According to Martin Haspelmath (in a recent blogpost), differential object marking (DOM) “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”. Haspelmath’s main thesis in this piece is that generative grammar (GG) focuses on finding new solutions to problems already solved within the functionalist tradition. More specifically, he focuses on the remarkable contribution of Kalin (2018) and suggests that, since the functionalist authors mentioned above had already proposed that the DOM is the consequence of “the universal tendency for prominent objects to get special marking”, neither Kalin’s contribution, nor the other alternatives offered by GG (on which her contribution is based) constitute an improvement, because the problem has already been solved.
Yet such a claim has two serious problems, and these are clearly related. On the one hand, it is not true that Kalin is unaware of this explanation. In fact, as early as page two of her article she mentions those studies which, according to Haspelmath, had already explained the phenomenon (although in reading Haspelmath’s piece one might be led to believe that Kalin does not know these studies, something which it seems cannot be the case). Indeed, Kalin mentions Comrie 1979, 1981; Croft 1988, 1990 and Bossong 1991 (which Haspelmath considered to be the authors mainly responsible of the explanation of the phenomenon). Yet Kalin has not overlooked those explanations; she simply does not consider that the generalization therein is an explanation. And this leads us to the second problem: an apparent inability to understand that the objective of generative grammar is different from the objective of functionalist linguistics, probably because their very objects of study are inherently different.
Indeed, there is a considerable difference between establishing a generalization (for example, that the DOM follows from “the universal tendency for prominent objects to get special marking”, in Haspelmath’s words) and specifying the computational mechanisms underlying these facts, which is what Kalin tries to do. Obviously, these are not incompatible tasks; rather, they are complementary. Haspelmath assumes that the DOM is already explained, and that it is sufficiently understood, but such a conclusion is inadmissible from the point of view of those whose aim is to reveal the internal computational mechanisms that derive the grammatical forms that we observe. Allow me a comparison: the discovery that a physical feature of a living being (a wing, for example) developed through evolution from another (a leg) does not guarantee the understanding of what genetic and biochemical mechanisms made the process possible and determined the ontogenetic development of wings in a given organism.
On the other hand, in any science it is dangerous to identify a descriptive generalization with an explanation. When this happens, science stops. What Haspelmath considers an explanation is actually another explanandum for the generativist approach (Kalin’s in this case). Of course, I do not mean that Kalin’s proposal is a complete explanation, nor that it cancels out the functionalist explanation (if only because science never achieves complete explanations), but it does seem to offer a more complete understanding of the phenomenon. Her proposal is not an alternative to the functionalist explanation, but is intended to be a progress in our understanding of the faculty of language. We might ask whether Newton’s physics offered a complete explanation of reality, and the answer is no, although it did offer a more complete explanation than Galileo’s. In turn, Einstein’s physics did not offer a complete explanation of reality, but rather a better understanding of it than Newton’s (without discounting the former). Kalin’s proposal does not offer a full explanation of DOM, but it does offer a more complete and profound understanding of the phenomenon than that offered by functional linguistics in the past.
As I have suggested, the source of the disagreement lies in the different conception of the object of study. For a functionalist, languages are cultural objects that are adaptively moulded as a consequence of their use for communication; for a generativist, languages are different (historically modified) external manifestations of a single computational system that creates internal derivations which are systematically projected in the conceptual-intentional and sensorimotor components. It is unsurprising that our expectations as to what constitutes an explanation in different linguistic theories are themselves different.
Haspelmath insists on several occasions that the explanation of DOM (and in general of grammatical structure) must be a functional-adaptive one. However, if we return to the example of evolutionary theory it becomes clear that these types of explanation (even when they are correct) are clearly insufficient. Darwin was able to found modern evolutionary theory while knowing almost nothing about genetics and nothing at all about molecular biology. Of course, Darwin’s theory works: he convinced us that natural selection produced wings from legs. But this does not imply that the theory provided us with an explanation of the evolutionary mechanism and the anatomy and physiology of wings. Without knowing in detail the genetics and molecular biology which underlie the transmission of traits and the development of the cells that shape organisms, the understanding of the phenomenon of how wings arise, and how they are formed from the embryo, is manifestly incomplete. And it would not be justified to affirm that since Darwin there has been no progress in our understanding of these processes.
The same happens in our case (recall that the title of Haspelmath’s blogpost is “No progress on differential object marking”). Let us assume that the functional adaptive explanation is reasonable and that DOM has its origin in the use of language in discourse (and, of course, in linguistic changes). Does that mean that the statement “prominent objects tend to get special marking” is an explanation of the computational mechanisms that underlie the linguistic behaviour of speakers? How is it represented in the internal grammar of speakers so that in certain constructions the object takes accusative case, dative case, or no case at all? Or is it enough to say that speakers “know” that prominent objects need special marking?
Generative grammar is in this sense analogous to the molecular biology that underlies our understanding of any adaptive process in natural evolution. How did the descendant of an animal with legs “know” that it has to develop wings? Darwin did not explain this question (nor did he need to), but it is obviously a question that requires an explanation if we want to understand the phenomenon from a scientific point of view.
Haspelmath concludes that “Kalin’s proposal fails to provide an explanation of the most basic generalization: that special coding is reserved for prominent objects”. But such a conclusion reveals a lack of understanding of what the objective of Kalin’s research is. Her model explains (very elegantly, in my opinion) why there is DOM with specific and animated objects yet not the other way around; it also predicts the sensitivity of the phenomenon to the animacy/person and specificity/definiteness scales formulated by functionalist authors. In simplified terms, Kalin’s model explicitly formulates a model of the derivative mechanism in which the inclusion (in certain parts of nominal structure) of a feature that requires licensing causes differential case and/or agreement marking to emerge. The explicitness of the model even makes it possible to formally represent the differences between the languages in which there are different types of DOM and those in which there is no DOM at all. Such formal explicitness is precisely what makes it possible to begin an exploration of how the mind and the brain process these constructions in real time, or how the mind of a child who is acquiring a language can internalize these patterns; also, and this is especially relevant in the present context, it offers an explicit framework in which to specify which linguistic changes could have given rise to DOM grammars and, therefore, to the typology of the phenomenon itself. Of course, as Haspelmath suggests, the history of each language, and not UG, determines what type of DOM (if any) a language will have, but it is only when we have an explicit model of what formal mechanisms generate constructions with DOM that we can understand what linguistic changes might give rise to grammars with this grammatical phenomenon and its typology.
Haspelmath is an admirable linguist and has made notable contributions to linguistic typology and to the theory of grammar itself, but he seems to have a blind spot regarding the meaning of generative grammar as a scientific enterprise. It is understandable, and perhaps to be expected, that he sees not real progress in a theoretical model of the computational mechanisms of case assignment and agreement (although he recognizes the merit of Kalin’s work), because from the functionalist point of view the explanation of a phenomenon by definition ends when you find its function. From the point of view of generative grammar, of course, this is an incomplete research program. A deeper understanding of DOM (as with any other grammatical phenomenon) does not only imply a why, but also a how.