Lexical and Structural Etymology: Beyond Word Histories 9781614510581, 9781614510598

Citation preview

Robert Mailhammer (Ed.) Lexical and Structural Etymology

Studies in Language Change

Edited by Cynthia Allen Harold Koch Malcolm Ross

Volume 11

Lexical and Structural Etymology Beyond Word Histories

Edited by Robert Mailhammer

ISBN 978-1-61451-059-8 e-ISBN 978-1-61451-058-1 ISSN 2163-0992 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.dnb.de. 6 2013 Walter de Gruyter, Inc., Boston/Berlin Cover image: iStockphoto/thinkstock Typesetting: RoyalStandard, Hong Kong Printing and binding: Hubert & Co. GmbH & Co. KG, Göttingen ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Contact details and affiliations of contributors Robert Mailhammer, University of Western Sydney [email protected] Luise Hercus, Australian National University [email protected] Harold Koch, Australian National University [email protected] Meredith Osmond, Australian National University [email protected] Patrick McConvell, Australian National University [email protected] Theo Vennemann, Ludwig-Maximilians Universität München [email protected] Alexandre François, CNRS–LACITO; Labex ‘Empirical Foundations of Linguistics’ Australian National University [email protected] Rachel Hendery, Australian National University [email protected] David Nash, Australian National University [email protected] John Giacon, Australian National University [email protected]

Abbreviations AAWE ACC Adm Aly AND ART AW BW C CP CPLT CPLZR DAT DEP Di DRG DU E ECAr ERG eZ F FB FBS fDC FF FFZ FM FMB FOC Fr FUT FZ FZD G GEN GR

Australian Aboriginal words in English accusative Admiralties Alyawarr The Australian National Dictionary article Arabana-Wangkangurru brother’s wife consonant complementizer phrase complete aspect complementizer dative dependent clause Diyari Dorig dual English Eastern and Central Arrenrnte ergative older sister Feminine father’s brother father’s brother’s son female ego’s daughter’s child father’s father father’s father’s sister father’s mother father’s mother’s brother focus marker French future father’s sister father’s sister’s daughter German genitive Gamilaraay

viii Hiw HRLM HZ Ka Kay KKY KLY KRO LHI LKN LMG LOC LTG LYP M M/MASC MB MBS mDC MFMD MFZ MFZDD Mi MM MM MMB MMBDD MMMZD MMZ MRL MSN MTA MTP MZ N NCV NE NEG Ng NNG

Abbreviations

Hiw Hunter River and Lake Macquarie husband’s sister Karuwalim Kaytetye Kalaw Kawaw Ya Kala Lagaw Ya Koro Lehali Lakon Lemerig locative Lo-Toga Löyöp mother masculine mother’s brother mother’s brother’s son male ego’s daughter’s child mother’s father’s mother’s daughter mother’s father’s sister mother’s father’s sister’s daughter’s daughter Mithaka Meso-Melanesian mother’s mother mother’s mother’s brother mother’s mother’s brother’s daughter’s daughter mother’s mother’s mother’s sister’s daughter mother’s mother’s sister Mwerlap Mwesen Mota Mwotlap mother’s sister noun North-Central Vanuatu Northeastern negation Ngamini North New Guinea

Abbreviations

NOM NP NPNy NUM Obj OE OED PAN/PAn PAST pCK PGmc. PIE Pitj PL/pl PN Pn PNCV POc POSS POT Pp pPN PRES PRF PROH PT Pun pWK RC REFL SE SES Sg/sg Subj SUBJ V VLW VRA VRS WAr

nominative noun phrase Non-Pama-Nyungan Nume object Old English Oxford English Dictionary Proto-Austronesian past tense Proto-Central Karnic Proto-Germanic Proto-Indo-European Pitjantjatjara plural Pama-Nyungan Polynesian Proto-North Central Vanuatu Proto-Oceanic possessive potential Pitta-Pitta Proto-Pama-Nyungan present tense perfect prohibitive Papuan Tip Punic Proto-Western Karnic relative clause reflexive Southeastern Southeast Solomonic singular subject subjunctive vowel/verb Volow Vera’a Vurës Western Arrernte

ix

x WM Wm WN WTS Ya YG Yi YR Yw YY

Abbreviations

wife’s mother Wangkumara Wangaaybuwan Western Torres Strait Yandhrundwandha Yuwaalaraay Gamilaraay Yarluyandi Yuwaalaraay (YR) Yawarrawarrka Yuwaalayaay

Table of contents Contact details and affiliations of contributors Abbreviations vii

v

Robert Mailhammer Introduction: Etymology beyond word histories Robert Mailhammer Towards a framework of contact etymology

1

9

Harold Koch and Luise Hercus Obscure vs. transparent cognates in linguistic reconstruction

33

Harold Koch The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered Rachel Hendery Constructional etymology: The sources of relative clauses Theo Vennemann Concerning myself

83

121

Patrick McConvell Granny got cross: semantic change of kami ‘mother’s mother’ to ‘father’s mother’ in Pama-Nyungan 147 Alexandre François Shadows of bygone lives: The histories of spiritual words in northern Vanuatu 185 Meredith Osmond Markers of the spirit world in Oceanic languages John Giacon Etymology of Yuwaalaraay Gamilaraay bird names

245

251

David Nash The smuggled budgie: case study of an Australian loanblend

293

Luise Hercus Archaisms in placenames in Arabana-Wangkangurru country

313

Subject index

323

53

Robert Mailhammer

Introduction: Etymology beyond word histories 1 Etymology and historical linguistics Etymological research naturally is fundamental to the historical investigation of a language. Etymologies permit generalizations about historical developments, for instance the formulation of sound laws, or about a particular synchronically attested phenomenon, such as the occurrence of a particular type of stem formation. Moreover, etymologies are also used as arguments to answer questions about historical relationships among languages. Consequently, the historical investigation of a language generally begins with a comparative analysis of single lexemes, which leads to etymologies via the reconstruction of a protoform. Etymologies then serve as basis for systematic generalizations. Conversely, etymological data are also used to verify hypothesized historical connections. In these cases etymologies do not serve a self-fulfilling purpose, i.e. to find out about the origin of a word, rather they are arguments in a historical discussion. As a result, etymology can be regarded as something like a fundamental auxiliary discipline of historical linguistics. However, etymological research can be focused more narrowly on aspects of historical word formation. From this perspective, more general statements about historical developments in a language are not explicitly part of the investigation. Rather, the aim is to pinpoint the origin of a word, its Urschöpfung, as Seebold (1980: 431) points out. Wenn wir ein Wort etymologisch untersuchen, dann richten wir unser Hauptaugenmerk auf seine Entstehung: Wir suchen zu zeigen, daß es aus einem nachweisbaren Grundwort nach dem Muster eines ebenfalls nachweisbaren Wortbildungstyps geprägt wurde. Solange der Zusammenhang zwischen Grundwort und abhängigem Wort noch voll durchsichtig ist, gehört diese Fragestellung in die Wortbildungslehre der betreffenden Sprachstufe; sobald aber geschichtliche Belege und geschichtliche Überlegungen eine Rolle spielen, befinden wir uns im Bereich der Etymologie. (Seebold 1980: 431) [If we examine a word etymologically, we focus on its origin. We attempt to show that it was coined based on an evidenced source lexeme following the model of a type of word formation that is likewise attested. As long as the connection between the base word and the derived word is still fully transparent, this is a problem belonging to the domain of word formation of the language examined; as soon as historical attestations and historical considerations come to play a role, we are in the domain of etymology.] 1

1 All translations are mine unless indicated otherwise.

2

Robert Mailhammer

Usually, such studies are undertaken once sufficient historical knowledge has been acquired, as the investigation of a word’s history is based on generalizations and observed developments. However, detailed studies of a word’s history are subsequently used to test hypothesized generalizations, which in turn leads to improved etymologies and so forth. As a result, there are in principle two applications of etymological research, which differ in their immediate aims. One is focused on the linguistic history of a language and the other on the linguistic history of a linguistic element. In practice, however, these two perspectives cannot always be strictly separated, because every etymology naturally points beyond the mere origin of a word. But conceptually it makes sense to differentiate between the two, without claiming that this separation is necessary rigid in practice.

2 The etymological approach The separation between etymology as an auxiliary discipline of historical linguistics and a discipline in its own right focusing on the origin of one item rather than on the history of a language highlights a methodological difference, which can be captured in the notion of the etymological approach. Describing the history of a language or even a word – in the tradition of histoire des mots (see Mailhammer 2007: 142–143) – assumes a perspective forward in time, from the point of origin to the present. By contrast, etymology as the investigation of the point of origin assumes a perspective backwards in time, a “retrograde” perspective (Mailhammer 2007: 143). Its purpose is not to explain the present directly through history but indirectly, by showing how and when something originated. A historical study would be focused on the question “how did this word arrive at this particular form and meaning?”, with associated implications about formal and semantic changes that make statements about the history of a language. For instance, the formal changes that have been captured in the formulation of the High German sound shift and the semantic developments since the earliest attestations would be foregrounded. By contrast, an etymological study in the spirit of the “etymological approach” is concerned first and foremost with the question “what is the origin of a word?”. That both perspectives are complementary and may in fact incorporate the same information is correct but irrelevant in this connection, because the purpose of the investigation is different. A historical investigation aims at finding out about the present or a development from a chronologically earlier to a chronologically later stage of a language, whereas an etymological investigation primarily is about getting to the point of where something comes into being. Of course, the historical perspective should incor-

Introduction: Etymology beyond word histories

3

porate this moment as starting point of the history, but etymology in this narrow sense is not concerned with the more recent history, but only with a point in time, it is in fact possibly even ahistorical. Etymological research starts from the earliest attestation and works its way back in time (if necessary), whereas a historical investigation works its way forwards from the point that etymological research established. Thus the etymological approach can be defined as a quest for the origin of something rather than its history. It can be extended beyond linguistics. For instance, a linguistic etymology of the English verb bear will give the information that this word was formed in Proto-Indo-European as a thematic present with the meaning ‘carry’, but it is less concerned with subsequent changes. In a similar vein, a non-linguistic etymological investigation may ask about the origin of a historical event, such as the Second World War, or a political entity, such as the British Empire, without concern for subsequent developments. Of course, the etymological approach feeds on historical information, but its goal is different from a historical approach. This is why etymology is both the backbone of historical linguistics and an application at the same time.

3 Lexical and structural etymology Traditionally, etymology has been mainly perceived as being concerned with words, but there is no reason why this should necessarily be the case. In principle anything can be studied etymologically, simply by asking the question, “where did that come from?”. The items investigated may be a phoneme or an entire text, and thus the question might be considered part of different disciplines, not even linguistic ones in the case of texts. Thus, one might wonder where affricates in High German dialects come from, what the origin of the 3sg present tense marker of Modern English full verbs is, what the origin of Fr guerre is, or when the plot of Shakespeare’s The Merchant of Venice originated. As a result, it seems unproblematic to extend the notion of linguistic etymology and etymological research beyond words and word histories to anything that can be linguistically described. From a historical perspective, linguistic units with a relatively high time stability may appear more interesting than more short-lived entities that can be created spontaneously, such as sentences. But this is clearly a matter of personal taste rather than of principle. As suggested in its title, this volume explicitly attempts to go beyond the more traditional notion of etymology as focusing on words and their history in extending the etymological approach to linguistic units outside the classical lexical domain. In two senses these extensions can be seen as structural rather

4

Robert Mailhammer

than lexical. First, constructions, such as relative clauses, are defined by their structure, and the origin of such a structure is the object of a subfield of etymology that can be called “structural etymology”. Second, structural components of language, for instance, grammatical morphemes, such as pronouns, TAM markers, and periphrastic forms, e.g. the English passive, can be investigated etymologically by applying the same research strategy. Many of the studies assembled in this volume follow one of these two pathways beyond classical lexical etymology and apply the etymological approach to structures or structural units. To sum up, the etymological approach can be applied to any linguistic entity, phonemes, lexical items, morphemes and phrasal expressions and other phraseological units and even sentences (e.g. as constructions in the sense of e.g. Construction Grammar, see Croft 2001 and Hendery, this volume). It is the focus on the origin that defines the etymological perspective rather than the subject of investigation. However, for the sake of systematicity and clarity a subdivision of etymology into lexical and structural etymology is advocated in this book.2

4 The studies in this book The etymology of this book – in the sense of the concept developed in the preceding sections – is a symposium at the Kioloa campus of the Australian National University in April 2010, the First Kioloa Symposium on Etymology. The idea was to bring together researchers from vastly different linguistic backgrounds, language focus and linguistic orientation but united in their interest in investigating the origin of phenomena they encountered in the languages they studied. This mix spawned interesting discussions about pathways of investigation, about methodologies, and about the origins and the genesis of words and structures in very different languages ranging from Indo-European to Austronesian and Australian languages. All contributions to this volume underwent a peer review process. A substantial amount were presented at the First Kioloa Symposium on Etymology, though some were solicited by invitation. In two cases presenters submitted papers different to their conference papers, and one contribution had to be withdrawn just before the volume went to press. As a result, this book comprises eleven chapters, each of which illustrates the etymological approach from different viewpoints and with different phenomena in different languages.

2 To my knowledge, the terms “structural etymology” and “lexical etymology” were coined by Theo Vennemann (2000); see discussion in Mailhammer (2007: ch. 3).

Introduction: Etymology beyond word histories

5

The chapter by the editor opens the volume outlining a framework for proposing loan etymologies that aims at removing a frequently observed bias against such etymologies by proposing a transparent and rigorous set of guidelines. This takes up a little-noticed message from a seminal paper by Milroy and Milroy (1985), which is that what is “normal transmission” of a language depends greatly on the situation, and cannot always be equated with trans-generational transmission within a homogenous community. There are situations in which a contact-influenced transmission is actually more probable, and this needs to be reflected in etymological investigation. In a similar vein, chapter two, by Harold Koch and Luise Hercus, proposes a break with tradition in an effort to achieve a breakthrough in etymological research in Australian languages, which requires a greatly increased pool of cognates. Reworking a principle by Oswald Szemerényi, Koch and Hercus develop the point that the reflexes of an established protoform may not necessarily be attested with the same semantics or formal characteristics. Chances are that there are cognates that appear in disguised shape or with a shifted meaning, and incorporating these into etymological equations results in a much bigger number of cognates. Accordingly, the task is not to find only the semantically equivalent and transparent cognates but all cognates. The third paper (by Harold Koch) is an application of the etymological approach focusing on the etymology of the 3sg feminine pronoun in the PamaNyungan languages, the largest Australian language family. Koch illustrates not only his method of morphological reconstruction (Koch 1996 et passim), but also showcases etymology going beyond words in uncovering the intricate shifts in meaning of grammatical forms and the formal developments of morphological markers from syntactic constructions and the absorption of morphological elements into lexical stems. One important methodological question Koch’s paper raises is how many cognates and in what distribution across a family are needed in order to “nail” an etymology, i.e. to demonstrate that the forms under comparison are more likely to reflect common ancestry than borrowing or chance. In her contribution, Rachel Hendery presents her take on structural etymology. She uses a construction-grammar based approach to trace and typologize developmental paths of relative clauses in a range of languages. In the spirit of sections 2 and 3 above, she focuses explicitly on the origin of constructions in an etymological sense. Chapter five is situated at the juncture between lexical and structural etymology, as it tracks down the origin of a word as well as that of a paradigm beyond what is common in etymology, namely across different languages. Theo Vennemann homes in on the history of the English reflexive pronoun, myself, yourself, etc., for which recent literature has suggested an origin under Celtic influence.

6

Robert Mailhammer

Now, for Vennemann this is not the end of the story, because one has to wonder why Celtic expressed the reflexive in such a way. It turns out that among the Indo-European languages only English and Celtic share this kind of reflexive, and Vennemann proposes that in Celtic too, it is the result of language contact, namely with Phoenician-Punic. This is an illustration of what Vennemann (2002) called the “transitivity of language contact”. It shows that etymological research does not stop within one language, but that in the case of a contact etymology one has to pursue the matter further. Chapter six, by Patrick McConvell, illustrates the key point made by Koch and Hercus, namely that cognates can survive with drastically different semantics, in this case the shift from ‘mother’s mother’ to ‘father’s mother’. In Australian kinship systems the conceptual difference between the patriline and the matriline is highly significant, so such a shift is by no means trivial. That it can happen nevertheless demonstrates the validity of Koch and Hercus’s etymological research principle. The next two papers, by Alexandre François and Meredith Osmond respectively, investigate pathways of semantic change in order to reveal something about words connected to the spirit world in Oceanic languages. François assembles an array of languages in detail in order to systematize avenues of semantic change in this semantic domain. Similarly, but within the limit of a much smaller paper, Osmond picks up a suggestion by Robert Blust on the origin of spirit marker terms in Oceanic languages and ties it together from an Austronesian perspective. In his contribution, John Giacon, assembles a detailed summary of etymologies of bird names in a group of Australian languages, and gives a rich appendix illustrating the need of philologically fine-grained sifting through the evidence in order to arrive at sound conclusions about the origin of linguistic items. The next chapter by David Nash zooms in on the name on one particular bird, the budgerigar. In minute detail Nash shows the path to discovering how this word came into English, positing that it is a case of loan blend rather than the borrowing of a word from one language. The final paper in the volume is by Luise Hercus. Delving into her intimate knowledge of Arabana-Wangkangurru country, she shows how cognates have survived in place names in non-transparent form, but as valuable evidence in historical research, thus showing another successful application of the etymological research principle developed in chapter two. The studies assembled here demonstrate the essence of the etymological approach and its relationship to historical linguistics, as outlined in sections 1 to 3 above. They show that a question about the origin of something is bigger than the etymology of a word, that in fact it is an investigative perspective in

Introduction: Etymology beyond word histories

7

its own right, and that it is in a symbiotic relationship with the question about why languages are the way they are, which crucially goes beyond a mere description, and can frequently only be answered from a historical viewpoint.

5 Acknowledgments This book would not have been possible without the kind and efficient help of a great many people. I am very grateful to the series editors, Cynthia Allen, Harold Koch and Malcolm Ross, for opening up the series Studies in Language Change, and to everyone at De Gruyter Mouton, especially Emily Farell, Wolfgang Konwitschny and Uri Tadmor, for their patience and all their help. My heartfelt gratitude also goes out to the reviewers, who did a great job in helping to ensure the academic quality of the contributions. Last, but not least, I would like to thank all the contributors for the insights they offer to the linguistic community, and for their patience and co-operation. Finally, thanks also to my family for putting up with the many hours behind the desk and for their moral support.

References Croft, William. 2001. Radical Construction Grammar. Syntactic Theory in Typological Perspective. Oxford: OUP. Koch, Harold. 1996. Reconstruction in Morphology. In: Mark Durie & Malcolm Ross (eds.), The Comparative Method Reviewed. Regularity and Irregularity in Language Change, 218–263. New York & Oxford: Oxford University Press. Mailhammer, Robert. 2007. The Germanic Strong Verbs: Foundations and Development of a New System (Trends in linguistics Studies and monographs 183). Berlin/New York: Mouton de Gruyter. Milroy, James & Leslie Milroy. 1985. Linguistic change, social network and speaker innovation. Journal for Linguistics 21. 339–384. Seebold, Elmar. 1980. Etymologie und Lautgesetz. In: Manfred Mayrhofer, Manfred Peters & Oskar E. Pfeiffer (eds.), Lautgeschichte und Etymologie: Akten der VI. Fachtagung der Indogermanischen Gesellschaft, Wien, 24.–29. September 1978, 431–484. Wiesbaden: Dr. Ludwig Reichert. Vennemann, Theo. 2000. Zur Entstehung des Germanischen. Sprachwissenschaft 25. 233–269. Vennemann, Theo. 2002. Semitic → Celtic → English: The transitivity of language contact. In: Markku Filppula, Juhani Klemola & Heli Pitkänen (eds.), The Celtic Roots of English (Studies in Languages 37), 295–330. Joensuu: University of Joensuu, Faculty of Humanities.

Robert Mailhammer

Towards a framework of contact etymology* 1 Introduction Analogically speaking, an etymology is like a biography. It tells the story of a word from when it came first into existence until its most recent attestation, which can be in the past or the present. It includes when and how, as well as semiotic aspects of it (form, meaning, combinatorics), changed, and ultimately involves the reconstruction of its original creation (Urschöpfung), beyond its earliest attestation, if necessary. Wenn wir ein Wort etymologisch untersuchen, dann richten wir unser Hauptaugenmerk auf seine Entstehung: Wir suchen zu zeigen, daß es aus einem nachweisbaren Grundwort nach dem Muster eines ebenfalls nachweisbaren Wortbildungstyps geprägt wurde. Solange der Zusammenhang zwischen Grundwort und abhängigem Wort noch voll durchsichtig ist, gehört diese Fragestellung in die Wortbildungslehre der betreffenden Sprachstufe; sobald aber geschichtliche Belege und geschichtliche Überlegungen eine Rolle spielen, befinden wir uns im Bereich der Etymologie. (Seebold 1980: 431) [If we examine a word etymologically, we focus on its origin. We attempt to show that it was coined based on an evidenced source lexeme following the model of a type of word formation that is likewise attested. As long as the connection between the base word and the derived word is still fully transparent, this is a problem belonging to the domain of word formation of the language examined; as soon as historical attestations and historical considerations come to play a role, we are in the domain of etymology.] 1

There are two basic etymological configurations. Either a word is native in the sense that it was created at a chronologically earlier stage of a language and handed down from one generation to the next, or it came into a language from another through some form of language contact.2 From a purely historical perspective this may not be relevant, because a borrowed word was also created at some point, albeit in a different language. Hence, the history of a word may involve two or more different languages, but ultimately the moment of creation is the starting point of that history. But from an investigative viewpoint the difference is crucial, because a word with a native etymology was always transferred

* I would like to thank Marc Pierce (University of Texas at Austin) for helpful comments. All remaining errors are my own. 1 All translations are mine unless indicated otherwise. 2 The first possibility also comprises onomatopoetic formations.

10

Robert Mailhammmer

by native speakers of one language to native speakers of the same language, or it was always used predominantly by native speakers; it is characterized by a more or less homogenous ecology. By contrast, a word with a contact etymology was at one or more points transferred to or used predominantly by non-native speakers (or at least bilingual speakers). If words change through the transfer from one generation to another or through the use of words by native speakers of one language, this is all the more true for situations of transfer to speakers of one or more different languages. It is a like a game of Chinese Whispers: every time a little bit of the message is changed, but considerably more so each time one player has no or little command of the language of the message. While native transmission can often display surprising results and does not always follow completely regular patterns, this is all the more true for contact transmission, where not only unexpected sound substitutions and unexpected ways of reanalysis occur, but where there can be a great amount of inconsistency involved (see e.g. the recent collection of papers in Calabrese and Wetzels 2009). The methodological consequence is that native transmission should be kept distinct from contact transmission. Although the fact that a transfer effect, i.e. a change, occurs is similar; the existence of a language contact situation adds another dimension of considerable complexity. As a result, a contact etymology is likely to be more complex and more volatile, and may seem less compelling than a native etymology that is based on relatively well-established sound laws. But this does not at all mean contact etymologies are pure speculation. As I will argue in this paper, there are clear methodological guidelines that should be followed if a convincing contact etymology is to be proposed. Most significantly, this involves a fine-grained linguistic and extralinguistic investigation of the contact situation with the aim of establishing at least a high degree of plausibility for the proposed outcome of the contact transmission if there is no direct evidence. If this is right, then contact etymologies should be able to get rid of a certain degree of speculative stigma that they still frequently seem to carry, and bring to bear their full explanatory potential. This paper is structured as follows. Section two discusses the question when the investigation of a contact etymology is warranted by the data and research history. Section 3 introduces the so-called Blueprint Principle, which is used to reconstruct possible contact languages in situation in which the contact language is unknown. In section 4 the core elements of a contact explanation are discussed, and section 5 outlines a methodological framework for the position of contact etymologies. The main points of this paper are summarized in section 6.

Towards a framework of contact etymology

11

2 When to propose a contact etymology The considerations presented in this and the following sections form the methodological foundation of an etymological sub-discipline, which could be called contact-etymology (see also Mailhammer 2010). At the beginning of the investigation a linguistic element (i.e. a word or a particular structure) is identified for which a contact-induced origin is suspected.3 However, it is important to emphasize that this does not automatically presuppose that such an element is unexplained. There may already exist a different explanation regardless of its quality. It is well known that historical linguists often consider an internal development the privileged null hypothesis, probably because it is assumed that normal transmission (in the sense of Thomason and Kaufman 1988), i.e. without break or interference from one generation to the other, is indeed the normal case. Frequently, it is advocated to consider non-internal explanations only if internal attempts have failed to account for the origin of a word or structure (see e.g. Thomason and Kaufman 1988: 57–62 and Farrar and Jones 2002, as well as Filppula 2003: 161–163 for critical discussions). And if a development has two potential explanations, the internal one is typically considered the more convincing. I will call this view the Internal Development Bias, and argue that it is in fact a fallacy, because what is “normal transmission” actually depends on the situation a given language is in. Under the Internal Development Bias “normal transmission” is equated with “transmission in a quarantined community.” It has long been known that language cannot be considered separate from its speakers. Language change is first of all an innovation by one or more speakers (Milroy and Milroy 1985). It is somewhat obscure why a speaker with connections to other communities, perhaps even with knowledge of other languages should innovate predominantly only using the material from their own native language. Likewise, it is implausible that a child exposed to other languages does not use this knowledge. In fact, children growing up bilingually demonstrate transfer from one into the other language on a daily basis. If a German-English bilingual child forms the word butter bread, it is clear that this could be an English-internal innovation. The rules of English word formation allow this innovation and this may actually exist somewhere as completely internal innovation. But it is also clear that the fact that there is an exact blueprint for this in G Butterbrot, which this particular child also knows, cannot go unnoticed. It seems misguided to 3 As argued in Mailhammer (2007: 144) and in the introduction to this volume, words as well as structures can be investigated etymologically, in the sense that their origin is examined. Hence, one could differentiate between lexical and structural etymology.

12

Robert Mailhammmer

suggest that external influence, in this case from German (on English), has to be disregarded as secondary possibility, only because butter bread could be an English word, and could therefore have arisen without influence from German. Research on bi-/multilingual first language acquisition has shown that interference between the available languages is normal and frequent (see e.g. Genesee and Nicoladis 2007 with references). This does not mean that butter bread has to be due to German influence, but it is likely or at least possible that the contact situation played a role in the formation of this word. It is in fact likely that this is a case of multiple causation; but to rule out language contact just because a completely internal development is possible is exactly what I call it here, a bias. This becomes especially clear from considering a hypothetical situation in which a monolingual child without any contact to a language with a blue print for butter bread forms this word. It is plain to see that suspecting language contact as the reason for this formation is completely off the mark in this case. Likewise, to claim that the abovementioned bilingual child forms butter bread without using its bilingual brain not only disregards results from research on bilingual language acquisition, it also lacks common sense. Generalizing from this, quite obviously any statement about “normality” in transmission can only be valid for a particular social situation type. In their seminal paper on language change Milroy and Milroy (1985) identify types of communities that are particularly susceptible to change while others are not. So, while for a close-knit community with strong intra-group ties and no outgroup ties (quarantine situation) “normal transmission” in the above sense may be normal, this may not be the case for a community with lots of mobility and lots of weak in- and out-group ties. Milroy and Milroy (1985) illustrate this quite nicely with the cases of Icelandic and English. While change and in particular external change would be expected for English, the reverse is true for Icelandic. As a result, both languages are at opposite ends with respect to innovative vs. conservative character within the Germanic languages. Consequently, what is normal in one situation may not be normal in another, which means that only similar social situations can be compared under the same term. Sure, if external change is posited for a change in Icelandic this would be more controversial than if the same was proposed for English. But if a change in English is attributed to internal development, double-checking is advised, given the history of the language. In this and other, more extreme cases, external influence may in fact be more normal. To sum up this discussion, what constitutes the default or normal way of transmission crucially depends on the social situation. In principle innovation, i.e. the pre-stage of language change, can be triggered by internal or external factors. Which of the two is more common in a given situation type may enjoy

Towards a framework of contact etymology

13

the position of the null hypothesis, but it would be unwise to give precedence to one or the other without considering the situation. In analogy, one might say on battlefields external causes of death are frequent, at universities less so. With respect to externally influenced change, it is clear that this is neither infrequent nor less likely in all situations. There have probably been only a handful of languages in the history of human language completely free from external influence. Even parallel cases for an internal development cannot be a convincing argument; at best they advocate a research strategy (Filppula 2003: 171). In addition, there are cases in which an internal development is a priori rather unlikely, and for which therefore a preference for an internal explanation would actually be a methodological blindfold. Such cases are e.g. terms for indigenous plants and animals in colonial languages, such as English wombat, kangaroo, etc. borrowed from Australian indigenous languages, instances of metatypy (Ross 1996), and also particular types of networks that favour external change, typically communities with loose ties (Milroy and Milroy 1985). As a result, there is no theoretical reason to assign internal explanations a privilege position over other explanations; it is also methodologically imprudent, because it limits the investigative horizon (see also Farrar and Jones 2002: 3). Nevertheless, if the phenomenon under consideration is not a priori more likely to be due to external influence, it makes methodological sense to investigate the possibility of internal causation first, simply because contact explanations involve considerably more effort, as will become clear further below.4 Having established that internal and external explanations have an equal theoretical status, the question remains when to investigate alternatives to internal developments. The solution Thomason and Kaufman (1988: 60) advocate, namely that contact explanations are appropriate if external influence can be found on more than one linguistic subsystem seems arbitrary. Moreover – as they also admit – this does not hold for borrowing situations. As is implicit in the results from the discussion about the status of the different types of explanations, it is also problematic to postulate that external explanations are only to be considered once internal explanations have failed. Methodologically, the soundest path seems to be to use what we know about 4 Farrar and Jones (2002) are correct in pointing out that language change is not infrequently subject to multiple causation, and that determining whether one kind of change can always be assigned unambiguously to the internal or the external domain is doubtful. Though this is another nail in the coffin of the Internal Development Bias, it is less relevant for etymological research, because typically words are either from another language or they are not. However, there are cases in which multiple causation cannot be ruled out or is likely (see Zuckermann 2003 for examples).

14

Robert Mailhammmer

language contact as a research strategy in order to hypothesize a scenario under which a given linguistic unit could have been developed as a result of language contact. This is not the same as to suggest using knowledge about language contact as diagnostic (see e.g. Thomason and Kaufman 1988: 63; Heine and Kuteva 2005: 22; Ross 2003: 176). The reason is that not all changes that are often the result of language contact could not also be caused internally. For instance, Ross (2003: 193) notes simplification as a symptom for the adoption of a lingua franca, which may well be the case, but simplification can of course also occur without having been caused by language contact. But even if what is known about language contact cannot be used as to diagnose language contact in the narrow sense of the word, it surely permits the reconstruction of a language contact situation that would have caused a particular development. This is what I would like to call the Blueprint Principle, i.e. using language contact theory to backtrack. If it is assumed that an item is indeed due to contact, a contact situation is reconstructed in which the word or structure could have originated. Essentially, this is means forming a hypothesis, whose value is first and foremost derived from its logical, i.e. its linguistic, plausibility. This makes it linguistically testable, independent of the extralinguistic evidence. Of course, the case for a contact explanation can be strengthened by independent evidence for a contact situation, but as I will argue further below, this is only a supplementary argument, especially in historical scenarios, because obtaining historical proof that something has happened is actually more difficult than it seems at first glance (see Honeybone 2012 for a detailed discussion including references). But this is valid for any reason for language change, internal and external. Ultimately, even if it may be impossible to prove that a particular linguistic innovation is due to language contact or that language contact took place in a historical situation, a probabilistic assessment can be given. From this it follows that a contact etymology seems a reasonable option if the word or structure in question is either an unambiguous or likely (based on parallel cases) result of language contact, or if there is no convincing internal explanation. The next section examines the necessary “ingredients” to proposing a contact explanation, which then form the basis of the Blueprint Principle, which is used to construct a hypothetical language contact situation that can explain the origin of a particular word or structure.

3 The Blueprint Principle The basic idea of the Blueprint Principle is to interpret the items that are hypothesized to go back to external influence as traces of a contact situation.

Towards a framework of contact etymology

15

Using these hypothetical traces as “negatives”, a positive, i.e. the language contact situation, is reconstructed using the methodology of investigating language contact. (1)

The Blueprint Principle If it is assumed or known that an item X (lexical or structural) either in its systemic status or in its usage is due to language contact, then a hypothetical scenario is reconstructed that explains the occurrence of X by interpreting X as traces of this scenario.

Basically, this is abductive reasoning, i.e. the relevant contact scenario is inferred from knowing that X commonly occurs in particular situations of language contact combined with the assumption or knowledge that X in language A is actually due to language contact. Note that it is methodologically irrelevant whether it is actually known if X is due to contact. But from a theoretical perspective this is significant. If it is known that X is a contact feature, the Blueprint Principle reconstructs how X came into the language. This is, however, an important part of the explanation, because it effectively checks the knowledge basis. Even if there are linguistic changes or phenomena which can only be explained by language contact, and even if an external source for X has been confirmed by non-linguistic data, if a matching contact scenario cannot be reconstructed or if it cannot be substantiated, this casts doubts on this “proven” origin. For instance, if a historical document mentions that a word was borrowed from another language and this seems unlikely from a linguistic perspective, it may well be that the author of the document was wrong about the word’s origin. However, if the external origin of a word is only assumed, then, of course, the detailed reconstruction of a plausible contact scenario that can explain the occurrence of X in A is crucial to substantiating a contact etymology. Thomason and Kaufman (1988: 63) highlight key aspects of proposing a contact explanation.5 We can sum up this section by answering the question posed its title in the following way: an external explanation for a particular structural change is appropriate, either alone or in conjunction with an internal motivation, when a source language and a source structure in that language can be identified. The identification of a source language requires the establishment of present or past contact of sufficient intensity between the proposed source language and the recipient language. Sufficient intensity (especially in a past contact situation for which little sociolinguistic information is available) may be inferred from

5 Thomason and Kaufman refer to situations of imposition (or substratum influence; section 4.1 below for a brief discussion of terminology), but these methodological considerations are in principle valid for all contact situations.

16

Robert Mailhammmer

the presence, in different grammatical subsystems of the recipient language, of structural innovations that may reasonably be attributed to that source language. The proposed source-language structures need not be, and frequently are not, identical to the innovated structures in the recipient language, but a successful claim of influence must of course provide a reasonable account of any reinterpretation of generalization that has occurred as a result of the interference. (Thomason and Kaufman 1988: 63)

From this three core elements can be extracted, namely a source language, a source structure and an appropriate contact situation. As argued in section 5 below, an application of the Blueprint Principle would probably seek to identify the kind of contact situation and reconstruct it in as much detail as possible and then attempt to identify a possible contact language and contact structure. The following section discusses each of these elements. Particular attention will be given to detail, because this is commonly where proposed explanations positing external influence often exhibit problems.

4 Contact scenario, contact language and contact structure This section addresses some of the most influential factors that have to be taken into account when investigating language contact, and when proposing an explanation based on language contact, such as a contact etymology. For want of space I cannot cover everything; therefore I refer the reader to the relevant literature. The purpose of this section is to systematize existing knowledge in order to work towards a coherent framework of contact etymology.

4.1 The contact scenario In describing the contact situation, it is of crucial significance to capture the linguistic dimension of the contact as well as its social dimension. It is preferable to do so within a framework that keeps these two aspects terminologically and taxonomically distinct, such as that proposed by Frans van Coetsem (2000).6 There are two basic linguistic contact situations, namely BORROWING 6 The basic reason is that e.g. although it is often the social substratum that shifts to the social superstratum thereby exercising substratum influence, this need not be the case, as cases of superstratum shift have been recorded (see Mailhammer (2006: 20–21) for a discussion, including references). Hence, if such a case were to be described using the traditional terminology one would have to say that a superstratum exercised substratum influence, and it is easy to see how this could be confusing.

Towards a framework of contact etymology

17

and IMPOSITION , as well as a composite case of MIXED LANGUAGES .7 The first refers to the importation of items of language B by language A, either covertly or openly. The second describes the influence language C has on language D in a process of language acquisition where speakers of C acquire D, while mixed languages are generally the result of a combination of the two. The basic difference between the two basic constellations is that speakers of language A generally have no real desire to learn B or to give up their native language, whereas the speakers of C aim at learning D (regardless of whether they end up giving up their native language or not). Consequently, the linguistic effects of the two processes are different. Borrowing integrates foreign elements into a language, which is why the most “mobile” items (words, especially nouns) are borrowed most easily and the least mobile items (inflectional morphology) are most difficult to borrow and presuppose a particularly intimate contact situation with a high degree of bilingualism.8 Imposition, by contrast, is more or less an accident in the acquisition process, especially if that takes place without instruction, which is why particularly well-entrenched parts of the language (phonology, syntax but also pragmatics and phraseology) are likely to carry over into the learner variety of the target language, commonly called interlanguage, and, if the circumstances are right (cf. next paragraph below), on into the target language as spoken by native speakers.9

7 This also comprises complex cases of convergence or pattern replication (see Matras 2009: 234–274 for a detailed overview). See van Coetsem (2000) for the basic model and terminology used here. In traditional frameworks “imposition” is usually called “substratum influence”, a term which is problematic because it does not distinguish the linguistic from the social contact situation (see also the preceding note). Note also that imposition is not the same as relexification, because it involves language learners whose aim is full proficiency in the target language. Relexification, by contrast, generally is a situation of extreme borrowing (which equals a situation of very imperfect language acquisition), so that nearly the whole vocabulary is swapped for that of the lexifier language. The difference is that imposition can influence the target language, but relexification usually does not. 8 This is, of course, a simplification. See Matras (2009: 153–165) for a detailed discussion of borrowing hierarchies. 9 Commonly, positive and negative transfer are distinguished; the former refers to a situation in which a transfer from the native language to the target language produces the correct result, while the in the latter case a transfer leads to an incorrect result. Consequently, positive transfer may reinforce existing structures to a point in which they may be overgeneralized at the expense of others. However, the more dramatic effects tend to occur in cases of negative transfer, because the mismatch can e.g. affect the acquisition or transfer of functional categories, so that new categories can be created, categories can be lost or changed and so forth.

18

Robert Mailhammmer

It is necessary to draw attention to a point that has generally been neglected in explanations based on language contact (though mentioned e.g. in Thomason and Kaufman 1988: 43 and Winford 2003: 17), and this is to account for why a particular item that comes into a language from outside should be adopted by the majority of native speakers so that it takes hold in the language and is transmitted to the next generation of speakers. There is a social side to this point (see also further below) but also a linguistic one. A new word or structural item that is incompatible with the existing linguistic system may not make it into the language as opposed to one that fits in or fills a need. This does not mean that an incompatible element cannot be integrated into a language; it just means that contact etymologies have to cover this point. As a matter of fact, the step from the interlanguage to the target language as spoken by native speakers may be a separate contact situation. Speakers of the target language may borrow from the interlanguage, but they may also effectively shift to it. The social dimension comprises the social relationship of the languages involved and the characteristics of the speech community. The former refers to whether one of the languages is spoken by a community which exerts a military or economic dominance (superstratum), and therefore enjoys a privileged position and, whether it is dominated (substratum), and therefore does not enjoy a privileged position, or whether neither language is in a socially higher position (adstratum; see Vennemann 2011 and this volume for a more fine-grained subdivision of adstratum types). Connected to the social constellation is the issue of attitudes of the contact languages to each other. For instance, Lutz (2008: 143) demonstrates how borrowing in what she calls a “forced contact” situation differs from borrowing under non-biased circumstances. Additionally, the numerical relationships between speakers of the languages involved (Thomason and Kaufman 1988: 47), and other factors come into play here. Among the characteristics of the speech community, i.e. the “network of relationships among speakers” (Ross 2003: 176), clines such as the degree of openness of the community (open vs. closed) and the degree of tightness (tightknit vs. loose) have been proposed (Ross 2003: 179–180; see also Enfield 2003, 2008 for a more detailed framework). In their classic study Milroy and Milroy (1985) showed that speakers with loose ties to a community tend to be most susceptible to external influence and that the crucial step of adoption is made by the core of the community, the so-called “first adopters”. However, the key prerequisite is that the innovation in question is perceived as positive by these first adopters. This parameter of prestige has remained somewhat vacuous despite the fine-grained nature of the model by Milroy and Milroy (1985) and subsequent models. However, at least to factors seem to contribute to the prestige of an innovation (external or internal).

Towards a framework of contact etymology

19

One is the degree of bi-/multilingualism in a community (see also Thomason and Kaufman 1988: 47–48), which is likely to be an important factor in determining the result of language contact. A multilingual speaker community that is e.g. familiar with the language words are borrowed from will be more sensitive to structural properties, such as phonology, morphology, etc. Lutz (2009), for instance, shows that the position of word stress in the integration of French loanwords into Medieval English and German correlates with the degree of education in the borrowing speech community. Whereas in England a large proportion of the speakers – mostly relatively uneducated – adopted French loanwords, “in Germany, French and Latin influence remained largely restricted to the language of scholars and upper ranks of society for a long time” (Lutz 2009: 283).10 A second factor is the compatibility of the innovation with the receiving language. There are various ramifications from the various possible social constellation of the contact situation, and the literature reiterates the significance of the social circumstances of language contact. And since, “practically any linguistic feature can be transferred from one language to another if the circumstances are right” (Winford 2003: 25), it is of such decisive significance to work out the contact scenario in as much detail as possible. Especially in historical scenarios, in which ultimate proof that an item is due to contact may not be possible, it is crucial to increase the plausibility of a contact etymology, by showing exactly how this item got into the language and how it took hold and persevered. Parallel cases are obviously excellent support, but first and foremost the relevant linguistic and sociolinguistic parameters of the contact situation need to be spelt out. As far as the extra-linguistic reality of the contact situation is concerned, it is apparent that this renders support to a contact etymology. However, I do not regard a proven contact situation as crucial to a contact explanation for the following reasons. First, it is obvious that the knowledge alone that two languages were in a contact situation is insufficient to prove that given items really go back to language contact. What is required at any rate is a linguistic explanation that

10 A reviewer remarks that the contact situation between English and French was presumably much more intense than that between German and French. First, this is debatable, given that French was the cultural superstrate through the second part of the Middle Ages for much of Europe. Second, even if that was true, it is difficult to see how a more intense situation of contact should lead to more adaptively integrated loanwords. From what is known about language contact situations, more intimate contact leads to more structural borrowing, i.e. more imitative borrowing. Therefore one would expect English to have borrowed the French words together with their accentuation like in German. But instead, as Lutz (2009) shows, prestige and bilingual knowledge are the decisive factors here.

20

Robert Mailhammmer

shows how that item was transferred from one language to the other. Second, even if there is no extra-linguistic record of contact between two languages, a detailed reconstruction of a contact situation demonstrating how external influence accounts for the existence of a linguistic element can support a contact etymology with a sufficiently high probability so that the contact between two languages can be inferred with reasonable certainty. Especially in historical scenarios, in which there is little record about the extralinguistic reality, the reconstruction of a detailed contact situation can actually present the key to unsolved puzzles, if this demonstrates that contact could actually account for the presence of a linguistic element in a given language. However, it is apparent that this has to be done carefully and within the limits of reasonable assumption to prevent contact scenarios that feature aliens or contact between languages that cannot have been in any contact situation, such as Old English and the Australian Aboriginal language Amurdak. But in principle the reconstruction of the language contact situation is independent of the extralinguistic reality, though it is clear that an extralinguistic confirmation of the contact (at least in probabilistic terms) renders strong support to a contact explanation.

4.2 The contact language In order to propose a convincing contact etymology, the contact language has to be identified as exactly as possible. Vague statements, such as “a non-IndoEuropean language” or the name of a language family, or simply “unknown substratum”, can sometimes be all that can be said about a word, but it is obvious that this can hardly be called an etymology. In addition, any statement has to be carefully supported. For instance, many etymologies for Indo-European words are attributed to substrate languages without justifying why the implicit social contact situation is assumed. Often the semantic domain of the word suggests that the word came into the language from a superstrate or adstrate language. This is the central point of Vennemann (1984), showing that a good deal of the non-etymologized Germanic vocabulary is much more likely to have come from a superstrate language than from a substrate language, judging from the semantic domains it belongs to. It is important to think about the contact situation, because that has ramifications for the reconstruction of the contact language. Because borrowing does not presuppose a significant degree of bilingualism, the adaptation and integration of a new item can sometimes have drastic consequences to the formal side of an item. Thus, calquing and other forms of covert borrowing make it hard to detect a foreign origin (see e.g. Mailhammer 2008). But imposed terms are rarely calqued and often retain formal character-

Towards a framework of contact etymology

21

istics that are alien to the target language, because of the imposing speakers’ bilingual abilities (see 4.1 above for the example of borrowed French loanwords in English and German involving different degrees of bilingualism). Consequently, one would expect a higher chance that a foreign word coming into the language via imposition retains more foreign features than if it is borrowed (unless, of course, the borrowing speakers are bilingual). The typological characteristics of the languages in contact seem to determine the result of language contact to some degree as well, a fact that has to be taken into account when proposing an etymology, as e.g. Thomason and Kaufman (1988: 53–54) point out. In cases of heavy borrowing and shift interference, at least, the effects of typological distance on the expected kinds of interference features seem limited, if not negligible. … Nevertheless, although we have no definitive evidence to suggest it, our tentative hypothesis is that in cases of light to moderate structural interference, the transferred features are more likely to be those that fit well typologically with corresponding features in the recipient language. That is, we believe (with Meillet, Weinreich, Vidomec, and others) that one structure will more readily replace another if they already match rather closely in function. (Thomason and Kaufman 1988: 53–54)

This means that the typological distance of the languages involved mainly impact on structural transfer, but that lexical transfer, which is most common in borrowing and thus likely to be relevant to lexical etymologies, seems to be largely unaffected by typological characteristics of the languages involved. However, as will become clear in the next section, these characteristics play a significant role in the formal side of an etymology, i.e. they cannot be neglected at all. If the contact language cannot be identified, this does not rule out that an item is due to language contact. It may still be possible to give characteristics of a hypothetical contact language based on the items assumed to be of external origin. For instance, if the overgeneralisation and systematisation of ablaut in Germanic (Mailhammer 2007) is interpreted as a hypothetical contact feature, it makes sense to look for a contact language in which ablaut is a prominent feature of verbal stem formation (see 5. below). Or loanwords in English with the initial cluster [ ʃn], such as schnapps, permit the inference that the donor language possessed this cluster at least in these words. Conversely, when looking for a potential donor language, languages without this cluster (or at least clusters that may be misperceived by Anglophones as [ ʃn]) as a possible wordinitial syllable head can be ruled out. To sum up, though the identification of the contact language is a key element in the proposal of a contact etymology, frequently there is insufficient data. In such a case the item suspected of external origin can be used to infer characteristics of a hypothesized contact language.

