Syntax Research Paper

1. Functions Of Syntax

Loosely speaking, the variations in form and position that constitute syntax fulfill a number of functions: The words and phrases of a sentence stand in various meaning relations to each other, and these relations are encoded by syntactic differences. A basic function is the expression of so-called ‘thematic relations’ or ‘argument linking’: a verb such as kill expresses a two-place predicate KILL (x, y), a relation between an agent x and a patient y in an event of killing. Such argument roles of a predicate must be linked to the argument expressions in a sentence, and at least in the ideal case, this linking is carried out systematically in natural language, for example by word order, as in English (1), or by case, as in German (2):

(1) the tiger killed the buffalo.

agent predicate patient

(2) dass der Tiger den Buffel totete

that the.nom tiger the.acc buffalo killed.

Argument linking is not the only semantic-pragmatic function figuring in syntax. Argument places of verbs can, for example, be questioned (who did he see?), and languages may require that question phrases appear in specific positions (e.g., clause initially, as in English). Information structure (‘new vs. old information,’ ‘focus vs. topic’) is a relevant dimension for sentence construction too, and may influence word order, for example. Unlike (2), German (3) seems to be fully felicitous only in a situation in which the crucial new information is the identification of the tiger as the agent of killing.

(3) dass den Buffel der Tiger totete.

A characterization of the functions that syntax encodes must be complemented by an identification of the means of encoding. Word order and shape variation are two options. The latter comes in two varieties. The fact that a noun is the agent of a verb may be encoded by shape variation of the noun ( = case), or by shape variation of the verb (agreement markers)—or a combination of both. ‘Word order,’ on the other hand, seems to be the result of mapping the two-dimensional hierarchical structure of a sentence onto the unidimensional sound (or letter) stream (see Sect. 4 below). Intonational differences often serve to express pragmatic and semantic functions, but they seem to be employed only rarely in argument linking.

2. Argument Roles

In a sentence such as (4), we can distinguish words and phrases that are linked to argument places (in italics) of the predicate (boldface), and words and phrases that are not; which have an adjunct function (underlined).

(4) Jack saw the girl in the park yesterday.

Let us confine our attention to argument roles. They are linked to (the meaning of ) the predicate hierarchically. For a verb such as kill, the agent is higher in this hierarchy than the patient. Grammatical differences reflect these hierarchies systematically. When two arguments of a predicate refer to the same entity, the higher argument will, for example, be realized in its usual shape, whereas the lower argument takes a special form (it is, e.g., realized as a reflexive pronoun) but not vice versa.

(5) (I expect) Bill to wash himself.

(I expect) *himself to wash Bill.

[Ungrammatical strings are preceded by a*]

For prototypical instances of ‘transitive’ relations, that is, for verbs expressing instantaneous causation of an event affecting a patient by an agent (kill, beat), the hierarchy seems to be the same in all languages, with agents always being more prominent than patients. Potential exceptions to this (Dyirbal, Arctic Eskimo; see e.g., Marantz 1984) can presumably be explained away. For less prototypical relations—e.g., for psychological predicates—grammatical hierarchies are linked to semantics less systematically, as the difference between like and please suggests: the ‘experiencer’ of the psychological state is encoded as the subject (as the higher role) with like, but as an object (the lower role) with please. Languages may differ in this respect—in some, nearly all psychological predicates follow the model of please (Basque), in others, all relevant verbs pattern with like (Lezghian), and in still others (like English), both construction types occur (see Bossong 1998).

In addition to identifying their place in the hierarchy, argument roles have often been classified in terms of notions that we have already used informally, namely/thematic (‘theta’) roles (Gruber 1965, Dowty 1991, Jackendoff 1990; but also Hjelmslev 1972, 1935–7, Jakobson 1971, 1936) such as ‘agent,’ ‘instrument,’ ‘goal,’ ‘experience,’ ‘source,’ ‘patient,’ etc. One needs to refer to these notions in the context of selecting the proper preposition or semantic case, but for capturing core syntactic laws, the necessity of integrating theta roles (in addition to hierarchy statements) into grammatical descriptions has yet to be established (apart, perhaps, from the special role played by the agent function). At least, most if not all current grammatical models can do without them easily (Fillmore 1968 and approaches following him are notable exceptions).

