Effects of relation strength and semantic overlap on retrieval and comparison processes during sentence verification

Three experiments were performed to distinguish two theoretical accounts of relatedness effects in the sentence verification task. There were three important results: First, subjects were generally faster to make both positive and negative responses when the subject and predicate words of the "sentences" were strongly related than when they were weakly related. This was the case as long as the words in the subject-predicate pairs were meaningfully related and presented in a meaningful context. Second, subjects were faster to respond to unrelated false items than to related false items. Finally, relatedness effects for meaningfully related items were reduced when the items were presented in a context of unrelated items; the relatedness effect on negative RT in one condition was actually reversed when unrelated items were added to the stimuli. The results suggest an integration of the findings of many previous investigations: Two kinds of relatedness manipulations must be distinguished in previous studies employing the sentence verification task. Studies involving a comparison of unrelated items with meaningfully related false items have found results consistent with a feature comparison processing strategy (e.g., E. E. Smith, E. J. Shoben, & L. J. Rips, Psychological Review, 1974, 81, 214-241); studies comparing only meaningfully related items differing in relatedness have produced results consistent with retrieval models (A. M. Collins & M. R. Quillian, Journal of Verbal Learning and Verbal Behavior, 1969, 8, 240-247). It is proposed that subjects base their responses on the retrieval of different types of information in these two situations.