Towards a radical construction grammar approach to clitic

Page 1

Towards a Radical Construction Grammar approach to clitic phenomena A case study of some Southwestern Amazonian languages Adam J.R. Tallman Texas Linguistics Society 17


Goals • Discuss a perspective on clitics inspired by Radical Construction Grammar (RCG) • Describe some clitic phenomena in Southwestern Amazonian (SWA) languages, specifically “word-internal” clitics that support this perspective. • Discuss methodologies for studying clitic phenomena that avoid methodological opportunism using SWA languages as a case study.


Radical construction grammar • There is no universal syntactic structure other than a part/whole relations. • Constructions are mappings of form to meaning; they are symbolic units. • All grammatical categories are language specific. • Croft (2003, 2010)

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Cross-linguistic methodological opportunism • The language specific view of categories comes out of a critique of current practice in syntactic theory called cross-linguistic methodological opportunism. • Cross-linguistic Methodological opportunism: In order to establish the existence of some construction or category, one uses only the criteria that point in a positive direction – negative evidence is not considered. • It is ignored that certain cross-linguistic categories are built out of ostensibly different empirical phenomena (i.e. justified according to different criteria). Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Radical construction grammar • The morphology-syntax distinction is gradient - there is a continuum between the lexical and the schematic. • Croft (2003: 17), Langacker (1987: 25)

• But there is an unanswered empirical question within RCG concerning the status of the morphology-syntax continuum on a language by language basis. • Are the constructions of languages organized in a random fashion along this continuum or is there some discontinuity in the transition from morphology to syntax? Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Croft (2003) I agree that what is found in the morphology is a continuation of the syntax to some extent, and that the syntax-morphology distinction is gradient. However, gradience need not imply continuity. That is, a phenomenon relevant to syntax such as headhood may not be continued all the way into the depths of the the word, although the boundary between syntax and morphology may be unclear. (Croft 2003: 267)

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Haspelmath (2010) In order to show that a fuzzy concept of word is theoretically significant, one would have to demonstrate that grammatical units are not randomly distributed over the continuum between fully bound and fully independent units, but that they cluster significantly.

(Haspelmath 2010: 63)

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Radical construction grammar and clitics • Clitics are morphemes that are “somewhere in the middle” of the affixword cline (Halpern 2001) • This paper is concerned with what clitic phenomena say about the (dis)continuity question in RCG. • The issue of clitics has not been brought to bear directly on the issue of gradience. • (but see Haspelmath [2015])

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Clitics • Position: “clitics” are just rhetorical short-hands for residue from a grammarian’s imposition of word boundaries or a morphology-syntax distinction generally – a natural consequence of imposing boundaries on a gradient scale. Affix

Morphology

“clitics” Words

Syntax

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Clitics • When the residue is high it should make us question the validity of the morphology-syntax distinction even on language specific grounds. Affix

Morphology

“clitics” Words

Syntax

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Clitics • General properties of clitics; • There are no properties that uniquely identify clitics (Spencer and Luis 2012) • Clitics cannot be motivated as a natural class even on language specific grounds. • Boundaries between clitics and affixes (morphology) and clitics and words (syntax) can only be posited a priori.

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Clitics • Other approaches to clitics try to place them in one or the other grammatical component or at the interface between components • (Anderson 1992, 2005; Selkirk 1995; Peperkamp 1997; Halpern 1995; Booij 1997, inter alia)

• However, its not clear why clitics cannot simply be seen as challenges to the components to begin with. • Croft (2003, 2010); Langacker (1987, 1990); Haspelmath (2010, 2015)

• Consider some clitic phenomena in Southwestern Amazonia… Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Southwestern Amazonian (SWA) languages • Southwestern Amazonian languages have been noted to contain a number of patterns that make them intriguing with respect to the morphology-syntax divide (Epps & Tallman forthcoming) • Syntax-like morphology (Payne 1990) • Open class grammatical categories (noun classifiers) (Krasnoukhova 2012) • Highly agglutinative (Dixon & Aikhenvald 1999; Aikhenvald 2012)