3. Grammatical Functions

One issue of syntactic theory is whether argument roles can be linked directly to the formal means of encoding, or whether the rules of encoding crucially employ ancillary notions, namely, the grammatical functions ‘subject,’ ‘direct object,’ ‘indirect object,’ etc. Some grammatical models (Relational Grammar, Perlmutter (1983); Lexical Functional Grammar, (Bresnan 1982)) assume that such grammatical relations are indispensable primitive notions of syntax. Other approaches, such as the Standard Theory (Chomsky 1965), or the Government and Binding (GB) framework (Chomsky 1981) take grammatical functions (at best) to be mere classificatory concepts of little explanatory value, which can be defined in terms of structure and hierarchy: a ‘direct object’ is a noun phrase ‘governed’ by the verb, or it is a noun phrase which is a structural sister of the verb.

Thus, Lexical Functional Grammar assumes that sentences are linked to a functional structure, in which, e.g., the verb kill is linked to two abstract grammatical functions, subject and object. How these grammatical relations are spelled out is a function of language-particular rules. In English, grammatical functions are encoded by word order (more precisely, hierarchically)—the direct object has to follow the verb immediately (1). In German (2, 3), the grammatical functions seem encoded by case marking (direct objects bear accusative case). The mapping from argument structure to formal means of expression would thus proceed as in (6):

(6) kill (role 1, role 2)

kill (subject, object)

kill (nominative, accusative)

Lexical Decomposition Grammar (Wunderlich 1997) translates argument hierarchies directly into morphological markings or order relations; that is, the second step in (6) is skipped. The Standard Theory and the GB framework assume that sentences are linked to an abstract ‘deep structure,’ in which the words and phrases combine hierarchically, with argument role hierarchies being reflected by argument expression hierarchies in deep structure, as in (7). Such deep structures are mapped on to audible representations in various ways, e.g., by case marking, reordering, etc.

(7) [_Infl-Phrase Infl [_VP [the tiger] [killed [the buffalo]]]]

In other words, there is no consensus concerning the status of grammatical functions. It seems fair to say however, that no universally applicable definitional criteria for, e.g., subjecthood, have been found so far. This suggests that grammatical functions must either be considered fuzzy concepts, or that they reflect a grammatical perspective that may be adequate in some languages only.

The nature of grammatical functions has often been discussed in the context of the claim that one type of linking argument expressions to argument places, namely the hierarchical-positional (‘configurational’) one that we find in English, is more fundamental than the others (such as case marking) in the sense that it underlies the grammatical system of all languages: at deep structure level, all languages follow roughly the constructional logic of English. This is the position held by the mainstream version of the GB framework, and has been taken over in the Minimalist program (Chomsky 1995). The contrary view that there are no privileged systems has been a minority position in GB (e.g., Hale 1983), and is the position held by Lexical Functional Grammar and Relational Grammar. Debates in the 1980s primarily contrasted languages with rigid word order and little case marking (English, Chinese) with languages with a relatively elaborate case system and free constituent order (German, Japanese). One outcome is the insight that a strong hierarchy-based configurational component can be identified in the grammar of languages with a fairly rich case marking too (e.g., Haider 1993).

On the other hand, it may also be uncontroversial that polysynthetic languages (such as Mohawk) with a very rich system of verbal agreement have a grammatical system that differs radically from, say, English and Japanese. It appears as if nominal expressions in the polysynthetic languages do not show a grammatical behavior that is even remotely reminiscent of subjects and objects—rather, their grammar is com- parable to what holds for adjuncts (like adverbs) in English. Argument linking and other important grammatical factors seem to be dealt with in the morphology only. Why this is so, and whether this means that there are languages without a syntax proper (a position strongly argued against by Baker 1996) is an open issue.