22

Robert Mailhammmer

4.3 The contact structure As mentioned in section 1 above, etymology involves historical word formation. In a contact etymology the crucial point, however, is not that of Urschöpfung, because that is an exercise is “conventional” etymology. What makes or breaks a contact explanation, of course, is the account of the point at which the word or structure is transferred from one language into the other. After establishing the situation and the contact language, it is necessary to pin down the source word or structure and explain any changes on the way into the receiving language. What happens after the item has come into the language falls then again within the domain of conventional etymology. For instance, E cockroach was probably borrowed from Sp cucaracha; its initial shape is recorded in the OED as Cacarootch. Later, however, it was reanalysed folk-etymologically as cock + roach.11 So, in this case the contact part of the etymology has to explain the form and meaning of the initial attestations, while the modern form is handled by conventional etymology. It must be emphasized at this point that it is precisely the partly unpredictable nature of the way a word (or structure, though these tend to be transferred more faithfully) may show up in the receiving language (see section 1 above), why this technical part of the etymology needs to be particularly well-researched, despite its difficulty. If a contact etymology competes with a native etymology, then this point is especially important, because a native etymology tends to be constrained a great deal more due to the rigid nature of sound laws that support an etymology. The way a linguistic item arrives in the receiving language is conditioned by the contact situation and by the systemic properties of the languages involved. The contact situation is pivotal in determining whether the goal in the process of transfer is adapting or imitating the foreign item (see van Coetsem 2000 for a detailed discussion of imitation and adaptation in language contact). Borrowing generally aims at adapting a foreign item integrating it into the system of the receiving language, thereby changing it more or less. There are imitative borrowings, particularly if the source language enjoys a high level of prestige and if the speakers of the receiving languages possess a degree of proficiency in the source language, but these are not the typical cases. By contrast, the objective of language acquisition is to speak the target language with native-like proficiency, which is rarely mastered by adult learners. Imposition is a by-product of second language acquisition, as the native grammar 11 I am grateful to Theo Vennemann (Munich, p.c.) for pointing this out to me. It corrects Mailhammer (2008: 183).

Towards a framework of contact etymology

23

interferes with the grammar that is to be acquired, resulting in a learner variety, the so-called interlanguage. Adaptation and imitation of the foreign item do not play a role in this first step of imposition proper because it is the target language that is to be imitated and imposition is just an accident. In a second step the changes imposed on the interlanguage are carried over into the target language, as they are taken on board by native speakers of the language, thus changing the target language.12 As pointed out above, this second step is a language contact situation by itself (most frequently a borrowing situation), which needs to be addressed in a contact explanation (also in terms of imitation and adaptation), because further changes may occur so that the original result of imposition actually shows up differently in the target language or a variety of it. The question whether a linguistic item is transferred aiming at adaptation or imitation is linked to the question of bilingualism among the speakers of the receiving language. If e.g. a foreign word is borrowed imitatively as much as possible of its form and meaning will be retained, resulting in new formal and semantic elements in the receiving language, especially if the borrowing speakers are bilingual. Contemporary English loanwords in German, for instance, tend to be much less phonologically integrated than older loanwords. Consequently, German phonology is acquiring sounds it did not acquire together with earlier loanwords, such as [ɻ ] and [ʤ]. The reason is that most Germans now are sufficiently proficient in English to be able to pronounce these sounds and do so also when using the words in German, thus effectively borrowing new sounds along with these words.13

12 Alternatively, the interlanguage forms a separate variety of the target language that stabilizes and is passed down to the next generation. Examples of this kind are the various (post-)colonial Englishes, e.g. Irish, Scottish Singapore, Cameroon and other Englishes. If speakers of these languages are bilectal, i.e. proficient in the target language as spoken by original native speakers, they may well transfer items from the former interlanguage to the “standard” variety via borrowing, in which case a bilectal contact situation arises. 13 A reviewer remarks that “the phonological integration of English sounds into German is presumably due to the increased amount of English loanwords in German, not to Germans’ increased proficiency in English.” This is implausible, because no matter how many loanwords a language receives, if the native speakers of the receiving language cannot pronounce certain sounds, they simply replace them regardless of their frequency. There are numerous examples that illustrate this fact. The numerous French loanwords in English did not lead to the adoption of nasal vowels in English, the many English loanwords in Romanian have not contributed one single phoneme to Romanian, and a similar situation exists in Japanese. Furthermore, while German imports new phonemes from English, it does enforce final devoicing in words like Job. This could not explained on the basis of frequency but if it assumed that Germans have not become that proficient in English that they can “turn off” final devoicing (it is well-known that L1 phonology impacts on perceiving and processing L2 in the acquisition process, see e.g.

24

Robert Mailhammmer

The systemic characteristics of the languages in contact constitute the second important technical issue that a contact etymology needs to pay attention to. Phonological differences lead to sound substitutions, re-syllabifications, and so forth, differences of word order can lead to word order changes etc. Differences in grammatical categories can lead to complex restructurings that can change the typological characteristics of the receiving language. It is therefore paramount that both contact languages are well known in order to assess what happens when they meet linguistically. Identifying the source structure requires a detailed investigation of all relevant aspects. For instance, it is not enough to present a root in a hypothesized contact language that is assumed to be the source for a hypothetical loanword. In order to present a convincing suggestion, one would have to give the fully inflected source form and range of meanings, including attestations and, if possible, uses, accounting for all differences between the original and the loan. If no direct evidence is available, reconstructions are vital in order to demonstrate the viability of the proposal. To sum up, the technical side of a contact etymology has two components, namely identifying the source item and accounting for any changes occurring in the process of transfer. In order to work this out the social and linguistic contact situation as well as the systemic properties of the languages in contact have to be taken into account.

5 Towards a framework of contact etymology 5.1 A step-by-step guide From the discussion in the preceding sections it has become clear that the point of transfer is both the defining as well as the key part of a contact etymology. This section outlines a tentative methodology for explaining items as contact features. It is apparent that – just like a native etymology – a contact etymology can be extremely complex, especially if more than one “contact point”, i.e. multiple contacts are involved (what Vennemann 2002 calls “the transitivity of language contact”). Any etymological investigation – in its traditional sense (see the discussion in Mailhammer 2007: 142–144) – begins with the earliest attestation of a word or structure in a language. This presupposes an exact identification of the item Broersma and Cutler 2010), this finds a natural explanation. Furthermore, older speakers of German with less proficiency in English never import/create those new phonemes, similarly speakers with less education in English.

Towards a framework of contact etymology

25

under investigation, because it will serve as the basis for the reconstruction of the contact situation. Is it a word, structure, a pattern of usage, a grammatical category, a meaning, a word order pattern, to name but a few phenomena that can be transferred through language contact? I will use the terms element and item synonymously to capture all of these. There may well be direct evidence for a contact-origin of the element identified, such as violations of sound laws (e.g. words with /p/ in High German dialects, as Pre High German /p/ normally shows up either as fricative or as affricate), other unambiguous effects of contact, e.g. metatypy (see Ross 2003), or a record of the contact. But frequently, there will be none of this sort. Often there will even be an internal explanation, sometimes even quite a good one. In a number of cases, nevertheless, there will either be no account or merely a description of a hypothesised development. At any rate, as soon as an item is suspected to have come into a language, one can begin to investigate the possibility of a contact etymology. The first step is the establishment of the contact situation, because this impacts strongly on the reconstruction of the contact language and the source structure. It is probably best to start with the contact type, i.e. borrowing or imposition, and this is gauged from the nature of the item in a rule-of-thumbway. Generally, words are more likely to be borrowed than imposed, whereas structural elements, usage patterns etc. are more likely to be the result of imposition. This is partly connected to other items in the language that may have been due to contact with the same language. For instance, if words as well as structural items are suspected of contact, then this may well point to a situation of imposition, particularly if the words are from matching semantic domains. It is obvious that this is only a first stab in the dark, which is subject to refinement as the reconstruction progresses. If possible, other aspects of the contact situation, such as the social relationships of the languages can be considered at this point. Also, the characteristics of the community can often be researched from historical data. Once the contact situation and its basic determinants – contact type, social relationship of the languages involved and degree of bilingualism – have been outlined, it is time to think about the contact language, if it is not already known from external data. The application of the Blueprint Principle in this step involves backtracking from the item under investigation to a language from which it was transferred. In situations of imposition this means reconstructing a complex scenario of second language acquisition, by asking the question: “Which language as L1 is likely to have caused X in the acquisition of the language in which X occurs?” The main way to investigate this question is to examine well-documented parallels, but sometimes more speculation is

26

Robert Mailhammmer

involved. For instance, the initial step in looking for a contact language that could have triggered the functionalization and systematisation of ablaut was to look for a language or language family that makes extensive use of ablaut in verbal inflection (see Vennemann 1998: 42). Often the initial search yields only a vague result, but as the proposal is becoming more refined, the source language should be identified as exactly as possible. The final step is the identification of a source structure and the reconstruction of the transfer. This involves also the question of why the transferred item survived in the receiving language. These technical details are not to be taken lightly, because a careful and fine-detailed renders powerful support to the etymology. For words, for example, this means explaining the sound correspondences and substitutions, meaning shifts, as well as morphosyntactic properties. In interference scenarios the grammatical systems of the languages have to be compared in detail. To sum up, in an ideal case a contact etymology demonstrates the transfer of the item in a fine-grained reconstruction that supports the assumption of a contact-induced origin for an item so strongly that an alternative is less plausible. Methodologically, it seems sensible to reconstruct the contact situation, proceed then to the contact language and to flesh out the proposal by adding the technical information about source structure and transfer.

5.2 Sketching an application: etymologizing PGmc. *pleganan ‘wager, dedicate, employ’ The step-by-step model outlined in the previous section is followed up here with a sketch of an application. For a full etymology and details on the contact situation and the transfer itself I refer the reader to more in-depth investigations (Mailhammer 2006, 2007: 202–208). PGmc. *pleganan ‘wager, dedicate, employ’ has no Indo-European etymology, and all existing suggestions have one or more fatal flaws. The hypothetical ProtoIndo-European root to be reconstructed from the attestations in the Germanic daughter languages according to regular sound laws, PIE *blek-, which is required to explain attestations like OE plēon (*(a)re in Arandic). The forms given in Table 6, on the other hand, demonstrate by the example of words including *arre ‘mouth’, that the Arandic languages do indeed make use of the morphological patterns, consisting of compounds and derivatives involving bodypart nouns, that are assumed in our etymological proposal that forms in (a)tn- and (a)rin Tables 2 and 4 respectively continue the Proto Pama-Nyungan terms terms *tha/ina 6 and *mara. The results in this section show that it would be rash to claim that an etymon which is widely attested in a language family, such as Pama-Nyungan or Indo-European, is totally absent from a particular subgroup, such as Arandic 5 Single quotes indicate that the word is cited from early sources (e.g. Teichelmann and Schürmann, Reuther) in the original spelling rather than in a modern phonemic notation. 6 Some Pama-Nyungan languages have *thana instead of *thina. We are assuming that the Arandic reflexes better reflect an earlier *thana – although at this stage the variability of wordinitial vowels in the Arandic languages is not totally understood – for some of the issues see Breen (2001).

39

Obscure vs. transparent cognates in linguistic reconstruction

language

word

gloss

Pattern

WAr Aly ECAr:NE

tnerurrke 7 atnererrke atnike

heel tendon at the back of the ankle, heel heel

N-N N-N N-N

ECAr ECAr Aly

atnware 8 atnalthe atnalthe

back of foot, heel a lot of tracks in an area place where there are a lot of footprints (-althe INTENS)

N-sfx N-sfx N-sfx

ECAr WAr ECAr Aly

arnpetnarnpeatnarnpeatnarnpe-

put foot down; head off jump/step/climb down jump/step down jump; get down

V N-V N-V N-V

ECAr WAr

urntetnurnte-

dance dancing with a stick

V N-V

ECAr ECAr

iweatniwe-

throw put s.o. on their feet

V N-V

ECAr WAr ECAr

ulketnulke-9 atnulke-

jerk, jump; get a shock get a fright, jump with fright jerk, jump; get a shock

V N-V N-V

Table 2: Arandic possible reflexes of ‘foot’ (a)tne < *thatne < *thana

or Italic. Szemerényi’s methodological principle D, quoted above, would thus appear to be valid for the Pama-Nyungan languages, especially those for which we have plentiful documentation, and those which have undergone drastic sound changes. At the same time, the discovery of reflexes in further subgroups (such as Arandic) strengthens the case for the reconstruction of the lexemes in the highest-level protolanguage.

3 A disguised Thura-Yura and Yarli cognate in Karnic The word pari ‘creek’ (which always has a retroflex r in those languages for which we have recordings10) is widespread in the Thura-Yura languages: Kaurna 7 This word also occurs in collocation with the contemporary ingke ‘foot’, in a Whole Part construction. 8 As for note 7: the collocation ingke atnware occurs. 9 In this word the initial element (a)tn- possibly reflects rather atne ‘stomach/guts’. 10 In sections 3 and 4, we use the following spellings for rhotic consonants: r for the retroflex glide, r for the apical tap, rr for the apical trill.

40

Harold Koch and Luise Hercus

Pintupi Pitjantjatjara

compounds

literal gloss

gloss

tjina yanyu-wana tjina tju-

footprints same along foot put

along the same track release, set free

Table 3: Western Desert derivatives and collocations with tjina ‘foot’

Pintupi Pintupi Pitj Pintupi/Pitj

compounds

literal gloss

gloss

mara ngurrkumara pirrtjimara-katimara-mara

hand scoop hand move hand-carry hand-hand

draw water from a hole with cupped hands wave the hand put out your hand child at crawling stage

Table 4: Western Desert derivatives and collocations with mara ‘hand’

Aly ECAr WAr ECAr Aly ECAr Kay

akeakerakerakerake-, irakearethape arengkelane-

pick, collect pick fruit, leaves, etc. from a plant rob, snatch away, grab away from s.o. grab, snatch s.t. away from s.o. grab, take s.t. from s.o. baby up to crawling stage crawl

V V N-V N-V N-V N-N?

Table 5: Arandic possible reflexes of ‘hand’ are < *mara

‘parri’ (cf. the South Australian placename Onkaparinga11), Nukunu pari, Wirangu bari, Adnymathanha vari, Parnkalla ‘parri’, Kuyani pari. Moreover the word pari ‘creek’ is found in all three Yarli languages: Malyangapa, Wadigali and Yardliyawara. The word pari is also used in a compound noun meaning ‘the Milky Way’, as in Kaurna “wodliparri the Milky Way, which the natives believe to be a large river” (Teichelmann and Schürmann [1840] 1982: 57). The Adnyamathanha (Flinders Ranges language) word is similar, warli-vari, lit. ‘the wurley (hut) creek’, as the Milky Way was envisaged as a river with innumerable campfires of ancestors near its banks. On the basis of these cognates, displayed in Table 7, we might think that the word pari results from a common innovation in the Thura-Yura and Yarli 11 The name Onkaparinga represents Kaurna ngangki ‘woman’, parri ‘creek’ followed by the locative suffix -ngka (cf. Teichelmann and Schürmann 1840: 30, 36, 75 and Wyatt 1879: 179).

Obscure vs. transparent cognates in linguistic reconstruction

Kay ECAr ECAr ECAr:NE ECAr ECAr ECAr:SE ECAr ECAr ECAr ECAr ECAr ECAr ECAr ECAr ECAr ECAr ECAr ECAr ECAr ECAr ECAr ECAr:NE

arre arrakerte arrilye arrirle arralte arre-yenpe arrimpimpe arre pmwarepmware arre-iltwe arre-anteke arre-pweke-pweke arranyenganyenge arrinte arrerlkarte, -erte arriwe arrangkearrerlpearrinkearralkearralthearrawenthe-irrearretyearrirte-

mouth mouth cheek, side of face cheek, side of face hair on face; beard, whiskers lips lips roof of the mouth in the side of the mouth big/loud-mouth first hair on teenager’s face toothache wooden toothpick with a strong taste in your mouth entrance, opening cry, bawl chew like, want, enjoy (food) open mouth wide; yawn shout at s.o. in a rude or angry voice whistle whisper, talk softly whisper, talk softly

-akert HAVING

alte hair ‘mouth + skin’ pmware wooden scoop anteke wide

angke- talk

41

N N-sfx N-sfx N-sfx N-N N-N N-N N-N N-N N-N N-N N-N N-N N-N N-N N-V N-V N-V N-V N-V N-V N-V? N-V?

Table 6: Arandic derivatives and collocations with arre ‘mouth’

languages, is specific to these adjoining subgroups, and is reconstructible to a common ancestor of Proto Thura-Yura and Proto Yarli. Careful comparison, however, reveals that it may go back even further in history; for there is one clear cognate in the Karnic subgroup: Pitta-Pitta paripi ‘creek’ (with an extra syllable increment). There are several further cognates among the Karnic languages, but these are hidden away in compounds. In Arabana-Wangkangurru and in Wangka-Yutjurru, its immediate neighbour to the north-east in Queensland, the word for ‘creek’ is the obviously non-cognate form karla. In Arabana-Wangkangurru pari does exist, but it is restricted to compounds: warru-pari ‘Milky Way’ (lit. ‘white creek’12) and -pari-pari ‘a white streak, a “flash”’, which itself is found only in compounds such as mangu-paripari ‘having a white streak on the forehead (like some wallabies – and horses)’ – a fairly frequently used term in Arabana-Wangkangurru. 12 Compounds where the adjective precedes the noun are unusual, but do occur in AW, as in the word for ‘emu’, warrukathi, lit. ‘white-meat’.

42

Harold Koch and Luise Hercus

creek Wirangu Kaurna Nukunu Pankarla Adnyamathanha Kuyani Proto Thura-Yura Wadikali, Malyangapa Yardliyawara

bari ‘parri’ pari ‘parri’ vari pari *pari pari pari

compounds

literal gloss

gloss

‘wodliparri’

hut creek

Milky Way

warli-vari

hut creek

Milky Way

Table 7: ‘creek’ in Thura-Yura and Yarli languages

creek compounds Pitta-Pitta Wangka Yutj AW

paripi karla karla X-pari-pari warru-pari Diyari karirri ‘pariwutju’ Diyari pariwilpa Diyari karirripariwilpa Yandruwandha karirri Kadrripariwirlpa

literal gloss

gloss

X-streak white creek X-long and thin creek-opening/gap celestial waterway creek-sky

having a white streak Milky Way narrow, deeply washed-out creek sky Milky Way Milky Way, also Cutrabelbo waterhole

Table 8: *pari in Karnic languages

In Diyari and closely related languages the word for ‘creek’ is karirri (Reuther 1981 writes ‘kajari’ and ‘kajiri’ 13); yet even here we find a fixed locution: “pariwutju (n) ‘a narrow, deeply washed-out creek’ – wutju means ‘long and thin’” (Reuther 1981 IV 2414). Furthermore pari is found in a much-used word pari-wilpa ‘sky’ (cf. Austin 1994: 132). This word for ‘sky’ in Diyari and neighbouring languages, including Yaluyandi and Yandruwandha, is a compound containing the same word pari followed by wilpa, which means ‘opening, gap’, also ‘open’. The semantics behind this are not clear, perhaps ‘(where) the Milky Way has a gap (a dark area)’, i.e. ‘the night sky’. This descriptive term for the night sky was presumably generalized as ‘sky’. This word is further compounded in a word for the Milky Way: “Kadri (or Kajaripariwilpa) ‘the Milky Way’ (or) ‘the celestial waterway’” (Reuther 1981: IV. 109.35), Kadripariwirlpa in Yandruwandha (Breen 2004: 19). 13 There is a lot of fluctuation between r and y in Karnic languages, especially before i.

Obscure vs. transparent cognates in linguistic reconstruction

43

On the evidence of these Karnic cognates, which are summarized in Table 8, we conclude that pari ‘creek’ belonged to Proto Karnic as well as to Proto ThuraYura and Proto Yarli. A cognate set like this is relevant to establishing whatever higher-order grouping these subgroups partake in.

4 Disguised Karnic cognates in ArabanaWangkangurru 4.1 Transparent cognates between Arabana-Wangkangurru and Central Karnic A genetic grouping of languages called Karnic has been recognized since Breen (1971), who based his conclusion on lexicostatistical principles. Four internal subdivisions were recognized, which we can refer to here as Eastern Karnic, Central Karnic, Northwestern Karnic (or Pittapittic), and Arabana-Wangkangurru. Austin (1990) compared vocabulary, established internal sound changes, and constructed a family tree. He claimed, inter alia, that the closely related dialects Arabana and Wangkangurru do not belong in the Karnic group, but may belong in a more inclusive group that includes Karnic. Bowern (1998), basing her argument on morphological reconstruction, did include Arabana-Wangkangurru in the Karnic family. There has been extensive recent discussion on the classification of Karnic languages, the latest being in Breen (2011). Austin’s main reason for excluding Arabana-Wangkangurru has to do with the presumed scarcity of vocabulary shared between Arabana-Wangkangurru and the (other) Karnic languages. There are 160 words listed in Austin (1990) which he has reconstructed as Proto Central Karnic, on the basis of Diyari, Yaluyandi, Ngamini, Yandruwandha, Yawarrawarrka, Mithaka and Karuwali. All but thirteen of these are said to have no cognates in Arabana-Wangkangurru (Austin 1990: 183). The absence of some Arabana-Wangkangurru equivalents is due simply to inadequate information being available at that time for this language. There are in fact many more cognates: a sample of s-cognates beginning with k is given in Table 9. Other Arabana-Wangkangurru cognates can be presumed to have been omitted from Austin’s (1990) list (a) on account of semantic changes: *kari, *mama-, *ngathara; (b) because they are only partial cognates, which included additional suffixes in Arabana-Wangkangurru: *kantha, *nharra-; or (c) because of other variations from Proto Central Karnic: *milamila. See Table 10.

44

Harold Koch and Luise Hercus

pCK

gloss

AW

gloss

*kalka *kantu *kaparra *karrtyi*kukunka *kuma *kunampira *kurluwa *kura *kungka-

yesterday rock wallaby come here to turn kite hawk to dance (of women) wild tomato needlebush hoarse to limp

kalka kantu14 kaparra karrtyikukunka kubmakudnampira kudluwa kura kungka-

evening, the other day rock wallaby come on! to turn black kite to dance (of women) ruby saltbush hakea indistinct, hoarse to limp

Table 9: Arabana-Wangkangurru and Central Karnic s-cognates

pCK

gloss

AW

gloss

*kari *mama*ngathara

chase to take back younger sibling

*kantha *nharra*milamila

grass to drive away mirage

karimamangatharu ngatha-ngathanganha kantharra nharrayi nyilanyila 15

to look for, find to grab behind, afterwards coming from behind, i.e. the youngest child grass to chase away mirage

Table 10: Arabana-Wangkangurru and Central Karnic: non-obvious cognates

Some quite common Karnic words appear to be absent from ArabanaWangkangurru; they are however still there, but hidden away. These are discussed in the following subsections.

4.2 Obscure cognates between Arabana-Wangkangurru and Central Karnic 4.2.1 *ngapa ‘water’ Austin (1990: 194) reconstructs *ngapa for his “Proto Karnic”. In ArabanaWangkangurru and the neighbouring western Pitta-Pitta dialects the Karnic word 14 Cf. also Adnyamathanha (Thura-Yura) andu. 15 Cf. Yandruwandha nhilanhila.

Obscure vs. transparent cognates in linguistic reconstruction

45

ngapa ‘water’ has been replaced by kutha, presumably a borrowing from Arrernte (cf. kwatye and kwethe-kwethe for ‘water’). There is, however, still plenty of evidence in Arabana-Wangkangurru of the presence of the word ngapa. Derivatives and compounds which incorporate an erstwhile ‘water’ term are displayed in Table 11. Note that several are placenames occurring in Arabana country (there are no ‘water’ names in the Simpson Desert in Wangkangurru country). The presence of these partial cognates is easily explained by the hypothesis that Arabana-Wangkangurru inherited the *ngapa root from Proto Karnic (it is also found in other Pama-Nyungan subgroups16).

compounds

literal gloss

gloss

ngapa-wirara ngapa-tya-mangapungka-

water-moving.about water-having-make

dragonfly to initiate someone as a doctor and clever man17 wade about in water, bathe18

Table 11: AW reflexes which include *ngapa ‘water’

The word ngapu ‘a floating fishing net’ may also be related. Placenames containing the word ngapa are discussed by L. Hercus (this volume).

4.2.2 *ngandi ‘mother’ Austin (1990: 191) reconstructs *ngandi ‘mother’ for Proto Central Karnic. The term for ‘mother’ in Arabana is the unique form lhuka (of uncertain origin19). Yet there is some evidence that *ngandi must have once been the term for ‘mother’ in Arabana as well; it survives in compounds in Arabana-Wangkangurru, as shown in Table 12. In these compounds the word has not undergone the loss of the initial velar nasal that is characteristic of kinship terms and pronouns in Arabana-Wangkangurru (see Hercus 1979). 16 The survival of *ngapa in AW is therefore compatible also with a hypothesis that would treat this word as the result of independent inheritance apart from a Karnic subgroup. 17 This may not be a genuine AW word as ngapatya ‘having (internal) water’ appears to be formed with the ‘having’ suffix -tya, which is not AW. 18 This verb is also found in Diyari. 19 It is likely that lhuka is derived from a word kadluka ‘woman’ which is attested as for Charlotte Waters, SA in Taplin (1879: 147), and as and respectively for Wangkangurru and Kuyani in Reuther (1981 vol. 4, Wonkanguru and Kujani, item no. 211 ‘a woman who has an inferior position’).

46

Harold Koch and Luise Hercus

compounds

literal gloss

gloss

mara-ngandi thidna-ngandi ngandi-thirri

hand-mother foot-mother mother-aggression

thumb big toe heavy fighting club20

Table 12: AW reflexes of *ngandi ‘mother’

This form *ngandi must therefore have been an innovation within the Karnic family at some level higher than Proto Central Karnic. It should be noted that there is evidence that an even older, Proto Pama-Nyungan form *ngama (Alpher 2004: 486–87) survived among the Karnic languages, in the form of: Wangkumara ngamatya, Pitta-Pitta ngamari, (ng)ama in Wangkangurru and the neighbouring western Pitta-Pitta dialects, and even in the Central Karnic language Yandruwandha, with a shifted sense, as ngama ‘mother’s brother’ (Breen 2004: 57).

4.2.3 *mawa ‘hunger’ This and the next example show partial cognates of Central Karnic nouns, which occur in Arabana-Wangkangurru in the form of verbs that have been presumably derived from erstwhile nouns. Austin (1990: 190) reconstructs *mawa ‘hunger’ for Proto Central Karnic. There is an apparent cognate form in Arabana-Wangkangurru, which was rare and used only as a verb: mawi- ‘to long for’, ‘to lust after’. This verb, which was intransitive, is not attested in the meaning of ‘being hungry for food’. An expression that was used several times by Wangkangurru speakers was mawinhangkarda- ‘he is continually after (women)’; here -nhangkarda is explained as a marker of continuous aspect. There is furthermore a placename Mawirani, a site near the Kallakoopah bend in the Simpson Desert, in Wangkangurru country, but very close to the Ngamini boundary. The interpretation of this name remains uncertain, however. In Diyari and Yandruwandha the stem was mawa-, not ‘mawi-’. There was a noun mawa ‘hunger, famine’ (Reuther ) and an adjectival derivative mawali ‘hungry’ (Reuther ). There are, however, in the Reuther’s Diyari examples instances where the meaning is figurative, as in Arabana-Wangkangurru.

20 The term is also used to refer to offensive weapons in general.

Obscure vs. transparent cognates in linguistic reconstruction

47

(1) ngani mauarli potuni = I am hungry for possessions, i.e. ‘I would like to have more possessions’ (2) mauarli narini = hungry for the deceased, i.e. ‘desirous to see the grave of the deceased’ (3) noani mauarli = hungry for one’s marriage partner, i.e. ‘desirous to have him or her present’ These expressions leave us with little doubt that Arabana-Wangkangurru mawi- is indeed cognate with Diyari and Yandruwandha mawa and with the corresponding Yaluyandi and Ngamini words for ‘hungry’, muwandu, muwanu.21 This lexical innovation must therefore go back further than Proto Central Karnic. 4.2.4 *kangi ‘jester’ Austin (1990: 189) reconstructs *kangi ‘jester’ for Proto Central Karnic. This word was recorded by Peter Austin in Diyari and is listed by Reuther also for Yawarrawarrka and Yandruwandha. There is a transitive verb kangi-ma- ‘to tease’ in Arabana-Wangkangurru, which is presumably derived from kangi, and is therefore a partial cognate with the Diyari, Yandruwandha and Yawarrawarrka nouns. In the light of these cognates, this term, like *ngandi ‘mother’ and *mawa/i‘hunger’, must have been an innovation within Karnic at a level higher than Proto Central Karnic.

4.2.5 A disguised pronominal form anhari Arabana-Wangkangurru shares much linguistic material with the Northern or Pitta-Pittic branch of Karnic. Within the paradigms of first person singular pronoun, shown in Table 1322, however, the dative forms appear to differ between the two languages. Arabana-Wangkangurru anthirda is obviously not cognate with Pitta-Pitta nganyari. It is worth noting that the Arabana-Wangkangurru form doubles as a locative; in this language dative and locative are syncretized in pronouns, and the -irda suffix is normal for pronouns. So it is plausible that this form originated as a locative case form. There is in fact a trace of an earlier dative form that agrees with that of Pitta-Pitta (except for the probably analogical change of nh to ny in the latter, for which see Bowern (1998: 117), and the regular 21 The variation between a and u in the first syllable, and between a and i in the second syllable of the cognates is not, however, understood. 22 The Proto-Karnic reconstructions are cited from Bowern (1998: 125).

48

Harold Koch and Luise Hercus

Proto Karnic Pitta-Pitta AW

Ergative

Nominative

Accusative

Dative

Locative

*ngathu ngathu athu

*nganyi ngantya antha

*nganha nganya anha

*ngantya nganyari anthirda

ngantyira anthirda

Table 13: 1Sg case paradigm in Pitta-Pitta and AW

loss of initial ng in the former): this is the adverbial form anhari ‘this way, towards me’. The apparent explanation is that Arabana-Wangkangurru has shared with Pitta-Pitta the innovation of creating a dative form based by adding -ri to the accusative form; Arabana-Wangkangurru has subsequently extended its locative case forms to cover the dative function as well, and preserved anhari as an extra-paradigmatic adverb. 4.2.6 A disguised verb nhangka- ‘sit’ In Arabana, nhangka- means ‘to lie down or sit down, to rest, to be slow, sluggish in doing something’. This verb has a clear cognate in Pitta-Pitta nhangka- ‘to sit’, ‘to be’, ‘to live’, with a derivative, in the Western dialects, nhangkali ‘alive’. It is also attested as an archaic form for Yaluyandi: nhangkarda means ‘sitting down’ and occurs in the traditional song of the Swan. The normal Yaluyandi word for ‘to sit’ is nguni-. (4) nhana puka’i nhangka’i she lignum sit ‘She sits down in the lignum.’ (This verb nhangka- may well be the same as a more widely used Karnic word meaning ‘to press down on’, ‘to weigh down on’, ‘to tread on’, which is found in Diyari, Wangkumara, etc.) This verb is hidden away in Wangkangurru and Arabana in the continuous aspect marker -n(h)angka-, as in wilpa-nhangkarda whistle-CONT-PRES ‘goes on whistling’ and in mawi-nhangka-rda ‘be continually lusting after women’, mentioned in Section 4.4. Cognates of the derived adjective nhangkali ‘alive’ of Pitta-Pitta are also found, but without any suffix, in Wangkangurru nhangka ‘alive’ and Arabana angka. There is an unexpected semantic extension of this word in Arabana: it can be said of a very greedy person: ipi-ipi angka tharnilira ‘he could eat a live sheep, i.e. an entire sheep’; in this context the word has come to mean ‘complete’. Reflexes of *nhangkaare shown in Table 14.

Obscure vs. transparent cognates in linguistic reconstruction

49

Pitta-Pitta Wangka-Yutjurru Yaluyandi Arabana

nhangkanhangkanhangkanhangka-

sit, stay, remain sit, live sit, stay lie down or sit down to rest, be slow, sluggish in doing something

AW Wangka-Yutjurru Wangkangurru Arabana

-n(h)angkanhangkali nhangka angka

CONTINUOUS ASPECT

alive alive alive

Table 14: Karnic reflexes of *nhangka- ‘sit’

5 Summary and Conclusion Our etymological discussions have illustrated Szemerényi’s principles with Australian data. They underline the fact that it is rarely safe to argue that a cognate is totally absent from any languages (this is especially the case with Australian languages, whose documentation is often incomplete and dates from a time when many languages were already in decline). In Section 2 it was shown that the widespread and solidly reconstructed Pama-Nyungan bodypart terms *mara ‘hand’ and *thina ‘foot’ are present in disguised, erstwhile compound and derivative forms in the Arandic subgroup, contrary to earlier beliefs. Section 4 similarly demonstrates that a number of Karnic words are present in ArabanaWangkangurru, and hence reconstructible to a level higher than Austin’s Proto Central Karnic or his “Proto Karnic” which is not treated as ancestral to ArabanaWangkangurru. Many of the Arabana-Wangkangurru forms cited are cognate with Central Karnic languages, while at least the last two are shared with Northern Karnic. Most of the cognates appear to represent lexical innovations within the Karnic subgroup or some subset of Karnic languages. They therefore support a scenario that puts Arabana-Wangkangurru squarely within the Karnic subgroup. The ‘creek’ words, discussed in Section 3, demonstrate that Karnic languages share this lexeme with the Thura-Yura and Yarli subgroups. This provides prima facie evidence for an early grouping that involves these three subgroups in a lexical innovation whose significance will have to be assessed in the light of further information. These historical conclusions have only been possible because our lexical comparisons have gone beyond exact translation equivalents (our “s-cognates”). We have compared vocabulary items that have shifted their sense, including grammaticalizing (from ‘sit’ to Continuous Aspect) or demorphologising (from

50

Harold Koch and Luise Hercus

‘1SgDAT’ to an adverb meaning ‘towards me’). We have also found many instances of partial cognates, where the earlier lexeme survives in a compound or an extended derivational form. Most of these cognates have been obscure in some sense. Our experience suggests that the stock of cognate sets among Australian languages can be greatly increased if comparativists look beyond the obvious cognates. We suggest that this is not only desirable but necessary if we are to gain a truer picture of how the languages are related historically. An adequate stock of cognates is required to establish (a) which subgroups of a language family (such as Pama-Nyungan) contain reflexes of a given etymon, (b) to what level (protolanguage or early areal group) a given etymon is reconstructible, and (c) what is the proper genealogical classification of the languages. Such a database of cognates can only be established by using a proper etymological method, one which goes beyond the obvious comparisons and seeks out the obscure cognates for the light they can provide on the linguistic past.

References Alpher, Barry. 2004. Pama-Nyungan: phonological reconstruction and status as a phylogenetic group. In: Claire Bowern & Harold Koch (eds.), Australian languages: classification and the comparative method (Current Issues in Linguistic Theory 249), 93–126, 387–574, 681– 687. Amsterdam: John Benjamins. Arlotto, Anthony. 1972. Introduction to historical linguistics. Boston: Houghton Mifflin. Austin, Peter. 1981. Proto-Kanyara and Proto-Mantharta historical phonology. Lingua 54. 295– 333. Austin, Peter. 1990. Classification of Lake Eyre languages. La Trobe Working Papers in Linguistics 3. 171–201. Austin, Peter. 1994. Diyari. In: Nick Thieberger and William McGregor (eds.), Macquarie Aboriginal words, 125–143. Sydney: The Macquarie Library. Austin, Peter. 1997. Proto Central New South Wales phonology. In: Tryon and Walsh (eds.), Boundary rider: essays in honour of Geoffrey O’Grady (Pacific Linguistics C-136), 21–49. Canberra: Australian National University. Bowern, Claire L. 1998. The case of Proto Karnic: morphological change and reconstruction in the nominal and pronominal system of Proto Karnic (Lake Eyre Basin). Canberra: Australian National University BA Honours thesis. Bowern, Claire L. 2001. Karnic classification revisited. In: Jane, Simpson, David Nash, Mary Laughren, Peter Austin & Barry Alpher (eds.), Forty years on: Ken Hale and Australian languages (Pacific Linguistics 512), 245–261 Canberra: Australian National University. Bowern, Claire & Harold Koch (eds.). 2004. Australian languages: classification and the comparative method (Current Issues in Linguistic Theory 249). Amsterdam: John Benjamins. Breen, J. G. 1971. Aboriginal languages of Western Queensland. Linguistic Communications 5. 1–88.

Obscure vs. transparent cognates in linguistic reconstruction

51

Breen, Gavan. 2001. The wonders of Arandic phonology. In: Jane, Simpson, David Nash, Mary Laughren, Peter Austin & Barry Alpher (eds.), Forty years on: Ken Hale and Australian languages (Pacific Linguistics 512), 45–69. Canberra: Australian National University. Breen, Gavan. 2004. Innamincka Words: Yandruwandha dictionary and stories (Pacific Linguistics 559). Canberra: Australian National University. Breen, Gavan. 2011. Reassessing Karnic: a reply to Bowern (2009). Australian Journal of Linguistics 31. 137–143. Capell, Arthur. 1956. A new approach to Australian linguistics. (Oceania Linguistic Monographs 1). Sydney: University of Sydney. Curr, Edward M. 1886–87. The Australian race. 4 vols. Melbourne: Government Printer/London: Trübner & Co. Dixon, R. M. W. 1980. The languages of Australia. Cambridge: Cambridge University Press. Dixon, R. M. W. 2002. Australian languages: their nature and development (Cambridge Language Surveys). Cambridge: Cambridge University Press. Hale, Kenneth. n.d. Arandic word list. Cambridge MA. unpublished manuscript. Harvey, Mark. 2003. An initial reconstruction of Proto Gunwinyguan phonology. In: Nicholas Evans (ed.), The non-Pama-Nyungan languages of northern Australia: comparative studies of the continent’s most linguistically complex region (Pacific Linguistics 552), 205–268. Canberra: Australian National University. Hercus, Luise A. 1979. In the margins of an Arabana Wangkangurru dictionary: the loss of initial consonants. In: Stephen A. Wurm (ed.), Australian linguistic studies (Pacific Linguistics C-54), 621–651. Canberra: Australian National University. Hercus, Luise A. 1988. Using other people’s words: a note on some compound nouns in Arabana, northern South Australia. In: T. L. Burton and Jill Burton (eds.), Lexicographical and linguistic studies: essays in honour of G. W. Turner, 73–80. Woodbridge, Suffolk: Boydell and Brewer. Hercus, Luise A. 1994. Grammar of the Arabana-Wangkangurru language, Lake Eyre Basin, South Australia (Pacific Linguistics C-128). Canberra: Australian National University. Hercus, Luise & Peter Austin. 2004. The Yarli languages. In: Claire Bowern & Harold Koch (eds.), Australian languages: classification and the comparative method (Current Issues in Linguistic Theory 249), 207–222. Amsterdam: John Benjamins. Hock, Hans Henrich. 1986. Principles of historical linguistics. Berlin: Mouton de Gruyter. Koch, Harold. 1997. Pama-Nyungan reflexes in the Arandic languages. In: Darrell Tryon & Michael Walsh (eds.), Boundary rider: essays in honour of Geoffrey O’Grady (Pacific Linguistics C-136), 271–302. Canberra: Australian National University. Koch, Harold. 2001. Basic vocabulary of the Arandic languages: from classification to reconstruction. In: Jane Simpson, David Nash, Mary Laughren, Peter Austin & Barry Alpher (eds.) Forty years on: Ken Hale and Australian languages (Pacific Linguistics 512), 71–87. Canberra: Australian National University. Koch, Harold. 2003. Morphological reconstruction as an etymological method. In: Barry J. Blake and Kate Burridge (eds.), Historical Linguistics 2001: selected papers from the 15th International Conference on Historical Linguistics 13–17 August 2001, 271–291. Amsterdam: John Benjamins. Koch, Harold. 2004a. A methodological history of Australian linguistic classification. In: Claire Bowern & Harold Koch (eds.), Australian languages: classification and the comparative method (Current Issues in Linguistic Theory 249), 17–60. Amsterdam: John Benjamins.

52

Harold Koch and Luise Hercus

Koch, Harold. 2004b. The Arandic subgroup of Australian languages. In: Claire Bowern & Harold Koch (eds.), Australian languages: classification and the comparative method (Current Issues in Linguistic Theory 249), 127–150, 575–580. Amsterdam: John Benjamins. Menning, Kathy & David Nash. 1981. Sourcebook for Central Australian languages. Alice Springs: Institute for Aboriginal Development. Miceli, Luisa. 2004. Pama-Nyungan as a genetic entity. In: Claire Bowern & Harold Koch (eds.), Australian languages: classification and the comparative method (Current Issues in Linguistic Theory 249), 61–68. Amsterdam: John Benjamins. O’Grady, Geoffrey N. & Terry Klokeid. 1969. Australian linguistic classification: a plea for coordination of effort. Oceania 39. 298–311. O’Grady, Geoffrey N. & Darrel T. Tryon (eds.). 1990. Studies in comparative Pama-Nyungan (Pacific Linguistics C-111). Canberra: Australian National University. O’Grady, Geoffrey N., C. F. Voegelin & F. M. Voegelin. 1966. Languages of the world: Indo-Pacific Fascicle 6. Anthropological Linguistics 8 (2).1–199. O’Grady, Geoffrey N., Stephen A. Wurm & Kenneth L. Hale. 1966. [Map of] Aboriginal Languages of Australia (a preliminary classification). Victoria, B.C.: Department of Linguistics, University of Victoria. Reuther, J. G. 1981. The Diari. Transl. P. A. Scherer. Canberra: Australian Institute of Aboriginal Studies. microfiche. Simpson, Jane, David Nash, Mary Laughren, Peter Austin, Barry Alpher (eds.). 2001. Forty years on: Ken Hale and Australian languages (Pacific Linguistics 512). Canberra: Australian National University. Simpson, Jane & Luise Hercus. 2004. Claire Bowern & Harold Koch (eds.), Australian languages: classification and the comparative method (Current Issues in Linguistic Theory 249), 179– 206, 581–646. Amsterdam: John Benjamins. Sommer, Bruce A. 1969. Kunjen phonology: synchronic and diachronic (Pacific Linguistics B-11). Canberra: Australian National University. Szemerényi, Oswald. 1977. [1962] Principles of etymological research in the Indo-European languages. In: Rüdiger Schmitt (ed.), Etymologie (Wege der Forschung 373), 286–346. Darmstadt: Wissenschaftliche Buchgesellschaft. Taplin, George (ed.). 1967. [1879] The folklore, manners, customs and languages of the South Australian Aborigines. New York: Johnson Reprint Corporation. Teichelmann, C. G. & C. W. Schürmann. 1982. [1840] Outlines of a grammar, vocabulary & phraseology of the Aboriginal language of South Australia, spoken by the natives in and for some distance around Adelaide. facsimile edn. Largs Bay, Adelaide: Tjintu Books. Tindale, Norman B. 1976. Tribal boundaries in Aboriginal Australia [Map]. Canberra: Australian Institute of Aboriginal Studies. Trask, Robert L. 2000. The dictionary of historical and comparative linguistics. Edinburgh: Edinburgh University Press. Tryon, Darrell & Michael Walsh (eds.). 1997. Boundary rider: essays in honour of Geoffrey O’Grady (Pacific Linguistics C-136). Canberra: Australian National University. Wurm, Stephen A. 1972. Languages of Australia and Tasmania. The Hague: Mouton. Wyatt, William. 1879. Vocabulary of the Adelaide and Encounter Bay tribes, with a few words of that of Rapid Bay. In: J. D. Woods (ed.), The Native Tribes of South Australia, 169–182. Adelaide: E. S. Wigg & Son.

Harold Koch

The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered 1 Introduction 1.1 Etymology and reconstruction Etymology is intrinsically involved in the process of comparative reconstruction. To establish a set of cognates and reconstruct their proto-form is to provide an etymology for each of the cognates. A given word is etymologized if it can be derived from a proto-form in an ancestral language with differences from the proto-form being accounted for by plausible (phonological, semantic, and morphological) changes.1 Etymological explanations apply both to general lexemes and to grammatical words and affixes. Reconstruction of grammatical paradigms from a set of related languages requires more than a simple comparison of the forms that occur in equivalent paradigmatic cells of functionally equivalent lexemes (although it is wise to start from these); successful reconstruction further depends on the consideration of the kinds of grammatical changes that are likely to affect the grammatical words, which in turn helps in identifying as cognates forms which are not found in the equivalent functional slots (Koch 2003a2, 20093). This paper 4 will discuss the evidence for reconstructing a 3rd Singular Feminine personal pronoun *nhan (and its inflectional paradigm) in Proto PamaNyungan (pPN). Doubts have been expressed about the strength of this evidence. The comparative evidence will be expanded by introducing further cognate data

1 Two other ways of accounting for the etymology of a word are to explain its origin as being borrowed from another language or as being formed by word-formation processes in an earlier stage of the language. 2 This paper emphasizes the relevance to grammatical reconstruction of: shifts in the meaning of grammatical forms, the development of morphological markers from syntactic constructions, and the absorption of morphological elements into lexical stems. 3 This paper exemplifies, by means of Pama-Nyungan deictic/anaphoric stems, the reanalysis of inflected forms as stems and the separation of particular inflected forms from their paradigm to become adverbs, particles, etc. 4 An earlier version of this paper had a different focus and was circulated under the title “Wide-scope vs. narrow-scope approaches to linguistic reconstruction: The Pama-Nyungan 3SgF pronoun reconsidered”. I am grateful for comments on this and/or the earlier version to Barry Alpher, Gavan Breen, Claire Bowern, Alan Dench, Mary Laughren, and Amanda Lissarrague. I take full responsibility, however, for the way I have used their comments.

54

Harold Koch

and justifying its relevance by showing that the prehistoric changes they presuppose are consistent with the obvious changes that can be observed within specific languages closer to the present.

1.2 The Pama-Nyungan language family The Pama-Nyungan languages of Australia are widely accepted as constituting a language family (Blake 1988; Koch 2003b; Alpher 2004; O’Grady and Hale 2004; Evans 2005; Campbell and Poser 2008). There is widespread agreement on which languages belong to this family and how most of the languages fit into low-level subgroups (see Bowern and Koch 2004). But how these subgroups fit into higher-level groupings is a matter of continuing uncertainty.

