Verbal Cues, Language, and Meaning in Selective Attention

Abstract

In 1953, Cherry did a series of experiments requiring a selective response to one of two verbal messages presented either binaurally or dichotically.1 While the irrelevant message caused negligible interference with the repeating back or 'shadowing' in the dichotic condition, it made selection extremely difficult when both messages were presented binaurally, at equal intensities and in the same voice. The Ss needed many attempts at the same message and pieced it together only gradually, making errors and transpositions which were influenced by the probability-structure of English. They did eventually, however, succeed in separating out one of the two messages when these were normal prose, but failed when presented with strings of cliches with transitional probabilities which were high within phrases but fell to zero between them. This performance seems qualitatively different from that in the dichotic condition, and may not involve selective attention at all. The Ss might have heard both messages and retrospectively put together what made sense (in fact they did much better if allowed to write), whereas in the dichotic condition they succeeded on the first trial and could report none of the verbal content of the irrelevant message, nor even that in one case it changed to a foreign language. Broadbent put forward a general theory of selective attention which accounts for the finding with dichotic presentation by postulating a filter which selects signals arriving on one of two or more input channels and rejects others before they are fully analyzed.2 The aim of the present experiment was to investigate selective attention when it depends solely on the identification of verbal or linguistic features, using a task where the messages are presented only once and require an immediate and continuous response. In more detail, it was hoped: