This book "asserts that the origin and spread of languages must be examined primarily through the time-tested techniques of linguistic analysis, rather than those of evolutionary biology" and "defends traditional practices in historical linguistics while remaining open to new techniques, including computational methods" and "will appeal to readers interested in world history and world geography."
Review of The Jarawara Language of Southern Amazonia
Date: Thu, 23 Jun 2005 13:03:45 +0900 From: Mike Morgan Subject: The Jarawara Language of Southern Amazonia
AUTHOR: Dixon, R. M. W. TITLE: The Jarawara Language of Southern Amazonia PUBLISHER: Oxford University Press YEAR: 2004
Michael W Morgan, Kobe City University of Foreign Studies
This book is a thorough descriptive grammar of the Jarawara language, or the Jarawara dialect of the Madi language, depending on which definition of language (p 7) you subscribe to. In either case, it is not the language of any large group of people (the dialect is spoken by less than 200 people, the larger language by less than 600). Nor is it a member of a large and wide-spread language family; taking the second sense of language, Madi is one of only five members of the Arawan (Arauan) language family, all (other members being Sorowaha, Paumari, Kulina-Deni, and the extinct Arawa) spoken around the Jurua and Purus rivers roughly in Brazil's Amazonas state and vicinity (Dixon includes maps, a general picture of Jarawara and its linguistic neighbors can be seen at Proel (2004)). However, though numerically and geographically "minor", Jarawara is a linguistically very interesting -- and thus, significant -- language, and Dixon's treatment of it is an important work to be added to the libraries not only of people interested in Arawan languages in specific, or Amazonian languages in general, but of ANYONE doing comparative and/or typological study of languages, or as an excellent example for anyone attempting to write a descriptive grammar of any language.
Dixon's book is not the first book-length introduction to aspects of Jarawara grammar; Vogel's (2003) 254-page University of Pittsburgh doctoral thesis has that honor. The latter, however, has more limited scope (verb classes), while Dixon's is a comprehensive grammar of the language. Although I have only just started reading Vogel's work, it appears to offer an alternative analysis to Dixon's on several issues. Dixon (2000) is also an earlier treatment of one area not covered by Vogel (2003). Also of interest for comparative purposes (the book under review itself provides a good deal of comparison) are the various works of Everett (n.d., 1999, 2003) on other Madi dialects and also Paumari, and Salzer & Chapman's (1999) dictionary of Paumari (which also includes a grammatical overview).
The length limitations of this review preclude a thorough presentation of the contents of each chapter, or even an examination of ALL the interesting things that are going on in this language; there are simply TOO many of them. The body of the book consists of 27 chapters, not specifically grouped into the normal sections of Phonology, Morphology, Syntax, but generally organized so that the contents of one chapter are related to those before and after. Chapter 1 (pp 1-15) starts off with a description of the "linguistic type" of Jarawara, which reads like a list of "100 reasons why you really should find this language fascinating". It then goes on to give an overview of who the Jarawara are, and where their language fits in the physical, cultural and linguistic world. It also provides a very short summary of the sources for the material analyzed, and a section on "Chances of Survival".
Chapter 2 (16-69) is one of the longer single chapters and deals with all issues of phonology. Of particular note for Jarawara is the clear need to distinguish between grammatical and phonological word. A description and discussion of a variety of phonological rules take up perhaps 2/3 of the chapter. As in all later chapters, comparisons and differences between Jarawara and the other Madi dialects and also between the Madi language and the other Arawan languages are included. One interesting phonological feature of the Jarawara dialect that distinguishes it from the other Madi dialects is a change in the syllable-counting rule for stress assignment from the underlying proto- Madi system of counting from the beginning of the word to a more recent system of counting from the end found only in Jarawara. In Jarawara certain phonological rules (e.g. the deletion of the initial syllable -he-/-ha- in the reported suffix) follow the older system, while other rules (e.g. actual stress placement) follow the newer one.
