Lexical Flexibility: A bibliography

Lexical flexibility is the ability for a lexeme to function as multiple parts of speech with little to no structural or behavioral differences between functions. This bibliography covers publications which discuss the concept of lexical flexibility directly, as well as relevant publications which focus on related topics such as conversion / zero-derivation or lexical categories (both crosslinguistically and in specific languages).

Total references: 23

Last updated: February 2, 2019


  1. Bach, Emmon. 1968. Nouns and noun phrases. In Emmon Bach, & Robert T. Harms (eds.), Universals in linguistic theory. Holt, Rinehart & Winston.


    Bach argues that there is no underlying difference between nouns, verbs, and adjectives at the level of deep structure. All three are instead subsumed under the term contentives, which is meant to represent something similar to the idea of a predicate in logic, and contrasts with terms, i.e. noun phrases or the arguments to a predicate. In short, Bach revises the machinery of the transformational grammar of his day so that the deep structure of sentences more closely resembles the logical systems used in philosophical logician by philosophers such as Rudolph Carnap. Bach notes:

    In particular, such systems do not have any subdivision of ‘lexical items’ into nouns, verbs, and adjectives. Much more basic is the distinction between variables, names, and general ‘predicates’ which can be n-placed with respect to the number of terms that can occur as their arguments. (p. 121)

    This work is notable for being the first (as far as I am away) to use the term contentives to refer to a category that subsumes the traditional major lexical categories of noun, verb, and adjective. Although Bach is not cited as a source, Bach’s usage of the term contentives in this manner was apparently adopted by Kees Hengeveld (2004) in his approach to flexible categories.

  2. Baker, Mark C. 2015. Nouns, verbs, and verbal nouns: Their structure and their structural cases. In Joanna Błaszczak, Dorota Klimek-Jankowska, & Krzysztof Migdalski (eds.), How categorical are categories? New approaches to the old questions of noun, verb, and adjective (Studies in Generative Grammar 122). Mouton de Gruyter. DOI: 10.1515/9781614514510-003.


    Baker begins the chapter by noting that questions about categoriality can apply to roots, whole words, or syntactic elements, and that his research explores just the status of syntactic elements. In his 2003 book Lexical categories, Baker argues for a distinction at level of syntactic elements between nouns, verbs, and adjectives. This chapter extends that analysis, claiming that nominal projections do not allow specifiers whereas verbal projections do, and that this fact explains certain behaviors of verbal nouns (gerunds). Thus Baker’s take on lexical flexibility is that, at the level of syntactic units at least, lexical categories are strictly delineated in terms of a number of well-defined syntactic properties. That is, lexical words are not flexible at the level of syntactic units, but fall into clear categories.

  3. Boas, Franz. 1911. Introduction. In Franz Boas (ed.), Handbook of American Indian Languages, Part 1 (Bureau of American Ethnology Bulletin 40). Smithsonian Institution.


    Boas’ introduction to the Handbook of American Indian languages is best known for debunking scientific racism and evolutionary anthropology, advocating an objective, non-normative approach to anthropology instead. However, Boas also discusses the grammatical features of Native American languages at a high level, and is sometimes quoted in discussions of lexical flexibility for being one of the first to note the incommensurability of categories across different languages. Two quotes in particular are commonly cited:

    In the treatment of our noun we are accustomed to look for a number of fundamental categories. […] None of these apparently fundamental aspects of the noun are necessary elements of articulate speech. (p. 36)

    We conclude from the examples here given that in a discussion of the characteristics of various languages different fundamental categories will be found, and that in a comparison of different languages it will be necessary to compare as well the phonetic characteristics as the characteristics of the vocabulary and those of the grammatical concepts in order to give each language its proper place. (p. 43)

    However, these quotes are typically taken out of context: Boas in these quotes is discussing grammatical rather than lexical categories, and thus these statements are not directly relevant to discussions of lexical flexibility. Boas does discuss lexical categories in passing, noting that many Native American languages treat adjectives as verbs, but he does not discuss this phenomenon or lexical categories in any detail.

  4. Booij, Geert & Jenny Audring. 2018. Category change in construction morphology. In Kristel Van Goethem, Muriel Norde, Evie Coussé, & Gudrun Vanderbauwhede (eds.), Category change from a constructional perspective (Constructional Approaches to Language 20). John Benjamins. DOI: 10.1075/cal.20.08boo.