1.3 Pama-Nyungan pronoun reconstruction The system of personal pronouns for Proto-Pama-Nyungan has been reconstructed with a considerable degree of certainty. The supporting forms and discussion are given in Dixon (1980), Blake (1988), and Dixon (2002). The singular forms with their reconstructible cases are given in Table 1, quoted from Koch (2003b).5 The non-singular forms, given in Table 2, are quoted from Sutton and Koch (2008). A major point of uncertainty is the 3SgFem. A form *nhan (reflected as nyan in languages which lack a contrast between lamino-dentals and lamino-palatals) has been recognized, but its reflexes have been considered to be too limited in number and geographical distribution to inspire certainty that the form goes back as far as Proto-Pama-Nyungan rather than to a subset of Pama-Nyungan such as “eastern Pama-Nyungan”.

1Sg 2Sg 3SgM 3SgF

ERG

NOM

ACC

DAT

*ngathu *nyuntu *nhulu *nhantu

*ngay *nyun *nhu *nhan

*nganha *nyuna *nhunha *nhana

*ngatyu *nyunu *nhungu *nhanu

Table 1: Reconstructed Pama-Nyungan singular pronouns

5 Note on orthography: long vowels are written double; the velar nasal is spelled ng; laminodentals are spelled th, nh; lamino-palatals are spelled ty, ny; r represents an approximant and rr a tap/trill; stops, which do not distinguish voicing, are spelled with voiceless symbols. Spellings of the sources are standardized into this system, except where indicated otherwise.

55

The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered

Dual Plural

1st

2nd

3rd

*ngali *ngana

*nhumpalV *nhurra

*pula *thana

Table 2: Reconstructed Pama-Nyungan non-singular pronouns

1.4 Aim of this paper In this paper I will reconsider the evidence for this proto-paradigm, expand the comparative data, and discuss methodological issues involving what can be compared (and what kinds of changes are implied) and how much evidence is required for a successful reconstruction.

2 The current state of knowledge 2.1 Previous approaches to the problem Dixon (1980: 277) comments on the rarity of distinct masculine and feminine pronouns in Australian languages, but indicates their presence in a number of languages now recognized as Pama-Nyungan: Yanyuwa, Diyari, and the Western Torres Strait (WTS) language. He later mentions that the feminine form matching a 3Sg Masculine pronoun *nhu- is nyaankan in Bandjalang, and nhan- in Diyari (Dixon 1980: 359), with a parallel nhan(pa)- in Pitta-Pitta (Dixon 1980: 494). Alpher (1987) discusses the default uses of the Kala Lagaw Ya (the Western Torres Strait language) 3SgF pronoun na vis-à-vis the masculine nuy (as well as demonstrative forms built on the pronoun stem). He finds a similar usage in the Wangkumara third-person non-masculine form (in the ergative case) nhandrru, which is opposed to a 3SgM (ergative) nhulu (Alpher 1987: 174). He points to cognate pronouns in Diyari (ergative nhandru), Pitta-Pitta, and Yandruwandha. Alpher explicitly proposes the reconstruction of a 3SgF for pPN. . . . the shapes of the third person singular pronouns in these and other eastern languages, like Bandjalang . . . show such resemblances to those of Kala Lagaw Ya as to suggest strongly the reconstructions *nyu- (masculine) and *nya- (feminine) for the common protolanguage. . . . Because language diversity within the Pama-Nyungan family is greatest in the eastern half of the continent, it is highly probable that these reconstructions are (at least) Proto-Pama-Nyungan in age. (Alpher 1987: 174)

Blake (1988: 6) presents a set of reconstructed pronouns as “characteristic of Pama-Nyungan”; these are “largely based on the set of pronouns reconstructed

56

Harold Koch

in Dixon 1980 as ‘proto-Australian’”. These are reproduced in Table 3. Blake notes that the 3Sg forms are less widely attested than the other forms, because of the prevalence of zero anaphora and the use of deictics instead of genuine 3Sg pronouns. Two forms, possibly etymologically related, are presented for the 3Sg, which functions as a masculine in languages which have a gender contrast.6

1 2 3

singular

dual

plural

*ngay *ngin *nyu (east) *ngu (west) *nyan (east feminine)

*ngali *nyuNpalV *pula

*ngana *nyurra *tyana

Table 3: Blake’s reconstructed Pama-Nyungan pronouns

Note that Blake indicates the 3Sg *nyan (east feminine) with the comment: “The feminine form nyan is opposed to nyu in only a few languages such as Pitta-Pitta. . . , Diyari and Bandjalang”. Pitta-Pitta and Diyari (as well as the Wangkumara and Yandruwandha cited by Alpher) belong to the Karnic subgroup;7 Bandjalang is a multi-dialect language of the east coast. Blake also finds the pronominal root *nhan (his *nyan) reflected in two languages of the Warluwarric subgroup, Wagaya and Yanyuwa8 (Blake 1988: 22, 65). A reconstruction to a Proto-Eastern-Pama-Nyungan is implied by his later identification of a Yolngu oblique stem *nhan- (of a 3Sg non-gendered pronoun) as “the feminine root of the eastern Pama-Nyungan languages” (Blake 1988: 26) – see Section 3.6 below on the Yolngu languages. In sum, the comparative evidence cited by Blake comes from four subgroups: Karnic, Warluwarric, Bandjalang, and Yolngu (where the meaning is no longer feminine). It is worth noting that no one has explicitly justified an “Eastern Pama-Nyungan” subgroup, and it is doubtful whether the Warluwarric and Yolngu subgroups would belong to it if there

6 The comparative evidence for the paradigms of *nhu (Blake’s *nyu) and *ngu is discussed in Koch (2009). 7 The classification of Karnic languages has been controversial: see Austin (1990), Bowern (1998, 2001, 2009), Dixon (2002: xxxvii), Breen (2007, 2011). I follow Bowern’s subgrouping, since it is based most explicitly on the criterion of common innovations. 8 Dixon (1980: 277) had mentioned Yanyuwa as a language with gender-distinguished 3Sg pronouns and Alpher (1987: 174) had cited the Western Wagaya non-masculine yamb as an example of feminine as the default gender, but neither claimed that these forms were cognate with the *nhan found in the Karnic languages, Bandjalang, and Western Torres Strait language.

The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered

57

were such a higher-level grouping. The clear implication of Blake’s comparative evidence is that the pronoun should be reconstructed to the Proto-PamaNyungan level, which Alpher had also considered to be “highly probable”. Dixon (2002: 305–6, 461–3) discusses 3SgF pronouns in the set of languages usually called Pama-Nyungan (although he does not accept Pama-Nyungan as a genetic group). He notes that a distinctive feminine pronoun is found in just over a dozen of the approximately 190 Pama-Nyungan languages – occurring only in languages in the south-eastern and east-central part of the Pama-Nyungan area. He also notes that in many of these languages the 3SgF form can be derived from an earlier form *nha(a)n, and furthermore that the inflectional suffixes ergative -tu and accusative -(nh)a can be reconstructed for this stem. The clearest evidence for *nhan is from the Karnic subgroup (Dixon’s WA areal grouping), forms of which are shown in his Table 7.5 (Dixon 2002: 305); the Warluwarric (his WM) subgroup, where through lenition of the laminal nasal it appears as *yan; the eastern language Bandjalang (his M group), where nyaan-kan includes an explicit productive feminine suffix -kan, which is widely found in east coast languages; nan from the far north-eastern Western Torres Strait (WTS) language (in his A group) is “conceivably” related (Dixon 2002: 462). In addition to these two subgroups plus two separate languages (which as isolates within PamaNyungan count as separate subgroups), he gives possibly related forms (there are doubts about their transcription) from three more languages of the southeast, from coastal New South Wales south of Bandjalang – Gathang, Awabakal, Darkinyung – without providing an analysis of how the forms relate to the reconstructed paradigm. These languages have been classified into a Kuri subgroup (Wurm 1972: 137–138). Altogether *nha(a)n occurs in six of Dixon’s 37 “areal groups” that comprise the traditional Pama-Nyungan family. I display his forms in Table 4 – ordered according to the reconstructed forms he implies. Given Dixon’s rejection of the Pama-Nyungan family and his reluctance to reconstruct, it is difficult to see what conclusion can be drawn. It appears that he suggests that a form *nha(a)n is ancestral to at least the east-central Warluwarric and Karnic groups plus Bandjalang in the east; possibly also to some languages to the south of Bandjalang, and “conceivably” the Western Torres Strait language, although he elsewhere states that this is a “Papuan” language with a “significant Australian substratum” (Dixon 2002: xxx). It seems pretty clear that he would not see the form as ancestral to all the Pama-Nyungan languages. The tenor of this book would rather suggest that he would seek an explanation in terms of diffusion of the form (or paradigm of forms) across a geographically limited area. Bowern (1998: 159), as background to her reconstruction of the Karnic pronouns, brings further forms into comparison. In a later paper she explicitly

58

Harold Koch

Group

Language earlier form

ERG *nha(a)ntu

NOM *nhaan

ACC *nha(a)nnha

WA WM M N N O A

Karnic pWarluwarric Bandjalang Kattang Awabakal Darkinyung Western Torres Strait

nhantu

nhan*yannyaan[kan] nyunh-

nhan(h)a

puwantuwa nun-tuwa nan

Table 4: Dixon’s evidence for *nha(a)n

posits a pPN 3SgF paradigm based on a stem *nhan (Bowern 2001: 252). “Some other languages also have a third person singular stem *nhan-, such as Djaru (Tsunoda 1981) and Yolngu (Waters 1989). These languages would thus appear to show a relic of an earlier feminine stem which has been bleached of gender specification and generalized to the third person singular” (Bowern 1998: 159). In Koch (2003b) I added cognate forms from several more languages and suggested four case-forms of its inflectional paradigm in Proto-Pama-Nyungan – though without giving full justification. Some of this evidence will be expanded in the following sections. Another pronominal stem ending in n [in addition to 2Sg *nyun and *nga(a)n ‘who’] is the 3rd person singular feminine stem *nhan- (given in Table 9). It appears to continue an inflectional pattern identical to that of 2Sg. Reflexes of this paradigm are widespread enough to allow reconstruction to a proto-language ancestral to languages from most parts of the continent. . . . The Walmajarri and Warlpiri forms . . . point to the former presence of the paradigm in the Ngumpin-Yapa subgroup (in the Northwest), although the contrastive meaning of feminine has been lost. The same explanation holds for Djapu (of the Yolngu subgroup in the north) and Biri (of the Maric subgroup in the inner east). (Koch 2003b)

2.2 Outstanding issues: reasons for uncertainty It is worth considering what are the reasons for a lack of certainty about reconstructing *nhan to Proto-Pama-Nyungan, especially on the part of Dixon and Blake. I suggest that it is related to the number and geographical distribution of the subgroups and language isolates in which reflexes of *nhan are found. Dixon restricts himself to pronouns that function synchronically as a feminine, and finds reflexes in only five subgroups (according to my classification): Karnic, Warluwarric, Bandjalangic, Kuri, and Western Torres Strait – with some doubt about the Western Torres Strait. These subgroups have an easterly distribution, although Warluwarric extends into the Northern Territory. Blake compares

The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered

59

cognates only from the first three of these, but adds Yolngu – not an eastern subgroup – even though the meaning is no longer feminine. Bowern mentioned further cognates – also without feminine function – from the western subgroup Ngumpin-Yapa: she mentioned Jaru (Djaru), to which Koch added Walmajarri and Warlpiri. If all of these subgroups are taken as attesting to reflexes of *nhan, this is still only seven subgroups out of some thirty or more that constitute the Pama-Nyungan family. One might ask whether this evidence is sufficient to reconstruct the form to Proto-Pama-Nyungan. How many independent witnesses do we need? What areal distribution should they display? Another issue is whether it is legitimate to include as cognates forms that do not have a feminine function, although their phonological form would allow them to be cognate. And how do we account for the fact that the majority of subgroups do not make a distinction between masculine and feminine in the pronoun? There are in fact only two Pama-Nyungan languages (to my knowledge) that distinguish gender and whose feminine pronoun is not related to *nhan. In Yaygirr on the east coast of New South Wales the form of the 3SgF is ngangkula-wan and this is clearly built on the 3SgM ngangkula using what is probably a reflex of the eastern feminine suffix -kan (Dixon 2002: 461–2). Dyirbal, in the north-eastern rainforest region, makes a gender distinction in article-like “noun markers”; the feminine of ‘that-Visible’ is pala-n, where the final n marks feminine. This feminine marker -n contrasts with -m, which marks an edible class, and -l (-yi in Absolutive case) which marks masculine. These markers appear after the case suffix on “noun markers”, i.e. demonstratives and words for ‘somewhere’ and ‘where’ (see Dixon 1972: 44–47, 254–59). Dixon has not explained the origin of the gender suffixes, except that he assumes they were added “at a rather late stage” to case-inflected demonstratives and suggests that -m may somehow continue a cliticized form of the widespread generic food term mayi (Dixon 1972: 353). Alpher (1997) presents a scenario in which the Dyirbal -n (in noun markers, as well as certain noun stems in a number of North Queensland languages, is ultimately a reflex (via -*ny) of a cliticized feminine pronoun of the shape *nya(a), but admits that this hypothesis becomes more difficult if the original pronoun was rather *nyan (as argued here). Alpher’s hypothesis is conceivable if the -n represents rather the final consonant of the pronoun stem, which lost its initial CV in the cliticization process; i.e. *pala=(nya)n. An alternative etymology would be to derive the -n from the nominal feminine derivational suffix -kan that occurs in many languages of the east coast of Australia.9 Both the Dyirbal and the Yaygirr 9 An earlier *pala-kan may have developed by sound change to palan just as *kutaka resulted in *kuta ‘dog’.

60

Harold Koch

feminine forms thus appear to be recent creations and hence have no bearing on the reconstruction of *nhan. The issues canvassed here will be revisited in Section 4.

3 Reconsideration and expansion of the evidence It is worth reconsidering the cognate forms and the historical arguments that can be drawn from them. The case for reconstructing *nhan back to ProtoPama-Nyungan can, I believe, be strengthened. The evidence can be expanded in two dimensions: further inflectional forms can be added to the comparison, and the number of subgroups where the pronoun is attested can be increased, especially beyond those cited by Dixon, if we add instances of forms attested with a changed meaning and/or analysis. In this section I will discuss the evidence according to subgroups, beginning with those that attest reflexes of *nhan as a feminine pronoun, then adding those where it has a different sense. I shall call attention to the kinds of changes that the forms have undergone in each subgroup, since these will be of importance in judging the cognacy of more doubtful forms. The results of this revised comparison are collected on Table 11 in Section 4.1.

3.1 Attestation and changes in the Karnic subgroup The best evidence for *nhan from a single subgroup comes from the languages classified as belonging to the Karnic subgroup. This includes Diyari and PittaPitta (cited by both Dixon 1980 and Blake 1988) as well as Wangkumara and Yandruwandha (cited by Alpher 1987 and Dixon 2002). Austin (1990: 177, 194) finds cognate forms in Ngamini, Yaluyandi, Yawarrawarrka as well. Bowern’s comparison (1998: 173, cf. 2001: 252) further adds Punthamarra. Although these scholars differ on the membership and status of this group (Austin includes fewer languages than Bowern, and Dixon more, but treats them as an areal rather than a genetic group), they are in rough agreement about the reconstruction of the three basic case forms (see Table 5). All reconstruct *nhantu as the ergative. Bowern and Dixon reconstruct the nominative as *nhan, and assume that different languages have added an augment -i or -pa, whereas Austin reconstructs *nhani. For the accusative Bowern reconstructs *nhana, even though more languages have nhanha (which Austin reconstructs), and suggests that the latter represents a replacement on the analogy of more common forms with accusative -nha. Dixon suggests an earlier *nhannha, with separate reductions

61

The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered

of consonant cluster n + nh to either n or nh in the different lects; this reconstruction is implausible given the rarity of clusters of this kind in Pama-Nyungan languages. Bowern further posits a locative form *nhanta and a dative form *nhanku, on the basis of a single language Pitta-Pitta. While nhanta has the expected form for a locative of a Proto-Pama-Nyungan pronominal stem ending in n (Koch 2003b), nhanku appears to be an analogical innovation (Bowern 2001: 252), and presumably replaced the expected *nhanu posited in Koch (2003b).

Bowern 1998, 2001 Dixon 2002 Austin 1990

ERG

NOM

ACC

*nhantu *nhantu *nhantu

*nhan *nhan *nhani

*nhana *nhannha *nhanha

Table 5: Karnic reconstructions

The Karnic evidence thus yields three case forms arrived at by reconstructions within a small language family – ERG *nhantu, NOM *nhan, and ACC *nhana. I follow Dixon, Bowern (as well as Blake 1988 and Alpher 1987) in treating these as potential reflexes of a much earlier ancestral language. It should be noted that Austin (1990: 177) rather treated the “development of a gender contrast in third person singular pronouns” as an innovation that served to define his Proto-Karnic (which excluded Arabana-Wangkangurru).10 It is worth noting what subgroup-internal changes are involved, if Proto Karnic had the paradigm we suggest. First, individual members of the paradigm underwent some changes in form. The nominative *nhan was reinforced by the addition of an increment -pa (in Pitta-Pitta) and -i (in a number of languages). The accusative *nhana altered its consonant to nh, in imitation of the -nha in the 3SgM pronouns, in several languages. The inherited dative *nhanu was replaced in all the languages by forms, usually nhangka or nhanka, of analogical origin (see Bowern 1998: 170–173 for the developments in this paradigm). Second, there were semantic changes: the 3SgF category extended its reference to include other non-masculine referents in the eastern Karnic language, Wangkumara, as noted by Alpher (1987): “In Waŋkumara (Garlali) there is no consistent use of ‘masculine’ form. . . . If a gender distinction exists in Waŋkumara (Garlali) it is very marginal – the non-masculine forms appear to have generalized to the normal use” (McDonald and Wurm 1979: 29). In the closely related 10 In a later unpublished manuscript, Austin and Hercus (1993), Arabana-Wangkangurru was included within Karnic and *nhan was reconstructed to the level of the revised Proto-Karnic.

62

Harold Koch

Punthamara, the feminine pronoun, extended by means of deixis markers, is used as an “impersonal demonstrative” to refer to non-animates (Holmer 1988: 141). Also interpretable as a semantic change is the fact that some languages lost the masculine vs. feminine contrast completely. Wangka-Yutjuru, the western neighbour of Pitta-Pitta, created a single paradigm for a genderless 3Sg, which includes erstwhile members of both the 3SgM and the 3SgF paradigms: nhuluand nhungu- from the former, nhanpa-, nhana- and nhanka- from the latter (see Blake and Breen 1971: 78–81). Arabana-Wangkungurru lost the feminine forms without a trace, replacing both the masculine and the feminine 3Sg by a new form uka, whose origin is uncertain (Bowern 1998: 160–161). A third kind of change affected the syntactic status of the pronoun. In a number of Karnic languages, the 3SgF pronoun, along with the 3SgM, underwent a split in function; when used as a definiteness or specificity marker (a kind of article) with an accompanying noun, the pronoun occurred cliticized to the last word of the noun phrase, was reduced formally through the loss of initial consonants, and ended up as a case clitic marking a combination of gender, number, and case; e.g. Punthamara walka-andru ‘woman-Sg.F.ERG’ (Holmer 1988a: 137). This change is an instance of grammaticalization, involving the bleaching of semantic features (here the definiteness component), the loss of phonological independence, and the reduction of its phonological bulk. We shall see that similar kinds of change recur in other Pama-Nyungan subgroups. These will inform our judgments about what forms can be treated as cognate and why related forms may be absent in some languages.

3.2 More forms of the Bandjalang paradigm Bandjalang has more to offer than the erstwhile nominative form *nyaan, augmented by the productive feminine suffix -kan, mentioned by several scholars (cf. 2.1). This form nyaankan is found in three dialects (Sharpe 1996: 59), and in Waalubal it serves as the inflectional base for other forms of the paradigm: ERG nyaankan-tu, ACC nyaankan-i, GEN nyaankan-aa, DAT/ALL nyaankan-paa (Crowley 1978: 78–9). An unaugmented nominative nyaan, said to be a variant of nyaankan, is cited by Smythe (1978: 371–372), a grammar which was based on information recorded in the 1940s from the Casino district (Crowley 1978: 251). The same form (spelled nyan) is given by an early (1892) source (Livingstone) for the Minyangbal dialect (Sharpe 1998: 25) and probably underlies the Yugambeh dialect form n’yahm given as ‘female’ by Hanlon in 1935 (Sharpe 1998: 145). The fact that nyaan is attested from three different dialects and in relatively early sources

The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered

63

guarantees it as a genuine old pronominal form within the Bandjalang linguistic group. An ergative case form nyaantu is also citable from Smythe (1978: 371– 372), alongside an accusative nyaani and an oblique stem (used as a base for the dative and possessive) nyaana-, which appears to be cognate with the Proto-Karnic accusative form. An examination of Smythe’s Text 4 yields a whole paradigm based on nyaan, as demonstrated in Table 6 (using his spellings). case

form

sentence

NOM ERG ACC DAT POSS

nja:n nja:ndu nja:ni nja:nagei nja:nagu

2, 5, 6, 8, 15, 21, 22, 23 12, 14, 21 7, 13, 16, 17, 18, 24, 25 19 22

Table 6: Paradigm of Bandjalang nyaan 3SgF from Smythe’s Text 4

The Bandjalangic subgroup thus provides evidence for three inherited forms of the paradigm, NOM nyaan, ERG nyaantu, and an oblique stem nyaana-, which match the Proto-Karnic NOM *nhan, ERG *nhantu, and ACC *nhana (the correspondence of ny to Karnic nh is regular in a language that has only a single laminal series). The following changes have taken place: NOM nyaan has the explicit feminine suffix -kan added to it in several dialects; the inherited ACC *nyaana has new case forms built on it; a new ACC nyaani is formed, using a suffix -i that is regular for nominals in the language. The Yugambeh dialect has largely replaced the feminine stem nyaan by a new pronoun created by adding the feminine suffix –kan to the 3Sg Masculine pronoun nyula, which continues the pPN 3SgM Ergative form *nhulu; a whole paradigm is built on the new inflected stem of the 3SgF, nyulakan (Sharpe 1998: 27); the same situation obtains in the Gidabal dialect (Geytenbeek and Geytenbeek 1971: 15). This process resembles the Yaygirr creation of a feminine pronoun from its corresponding masculine form (cf. 2.2).

3.3 Refinement of the Western Torres Strait evidence The Western Torres Strait (WTS) Island language exists in two dialects called Kala Lagaw Ya (KLY) and Kalaw Kawaw Ya (KKY). Although the relation of this language to other Australian languages has been disputed, Evans (2005: 254–256) and Alpher, O’Grady and Bowern (2008) provide strong evidence for its membership in the Pama-Nyungan family. The latter claim, in fact: “The pronoun

64

Harold Koch

paradigms . . . constitute some of the most striking evidence for the PN membership of WT” (Alpher, O’Grady and Bowern 2008: 15). Dixon (2002: 462) cited the 3SgF pronominal form nan as a possible cognate to the 3SgF forms he compared. The initial consonant n is regular as a reflex of pPN nh; cf. 3SgM nu- reflecting pPN *nhu- (Alpher, O’Grady and Bowern 2008). The case forms in both dialects are actually NOM na, ERG nadh, ACC nan, and Possessive nanu. The ERG and ACC suffixes -dh and -n respectively recur in the 2Sg and 3SgM pronouns as additions to the stem found in the NOM, so may be language-internal innovations.11 This leaves NOM na and POSS nanu as possibly inherited forms: nanu is a likely reflex of a pPN Dative *nhanu (as the KLY 3SgM POSS nungu apparently reflects the pPN Dative *nhungu); the NOM na shows loss of final n, a change which recurs in 2Sg ni, which Alpher, O’Grady and Bowern (2008) derive from pPN *nhun.12 Curiously, however, these authors do not posit an n-final stem for the 3SgF proto-form, but reconstruct it as *nha. Another pPN n-final stem, *nga(a)n ‘who’ is parallel to the 3SgF paradigm in lacking the n in the NOM, having final n with no following vowel in the ACC, and having a POSS form with final u: Bani and Klokeid (1972: 76) give the NOM, ACC, and GEN for KLY as respectively ngaa, ngan, and ngunu and Holmer (1988b: 28) gives the NOM and POSS of KKY as nga and ngunu (both dialects showing a obscure vowel change in the GEN/POSS). The 3SgF pronoun of the Wester Torres Strait language has undergone the following changes. In terms of inflectional forms, it has created a new ERG na-dh and a new oblique stem nabi- (KLY) or nabe- (KKY), and built an Ablative form on the inherited Dative form nanu, which itself serves as a Possessive. Semantically it has expanded the scope of feminine to include third person singulars other than masculines, as explained by Alpher (1987) and illustrated in Bani and Alpher (1987). This resembles the situation in the Eastern Karnic language Wangkumara.

3.4 More detail on forms from the Kuri languages The languages of the New South Wales central coast, mentioned briefly by Dixon as providing possible support of *nhan, have more to offer than has been exploited so far. These languages were classified as a Kuri group by Schmidt 11 The 3SgM ACC nuyn clearly has added an inflectional -n to the NOM Sg nuy; however, the Accusative forms of 2Sg and 3Sg, nin and nan respectively, could continue the Proto-PamaNyungan forms that end in -na, since the evidence for final vowel loss is ambiguous (Alpher, O’Grady and Bowern 2008: 28). 12 The corresponding KKW 2Sg ngi is interpreted as having its initial consonant analogically adjusted to conform to first person pronouns, which all begin with ng.

The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered

65

(1919) – but their subgroup status has not yet been established by common innovations. Dixon (2002: xxxiv–xxxv) however, accepts Gathang and Awabakal as forming one genetic group and Darkinyung and Dharruk as constituting another. In my opinion it is likely that these two sets of language do in turn constitute a subgroup, which following Schmidt we could call the Kuri subgroup. These languages have been the subject of recent consolidated accounts, which attempt to provide a plausible phonemic and morphological interpretation for all the lexical and grammatical materials available in old sources. These languages all possess distinctive masculine and feminine paradigms for 3Sg pronouns. The forms for some cases of these pronouns (especially free form pronouns) end in -wa, which recurs in 1Sg, 2Sg, and 3SgM pronouns and may have its origin in a contrast-marking enclitic (possibly originally *-pa). We begin with Darkinyung. The relevant forms are quoted from Jones (2008), who has proposed phonemic representations of originals given in the grammatical sketch by Mathews (1903: 272–273). The members of the paradigm are: ERG/NOM nhanta (nonda)13, ACC nhanawan (nonoan), POSS nhanangkayi (nonanggai).14 The first form, nhanta, obviously continues *nhantu, with a change to the final vowel; another form, spelled nondwa by Mathews, may continue *nhantuwa with the original u vowel maintained in non-final position. Note that the ERG form has extended its use to include that of the expected NOM. This is a relatively common change in Australian pronouns (Blake 1979). The ACC contains the increment -wa with a further final nasal (possibly to explicitly signal ACC), presumably built on an earlier ACC *nhana, which may also have provided the base for the POSS nhanangkayi (all singular possessive pronouns involve the suffix -ngkayi). Another form nhana (nana) is glossed by Mathews as ‘him’; this could be the unaugmented inherited 3Sg Feminine ACC form – either misunderstood as a masculine or extended to the other gender in the manner familiar from some Karnic languages. Closely related to Darkinyung is the language of Sydney, which was documented – as the language of the Eora – by William Dawes beginning in 1790 (Dawes 2009) and as Dharruk by Mathews (1901a). The only form of the 3SgF paradigm given by Dawes is naanóongí ‘his or hers’ (Dawes 2009: 32), which can probably be interpreted as /nhaanu-ngay/. This would represent the same

13 The forms in parentheses are the original spellings; the other form represents Jones’ phonemicization expressed in the standard orthography used in this paper. 14 According to Gavan Breen (p.c., 27 September 2005) it is quite possible that Mathews’ o represents /u/ rather than /a/, and Dixon (2002: 461) interprets the longer nominative form as nun-duwa. If this is correct, the feminine pronoun would have undergone the same vowel change under the influence of the masculine *nhu that took place in Gadhang (see below).

66

Harold Koch

Possessive suffix as Darkinyung -ngkay(i)15, except that it would reflect the operation of the nasal cluster reduction rule of the coastal dialect of the Sydney language (for which see Nash 2011). The base to which this suffix is added appears to be not *nhana- as in Darkinyung, but a form nhaanu-, which could directly reflect the pPN Dative. Mathews (1901a: 156) also cites a possessive form nannunggai ‘his’; this could represent /nhanu-ngkay/ without the cluster reduction or possibly be identical to the Darkinyung /nhanangkay(i)/ – with a low vowel in the second syllable. Mathews gives the Nominative form for the 3Sg as nanu, and the Ergative (his “nominative agent”) as nanudya. This 3Sg stem may possibly reflect the inherited Dative form, being used as a subject form (in a language obsolescence situation?).16 Both Eora and Dharruk thus provide evidence for an inherited form that reflects the pPN Dative *nhanu. Both Dawes’ and Mathews’ glosses suggest that the language has generalized the feminine as a default form to include masculine reference; this would be consistent with Mathews’ Darkinyung nana ‘him’. Note, by the way, that the Sydney language alone provides support for the long vowel of the root that occurs in Bandjalang. In Gathang, as in Darkinyung, the subject form of the 3SgF has for the NOM a short and a long form, nyunta and nyuntuwa, which parallel the 1SgNOM ngatha and ngathuwa. The accusative nyunkung and the possessive nyunkupa or nyunkumpa (Holmer 1966; Lissarrague 2010: 64) both appear to have been built on a form in -ku, which probably represents an earlier dative case-form: the suffix -ku recurs in the stem nyuku- used in the ACC and POSS of the 3SgM, whose NOM is nyuwa. I propose that the feminine forms *nyuntu- and *nyunkuhave changed their stem vowel from a to u by analogy with the 3SgM nominative nyuwa and accusative nyukuwang (or earlier *nyuku) and hence derive from earlier *nyantu and *nyanku respectively.17 The suffix -ku is widespread as the dative formative in Pama-Nyungan languages, especially in nouns, and in many languages has found its way into the paradigm of 3Sg pronouns, where it is used to create new Datives *nyanku and *nyuku in place of the pPN *nyanu and *nyungu respectively. The earlier form *nyantu (or *nyuntu) must have changed functionally from an Ergative to an Ergative/Nominative; and new

15 In my opinion the suffix may be phonemically either -ngkay or -ngkayi. 16 Nanu is unlikely to continue the Ergative form *nhantu with the effect of nasal cluster reduction, since that rule applied only to the coastal dialect, and is not manifested, for example, in Mathews’ 2SgNOM, which is spelled ngindi. 17 Dixon (2002: 462) suggests, as one possible explanation for the u vowel, that it “may be due to assimilation to the following vowel”.

The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered

67

Accusative and Possessive forms would have been built on the earlier Dative *nyanku (or *nyunku). The data for the Hunter River and Lake Macquarie (HRLM) language has been compiled in Lissarrague (2006), from which the forms discussed here are cited. The earliest documentation was by Threlkeld from the 1820s; the language of his data has been called Awabakal. This language shows indirect reflexes of *nhantu and *nhanu (with the root vowel possibly altered to u as in Gathang). These forms would have been extended with the -wa increment (as in pronouns of Gathang and Darkinyung). The functions of these pronouns have been extended from Ergative to Ergative/Nominative and from Dative to Accusative/ Dative respectively, the latter adding a final -n, which marks the Accusative in most personal pronouns. The resulting forms *nhantuwa and *nhanuwan suffered the truncation of their initial consonant and vowel when the forms occurred cliticized to other elements, yielding *-ntuwa and *-nuwan18. The object clitic nuwan (spelled noun) is also found as the second element of “compound pronoun” forms, combined with subject markers: panuwan (bá-nó-un) ‘1SgSubj + 3SgFObj’ and pinuwan (bi-nó-un) ‘2SgSubj + 3SgFObj’. The free form is puwanuwan (bo-un-no-un) ‘3SgF Accusative’, where the enclitic –nuwan is combined with an element puwa- (of unknown origin19). The free-form subject pronoun, puwantuwa (bountoa) ‘3SgF Ergative-Nominative’ consists of the same first element plus -ntuwa, a cliticized form of *nhantuwa. The same second element occurs in the compound pronoun pintuwa (bín-tó-a) ‘2SgObj + 3SgFSub’, where pin is the bound form of the 2SgAcc. It is possible that the compound pronouns should each be seen as a string of enclitics attached to a preceding word, since they never occur initially in a sentence (Lissarrague 2006: 47–48). To sum up, the “Kuri” languages seem to provide solid evidence for an ergative (>ergative/nominative) form *nhantu, plus plausible evidence for the dative form *nhanu and weak evidence for an accusative *nhana – the latter based on a single language, Darkinyung. (It is questionable whether a reflex of the NOM *nhan would have been required to motivate the form of the analogical nyunku- in Gathang.) These forms are collected in Table 7, where they are arranged – the order of languages is from north to south – according to their presumed Proto-Pama-Nyungan and Proto-Kuri origins. Of these forms, the only ones previously compared to pPN *nhan (in Dixon 2002: 461) were the Gathang stem nyun- (with a comment on its changed root vowel) and the Darkinyung and

18 If the object form were rather phonemicized as -nawan (depending on how the orthographic o is interpreted), these forms would descend from the pPN ACC *nhana rather than the DAT *nhanu. 19 It may possibly be related to a 3Sg Masculine enclitic object form -pun.

68

Harold Koch

ERG

NOM

ACC

DAT

*nhan

*nhana *nhana(wa)

*nhanu *nhanu(wa)

HRLM

*nhantu *nhantu(wa) nyunta, nyuntuwa bountoa, bín-tó-a

Darkinyung

nonda, nondwa

pPN Proto-Kuri Gathang

nyunku-

nana, nonanggai, (nonoan)

Eora/Dharruk

-noun, bo-un-no-un, bá-nó-un bi-nó-un (nonoan) nanu, naanóongí, nannunggai

Table 7: Possible reflexes of *nhan in the Kuri languages

HRLM subject forms cited as nun-duwa and buwanduwa respectively (without any further analysis).

3.5 The evidence from Warluwarric As mentioned in 2.1, Blake (1988: 22) had found reflexes of *nhan in Wagaya and Yanyuwa of the Warluwarric subgroup, where as a result of lenition of the initial consonant it appears as yan-; Dixon (2002: 306) accepts this interpretation.20 It is not clear whether this lenition is a general sound change. It appears that all 2nd and 3rd person pronouns which originally began with a laminal nasal (nh or ny) became y in Warluwarric (see the table of pronouns in Blake 1988: 65) – which might suggest that some instances of the initial y are the result of irregular, perhaps analogical, changes affecting pronouns.21 However, a general sound change initial nh to y would appear to be supported by Warluwarric *yanga‘see’ (Carew 1993: 70, Breen 2004: 657), which is probably an extended form of the Proto-Pama-Nyungan verb root *nhaa-. On the other hand Warluwarric *nhangali ‘what’ (Carew 1993: 77) fails to show this change, if it is a reflex of the widespread Pama-Nyungan interrogative root *nhaa- (Dixon 2002: 333). A Nominative form *yan-pa, reinforced with -pa, as in Pitta-Pitta, is behind the Wagaya and Injilanji forms yamb (Breen 1976: 591) and yambi (Brammall

20 Forms beyond those mentioned by Blake are taken from Breen (1976), Kirton and Charlie (1996), and Brammall (1991). 21 In a number of languages, e.g. Marrgany-Gunya (Breen 1981: 333–4) all 2nd person pronouns – the reflexes of 2Sg *nyun-, 2Du *nhumpala, 2Pl *nhurra – have lenited their initial consonant to y.

The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered

69

1991: 22) respectively, which further involve place assimilation of the nasal and final vowel loss, and the subsequent remarking with a new overt feminine suffix -i in Injilanji. The West Wagaya ACC yani appears to be built on a stem yan-. The ERG yand presumably is a reflex of *yantu (cf. the 2SgERG inda beside NOM imb). Apparently cognate with the Wagaya yand is the NOM form anda of Yanyuwa, assuming the ERG has expanded its function to include the NOM, as is common for personal pronouns. A Warluwarric Dative *yanku- seems to be reflected in the oblique stem yanggu- of W Wagaya and yangg- of Injilanji and in the Dative anku of Yanyuwa. This would represent the same recreation of the inherited Dative by means of the normal suffix -ku that occurred in other PamaNyungan languages and which is paralleled in Warluwarric by the 3SgMasculine yiku of Yanyuwa and yugu- of Wagaya. A summary of the Warluwarric cognates is given in Table 8.

pWwc WWagaya Injilanji Yanyuwa

ERG

NOM

DAT

*yantu yand

*yan yamb yani ACC yambi

*yanku yanggu- OBL yangg- OBL anku DAT

anda NOM

Table 8: Warluwarric reflexes of *nhan

The Warluwarric subgroup shows evidence of semantic changes. Alpher (1987: 174) pointed out that Western Wagaya classifies nouns as masculine vs. feminine/ neuter. The 3Sg pronouns (paradigms of yu- and yamb) are described by Breen (1976: 591) in these same terms. So this dialect has expanded the functions of the inherited 3SgF pronoun in the same manner as Wester Torres Strait and Wangkumara. The eastern dialect of Wagaya has lost the gender contrast and the forms of the yan- paradigm. Warluwarra and Bulanu lack a gender contrast and use the inherited masculine forms as general 3Sg pronouns.22 There is also in this subgroup evidence that the Nominative was reinforced with an increment and that new case-forms were created: a Dative in -ku perhaps already in ProtoWarluwarric and a new ACC in –i in Wagaya.

22 Brammall (1991: 23) rather offers a historical solution in which Yanyuwa and the WakayaInjilanji branch independently acquired their feminine forms from neighbours – but he doesn’t provide any etymological support for this hypothesis.

70

Harold Koch

3.6 Evidence from the Yolngu subgroup As already mentioned by Blake and Bowern, the Yolngu languages provide further plausible cognates, albeit without the feminine meaning. The Yolngu subgroup is a Pama-Nyungan enclave surrounded by non-Pama-Nyungan (nPN) languages in Arnhem Land of the central northern part of Australia. These languages lack a masculine vs. feminine contrast in 3Sg pronouns. The 3Sg pronoun reflects a root *nhan, which Blake (1988: 26) relates to the Pama-Nyungan feminine pronoun.23 In the western languages, Djinang, Djinba, and Yan-nhangu, a reflex of this stem is found in the whole paradigm; in the eastern languages such as Djapu, Djambarrpuyngu, and Ritharrngu, however, the paradigm is suppletive, with nhan- as the oblique stem and an unrelated form ngayi as the Nominative (and the Accusative based on the Nominative). Blake (1988: 26) is surely right in suggesting that ngayi is an innovation borrowed from non-Paman-Nyungan languages. Some oblique case forms are built on a stem nhan-, e.g. Yan-nhangu DAT nhan-ku, Ritharrngu GEN/DAT nhan-ngu, Djapu DAT2 nhan-bal. These presuppose a former NOM *nhan. A different oblique stem nhanu- is found in Ritharrngu and the Djapu Oblique case form nhanu-kal. The nominative form in the western languages seems to reflect *nhani; cf. Yan-nhangu nhani, Djinang and Djinba (in which nh has merged with ny) nyani. This *nhani appears to be a reinforced form of the pPN Nominative *nhan. (The final vowel i of Djinang and Djinba could continue an earlier vowel a or u, but this appears not to be the case for Yan-nhangu.) The Yolngu languages thus reflect two Pama-Nyungan case-forms of the 3SgF paradigm: NOM *nhan and DAT *nhanu.24 They provide another instance in which the feminine has changed its function to include all 3Sg references. The pPN masculine 3Sg form *nhu is absent from the Yolngu languages, unless a trace of it is to be seen in the 1Du Exclusive ngalinyu (vs. ngali Inclusive), if it consists etymologically of ‘we two’ + ‘he’. The creation of new case forms on the inherited Dative *nhanu and on the synchronic Nominative *nhan are expected kinds of paradigm renewal.

23 Sources from which data are cited are: Waters (1989) for Djinang and Djinba, Baymarrawaŋa et al. (2006) for Yan-nhangu, Morphy (1983) for Djapu, and Heath (1980) for Ritharrngu. 24 Since there was a sound change in Yolngu that reduced nasal + stop clusters to just nasals (Sutton and Koch 2008: 489), the oblique stem nhanu- could conceivably be a reflex of the pPN Ergative *nhantu, provided that this form had first expanded to include the NOM function. But the (eastern) languages which manifest this oblique stem rather have ngayi as their NOM. It is conceivable that the ambiguity of *nhanu – between ERG and DAT – caused by the nasal cluster reduction was a factor in the borrowing of the new form ngayi.

The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered

Djinang Djinba Yan-nhangu Djapu Ritharrngu

71

NOM

Oblique stem

Oblique stem

nyani nyani nhani

nhannhannhan-ku nhan-bal nhan-ngu

nhanu-kal nhanu-

Table 9: Yolngu 3Sg pronominal forms in *nhan-

3.7 Evidence from the Ngumpin-Yapa subgroup One further subgroup, this time from the west, provides more paradigmatic evidence for this pronoun, provided we assume a similar loss of gender-specific meaning. The Ngumpin-Yapa subgroup is spread widely across north-western Australia and consists of two branches, the Ngumpin languages and the Yapa languages (for justification of this subgroup see McConvell and Laughren 2004)25; in these languages there is no laminal contrast, so *nhan descends as nyan. A bare form nyan does not occur, but it is presupposed by the Jaru 3rd dual nyanpula,26 where the dual-marker pula is appended to what must be a 3rd person stem nyan-. This root appears in an extended form nyani as the 3rd person pronoun stem in Mudburra and, with further extensions, in Warlmanpa. A reflex of the earlier ergative form, nyantu, occurs in Gurindji, Jaru (as the nominative form of the 3rd singular and plural form), and Walmajarri (where it is a stem for all numbers of the 3rd person – 3Du is nyantu-jarra and 3Pl nyantu-warnti). Shifts from ergative to nominative function are common in Australian pronouns (Blake 1979) – and were presupposed in the Kuri language above. A (clitic) case form nyanta (matching Bowern’s Proto-Karnic *nhanta) also occurs in the Ngumpin languages: it is labelled locative in Jaru and accessory in Walmajarri. This is the expected form of the locative of an n-stem nominal in Pama-Nyungan. The pPN Dative *nhanu is well reflected in both the Ngumpin and Yapa branches. The oblique stem of the pronoun is nyanu- in Jaru; it is common for oblique cases of pronouns to be built on the dative. In most of the NgumpinYapa languages nyanu occurs as an enclitic reflexive object marker for all persons and numbers (in some lects as nyunu, with a harmonized vowel). The reflexive forms apparently derive from a 3Sg object form. Their origin as 3rd person forms 25 The forms are taken from Hudson (1976), Tsunoda (1981), Nash (1979) and Nash (1996). 26 This form should not be confused with the widespread nyumpalu (or similar), which is the reflex of the pPN 2DU pronoun (cf. Table 1).

72

Harold Koch

is still to some extent reflected in Warlpiri. “Reflexive/reciprocal -nyanu never occurs with 1 sg subject, and usually not with the singular imperative; the singular object enclitic occurs instead” (Nash 1996: 124). The apparent development have been: nyanu was originally a dative pronoun; its function was extended to accusative case (a common change in Australian pronouns); it was used as a reflexive marker first for 3rd person objects (non-reflexive 3rd person objects in these languages have zero clitics); then it was generalized to indirect objects and reflexive and reciprocal objects of non-3rd persons.27 These Ngumpin-Yapa forms are displayed in Table 10, where they are aligned according to their presumed etymological case form. The languages in the top half of the table are Ngumpin, while the bottom three belong to the Yapa branch. The changes presupposed by these cognates are by now familiar: loss of distinctive feminine meaning of the root, generalization of the Ergative form to the Nominative, extension of the bare form with -i; creation of new case forms on the basis of the Dative; use of the Dative form for direct objects; reduction of a free form pronoun to enclitic status. The semantic development from a 3rd person object to a reflexive object is a change not encountered before with this root.

pPN pNgY Walmajarri Jaru Gurindji Mudburra Warlmanpa Warlpiri Ngartily

*ERG

*NOM

*DAT

*LOC

*nhantu *nyantu nyantu nyantu nyantu

*nhan *nyan(i)

*nhanu *nyanu =nyanu nyanu- =nyanu, nyunu =nyunu =nyanu, nyunu =nyanu, nyanungu =nyanu, nyanungu =nyanu, nyanungu

*nhanta *nyanta =nyanta =nyanta

nyan-pula nyani nyani-

Table 10: Ngumpin-Yapa reflexes of pPN *nhan

3.8 Evidence from Pilbara languages A suffix -nhanu, cognate with Ngumpin-Yapa nyanu, is found in at least two other languages of the west, Martuthunira and Tharrkari.28 Martuthunira belongs 27 The same form, nyanu, may have served as the base for a free-form pronoun nyanungu ‘the aforementioned’ in the Yapa languages. These might, however, rather reflect a deictic root of similar form. 28 I am grateful to Alan Dench (p.c., 25 October 2005) for calling my attention to these forms; he in turn was alerted to their relevance by Mary Laughren.

The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered

73

to either the Ngayarta or the Mantharta subgroup and Tharrkari to the Mantharta subgroup; since these two groups, plus, the Kanyara subgroup, are closely related, it is sufficient for our purposes to refer to a possible “Pilbara” subgroup of Western Australian languages.29 The suffix is found as a third person possessor in kin nouns, e.g. kurntal-nhanu ‘daughter-3POSS’, mayili-nhanu-ngu ‘father’s father-3POSS-ACC’; the 1POSS suffix after consonants is -yu, e.g. kurntal-yu ‘daughter-1POSS’, where -yu is presumably derived from the 1Sg clitic =tyu of other languages of the region (Dench 1995: 106–107). I interpret this =tyu as a reduced form of the 1Sg Dative pronoun *ngatyu. The parallelism with 1POSS strongly suggests that -nhanu also continues the dative form of a 3Sg pronoun, which is undoubtedly the erstwhile 3Sg feminine pronoun. The same suffix appears in Tharrkari: Klokeid (1969: 21) glosses -nhanu as ‘one’s own’ and says it “is apparently suffixed only to kinship terms, and seems to indicate one’s biological kin, as opposed to classificatory relatives”. Note that if these instances of -nhanu are indeed reflexes of the pPN 3SgFem DAT, the changes undergone include: generalization of the Feminine to general 3rd person reference, change in grammatical status from free word via enclitic to suffix, and the separation of an individual form from the remainder of its paradigm. The reduction of Dative free pronouns to possessive affixes on kin nouns has a parallel in the languages of the Arandic subgroup of Central Australia (Koch 2004b), and possibly in some of the Yuin languages as well (cf. 3.9.3). The isolation of an erstwhile case-form from its paradigm is illustrated in Koch (2009), where the process is described as de-paradigmatization.