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Clitics in Southwestern Amazonia languages • There are a number of properties of clitics that are problematic typologically, but I will only focus on one to make my point; • Variable prosodic word projection • (Aikhenvald 2002; Olawsky 2006; Zariquiey 2011; Valle 2017; Tallman in progress)

• Non-scopal variable affix/clitic ordering • (Mihas 2015; Tallman in progress)

• Word-internal clitics Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


“Word-internal” clitics • Word-internal clitics refer to cases where clitics occur inside a grammatical word or even a stem. • There are two ways of describing such cases; • Endoclisis: the clitic is actually inside the grammatical word • Affix-clitic homophony: the affix and clitic are homophones with the same semantics.

• Endoclisis is rare typologically, but (if we follow) the available descriptive grammars, not so much in Amazonia. Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


“Word-internal” clitics • The phenomena are very commonly described in Panoan; consider the distribution of –pacho ‘always’ from Shipibo (Panoan) (Valenzuela 2003: 292) (1)

Ja r-iki chiponki-a-x jo-ai no-n ninká-pacho-ai. 3:ABS EV-COP down.the.river-ABL-S come-PP1 1p-ERG hear-always-INC ‘He is the one (about whom) we always heard was coming from the lower Ucayali.’ () … ja oin-a-ki awakan bai-shaman-ki pues 3:ABS see-PP2:HSY2 tapir:GEN way-INTENS:ABS-HSY2 pues

(2)

ja-n-pacho

ka-[a]i. that-LOC-always go-INC ‘… they found the tapir’s way pues, along which he always walks.’ Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


“Word-internal” clitics • This is even true of “second position clitics” in Valenzuela (2003:40); for instance the interrogative =ki in Valenzuela (2003) can occur “inside” the verb complex in (c) . (3)

(4) (5)

mi-a=ki xawi wai-nko ka-[a]i ? 2-ABS=INTER sugar.cane garden-ALL go-PP1 xawi wai-nko=ki mi-a ka-[a]i ? sugar.cane garden-ALL=INTER 2-ABS go-PP1 Ka=ki-ai mi-a xawi wai-nko ? go=INTER-PP1 2-ABS sugar.cane garden-ALL ‘Are you going to the sugar cane garden?’

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


“Word-internal” clitics • Arawak languages (especially those from the Campa branch) have some of the most complex valence systems in the world. • In some languages, postpositions “incorporate” into the verb and the distinction between a valency marker and a postpositional clitic is difficult or impossible to make. • A striking example comes from Paresi (Brandao 2014).

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


“Word-internal” clitics • The comitative/reciprocal morpheme =kakoa has the same function regardless of its position. • Here there is analytic indeterminacy between the homophony vs. incorporation/endoclitic analysis. (13)

wakoakare=kakoa Ø=aitsa-kakoa-ha minita hoka Indian=COM 3SG=kill-RECP-PL always CON kazaihera-ty-oa-heta be.invisible?-TH-MM-PERF ‘They were always fighting with each other, with the Nambikwara, and he became invisible.’ Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


“Word-internal” clitics • Further cases have been noted in Tacanan (Guillaume 2016), in Aguaruna (Overall 2017) and in Baure (2007) among others. • Ask me about these!

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


“Word-internal” clitics • Valenzuela (2003), Danielson (2007), Overall (2017) describes “wordinternal” clitics as endoclitics. • Zariquiey (2011) and Guillaume (2016) describe them as “clitics that function as suffixes”. • Other authors, use such cases to reanalyze the position of word boundaries (Valle 2017; Tallman forthcoming).

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Word-internal clitics • While there are a number of approaches to endoclisis in the clitic literature from the componential perspective. • (Luis 2002; Harris 2002; Anderson 2005).

