1 We are indebted to Greg Carlson for his comments on an earlier version of this paper. His suggestions have led to numerous substantial improvements. Naturally, we are responsible for any remaining deficiencies. To appear in Yearbook of Language Variation, Vol. 2, ed. Johan Rooryk and Pierre Pica, John Benjamins, Amsterdam. MARKEDNESS, ANTISYMMETRY AND COMPLEXITY OF CONSTRUCTIONS 1 PETER W. CULICOVER AND ANDRZEJ NOWAK The Ohio State University and University of Warsaw 1. Introduction One of the strongest arguments for the thesis that the human mind possesses a Universal Grammar (UG) with specific grammatical properties is that languages do not appear to have arbitrary and uncorrelated properties. What we find, rather, is that the properties of languages cluster, and that there are asymmetries among the logical possibilities. For example, VSO languages are always prepositional, and SOV languages are usually postpositional (Greenberg 1963:78-79). There are languages that express wh-questions using leftward movement to a peripheral position in the clause, and there are languages that express wh-questions without overt movement. But there do not appear to be languages that express wh-questions using rightward movement to a peripheral position in the clause. It is natural, given observations such as these, to posit that they are direct reflections of UG, which the language learner draws upon in choosing or constructing grammars. However, there are two other possibilities that have to be ruled out before such a conclusion can be drawn. First, the clustering of properties and the absence of certain logical possibilities may be due to social forces. In such a case we would not expect to find the same asymmetries in different parts of the world where languages are not genetically related or in contact. Second, these asymmetries may be due to the interaction between the grammatical or processing complexity of certain constructions and social forces. On this view, all of the logical possibilities are linguistic possibilities, but those that are more complex tend to lose out over time to their less complex competitors as linguistic knowledge is transmitted from generation to generation in a network of social interactions. The intention of this paper is to explore and make somewhat more precise