Chapter 3 (70-94) provides a Grammatical Overview of the language. If time only permits reading a 25-page excerpt of the book, these are the pages to read. They expand on the list given at the outset of the book, and all are treated in detail in subsequent chapters. This is, however, the place in the book where Dixon treats the Janus issues of homonymy and redundancy most directly: Jarawara has a fairly high level of homonymy (especially where it matters most), yet a fairly low level of redundancy -- which information-theory tells us would otherwise be one way to compensate for this redundancy. On theoretical grounds I personally have my suspicions that some of the proposed examples of homonymy are in fact merely different (context- motivated) senses of one and the same item. Although a single reading of the book (and only a couple weeks to absorb it all) did not give me enough information to propose a Gesamtbedeutung for any particular case, at times the connection seems probable ... and in fact at times Dixon himself hints that this is the case. Still Dixon is probably right: homonymy is at a higher level and redundancy at a lower level than in MOST languages.
Chapters 4 through 9 (95-279) introduce the predicate structure of Jarawara. As Dixon states at the outset of this "section", "[t]he predicate is the most complex part of Jarawara grammar" (95). The predicate, which is treated in chapter 4, is quite complex, and is composed of eleven slots (some with "sub-slots"), as follows: A) first pronominal slot, B) second pronominal slot, C) prefixes, D) verb root, E) auxiliary root, F) miscellaneous suffixes, G) tense-modal suffixes, H) third pronominal slot, I) secondary verbs, J) mood suffixes, and K) post-mood suffixes.
While this system already seems somewhat complex, it is in fact even more complex, in that the items that go into many slots are more numerous (or at the very least, different) than those in most the languages that most of us are probably familiar with. For example, slot G includes nine tense-modal suffixes, but of these no fewer than six are different past tense markers.
To briefly mention a few of the interesting phenomena going on in Jarawara, the first has to do with pronominal slots. First, in fact, is that there is a fourth pronominal slot, in that slot C) includes first and second person singular subject prefixes. For Jarawara, the first pronominal slot is for the O argument, and the second pronominal slot is for the S or A arguments (or the copula subject) ... a seemingly straightforward accusative system. Which of the two previous pronominal slots the item in the third pronominal slot repeats is, however, much more complexly determined, and unlike the first two pronominal slots, is not obligatory. It is in fact, too complex to get into here.
Another interesting morphophonological phenomenon has to do with gender agreement. Since nouns in Jarawara are masculine or feminine (which is the unmarked gender, and the gender of all pronouns for the purposes of agreement), there is gender agreement in the predicate, which involves (generally) final vowel alternation. The interesting thing is that gender agreement can occur in slots D, E and F (when word-final) and in slots G, J, and K (regardless of position), and there are a number of different alternations. If the alternation occurs in slots D through GIa, then the masculine is higher and fronter; if in slots GIb, G-2, J, or K, then the opposite is the case. And agreement alternation is in the negative suffix. If it follows a tense- mood suffix or a secondary verb, then it is in slot K and acts accordingly. If no such item precedes, then it falls back to slot F6d, and is in the area where the other type of vowel-alternation occurs.
Also of special interest are the four auxiliaries (which provide a place to attach the various suffixes, and, with non-inflecting verbs, a place to attach the prefixes and suffixes that would normally go on an inflecting verb), and an extensive series of more than fifty "miscellaneous" suffixes, which are organized into six echelons (each averaging three slots), both of which are treated in chapter 5, one of the longer single chapters. Assignment of a given suffix to an echelon "slot" is generally determined by the relative order when suffixes occur together in a single predicate. Items within given echelons are not, generally speaking, semantically related; for example, second echelon slot F2 includes three suffixes: 1) "two participants, a pair", b) "in the morning, tomorrow", and c) "do first". Their morphosyntactic properties do, however, tend to be similar; for example, with all three of the F2 suffixes, the -na- auxiliary, when present, is retained.
Although I can't get into any of the details, chapter 6 deals with the tense-modal system (which includes eleven items, half of them providing information about evidentiality), chapter 7 with a miscellany of elements that close out the predicate (secondary verbs, mood and negation), chapter 8 with verbal derivation, and chapter 9 with verbal reduplication (of which there are three kinds, each with its own semantics, and they can co-occur).
Chapters 10 through 12 (280-376) treat the noun phrase, and related issues. Chapter 10 deals with general noun phrase structure. Of interest is the fact that, of the two genders, feminine is the unmarked member, and, for example, if any element in the predicate agrees with the subject and that subject is a pronoun, then agreement is feminine. On the other hand, as in many languages, women can be accorded masculine gender as a sign of respect. And such traits as gender, number, animacy, etc. are not generally marked in the noun phrase itself, but are only manifest by agreement in the predicate. Also, while determination of the head of a noun phrase might be straightforward in most languages, one interesting phenomenon is noted for Jarawara: for example, while the noun "mano" 'arm' is masculine, in the sentence (part of example 10.10) "o-mano koma-ke" 'my arm is sore' agreement is feminine, due to the inherent feminine gender of the pronominal possessor "o-" 'my' (here we have to do with inalienable possession), and indicated by the form of the declarative suffix "-ke" (which would be "-ka" if masculine). Chapter 11 goes on into greater detail about possessed nouns, and chapter 12 treats demonstratives and related forms.
