奈良教育大学学術リポジトリNEAR
Histocompatibility in Syntax
著者 TAKAHASHI Koji
journal or
publication title
奈良教育大学紀要. 人文・社会科学
volume 35
number 1
page range 1‑6
year 1986‑11‑25
URL http://hdl.handle.net/10105/2103
Histocompatibility in Syntax
Koji TAKAHASHI
(Department of English Linguistics, Nara University of Education, Nara 630, Japan) (Received April 16, 1986)
As the main topic in developmental biology in general is the uniformity of development abstracting away from genotypic variation, so in generative linguistics we idealize to the genotypical invariance of Universal Grammar aiming at a genetic theory of language acquisition. Nevertheless, this attitude does not deny the significance of the problem on grammatical variation in the same speech community. In fact, in my paper (1978) I have already reported a generative perspective on "Cognitive Linguistics , the core issue of which is about the phenomena of structure disruption at certain levels. Suppose that a slightly di庁erent brain is doing things a normal brain would not, this will mean that an insult affected mentally or surgically to some area of the brain results in a selective loss of the grammatical/pragmatic ability to understand and produce sentences. The case of selective impairment to language me‑
chanisms, either a result of a loss of grammatical knowledge or a disruption to some process that brings this knowledge to use, must obviously be compatible with the internal organization of the theory of syntax. Concerning these aspects of grammar and the algorithms Grodzinsky's (1985) assessment of competing theories of syntax could be seen as encouraging support to my previous metho‑
dology.
The present purpose of this paper is to reexamine and check the essen‑
tial properties of "breakdown‑compatibility'一within the Pisa‑framework, where
independently motivated principles are to contribute to broadly construed Linguistic Pathology.
1. Case and the Theta Criterion
I will first review the Case Filter and the theoretical connection between the Filter and the ^‑Criterion. Chomsky in LGB (p.334) modifies the Case Filter, which requires an NP that has a phonological matrix also to have Case, to (1) incorporating the concept of Chain:
(1) Every lexical NP is an element of a Chain with Case. Also the ^‑Criterion is stated in terms of the concept of Chain (LGB, p. 334f.). Informally it says, that every thetaィ01e is assigned to one and only one chain containing exactly one argument, and that every chain containing exactly one argument must have one and only one theta‑role.
If the ♂‑Criterion is a principle which every grammatical theory has to include in some way or other, and if any statement about the structure of language is said to be a statement about the structure of the mind, there must be a device that makes a chain visible for the Theta Criterion. The Case Filter is a necessary stipulation for the structural
1
2 Kou Takahashi
information in this sense. We should understand this internal image within the GB‑
theory when we say that the Case Filter is derivable from the ♂‑Criterion. For instance, sentences like
(2) *It seems [.John to capture the essence]
will be ruled out by the ^‑Criterion; John is assigned no Case in this construction, as it is the subject of an in丘nitive and therefore in an ungoverned position. Lacking Case as a structural information, John is invisible for the ^‑Criterion. It receives no 6ィole, and the sentence is out by virtue of the Criterion, since it contains a chain, with John not assigned a 0‑role, and since the subject 0‑role of the VP infinitive is not assigned to any chain. Sentences like the one in (3), on the other hand, are not ruled out by the Theta
Criterion. In (3) there is a chain composed of antecedent and trace (John,‑, t,), which is visible to the Criterion, as John is in a Case‑marked position, and is an argument.
(3) John,‑ seems [,t,‑ to capture the essence]
Given our Lattice Theory (1984), all S‑structure representations of the sentences displaying "Monadic ♂‑sharing Tendency'contain a trace in subject or object position via Case Absorption. Following Grodzinsky, suppose that every trace at S‑structure had been deleted with its associated index. Theta‑role assignment will now be restricted to lexical NPs in GF‑^‑positions. NPs in non‑#‑positions will not be able to inherit a 6ィole by virtue of their being in a chain with a trace, since the trace had been deleted. In this case, every NP in a non‑0‑position that had not been assigned a 0‑role structually will be assigned the role of Agent by default if it appears in a clause‑initial position.
The result for the syntactic passives and object relatives will be the following:
(4) a. The boy was pushed by the girl
1 l
Agent Agent
b. The boy that the girl was pushing was tall
J J
Agent Agent
c. Show me the boy that the girl pushed
L I
Agent Agent
We may safely conjecture that this theta‑role conflict is invoked for lack of trace‑Memory,
since the two correlated properties of β一marking the subject and Case‑marking the object
are not captured by the learner, and the unacceptable thematic structure emerges. In the active set of cases, where no trAces appear, hence none are deleted, no such thing occurs.
However, the learners are said to get the right interpretation for the wrong reasons, so he continues to argue. In order for the syntactic theory to be breakdown‑compatible, the distinctions that brain damage makes must be statable in the model as a descriptive
generalization over impairment patterns, and this actual argument seems to support the relevance of Chomsky s Extended Projection Principle.
2. Dual Visibility
Everett (1985) proposes an integrational theory of the crosslinguistic contrasts in clitic behavior based on a view of clitics (and morphological categories in general) as bundles of features. Clitics are assigned the features [C(ase)], [A(rgument)], [pronoun], [anaphor]
in Universal Grammar with values士 and α (‑open) for each feature determined by the interaction of independent principles of UG (e. g. the ^‑Criterion or Binding Condi‑
tions) and specific properties of individal languages. As an innovation to my knowledge, English passive ‑en is treated as an example of a [‑C/+A] clitic:
(5) a. John,‑ was beaten t,‑ (S‑structure)
b. [s[NP] [INFL" [CL ‑en] [INELα‑1 to be十past] [VP beat John]
(D‑structure)
As a +A element in INFL, ‑en as a nonmaximal projection must receive the external 0‑
role of beat. Also, ‑en must receive the index of the subject position to which its 0‑role is assigned. This is the reason why pleonatic elements cannot be inserted to derive:
(5) c. *There was beaten John.
d. *It was beaten John.
Then the external βイole is absorbed but nominative Case is still assigned to [NP, S]
since ‑en is [‑C], thus allowing John to raise as a realization of Monadic ^‑sharing Tendency already mentioned. If this analysis is correct, this bound morpheme is still visible for the Theta Criterion, in spite of having no Case. This leads to their modification of Visibility.
(6) Visibility:
(i) α is position‑visible (one of the Position‑Dependent Properties) iff for some β (β‑Ⅹ仇蝣・). β has Case and β‑α or β is a projection of α;
(ii) α is head‑visible iff for some γ (γ‑X‑), γ contains α and γ is position‑
visible.
Thisバdual recognition model as polycategorial syntax will serve to explain various levels
of syntactic misgeneration ("agrammatism ).
3. Theta Complex
In Koopman & Sportiche (1985) their new suggestion on long distance wh‑extraction out of wh‑islands is that these asymmetries within a language, depending on the categonal type and position of the extraction site, and across languages, with respect to a given position and /or category, all result from the interaction of a single condition with inde‑
pendently motivated parameters of 0‑role and Case assignment.
Concerning the basic condition governing extractions out of wh‑islands, the lack of typological variations indicated in recent studies suggests the following generalization:
4 Kou Takahashi
(7) If α is a non‑♂‑position, α cannot be a long wh‑extraction site.
As the counterpart of (7) an extended generalization is obtained which seems, they assert, to hold with overwhelming typological regularity:
(8) α is a possible wh‑extraction site (out of a wh‑island) if α is a ^‑position.
A cosequence this generalization captures is, for example, that direct objects are extrac‑
table out of wh‑islands. We may observe prima facie counterexamples in which cases of 6‑positions are not possible extraction sites (subjects in many languages, object in some), but they argue that independently motivated assumptions on the structure of 0‑marking and Case‑marking, again with respect to an internal image of Pisa Theory, will show that these positions are always non‑e‑positions. Besides the recent proposals as to the Adjacency Condition on Case assignment and directionality of Case and 0‑marking (cf. my simul‑
taneous symmetrical directionality in TAKAHASHI (1984)), they introduce the 0‑complex connectedness condition :
a e‑position must be adjacent to its 0‑marker M or to a 0‑position (^‑dependent on M.
In addition to the archetypal Dutch data, where direct objects are not in a ^‑position via movement to a non‑^‑position because of Case, the treatment is extended to subjects in the languages in which they are not extractable. When they argue that English subjects are in a non‑0‑position of the surface subject position, we notice in retrospect that in Brody (1984) launching sites of movement and phonetiaclly unrealized NPs with the interpretation of roughly "for someone are shown to be in complementary distribution as α and (ll) illustrate.
It is illegal e to go there (‑"It is illegal for somebody to go there ) a. *Who is it illegal e to go there
b. *Tom is illegal ♂ to go there
As an attempt to develop some deep principle of UG, he attributes the default computa‑
tion to a ^‑Criterion violation, namely, since the Empty Categories are ungoverned there emerge multiple pernicious chains against Chain‑visibility principle (cf. (6)).
Koopman and Sportiche continue to argue that the French sentences 'Jean voit Pierre has the structure in
[[NP Jean] INFL [X[NP*t] VP]]
where NP is the surface subject position as (Spec, INFL) and NP* the underlying sub‑
ject position meeting the 0‑complex connectedness condition (9), the X‑constituent being a small cause with VP head.
