From my perspective, differential object marking (DOM) – the universal tendency for prominent objects to get special marking – has been well-understood since the 1980s, but even though the explanation was clearly stated in Comrie (1989) (and also formulated clearly in Croft (1988) and Bossong (1991)), many linguists seems to have forgotten about it.
In an earlier blogpost, I commented on the recent trend for generative linguists to propose new explanations for DOM, and here I discuss a particularly ambitious attempt by Laura Kalin at not only providing an analysis for a kind of special object treatment in the Neo-Aramaic language Senaya, but also to explain DOM phenomena more generally (Kalin 2018). On 47 dense pages, she provides a high-tech Minimalist proposal which must have required an enormous effort, but which sadly does not lead to any new insights.
Recall that the DOM tendency is just a special instance of the general tendency for languages to provide special (asymmetric) coding for unexpected meanings: referentially prominent objects (i.e. animate and definite objects, but also locuphoric (first/second person) and topical objects) are unexpected as objects, and hence they tend to get special markers. Language structures often reflect an efficient trade-off between the hearers’ need for explicit coding and the speakers’ need to minimize the encoding effort. This is a prime example of a functional-adaptive explanation that has stood the test of time.
Kalin says that her proposal also contains an “economy” element, because object marking may be due to an “optional licenser” that “will only merge and be active” when an object that needs to be licensed would not otherwise be licensed. But the “economy” is not related to language use, so it remains at a very abstract level, and Kalin does not want to make the proposal fully concrete in order to keep it compatible with a range of other generative proposals (e.g. both with the traditional view that abstract Case is assigned by a functional head and with more recent dependent-case views). So readers who are not immersed in the recent generative literature will understand little, and I do not claim that I understand everything. There is a very large number of ancillary assumptions that need to be made in order to get the proposal to work, and these assumptions are often motivated primarily by referring to other people’s work that I have not read. But I think I understood that Kalin’s proposal fails to provide an explanation of the most basic generalization: that special coding is reserved for prominent objects. In her proposal, the prominence scales that determine DOM (especially animacy and specificity) are included in each nominal’s functional projection: Every nominal crucially includes a functional sequence like NameP > DefP > SpecP, and every nominal has a sequence like PersonP > HumanP > AnimP – but she gives no explanation why these functional features should occur in this order, and why there are no features like “IndefP” or “InanimP”. Of course, there are “entailments” (and ‘human’ entails ‘animate’, etc.), but why should semantic entailments play a role in a functional sequence? (Nanosyntax also makes extensive use of functional sequences to explain cross-linguistic generalizations, e.g. for case-marking, as recently discussed here, but “entailments” do not seem to be relevant in these cases.) Kalin wants to somehow express the intuition that object marking of a specific nominal is less likely than object marking of a definite (or proper name) nominal, but since she ignores the functional-adaptive explanation, she needs to do it in a technical way requiring a move that makes little sense in her framework (as far as I can see). Moreover, as we know that indefinite and inanimate nominals get special treatment elsewhere (e.g. English requires a special adpossessive construction when the possessed nominal is indefinite: my friend vs. a friend of mine), Kalin still needs to explain why there is no language with an IndefP projection and special object marking for indefinite nominals.
But perhaps the biggest gaps in Kalin’s proposal are (a) the failure to account for the mirror-image differential agent marking phenomena, and (b) the failure to see (and account for) the substantial differences between differential object flagging and differential object indexing (Kalin simply conflates the two).
Differential agent marking (or differential subject marking) shows a range of mirror-image effects, as has been known since the 1970s: In transitive subjects, special (ergative) coding is generally reserved for aliophoric pronouns (Kalin would say “non-Participant”, i.e. 3rd person), or inanimate nominals, or non-topical nominals. This is not as well-known as the generalizations of differential object marking, because ergative coding is not as widespread as accusative coding, and there are no major world languages with differential ergative marking (unlike Spanish and Hindi-Urdu, which are major world languages with DOM). But the mirror-image pattern (clearly stated in Dixon 1979; and recently summarized again in Haspelmath 2019) is no less real, even though some authors have tended to emphasize the ways in which it is not a perfect mirror image. In Kalin’s story, one would expect subjects to be specially marked also when they are animate or definite (as she says clearly on p. 154), but we actually find the opposite. (It is true that special ergative marking of indefinite subjects seems to be rare, but of course languages often ban indefinite transitive subjects entirely and require speakers to do something else, such as passivization – and this can be seen as a kind of special ergative marking.)