4. Movement

If more than one semantic–pragmatic relation has to be expressed, and if the same encoding mechanism is used for the different functions, the encoding rules may impose incompatible requirements on a phrase in a sentence. In English, the positional system of identifying arguments requires that objects follow the verb, whereas question phrases have to appear at the beginning of a clause. These requirements cannot be met simultaneously:

(8) (a) *(I do not care) you like who

(b) (I do not care) who you like

One way of dealing with this problem has been the postulation of various levels of representation, such that different requirements have to be met at different levels. In the Chomskyian tradition (see e.g., Chomsky 1957, 1965, 1981) a level called deep structure/d-structure, representing the thematic relations among arguments and predicates has played a prominent role until quite recently. This d-structure (corresponding to (9a)) would then be mapped on to representations coming close to the phonetic ‘surface’ by a set of structure-changing operations (‘transformations’), of which the movement (more precisely: fronting) of phrases has been the most important one (relative to (9a), who has moved to clause initial position in (9b)). The law that requires that complement questions begin with a question word applies at ‘surface structure’ (s-structure) only, not at deep structure.

(9) (a) you like who?

(b) who do you like?

Details changed in the 1990s, but the distinction between structure building operations encoding argument linking and structure-changing operations (movement) is still a key property of all grammatical models inspired by Chomsky. Most alternative approaches (e.g., Generalized Phrase Structure Gram- mar; Gazdar et al. 1985) differ at a purely technical level only. Optimality Theory borrowed from phonology (Prince and Smolensky 1993; see also Grimshaw 1997, Kager 1999, Muller 2000) allows the handling of conflicts between incompatible grammatical requirements in a quite different way: the grammatical principles are ordered, and lower principles (such as: the object appears to the right of the verb) may be violated in the interest of higher ones (such as: question phrases appear clause-initially). Whether such options will eventually reduce the importance of movement metaphors in grammatical descriptions is an open question at the time of writing.

Beginning with the seminal work of Ross (1967), investigations on syntactic movement have brought to light a new type of syntactic law. There is a set of ‘locality conditions’ on movement, in the sense that words and phrases cannot be moved out of certain kinds of construction. Thus, in English, simple complement clauses are transparent for movement, while complement questions or relative clauses are not. (The place where a phrase has moved from is indicated by.)

(10) (a) you think that she likes someone

who do you think that she likes

(b) you wonder when she kissed someone

*who do you wonder when she kissed

(c) you know a man who kissed someone

*who do you know a man who kissed

In (10b) and (10c), we observe an ‘intervention’ effect. The phrase who should be placed in the position indicated by at d-structure (because of the laws of argument linking), and in sentence initial position at s-structure because of the laws of question formation. The ‘preposing’ of who yields ungrammaticality in (10b) and (10c) because phrases cannot cross similar phrases (like when in (10b)) when they move to their surface position.

A similar intervention effect can be observed in (11) when a choice must be made between different phrases that could be placed into sentence-initial position: again, a phrase of type X must not cross a phrase of the same type in this context. This is the so-called superiority effect (see Chomsky 1973, also Chomsky 1995, Rizzi 1990).

(11) (a) who do you expect to buy what

(b) *what do you expect who to buy

Furthermore, only complement clauses with an object function are transparent for movement, but not subject clauses (12a) and adjunct clauses (12b):

(12) (a) *who is to kiss fun?

(b) *who do you weep because you met_ ?

The Chomskyan approach (in the narrower sense) tries to identify a set of purely formal and universally valid syntactic laws that characterize such ‘island’ effects (see Chomsky 1981, 1986). The empirical challenge is that the constraints proposed so far are not being respected in all languages in the same way.

5. Logical Forms

Not all languages require that constituent questions be formed by the fronting of a question word or phrase (wh-phrase). In Chinese, wh-phrases are realized in the position where they should appear according to the laws of argument linking.

(13) ni xihuan shei

you like who

‘who do you like?’

Interestingly, the scope options of Chinese wh- phrases are constrained, and in a way reminiscent of what we saw in the preceding section for English. Thus, a wh-phrase which takes scope over the whole matrix clause may appear as part of an object clause (14), but cannot show up in a subject clause (15). This is identical to what we have just observed for English (10a), (12).

(14) Zhangsan xiangxin shei mai-le shu

Zhangsan believe who bought books

‘who does Zhangsan believe bought books?’

(15) *Zhangsan tao-le shei zhen kexi

Zhangsan marry who real pity

‘for which x: that Zhangsan married x is a real pity’

Such observations gave rise to the idea (Huang 1982), quite popular in the GB framework, that shei undergoes movement to the clause initial position corresponding to its semantic scope as well (so that it is clear why its scope behavior is constrained by restrictions on movement), but ‘invisibly’ so. There are two ways in which the relevant observations may be interpreted, one that gives priority to syntax, and one that does not.

