Skip to main content

The lexical semantics of lexical categories

Periodic Reporting for period 2 - LexsemLexcat (The lexical semantics of lexical categories)

Reporting period: 2020-03-01 to 2021-08-31

The distinction between the major lexical categories of noun, verb and adjective figures into myriad linguistic generalizations and has been a center of gravity in the study of language since antiquity. Notwithstanding their importance, lexical categories are poorly understood (see e.g. Baker & Croft 2017). Outstanding is whether there are generalizations about the meanings words in the major categories have. Many have claimed there are, and proposed theories linking meaning and category, in a one to one fashion. Such theories have been criticized, however, in light of clear counterexamples, and consequently the search for a universal link between meaning and category is perceived by many to have been unsuccessful (see von Fintel & Matthewson 2008). This project recasts the search for a link, in the spirit of recent work (Francez & Koontz-Garboden 2017: Chapter 5), not as one for a one-to-one mapping, but for constraints on meaning induced by category. The project targets a domain where the set of relevant meanings is small, but where there is variation in category: property concept sentences—sentences like (1) He is very clever, whose main predicate is an adjective or, (2) akwai shi da waayoo `He is very clever (lit: He exists with cleverness; Hausa; Newman 2000:179)', whose main predicate is not an adjective, but is translated by a sentence whose main predicate is an adjective in languages with a large class of them. Although (1) and (2) have the same meaning, their component parts do not. Recent work shows that the words in property concept sentences that introduce the descriptive content (clever in (1), waayoo (2))--property concept words--vary in meaning, not just in category (Dixon 1982). With three postdocs, this project draws on a typological survey and in-depth fieldwork to examine the crossclassification of meaning and category in property concept words to shed light on the semantic nature of nouns, verbs, and adjectives generally.
The project began with work on WP1, devoted to devising a detailed questionnaire used to collect data from existing grammars and dictionaries on the category and meaning of `property concept’ words (words with the meanings of adjectives across languages). Using this, we have uncovered data showing that not only verbal property concept lexemes, as in Washo, but adjectives (as in Tswefap, see Clem 2019, for example) can have a class of meaning which the record previously showed only verbs to have (so-called ‘degreeless meanings’).
Second-hand data evaluated during this phase, and later followed up on as part of WP4 (see Bowler 2020), suggest that in Warlpiri (WP4) nouns too can have this same kind of meaning. A clear result of WP1, therefore, is that this particularly class of meaning is found among nouns, adjectives, and verbs. This finding led us to investigate the nature of degreelessness generally. The results of this investigation
show that the crosslinguistic data do not support the orthodox view of the Degree Semantics Parameter as binary parameter which languages simply have a positive or negative setting for. Instead, we have recast degreefulness as tied to variation in whether languages do or do not have particular elements which support that kind of meaning, such as constructions of measurement, particular kinds of comparative constructions, etc.

This work, in turn, has raised questions about the status of degrees in languages where there is little evidence for them. In order to investigate this question, we are undertaking detailed investigations in the contexts of WP3 and WP4 of Washo and Warlpiri respectively. In the case of Washo, we have found that while the language remains degreeless on a wide-scale found among few other languages (Bochnak in prep), the language invokes possession in the predication of property concept lexemes, much like Hausa does. This work
has therefore opened up a hither-to unknown corner of linguistic typology, showing that degreelessness and property concept possession crosscut one another. The work on degreelessness in Warlpiri has focussed on the comparative construction, which is like that found in Washo in being conjoined, so that ‘Kim is taller than Sandy’ is rendered as ‘Kim is tall and Sandy is short’. Warlpiri is unlike Washo and more like English, however, in allowing such a comparative to be used in contexts where, for example, Kim and Sandy differ only minimally in their height. Bowler has shown that this fact can be understood with reference to particular aspects of the nominal nature of property concept words in Warlpiri and the nature of noun phrases in Warlpiri generally, raising important questions about the ways in which comparatives can vary across languages both in their form and meanings. Another major strand of work in WP3 has collected additional data to expand on the database `Verbal Roots Across Languages’ to show that a property concept word (e.g. red) is more likely to have the same form as a change into that property concept (e.g. redden) just in case the property concept word is a verb, rather than a noun or an adjective. This finding is confirmatory evidence for a key project hypothesis about the nature of verb meaning: that only verbs can describe changes of state.

In WP2, we have investigated adjectives in Fijian and Basaa. Fijian was selected early on, due to the claim that it was a language with adjectives that were degreeless. We have shown that while the property concept lexemes are indeed adjectival, the facts do not support a degreeless analysis, at least with the constructions investigated. Investigation of Basaa has offered rare evidence for the hypothesis that the meanings of adjectives (e.g. strong) is the same as the meanings for have + mass property concept noun constructions (e.g. have strength), allowing us to shed light on underlying meanings of property concept words and offering some evidence for a universal underlying semantic core for them.

This result has fed into work which falls under WP5. Drawing on crosslinguistic results of WP1, we have collected data from five diverse languages that show that there are affixes that create property concept words that are nouns, adjectives, or verbs that are at the same time possessive. Alongside results of WP2 and WP3, this result provides additional evidence for a universal underlying core meaning to property concept words. Another strand of word under WP5 considers what it means when property concept words and words that are not property concept words appear in the same grammatical contexts. Work on Logoori, for example, has shown that a single semantic generalization underlies the use of a wide range of word classes with the verb kudoka ‘arrive/be enough/reach/must’, thereby offering evidence for a particular meaning component in the meanings of property concept words. Additional work considers what can be learned from the derivation of change of state verbs from property concept words, both in English and crosslinguistic context, about the underlying meaning of property concept words as against adjectives with other kinds of meanings.
Much of our work on individual languages draws on the traditional methods of linguistic fieldwork and theoretical syntax and semantics. Much of it also, however, uses language typology---the comparison of many languages to one another in search of commonalities and differences---as a tool of formal analysis. The way in which we deploy it is not entirely unique to our project, but is arguably at the cutting edge, in that it is not frequently used as a tool for the development and testing of hypotheses in formal semantics. The way in which we have brought it to bear on the issue of lexical categoryhood is at least unconventional, if not novel, in that it draws on larger scale typological survey to bear on detailed semantic questions of a kind which have traditionally be considered only from the perspective of single languages. Our approach maintains this focus on individual languages, but seeks to use larger scale crosslinguistic surveys as further tests on our hypotheses. The fruits of this approach are seen particularly clearly in (i) our work on degreelessness (Bochnak et al 2020), (ii) the work on categorization (Hanink and Koontz-Garboden 2021), (iii) the work on root meanings (Beavers et al in press), and (iv) the work on verbal polysemy and verbhood (Koontz-Garboden et al 2019).