You are on page 1of 10

Lecture 1 August 2006 TOTAL UNIFORMITY A Non-parametric Theory of the Faculty of Language (Miyagawa, in preparation) Shigeru Miyagawa 1 Introduction:

Setting the Stage (1) Total Uniformity (Miyagawa, in preparation) The linguistic system responsible for creating the expressions of a language is invariant across all human languages. (2) FL is non-parametric, hence invariant across all languages (Miyagawa, in prep.) a) Universal Grammar: initial state of the Faculty of Language (FL) (Chomsky 2006) b) UG, as initial state, is also the final state of complete mastery of a language. c) The perceived differences for the linguistic systems of various languages is an ILLUSION. d) There is no such a thing as language acquisition. e) The only differences are the words of a language lexicon (Borers earlier work). f) The only acquisition involved in learning a language is simply learning the lexical items of a language g) Theories that crucially assume learning some part of the linguistic system cannot be right, viz, principles-and-parameters, Optimality Theory. The latter assumes that the language learner must learn the ordering of the universal stock of grammatical constraints. This may be fine for phonology, but not for narrow syntax. (3) Some questions that immediately arise: a) What is the function of agreement? Why does it appear in some languages but not in others? E.g, Indo-European vs. Japanese b) What is the function of movement? Why does it occur in some instances but not in others? E.g., wh-movement vs. wh-in-situ languages c) Why does the basic word order vary from language to language? E.g., SOV vs. SVO

2 What is the real subject matter of linguistic theory? (4) To account for how a language makes infinite use of finite means (Humbold 1836; Chomsky 1955/75, 1965).

Beyond this single constant, linguistic theories from the outset have been guided by two considerations. (5) (i) We must construct descriptively adequate grammars for particular languages. (ii) We must give a general theory of linguistic structure of which each of these grammars is an exemplification. (Chomsky 1955/75: 77) Chomsky (1965: 25-27) refers to (1i) as descriptive adequacy and (1ii) as explanatory adequacy. 3 Standard Theory, Government and Binding Theory, and the Minimalist Program 3.1 (6) (7) Standard Theory: Emphasis on Descriptive Adequacy (e.g., Chomsky 1957, 1965, 1968) The universality of language is expressed as types of rules available to language. Phrase Structure Rules

The base component is a system of rules that generate a highly restricted...set if basic strings, each with an associated structural description called a base Phrase-marker (Chomsky 1965: 17). The rule type in the base are phrase structure rules, which are rewriting rules that give rise to the elementary phrase structures. The creative aspect of language arises, in part, from the recursive nature of phrase structure rules (Chomsky 1955/1975: 171). (7) VP > V (NP) (NP) (PP) (e.g., English) (8) VP > (PP) (NP) (NP) V (e.g., Japanese) (9) Transformational Rules

A transformational rule is a single-valued mapping of certain strings in P into P (Chomsky 1955/1975: 309). A transformational rule often involves permutation, which carries a1 - a2 - a3 into a3- a2-a1, for any string a1, a2, a3 (Ibid.). (10) Standard Theory: the emphasis is on descriptive adequacy

2.2

EST/GB: Emphasis on Explanatory Adequacy Theory (Chomsky 1977, 1979, and especially 1981)

