At an inspiring recent conference at the University of Mainz, quite a few talks discussed the notion of serial verb construction (SVC). I had written about SVCs in Haspelmath (2016), so I followed and took part in the discussion with interest. In this post, I share some of the thoughts that came up, in part presented in my talk (the handout is available here).
Let us start with the relation between verb compounding and SVCs. According to Alexandre François, nuclear-level SVCs in Oceanic languages, such as Mwotlap -yim matmat in (11) (from François 2006: 228), are not verb-verb compounds.
François placed different types of SVCs in Oceanic languages that he had studied on a scale going from biclausal combinations to verb-verb compounds, arguing that they occupy the middle ground – not really biclausal, and not compounds either. He generalized this to all “verb-based complex predicates”, a broad group that also comprises auxiliary verb constructions and “light verb constructions” of various types (this phenomenon is the topic of an ongoing ANR-DFG project led by Walter Bisang, Alexandre François, Andrej Malchukov and Martine Vanhove).
But why should serial verb constructions not include verb compounds? According to my (2016) definition, verb-verb compounds are not excluded:
Serial verb construction (Haspelmath 2016: 296)
“A serial verb construction is a monoclausal construction consisting of multiple independent verbs with no element linking them and with no predicate–argument relation between the verbs.”
Compounds are generally regarded as complex words rather than phrases, but in this definition, nothing is said about the word vs. phrase distinction of the construction, for good reasons: There is little agreement among linguists on how to delimit words from phrases in general (Haspelmath 2011), and the distinction does not seem to be necessary. In the above definition, when I say “verb”, I actually mean “verb root” (maybe I should have made this clearer). A root can be part of a compound word (containing two or more roots), or it can occur as the only root of a word.
In Alexandra Aikhenvald’s recent book (2018), she offers a definition on the first page that does not exclude compounds either:
“In many languages of the world, a sequence of several verbs act together as one unit. They form one predicate, and contain no overt marker of coordination, subordination, or syntactic dependency of any other sort. Such series of verbs are known as serial verb constructions, or serial verbs for short. Serial verbs describe what can be conceptualized as a single event. They are often pronounced as if they were one word.” (Aikhenvald 2018: 1)
And in fact, later in the book she specifically says that “a serial verb construction may form one or more grammatical words” (Aikhenvald 2018: §42, p. 93), citing examples of “single-word serial verbs” from Alamblak, Yimas, Igbo, and Olutec. A further example comes from Cantonese (cited from Matthews 2006: 82):
So why does A. François think that the above example from Mwotlap is not a compound? At the Mainz conference, he cited phonological reasons: The two verb roots are two different phonological words, and one would expect them to be a single phonological word if they constituted a compound. However, it is not clear to me that one would expect this, given that even nominal compounds are often said to constitute a single grammatical word but two phonological words (e.g. Hall 1999: 2-4). And conversely, “a serial verb construction can … consist of one phonological word which is made up of several grammatical words”, according to Aikhenvald (2018: 94).
But while I agree with Aikhenvald on the issue of single-word SVCs, I see several problems with her definition of SVCs: (i) Saying that they “act together as one unit” is extremely vague and does not really single out comparable phenomena across languages; (ii) Aikhenvald includes the notion of “predicate” in her definition without saying what she means by it; (iii) the definition does not exclude control constructions (such as “want to do”, or “ask to do”), which are not normally regarded as SVCs; and (iv) we do not know what “a single event” is. At the Mainz conference, Robert Van Valin presented the macro-event property (MEP) as identifying constructions that denote a single event (see Bohnemeyer & Van Valin 2017). But MEP constructions may consist of several clauses (as noted by Bohnemeyer et al. 2007), while everyone agrees that SVCs are monoclausal. It therefore seems to me that there is no need to include the MEP (or the vaguer “single event”) criterion in the definition.
But in turn, Aikhenvald (2018: 18) also sees problems with my (2016) definition, so let’s look at what she has to say (emphasis added):
“[Haspelmath’s] definition is fairly problematic. … The obscure formulation ‘predicate–argument relation between the verbs’ arbitrarily excludes serialization of complement-clause taking predicates and causative serial verb constructions (widespread types of serial verbs in many languages…). Sharing tense, aspect, modality, and mood [TAM] are deemed to be ‘unnecessary’ criteria for serial verb constructions.”
Aikhenvald is right that my 2016 definition is (partly) “arbitrary”, like any other proposal for a definition. But it is well-motivated, because complement and causative relations are not normally treated as SVCs in the literature (as I noted above). And I do not want to have a definition of a well-known grammatical term that does not correspond to the actual usage of the term. And yes, one could add TAM sharing as a criterion, but such an additional criterion would exclude only very few further cases (Lovestrand 2021: §2.2.2 mentions Ewe and Paamese). Not including it in the definition makes it possible to state the sharing of tense and mood as empirically testable generalizations (Haspelmath 2016: §4).
Perhaps Aikhenvald’s unease with my definition also has to do with her view of what definitions should do: In her own characterization on the first page of the book, she adds a feature that is often – but not necessarily – present: “They are often pronounced as one word”. But if this feature is not necessary or sufficient, it has no place in a definition. And in fact, Aikhenvald does not say explicitly that this is a definition – so maybe she does not think of it as a way to delimit SVCs from other phenomena.
Why would one insist on a definition that not only gives a rough characterization, but also delimits a set of phenomena from others in a way that may not do justice to the continuities that we actually observe? After all, it is questionable that “serial verb construction” is a concept that cuts nature at its joints, and it seems very likely that the phenomena of the world’s languages actually form a continuum, rather than clustering neatly in the area that we call “SVC”. Lovestrand (2021: §5) discusses the fairly popular “prototype approach” which “allows linguists to make broad generalizations across languages … without having to rigorously test the properties of every putative SVC.”