• Its not clear why the clitics do not challenge the validity of the word they interrupt. • How many interruption tests does a purported word constituent need to fail before one regards it as a phrase?. Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Clitics and morphological autonomy • In approaches to morphology that posit that morphology is “autonomous”, morphology-syntax indeterminacy are considered “limiting cases” that do not challenge the componential perspective. • (Blevins 2006, 2016)

• But the properties of clitics in Amazonian languages (ubiquity, heterogeneity) challenge this idea. • At what point do the boundary phenomena overwhelm the boundaries? Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Clitics Clitics as residue from language specific modes on an affix-clitic cline.

Affix

Morphology

“clitics” Words

Syntax

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Clitics • If affixes and words are just distributions along a criterial wordhood cline, then their standard deviation can vary from language to language; imagine a language with more “clitics”, by virtue of having fuzzier affixes and words. Affix

Morphology

Words

Syntax

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Clitics • Multimodality in statistical structure is only ever present to a certain extent; does it account for the data better than a unimodal distribution? Affix

Morphology

Words

Syntax

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Clitics • Furthermore, multimodality in statistical structure is a theoretical abstraction over quantitative data. Affix

Morphology

Words

Syntax

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Clitics • It is an empirical question whether a multimodal distribution is motivated in any given case – a ubiquity of “clitics” actually could suggest that the distinction between morphology and syntax might not be there at all for a given language and it is imposed by the grammarian. Grammar

Morphology

Syntax

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Two empirical questions • Are we justified in saying that languages even make a fuzzy distinction between morphology and syntax? • If languages vary in terms of morphological autonomy, how do we characterize or study this variation?

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Methodological challenges • We need some way to rank morphemes and constructions on a scale from affix-like to word-like. • The main methodological challenge is translating qualitatively distinct wordhood properties to a bidimendional quantitative scale. • Challenges: • Development of a database from criterial wordhood properties. • Tallying or correlating the criterial wordhood properties in ways that are meaningful for the research question. • See Epps and Tallman (submitted) Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Statistical methods • Haspelmath (2010) states that “statistical methods” could be used to determine this question; • I introduce one method using pairwise correlations between criterial wordhood properties (Epps & Tallman forthcoming); this method only addresses the second question. • I mention more direct methods of testing this problem in the conclusion. (testing multimodality); this method addresses the first question.

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Methodology and Languages • Languages of this study; • We chose languages with relatively comprehensive descriptions; the descriptions have to be detailed enough for us to determine the wordhood properties of the coding of grammatical categories we investigated. • Time constraints have so far prevented us from looking at more than 11 Amazonian languages (Hup, Chácobo, Kotiria, Tariana, Perene, Cavineña, Jarawara, Urarina, Movima, Paresi) • We looked at four domains (nominal classification, evidentiality, tense, valency) Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Methodology and Languages • We have been developing a database that codes for each morpheme, its word-like vs. affix-like properties. • We coded 24 properties for each morpheme; it is outside the scope of this paper to review all of these, but some of the important ones are listed below. • • • • • •

Number of allomorphs (0 to 5) Suppletive allomorphy (yes = 1, no =0) Culminative expression (yes =1 , no =0) Bounded (yes =1, no =0) Prosodic independence (0 =always, 1 =sometimes, 2 =never) Contiguity (yes =1, no =0)

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Control language • We use Central Alaskan Yupik (CAY) as an outlier language against which to compare Amazonian languages. • It is an appropriate outlier given its well-known status as displaying canonical, obvious and/or old polysynthesis. • Languages with a fuzzy boundary between morphology and syntax are not expected to appear canonically polysynthetic.

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Exponence complexity (EC) • Recall there are three variables of exponence complexity in this study (Anderson 2015). • Segmental allomorphy • Suppletive allomorphy • Exponence

Exponence Complexity (EC) = Segmental allomorphy + Suppletive allomorphy * 2 + Exponence

Introduction – System complexity – Exponence Complexity – Morphological autonomy – Conclusion


Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Exponence complexity and its covariates • The first approach tests how well criterial wordhood properties correlate with each other. • We take the EC score and measure is against a number of other criterial wordhood properties, such as Contiguity, Boundedness and Prosodic Independence. • A high correlation to us suggests higher morphological autonomy; our results confirmed our general impression that Amazonian languages were less morphologically autonomous. Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Boundedness • Boundedness is also a criterion for morphological status. (Bloomfield 1933; Hockett 1956)