Perhaps the biggest error that Kalin makes is to basically equate differential flagging and differential indexing of objects. In §2 and §4, she focuses on Senaya (Neo-Aramaic), which has no accusative flagging, only object indexing that is limited to specific objects. Her technical account is framed in terms of a an abstract Case feature, so in this sense, she claims that Case is relevant for Senaya as well, but all these complications are beside the point, because Kalin assumes the wrong generalization: differential indexing of objects is quite unlike differential flagging, because differential indexing is always conditioned by argument prominence. We see exactly the same pattern with subjects (including transitive subjects): When an argument is referentially prominent (definite, animate, pronominal, locuphoric (1st/2nd person), discourse-given, non-focused), it tends to be indexed on the verb – regardless of whether it is a subject or an object. This was formulated very clearly by Croft (1988), and the basic generalization has been known since Moravcsik (1974) and (1978a; 1978b) (see also Lehmann 1982 for a very clear exposition): Argument indexing is favoured on prominent arguments and highly ranked roles (subject > object > oblique), whereas argument flagging is favoured on low-ranked roles (oblique > object > subject), and is used on subjects and objects primarily if the prominence level deviates from the expectations. It is sad to see that these strong and well-understood generalizations have not made it into the general canon of linguistics, and that smart young scholars like Kalin are being led astray by an intellectual environment that does not seem to be able to integrate these fundamental insights into a larger theory of the world’s grammatical patterns.
Kalin is not completely averse to functional thinking, and she offers a number of speculative remarks, e.g. the idea that some features of nominals (like definiteness and specificity) “need to be anchored in the speech act to be intepreted” (p. 154). But even the most basic point, that “abstractly licensed” nominals must have special marking (rather than not requiring any marking and thus conceivably contrasting with marked unlicensed nominals), is not intuitive at all.
But there is something positive that I can say about Kalin’s paper: In §2.2, she criticizes two classes of previous generative explanations of DOM, represented by about 20 different earlier papers – and these criticisms are justified. The two classes are visibility-based explanations (which claim that specially marked objects are DPs, and other objects and NPs) and raising-based explanations (which claim that only some objects raise to higher arboreal positions, and that only these are case-marked). The visibility-based explanations do not work in general because there is no clear correlation between case-marking and referential-prominence across nominals (subjects often behave differently from objects for case, even though they may have the same kinds of referential interpretations). And the raising-based explanation does not work in general because there are many languages, including Senaya, where the position of definite or animate objects is not any different from the position of objects that lack special marking (this was also one of my points of criticism of Baker’s (2015) book “Case”, Haspelmath 2018).
So why are there so many generative attempts to explain DOM, but none of them has been successful? It seems to me that the fundamental problem is that generative linguists try to achieve two things simultaneously: Elegant description of a particular language, and UG-based explanation of cross-linguistic trends. The focus on one particular language (Senaya in Kalin’s paper, Palauan in T. Levin’s (2018) paper discussed in the earlier blogpost, Tatar in Lyutikova & Pereltsvaig (2015), and so on) typically means that the authors do not see what is truly general (and thus can be explained at the level of Human Language), and what is particular to the language (and thus needs a parochial description, plus perhaps a historical explanation). In Kalin’s paper, the contrast between the language-particular story (in §2 and §4) and the more general theory (in §3 and §5) is particularly striking. The Senaya facts play no real role for the general theory, other than serving as illustrations of some points that could also have been made by using more familiar languages, but the technical “implementation” is so complex that the author may well have lost sight of the bigger picture because of the focus on the language-particular technicalities.
On the one hand, we must of course be happy that studying small, out-of-the way languages has become very prestigious in many corners of linguistics – so in this sense, Greenberg’s empirical universalist programme (first taken up by Ken Hale in the generative tradition) has been a resounding success (compared to the 1970s and 1980s, current generative linguistics has become thoroughly typological). But on the other hand, why should one expect that a study of Senaya would lead to deeper insights into the workings of object indexing and object flagging more generally? It seems that the Chomskyan focus on the language faculty is wrongly leading many linguists to think that by studying a single language, they can make a contribution to general linguistics.
