(Non-)universality of word-classes and words: The mid-20th century shift

Martin Haspelmath
Max Planck Institute for Evolutionary Anthropology, Leipzig

While looking at a range of views by grammarians on word-class distinctions (noun, verb, adjective etc.) and word division in two recent papers (Haspelmath 2011; 2012a), I was struck by what appears to have been a major shift of perspective: While the first half of the 20th century emphasizes the uniqueness of languages and the categorial differences between them, the second half starts out from the assumption that languages do not differ in their basic categories. (Elsewhere I called this distinction categorial particularism and categorial universalism; Haspelmath 2010.) There are some signs that the perspective adopted in the first half of the 20th century is now getting more attention again.

Earlier history

Let us start with some earlier history. According to standard accounts, language thinkers in the 17th and 18th centuries did not show particular interest in structural differences between languages. Even though Western scholars became better and better acquainted with some of the major Oriental languages during this period, the prevailing approach among grammarians and philosophers of grammar was to (try to) use the same concepts for all languages, and to assume that categories such as “noun”, “verb”, “preposition”, as well as “word” or “root”, were cognitively grounded and hence could be applied universally. In textbooks on the history of linguistics, this view is particularly associated with the grammaire générale tradition. Thus, according to Robins (1997: 141), in the Port Royal grammar, “the six cases of Latin are … assumed in other languages, though some of them were expressed by prepositions and word order”. But even though the spirit of Renaissance linguists in the 16th century may have been somewhat different, emphasizing the peculiarities of the modern European languages (e.g. Ramée’s recognition that nouns in French are recognized by number rather than by case), the classical teaching of nine parts of speech was rarely questioned (pronoun, noun, verb, participle, adverb, article, preposition, conjunction, interjection).

The 19th century gradually brought some changes, also influenced by the wider cultural movement of Romanticism. Languages were no longer thought to be mere expressions of human thinking, but different languages were often considered as manifestations of different worldviews and different cultural predilections of their speakers. This did not lead to very striking innovations in grammatical practice, and of course most grammarians were interested in the prestige languages Latin, Greek, German, French, etc. Serious interest in non-(Indo-)European languages among general linguists became widespread only in the 20th century. However, some 19th century linguists did observe that non-Indo-European languages are organized rather differently with respect to their word-classes (or “parts of speech”, using a calque from Latin partes orationis). In particular, linguists observed that some languages such as Hungarian and Turkish had conjugation-like person markers on their nouns as well as on their verbs, and they hinted that these languages did not make the same kind of sharp distinction between nouns and verbs.

Similarly, scholars began to doubt that the notion of “word” is universal. I am not aware of any doubts of the universality of the word in the 19th century, but the foundational concept “morpheme”, which makes the word less important than it was in the past, was coined by Jan Baudouin de Courtenay in 1880 (cf. Mugdan 1986). Moreover, the distinction between “agglutinative” and “flective” languages, which points to rather different word-building principles in different languages, goes back to the early 19th century (Friedrich von Schlegel and Wilhelm von Humboldt, cf. Greenberg 1974).

Particularism in the 20th century

In the first half of the 20th century, such (categorial particularist) views became mainstream. Franz Boas is particularly well-known for emphasizing that North American languages should not be squeezed into the Procrustean bed of European categories, that the categories chosen for description “depend entirely on the inner form of each language”, and that languages should be described “in their own terms” (Boas 1911). Otto Jespersen described English in a way that highlighted its “progress” over the ancient, more cumbersome languages, and along the way Jespersen coined a fair number of new terms for concepts of English grammar (e.g. extraposition, adjunct). Bloomfield (1933) noted that articles, demonstratives and possessives formed a single substitution class in English and created the new “determiner” category for them. A less well-known figure is Erich Drach, who first described German sentence structure in terms of a template consisting of a prefield, a left bracket, a middle field, a right bracket, and a postfield (Drach 1937). In main clauses, the auxiliary and the main verb occupy the left and right bracket, respectively, while the filling of the prefield is relatively free. This model of German sentence structure is still used by German syntacticians today (e.g. Müller 2010).

