Recognition Failure of Prose-Embedded Words

In previous experiments on recognition failure of recallable words, the cue-target information has typically been confined to word pairs. The present series of 8 experiments was designed to transcend these confines. By using prose-embedded targets, it was possible to evaluate the effects of manipulations of semantic involvement during story acquisition and retrieval: (a) For immediate testing conditions, the probability of recognition given recall adhered to the Tulving-Wiseman (1975) function, (b) the number of trials for story acquisition, type of instructions for acquisition, response mode, as well as type of proposition embedding the target did not cause any systematic deviations from the Tulving-Wiseman function, (c) deviations above the Tulving-Wiseman function were not a replicable finding, whereas (d) deviations below were replicable for longer delays and for targets that were thematically related to the overall contents of the story, and (e) false recognitions given recall were biased to semantic lures in 93% of the failures made, across experiments. The discussion centers around the generalizability of the episodic retrieval independence model (Flexser & Tulving, 1978) and the invariance of the Tulving-Wiseman function despite the semantic bias in the failures made. To account for deviations two rival semantic, schema-based hypotheses were proposed: Jones's (1978, 1983) reconstructive hypothesis and the assimilation hypothesis (cf. Graesser, Woll, Kowalski, & Smith, 1980; Yekovich & Walker, 1986). It was concluded that the latter hypothesis gained most support.