Projekt A3:
Suboptimal Syntactic Structures

Some previous studies

Word order in the Mittelfeld

The question of the order of external arguments in the middle field of German sentences has been subject to a great deal of research and discussion from Lenerz (1977) and Uszkoreit (1986) onwards. This assumes that the phenomenon is one of competing constraints interacting to determine linear precedence (see discussion in Keller 2000 and others) and uses the magnitude estimation methodology to establish empirically which factors play a role in this ordering. This initial experiment tests only animacy (+animate, -animate), case (dat>acc, acc>dat), definiteness (definite, indefinite) and verb position (verb second, verb final). The results so far support the constraint competition model of argument order and show relatively strong effects for case order and definiteness, weaker effects for animacy, but no effect at all for verb position (in contradiction to certain claims in the literature). Further tests are planned in which the factors heaviness and context will be examined.

Superiority, that-trace and d-linking in German

Many German linguists have claimed that German differs from English in wh-movement constructions and that this provides evidence that the clause structure of German is different from the English. In particular, it has been repeated questioned whether German exhibits Superiority and d-linking effects as English does. We tested this empirically in this study using magnitude estimation to obtain more finely grained judgements from informants. The results show a clear superiority effect when a subject is in situ in a multiple wh-question, but no effects for other argument types. This result is consistent with accounts of the phenomenon which attribute it to a different position of a subject in a clause, perhaps in spec-IP. It is not consistent with theories which assume different clause structures in German and English.

Additionally this experiment tested the description of d-linking effects put forward by Pesetsky (1987), who argues that an in situ wh-item which would otherwise trigger a superiority violation does not do so if it is of the form "which NP", ie it is "discourse-linked". This effect has also been the subject of considerable controversy, particularly in German. Not surprisingly, authors who deny superiority will also question d-linking in German. We found robust reflection of this claim: the d-linked status of the in situ item annulls a potential superiority effect also in German just as in English.

In a related study we addressed the question of that-trace effects in German. These have frequently been denied in the literature, since many speakers find extraction from complement clauses with a complementizer very marked or even ungrammatical. Our results show a consistent and robust that-trace effect in spite of the marginal grammaticality of the structures.

In a rather more speculative study, we tested the ability of different structures to contain a wh-item in situ, The data shows a clear distinction between in situ wh-items in mono-clausal structures and biclausal structures. While an in situ wh-item in a single clause is readily acceptable, and only a little worse than fully grammatical controls, in situ wh-items in embedded clauses are much worse. They do show marginal acceptability, however, as long as operator movement to the sentence-initial operator position is occupied by another constituent. When this position is empty, however, and nothing blocks the operator from movement, the sentence worsens still further. This seems to show that any constituent can block operator movement, at least to some extent.

These results, taken together, provide clear support for a Universal Grammar and the modules generally included within it, since they show that the constraints assumed for English exist in German too, in spite of claims to the contrary, which makes them realistic candidates for universals. Our results also support the view that at least a proportion of grammatical constraints are not "hard" and inviolable as has traditionally been assumed, but rather "soft" and violable. This is consistent with constraint-ranking models of syntactic description.

Bridge verbs and V2 complements

Verbs which permit extraction from a complement clause (with dass) are often called "bridge" verbs. Perhaps the most basic question about them concerns the nature of the factor which makes one verb a good bridge verb and another a poor one. In this study we examine one of the most frequently advanced suggestions, namely that bridge verbs are exactly the same set of verbs as those which permit V2 complements. In these experiments we tested eight verbs which vary on the parameter of "bridgeness". The results show that the acceptability of a V2 complement and acceptabilty of extraction from a complement clause show considerable similarity, but do not precisely co-vary. We conclude that the set of verbs which permit V2 complements and those which permit extraction from a complement clause are not identical.

Binding in double object structures

Our most recent work concerns the binding possibilities in a double object structure. We tested whether a range of factors were concerned (antecedent type, anaphor type, case order, selbst insertion) and have developed an account of the results which derives the constraints which are experimentally confirmed from more general principles. We therefore account for the marginality of such structures in an economic and principled way. We intend to extend this work to other binding structures in the hope that our explanatory model will prove itself to be applicable to a wider range of data.

French wh-in-situ questions and syntactic optionality

A particular characteristic of French interrogative sentences consists in the number of possible word order variants. (1a, b) show, for example, two different ways to construct a sentence like "Where do you go ?".

(1a) Tu   vas où ?

        you go  where

(1b) Où     tu   vas ?

       where you go ?

In traditional linguistics, different word order variants in French interrogatives have often been described in variational terms. Most people learning French as a second language are, for example, taught in which situations to use variant x and in which situation to use variant y. Apparently, people can choose among different possibilities to express a question in French and they make their choice with regard to variational aspects, above all considerations about the appropriate register.

Assuming the movement to be triggered by pragmatic operations located at a syntax-pragmatics- interface, we integrate the variational description of traditional linguistics in our optionality analysis. Stylistic differences between different word order variants come into play at this level. This view is in line with the general assumption of Haider & Rosengren (1998) that optional movement is exploited at the interface level of syntax.

We hope that this approach accounts in a more satisfying manner for the phenomenology of word order variants of French wh-questions.

Distinction and syntactic optionality

Adli (2001b) considers stylistics and speech register as a sort of connecting link between grammar and social structure. In this work, the sociological dimension was operationalised by the concept lifestyle (Bourdieu, 1979) using a questionnaire with 68 items. Multivariate techniques of data reduction (factor and cluster analyses) led to three clusters of lifestyle. An analysis of variance shows their significant effect on the grammaticality judgments of those structures assumed to allow optional movement. Sociological factors are supposed to influence the choice between different optional structures in French.