While standard theory emphasized descriptive adequacy, whereby grammars must meet the requirement of describing all and only the grammatical strings of a particular language, expressed in the universal rule types of phrase structure and transformational rules, in GB, the emphasis shifts to explanatory adequacy (Chomsky 1995: 7). Explanatory adequacy requires that we give a general theory of linguistic structure of which each of these grammars is an exemplification (Chomsky 1955/75: 77). The theory of UG must meet two obvious conditions. On the one hand, it must be compatible with the diversity of existing (indeed, possible) grammars. At the same time, UG must be sufficiently constrained and restrictive in the options it permits so as to account for the fact that each of these grammars develops in the mind on the basis of quite limited evidence (Chomsky 1981: 3). (11) Parameters In GB, the universality of language follows from the view that UG is composed of a number of fundamental principles that sharply restrict the class of attainable grammars...but with parameters that have to be fixed by experience (Chomsky 1981: 34). The set of principles in UG includes the following (Chomsky 1981: 5). (12) (i) bounding theory (ii) government theory (iii) !-theory (iv) binding theory (v) Case theory (vi) control theory Example of parameter for bounding theory (Rizzi 1978) (13) *Your brother, to whom [S I wonder [S which stories [S they told t]]], was very troubled. (14) tuo fratello, [S a cui mi domando [S che storie [S abbiano raccontaro t]]], era molto preoccupato. Rizzi (1978) suggests that in Italian, the value for the bounding nodes are fixed as S and NP, and in (14), only one S is crossed so that there is no violation of the bounding theory. (15) Difference between Standard Theory and GB The principles-and-parameters approach is a much more restrictive theory of UG in that, while the rule types of the standard theory allow numerous variations in the way the rules

are formulated for any given language, in the principles-and-parameters theory, the hope is to find that the differences between any two languages are reducible to a single parameter, fixed in one or another way (Chomsky 1981: 6). Thus, the difference between English and Italian is not that they are subject to two different bounding theories, but rather, it is the same bounding theory with the parameter fixed as S (and NP) for English and S (and NP) for Italian. At the initial state, the principles are given as UG, and when the parameter of each principle is fixed, a particular grammar is determined... (Chomsky 1981: 7). 3.3 Minimalist Program: Beyond Descriptive and Explanatory Adequacies (16) In the standard theory the evaluation metric focused on whether a particular theory adequately describes all and only the grammatical strings of a language within the confines of the universal rule types, phrase structure rules and transformational rules. This is the evaluation metric of descriptive adequacy. (17) In GB, the emphasis shifts to hypotheses about a universal set of principles, each principle parametrized along restricted values, that together adequately explain the phenomenon of human language. This is the evaluation metric of explanatory adequacy. (18) In the Minimalist Program (MP), there is yet another shift, a shift that is best characterized as conceptually breaking from the original view that linguistic theories are internally motivated and evaluated, as outlined in Chomsky (1955/1975). What is required of the linguistic theory in MP is no longer an internal matter, but rather, the requirements on it are imposed from the outside in the form of bare output conditions at the interface levels of PF and LF, or the articulatory-perceptual and conceptual-intentional interfaces (Chomsky 1995: 199, 219, 221). What has not changed is that the computational system pairs sound and meaning of an expression (Chomsky 1955/1975: ?). In the standard theory and GB theory, this pairing is mediated by an independently motivated complex system of rule types (standard theory) or universal principles (GB). In MP, there are no rule types or universal principles to evaluate. In MP, UG, or the Faculty of Language (FL), is a computational system that links the two interfaces in the most optimal fashion, that is, the computation is maximally economical (Chomsky 1995: 224). If what is produced by the computational system is compatible, it converges at that interface, and the output of a particular computation is legible, meaning that it is usable by the external component (Chomsky 2000: 95). 4 No Principles, No Parameters: Total Uniformity (Miyagawa, in prep.) (19) Strong Minimalist Thesis (Chomsky 2005b: 3; cf. also Chomsky 2000, 2001, 2005a) ...language is an optimal solution to interface conditions that the Faculty of Language must satisfy.

(20)