The GB framework postulated a grammatical architecture as in (16): by putting together words from the lexicon, a deep structure arises. This d-structure undergoes a number of transformations, mainly movement, in order to reach the interface levels Phonetic Form (PF) (interface with the articulatory–perceptory system) and Logical Form (LF) (interface with the cognitive system). The first steps, so-called ‘overt’ syntax, are common to these two derivations targeting PF and LF; they yield surface structure (s-structure). ‘After’ s-structure, the two paths split. The syntactic derivation of LF proceeds in the standard way, but no longer affecting phonetic material (covert syntax).

In this model, the Logical Form of (14) is (17), i.e., question words are preposed in Chinese too—the only difference from English being that the movement of shei has applied ‘after’ s-structure, so that it lacks a phonetic effect. All languages, it is claimed, have the same Logical Form, but they may make (among other things) different decisions as to whether a certain movement operation (like the fronting of wh-phrases) applies in overt or covert syntax.

(17) shei Zhangsan xiangxin mai-le shu

who Zhangsan believe bought books

Note that the major argument for such a view is the empirical observation that conditions that constrain c visible movement in some languages (see (12)) restrict the scope-taking behavior of certain phrases that have not been moved visibly (see (15)) in other languages. If one presupposes or shows that the relevant constraints c are formal-syntactic in nature, the conclusion that there is a syntactic component with invisible movement is quite compelling—but one must keep in mind that the overt movement operations constrained by c serve the purpose of scope-taking too.

Thus, one can also take quite a different perspective on facts as represented by (12) and (15): language could be governed by an elaborate system of semantic constraints on scope taking, and apparent syntactic conditions on movement might turn out to be superficial consequences of these constraints. Such semantic constraints affect the distribution of, for example, question words quite independently of whether they are moved syntactically or not. Proposals that can be related to this idea can be found in the literature (see, e.g., Erteshik-Shir 1977, Szabolsci and Zwarts 1997), but a full semantic account of constraints on movement has yet to be developed.

6. Structural Hierarchies

The previous section illustrated the difficulty of disentangling syntax and semantics in the domain of constraints on movement. This difficulty reappears when one considers the role which the geometry of abstract syntactic structures plays for quite a variety of relations and processes. (18)–(20) illustrate probably universal asymmetries that are cases in point: the pronoun he can be bound semantically by no one in (18a) only, but not in (18b): (18b) cannot have the interpretation sketched in (18c). E er in (19) is a socalled ‘negative polarity item.’ It can appear in a sentence only if there is a negating expression present in that sentence. Nobody manages to license the polarity element e er in (19a), but not in (19b). (20) illustrates the fact that question (wh-) phrases are almost always moved to the left (if at all), but not the right.

(18) (a) No one thinks that he is incompetent.

(b) *He thinks that no one is incompetent.

(c) Vx (x thinks: x is incompetent)

(19) (a) Nobody thinks he will ever have enough money.

(b) *Bill ever thinks nobody will have enough money.

(20) (a) you tell me that she likes who

who do you tell me that she likes?

(b) you tell who that she likes Mary

*you tell that she likes Mary who

Such asymmetries were noted very early in modern syntax theory. After several unsuccessful attempts involving crucial reference to precedence, Culicover (1976) and Reinhart (1976) discovered the relevant purely structural factor at more or less the same time. The term c-command goes back to Reinhart (1976). We have reformulated it here as in (21), in a way that fits easily into present-day assumptions.

(21) a c-commands b if b is part of the structural

sister of a

Recall that argument hierarchies are mapped, e.g., on to hierarchical abstract deep-structures, which are then transformed to surface structures by further operations. For (22), one might postulate a surface structure such as (23) in the bracket notation, and (24) in the tree notation.

(22) the dangerous tiger has killed the old buffalo

(23) [the [[dangerous] [tiger]]] [has [[killed]

[the [[old] [buffalo]]]]]]

the dangerous tiger has killed the old buffalo

We have labeled the nodes in the tree (24) for ease of reference. We say that A dominates B iff there is a uniquely downwards directed path in the tree from A to B. Thus, s dominates x, but neither y nor z dominate w. A immediately dominates B iff A dominates B and there is no C that dominates B but not A. A is a sister of B if there is a C such that C immediately dominates A and B. Therefore, (21) implies that z c-commands w but not vice versa.