3.9 Further possible cognates Once a reconstruction has been made with some certainty, we expect to find traces of the original in further languages, where its presence may not at first have been obvious.30 A few more potential cognates are described in the following subsections. 3.9.1 Evidence from Kungkari Kungkari is a language of inland Queensland, located between the large Karnic and Maric subgroups. Kungkari is one of a number of languages of dubious 29 For issues in the classification of Pilbara languages see Dench (1998, 2001), Koch (2004a: 34–37). 30 Cf. Koch and Hercus (this volume), where the same principle is explored with reference to lexical cognates in Australian languages.

74

Harold Koch

classification which may possibly be a member of the Karnic subgroup (Bowern 2001: 247), in which case its evidence merely supplements that of the Karnic subgroup. Kungkari has a (rare) 3SgNom form nhana-, which, like third person pronouns in Karnic, is always extended by a deictic marker (Breen 1990: 32). This could be a reflex of the accusative form of an erstwhile feminine stem *nhan, hence cognate with the Pitta-Pitta (Karnic) form nhana. The case for deriving it from a former accusative form is strengthened by the parallel of the 1Sg and 2Sg nominative/accusative forms, nganha and ngina respectively, which are likewise reflexes of earlier accusative forms – the accusative forms having expanded their function to include that of the nominative (for this change in Australian languages see Blake 1979). If this form is authentic, it may represent a functional loss of the masculine vs. feminine contrast which is not fully matched by a loss of one set of forms. This would also be the mechanism behind the mixed paradigm of Wangka-Yutjurru mentioned in 2.1.

3.9.2 Evidence from Maric A most interesting form is nhani, which marks a non-gendered 3SgNom in the Wirri dialect of Biri, a language of the eastern branch of the Maric subgroup of inland Queensland (Terrill 1998: 72). The final vowel i looks like the increment found in some Karnic languages (Diyari, Yandruwandha, Wangkumara but not Pitta-Pitta – for the forms see Dixon 2002: 305). The same increment i occurs in the stem of ngani ‘what’ ( heaven in the sense of the ‘upper regions of the air’. This sort of etymology presumes that there was a continuous chain of speakers from a continuously-existing speech community using form-meaning pairs that gradually changed from A1 to An via A2 A3 A4, etc.3

2 All etymologies in this paragraph are taken from the Oxford English Dictionary, online edition. November 2010. Oxford University Press. Sighted 16 January 2011. 3 While I refer to these relatively simple etymologies as ‘prototypical’, I do not mean to imply that lexical etymologies are always this straightforward. Robert Mailhammer (p.c.) points out that even lexical etymologies can involve multiple sources, as in the case of blends, calques or taboo replacements. However, it seems to me that in the few cases where constructional etymology is discussed in the literature in a theoretical sense, the assumption is that rather straightforward ‘source-outcome’ relationships exist: a model reflected more in the most basic sorts of (lexical) etymology.

Constructional etymology

87

This prototypical sort of etymology can also be involved in constructional change.4 A construction C1 is a form-function pair where the form consists of a word order pattern, fixed morphological or lexical marking, and a syntactic relationship to other phrases or clauses in the sentence. Changes in any of these and/or extension of the construction to a new function can create a new construction C2, which we could say is etymologically related to C1. We would want to describe C1 and Cn this way if the changes in form and function were gradual such that a continuous chain of etymologically related constructions C1, C2 C3 . . . Cn had existed. This sort of change is schematized in Figure 1.

Figure 1: Schematization of a source-outcome relationship

However, as I will show in the later sections of this paper, with relative clauses at least, this source–outcome model seems almost never to be the case. Instead, we find different sources for the changes in form and the changes in function, or abrupt borrowing or calquing, so that C1 and C2 are not related in this sort of gradual etymology. These sorts of changes are schematized in figures 2 and 3. In the case of analogical change, the “source” continues to exist alongside the “outcome”. In the case of borrowing or calquing, the source continues to exist in the donor language or dialect, or if it does not, its disappearance is not directly connected to the fact it was the source of the new construction. Moreover, F2 is usually wildly different from F1 because it is frequently a calque, or in any case heavily adapted to the recipient language. The morphological forms, lexical forms, or word order may be different (although not usually all three of these – otherwise we would not identify the construction as based on that of the

4 See also Mailhammer (2007: chapter 3) for proposing etymology as a research perspective that can be applied to lexical as well as to structural items (Lexical vs. Structural Etymology, see Vennemann 2000), which can easily be extended to constructions.

88

Rachel Hendery

Figure 2: Schematization of analogical change

Figure 3: Schematization of borrowing or calquing

Figure 4: Schematization of reanalysis

donor language at all). M1a on the other hand is usually identical or extremely similar to M1 (which is why I have labelled it M1a instead of M2). Finally, reanalysis can lead to a new construction or term, but again this is different from the sort of scenario represented in figure 1, since construction 1 often continues to exist once construction 2 arises. With this outline of the various possibilities in mind, I will now move on to a discussion of the actual sources of a particular construction, the relative clause.

Constructional etymology

89

3 Relationship between sources of constructions and sources of lexemes that mark that construction In order to understand the etymologies or sources of relative clause constructions, we can begin with the simpler question of the sources of relative clause markers. While I will argue that the relationship between these two sorts of etymologies is not a straightforward equation, one nonetheless provides a clue to the other. Relative clause markers are defined for the purpose of this paper as grammatical items that serve the purpose of identifying a clause as having a relative clause function.5 These markers have a variety of etymological sources, the most common of which are pronouns (interrogatives and demonstratives) or elements commonly found in non-relative NPs: classifiers, possessive markers, “generic” nouns, even definite articles. In this section I will discuss how these various types of markers came to be used in relative clause constructions. It will be seen that the etymological relationship is generally not one of “grammaticalization” but nor is it as straight-forward as the sort of development discussed at the start of section two. The simplest and by far most common approach to the etymology of relative clause markers is to look for obviously cognate forms in other constructions: demonstratives, interrogatives, and other clausal constructions such as NP and VP complementation, adverbial clauses, etc. This is the approach taken, for example, by Heine and Kuteva (2002). Under the assumptions of “grammaticalization”, such markers are thought to have gradually become part of the relative clause construction through such mechanisms as phonological reduction, fusion, and semantic bleaching. In fact, however, these are strange mechanisms to assume for the development of relative clause markers, which, unlike the prototypical cases of grammaticalization, are most frequently already grammatical items when they are co-opted for use in relative clauses. A survey of known sources of relative clause-marking elements is still worthwhile, but as an end in itself it is rather uninteresting, because it does not tell us how the etymological change proceeds. For this we have to turn to case studies where a change has happened during a period of a language for which we have written attestation or otherwise strong evidence for the directionality of the change and the steps it took on the way.

5 In other words, I am subsuming relative pronouns, relative complementizers, and relative verb affixes under one category here.

90

Rachel Hendery

In Table 1 I list the sources of relative clause markers that are best documented as historical developments. These are not simply based on apparent cognates, but are cases where the directionality of the change is well-evidenced, usually because the development occurred during a period of the language with written history. More details on these developments can be found in Hendery (2012). Other possible sources that are less well documented include comparatives, possessive pronouns, and indefinite pronouns. Source of RC marker

Evidenced in

interrogatives demonstratives discourse markers (e.g. topic or focus markers) generic nouns classifiers personal pronouns

many languages many languages Quechua, Basque, Tok Pisin, PIE, and possibly many others Malay, Lhasa Tibetan, Newari, Japanese, Hebrew, Korean Thai, various Amazonian languages, Newari, Cantonese Gothic

Table 1: Sources of relative clause markers

From a comparison of these case studies, it can be seen that development of new relative clause markers generally proceeds in one of two different ways, the first of which is through reanalysis: fossilization of an element that happens to appear in relative clauses or on the clause boundary, but is not originally part of the relative clause construction per se. This is the case for some generic nouns, classifiers, and various discourse markers, and can be illustrated with the case of Tibetan, as outlined in DeLancey (1986). Among various other relative clause markers, Lhasa Tibetan has one, sa, that literally means ‘earth’ or ‘place’, while Newari has mha, meaning ‘body’. DeLancey proposes that these may have been ‘kidnapped’ into the relative clause from an appositive construction, something like example (1). (1) [ [ jĩ-ĩ nyan-a ]-mha ] [ NP [ CP I-ERG buy-PART]-live.thing]

[nya ] [NP fish]

‘the fish, the live thing I bought.’ If the clause boundaries on example 1 were reanalysed, the resulting construction is the modern-day Newari relative clause illustrated in example (2). (2)

[ [ jĩ-ĩ nyan-a -mha ] [ NP [ CP I-ERG buy-PART-live.thing] ‘the fish, the live thing I bought.’

nya] fish]

Constructional etymology

91

A second way for relative clause markers to arise is through analogy with other constructions that resemble the relative clause, but even the clearest examples of this pathway, examined in Hendery (2012), are ambiguous in terms of directionality. Possessive constructions, for example, resemble relative clauses in many languages, and in Japanese (no), Malay (punya) and the Semitic languages (various markers), there is enough historical evidence to trace the development. The problems with these examples are that modern Japanese no does not function as a relative clause marker, except in child language; Malay punya is not yet unambiguously a relative clause marker, although it seems to be moving in that direction, and the Semitic particles are all related to demonstratives as well, so that both the relative and possessive function might have independently developed from this third source. Certainly as a general rule, markers can be extended to new constructions by analogy, and Hendery (2012) shows how this is frequently the case with the extension of relative clause markers to new constructions (adverbial clauses, general subordination, adjective phrases, among others). That the reverse is not strongly evidenced is probably merely an accident of the historical record. A few sources, such as interrogatives, appear to have primarily spread through language contact, which can also be viewed as a sort of analogical process, as the first speakers to bring them into the new language are presumably doing so on analogy with the relative clause constructions in their other language. Interrogatives in colloquial Basque relative clauses, for example, are thought to have developed on analogy with the French and Spanish constructions (Hurch 1989). A language contact cause has also been suggested for English interrogative relative clauses (with French as the source), and is also thought to have been the case in Nahuatl (Hill and Hill 1981), Quechua (Lefebvre and Muysken 1982: 45; Karttunen 2000: 403), and the Uralic (Comrie 1988: 474–475) and Turkic languages (Johanson 1998: 333, 335) that have interrogative-based relative clause markers. It has even been suggested that this is nearly always the case when the interrogative-based relative pronoun construction is found in nonIndo-European languages (Comrie 1998: 61). It is therefore the case that even the etymologies of relative clause markers can be quite complex. They can have more than one source, as in the Basque example, where the Basque interrogative zein, the French relative pronoun qui, etc, and the Spanish relative pronouns que, etc are all “sources” (inputs) in some sense for the new Basque relative pronoun zein. The sorts of elements that become relative clause markers do so partly because they are well suited to the role, in the sense that they already perform one or more of the characteristic operations involved in relativization: subordination, attribution or pronominalization (cf. Lehmann 1984: chapter 4). For example in

92

Rachel Hendery

English, even if the demonstrative that were not a relative clause marker, it would still perform all of these functions: it can stand in a modifying relationship to a noun in the NP (that man), it can introduce a subordinate clause (I said that he left) and it can stand alone as a referential pronoun (I saw that). This sets it up perfectly to function as a relative clause marker, since it already plays all the necessary roles. In this sense, the structural features of a source construction shape the path of change taken by the marker, which is the sort of change that most closely resembles the prototypical single-source etymological path. It is possible that some pathways of development toward relative clause markers are not unidirectional. Possessive markers, comparatives, focus markers, and indefinite pronouns are all elements which might be both source and outcome of relative clause marker change. In Hendery (2012), however, I show that none of these is well documented for both directions. For possessive markers becoming relative clause markers, there is strong evidence of direction, but for the reverse direction, the evidence is weaker and there are plausible alternative explanations. The same is true of indefinite pronouns. Focus markers are wellattested as sources of relative clauses, but not so strongly as outcomes. For comparatives, neither direction is well-documented or without alternative possible explanations. If there are markers for which both directions of change occur, however, this is unsurprising since the developments discussed here are not “grammaticalization” in the strict sense of the term, but rather are the extension of a grammatical marker from one grammatical function to another. In this section we have seen that the etymologies of relative clause markers can resemble the more prototypical lexical etymologies, in that they can have a single source, consist of semantic and formal changes, and involve loss of the element in its original function so that it really does appear that a change has occurred. On the other hand, when relative clause markers arise through analogy, whether it is analogy with another construction in the same language, or with relative clauses in other languages of the region, their etymology is more complicated, with multiple “sources”. As we turn to the etymology of whole constructions in the next section, we will find this latter type of relationship between source and outcome constructions, and even more complex variations where different components of the construction have different source relationships.

4 Relative clauses: constructional change and etymology As the examination of relative clause markers above suggested might be the case, the diachronic interaction between relative clauses and other constructions

Constructional etymology

93

is often a matter of analogical change rather than a source–outcome relationship. In cases of constructional borrowing or calquing, the mechanisms involved for whole constructions are also similar to those that lead to transfer of a relative clause marker from one language to another.

4.1 “Sources” of relative clauses Just as is the case for relative clause markers, a cursory glance at relative clause constructions in historical perspective may lead to the impression that a simple source–outcome relationship exists. In the Ethiopian Semitic languages, for example, the modern prenominal RC seems to have developed directly out of earlier postnominal constructions. In these Ethiopian languages a drastic restructuring of the word order and phrase order rules took place, presumably due to influence of the surrounding Cushitic languages. The relative clause construction was literally flipped around the head noun as part of this reordering. It can therefore be said to have developed directly out of the older construction, which did not persist. However, one could also argue that the various Cushitic relative clause constructions were sources of the new Semitic one as well, but note that this is not a one-to-one relationship, as it is not one specific construction in a single Cushitic language that is said to be the source. In other cases, a source–outcome relationship between two different constructions might exist, but it is only one of many relationships that the construction participates in. For example, in 12th century Middle English, a relative clause construction developed that was introduced with an interrogative pronoun. This could be said to be a new relative clause that had its etymological source in embedded questions. Certainly it resembles these closely. However, the construction also has diachronic relationships6 with other constructions: pre-existing relative clause constructions in English, and relative clause constructions in French. These relationships are evidenced by the fact that it resembled the pre-existing relative clauses in terms of position, word order, and other features, and even participated in blended constructions, such as whom that relative clauses, illustrated in example (3).

6 And it presumably also had synchronic relationships with these other constructions. Speakers would have seen the set of constructions involving interrogatives as ‘similar’ to each other in some way, and saw the set of constructions that served the same function (relativization) as ‘similar’ to each other in some other ways. It is this connection that led to the transfer of features among the various constructions in the first place.

94 (3)

Rachel Hendery

Only the sighte of hire, whom that I serve, . . . wolde han suffised right ynough for me ‘Only the sight of her, whom (that) I serve, . . . would have sufficed right enough for me’ (Knight’s Tale 1231: line 373–5)

The adoption of interrogative-marked relative clauses may also have been triggered or accelerated by the fact that this “constructional polysemy” already existed in the Norman French spoken by many residents of England at that time, as illustrated in example (4). It is a development which may have occurred anyway, however, as wh-pronouns were already occurring in the heads of “free” relative clauses, as in example (5), which meant a simple reanalysis could have turned them into relative pronouns even without an exact model. (4) Irieement parla li lous, ki mult esteit cuntrarious Angrily spoke the wolf, who very was quarrelsome (Lais of Marie de France 2: 5) (5)

Soðes is þe sylle swa hwæt swa [ þu me byddest] Truly I you give so what-ACC as [you me ask] ‘Truly I give you whatever you ask of me’ (Gospel of St Mark in the West-Saxon Gospels 290, from Allen 1980: 278, ex. 51).

Part of the problem with thinking of constructions in terms of the source– outcome etymological model is that there are usually multiple types and subtypes of a construction in a single language. Therefore, the construction already participates in a kind of network of other constructions, which are treated by speakers as similar to varying degrees. This means that changes can enter a construction from multiple sources in this network. (For example in the Middle English example, a new construction was created with the function of one, the markers of another, word order and other restrictions from several more, and on analogy with yet another construction that existed in the grammars of some bilingual speakers.) Note that the connections in Figure 5 are not meant to represent causal relationships, or one-way source–outcome connections. Rather they represent relationships of any sort, including source–outcome relationships, but also analogies, constructions that were both simultaneously present in a speaker’s grammar and resemble each other, constructions that participated in blends, or competed with each other (e.g. one was extended into the other’s domain). This is the field of constructions that may have participated in the etymology of the Middle English interrogative-marked relative clause. The diagram shows how

Constructional etymology

95

Figure 5: Middle English interrogative-based relative clauses

much of the complexity of the context is lost when we try to reduce it to a simple one-to-one source–outcome relationship. Because multiple types of construction with the same or similar function exist, sometimes nothing is lost when a new construction is created. The new construction can co-exist side-by-side with other pre-existing constructions that fulfil the same function, and with its source or sources. This was the case with the English interrogative-based relative clauses: the construction with that persisted alongside the new alternatives. Although the distribution of each type is not identical, there are some contexts in which either can occur. This means that it does not really make sense to talk about it as a “change” in any one relative clause construction: the interrogative relative clause is not a change to the demonstrative-marked construction. Rather these sorts of changes are changes to a system, through a sort of (non-identical) duplication of the form of a construction, sometimes with extension to a new function. This is unlike the simple kind of etymology illustrated by friend, as detailed in section two, since in that case the related verb was lost, and the morphology as a present participle became opaque. The example of ME wind-oge ‘windeye’ > window, is also different, because in that case the ‘wind-eye’ term with its transparent meaning and exact morphology and phonological form is (gradually) lost and replaced by the new single-morpheme form. It is also unlike the sort of etymology illustrated by ONorse sky ‘cloud’ > MnE sky, where the old meaning is lost and replaced by the new meaning to which it was extended.

96

Rachel Hendery

Rather it is more similar to the sorts of etymologies exhibited in some cases of zero derivation: English to mouse ‘to use a (computer) mouse’ was recently extended to this new meaning/function, while the language retains both the noun mouse, meaning a type of computer peripheral, and the original word denoting the animal. Grammatically and morphologically the new mouse behaves like other verbs (mousing, moused). Interestingly, the noun mouse in the sense of a computer peripheral seems to be distinct in speakers’ minds from the original mouse, as, for at least some speakers, the two have different plurals: mice versus (computer) mouses. The difference between this particular example, and our examples of relative clause “etymologies” is that a verb with the same meaning as to mouse did not previously exist in the language, while most languages already have RC constructions before a new one develops, and the old construction often continues to exist alongside the new one.7 In Hendery (2012) I categorized a large number of relative clause changes into types, which are summarized in Table 2.8 None of these types proved to involve simple, direct changes of one construction into another. In no case can we say that a given construction has an etymological source. In order to illustrate this etymological complexity, in the rest of this section I will discuss examples of each of these types of change.

4.1.1 Extension (a) When a relative clause marker is extended to new types of construction, this essentially creates a new relative clause. The new construction has the syntactic form of one older construction, and the morphological marking of another. This makes it difficult to say exactly which older construction is the etymological source of the new one. Such cases are common, but the history of Russian provides a particularly clear illustration. Russian has several different relative clause markers which have changed their distribution. The element izhe, originally inflecting for gender, number 7 Sometimes when two or more relative clauses continue to co-exist in a language, they each become used in different contexts, i.e. one will become most commonly used for animate nouns, another for inanimates, and/or one will be used for definite nouns, and one for indefinites, and/or one for subject relative clauses and another for objects. These sorts of developments of complementary or near-complementary distribution have occurred in the Romance languages, English, and Russian, among others. See Hendery (2012: 133–134 onwards) for further details. 8 The languages listed in the right-hand column are not an exhaustive list, but they are those for which there is the most evidence of change and information about what exactly occurred.

Constructional etymology

Change

Languages affected

a.

Extension (and/or restriction) of the ‘types’ of RC a marker can be used in

b.

Grammatical change in category of the RC marker (e.g. affixation, loss or development of inflection) Development of a RC with a ‘balanced’ verb construction, where previously only deranked constructions existed

Many languages, including English, Quechua, Turkish, the Celtic languages, Russian, the Romance languages, etc. English, Celtic, Bantu languages, Akkadian, Russian, Egyptian.

c.

d.

Development of a RC with a deranked verb construction, where previously only balanced constructions existed

e.

Development of a prenominal RC where only postnominals, correlatives or internally headed RCs had previously existed Development of a correlative RC where one had previously not existed in the language Development of a postnominal RC where these had previously not existed in the language

f. g.

97

Japanese, Hebrew, Turkish, Hungarian, Finnish, various Mongolian languages, possibly Egyptian, possibly proto-IndoEuropean. Armenian, possibly Egyptian, possibly proto-Indo-European (we do not know the directionality of the change in the latter two) Armenian, perhaps various other branches of Indo-European (pre-historic) Some L2-based New Englishes, possibly the Dravidian languages. Quechua, Basque, perhaps Turkish, Telugu, and most branches of Indo-European (prehistoric).

Table 2: Structural changes in relative clauses

and case, but later used invariantly, was restricted to postnominal, embedded relative clauses, but was later extended to correlatives as well (Leckey 1992: 16– 17, 21). The element kotoryi, which inflects for number, gender and case, underwent the opposite expansion. It was originally used only in correlatives, but then extended to postnominal embedded relative clauses (Leckey 1992: 253, 259–260).9 Leckey suggests that this was an analogical development, meaning that in some sense the two different relative clause constructions were both sources of each of the two “new” constructions: The increasing frequency of the preposed RC with izhe (from the time period attested by Ickler to that attested here) was an analogical development. As kotoryi began functioning in environments typical for izhe, izhe became eligible for environments typical of kotoryi. (Leckey 1992: 259–260)

9 The two elements were not in competition for very long, as izhe was then lost altogether.

98

Rachel Hendery

The diagram in figure 6 schematizes this development.

Figure 6: Medieval Russian relative clauses (based on Leckey 1992)

4.1.2 Change in category of relative marker (b) The change of grammatical category of a relative clause marker also creates a new construction. Rather than a clause introduced by a relative pronoun, the new construction might be marked with an invariant subordinator, or even with a special verb form, in cases where the relative clause marker cliticized and then fused to the verb (as is thought to have occurred in Nguni, for example (Zeller 2004).) In English the relative marker that, while not originally a relative pronoun, is becoming more pronominal in current English usage (Seppänen and Kjellmer 1995; Seppänen 1997). A genitive form that’s has come into usage in some varieties. In English it is clearly an analogy between who – whose leading to that – that’s. It is hard to imagine that such a development would occur in a language that did not already have a pre-existing inflecting relative pronoun. We can say that the construction in (6a) has at least three “sources”, which are given in (6b–d). The first source is the construction that is actually being replaced in the dialect of the speaker of (6a). The second is the standard that construction, which is contributing the marker to the new construction. And the third is the wh-relative-clause, where there is minimally a distinction between possessive and non-possessive forms, and which is therefore an analogical source for the new construction. (6)

a.

The table that’s leg is broken

b.

The table, the leg of which is broken

c.

The table that I painted

d.

The boy whose leg is broken

Constructional etymology

99

4.1.3 Development of a new “balanced” relative clause (c) The development of a new “balanced” relative clause construction can be illustrated by Turkish. Turkish has a standard deranked relative clause construction, but also a minor secondary strategy involving the marker ki and a balanced postnominal relative clause. In some dialects this has become the main relative clause construction (Johanson 1998: 333, 335). There are at least two sources of this new construction. The first is the fact that some verbs in the deranked paradigm happen to be identical to the corresponding main clause forms (Antinucci, Duranti, and Gebert 1979: 162–163). This means that some standard deranked relative clauses looked like balanced constructions, and the new, truly balanced construction may have had its roots in an extension of these. The second source, however, is thought to have been heavy contact with Iranian and/or Slavic languages, as the dialects in which the balanced construction became standard are those in closest contact with these models. The relative clause marker ki itself is from Persian ke (Johanson 1998: 335).

4.1.4 Development of a deranked prenominal relative clause (d and e) Armenian has taken the opposite route, and developed a deranked, prenominal relative clause, where it previously had only had postnominal and correlative constructions. In Classical Armenian, relative clauses were expressed with postnominal or correlative constructions (Hewitt 1978: 100–113). In Modern Armenian it is also possible to use prenominal relative constructions with deranked verb forms: participles, otherwise used as gerunds (Hewitt 1978: 128–129, fn. 20). These three types of relative clause are illustrated in example (7). (7)

a. əntreac̣ erkotasans, z-or-s ew aṙakeals anouaneac̣ he-chose twelve whom indeed apostles he-named ‘he chose twelve, whom he named apostles’ (Classical Armenian, Luke 6: 13, in Hewitt 1978: 102). b. or oḳ I noc̣anē vał aṙis hasṣē, zna who ever from them quickly to me reaches-SUBJ, him ṭagawor araric̣ king I-shall-make-SUBJ. FUT ‘Whichever of them reaches me the sooner, him shall I make king’ (Classical Armenian, Eznik, in Hewitt 1978: 105).

100

Rachel Hendery

c.

c̣ourtēn tandž-ou-oł ajs xełč mardə by-the-cold-ABL being-tormented this poor man-the ‘this poor man who is being tormented by the cold . . .’ (Modern Western Armenian, Feydit 1948: 127 in Hewitt 1978: 124)

It is usually thought that the deranked prenominal construction is a calque on the Turkish relative clause (Feydit 1948: 352) or due to influence from literary Russian (Stilo 2004: 46), but Hewitt argues that it could have developed in a series of gradual changes from the pre-existing postnominal construction. The scenario he outlines is that the auxiliary verb in a sentence like “the church which was built by Gregory the Great . . .” was lost, and then the relative clause marker was also eliminated, since the clause began to be reinterpreted as phrasal. This creates a construction like the English participial relative clause: the church built by Gregory the Great. . . . These intermediate constructions are attested in Classical Armenian. Finally, the participle, now analogically seen as an adjective, moved to the same position in which adjectives in Armenian are most common: prenominally (Hewitt 1978: 128). Since we know that Armenian has been heavily influenced by Turkish, it is probable that both Hewitt’s suggested path of change and language contact were involved: that Armenian recreated the Turkish relative clause strategy using resources already available to it. The question of whether it was a language-internal development hastened and encouraged by outside influence, or a language contact phenomenon supported by internal typological developments is simply a matter of perspective. The deranked prenominal Armenian relative clause therefore may have had three sources: the Armenian postnominal relative clause, the adjective phrase (an analogical source), and the Turkish relative clause. 4.1.5 Development of a correlative construction (f) The development of a correlative relative clause is less commonly attested than other developments. There are no documented cases of the gradual, organic evolution of a correlative construction out of any other type of relative clause. What is known to occur, however, is the transfer of correlative patterns from L1 to L2 when speakers acquire a new language. In at least one case, South African Indian English, this has led to the widespread use of correlatives in a newly created dialect (Mesthrie and Dunne 1990). An example of this is given in example (8). (8) Which-one I put in the jar, that-one is good (Mesthrie and Dunne 1990: 37, ex. 13)

Constructional etymology

101

The use of an interrogative pronoun suggests some relationship to the Standard English relative clause. Hindi, the “donor” language for this construction10, does not use an interrogative here, but rather a specifically relative pronoun. So once again, there are at least two sources contributing to this new construction.

4.1.6 Development of a new postnominal construction (g) Finally, to illustrate the development of a new postnominal relative clause construction, I will take the example of Quechua. In Quechua postnominal relative clauses such as the one illustrated in example 9 are assumed to be modelled on Spanish. These are clearly a recent development, as according to Lefebvre and Muysken (1982: 45) they are not described in the otherwise detailed early grammars such as that by Holguín ([1609] 1842), or that by Ricardo ([1586] 1970). Nor are they found in the conservative modern dialects of the language, but only in Cuzco Quechua. (9)

warma-wan puklla-ra-ni [pi-wan-mi Joseca rima-sqa-n] girl-with play-PAST-1 [who-with-AF Jose talk-PAST-3] ‘I played with the girl that Jose talked with’ (Lefebvre and Muysken 1982: 3, ex. 3)

This example contrasts with the relative clause constructions that are more widely used among the various Quechua dialects, i.e. internally headed relative clauses, prenominal and correlative RCs with nominalized verbs (Lefebvre and Muysken 1982; Cole 1985; Grosu 1998). A correlative construction with a interrogative-based relative marker that looks a little like some sort of intermediate stage between the older and more recent relative clause types also exists and is illustrated in example (10). (10)

pi-ta-n khuya-nki chay-ta-n ñoqa-pas khuya-ni who-acc-af love-2 that-acc-af I-too love-1 ‘who you love, I love too’. (Ricardo, 1970)

10 South African Indian English speakers may have either a northern or southern Indian background. Hindi is the original L1 of the speakers with a northern background, and it seems to be members of this subset of the community that use the correlative construction. Those with a southern background, speakers of Dravidian languages, use a Dravidian-style participial relative clause instead (Mesthrie and Dunne 1990: 35–38).

102

Rachel Hendery

This construction was attested as early as 1586, before the postnominal construction appears. This suggests that the interrogative-based relative pronoun (possibly calqued from Spanish relative clauses) developed first in the correlative construction, and then was extended or reborrowed into a new construction that was a more direct copy of the Spanish one. Even those who argue that the interrogative-marked construction is a Quechua-internal development, not a calque (e.g. Lefebvre 1984; Appel and Muysken 2006: 161), admit that the postnominal position of the new construction is at the very least most likely a result of word order changes triggered by Spanish influence (cf. Appel and Muysken 2006: 161). So once again, it seems that the etymology of the new construction is complicated: both the traditional Quechua correlative construction and the Spanish relative clause have fed into the new postnominal RC. In this section we have seen that for each of the types of new relative clause a language can develop, the “etymology” is complex, with multiple sources playing a role. The examples chosen here were not deliberately selected because they illustrate this point, but only because they are the best-documented cases of the various changes, so we can be most certain of the historical details. In any of the examples in Table 2 for which we have sufficient information to enable close examination, we find the same complexity.

4.2 Non-RC constructions as sources of relative clauses In the previous section, most of the sources of new relative clause constructions discussed were other types of relative clause construction in the language, and relative clauses in other languages with which speakers were in contact. In this section I will discuss the few non-relative constructions that are found as sources of relative clauses. The constructions in question include simple noun phrases, possessives, adverbial clauses, and comparatives. Once again, it will be seen that the new construction is rarely the result of one simple change operating on a single source.

4.2.1 Simple noun phrases Simple noun phrases frequently serve as points of analogical extension to the relative clause. Classifiers, demonstratives, and possibly articles are all elements whose prototypical function belongs to the simple noun phrase, but can be adopted as relative clause markers. Even then, however, the simple noun phrase itself is not the sole source of the new relative clause construction. Take, for

Constructional etymology

103

example, the origin of relative clause constructions that are marked with classifier morphemes. Such constructions are found in various Amazonian languages (cf. Aikhenvald 2000: 88–94); Newari (DeLancey 1986), Cantonese (Shi and Li 2002); and Thai (Sornhiran 1978). An example of an Amazonian classifiermarked relative clause construction (from the language Tuyuca) is shown in example (11). (11) tibãré ãdõ- pé that- CL : PATH - RE here- THEM .CONTR [kĩí atíaribãpi] [3MSG come- recently- SG . NOM - CL : PATH - LOC ] hoáwa- yigí start.down.path- go- 3MSG . PAST. EVIDENTIAL ‘He started down that path over here [that he had recently come on]’ (Aikhenvald 2000: 93) Aikhenvald (2000: 88) suggests that the characteristic of classifiers that leads them to develop into relative clause markers may be their ability to function anaphorically. An example of this anaphoric use can be seen in the example from the Akateko language in example (12). (12) yeeṣ̌in ši nax tṣ̌onwom šʔey tšotan nax all.right said CL : MAN merchant sat down CL : MAN smaxa ṣ̌yetšmane nax sat k’al k’am tšen tumin waiting waited CL : MAN long.time there.was.no CL : ROCK money ‘All right, said the merchant and sat down to wait, he waited, he waited a long time, but there was no money.’ (Aikhenvald 2000: 88) It seems, in fact, that all the examples of languages in which classifiers appear to have become relative clause markers also demonstrate a similar anaphoric use of classifiers. So these anaphoric classifier constructions may well be another source contributing to the evolution of the classifier-relative-clause construction. It is also the case that classifiers occur with a number of other components as well as in a simple noun phrase. While the most well known classifiers occur with numerals, in some languages there are also classifiers that accompany possessives and/or verbs. Cantonese is a language that illustrates the wide range of contexts in which such classifier elements can appear. Historically, the Cantonese element ge had both a demonstrative use and a classifier use (origi-

104

Rachel Hendery

nally denoting lengths of bamboo, later extended to arrows, then candles and certain animals, then to fruit, birds and people, and now used as a general “default” classifier (Aikhenvald 2000: 401). It also came to be used in possessives, adjective phrases, adverbial phrases and to link demonstratives to the corresponding noun (Matthews and Yip 1994: 88–90). But because the order of its historical spread into the various functions is unclear (Shi and Li 2002), we cannot be certain whether the relative clause construction with ge was based on one or more of these other constructions, on its demonstrative use, or original classifier use. At the very least, both the demonstrative and numeral classifier functions of ge already existed when the relative clause construction developed, so it is possible that these both played a role in the evolution of the latter. If more of the other ge constructions already existed, in particular the possessive, adjective and adverbial “linking” functions, it is possible that ge simply became reanalysed as a general-use linking particle, and the new relative clause construction was created on this basis. Classifier constructions are therefore good examples of cases where multiple other constructions also seem to serve as sources for a new relative clause. 4.2.2 Possessives Although possessive constructions and relative clauses share markers and other features in many languages, it is hard to pin down cases where one is undeniably the source of the other. Generally there are other equally plausible explanations for the similarity, or a third type of construction provides an intermediate step between the two. Japanese is a good illustration of this messiness. In Japanese since the 16th century, a headless relative clause construction has existed, which is exemplified in example (13). This construction is discussed in depth in Yap, Matthews, and Horie (2004), who call it an “S-pronominal construction”. (13) Taroo-ga kat-ta no takakat-ta Taroo-NOM buy-PAST NO expensive-PAST ‘the one Taro bought was expensive’ (Yap, Matthews, and Horie 2004: 142, ex. 12) In the late 19th century until the early 20th century, it was possible to use a similar construction for headed relative clauses. This sort of construction still occurs in child language, so it is clearly a natural extension of example (13). But where did (13) come from? There are several theories about this. Yap, Matthews, and Horie (2004) suggest that the no in this construction is ultimately the same no used as a possessive

Constructional etymology

105

marker in Japanese since the 8th century. Martin (1991) has an alternative theory: that this no was originally mono ‘thing’, which explains why it was first used only in headless relative clauses. No matter which of these is the ultimate source, both agree that the relative clause construction was mediated by at least one other construction: what Yap, Matthews, and Horie call the “N-pronominal”. This construction is shown in example (14), and can be used in such ellipses as we would use the corresponding ’s in English: e.g. John’s is over there or Whose is it? Sarah’s. Martin (1991: 282–283) suggests that this no here, too, is a reduction of mono ‘thing’. (4) Taroo no Taroo no ‘Taroo’s’ (N-pronominal function (Yap, Matthews, and Horie 2004: 141 ex. 7)) In Yap, Matthews, and Horie’s explanation, then, we have both the possessive construction (Taroo no hon ‘Taroo’s book) and the ‘N-pronominal’ function both feeding into the creation of a new (headless) relative clause construction. In Martin’s scenario, only the construction with (mo)no in the N-pronominal construction is the source of the relative clause. Since no existed at the same time in a possessive use, however, as soon as mono was reduced to no, speakers would have been able to see the possessive and relative constructions as related. The evidence that they did so, or at least that they lost the connection with the mono construction is the extension of the no relative clause to headed constructions in the late 19th century.

4.2.3 Adverbial clauses The relationship between adverbial clauses and relative clauses is a little more complicated. This is because extension seems to occur in either direction (cf. Hendery 2012: chapter 2). It is also because a boundary between true adverbial clauses and true relative clauses is not straightforward to draw: adverbial clauses introduced with where in English, for example, are simply headless relative clauses, and the same holds for the locative adverbial marker in many other languages. In this section I will summarize the development of the poumarked relative clause construction in Greek as presented by Nicholas (1998), since this is perhaps the best-documented case available. From the earliest attested period where relative clauses with (ho)pou appear in Greek, (ho)pou is also found in adverbial clauses. Nicholas (1998: 178) argues

106

Rachel Hendery

that this is because (ho)pou was created on analogy with the pre-existing hostis and therefore it inherited the multiple functions of this element, including indefinite headless relativization and embedded questions. The indefinite headless relatives (which, when locative, are adverbial clauses), and the “true” relatives are compared in examples (15)–(18). Headless indefinite RC with hostis (15)

dó:so: gàr díp hron te dúo: Give.FUT.1S for chariot.ACC and two híppous . . . t’eriaók h enas and arching-neck.MASC . ACC . PL HORSE . MASC . ACC . PL hóstis ke tlaíe: hoî t’autô:I REL . M . NOM . S particle dare.OPT. PRES .3S 3S .DAT and’self.3MS kû:dos ároito. glory.ACC win.AOR .OPT. MID.3S ‘For I will give him a chariot, and two horses of arching neck, . . . to whosoever shall dare the deed, and for himself shall win glory.’ (Iliad X 305–7, via Nicholas 1998: 172 ex. 8b)

Adverbial clause (locative headless indefinite RC) with hopou (16) hópou d’apóllo:n skaiòs ê:I Where for’Apollo ignorant.NOM . MS be.3S . PRES tínes sop h oí? anyone.M . PL . NOM wise.M . PL . NOM ? ‘Where Apollo is ignorant shall men be wise? (Euripides Electra 972 via Nicholas 1998: 178 ex. 11a) True relative clause with hostis (17)

áp h ron dè: keînós ge kaì foolish.MS . NOM indeed DEM . MS . NOM at.least and outidanòs pélei ané:r, hóstis worthless.MS . NOM be.3S . PRES man.MS . NOM REL kseinodóko:I érida prop hére:tai aét h lo:n host.MS . DAT strife.FS . ACC propose.3S . PRES . SUBJ contest.MPL .GEN ‘That man indeed is foolish and worth nothing, whoever proposes a strife in contests with his host’ (Od VIII 210, via Nicholas 1998: 173, ex. 8c)

Constructional etymology

107

True relative clause with hopou (18)

eîta en tê:I toiaúte:I pólei hópou mè: then in a.FS . DAT such.FS . DAT state.FS . DAT where not lógo:I érgo:I te hikanoì thought.MS . DAT work.NS . DAT and sufficing.MPL . NOM p h ulakes eîen . . . guardian.MPL . NOM be.3PL . PRES .OPT ‘In such a state, where there are no custodians competent in act as in thought [ . . . ]’ (Plato, Laws. 964C, via Nicholas 1998: 178, ex. 11c)

According to Monteil (1963: 387, cited in Nicholas 1998), at this stage hostis was also just starting to appear in some examples that are interpretable as specific (restrictive) relative clauses, such as the example in (17). The new form hopou, formed from ho- (article/demonstrative) and -pou (locative interrogative), just as hostis was formed from ho- and -tis (nominative interrogative), took on the same functions with a locative (and hence adverbial) flavour: marking of indirect questions about location, generalising locative adverb (‘wherever’), and marker of oblique relative clauses (Nicholas 1998: 178). Nicholas, and before him, Tzartzanos (1991) reconstructed a subsequent pathway of development that led to pou being used in more abstract spatial relative clauses/adverbials, adverbials of cause and then temporal adverbials. The extension of pou to core relatives (subject and direct object NPrels) occurred in Late Middle Greek or Early Modern Greek. There are three clear examples from the fifth, sixth and seventh centuries (Nicholas 1998: 207), and then none at all until 980 AD. The fifth century example is given in (19). Non-locative RC with hopou (Middle Greek) (19) tòn adelphòn hópou eîkhe pròs tè:n the.MS . ACC brother.MS . ACC REL have.3S towards the.FS . ACC . autòn lúpe:n him.MS . ACC sorrow.FS . ACC ‘the brother who had distressed him’ (Apophthegmata Patrum 300B, via Nicholas 1998: 207, ex. 34a) The Greek data provide a demonstration of how constructional polysemy can arise without gradual extension. The pou construction was polysemous from the start, because pou was created on analogy with another element which had

108

Rachel Hendery

undergone extension and inherited the polysemy of its source. The etymological sources of the pou-marked relative clauses in Greek are therefore simultaneously but to varying degrees: indirect questions, adverbial clauses, generalising indefinite (headless) relative clauses, articles/demonstratives, and interrogatives. These relationships are depicted in the diagram in Figure 7.

Figure 7: Sources contributing to Greek relative clauses with (ho)pou

4.2.4 Comparatives The diachronic relationship between comparative and relative constructions most frequently seems to be a matter of relative clauses being a source of comparatives (Haspelmath and Buchholz 1998: 290), but the other direction probably occurred in English and Old Norse (Faarlund 2004: 259). I will take Norse as an example here. The first occurrence of the comparative marker sem in a Norse relative clause construction was in an inscription where it was combined with another marker es, shown here in example 20. This construction therefore looks like a blend between the comparative and a pre-existing relative clause: so once again we have at least two sources contributing to the construction’s etymology. (20)

sa hit aki sims (= sem es) uti furs This was.called Aake, sims (= REL REL) abroad traveled.REFL ‘This was Aake, who traveled abroad’ (= died) (Lindblad 1943: 113)

As well as a primary comparative function, the marker sem was also found in adverbials of place, which may have provided a bridging construction to the relative clause in some contexts.

Constructional etymology

109

A pre-existing relative clause construction in Norse, marked with er, was also multifunctional, and could be used in comparative constructions such as example 21. This means it was almost certainly a model for the new sem construction. (21) ƿá fundu menn hans í gamma einum then found.3 PL men.NOM his in hut.DAT one.M . DAT konu ƿá er ƿeir hǫfðu enga eét jamvæna woman.ACC that REL they. M had.3 PL not.F. ACC seen even-beautiful.F. ACC ‘Then his men found in a hut a woman so beautiful that they had never seen anyone like her’ (Faarlund 2004: 261) Finally, it is thought that sem developed relative clause marking functions first in the Norwegian dialects of Old Norse, spreading later to Iceland, Sweden and then Denmark (Noreen 1923: 319; Faarlund 2004: 259). In these latter varieties, the comparative construction with sem existed when the relative construction was adopted, so we could say there were three sources contributing to the new construction in these cases: the sem-comparative/adverbial, the preexisting relative clause (with er), and the ‘foreign’ sem relative clause in the dialect from which it spread. In the previous section we saw that relative clause constructions that develop out of other relative constructions had complex, multi-faceted etymologies, and in this section we have seen that the same holds for relative clauses that develop out of non-relative constructions. In fact, even describing the changes as one construction “developing out of” one another is misleading. It is better to visualize a construction as participating in a web of diachronic interactions – and the way in which we carve up this web and declare everything in one space to be one construction and everything in another to be a different one, is fairly arbitrary. This becomes very clear when we try to represent constructional etymologies visually. In the following section I will discuss the ways in which constructional etymologies are sometimes represented visually and will suggest that these representations can be improved if we do not gloss over the sorts of complex relationships that I have discussed above.

5 Conceptual space representations I have shown in the previous sections how inclusion of multiple sources, potential sources, and language contact influences in constructional etymologies leads to a more accurate understanding of how and why relative clauses end

110

Rachel Hendery

up looking the way they do. One difficulty with these more complex etymologies, however, is that they are much more difficult to represent visually (and twodimensionally) than simpler source-outcome models. In this section I will begin by introducing the sorts of representations of constructional change that are frequently found in the literature, and will then move on to presenting some suggestions for how they can be extended to represent the more complex etymologies that I have discussed above. Diagrams of semantic or functional change are common in the grammatical literature, but sometimes consist of rather linear, unidirectional pathways, where each function is a step mediating between a source and an outcome, as in the example from Van der Auwera and Plungian (1998: 88) reproduced in Figure 8.

Figure 8: Grammaticalization of possibilities, from Van der Auwera and Plungian (1998: 88)

The diachronic semantic map in Figure 8 shows the “grammaticalization pathways” of words that encode possibility. (The “deontic possibility” set is located within the set of “participant-external possibility” because it is seen as a subtype of the latter.) Van der Auwera and Plungian (1998) do also use more complex maps, with many-to-one and one-to-many connections. However, these connections are not meant to suggest multiple sources, but rather alternative sources or outcomes of change in different languages. Van der Auwera and Plungian’s diachronic maps, like many others in the literature, are intended to present a cross-linguistic summary of all possible pathways of change. Diagrams intended to represent typological reality rather than the history of an individual language sometimes have bidirectional links, representing the fact that change in either direction is possible. This is the case for the map of nominalizers/pronominalizers in Yap, Matthews, and Horie (2004: Fig. 3), reproduced here as Figure 9. The connections between each step of the pathway are bidirectional. Yap, Matthews, and Horie’s diagram still glosses over the fact that one construction does not smoothly “become” another. In this map that is not a serious problem, as each change is of approximately the same type: simple extension of a construction to a new function. This diagram leaves out, of course, any con-

Constructional etymology

111

Figure 9: “Grammatical functions centered on (pro)nominalizers as an implicational map. Note that x represents the (pro)nominalizer morpheme under investigation (e.g. Japanese no, Mandarin de, and Malay empunya).” (From “From pronominalizer to pragmatic marker: implications for unidirectionality from a crosslinguistic perspective”. In Olga Fischer, Muriel Norde & Harry Perridon (eds.), Up and Down the Cline – The Nature of Grammaticalization, 2004, pp 137–168. With kind permission by John Benjamins Publishing Company, Amsterdam/ Philadelphia. www.benjamins.com)

nections to pre-existing constructions filling those functions, as well as any other models for the functional polysemy that may have existed. To represent the full story of the etymology of some relative clause constructions we would need to include not only these complicating factors, but also different kinds and different strengths of links between them and their sources. For example, we may have borrowing of a marker from one language to another, calquing without lexical borrowing, reanalysis, extension of a marker by analogy, or transfer of word order patterns. Semantic maps and “conceptual spaces”, both for individual lexemes and for constructions, allow the mapping of the terms/constructions of an individual language onto the typological diagram, showing at the same time synchronic and diachronic reality. Croft (2001: 92–95; 2010) discusses these sorts of maps. An example for Romanian indefinite pronouns overlaid onto Haspelmath’s semantic map of indefiniteness is reproduced in Figure 10. Croft argues that such maps should be data-driven: derived from comparison of multiple exemplars. In other words, the Romanian and other language data should first be graphed, to show which senses/functions are expressed uniquely, and which are expressed by polysemous forms. When sufficient lan-

112

Rachel Hendery

Figure 10: Croft’s mapping of Romanian and Kazakh indefinite pronouns onto Haspelmath’s semantic map (Croft 2010: Figure 2)

guages are included, this allows the automatic generation of a semantic map such as Haspelmath’s, which underlies Figure 10. The advantage of semantic maps is that the overlay of several languages would allow us to see the models that exist in one language for new constructions in another. For example, if we were to create a semantic map of pronouns, we would see that interrogative and relative functions in French and Spanish are both performed by the same function, and that in older Basque this was not the case, but in modern Basque it was. To see this on one map we would of course need to treat multiple stages of a language as separate languages. The difficulty with such maps for our purposes is that when we move beyond lexical items to entire constructions, polysemy is not the only relationship we wish to represent. In the discussion in the previous sections we have seen that numerous types of diachronic and synchronic relationship can obtain between constructions: even various different types of polysemy. Two constructions can be marked with the same marker – this is a type of polysemy. But two constructions that are both postnominal, with balanced verbs and the same marker are much more similar than two that share a marker, but do not have the same position or verb type. Even when only examining the markers themselves, while analogical extension, reanalysis and even calquing can leave evidence in the shape of polysemy, direct borrowing of a marker from one language to another does not. And even when we see polysemy in a semantic map, we cannot be certain whether that polysemy is the result of extension, reanalysis

Constructional etymology

113

or calquing. Semantic maps, with bidirectional relationships between elements, where necessary, are a good starting point. But the ideal representation of the diachronic relationship between constructions would also be able to represent multiple languages, multiple stages of each language, and different types and strengths of relationship between constructions. Such a diagram allows us to represent something more closely resembling the speakers’ reality: the constructions that speakers of the language have considered similar, including relationships between constructions in different languages spoken in the speech community. The difficulty with showing multiple languages, multiple types of connection, and many-to-many links on the same diagram is that it quickly becomes unwieldy. The key is to treat them not as cross-linguistic typological maps, but rather to represent only a small number of languages at any one time: a single multilingual situation rather than some sort of universal truth. As an example of this sort of “diachronic construction map”, the Old English interrogative-marked relative clause(s) are represented in Figure 11.

