We are exploring different aspects of the general schema for dependent case assignment given in (1):

(1) If XP bears c-command relationship Y to ZP in local domain WP, then assign case V to XP.

In the last chapter, I presented some evidence that c-command is indeed crucial to dependent case assignment, and I surveyed the different c-command relationships that languages can make use of in their case-assigning rules. A second aspect of this schema that calls for clarification – and where we might look for parameterization – is the local domain WP which must contain both XP and ZP in order for dependent case to be triggered. That is the topic of this chapter.

Marantz’s (1991) original outline referred to domains in two ways. First, he recognized that case assignment is local to (roughly) the clause. For example, the subject and the object of a single clause interact case theoretically, but the subject of a matrix clause and the subject of an embedded clause typically do not. More specifically, he said that one NP causes dependent case to be assigned to another when both are “governed by the [same] VþI complex.” Second, he envisioned the possibility that different cases might be assigned within different kinds of phrase. In particular, the unmarked case for a single NP in a clause might be nominative/absolutive, whereas the unmarked case for a single NP in a complex nominal might be genitive. Generalizing on this, we might expect two sorts of domain effect. First, there might be some variation across languages in whether phrase WP counts as a domain for dependent case marking at all. Second, given that WP counts as a domain, the particular cases that are assigned in relevant c-command configurations might vary with the category of WP. Indeed, we could see this type of variation in dependent case rules as well as in unmarked case rules.

Having characterized in Chapter 1 what I take to be the overarching puzzles of structural case, I proceed by taking stock of what theoretical resources are on hand for addressing it.

The baseline, “lingua franca” idea in the recent generative literature has been that structural case is assigned to an NP by a nearby functional category F when Chomsky's (2000, 2001: 137) relation of Agree holds between F and NP. Moreover, the properties of Agree are fairly well defined in the literature originating with Chomsky (2000, 2001) (see also Baker [(2008, 2013a, c)] for overviews, among many others). First, F must c-command NP in order to agree with it. Second, F and NP must “Match” in the sense that they are specified as having the same feature attributes, valued on NP and unvalued on F. Third, there must be no other NP with the feature attributes that F is looking for that intervenes between F and the NP that it agrees with. Fourth, there must not be a spell out domain (phase) that contains NP but not F. Fifth, NP must be “active” for agreement by having an unvalued case feature. There are also proposals about how Agree might be parameterized across languages. For example, in Baker (2008) I argued that some languages require that an NP c-command F for agreement to happen between them, rather than that F c-command NP, and also that the activity condition holds in some languages but not others. One aspect of this cluster of ideas that is less clear is what the range of agreeing and case-assigning functional heads is, and why. But the usual suspects include finite T/Infl, which assigns nominative case under Agree, and active v/Voice, which assigns accusative. A natural generalization is to say that possessive D assigns genitive case under Agree inside NPs.

The three preceding chapters have explored in detail the schema for dependent case assignment repeated in (1). Instantiations of this schema characterize what undergoes this type of structural case assignment, and where it must be relative to a case competitor in order to do so.

(1) If XP bears c-command relationship R to ZP in local domain WP, then assign case V to XP.

However, (1) does not say anything explicitly about when in the course of a complex derivation a given XP receives dependent case. That is something one should try to be clearer about, and this chapter takes up the matter.

Indeed, there already have been controversies about this issue in the literature on dependent case assignment. Marantz’s original version was linked to the idea that morphological case is determined at PF, after the syntax – a view also adopted by Bobaljik (2008: 150). However, Legate (2008: 85–91) criticizes this view, asking (among other considerations) why movement such as scrambling the object past the subject does not affect which argument gets accusative or ergative case, if case is assigned at PF. For this and other reasons, Baker and Vinokurova (2010) claimed that dependent case was assigned in the syntax; indeed, we assumed that dependent case assignment happened immediately, as soon as a relevant configuration is created by merging an NP into a domain that already contains another NP. See also Preminger (in press: ch. 9) for similar arguments and conclusions. Given these disagreements, we should consider the timing of dependent case assignment within the broader empirical context considered here, together with the related question of how case assignment interacts with movement.

In the previous chapter, I showed that case values can be assigned to NPs by configurational rules of dependent case assignment. This configurational mode of case assignment is an alternative to case being assigned by agreement with a functional head, and it seems to be no small part of the overall theory of morphological case. I have reasoned that all languages that have structural ergative case use this mode of case assignment, including ergative languages proper and tripartite languages. Furthermore, most of the languages with overt accusative case that I have worked on also seem to be of this type, in that they either do not have object agreement or object agreement does not track accusative case very closely. We have also seen that agreement on some functional heads depends on the agreed-with NP already having a particular case through a rule of dependent case assignment or unmarked case assignment.

While agreement-assigned case is arguably reasonably well understood, given the extensive literature on Agree and its properties, the notion of dependent case is less familiar. It requires further development in order to know what contribution it can make to understanding the case properties of the languages of the world. In particular, some conceptual analysis is in order to identify the likely range of parametric variation within the general domain of dependent case marking. Ideally, this parametric variation should be rich enough to be descriptively adequate, and restricted enough to be explanatorily adequate. Exploring the rules of configurational case assignment in detail, then, is the focus of the next three chapters.