The polycategoriality parameter: Noun-verb similarities in some languages of the Americas

1 Workshop “Effects of grammatical system on the lexicon”, University of Sonora, 2025 November 13-14 The polycategoriality parameter: Noun-verb similarities in some languages of the Americas MARTIN HASPELMATH Max Planck Institute for Evolutionary Anthropology (Leipzig) 1. Big-picture comparison of languages a widespread 19th century view (naive evolutionism): languages of “less advanced” peoples are somehow “primitive”, perhaps reflecting different degrees of intellectual development a widespread early 20th century view (cultural particularism): languages on different continents may be very different from each other, partly reflecting the very different cultures of their speakers a widespread late 20th century view (cognitive universalism): languages from around the world share the same (possibly innate) blueprint and vary just slightly from each other Are we condemned to presenting views of languages that reflect the prevalent ideology in our societies? Or is an objective perspective on language structures possible? briefly on the notion of “lexicon” vs. “grammatical system”: It seems to me that language structure is best conceptualized in terms of the notion of inventorium rather than “lexicon”. “Lexicon” (or “lexical”) has several different meanings (Haspelmath 2024), and there are no good reasons to think that there exists a “component” of grammars that can be called “lexicon”. Languages consist of roots and affixes and constructions and idioms, and all of these must be learned and remembered – they constitute the inventorium (= Jackendoff’s “extended lexicon”) Using the example of grammatical noun-verb similarities in a range of North American languages, I will argue • that cognitive universalism may lead us unwittingly toward Eurocentrism • that the noun-verb similarities of some North American languages are indeed striking and worth highlighting as unusual among the world’s languages, BUT that they are not exceptions to robust cultural universals of grammatical encoding 2 So Jelinek & Demers (1994) were not wrong when they claimed that Salishan languages have “no noun-verb distinction” in some sense (contra Davis et al. 2014: e199). 2. Polycategoriality in Nahuatl, Salishan, Wakashan, Siouan, and Mayan Some North American languages are famous for allowing action roots and object roots to occur in predicating function and in referring function without special function indicators – i.e. without copulas or relativizers. (1) Classical Nahuatl (Uto-Aztecan; Launey 1994) a. Tzàtzi in pilli. cry DEF child ‘[The Ø child] [Ø cries].’ b. Pilli in tzàtzi. child DEF cry ‘[The one who cries] [is a child].’ (4) a. ø-tzàtzi [ in coyo-tl ] (from Peralta Ramírez’s abstract) 3SUJ-gritar DET coyote-ABS ‘El coyote grita’ b. ø-coyo-tl [ in ø-tzàtzi ] 3SUJ-coyote-ABS DET 3SUJ-gritar ‘Es coyote el que grita’ (2) Lillooet (= Stháthimcets, Sƛ’aƛ’imxǝc; Salishan; Davis et al. 2014: e196) a. action root → predicating; object root → referring ƛ’iq ta=kʷúkʷpiʔ=a arrive DET=chief=EXIS ‘[The Ø chief] [Ø arrived].’ b. object root → predicating šmúɬač ta=kʷúkʷpiʔ=a woman DET=chief=EXIS ‘[The Ø chief] [is a woman].’ c. action-root → referring kʷúkʷpiʔ ta=ƛ’iq=a chief DET=arrive=EXIS ‘[The one who arrived] [is a chief].’ 3 (3) Nuuchahnulth (Wakashan; Swadesh 1939: 78) a. Mamo:k-ma qo:ʔas-ʔi. work-PRS.3SG man-DET ‘[The Ø man] [is working].’ b. Qo:ʔas-ma mamo:k-ʔi. work-PRS.3SG work-DET ‘[The one who is working] [is a man].’ (4) Lakhota (Siouan; Van Valin 2019: 4) a. Wíŋyaŋ kiŋ hená wačhí-pi-kte-šni. woman DEF those dance-PL-FUT-NEG ‘[Those Ø women] [will Ø not dance].’ b. Wačhí-pi kiŋ hená wíŋyaŋ-pi-kte-šni. dance-PL DEF those woman-PL-FUT-NEG ‘[Those dancers] [will not be women].’ (5) Yucatec Maya (Vapnarsky 2013: 44) a. T-u-hantah le kay-o‘. CPL-3.ERG-eat DEF fish-DET ‘She [Ø eats] [the Ø fish].’ b. Kay le t-u-hantah-o‘. fish DEF CPL-3.ERG-eat-DET ‘[Fish is] [what she eats].’ How are these languages different from English and other European languages? This can perhaps be formulated as follows (inspired by Vapnarsky & Veneziano 2017): (6) The polycategoriality parameter Value A: Object roots require a copula in predicating function, and action roots require a relativizer in referring function. Value B: Object roots do not require a copula in predicating function, and action roots do not require a relativizer in referring function. Value A languages: English, Spanish… Value B languages: Nahuatl, Lillooet, … (Alternatively perhaps: Omnipredicativity parameter, Launey 1994; 2004) Value B is a kind of “marker economy” – • these languages lack function indicators where other languages have them: • the propositional act function (predicating vs. referring, cf. Croft 2000) remains implicit and is inferred from the context (e.g. the word order). Marker economy is frequently found elsewhere, because many grammatical meanings can be inferred from the context, e.g. 4 number marker economy many languages do not make a singular-plural distinction in nouns, especially when combined with numerals cf. Haspelmath (2005) in WALS tense marker economy many languages do not make a present-future distinction in verbs, especially when combined with time adverbials cf. Dahl & Velupillai (2005) in WALS evidential marker economy many languages do not make an evidential distinction in verbs cf. de Haan (2005) in WALS Thus: Nahuatl, Lillooet, Nuuchahnulth, Lakhota, and Yucatec exhibit function indicator economy, of a kind that one might call “polycategoriality” (cf. Vapnarsky & Veneziano (eds.) 2017). But does this mean that there is “no noun-verb distinction” in these languages? A controversy in the literature: Jelinek & Demers (1994): Straits Salish lacks a noun-verb distinction Davis et al. (2014: e199): this is wrong – these languages HAVE such a distinction BUT: What exactly is a “noun-verb distinction”? 