• Boundedness: Does the morpheme pass the minimal free form test? (yes = 0, no = 1)

• Chácobo is a language with a lot of bound morphemes, even though these morphemes are syntactic according to other criteria. (Tallman 2016) Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Bounded (11) a.

b. e. f. g. h. i.

oʂa-mis=tɨkɨn=kas=ʔitá=kɨ=rɨ ́ sleep-A N T IP A S S =A G A IN =D E S =R E C P =D E C :P =L A M E N T ‘What a shame that he only wanted to sleep yesterday.’ *-mis *=tɨkɨn *=ria *=ʔitá *=kɨ *=rɨ

(bound = 1) (bound = 1) (bound = 1) (bound = 1) (bound = 1) (bound = 1)

Chácobo

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Statistical methods • We use Kendall’s tau 3, a non-parametric rank statistic to assess the correlation between EC complexity and each of the wordhood properties. In R: cor.test(x, y, method=“kendall”). • The reason we use Kendall’s tau is that the data are not normally distributed.

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Boundedness ~ EC • The only language that displays a statistically significant and positive correlation is CAY. • This is because bound elements in CAY are more morphology like in terms of EC – whereas in Amazonian languages there is no relationship. • Epps & Tallman (submitted) for discussion

tau correlation

p-value

CAY

0.543

0.004

Kokoma

0.449

0.017

Jarawara

0.329

0.139

Paresi

0.246

0.166

Tariana

0.189

0.038

Hup

0.183

0.143

Perene

0.119

0.231

Chácobo

0.099

0.445

Cavineña

0.085

0.671

Urarina

0.037

0.858

Kotiria

-0.342

0.050

Movima

-0.716

>0.005

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Testing for multimodality in the affix-word cline • The problem with the case by case methodology is that it does not test Haspelmath’s question directly, it only tells us to what degree a a language instantiates morphological autonomy along a few parameters a time.

• The methodology of Epps and Tallman (forthcoming) is question begging to a certain extent; maybe there is no morphological autonomy to score to begin with. • How do we test Haspelmath’s question directly? Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Global bondedness • We need to be able to score every item in the database according to a global bondedness measure; Global bondedness = 5 – Exponence complexity – Boundedness –Contiguity -Prosodic dependence –Obligatoriness – Fixedness • The 5 is just to get the data centered around 0; low values = really affix like, high values = really word like. Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Affix-word scales • Assuming we accept the weighting of the criterial wordhood properties (which could always be up for debate) – we can now map the coded morphemes on a scale from most affix/morphology to most word/syntax.

• Density distributions can provide a graphic representation that we can use to gauge whether a distribution is multimodal.

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Density distribution • For instance, in CAY it looks there is a multimodal distribution.

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Density distribution • For instance, in CAY it looks there is a multimodal distribution. • Compare this with Perene, which we found had weaker morphological autonomy according to our rank correlation tests.

Perene

Central Alaskan Yupik

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Statistical tests • It is not really clear whether Perene is best described as having a multimodal or unimodal distribution. • Even for CAY, we do not know if the distributions we find are just accidents based on the data we chose to code (is the multimodality statistically significant?)

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Statistical tests • There are a number of exploratory ways of testing for multimodality (there are no generally accepted “statistical tests” that just test for bimodality without complications) • The simplest way is to use the Hartigan’s dip statistic (HDS) (Hartigan and Hartigan 1985; Freeman & Dale 2012; Maechler 2015). • HDS tests the null hypothesis of unimodality against the alternative hypothesis of multimodality. Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Hartigan’s dip statistic • The tests show that we do not have enough evidence to reject the hypothesis that Jarawara, Cavineña and Urarina have a unimodal distribution. • Perene has a bimodal distribution, but its D statistic is lower, showing a weaker distinction (lower morphological autonomy)