SOUND

MEANING

Faculty of Language (21) Uniformity Principle (Chomsky 2001: 2) In the absence of compelling evidence to the contrary, assume languages to be uniform, with variety restricted to easily detectable properties of utterances. (22) Silent Principle: all languages share exactly the same features; some features are simply not pronounced (Sigurdsson 2003). (23) Total Uniformity (Miyagawa, in preparation) The linguistic system responsible for creating the expressions of a language is invariant across all human languages. (24) The only source of differences can only be lexical items, which must be learned for each language that is a given and cannot be avoided. This means that there is no language acquisition, aside from learning of words of the language. Cf. early work by Borer, also the notion of microparameter in Manzini and Roussou 2002, in Subjects, expletives, and the EPP, P. Svenonious, ed. 5 Total Uniformity Exemplar I: Source of formal agreement features, and why dome languages dont have them Chierchias (1998, NLS): Nominal Mapping Parameter (25) NPs are mapped differently depending on the language, and this difference is expressed with the [arg] and [pred] features. (26) The combination, [+arg, pred], is for a language in which the NP denotes a kind, and it is able to function as an argument without requiring an article. When one has to count, one must use numeral classifiers. a. san three b. *san hon three books c. san-satu-no hon three books 3-CL-GEN book A numeral cannot directly count the denotation of NP in Japanese-type languages, since the NP denotes a kind. (27) In contrast, Indo-European languages, for example, typically have the feature bundle [+arg, +pred] associated with the NP, which allows the NP to occur as an

argument or a predicate. This type of language contains count and mass nouns, allowing for singular/plural distinction, in turn predicting that this type of language would have number agreement. (28) What about other kinds of "-feature agreement, such as person and gender? Chierchias theory hints at an approach we could explore. The two types of languages differ in whether the NP takes on a DP or not; it does in Indo-European and other "-feature agreement languages, but it does not need to in a [+arg, pred] language such as Chinese, Japanese, and Korean. Suppose that D is the locus of "feature agreement, so that in the Indo-European-type languages, which have a DP, there is a place for "-feature agreement to merge, on D, and the Stem Feature would select it as its goal. But in a [+arg, pred] language, there is typically no D needed, so that there is no appropriate head for "-feature agreement features to merge, leaving the NP without any "-feature agreement. (29) All of the information needed for the Nominal Mapping Parameter is in the lexicon, including whether there is D in the language in the relevant sense or not. 6 Total Uniformity Exemplar II: Agreement and agreement-less languages (Miyagawa 2005, to appear) In agreement languages, agreement is typically associated with the EPP: the goal of the probe-goal pairing is also the target of the EPP. Is there a counterpart of agreement in agreement-less languages? FOCUS. -mo (Hasegawa 1991, 1994, Kuroda 1965, 1969/70) A mo expression always carries focus stress. (30)a. Taroo-wa HON-o Taro-TOP book-ACC Taro bought a book. katta. bought

b. TAROO-mo hon-o katta. Taro-also book-ACC bought Taro also bought a book. The mo phrase gets interpreted outside the scope of negation Hasegawa (1991, 1994). (31)a. John-mo ko-nakat-ta. John-also come-NEG-PAST John (in addition to someone else) did not come. b. John-ga hon-mo kaw-anakat-ta. John-NOM book-also buy-NEG-PAST A book is one of the things that John did not buy. 6

(31a) only has the interpretation that there is at least one person who did not come besides John. It does not mean that someone came, but John didnt come as well, which would be the interpretation if the mo phrase is inside the negative scope. Likewise, (31b) only means that John did not buy something besides a book; it does not mean that John bought something but not also a book. (32) Taroo-ga [Hanako-ga suteeki-mo tabeta to] omotte-i-nai. Taro-NOM [Hanako-NOM steak-also ate C] think-NEG O-susi-dake-da. sushi-only-COP Taro doesnt think that Hanako also ate steak. Just sushi. (33) The EPP targets the focus feature (Miyagawa 2005, to appear; Hasegawa 2005) (34) SUB-moi FOCUS vP ti VP ...Object... (35) OBJ-moi FOCUS vP ... VP ti (36) EPP and Agree (cf. Chomsky 2001, Miyagawa 2005) The EPP targets the goal of Agree. Goal: agreement feature (IE) or goal (Japanese) What and where is the probe? v' v TP T' Neg TFOCUS/EPP v' v TP T' Neg TFOCUS/EPP