    Morphological constructions can be formalized as schemas that specify semantic and formal output properties of complex words. Such schemas impose these output properties on their constituent words through various coercion mechanisms. In this article we focus on coercion-by-override and the concomitant category change. Our data are mainly from Dutch. The meaning of a syntactic or morphological construction can override the lexical meaning of a word in that construction. Morphological schemas may therefore change the semantic class of the base word. Semantic coercion may be accompanied by changes in word class. Morphological schemas may receive a higher degree of productivity within certain syntactic constructions, a phenomenon known as embedded productivity. Thus, morphological schemas contribute to the creativity and flexibility of the language system.


    Booij & Audring summarize the way in which category change is handled within the framework of construction morphology, and exemplify this process with data from Dutch. In the construction morphology approach of Booij (2010a, 2010b), items of one syntactic category may appear in slots for words of other syntactic categories. This is taken to hold for both syntactic and morphological constructions. This approach implies that words are prespecified for their syntactic category, but that this category may be “overridden” in certain contexts, and that these categories may be cleanly delineated. Thus a construction morphology approach to lexical flexibility still contains latent assumptions stemming from classical parts of speech theory.

  5. Braithwaite, Ben. 2015. Nuu-chah-nulth nouns and verbs revisited: Root allomorphy and the structure of nominal predicates. In Joanna Błaszczak, Dorota Klimek-Jankowska, & Krzysztof Migdalski (eds.), How categorical are categories? New approaches to the old questions of noun, verb, and adjective (Studies in Generative Grammar 122). Mouton de Gruyter. DOI: 10.1515/9781614514510-004.


    Braithwaite argues that allomorphy in free vs. bound forms of certain roots in Nuuchahnulth (a.k.a. Nootka; Wakashan > Southern Wakashan, British Colombia) provides evidence for positing a null copula, which blocks allomorphy in nominal predicates, thus explaining the differential behavior of free vs. bound forms. The differential behavior of interest in this chapter is the fact that verbal roots show allomorphy when in free vs. bound position, while nominal roots do not. However, the presentation of the data in this chapter downplays the fact that seemingly nominal roots do in fact show allomorphy in free vs. bound contexts; this allomorphy is simply associated with certain semantic shifts, and is often lexicalized with a new, idiosyncratic meaning. Thus while this chapter does point out different behavioral tendencies between one class of roots and another, no explanation is offered for instances where the two classes of roots behave similarly.

  6. Broschart, Jürgen. 1997. Why Tongan does it differently: Categorial distinctions in a language without nouns and verbs. Linguistic Typology 1(2): 123–165. DOI: 10.1515/lity.1997.1.2.123.


    This article appeared in the second issue of Linguistic Typology, at a time when lexical categories was receiving a good deal of attention in the typological literature (Frans Plank’s bibliography of word classes having been just published in the first issue of the same journal). As such, this article is influential in the literature on lexical flexibility, and the term “Broschartian language” is still used when discussing the possible ways to analyze cases of lexical flexibility.

    Broschart argues that there are three possible ways to analyze the seeming lack of a noun-verb distinction in Tongan: 1) the conversion hypothesis argues that the lexical categories of noun and verb are distinct, but simply unmarked, so that all cases of lexical flexibility are simply conversion; 2) the nominal vs. verbal syntax hypothesis states that lexical categories are not lexically specified, but instead are syntactic slots; and 3) the structural alternative hypothesis argues that Tongan does not possess the categories noun and verb, on either the lexical or syntactic levels, and that another distinction is relevant instead.

    Broschart adopts the third approach, and argues that Tongan is a type/token language, i.e. a language in which all lexical items are “predicates” as the term is meant in predicate logic. Any lexeme in the language may be used to predicate, and what the grammar of the language encodes is instead a distinction between whether a lexical item is being used to predicate or to indicate a term which is being predicated over. Broschart’s approach thus mirrors that of Bach (1968), who argued for the same distinction but at the level of deep structure in early formulations of generative grammar. In some ways this approach is also similar to calling Tongan an omnipredicative language. Though well-known and widely discussed, I know of no other researchers who have adopted a Broschartian analysis of languages purported to exhibit lexical flexibility.

  7. Błaszczak, Joanna & Dorota Klimek-Jankowska. 2015. Noun and verb in the mind: An interdisciplinary approach. In Joanna Błaszczak, Dorota Klimek-Jankowska, & Krzysztof Migdalski (eds.), How categorical are categories? New approaches to the old questions of noun, verb, and adjective (Studies in Generative Grammar 122). Mouton de Gruyter. DOI: 10.1515/9781614514510-005.