3. Language-particular classes What does it mean to say that a language “has a noun-verb distinction”? Clearly, it makes a general statement about the language, which goes beyond its own language system – it is a comparative claim. But morphosyntactic word classes are language-particular, created in order to describe a language in a complete way with respect to its morphosyntactic constructions. 5 e.g. Maricopa (Gordon 1986: 29; 90): Pluralizables: Nonpluralizables: humar/humaar ‘child(ren)’ avhay ‘dress’ mhay/mhaa ‘boy(s)’ shviily ‘feather’ ‘iipaa/’iipash ‘man/men’ vii ‘rock’ hmii/hshmee ‘be tall’ hly’a ‘month’ uukup/shuukuup ‘dig’ kwnho ‘basket’ nak/anaak ‘sit’ nyik ‘rope’ e.g. Godoberi (Kibrik (ed.) 1996) Gendered Words: Genderless Words: (b-)ičã- ‘sell’ bit’i- ‘tear’ q’aruma(-b) ‘greedy’ mik’isi- ‘young’ e.g. English Much-Premodified Words: Non-Much Words: money pay admire love regret hate We don’t have much money. *We don’t much pay. We much admire your technique: *We much hate your uncle. We much regret the inconvenience. *We much love our teacher. e.g. French Article Person Words Articleless Person Words la mère ‘the mother’ Hélène le frère ‘the brother’ Jacques In what way can such language-particular classes be associated with general classes? This is not clear – the literature usually focuses on the labeling question, e.g. Schachter (1985: 7): “The label noun is assigned to the class of words in which occur the names of most persons, places and things.” This is indeed how we often label our classes – and if there is no class called “noun”, we often create one by lumping some classes. (e.g. “Common Noun”, “Proper Noun”) But sometimes we do not follow this advice for labeling – cf. Maricopa, where we DO NOT equate the Nonpluralizables “nouns”, even though most object roots are in that class: Pluralizables: Nonpluralizables: ‘iipaa/’iipash ‘man/men’ vii ‘rock’ hmii/hshmee ‘be tall’ hly’a ‘month’ uukup/shuukuup ‘dig’ kwnho ‘basket’ nak/anaak ‘sit’ nyik ‘rope’ 6 Regardless of the labeling, how can we tell in general whether there is a “noun-verb distinction” in a language? 4. Why the “noun-verb distinctness question” cannot be answered Many linguists have formulated the issue concerning noun-verb similarities as follows: Does language X have a noun-verb distinction? cf. Bloomfield (1933) Lyons (1977) Chung (2012) Davis et al. (2014) ... and many others But in practice, this question cannot be answered in a way that does not already presuppose the answer (see Haspelmath 2023). To answer it in an empirical way, we would need either (A) or (B): (A) a way to identify “noun” and “verb” by the same positive criteria way in all languages (B) complete knowledge of all the possibly relevant morphosyntactic constructions of the language But do we have either (A) or (B)?? Ad (A) (positive criteria): • Since each language has different morphosyntactic constructions, “noun” and “verb” cannot be identified uniformly in all languages – unless we limit the choice of valid criteria to those that are potentially relevant in all languages. • Features such as articles, gender, number, tense, aspect are very variable across languages and cannot be used to identify nouns and verbs across languages. (• Only function indicators might be a possible basis for comparison: Copulas are characteristic of nouns in predicating function, and relativizers are characteristic of verbs in referring function.) Ad (B) (complete knowledge): • One might object: No, other criteria are valid to establish a noun-verb distinction as well – we just need to look for them very hard, because the distinction is often “subtle”. 7 • BUT: We can never be sure that we have examined all the potentially relevant constructions across languages – which means in practice that it is not possible to demonstrate the lack of a noun-verb distinction. “absence of evidence is not necessarily evidence of absence (of a noun-verb distinction)” (Davis et al. 2014: e197)? – but in practice, on this view, the absence of a noun-verb distinction cannot be demonstrated • MOREOVER: One could always claim that an alleged distinguishing criterion merely establishes subclasses. So can we take the fact that the Verbs of Straits Salishan can follow Auxiliaries as “evidence for a noun-verb distinction” (cf. Montler 2003)? NO – because one could always say that Straits Salishan has several subclasses of its “Full Words” class. [In fact, one might even claim that English has a class of “Full Words”, characterized by the possibility of a Number suffix -s (with Plural meaning in one subclass, e.g. cat-s, and Singular meaning in another subclass, e.g. eat-s). This sounds absurd, but it isn’t impossible.] Thus, all we can do, in practice, is ask the following questions: (A) Are there universal tendencies in the way languages treat object roots and action roots in referring function and in predicating function? (B) Are there salient differences between languages in this respect? We already saw that there is a salient difference – the polycategoriality parameter. It is not clear to me that this reflects any kind of really “deep” difference between languages (cf. Launey’s “omnipredicativity”), but it is certainly “salient” – and apparently cross-linguistically uncommon. Absence of a copula is common – cf. Stassen (2005) in WALS. But absence of a relativizer does not seem to be common (though I do not know of any systematic cross-linguistic research). 