(37) Culicover and Rochemont (1983), for example, argue that the focus feature is in the CP domain, for example. (38) Question: is the probe that ultimately gets valued for formal agreement at T (IE) while the probe that ultimately gets valued for goal at C (as in Japanese)? Are they different probes? It should all the exactly the same, under Total Uniformity. 6.1 Evidence that the Focus feature for mo occurs (ultimately) on T (Miyagawa 2005, to appear)

Weak Crossover (39) Hotondo-daremo-oi [sono-hitoi-no tomodati]-ga ti suisensita. almost-everyone-ACCi [hisi-GEN friend]-NOM ti recommended Lit: Almost everyone, his friend recommended. This parallels what we find in English (cf. Mahajan 1990). (40)a. ??Whoi does hisi mother love ti? b. Whoi ti seems to hisi mother ti to be smart? (41) a. Hotondo-daremo-oi [sono-hitoi-no tomodati]-gati suisensita. almost-everyone-ACCi [hisi-GEN friend]-NOM ti recommended Lit: Almost everyone, his friend recommended. b.??Hotondo-daremo-oi [sono-hitoi-no tomodati]-mo almost-everyone-ACCi [hisi-GEN friend]-also Lit: Almost everyone, his friend also recommended. Idioms Only A-movement scrambling can move an idiom chunk (Miyagawa 1997) (42) idiom: kosi-o orosu hip-ACC lower idiom: sit down benti bench ti suisensita. ti recommended

(43)a. kosi-oi Taroo-ga ti orosita hip-ACCi Taro-NOM ti lowered the bench where Taro sat down

b. *?kosi-oi Hanako-ga [Taroo-ga ti orosita to] omotta bentii hip-ACCi Hanako-NOM [Taro-NOM ti lowered C] thought bench the bench where Hanako thought that Taro sat down

(44) a. Taroo-mo kosi-o orosita benti Taro-also hip-ACC lowered bench the bench where Taro also sat down b. kosi-oi Taroo-ga ti orosita hip-ACCi Taro-NOM ti lowered the bench where Taro sat down benti bench benti bench

c. ???kosi-oi Taroo-mo ti orosita hip-ACCi Taro-also ti lowered the bench where Taro also sat down (45) idiom: te-o nobasu hand-ACC extend get involved in

a.

Taroo-ga hoteru-gyoo-ni te-o nobasita Taro-NOM hotel-industry-DAT hand-ACC extended the rumor that Taro became involved in the hotel industry te-oi Taroo-ga hand-ACCi Taro-NOM hoteru-gyoo-niti nobasita hotel-industry-DAT ti extended

uwasa rumor uwasa rumor

b.

c. ???te-oi Taroo-ga hoteru-gyoo-ni-mo ti nobasita uwasa hand-ACCi Taro-NOM hotel-industry-DAT-also ti extended rumor

7 Total Uniformity Exemplar III: Word Order and Linearization (Miyagawa, in prep.) (46) Total Uniformity: there cannot be any variation in word order across languages in the core computational system. Therefore, there is only a hierarchical relation. (47) However, to pronounce a sentence, the lexical elements must be linearized (Kayne 1994, Perlmutter and Postal 1983). (48) Three kinds of linearizations in the literature: A. Relational Grammar: linearization takes place at the end, in the final stratum, after all the rules have applied (Perlmutter and Postal 1983). B. Antisymmetry (Kayne 1994): LCA applies at the beginning of the derivation, when elements are merged. C. Head parameter (cf. Chomsky 1981, Stowell 1981): head-initial, head-final Also: Fox and Pesetsky (2005): a combination of C and linearization statements for each spell-out domain 9

(49): Given Total Uniformity, the Relational Grammar approach must be right. Within core/narrow syntax, there is no word order, but only hierchical order. Linearization comes in only at the point of pronounciation, as the string is transfered to PF. (50) Linearization at PF: head parameter, encoded into the lexical items of a language as X[+/- FINAL]. (51) This leaves the question of specifiers, which are pronounced to the left of the head in virtually all cases. Why? We will answer this question later.

10

You might also like