Four experiments were conducted on the interrelationship between semantic and syntactic structures. The former were specified by the method of restricted associations, and the latter by traditional grammatical analysis. In all experiments subjects were required to generate sentences from sets of given words. The generated sentences were studied as a function of (1) number of words given; (2) types of relations (free associations, restricted associations, or random words); (3) types of restrictions (Superordinates, Coordinates, Subordinates, Similars, Locations, Parts, Foregoing Words, or Following Words); (4) directions of relations (forward or backward); (5) strength of associations; (6) frequency of stimuli; (7) information given; and (8) instructional sets (creative, common, or control sentences). With the exception of the stimulus frequency, all independent variables had differential effects upon the sentences. In general, sentences were longer and given words were less often used "correctly" for large sets of given words, random pairs, Similars, backward pairs, low strength associates, known relationships, and the creative condition. Numerous interactions also were significant. (Author/KB)