Scales of referential prominence (animate > inanimate, definite > indefinite, locuphoric (1st/2nd) > aliophoric (3rd), topical > non-topical) are known to play a role in differential object marking generalizations. But what exactly is their role? Do they merely “capture descriptive generalizations”, or is there “explanatory power” in theories that invoke them (such as Aissen 2003; Haspelmath 2021)?
The cross-linguistic facts have been well-known since the 1980s (e.g. Bossong 1991; Lazard 2001) and need not be repeated here: Many languages have object flagging splits such that there is a marker for referentially prominent objects, but no marker for less prominent objects. For example, Ohrid Macedonian has Go vidov na čovekot [him I.saw ACC man.the] ‘I saw the man’, with an accusative preposition na that is only used with animate objects (Bužarovska 2020). The reverse does not occur, so we there is a set of striking implicational universals.
Ormazabal & Romero (2019) discuss Spanish differential object marking from a Chomskyan perspective, and before they get to their framework, their paper has an interesting discussion of “the functionalist approach” (they cite Silverstein 1976; Comrie 1981; Croft 1990; Aissen 2003), which they end up rejecting. We do not often see generative grammarians discussing functional alternatives to their framework, so this is a very welcome contribution to the debate which is worth replying to – also because I actually agree with part of their criticism (see two paragraphs down).
But I don’t agree with their characterization of the contrast as “scale-based theories” vs. “syntactic accounts” (§0), because there is no question that we are dealing with a syntactic phenomenon. I had a similar problem with Anagnostopoulou’s (2017) opposition between a “frequency-based approach” and a “grammar-based approach” (we discussed this earlier here).
Generative syntactician have a tendency to use the words “grammar” or “syntax” for “the apparatus of generative syntax that is assumed to be available to all languages”, but unless we assume that these are innate, we do not really have a general explanation. Of course, Spanish has syntactic rules that must be described, but it is an open question how the rules of Spanish relate to those of Macedonian, Persian, and so on. Grammatical/syntactic rules may simply be language-particular and accidental. Even many Chomskyans are not sure about innateness of the building blocks anymore. So I will leave aside Ormazabal & Romero’s approach (which may well be correct as a description of Spanish), and I focus on what is clearly non-accidental: the implicational universals.
However, I do agree with Ormazabal & Romero that ambiguity avoidance is not a good explanatory factor for explaining the DOM universals. The idea of this proposal is that those objects which are the most subject-like get special coding so as to allow better distinguishability of subjects and objects (this is an old idea, but it was prominently advocated by Aissen (2003: 437): “it is those direct objects which are most in need of being distinguished from subjects that get overtly case-marked”).
They give two good arguments against ambiguity avoidance:
(I) The diachronic development of DOM in Spanish begins with pronouns, and “this happens in spite of the fact that pronouns are the only elements … that have an independent form for accusative and nominative” (§1.1). So it begins precisely in those cases where there is no ambiguity!
(II) While ambiguity avoidance might seem to make the right predictions for direct-object marking (because definite and animate objects are “more subject-like”), it makes wrong predictions for recipient/goal arguments. As is well-known, one can say I sent a book to Mary/Paris (with the preposition to), and the preposition can sometimes be dispensed with – but only if the recipient is animate:
(1) I sent Mary a book.
(2) *I sent Paris a book.
Here it is precisely the non-subject-like inanimate goal (‘Paris’) that requires the indirect-object marking, which is the opposite of what one would expect if the driving force were ambiguity avoidance.
But I have claimed in a recently published paper (Haspelmath 2021) that there is a highly general functional-adaptive explanation underlying a large number of argument coding splits: Rather than striving to avoid ambiguity, speakers allocate their resources efficiently in a very general way: They use argument flags (case-markers and adpositions) preferentially in situations that are not expected. This can be summarized as follows:
Expected associations of role rank and referential prominence: Higher-ranking roles (A, R) are associated with referential prominence, and lower-raking roles (P, T) are associated with referential non-prominence.
So languages use differential marking in unexpected or unusual situations, even when there is no ambiguity. This explains, for example, the contrast between (1) and (2):
(3) I sent Mary a book.
(4) I sent the book to a professor.
The construction with the preposition to in (4) is preferred when the R (the recipient) is indefinite (i.e. referentially non-prominent), while the double object construction in (3) is preferred when the R is a definite (especially when it is a personal name or pronoun). Crucially, indirect objects are more like subjects: They tend to be animate and definite, unlike direct objects (P-arguments). So the use of to is fully in line with the general pattern (called the role-reference association universal in Haspelmath 2021).
The use of an extra accusative marker on Spanish personal pronouns (e.g. a mi, a tú ‘me, you’) is also in line with the general pattern. Ambiguity alone does not motivate it, but it is motivated by the tendency to use extra marking in unexpected combinations.
Ormazabal & Romero (2019: §1.2) also discuss a functionalist alternative that does not appeal to ambiguity avoidance, and which is roughly what I laid out above (they refer to my 2018 blogpost, where I cited an earlier version of my 2021 paper): “In this type of proposals, any argument that deviates with respect to [frequency-induced canonicity] may end up being marked.” This is indeed roughly what I say, though it needs to be added that the reverse marking pattern is predicted not to occur. Ormazabal & Romero say that it is not completely clear to them what one should count, and they observe that with a verb like Spanish ver ‘see’, only two thirds of the objects are inanimate.