Figure 11: Old English interrogative-marked relative clauses

The two main boxes group together those constructions that belong to the same language. These boxes are labelled with the language names. The spatial arrangement of the constructions represents time: older constructions are further to the left; newer ones are further to the right. Constructions that are from the

114

Rachel Hendery

same time period are lined up underneath each other. If desired, the columns could be labelled with the time period (Old English, Middle English and Late Middle English in this case). Lines connecting constructions represent a formal relationship. I have used lines ending in an arrow to represent the relationship “developed directly into”. This is the case where a marker merely underwent sound change, or where one marker is just a differently case-marked version of another. Lines not ending in arrows represent a less direct relationship. This includes reanalysis, analogy and potential analogy, i.e. very similar forms occurring in similar functions, or constructions in the same time period that use the same markers for different functions. Finally, the dotted line in this case represents a potential cross-language analogy. I would suggest this could be used to represent the type of relationship that exists between e.g. two markers that are phonologically very different, but that belong to the same word class and perform the same function. This will usually be the case with calques, but may also occur within a language. The dotted line surrounds constructions that have the same function, i.e. here: relative clauses (excluding the indefinite “whoever” type). Because these are grouped into a labelled set, I have not written “Relative clause with X marker” for each construction, but for simplicity’s sake have left the marker to represent the whole construction. In cases where one wishes to represent multiple variables such as marker, clause order, verb type, etc, this can be done as in Figure 12. Listing each variable on a separate line makes it possible to connect two constructions by means of only one variable – e.g. draw a line from the

Figure 12: Russian postnominal and correlative relative clause constructions.11

11 In Figure 12 all the constructions of interest have the same (relative) function, and are from the same language, so the conventions I used in Figure 11 to group constructions by language or function are redundant here and have not been used.

Constructional etymology

115

marker in construction 1 to the marker in construction 2 to show the marker was adopted, or to draw a line from “postnominal” in construction 1 to “postnominal” in construction 2, to show that the same word order is used. An advantage of these sorts of visual representations of constructional etymologies is that they show the difference between constructions that “develop into” another construction in a later period (in the same way that a lexical item like frēond ‘loved, favoured’ becomes friend), and constructions that spawn a new function or form, but also remain in their previous incarnation alongside the new construction. We can see this latter sort of development quite clearly in the map of Russian in Figure 12: instead of two constructions in the first column developing gradually into two other constructions in the second column, we find four constructions in the later stage of the language: two retentions and two new innovations. It is also possible to “read off” such diachronic construction maps the following information: – the number of possible influences on the development of the construction (how many lines connect to the construction in question? How many connect indirectly (i.e. connect to a construction that in turn connects to the one of interest?) How strong are these connections? – the number of other constructions in the language performing the same function (grouped within the dotted line). This has implications for how much competition the new construction has, and hence what is likely to happen next – loss, redistribution, etc. – the influences from other languages – the centrality or ‘fringe’ status of various constructions: are there many lines connecting them to others or only a couple? Do they belong to the same sets as most of the others? The same language? The more tightly embedded into the “network” a given construction is, the more likely it is that speakers would consider it a “type” of the construction that the diagram focuses on. Note that this degree of embeddedness in the network is not absolute, since each diagram is only a snapshot of the network centered on one particular construction – in these cases, the relative clause. In Figure 11, the embedded question construction looks very “fringe”, but if we had instead created a “questions” map, it would no doubt be much more connected. Finally, such diachronic construction maps provide an alternative to having to choose which variable to privilege in a taxonomy of constructions. For example, in Russian we could present a taxonomy of medieval relative clause types (based purely on those in Figure 12) as either the left-hand hierarchy or the right-hand hierarchy in Figure 13.

116

Rachel Hendery

Figure 13: Two alternative taxonomies for Medieval Russian relative clauses

In the map in Figure 12, however, it is apparent that there are relationships between the late medieval constructions both in terms of the clause position and in terms of the marker, and we do not have to treat one as a subtype of the other unless we wish to. The fuzzier taxonomy is less misleading with regard to representing speakers’ own intuitions about the relationship between the constructions. Unless we have psychological evidence to the contrary, there is no reason to think that speakers perceive one construction as a subtype of another, but rather they would merely be aware of (some) formal and functional similarities and differences: i.e. the relationships in the right-hand column of Figure 12. The sorts of maps presented in this section are not ideal for all purposes. There are situations for which a hierarchical taxonomy, or a more typical semantic map or grammatical pathway diagram is more appropriate. When we want to represent a fuller picture of the diachronic situation, however, especially if language contact is a relevant factor, it is worth considering maps like those in Figure 11 or 12.

6 Conclusions The sorts of diachronic relationships that can exist between constructions are much more complex than simple source–outcome relationships of the sort found in what we might think of as “prototypical” lexical etymologies. The types and subtypes of construction that linguists may identify are not always the same constructions that speakers treat as “similar” and “different”, yet processes of change such as analogy, borrowing or calquing are the direct result of speakers’ intuitions. As far as relative clauses specifically are concerned, I have shown here that the markers, verb types, and position of the construction are all relatively independent variables. One or more of these can be borrowed from another language, or changed on analogy with other constructions in the language. These other

Constructional etymology

117

constructions that can participate in analogy or other diachronic interactions include other relative clauses, but also simple noun phrases, possessives, adverbial clauses, and comparatives. By working through examples of each of the types of attested change in relative clause constructions (see Table 2), I showed that none of these is straightforward. Even the development of a “new” construction, such as a prenominal relative clause where previously only postnominals have existed, or vice versa, is not a case of a single source construction being adapted to a new function. Rather there are frequently multiple other constructions that feed into the new one, often including constructions from another language. For this reason, we would miss important generals by limiting a typology to a subset of constructions that share one function; nor is it ideal to look for a simple chain of constructions where the original function F1 is extended to F2, then F3, etc. To understand constructional etymology and change it is necessary to view them in the context of the whole web of other constructions that surrounds them, including those which only share a marker, or only share a function, or only share one other feature such as word order. Such networks of constructions make for very complex typologies, which cannot be represented using current diagrammatic conventions. I therefore turned in the final section of this paper to the question of whether it is possible to capture such complex diachronic relationships visually. I outlined a suggestion for a diachronic construction map that allows a researcher to display complex and nuanced findings about synchronic relationships, language contact influences and the different sorts and strengths of diachronic relationships.

References Aikhenvald, Alexandra. 2000. Classifiers: a Typology of Noun Categorization. Oxford: Oxford University Press. Allen, Cynthia. 1980. Topics in Diachronic English Syntax. New York & London: Garland Publishing. Antinucci, Francesco, Alessandro Duranti & Lucyna Gebert. 1979. Relative clause structure, relative clause perception, and the change from SOV to SVO. Cognition: International Journal of Cognitive Psychology 7. 145–175. Appel, René & Pieter Muysken. 2006. Language Contact and Bilingualism. Amsterdam: Amsterdam University Press. Comrie, Bernard. 1988. General features of the Uralic languages. In: Denis Sinor (ed.), The Uralic Languages: Description, History and Foreign Influences, 451–477. Leiden: Brill. Comrie, Bernard. 1998. Rethinking the typology of relative clauses. Language Design 1. 59–86. Cristofaro, Sonia. 2005. Subordination. Oxford: Oxford University Press.

118

Rachel Hendery

Croft, William. 2001. Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford: Oxford University Press. Croft, William. 2010. ‘Relativity, linguistic variation and language universals’, CogniTextes (Online) 4. http://cognitextes.revues.org/303 (accessed 25/2/2011). DeLancey, Scott. 1986. Relativization as nominalization in Tibetan and Newari. Paper presented at the 19th International Conference on Sino-Tibetan Languages and Linguistics. http://www.uoregon.edu/~delancey/papers/relnom.html (accessed 24/2/2011). Durkin, Philip. 2009. The Oxford Guide to Etymology, Oxford: Oxford University Press. Faarlund, Jan Terje. 2004. The Syntax of Old Norse. Oxford: Oxford University Press. Feydit, Frédéric. 1948. Manuel de Langue Arménienne (Arménien Occidental Moderne). Paris: Klincksieck. Fried, Mirjam. 2009. Construction Grammar as a tool for diachronic analysis. Constructions and Frames 1(2). 262–291. Grunberg, Madèleine (ed.). 1967. The West-Saxon Gospels. Amsterdam: Scheltem & Holkema. Heine, Bernd & Tania Kuteva. 2002. World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Hendery, Rachel. 2012. Relative Clauses in Time and Space (Typological Studies in Language). Amsterdam & Philadelphia: John Benjamins. Hewitt, Brian. 1995. Georgian. Amsterdam and Philadelphia: John Benjamins. Hill, Jane H. & Kenneth C. Hill. 1981. Variation in relative clause construction in modern Nahuatl. In: Frances E. Karttunen (ed.), Nahuatl Studies in Memory of Fernando Horcasitas (Texas Linguistic Forum 18), 89–104. Austin: Dept. of Linguistics, University of Texas at Austin. Johanson, Lars. 1998. Code copying in Irano-Turkic. Language Sciences 20(3). 325–337. Haspelmath, Martin & Oda Buchholz. 1998. Equative and similative constructions in the languages of Europe. In: Johan van der Auwera & Dónall Baoill (eds.), Adverbial Constructions in the Languages of Europe, 277–334. Berlin & New York: Walter de Gruyter. Holguín, Diego G. 1842. [1609]. Gramática y arte nueva de la lengua general de todo el Peru, llamada lengua quichua o lenqua del Inca. New revised and corrected ed. by Francisco de Canto, Lima. Hurch, Bernhard. 1989. Hispanisierung im Baskischen. In: Vielfalt der Kontakte, Norbert Boretzkey & Birgit Igla (eds). Essen: Brockmeyer. Karttunen, Frances. 2000. Language contact in Latin America. In: Thomas L. Willett (ed.), UtoAztecan: Structural, Temporal & Geographic Perspectives: papers in memory of Wick R. Miller by the Friends of Uto-Aztecan, 387–410. Hermosillo, Son: Universidad de Sonora. Kikuta, Chiharu Uda. 2002. Clausal complement or adverbial clause?: toward an integral account of the Japanese internally-headed relative clause. In: Frank van Eynde, Lars Hellan and Dorothee Beermann (eds), Proceedings of the 8th International HPSG Conference, Norwegian University of Science and Technology (3–5 August 2001), 202–220. CSLI Publications. http://csli-publications.stanford.edu/202 (accessed 24/2/2011). Leckey, Celia Elizabeth. 1992. Relative clauses in medieval Russian texts. Berkeley: University of California dissertation. Lehmann, Christian. 1984. Der Relativsatz. Tübingen: Gunter Narr. Lefebvre, Claire. 1984. Grammaires en contact: définitions et perspectives de recherche. Revue Québécoise de Linguistique 14(1). 11–48. Lefebvre, Claire & Pieter Muysken. 1982. Relative Clauses in Cuzco Quechua: Interactions between Core and Periphery. Bloomington: Indiana University Press.

Constructional etymology

119

Lindblad, Gustaf. 1943. Relativ Satsfogning i de Nordiska Fornspråken [Relative clauses in the ancient Nordic languages]. Lund: Walter Ekstrands. Mailhammer, Robert. 2007. The Germanic Strong Verbs: Foundations and Development of a new System. Berlin & New York: Mouton de Gruyter. Martin, Mary L. (ed.). 1984. The Fables of Marie de France. Birmingham, AL: Summa Publications. Martin, Samuel E. 1991. Recent research on the relationships of Japanese and Korean. In: Sydney M. Lamb and E. Douglas Mitchell (eds.), Sprung from Some Common Source, 269– 292. Stanford: Stanford University Press. Matthews, Stephen & Virginia Yip. 1994. Cantonese: A Comprehensive Grammar. New York: Routledge. Mesthrie, Rajend & Timothy T. Dunne. 1990. Syntactic variation in language shift: The relative clause in South African Indian English. Language Variation and Change 2. 31–56. Monteil, Pierre. 1963. La Phrase Relative en Grec Ancien. Paris: Klincksieck. Nicholas, Nick. 1998. The story of pu. Melbourne: The University of Melbourne dissertation. Noël, Dirk. 2007. Diachronic construction grammar and grammaticalization theory. Functions of Language 14 (2). 177–202. Ricardo, Antonio. 1970. Reprint. Arte y vocabulario en la lengua general del Peru llamada quichuay en la lengua espanola. ed. Rafael Aguilar Paez, Lima. Original edition, Lima, 1586. Romaine, Suzanne. 1980. The relative clause marker in Scots English: diffusion, complexity and style as dimensions of syntactic change, Language in Society 9. 221–249. Seppänen, Aimo. 1997. The genitives of the relative pronouns in present-day English. In: Jenny Cheshire and Dieter Stein (eds.), Taming the Vernacular: From Dialect to Written Standard Language, 152–169. London: Longman. Seppänen, Aimo & Göran Kjellmer. 1995. The dog that’s leg was run over: on the genitive of the relative pronoun. English Studies 76. 389–400. Shi, Yuzhi & Charles N. Li. 2002. The establishment of the classifier system and the grammaticalization of the morphosyntactic particle de in Chinese. Language Sciences 24. 1–15. Sornhiran, Pasinhee. 1978. A transformational study of relative clauses in Thai. Austin: University of Texas at Austin dissertation. Stilo, Donald. 2004. Iranian as a buffer zone between the universal typologies of Turkic and Semitic. In: Éva Ágnes Csató, Bo Isaksson & Carina Jahani (eds.), Linguistic Convergence and Areal Diffusion: Case Studies from Iranian, Semitic and Turkic, 35–63. London: Routledge-Curzon. Van der Auwera, Jan & Vladimir A. Plungian. 1998. Modality’s semantic map. Linguistic Typology 2 (1). 79–124. Vennemann, Theo. 2000. Zur Entstehung des Germanischen. Sprachwissenschaft 25. 233–269. Yap, Foong Ha, Stephen Matthews & Kaoru Horie. 2004. From pronominalizer to pragmatic marker: implications for unidirectionality from a crosslinguistic perspective. In: Olga Fischer, Muriel Norde & Harry Perridon (eds.), Up and Down the Cline – The Nature of Grammaticalization, 137–168. Amsterdam: John Benjamins. Zeller, Jochen. 2004. Relative clause formation in the Bantu languages of South Africa. Southern African Linguistics and Applied Language Studies 22 (1–2). 75–93.

Theo Vennemann

Concerning myself 1 1 Purpose The English reflexive and intensifying pronoun myself, yourself, etc. has received much attention, most recently in a contact linguistic (Vezzosi 2003), in a comparative (König and Vezzosi 2007), and in an historical perspective (Lutz 2002).2 The most specific recent explanatory proposal for the origin of the English paradigm was made by Vezzosi who “put forward [the] hypothesis that the M[iddle] E[nglish] self-form grammaticalized under the strong substratal influence of Insular Celtic” (Vezzosi 2005: 240).3 The present paper generalizes this explanation in two directions: It proposes a contact explanation for the Insular Celtic way of expressing intensification and reflexivization; and it proposes a contact explanation for the fact that Old English – more generally: Ingvaeonic4 – and Insular Celtic lack a reflexive pronoun in the first place.

1 I would like to thank Prof. Lutz Edzard (University of Oslo), Dr. Stephen Laker (Kyushu University, Fukuoka), Prof. Angelika Lutz (University of Erlangen), Prof. Robert Mailhammer (then Arizona State University, Tempe, now University of Western Sydney), and two anonymous referees for valuable corrections and suggestions. I also have to thank Prof. Peter Schrijver (Utrecht University) for a letter of 20 January 2011 which I will cite further below, and for additional advice in an e-letter of 4 October 2011. Prof. Cor van Bree (University of Leiden) kindly sent me much information on the reflexive in the history of Low Franconian, of which some is cited in the Appendix below. Prof. Rainer Voigt (Freie Universität Berlin) corrected a number of forms in the Semitic part of the paper, for which I owe him thanks. While some of these colleagues whose assistance I gratefully acknowledge have signaled agreement with the interpretations offered in this paper, there is no implication that the same is true for all of them, or for all parts of the paper. 2 A concise introduction to the terminology and typology of this topic is provided in König and Siemund (2000), König (2001). Explanation for the correlation between reflexives and intensifiers are attempted e.g. in Lyutikova (2000, Ms.) and in Gast and Siemund (2006). 3 Cf. the review of Vezzosi’s proposal in Filppula, Klemola, and Paulasto (2008: 95–97), resulting in the following positive assessment: “The case of the self-forms is an interesting addition to the list of those features that set English apart from its Germanic sisters and at the same time link it with its Celtic neighbours.” 4 I.e. the North Sea Germanic languages: Old English, Old Frisian, and to some extent Old Saxon and Old Low Franconian as a subgroup within West Germanic. Cf. Maurer (1952: 135), also the Internet site “Ingvaeonic languages”, http://en.wikipedia.org/wiki/Ingvaeonic_ languages (accessed 26 January 2012), where the loss of the Germanic reflexive pronoun is listed as the second of six unambiguous defining criteria. See also the Appendix below.

122

Theo Vennemann

2 The reflexive in English and Welsh The proposal that the English reflexive is modeled on the Welsh is not entirely new. The following is Preusler’s (1956) account: Im A[lt]e[nglischen] und noch überwiegend bei Shakespeare und in den heutigen mundarten dient das einfache personalpronomen als reflexiv. Daneben kommt schon ae. verstärkung mit self vor, doch erst seit der ersten hälfte des 13. jhs das possessiv mit self (Einenkel [1916: §49 α and ζ]). Jespersen bringt für das letztere schon einen einzelnen beleg aus Aelfreds zeit, Oros. 186.7 þæt he heora self onseon nolde ([1940:] II 8.4.3.1)5; nach Einenkel ist es häufiger erst seit der mitte des 13. jhs. Aus Chaucer bringt Jespersen einen fall, wo 3 m[anuskripte] us self und 3 m[anuskripte] our self schreiben [Canterbury Tales E 107f. “. . . that we / Ne coude nat vs self deuysen how. . .”]. – Das Kymrische kennt seit alters beides, my hun ich selbst neben fy hun mein selbst. Die neigung zum possessiv ist hier so stark, daß es sich in allen personen durchgesetzt hat. Die pluralformen zeigen auch den plural von hun: ein hunain wir selbst usw.; im Englischen treten die pluralformen nach Jespersen [1940: II 8.4.3.1] im 16. jh. auf (More, Utopia, vom j[ahre] 1516). (Preusler 1956: 345)6 [In Old English, and still predominantly in Shakespeare’s works and in the present-day dialects, the simple personal pronoun serves as reflexive. In addition, strengthening with self occurs as early as Old English, but the possessive with self only from the first half of the 13th century onward (Einenkel 1916: §49 α and ζ). For the latter construction, Jespersen (1940: II. 8.4.3.1) offers a single attestation even from the time of Ælfred (Orosius 186.7 þæt he heora self onseon nolde). According to Einenkel it does not become more frequent before the middle of the 13th century. Jespersen (1940: II. 8.4.3.1) cites a case from Chaucer where three manuscripts write us self and three, our self [Canterbury Tales E 107f. “... that we / Ne coude nat vs self deuysen how...”]. – Welsh has known both constructions from the earliest times, my hun ‘[lit.] I self’ alongside fy hun ‘my self’. The tendency to use the possessive was here so strong that it asserted itself in all persons. The plural forms also show the plural of hun: ein hunain ‘we selves’ etc. According to Jespersen (1940: II. 8.4.3.1), the plural forms first appear in English in the 16th century (Thomas More, Utopia, a. 1516)]7

The proposal of Celtic influence in the development of the English reflexive has also been made, with various degrees of conviction, by Tristram (1999: 24), Lange (2006), and Poppe (2009). Such recourse to language contact does appear necessary in the case at hand; for whereas German, as a closely related Germanic language, only has a

5 This is in harmony with the observation that early Celtic influence generally is stronger in West Saxon than in the other Old English dialects; cf. Lutz (2009: 235). 6 Cf. similarly already Preusler (1938: 186–187). 7 Translations in this paper (marked by brackets) are mine.

Concerncing myself

123

general adverbial intensifier selbst/selber 8 and a single third person reflexive pronoun, sich ‘him-, her-, itself, themselves’9, using otherwise the personal pronoun in reflexive function, English has like Welsh a complete paradigm of composite pronouns functioning both as reflexives and as intensifiers: English myself yourself himself herself itself

Welsh ourselves yourselves themselves

fy hun dy hun ein hun

ein hun eich hun eu hun

Table 1: Composite reflexive/intensivier pronouns in English and Welsh

There exist regional and stylistic variants: Besides the Welsh paradigm above, Jones and Thomas (1977: 200) list a second paradigm, yn/fy hunan, dy hunan, ’i hunan; yn hunain, ych hunain, ’i hunain and a third, yn/fy hun, dy hun, ’i hun; yn hunan, ych hunan, ’i hunan. Cf. also King (2003: §132) for paradigms and examples. – The Welsh construction is old. Essentially the same forms are found in Middle Welsh: Fy hun ‘myself’, &c. The reflexive pronoun consists of the prefixed possessive pronoun, followed by hun or hunan, which are forms of the numeral un.10 The pl[ural] is hunein, but hun is also found: eu lles e hun. . . , wynt e hun [‘themselves’]. . . . The form of the personal

8 In the history of German, selb- has never developed any particular relationship to the construction of reflexives, except that it often accompanies reflexive pronouns as an intensifier; cf. Behaghel (1923: chapter “Die Verwendung von selb”, §§212–216). In Old High German, selbwas a regularly inflected adjective (Braune 2004: §290). 9 Proto-Germanic had a set of reflexive pronouns only for the 3rd person (cf. Goth. gen. seina, dat. sis < +siz, acc. sik; ON sīn, sēr, sik), the same forms in the singular and in the plural, as in other Indo-European languages as well (e.g. Lat. suī, sibī, sē); cf. Seebold (1984: 57–58, 73– 79). English, Frisian, Low German and probably Old Low Franconian lost the entire set of reflexive pronouns in prehistoric times; “only the old genitive sin survived as a possessive of limited occurrence” (Kisbye 1972: II. §S1-1). Old High German lost the dative form +sir, for which the accusative sih (NHG sich) stepped in after the Middle High German period. (Cf. the succinct description of the Germanic situation in Krahe 1969: §34.) I will return to the Ingvaeonic loss of the reflexive pronoun in section 7 below. 10 The numeral un ‘one’ “is freely used as a pronominal; it may be substantival or adjectival” (Evans [1964] 1970: 87), exactly as self in Middle English; see below. In the latter function it may mean ‘same’ (Evans 1970: 88). As to its etymology, Prof. Schrijver writes in his letter of 20 January 2011 (cf. note 1 above):

124

Theo Vennemann

pronoun in the 1. pl. is ny: ny hunein. . . , also ny hûn. . . . Sometimes the -y of the prefixed pronoun 1. and 2. sing. appears as -u. . . . The forms mi hun, mu hunan. . . , tu hun are also found. (Evans [1964] 1970: 89)11

3 The “exceptional” nature of the English and Welsh reflexive Preusler’s and Vezzosi’s is in my view an excellent explanation for the development of the un-Germanic English reflexive and intensifying pronoun, because the similarity between the English and the Welsh ways of expressing reflexivity and intensification are indeed very similar, and both very different from the Germanic way. In an Anglicist perspective the inquiry ends here. But in a wider perspective, of course, the inquiry has to be continued, because now the question arises how Welsh acquired its reflexive and intensifying pronoun. From an Indo-European point of view this question is exactly as important as Preusler’s and Vezzosi’s proposal concerning English, because the Welsh method is just as un-Indo-European as the English way is un-Germanic.12 According to Haspelmath (2001) and Siemund (2002, 2003), English and Celtic are “exceptional among the European languages in that the reflexive and the intensifier are formally identical” (Lange 2007: 259).13 Why English and Insular Celtic stand alone within the Indo-European language family regarding this section of grammar would naturally be best explained by the special history of these languages. For English this is precisely the approach Preusler and Vezzosi chose. The best explanation for Welsh would therefore be one with the same ground plan, i.e., contact with a language that could function in the same way for Celtic as Celtic functioned for English, namely as a substratal contact language itself possessing the property at issue, formally identical reflexives and intensifiers. Das h- von wal. hun(an), pl, hun(ein) und bretonisch hunan führt man wenigstens seit Morris Jones (1913: 307) auf den Stamm des idg. Reflexivpronomens *sw- zurück. Entweder wurde dies vor das Zahlwort für ‘eins’ präfigiert (so Morris Jones), oder es liegt eine Pronominalform *swoi-no- zugrunde (vgl. slaw. svojĭ ‘eigen’), das unter den Einfluss des Zahlworts geraten ist, wie ich selber vorgeschlagen habe [in Schrijver 1997: 83]. [The h of Welsh hun(an), plur. hun(ein) and Breton hunan has been explained as deriving from the stem of the Indo-European relative pronoun *sw- at least since Morris Jones (1913: 307). Either this was prefixed to the numeral for ‘one’ (thus Morris Jones), or the base is a pronominal form *swoi-no- (cf. Slav. svojĭ ‘own’) which was influenced by the numeral, as I myself have proposed in Schriver 1997: 83.] 11 In Old Irish the concept of ‘self’ is expressed differently from Welsh; cf. section 6 below. 12 An overview is presented in Puddu (2007). 13 Finnish belongs to this group, for which I know no explanation.

Concerncing myself

125

4 The reflexive in Semitic Fortunately we do not have to look very far to find such a language. Several scholars, among them Morris Jones (1900), Pokorny (1927–30), Preusler (1956: 350), Gensler (1993) and myself (2001, 2002a, b), have held the view that the non-Indo-European features of the Insular Celtic languages developed on a Hamito-Semitic substratum. Jongeling (2000) has shown that the syntactic systems of Hebrew and Welsh are very similar, bordering on identity.14 Since Hebrew and Phoenician are closely related North West Semitic, viz. Canaanite, dialects, this similarity should extend to Phoenician and Welsh, more generally: to Phoenician and Insular Celtic. The Isles are known to have been in the scope of Phoenician economic activities; best known are the tin trade with the Tin Islands (Strabo), generally understood as having included Cornwall, and the copper trade with Copper Island, i.e. Ireland (Vennemann 1998; Broderick 2009). Such extended economic contacts are known from history to be able to introduce the language of traders and their trading factories into the surrounding country as a lingua franca, and the simplest assumption concerning the substrate should therefore be that it was a variety of Phoenician the Celtic invaders found in the Isles.15 This assumption is in harmony with the analogous theory concerning the Semiticisms of Germania (cf. Vennemann 2004; Mailhammer 2006, 2007), except that there Phoenician played the role of a superstrate to the pre-Germanic IndoEuropean of the north. Let us therefore look into Semitic, and Phoenician in particular, to see if the Semitic way of expressing reflexivity and intensification may have served as a model for the Celtic of the Isles.16

14 Unfortunately Jongeling does not treat the reflexives, possibly because they are so similar in Hebrew and Welsh, and seemingly uninteresting in an English language perspective. I have only found a single example presented in a totally different context, sentences with expressed pronominal subjects: dibbartī ʔăni ʕim libbī “Spoke-I with heart-my”, ‘I said to myself’ (Jongeling 2000: 94). 15 The original supporters of the theory of a Semitic substrate of Insular Celtic (Morris Jones 1900, Pokorny 1927–1930, also more recently Gensler 1993), were more inclined to assume a northwest African Hamito-Semitic, e.g. Berber, substrate. Two reasons have led me to think of a more recent, namely genuinely Semitic substrate: first, the clearly Semitic, namely Phoenician (Carthaginian) influences in Germanic, and second, the virtual identity of Welsh and Hebrew in syntactic matters demonstrated by Jongeling (2000). 16 With Irish, however, developing its new method of intensification and reflexivization in a formally different manner, cf. section 6 below.

126

Theo Vennemann

Semitic has no genuine (inherited, reconstructable) reflexive pronoun corresponding to German 3rd person sich (cf. Lipiński 2001: §36.28; Rubin 2005: 19). Instead, the usual suffixed pronoun is used even when a reference to the subject is intended: (1)

Reflexive construction in Arabic and Hebrew (Lipiński 2001: §36.28) a. Arabic ʔilā Marwāna fada ʕāhu ʔilay-hi ba ʕata (he)-sent for Marwān summoned-(he)-him to-him ‘he sent for Marwān and summoned him to himself’ b.

Hebrew ʔĒhūd ḥereb way-ya ʕaś l-ō and-(he)-made for-him Ehud sword ‘Ehud made a sword for himself’

(Judg. 3,16)

This is similar to the way of coding reflexivity in Old English where the freestanding personal pronoun was used, leaving it to the hearer or reader to decide whether the pronoun (in 3rd person constructions) referred back to the subject or to some other referent, cf. the examples in Kisbye (1972: §S1–2; Lutz 2002: 46), and in section 7 below. Kisbye explicitly says: The ambiguity caused by the use of the personal pronoun to express reflexivity is evident from cases like he acwealde hine [‘he killed him/himself’], he hine aheng [‘he hanged him/ himself’], heo hie ofslog [‘she slew her/herself’], etc. where it is not immediately obvious whether the meaning ‘he killed him’ or ‘he killed himself’, etc. is intended. . . . The addition of the adjective self already comparatively early in O[ld] E[nglish], however, precludes the possibility of misunderstanding: se cyning Hasterbald hine selfne acwealde, and his wif mid hiere twæm sunum hie selfe forbærnde [‘King Hasdrubal killed himself, and his wife with her two sons burned herself’] (Alfred: Orosius). (Kisbye 1972: §§S1–3, S1–4)

Where Old English here used self for clarity, various nouns are used for the same purpose in the Semitic languages. Thus, after the above two Semitic examples Lipiński continues: However, Semitic languages prefer to employ the noun raman-, ‘self’, or napš-, ‘person’, with the required pronominal suffix. In East Semitic, besides the generally recognized ramanu (e.g. ana ramanišu, ‘for himself’), one finds sometimes another noun, as qaqqadu, ‘head’ (e.g. qaqqassa ana šīmim iddin, ‘she sold herself’), or pagru, ‘body’ (e.g. pagaršu ina šīmim iddin, ‘he sold himself’). A similar use is attested in other Semitic languages, not only with napš- (e.g. Hebrew nišbaʕ bə-napšō, ‘he swore by himself’), but also with nouns meaning ‘head’, ‘belly’. . . , ‘heart’ (e.g. Hebrew ʔāmar bə-libbō, ‘he said to himself’), ‘bone’ (e.g. Syriac ʕal garmah, ‘about herself’, lit. ‘about her bone’; Sabaic grmk, ‘you yourself’). . . .

Concerncing myself

127

Phoenician, Punic, Syriac, and Samaritan Aramaic use also the noun qnūm-, ‘person, being’; e.g. Neo-Punic p ʕ l mqr . . . l-qn ʔm17, ‘Maqer made it for himself’; Syriac ba-qnūmeh, ‘by himself’. (Lipiński 2001: §36.28)18

Rubin (2005: 19–22) lists the individual Semitic languages in which nouns of these meanings – plus words for ‘eye’ (Classical Arabic ʕayn-), ‘owner’ (Tigrinya baʕl-), and ‘state’ (Syrian Arabic ḥāl-) – are used as reflexive markers, and illustrates the use of some of them. He also shows the use of these markers as intensifiers in some of the languages. Phoenician is not mentioned in his account. Probably the most detailed and exemplary description of intensification and reflexivization in a single Semitic language is Goldenberg’s (1991) article on Amharic.19 Needless to say this South Semitic language is outside the contactlinguistic scope of the present paper. Still the fact that Amharic alone offers a slew of nouns of the kind cited by Lipiński for constructing intensification may be of general linguistic interest: ras ‘head’, qəl (of uncertain meaning, perhaps ‘gourd, head’), baläbet ‘house owner’ and, with different syntax, säwənnät ‘person, body’, näfs ‘soul’, hod ‘belly’, ləbb ‘heart’, anǧät ‘intestines’. For reflexivization, ras ‘head’ seems to be the most important noun; all the examples used by Goldenberg (1991: 534–535) to illustrate reflexivization are based on it. The “noun strategy” applied by the Semitic languages to express intensification and reflexivization is also well-known from creoles using nouns meaning ‘body’ and ‘head’, but also others: “Much more rarely, it may be nouns meaning ‘person,’ ‘owner,’ ‘self,’ ‘reflection,’ ‘spirit,’ or body parts such as ‘bone,’ ‘skin,’ or ‘eye’ that may be recruited”. (Heine 2005: 216 note 18). Heine also mentions the rare ‘belly’ (228). However, the situation is not very different for non-creole languages:

17 The Punic 3rd person masculine singular possessive pronoun suffix is -m. Two identical consonants that are not separated by a vowel are written by a single consonant symbol. 18 Prof. Voigt (cf. note 1) has suggested a number of corrections or qualifications for this quotation. East Semitic ramănu(m) is Assyrian; Akkadian has ramānu(m), cf. von Soden (1972: II. s.v.). The Syriac form for ‘about herself’, lit. ‘about her bone’ is ʕal garmåh. As to Sabaic grmk ‘you yourself’, Prof. Voigt writes: “Für das sabäische grm kenne ich keine Belege.” [For Sabaic grm I know of no attestations.] And as to the noun qnūm-, ‘person, being’, he suggests for Syriac the form qnom-, qnum- being Western Syriac; and for Phoenician he writes: “Für das Phönizische kenne ich nur qnmy. Bei Friedrich / Röllig / Amadasi Guzzo findet sich keine andere Form.” [‘For Phoenician I only know qnmy. No other form can be found in Friedrich and Röllig 1999.’] 19 Reference owed to Prof. Edzard (p.c.), cf. note 1.

128

Theo Vennemann

We remain in fact with two main sources used in the noun strategy: ‘body’ and ‘head,’ where the former is clearly more common than the latter. This is exactly also the situation found in non-creoles: ‘body’-reflexives are by far the most common ones in the languages of the world, followed by ‘head’-reflexives, while other nominal sources, such as ‘soul,’ ‘spirit,’ ‘heart,’ ‘bone,’ etc. are far less common (see Schladt, 2000). (Heine 2005: 228)

This use of concrete nouns (body part nouns) also occurs in Welsh, where pen ‘head’ (with lenition ben, with nasal mutation mhen, with aspirate mutation phen) functions in the expression of the idea of ‘by oneself’, ‘on one’s own’: ar ’y mhen ’yn hun(an) ‘on my own’, ar dy °ben dy hun(an) ‘on your own’, ar ei °ben ei hun(an) ‘on his own’, ar ei h phen ei hun(an) ‘on her own’, etc. (King 2003: §133). Germanic +selb- (as in English self ) appears to have a similarly concrete origin; cf. Kluge/Seebold (2002: s.v. selber (selbst)): “Außergermanisch vergleicht sich zunächst venet. sselboisselboi (mit einer Verdoppelung, die auch in ahd. selbselbo bezeugt ist). Weiter zu air. selb, kymr. helw ‘Besitz’ aus (ig.) *sel-wo(vgl. die Verwendung von ↗eigen). Weiter zu ig. (eur.) *sel- ‘ergreifen’ (vgl. ne. sell).” [Outside Germanic, Venetic sselboisselboi may be compared (with a reduplication that is also attested in Old High German selbselbo). Related to Old Irish selb, Cymric helw ‘property’ from (Proto-Indo-European) *sel-wo- (compare the use of eigen ‘own’). Derived from Proto-Indo-European (European) *sel- ‘seize’ (compare Modern English sell).] In Kluge/Seebold (2011: s.v. selber [selbst]), Oettinger (1999) is cited with a reconstruction of +selb- as “reflexive[s] *s(w)e- + *-el als Zugehörigkeitsbildung (vgl. den hethitischen pronominalen Genetiv) + Adjektivableitung -bho-” [reflexive *s(w)e- + *-el as attributive formation (cf. the Hittite pronominal genitive) + adjectival derivation with -b ho-]. It may be doubted, however, that this is an improvement over the earlier reconstruction: +selb- does not by itself have a reflexive meaning. Even though it too struggles with various reconstructions, the Etymologisch Woordenboek van het Nederlands, s.v. zelf, may for the time being offer the best summary of the situation, after presenting the Germanic evidence: “Verdere herkomst onduidelijk. Er zijn in elk geval geen zekere verwante woorden met vergelijkbare betekenis.” [Further origin unclear. In any case there exist no securely related words with a comparable meaning.]. Among the Semitic languages, Phoenician has an intensive and reflexive nominal particle of its own, BT- (bitt-) with its free variant BNT- (binat-). Krahmalkov writes: “The etymology of this pronoun, the basic meaning of which is ‘self, one’s self,’ is obscure. The pronoun is used in Phoenician and Punic with (i) the independent personal pronoun, (ii) the suffixal possessive pronoun and (iii) the anaphoric pronoun to form other pronouns” (Krahmalkov 2001: 116).

Concerncing myself

129

Krahmalkov (2001: 116–117) provides three sets of examples for the three basic functions of BT- (bitt-) / BNT- (binat-)20: (2)

The three basic functions od Punic BT- (bitt-) / BNT- (binat-) a. ‘one, one’s self’, following and complementing an independent pronoun21: W-B ʕL ḤRŠ H ʔ BT-M and-master craftsman he self-his ‘He himself (was) the architect’ b.

‘one’s own’, following and complementing a suffixal possessive pronoun: ʔYT H-Š ʕR Z . . . P ʕLT B-TKLT-Y BNT-Y ACC the-gate this built-I at-expense-my self-my ‘I built this gate at my own expense’ ʔDM MLK BŠR-M BN ʕT-M sacrifice human-being flesh-his self-his

‘the sacrifice of a human being of his own flesh’ ʔDM BŠ ʕR-M BT-M MLK sacrifice human-being flesh-his self-his

‘the sacrifice of a human being of his own flesh’ MZBḤ W-P ʔDY P ʕL LMB-MLKT-M BT-M altar and-podium (he)-built from-property-his self-his ‘He built the altar and the podium at his own expense’ c.

‘that very, that same’, following and complementing an anaphoric pronoun: WYṢ ʔ . . . LʔGD L-M MLḤMT and-(they)-came-forth to-wage with-them battle B-MQM ʔZ B-YM H ʔ BNT-Y at-place this on-day that self-his ‘They came forth to do battle with them at this place on that same day’

20 I have separated the articles, particles, prepositions, and pronouns from their bases by a hyphen for perspicuity. – Friedrich and Röllig (1999: §254.III) treat BT/BNT rather briefly under “Particles”. They show their vocalization bit(t)ē- in the Latino-Punic baiaem bithem (= bithe-m ‘his own’) ‘zu seinen eigenen Lebzeiten’ [‘during his own lifetime’], and perhaps b(i)nat- in bn‘t-m (if ‘Ayin here stands for a, as commonly in Late Punic; cf. also the third of Krahmalkov’s examples below). 21 See also the quotation at the end of this section.

130

Theo Vennemann

For the (b) group Krahmalkov refers to additional examples in Krahmalkov (2001: 63–65). There he (2001: 63) writes under the heading “Expressing the Reflexive Possessive (with BT- or BNT-)”: Phoenician possessed a reflexive possessive pronoun (cf. Latin suus and ipsius) ‘his own’) expressed by the suffixal possessive pronoun followed by the particle BNT (or its free variant BT ) + Form B of the third person suffixal pronoun. The particle BT had the shape bitt-; and BNT the shape binat-. The etymology of the particle is obscure but in its syntax and use it may be properly compared with the Egyptian particle ds in the common Middle Egyptian construction pr.f ds.f ‘his own house’22. . . . The expression ‘at one’s own expense,’ employing the reflexive possessive, is common in the Neo-Punic inscriptions. Neo-Punic normally employs the variant form BT of the particle in this particular expression although it also knows BNT, used in free variation with BT, in other formulaic expressions. . . : B-T ʔR-M BT-M P ʕL W- ʔ YQDŠ [from-money-his self-his (he)-made and-dedicated] ‘He built and dedicated at his own expense.’

As to the use of Phoenician and Punic BT- or BNT- for intensification (cf. his set (a) above), Krahmalkov writes more fully earlier in his book (2001: 47): The intensive independent personal pronoun (“I myself,” Latin ego ipse) is expressed in Phoenician and Punic by the independent person pronoun followed by the particle BT(BNT-) + third person suffix pronoun: N[eo]Pu[nic] 72 B lines 1/4 P ʕ L WNDR WḤDŠ ʔYT T-GZT ST ʕ BD ʔ ŠMN BN ʕZRB ʕL HKHN LRBBTN LTNT ʔDRT WHGD WB ʕ L ḤRŠ H ʔ BTM, ‘Abdesmun son of Hasdrubal the Priest made this gzt for Our Lady Great Tinnit. He himself was the gd (?designer) and architect.’

Here ‘he himself’ translates Neo-Punic H ʔ BT-M, vocalized probably hū bittēm.23 He continues as follows: The particle BT bitt- and its free variant BNT binat- were commonly uesd in Phoenician and Punic after the possessive pronouns to express the reflexive possessives (‘my own,’ ‘your own,’ etc.), and with the anaphoric pronoun to express the emphatic anaphoric pronoun (‘the same,’ ‘the very’). (Krahmalkov 2001: 47–48)

5 Semitic → Welsh → English Forms such as (1) Phoenician bnt-y, (2) Sabaic grm-k, (3) Hebrew napš-ō, Phoenician qn ʔm (+qn ʔm-m24), bt-m, etc.25, despite their mirror-image arrangement,

22 23 24 25

Literally ‘house-his self-his’; see the first example under (b) above. Cf. Krahmalkov (2001: 40) and note 20 above. Cf. note 17 above. But see note 18 above.

Concerncing myself

131

correspond exactly to Welsh fy hun, dy hun, ei hun, etc.26 They consist of a pronominal element indicating possession and a nominal element of various etymologies. The English case is somewhat different, owing to the adjectival origin of the intensifier +selb-, which only became a noun in the course of the contact-induced English development. But with this categorial change, Late Middle English / Early Modern English my self, thy self, his self, etc. also clearly agree with that model. The nuclear part of the English development is concisely described by Brunner: Im Lauf der m[mittel]e[nglischen] Zeit wird self als Substantivum aufgefaßt, doch hält sich der adjektivische Gebrauch daneben. Das führt dazu, daß anstelle des Obliquus der Personalpronomina das Possessivpronomen gesetzt wird und durch self verstärkte Substantiva im Genitiv erscheinen. Der Gebrauch der Possessiva hat sich aber bloß in der 1. und 2. Person gemeinsprachlich durchgesetzt, in der 3. Person kommt zwar hisself und theirself (-selves) me. und frühne. vor, sie sind aber heute nur mehr mundartlich, so wie auch Bildungen in der 1. und 2. Person mit dem Obliquus des Personalpronomens (meself, usselves) in Mittel- und Südengland mundartlich noch erhalten sind. (Brunner 1962: 121–122) [In the course of the Middle English period, self is understood as a noun, even though the adjectival use continues – with the result that the possessive is employed instead of the oblique forms of the personal pronouns, and nouns intensified by self appear in the genitive. But the use of the possessives only became common in the 1st and 2nd persons. In the 3rd person, hisself and theirself (-selves) do occur in Middle and Early Modern English, but nowadays they exist only in dialects, as formations in the 1st and 2nd persons with the oblique forms of personal pronoun (meself, usselves) have been preserved in dialects of Central and Southern England.27]

Summing up, I would like to propose that just as English myself etc. is calqued on Welsh fy hun etc., Welsh fy hun etc. is itself calqued on Phoenician BNT-Y etc., so that the development of the intensifier/reflexive in the Isles is recognized as one further instance of transitive importation, viz. positive importation28, Semitic → Celtic → English (cf. Vennemann 2002b). 26 One of the anonymous referees (cf. note 1 above) adds the following comment: “In each case one has the normal order of possessor and possessum in the language concerned. Speakers shifting language would have had access to this information, indeed even under the extreme low level of access to the target language that characterizes creolization constituent order patterns of the target language usually get through.” 27 Cf. note 5 above. – The seemingly unsystematic use of monosyllabic and disyllabic forms of self (inflected singular as well as plural forms) in Middle English is explained as rhythmically (metrically) motivated in Lutz (2002). 28 Cf. Vennemann (2011) for a synopsis of language contact terminology with special reference to English. “The term importation . . . refers to the transfer of features in language shifting, namely from the language SL of the language shifters to the language TL, the target language

132

Theo Vennemann

6 The reflexive in Irish “[In Old Irish] there are no special reflexive pronouns; any infixed or suffixed pronoun can refer to the person or thing forming the subject of the clause” (Thurneysen 1946: §401). The concept of ‘self’ is expressed, differently from Welsh, with forms of féin/fadéin and céin/cadéin. Thurneysen tabulates all of these forms and writes: ‘-self’ is expressed by a great variety of uninflected forms which, except when they accompany a noun with the article (or a proper name) or form the subject of a clause, are combined with a personal or possessive pronoun. . . . Examples: caraid cesin ‘he himself loves’; da.berid-si féissne ‘ye yourselves give it’; don chrunn fésin ‘to the tree itself’; mé féin ‘I myself’; frinn fanisin or frinn fesine ‘against ourselves’; far m-bráthir fadisin ‘your own brother, uester ipsorum frater’. . . . In Mod. Ir. féin has become the universal form. (Thurneysen 1946: §485)

All of Thurneysen’s examples only show the intensifying use of féin etc. For Modern Irish, an emphatic and a reflexive usage are distinguished. Thus, the Wikipedia sketch of Irish grammar says: The word féin (/f j eːn j/ or /heːnj/) ‘-self’ can follow a pronoun, either to add emphasis or to form a reflexive pronoun. Rinne mé féin é. ‘I did it myself.’ Ar ghortaigh tú thú féin? ‘Did you hurt yourself?’29

Here féin is added to the personal pronouns, mé and tú (thú), in both usages. As to the analysis of the complex Old Irish expressions for ‘-self’, Lewis and Pedersen write: ‘Self’ in [Old] Ir[ish] . . . was expressed by a word consisting of a form of the verb ‘to be’ ( f-, fad-) or cia, cid ‘although’ (reduced to c-, cad-) + a personal pron[oun] as predicate + a demonstrative as subject. There have however been many modifications and readjustments. The forms which occur are as follows (the component parts being hyphenated in those cases where the composition is etymologically regular): sg. 1. féin, fadéin, céin[,] cadéin; 2. féin, fadéin; 3. m. fad-e-sin, f-e-ssin, f-e-in, cad-e-ssin, c-e-sin; 3. f. fad-i-sin, f-issin. . . , fesine, féisne; plur. 1. fa-ni-sin, ca-ni-sin, fesine; 2. fad-i-sin, (with analogical lenition of s- of si ‘you’ to h > nil, instead of to f > ƀ), féisne, fadéisne; 3. fad-e-sin, fad-e-sine, fad-é-sne, f-e-sin, f-e-sine, f-ei-ssne, cad-é-sin, cad-e-sin, cad-e-sne. – In M[oder]n Ir[ish] the form for all persons is féin. (Lewis and Pedersen 1989: §346.1)30 the language shifters are trying to shift to” (p. 221). “The term positive importation refers to those cases in which the shifters carry properties of their SL into their ILs [interlanguages] that do not exist, or exist differently, in TL” (p. 223). 29 Cf. “Irish grammar”, http://en.wikipedia.org/wiki/Irish_grammar#Possessive_pronouns (29 July 2011). 30 The account translates the original of Pedersen ([1913] 1976: §490).

