Symposium on Endangered Languages of Amazonia Talk Summaries

The following is a summary of talks given at the “Symposium on Endangered Languages of Amazonia”, held at the University of Texas at Austin on February 16-17, 2007. That’s now a while ago, but I figure this is a good a place as any to archive these summaries.

I apologize in advance for any inadvertent misrepresentations due to the need for brevity, or due to misunderstandings on my part. I have suppressed most citations.

Nilson Gabas Junior (Keynote). Ideophones and evidentiality in Karo (Tupi).

Gabas presented two talks: one on ideophones and a second on evidentiality in Karo, a Tupian language spoken in the Brazilian state of Rôndonia, near the border with Bolivia. Karo has approximately 150 speakers living in two villages, and the language is currently vital.

Ideophones in Karo

Gabas argues that ideophones forms a distinct, open word class in Karo, on par with nouns, verbs, adjectives, etc. In Karo, ideophones are semantically similar to verbs, in that they convey actions or movements, are morphologically similar to particles in that they display no internal structure, and take no inflectional or derivational morphology, as well as to adjectives and verbs, in that they can take an adverbializer clitic. They display a similar syntactic distribution to adverbs. Phonologically, Karo ideophones are phonologically exceptional in exhibiting a doubly-articulated alveolar-labial nasal.

Syntactically, ideophones appear in all clause types, and are usually acompanied by an auxiliary, except in the case of negative imperative clauses, nominalizations, and time subordinate clauses. In any clause type, however, an ideophone or an ideophone + auxiliary combination may replace a verb to indicate the action of the clause. Note that ideophones may also co-occur with verbs of the same or similar meaning.

Gabas argued Karo ideophones display some obvious onomatopoeic features, but do not exhibit the one-to-one sound-meaning correlation characteristic of sound symbolism. Moreover, some ideophones have meanings that correspond to regular Karo verbs, such as `look’, `miss’, `throw’, and `disappear’, and can replace them via the constructions mentioned previously.

Gabas has found that Karo ideophones are more common in narratives than in everyday conversation, and he associates them with `expressiveness’ and the notion of involvement stemming from Tannen, Chafe, and others.

Gabas closed with comments on the typology of ideophones, remarking that there is not yet an ageed upon definition of ideophone, other than its basis in onomatopoeia and sound iconicity. Gabas proposed that a cross-linguistically valid definition of ideophones may be prototype-based, involving central members with the following characteristics: sound iconicity; the use of sounds rare or absent in other categories; the use of reduplication to indicate iteration, progressive, or continuative aspect; syntactic distribution different from other word classes and inclined towards clausal edges, and their own intonational unit; and an association with narrative peaks.

Evidentiality in Karo

Gabas described a set of eleven particles in Karo that would be considered evidentials under the `broad’ sense of evidentiality (i.e. source-of-information, reliability (i.e. epistemic modality), and possibly, mirativity). These particles are not obligatory for morphosyntactic wellformedness (i.e. do not form an inflectional category), but are very common in Karo discourse. Karo evidentials appear either in sentence-final position or NP-final position, which are distributional characteristic that distinguish them from adverbs.

Karo evidentials appear in declarative, interrogative, focal, negated, predicate adjective, and predicate nominal clause types, but do not appear in imperative clauses, or any future clauses. More than one evidential may appear in a clause, providing their meanings are compatible (e.g. source-of-information and reliability), but Gabas pointed out that fully determining the co-occurence patterns in this intricately related set of particles remains an issue for further research.

The semantics of Karo evidentials appears quite complicated and subtle. Gabas identifed five clear source-of-information evidentials: visual, non-visual, hearsay (reportive), and two inferentials distinguished by whether or not the proposition is counter-expectational or not. Another particle is used to indicate `assumption’ from patterns of previous activity (this could also arguably be considered an inferential based on habitual behavior). In addition, Gabas identified three particles that indicate high, medium, and low reliability. Finally, Gabas identified two evidential particles identified as `counter-expectation’ and `subject confirmation’ which indicate that the proposition does derive from *some* source of information (unspecified as to its nature), and that the proposition based on that source of information is, respectively, in fact counter to expectation or other information sources, or confirms a possible proposition regarding the subject of the clause.

Gabas closed his talk with a review of the typological discussions of evidentiality by Dixon and Aikhenvald (1998) and Aikhenvald (2004), a review of the areal distribution of evidentiality in the Amazon (which shows a greater tendency towards larger numbers of evidential distinctions in the Northern Amazon than in the Southern Amazon), and mention of other languages, apart from Karo, in the Southern Amazon and in the Tupian family that exhibit evidentiality (Southern: Kamayura, Nambiquara, Madi, and Panoan languages; Tupian: Gavião and Sakuriabat)

Lev Michael. Mood and Negation in Nanti (Kampan, Arawak).