Chapters 13 through 19 (377-487) deal with a wide range of clause types, which include in addition to copula clauses (chapter 13), the verbal main clause (chapter 14), complement clauses (chapter 17), dependent clauses (chapter 18), and nominalized clauses (chapter 19). Chapter 15 is a discussion of commands and questions, two seemingly unrelated topics, which also seem a bit out of place in a sequence of chapters dealing NOT with morphology per se, but with clause and construction types.
Of these chapters, chapter 16 (a "mere" thirty pages and so NOT the longest chapter in the book) is called "a central chapter in the grammar of Jarawara" (417). To a certain extent this might be viewed as Dixon, author of Ergativity (1994), going on once again about his pet topic. But in fact, it is much more. In general, Jarawara can be said to have two types of verbal-clause constructions: A-constructions (where the A(gent) argument is used as a pivot), O-constructions (where the accusative or O(bject) argument serves that function). There are multiple contrasts between the two constructions (as summarized in table 16.1 on page 420), but no single morphological or syntactic marker of the difference. Some of the features are cut and dried (e.g. the prefix "hi-", used where both A and O are third person, is not used in the A-construction); others are mere tendencies (e.g. the preferred order of A and O are statistically in opposition: 85% A-O in A-constructions, 73% O-A in O-constructions). From the overall list, though, it is generally possible to determine whether a given clause is an A- or O-construction. In fact, we do get minimal pairs: for instance, the two examples in 16.13: A-construction: "mee o-wa.katoma-ra o-ke" versus O-construction "mee o-wa.katoma-ra-ke", both meaning 'I stared at them', but distinguished by the absence of the repeated "o- " 'I' in the third pronominal slot in the second O-construction example. The structure of A- and O-constructions is analyzed in great detail later in the chapter, but for the general reader, perhaps the most interesting part of the chapter is where Dixon gives examples of the discourse roles of these constructions. Sixteen short passages are presented, and the role of the constructions to provide a pivot in discourse is demonstrated. Although I can imagine that it is not a totally uncommon occurrence for a given clause to be indeterminate out of context (though I could not find any clear examples in a quick search), context would most likely disambiguate.
These are followed by a series of "minor" and miscellaneous chapters. This includes a discussion of peripheral markers in chapters 20 (488- 497) and 21 (498-508), and a relational noun in chapter 22 (509-516), all of which in more traditional grammars might be dealt with in a chapter on postpositions and conjunctions. Of interest is the presence of a very generic postposition "jaa" with a full range of locative, allative and ablative uses with noun phrases, and temporal, conditional and causal senses nominalized clauses. Context (and the semantics of the verb) disambiguates.
Chapter 23 returns to a last type of construction, the list construction. It is similar to the preceding three chapters in that it is a construction type which occurs either with noun phrases or with clauses. The interesting feature of this construction with clauses is that otherwise inflecting verbs do not inflect as usual.
Chapter 25 (532-538) returns to morphology, in this case derivational morphology.
Chapter 26 (539-565), the penultimate chapter, deals with a range of topics in semantics; in many ways, it is a hodge-podge chapter. For nouns, notable are the concreteness of the lexicon, the relative lack of abstract nouns, and also a relatively small set of generic nouns. For verbs, things are bit more interesting: there is a good deal of suppletion. For example, corresponding to English "lie", Jarawara distinguishes between whether the thing doing the lying is singular, dual or plural. In addition, if it is singular (but NOT in the dual or plural), it distinguishes between whether it is lying on the ground, on a raised surface or in the water. While this is perhaps an extreme case of suppletion, the distinction of singular versus plural is reasonably widespread: for subject argument for intransitives, and for the object argument for transitive verbs. This is basically an ergative alignment (S=O). In this chapter, Dixon also examines the sub-classification of verbs into inflecting versus non-inflecting, and into intransitive versus intransitive, plus S=A and S=O ambitransitive sub-classes. It is a remarkable feature of Jarawara that of the last four classes, S=O ambitransitives are the second largest group (28% of verbs). This leads to a discussion of semantic roles and syntactic functions, and finally to an interesting typological characterization of Jarawara: Jarawara is an action-oriented (as opposed to an argument-oriented) language, and the relationship between semantic roles and syntactic functions is much more fluid. (For the record, the example Dixon gives of an argument-oriented languages is Dyirbal, another language for which he wrote THE descriptive grammar. English, Dixon says, falls in the middle.) All this leads to where, for example, a verb mii -na- 'shit' can be used intransitively (ex. 26.26), transitively with a normal accusative-like object argument ('blood' in ex. 26.27) or with a locative- like object argument ('top of the table' in ex. 26.28); structurally the last two are the same.
