Abstract
One of the most intriguing questions science has recently embarked on is how our capacity for language is embedded in our cognitive system. To fruitfully approach this question and bridge the gap between linguistics and cognitive science in general it is crucial to know how ?special? language is. It is this question this project will address, and more specifically the status of language universals. As should be uncontroversial, language acquisition is impossible without innate constraints on what is a possible language. However, it is an empirical question whether such constraints are specific to language, or reflect general properties of our genetic endowment. This question became urgent once more in recent discussions of the evolutionary origin of language. Constraints on evolutionary processes require language to have arisen in relatively small steps. Small, even if one of these steps - recursion - should be considered the ?clincher? (see Hauser, Chomsky, Fitch 2002). This makes it necessary to reassess language universals that have been proposed. This project investigates universals capturing cross-linguistic similaries in conditions on anaphoric relations. All languages have elements that for their interpetation may or must depend on another element (their antecedent). Anaphors are elements that must depend on an antecedent. English (1) exemplifies a condition on the domain in which a reflexive such as himself must find an antecedent. (1) Bill expected [George/*Cindy to admire himself] |__________________X_________________| Himself needs an antecedent, but is not allowed to skip the nearest subject, even if not skipping it makes the sentence ill-formed, as when it is realized as Cindy. This condition is known as the ?specified subject condition? (SSC). The pattern it captures is cross-linguistically pervasive. Chomsky (1980) proposes that the SSC is innate. However it is hard to conceive how such a principle could be implemented in our ?wet-ware?. It is also unclear how it could be explained from an evolutionary perspective. The SSC and all linguistic principles of this type represent so-called ?macro-universals?. For a true understanding, such macro-universals have to be decomposed into more elementary principles, based on properties of operations that can conceivably be implemented and evolve. This relates to another issue. As knowledge of cross-linguistic variation increased it was found that macro-universals such as the SSC, or further conditions that were proposed, face a range of exceptions. They are in a sense too bad to be true; yet they are very good approximations, which makes them too good to be false. The expectation is, that decomposing these macro-universals serves two purposes: i) explaining the exceptions as consequences of the way elementary rules and principles interact; ii) bridging the gap with the cognitive sciences. If one investigates the computational system of natural language grammar one may expect the following sources of universals: " Type 1: Necessary properties of computations, modulo a medium in which they take place " Type 2: Economy of computation, modulo resource types and restrictions " Type 3: General properties of computations specific to language An important subclass of anaphoric dependencies involves conditions on reflexivization; in cases where a subject is the antedent of an object (as in George admires himself) the predicate is called reflexive. In many languages reflexivization requires a special marker (self as in himself, a bodypart as in his body, his spirit, a special marker on the verb, etc.). The question is why this is so. Reuland (2005b) shows that this pattern is a good candidate for a type 1 universal, and the SSC for a type 2 universal. This project will conduct a systematic cross-linguistic investigation of reflexivization strategies. In particular it will: * Investigate the prevalence of reflexivization in which one argument binds another without any special marker, and to the extent in which this type of reflexivization is found, determine its typological correlates. (i) - Languages license reflexivity in at least two ways: (a) a special operation on the (argument structure of) the verb; (b) a special (protective) marker on the argument, - Arrive at a comprehensive cross-linguistic inventory of morphemes signaling an operation on the verb and an analysis of the types of contributions they make (ii) - Arrive at a comprehensive cross-linguistic inventory of morphemes (self, bodypart) protecting an argument in reflexives and an analysis of their possible types. (iii) * In many, but not all languages the element licensing reflexivity also enforces it, as in as in Bill expected George to admire himself [himselfBill]: - Arrive at a comprehensive cross-linguistic inventory of morphemes as found in Malayalam or Peranakan Javanese licensing reflexivity without enforcing it, including an analysis of their possible types. (iv) - Arrive at a comprehensive cross-linguistic inventory of semantic side effects of reflexive markers and an analysis of possible correlations between such side effects and the obligation to enforce reflexivity. (v) - Arrive at an understanding of the mechanisms underlying enforcing reflexivity. (vi)