Nanti is an Arawak language of the Kampan sub-family, which includes languages such as Asháninka, Ashéninka, and Matsigenka, and is spoken in lowland southeastern Peru. There a 500 speakers of Nanti, and the language is extremely vital, most Nantis being monolingual.

Michael presented a description and analysis of the mood system of Nanti, focusing on the interaction of mood and negation. In Nanti, mood is an inflectional category that exhibits a binary contrast between `realis’ and `irrealis’. Michael reviewed the debate on the validity of realis and irrealis as typological grammatical categories (Marianne Mithun, Talmy Givón, Jennifer Elliot, William McGregor and Tamsin Wagner come down in favor of it, Joan Bybee, Revere Perkins, and William Pagliuca reject it), and argued that the concepts appear to have a great deal of descriptive utility, and that they should not be abandoned at this time.

Returning to Nanti, Michael showed that in affirmative main clauses, realis marking is found in non-imperative clauses with non-future temporal reference, whereas irrealis marking is found in imperatives, counterfactuals, conditionals, and clauses with future temporal reference. Michael pointed out that this patterning is common for languages with a realis/irrealis contrast. In Nanti, irrealis marking is also found in non-finite complements, temporal coordination clauses, and purpose clauses.

In negated clauses, mood-marking is more complicated. First, Nanti exhibits two negators, te- and ha-. The former is used to negate otherwise semantically realis clauses, and the latter is used to negate otherwise semantically irrealis clauses. That is, te- is used to negate non-imperative clauses with non-future temporal reference, whereas ha- is used to negate imperatives, counterfactuals, conditionals, and clauses with future temporal reference.

Te-negated clauses take irrealis marking, and ha-negated clauses take realis marking. Irrealis marking of negated clauses is common in languages with realis/irrealis systems, making the behavior of te-negated clauses unremarkable. However, the fact that ha-negated clauses are realis-marked is unexpected. If anything, these clauses seem like they are, semantically speaking, doubly irrealis.

Michael then argued that in fact, it is not only the mood-marking of ha-negated clauses that seems `backwards’, but also that of te-negated clauses. Michael argued that in order to understand Nanti mood marking in negated clauses it is necessary to go back to the notional definition of `realis’ given by Mithun, and note that part of the definition refers to `knowability through direct experience’. On this basis, Michael argued, te-negated clauses (which necessarily have non-future temporal reference) can be considered to be semantically realis, just like their non-negated counterparts, since it is possible to directly experience the non-occurrence of something in past or present. If this is given, both te-negated and ha-negated clauses exhibit mood marking opposite to what one would expect on notional grounds.

Inspired by certain analyses of counter-intuitive mood marking of Mithun’s, Michael proposed that mood-marking discrepancy exhibited in Nanti negated clases can be resolved by considering the scopal interaction of negation and mood in Nanti. Michael’s argument involved three steps: first, we accept work that has argued that negation can have scope of mood in particular languages and assume that this is the case in Nanti. Second, we accept work that shows that in certain languages the total modal value for a clause is the result of the interaction of negation and mood values. And third, we explictly formulate the relationship between relationship between realis and irrealis as: irrealis = NOT realis, and realis = NOT irrealis.

From these starting points, Michael argued, we can easily explain `backwards’ marking of mood in Nanti negated clauses. In te-negated clauses, which are semantically realis according to the above argument but are *marked* irrealis, we would analyse the total modal value of the clause as NEG(irrealis) = realis. In other words, the negation `flips’ the irrealis modal value marked on the verb, resulting in overall realis modal value for the clause. In ha-negated clauses, which are semantically irrealis, but are marked realis, the total modal value of the clause is given by NEG(realis) = irrealis. By this argument, then, the overall modal value of the negated clauses are what one would expect on notional grounds.

I-Wen Lai. Conditionals and Counterfactuals in Iquito

Iquito is a Zaparoan language spoken in northern Peruvian Amazonia, near the present-day city of Iquitos. The language is moribund, with only 25-40 remaining speakers, all over the age of 60.

