LECTURETTE #16: INTRODUCTION TO CASE THEORY: THE CASE FILTER In traditional grammar, 'case' refers to a category of morphophonological alteration related to a nominal's syntactic and/or semantic function. Languages are regarded as varying in the number of distinct cases defined by their grammars, from languages like Finnish and Hungarian with (at least by some tabulations) close to twenty cases to languages like Chinese with none at all. Obviously, the feature 'case' so defined is assumed to be relevant only to those languages that actually exhibit such alterations in the surface forms of nominals. In the late 60's, Charles Fillmore developed a functional theory of syntax based on the notion that there is a universal set of syntactic-semantic roles that different languages may encode in various ways. These roles Fillmore refers to as 'Cases' (note that, whereas in its traditional, morphological usage 'case' is typically printed just like any other noun, in this more abstract usage it is capitalized); as i noted back in Lecturette #12, what Fillmore calls 'Cases' are in REST referred to as 'theta roles'. In REST, 'Case Theory' (note that again, abstract 'Case' is capitalized) has to do not with deep semantic or even syntactic roles but with the surface licensing of nominals: Why can NPs surface in certain positions but not others? For instance, (1a) is a perfectly good English clause. Why does replacing the verb 'bought' with an approximately synonymous noun as in (1b) require the addition of extra material such as the /s/ suffix and the preposition 'of' as in (1c)? Or why is the construction in (2a-b) acceptable when the subordinate verb is finite but when it is infinite a slightly different construction as in (2c) is required? (1) a. Sam bought a new bookcase. b. *Sam purchase a new bookcase c. Sam's purchase of a new bookcase (2) a. That Sam bought a new bookcase surprised Terry. b. *That Sam to buy a new bookcase surprised Terry. c. For Sam to buy a new bookcase surprised Terry. The answer to questions like these within the REST framework was gradually worked out during the 70's and early 80's, mostly by Dorothy Siegel in her 1974 MIT dissertation Topics in English Morphology and by Rouveret & Vergnaud in their 1980 paper 'Specifying Reference to the Subject: French Causatives and Conditions on Representations' (Linguistic Inquiry 11:97-202); Vergnaud further developed his views in his 1982 Paris dissertation D'ependences et niveaux de repr'esentation en syntaxe (published in 1985 by Benjamins). Their solution was that although overt morphological case is limited to only certain languages (note that Siegel, Rouveret, and Vergnaud focussed on two languages, English and French, which have at best rudimentary overt case systems), all human languages exhibit an abstract feature 'Case', and that Universal Grammar includes a general 'Case Filter' which forbids any NP from surfacing without appropriate Case-marking. Specifically, the Case Filter operates at PF, and asserts that an NP that is 'visible' at PF must be Case-marked. In general, PF 'visibility' means having some overt phonological manifestation; it will be seen in Lecturette #21 that it is possible for an NP to be itself unpronounceable, and to lack overt phonological manifestation in that sense, but nevertheless to have an observable affect on the phonology of the string as a whole. Whether such an NP is subject to the Case Filter is debatable. According to REST's Case Theory, Case is a manifestation of the relationship between an NP and a governing head which has the power to 'assign' Case. We shall get into the details of the question of what constitutes a legitimate Case-assigner in the next lecturette; for the moment, we'll just assume that verbs and prepositions (i.e., -N constituents) have the power to assign Case. Case is assigned under the normal conditions of government, i.e. c-(or m-)command in the absence of a barrier. Case Theory accounts for the facts in (1) above by saying that a transitive verb like 'bought' can assign 'objective' (a.k.a. 'accusative') Case to its object NP, in (1a) 'a new bookcase', thereby enabling that NP to pass the Case Filter. Because the clause as a whole is finite (i.e., the verb is explicitly marked for Tense, etc.), it is assumed that Infl, or Agr, or whatever you regard the head of S to be, is also able to assign 'subjective' (a.k.a. 'nominative') Case to the subject NP 'Sam', which is therefore also able to pass the Case Filter. But a noun can't do either, so replacing 'bought' with 'purchase' in (1b) without making any other changes results in an unacceptable string. The 'object' NP can be salvaged as in (1c) by means of a 'dummy' preposition 'of' whose only function is to get it past the Case Filter. (I'll say a little bit about precisely what 'Case' is involved here in the next lecturette.) The overt 'genitive' marking on 'Sam's' in (1c) also serves to get that NP past the Case Filter in the absence of any alternative. Note, by the way, that the exact same subterfuges would be necessary if instead of 'purchase' we used a noun more directly related to the verb 'buy', as in (3); likewise, there is in English a verb 'purchase' which functions just as well as the verb 'buy' with regard to Case Theory. (3) a. *Sam buying a new bookcase b. Sam's buying of a new bookcase c. Sam purchased a new bookcase. {Note, however, that (3a) might be acceptable in a quasi-'telegraphic' style suitable for, say, a picture caption or a newspaper headline. In such a context, the (non-finite) verb 'buying' might be said to retain at least some of its Case-assigning power. As to where the subject would be getting Case, the best answer that occurs to me at the moment is a rather unsatisfactory one involving an ellipted introductory string such as 'a picture of ...', involving a Case-assigning preposition.} The subordinate clause in (2a) is identical to (1a), and this sentence in and of itself is unremarkable for our discussion. But (2b) shows that while 'buy' is able to Case-mark its object even when non-finite, its subject is now vulnerable. There is no way for 'Sam' to get Case in (2b), which is why this string is unacceptable, according to REST. The alternative in (2c) salvages it by means of a different complementizer. Unlike 'that', which is etymologically a pronoun, 'for' is etymologically a preposition, and as such is a possible Case-assigner. The important thing to note here is that within a non-finite clause, the subject NP has no Case-assigner; if it is to pass the Case Filter, it must find a Case-assigner outside the clause. (As we shall see in the discussion of empty categories and the Binding Theory in Lecturette #21, at least some phonologically null NPs are not subject to the Case Filter, and can therefore serve freely as subjects of infinitives.) Fortunately, S is not a barrier to government for its subject, and so an NP in subject position is properly governed by a complementizer, which can therefore assign it Case if it has the power to do so. In REST, the Case Filter is traditionally used to explain movement from object to subject position in passive clauses. In a typical active transitive clause like (1a = 4a), the verb has two theta-roles to assign, and the DS provides it with two NPs to bear them. It also has the power to assign 'objective' or 'accusative' case to its object; as noted earlier, the subject NP gets Case from the head of S. In the passive version of this clause, the verb, being passive, has only one theta-role to assign, hence at DS has a single argument; the subject position is empty as in (4b). (I've included an agent phrase in parentheses; in such a construction, the 'Agent' theta-role is assigned not by the verb but by the preposition 'by', and the agent NP gets its Case from the same governor.) However, while a passive form of a transitive verb as in (4b) is still able to assign an appropriate theta-role to its object, it is *unable* to assign Case to the object position, so a string like (4b) includes an overt NP that, as it stands, cannot pass the Case Filter. Thus, as it stands, this string is ungrammatical. But movement, provided it doesn't violate any constraints, is always an option in REST. In (4b) there is an empty position that can host an NP and, most importantly, because the clause is finite *this position is eligible to receive Case*. So the NP 'a new bookcase' can move to subject position as in (4c), where it receives 'subjective' or 'nominative' Case from the head of S. (4) a. Sam bought a new bookcase. b. {e} was bought a new bookcase (by Sam). c. A new bookcase was bought (by Sam). It's important to understand at this point the perspective on this analysis that has been implicit all along in REST's history but has become increasingly explicit in recent 'Minimalist' work. It is strictly speaking incorrect that in (4c) the NP 'a new bookcase' has moved to the subject position *in order to* pass the Case Filter. Rather, it has moved there because there was nothing to prevent its movement, *and having done so is able to pass the Case Filter*. Indeed, such movement is the only escape hatch for a construction like (4b), so that such a DS can only surface through an SS in which the relevant NP has moved in the indicated fashion. The claim that passive verbs have one less theta-role to assign than their active counterparts, and unlike the latter are unable to assign Case to their objects, is to the best of my knowledge simply asserted in REST; indeed, it might be regarded as the fundamental definition of the active/passive distinction within this framework, since in connection with Case Theory it motivates the movement of the 'underlying' object to subject position that is the structural manifestation of the passive construction. As outlined above, the basic idea is that the verb's theta-grid requires an object, which is therefore generated at DS in object position in order to receive the appropriate theta-role and thus satisfy the Theta Criterion. But the verb is unable to assign Case to its object, which therefore must move into a different position in order to pass the Case Filter. Crosslinguistically this must be regarded as an oversimplification; there are plenty of languages that generate passive constructions involving intransitive verbs. And the ability of analysing passive constructions in English by means of 'mismatches' between theta-role and Case assignment invites similar analyses of other 'valence-changing' phenomena in other languages, such as applicatives, etc. Details of such analyses are a fruitful area of recent research. Before i close this lecturette and proceed in the next one to a more detailed discussion of how Case is assigned, i think this is as good a time as any to introduce the notion of 'chains'. The set of all structural positions occupied during the course of a derivation by a given constituent is called a 'chain'. Obviously, such a set may have only one member, if the constituent in question remains in situ throughout the derivation. On the other hand, such a set can be quite extensive, as in the derivation of a rather complex sentence like (5a); the relevant positions are represented overtly in (5b) by traces (abbreviated 't'). (5) a. That's the new bookcase that Terry asked Morgan whether Hilary believed Tiffany to have seen Sam buy. b. That's the new bookcase that Terry asked Morgan whether (t) Hilary believed Tiffany (t) to have seen (t) Sam buy (t). Properly speaking, although we typically speak of theta-roles and Case as being assigned to NPs they are really assigned to chains, in the sense that a given chain can only receive one theta-role and one Case, and therefore must include exactly one theta-position and one Case position. As already indicated in the case of the analysis of passives, these need not be the same position. In general, the 'lowest' position in the chain (i.e., the one that is c-commanded by all the others) is typically the one theta-position and the 'highest' (i.e., the one that c-commands all the others, the one that the NP occupies at SS or PF) is the Case position, but this is not necessarly always true; in fact, in (5b) the verb 'seen' is quite capable of assigning Case to its object, and unless i'm much mistaken that's where the second 'that' in the sentence gets Case, in the second position it occupies in the derivation. > as for dr. schaufele, personally, > thanx, the lecturettes are very interesting and i had recommended and > passed arouond to a few friends, thanx again... You're very welcome. I'm always glad to hear from people who seem to be benefitting from SYNTHINAR. I'm currently in the process of getting the SYNTHINAR lecturettes ready to give to my own students here in Taiwan, and in the process updating and adding to the collection. Best, Steven -- Steven Schaufele, Ph.D., Asst. Prof. of Linguistics, English Department Soochow University, Waishuanghsi Campus, Taipei 11102, Taiwan, ROC (886)(02)881-9471 ext. 6504 fcosw5@mbm1.scu.edu.tw http://www.prairienet.org/~fcosws/homepage.html ***O syntagmata linguarum liberemini humanarum!*** ***Nihil vestris privari nisi obicibus potestis!***