    This chapter is a thorough and very useful review of the psycholinguistic and neurolinguistic work on the distinction between nouns and verbs, including processing, language acquisition, aphasia, and neuroimaging studies. These studies locate this distinction at the conceptual, lexical, or morphological level, but the authors suggest that information about the distinction is organized into a distributed network that functions at all three levels. This perspective aligns well with Taylor’s (2003) view that the prototype structure of lexical categories can be fruitfully examined at both the conceptual and lexical levels.

  8. Cetarnowska, Bożena. 2015. Categorial ambiguities within the noun phrase: Relational adjectives in Polish. In Joanna Błaszczak, Dorota Klimek-Jankowska, & Krzysztof Migdalski (eds.), How categorical are categories? New approaches to the old questions of noun, verb, and adjective (Studies in Generative Grammar 122). Mouton de Gruyter. DOI: 10.1515/9781614514510-006.


    This chapter concerns itself with the categorial status of relational adjectives, which are adjectives derived from nouns and whose meaning can be paraphrased as ‘relating to N’. They contrast with qualitative adjectives, which denote the quality of some entity, and may be morphologically simplex, or derived from nouns, verbs, or other adjectives. Cetarnowska argues that relational adjectives in Polish are pseudo-adjectives, i.e. “deep nouns”, which are dominated by a nominal projection.

  9. Constantinescu, Camelia. 2015. Degree modification across categories: Nouns vs. adjectives. In Joanna Błaszczak, Dorota Klimek-Jankowska, & Krzysztof Migdalski (eds.), How categorical are categories? New approaches to the old questions of noun, verb, and adjective (Studies in Generative Grammar 122). Mouton de Gruyter. DOI: 9781614514510-006b.


    Constantinescu seeks to answer in this chapter whether nouns show gradability in the same way as adjectives. Consatantinescu states that the nominal domain does not show such gradibility, and that this constitutes a fundamental distinction between the two classes of words. The argument provided in evidence of this analysis is that words like big and real are in fact not degree modifiers, but always size adjectives and epistemic adjectives respectively, and can and should be analyzed this way even when they seemingly act as degree modifiers. Constantinescu thus denies any meaningful overlap between nouns and adjectives.

  10. Denison, David. 2018. Why would anyone take long? Word classes and Construction Grammar in the history of long (Constructional Approaches to Language 20). John Benjamins.


    I review the word classes proposed for long in such idiosyncratic English usages as I won’t be/take long, all night long. Although adverb fits most of the contentious data best, sometimes the word class is underdetermined. I suggest that long exhibits adjective ~ adverb underspecification from Old and Middle English onwards and can also be a semi-grammatical, decategorialised word. We need not assume that every word in every grammatical sentence must belong to one and only one word class (Denison, 2013). At the phrasal level the distribution is less anomalous and correlates with semantic and pragmatic features. Accordingly, it is sensible to describe the history of such usages in Construction Grammar terms. Recent Danish developments make an intriguing comparison.


    Denison argues that the word long in the history of English has undergone partial grammaticalization in certain of its senses, and in doing so has undergone a process of partial degrammaticalization, so that in certain constructions it exhibits adjective~adverb underspecification. Denison lays out data from diachronic corpora to show the trajectory by which long developed each of its known uses. Denison thus falls into the class of theorists who explain lexical flexibility via appeal to categorial underspecification.

  11. Gil, David. 1994. The structure of Riau Indonesian. Nordic Journal of Linguistics 17: 179–200.


    This article is well-known in the literature on lexical flexibility, because in it Gil adopts the radical claim that Riau Indonesian has no lexical categories - all lexical words belong to a single lexical category S. As such, Riau Indonesian is rife with zero conversion (although Gil points out that the use of the term “zero conversion” is inaccurate because there are not distinct lexical categories in the lexical to convert between). Gil provides a number of examples (taken from naturally-occurring conversation, with analysis based on extensive fieldwork) aiming to show that any lexical item can occur in any position, and that the resultant meaning of any combination of items is straightforwardly compositional. While Gil claims that lexical items “have no typical or unmarked ontological category” (p. 190), he also notes that single-word expressions most typically have a default interpretation, even though other interpretations are sometimes available. Under a prototype-based approach to word meaning, it is expected that certain senses of a word are more central, and less semantically marked, than others. Based on cognitive research on prototypes, it would therefore be surprising if Gil’s extreme claim - that lexical items do not have a default ontological type / category - were true. Nonetheless, data from Riau Indonesian provides a strong challenge to all languages have lexical categories; at the very least, it requires that part-of-speech typologies significantly broaden their understanding of the ways in which those categories are realized, and the crosslinguistic variability.

  12. Gil, David. 1995. Parts of speech in Tagalog. In Mark Alves (ed.), Proceedings of the Southeast Asian Linguistics Society.Program for Southeast Asian Studies, Arizona State University.