Lai presented a description of the morphosyntax of conditional and counterfactual constructions in Iquito and an analysis of the morphosyntactic composition of conditional and counterfactual semantics in these constructions. The analytical framing of Lai’s analysis involved a contrast between languages in which conditionality and counterfactuality is conveyed by implicature, on the basis of tense/aspect/mood (TMA) morphology, and languages in which conditionality and counterfactuality is expressed by morphology dedicated to this meaning. Significantly, in languages of the former type, like English and Modern Greek, TMA morphology does not receive its normal interpretation in conditional and counterfactual constructions (resulting on so-called “fake” TMA), while in languages of the latter type TMA morphology still receives its normal interpretation (so-called “real” TMA). Lai argued that Iquito is a language of the latter type.

Lai then presented descriptions of Iquito non-counterfactual conditional constructions (habitual patterns and situations, epistemic conditionals, and future conditionals) and counterfactual conditionals. Non-counterfactual conditionals use the bipartite non-assertive morpheme `sa-cari’ and realis word order in the antecedent (In Iquito, realis and irrealis mood are distinguished by word order; in realis order, no elements intervene between the subject verb, whereas in irrealis order, other elements, such as object NPs, determiners, and adverbs typically intervene between the subject and the verb). Tense and aspect appear between the two parts of the bipartite non-assertive morpheme.

The habitual conditional (e.g. If he sees you, he does not look for you) is characterized by imperfect aspect and realis order in both the antecedent and consequent clauses. Epistemic conditionals (e.g. If he drank it earlier today, he is probably recovering now) also employ realis order in both the antecedent and consequent clauses and always employs the epistemic adverb `cuuta’ (perhaps) in the consequent clause. Tense and aspect morphology varies with the particulars of the situation being described. Future conditionals are characterized by realis order in the antecedent and irrealis order in the consequent. Aspect used in the antecedent varies with the particulars of the situation being described; the consequents always employ one of a large set of aspect morphemes which incorporate perfective meanings. The particular choice of perfective depends on the remoteness of the future being indicated.

Counterfactual conditionals are characterized by the use of either the non-assertive `sa-cari’ and and realis word order in the antecedent *or* the counterfactual morpheme +t+ (+ = high central unrounded vowel) with irrealis order. Consequent clauses always bear the counterfactual morpheme and exhibit irrealis order. Lai then presented a large set of examples of counterfactual conditional constructions with verbs of different Aktionsart classes (statives, activities, and telics), tenses (recent past, distant past, and zero tense), and aspects (perfective and imperfective).

Lai then proceeded to the main analytical part of her talk, in which she showed that the TMA morphology in both counterfactual and non-counterfactual conditionals receives the same interpretation in these constructions as in non-conditional constructions. A pair of examples illustrates the basic point: consider the English sentence `If he had been knowledgeable a long time ago, he would be a teacher now’. This sentence is compatible in English with either perfective of imperfect readings in the antecedent clause. To make the perfective sense clear, in contrast to the imperfective sense, one would have to reword the sentence as something like: `If he had *become* knowledgeable a long time ago, he would be a teacher now.’ In Iquito, however, the perfective and imperfective senses can be distinguished by substituting perfective for imperfective aspect in the antecedent clause, a possibility ruled out in English due to the fact that TMA morphology is “fake” in English counterfactual conditionals. Lai’s final point was that counterfactual senses in Iquito are not the consequence of inferences based on TMA morphology, as they are in English, but are conveyed by the specialized counterfactual morpheme +t+.

Mily Crevels (Keynote). How a language without nominal number expresses plurality: the case of Itonama [Isolate, Bolivia].

Itonama is an isolate spoken in the northeastern Bolivian Amazonian lowlands by fewer than five very old speakers. Crevels described how number is expressed in Itonama, despite the fact that the language lacks any systematic nominal number morphology.

Crevels presented a typological overview of Itonama: Itonama is a head-marking, polysynthetic language that exhibits body-part incorporation. It exhibits VSO basic order and an inverse system in independent bivalent predicates (e.g. main clause transitive verbs). Itonama also exhibits two sets of classifiers (a verbal/deictic set and a numeral set), directionals, end evidentials.

Crevels began by showing that there is no nominal number morphology in Itonama, although there are a very small number of human nouns which mark plurality suppletively. The language does, however, exhibit two other productive resources for marking number: verbal/deictic classifiers and pluractional markers.

Crevels listed 17 verbal/deictic classifers, which encode information about animacy, orientation, and shape; there is also fluid classifier. Crucially, many Itonama verbal classifiers also encode number (classifiers for which a count interpretation is not very natural, like the fluid classifier, do not encode number). Itonama verbal/deictic classifiers are employed in existential constructions, and with predicates of possession, location, or manipulation. In existential constructions, these classifiers may appear on existential verbs, or on nouns, if a predicate nominal existential construction is employed. Otherwise, the classifiers appear on the associated verb in the corresponding construction types. In addition, these classifiers can appear with demonstratives. Classifiers are associated with the subjects of intransitive verbs and the objects of transitive verbs.