The following generalizations then hold for natural languages:

(25) A noun phrase A can bind a pronoun B only if

A c-commands B.

(26) A negating expression A licenses a polarity

element B only if A c-commands B.

(27) An element can move from position A to

position B only if A c-commands B.

Let us exemplify the impact of these statements with (25) only. The examples in (18) translate into the simplified partial structural representations (28). It is easy to figure out that no one c-commands he in (28a), but not in (28b), so that (25) correctly predicts that no one can bind the pronoun in (18a) only.

(28) (a) [no one [thinks [that [he [is [incompetent]]]]]]

(b) [he [thinks [that [no one [is [incompetent]]]]]]

Reference to the structural notion c-command in (25) cannot be replaced by precedence, because such an account would fail to explain why nobody can neither bind the pronoun nor license the polarity element in (29).

(29) *a letter to nobody would ever please him Presumably, the establishment of syntactic relations between A and B always presupposes that A c-commands B, or vice versa.

Kayne (1994) has proposed that elements α that asymmetrically c-command β precede β in all natural languages. If his approach is correct, we would understand why the basic orders subject–verb–object (SVO, English), SOV (Japanese), and VSO (Irish) predominate among the world’s languages; subjects occupy a hierarchically higher position in sentence structure than objects, and should therefore precede them if Kayne is right.

7. Operations Affecting Grammatical Relations

That the formation of questions, relative clauses, topicalization, etc. involves a preposing operation in certain languages such as English seems to be accepted quite generally, There may even be psycholinguistic evidence suggesting that such movement dependencies are processed in a special way by the human parser (see, e.g., Frazier and Clifton 1993).

In addition to such dependencies, in which phrases are displaced in order to encode semantic or pragmatic functions (so-called A-bar-movement), there are grammatical operations such as ‘passive’ (as in (30)) or ‘raising to subject’ (as in (31)) which primarily affect the grammatical functions of arguments: in a thematic sense, a moose in (30c) has properties of an object (it expresses the patient role); in other respects, it behaves like a subject (in terms of position and agreement with the verb).

(30) (a) one shot a moose

(b) there was a moose shot

(c) a moose was shot

(31) (a) it seems that he likes Mary

(b) he seems to like Mary

That such operations affecting grammatical functions should be expressed in terms of movement (see (32)) is less obvious (at least for languages in which grammatical functions are not straightforwardly en- coded positionally) and less widely accepted.

(32) d-structure

_was shot a moose

s-structure

a moose was shot_

Although the proper analysis of these processes is still subject to some debate, a number of locality restrictions on the process have, however, been un-covered, which are illustrated in (33).

(33) (a) I shoot a moose

a moose was shot

(b) I laugh at him

he was laughed at

(c) I expect her to win

she was expected to win

(d) I expect she will win

*she was expected will win

(e) I prefer very much for you to kiss Mary

*Mary was preferred very much for you to

kiss

The passive operation may map a direct object (33a), an object in a prepositional phrase (33b), and the subject of a nonfinite complement (33c) to the subject position, but fails to be able to establish such a relation for the subject of an embedded finite clause (33d), or any object in an embedded clause (33e).

The restrictions may be slightly different in other languages, as the contrast between English and German in (34) shows, but the crucial observation is that similar locality restrictions can be observed with the binding of reflexive pronouns, at least in the most restrictive subcase of this domain, as it is realized in English; see (35).

(34) (a) he forgot to repair the car

*the car was forgotten to repair

(b) er vergass den Wagen

he.nom forgot the.acc car

zu.repaieren

to repair

der Wagen wurde zu reparieren

the.nom car was to repair

vergessen

forgotten

(35) he shot himself

he laughs at himself

he expects himself to win

*he expects himself will win

*he prefers very much for her to kiss himself

The set of constructions to which these locality restrictions common to passivization and the grammar of reflexives apply is larger: it includes quantifier scope. A quantifier s can always take scope over a quantifier t if s c-commands t at s-structure. Thus, since subjects are higher in the structural hierarchy than objects, sentence (36a) allows a reading in which the existential quantifier takes scope over the universal quantifier