Concerncing myself

133

This is neither an inherited method nor one for which I have seen a possible source in any conceivable prehistoric contact language. If correct, the fact that the complex syntactic structure is still transparent in the univerbated attested forms makes it appear to be a fairly late pre-Old Irish, and strictly Irish, innovation, which sheds no light on the history of the Welsh expression of ‘-self’. It does not seem to be correct, however; for Peter Schrijver writes in reference to the above quotation: Die von Ihnen zitierte Herleitung des altirischen féin ‘selbst, eigen’, usw., die auf Pedersen zurückgeht, wird wohl allgemein angezweifelt, sicherlich seit Greene [1969], der die Form im Titel [d.i. fa-deisin] auf *swe-deksini ‘with one’s own right hand’ zurückführt. Das ist sicherlich falsch. Ich habe im genannten Buch [Schrijver 1997: 72–83] ausführlich über Etymologie und Verwendung der Formen des äußerst komplizierten Paradigmas geschrieben. Meines Erachtens ist die Grundstruktur: *swe-sin(-) (>air. féin ‘selbst’). In der Mitte können die Formen des betreffenden Personalpronomens (z.B. *swe-en-sin(-) ‘er/ihm, usw., selbst’ > air. feissin) und die anaphorische Partikel *de erscheinen (z.B. *swe-de-en-sin(-) ‘er/ihm, usw., selbst’ > air. fadeissin). Diese Formen haben vermutlich eine etymologische Beziehung zu den mittelwalisischen sogenannten ‘conjunctive pronouns’ des Typs ynteu ‘er aber, er seinerseits, der andere, sogar er, usw.’ < *en-de-suwe (Schrijver 1997: 83–90).31 [The derivation of Old Irish féin ‘-self, own’ etc. which you cite and which goes back to Pedersen, is probably doubted by everyone, and certainly so since Greene (1969), who traces the form in the title of his note [viz. fa-deisin] back to *swe-deksini ‘with one’s own right hand’. That is definitely incorrect. In my book (Schrijver 1997: 72–83) I have written in detail about the etymology and the use of the forms of the extremely complicated paradigm. In my view the basic structure is *swe-sin(-) (> OIr. féin ‘-self’). Inside this structure the forms of the appropriate personal pronoun (e.g. *swe-en-sin(-) ‘he/him, etc., -self’ > air. feissin) and the anaphoric particle *de may appear (e.g. *swe-de-en-sin(-) ‘he/ him, etc., -self’ > OIr. fadeissin). It would seem that these forms are etymologically related to the Middle Welsh so-called “conjunctive pronouns” of the type ynteu ‘he however, he for his part, the other, even he (etc.)’ < *en-de-suwe (Schrijver 1997: 83–90).]

It seems that the use of féin and the related forms started out as a kind of optional clarification. That would make the Irish development similar to the English one. In Peter Schrijver’s words: Das Altirische benutzt zum Ausdruck des Reflexivums die ingwäonische Struktur: ein Objektspronomen im Akkusativ: at-cí ‘er sieht ihn = er sieht sich’. Die reflexive Funktion kann durch Hinzufügung des Pronomens féin, fadein betont werden: condid tánicc fessin ‘until he has reached himself’ (Würzburger Glossen, 12b34a). That is very much like English, of course.32

31 Prof. Schrijver in his letter of 20 January 2011 (cf. note 1 above). 32 Prof. Schrijver in his letter of 20 January 2011 (cf. note 1 above).

134

Theo Vennemann

[Old Irish uses the Invaeonic structure for expressing the reflexive: an object pronoun in the accusative: at-cí ‘he sees him = he sees himself’. The reflexive function can be stressed by adding the pronoun féin, fadein: condid tánicc fessin ‘until he has reached himself’ (Würzburg Glosses, 12b34a). That is very much like English, of course.]

Where Irish seems to differ from the Welsh and the English development is that no nominalization of féin, fadein with concomitant use of a possessive pronoun of the type fy hun, myself to form the reflexive pronoun has occurred.

7 Why did Old English and Insular Celtic not have a reflexive pronoun? One defining criterion for the “Ingvaeonic” subgroup of West Germanic, comprising English, Frisian, and, to some extent, Old Saxon and Old Low Franconian, is the loss of the set of reflexive pronouns in prehistoric times, with only the old genitive sīn surviving as a possessive of limited occurrence.33 The reasons for these divergent developments seem to be unknown. Campbell wrote in 1959: It is a peculiarity of the Ingvaeonic languages, distinguishing them from O[ld] H[igh] G[erman], that the pronoun of the third person is also used as the reflexive and reciprocal pronoun of the third person, e.g. O[ld] E[nglish] ða beþohte he hine ‘then he bethought himself’ [lit. ‘. . . him’], hie . . . hie ġemetton ‘they met one another’. (Campbell 1959: §704)34

If I am not mistaken, it is still considered as peculiar as it was in 1959. I have been unable to find a single attempt to deal with this particular feature in the literature, or indeed anything suggesting an awareness of this not just as a peculiarity but as something in need of an explanation. As shown in sections 1 and 6 above, not only did Old English lack a reflexive pronoun but so did Welsh and Irish; they all used the same circumlocutions that were also, and originally, used as intensifiers. The inherited Indo-European pronoun35 had obviouly fallen out of use in its reflexive function. Since the explanation for the new English intensifying and reflexive pronoun given by Preusler, 33 Cf. notes 4 and 9 above as well as the Appendix. 34 Whereas English has developed its new reflexive pronoun, the Continental Ingvaeonic dialects have either borrowed the old reflexive pronoun from dialects that preserved them, e.g. from High German (cf. the Appendix below), or continued the Ingvaeonic state of affairs. Thus, Modern Frisian not only uses the personal pronoun in reflexive use in the 1st and 2nd person but also in the 3rd person: West Frisian Ik skamje my ‘I am ashamed of myself’, Hy wasket him ‘He washes himself’ (Sipma 1913: §§226–227, 234). 35 Dat. +soi ̯ / +sṷoi ̯, acc. +se / +sṷe, cf. Meier-Brügger (2000: §F401.3).

Concerncing myself

135

Vezzosi, and others is substratal language contact with Welsh, the assumption within this context for the loss of the inherited reflexive pronoun in English likewise has to be substratal influence of Celtic: The Celts learning Anglo-Saxon, not having a category of reflexive pronoun in their own language, failed to acquire the reflexive pronoun of their target language, with the natural consequence that their resulting interlanguage36 did not have a reflexive pronoun either. This answer can be generalized to Ingvaeonic as a whole: Since there are other reasons for assuming a Celtic substratum not only for the Isles but for the opposing coastal regions as well (Schrijver 1999; Vennemann 2009), the proposal just made for Old English may be extended to Frisian and Old Saxon, and also to Old Low Franconian if the assumption is correct that the reason why this dialect in its eastern regions borrowed its reflexive pronoun from neighboring Old High German is that it had lost the inherited Germanic pronoun.37 That leaves us with the question of how Celtic itself lost the reflexive pronoun. The best answer, as always within the theory of language contact, would be one based on the same model as that used for Ingvaeonic in relation to Celtic, in this case: pre-Celtic substratal influence from languages likewise lacking a reflexive pronoun. As shown in section 4, the West Semitic languages were of this kind: They all lacked a reflexive pronoun but used the same strategy based on an intensifier accompanied by a possessive/personal pronoun as Celtic and Old English. Summing up, I would like to propose that just as the English lack of a reflexive pronoun is a consequence of the same property of the Insular Celtic substratum, a case of negative importation38, the Celtic lack of a reflexive pronoun is itself a consequence of the same property of the Semitic substratum of the Isles, likewise a case of negative importation. This iterated contact influence is thus yet a further instance of transitive importation, Semitic → Insular Celtic → English or, more generally, Semitic → North Sea Celtic → Ingvaeonic.

36 Cf. Vennemann (2011: 221): “In trying to acquire TL [the target language] the shifters use their previous experience with their native SL [the shifters’ language] to express themselves with means provided by TL. The result is an interlanguage IL ‘between’ SL and TL, namely a language which is like TL except for more or less systematic deviations stemming from the shifters’ unintentionally applying their native linguistic habits to the language they are trying to speak. . . . Ideally, TL will eventually be reached by the shifting population. But it often happens that TL will never quite be reached. In that case, the ILs may stabilize somewhere between SL and TL, a stage we may call the resulting interlanguage, RIL.” Cf. also note 28 above. 37 Cf. note 4 above and the Appendix. 38 “The term negative importation refers to those cases in which the [language] shifters fail to acquire a feature of TL [the target language], so that it is lacking from their ILs [interlanguages] and from RIL [the resulting interlanguage]” (Vennemann 2011: 223).

136

Theo Vennemann

8 Myself and reflexivization typology The changes analysed above can be succinctly captured in terms of the following reflexivization typology proposed in Heine (2005: 206–214); it is there used for the typological description of creoles but is well suited also for non-creole languages: – Type A, i.e. systems where the personal pronouns of all three persons are used to express reflexivity; – Type B, i.e. systems with the personal pronoun functioning as reflexive only in the 1st and 2nd person but with the 3rd person personal pronouns being “anti-reflexive”, a special reflexive stepping in to express reflexivity; with subtype BR, possessing a reflexive marker R that is used exclusively as a third person reflexive marker (comparable to French se, German sich); – Type AI, i.e. systems that are like Type A but with an intensifier (I) which is added to the personal pronoun to indicate that that pronoun is coreferential with the subject; – Type PI, i.e. systems where the intensifier (I) may be preceded by a possessive attribute (P) coreferential with the subject; – Type PN, i.e. systems where the reflexive marker has the structure of a possessive noun phrase, consisting of a head noun (N) and a possessive attribute (P) coreferential with the subject. One type which is of obvious significance in the decription of the systems treated above is not mentioned in Heine (2005), probably because it did not occur in Heine’s sample of creole languages: – Type C, i.e. systems in which all personal pronouns are “anti-reflexive”, special reflexives stepping in to express reflexivity; with subtype CR, possessing special reflexive markers R for the different persons and numbers. The Celtic development began with a change from the Proto-Indo-European Type B, subtype BR, to Type A, explained above as occurring on a Semitic substrate lacking a reflexive pronoun. Then Irish moved on to Type AI, Welsh to a variant of Type PN where the element hun, a nominal element without lexical meaning, serves as N39; this change is likewise explained above, at least in the case of Welsh, as occurring on a Semitic substrate which was partially of Type 39 Heine’s criteria for an element to count as a noun in this context are weak enough to cover both Welsh hun and Punic bitt-/binat-: “The label ‘noun’ is used for units that are either unambiguously nouns or else have a number of nominal properties, such as taking possessive modifiers and/or having uses as lexical nouns in other contexts” (Heine 2005: 206).

Concerncing myself

137

PN, namely, if the substrate is identified with Punic, of a variant whose N, bitt-/ binat-, was also a nominal element without lexical meaning.40 The English development began with a change from the Proto-Germanic Type B, subtype BR, to the Ingvaeonic Type A, explained above as occurring on a Celtic substrate lacking a reflexive pronoun. Then English, after a period of moving in the direction of type AI (with I = self ), nominalized its intensifier and became of Type PN (N = self ), its standard varieties winding up with a hybrid system mixing the innovated PN features with the older AI features (or rather AN features, after the change of self into a noun). By further grammaticalization of the AI/PN reflexive markers, standard varieties of English eventually became of Type C, subtype CR, with the special feature that the R elements are at the same time used as intensifiers (I washed myself / I myself did it / I did it myself, etc.). English has thus run the gamut from Type B to Type A and on to Type C, and this in the short time of two and a half thousand years, a time during which High German has stayed solidly within type B, i.e., did not change its reflexivization type at all. Since the Germanic reflexivatization type is reconstructed as the same as that of Proto-Indo-European, German even has a reconstructed reflexivization history of at least five thousand years without a single change of type.41 The same holds true for the Romance languages. Since I do not know of any systematic investigation into the question of how and at what rates languages change their ways of expressing reflexivity, I cannot objectively determine which of the two developments is typical, or more natural: the English volatility or the German stability; but my guess would be stability. If true, then the very rapidity of change of reflexivization type that English has gone through would suggest that it has not experienced natural, internally motivated change but change caused by language contact. This then would prove the reflexivization history of English to be in harmony with other aspects of the language’s grammatical history. E.g., the English history of word order is likewise the result of radical change, in this case from the Proto-Indo-European SOV type to the Germanic TVX type and on to the modern SVO type, English in fact being the only Germanic language with SVO syntax42 – changes which in such short periods of time can be made likely only to be possible by language contact (cf. Vennemann in prep.). 40 More precisely: without known lexical meaning. 41 High German lost the Germanic dative +siz between Proto-Germanic and Old High German and generalized the accusative +sik in Early New High German (see note 9 above); but all of this did not move the language away from Type B. 42 S: subject, V: verb, O: object, T: topic, X: further specifiers of the verb. The Germanic TVX type is characterized by “split word order”, i.e. verb-early position at the sentence level but verb-late position in subordinate structures.

138

Theo Vennemann

9 Summary and conclusion This paper addresses four questions: (1) Why did English develop a new method of expressing reflexivization by utilizing an intensifier construction (based on nominalized self )? (2) Why did Insular Celtic develop new methods of expressing reflexivization by utilizing intensifier constructions? (3) Why did Old English – more generally: Ingvaeonic – not possess the Indo-European reflexive pronoun? (4) Why did Insular Celtic not possess the Indo-European reflexive pronoun (in its reflexive function)? Only question (1) has been answered in the literature: English developed its method of expressing reflexivization by utilizing an intensifier construction (based on nominalized self ) on its Celtic substratum by positive importation, cf. section 1 above. Answers to questions (2) to (4) are given in this paper. Ad (2): Insular Celtic developed its methods of expressing reflexivization by utilizing intensifier constructions on its Semitic substratum, likewise by positive importation, cf. sections 3 to 5 above. Ad (3): Old English lost the Germanic reflexive pronoun through negative importation on its Insular Celtic substratum; more generally: Ingvaeonic – i.e. Old English, Old Frisian, Old Saxon, and Old Low Franconian43 – lost the Indo-European reflexive pronoun through negative importation on its North Sea Celtic substratum, cf. section 7. Ad (4): Insular Celtic – more generally: North Sea Celtic – lost the Indo-European reflexive pronoun (in its reflexive function) on its Semitic substratum, the Semitic languages not possessing a reflexive pronoun but only a strategy of reflexivization using constructions of intensification, cf. section 4.44 Finally, in section 8, these developments are described succinctly within Heine’s reflexivization typology. By way of conclusion we may say that the features of English treated here, which are “exceptional” in a purely Germanic and Indo-European perspective (cf. section 3), are natural results of language change within a contact scenario that has been developed on independent grounds since the beginning of the twentieth century, a scenario where North Sea Celtic has developed on a Semitic substratum and Ingvaeonic including English on a North Sea Celtic substratum, a scenario we may flag with the transitive formula Semitic → Insular Celtic → English, more generally Semitic → North Sea Celtic → Ingvaeonic. 43 For Old Saxon and Old Low Franconian see the Appendix below. 44 If I am not mistaken, Pokorny (1927–1930), the most comprehensive study of Semiticisms in Insular Celtic, does not treat reflexivization. The loss of a proper reflexive pronoun and the development of a new reflexivization strategy based on intensification may therefore be added to the list of 64 Semiticisms I extracted from Pokorny (1927–1930) in Vennemann (2002b: 323– 326) [= Vennemann forthc.: 208–211], the “Pokorny List”.

Concerncing myself

139

Appendix: Ingvaeonic and the reflexive It is customary only to include Old English and Old Frisian fully among the Ingvaeonic, or North Sea Germanic, languages but to relativize judgements in the case of Old Saxon and Old Franconian.45 However, judging by the criterion ‘loss of the Germanic reflexive pronoun’ alone, Old Saxon and Old Franconian may be recognized as totally Ingvaeonic, the historical occurrence of the Germanic reflexive pronoun there being without any doubt attributable to language contact. In the case of Low German there is some debate among specialists about the source of the Middle Low German reflexive pronoun sik but not about the fact that its occurrence at this stage of the development of the language is an innovation owed to language contact: It could not be different in view of ‘the complete absence of secure traces of the old reflexiv pronoun in the Old Saxon written tradition’46. As to the sih in the Essen Gospel Glosses, it clearly is taken over from High German.47 Rauch writes: O[ld S[axon] clearly belongs to the Ingvaeonic wing of the West Germanic dialects. . . . The relationship of these three dialects [i.e. Old Saxon, Old Frisian, and Old English] may have been one of ancient adstrata on the northwestern coast of the continent. (Rauch 1992: 108)

As for Old Low Franconian, matters seem to be less clear-cut, as shown by the fact that Rauch does not even mention it in this connection. But a closer look at the reflexive may show that this lack of clarity may only be apparent. Robinson (1992: 257) uses the reflexive pronoun criterion to identify Old English, Old Frisian, and Old Saxon as Ingvaeonic, but he repeatedly (pp. 214, 257) draws attention to the fact that the Old Low Franconian reflexive pronoun sich was taken over from Old High German; he cites psalm 65 (cf. 66 of the King James Bible) of the East Low Franconian Wachtendonck Psalm fragments, thia uuitherstridunt, ne uuerthint irhauan an sig seluan.48 The second Old Franconian text, the Leiden Willeram, also has the reflexive pronoun. Quak and van der Horst 45 Cf. note 4 above. 46 “Das vollständige Fehlen sicherer Spuren des alten Reflexivpronomens in der altsächsischen Tradition” (Krogh 1996: 325). Also Rauch (1992: 191) writes, “[In Old Saxon] the reflexive pronoun [is] isomorphic with the dative and accusative of sets (1) and (2) [i.e. the personal pronouns of the 1st and 2nd person and of the 3rd person, respectively]. The lack of independent reflexive morphology is shared with O[ld] E[nglish] and O[ld] F[risian].” 47 “Sih in den Essener Evangelienglossen ist hochdeutsch” (Krogh 1996: 325). 48 Robinson (1992: 206, 209) suggests that the accusative form is here used in the function of dative plural. But in Old High German itself, sih was only used in the accusative function; in the dative function the personal pronoun was used. Cf. Braune/Reifenstein 2004: §282 note 1, and note 9 above.

140

Theo Vennemann

(2002: 44) consider the sich of both texts as deriving from the High German prototype; Cor van Bree (p.c.)49 considers this only necessary for the Leiden Willeram, which stems from Egmond Abbey in North Holland, a typically Ingvaeonic, even Frisian region, but not for the Wachtendonck Psalms, whose reflexive pronoun may either derive from the High German prototype or from the local Limburg dialect which borders on High German territory. Middle Dutch did not have a separate reflexive pronoun, except for sich, sik in eastern texts and (rare) sick in north-eastern texts. Modern Standard Dutch sich was borrowed from High German in the 16th century (van der Wal and van Bree 2008: 214–215). For the dialect situation around the middle of the 20th century cf. Ureland (1979). – In any event, adopting the reflexive pronoun from a neighboring dialect makes it appear likely that the receiving dialects did not have a reflexive pronoun of their own. Hence, by the criterion ‘loss of the Germanic reflexive pronoun’, Old Low Franconian belongs with Ingvaeonic. Other Ingvaeonisms found in the Wachtendonck Psalms, among them the loss of nasals before fricatives and the merger of dative and accusative singular of the 1st and 2nd person of the personal pronoun (only mi, thi), suggest that Limburg and the Lower Rhineland too originally were Ingvaeonic territory (Cor van Bree, p.c.50). That Old Saxon and Old Low Franconian may have started out as fully Ingvaeonic languages that were later in part “de-Ingvaeonized” by contact with High German is, to be sure, not a new idea: In order to delimit the extent of the Ingvaeonic/North Sea Germanic speech area . . . it is important to examine the relationship of Old Saxon to Old Frisian and Old High German in close detail. Many scholars, among them Nils Århammer [1990], think that Low German and Low Franconian were deingvaeonicized and that therefore the numerous parallels shared exclusively by Old English and Old Frisian give a somewhat distorted picture of the dialectal situation in Northern Germany and the Netherlands at the time of the AngloSaxon emigration. (Nielsen 1989: 148)51

Nevertheless, with this larger picture of Ingvaeonic, an old fourfold partition of Germanic becomes recognizable: East Germanic or Gothic, separating from the remaining Germanic – Northwest Germanic – at a time shortly after the Proto-Germanic period; North Germanic or Scandinavian; North Sea Germanic or Ingvaeonic (Old English, Old Frisian, Old Saxon, and Old Low Franconian), and 49 Cf. note 1 above. 50 Cf. note 1 above. 51 Cf. also Nielsen (1989: 103 note 5): “There has been some controversy as to the Ingvaeonic character of Old Saxon, but since Rooth’s investigation of the Merseburg texts most scholars have tended to believe that Old Saxon became deingvaeonicized only as a result of massive High German (Franconian) influence” [with references].

Concerncing myself

141

Elbe Germanic or High Germanic (the Old High German dialects and Lombardic, both differentially Low Germanized in more recent historical events by Low Franconian and Low Saxon influence, respectively52). The question of a still earlier grouping of Ingvaeonic with Scandinavian or with High Germanic is a very controversial issue and too complex to be discussed here.53

References Århammer, Nils. 1990. Friesisch und Sächsisch: Zur Problematik ihrer gegenseitigen Abgrenzung im Früh- und Hochmittelalter. Amsterdamer Beiträge zur älteren Germanistik 31/32. 1–25. Behaghel, Otto. 1923. Deutsche Syntax: Eine geschichtliche Darstellung. Vol. 1: Die Wortklassen und Wortformen (Germanische Bibliothek, I. Sammlung germanischer Elementar- und Handbücher, 1. Reihe: Grammatiken 10). Heidelberg: Carl Winter. Braune, Wilhelm. 2004. Althochdeutsche Grammatik I: Laut und Formenlehre (Sammlung kurzer Grammatiken germanischer Dialekte, A. Haupreihe 5/1). 15th ed. Ingo Reiffenstein. Tübingen: Max Niemeyer. Broderick, George. 2009. The names for Britain and Ireland revisited. Beiträge zur Namenforschung 44. 151–172. Brunner, Karl. 1962. Die englische Sprache: Ihre geschichtliche Entwicklung. Vol. 2: Die Flexionsformen und ihre Verwendung (Sammlung kurzer Grammatiken germanischer Dialekte, B. Ergänzungsreihe 6). 2nd rev. ed. Tübingen: Max Niemeyer. Campbell, Alistair. 1959. Old English grammar. Oxford: Clarendon Press. Edzard, Lutz. 2011. Biblical Hebrew. In: Stefan Weninger, Geoffrey Kahn, Michael P. Streck and Janet C E. Watson (eds.), The Semitic languages: An international handbook (Handbooks of Linguistics and Communication Sciences 36), 480–514. Berlin: De Gruyter Mouton.

52 Cf. Vennemann (2008) for the partial Zurückdrängung (‘repression’) of the results of the High Germanic Consonant Shift in High German under Old Low Franconian, and in Lombardic under Old Saxon, influence. 53 Cf. Nielsen (1989: chapter IV) for numerous attempts at grouping the early Germanic languages, none, as far as I can see, quite the same as the simple division identified above. As far as the major branches go, this division is most similar to the oldest one presented by Nielsen (1989: 67), that of Rasmus ([1818] 1999: 63–72), who assumes the following four branches: Mœso-Gothic, High German, Saxon (with Low German, Dutch, Frisian, and AngloSaxon), and Nordic (with Icelandic, Faroese, Norwegian, Danish, Swedish, and Dalecarlian [a group of dialects in Dalecarlia, Sweden]). His four branches, Mœso-Gothic, High German, Saxon, and Nordic, may in essence be equated with the above East Germanic, Elbe Germanic, North Sea Germanic, and North Germanic, respectively. The way Rask groups his four branches is, however, quite different and, from a modern point of view, clearly wrong: “[He] combined High German and Mœso-Gothic into a German group which together with [the] Saxon group . . . constituted the Germanic branch of the Gothic language family, whose other member was [the] Nordic (Scandinavian) branch” (Nielsen 1989: 67).

142

Theo Vennemann

Einenkel, Eugen. 1916. Geschichte der englischen Sprache II: Historische Syntax. 3rd ed. Straßburg: Karl J. Trübner. Evans, D. Simon. 1970. [1964] A Grammar of Middle Welsh (Medieval and Modern Welsh Series, Supplementary volume.) Dublin: The Dublin Institute for Advanced Studies. Filppula, Markku, Juhani Klemola & Heli Paulasto. 2008. English and Celtic in contact. New York: Routledge. Friedrich, Johannes & Wolfgang Röllig. 1999. Phönizisch-punische Grammatik (Analecta Orientalia 55). 3rd ed. rev. by Maria Giulia Amadasi Guzzo in co-operation with Werner R. Mayer. Rome: Editrice Pontificio Istituto Biblico. Gast, Volker & Peter Siemund. 2006. Rethinking the relationship between SELF-intensifiers and reflexives. Linguistics 44. 343–381. Gensler, Orin David. 1993. A typological evaluation of Celtic/Hamito-Semitic syntactic parallels, Berkeley: University of California, Berkeley dissertation. (Available from University Microfilms International, Ann Arbor, Michigan, no. 9407967). Goldenberg, Gideon. 1991. ‘Oneself’, ‘one’s own’ and ‘one another’ in Amharic. In: Alan S. Kaye (ed.), Semitic studies: In honor of Wolf Leslau: On the occasion of his eighty-fifth birthday, November 14th, 1991. Vol. 1, 531–549. Wiesbaden: Otto Harrassowitz. Greene, David. 1969. Varia ii.4: fa-deisin ‘himself’. Ériu 21. 93–94. Haspelmath, Martin. 2001. The European linguistic area: Standard Average European. In: Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible (eds.), Language typology and language universals: An international handbook (Handbooks of Linguistics and Communication Sciences 20.1). Vol. 2, 1492–1510. Berlin: Walter de Gruyter. Heine, Bernd. 2005. On reflexive forms in creoles. Lingua 115. 201–257. Jenni, Ernst. 1981. Lehrbuch der hebräischen Sprache des Alten Testaments. 2nd ed. Basel: Helbing & Lichtenhahn. Jespersen, Otto. 1940. A modern English grammar on historical principles, Vol. 2/1. London: George Allan & Unwin. Jones, Morris & Alan R. Thomas. 1977. The Welsh language: Studies in its syntax and semantics. Cardiff: University of Wales Press. Jongeling, Karel. 2000. Comparing Welsh and Hebrew (CNWS Publications 81). Leiden: Research School of Asian, African & Amerindian Studies, University of Leiden. King, Gareth. 2003. Modern Welsh: A comprehensive grammar. 2nd ed. London: Routledge. Kisbye, Torben. 1972. An historical outline of English syntax. Vol. 2. Aarhus: Akademisk Boghandel. Kluge, Friedrich. 2002. Etymologisches Wörterbuch der deutschen Sprache. 24th ed. by Elmar Seebold. Berlin: Walter de Gruyter. Kluge, Friedrich. 2011. Etymologisches Wörterbuch der deutschen Sprache. 25th ed. by Elmar Seebold. Berlin: Walter de Gruyter. König, Ekkehard. 2001. Intensifiers and reflexive pronouns. In: Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible (eds.), Language typology and language universals: An international handbook (Handbooks of Linguistics and Communication Sciences 20.1). Vol. 2, 747–760. Berlin: Walter de Gruyter. König, Ekkehard & Peter Siemund. 2000. Intensifiers and reflexives: A typological perspective. In: Zygmunt Frajzyngier & Traci S. Curl (eds.), Reflexives: Forms and functions (Typological Studies in Language 40), 41–74. Amsterdam: John Benjamins. König, Ekkehard & Letizia Vezzosi. 2007. Possessive adjectives as a source of intensifiers. In: María José López-Couso & Elena Seoane (eds.), New Reflections on Grammaticalization 3, 183–206. Amsterdan/Philadelphia: John Benjamins.

Concerncing myself

143

Krahe, Hans. 1969. Germanische Sprachwissenschaft II: Formenlehre (Sammlung Göschen 2233). Berlin: Walter de Gruyter. Krahmalkov, Charles R. 2000. Phoenician-Punic dictionary (Studia Phoenicia 15). Leuven: Peeters. Krahmalkov, Charles R. 2001. A Phoenician-Punic grammar (Handbook of Oriental Studies, Section one: The Near and Middle East 54). Leiden: Brill. Krogh, Steffen. 1996. Die Stellung des Altsächsischen im Rahmen der germanischen Sprachen (Studien zum Althochdeutschen 29). Göttingen: Vandenhoeck & Ruprecht. Lange, Claudia. 2005. Reflexivity and intensification in Irish English and other New Englishes. In: Hildegard L. C. Tristram (ed.), The Celtic Englishes IV, 259–282. Potsdam: Potsdam University Press. Lewis, Henry & Holger Pedersen. 1989. A concise comparative Celtic grammar, 3rd ed. 2nd impression with the supplement of 1961 by Henry Lewis. Göttingen: Vandenhoeck & Ruprecht. Lipiński, Edward. 2001. Semitic languages: Outline of a comparative grammar. 2nd ed. (Orientalia Lovaniensia analecta 80). Leuven: Peeters. Lutz, Angelika. 2002. The reflexive in Middle English: Loose ends in the pronominal system. In: Peter J. Lucas & Angela M. Lucas (eds.), Middle English from tongue to text: Selected papers from the Third International Conference on Middle English: Language and Text, held at Dublin, Ireland, 1–4 July 1999 (Studies in English Medieval Language and Literature 4), 45–68. Bern: Peter Lang. Lutz, Angelika. 2009. Celtic influence on Old English and West Germanic. English Language and Linguistics 13. 227–249. Lyutikova, Ekaterina. 2000. Reflexives and emphasis in Tsakur (Nakh-Dagestanian). In: Zygmunt Frajzyngier & Traci S. Curl (eds.), Reflexives: Forms and functions (Typological Studies in Language 40), 227–255. Amsterdam: John Benjamins. Lyutikova, Ekaterina. Ms. Compound reflexives: Morphology reflecting semantics. http://www. univie.ac.at/linguistics/tagungen/morphologie/abstracts/lyutikova.pdf (accessed 5 January 2011). Mailhammer, Robert. 2006. On the origin of the Germanic strong verb system. Sprachwissenschaft 31. 1–52. Mailhammer, Robert. 2007. The Germanic strong verbs: Foundations and development of a new system (Trends in Linguistics. Studies and Monographs 183). Berlin: Mouton de Gruyter. Maurer, Friedrich. 1952. Nordgermanen und Alemannen: Studien zur germanischen und frühdeutschen Sprachgeschichte, Stammes- und Volkskunde (Bibliotheka Germanica 3). 3rd rev. and expanded ed. Bern: A. Francke. Meier-Brügger, Michael. 2000. Indogermanische Sprachwissenschaft (de Gruyter Studienbuch). 7th ed. Berlin: Walter de Gruyter. Morris Jones, John. 1900. Pre-Aryan syntax in Insular Celtic. In: John Rhys & David BrynmorJones, The Welsh people, 617–641. London: T. Fisher Unwin. Morris Jones, John. 1913. A Welsh grammar, historical and comparative: Phonology and accidence. Oxford: Clarendon Press. Nielsen, Hans Frede. 1989. The Germanic languages: Origins and early dialectal interrelations. Tuscaloosa: University of Alabama Press. Oettinger, Norbert. 1999. Zum nordwest-indogermanischen Lexikon (mit einer Bemerkung zum hethitischen Genitiv auf -l). In: Peter Anreiter & Erzsébet Jerem (eds.), Studia Celtica et Indogermanica: Festschrift für Wolfgang Meid zum 70. Geburtstag (Archaeolingua: Main series 10), 261–267. Budapest: Archaeolingua Alapítvány.

144

Theo Vennemann

Pedersen, Holger. 1976. [1909, 1913]. Vergleichende Grammatik der keltischen Sprachen. Vol.: 1: Einleitung und Lautlehre. Vol. 2: Bedeutungslehre (Wortlehre). Göttingen: Vandenhoeck & Ruprecht. Philippa, Marlies, Frans Debrabandere, Arend Quak & Tanneke Schoonheim (eds.). 2009. Etymologisch Woordenboek van het Nederlands. Vol. 4 (S–Z ). Amsterdam: Amsterdam University Press. Pokorny, Julius. 1927–1930. Das nicht-indogermanische Substrat im Irischen. Zeitschrift für celtische Philologie 16. 95–144, 231–266, 363–394, 17. 373–388, 18. 233–248. Poppe, Erich. 2009. Standard Average European and the Celticity of English intensifiers and reflexives: Some considerations and implications. English Language and Linguistics 13. 251–266. Preusler, Walther. 1938. Keltischer Einfluß im Englischen. Indogermanische Forschungen 56. 178–191. Preusler, Walther. 1956. Keltischer Einfluß im Englischen. Revue des Langues Vivantes 22. 322– 350. Puddu, Nicoletta. 2007. Typology and historical linguistics: Some remarks on reflexives in ancient IE languages. In: Matti Miestamo & Bernhard Wälchli (eds.), New challenges in typology: Broadening the horizons and redefining the foundations (Trends in Linguistics: Studies and Monographs 189), 249–268. Berlin: Mouton de Gruyter. Quak, Arend & J. M. van der Horst. 2002. Inleiding Oudnederlands (Ancorae: Steunpunten van studie en onderwijs 16). Leuven: Universitaire Pers Leuven. Rask, Rasmus K. 1999. [1818]. Undersøgelse om det gamle Nordiske eller Islandske Sprogs Oprindelse: Et af det Kongelige Danske Videnskabets-Selskab kronet Prisskrift. London: Routledge. Rauch, Irmengard. 1992. The Old Saxon language: Grammar, epic narrative, linguistic interference (Berkeley Models of Grammars 1). New York: Peter Lang. Robinson, Orrin W. 1992. Old English and its closest relatives: A survey of the earliest Germanic languages. Stanford, CA: Stanford University Press. Röllig, Wolfgang. 2011. Phoenician and Punic. In: Stefan Weninger, Geoffrey Kahn, Michael P. Streck and Janet C. E. Watson (eds.), The Semitic languages: An international handbook (Handbooks of Linguistics and Communication Sciences 36), 472–479. Berlin: De Gruyter Mouton. Rooth, Erik. 1932. Die Sprachform der Merseburger Quellen. In Niederdeutsche Studien: Festschrift für Conrad Borchling zum 20. März 1932, dargebracht von Freunden und Mitarbeitern und dem Verleger, 24–54. Neumünster: Wachholtz. Rubin, Aaron D. 2005. Studies in Semitic grammaticalization (Harvard Semitic Studies). Winnona Lake, IN: Eisenbrauns. Schladt, Mathias. 2000. The typology and grammaticalization of reflexives. In: Zygmunt Frajzyngier & Traci S. Curl (eds.), Reflexives: Forms and functions (Typological Studies in Language 40), 103–124. Amsterdam: John Benjamins. Schrijver, Peter. 1997. Studies in the history of Celtic pronouns and particles. (Maynooth Studies in Celtic Linguistics 2). Maynooth: Department of Old Irish, National University of Ireland. Schrijver, Peter. 1999. The Celtic contribution to the development of the North Sea Germanic vowel system, with special reference to coastal Dutch. Northwestern Language Evolution 35. 3–47.

Concerncing myself

145

Seebold, Elmar. 1984. Das System der Personalpronomina in den frühgermanischen Sprachen: Sein Aufbau und seine Herkunft (Ergänzungshefte zur Zeitschrift für Vergleichende Sprachforschung 34). Göttingen: Vandenhoeck & Ruprecht. Siemund, Peter. 2002. Reflexive and intensive self-forms across varieties of English. Zeitschrift für Anglistik und Amerikanistik 50. 250–268. Siemund, Peter. 2003. Varieties of English from a cross-linguistic perspective: Intensifiers and reflexives. In: Günter Rohdenburg & Britta Mondorf (eds.), Determinants of grammatical variation in English (Topics in English Linguistics 43), 479–506. Berlin: Mouton de Gruyter. Sipma, Pieter. 1913. Phonology and grammar of Modern West Frisian: With phonetic texts and glossary (Publications of the Philological Society 2.) London: Oxford University Press. Thurneysen, Rudolf. 1946. A Grammar of Old Irish, revised and enlarged edition. Transl. D. A. Binchy & Osborn Bergin. Dublin: Dublin Institute for Advanced Studies. Tristram, Hildegard. 1999. How Celtic is Standard English? ([Publications of] Institut lingvističeskich issledovanij, Rossijskoj akademii nauk.) St. Petersburg: Nauka. Ureland, Sture. 1979. The development of Dutch and West-Frisian reflexives between 1879 and 1979. In: Marinel Gerritsen (ed.), Taalverandering in Nederlandse dialekten: Honderd jaar dialektvragenlijsten: 1879–1979, 250–263. Muiderberg: Coutinho. van der Wal, Marijke & Cor van Bree. 2008. Geschiedenis van het Nederlands. 5th impr. Utrecht: Het Spectrum. Vennemann, Theo. 1998. Zur Etymologie von Éire, dem Namen Irlands. Sprachwissenschaft 23, 461–469 (Reprinted as chapter 22 in Vennemann 2003). Vennemann, Theo. 2001. Atlantis Semitica: Structural contact features in Celtic and English. In: Laurel Brinton (ed.), Historical Linguistics 1999: Selected Papers from the 14th International Conference on Historical Linguistics, Vancouver, 9–13 August 1999 (Current Issues in Linguistic Theory 215), 351–369. Amsterdam: John Benjamins (Reprinted as chapter 5 in Vennemann forthc.). Vennemann, Theo. 2002a. On the rise of ‘Celtic’ syntax in Middle English. In: Peter J. Lucas & Angela M. Lucas (eds.), Middle English from tongue to text: Selected papers from the Third International Conference on Middle English: Language and Text, held at Dublin, Ireland, 1–4 July 1999 (Studies in English Medieval Language and Literature 4), 203–234. Bern: Peter Lang (Reprinted as chapter 10 in Vennemann forthc.). Vennemann, Theo. 2002b. Semitic → Celtic → English: The transitivity of language contact. In: Markku Filppula, Juhani Klemola & Heli Pitkänen (eds), The Celtic roots of English (Studies in Languages 37), 295–330. Joensuu: University of Joensuu, Faculty of Humanities (Reprinted as chapter 11 in Vennemann forthc.). Vennemann, Theo. 2003. Europa Vasconica – Europa Semitica (Trends in Linguistics: Studies and Monographs 138). ed. Patrizia Noel Aziz Hanna. Berlin: Mouton de Gruyter. Vennemann, Theo. 2004. Phol, Balder, and the birth of Germanic. In: Irma Hyvärinen, Petri Kallio & Jarmo Korhonen (eds.), Etymologie, Entlehnungen und Entwicklungen: Festschrift für Jorma Koivulehto zum 70. Geburtstag, 439–457. Helsinki: Société Néophilologique (Reprinted as chapter 19 in Vennemann forthc.). Vennemann, Theo. 2008. Lombards and lautverschiebung: A unified account of the High Germanic Consonant Shift. Sprachwissenschaft 33. 213–256. Vennemann, Theo. 2009. I and U: Glide accretion in Ubian, English, and Celtic. Anglia 127. 208–237.

146

Theo Vennemann

Vennemann, Theo. 2011. English as a contact language: Typology and comparison. Anglia 129. 217–257. Vennemann, Theo. forthc. Germania Semitica, ed. Patrizia Noel Aziz Hanna. Vennemann, Theo. in prep. Radical word order change. Munich: University of Munich unpublished manuscript. Vezzosi, Letizia. 2003. Where himself comes from. Linguistica e Filologia 17. 205–240. Vezzosi, Letizia. 2005. The development of himself in Middle English: A ‘Celtic’ hypothesis. In: Nikolaus Ritt & Herbert Schendl (eds.), Rethinking Middle English: Linguistic and literary approaches (Studies in English Medieval Language and Literature 10), 228–243. Frankfurt am Main: Peter Lang. von Soden, Wolfram. 1972. Akkadisches Handwörterbuch. Unter Benutzung des lexikalischen Nachlasses von Bruno Meissner (1868–1947) bearbeitet von Wolfram von Soden. 3 vols. Wiesbaden: Otto Harrassowitz.

Patrick McConvell

Granny got cross: semantic change of kami ‘mother’s mother’ to ‘father’s mother’ in Pama-Nyungan 1 Introduction One of the more problematic aspects of linking etyma is where there is a difference in meaning between them. As is widely acknowledged, the theoretical underpinnings of work in semantic change do not have the strength of those in phonological change, which are supported by much theoretical phonology and awareness of which are the most common and natural sound changes. Kinship however is an area in which there has been a great deal of exploration in comparative semantics, and theoretical frameworks have been proposed. Change in kinship meanings has been less in focus but there has been some research in that area utilizing the general structural models proposed (see e.g McConvell, Keen, and Hendery in press for some examples). One notion which is commonly used in kinship semantics is that of cross versus parallel. A cross kintype is one in which, for case of a kintype having two links, one of the links is between two relatives of different sex; ‘cross cousin’ for instance is a mother’s brother’s child or father’s sister’s child. By contrast in a parallel kintype, the link is between relatives of the same sex, for instance parallel cousin is a mother’s sister’s child or father’s brother’s child. A very important distinction in the typology of kinship systems is whether they instantiate the cross-parallel distinction in their terminology or not. In Australia, unlike in most European and many other languages, most (but not all, see Dousset in press; McConvell and Gardner to appear) do have the crossparallel terminological distinction not only in cousins, but also in the grandparental generation. For instance most systems have different terms for ‘father’s mother’ (FM: cross) and ‘mother’s mother’ (MM: parallel). Only a few Australian systems like Western Desert neutralize this distinction and have a single ‘grandmother’ term as in English. The root *kami in proto-Pama-Nyungan, the ancestor of the biggest language family in Australia, has the meaning MM. However in some areas it has the meaning FM. These two meanings are quite distinct in most of Australia, falling on opposite sides of the cross-parallel divide. In this paper an explanation is sought for how and why this change occurred. First the generalization that

148

Patrick McConvell

semantic change occurs via an intermediate stage of polysemy (Evans 1992) is brought into play: evidence is assessed as to whether there was a transitional stage of MM = FM (ie a cross-parallel neutralized ‘grandmother’ term). Other linguistic and cultural characteristics of the area where the change occurred, such as the existence of matrilineal institutions, are investigated to probe whether these had an impact on the change.

2 Semantic change and reconstruction 2.1 Theories of semantic change applied to kinship From the matrix of semantic field theory, in the early twentieth century, more attention focused in the 1950s–1970s on how the differences between the words within the field could be more systematically described. Kinship became a playground for the development of the new “componential” approach since kinship terms seemed especially amenable to decomposition into a small number of semantic features which were perhaps universal, but combined in different ways to form different terms (Allan in press). Different combinations of the same basic features could represent terms in different languages with different meanings, at least ideally. This development was almost entirely based on synchronic data and while a few attempts were made to apply componential semantics to historical change in kinship they were not notably successful (Frisch and Schutz 1967; and see below). “Structural semantics” – the componential or distinctive feature approach – has been adjudged to be quite inadequate for historical work by some scholars. Fritz (forthc.) lists several criticisms levelled at the approach: (1) it is weak on parole (speech), the use of language; (2) it draws a strict boundary between lexical and encyclopaedic knowledge, which makes it difficult to explain innovations due to metaphor and metonym; (3) it is not well-equipped to deal with polysemy – a key element in meaning change; (4) its definition of features as necessary and sufficient excludes fuzzier mechanisms such as prototypes and “family resemblances”. In sum, it neglects the context, contextual use and pragmatics so important for an understanding of change. Wierzbicka (1987) also points out the inadequacy of componential approaches to kinship in several respects, including lack of a proper account of polysemy. She uses examples from componential analyses of Australian indigenous kinship terms in such works as Burling (1969). While criticizing Scheffler’s use of a componential framework she applauds his approach to polysemy which distinguishes the core meanings of such terms as ‘mother’ and ‘father’ from exten-

Granny got cross

149

sions to wider kin (Wierzbicka 1987: 140; Scheffler 1978). The Natural Semantic Metalanguage approach of Wierzbicka and colleagues has not been applied to semantic change. In the field of kinship, Coseriu (1964) did use a componential type of approach to analyse the change from what anthropologist would call a Sudanese or Lineal system for uncles and aunts in Latin, where mother’s sibling is distinguished from father’s, to an Eskimo system in descendant languages like French. There is loss of a distinctive feature paternal/maternal in this process. Besides yielding a neat description in this case at least, this kind of approach can make predictions – if for instance one has a hypothesis that change only occurs where such features are say, added or dropped – provided the features are well justified, for instance universal. Now in this case the maternal/paternal opposition was seen to be the most salient. However, when we approach it from the diachronic viewpoint another opposition sometimes called “parity” – the one we are discussing in this paper between MM and FM – between parallel (same-sex siblings: MZ and FB) and cross (opposite sex siblings: MB and FZ) comes to the fore. The cross-parallel distinction was lost when Latin amita changed from meaning ‘father’s sister’ to meaning also ‘mother’s sister’ (which was originally matertera) and similarly avunculus lost its cross meaning of ‘mother’s brother’ to become ‘uncle’ including ‘father’s brother’ (originally patruus). Dealing with such diachronic issues built up the case for adding different approaches which might predict the direction of change to the arsenal of semantic theories alongside the componential approach. The absence of social contexts of use in theories such as componential analysis is telling also in this case too. What intrigues about this loss of crossness in ‘uncle’ and ‘aunt’ terms is not just the simplification of the semantic field but the reasons why it happened. This is still debated in the history of European kinship, (cf. Goody 1983, 2000) but it appears to be related to changes in family structure and descent. We will touch on similar issues in our discussion of motivations of change of meaning of grandparental terms in Australia. Explanation of kinship in terms of Marking (or “Markedness”) was an important contribution by Greenberg (1980; see also Kronenfeld 1996, in press). While Greenberg hinted at the implications of his Marking findings for a diachronic theory of kinship semantic change, it was Hage especially who drew the “implicit” diachronic implications from Greenberg’s theory out and provided empirical evidence (1999). Hage (1999: 425) derives several principles about the behaviour of marked and unmarked terms from Greenberg which are relevant to kinship.