D

p-value

Movima

0.176

*<0.005

Hup

0.161

*<0.005

Central Alaskan Yupik

0.160

*<0.005

Paresi

0.156

*<0.005

Kotiria

0.147

*<0.005

Kokama-Kokamilla

0.112

*0.003

Perene

0.092

*<0.005

Chácobo

0.090

*<0.005

Jarawara

0.087

0.242

Tariana

0.084

*<0.005

Cavineña

0.077

0.233

Urarina

0.063

0.650

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Density distributions • However, the data display an interesting anomaly, at least with respect to the results presented in Epps and Tallman (forthcoming)

Hup

Movima

• Movima and Hup display a greater D statistic than CAY, which means their distributions are more emphatically bimodal. Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Possible interpretations • HDS assumes that the distributions are normal (its parametric), which could be problematic – D stat could be high and show multimodality just because of skewing for instance. • The languages vary in terms of sample size (some languages just have more tokens than others). • HDS finds the density of modes more important than the distance between them (is this a desirable result?). Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Conclusion: theoretical • From an RCG perspective clitics are the overlapping area on language specific modes on the affix-word cline. • The ubiquity of clitics and their structural heterogeneity should make us question the boundary that they actually blur rather than attempting to view them as some aberrations on a presupposed Platonic ideal.

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Conclusion: methodological • However, taking the issue of the morphology-syntax distinction as an empirical question seriously, is an extremely vexing methodological question. • It is difficult to globally quantify morphological autonomy – and it is even more difficult to test whether there is a multimodality is present at all in the affix-word cline. • More research needs to be conducted on how to actually construct a quantitative affix-word cline out of qualitatively distinct criterial wordhood properties and what statistical methods should be used. Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Questions?


“Prosodic word” clitics • A number of clitics in Amazonian languages project prosodic words – or are not prosodically dependent (at least in certain syntactic contexts). • This is despite being “bound” in the sense of failing the minimal free form test. • Sometimes results in apparent cases of “words” with more than one “stress” or marker of prominence. • It is difficult to classify such cases as canonically word like in terms of their prosodic behavior.


“Prosodic word” clitics • In Kakataibo, second position clitics sometimes surface even when the first element in a null pronoun. • In such cases, normally bound and prosodically deficient forms apparently project prosodic words (Zariquiey 2011; Valle 2017) • According to Zariquiey (2011) and Valle (2017), in such cases, the clitic bears stress and has all of the properties of a phonological word.


“Prosodic word” clitics • Valle (2017: 245-246); “it is not uncommon for second-position enclitics to occur as the first element of the clause when no constituent precedes them. When this happens, the second-position enclitic or group of them (second-position clitic complex) forms a phonological word, carrying a main stress….” (1)

(2)

ñu nami këñu ‘arunuxudapika biáxa. ñu nami këñu ‘aru-nuxun=dapi=ka=a thing flesh with cook-A/S>A:FE=DUB=VAL=3A/S ‘He grabbed (it) in order to cook with meat.’ ka ñu tuiáxi. =ka=a ñu tuin-a-x-i =VAL=3A/S thing hold-PFV-3-PROX ‘(S/he) is holding something.’

bis-a-x-a grab-PFV-3-N.PROX


“Prosodic word” clitics • In Urarina, Olawsky (2006: 80-90) identifies clitics as bound, variable position in the clause and/or their ability to be ordered before another clitic. • “From a cross-linguistic point of view, clitics tend to be prosodically deficient: i.e. they typically do not have stress or tone. However, about half of the Urarina enclitics do follow this tendency in that they carry a H tone in addition to the tone pattern on the word they are attached to” (Olawsky 2006: 87)


“Prosodic word” clitics • In Chácobo (Pano), obligatory high tones are a property of phonological words, and the vast majority of clitics (and bound forms generally) bear their own lexical high tone. • What’s more interesting is that clitics seem to variable project a stress domain depending on their syntactic position. • This is true of “graded tense” morphemes in Chácobo, which have been described as “functioning as auxiliaries” (Cordoba et al. 2012; Epps & Tallman submitted)


