How typology has solved its “comparability problem”: Some comments on Evans (2020)

Language comparison was long restricted to the question of phylogenetic inheritance (e.g. Bopp 1816; Schleicher 1860), but since authors such as Humboldt (1822) and von der Gabelentz (1891), linguists have also been interested in an ahistorical kind of comparison that is now often called “linguistic typology” – and which has become increasingly prestigious since Greenberg (1963), Chomsky (1981), and the 1997 foundation of the journal Linguistic Typology (LT).

Linguistics is thus similar to biological comparison in that it has both homologies (= similarities due to inheritance) and analogies (= similarities due to common function, or some other cause), something that our discipline has actually known for longer than the biologists (it was only with Darwin’s new theory that biologists accepted this distinction in the later 19th century; earlier there was a lot of speculation about “ideal forms”, as in the writings of morphists like Goethe and Geoffroy St. Hilaire). Bromham (2020) presents a beautiful account of the well-known distinction in biology between analogous and homologous similarities.

In order to investigate “analogies” systematically across languages, typologists must of course investigate the features of diverse languages on the basis of clear and objective definitions, as is the case in anthropology, geography, economics and other fields that study human cultural patterns in a worldwide perspective (musicology, religion studies, cross-cultural psychology, etc.). This presents no special problem, because all comparisons require clearly defined comparative concepts. For example, when psychologists compare the behaviour of different experimental subjects, they must make sure to apply the same coding procedures in all cases.

So why should there be a “comparability problem” in linguistics, which according to Evans (2020: 1) even “lies at the heart of linguistic typology”? In his introduction to a special section of the journal LT, Evans goes so far as to say that “problems of comparability will never go away” (2020: 8). But why should this be? What is the nature of the alleged problem?

The goal of this blogpost is to point out that if there ever was a comparability problem, typology has solved the problem. We can move on to other methodological challenges and need not be detained by this one further. But why did the impression arise that there is a problem (cf. also Croft 1995: 88; Stassen 2011)?

Problems of comparison may arise when we think that our stereotypes are sufficient for comparison of the whole domain, but once a critical scientific perspective is adopted, they usually disappear. For example, when geographers compare different regions of the world with respect to urbanization, they need to define the terms “urban” and “rural”. We have stereotypes about towns and villages, but a geographer will of course soon realize that some human settlement patterns do not fall clearly under either of these categories. So before embarking on such a study, we need to define the comparative concepts clearly, going beyond the stereotypes. This is quite trivial, and it also happens in other disciplines, of course. As Evans (2020: 3) notes, biologists must clearly define “wing” or “leaf” (and cannot rely on immediate understanding of their precise boundaries) because some trees have structures that do not clearly fall under the notion of a stereotypical “leaf” (e.g. phyllodes); likewise, anthropologists must clearly define “money”, because while we have a good sense of stereotypical money, the boundaries are not immediately clear (should gold bullion be regarded as money?)

But there is no scientific problem here, because scientists know that they cannot study the world systematically on the basis of everyday stereotypes. So it is not true that “the definition of money is a tricky problem” (Evans 2020: 3). Defining is a methodological prerequisite to scientific comparison, but it is not a “problem”. Acquiring research funds, getting good data, developing the right theories – all these are difficult challenges, but defining comparative concepts is easy and unproblematic. (Of course, not all comparisons will yield equally enlightening results – but this falls under the difficulty of developing the right theories, and is an entirely different matter.)

Now it is true that in linguistics, traditional stereotypes have had particularly long lives. Examples of such long-standing stereotypes are the idea that all sentences of all languages can be broken up into “words”, or that all languages have “syllables”, “subjects”, “adjectives”, and so on. Why have so many linguists persisted in holding on to these stereotypes?

It seems that the main reason for this is that most linguists are not comparative linguists. But even when they study a single language, many think that they can contribute to the general study of Human Language. They make highly general claims about phenomena such as “object case” even though they have compared only very few languages (e.g. Italian and English, Burzio 2000). This approach is not promising, just as we cannot make general statements about mammals on the basis of only two or three species (say, wombats, koalas, and possums). By studying a single language, one may well get deep insights into its particular intricacies, but one may also be tempted to prematurely generalize, and thus to reinforce old stereotypes such as “word” or “syllable”. However, typologists have long been aware of this source of bias, and the whole point of their enterprise is to avoid it – we want to compare languages scientifically, not via stereotypes that come from the big European languages.

So why does Evans (2020) think that there are “analytic dead-ends that [typology] risks becoming trapped in” (p. 5), or that we should “develop new tools for tackling this fundamental problem of linguistic typology” (p. 8)? I have no good explanation, because in earlier work, Evans himself has clearly expressed the solution: In addition to the language-particular categories of each language, typologists need an extra set of comparative concepts, like the other sciences:

“Language-specificity of categories raises problems for [comparison and explanation], as the typologists have become increasingly aware (Dryer 1997; Haspelmath 2007; 2010). It doesn’t follow that comparison is impossible, only that it has to be undertaken in an auxiliary language designed to generalize over language-specific categories.” (Levinson & Evans 2010: 2737).

(I also discussed Nick Evans’s published views in an earlier blogpost.)

Bromham’s (2020: 11-12) formulation is also very clear: As part of her discussion of biological comparison, she notes that typological concepts in linguistics are similar to analogous concepts like “wing” or “leaf” in biology, in that they are defined by their functions, and not structurally:

“A morpheme such as a dative might be doing the same “work” in two different languages, but it does not necessarily imply that they are derived from the same ancestral source (Haspelmath 2010). It is a functional category, not a marker of descent. We can refer to the wing of a pterodactyl, the wing of a bat, or the wing of a dragonfly, using “wing” as a functional category to describe an appendage used for active flight, not as a class of structures that are similar by descent.”

Curiously, the really deep problem of linguistics is not even mentioned by Evans: The fact that generative linguists do not worry about comparability but simply assume that all languages make use of the same innate building blocks (the “alphabet of human language”, the substantive universals of UG). The generative approach has of course been very influential, and especially from the 1980s until about 2010 (Chomsky 1981; Roberts 1996; Baker 2010), many typologists worked with assumptions from the Chomskyan tradition. Croft (2021) calls this the “skeleton model” (= “the assumption that the theoretical entities in Universal Grammar are of the same type as the entities in particular language grammars”). This view does indeed present a problem for comparison, because generative grammarians think that different criteria can be used for different languages, and “depth of analysis” is a prerequisite of comparison. On this view, comparison of grammars is like comparison of chemical compounds: We first have to break them down into their constituent elements before we can compare them properly. The hypothesized “bones” of the skeleton model are natural kinds (Aronoff 2016), like the elements of chemistry or the particles of physics.

Croft notes that most of modern typology does not adopt the skeleton model, and it plays no role in Evans’s considerations, but strangely, he still sees “problems which have appeared intractable, or even to undermine the whole typological enterprise” (p. 3). Maybe Evans means the slight nuisance that we often use the same terms in different senses for different purposes, e.g. “genitive” both for a general case type (= as a comparative concept) and for a specific category of a particular language (e.g. the Genitive case of Russian, whose functions go beyond the general genitive functions). But at least since Comrie (1976) and Bybee (1985), typologists have been acutely aware of these differences and have often signaled them by using upper case for language-particular categories such as the Russian Genitive.

In fact, the double duty of some of our terms is more than a nuisance: It often makes good sense to use a general terms also for language-particular categories, simply for mnemonic purposes (a term like “Case II” for the genitive would not be transparent; see Haspelmath 2021: §3). Evans is right that typologists (who are often also language describers) “move back and forth between descriptions of individual languages and synthetic cross-linguistic surveys” (p. 6), but as they are aware of the contrast between comparative concepts and language-particular categories, this presents no problem.

In two passages, Evans seems to hint that he feels that comparative concepts should not be “arbitrary” (to use Lazard’s 2005 drastic formulation). He says that typologists should “work out the best conceptual cuts” (p. 1), and that “good categories should ‘earn their status’, [e.g.] by exhibiting clustered properties” (p. 7). So he seems to feel that gram-types of the Bybee-Dahl type (Bybee & Dahl 1989; Dahl 2000) are good comparative concepts (in this connection, Evans refers to Spike 2020, whose paper I discussed here and here). However, observed clusters are empirical research results, whereas comparative concepts are prerequisites to comparison. In order to find clusters, we need to have uniformly defined comparative concepts (e.g. the questionnaire contexts of Dahl’s 1985 study of tense and aspect; see also Dahl’s comment on an earlier post).

It may be that broader comparative concepts such as “genitive”, “instrumental”, “sex-based gender”, “perfective aspect”, “adjective”, “relative clause”, or “antipassive” are more useful when they stand for cross-linguistic clusters. Thus, a WALS map such as Corbett’s (2005) map on sex-based grammatical gender would be more informative if the concept “sex-based gender” were not completely arbitrary. However, to the extent that the concept is easy to understand and already well-known, it is certainly worth having such a cross-linguistic study, even if we do not know whether sex-based gender forms some kind of “cluster”, or can “be used as a basis for successful inferences”. Empirical results are always preliminary (our samples may be too small, or not representative), and “success” is rarely certain – but methdological rigour requires objectively defined comparative concepts. Thus, it would not make sense to require that comparative concepts must be based on successfully demonstrated clusters. It cannot hurt if they are (as with Dahl’s “perfective” and “perfect”), but the range of possible concepts for comparison is very broad (even Bible verses may serve as comparative concepts, as in parallel-text studies).

Typologists have started to think more about sign languages over the last two decades, and Evans mentions Zeshan & Palfreyman’s (2020) article, but again, there is no problem of comparability. There may be a “well-known controversy over whether sign languages have true personal pronouns” (as Evans puts it), but such “controversies” are merely discussions of terminology (unless one adopts the UG-based skeleton model). “Personal pronoun” is a general term, and depending on how one uses it (=which comparative concept one takes it to stand for), it includes or does not include the “points to participants” in sign languages.

Before ending, I should be honest and admit that until about 2005, I was myself a bit confused over the distinction between comparative concepts and cross-linguistic categories. In 1995, I published a paper on “the converb as a cross-linguistically valid category”, in which I did not make it clear enough that I meant “converb as a comparative concept”. And in fact, I probably thought that “converb” is somehow given in advance (“pre-established), and that each language that has a converb has a converb in the same sense (see also this earlier blogpost). It was only with Dryer’s (1997) paper and Croft’s (2001) book that it became increasingly clear to me that each language-particular category is defined in a language-particular way, not as matching a pre-existing general category (see also Levinson & Evans 2010, from which I quote above). So yes, a comparability problem used to exist for me (and probably for others), apparently as a result of too much influence from the generative skeleton model. However, in 2020 the situation is very different: The contrast between descriptive categories and comparative concepts is very well known and can be said to be generally accepted. Some of Evans’s comments in his 2020 paper seem to reflect the pre-2010 situation in typology. Having overcome these ealier confusions, the field can now approach its other challenges.


Baker, Mark C. 2010. Formal generative typology. In Heine, Bernd & Narrog, Heiko (eds.), The Oxford handbook of linguistic analysis, 285–312. Oxford: Oxford University Press. (Accessed July 6, 2015.)

Bopp, Franz. 1816. Über das Conjugationssystem der Sanskritsprache. Frankfurt am Main: Andreäische Buchhandlung.

Bromham, Lindell. 2020. Comparability in evolutionary biology: The case of Darwin’s barnacles. Linguistic Typology. De Gruyter Mouton 1(ahead-of-print). (doi:10.1515/lingty-2020-2056)

Burzio, Luigi. 2000. Anatomy of a generalization. In Reuland, Eric J. (ed.), Arguments and case: Explaining Burzio’s Generalization. Amsterdam: Benjamins.

Bybee, Joan L. 1985. Morphology: A study of the relation between meaning and form. Amsterdam: Benjamins.

Bybee, Joan L. & Dahl, Östen. 1989. The creation of tense and aspect systems in the languages of the world. Studies in Language 13(1). 51–103.

Chomsky, Noam A. 1981. Lectures on government and binding. Dordrecht: Foris.

Comrie, Bernard. 1976. Aspect: An introduction to the study of verbal aspect and related problems. Cambridge: Cambridge University Press.

Corbett, Greville G. 2005. Sex-based and non-sex-based gender systems. In Haspelmath, Martin & Dryer, Matthew S. & Gil, David & Comrie, Bernard (eds.), The world atlas of language structures. Oxford: Oxford University Press. (

Croft, William. 1995. Modern syntactic typology. In Shibatani, Masayoshi & Bynon, Theodora (eds.), Approaches to language typology, 85–144. Oxford: Oxford University Press.

Croft, William. 2001. Radical Construction Grammar: Syntactic theory in typological perspective. Oxford: Oxford University Press.