150

Patrick McConvell

Greenberg also had concrete hypotheses about where unmarked and marked divisions would fall in kinship: (a) genealogically close relations are more unmarked than distant relations, and consanguineal relations are unmarked vis a vis marked affines; (b) first ascending generations are unmarked as against all other generations, and ascending generations are unmarked as against the corresponding descending generations. Ehret (2008) also uses a similar kind of generalization as a guide for deciding which direction of change in kinship terms is more likely where other evidence is equivocal. Hage proceeds through an application of the Marking theory interpreted diachronically to the Salishan family of languages in north-western USA. He was unusual among contemporary anthropologists in taking very seriously the results of historical linguistics in reconstruction of forms and the implications of geographical distribution of cognate forms. Among other principles, Hage (1999: 433) mentions the interaction of marking and salience and marking and prototype effects. The combined effect of the marking rules, as interpreted diachronically by Hage for Salishan, seems to be an evolution towards simpler systems with less terms, while some systems in the family remain with more terms reflecting earlier stages. This is a powerful demonstration of this trend, but one has to question if it is in part built into the scheme by the way Marking theory is put into operation. Marking theory predicts that the marked members of pairs will be lost first, but there does not appear to be any complementary theory of complexification, or of a change such as the one we are considering here, MM → FM, which may have proceeded in stages, first simplification/neutralization MM → MM = FM, then complexification, yielding the FM meaning. There are limitations to the componential view of lexical items in general, and of kinship terms in particular. The notion of extension once again became important for the semantics of kinship. Kinship terms do not typically have a fuzzy halo around a core meaning like other types of words, but have a specific extended meaning (see Kronenfeld 2008, in press). So in systems of the classificatory type where for instance ‘father’ is the core meaning, the same term also refers to ‘father’s brother’, a clearly separate and definable kintype. This is an example of polysemy in my (and Wierzbicka’s, as discussed) view. Lounsbury built a new framework for analysis in which extension was central, including transformational equivalence rules, extending central basic kintypes to others. This approach was elaborated by Scheffler (1972), who produced a masterly work on Australian kinship classification (1978). Comparative and historical studies of variation and change in kinship terms have been carried out using ethnography and historical documentation. In North America, Fred Eggan was early in analysing change in Native American kinship

Granny got cross

151

(1937), followed by Spoehr (1947) and others. Utilizing literary and legal sources, large amounts of data on kinship terms and their changing meaning and usage have been retrieved in Europe and elsewhere. This is an important area of study since it can provide us with detail of the ordering of usage and meaning over hundreds of years. As well as illuminating many issues of socio-historical change in the societies concerned, such research also provides us with documented evidence of how complex changes proceed, which can serve as models for our reconstruction of prehistoric change where there is no such direct evidence. For instance Jones (1993) carried out a meticulous study of the change in usage and meaning of kinship terms through the history of German and found abundant evidence for extension of meaning as a prime mechanism of change (1993: 181). The extensions sometimes involved “semantic encroachment” on the domain of other terms, without necessarily any negative impact on the terms that are overlapped, although ultimately such processes did cause upheavals and reorganization of the field of kinship terminology One of the late medieval changes was the collapse of the bifurcate structure in German, which, like Latin, originally distinguished all four aunts and uncles. This is an example of the loss of the paternal/maternal (or possibly cross/parallel) distinction in uncle/aunt terms, converting a Sudanese pattern to an Eskimo or Lineal one. We have encountered this transformation previously, and it has been attributed to Marking by Hage (1999).

2.2 Semantic reconstruction of kinship terms In this section an example is presented of an explicit method of reconstruction of meanings of proto-language kinship terms, that of Dyen and Aberle (1974), especially their treatment of ‘grandmother’ terms in Athapaskan. We then move on to consider issues in the reconstruction of kinship terms in Australian languages. Dyen and Aberle’s work on the lexical reconstruction of proto-Athapaskan kinship terminology (1974) focuses on the task of reconstruction of the prototerminology using lists of terms (related forms, assigned codes) and meanings (also coded). Establishing a sub-grouping of the family is important in their method, as the key concept of “retained candidate” is defined using the notion of sub-grouping (1974: 17). The requirement for a retained candidate is that is cognates be different sub-groups (of the highest order) and that it cannot be (according to the evidence) an innovated candidate. . . . As it were an innovated candidate results from the independent ‘borrowing’ of a word from another list in which it is a retained or innovated candidate. Each such independent ‘borrowing’ is a TRANSFER [emphasis in original].

152

Patrick McConvell

This notion of “transfer” is in fact about what we will be talking of as different meanings in different languages, or meaning change, although the language used by Dyen and Aberle makes this somewhat opaque. They go on to introduce the notion of “inclusion” (1974: 18) If a candidate c appears in a list of subgroups L1 in meaning M1 and also in a different list of subgroups L2 in meaning M2, then if L2 contains all of the subgroups of L2 and at least one subgroup in addition, L2 is said to include L1; our practice is to say that the particular candidate in M2 INCLUDES [emphasis in original] the candidate c of M1. For example 65 appears in the meaning FM in Pac . . . and in Ap. Since 65 appears in the meaning mSS in all subgroups but Tan, mSS includes FM. If one candidate includes another and transfers are involved, the direction of any transfers is probably from the including candidate to the included candidate.

Further heuristics are proposed including the following which refers to an example of an etymon which has the meaning MM in some languages and FM in others, thus similar to the case which we are going to examining of *kami in Pama-Nyungan. An included candidate found in two sub-groups is immediately regarded as originating from transfers if in the same meaning there is found an unincluded candidate or an included candidate found in at least three subgroups. Thus the candidate MM 3, found in Ing and Can, is inferred to originate entirely from transfers from FM, found in Ing, Can, Pac and Ap.

Figure 1: Divergence of meaning in a proto-Athapaskan grandmother root

Granny got cross

153

Dyen and Aberle’s approach to transfer (meaning change) ingeniously attempts to determine the direction of change entirely by the structure of subgrouping and the numbers of branches with particular forms and meanings. This approach does not delve into the semantics of the terms including the role of markedness or of polysemy, nor the socio-cultural embedding of the terms. As Blust noted (2009): Dyen generally showed little interest in semantics. In citing cognate sets and their reconstructed prototypes, for example, he almost invariably left the reconstruction unglossed, listing only the meanings of the reflexes. This is perhaps understandable for someone trained in the American Structuralist tradition. . . . In this theoretical tradition, meaning stands largely outside the realm of linguistic analysis, and the process of semantic reconstruction stands at a correspondingly greater remove from analytic certainty.

Others as well as Blust have criticized what they see as an over-mechanistic approach. Whistler (1980: 280) writes of a basic methodological flaw of Dyen and Aberle’s approach to kin classificatory reconstruction. . . . Thus, for example, they ascribe proto-language status to a reconstructed kinterm and specify its meaning in terms of a mechanical ‘majority rules’ algorithm which counts representations across significant branches in the family.

Whistler’s admirable reconstruction of the semantics of proto-Wintun kinship drawing on the approach of Lounsbury and Scheffler “eschews any such formulaic approach; meanings for proto-forms are specified within their specific contexts of equivalence rules, kin super-classes and morphological system” (Whistler 1980: 276–277). Different Wintun subgroups can be characterized by presence or absence of particular Lounsburian equivalence (reduction) rules.1 Unfortunately, it seems this excellent methodology was not followed in succeeding years, probably because of the decline in interest in kinship and particularly the history of kinship systems, and attacks on structuralist approaches.

2.3 Change via polysemy One theory of semantic change which has currency in Australia, including in kinship etymology, is the idea that there must have been a bridging case of polysemy between two disparate meanings of a root, and sometimes we are lucky enough to find the predicted polysemy in the right area still extant.

1 For other attempts at implementing a diachnronic approach using Lounsburian analysis, see Bright and Minnick (1966).

154

Patrick McConvell

The Australian work on semantic change using this kind of approach (e.g. Evans 1992; Riemer 2005) is related to other work which is oriented towards notions of polysemy and pragmatics (e.g. Gerraerts 2002; Vanhove 2008) which, as we have seen, are difficult to integrate into a framework of componential semantics, even with the addition of Marking theory. One view is that semantic change in general is never directly from A to B but must proceed by way of an intermediate polysemy A = B. This idea fits well with the known set of common polysemies (equations) in kinship terms universally, and also in regions of the world. This parallelism yields a method of testing the plausibility of a reconstructed change in a kinship term. If the change is thought to be from say mother’s brother to mother’s brother’s son then we can fairly confidently say that there must have been an intermediate stage of MB = MBS. We know such equations exist and are known as Omaha skewing (McConvell and Alpher 2002; McConvell in press). Further tests of the plausibility of the hypothesis include the fact that such equations usually only occur in societies with a patrilineal emphasis. Examples of actual recent Omaha equations in the region where this is thought to have been the pattern strengthen the case for the hypothesis. This is only the first part of the testing of the change hypothesis: A > A = B. The second phase, the loss of A, is harder to deal with. It is not a result of Marking theory, at least not of the kind used by Hage (1999) based on Greenberg. In the example of change via Omaha in northern Australia (McConvell and Alpher 2002; McConvell and Keen 2010; McConvell in press) the original MB sense is taken over in Yolngu by a loanword from a neighbouring language. In the case of older German veter, the extension of oheim from MB to MB = FB (loss of the maternal/paternal or cross/parallel feature) eventually removes the original meaning of veter, FB, and facilitates its extension to FBS (which may have been a secondary meaning from early on), and ultimately ‘cousin’ in general as in modern German Vetter (Jones 1990). These are more the mechanisms – borrowing and extension of another term – rather than the motivation of the second stages of change however. There are also diachronic interactions of meaning changes similar to push- and drag-chains in sound change, of which one must take account. One process which we explore in this paper is resistance to semantic overlap in two terms leading to loss of the sense which is duplicated by two terms. In the case of loss of Omaha skewing in Australia, and of the more unusual type of skewing in German, it may be the case that skewing is a highly marked kind of polysemy – and is thought to be so even by groups which have it (Kronenfeld 2008) – and is an ‘overlay’ which can come and go relatively easily in history. If we pursue the idea of a two stage process of MM > FM however, the intermediate stage of polysemy MM = FM cannot be considered marked in

Granny got cross

155

general in world kinship systems – it is the most common meaning combination ‘grandmother’. However as discussed in the next section, it is perhaps ‘marked’ in the Australian context.

2.4 Constraints on polysemy in kinship Some polysemies (or equations, as they are better known in the anthropological literature) are very common world-wide. In societies with what have been called “classificatory” kinship systems, same sex siblings of a person A are called by the same term as person A. Further in many such societies parallel cousins of A are termed siblings of A and if same sex, are called by the same term as A. Let us consider grandparent terms in Australia, which is the topic of the present paper. These exhibit the above common kinds of equation mentioned above: all grandparents’ same sex siblings are referred to by the relevant grandparent term eg MMZ = MM. Since a parallel cousin is classified as a sibling in most groups, by two ordered equivalence rules (in Lounsbury’s approach) a grandparent’s parallel cousin is also referred to as the relevant grandparent e.g. MMMZD = MM. Additionally though there are other types of equivalence or merging which are found in only some parts of the world, and some parts of Australia. These are shown in Table 1. If it is true that changes in kinship term meanings which are from one meaning to another actually go through a polysemous stage, then polysemies either near-universal, or common in the region, which match the transitional polysemy add weight to the case for the change. However this method in itself does not necessarily solve the problem of the direction of the change, which is of course the key to deciding on the earlier proto-meaning of the form. In the case of the change via Omaha skewing discussed below, we can be fairly sure of the direction of change because the senior term (eg MB) extends to that in the lower generation (eg MBS), not as far as we know, in the opposite direction, in Australia.2 The other equivalence rules (eg A–F above) can also be interpreted in terms of diachronic change. If there are directional constraints on change this would be very helpful in proposing changes in particular cases. This is a task beyond

2 In Australia the Omaha rule found in nearly all cases is Type 1 (Lounsbury 1964; Scheffler 1978) which has this directionality. In just a few cases involving Type 3 the extension is upwards, for instance from MB to MF, as well as downwards. In discussion below the possibility of prior Type 3 (upward) Crow skewing in South-eastern Australia is also raised.

156

Patrick McConvell

Code

Name

Equation (example)

System type

Sheffler number: pp.

Scheffler name

A

Merger within parallel and within cross

FF = MMB ≠ MF = FMB

“Kariera”

4: 226

Parallel-cross neutralization

5: 226

Parallel-cross statusextension

B

Merger of same gender

FF = MF ≠ MM = FM

C

Skewing (adjacent generation)

e.g. Omaha, “Ngarinyin” mother = crosscousin etc

5: 404

Omaha skewing

D

Alternate generation equivalence

sibling = parallel grandparent; crosscousin = cross grandparent

7: 226; 376

AGA

E

Consanguinealaffinal

FZ = WM

8: 226

Spouseequation

F

Merger of opposite-sex siblings

MM = MMB

“Aluridja”

Table 1: Common Equations and Paths of Semantic Change in Australian Kinship Terms

the scope of this paper, but we will focus on one change which is the main topic of this paper. I introduced the paper with a confident statement that, in the case of PPNy *kami, the meaning change in some branches was *MM > FM. We shall look at arguments for this in detail below but here we shall briefly examine whether this falls out from the principle of transitional polysemy and constraints on directionality in the stages of the change. As we have noted and will be further explored below, the transitional polysemy hypothesis predicts that the intermediate phase of this change was MM = FM (‘grandmother’). While such grandmother terms without reference to the paternal/maternal of cross/parallel dimension are not common in Australia, they are found, and in regions which may be connected to the area where the root has the FM meaning. It is not so certain though, based on independent evidence and principles that the process began with the meaning MM and ended with FM. MM is the more widespread meaning and this certainly is a piece of evidence among

Granny got cross

157

others, but as we have noted above “majority rules” schemes are shaky when offered as prime evidence for directionality of change. Alternatives may be available in the varieties of “Marking theory” approaches, such as that of Hage (1999). Marking and salience: For instance sex of relative is more salient than sex of connecting (linking) relative, so systems which neutralize the latter and not the former are more common than those that neutralize the latter and not the former. For instance systems where MM = MF and/or FF = FM are rare but systems of MM = FM and FF = MF are common (e.g. the current European system and others in many places). The diachronic interpretation is that systems which distinguish all four of these grandparents (as in the proto-Salishan system) lose the linking relative sex distinction first, yielding a MM = FM ≠ FF = MF system.3 Marking and prototype effects: Assuming that FF and MM are the best examples or prototypes of grandparents, then if MF and FF are neutralized, it is predicted that the FF term will be the one retained for both, and the term for MM replaces that for FM. This is confirmed in Salishan in most cases. These principles would predict the transition from MM to MM = FM for parts of Pama-Nyungan also, and it follows that it is the original term for MM which also becomes FM if and when the MM meaning is replaced by another term.4 The final stage of replacement of MM = FM by the FM meaning does not fall out from these principles and in fact raises issues about whether the Marking principles are the correct or full explanation. As noted previously, making the terminological distinction between cross and parallel is a striking feature of Australian Aboriginal kinship systems in general, and one which is found even in the grandparental generation, where this distinction is missing in many parts of the world. It may be therefore that while the trend predicted by Hage from Marking Principles has been active, there is a countervailing trend to restore the cross-parallel distinction, such as distinguishing MM from FM, for instance to fit with neighbouring systems which do have the distinction. While the merging of kintypes under a single term requires only the loss or change of meaning of one of the terms, the restoration of the FM versus MM distinction requires the recruitment of a new term, either by borrowing of a term

3 The other possibility is the kind of Kariera system found in Australia or elsewhere, where MM = FF and MF = FM. As in the previous example of parity (cross/parallel) is more important here than either sex of relative or sex of linking relative. 4 This could be an example of Kurylowicz’s preservation of a secondary function after replacement of the primary function (4th law of analogy, as discussed in Arlotto 1981) the extended (non-prototypical) sense retains the older form.

158

Patrick McConvell

from a neignbouring language, or by the change or extension of meaning of a another term within the language – mechanisms which are not dealt with by the kind of Marking Theory offered by Hage, nor probably by diachronic adaptations of Lounsburian equivalence rules as they are known. This issue is dealt with when the proposed split of FM and MM is discussed, associated with the innovation of a new MM term.

3 The Pama-Nyungan root *kami ‘mother’s mother’ 3.1 Pama-Nyungan Pama-Nyungan (PNy) is the largest language family in Australia, stretching over most of the continent except the Central North and Tasmania. All the languages and sub-groups discussed here are Pama-Nyungan. See Map 1 for Australian

Map 1: Australian Languages

Granny got cross

159

languages showing Pama-Nyungan sub-groups according to O’Grady, Wurm, and Hale (1966). There is significant evidence for the linguo-genetic unity of Pama-Nyungan languages, and their separateness from the Non-Pama-Nyungan families. The most widely held view is that Pama-Nyungan split off from a grouping of NonPama-Nyungan (NPNy) languages and developed its common innovations at that time, including specific forms of pronouns, case suffixes and verbal morbology, which can be reconstructed to proto-Pama-Nyungan (pPNy). There are also some hundreds of lexical reconstructions for proto-Pama-Nyungan not found in Non-Pama-Nyungan (see Evans 2003, 2005; Alpher 2004; O’Grady and Hale 2004). There is some debate about a couple of groups of languages, the Tangkic family and Garrwa-Wanyi, as to whether they should be regarded as affiliated to Pama-Nyungan or not. This may indicate that they are Non-Pama-Nyungan branches sharing some close phylogenetic connection with Pama-Nyungan. This together with other evidence indicates a possible origin of Pama-Nyungan around the Gulf of Carpentaria, in Northern Queensland. Some lower-level subgroups have been established using the comparative method, building on initial assays with lexicostatistics. Some subgroups have been shown to exist which are mergers of two sub-groups originally proposed e.g. Ngumpin-Yapa (McConvell and Laughren 2004). Work is on-going on higherlevel sub-groups (Bowern and Koch 2004). It may be for instance that a number of western subgroups can be combined into a group somewhat like what O’Grady and Hale called Nyungic (cf. Wurm 1972), but which may have to be somewhat modified.

3.2 The root *kami ‘mother’s mother’: retention of original meaning There are a number of kinship roots which may be counted as proto-PamaNyungan. At the moment it appears that the system could have been Dravidian/ Kariera in nature; that is, that they involved a rigorous cross-parallel division in terms along with equation of parallel grandparents and of cross grandparents (e.g. MFZ = FM) and association with bilaterial cross-cousin marriage (for details see McConvell 2008; McConvell and Keen 2010). One of the most well-established of the proto-Pama-Nyungan kinship terms is *kami which has I believe the central meaning ‘MM’. If the proto-system was Kariera it would have also had the meaning FFZ (i.e. a “Kariera” extension) but this issue is not immediately relevant to the problem at hand. What is at issue is

160

Patrick McConvell

whether the proposal of the proto-meaning MM is justified given that there are a number of the languages where reflexes of the root mean FM, FM = MM, and in a smaller region, a number of other kintypes. Reflexes of *kami or *kaminy are found widely distributed with the meaning ‘mother’s mother’ in Pama-Nyungan languages (see Map 2 below). As already noted, this is not a clinching argument, but is part of a strong case for this option as opposed to, for instance, proposing FM or FM = MM as the protomeaning. This root is not found in Non-Pama-Nyungan languages with the exception of the Nyulnyulan family where Bowern (1998) reconstructs *kamirta as ‘mother’s mother’ to proto-Nyulnyulan (marked with black square on Map 2). It is undoubtedly a loanword from neighbouring Pama-Nyungan languages (perhaps proto-Marrngu), with the suffix -rta which may be related to the Pama-Nyungan kinship suffix *-rti (McConvell 2008). Within Pama-Nyungan the meaning MM is found in many parts of Queensland (Paman in Cape York Peninsula; Maric in southern interior Queensland and also on the Central coast, extending also to Bandjalangic in northern NSW, and in the Mayi languages south of the eastern Gulf of Carpentaria). It is found in the Thura-Yura languages of South Australia (exemplified on the map by Kaurna around Adelaide) and in the Marrngu languages of the eastern Pilbara in WA, and possibly in Arandic5. In the Western Desert area and in some dialects of Bandjalangic *kami reflexes have a meaning of both MM and FM – a cross-parallel neutralization.6 This is further discussed below. In some Paman languages of North Queensland the reflexes of *kami have a “Kariera” extension, meaning both MM and FF(Z) (see McConvell and Keen 2010). Complete meaning shifts to FM and other kin types are not shown on Map 2 but are shown on other maps below and discussed in later sections of this paper. In some places in the Northern Territory *kami reflexes are not found meaning MM but kami with an archaic suffix means its reciprocal fDC (see following section 3.3). The sections of the map where *kami MM reflexes are apparently not found are predominantly in the Lake Eyre Basin, the upper Murray, in New South 5 Koch (in press) argues against forms like ipmenhe MM in Arandic languages being descended from *kaminy. It is argued there that ipmenhe is a reflex of a Proto-Arandic *aylpmenhe, so not a reflex of *kami at all, but possibly *Calmi. 6 The MM = FM term in Western Desert has two forms, each in different dialects: kami, as discussed, descended from MM, and kaparli descended from an FM root. This supports the idea that Western Desert loss of the cross-parallel distinction is a later development from a Kariera system or similar (McConvell and Keen 2011).

Granny got cross

161

Wales and Victoria, and in western Pilbara and the south-west of Western Australia. However in some of these areas reflexes of *kami are found with other meanings which it is argued below are later developments from the MM meaning. In some of the other areas (notably the Lake Eyre Basin and a surrounding region, and the Western Pilbara) another root is found for MM *kanyi+ which itself may possibly ultimately be related to *kami, for instance descended from complex forms with a suffix and syncope *kaminy+ > kany+.

Map 2: Reflexes of *kami with the meaning ‘mother’s mother’

3.3 Senior and junior forms with additional morphology Although kami(ny) is not found in the Ngumpin-Yapa language as a term for MM a form with an archaic suffix kaminyjarr is in Gurindji and some other Ngumpin-

162

Patrick McConvell

Yapa languages as the junior reciprocal fDC.7 Forms cognate with both root and suffix of the Eastern Ngumpin form are found in a number of widely distributed régions. The form kamintyarr is also found in Djabugay on the north east coast of Queensland the meaning of reciprocal of MM and FFZ, following the Kariera equiation of these two relations found in that area, and a variation gamidharr far south in the Gidhabal dialect of Bandjalang (Geytenbeek and Geytenbeek 1971: 59 glossed ‘female ego’s grandchild’).8 In the Yolngu enclave of Pama-Nyungan languages surrounded by NonPama-Nyungan languages in North-east Arnhem Land, the form gaminyarr is a grandchild term, although no kami-related form is found as MM or other grandparent. The absence of ty here results from a regular sound change in the Yolngu languages. An interesting point is that the meaning is fSC or mDC, the reciprocal of FM or MF, not fDC the reciprocal of MM, possibly indicating that the term kami in this area, before it was lost, had changed to mean a cross grandparent (FM) as in some other areas. However matters may not be so straightforward. In Tulua on the Central coast of Queensland the base term kami, with a partially reduplicated form kamimi, retains a parallel meaning MM, but its derivatives in the grandchild generation, kamilam, and kamiyan are cross, the reciprocals of FM, not of MM. So differences in parity between the +2 and –2 generation – which I refer to as “twisting” – do not necessarily mean that the grandparent term has changed to cross first. Incidentally the use of partial reduplication to signal the senior member of a pair seems to be an old phenomenon found scattered in various places in PamaNyungan (McConvell 2008), as illustrated with the forms descended from *kami in Kaurna around Adelaide. In this language there is no ‘twisting’ of the relationship between the senior and junior forms – both are parallel. 7 The form for MM in these languages is jaju or jaja, but the etymology so far remains problematic. There are forms like tyatyi in north-central Queensland and into the Northern Territory with meanings like MF, and forms like tyatya in Victoria and central coastal NSW, meaning ‘elder sibling’, often ‘elder sister’. Of these the latter connection is preferable: the pattern of common equations (type D) suggests that eZ may have been the original meaning shifting to MM. However there is a long distance between these languages and Ngumpin-Yapa and there is no plausible link in a common higher level subgroup, leaving us with a suspect hypothesis of independent retentions from proto-Pama-Nyungan with a shift of meaning. Bridging polysemies were not found with this root, but they do exist for instance in Paakantyi kany(i)tya MM = eZ. 8 The basic meaning of kami in Gidhabal is given as ‘father’s mother’ in this source, contrasting with ‘mother’s mother’ baabany (Geytenbeek and Geytenbeek 1971: 75). See below for further discussion of this reversal of the proto-Pama-Nyungan meanings, and cross-parallel neutralization in the area.

Granny got cross

163

Map 3: *kami MM reflexes plus *kami + suffix fDC (>mDC)

4 Change in meaning of *kami to ‘father’s mother’ 4.1 Overall distribution of the FM meaning In various languages and groups of languages reflexes of *kami have the meaning ‘father’s mother (FM)’, which is probably a later development. This may not seem too strange to those attuned to European kinship but this entails a change from a “parallel” relative to a “cross” relative – two categories which are kept strictly apart in most Australian indigenous kinship systems. Map 4 shows the distribution of the FM meaning of this root. In the Karnic subgroup of Pama-Nyungan (Bowern 1998, 2001; Breen 2011) most of the

164

Patrick McConvell

languages have this meaning and another root is MM. This is discussed in the following section 5.2. In the Bandjalangic sub-group some dialects appear to have the MM and some the FM meaning, while others have both meanings. The Waka-Kabic subgroup to the north has something similar, but also other meanings as discussed in section 6. To the north of Waka-Kabic the MM meaning is found, with the exception of Warrungu, where it is FM. In the Western Desert dialects which have reflexes of *kami there is complete cross-parallel neutralization, with the meaning ‘grandmother’ as in English, combining MM and FM. This polysemy covers a wide area of the arid internal parts of Western Australia, south-west Northern Territory and north-west South Australia: only a couple of examples are marked on the map.

Map 4: *kami FM

Granny got cross

165

4.2 Karnic (Eyre Basin) The area where the cross (FM) meaning of kami is most clearly dominant, with no cases of FM = MM equations, is in the Karnic languages of the Lake Eyre Basin. If, as is argued here, MM was the original meaning, the change may represent a shared semantic innovation in the Karnic subgroup. In fact however the relevant grouping in which the FM meaning of *kami is found is not the entire Karnic subgroup, but corresponds to a large subgroup within Karnic identified by Austin.

Diagram 2: Internal subgrouping in Karnic (after Austin 1990)9

Bowern (1998), basing her argument on morphological reconstruction, did include Arabana-Wangkangurru in the Karnic family. There has been extensive recent discussion on the classification of Karnic languages, the latest being in Breen (2011); see also Koch and Hercus, this volume. The classification of Karnic languages has been controversial (see Austin 1990; Bowern 1998, 2001, 2009; Dixon 2002: xxxvii; Breen 2007, 2009). Bowern’s subgrouping is based most explicitly on the criterion of common innovations (cf. fn 7 of Koch, this volume). Austin (1990) reconstructs *kami FM only to proto-Central Karnic, not to proto-Karnic: *kami ‘father’s mother’ Di, Ng, Yl, Ya, Yw: kami. Others do consider Arabana and Wangkangurru as part of the Karnic subgroup, perhaps to be grouped in the Western Karnic branch (although pronoun evidence suggests that there is considerable distance between Arabana-Wangkangurru and all other

9 Proto-Western Karnic (pWK), daughter languages Diyari (Di), Ngamini (Ng), Yarluyandi (Yi); Proto-Central Karnic (pCK), daughter languages pWK, Yandhruwandha (Ya), Yawarrawarrka (Yw), Mithaka (Mi), Karuwalim (Ka); Proto-Karnic, daughter languages pCK, Pitta-Pitta (Pp), Wangkumara (Wm).

166

Patrick McConvell

branches of Karnic: Bowern 1998). If the classification of Arabana-Wangkangurru with Western Karnic is correct, then one might expect *kami to be inherited in Arabana-Wangkangurru also. There is a FM term amanyi in Arabana. This is unlikely however to be inherited from *kami(ny) as loss of initial k is not a sound change found in these languages. However, it is possible that this form is borrowed from Arandic (cf. ipmenhe, epmanhe; probably not a reflex of *kami, as discussed here). The Arandic term means MM however, so this might imply that this term underwent the change from MM to FM in line with other reflexes of *kami. In Wangkangurru the FM term is apirla. no doubt borrowed from Arandic aperle FM. In the languages outside the CK subgroup in the north-east, kami is not found. In Pitta-Pitta the FM term is ngathanha, no doubt descended from pPNy *ngatyV+ ‘MF/FMB’ (a Kariera equation) from which FM can be reached by a short step of loss of male gender specification.

4.3 The other Pama-Nyungan MM root *kanyi(CV) The term for MM in Karnic including Pitta-pitta in the north, and Arabana-Wangkangurru in the west, is kanyini and some variants (kanhini and with prestopping katnhini in Arabana-Wangkangurru). The exception is kanyitya in Wangkumarra in the south-east of the group. This term is shared with Paakantyi (and other related languages like Wilayakali) kanyitya/kanytya MM, to the south-east of Karnic, and likely to be a borrowing into Karnic from that direction. However kany(i)tya and kanyini are likely to be cognates going back further in Pama-Nyungan, possibly sharing a common root *kanyi together with different suffixes of obscure function. Kanyini is found in the Thura-Yura language Parnkalla with the meaning FM and in some sources MM. Thura-Yura *kanhini is also reflected in Adnyamathanha as atnyani MM and in Kuyani as katnyini MM, and Nukunu as katnyini MM. Kanyi does occur as a term for ‘daughter’s child’ (gender of propositus not specified) in Warluwarra, to the north-west of Karnic, and as (at least partially) a reciprocal of MM, is a probable cognate.10 In the western Pilbara there a number of terms of the form kantharri or kantyarri MM which could well be cognate, with the forms in the east which we have been discussing, with an added suffix -tyarri, and if so, the root would possibly

10 Elkins’ Pitta Pitta includes the meaning mSD for kanyini. Pitta Pitta is adjacent to Warluwarra. If correct this is a female reciprocal of FF but this may represent a Kariera equation of maternal and paternal parallel kin.

Granny got cross

167

be reconstructable to proto-Pama-Nyungan or some high-level subgroup of it.11 Map 6 shows the distribution of this cluster of forms which may represent a single MM root *kany(ty)V(rri). Semantically they are mostly the same or easily relatable. However the forms in the Mayi languages south of the Gulf of Carpentaria, meaning ‘father’ or ‘father’s sister’ are less tractable, being of a different generation (+1) and in the paternal line. In the absence of an equivalence rule or transitional meanings, these might have to be regarded as unrelated. The

Map 5: The rot kany(th)V(rri) MM

11 This would pose a problem of which of the roots *kami or *kanyi- to reconstruct to protoPama-Nyungan as MM. In another paper (McConvell in press) further grandparental terms are analysed which seem to offer two wide-scale regional distributions, and issues of whether this results from a sub-grouping or areal split early in Pama-Nyungan and/or the innovation of meanIng from another term in one of the terms. Another possibility in this case is that kanycould be a syncopated form of *kaminy found expecially where there are suffixes.

168

Patrick McConvell

Tablelands Yidiny form kanytyirri might be thought to fall under the polysemy principle D (alternate generation equivalence) but it is cross (cross-cousin) as opposed to parallel (MM), and it is very isolated from any other similar forms, so not to be seriously considered. Austin (1990) reconstructs *kanyini MM to proto-Karnic: *kanyini ‘mother’s mother’ Ng, Yl, Ya, Yw, Mi, Pp: kanyini; Wm: kanyidja; Di: kanhini.12 The puzzle therefore is to reconstruct the history in which Karnic retained this old root for MM while adopting the root kami in an innovated meaning FM. One element in a solution could be to propose that the prior existence of *kanyini ‘MM’ somehow forced the change in meaning of kami to FM. However this seems inadequate. There needs to be a more explicit hypothesis about the process which unfolded here. Harking back to earlier discussion, there might seem to be a need to invoke borrowing, or overlap of meanings, or both. One scenario could be the following: – Kami was borrowed into the proto-language of South-Central Karnic from a language in which it was polysemous between MM and FM. – Since there was already an inherited term for MM in these languages (kanyini) the term kami narrowed to FM. Alternatively, one need not assume borrowing of polysemous *kami but a sequence as follows (thanks to Harold Koch for this observation). I. Pre-Karnic *kami MM, FM II. Proto-Karnic *kanhini MM, *kami FM (innovation in basic sense, retention in secondary sense) III. Then replacement of FM in particular languages: W-Yutj ngapirla, W-Ng apirla FM borrowed from Arandic aperle; Pitta Pitta ngathanha; Wangkumara no information. One might expect to find traces of kami in the non-Central languages: WYutjura may indeed have a trace in kami ‘uncle’s child’ (MBC), which is an extension of FM by the alternate generation polysemy principle of the “Dieri system” type discussed further below. This solution has the advantage of resting on the transitional polysemy principle. It also draws on the idea that overlaps in meaning between kinship terms

12 This is in terms of Austin’s Karnic, not including Arabana-Wangkangurru. In terms of Bowern’s (2001) branching, it is reconstructable even higher, since it occurs in ArabanaWangkanurru. It does not occur however in Bowern’s eastern Karnic, where Wangkumara rather shows the Paakantyi loan kanyitya. It is also reconstructable for the (northern) Yura branch of Thura-Yura, which may have borrowed it from Karnic languages.

Granny got cross

169

are tolerated for some time, but eventually resolve themselves by the overlapping senses being eliminated, which seems to be borne out in languages with better historical documentation like German. There are of course a number of unanswered questions attached to this hypothesis. In the variant including borrowing of polysemous *kami one such unanswered question is the identity of the source language with the polysemous kami term (MM = FM). On the evidence of recently recorded languages, those which had such a polysemy were distant from the Karnic subgroup, such as Bandjalangic. The most likely donor languages for a kami loanword are to the south-east because the northern Karnic languages did not have this innovation and those in the south-west have cognates of that root if they have it at all, deformed by radical sound change such as initial dropping. Gunya, of the neighbouring languages to the south-east had the most common arrangement found in this part of Queensland, kami for MM and papiny for FM, both retaining the form and the meaning reconstructable to proto-Pama-Nyungan, so is not likely to be a source of a cross-parallel neutralized term MM = FM. Interestingly there is cross-parallel neutralization in the male grandparents where ngatyang (from a proto-Pama-Nyungan root meaning MF in all likelihood) is MF = FF, but there is no indication this pattern is found in the grandmothers. In Guwa, there is no record of the grandparental terms but kami means ‘elder sister’ and this is commonly equated with parallel grandparents, in this case MM. This contrasts with the pattern noted in Wangka Yutjura noted above, where kami is a cross-cousin, not a parallel cousin/sibling. These differences exemplify the effects of the distinctive “Dieri” kinship system in a region around Karnic, to be discussed below.

4.4 The “Dieri” kinship system Scheffler (1978: 365–384) also singles out what we know as the Karnic group of languages (plus a few neighbours in New South Wales) as having a distinctive kind of kinship system: the “Dieri” system, in which the FM term also functions as ‘cross-cousin’. This is in contrast to a much larger area of Australia in which the MF term has this polysemy. Another feature of this system is that the MM(B) term (kanyini) also functions as a sibling term (Scheffler 1978: 369), in a complementary fashion to the FM = cross-cousin polysemy. More common over wider areas is the equation of FF(Z) with siblings. Scheffler bases his explanation on the fact that the Alternate Generations Uterine rule (AGU) has priority over the Alternate Generation Agnatic rule (Scheffler 1978: 377) in this “Dieri System” area, reversing the more widespread

170

Patrick McConvell

priority ordering. The Alternate Generations Uterine rule merges terms in the matriline, and the Alternate Generation Agnatic rule rule operates on the patriline. Although Scheffler does not offer any diachronic interpretation, this is reminiscent of the way Whistler (1980) used the Lounsbury/Scheffler framework to track change in kinship in linguistic subgroups by proposing the addition of kin equations at nodes in a phylogeny. Another interesting aspect of Scheffler’s approach is that one equation does not actually replace another completely but makes changes in the priority ordering so that while a new rule dominates, the older one can still have some effect. This bears a resemblance to Optimality Theory developed in the 1980s–1990s in linguistics, and recently the application of Optimality Theory has been proposed for kinship (Jones 2010, including comment by McConvell). Anthropologists like Scheffler, unlike linguists, do not generally track the changes of meaning of specific terms as linguistic forms, however – for instance note the absence of this type of analysis in the groundbreaking volume by Godelier, Trautmann, and Tjon Sie Fat (1998). Marriage rules are also linked to this system. In Dieri (Diyari) a proper wife to a man is ‘woman whose mother is his mother’s cousin kami‘ (Scheffler 1978: 379) i.e his MFMD/MMBDD/MFZDD. However ‘mothers of a man and woman [who may marry]’ are mutually kamari (BW–HZ) rather than actually kami ‘cousin’ (Scheffler 1978: 380)13 The main diacritical feature [of the Dieri system] is that cross-cousins are classified with FM and wSC, not with MF and mDC as in Warlpiri & Aranda . . . systems of this kind are confined to northwest SA but they are similar to the Murawari and Wongaibon systems of NSW, in which it may be that the AGU rule also governs the kin-class statuses of crosscousins. (Scheffler 1978: 381)

The reference here to Wangaaybuwan in central northern New South Wales as being an easterly extension of this system does provide a possible link to the Bandjalangic systems, and other systems nearer the east coast to be explored below where kami either has the meaning FM, or FM in addition to MM, or other meanings apart from MM. In Wangaaybuwan and closely related languages which have this pattern, kami is apparently only present in the guise of kamiyan FZ/WM. This latter kintype is a person’s mother’s cross-cousin, not a cross-cousin as kami is in the Dieri system, and there is no warrant in terms of our known equations for shifting up one generation in this way.14 Recall though that Tulua, a Waka-Kabic 13 Kamantyarra is ‘right marriage partner’ in neighbouring Muruwari. 14 FZ = FZD is however a Crow skewing equation. Crow skewing is generally associated with thoroughgoing matrilineal institutions, which do not exist in full form in Australia. Crow skewing is only known from a small area of western Arnhem Land which has significant matriclans and matrimoieties alongside patrilineal descent.

Granny got cross

171

language, has the same form kamiyan in the meaning ‘son’s daughter’. If the propositus is assumed to be female, this is the reciprocal of FM.15

4.5 Matrilineal institutions This region of the “Dieri system” as described above is also the heartland of matrilineal clans and moieties. The “Dieri system” area is captured approximately on the AustKin map (Map 6) where the equation FM = cross-cousin is shown by the black dots (the black squares indicate MF = cross-cousin). This closely corresponds to the region of matrilineal institutions shown on Berndt’s map of distribution of social organization overlaid on a map of Australian language families and subgroups in Map 7 (courtesy of Claire Bowern). The region which kami is found in the Karnic languages (in the meaning FM) coincides almost precisely with the distribution of matrilineal moieties. The subgroups within Karnic that do not have kami (the northern languages not in the Central subgroup of Austin) do not have matrimoieties, and ArabanaWangkangurru is on the western edge of the matrimoiety distribution. Western Victoria (Kulin languages) also has the matrimoiety system, as shown on Map 7. The western Victorian language Dhauwurd Wurrung/Warrnambool (Blake 2003) is also within the matrimoiety area and has kamity, a probable cognate of kami, in the meaning of ‘cross-cousin’. In this case it is likely that the FM = cross-cousin polysemy of the “Dieri system” languages has gone to the next stage and lost the FM meaning.16 The central Thura-Yura language groups in South Australia also participate in the matrimoiety system, but the area around Adelaide in the east has patrilineal institutions and the western peripheral groups also do not have matrilineal institutions. These languages, like Kaurna, which we have encountered previously, retain *kami cognates in the original MM meaning. Parnkalla

15 Harold Koch (p.c.) observes that Tulua kamiyan is not necessarily a direct cognate, since this language has the regular -kan feminine suffix, and the -yan in kamiyan is likely to be a reflex of -kan. The -yan in CNSW languages may have another origin. 16 Kamity is glossed as a ‘female cross-cousin’ with female propositus. A form kamuty is given as ‘husband’s sister’ but this is possibly really the same form, given that female cross-cousin with female propositus would be a husband’s sister as an affine. Also in western Victoria kamuty is found as HZ in the ‘Mount Talbot’ language. FM is linyar in the Warnambool language, pointing to this form having taken over part of the field of the kami-related term leaving the cross-cousin meaning. There are no obvious cognates of linyar.

172

Patrick McConvell

does not have a *kami reflex meaning either FM or MM but does have kanyini as MM and ngapula as FM (cognate with a widespread Pama-Nyungan term for FM) with the “Dieri” polysemy FM = cross-cousin. The languages of the lower Murray like Ngayawang do not have cognates of kami but do have the FM = cross-cousin polysemy with an unrelated form nawulya. In northern New South Wales inland there is some influence of the matrilineal institutions and there is some FM = cross-cousin polysemy, recognized by Scheffler, showing up on the AustKin map in Yuwalaraay/Yuwaliyaay. In sum, there is a correlation between these matrilineal institutions and the ‘Dieri’ kinship system The MM rather than FF correlates with sibling because of the Alternate Generation Uterine equivalence which dominates. On the cross side, FM is equated with cross-cousin in the alternate 0 generation, rather than MF.

Map 6: FM = ‘cross cousin’ equations in southeastern Australia

Granny got cross

173

Map 7: Matrilineal institutions and linguistic subgroups in southeastern Australia (Language map from Claire Bowern with overlay of social institutions map from Berndt and Berndt 1977)

5 Coastal northern New South Wales and Queensland 5.1 Introductory The FM cross meaning of *kami reflexes is also found in other parts of Queensland and New South Wales. In some languages such as Bandjalangic it is found together with the MM meaning, frequently producing a MM = FM polysemy. In other languages there appears to have been a “swap” in meaning between two proto-Pama-Nyungan roots, *kami originally MM and *papi originally FM. These

174

Patrick McConvell

two developments could be linked and possibly linked to the switch to FM meanings farther inland. Additionally *kami reflexes have other meanings in languages north of Bandjalangic, such as Wakka-Kabic. The paths of development of these meaning innovations are complex and require deeper discussion of the particular unusual kinship systems in this area than we have space for here.

5.2 Bandjalangic In Bandjalangic there are dialects with the MM meaning, dialects with the FM meaning and dialects with the FM/MM polysemy. This appears to be a good case of the transition of one meaning to another via polysemy (1)

MM > MM = FM > FM

This might be interpreted as the transition which led to the MM > FM change more generally (in Karnic and the other areas mentioned). However the specific Bandjalangic change cannot realistically be the source of the whole set of MM > FM changes, because 1. the different areas/subgroups where this change took place are too distant from each and are not closely linked either linguo-genetically or by common membership of a diffusion area, as far as we know; 2. the differences between the Bandjalangic dialects are likely to be quite recent, since the dialects are similar, whereas the wider changes are likely to be much older; 3. it is possible that the languages involved in the change may have been closer together at some distant time so that the relevant polysemy and change might be part of some diffusion area. There is no other evidence for this that we are aware of at the moment. Given this scenario the Bandjalangic situation would be just a relic of some earlier variation rather than pointing to a source area 4. there is no evidence of bridging polysemy MM/FM in the Karnic region of its vicinity, or in the other cases (but see below for somewhat different signatures of a change from parallel to cross and transitional states). As noted in discussion of Karnic above, if the change in Karnic involved borrowing of a kami form which was polysemous into karnic follwed by subsequenct loss of the MM meaning because of overlap with kanyini, there is no candidate source language near to Karnic with this polysemy – indeed the languages to the area, if they have kami have it solely in the meaning MM.

Granny got cross

175

5.3 Switching MM and FM in Gumbayngirr *Kami has switched to FM meaning in Gumbaynggirr south of Bandalangic. In this case this is part of a swap of meanings between this and another protoPama-Nyungan root whose original meaning can reliably be reconstructed as FM: *papiny The Gumbaynggirr reflex baabany has changed meaning to MM.

PARALLEL

CROSS

PROTO-PAMA-NYUNGAN kami(ny)

MM

papi(ny)

FM

MM

gami < kami(ny)

FM

GUMBAYNGGIRR baabany < Papi(ny)

Diagram 3: Switching of proto-Pama-Nyungan meanings

It may be that a bridging polysemy, FM = MM, could be involved in this transition. But the transitional stage would presumably involve reflexes of both the proto-terms, *kami(ny) and *papi(ny), having the same meaning MM/FM, then changing becoming FM and MM in the next stage, opposite to the original meaning. This issue will have to be left for further research.

5.4 Further variation in north-central coastal New South Wales Kamiin in Kattang (Gathang) south of Gumbayggirr has the meaning ‘crosscousin’ which is likely to have developed from the shifted meaning FM of the root *kaminy via the “Dieri system” polysemy. The meaning MM has been occupied by the root *ngapi which is an old Pama-Nyungan alternative root for FM alongside *papi(ny). Thus the development in Kattang appears to have been as in Gumbaynggirr – a switch of the old roots for MM and FM, with FM > FM = cross-cousin > cross-cousin following, the last step being consolidated by the innovation of a new FM term kimpi. Another interesting feature of Kattang is the use of a suffixed form of kamiin, kamiintha as ‘wife’s mother’. This leads us to the rather more complex changes of meaning found elsewhere in the region.

