

department of the University of Auckland, and she holds the position of principal language planning advisor at the te Reo Māori Commission in Wellington, New Zealand. Maggie Chittspattio is from the Naskapi Nation from Kawawachikamach. She is a proud mother of three that keeps her going in life, and after her dreams. Maggie Chittspattio is currently an intervention worker and team leader at Resilience Montreal. Vincent Collette received his PhD in linguistics from Université Laval. He is regular professor of linguistics at Université du Québec à Chicoutimi. His fields of research include the semantics, morphology, language description, as well as the history of Algonquian (Cree, Innu) and Siouan (Assiniboine, Stoney) languages. David J. Costa is the program director for the Language Research Office at the Myaamia Center at Miami University of Ohio. He grew up in the San Francisco Bay Area and completed his PhD in linguistics at University of California, Berkeley in 1994. He has been studying the Miami-Illinois language since 1988. Daniel Dacanay is a graduate student in the Department of Linguistics at the University of Alberta, whose research primarily focuses on lexical semantics and documentation in the context of Plains Cree. Amy Dahlstrom is an associate professor of linguistics at the University of Chicago. Her research deals with issues of morphology, syntax, and information structure in Meskwaki and Plains Cree. Jaymee Dhein graduated from University of Wisconsin–Milwaukee with a BA in linguistics and German studies. She is now continuing her studies at the same university, pursuing an MA in translation and interpreting studies. Her focus is on German to English translation with a particular interest in literary translation. Nadia Ferrara is currently the senior director for Values, Ethics, Diversity and Inclusion at Parks Canada, and remains on faculty at McGill University as adjunct professor in the Department of Anthropology. Before entering public service in 2003, Nadia worked as an art therapist for sixteen years, specializing in cross-cultural

|

Contributors   333

psychotherapy with Indigenous Peoples in Quebec and Ontario. Nadia is also a published author and artist. Amélie La Rose Gauthier is Abenaki from Odanak and mother of two young children. She graduated from University of Ottawa with a BA in psychology and is working as a privacy advisor within the federal government. She is currently completing her MA in public ethics at Saint Paul University with her main research interests being data governance, indigenization of AI, and critical theories. Ashley Guanish is an Indigenous woman from the Naskapi Nation of Kawawachikamach. She is passionate about language, community, traditional crafts and lifelong learning. Ashley is determined for her children to appreciate the strength of being Naskapi and to respect being raised by a matriarch of that Nation. Christopher Hammerly is a mixed Anishinaabe-Norwegian man and a descendant of the White Earth Nation in Minnesota who grew up on Dakota lands and currently lives on Musqueam lands. He is an assistant professor of linguistics at University of British Columbia. Mary Jane Hannaburg, a Kanien’kehá:ka woman from Kanehsatake. She is an active member of her community working as a mental health support worker, a registered addictions specialist, certified crisis worker and a SafeTalk trainer. Mary has a background in advocacy being actively involved in the Quebec Native Women movement for a number of years, and most recently completed three consecutive mandates as the vice president for the organization. Anahi Morales Hudon is an associate professor in the School of Social Innovation at Saint Paul University. She holds a PhD in sociology from McGill University and specializes in the study of social movements, intersectionality, organizational dynamics and transformation of power relations. She has published in different journals such as the Canadian Journal of Political Science and the Journal of Latin American Studies. Hunter Johnson is of Norwegian and Swedish descent and grew up on Dakota and Ojibwe lands in Minnesota. He is currently a linguistics PhD student at the University of California, Los Angeles.