    In this article, Gil applies his well-known analysis of Riau Indonesian (which he claims does not distinguish parts of speech; Gil 1994) to Tagalog, making in effect the same claims for both languages. His most significant claim is that Tagalog, like Riau Indonesian, possesses a single open syntactic category S. Key to his point is the idea that “anything can go anywhere”, that is, words in Tagalog do not seem to be limited in their syntactic distribution. If word class membership is defined in terms of syntactic distribution, then there is no basis for distinguishing syntactic categories in Tagalog. Gil notes that defining parts of speech in Tagalog via morphological criteria yields the same result, since almost any word can take voice-aspect morphology. While not every word in Tagalog can take voice-aspect morphology, the set of words which can appears to be lexically-specified - that is, they do not share any syntactic properties in common. This leaves Gil’s claim open to criticism, since, as various others have argued, the ability to take or not take voice-aspect morphology could be considered a valid criterion for distinguishing between different parts of speech.

  13. Givón, Talmy. 2017. The story of zero. John Benjamins. DOI: 10.1075/z.204.


    This volume seeks to explain the functional motivations for zero reference, i.e. zero anaphora and cataphora, most especially from cognitive and diachronic perspective. It comprises two sections: discourse or ‘ungoverned’ anaphora/cataphora and syntactic or ‘governed’ anaphora/cataphora. The running theme of the book is to show that zeros signal maximum referential continuity, and that this principle helps explain the distribution of zero reference crosslinguistically. Givón explains the use of zero for referential continuity via two cognitive principles: informational predictability (anaphoric), wherein predictable information need not be mentioned, and informational importance (cataphoric), wherein unimportant information need not be mentioned.

    The book spends little time discussing lexical categories, but does note one highly relevant connection for the study of discourse and lexical flexibility:

    The overwhelming tendency in natural human discourse is to make perceptually salient, compact and temporally-stable entities – most commonly nouns – the old-information chunks that ground new information. As a result, natural human discourse seems to be ‘about’ such nominal entities, which are then coded by various grammatical means as the topical referents in the discourse, most typically the grammatical subjects and objects of clauses. (Givón 2017: 32)

    A prediction that would seem to stem from this is that (back)grounding information is more likely to be coded using referring constructions than predicative constructions.

    Givón also notes that nominalization is a common diachronic pathway into subordination (with the exception of adverbial clauses), motivated by the fact that the nominalized referent is functioning as an argument of the clause. This raises the question whether, in highly flexible languages, we should similarly expect to see lexical items used as arguments coded with referential rather than predicative constructions.

  14. Hopper, Paul J. & Sandra A. Thompson. 1984. The discourse basis for lexical categories in Universal Grammar. Language 60(4): 703–752. DOI: 10.2307/413797.


    Most linguists who have investigated linguistic categories from a universal viewpoint have accepted the existence of two basic parts of speech, noun and verb. Other categories are found to be only inconsistently represented; thus adjective is manifested in many languages as a class of stative verb. Furthermore, individual languages often have intermediate categories such as gerund, which cannot be unambiguously assigned to a single category. We suggest here that the basic categories N and V are to be viewed as universal lexicalizations of the prototypical discourse functions of ‘discourse-manipulable participant’ and ‘reported event’, respectively. We find that the grammars of languages tend to label the categories N and V with morphosyntactic markers which are iconically characteristic of these categories to the degree that a given instance of N or V approaches its prototypical function. In other words, the closer a form is to signaling this prime function, the more the language tends to recognize its function through morphemes typical of the category—e.g. deictic markers for N, tense markers for V. We conclude by suggesting that categoriality itself is another fundamental property of grammars which may be directly derived from discourse function.


    Hopper & Thompson argue that categoriality is not a property of linguistic forms per se, but rather a matter of their function in discourse. Linguistic forms are said to be acategorial, although Hopper & Thompson acknowledge that most forms, as a result of their semantic affordances, have a propensity for one category or another. They define prototypical nouns as items which are discourse manipulable, and prototypical verbs as items which are construed as reported events. They provide empirical data showing that, when an item is not functioning to introduce a new participant into the discourse or to report an event (e.g. a continuing topic or a complement clause) it is less strongly marked as a member of the category noun or verb. Conversely, items which perform these functions are more likely to be explicitly marked with morphology relating to that function. They also posit an implicational universal wherein languages require derivational morphology to derive a noun from a verbal root, but can derive verbs productively from either nominal or verbal roots with no overt morphology. In other words, all languages are omnipredicative but do not necessarily show bidirectionality between nouns and verbs (to borrow terminology from Evans & Osada (2005)).