(36) (a) some man loves every woman

(b) there is a man such that this man loves every woman

(c) for every woman there is a man (possibly a different one for each woman) such that that man loves the woman

In English, however, a structurally lower quantifier s in position S may additionally take scope over t in position T if S and T could interact in terms of, say, passive formation, or the binding of reflexive pronouns. Thus, (36a) allows the further interpretation indicated in (36c), in which the object quantifier takes scope over the subject quantifier. A comparison of (33), (35), and(37) exemplifies the parallelism between wide scope options for a hierarchically lower quantifier, and the possibility of passive formation binding of reflexive pronouns. See also Hornstein (1995).

(37) some woman laughs at every man

(wide scope for every man possible)

some woman expects every man to be a loser

(wide scope for every man possible)

some woman expects every man will be a loser

(wide scope for every man impossible)

some woman prefers very much for me to kiss every man

(wide scope for every man impossible)

If formal syntax is responsible for the restrictions exemplified in (33), the binding rules for reflexives and the scope taking behavior of quantifiers must be reducible to syntactic operations too. However, it is hard to exclude the possibility that some extrasyntactic constraints governing quantifier scope or reflexive binding have side effects on passivization or raising, so that (33) would find a nonsyntactic account. Note, furthermore, that the impression that the same factor constrains the three different types of process discussed here may be spurious. Thus, in many languages (e.g., Swedish) the domain for the binding of reflexive pronouns is much larger than the domain for passive formation.

8. A Note Of Autonomy

As a reaction to so-called Generative Semantics, the hypothesis of the autonomy of syntax governed much of the history of current syntactic theorizing (see Newmeyer 1980): syntax was considered to be the only ‘generative’ component, the results of which are merely interpreted by semantics and by phonology. The preceding sections have revealed that it is often difficult to draw a sharp boundary between syntax and semantics, and strong adherence to the autonomy hypothesis may nowadays reflect a rather parochial perspective.

The same may be true for the interaction of phonology and syntax. The laws of the placement of clitics (e.g., unstressed pronouns) as we find them in Romance and Slavic languages often make crucial reference to phonological conditions (no clitic element must appear in the initial position of an intonation phrase) and syntactic conditions (elements bearing a given case etc. must appear in the leftmost position of, say, the verbal projection) at the same time. Phonological conditions may even be responsible for the application of syntactic operations such as verb movement.

The idea of a completely autonomous and encapsulated syntactic component, in which all processes can be understood in purely formal syntax-based terms only, in which syntactic processes are driven by syntactic conditions only, which was most popular in the GB tradition, therefore has to be modified, at least.

9. Approaches To Universals

The discovery of syntactic universals and their explanation has always been the major goal of syntactic theory. The typological tradition tries to attain this goal by comparing a large and representative sample of languages with respect to specific grammatical properties. Often this presupposes that the constructions can only be looked at in less detail than in the generative tradition. In the generative approach, the idea is to uncover abstract universal laws by an in-depth analysis of a few languages only.

The GB approach (Chomsky 1981) tried to identify a small set of simple yet universally valid syntactic principles, defining a Universal Grammar. It was assumed, especially in later stages of the theory, that formal differences between languages reduce to differences in the lexical inventory. The quest for such simple yet universally valid syntactic generalization was not successful, however, so the GB framework was replaced by, essentially, two new models. The Minimalist Program (Chomsky 1995) tries to understand syntactic properties in terms of so-called interface conditions (phonology, semantics). It may have explanatory value, but its empirical coverage is sometimes limited. Optimality Theory (OT) (Prince and Smolensky 1993, Grimshaw 1997) assumes that there are universally valid and simple conditions. Phonology, syntax, and semantics are not necessarily isolated subcomponents. OT acknowledges that simple principles will be in conflict with each. Such conflicts are resolved by ordering the principles hierarchically. Hierarchies are different in various languages, the only source of grammatical differences.

In particular, the major principles assumed in OT may correspond to ‘ideal types’ that languages may come close to in different degrees (as a function of the ranking of the relevant principles), so OT may have moved the generative enterprise closer to typology. This may be true in yet another sense; in contrast to GB, OT declared explicitly itself as a theory of variation and its explanation.

