Referential redundancy and the integration of verbal information

Abstract

In two experiments, college students read pairs of messages describing an actor's portrayal of a particular emotion. They were then asked to identify the one of 24 target photographs about which each message-pair was written. When the respondents' selections were made immediately after reading the message-pair, accuracy was inversely related to the redundancy of the two passages. Performance on the nonredundant pairs deteriorated relatively rapidly, however, when an arithmetic task was interpolated between the receipt of the message-pair and the presentation of the referent array from which the target was to be selected; by contrast, the redundant pairs elicited a stable performance pattern that was essentially unaffected by the interpolated arithmetic task. In a third experiment, each subject served as a transmitter. He was shown a target-photograph together with a "given" description and was asked to "supplement" it with a second description to produce a message-pair that would enable a recipient to identify the proper target-photograph in a decoding task of the type used in Experiments I and II. Given a choice between two "additional" descriptions that produced identical hit-rates when presented singly, respondents generally selected the passage that was more redundant with the "given" description.