    This influential article was one of the first studies to prominently take a discourse-based approach to lexical categories, and served as a foundation for many other studies in functionalist linguistics, including especially work by William Croft. Croft’s typological-markedness approach echoes that of Hopper & Thompson in that it sees categoriality as prototypal rather than a property of specific lexical items which divide neatly into distinct categories. If Hopper & Thompson’s thesis is correct, it would seem to make two predictions: 1) In rigid languages (i.e. languages with more clearly delineated parts of speech), category-specific morphology should arise from the use of lexical items in discourse to either introduce a discourse-manipulable participant or report an event; 2) In flexible languages, the choice of category for a flexible wordform is dependent on its discourse function.

  15. Jacobsen, William H. 1979. Noun and verb in Nootkan. In Barbara S. Efrat (ed.), The Victoria conference on northwestern languages.British Columbia Provincial Museum.


    This article provides a thorough summary of the literature on the distinction between noun and verb as it pertains to Nuuchahnulth (a.k.a. Nootka; Wakashan > Southern Wakashan, British Colombia) and the surrounding languages. Jacobsen states that there have been two main reasons for claiming a lack of a distinction between noun and verb in these languages: 1) the fact that noun-like words exhibiting categories typically associated with verbs (tense, aspect, and mode); and 2) lack of evidence for distributional differences between the two categories; in particular, the fact that all words can act as either predicates or arguments. Jacobsen takes particular issue with the second claim, which he notes is the more substantive of the two. He provides evidence in these languages for a syntactic difference between argument and predicate, and shows that flexibility between noun and verb is often unidirectional: verbs may only appear as arguments in their nominalized form. However, Jacobsen also allows for instances of multiple class membership and lexicalization, and it is not clear how this concession affects his position.

    This article is especially useful for its thorough survey of the literature on this topic, and similarly thorough treatment of data from the relevant languages. It is influential in pointing out numerous empirical pieces of counterevidence to the claim that some northwestern languages completely lack a noun-verb distinction; however, it has little to say towards why the distinction is so blurred. The position in this article is thus not so different from Dixon’s (2010: 11a) approach to adjectives, which argues that this category can be found in any language provided one looks hard enough for the evidence.

  16. Klockman, Heidi. 2015. What are categories? Adjective-like and noun-like semi-lexical numerals in Polish. In Joanna Błaszczak, Dorota Klimek-Jankowska, & Krzysztof Migdalski (eds.), How categorical are categories? New approaches to the old questions of noun, verb, and adjective (Studies in Generative Grammar 122). Mouton de Gruyter. DOI: 10.1515/9781614514510-008.


    In this chapter Klockman presents an analysis of Polish numerals within the generativist framework. Polish numerals, like numerals in many languages (Corbett 1978) show a mix of properties of nouns and adjectives, with lower numbers being more adjective-like, and higher numbers being more noun-like. Klockman argues that this situation results from the fact that lexical categories are not syntactic primitives. Instead, lexical categories are derived from more basic phi-features. Elements with a full set of valued phi-features are nouns, and a full set of unvalued phi-features are adjectives. Items with various combinations of valued and unvalued (or incomplete) phi-features therefore exhibit a mix of properties. This is an especially interesting work because it treats lexical categories not as basic, but rather as derived from more foundational notions. The idea of lexical categories as emerging from the interactions of other, more basic properties is not one that appears elsewhere in the literature on lexical flexibility. In (most?) all other approaches, lexical categories within particular languages are viewed as either nonexistent, or existent with fuzzy boundaries and/or non-central cases.

  17. Mithun, Marianne. 2017. Polycategoriality and zero derivation: Insights from Central Alaskan Yup’ik Eskimo. In Valentina Vapnarsky, & Edy Veneziano (eds.), Lexical polycategoriality: Cross-linguistic, cross-theoretical, and language acquisition approaches (Studies in Language Companion Series 182). John Benjamins. DOI: 10.1075/slcs.182.


    The possibility of polycategorial lexical items, unspecified for category, raises interesting issues. One is the predictability of semantic relationships between polycategorial forms in predicating and referring uses. Another is the language- internal generality of the phenomenon, whether it necessarily holds of all units at a particular level. A third is its cross-linguistic generality and potential association with certain typological features. These issues are examined with examples from a language exhibiting striking similarities to the Mayan languages for which polycategoriality has recently been argued. Central Alaskan Yup’ik Eskimo appears to show robust polycategoriality at the root, stem, word, and clause level. A closer look indicates that the story is actually more interesting, and that the universality of polycategoriality remains an open question.


