To the editors:

David Berlinski and Juan Uriagereka’s (B&U) article strengthens the unifying and explanatory role of Case in grammar, as Vergnaud envisaged it. As they note, we moved from a stage with language-particular filters designed to capture why sentences like (1) or (2) are ruled out to a scenario where a unique Case Filter came to capture all the facts—hence, from n Particular Grammars to one Universal Grammar.

1. *Seems Trump to be a thug. (English)
2. *Parece Trump ser un matón. (Spanish)

The idea behind Vergnaud’s letter was elegant and theoretically sound: all languages have (abstract) Case, even if the morphology does not overtly reflect it. If correct, the proposal would also provide additional support for the so-called “Poverty of Stimulus” argument—children’s knowledge is easily activated even if there is weak (or no) evidence in the primary linguistic data.

B&U make reference to three points that can still be regarded as open issues within Case Theory.

3. 1. The [+/-N] feature determines what assigns / receives Case.
   2. Case is mandatory for overt subjects.
   3. Infinitives must have a (silent) subject.

Typically, questions like the ones above arise the moment the empirical base of a given phenomenon is widened, which signals a very interesting and familiar tension between description (coverage) and explanation (understanding). As Norbert Hornstein points out:

B&U show how classical case theory arose. Similar stories can be told for virtually every other non-trivial theory within linguistics. It arose not via the study of lots of languages but by trying to understand simple facts within a small number in some deep way. This is how bounding theory arose, the ECP, binding and more. So why the presupposition (visible in the replies we give to our critics that we do, really really do, study more than just English) that cross linguistic typological investigations are the only sure way to investigate [Universal Grammar]?

In 1986, few years after the introduction of Vergnaud’s suggestion into the Principles & Parameters framework, Chomsky drew a line between two types of Case: structural and inherent. As he noted, structural Cases are nominative and accusative (or ergative and absolutive, to which I return below), and inherent Cases “include the oblique Case assigned by prepositions and [...] also genitive Case, which we assume to be assigned by nouns and adjectives [...] inherent Case is associated with theta-marking.”¹ Why is this distinction relevant? Well, because it distinguishes Cases that make no semantic contribution (structural ones) from those that, presumably, do. Or at least that was the idea at the time.

Consider, in this respect, (4) and (5), where the NP Gandalf receives the same interpretation (the agent of a fighting event), although it receives two different Cases, nominative in (4) and accusative in (5).

4. Gandalf_NOMINATIVE fought in Khazad-dûm
5. I saw Gandalf_ACCUSATIVE fight in Khazad-dûm

Note that Case in (4) and (5) is abstract, although it is easy to show its morphological reflex in the pronominal paradigm:

6. 1. He_NOMINATIVE fought in Khazad-dûm
   2. I saw him_ACCUSATIVE fight in Khazad-dûm

The same is not true with inherent Cases (dative, genitive, locative, etc.), which were taken to assign a constant morphology-and-semantics to NPs. In fact, inherent Case can be seen as the instantiation of situations where morphology (Case) and interpretation (theta-role) go hand in hand.

This whole discussion raises the question of why structural Cases exist in natural languages; if they make no semantic contribution (and, in many languages, morphological either), they should not exist. Note that the discussion has a deeper philosophical background, one that revolves around the idea that formal languages lack the kind of nuances natural ones deploy (most significantly, morphology). Crucially, this also raises the bar of linguistic inquiry demands; by asking why a certain property exists in natural languages we leave the domain of explanatory adequacy (which is where Vergnaud’s approach to Case left us) to go beyond it. That is an unusual type of question, even for conventional hard sciences, as Chomsky has often noted.