But while such cluster concepts (along the lines of Bybee & Dahl’s (1989) gram-types) would be a very interesting outcome of cross-linguistic research, they cannot stand at its beginning. Dahl’s (1985) original comparative concepts were his questionnaire sentences, and these were sharply delineated rather than fuzzy prototypes. So if one wants to formulate and test cross-linguistic generalizations (such as the 10 universals discussed in Haspelmath 2016: §4), one needs clearly delimited concepts.
In their discussion of traditional concepts such as “infinitive”, “converb”, “auxiliary” and “SVC” (in their ANR-DFG project application text), Vanhove & Bisang (2021) note:
“If the diagnostics show that the distinction between certain types is in fact scalar, this would have repercussions for both typology and linguistic theory. Indeed, the typological relevance of many concepts in this domain remains currently unclear.”
Which concepts are the most “relevant” for typology will have to be the result of further research. Terms such as “serial verb construction”, “infinitive” and “converb” are traditional stereotypes, and we do not know whether they capture something real (quite possibly they do not). But if we treat them as fuzzy prototypes, we will never know. Instead, we need to do systematic cross-linguistic research, based on cross-linguistically comparable concepts, i.e. the phenomena must be identified across languages using the same criteria. The problems facing us are nicely described in Lovestrand’s (2021) closing paragraph:
“One point that the literature on SVCs makes clear is that traditional notions of complex sentences based predominantly on studies of European languages are inadequate from a more universal perspective, but this problem is not repaired by the misleading overgeneralization of lumping the diverse types of multiverb constructions into a handful of categories such as serial verb constructions, clause-chaining (Foley 2010), cosubordination (Bickel 2010), converb constructions (Haspelmath 1995), and consecutive constructions (Hyman 1971). There are many more degrees and layers of interclausal cohesion than have been explicitly accounted for. Further research into the myriad ways in which languages combine verbs is a significant opportunity for refining our understanding of the nature of clausehood…” (Lovestrand 2021: §6)
Which kinds of clusters will eventually emerge, and what kinds of concepts are the best for worldwide language comparison seems to be quite unclear at the moment. This is reason for casting the net wide (as urged by Lovestrand), but not for accepting vague or approximate meanings of technical terms.
Aikhenvald, Alexandra Y. 2018. Serial verbs. Oxford: Oxford University Press.
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.
Bohnemeyer, Jürgen & Enfield, Nicholas J. & Essegbey, James & Ibarretxe-Antuñano, Iraide & Kita, Sotaro & Lüpke, Friederike & Ameka, Felix K. 2007. Principles of event segmentation in language: The case of motion events. Language 495–532.
Bohnemeyer, Jürgen & Van Valin, Robert D. 2017. The macro-event property and the layered structure of the clause. Studies in Language 41(1). 142–197. (doi:10.1075/sl.41.1.05van)
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.
Dahl, Östen. 1985. Tense and aspect systems. Oxford: Blackwell.
Foley, William A. 2010. Clause linkage and nexus in Papuan languages. In Bril, Isabelle (ed.), Clause-hierarchy and clause-linking: the syntax and pragmatics interface, 27–50. Amsterdam: Benjamins.
François, Alexandre. 2006. Serial verb constructions in Mwotlap. In Aikhenvald, Alexandra & Dixon, R. M. W. (eds.), Serial verb constructions: A cross-linguistic typology, 223–238. Oxford: Oxford University Press.
Hall, Tracy A. 1999. The phonological word: A review. In Hall, Tracy A. & Kleinhenz, Ursula (eds.), Studies on the phonological word, 1–22. Amsterdam: Benjamins.
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. 2011. The indeterminacy of word segmentation and the nature of morphology and syntax. Folia Linguistica 45(1). 31–80. (doi:10.1515/flin-2017-1005)
Haspelmath, Martin. 2016. The serial verb construction: Comparative concept and cross-linguistic generalizations. Language and Linguistics 17(3). 291–319. (doi:http://doi.org/10.1177/2397002215626895)
Hyman, Larry M. 1971. Consecutivization in Feʔfeʔ. Journal of African Languages 10. 29–43.
Lovestrand, Joseph. 2021. Serial verb constructions. Annual Review of Linguistics 7(1). 109–130. (doi:10.1146/annurev-linguistics-031920-115317)
Matthews, Stephen. 2006. On serial verb constructions in Cantonese. In Aikhenvald, Alexandra & Dixon, R. M. W. (eds.), Serial verb constructions: A cross-linguistic typology, 69–87. Oxford: Oxford University Press.
Vanhove, Martine & Bisang, Walter. 2021. Complex predicates in languages: Emergence, typology, evolution (ComPLETE). Paris and Mainz: ANR-DFG Project application. (https://marama.huma-num.fr/ComPLETE/data/ANR-DFG_2021_ComPLETE_Project_Main.pdf)
I think I’ll back Aikhenvald here on the basis that I think that the verb combinations with no marking of subordination at least (won’t worry about coordination here) have a very interesting functional property, and so are worth a name, for which SVC would be closest to existing traditions. The fact that at deeper levels they can do a variety of things, but terminology such as causative, aspectual, etc SVC already does justice to that.
But is there a tradition of actually calling them serial verbs? For ‘I asked him to come’, Mandarin has something like “Wǒ qǐng tā lái” – this would fit the bill. But has anyone called this a serial verb construction? It seems to me that in practice, the SVC label has been used more restrictedly, and this is reflected in my definition. Of course one could argue that such constructions ar similar enough to b included, but this would involve a change in the traditional meaning, I think.