Crevels discussed four pluractional markers, a `distributive’, a `plural’, a `multiple’, and a `continuative’ . In addition, reduplication is a productive process for indicating iteration. In addition to these productive processes, certain verbs (e.g. `flee’ and `fall’) have suppletive singular and plural forms.

The distributive indicates multiple instances of the action or state indicated by the verb root, affecting or involving spatially separated (and hence, by implicature, multiple) objects. The plural, on the other hand indicates multiple instances of an action or state either affecting a single object, or multiple realizations of an action or state by a single subject in a single place. An illuminating example of the constrast between the distributive and plural involves fishing. Fishing with a fishhook (multiple throws in a single location) is associated with the plural, whereas net fishing (multiple throws in spatially separated locations) is associated with the distributive.

Crevels then turned to a discussion of distinguishing event and participant number with regard to the number-related resources she had just finished describing. The basic point of Crevels detailed discussion, which I cannot do justice to here in this brief summary, is that some resources (like the distributive and iterative) encode only event number, and convey participant number via implicature. Other resources, like the classifiers, directly encode participant number.

Crevels closed with a comparison of the Itonama number system with nominal and verbal number in other lowland Bolivian languages (Movima (isolate), Yuakaré (isolate), Mosetén (isolate), Baure (Arawak), and Yuki (Tupi-Guaraní)). Of these, Itonama is the only language which clearly lacks nominal number. In Movima, there is no marking on nouns per se, but there is on articles.

Taryne Hallet. A paradigm of event modality: the Iquito continuum.

Hallett presented a description of Iquito event modality constructions (abilitive, obligative, and imperative) and discussed the grammaticalization of obligative constructions from clauses containing abilitative verbs. Iquito event modality constructions depend on the polarity of the sentence, so Hallett distinguished affirmative from negative sentences in her discussion.

In Iquito, affirmative abilitive constructions consist of a main clause with an inflected form of the verb root /parii/ `be able’ and a non-finite complement. The only difference in negated abilitive constructions is the addition of the negation /caa/.

Iquito distinguishes strong and weak obligatives. Hallett noted that affirmative weak obligatives are formally identical to abilitive constructions (i.e. involve the verb root /parii/ in main clauses), but it is clear from example glosses that in certain discourse contexts, the construction is interpreted as a weak obligative, and not an abilitive. Negative weak obligatives are distinct from negative abilitives (and affirmative weak obligatives) in that the verb root of the main clause is not /parii/, but rather /paji/, which in non-obligative constructions is glossed as `learn’.

Strong obligative constructions are formed very differently. The affirmative strong obligative consists of an irrealis mood clause with perfective marking (/-qui/). Strictly speaking, this is ambiguous between strong obligative and simply future temporal reference, but the latter typically co-occurs with momentary aspect /-r++/ (+ = high central unrounded vowel). The negative strong obligative is distinguished from the affirmative by the presence of the prohibitive suffix /-cuma/.

The affirmative imperative is formed by omitting the subject person marker verbal clitic and any co-referential NP; the verb typically takes some form of perfective marking, including momentary aspect (which typically gives rise to inceptive readings in imperatives), and two suffixes with roughly allative and ablative senses. The negative imperative also involves the suffix /-cuma/, mentioned above, as well as subject omission.

Hallett then describes a `non-curative’ construction, which is basically a third person imperative with the added sense that the speaker will not interfere with the realization of the action in question. This construction may be used in a permissive sense to indicate that the third person referent is allowed to carry out a given action. This construction is characterized by the addition of a clause-initial particle /pa/ to a realis clause.

Hallett distinguishes the former from a first person permissive construction, which consists of an imperative construction in which the verb bears the causative derivational suffix /-t++/. Note that in Iquito the causative suffix is neutral with respect to causee desire and volition.

Having completed the description of Iquito event modality, Hallett turned to a discussion of the grammaticalization of the strong obligative construction from the future-reference construction, and the grammaticalization of the obligative construction from abilitive constructions. Hallett also pointed out the similarity between the non-curative /pa/ and the abilitive verbs /parii/ and /paji/.

Patience Epps. Verb compounding and antigrammaticalization in Hup (Nadahup, Brazil)

Hup is a language spoken in Vaupés region of the Brazilian state of Amazona, near the border with Columbia. There are approximately 1500 speakers of Hup, and it presently remains a vital language, although Epps remarked that it is possibe that Hup society is on the brink of profound cultural change, which poses significant dangers for the survival of the language.