Bibliography:

1. Baker M 1996 The Polysynthesis Parameter. Oxford University Press, New York
2. Bossong G 1998 Le marquage de l’experient dans les langues d’Europe. In: Feuillet J (ed.) Actance et Valence dans les Langues de l’europe. Mouton de Gruyter, Berlin, pp. 259– 94
3. Bresnan J (ed.) 1982 The Mental Representation of Grammatical Relations. MIT Press, Cambridge, MA
4. Chomsky N 1957 Syntactic Structures. Mouton, The Hague, The Netherlands
5. Chomsky N 1965 Aspects of the Theory of Syntax. MIT Press, Cambridge, MA
6. Chomsky N 1973 Conditions on transformations. In: Anderson S, Kiparsky P (eds.) A Festschrift for Morris Halle. Holt, Rinehart and Winston, New York, pp. 232–86
7. Chomsky N 1981 Lectures on Government and Binding. Foris, Dordrecht, The Netherlands
8. Chomsky N 1986 Barriers. MIT Press, Cambridge, MA
9. Chomsky N 1995 The Minimalist Program. MIT Press, Cambridge, MA
10. Culicover P 1976 A constraint on coreferentiality. Foundations of Language 14: 109–18
11. Dowty D 1991 Thematic protoroles and argument selection. Language 67: 574–619
12. Erteshik-Shir N 1977 On the Nature of Island Constraints. IULC, Bloomington, IN
13. Fillmore C 1968 The case for case. In: Bach E, Harms R (eds.) Universals in Linguistic Theory. Holt, Rinehart and Winston, New York, pp. 1–88
14. Frazier K, Clifton C 1993 Construal. MIT Press, Cambridge, MA
15. Gazdar G, Klein E, Pullum G, Sag I 1985 Generalized Phrase Structure Grammar. Blackwell, London
16. Grimshaw J 1997 Projections, heads and optimality. Linguistic Inquiry 28: 373–422
17. Gruber J 1965 Studies in lexical relations. Diss., MIT, Cambridge
18. Haider H 1993 Deutsche Syntax—generati . Narr, Tubingen, Germany
19. Hale K 1983 Warlpiri and the grammar of non-configurational languages. Natural Language and Linguistic Theory 1: 5–47
20. Hjelmslev L 1972 [1935–7] La categorie des cas. Etude de grammaire generale. W Fink, Munich, Germany
21. Hornstein N 1995 Logical Form. Blackwell, Oxford, UK
22. Huang C T J 1982 Move WH in a language without Wh movement. The Linguistic Review 1(4): 369–416
23. Jackendoff R 1990 Semantic Structures. MIT Press, Cambridge, MA
24. Jakobson R 1971, [1936] Beitrag zur allgemeinen Kasuslehre. Gesamtbedeutungen der russischen Kasus. Selected Writings II. Mouton, The Hague, pp. 23–71
25. Kager R 1999 Optimality Theory. Cambridge University Press, Cambridge, UK
26. Kayne R 1994 The Antisymmetry of Syntax. MIT Press, Cambridge, MA
27. Marantz A 1984 On the Nature of Grammatical Relations. MIT Press, Cambridge, MA
28. Muller G 2000 Elemente der optimalitatstheoretischen Syntax. Stauffenburg, Tubingen, Germany
29. Newmeyer F 1980 Linguistic Theory in America. Academic Press, New York
30. Perlmutter D (ed.) 1983 Studies in Relational Grammar 1. University of Chicago Press, Chicago, IL
31. Prince A, Smolensky P 1993 Optimality theory Ms., Rutgers University, NJ
32. Reinhart T 1976 The syntactic domain of anaphora. Ph.D. diss., MIT, Cambridge, MA
33. Ross J R 1967 Constraints on variables in syntax. Ph.D. diss., MIT, Cambridge, MA
34. Rizzi L 1990 Relativized Minimality. MIT Press, Cambridge, MA
35. Szabolcsi A, Zwarts F 1997 Weak islands and an algebraic semantics for scope taking. In: Szabolcsi A (ed.) Ways of Scope Taking. Kluwer, Dordrecht, The Netherlands
36. Wunderlich D 1997 Cause and the structure of verbs. Linguistic Inquiry 28(1): 27–68