    Mithun presents numerous potential cases of lexical flexibility at the root, stem, and fully-inflected word levels in Central Alaskan Yup’ik (Eskimo-Aleut > Eskimo > Yupik, Alaska), and argues that these items are specified for lexical category in each case, and therefore should not be considered flexible. As a diagnostic for identifying lexical flexibility, Mithun notes that zero-derivation presupposes a basic/derived relationship, even if diachronically, whereas polycategoriality assumes a single abstract meaning. In the case of polycategoriality, more specific meanings in different constructions must be accounted for.

    Mithun presents the following arguments in favor of zero-derivation for seemingly flexible forms in Yup’ik:

    • Most roots are only used as nouns or verbs. 35% are nominal, 53% are verbal, and only 12% may be used as both.

    • There are no semantic commonalities to which roots may be used for which function. The possible uses of a root must be memorized on a lexeme-specific basis.

    • The semantic relationships between roots being used for different functions are not predictable, suggesting that the items in each pair constitute separate lexical items.

    • The same kinds of semantic relationships hold between homophonous roots as appear in derived stems. Some roots can alternate classes with no derivation, while others cannot. This suggests that homophonous roots are actually the result of zero-derivation.

    • Yup’ik derivational affixes are fully specified for both the lexical catgory they attach to and the lexical category that results. If two meanings are possible for a root, a derivational affix will utilize the meaning most prototypical to the category it selects for.

    • Despite homophony between nominal and verbal uses of fully-inflected words, the two categories have different syntactic distributions.

    • The homophony between nominal and verbal fully-inflected words is the result of insubordination and other diachronic processes.

  18. Nakayama, Toshihide. 2002. Nuuchahnulth (Nootka) morphosyntax (University of California Publications in Linguistics 134). University of California Press.


    This volume is a grammatical sketch of the Nuuchahnulth (a.k.a. Nootka; Wakashan, British Columbia) language. It is a revised version of Nakayama’s Ph.D. dissertation at the University of California, Santa Barbara, covering phonology, morphology, word classes / word formation, and syntax / argument structure, with a strong focus on discourse throughout. The sketch is based on natural discourse data collected during Nakayama’s fieldwork.

    Nuuchahnulth and other Wakashan languages have featured prominently in the literature on lexical flexibility, primarily due to claims by Sapir and Swadesh (Sapir & Swadesh 1939; Swadesh 1939) that the language lacks lexical categories. Jacobsen (1979) later took an opposing view, stating that lexical items in Nuuchahnulth are specified for lexical category. In this grammar sketch, Nakayama adopts a functional approach informed by diachrony, and argues that, while lexical categories do exist in Nuuchahnulth, they are not strongly grammaticized, and there is a low degree of specialization for lexical items. Instead, he states that lexical categories in Nuuchahnulth are a matter of repeated discourse tendencies rather than a set of structural properties.

    Nakayama’s approach to lexical flexibility shows similarities to Hopper & Thompson’s (1984) approach to lexical categories more generally in claiming that lexical categories in Nuuchahnulth are imposed via discourse rather than lexically-specified. Nakayama’s idea that parts of speech in Nuuchahnulth are “less strongly grammaticized” than other languages also has an interesting implication for the historical development of lexical categories: it suggests that lexical specification for parts of speech may emerge from repeated discourse tendencies that become routinized over time, so that a given lexical item become solely or at least strongly associated with a particular pragmatic function, and the set of constructions which encode that function.

  19. Plank, Frans. 1997. Word classes in typology: Recommended reading. Linguistic Typology 1(2): 185–192. DOI: 10.1515/lity.1997.1.2.185.


    In the summer of 1996 some thirty experts, all on record with relevant publications, were asked to nominale up to ten titles of really recom- mended reading on word classes—articles and books (or also chapters of books) that in their opinion had shed light on the problem of how and why to distinguish word classes, in general or with reference to individual classes, and especially on typological Variation in this respect. About half of them replied by mid-November, sometimes providing longer lists.

    The following bibliography compiles the recommendations of Alexandra Aikhenvald, Jan Anward, D. N. S. Bhat, Jürgen Broschart, William Croft, Inga Dolinina, David Gil, Kees Hengeveld, Aleksandr Kibrik, Edith Moravcsik, Frans Plank, Regina Pustet, Jan Rijkhoff, Leon Stassen, Sandra Thompson, Petra Maria Vogel, Arnfinn Vonen, and Anna Wierzbicka. (Thank you all for your cooperation.)