Epps began with a review of grammaticalization, which overwhelmingly tends to be a unidirectional process of increasing morphological binding (word > clitic > affix), increasing semantic abstraction, shift from lexical to grammatical function, phonological reduction, and increasing textual frequency. Antigrammaticalization is extremely rare, but Epps argued that Hup provides a good example of this phenomenon.

In Hup, a set of verb roots have undergone a transition from verb root to suffix, e.g a verb meaning `make noise’ to a non-visual evidential suffix. A subset of these suffixes has subsequently antigrammaticalized, including the non-visual evidential, becoming an enclitic or particle.

Epps provided the following account of the grammaticalization process: Hup exhibits productive verb compounding, where multiple verb roots stack up inside any inflectional morphology (nuclear serialization). If a comound verb refers to a series of events, the roots appear in iconic temporal order. If the verb refers to a single event, the `main’ verb root is first, followed by dependent verbs that modify the main verb. Following the verb root(s) in any verb one find an `inner suffix’, with a variety of possible meanings. Synchronically, verb roots and inner suffixes are not fully distinct categories, in that there are elements which are clearly in transition between being lexical verbs and being inner suffixes. There are marginally grammaticalized verbs/inner suffixes like `play’, where V-play means `play at V, do V playfully’. These marginally grammaticalized verb roots exhibit little semantic or phonological change in comparison to the corresponding free verbs. Next on the grammaticalization continuum are elements that Epps refers to as `auxiliaries’ , such as `see’, where V-see means `try to V’. Auxiliaries exhibit significant changes in their meaning in comparison to the corresponding free verbs, may show phonological reduction, and the free verb forms may be rare or unattested. Inner suffixes, which are maximally grammaticalized, such as the frustrative suffix (developed from the verb `request, order’) or the inferred evidential (developed from a verb `be inside’) exhibit great semantic shifts, are highly productive as bound morphemes (though the corresponding free form may be rare or absent), and frequently exhibit phonological reduction. For some verbs, all three forms exist, as in the case of `make noise’ (free lexical verb) which has a corresponding modifying verb/auxiliary form `make noise doing V’ and an inner suffix form which is a non-visual evidential.

There is another set of morphemes that Epps dubs `fluid suffixes’ (alluded to above), which may appear either as an inner suffix or as an enclitic, usually with little change in semantics, as in the case of the visual evidential (whether the inner suffix form or the enclitic form is used depends on the choice of boundary suffix). Epps provides the following argument for the inner suffix being the source of the enclitic: of all the grammaticalized elements, only the inner suffixes display the phonologically reduced CV form (a consequence of a phonological process due to the adjacent vowel-initial `boundary suffix’.) Consequently, it is extremely unlikely that the non-visual enclitic (which is of CV shape) was the source of the inner suffix. Since it is clear that the ultimate source is the free verb (which has CVC shape), the only way for the enclitic to have the shape it does is if it developed from the CV-shaped inner suffix. Crucially, this is an example of antigrammaticalization, since it is an instance of decreasing binding.

Epps then presented a five stage diachronic explanation for the antigrammaticalization of the visual evidential. The key idea in the analysis involves a process by which certain erstwhile free verbal modifiers end up being reanalyzed, in the presence of particular boundary suffixes, as inner suffixes (with changes in meaning, e.g. a verbal modifier `do for a long time’ becomes an habitual inner suffix.). Consequently, these modifiers end being fluid suffixes, since they appear both as verbal modifiers outside the verb and inner suffixes. By analogy, other inner suffixes are reanalyzed as fluid suffixes, with the consequence that they can now appear as enclitics, in addition to being inner suffixes. The result is antigrammaticalization of certain inner suffixes. As Epps points out, this process has been quite productive in Hup, which apparently is otherwise cross-linguistically unattested.

Christine Beier. Putting the hunt into words: formal features of Nanti hunting stories in Montetoni (Southeastern Peruvian Amazonia)

Beier presented a description and analysis of the formal features of a Nanti discourse genre: hunting stories. Beier’s work on hunting stories is part of a broader effort to understand the non-referential components of meaning associated with the varied discourse genres.

Beier began by distinguishing speech styles from a speech (or discourse) genre. The former is defined as a type of language use attributable to single utterances, which can be characterized by a set of formal features (phonetic, phonological, morphological, lexical, etc.) which render it distinguishable from other types of speech. Speech styles can be inserted into various institutional settings and activity frames. Genres, on the other hand, is are more complex organizations of shared language use that encompass not only speech styles, but also setting, activity frame, participant structures, and other factors.