5. Cultural universalism and cognitive universalism (A) Are there universal tendencies in the way languages treat object roots and action roots in referring function and in predicating function? YES (Croft 1991; 2000): Languages show a very strong tendency to have function indicators for those functions that are not usually associated with semantic root types: 8 – relativizers for action roots in referring function – copulas for object roots in predicating function [WHO arrives first] gets a prize. (relativizer) Mother [WAS a linguist]. (copula) Languages do not have relativizers for the object roots in referring function, or copulas for action roots in predicating function: *[WHO mother] left for a conference. (relativizer) *Mother [WAS arrived first]. (copula) This generalization is part of a larger universal: (2) The grammatical form-frequency correspondence universal When two grammatical construction types that differ minimally occur with significantly different frequencies, the less frequent construction tends to be overtly coded, while the more frequent construction tends to be zerocoded (Haspelmath 2021). Function indicators occur in the less frequent construction types, i.e. when object roots occur predicatively or action roots occur in referring function. This universal is a cultural universal, because it has a good functional-adaptive explanation: Languages use short forms for frequently occurring meanings because speakers try to expend only as much energy as they must (Zipf 1935; Gibson et al. 2019). Cultures vary widely across human populations, and most anthropologists study cultural peculiarities – but whenever aspects of cultures are universal, they seem to have functional-adaptive explanations. By contrast: Cognitive universalism starts out with the idea that similarities between languages are due to a biocognitive blueprint – an innate toolkit of categories (“UG”). If there is a small set of innate categories, then it makes sense to use the Uniformity Principle as a heuristic: (7) Uniformity Principle “In the absence of compelling evidence to the contrary, assume languages to be uniform, with variety restricted to easily detectable properties of utterances.” (Chomsky 2001: 2) (But there is no good reason to assume that there is a rich innate toolkit of categories – most language universals can be explained without such a strong assumption, which is implausible anyway for biological reasons, cf. Berwick & Chomsky 2016). 9 If one adopts the strong biocognitive toolkit assumption as a leading asumption, it is easy to “find evidence” for it – any morphosyntactic property that remotely looks like a “noun-verb” distinction can be taken as evidence for it. The “noun-verb” distinction can look very different in different languages (e.g. Nonpluralizable/Pluralizable in Maricopa), but one could still say that all languages “have a noun-verb distinction” (even if it is very “subtle”). HOWEVER: this carries a danger of Eurocentrism, because in practice, it implies that we should look for those categories that happen to be salient in our current theories, which tend to be strongly influenced by English (plus Latin, German, Spanish and other influential languages). Each claim that “the distinction between X and Y is universal” is virtually impossible to prove wrong, because one cannot show (in practice) that a language completely lacks the distinction. (See Haspelmath 2012 on Chung 2012, which presents a rather Eurocentric view of Chamorro word classes.) A Maricopa linguist could claim: English is an example of category neutrality, because it lacks the important distinction between Pluralizables and Nonpluralizables – the counterparts of both classes are pluralizable in English: e.g. Maricopa (Gordon 1986: 29;90): Pluralizables: Nonpluralizables: humar/humaar ‘child(ren)’ avhay ‘dress’ mhay/mhaa ‘boy(s)’ shviily ‘feather’ ‘iipaa/’iipash ‘man/men’ vii ‘rock’ hmii/hshmee ‘be tall’ hly’a ‘month’ uukup/shuukuup ‘dig’ kwnho ‘basket’ nak/anaak ‘sit’ nyik ‘rope’ This would be a Maricopa-centric view of the world, though a Maricopa cognitive universalist could point out that if you look harder in English, you do find some words that are not pluralizable (e.g. milk, rice, should) – so this would be evidence for the cognitive universality of the two Major Parts of Speech after all. 6. Conclusion Jelinek & Demers (1994) were not completely wrong, and there was no reason for Eloise Jelinek to “recant” – Salishan languages are indeed strikingly different from English and other European languages. If one allows for cultural universals in addition to bicognitive universals, then there is no reason to start out with the assumption of cognitive universals.