    An asterisk marks titles nominated twice; two asterisks are for three to five nominations; and three are for seven or more—that is, word-class world-class.


    A bibliography of crucial works in the typological study of word classes. Many of these sources treat directly with the issue of lexical flexibility as well (those sources are included in this bibliography).

  20. Swadesh, Morris. 1933. The internal economy of the Nootka word. Ph.D. dissertation. Department of Linguistics, Yale University.


    In Nootka, it is necessary to distinguish the word as a lexical unit from the word as a syntactic unit. The lexical word is a close-knit phonetic and semantic unit built out of a stem and any number of formative suffixes. As used in the sentence, the word may have the form of the basic lexical unit or may consist of the latter with the addition of any number of incremental suffixes which are phonetically and, to a certain extent, semantically distinguishable from formative suffixes.

    Incremental suffixes have the function of rounding out the word into a predication or otherwise preparing it for its function in the sentence; they express ideas of time, voice, mode, and person. Formative suffixes have the function of building up the lexical or unextended word on the basis of the stem; they include aspect formatives and derivational suffixes.

    Formal word types (parts of speech) in Nootka are two: particles and normal words. Particles include syntactic, imperative, and interjectional elements; they are characterized by a simple structure, a lack of aspect implication, and a limited capacity for extension by means of incremental suffixes. Normal words are all verbal in fundamental nature; they are formally characterized by a potentially polysynthetic structure, a definite aspect implication, and unlimited capacity for extension by means of incremental suffixes.

    The unextended word is semantically a potential predication concerning the potential subject of the word. (The term “subject of the word” is justified in Nootka because subject pronouns in the form of incremental suffixes are attachable to any normal word.) Some words constitute a unified potential predication, others a compound potential predication, depending on the type of derivational suffix employed in the construction of the word. Thus, ther is one type of suffix that always makes the word into a unified potential predication; suffixes of this type may be called governing suffixes. Another type of suffix sets a secondary potential predication alongside the original one and builds thus a compound potential predication; suffixes of this type may be called restrictive suffixes.

    The units in derivation are the underlying theme and the derivational suffix and the meaning of the derived form is a function of the meaning of these two units. To understand the semantics of derivation in Nootka, it is necessary to classify underlying themes and derivational suffixes according to semantic content. To this purpose, the following semantic classes are proposed: substantive (e.g., “to be a house”), state (e.g., “to be erect”), action (e.g., “to jump”), location (e.g., “to be in the houes”), time (e.g., “to be at night”), quantity (e.g., “to be three”), specification (e.g., “to be this one”). It is important to note that these are semantic and not formal classes; for, aside from particles, there is only one formal class, essentially verbal. The classification holds for derivational suffixes as well as themes except that the distinction bewteen governing and restrictive derivational suffixes is of even more fundamental importance. In addition to themes belonging to one or another of these classes, there are restricted themes (built up by means of restrictive suffixes and expressing compound potential predications) belonging to two or more semantic classes at once.

    The derivational economy of Nootka tends on the whole to be mechanical in nature, but two interfering processes need to be recognized: implicit derivation and specialization. By implicit derivation, a theme takes on the meaning of a derivative of that theme (e.g., “knife in the hand” is used for “having a knife in the hand”). By specialization, a theme that should have a general reference is used to refer to something more particular (e.g., “big creature” is the regular expression for “whale”). While specialization and some instances of implicit derivation are sporadic and unpredictable, there are certain types of implicit derivation that are actually or almost regular.


    Swadesh’s Ph.D. dissertation at Yale under Edward Sapir, this work provides a description of word structure and lexical semantics in Nuuchahnulth (a.k.a. Nootka; Wakashan > Southern Wakashan, British Colombia). Swadesh famously makes the claim that “Normal words [in Nuuchahnulth] are all verbal in fundamental nature” (p. iv), and that they are all of one type (i.e. a single lexical category). However, Swadesh also notes various selectional restrictions for certain affixes which partition the lexicon, such as a strict separation between suffixes which may attach to numeral stems vs. non-numeral stems (p. 112), or that certain suffixes require a “substantive” stem (p. 113). As such, Swadesh’s categorization of the lexicon is based on semantic meaning, and comprises the following: substantive (entity), state, action, location, time, quantity, and specification (indication). Swadesh partitions the lexicon in this way on the basis of the existence of a set of seven “special reference stems” which express these categories, and assumes that these stems represent a fundamental set of categories in the lexicon, despite there being no formal evidence elsewhere in the language. (Here, Swadesh anticipates what Levinson & Burenhult (2009) refer to as a semplate, or a semantic template that operates in more than one area of the grammar.) Swadesh also discusses various idiosyncrasies in the lexicon, in particular implicit derivation and specialization. In implicit derivation, a derived stem takes on a meaning that is more than the sum of its parts, e.g. a derived form that literally means ‘knife in hand’ is in actual practice used to mean ‘having a knife in hand’ (p. iv; pp. 43ff.). Some of these cases fall into classes of regular patterns, and others are idiosyncratic. By specialization, Swadesh refers to the use of a derived stem to refer to something more specific than the meaning of its parts implies, e.g. the use of a form literally meaning ‘big creature’ to mean ‘whale’ (p. 54). Like implicit derivation, some cases of specialization follow certain general patterns, while others are idiosyncratic.