Hunting talk is a speech style principally employed by adult men, and indeed, the social transition between adolescence and adulthood for young mean is closely tied to their mastery and public performance of this speech style. Spatially, hunting talk is mostly employed in cooking huts, either during inter-household visits by men, or during manioc beer feasts.

The genre of hunting stories emerges when interactants actively orient themselves towards the sustained use of hunting talk and cooperate to build a narrative out of one or several hunting stories. Hunting stories emerge from the successful deployment of sequences of hunting talk, which are ratified and sustained and sustained largely through appropriate backchanneling and commentary from multiple participants.

Hunting talk is distinguished by its referential content, intonation contour, vowel lengthening, rhythm, rate of speech, and use of parallelism. Hunting talk has a distinctive intonation contour. In comparison with normal conversational talk, which has a fairly flat contour and then drops over the last syllable or two, hunting talk starts at a relatively high pitch, then drops melodically and dramatically across the first vowel (which is typically lengthened), stays at a low pitch, and then rises over to the last one or two syllables back to the iintial pitch. Beier played audio examples exemplifying this and showed pitch track data confirming her observations. As already intimated, the first vowel of a word uttered using the hunting talk style is typically lengthened, and the overall tempo of the utterance is faster than that of normal conversational talk. Just as some vowels are lengthened, the remaining vowels in an utterance are compensatorily shortened, creating a characteristic rhythm (Beier played additional audio examples to exemplify this). Finally, in any given strip of hunting talk, particularly relevant words tend to be repeated in this context of this rhythmic speech, setting up a parallelistic organization to a strip of talk (more audio examples were presented to illustrate this).

Beier also mentioned that there are other features characteristic of hunting talk of which she omitted lengthier discussion for reasons of time. These include increased nasalization, pervasive centralization of vowels, and increased breathiness.

Beier then discussed the non-referential communicative content of hunting talk. Beier explained that the distinctive formal features of Nanti hunting talk serves to constrain interpretations of the referential content of the talk, essentially serving as a metapragmatic marker that the talk in question is hunting talk, and all that entails socially and interactionally. Specifically, the formal features index the fact that the speaker is taking a particular kind of narrative stance, in which he is playing up a particular event for narrative purposes, and also indexes as appropriate a interactional pattern of multi-party talk and significant overlapping among participants, which is generally not considered appropriate in Nanti communicative interaction.

Beier concluded with a discussion of why descriptive linguists and linguistic anthropologists should care about the description and analysis of discourse genres. First, discourse styles and genres are phenomenologically significant to speakers of languages, consequently they should to language scientists as well, since such a robustly patterned aspect of language should form a part of the complete description of a language. Second, although the basic tools for grammatical description are relatively well-developed, those for the description of discourse styles are not. This means that this important aspect of language description largely remains the domain of impressionistic descriptions. Further development of tools to render such descriptions more rigorous and systematic depends on greater empirical engagement with these pervasive phenomena.

Cynthia Anderson. Hands, fingers, feet, and toes? Oh brother. The numeral system in Iquito.

Anderson presented a description of the Iquito numeral system and the implications of the certain features of this numeral system for language contact between Iquito and other Amazonian languages. Anderson characterized the Iquito numeral system as quinary (base five), with four lexicalized cardinal numerals (1-4), and set of highly variable expressions based on counting fingers, toes, hand and feet for higher numbers (5+). The lexicalized numerals 2, 3, and 4 display animacy agreement, although 1 does not (more generally in Iquito, only plural-marked elements display animacy agreement).

The numeral system is based on three numeral forming strategies, which presumably correspond to different historical stages in the development of the numeral system. 1 and 2 are plausibly reconstructible in proto-Zaparoan. 3 and 4 are different, not being reconstructible in Proto-Zaparoan, and at six syllables in length, are uncharacteristically long for monomorphemic words in Iquito (3= s++saramaj+taami, 4=suhuaramaj+taami). Moreover, further examination suggests an etymology. The first two syllables of 3 and 4 correspond to the words ‘bad’ (s++sa) and ‘good’ (suhua), respectively. The common remainder consists of a participialization of the verb root `to have a sibling (of the same sex)’ (aramaj+) with the instrumental/comitative suffix -ta. These numerals are examples of a `fraternal’ strategy.

5-20 are expressed by nonce expressions based on a body-part tally system, begining with a number of terms for `one hand’ and successively adding fingers, toes, and entire hands or feet, as needed.