“Prosodic word” clitics • Chácobo graded tense morphemes Project their own prosodic words. Chácobo (Panoan; Tallman in progress) (2) a. kako sani=ʔi (ka=ʔitá=kɨ)Pwd Caco fish=SS go=REC.PST=DEC:PST ‘Caco went fishing (yesterday or two days prior)’ b. sani=ʔi (kaa)Pwd káko (=ʔitá=kɨ)Pwd fish=SS go Caco =REC.PST=DEC:PST ‘Caco went fishing (yesterday or two days prior).’


Canonical typology • Another approach to the typology of clitics has been suggested in the literature. • Spencer and Luis (2012) attempt to develop a canonical approach to clitics. • In canonical typology, one finds a canonical instance of some construction based on a well accepted definition, and then finds divergences from it.


Canonical typology • However, Spencer and Luis (2012) show that clitics are problematic in this regard. • There are no criteria that distinguish clitics from affixes and words simultaneously. • Therefore they take for their point of departure the idea that a canonical function word and a canonical affix can be defined, and argue that a canonical clitic has the form of an affix but the distribution of a function word.


Canonical typology • Clitic: prosodically dependent (like affix) but positioned syntactically (like function word). • I argue that this approach presupposes the categories that clitics problematize. • It results in methodological opportunism, because its unclear what criteria should be brought to bear for prosodic dependence and syntactic positioning to begin with.


Clitics and morphological autonomy • Such a situation poses a potential problem for those who advocate that morphology is autonomous in some sense from syntax (Booij 2010; Blevins 2016).


On Canonical typology • Word-internal clitics challenge the division between morphology and syntax to begin with in the same way that


Clitics “A clitic is an element whose distribution linguists cannot comfortable consign to a single grammatical component” (Sadock 1995: 260) Affix

Morphology

“clitics” Words

Syntax


Radical construction grammar • Radical construction grammar is associated Framework free grammatical theory (Haspelmath 2010) – that rejects all a priori categories for language specific analyses. • In fact Croft (2010) suggests that it may be indistinguishable from the latter. • “Radical construction Grammar is essentially a framework free grammatical theory grammatical theory (Haspelmath 2010, perhaps the only one existing at the present.” (Croft 2010: 2)


Cross-linguistic methodological opportunism • There are many examples of it in the linguistic literature (see Croft [2003] for a review) • But most pertinent to our study is the critique of the category word done by Haspelmath (2010). • Haspelmath (2010) argues that linguists do not have any basis for identifying the category “word” cross-linguistically.


Cross-linguistic methodological opportunism Since there is not one single criterion that identifies words, and attempts at coming up with jointly necessary and sufficient conditions have not been successful either, in practice linguists have often adopted the strategy of persuasion via test batteries. In this strategy, a number of criteria are selected and applied, and n the published accounts usually all of them point in the same direction. The more criteria converge, the more persuasive the argument becomes, but the method is not rigorous, because the criteria can be selected opportunistically by the author. (Haspelmath 2011: 59)


Clitics and morphological autonomy • Morphological autonomy refers to the idea that morphological patterns are phenomenologically distinct from syntactic ones (e.g. Anderson 1992; Booij 1997, 2010; Blevins 2016). • In such approaches clitics or formatives on the boundary between morphology and syntax are not seen as problematic for morphological autonomy (e.g. Blevins 2006). • However, the ubiquity of clitics implies that morphological autonomy is at least a matter of degree, even if it is motivated.