However, a two-thirds majority is a clear majority, and it seems that this trend is found in all languages where it has been studied. In my (2021) paper, I give just a few examples, from Swedish, Movima and English, illustrating the predominance of animate/definite A-arguments and inanimate/indefinite P-arguments (cf. the table below).
Ormazabal & Romero make the interesting additional point that “most verbs are selective with respect to their complements, and in consequence, it is not clear to us that a global comparison is of any use”. They are right that verbs such as ‘learn’ and ‘practice’ only take inanimate objects, while verbs such as ‘welcome’ and ‘help’ mainly take animate objects – so why should we lump them together as “A-arguments” in making counts such as those just presented? I admit that there is no good answer to this other than the fact that most languages treat physical-effect verbs (‘break’, ‘kill’, etc.) alike in their coding, so that a generalized concept “P-argument” (or more traditionally, “direct object”) has proven very useful – and most languages also treat many other two-argument concepts in the same way as the physical-effect verbs (see Lazard 2002; Haspelmath 2011). There are some exceptions, and they go in the expected direction (e.g. languages like German and Russian require a dative object for ‘help’). So again, Ormazabal & Romero are making an interesting point that is not very much present in the literature and that should be discussed more (see also Haspelmath 2014 for discussion of the same issue). Indeed, we do not have a good explanation for the widespread unity of P-arguments, but if we take it as a given, then the efficiency-based explanation of the differential object marking universals holds up.
At the end of their §1, Ormazabal & Romero say that the scale-based approaches “cannot go further than capturing certain descriptive generalizations, lacking any explanatory power”. But we have already seen that the explanation works very well (given the P-argument premise), and it is hard to see how Ormazabal & Romero’s paper could provide a serious alternative, given that they only consider Spanish. The efficiency-based explanation makes use of highly general factors to explain highly general outcomes, whereas generative syntax papers often seem to highly general explanations (UG building blocks) for parochial phenomena.
And what do they man by “descriptive generalizations”? What other generalizations are there? It seems to me that “description” is best used in contrast to “comparison”, so that “descriptive generalizations” would be language-particular ones – but the efficiency-based explanation is intended to explain the implicational universals, not the descriptive generalizations in this sense. Perhaps by “descriptive generalizations” they mean “phenomenological generalizations”, i.e. generalizations over descriptions that do not make reference to innate building blocks. And maybe by “capturing” they mean “expressing”, and then they would be partly right: Indeed, implicational universals of the Greenbergian type express generalizations that can be made without assuming an innate UG. But the efficiency explanation goes beyond that, actually explaining the generalizations.
Aissen, Judith. 2003. Differential object marking: Iconicity vs. economy. Natural Language & Linguistic Theory 21(3). 435–483.
Anagnostopoulou, Elena. 2017. The Person Case Constraint. The Wiley Blackwell companion to syntax, 1–47. Wiley Blackwell. (doi:10.1002/9781118358733.wbsyncom101)
Bossong, Georg. 1991. Differential object marking in Romance and beyond. In Kibbee, Douglas & Wanner, Dieter (eds.), New analyses in Romance linguistics, 143–170. Amsterdam: Benjamins.
Bužarovska, Eleni. 2020. The contact hypothesis revised: DOM in the South Slavic periphery. Journal of Language Contact. Brill 13(1). 57–95. (doi:10.1163/19552629-bja10003)
Comrie, Bernard. 1981. Language universals and linguistic typology: Syntax and morphology. Oxford: Blackwell.
Croft, William. 1990. Typology and universals. Cambridge: Cambridge University Press.
Haspelmath, Martin. 2011. On S, A, P, T, and R as comparative concepts for alignment typology. Linguistic Typology 15(3). 535–567.
Haspelmath, Martin. 2014. On system pressure competing with economic motivation. In MacWhinney, Brian & Malchukov, Andrej L. & Moravcsik, Edith A. (eds.), Competing motivations in grammar and usage, 197–208. Oxford: Oxford University Press.
Haspelmath, Martin. 2019. Differential place marking and differential object marking. STUF – Language Typology and Universals 72(3). 313–334.
Haspelmath, Martin. 2021. Role-reference associations and the explanation of argument coding splits. Linguistics 59(1). 123–174. (doi:10.1515/ling-2020-0252)
Lazard, Gilbert. 2001. Le marquage différentiel de l’objet. In Haspelmath, Martin & König, Ekkehard & Oesterreicher, Wulf & Raible, Wolfgang (eds.), Language typology and language universals: An international handbook, vol. 2, 873–885. Berlin: Walter de Gruyter.
Lazard, Gilbert. 2002. Transitivity revisited as an example of a more strict approach in typological research. Folia Linguistica 36(3–4). 141–190.
Ormazabal, Javier & Romero, Juan. 2019. Prolegomena to the study of object relations. Lingvisticæ Investigationes 42(1). 102–131. (doi:10.1075/li.00031.orm) (https://www.researchgate.net/profile/Javier_Ormazabal/publication/334381698_Prolegomena_to_the_study_of_object_relations/links/5e0a51e2a6fdcc28374acfec/Prolegomena-to-the-study-of-object-relations.pdf)
Silverstein, Michael. 1976. Hierarchy of features and ergativity. In Dixon, Robert M. W. (ed.), Grammatical categories in Australian languages, 112–171. Canberra: Australian National University.