This treatment will make it possible to reinterpret the Italian data on subject extrac‑
tion. In Rizzi (1982) his basic empirical observation is that Italian allows a highly selec‑
tive class of violations of the Wh Island Constraint: a relative pronoun can be extracted from an indirect question. The solution in which the grammar admits the violation of both TSC and SSC (Opacity)‑ is empirically superior and the results arrived at there were summarized as follows :
i. Wh movement doesn t obey Tensed S Condition and Specified Subject Condi‑
tion.
ii. Wh movement does obey the Subjacency Condition.
iii. The bounding node which is relevant for subjacency is S, not S.
All the examples considered in support of his hypothesis involve the movement of a rela‑
tive pronoun from object position (direct object, indirect object, prepositional object ‑ all ^‑positions), but not from subject position. In the present theoretical context his pro‑
posal can be reformulated as a realization of ^‑opacity. Extraction out of domain D is possible iff D is a ^‑position with respect to some predicate that governs D.
4. CONFL Decoupling
We conclude this brief, but promissory note for more comprehensive work by examin‑
ing the di汀erence in the operator COMP vs. TENSE.
Given small amounts of paths for the category of movement rules, which can be in‑
duced by the directionality of the deductive link through some unique interactions within the subsystems of principles, certain newly designated elements (possible TENSE in this case), being accorded no privileged ontological status in these domains, must necessarily exploit the options already available. The tense‑interpretation differences among certain clausal complements (control complements, gerundive complements, exceptional Case mark‑
ing complements (ECM), and raising complements) are essentially the corollary of different functional activation of CONFL as a special instance of agreement system (COMP十INFL,
in an extended view of Platzack (1983)).
In以The Tense of Infinitives , Stowell (1982) argues, concerning the COMP position
of clausal structures, that gerunds lack the COMP position entirely, unlike infinitives and tensed clauses. In their tense‑interpretation, gerunds also differ from tensed and infini‑
tives with respect to the presence of a tense operator. Especially in contrast to the tense in the case of control complements, which must include a COMP position, the understood
tense of a gerund is completelyりmalleable to the semantics of the governing matrix verb,
which implies that gerundive complements have no internally determined tense and there‑
fore that their understood tense is externally determined. Infinitival clauses and 丘mte tensed clauses share the two properties that gerunds lack : they contain a clause‑internal COMP position, and they contain a tense operator which fixes the understood time frame of the complement clauses relative to the tense of the matrix. Thus, Stowell assumes that the COMP position is where tense operators must appear at some level of grammatical representation. Then the contrast between gerunds and infinitives with respect to the internal specification of tense would follow as an automatic consequence of the alleged fact
that gerunds have no COMP position. If COMP may appear only as a constituent of S′,
it follows that there is no COMP position in a complement clause, ､vhenever the govening verb assigns Case to its subject. This means that ECM complements (and Raising comple‑
KOii Takahashi
ments with Case Absorption) and gerundive complements are to pattern together with each other with respect to TENSE. He continues to state that only under the assumption that Case can only be assigned to the subject position of the infinitive directly by the matrix verb does it follow that the COMP position must be unavailable in precisely these structures where the tense operator is suppressed. However, the gerundive complements do contain "parasitic COMP nodes with an empty operator in spite of the apparent behavior as if they lack a COMP position in their first cyclic domain, which is irrelevant for the questions at issue in this paper. Certain syntactic features (Case, thematic roles, and tlle relevant understood TENSE) are therefore subject to the "histocompatiblity Component for the independent feature spreading devices. The contingent nature of S‑bar deletion is tentatively represented as an aspect of CONFL decoupling differentiation.
(14)
CONFL COMP INFL/＼
I I
±S′ 士TENSE
Control complements 十 Gerundive complements
(+) ECM complements Raising complements
References
Brody, M. (1984) …on the complementary Distribution of Empty Categories, " paper presented at the
1984 GLOW conference, Copenhagen.
Chomsky, N. (1981) Lectures on Government and Binding, Foris, Dordrecht.
Everett, D. (1985) The Parameterization of Nominal Clitics in Universal Grammar, " paper presented at the 1985 GLOW conference, Brussels.
Grodzinsky, Y. (1985) "Neurological Constraints on Linguistic Theories, " paper presented at the 1985 GLOW conference, Brussels.
Koopman, H., and D. Sportiche (1985) "♂‑Theory and Extraction, " paper presented at the 1985 GLOW conference, Brussels.
Platzack, C. (1983) Germanic Word Order and the COMP/1NFL Parameter, Working Papers in Scandi‑
navian Syntax.
Rizzi, L. (1982) Issues in Italian Syntax, Foris, Dordrecht.
Stowell, T. (1982) "The Tense of In丘nitives, " Linguistic Inquiry, 13.
Takahashi, K. (1978) 'A Bounded WH‑syntax against Overgeneration", Bulletin of Nara University of Education, 27.
(1984) "Domain‑preference and Gap Modulations, " Bulletin of Nara University of Education, 33.