Observations about clitics in Amazonia • Clitics are ubiquitous in many descriptions of Amazonian languages (often more common than affixes or function words). • Grammarians frequently point out that there is a lot of analytic ambiguity between affixes and clitics. • Grammarians vary in terms of whether clitics are described “in the morphology” or “in the syntax”, even where dealing with ostensibly similar phenomena. Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Haspelmath’s problem In the first hypothetical situation (clustering distribution), there are three clearly discernible clusters. If the dimension along which the units differ (the boundedness scale) can be quantified, the clustering can be demonstrated by statistical techniques. There are intermediate cases between the clusters of affixes, clitics, and independent words, but these are few and just exceptions to the rule. (Haspelmath 2011: 63)

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Haspelmath’s problem In order to show that a fuzzy concept of word is theoretically significant, one would have to demonstrate that grammatical units are not randomly distributed over the continuum between fully bound and fully independent units, but that they cluster significantly. (Haspelmath 2010: 63)


Clitics are heterogeneous: Kakataibo • Consider a recent description of Kakataibo; the classification demonstrates that the clitic classes literally have nothing in common (Valle 2017: 243).

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Heterogeneity of clitics • Another global property of clitics in SWA languages support the conception of clitics defended here. • Clitics are highly heterogeneous in terms of their word criterial properties (cf. Aikhenvald 2002) • Why are they even regarded as one category?

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Clitics are heterogenous: Urarina “Comparing the respective enclitics in (105) and the features assigned to each of them, it becomes evident that it is not practical to represent them as natural groups or classes of clitics. The distinctive features merely help to characterize the similarities and differences between the clitics, but these do not neatly coincide with each other” (Olwasky 2006: 89)

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


“Word-internal” clitics • Tacanan (Guillaume 2016); some emotive morphemes have wordexternal or word internal distributions. (6)

Enekita pa beu bute-ichenu-iti-a beu really QUOT PFV go_down-COMPAS-PVF-PST PFV ‘He climbed down (from the tree, the poor man who had been abandoned).’

(7)

Beidaji=ichenu beu daja y-a-ta-ani=su. happy=COMPAS PFV thus IPFV-tell-A3-IPFV.SITTING=TMP.DS ‘He was happy, the poor (man), when he was told (how to climb down from the tree).’

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


“Word-internal” clitics • In Baure, personal pronoun clitics can also be “closed in” by nominalizers that normally combine at the word-level (Danielson 2007:XX). (8) ro=komoroki-yi-wo=pi-no kove’ 3SGm-bite-LOC-COP=2SG-NOM1 dog ‘Where did the dog bite you?’ [RP-6/7/04-3]

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


“Word-internal” clitics • The question clitic in Aguaruna (Overall 2017: 390-391) (9)

(10)

amɨ=sha puha-mɨ=ka 2sg=Q.TOP live+IPFV-2sg=Q ‘are you living?’ (greeting formula) Mario Vargas Llosa Premio Nobel su-sa-aha-ma M. V. L. Nobel Prize give-PFV-PL-RECPST nu=sha dɨka-a-ma=ka-umɨ ANA=Q.TOP know-PFV-RECPST=Q-2sg ‘did you know they gave the Nobel Prize to Mario Vargas Llosa?’

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Contiguity~EC • Only CAY and Jarawara show a positive and significant relationship. • Again, CAY shows the highest effect size.

Language Central Alaskan Yupik Jarawara Cavineña Perene Chácobo Tariana Movima Urarina Kotiria Paresi Hup

Tau statistic 0.594 0.531 0.305 0.236 0.178 0.131 0.101 0.035 0.030 -0.053 -0.166

p-value 0.002 0.017 0.122 0.122 0.168 0.150 0.190 0.868 0.862 0.764 0.183

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Prosodic dependence ~ EC • Here we find the same pattern, the correlation is high and significant for Central Alaskan Yupik and nonsignificant for Amazonian languages. • Prosodic dependence does not correlate with EC in Amazonian languages.

Language Central Alaskan Yupik

Tau statistic p-value 0.543 0.004

Perene Tariana Cavineña Chácobo Urarina Hup Jarawara Movima Kotiria

0.187 0.155 0.151 0.129 0.122 0.018 -0.040 -0.089 -0.273

0.052 0.076 0.443 0.307 0.485 0.884 0.842 0.236 0.113

Radical Construction Grammar – Amazonian clitics – Quantifying morphological autonomy – Testing multimodality - Conclusion


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.