  21. Taylor, John R. 2003. Linguistic categorization: Prototypes in linguistic theory, 3rdth ed. Clarendon Press.


    The central aim of this book is to show that linguistic categories are prototypal, i.e. exhibiting the properties of prototype-based categories. By this he means that categories a) often lack clear boundaries, b) have an internal structure wherein some members of the category are better representatives of the category than others, and c) the members of a category often show a family resemblance structure, i.e. they are a radial category. Taylor shows how prototype theory applies to linguistic categorization in three domains: 1) the categorization of experience (applying linguistic labels to our experience of the world); 2) the categorization of linguistic experience (abstracting linguistic types and schemas from individual tokens); and 3) the categorization of linguistic concepts (technical concepts and terminology in the field of linguistics). Taylor shows that all of these exhibit the behavior expected of prototype categories.

    This book was central in introducing linguists to the prototype theory of Rosch and Lakoff, and making these concepts more broadly accessible to the field. The book explicitly discusses several issues of central relevance for lexical flexibility, including polysemy and compositionality. Regarding lexical categories, Taylor notes that they exhibit a prototype structure as well, with some members of, for example, the category Noun being better members of the category than others. Taylor also questions the utility of the distinction between polysemy and homonymy; a more useful framework is to consider the various meanings of a form as having a family resemblance structure (meaning chains), wherein some meanings are more closely linked or related than others. This approach would also seem to be applicable to lexical flexibility: rather than viewing zero-derivation / conversion as a case of homonymy (or polysemy), it may be more useful to view those particular meanings of a form that have crossed the traditional part-of-speech boundary as simply non-central members of the lexeme’s meaning.

  22. Thompson, Sandra A. 1989. A discourse approach to the cross-linguistic category ‘Adjective’. In Roberta Corrigan, Fred R. Eckman, & Michael Noonan (eds.), Linguistic categorization (Current Issues in Linguistic Theory 61). John Benjamins. DOI: 10.1075/cilt.61.16tho.


    In a well-known paper, Hopper & Thompson (1984) argue that the lexical categories noun and verb are not lexically-specified but rather imposed by discourse. In the present paper, Thompson builds on this work to provide a discourse-based explanation for the fact that property concepts crosslinguistically sometimes pattern like verbs, and other times like nouns, and in other languages constitute their own class (with combinations of these attested for individual languages as well). Why are property concepts distributed across different classes, rather than forming their own class in every language? Thompson answers this question by examining the use of property concepts in natural discourse data from English (which is generally thought to have a distinct adjective class) and Mandarin (in which property concepts largely pattern like verbs). Thompson finds that property concepts generally perform two functions in discourse: to predicate a property of an already-introduced referent, in which case the property concept appears in a predicative adjective construction; or to introduce a new referent into the discourse, in which case the property concept appears as an attributive (modifying) adjective with a (relatively) semantically vacuous nominal head, such as, for example, it’s a religious thing. In the latter case, the attributive use is used to introduce a new referent with the particular defining property indicated by the adjective. This functional division is what explains the crosslinguistic distribution of property concepts across both nominal and verbal categories, with some languages exhibiting a distinct adjective category since property concepts are neither prototypical nouns nor verbs.

  23. Trousdale, Graeme. 2018. Change in category membership from the perspective of construction grammar. In Kristel Van Goethem, Muriel Norde, Evie Coussé, & Gudrun Vanderbauwhede (eds.), Category change from a constructional perspective (Constructional Approaches to Language 20). John Benjamins. DOI: 10.1075/cal.20.11tro.


    This chapter summarizes the contributions of the volume, Category change from a constructional perspective (Van Goethem et al. 2018), with the aim of understanding the notion of a linguistic category in the perspective of construction grammar. It highlights the roles of schema formation and analogy in category change, and emphasizes the centrality of gradualness / gradience in this process. The chapter provides a useful overview of some of the major insights of the chapters in this volume.