Anderson then returned to the areal implications of the fraternal system for 3 and 4. Fraternal systems are typologically rare, but are common in an area of the northwest Amazon centered on the Vaupés region. In that region fraternal numeral strategies have, according to Epps, diffused from the Tucanoan languages to Tariana (Arawak) and the Nadahup languages. Further to the south, in the Peruvian/Columbian border region, Bora, Witoto, and Yagua display signs of fraternal numeral strategies. Iquito thus becomes the southwesternmost language displaying this areal feature. Anderson noted that Iquito formerly bordered on Yameo (now extinct), a language of the Peba-Yaguan family, which may have been the source of the fraternal strategy in Iquito.

Brianna Rauschuber. Tonal accent in Iquito.

Rauschuber presented a description of the stress system and tonal accent system found in Iquito nouns and adjectives. Rauschuber began with a discussion of tonal accent, which is an `abstract’ mark on a given syllable that gives it greater prosodic salience than sourrounding syllables. A tonal accent system is one in which abstract tonal accent marks correspond to a specific tonal realization across all accented tone-bearing units (TBUs). Rauschuber also discussed more recent work by Hyman that casts doubt on `pitch accent’ systems, which are frequently understood as based on tonal accent, as a coherent type of prosodic system distinct from stress systems and tonal systems. Rauschuber followed the position that pitch-accent systems are essentially tonal systems which make minimal use of tonal contrast (making Iquito an instance of a pitch-accent system).

Rauschuber then turned to a description of Iquito tonal accent. Iquito nouns and adjectives bear a single audible pitch peak per word, which Rauschuber takes to be the reflex of pitch accent. Some fraction of Iquito nouns and adjectives must be lexically marked for tonal accent, because of the existence of minimal pairs that are distinguished solely by the position of tonal accent. Rauschuber also argued that the TBU in Iquito is the mora, based on the existence of contrasting rising and falling tones, which are only found on long vowels.

Rauschuber argued that the position for non-lexical tonal accent is the penultimate mora of a word, based on evidence from morphologically complex words. When suffixes are added to roots with penultimate tonal accent (e.g. cúni `snake’), the tone shifts to the penultimate position of the morphological complex word (e.g. cuní-hua `snakes’). When suffixes are added to words with tonal accent in other positions, this shift does not occur (e.g. táasa `basket’, táasaca `baskets’).

Another relevant piece of evidence involves the first person subject marker quí-, which bears the high pitch for most words to which it is prefixed (e.g. quí-cuni `my snake’), but not in all words , since some retain pitch on the same syllable as in the correponding morphologically simplex form (e.g. qui-táasa, `my basket’). As it turns out, words that display pitch mobility with respect to suffixation also display pitch mobility with respect to quí- prefixation.

Rauschuber observes that, in cases in which the accent position varies with the morphological environment, the location of pitch in morphological simplex forms (e.g. cúni, `snake’) is predictable: it falls on the penultimate mora. Accent position is not predictable for words in which the accent position does not vary according to morphological environment, and hence must be lexically specified.

Rauschuber then turned to stress in Iquito nouns and adjectives, showing that the position of primary stress is predictable on the combined basis of wordshape and tonal accent. Relevant diagnostics for stress include lengthening of syllables bearing primary stress in emphatic speech and the centralization, devoicing, or omission of non-stressed vowels.

In words that do not bear a lexical tonal accent, primary stress falls on the penultimate syllable, if the final syllable is light, or the penultimate syllable, if the final syllable is long. As Rauschuber observed, this is consistent with the right-to-left parsing of words by moraic trochees. In words that bear a lexical tonal accent, primary stress fall on the same syllable as the tonal accent. Iquito consequently displays features of a tone-driven stress system, since stress is attracted to lexical tonal accent, as well as features of a stress-driven tone system, since tone coincides with metrically-assigned primary word stress.

Spike Gildea and Flávia Castro Alves (not present at symposium) (Keynote). Nominative-Absolutive alignment in Cariban and Jê description and reconstruction.

Gildea began by positioning Nominative-Absolutive systems in a typology of alignment. Nominative-Absolutive alignment systems are ones in which S arguments are simultaneously coded as, or treated as, P arguments (`absolutive’) and as A arguments (`nominative’) by different elements in a clause. This kind of alignment pattern should be distinguished from Split Intransitive systems (aka Stative/Active, Split-S, or Fluid-S systems), in which an S argument is treated *either* as an A or as a P, depending on the semantics of the verb or, less commonly, discourse pragmatics.

Gildea presented data on Nominative-Absolutive alignment in two Cariban languages (Panare and Katxuyana) and three Jê languages (Apinajé, Suyá, and Timbira), and then discussed the historical process by which which this system arose in these languages.