For the historical linguist, and for anyone who has made a careful reading of this book and noted all the instances where something quirky is going on in present-day Jarawara (either in and of itself, or when compared to the other modern Madi dialects), chapter 27 ("Prehistory", pp 566-582), is one of the most interesting. In this chapter, Dixon proposes a sequence of ordered diachronic developments which account for those points.
As mentioned above, the 27 chapters are followed by appendices, which include 3 "short" texts (averaging 8'42" delivery time; 583-611), references (612-614: Arawan languages have not been so widely studied as to have produced a long bibliography), a vocabulary of the approximately 1200 words occurring in examples, texts and discussions in the book, a list of affixes (630-632) and finally an index (633-636).
This is, in many ways, a near-perfect model of how a descriptive grammar of ANY language should be presented. Although the grammatical treatment and exposition are excellent, for me, one of the best parts of the book are the examples and how they are presented. Examples are rarely given singly to illustrate a grammatical point. They are most often presented in pairs; sometimes contrastive pairs, sometimes complementary pairs, but always jointly to better explicate the issue at hand. And when two do not suffice, Dixon gives a third and fourth, and then a cross-reference to further examples of the same or related phenomena given elsewhere in the book. In addition, examples are almost NEVER given without (detailed) context. As Dixon says (433): "the full significance of a clause in Jarawara can only be appreciated if it is considered in a discourse context". While Dixon makes it clear that this is the case for Jarawara, I think it should be clear that this is equally the case in EVERY language. While examples can be given that are more or less clear more or less independent of context, such are rarely really interesting examples that get to the heart of how the given language really works. All descriptive grammars should follow Dixon's lead and provide the illustrative context for the sentences they present. (This would also, no doubt, lead to a reduction the number of so- called "ungrammatical", starred sentences cited; in my experience many (perhaps most) such sentences simply show the linguist is too unimaginative (and too lacking in a convincing command of the creative potential of the language they purpose to analyze) to come up with the perfect context in which such sentences could in fact be used and make perfect sense.) And Dixon's exposition of the contexts provide the added bonuses of being an introduction into the cultural and cognitive world of the Jarawara (which, Victorians beware, includes not infrequent references to bodily functions and bodily fluids), and often explicative of how one does linguistic field work ... and are also just plain amusing anecdotes. Who said a descriptive grammar should not also be a good read?
All this said, a couple aspects of the book were disappointing.
The first one, a mere pet peeve, is the consistent intrusion of elements of analysis into the top line of the multilinear presentation of example sentences. While it is common, and I believe quite acceptable, for elements of phonemic analysis to be incorporated into this line, and even morphemic segmenting (IF the language allows you to do that without getting too far from the surface form), ALL other additions should be relegated to the second line (the gloss line) ... and an additional line if multiple layers of analysis are required or preferred. However, Dixon, in a work that is otherwise a model of descriptive and theoretic "neutrality", consistently and constantly inserts ONE element of analysis into the first line: the assignment of S, A and O roles. It does not, I think, really get in the way; in fact, some may argue that it adds clarity (perhaps Dixon's motivation). It is, however, I think NOT good practice.