|

334   Contributors

Denise Larocque is a proud Mi’gmaq 2spirit woman from Gesgapegiag, a mother and grandmother. She sat on the Board of Quebec Native Women for several terms. She is a contributing author, Kasala poetry teacher, and an addiction counselor. Her passion has been to support families as a traditional birth attendant, while also speaking publicly against violence of women. Angela Mesic is an associate lecturer of the first-year Ojibwemowin courses at the University of Wisconsin–Milwaukee and University of Wisconsin–Superior. She obtained her bachelor’s degree in the field of psychology and is currently working on a master of community psychology. At the Electa Quinney Institute for American Indian Education, she manages several projects in language revitalization with regional partners including the Indian Community School, several tribal nations, and universities throughout the United States. Mizuki Miyashita is a professor of linguistics at the University of Montana. The focus of her research is documentary linguistics in Blackfoot. She has recorded lullabies and other songs, narratives, conversations, and isolated words in Blackfoot. She engages in community-based research and her projects involve documentation and application of Blackfoot prosody. Anny Morissette is an assistant professor in the School of Conflict Studies at Saint Paul University. She holds a PhD in anthropology from Université de Montréal. An Indigenous studies specialist, her areas of research, publication and teaching focus on First Nations histories, politics, religious pluralism, Indian Day schools as well as the empowerment of First Nations women. Burak Oney is a second-year PhD student at Cornell University. His research focuses on theoretical and experimental syntax, and he is mainly interested in scrambling. He received his MA degree from Memorial University in 2021, where he worked on the acquisition of preverbs in Northern East Cree for his master’s thesis under the supervision of Julie Brittain. Marie-Louise Perron was born on her grandfather’s land in Saskatchewan and is a descendant of the Red River Métis and early French newcomers. She is a Métis Knowledge Keeper currently serving Saint Paul University in that capacity. Through

|

Contributors   335

several different careers, as a high school teacher, visual artist, author, archivist, and public servant, she has maintained the storytelling tradition of her people. Jolene Poulin is a software developer with the Alberta Language Technology Lab, whose work includes the production of a wide variety of computational tools and resources for Canadian Indigenous languages. Richard A. Rhodes is emeritus professor of linguistics at the University of California, Berkeley, interim director of Canadian Studies, and Thomas Garden Barnes Chair of Canadian Studies. He has studied Ojibwe intensively throughout his career. He has also contributed significantly to the documentation of other endangered languages, primarily Métchif and the Mexican language Sayula Popoluca. The focus of his work has ranged from morphology and descriptive syntax to lexical semantics and lexicography. He is currently working on historical linguistic questions in Algonquian and on cognitive approaches to lexical semantics. Mskwaankwad Rice is an L2 speaker/learner of Nishnaabemwin from Waasaaksing First Nation. He is a cofounder of the Eshki-nishnaabemjig immersion program and has worked on language initiatives in his home community, such as with on-air and online audio and video lessons for CHRZ Rez 91, and elsewhere in Ojibwe country. Mskwaankwad is currently undertaking a PhD in linguistics at the University of Minnesota and he seeks to apply his knowledge of linguistics to Nishnaabemwin/ Ojibwe reclamation efforts in order to better serve the movement as a whole. Katherine Schmirler recently completed a PhD in linguistics at the University of Alberta. She is involved with the Alberta Language Technology Lab’s 21st Century Tools for Indigenous Languages project, and is currently undertaking postdoctoral research on the computational modeling of Blackfoot at the University of Lethbridge. Susan Wade received a PhD in history from the University of Wisconsin–Milwaukee. Her dissertation focuses on Indigenous women and maple sugar production complementing an expertise in food history. She has taught introductory Ojibwemowin for Michigan State University and the University of Wisconsin–Milwaukee.

|

336   Contributors

Natalie Weber is an assistant professor in the Department of Linguistics at Yale University. Their main interests lie in phonology and its interfaces with other components of grammar. Since 2011, their empirical focus has been Blackfoot, an Algonquian language spoken in Alberta and northern Montana. Matthew Windsor is a linguist at SIL International and member of Wycliffe Bible Translators Canada. Since 2018 he has been working in the Oji-Cree community of Kingfisher Lake (Ontario) as resident linguist at Mishamikoweesh Corporation, where he provides support for local language and culture initiatives. Yadong Xu is a linguist whose research focuses on Algonquian agreement and cross-linguistic patterns. She earned her PhD in linguistics from the University of Manitoba in February 2023 and currently works at the Manitoba First Nations Education Resource Centre. Xu’s research explores the complex structures of Algonquian languages and the ways in which agreement works within these systems. Her work is grounded in the theory of generative linguistics and draws empirical evidence from multiple Algonquian languages, including Plains Cree, Ojibwe, Unami Delaware, and Maliseet-Passamaquoddy.