5.5 Historical variation in Waka-Kabic and other groups Change from MM to FM is part of the story in Waka-Kabic, a subgroup north of Bandjalangic. There is some evidence in parts of this area that this change may

176

Patrick McConvell

have been relatively recent. For Goreng Goreng, Roth records kami as MM at Miriam Vale in 1898 and Mathews as FM in 1913. This paper has mainly been on the change of meaning of reflexes of *kami from MM to FM in some areas, and has brought in the change to cross-cousin via the FM = cross-cousin polysemy in a wide area. That is challenging enough, but pales beside the kinds of changes in meaning in parts of southern Queensland, including FZ, M, MB, elder and younger sister. These are shown on Map 8 (not all shown on the map). Sibling meanings as in Butchulla are relatively transparently related to MM by the Alternate Generation Uterine rule (with matrilineal institutions in the background) and complement the cross-cousin = FM pattern. However the +1 generation forms are especially puzzling. FZ is quite a dominant meaning and might be even reconstructable to proto-Waka-Kabic. Its distribution is shown in a schematic way in Map 9. Now the equation FZD = FZ is one of the principles of what Lounsbury (1964) calls Crow Type I skewing, and as we have seen the Alternate Generation Uterine rule yields FM = FZD which can feed into this. Alternatively, FM = FZ directly is found in Crow Type III skewing. The problem of course is that the legitimacy of bringing Crow skewing into this development is in question. In general Crow skewing is associated with matrilineality on a world scale and classic matrilineality is not reported in Australia. Further actual examples of these equations/polysemies are not found in this area so must be conjectured to have existed. Nevertheless Crow skewing does exist in at least one other area of Australia: western Arnhem Land (Harvey 2001; Garde in press). In this area, while patrilineal institutions are present, matrilineal non-local clans or phratries and matrimoieities are present and are of considerable significance. This situation has parallels with that in the area of south-eastern Australia which we have been examining. The map from Berndt obscures the situation by casting most of Queensland and New South Wales as an area of “sections”. The latter is an overlay on a system of matrilineal totemic clans (known as “meats”) and matrimoieties are present throughout this part of the section area. The occurrence of *kami reflexes in the meaning ‘mother’s brother’ is more difficult to explain. Equations of the type MM or FM or cross-cousin = MB are rare generally not only in Australia17. Further work is needed on this interesting set of variations. The polysemy “uncle (MB)” = “elder brother” is found on the 17 Shapiro (1968: 51) finds FZS = MB in Siriono and Mundurucu in South America, associated with marriage between uncle (MB) and niece. MBS = MB is found in Omaha skewing but not reported in this area, and the focus of the FM polysemy is FZC not MBC.

Granny got cross

177

Map 8: Further meanings of *kami cognates in southern Queensland and northern New Sout Wales

NSW central coast in Dunghutti and Kattang with a different root piyang. However this root also has the meaning ‘cross-cousin’ (MBS) and can be better analysed as involving an overlap of the two types of polysemy Omaha skewing and cross-parallel neutralization.

178

Patrick McConvell

Map 9: kami: ‘father’s sister’ and ‘cross-cousin’

Polysemy between “mother’s brother” and “elder brother” or “father’s sister” and “elder sister” does exist as well as in Dunghutti and Kattang, in other regions: Yir Yoront, Mayi-Thakurti, Diyari, Ngatjumaya, Wilyakali, Wangaaypuwan, Yuwaalayaay, Gumbaynggir, Darumbal, so this appears to be a more general phenomenon which should be investigated.

Granny got cross

179

5.6 Meaning changes in the north *Kami reflexes in some north Queensland groups reflect a “Kariera” polysemy, where the term means both MM and FFZ. This is probably an original polysemy in proto-Pama-Nyungan which has been lost in other areas, not a later change. An unusual aberration is the presence of the FM meaning in a single northern Maric language Warrungu, far north of the large group of languages which have this meaning to the south. It is unlikely that this is a relic of a much wider distribution of this meaning, but the reason why the parallel to cross change took place apparently isolated from other similar changes elsewhere is unclear. However, we have already remarked on what I called “twisting” of the meaning in the northern Waka-Kabic language Tulua, in which the grandmother term is parallel but the grandchildren terms descended from *kami are cross. A quite similar “twisting” of the junior reciprocal seems to have taken place in the Yolngu languages much further north, and west, in Northeast Arnhem Land. The term gaminyarr ‘woman’s son’s child’ the reciprocal of FM is descended from *kami-ny-jarr whose reflexes generally mean ‘woman’s daughter’s child’ the reciprocal of MM – another instance of the change from parallel to cross.

6 Conclusions Faced with a range of meanings for a Pama-Nyungan root kami, the paper has proposed that the proto-meaning was ‘mother’s mother’. This meaning is found in the largest number of languages in diverse sub-groups throughout the PamaNyungan area. By contrast the areas where the root has the meaning ‘father’s mother’ are quite geographically restricted, mainly to part of the Karnic subgroup. It is proposed that this is an innovation in meaning. This change to FM is found in Karnic where the reconstructed form for MM kanyini also has widespread cognates in Pama-Nyungan. If we adopt the productive idea that semantic changes go through a transitional stage of polysemy between the old and new meanings, then this will predict a transitional polysemy of MM = FM – a semantic package like the word ‘grandmother’ in English. While not common, this type of polysemy is found, including in Bandjalangic on the New South Wales-Queensland border. The gap between this area and Karnic and the fact that the nearer neighbours of Karnic have *kami cognates in the meaning MM raises questions about the applicability of the transitional polysemy hypothesis. Whether or not transitional polysemy is involved, it seems likely that the prior presence of another root for MM in the

180

Patrick McConvell

Karnic subgroup precludes or erodes the adoption of *kami in the same meaning MM. Throughout the Karnic area and in a region surrounding it there is a particular distinctive pattern of polysemy related to Alternate Generation Uterine, the matrilineal of uterine variety of alternating generation equivalence, in which FM = cross-cousin and MM = elder sister. This in turn is an area where there are significant matrilineal institutions – matrilcans and matrimoieties. The pattern of lineality of the broader social categories fits with the patterning of polysemy in the kinship terms. Further changes in meaning of *kami cognates especially in southern Queensland seem to flow from the prior set of changes MM > FM > FZD, for instance the change to FZ perhaps due to Crow skewing having been active in the area, also explicable in terms of the matrilineal backgound. In more general terms this paper raises the issue of what kinds of constrained theories of lexical semantics and semantic change can be used in doing etymology and what role evidence external to language (in this case, attested or reconstructed matrilineal social institutions) should play. A small set of constraints operating on polysemy and change outcomes in Australia, and more generally, provides a robust framework for etymological work on kinship terms and their meanings. The resulting reconstruction of ancient kinship systems made up of these kinterm meanings and their geographical distribution provides the basis for hypotheses about how social as well as linguistic patterns evolved. Clearly there is a relationship between changes from parallel to cross meanings of terms and matrilineal institutions, but which came first and how their interaction played out is a theme to be explored further.

References Allan, Keith. In press. Semantics and pragmatics. In: Linda R. Waugh, John Joseph & Monique Monville-Burston (eds.), Cambridge History of Linguistics. Cambridge: Cambridge University Press. http://users.monash.edu.au/~kallan/papers/SemPragCHL.pdf (accessed 9 October 2010). Alpher, Barry. 2004. Pama-Nyungan: phonological reconstruction and status as a phylogenetic group. In: Claire Bowern & Harold Koch (eds.), Australian Languages: Classification and the Comparative Method, 93–126. Amsterdam: Benjamins. Arlotto, Anthony. 1981. Introduction to Historical Linguistics. Toronto: Madison Press. Austin, Peter. 1990. Classification of Lake Eyre languages. LaTrobe University Working Papers in Linguistics 3. 171–201. Berndt, Ronald & Catherine Berndt. 1977. The World of the first Australians. Melbourne: Lansdowne Press.

Granny got cross

181

Blake, Barry. 2003. The Warrnambool language: a consolidated account of the Aboriginal language of the Warrnambool area of the western district of Victoria based on nineteenth century sources. Canberra: Pacific Linguistics. Blust, Robert. 2009. In Memoriam Isidore Dyen, 1913–2008. Oceanic linguistics. 48. 2. 488– 508. Bowern, Claire. 1998. The Case of Proto-Karnic. Canberra: Australian National University, BA Honours Thesis. Bowern, Claire. 2001. Karnic Classification revisited. In: Jane Simpson, David Nash, Mary Laughren, Peter Austin & Barry Alpher (eds.), Forty years on: Ken Hale and Australian languages, 245–261, Canberra: Pacific Linguistics. Bowern, Claire. 2009. Reassessing Karnic: a reply to Breen. Australian Journal of Linguistics 29. 337–348. Bowern, Claire & Harold Koch (eds.) 2004. Australian Languages: Classification and the Comparative Method. Amsterdam: Benjamins. Breen, J. Gavan. 2011. Reassessing Karnic: A Reply to Bowern (2009). Australian Journal of Linguistics 31 (1). 137–142. Breen, J. Gavan. 2007. Reassessing Karnic. Australian Journal of Linguistics 25: 175–199. Bright, William & Jan Minnick. 1966. Reduction Rules in Fox Kinship. Southwestern Journal of Anthropology. 22 (4). 381–388. Burling, Robbins. 1969. Cognition and componential analysis: God’s truth or hocus-pocus?. In: Stephen Tyler (ed.), Cognitive Anthropology, 419–28. New York: Holt, Rinehart and Winston. Coseriu, Eugenio. 1964. Pour une sémantique diachronique structurale. Travaux de lingustique et de literature 1. 139–186. Dixon, R. M. W. 2002. Australian Languages: their nature and development. Cambridge: Cambridge University Press. Dousset, Laurent. In press. “Horizontal” and “vertical” skewing: similar objectives, two solutions? In: Thomas Trautmann & Peter M. Whiteley (eds.). Crow-Omaha: New Light on a Classic Problem of Kinship Analysis. Amerind Foundation Advanced Seminar series. Tucson: University of Arizona Press. Dyen, Isidore & David Aberle. 1974. Lexical reconstruction: the case of the proto-Athapaskan kinship system. Cambridge: Cambridge University Press. Eggan, Fred. 1937. Historical changes in the Choctaw kinship system. American Anthropologist 39. 34–52. Ehret, Christopher. 2008. Reconstructing ancient kinship in Africa. In: Nicholas J. Allen, Hilary Callan, Robin Dunbar & Wendy James (eds.), Early Human Kinship: From Sex to Social Reproduction, 200–231. Malden: Wiley-Blackwell. Evans, Nicholas. 1992. Multiple semiotic systems, hyperpolysemy, and the reconstruction of semantic change in Australian languages. Diachrony within synchrony: language history and cognition, Günter Kellermann & Michael D. Morrissey (eds.), 475–508. Frankfurt: Peter Lang. Evans, Nicholas. 2003. Introduction: Comparative non-Pama-Nyungan and Australian historical linguistics. In: Nicholas Evans (ed.), The non-Pama-Nyungan languages of northern Australia: comparative studies of the continent’s most linguistically complex region, 3–25. Canberra: Pacific Linguistics. Evans, Nicholas. 2005. Australian Languages Reconsidered: A Review of Dixon (2002). Oceanic Linguistics 44 (1). 242–286.

182

Patrick McConvell

Frisch, Jack & Noel W. Schutz. 1967. Componential Analysis and Semantic Reconstruction: The Proto Central Yuman Kinship System. Ethnology 6 (3). 272–293. Fritz, Gerd. forthc. Theories of meaning change; an overview. Maienborn, Claudia, Klaus von Heusinger & Paul Portner (eds.). Semantics: An International Handbook of Natural Language Meaning. Vol. 3. Berlin: Walter de Gruyter. Garde, Murray. In press. Crow skewing in Bininy-Gunwok. Geeraerts, Dirk. 2002. The theoretical and descriptive development of lexical semantics. In: Leila Behrens & Dietmar Zaefferer (eds.), The Lexicon in Focus. Competition and Convergence in Current Lexicology, 23–42. Frankfurt am Main: Peter Lang. Geytenbeek, Brian & Helen Geytenbeek. 1971. Gidabal Grammar and Dictionary. Canberra: AIAS. Godelier, Maurice, Thomas Trautmann & F. Tjon Sie Fat eds. 1998. Transformations of kinship. Washington: Smithsonian Institution Press. Goody, Jack. 1983. The development of the family and marriage in Europe. Cambridge: Cambridge University Press. Goody, Jack. 2000. The European Family: an historico-anthropological essay. Oxford: Blackwell. Greenberg, Joseph. 1990 [1980]. Universals of kinship terminology. In: K. Denning & S. Kemmer (eds.). On language: selected writings of J. H. Greenberg. Stanford University Press. Hage, Per. 1999. Marking universals and the structure and evolution of kinship terminologies. JRAI 5(3). 423–441. Harvey, Mark. 2001. Oenpelli Kunwinjku kinship terminologies and marriage practices Oceania 72. 117–142. Jones, Doug. 2010. Human kinship, from conceptual structure to grammar Behaviour and brain sciences 33. 367–416. Jones, Douglas & Bojka Milicic (eds.). Kinship, language and prehistory: Per Hage and the renaissance in kinship studies. Salt Lake City: Unviersity of Utah Press. Jones, William Jervis. 1990. German kinship terms (750–1500) Berlin & New York: Walter de Gruyter. Koch, Harold. The etymology of a paradigm: the Pama-Nyungan 3SgF reconsidered. This vol. Koch, Harold and Luise Hercus. Obscure vs. transparent cognates in linguistic reconstruction. This vol. Koch, Harold. In press. The reconstruction of kinship terminology in the Arandic languages of Australia. In: Patrick McConvell, Ian Keen & Rachel Hendery (eds.), Kinship Change and Reconstruction. Salt Lake City: University of Utah Press. Kronenfeld, David. 1996. Plastic Glasses and Church Fathers: semantic extension from the ethnoscience tradition. New York: Oxford University Press. Kronenfeld, David. 2008. Fanti Kinship and the analysis of kinship terminologies. Champaign: University of Illinois Press. Kronenfeld, David. In press. Kinship terms: typology and history. In: Patrick McConvell, Ian Keen & Rachel Hendery (eds.), Kinship Change and Reconstruction. Salt Lake City: University of Utah Press. Lounsbury, Floyd G. 1964. A Formal Account of the Crow- and Omaha-Type Kinship Terminologies. In: Ward H. Goodenough (ed.), Explorations in Cultural Anthropology: Essays in Honor of George Peter Murdock, 351–393. New York: McGraw-Hill. Maienborn, Claudia, Klaus von Heusinger & Paul Portner (eds.). forthc. Semantics: An International Handbook of Natural Language Meaning. Vol. 3. Berlin: Walter de Gruyter.

Granny got cross

183

McConvell, Patrick. 2008. Grandaddy Morphs: the importance of suffixes in reconstructing Pama-Nyungan kinship. In: Bethwyn Evans, Claire Bowern and Luisa Miceli (eds.). Morphology and language history: in honour of Harold Koch, 313–327 Amsterdam: Benjamins. McConvell, Patrick. In press a. Omaha Skewing in Australia: Overlays, Dynamism and Change. In: Thomas Trautmann & Peter M. Whiteley (eds.). Crow-Omaha: New Light on a Classic Problem of Kinship Analysis. Amerind Foundation Advanced Seminar series. Tucson: University of Arizona Press. McConvell, Patrick. In press b. Proto-Pama-Nyungan kinship and the AUSTKIN project: reconstructing terms for proto-Mother’s Father and their transformations. In: Patrick McConvell, Ian Keen & Rachel Hendery (eds.), Kinship Change and Reconstruction. Salt Lake City: University of Utah Press. McConvell, Patrick & Barry Alpher. 2002. The Omaha Trail in Australia: tracking skewing from east to west. Anthropological Forum 12 (2). 159–176. (Special number on “Kinship Change” eds. P. McConvell, L. Dousset and & F. Powell). McConvell, Patrick & Helen Gardner. To appear. The Descent of Morgan in Australia: Kinship Representation from the Australian Colonies. Ethnology McConvell, Patrick & Ian Keen. 2010. The Transition from Kariera to an Asymmetrical System: Cape York Peninsula to North-East Arnhemland. In: Douglas, Jones & Bojka Milicic (eds.). Kinship, language and prehistory: Per Hage and the renaissance in kinship studies, 99– 132. Salt Lake City: Unviersity of Utah Press. McConvell, Patrick & Ian Keen. 2011. The Austkin Project: Modelling Evolution Of Kinship Systems In Australia Using Linguistic Evidence. Paper to Society for the Anthropological Sciences meeting. Charleston. McConvell, Patrick, Ian Keen & Rachel Hendery (eds.). In press. Kinship Change and Reconstruction. Salt Lake City: University of Utah Press. McConvell, Patrick & Mary Laughren. 2004. Ngumpin-Yapa Languages. In: Claire Bowern & Harold Koch (eds.), Australian Languages: Classification and the Comparative Method, 151–178. Amsterdam: Benjamins. O’Grady, Geoffrey & Kenneth Hale. 2004. The coherence and distinctiveness of the PamaNyungan language family within the Australian linguistic phylum. In: Claire Bowern & Harold Koch (eds.), Australian Languages: Classification and the Comparative Method, 69–92. Amsterdam: Benjamins. O’Grady, Geoffrey, Stephen Wurm & Kenneth Hale. 1966. Australian Language Families, Victoria, British Columbia (Map with comments). Riemer, Nick. 2005. The Semantics of Polysemy: Reading Meaning in English and Warlpiri. Berlin: Mouton de Gruyter. Scheffler, Harold. 1972. Kinship Semantics. Annual Review of Anthropology 1. 309–328. Scheffler, Harold. 1978/2009. Australian Kin Classification. Cambridge: Cambridge University Press. Shapiro, Warren. 1968. Kinship and marriage in Sirioný society: a re-examination. Bijdragen tot de Taal-, Land- en Volkenkunde 124 (1). 40–55. Spoehr, Alexander. 1947. Changing kinship systems. Anthropological Series, Field Museum of Natural History 33. 159–235. Trautmann, Thomas and Peter M. Whiteley (eds.). 2010. Crow-Omaha: New Light on a Classic Problem of Kinship Analysis. Amerind Foundation Advanced Seminar Series. Tucson: University of Arizona Press.

184

Patrick McConvell

Vanhove, Martine (ed.). 2008. From Polysemy to Semantic change.: towards a typology of lexical semantic associations. Amsterdam: John Benjamins. Whistler, Kenneth. 1980. Proto-Wintun kin classification: a case-study in the reconstruction of a complex semantic system. Berkeley: Univeristy of California, Berkeley dissertation. Wierzbicka, Anna. 1987. Kinship semantics: Lexical universals as a key to psychological reality. Anthropological Linguistics 29 (2), 131–156. Wurm, Stephen. 1972. Languages of Australia and Tasmania. Den Haag: Mouton.

Alexandre François

Shadows of bygone lives: The histories of spiritual words in northern Vanuatu 1 Studying etymology with unwritten languages Many linguists are familiar with etymologies in Indo-European languages, and how they help us discover invisible threads between words that have become widely separated in form or in meaning.1 Etymology teaches us that miracle, marvel and mirror all stem from a single Latin verb mīrāri ‘wonder, admire’ – respectively via its derived nouns mīrāculum, Late Latin mīrābilia, and Old French mireor. The deep-reaching insights of Proto Indo-European reconstruction even allow us to find connections between the words idea, view, advice, story and witty: they all stem ultimately from a single PIE root whose form can be reconstructed as *weyd ‘see, know’ (cf. Chantraine 1968: 796). Similar etymological links can be drawn between quick, vital, bio and zoo VRS œtœ‑k ‘my soul’; *atá‑na > ata‑n ‘his/her soul’; *atá‑i > ɛtɛ ‘soul of’ (François 2005: 484–488). As for the sequence (C)e(C)ə in Lo-Toga, it is the regular reflex of an earlier string *(C)a(C)a (François 2005: 490). Proto Torres–Banks *ata ‘soul, spirit’ is the same word as PNCV *ʔata, whose sense ‘soul, spirit’ is found in other languages of Vanuatu (Clark 2009: 76). The three languages of Vanikoro, in the Solomon Is, also reflect it: Teanu ata, Lovono ala, Tanema ae ‘soul, spirit’ (François 2009: 111). Cognate forms in other Oceanic languages have a meaning ‘shadow, reflection’, as in Tongan ʔata (Ross, Pawley and Osmond, in prep.) – a sense which is lexified differently in north Vanuatu. All this points to the reconstruction of a POc form *qata ‘shadow, reflection; soul, spirit’. It is not entirely clear whether POc *qata ‘person, human being’ and POc *qata ‘shadow, reflection; soul’ formed a case of polysemy of a single word – or if they were mere homophones. The semantic connection between ‘person’ and ‘soul, essential living component of a person’ does not seem implausible, and one could propose that the two roots may be ultimately the same word. However, a discussion with several specialists of the domain (R. Blust; A. Pawley; M. Ross; M. Osmond, p.c.) yielded the conclusion that POc *qata is more

The histories of spiritual words in Vanuatu

211

probably a case of homophonous terms arising from different sources (see also Osmond, this volume). Table 2 sums up the steps that can be reconstructed for the shifts in meaning and sound of the two roots, which are further commented upon below.

‘outsider, alien’

PAN

Pre‑POc

POc

PNCV

Torres– Banks

e.g. Lehali

*qaRta >

*qata ↓ *qata >

*qata >

*ʔata >

*ata >

(11) n‑at

*qata ↓ *qata >

*ʔata >

*ata >

(38) n‑ɛta‑n

‘person’ ‘shadow, reflection’

*qantad

>

‘soul, spirit’

Table 2: The two homophones *qata of Proto Oceanic

The starting point for the first etymon is a Proto Malayo Polynesian root *qaRta, which Blust (1972, 2012) reconstructs as ‘outsider, alien’; it shows a shift to ‘human being, person’ even before Proto Oceanic, in various Central MalayoPolynesian languages (Blust 2012). At the level of POc, the form had changed to *qata, and the original meaning ‘outsiders, alien people’ had evidently been lost, as it is attested nowhere in Oceanic languages; the word had broadened its meaning to ‘person, human being’, for which *qata then competed with *tau (Ross, Pawley and Osmond, in prep.). The Banks and Torres forms cited in (11), illustrated in Table 2 with Lehali, reflect regular sound change from *qata. As for the second *qata, it has been linked (Ross, Pawley and Osmond, in prep.) with a PAN root *qantad ‘shadow, reflection, image, likeness’ (Dahl 1981). The semantic connection between ‘soul, spirit’ and ‘shadow’ is relatively widespread in the Austronesian word, whether in relation to this root *qata or with POc *qanunu ‘shadow, reflection, soul’ (Blust 2012). Common to all these notions is the reference to a person’s particular “presence”, when it contrasts with the physical reality of their body. The spirits of our ancestors are present amongst us, in our personal memories and in our collective representations – like the shadows of bygone lives. 4.3.2 A second etymon for ‘soul’ Besides the form *qata given in (38), some Banks languages lexify ‘soul, spirit’ with a different protoform *tala (‑na):

212

Alexandre François

(39) *tala (‑na) → ‘(his/her) soul, spirit’ LYP n‑tala-n; VLW n‑tala-n; MTP na‑tala-n;

LMG

n‑ʔalɒ‑n; LKN tala/tɪ‑n

The origin of *tala is unclear. It may ultimately be derived from *ata (?), plus an element *‑la‑ of unknown origin.10 My Mwotlap informants were specific about what has a soul and what does not. Plants, as well as smaller animals such as insects, mice, or fish, are not endowed with a soul. By contrast, higher animates have a soul, including pigs (an important animal in traditional society), whales, dolphins, and the dreaded sea-snakes – but not sharks or turtles.11 This root *tala can be found in the compound *tala-mauri – literally ‘soul alive’ ( DRG Watɣʊrɣʊr, LKN Wɪtaɣɪːɣɪː), a word with no known etymology, and no relation with the names used in Mwotlap. It would be an interesting topic to compare the names and attributes of various creatures of the folklore in this region, and beyond.

13 See François and Stern (forthcoming). 14 Cf. the Kakamora creatures of Makira, in the Solomon Islands (Fox and Drew 1915; Fox 1962).

The histories of spiritual words in Vanuatu

219

4.4.5 The primal deities Despite the variety of terms, one root stands out as a widespread and important word, besides *[a]tamate. This term is *βui, which may perhaps be glossed ‘primal spirit’ or ‘deity’: (55a)

*βui → ‘primal spirit; deity’: HIW wʉ; LTG wʉ; LHI n‑βu; LYP n‑βu; VLW n‑βu; MTP nu‑βu; LMG n‑βu; VRS βü; MSN βu; MTA βui; DRG βu; KRO βu; LKN βuː; MRL nu‑βu

Codrington (1891), and after him Ivens (1931), discuss the reference of Vui in the “religion of Mota”; their observations are mostly confirmed by what I heard from other languages, and can safely be assigned to the protoforms. On the one hand, *[a]tamate primarily refers to “ghosts”, i.e. spirits emanating from dead mortals; they are mostly seen as dangerous creatures, who would be invoked, for example in malevolent sorcery. On the other hand, *βui designates the eternal spirits of the place, who were present even before mankind, and still inhabit the forest. These primal spirits created the world as demiurgic forces: they can legitimately be compared to deities or “gods” (see §4.4.6), whose prestige and aura rank much higher than *[a]tamate. Ghosts (*[a]tamate) are normally visible, and keep or take human shape to deceive their victims. By contrast, primal spirits *βui are immortal and invisible, present in the very fabric of the land. The etymology of *βui is unclear. Among the ones discussed by Ivens (1931), the most promising may be POc *puqun ‘base (of tree); root, origin’ – because these deities were present at the very beginning of Time. This tentative hypothesis is supported by the existence of similar metaphors in modern languages – ͡ wɪtɪ βʊnʊ ‘the Origin of the World’, literally ‘at the root of land’ e.g. Koro l‑kp (François forthc.). It is possible that *βui can further be analysed as a radical *βu ( mbønø). But why? No language of the Torres–Banks area seems to associate *mbanoi specifically with volcanoes. In fact, this connection is found in other languages of Vanuatu further south (Clark 2009: 80): (72b)

*mbanoi → ‘volcano’: Paama vanei; Namakir mbane; Nguna na‑panoi

(72c)

*mbanoi → ‘volcanic ash’: Tamambo mbanoi; Uripiv mbenu; Lewo pani

The histories of spiritual words in Vanuatu

229

Clearly, ancient cultures of Vanuatu have long associated the abode of the dead with volcanoes. Should one observe these words on a synchronic basis in each language taken separately, this connection would appear nowhere; it only comes to light thanks to language comparison. In sum, I propose to reconstruct the two following meanings for these two words: – *mbanoi ‘volcano; esp. the Volcano where the dead abide, the Underworld’; – *asura ‘a long and narrow cavity in a mountain or volcano; esp. the long and narrow corridor leading to the Underworld (*mbanoi)’ Nowadays, the association of *mbanoi with volcanoes has been mostly lost. In Hiw, the word Penɵ! has become a respectful salute when parting with someone: one says ‘[See you in] the Other World!’ to make sure that, should anything bad happen to one’s friend, their soul will safely reach the abode of the dead, rather than wander around the world like a lost soul. Modern reflexes of *mbanoi refer to an abstract location – Hell – which most modern speakers view as a remote place, often with no further specification. Except for Gaua island where the Gharet volcano is still a major landmark, elsewhere the abode of the dead is seldom identified in the actual geography of the islands; it is usually described as a ‘very remote’ place – to the point that its name is sometimes used, jokingly, to mean ‘somewhere very far’. Some speakers of Mwotlap told me that Amnʊ may be somewhere in the ocean – in conformity with the traditional link, heard in some stories, between ghosts and the sea. The most reliable information I collected was in an interview with Sesil Pilageliqe, a highly knowledgeable shaman (§4.6) of Toga island who is still active, and has travelled oftentimes to the Other World. According to him, Pənə ( bm, l > dl), following the main stress accent in the first syllable was a relatively recent diffusion phenomenon in the Lake Eyre basin (Hercus 1972). Absence of pre-stopping might therefore be regarded as a good indication of archaism in placenames. There are however no clear examples of it. This is because of the rules of pre-stopping:

Archaisms in Placenames in Arabana-Wangkangurru country

317

it is not found in the second member of compounds, as these do not carry the main stress accent. It is optional in words of more than three syllables. As most placenames are compounds we therefore cannot prove that placenames like the following are examples of archaisms, though they obviously contain words that normally have pre-stopping: – Kuluwa-nha ‘(a place with) kudluwa needlewood Hakea trees’. This is not a compound, but with the inclusion of the proper noun marker, nha, it could be regarded as containing more than 3 syllables. In the Mythology the Seven Sisters found a grove of needlewood trees at this place. – Nhili-paka-nha ‘digging for nhidli’ (a mouse species, probably Sminthopsis crassicaudata, once common and considered ‘good eating’). In the Mythology, that is what the Seven Sisters were doing at this site, near Lake Palankarinna, east of Lake Eyre. – Kini-thaka ‘striking with its tail, kidni’, i.e. ‘Scorpion’ is the name of a location in the Simpson Desert, mentioned in the Crane History.

3 Obsolete words surviving in placenames 3.1 General As we do not have any old records of the language it is difficult to be sure that a word is archaic. One way of knowing is when a word survived long enough for there to be some distant memory of it: ‘old people used to say that, but we don’t say it any more’. This was the case with the word nginya, ‘green’, which was heard only in the name of a Simpson Desert Emu site, Papu-nginya ‘Green Egg’. The word kupa ‘little’ is used in Arabana, but in Wangkangurru it has been replaced by nyara ‘little’. The word survives however in Wangkangurru country in the name of the Kallakoopah, the Karla kupa ‘little creek’, a creek that winds out west from the Diamantina through the desert and joins the lower Macumba on its way to Lake Eyre. The fact that the word kupa is no longer used in Wangkangurru means that the kinship term kuparli ‘younger sibling‘ has become unanalysable in terms of modern Wangkangurru. Ngapa ‘water’ is found in most Karnic languages, (the subgroup to which Arabana and Wangkangurru belong) and in the Yarli languages (Yardliyawara, Wadigali and Malyangapa ‘clay-water’). In the two languages Arabana-Wangkangurru and Wangka-Yutyurru3 immediately bordering on Arandic, it has been replaced by kutha, a borrowing from Arandic. One might question whether kutha really could be a direct borrowing from Arrernte kwatye, but as pointed out by 3 Details of Wangka-Yutyurru vocabulary are known only from a word-list by Gavan Breen.

318

Luise Hercus

Illustration 2: The Kallakoopah winding across the desert in a good season. Photo by Joseph Schofield

Harold Koch (pers.com.), there was also an Arrernte word kwethe. This is listed by Henderson and Dobson (1994: 456) in the following entry: kwethe-kwethe (SE, NE [dialects]) water. Only used when talking to a dog, e.g. calling it for a drink. See also kwatye. [emphasis in orginal]

In Arabana-Wangkangurru ngapa survives only in placenames and in derivatives. The following placenames in Arabana and Wangkangurru country appear on modern maps. The word for ‘water’ is usually first in these compound names in Arabana-Wangkangurru: Ngapa-marra ‘fresh water, Ngapamura, the lower Neales River; Umbum waterhole was traditionally called Ngapa-murirla ‘water dried out’. The following sites are known only from traditions and do not appear on modern maps: – Ngapa-wayarla ‘shrub-water’, water where wayarla bushes grow. This plant grows only in favourable locations in gullies, and was used for thatching humpies 4. 4 The work of Reuther (1981) interprets wayarla as meaning ‘wide’, Vol. VII no. 1364, but it was known to Mick McLean as the name of a bush.

Archaisms in Placenames in Arabana-Wangkangurru country

319

– Ngapa-Palkura ‘wattle water’, palkura is the ‘creek wattle’, Acacia stenophylla. – Ngapa-pirla-pirla ‘very black water’. There are however exceptions and ngapa can be the second the second member of the compound as in Thipa-ngapa ‘beetle water’, Teepagnuppa Creek and Springs, which is a major traditional site shown on the Warrina map-sheet, and in Pula-Ngapa ‘Double water’, the name of an un-mapped waterhole on the lower Diamantina. The newer word kutha has so far been found in only one placename appearing on maps: Kutha Parkulu ‘Two waters’, ‘Cootabarcoolia bore’, originally a double spring near Algebuckina. There are two other names on maps that might contain the word kutha ‘water’, but their derivation is not clear: Cootabungoodana Swamp, and Cootakartana waterhole, both in Arabana country. There are however more names with kutha ‘water’ listed by Reuther (1981: vol. VII): – “Kudamiri”, (Reuther 1981: VII, 687) translated by him as ‘the whirlpool in a waterhole’, and with a variant spelling. This word is more likely to represent kutha-mirri ‘a lot of water’. – “Kudanguru” (Reuther 1981: VII, 534) translated by Reuther as ‘clear water’. There is however no evidence of a word ‘nguru’ meaning ‘clear’ in the available Wangkangurru data. – “Kutamantara” (Reuther 1981: VII, 775) i.e. Kutha-mantarra ‘Sandhill-wattle water’. – “Kutawanngiri” (Reuther 1981: VII, 811), ‘tufts of wanngiri grass at the water’s edge.5 ’ – “Kutakudlangura” (Reuther 1981: VII, 524) which is analyzable as Kuthakurdalangura ‘water was lying there’. From this evidence it is clear that the distribution in placenames of the old word ngapa and the new borrowed word kutha reflects an earlier phase of the language, when kutha was taking over from ngapa. The placename ‘Mundoorana’, which refers to a spring created by the Two Ancestral Snakes in the Davenport Range, is based on mandura ‘two,’ cf. Diyari mandru. This word is no longer found in Arabana-Wangkangurru but is used in related languages east of Lake Eyre, in Diyari, Thirrari and probably Pirladapa, and the Yarli language Wadigali, which is immediately adjacent to Diyari. The name of the Simpson Desert well Kadlalumpa was said to represent an old word for ‘urine’. The site belongs to the myth of the Seven Sisters.

5 This should be wanngira, nitre-bush, Nitraria Schoberii.

320

Luise Hercus

3.2 Flora and Fauna There are plants and many animals that are now no longer known, and even their exact identity 6 is uncertain, but their name is still enshrined in placenames. Of special importance was a renowned and most powerful healing plant called wakimpa. The last elders who knew it had not seen one for many years and were sure it had been wiped out by rabbits: it is still known in songs and in placenames as for instance Wakimpa-puntaka ‘they broke down the wakimpa bushes’ (to harvest the leaves). The creeping plant named wari-wari (species unknown) was used for making carrying pads. There are six sites with this name listed by Reuther (Reuther 1981) but we have recorded only one on the lower Kallakoopah, where the Ancestral Grinding Stone Men were said to have collected these plants because their carrying pads had become worn out. Kawari probably Dasyuroides byrnei ‘the Kowari’ was once common and its name is preserved in Cowarie Station, on the lower Diamantina in SA and Kauri waterhole far to the north, adjacent to Eyre’s Creek in Queensland. Kadnungka the hare wallaby, Lagorchestes sp. with big rings around its eyes, has been extinct in the area for a long time, but it is remembered as ‘the Ancestor who creates heavy colds’, and a kind of flu, and Kadnungka-ngura ‘the Hare Wallaby Camp’ is the name of a sheltered valley just south of William Creek in Arabana country. Ngudlukanta/Ngulukanta the Desert Rat Kangaroo, Caloprymnus campestris (Johnston 1943), which lived among bushes, was a major totemic Ancestor. The species was in decline more than a century ago, as was the similarly named burrowing animal Ngudluwaltu, probably Bettongia lesueuri. Their names survive in the placenames Ngudluwaltu-ngarru ‘(headgear) made from the rat-kangaroo’ (see Reuther 1981: VII, 1463 ‘Ngurluwordungaru’) and Ngudlukanta-ngura ‘Ratkangaroo Camp’ (Reuther 1981: VII, 1427 ‘Ngurlukantangura’). Wilikinti, Notomys fuscus, which is better known and is sometimes equated with kawari, has given its name to two Simpson Desert sites, one is simply called Wilkinti, the other is Wilkinti-thara ‘rat-leg’. It is ironical that Kudnayapu, the extinct ‘lesser sticknest rat’ is recalled just in the one placename Kudnayapu-thipi ‘sticknest rat alive’ (cf. Reuther 1981: VII, 16).

6 Some of the Aboriginal names have been identified by Johnston (1943).

Archaisms in Placenames in Arabana-Wangkangurru country

321

3.3 Material Culture There are also old items of material culture, and practices that are no longer clearly recalled, but which are still visible in the placenames. Examples are: Walka, walka-walka (Reuther’s ‘wolka’) is a kind of fence made of bushes and put across creeks to trap fish, and thita is the Arabana word for a ‘mobile fence’ or ‘broom’ made of bushes and held by many people: it served the same purpose. There are a number of places where Ancestors had these fish-drives and this is shown in the placenames: there are sites named Walka-walka (Reuther’s ‘Wolka-Wolka’, 1981: VII, 2255) both on the lower Kallakoopah and on the Diamantina. In Arabana country there are the places named Thita-purrunha ‘(they) have got a fish-fence’ and Thita-puntakanha, ‘the fish-fence broke’. Warrtha ‘a large head-dress made of coloured feathers’. This word is preserved in the placename Warrtha-thakali ‘(they) were making a large head-dress’, ‘Wodadakarli’ (Reuther 1981: VII, 2447), shown as ‘Wathacully Yard’ on modern maps, near Old Alton Downs. The name belongs to the myth of the Two Boys. The Two Boys also made a small feather head-dress warrtha-warrtha7 at a place of that name (Reuther 1981: VII, 2407 ‘Wodawoda’) which was said to be in Goyder Lagoon.

4 Placenames re-analysed There are examples of well-known placenames, which were felt to be archaic, being re-analysed and given popular etymologies. A striking example is the name Mirra-Mitta, the site of a bore on the Birdsville track in Ngamani country, adjacent to Wangkangurru country. This was said to mean ‘Rat-place’ mayarru ‘long-haired rat, Rattus vilosissimus’ and mitha ‘place, country’. This name was unique: it was the only placename far and wide that actually contained the word for ‘place’. However the senior Wangkangurru speaker, Mick McLean, on our many trips past this site, never pronounced the name as Mayarru-mitha, he always said Mayarru-mithi. The data in Reuther’s work (Reuther 1981: VII, 2174) show that this was indeed the original name, and it meant ‘Rats shining’ and referred to the silky fur of the mayarru rats. Another example is the name of Kalamurina Station (this was originally downstream of its present location). The old station site and the nearby lake were called Karla-marra-nha ‘creek-new’ with the use of the optional proper

7 Repetition in nouns served a diminutive (Hercus 1994: 98).

322

Luise Hercus

noun marker -nha. This was a historical name because the Diamantina/Warburton had long ago broken through a sandhilll during a flood and carved out a new course at this place. This name is now being re-interpreted as Karla-marna ‘Creek-mouth’. The placenames thus give us a glimpse into the past in the use of older forms of words (palatals instead of dentals, and the possible absence of prestopping). The placenames also contain archaic words, and references to the fauna, flora and material culture of the past. This gives us time-depth for an area where we have no old historical records.

References Alpher, Barry. 2004. Pama-Nyungan: Phonological reconstruction and status as a phylogenetic group. In: Claire Bowern, and Harold Koch (eds), Australian Languages, Classification and the Comparative Method, 93–126. Amsterdam and Philadelphia: John Benjamins. Breen, Gavan, n.d. Wangka Yutyurru wordlist. Electronic file. Dixon, R.M.W. 1970. ‘Proto-Australian Laminals’ Oceanic Linguistics 9. 79–103. Dixon, R.M.W. 2002. Australian Languages. Their Nature and Development. Cambridge: Cambridge University Press. Henderson, John, and Veronica Dobson. 1994. Eastern and Central Arrernte to English Dictionary. Alice Springs: Institute for Aboriginal Development. Hercus, Luise A. 1972. ‘The pre-stopped nasal and lateral consonants of Arabana-Wanganuru’. Anthropological Linguistics 14 (8). 293–305. Hercus, Luise A. 1994. A Grammar of the Arabana-Wangkangurru Language, Lake Eyre Basin, South Australia (Pacific Linguistics Series C-128) Canberra: Australian National University. Hercus, Luise A. 2010. Song Language as used in the Arabana-Wangkangurru Emu History and the Crane History. Canberra: Australian National University manuscript. Johnston, T. Harvey. 1943. Aboriginal names and utilization of the fauna in the Eyrean region. Transactions Royal Society of South Australia 67. 244–311. Koch, Harold, and Luise Hercus (eds.). 2009. Aboriginal Placenames, naming and re-naming the Australian landscape (Aboriginal History Monograph Series 19). Canberra: ANU e-press. Reuther, Johann G. 1981. [1904]. The Diari. Transl. Philipp A. Scherer. Canberra: Australian Institute of Aboriginal Studies microfiche. Spencer, Sir Baldwin, and Gillen, Francis J. 1912. Across Australia. Vol. 1. London: Macmillan. Spencer, Sir Baldwin. 1903. MS Urabunna, Old Peake Station. Melbourne: National Museum of Victoria manuscript.

Subject index analogy 35, 60, 66, 86, 91–92, 94, 98, 106– 107, 111, 114, 116–117, 157 Arabana-Wangkangurru 6, 36–37, 41, 43– 49, 61, 165–166, 168, 171, 313–314, 316–317, 319 Arandic languages 37–38, 160 Arrernte 36, 38, 45, 317–318 Australian English 293, 302, 305–307

gender 56, 58–59, 61–62, 65, 69, 75–76, 97, 156, 166, 199, 201–203 Gamilaraay 251–259, 261, 263–272 Germanic 12, 20–21, 26–28, 86, 121–128, 134–135, 137–141

bilingual 10–12, 17, 19–21, 23, 25, 94 Blueprint Principle 10, 14–16, 25, 29 borrowing 5–6, 13, 16–23, 27, 37, 45, 70, 76, 83, 86–88, 93, 111–112, 116, 151, 154, 157, 166, 168–169, 174, 240, 261, 294, 306–307 budgerigar 6, 267, 281, 293–307

importation 17, 131–132, 135, 138, 306 Internal Development Bias 11, 13 Irish 23, 124–125, 128, 132–134, 136

Celtic influence 5, 121–122 Central Karnic languages 49, 165, 168 classifier 83, 89–90, 102–104 Compound 35, 37–38, 40–42, 45–46, 49– 50, 67, 86, 198, 200–203, 207–209, 212–214, 216–217, 231, 234, 251, 255– 257, 259–261, 269, 277, 280, 283–284, 286-288, 294, 307, 314, 317–319 comparative 1, 34–36, 38, 53, 55–57, 76–77, 90, 92, 102, 108–109, 117, 121 contact etymology 6, 9–11, 14–16, 19–22, 24–29 correlative 97, 99–102 , 114 Crow skewing 155, 170, 176, 180

historical linguistics 1–3, 6, 33–34, 87, 150, 185

Kariera system 156–157, 159–160, 162, 166, 179 Karnic languages 36–39, 41–49, 56–58, 60– 65, 73–74, 77–78, 163, 165–166, 168– 169, 171, 174, 179–180, 313, 317 kinship 6, 45, 73, 75, 147–151, 153–157, 159– 160, 163, 168–170, 172, 174, 180, 317 laminal consonants 313, 316 language contact 6, 9–10, 12, 14–16, 18–26, 29, 84, 91, 100, 109, 116–117, 122, 131, 135, 137, 139, 307 lexical etymology 4, 78, 86 loanword 19, 21, 23–24, 27–28, 154, 160, 169, 306–307 Malyangapa 40, 42, 317 normal transmission 5, 11–12, 28

Dieri system 168–172, 175 Diyari 38, 42–43, 45–48, 55–56, 60, 74, 165, 170, 178, 319 etymological approach 2–6 extension 48, 71–72, 85, 87, 91–92, 95–97, 99, 102, 104–105, 107–112, 150–151, 154–156, 158– 160, 168, 170, 215–216, 247 feminine 5, 53, 55–60, 62–63, 65–66, 69– 70, 72, 74, 75, 77–78, 171

obscure cognate 35–36, 44, 50 Oceanic 6, 186–187, 189–192, 194, 200, 203, 201, 221, 225, 229, 240, 245–248 Omaha skewing 154–156, 176–177 onomatopoeia 257–258, 263 Pama-Nyungan 5, 33–34, 36–39, 44–46, 49–50, 53–63, 66, 68–71, 75–76, 78, 152, 157–163, 167, 172–173, 175, 179 paradigm 5, 47–48, 53, 56–58, 61–66, 69– 70, 73–74, 76–78, 99, 121, 123, 133, 202

324

Subject index

partial cognate 35, 43, 45–47, 50 Phoenician/Punic 6, 27–28, 125, 127–130, 136–137 placenames 6, 40, 45–46, 256, 267–269, 270, 273, 290, 299, 313, 315–322 polysemy 94, 107–108, 111–112, 148, 150, 153–156, 164, 168–169, 171–180, 197– 198, 207, 210, 214, 216 possession 47, 75, 131 pronoun 4–5, 45, 47, 53–78, 83, 89–94, 98, 101–102, 111–112, 121–124, 126–140, 159, 165, 202 reanalysis 10, 53, 86, 88, 90, 94, 111, 112, 114, 300, 306 reduplication 128, 162, 251, 259–260, 268– 269, 272 reflexive 5–6, 71–72, 78, 121–140 relative clause 4–5, 83–85, 87–109, 111, 113–117 semantic change 2, 6, 37, 43, 61–62, 69, 147–149, 153–154, 156, 179–180, 187, 191, 196–198 Semitic 91, 93, 121, 125–128, 130–131, 135– 136, 138

spirit 6, 127–128, 185–186, 198–200, 203, 209–223, 225–227, 229–234, 238–241, 245–246, 279, 288, 290 structural etymology 3–5, 11, 87 substitution 10, 24, 26, 28, 306 substratum 15–18, 20, 57, 125, 135, 138 supernatural 186, 201, 212, 215, 217–218, 220–222, 232, 234, 237–240, 245, 247 Szemerényi 5, 34–35, 37, 39, 49 Thura-Yura languages 33, 36–44, 49, 160, 166, 168, 171 transfer 9–11, 17, 19–28, 83, 85, 93, 100, 111, 131, 151–153 typology 83, 117, 121, 136, 138, 147, 197 Vanuatu 185–191, 194–195, 197–199, 203– 204, 207, 209–210, 214–215, 220–223, 228–230, 234, 237, 239–240 Wadigali 40, 317, 319 Warluwarric 56–58, 68–69, 77 Welsh 122–125, 128, 130–136 Yardliyawara 40, 42, 317 Yarli languages 36–37, 39–43, 49, 317, 319 Yuwaalaraay 251–259, 261, 263–272