Of course, things are more complex if we go into the details. Let me discuss a well-known example. It concerns dative Case, which at first glance seems to qualify as inherent—it is assigned by a preposition. The literature has noted that dative can be affected by processes like passivization or P-suppression / incorporation, just like structural Cases.² Thus, a dative-marked NP can become both accusative (in the so-called Double Object Construction, see (7b)) and nominative (in regular passives, see (7c)):

7. 1. The goonies gave the gems to Mr. Perkins_DATIVE.
   2. The goonies gave Mr. Perkins_ACCUSATIVE the gems.
   3. Mr. Perkins_NOMINATIVE was given the gems.

The structural-like behavior of dative Case is troubled by the fact that it is absent in languages of the Spanish sort. More specifically, dative-marked NPs cannot become subjects in standard sentences (see the examples in (8)):³

This said, raising-to-subject seems possible in constructions involving infinitival clauses in Peruvian varieties, as shown in (9):⁴

The data are further compounded by the fact that Spanish manifests Person Case Constraint (PCC) effects, which bars the combination of a non-third person accusative clitic with a third person dative clitic, thus showing that datives somehow participate into agreement relations, like structural Cases.

These matters have been on the table for decades, posing a puzzle that is not easy to solve: on the one hand, the structural / inherent cut appears to swing from language to language; on the other, even within the same language, one finds hybrid properties associated to the same Case. With the advent of minimalism, in 1993, Chomsky argued that structural Cases should be related to another prima facie imperfection of natural languages: the presence of phi-features (person, number and gender) in verbs.⁵ In particular, Chomsky suggested that the two imperfections should be one and the same, thus reducing structural Cases to a situation in which the phi-features of a functional category are valued by an NP, which gets structural Case in return.

If we put all the pieces together, we get the structure in (11) for a transitive clause like Jason found the golden fleece, where X and Y stand for the elements endowed with subject-oriented and object-oriented phi-features, the loci of nominative and accusative Cases:

11. [ X [ Jason Y [ found [ the golden fleece ] ] ] ]

The position X in (11) is labeled “T” (or “I,” a shorthand for inflection) and can be occupied by in modals and auxiliaries in English. A remarkable property of this language is that it requires for the subject Jason to move to the left of X (from its base position, where we leave a trace), as in (12):

12. [ Jason will (X) [ t Y [ found [ the golden fleece ] ] ] ]

If T is the source of subject-oriented phi-features, then what is Y? In 1995, Chomsky took Y to stand for a functional head responsible for (i) introducing external arguments (agents / experiencers), and (ii) assigning accusative Case to internal arguments (themes / patients).⁶ The proposal sought to provide an explanation to the empirical generalization, made by Luigi Burzio in the early 80s, that passive and unaccusative verbs fail to license accusative Case as much as they fail to introduce external agents.

13. Burzio’s Generalization⁷

There is a universal correlation between [...] assignment of theta-role to the subject position, and [...] assignment of [...] accusative Case.

Chomsky called Y v (little v). I emphasize this because v is, just like the structural / inherent distinction, a key bridge between morphology (Case) and semantics (theta roles)—between sound and meaning, if you prefer. Notice that the connection is not just between external arguments and accusative Case; the external argument must be an agent, or else accusative would not be assigned. This seems to be correct with psych predicates of the Catalan agradar type (Eng. “like” in (14)). Since they select an experiencer, we expect accusative Case to be unavailable, and it is.

The examples in (12) and (14) illustrate the abstract structures in (15), which are in accord with Burzio’s observation:

15. 1. [ NP_AGENT [ v [ V NP_ACC ] ] ]
    2. [ NP_EXPERIENCER [ (clitic) v [ V NP_NOM ] ] ]

In (14), Juan is interpreted as an experiencer and receives dative (presumably non-structural) Case from the clitic-v cluster, whereas the NP los toros (Eng. “the bulls”) gets nominative after agreement with the verb. The same holds, trivially, for other cases where the NP interpreted as an agent can or must be dispensed with: passive and unaccusative VPs. The case of passives is particularly interesting, especially if Collins⁸ is right in treating the external argument as an inherently marked NP, just like Juan is in (14):

16. [ was [ by Jason v [ found [ the golden fleece ] ] ] ]

This raises an issue that reinforces Chomsky’s 1995 proposal in interesting ways: if v can assign a theta-role to the external argument, then it should be able to assign inherent Case to it too.⁹ This is, one could argue, what happens in (14) and in (16) if v assigns inherent Case to the external argument, the prepositions by and a being nothing but its realization.