The Leningrad linguist Lev V. Ščerba described Russian syntax in the 1920s using a completely novel category, the “state category” (in Russian kategorija sostojanija), e.g. lexical items such as vidno ‘one can see’, bol’no ‘it hurts’, dosadno ‘it is annoying’. These are predicative words that are in some ways like adjectives or adverbs, but they occur only predicatively, like verbs, and some of them can even take objects. In his last paper, Ščerba (1945: 186) demanded that underdescribed languages should be studied “concretely, without seeing them through the prism of the researcher’s native language, or another language with a traditional grammar, which distorts the grammatical reality…” Thus, by this time Western linguists had emancipated themselves from the tradition of antiquity. Even though they still used many of the old terms, they became used to asking for justification before applying a traditional concept, and they were willing to posit novel concepts, and to recognize certain categories as unique to particular languages. This approach to language structure became known as “structuralism” in the second half of the century. Whereas 19th century linguists invested most of their efforts into elucidating historical relationships between languages, the study of individual synchronic language systems began to take centre stage in the 20th century.

One accompanying feature of this activity was the coining of new terminology, of which a few examples are given above. When each language has its own categories, in principle one would need new terms for each language. And indeed, we do find some thoroughly language-specific terminology such as “middle field”, “determiner” and “state category”.

What’s in a word

The first half of the 20th century also saw people asking for justification of the “word” concept, apparently for the first time. While the morpheme was easy to define (as a minimal sign, a minimal combination of form and meaning), it turned out that the word was more difficult to characterize. Bloomfield (1933: 178) attempted to define the word as “a minumum free form”, but this definition was not widely adopted, and some researchers doubted whether all languages had words. Thus, Hockett (1944: 255) claimed that “there are no words in Chinese”, and Milewski (1951) made similar remarks about North American languages. In the closing session of the 6th International Congress of Linguists, held in Paris in 1948, congress president Joseph Vendryes remarked that modern linguistics was in a crisis, and that linguists were not even in agreement on what a word is, one of the fundamental concepts of their object of studies (cf. Togeby 1949: 97; Haspelmath 2011: 71). The basic problem was that linguists did not find a good way to justify the distinction between affixes (prefixes and suffixes) and short function words (often clitics). It was often observed, for instance, that French subject person forms (je viens, tu viens, il vient, etc. for the verb ‘come’) behave like prefixes, even though they are written separately. Likewise, compounds and phrases are often impossible to tell apart (is French chemin de fer [path of iron] ‘railway’ a compound word or a phrase?).

Doubts about the universality of the “word” concept were very widespread in the structuralist period. Schwegler (1990: 45) summarizes his detailed overview of the view of this period:

After generations of 19th- and 20th-century linguists had taken the ‘word’ largely for granted, structuralists set out to define what in popular as well as in scientific circles was regarded as the basic unit of speech . . . this lively discussion eventually led to the now generally accepted conclusion that (a) the ‘word’ cannot be defined by a single (or for that matter, multiple) common denominator, and (b) not all segments of speech are ‘words’ in the proper sense of the term.

The more optimistic linguists of this period claimed that while there was no universally applicable definition of the word, words could be defined by language-specific criteria. Thus, Bazell (1958: 35) noted: “Now there is perhaps no unit over which there is less agreement than the word. If there is any agreement at all, it is that the word has to be differently defined for each language analysed.”

Thus, we see great efforts being made in the first half of the century (the structuralist period) to carefully define the categories that are used for description and to justify them for each language individually. In principle, this is still what our textbooks teach our students. But the practice of the second half of the 20th century was very different.

Generative universalism

In the wake of Chomsky (1957), linguists increasingly lost interest in the question of how to justify categories. The focus was now on stating rules formally, on rule interactions (derivations), and on the relations between “underlying structures” and actual forms. But the new enterprise of transformational grammar was at the same time intended as a way of explaining the possibility of language acquisition despite the poverty of the stimulus, and the proposed solution was the innate universal grammar. On this view, categories (or at least their constitutuent features) were thought to be universal.

Now one might have expected linguists to embark on an empirically grounded research project asking what these presumed universal categories are. Perhaps surprisingly, nothing of this sort happened. Linguists simply abandoned the practice of seeking the best categories for different languages and justifying them. Instead, they focused on their new tasks of expressing generalizations through abstract representations and ignored the insights of their structuralist predecessors. Since the late 1960s, students were increasingly discouraged from reading the earlier literature, so the categorial particularism of the structuralist period was largely forgotten. Instead of coming up with a new original set of categories, they fell back on what the school grammars offered all along: the traditional concepts inherited from Priscian and Donatus through the Middle Ages and the early modern era.

Nouns, verbs and adjectives

When Chomsky’s proposals from the 1960s were extended to other languages, few linguists proposed word-classes other than those used by Chomsky for English. And these were all very traditional. In his 1970 paper “remarks on nominalization”, Chomsky proposed that the four categories N (nouns), V (verbs), A (adjectives) and P (prepositions) were universal, mostly on the basis of a few observations on English. There was little discussion of other possible categorizations in the subsequent decades, and Baker (2003) essentially confirmed Chomsky’s views, using data from a wide variety of languages (but leaving aside prepositions, which are grammatical rather than lexical words).

