Differential coding in property words: A typological study
Y`eJ`ıngt´ıng
Leipzig University
5.3.2020
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 1 / 53 Outline
1 Introduction
2 Comparative concepts
3 Differential coding types and the attributive prominence hierarchy
4 The semantic map of property concepts: an MDS analysis
5 Conclusions
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 2 / 53 Outline
1 Introduction
2 Comparative concepts
3 Differential coding types and the attributive prominence hierarchy
4 The semantic map of property concepts: an MDS analysis
5 Conclusions
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 3 / 53 Introduction
The phenomenon: Subclass-conditioned differential coding in property words For instance:
(1) Savosavo (Solomons East Papuan, Papunesia) a. pa ngai vaka=na [No marker] one big ship=nom ’a big ship’ (Wegener 2012, 69) b. pa savanga sua kola [Marked by a relativizer] one be.long rel stick ’a long stick’ (Wegener 2012, 336)
In the previous literature, such phenomena are called ‘split adjectives’ (Wetzer 1992, 1996; Stassen 1997) or ‘adjective classes’ (Dixon 2004a), but have not been subject to much inquiry.
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 4 / 53 Aims and methodology
This talk aims at: providing a systematic typological investigation of subclass-conditioned differential coding in property words. building semantic maps for property concepts identifying universal tendencies in the mapping between form and meaning. The methodology Grammar mining → Construction-specific (attributive and predicative constructions) → Sample: 232 constructions of 50 geographical and genealogical diverse languages worldwide.
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 5 / 53 The sample
50 languages 39 language families 6 macro-areas – Africa (13) – N.America (7) – S.America (7) – Papunesia (9) – Eurasia (8) – Australia (6)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 6 / 53 Outline
1 Introduction
2 Comparative concepts
3 Differential coding types and the attributive prominence hierarchy
4 The semantic map of property concepts: an MDS analysis
5 Conclusions
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 7 / 53 Adjectives and Property words
The adjective has long been regarded as a mixed category, with [+N] and [+V] as its characterizing features.(Chomsky, 1970) Crosslinguistically, substantial variation is observed regarding the size, syntactic behaviour, and semantic scope of adjectives. In many Indo-European languages, adjectives are considered to constitute an open word class, while in some other languages, only a restricted set of words are regarded as adjectives. For instance, in Zialo there are only five adjectives: g`Ol`a ’big’, k´ul´o ’small’, k´ul´ok´ul´o ’very small’, n´ı´ın´a ’new’, p˜ad`a ’good, beautiful’, p´Ol´O ’old’ (Babaev, 2010).
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 8 / 53 Adjectives and Property words
Dixon’s seven semantic types
Despite the crosslinguistic variation, Dixon (1982) suggests that there are 7 semantic types that tend to be included in the word class of adjectives: dimension (big, small. . . ) age (old, young. . . ) value (good, bad. . . ) colour (white, black. . . ) physical property (hot, cold. . . ) human propensity (happy, sad. . . ) speed (quick, slow. . . ) Hallonsten Halling (2018) claims that speed is a core semantic type for adverbs. → speed is excluded in the present study.
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 9 / 53 Semantic targets
Property words A property word is a lexeme that denotes a descriptive property, such as age, dimension, colour, value, physical property and human propensity, and that can be used to narrow the reference of a noun. (Based on Dixon 1982,Haspelmath 2010) 28 Semantic targets used in this study age young, new, old (animate), old (inanimate) dim big, small, long, short col white, black, red, green val good, bad, beautiful, ugly phy hot, cold, soft, heavy, dry hum happy, sad, clever, stupid, angry, crazy, afraid
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 10 / 53 Substantive marker and function indicator
Grammatical markers are morphemes which have less concrete referential meaning than content words (Meillet 1912; Traugott 1980). Two types of grammatical markers: Substantive markers Function indicators → Based on the distinction between structural coding and behavioural potential in Croft (1991, 2001).
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 11 / 53 Function indicators (structural coding): – indicate the unusual association between the semantic class and pragmatic function. – no substantive information – typical types of function indicators 1 attributivizer 2 predicativizer (copular) e.g. Das Kleid ist sch¨on. Substantive markers (behavioural potential): – bear substantive information – typical types of substantive markers 1 nominal substantive marker: related to nominal domain (e.g. gender/number/case/definiteness) 2 verbal substantive marker: related to verbal domain (e.g. TAM) e.g. ein sch¨ones Kleid
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 12 / 53 Croft’s theory
According to Croft, the coding of lexical categories varies according to the prototypicality of the correlation between semantic classes and pragmatic functions.
Table 1: Semantic properties of prototypical lexical classes (Croft, 1991, 55)
syntactic category noun adjective verb semantic class object property action pragmatic function reference modification predication
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 13 / 53 When the association between semantic classes and pragmatic functions is not a prototypical one, languages may show function-indicating morphosyntax.