The pattern we are considering in (14) and (16) is a trait of so-called ergative languages (Basque, Mayan, Dyirbal, etc.), where external arguments (EAs) receive ergative Case and internal ones (IAs) receive absolutive Case. The literature has emphasized that the key distinction between ergative and accusative languages concerns the Case marking strategies deployed for the vP arguments. This has led to classify languages according to their alignment; assuming the distinctions in (17), the different alignments are as depicted in (18):

17. 1. A = EA of a transitive vP (e.g., Bob Dylan won the Nobel)
    2. O = IA of a transitive vP (e.g., Bob Dylan won the Nobel)
    3. S = EA of an unergative vP (e.g., Bob Dylan won)
18. 1. accusative alignment
       
    2. ergative alignment
       

It is not immediately obvious how to understand, formally, the relevant asymmetry in (18), especially so in Chomsky’s (2000, 2008) terms.¹⁰ Putting aside the nature of the alignments themselves (which are subject to much more fine-grained subtleties I put aside), it seems that even Romance languages can manifest corners of ergativity, at least in dative-nominative verbs (psych predicates of the agradar type) and passives, as noted above.

There is a consensus that absolutive Case is structural, but the nature of ergative Case is, like that of, e.g., dative, debated.¹¹ This brings us back to square one: the cut Chomsky established, which was meant to tease apart those Cases that arise as a byproduct of agreement from Cases that piggy-back on configurational properties.¹² The latter were said to be stable, for the very reason configurational information is (an NP interpreted as a pacient is always a pacient, regardless of the transformations it is subject to), but we know that even robustly inherent Cases can be lost (see the pseudopassives in (19)), just like we know that the alleged thematic stability is not so strong under a closer inspection (the NP Assange in (20) can be interpreted as an agent, a theme or a possessor):

19. The situation was talked about
20. The pictures of Assange

The discussion could go on, but I think this is enough to stop. In the previous lines I have focused on one of the many puzzles that arose after Vergnaud’s recovery of Case, namely the distinction between structural and inherent Case—a cut already noted by classical grammarians, and revamped by Chomsky in 1986.¹³ Inherent Cases do not pose a significant puzzle; they match theta roles, so they are welcome inasmuch as they are licensed by the Conceptual-Intentional systems. Structural Cases, like uninterpretable morphology more generally, are not.

Chomsky hinted as a possible way out already in 1986 by conjecturing that Case morphology made NPs visible for theta role assignment, but it was never clear what visibility meant—certainly not a phonetic property alone, as silent NPs receive an interpretation too.¹⁴ Years later, Chomsky tried to relate uninterpretable morphology with movement (two apparent imperfections of natural languages), but the existence of Case assignment without overt movement (so-called long distance agreement) quickly undermined that option.¹⁵ In the last years, Chomsky has entertained the possibility that phi-features (the source of structural Cases) exist to motivate specific syntactic domains (the phases) that roughly correspond to the vP and the CP—the loci of predication and sentence phenomena (or Deep Structure and Surface Structure, to use earlier terminology).¹⁶

Whatever the ultimate account of the distinction between structural and inherent Cases, the thin frontier separating them and their non-trivial parametric nuances, Generative Grammar has always aimed at a principled explanation—be it connected to interface conditions or efficiency considerations. The answers to all these matters are far from clear, but the understanding we have achieved by addressing them can hardly be described, and all of it would not have happened if Vergnaud had not sent that letter to Noam Chomsky and Howard Lasnik back in 1977, as B&U remind us.

Ángel Gallego