When Miller (1973) discussed the phenomena that had led Ščerba a few decades earlier to posit a new category for Russian (as mentioned above), his wording was characteristic of the new thinking of the generative era:

Many [Soviet linguists] have supposed that it was necessary to set up another part of speech, the “category of state”… The traditional Soviet account is not very satisfactory, since one either has to accept a new part of speech or some loose ends in one’s taxonomy… It will be assumed in this paper that verbs and adjectives are surface-structure categories which derive from a single deep-structure category which will be called Predicators.

The positing of a new category was not considered an insight into the peculiar nature of Russian, but as a defect of the description that was to be remedied. Characteristically, instead of justifying his categories, Miller starts by making assumptions and then builds an analysis on these assumptions.

While the first half of the 20th century saw a lot of new terms coined by linguists, the generative era adopted a very different approach: The use of new terms for categories was limited, and the main way in which conceptual innovations could take place was by giving new meanings to existing terms (e.g. government, Fillmorean or Chomskyan case, adjunct, complement).

An interesting recent paper on word-classes is Chung’s (2012) study of Chamorro, which concludes that Topping’s (1973) earlier (structuralist) account in terms of two main classes (class I and class II), which differ radically from the familiar European classes, could be reanlyzed in terms of the familiar categories, V, N, and A. Topping set up two large classes mainly on the basis of the expression of the pronominal subject: Class I, comprising transitive action words, and Class II, comprising intransitive action words, property words, and thing words. Chung notes that Class II is not quite internally homogeneous, and that transitive and intransitive action words behave alike syntactically in some ways. Thus, if one wants, one can say that Chamorro has nouns, verbs and adjectives like English, even though Topping’s Class I vs. Class II division is more salient if one leaves English aside (as he did). Chung inserts her proposal for analyzing Chamorro in the context of the general discussion of word-class universality, but she does not seem to realize that the idea of the universality of V, N and A is derived from the tradition of English grammar since the 18th century (before that, nouns and adjectives were actually not clearly distinguished, following the Latin tradition of Priscian and Donatus). It is true that in many languages, one can draw similar distinctions if one wants, but one could probably also find such evidence for a Class I vs. Class II distinction and argue that it is the Chamorro distinction that is universal (cf. Haspelmath 2012b).

The word as a universal category

The word was never discussed seriously by Chomsky, and in fact the mainstream generative literature both of the 1960s and of the 21st century tends to assume that there is no strict division between words and phrases, and between morphology and syntax. However, between the 1970s and the 1990s, a very widespread view was that the word (or the lexicon) represents an important level of analysis, and that lexical and syntactic rules are quite distinct (this is called lexicalism; e.g. Anderson 1992; Bresnan & Mchombo 1995). Curiously, lexicalists never say what they mean by a word and how they identify it. In particular, they do not say how they might distinguish clitic words from affixes, and phrases from compound words, in a way that works across languages.

In this regard, lexicalism reminds one of the pre-structuralist time, when the word concept was simply taken for granted and not seen as a problem. It is striking that textbooks of the structuralist era usually include a lengthy discussion of how to delimit words (e.g. Hockett 1958; Gleason 1961), while more recent textbooks say very little about identifying words (the last general linguistics textbook with a long discussion was Langacker (1972), apparently still under the influence of structuralist particularism). But what happens if one simply pretends that words are universal, without saying how they can be identified universally? As in the case of nouns, verbs and adjectives, in practice this means that linguists go back to the traditional views of our school books. And traditionally, the word is regarded as a string of letters between spaces. As I showed in Haspelmath (2011), there is no other word identification procedure that linguists were able to rely on.

Of course, if one assumes that the categories of languages are given in advance (i.e. that we are born with them), there is no need to define them, anymore than we need to define other natural kinds such as potassium, dandelion or kidney (cf. Haspelmath 2014 on the distinction between defining and diagnosing categories). However, one needs to be aware that in practice that may mean that the consensus converges on the categories that we have been familiar with since our school days.