On the much more serious side, the index is EXTREMELY disappointing (without a doubt the MOST disappointing thing about the book) ... not quite useless, but far less useful than it could and SHOULD have been. Produced in the age of computers, one would think that a more or less thorough index could be provided with minimal effort; for a book of more than six-hundred pages, it really is a necessity. The index found on pages 633-636 is, however, extremely superficial and incomplete. When, for example, a mention on p 571 of how Banawa (another Madi dialect) speakers exaggerate sentence- final nasalization when they imitate Jarawara speakers, I had forgotten Dixon's ever mentioning final nasalization in Jarawara -- since none of the examples have it indicated. Yes, the index, when consulted under "nasalization", did in fact lead me back to the original statement (with three examples where nasalization IS indicated) on page 28, but the instance on p 571, more rememberable because of the context, is itself NOT listed in the index under nasalization. It can be found by following ALL the leads given under the heading Banawa dialect, where it seems to fall under the general sub-entry "grammar", where the reference is to a longish 12-page section which mentions a score of various phenomena, none of them remotely phonetic in nature (the subentry heading IS grammar), EXCEPT the one in question. While reading the book and trying to follow up links between related topics, in general my own memory of what I had read and where was much more helpful than the index ... and that simply should NOT have to be the case.
In addition, there are a few things which I would have found desirable, though their absence does not mark Dixon's book as falling in any way short of what passes as acceptable; they are merely among my list of desiderata for a more ideal descriptive grammar. I will not give a complete list, just one example. Although the vocabulary in fact appears to be complete (at least in the sense of accounting for all the Jarawara words that occur in this book), and the glosses are acceptable (though not "complete" dictionary entries, which Dixon readily states), and though sufficient grammatical information is probably given for all items, only about 2% of the entries have any cross reference to a sub-section in the body of the text where they are discussed. While cross-referencing every item to every possibly connected discussion in the book is too much even for me to ask, it would be instructive, and helpful, if there WERE a cross reference for EVERY item to its occurrences not only in grammatical discussion but even more so in examples. Whether it is an issue of semantics (and range of usage) or grammar (and range of usage), being able to see and compare different occurrences of the same items is something any inquisitive reader would want to do. Within the body of the grammar, as mentioned above, Dixon DOES have frequent cross- references to further examples. It would just be more helpful if this information could also be found based on individual lexical items. Of course for items occurring hundreds of times, there need not be a list of ALL the citations. However, I would think up to five occurrences for each and every item would not be unreasonable in this age of computer-generated cross-references.
Lest I close on a negative note about a book that I feel extremely positive about, and to make one last attempt to interest anyone not yet so inclined in this fascinating book and this fascinating language, I will give one of MY (dozens and dozens of) favorite examples: the Jarawara description of "whiteout" correction fluid is: "jama hani mese ke-teha-ni fawa", literal morpheme-by-morpheme translation: 'thing(f) writing(f) top.surface.of applicative-put.on- immediate.past.noneyewitnessed(f) disappear' ... an apt description indeed!
Dixon, R.M.W. (1994) Ergativity. Cambridge: Cambridge University Press.
Dixon, R.M.W. (2000). Categories of Noun Phrase in Jarawara. J. Linguistics 36, 487-510.
Everett, Daniel L. (n.d.) Optimality Theory and Arawan Prosodic Systems: Chapter One: Prosodic Levels and Constraints in Banawa and Suruwaha. http://ling.man.ac.uk/info/staff/DE/suruwaha.pdf
Everett, Daniel L. (1999) Syllable Integrity. WCCFL 16 Proceedings, CSLI/University of Chicago Press. http://roa.rutgers.edu/files/121- 0496/121-0496-EVERETT-0-0.PDF
Everett, Daniel L. (2003) Iambic Feet in Paumari and the Theory of Foot Structure. Linguistic Discovery 2/1:22-44. http://journals.dartmouth.edu/webobjbin/WebObjects/Journals.woa/1/x mlpage/1/article/263
Salzer, Meinke & Shirley Chapman (1999). Dicionario Bilingue nas Linguas Paumari e Portuguesa. http://orbita.starmedia.com/~i.n.d.i.o.s/paumari/portpmdc.htm
Vogel, Alan R (2003) Jarawara Verb Classes (Univ. of Pittsburgh PhD dissertation). http://www.sil.org/americas/brasil/PUBLCNS/LING/JAVerb.pdf
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Michael W Morgan has a doctorate in Slavic Linguistics (with a minor in Indo-European Studies) from Indiana University (1990), and his current research interests focus on historical, comparative and typological topics, especially but not solely relating to sign languages, and also various issues in a Sign Theoretic (aka Jakobsonian-van Schooneveldian) approach to linguistic analysis. He also currently teaches at Kobe City University of Foreign Studies.