Clause union, the stratal uniqueness law and the chômeur relation

ConclusionSection 1 described this study as representing the convergence of two seemingly unrelated lines of research. One of these focused on the behavior of embedded 3s in Romance and other languages. The second concerned counterexamples to the ergative analysis provided by languages which did not fit the Romance pattern. A resolution of each of the problems in both these areas is possible through the Inheritance Principle and the recognition of the role of the SUL in the clause union construction. In languages with French-type clause union the failure of certain complement 3s to behave as 3s in the union clause has been shown to follow from their chômage, which is in turn predictable from the interaction of the Inheritance Principle and the SUL. The variation across languages in the clause union construction is restricted to the single parameter of mapping embedded 1s into matrix Objects. Once this mapping has been specified, the remaining properties of the clause union construction, including the distribution of chômeurs, are completely determined by the interaction of the Inheritance Principle and the SUL. Note that the collapsing view of clause union is crucial for our analysis, especially for the case in which an embedded nominal is a chômeur at the union stratum. Our results in this area could not be attained under a conception of the clause union construction in which it is biclausal at all levels, since under such an analysis the SUL would not be relevant.With respect to all GRs except the 1-relation our proposal is extremely restrictive. For example, in a language where no embedded 1s are union stratum 3s we claim that it is impossible for any embedded final 3s to be union stratum Chômeurs. Conversely, in a language in which some embedded transitive 1s are union stratum 3s, it is impossible for all embedded final 2s to be union stratum Chômeurs.Our system allows for considerable variation, however, in the mapping of embedded 1s into matrix Objects. We do not yet have sufficient empirical evidence bearing on the issue of limiting these possibilities to offer any proposal more restricted than our union parameter (21). Available evidence suggests that the number of attested realizations of (21) will be quite small. As an example, we know of no language in which all embedded 1s are union stratum 3s, an instantiation allowed by (21), though we do not have as yet any motivated principle for excluding this or other possibilities. However, it is clear what type of structure should occur if in some language the embedded 1 happened not to fall under some language-particular union rule. In that case, the embedded 1 would follow the dictates of the Inheritance Principle, clashing with the matrix 1. This conflict would then be resolved through the chômage of the embedded 1 in the clause union construction. We have not been able to discover any clear example of such a language (see Rosen, 1983).The question of how many patterns exist for the clause union construction is addressed in Comrie (1976), where the hierarchy of syntactic positions in (i) is claimed to be relevant for characterizing the causative construction. (i) subject - direct object - indirect object - other oblique constituentComrie suggests that in the clause union construction the embédded 1 is shifted to the right along this hierarchy to the first open position. In some respects, our view of clause union is closer to Comrie's than to the ergative analysis of Perlmutter and Postal. However, the precise nature of his claims is sufficiently difficult to ascertain to make comparison difficult.