Reconciling particularism and universalism in the 21st century

While Greenbergian typology of the 1960s and 1970s seemed to be similarly universalist in outlook to the generative paradigm (and was thus embraced by many Chomskyans in the 1980s and 1990s, e.g. Baker 2001), more recently many linguists coming from the Greenbergian tradition have made clear statements advocating a return to particularist views when it comes to describing individual languages (Dryer 1997, Croft 2001, Lazard 2006, Haspelmath 2007). Interestingly, some of them are probably continuing structuralist views that they always held, influenced by pre-generative structuralists. (Thus, Dryer was a student of H. Allan Gleason’s at the University of Toronto in the 1970s, and Lazard was a student of Émile Benveniste’s in Paris in the 1940s and 1950s.)

But have these people been arguing for a return to the older practice of concentrating exclusively on language-specific system, to the exclusion of cross-linguistic generalizations? Not at all: The particularism concerns language-specific analyses, but this is not incompatible with (even large-scale) cross-linguistic comparison. One just needs to realize that grammatical comparison does not start out from language-specific categories, but from comparative concepts that are quite distinct from descriptive categories (Haspelmath 2010). Grammatical comparison is thus not different from lexical comparison (e.g. Swadesh lists) – everyone recognizes that word meanings in different languages cut up the conceptual space in different ways, but we still compare words across languages.

Thus, when comparing languages with respect to nominal, verbal and adjectival behaviour, we just need to look at thing words, action words and property words. These can be readily identified across languages, and interesting comparisons can be made (Haspelmath 2012a). In effect, a Greenbergian word-order universal concerning adjective-noun order refers to the order of property words and thing words. Likewise, when comparing languages with respect to words vs. affixes, one needs to work with comparative concepts that work across languages, such as grammatical morph vs. lexical morph (Haspelmath 2014), mobile vs. fixed grammatical morph, etc.

Thus, there is no need to abandon the structuralist concern for language-internal justification of categories, or large-scale cross-linguistic comparison. Once one recognizes that these are different enterprises, both can continue as originally envisaged.


Anderson, Stephen. 1992. A-morphous morphology. Cambridge: Cambridge University Press.

Baker, Mark C. 2001. The atoms of language. New York: Basic Books.

Baker, Mark C. 2003. Lexical categories: verbs, nouns, and adjectives. Cambridge: Cambridge University Press.

Bazell, C. E. 1958. Linguistic typology: An inaugural lecture delivered on 26 February 1958. School of Oriental and African Studies, University of London.

Bloomfield, Leonard. 1933. Language. New York: H. Holt and Company.

Boas, Franz. 1911. Introduction. In Franz Boas (ed.), Handbook of American Indian Languages, 1–83. Washington, DC: Bureau of American Ethnology.

Bresnan, Joan & Mchombo, Sam M. 1995. The lexical integrity principle: evidence from Bantu. Natural Language and Linguistic Theory 13: 181-254.

Chomsky, Noam A. 1957. Syntactic structures. ’s-Gravenhage: Mouton.

Chomsky, Noam A. 1970. Remarks on nominalization. In R.A. Jacobs & Peter S. Rosenbaum (eds.), Readings in English transformational grammar, 184–221. Waltham, MA: Ginn.

Chung, Sandra. 2012. Are lexical categories universal? The view from Chamorro. Theoretical Linguistics 38(1-2). 1–56. doi:10.1515/tl-2012-0001 (17 September, 2012).

Chung, Sandra. 2012. Are lexical categories universal? The view from Chamorro. Theoretical Linguistics 38(1-2). 1–56. (doi:10.1515/tl-2012-0001)

Croft, William. 2001. Radical construction grammar: Syntactic theory in typological perspective. Oxford: Oxford University Press.

Drach, Erich. 1937. Grundgedanken der deutschen Satzlehre. Frankfurt am Main: Diesterweg.

Dryer, Matthew S. 1997. Are grammatical relations universal? In Joan L. Bybee, John Haiman & Sandra A. Thompson (eds.), Essays on language function and language type: Dedicated to T. Givón, 115–143. Amsterdam: Benjamins.

Gleason, Henry Allan. 1961. An introduction to descriptive linguistics. New York: Holt, Rinehart and Winston.

Greenberg, Joseph H. 1974. Language typology: A historical and analytic overview. The Hague: Mouton.

Haspelmath, Martin. 2007. Pre-established categories don’t exist: consequences for language description and typology. Linguistic Typology 11. 119 – 132.

Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in crosslinguistic studies. Language 86(3). 663-687.

Haspelmath, Martin. 2011. The indeterminacy of word segmentation and the nature of morphology and syntax. Folia Linguistica 45(1). 31–80.

Haspelmath, Martin. 2012a. How to compare major word-classes across the world’s languages. In Thomas Graf, Denis Paperno, Anna Szabolcsi & Jos Tellings (eds.), Theories of everything: in honor of Edward Keenan, 109–130. (UCLA Working Papers in Linguistics 17). Los Angeles: UCLA.