Table 2: Function-indicating morphosyntax (Croft, 1991, 67)
Reference Modification Predication Objects unmarked genitive, adjectiviza- predicate nominals nouns tion, PP Properties deadjectival unmarked adj predicate adjectives nouns Actions action nomi- participles, relative unmarked verbs nals, comple- clause ments, infini- tives, gerunds
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 14 / 53 Table 3: English examples (Croft 1991)
Reference Modification Predication Object vehicle vehicular be a/the vehicle Property whiteness white be white Action destruction destroying/destroyed destroy
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 15 / 53 Outline
1 Introduction
2 Comparative concepts
3 Differential coding types and the attributive prominence hierarchy
4 The semantic map of property concepts: an MDS analysis
5 Conclusions
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 16 / 53 Differential coding illustrations
Type 1. Differential coding of function indicators in the attributive position:
(2) Cavine˜na(Pano-Tacanan, South America) a. badi nana moon young ’new moon’ (Guillaume, 2008, 469) b. wika arida=ke hook big=rel ’the big hook’ (Guillaume, 2008, 360)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 17 / 53 Differential coding illustrations
Type 2. Differential coding of function indicators in the predicative position: In some languages, a copula only occurs under certain conditions (e.g. past tense). For example, in Cherokee (Iroquoian, North America):
(4) The past tense (3) The present tense a. a-aliihe´el`i´ik-v´vPi a. a-aliihe´el`i´ika 3-be.happy-pst 3-be.happy.prs ’He was happy.’ ’He is happy.’ b. ka-alij´o´ohita jikees-v´vPi b. aki-alij´o´ohita 3-fat cop-pst 1-fat ’He was fat’ ’I am fat.’ (Montgomery-Anderson, 2008, 269, 279, 493)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 18 / 53 Differential coding illustrations
Type 3. Differential coding of substantive markers in the attributive position
(5) Neverver (Austronesian, Papunesia) a. niterikh lele ang child small the ’the small child’ (Barbour, 2012, 112) b. netan i-rvikh thing 3.real.sg-good ’good thing’ (Barbour, 2012, 391)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 19 / 53 Differential coding illustrations
Type 4. Differential coding of substantive markers in the predicative position
(6) Yimas (Lower Sepik-Ramu, Papunesia) a. yan imprampat yua-ra arak. those basket.vii.pl good-vii.pl cop.vii.pl ’Those baskets are good.’ (Foley, 1991, 189) b. na-na-kuck-n 3sg-definitive-be.happy-prs ’He is happy.’ (Foley, 1991, 446)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 20 / 53 Semantic types and grammatical markers
Figure 1: The semantic variation of Figure 2: The semantic variation of attributive function indicators predicative function indicators
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 21 / 53 Semantic types and grammatical markers
Figure 3: The semantic variation of Figure 4: The semantic variation of nominal substantive markers verbal substantive markers
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 22 / 53 The attributive prominence hierarchy
Based on this, the following attributive prominence hierarchy can be proposed: age> dimension, colour> value > physical property > human propensity From left to right, the attributive prominence decreases. A property word is ‘attributive prominent’ means: 1 it is less likely to occur with an attributivizer 2 it is less likely to occur with a verbal substantive marker This hierarchy is different from Stassen (1997)’s adjective hierarchy in the following respects: The position of value is different. Stassen’s adjective hierarchy only predicts the occurrence of verbal encoding, but does not predict the occurrence of an attributivizer.
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 23 / 53 Stassen’s adjective hierarchy
Stassen (1997, 168) compared the predicative use of adjectives and proposed the adjective hierarchy: The adjective hierarchy (Stassen, 1997, 169) Human Propensity / physical property / dimension, color / value, age, form / material, gender This hierarchy predicts that when languages have a split in the encoding of adjectives, the property concepts to the right of the “cut-off point” are more likely to show non-verbal encoding and the property concepts to the left of the “cut-off point” are more likely to show verbal encoding (Stassen, 1997, 169).
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 24 / 53 Outline
1 Introduction
2 Comparative concepts
3 Differential coding types and the attributive prominence hierarchy
4 The semantic map of property concepts: an MDS analysis
5 Conclusions
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 25 / 53 Some notes on semantic maps
The semantic map model is a method for describing and illustrating the patterns of multifunctionality of grammatical morphemes that does not imply a commitment to a particular choice among monosemic and polysemic analyses. The semantic-map approach is firmly rooted in empirical observation of individual languages, but through systematic cross-linguistic comparison we can arrive at well-motivated structural patterns in the universal conceptual space.
