Configural properties in sentence memory

Two theories of human memory, Gestalt theory and assoclationism, are contrasted with respect to their predictions about the cued recall of sentences. The Gestalt theory, with its assumption of emergent propemes, predicts that test probes which maintain the configural properties of the studied sentences should be superior to probes that do not. The associative theory, with its assumption of independent associatlons, is shown to yield just the opposite prediction. The associative theory is confirmed in Exps. 1 and 2, but Gestalt theory is supported in Exps. 3 and 4 where the Ss were required to generate continuations to the sentence as part of the study task. What is the structure of human memory? Historically, psychology has known two major theories on this subject which stand in profound opposition to one another. The first is assoclationism, dating back to Aristotle, which attempts to reduce memory to a set of base elements called ideas, words, sense data, or, more recently, memory nodes. Knowledge is encoded in the form of connections among the base elements. The structure of these connections is very simple and mechanistic; there are bonds or associations linking pairs of elements. In recent times, the naked associations have been clothed with labels for semantic relations (e.g., Quillian, 1969; Anderson, 1972; Rumelhart, Lindsay, & Norman, 1972), but even so, one ts left with a rather simplistic structure for human memory. The second theory is Gestalt or organizational psychology which proposes a strikingly different conception of human memory. Compared to associationism, it is a relative newcomer to the scene, receiving its first systematic formulation in Koffka's (1935) classic book. The Gestalters completely abandoned the associatlonists' attempt to