Senaya is unusual in the way it uses two sets of object indexes (the S-suffixes and the L-suffixes) for indexing subjects and objects in different aspectual forms. But a general UG-based theory cannot contribute to understanding such unusual phenomena – for this, one needs parochial theories, e.g. about the surrounding Iranic languages influencing the Northeastern Neo-Aramaic languages (as recently discussed by Coghill 2016: §6.2). One surprising feature of Senaya (and other Neo-Aramaic languages) that Kalin mentions is the fact that the L-suffixes are invariant when used as subject indexes, but differential (restricted to specific arguments) when used as object indexes. It so happens that there is a recent paper on the history of object indexing in Iranic, Haig (2018) (discussed on this blog), which makes a similar point to Kalin’s about Iranic from a diachronic perspective. Haig ends up saying something rather similar to Kalin: that the different behaviour of object indexes and subject indexes cannot be due to their history (because they derive from the same source elements, dative/genitive pronouns), but there must be some kind of “attractor state” which makes subject indexes tend to be invariant, while object indexes tend to be variable (in view of what I said above, one may wonder how strong this is, because even subject indexing is omitted in many languages when the subject is focal or postverbal). Thus, we can partially understand Iranic (and Neo-Aramaic) languages from the perspective of their histories, and partially from the perspective of universal trends – but to identify the universal trends, we need to examine and compare a lot of languages from around the world.
Baker, Mark C. 2015. Case: Its principles and parameters. Cambridge: Cambridge University Press.
Bossong, Georg. 1991. Differential object marking in Romance and beyond. In Douglas Kibbee & Dieter Wanner (eds.), New analyses in Romance linguistics, 143–170. Amsterdam: Benjamins.
Coghill, Eleanor. 2016. The rise and fall of ergativity in Aramaic: Cycles of alignment change (Oxford Studies in Diachronic and Historical Linguistics). Oxford: Oxford University Press.
Comrie, Bernard. 1989. Language universals and linguistic typology: Syntax and morphology. Oxford: Blackwell.
Croft, William. 1988. Agreement vs. case marking in direct objects. In Michael Barlow & Charles Ferguson (eds.), Agreement in natural language: Approaches, theories, descriptions, 159–180. Stanford: CSLI.
Dixon, R.M.W. 1979. Ergativity. Language 55. 59–138.
Haig, Geoffrey. 2018. The grammaticalization of object pronouns: Why differential object indexing is an attractor state. Linguistics 56(4). 781–818. doi:10.1515/ling-2018-0011.
Haspelmath, Martin. 2018. Review of “Baker, Mark. 2015. Case. Cambridge: Cambridge University Press.” Studies in Language 42(2). 474–486.
Haspelmath, Martin. 2019. Role-reference associations and the explanation of argument coding splits. https://ling.auf.net/lingbuzz/004047
Kalin, Laura. 2018. Licensing and differential object marking: The view from Neo-Aramaic. Syntax 21(2). 112–159. doi:10.1111/synt.12153.
Lehmann, Christian. 1982. Universal and typological aspects of agreement. In Hansjakob Seiler & Franz Josef Stachowiak (eds.), Apprehension: Das sprachliche Erfassen von Gegenständen, 201–267. Tübingen: Narr. https://www.christianlehmann.eu/publ/lehmann_agreement.pdf.
Levin, Theodore. 2018. On the nature of differential object marking: Insights from Palauan. Natural Language & Linguistic Theory 1–47. doi:10.1007/s11049-018-9412-5.
Lyutikova, Ekaterina & Asya Pereltsvaig. 2015. The Tatar DP. Canadian Journal of Linguistics/Revue canadienne de linguistique 60(3). 289–325. doi:10.1017/S0008413100026232.
Moravcsik, Edith A. 1974. Object-verb agreement. Working Papers in Language Universals 15. 25–140. doi:http://doi.org/10.5281/zenodo.1423196.
Moravcsik, Edith A. 1978a. Agreement. In Joseph H. Greenberg (ed.), Universals of human language, vol. 4: Syntax, 331–374. Stanford: Stanford University Press.
Moravcsik, Edith A. 1978b. On the case marking of objects. In Joseph H Greenberg (ed.), Universals of human language, vol. 4: Syntax, 249–289. Stanford: Stanford University Press.