(Haspelmath, 2003)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 26 / 53 The MDS analysis
One way of generating semantic maps is to use the multidimensional scaling (MDS) model proposed by Croft and Poole (2008a). In this method, the multidimensional scaling (MDS) model is used to calculate (dis)similarities between various factors in psychology and other social sciences. The MDS semantic map is thought to be more suitable for large-scale data analysis. This method has been applied in many linguistic studies, such as Narrog and Ito (2007)’s semantic map for commitative and instrumental cases, W¨alchliand Cysouw (2012)’s semantic map for motion verbs, Rogers (2016)’s semantic map for parts of speech systems.
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 27 / 53 The MDS analysis
“Multidimensional scaling produces a spatial representation of similarity. As applied to linguistic phenomena, it produces a spatial representation of similarity for a set of functions as determined by their grouping under a single word form or construction in a language, generalized across different forms and across different languages. ”(Croft and Poole, 2008b, 14)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 28 / 53 The semantic map connectivity hypothesis
The semantic map connectivity hypothesis: Any relevant language-specific and construction-specific category should map onto a connected region in conceptual space. (Croft, 2001, 96)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 29 / 53 My data on property words
I have documented the attributive and predicative use of the 28 property concepts in 50 languages (232 constructions and 6496 data-points). a property concept can occur in a construction → “1” a property concept can occur in a construction → “0” no information → “NA” Based on the formal similarity of 28 different property concepts at a constructional level, various MDS analyses are generated.
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 30 / 53 The result
Figure 5: The MDS analysis of differential coding of in property concepts (all constructions)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 31 / 53 The clustering of different semantic types
Figure 6: The clustering of different semantic types in the MDS
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 32 / 53 The result
Figure 7: The MDS analysis of differential coding of function indicators in property concepts
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 33 / 53 The result
Figure 8: The MDS analysis of differential coding of substantive markers in property concepts
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 34 / 53 The result
Figure 9: The MDS analysis of attributive constructions
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 35 / 53 The result
Figure 10: The MDS analysis of predicative constructions
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 36 / 53 How to interpret the MDS semantic map?
The MDS semantic map shows the general trends of similarities observed in the coding of property words. The closer two points are located on the semantic map, the more likely that they use similar coding strategies in languages. The MDS semantic map represents the universal semantic space, and different languages carve up the semantic space. Given that Croft’s ’connectivity hypothesis’ is correct, it should be possible to draw a cut-off line on the semantic map for each language. On this semantic map (for a specific language), property words on the one side of the cut-off line tend to show one coding type, while property words on the other side tend to show the other coding type.
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 37 / 53 Illustration (1): Cavine˜na
(7) Cavine˜na(Pano-Tacanan, South America) a. badi nana moon young ’new moon’ (Guillaume, 2008, 469) b. wika arida=ke hook big=rel ’the big hook’ (Guillaume, 2008, 360)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 38 / 53 Illustration (1): Cavine˜na
Figure 11: The semantic map for Cavine˜na
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 39 / 53 Illustration (2): Jarawara
(8) Jarawara (Arawan, South America) a. Jowao botee amaka waha. name old cop now ’Jo˜aois now old.’ (Dixon, 2004b, 340) b. tika hijari amosa-tee ama ti-ke. 2sg.poss storytelling be.good-hab extent 2sg-decf ’Your storytelling is customarily good.’ (Dixon, 2004b, 192)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 40 / 53 Illustration (2): Jarawara
Figure 12: The semantic map for Jarawara
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 41 / 53 Illustration (3): Chabu
(9) Chabu (Isolate, Africa) a. t’a d`a`a ji. food good cop ’Food is good.’ (Tsehay, 2015, 302) b. ENgEta-e-N budZ-e. younger.sister.in.law-def-fem angry-pst ’The younger sister in law was angry.’ (Tsehay, 2015, 137)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 42 / 53 Illustration (3): Chabu
Figure 13: The semantic map for Chabu
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 43 / 53 Illustration (4): Barupu
In Barupu, differential coding is found in the attributive position.
(10) Barupu (Sko,Papunesia) a. ˆai koka p´ako rewo tree trunk big very ’a very big tree trunk’ (Corris, 2005, 143) b. ne taipˇe-r´e magic bad-3pl.fem ’bad magic’ (Corris, 2005, 141)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 44 / 53 Illustration (4): Barupu
Figure 14: The semantic map for Barupu attributive constructions (based on the semantic map for attributive constructions)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 45 / 53 Illustration (5): Ma Manda
(11) Ma Manda (Nuclear Trans New Guinea, Papunesia) a. nak-nga=nang yot udu kusamba. 1sg-emph-gen house that big ‘My house is big.’ (Pennington, 2016, 133) b. na wa=lˆu gulˆu daampa-be-k. man that=nom alright happy-irr.sg-3sg ’That man will be blessed.’ (Pennington, 2016, 128)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 46 / 53 Illustration (5): Ma Manda
Figure 15: The semantic map for Ma Manda predicative constructions (based on the semantic map for predicative constructions)
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 47 / 53 Semantic space and universal tendencies
The semantic map is generated based on linguistic forms and shows the ,universal conceptual space of property concepts. → Universal tendency On the semantic map, property concepts located closer are more likely to show similar coding strategies. The semantic map represents the universal semantic space, which can be carved in different ways for various languages.