Haspelmath, Martin. 2012b. Escaping ethnocentrism in the study of word-class universals. Theoretical Linguistics 38(1-2). doi:10.1515/tl-2012-0004.

Haspelmath, Martin. 2014. Defining vs. diagnosing linguistics categories: A case study of clitic phenomena. In Joanna Błaszczak (ed.), Categories. Berlin: De Gruyter Mouton.

Hockett, Charles F. 1944. Review of Linguistic interludes and Moprhology: The descriptive analysis of words (1944 edition) both by Eugene A. Nida. Language 20.252-255.

Hockett, Charles F. 1958. A course in modern linguistics. New York: MacMillan.

Langacker, Ronald W. 1972. Fundamentals of linguistic analysis. New York: Harcourt Brace Jovanovich.

Lazard, Gilbert. 2006. La quête des invariants interlangues: La linguistique est-elle une science?. Paris: Champion.

Milewski, Tadeusz. 1951. The conception of the word in languages of North American natives. Lingua Posnaniensis 3: 248–268.

Miller, Jim. 1973. A generative account of the “category of state” in Russian. In Ferenc Kiefer & Nicolas Ruwet (eds.), Generative grammar in Europe, 333–359. (Foundations of Language 13). Dordrecht: Reidel.

Mugdan, Joachim. 1986. Was ist eigentlich ein Morphem? Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 39: 29-43.

Müller, Stefan. 2010. Grammatiktheorie. Tübingen: Stauffenburg.

Robins, R. H. 1997. A short history of linguistics. London: Longman.

Ščerba, Lev V. 1945. Očerednye problemy jazykovedenija. Izvestija Akademija Nauk SSSR 1945.4(5): 173-186.

Schwegler, Armin. 1990. Analyticity and syntheticity: a diachronic perspective with special reference to Romance languages. Berlin: Mouton de Gruyter.

Togeby, Knud. 1949 [1978]. Qu’est-ce qu’un mot? In Michael Herslund, ed. Knud Togeby: Choix d’articles 1943–1974. Copenhagen: Akademiske Forlag, 51–65.

Topping, Donald. 1973. Chamorro reference grammar. Honolulu: University Press of Hawaii.

How to cite this post

Haspelmath, Martin. 2014. (Non-)universality of word-classes and words: The mid-20th century shift. History and Philosophy of the Language Sciences. https://hiphilangsci.net/2014/10/08/non-universality-of-word-classes-and-words-the-mid-20th-century-shift

Tagged with: , , , ,
Posted in 20th century, History, Linguistics, Typology, Uncategorized
4 comments on “(Non-)universality of word-classes and words: The mid-20th century shift
  1. John Joseph says:

    Thank you for this interesting post, and I hope you will continue looking in detail at just what changes happened when and where. It may be no coincidence if the shift you describe occurred at the same time as the one discussed in Joseph & Newmeyer, ” ‘All Languages Are Equally Complex’: The Rise and Fall of a Consensus”, Historiographia Linguistica 39 (2012), 341-68.

    On a minor point, while it’s true that “nouns and adjectives were actually not clearly distinguished” in linguistic analysis before the 18th c., there was a tradition of characterizing the difference between them in terms of abstract and concrete, with (surprisingly from today’s perspective) nouns aligning with the abstract and adjectives with the concrete. Thus, for Locke, simple ideas “have all abstract, as well as concrete names: the one whereof is (to speak the language of grammarians) a substantive, the other an adjective; as whiteness, white; sweetness, sweet”. But you’re right that this was never “clear”.

  2. […] languages are unique, and that each language has its own categories. But later in the 20th century, the view came to prevail that all languages basically make use of the same categories. The particularism of Franz Boas and European structuralists such as Hjelmslev and Martinet is well […]

  3. […] R: La primera respuesta es “¡ahora y siempre resistencia!”. Mark Aronoff ha demostrado que existe una dimensión “morfómica”. Creo también que el nivel de análisis de “palabra” es importante y legítimo. Para mí el problema de la soberanía de la palabra es una perspectiva de retaguardia, debido a la supuesta hegemonía generativista. Me parece más interesante el debate propuesto recientemente por Martin Haspelmath sobre la (no)-universalidad de las clases de palabras y de la palabra como categoría universal (cfr. https://hiphilangsci.net/2014/10/08/non-universality-of-word-classes-and-words-the-mid-20th-century-s…). […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: