Where I went wrong (II): Converbs are not “non-finite verbal adverbs”

Another way in which I went wrong in the 1990s was by proposing a definition of converb as a “non-finite verbal adverb” (see this earlier blogpost about my error about iconicity). At the time (Haspelmath 1995), I was quite naïve about the traditional notions of “finiteness” and “adverb”, and I now think that defining “converb” in this way was not a good idea.

Of course, the 1995 paper was not wrong in drawing wrong conclusions or making wrong claims – the mistake was at the conceptual level. The paper (along with the volume in which it appeared) was extremely successful in that it put the term “converb” on the map – while this term was known almost only to Altaicists (specialists of Mongolic and Turkic) at the time, it has become well-known, and even entire conferences are being devoted to converbs (e.g. this 2022 conference in Verona). Our reason for promoting this term (following Nedyalkov & Nedyalkov 1987) was that we felt that Romance “gerundios”, Slavic “adverbial participles”, and the various forms called “deepričastie” in Russian were so similar that they should be treated together.

In the first period of modern typology (1970s-1990s), one common thread in the literature is that some phenomena that had been regarded as unique in different languages were found to be so similar that they should be seen as “typologically the same phenomenon”. For example, special “transitive nominative” cases (Pott 1873) had different names in different languages (relative in Eskimo, narrative in Georgian, instrumental in Basque, and so on), but by the 1960s, the term ergative had become well-established (Žirmunskij (ed.) 1967; Comrie 1973), and this led to a flourishing of research on ergativity (Dixon 1979; Plank (ed.) 1979; and much subsequent work). Clearly, it was productive to treat these phenomena together – as if they were all “the same phenomenon”.

Another example is the term ideophone, which was used originally for African languages, for a phenomenon quite similar to what had been called mimetics or expressives in Asian languages. As Dingemanse (2019) notes, the term ideophone is now considered as a general term:

” ‘Ideophone’ has come to be widely used and has shed its Bantu connotation. It is used in descriptions of languages around the world, from Awetí, Basque and Chintang to Xhosa, Yir-Yoront and Zuni. The alternative labels ‘expressives’ and ‘mimetics’ continue to be current in the prolific research traditions of South-East Asian and Japanese linguistics, but there seems to be broad agreement that these point to essentially the same phenomenon.” (Dingemanse 2019: 14-15)

Thus, having terms that pick out comparable phenomena in languages around the world is often seen as useful, and in this sense, the converbs volume (Haspelmath & König (ed) 1995) was progress.

However, it was wrong to talk about “the converb as a cross-linguistically valid category” (even in the title of my 1995 paper), because categories are never “cross-linguistically valid”. When we say that two categories of two different languages are “the same phenomenon”, we really mean that we want to subsume them under the same comparative concept. But our comparative concepts are methodological tools and do not (necessarily) have any reality beyond that. Thus, we need to define them in a clear way, and we cannot hope to find out more about them simply by doing more research. This is different with natural kinds, such as fish and mammals. Pullum (2015; 2017) emphasizes the similarities between adverbial subordinators such as English after (as in after we had lunch) and prepositions (such as after in after lunch), and it is surely possible to say that after is a preposition in both cases. But this is not a discovery akin to the finding that whales are part of the mammal lineage (which Linnaeus recognized in 1758). Prepositions and other classes of forms are cultural categories, and there is no unique way of classifying linguistic forms. And it seems extremely unlikely that categories such as “adposition” or “conjunction” might be innate elements of our grammar blueprint – so when we talk about them in general terms, we talk about comparative concepts, not natural kinds. (There is thus no unique “catalogue of categories”, analogous to the catalogue of animal species that Linnaeus drew up; its modern version is called Catalogue of Life.)

When I wrote about converbs back in 1995, I was simply naïve about all this – I took it for granted that “converb” had a reality outside of my comparative perspective, because I was struck by the similarities between languages that others seemed to have overlooked. However, I now realize that others might draw different comparisons and see different similarities – my 1995 article did not make any testable claims, so it may have been a subjective exercise. Many people shared my enthusiasm about studying converbs, and the book was widely noted – but they may have had the same subjective biases (e.g. an undue emphasis on Eurasian languages).

In fact, some problems with the “converb” notions were apparent early on: The definitions that were used by Vladimir Nedjalkov and by myself in our two overview chapters were not identical (as was noted by perceptive reviewers such as Bickel 1998). But it did not seem to matter much to me, because I (wrongly) thought that converbs somehow exist independently, and that finding a good definition was a research question. I preferred my own definition because it included the concept of “nonfiniteness”, which I though should be included, because the most typical converbs do not show person marking. For example, Mongolian converbs (for which Ramstedt  coined the term converb in 1903) and other Altaic converbs are non-finite, and so are Romance gerunds and Slavic adverbial participles.

But what is “(non-)finiteness”? I was aware with the problems of this term (highlighted later by Cristofaro 2007), but again I thought that the notion existed independently and its precise nature was a matter of further research. I associated it primarily with the presence of person marking and/or tense marking, but I. Nedjalkov (1998: 421) defined it very differently (“Nonfiniteness is understood here as the impossibility of a verb form … being used as the only verb form in a simple nonelliptical sentence”). For this concept, I used the term dependent verb, which I also distinguished from subordinate.

The original definition of the term converb was formulated by Nedyalkov & Nedyalkov (1987: 75):

“a verb form which does not occur in the position (i) of the predicate of a simple sentence, (ii) of the attribute to a noun, (iii) of the predicate actant, (iv) of the subjective actant. In the above mentioned four positions the following prototypical verb forms occur: position (i) – a finite form, position (ii) – a participle, position (iii) – an infinitive, position (iv) – a gerund.”

This was thus a purely negative definition, and I felt that that was too broad, because it would include all kinds of dependent verb forms that have never been called “converbs”, such as dependent verb forms of Swahili (verbal prefix ki-) and Eskimo (verbal suffix -mmat):

Moreover, I felt that it would be better to have a positive definition: Just as participles are verb forms specialized for adnominal/adjectival use, and gerunds/infinitives (I called them “masdars”) are verb forms specialized for argumental/nominal use, I felt that converbs should be regarded as verb forms specialized for adverbial use, and I drew up the following table (Haspelmath 1995: 4):

But I did not have any positive characterization of “adverbial” as a “syntactic function”, and after reading Croft (1991; 2000), I later came to realize that there is far too much internal and cross-linguistic diversity, both functionally and formally, within the expressions that are often subsumed under “adverbial”. (Again, I had been aware that “adverb” was not a simple concept, also from Ramat & Ricca’s (1994) paper, but I seem to have thought that the problems could be addressed by further research.)

Some of these definitional problems were discussed quite insightfully by van der Auwera (1998), who discussed both the broader definition used by Nedyalkov & Nedyalkov and my narrower definition, as well as ingredient concepts such as “± dependent”, “± subordinate”, “± finite”. He came up with a summary table of different concepts, where “converb in a struct sense” is my (1995) definition, and “in a broad sense” is the Nedjalkovian definition:

Later papers discussing the definition of “converb” are Ylikoski (2003), Coupe (2006), and Rapold (2010), but the most widely cited paper on converbs is still my (1995) paper.

Where does this leave us? It is clear, of course, that broad terms such as “converb” will never be sufficient to express the variation across languages that we find, and multivariate approaches that use a rich set of cross-linguistic variables are needed (Bickel 2010; Rapold 2010). However, linguists will always use broad concepts as well (in addition to more fine-grained concepts), and these concepts should be defined clearly and uniformly, too.

Thus, I am now proposing a new definition of converb that is not too far from my 1995 definition (because that paper has been the most widely cited paper), but that does not make use of the unclear concepts “non-finite” and “adverbial”. The new definition goes as follows:       

converb: A converb is a form that consists of a verb stem and a special affix but no subject person indexing and that can occur as the predicate of a subordinate clause that is neither a relative clause nor a complement clause.

This definition does not include the Swahili and Greenlandic cases cited above that I wanted to exclude in 1995, so it is not extensionally different from the earlier definition. But it avoids the “adverbial” notion by adopting the “negative” definition from Nedyalkov & Nedyalkov (1987), where converbs are explicitly distinguished from participles and verbal nouns (“gerunds”, “infinitives”, “masdars”) which occur in relative clauses and complement clauses, respectively. The ”nonfinite” notion is replaced by specific reference to absence of subject person indexing.

The new definition is more complex than the one I used in 1995, and it is clearly unnatural – there is no real reason for the absence of subject person indexing (or even for the presence of a special affix). However, I have come to realized that unnatural definitions of traditional terms are necessary if we want to make progress in understanding language structures. For example, I proposed a very unnatural definition of “word” (Haspelmath 2023), which is primarily intended to make clear that our theoretical understanding of morphosyntax can hardly be based on a morphology-syntax division. (The terms ergative and ideophone that I mentioned at the beginning seem to be less less unnatural, but we should not have the illusion that these concepts are “discoveries” in the sense that it is a discovery that whales share a unique ancestor with the other mammals. Just like converb, the terms ergative and ideophone allow us particular perspectives on languages, which we hope will be productive.)

This is a rather sobering way to end this blogpost, but it should be clear that it is much easier to make apparent progress by introducing a new buzzword (“converb”) than to make real progress by creating new testable theories that have a good chance of being confirmed by subsequent research. It seems that comparative subordination research in linguistics is still in its infancy.


Bickel, Balthasar. 1998. Review article of Haspelmath & König, eds., Converbs, Berlin: Mouton de Gruyther 1995. Linguistic Typology 2. 381–97.

Bickel, Balthasar. 2010. Capturing particulars and universals in clause linkage: A multivariate analysis. In Bril, Isabelle (ed.), Clause-hierarchy and clause-linking: the syntax and pragmatics interface, 51–102. Amsterdam: Benjamins.

Comrie, Bernard. 1973. The ergative: Variations on a theme. Lingua 32(3). 239–253. (doi:10.1016/0024-3841(73)90044-2)

Coupe, Alexander R. 2006. Converbs. Encyclopedia of languages and linguistics (2nd edition), vol. 3, 145–152. Oxford: Elsevier.

Cristofaro, Sonia. 2007. Deconstructing categories: Finiteness in a functional-typological perspective. In Nikolaeva, Irina (ed.), Finiteness: Theoretical and empirical foundations, 91–114. Oxford: Oxford University Press.

Croft, William. 1991. Syntactic categories and grammatical relations: The cognitive organization of information. Chicago: University of Chicago Press.

Croft, William. 2000. Parts of speech as language universals and as language-particular categories. In Vogel, Petra M. & Comrie, Bernard (eds.), Approaches to the typology of word classes, 65–102. Berlin: Mouton de Gruyter.

Dingemanse, Mark. 2019. ‘Ideophone’ as a comparative concept. In Akita, Kimi & Pardeshi, Prashant (eds.), Ideophones, mimetics, and expressives (Iconicity in Language and Literature), 13–33. Amsterdam: Benjamins. (https://doi.org/10.1075/ill.16.02din)

Dixon, R.M.W. 1979. Ergativity. Language 55. 59–138.

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. (https://zenodo.org/record/227108)

Haspelmath, Martin. 2023. Defining the word. WORD 69(3). 283–297. (doi:10.1080/00437956.2023.2237272)

Nedjalkov, Igor V. 1998. Converbs in the languages of Europe. In van der Auwera, Johan (ed.), Adverbial constructions in the languages of Europe (Empirical Approaches to Language Typology: EUROTYP 20–3), 421–455. Berlin: Mouton de Gruyter.

Nedjalkov, Vladimir P. 1995. Some typological parameters of converbs. In Haspelmath, Martin & König, Ekkehard (eds.), Converbs in cross-linguistic perspective, 97–136. Berlin: Mouton de Gruyter. (https://zenodo.org/record/227108)

Nedyalkov, Vladimir P. & Nedyalkov, Igor V. 1987. On the typological characteristics of converbs. In Help, Toomas & Murumets, S. (eds.), Symposium on language universals, 75–79. Tallinn: Academy of Sciences of the Estonian SSR. (https://zenodo.org/records/11401365)

Plank, Frans (ed.). 1979. Ergativity: Towards a theory of grammatical relations. New York: Academic Press.

Pott. 1873. Unterschied eines transitiven und intransitiven nominativs. Beiträge zur vergleichenden Sprachforschung auf dem Gebiete der arischen, celtischen und slawischen Sprachen. JSTOR 7(1. H). 71–94.

Pullum, G. 2015. Prepositions as Conjunctions, Whales as Fish. The Chronicle of Higher Education. 8th Nov.

Pullum, Geoffrey K. 2017. Language Log » Ask Language Log: with + nonfinite clause? (https://languagelog.ldc.upenn.edu/nll/?p=35749)

Ramat, Paolo & Ricca, Davide. 1994. Prototypical adverbs: On the scalarity / radiality of the notion  of ADVERB. Rivista di Linguistica 6. 289–326.

Rapold, Christian. 2010. Defining converbs ten years on: A hitchhiker’s guide. In Völlmin, Sascha & Azeb Amha & Rapold, Christian & Zaugg-Coretti, Silvia (eds.), Converbs, medial verbs, clause chaining and related issues (Frankfurter Afrikanistische Blätter 19 (2007)), 7–30. Köln: Köppe.

van der Auwera, Johan. 1998. Defining converbs. In Kulikov, Leonid I. & Vater, Heinz (eds.), Typology of verbal categories, 273–282. Tübingen: Niemeyer.

Ylikoski, Jussi. 2003. Defining non-finites: Action nominals, converbs and infinitives. SKY Journal of linguistics 16(1). 185–237.

Žirmunskij, V.M. (ed.). 1967. Èrgativnaja konstrukcija predloženija v jazykax različnyx tipov: Issledovanija i materialy. Leningrad: Nauka.

Cite this blog post
Martin Haspelmath (2024, May 31). Where I went wrong (II): Converbs are not “non-finite verbal adverbs”. Diversity Linguistics Comment. Retrieved June 13, 2024, from https://doi.org/10.58079/11r1d

8 thoughts on “Where I went wrong (II): Converbs are not “non-finite verbal adverbs”

    • No, it doesn’t. A “deranked verb” in Cristofaro’s work is similar to “dependent verb form” in my 1995 paper, I think. But I don’t use “deranked”.

  1. Are ALL terms in your definition welldefined (if so, pls, indicate where)?
    What is a “special” affix?
    Why “can occur”, and not “occurs”?

    • They should all be well-defined – for example, see Haspelmath (2013) for person indexing. The reason the definition includes “can occur” rather than “occurs” is that “combined converbs” (in Nedjalkov’s terms) are not excluded.

  2. Though there are Romance varieties in which the so-called ‘gerundio’ has a subject affix. Old Neapolitan is one that has been quite widely discussed in the specialist literature and a similar phenomenon has also been reported for non-standard varieties of European Portuguese and Galician.

    • These would no longer count as converbs, according to the definition. And if I were to propose a general definition of “infinitive”, then surely Portuguese and Old Neapolitan “infinitives” would not be included.

      • Of course they don’t count according to the new definition but to me that suggests the new definition needs rethinking.

        • Our whole conceptual apparatus needs rethinking – but I don’t think that traditional terms should change their meaning when we make new discoveries. If we realize that our old concepts are insufficient for a richer reality, we need new concepts and new terms.

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.