By and large, property concepts on the right of the semantic map tend to be less attributive prominent, and the ones on the left tend to be more attributive prominent.
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 48 / 53 Outline
1 Introduction
2 Comparative concepts
3 Differential coding types and the attributive prominence hierarchy
4 The semantic map of property concepts: an MDS analysis
5 Conclusions
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 49 / 53 To sum up
The talk: provides a systematic typological overview of differential coding in property words proposes the attributive prominence hierarchy: age > dim,col > val >phy > hum generates various MDS semantic maps for property words based on cross-linguistic comparison claims that these semantic maps represent the universal conceptual space of property words, and each language carves up the conceptual space in a particular way.
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 50 / 53 ReferencesI
Babaev, K. V. (2010). Zialo: The newly-discovered Mande language of Guinea, volume 82. Lincom Europa. Barbour, J. (2012). A grammar of {Neverver}. Mouton de Gruyter, Berlin. Chomsky, N. (1970). Remarks on nominalization. In Readings in English transformational grammar, pages 184–221. Corris, M. (2005). A grammar of Barupu: A language of Papua New Guinea. PhD thesis, University of Sydney. Croft, W. (1991). Syntactic categories and grammatical relations: The cognitive organization of information. University of Chicago Press. Croft, W. (2001). Radical construction grammar: Syntactic theory in typological perspective. Oxford University Press on Demand. Croft, W. and Poole, K. (2008a). Multidimensional scaling and other techniques for uncovering universals. Theoretical Linguistics, 34(1):75–84. Croft, W. and Poole, K. T. (2008b). Inferring universals from grammatical variation: Multidimensional scaling for typological analysis. Theoretical linguistics, 34(1):1–37. Dixon, R. (1982). Where have all the adjectives gone? In Where have All the Adjectives Gone?: And Other Essays in Semantics and Syntax, number 1, pages 1–62. De Gruyter Mouton. Dixon, R. M. (2004a). Adjective classes in typological perspective. Adjective classes: A cross-linguistic typology, pages 1–49. Dixon, R. M. W. (2004b). The Jarawara language of southern Amazonia. Oxford University Press. Foley, W. A. (1991). The Yimas language of New Guinea. Stanford University Press. Guillaume, A. (2008). A grammar of {Cavine˜na}. Mouton de Gruyter, Berlin. Hallonsten Halling, P. (2018). Adverbs: A Typological Study of a Disputed Category. Stockholm University. Haspelmath, M. (2003). The geometry of grammatical meaning: Semantic maps and cross-linguistic comparison. In The new psychology of language, pages 217–248. Psychology Press. Haspelmath, M. (2010). Comparative concepts and descriptive categories in crosslinguistic studies. Language, 86(3):663–687. Meillet, A. (1912). L’´evolutiondes formes grammaticales. Scientia, 6(12):384.
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 51 / 53 ReferencesII
Montgomery-Anderson, B. (2008). A Reference Grammar of Oklahoma Cherokee. PhD thesis, University of Kansas. Narrog, H. and Ito, S. (2007). Re-constructing semantic maps: the comitative-instrumental area. STUF–Sprachtypologie und Universalienforschung, 60(4/2007):273–292. Pennington, R. (2016). Ma Manda: A Papuan language of Morobe Province, Papua New Guinea. PhD thesis, James Cook University. Rogers, P. (2016). Illustrating the prototype structures of parts of speech: A multidimensional scaling analysis. Stassen, L. (1997). Intransitive Predication. Oxford University Press. Traugott, E. C. (1980). Meaning-change in the development of grammatical markers. Language Sciences, 2(1):44–61. Tsehay, K. (2015). Documentation and Grammatical Description of Chabu. PhD thesis, Addis Ababa University. W¨alchli,B. and Cysouw, M. (2012). Lexical typology through similarity semantics: Toward a semantic map of motion verbs. Linguistics, 50(3):671–710. Wegener, C. U. (2012). A grammar of Savosavo. Mouton de Gruyter, Berlin. Wetzer, H. (1992). “Nouny” and “verby” adjectivals: A typology of predicative adjectival constructions. In Meaning and grammar: Cross-linguistic Perspectives. De Gruyter. Wetzer, H. (1996). The typology of adjectival predication. Walter de Gruyter.
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 52 / 53 Thank you!
Y`eJ`ıngt´ıng (Leipzig University) DGfS, University of Hamburg 5.3.2020 53 / 53