Panare exhibits a complex alignment system involving the following subsystems, which are conditioned by tense-aspect inflections: Split-S/Inverse, Nominative-Accusative, Passive/Ergative-Absolutive, and Nominative-Absolutive. Nominative-Absolutive alignment is conditioned by Future, Desiderative, and “Nonspecific Aspect”. Focusing on the Nominative-Absolutive context , absolutive marking is realized via person-marking verbal prefixes, while nominative marking is realized via agreement on auxiliaries and an optional free pronoun.

Katxuyana also displays several tense-aspect splits in its alignment system: Split-Intransitive/Hierarchical (everything but distant past and imperfective), Ergative-Absolutive (one of several pasts – due to contact with Tiriyó), and Nominative-Absolutive (imperfective). In the Nominative-Absolutive context, absolutive marking is realized by verbal person-marking prefixes, and nominative marking by free pronouns.

Apinajé generally displays Split-Intransitive alignment, with Ergative-Absolutive alignment conditioned by prospective aspect, and Nominative-Absolutive conditioned by progressive and continuative aspect, and negative clauses. In these latter contexts, absolutive marking is realized by verbal person-marking prefixes, and nominative marking by free pronouns.

Suyá displays Nominative-Absolutive marking in negated and future tense clauses, in which nominative marking is realized by free pronouns and absolutive marking by person-marking affixes.

Timbira generally displays Split-Instransitive alignment, with Ergative-Absolutive alignment in the past tense, and Nominative-Absolutive alignment in a large set of clause types headed by non-finite verb forms with post verbal auxiliaries. These include clauses with any of four evaluative modes, four iterative aspects, four continuative aspects, four perfect aspects, and the non-past negative. In Nominative-Absolutive clauses, nominative marking is realized by free pronouns and absolutive marking by person-marker affixes.

Having completed this empirical introduction, Gildea then set out a historical explanation for the development of Nominative-Absolutive alignment in the Cariban and Jê languages. There are two key observations. First, ergative case-marking and absolutive person-marking are both present in subordinate clauses in both language families. And second, the general process by which new, complex TAM distinctions develop involves the reanalysis a complex clause (e.g. a verb with a non-finite complement) as a simple main clause with a complex predicate consisting of an auxiliary (the former main verb) and a main verb (the former non-finite complement). The result is nominative marking on the auxiliary (or associated with the auxilliary) and ergative-absolutive alignment on the new main verb.

The other element in the development of nominative-absolutive alignment is the loss of ergative marking on the main verb. This the consequence of a process of NP equi-deletion in the Cariban and Jê languages: when the S of the matrix verb is co-referential with the A of the subordinate clause, the A is deleted. This leaves nominative marking in the matrix clause and absolutive marking in the subordinate clause. When the reanalysis described above takes place, the result is nominative-absolutive marking in a single clause.

Gildea then turned to consider the status of nominative-absolutive alignment in the typology of alignment systems, defending the tentative conclusion that nominative-absolutive and ergative-absolutive are different alignment types. Gildea first observed that in terms of the traditional definition of ergative alignment, nominative-absolutive alignment would be considered a subtype of ergative alignment, since an ergative pattern is any morphological pattern that treats S and O as a category distinct from A.

Glidea then presented a number of reasons favoring distinguishing nominative-absolutive from ergative-absolutive alignment. First, nominative-absolutive marking displays two characteristics that are counter-universal for ergative constructions: A. If there is a split between case-marking and person-marking affixes, then the bound forms will be accusative and the free forms ergative. This is contradicted in the nominative-absolutive systems of Katxuyana, Apinajé, Suyá, and Timbira. B. An ergative system is less likely to employed when the clause refers to something that has not yet happened (future ternse), is not complete (imperfective) or did not happen (negative polarity), or emphasis is on the agent’s role (imperative and hortative). But in multiple languages, nominative-absolutive alignment is conditioned by future tense, imperfective aspect, and negative polarity.

Second, we see different consequences in the diachronic loss of ergative case markers in nominative-absolutive versus ergative-absolutive systems. When nominative-absolutive systems lose their ergative case marker, the resulting system is “less ergative”. Indeed, since case-marking is frequently the only ergative pattern in the language, the loss of the case marker is sufficient to shift the entire affected construction out of the ergative alignment category.

Gildea suggested that the typology of alignment systems might be profitably reformulated by redefining ergative alignment as a characteristic of clause types that contained a *marked ergative*. Under this definition, nominative absolutive constructions are no longer `ergative’ and most of the tense-aspect violations to universal ergative systems, mentioned above, would disappear. Such a redefinition would also make identifying nominative-absolutive sytems easier, allowing linguists to more easily build a typology of correlations to nominative-absolutive alignment.