Diachronic syntax – free relatives in Romance

Abstract

Maria Luisa Rivero (1984) puts forward certain hypotheses about linguistic change, with illustrations from Spanish, and raises interesting issues. Here I shall concentrate on just a few. Making use of Henning Andersen's (1973) 'abduction' theory of grammatical change, Rivero develops ideas she has previously advanced (e.g. Hirschbiihler & Rivero, I98I, I982, I983) about the contribution the concept of 'core grammar' can make to diachronic linguistics. To put it simply, she postulates that the child learning his first language would require to have access to positive evidence to 'abduct' marked or auxiliary rules, but 'negative' evidence (i.e. absence of counter-examples) would be sufficient for the construction of unmarked ('core', 'universal') rules of grammar. Thus, if linguistic changes over time can be shown to affect only parameters of variation, and to move in the direction of core grammar, then support would be lent to the notion of core grammar, and the general theory in which it is embedded. Perhaps the first thing to establish in testing Rivero's hypothesis is precisely what changes have taken place over time: this is not always as easy as it seems because linguistic documentation for past periods is often unreliable, and comparable evidence for such aspects as register and dialect is not always available at widely different periods. Rivero discusses changes in the structure of' free' (headless, independent) relatives in Spanish between the I 3th century and the i6th century. I shall return to discussion of her data' after a brief survey of problems that arise in the analysis of free relatives in Romance. I shall consider, for brevity's sake, only NP-relatives. Rivero's analysis suggests that in Modern Spanish, as in Catalan and French, free relatives are similar to NP-headed relatives (with the WH-relative in COMP), in contrast to a more traditional analysis which sees the WH-relative as substantival (the 'HEAD' analysis). Whether there is WHmovement in free relatives (cf. Bresnan & Grimshaw, 1978) is not an issue