Croft, William. 2016. Comparative concepts and language-specific categories: Theory and practice. Linguistic Typology 20(2). 377–393. (doi:10.1515/lingty-2016-0012)

Croft, William. 2021. Word classes in Radical Construction Grammar. to appear in a handbook.

Dahl, Östen. 1985. Tense and aspect systems. Oxford: Blackwell.

Dahl, Östen (ed.). 2000. Tense and aspect in the languages of Europe. Berlin: Mouton de Gruyter.

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

Evans, Nicholas. 2020. Introduction: Why the comparability problem is central in typology. Linguistic Typology (ahead of print). De Gruyter Mouton. (doi:10.1515/lingty-2020-2055)

Greenberg, Joseph H. 1963. Some universals of grammar with particular reference to the order of meaningful elements. In Greenberg, Joseph H. (ed.), Universals of language, 73–113. Cambridge, MA: MIT Press.

Haspelmath, Martin. 1995. The converb as a cross-linguistically valid category. In Haspelmath, Martin & König, Ekkehard (eds.), Converbs in cross-linguistic perspective, 1–56. Berlin: Mouton de Gruyter. (

Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in crosslinguistic studies. Language 86(3). 663–687. (doi:10.1353/lan.2010.0021)

Haspelmath, Martin. 2021. The structural uniqueness of languages and the value of comparison for description. Asian Languages and Linguistics (to appear).

Humboldt, Wilhelm von. 1822. Über das Entstehen der grammatischen Formen und ihren Einfluss auf die Ideenentwicklung. Abhandlungen der Akademie der Wissenschaften zu Berlin 31–63.

Lazard, Gilbert. 2005. What are we typologists doing? In Frajzyngier, Zygmunt & Hodges, Adam & Rood, David S. (eds.), Linguistic diversity and language theories, 1–23. Amsterdam: Benjamins.

Levinson, Stephen C. & Evans, Nicholas. 2010. Time for a sea-change in linguistics: Response to comments on ‘The myth of language universals.’ Lingua 120(12). 2733–2758.

Roberts, Ian. 1996. Comparative syntax. London: Edward Arnold.

Schleicher, August. 1860. Compendium der vergleichenden Grammatik der indogermanischen Sprachen. Weimar: Böhlau.

Spike, Matthew. 2020. Fifty shades of grue: Indeterminate categories and induction in and out of the language sciences. Linguistic Typology (ahead of print). De Gruyter Mouton. (doi:10.1515/lingty-2020-2061) (Accessed September 8, 2020.)

Stassen, Leon. 2011. The problem of cross-linguistic identification. In Song, Jae Jung (ed.), The Oxford handbook of language typology, 90–99. Oxford: Oxford University Press.

von der Gabelentz, Georg. 1891. Die Sprachwissenschaft, ihre Aufgaben, Methoden und bisherigen Ergebnisse. Leipzig: C. H. Tauchnitz. (

Zeshan, Ulrike & Palfreyman, Nick. 2020. Comparability of signed and spoken languages: Absolute and relative modality effects in cross-modal typology. Linguistic Typology (ahead of print). De Gruyter Mouton 1. (doi:10.1515/lingty-2020-2059)


2 thoughts on “How typology has solved its “comparability problem”: Some comments on Evans (2020)

  1. Martin, I wonder if you could explain what you mean when you say that a Bible verse can serve as a comparative concept in a parallel corpus. I guess you want to say that it gives you a basis for a comparison in that you have a number of short texts which are translational equivalents of each other. But in my view, a Bible verse in a parallel corpus is a set of data points, which cannot be treated as any kind of comparative concept without making a lot of assumptions and without breaking it down into simpler parameters.

    • Yes, to be really useful in practice, a Bible verse usually needs to be broken down (syntagmatically, as it were) into further comparative concepts, and matching these across languages is often a challenge. But this is the same with (Nijmegen-style) nonverbal stimuli, and actually with all kinds of other comparative concepts. For example, “dative” needs to be broken down into “recipient” and “addressee” (paradigmatically, as it were) for really detailed study. There’s no principled upper limit on granularity, and no lower limit on coarse-grainedness – it all depends on one’s theoretical questions. For some purposes, very coarse-grained comparisons are very useful (people like Fortescue would argue this for the term “polysynthesis”, for example).

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.