Broadbent (1983) has suggested that the influence of unattended speech on immediate serial recall is a perceptual phenomenon rather than a memory phenomenon. In order to test this, subjects were required to classify visually presented pairs of consonants on the basis of either case or rhyme. They were tested both in silence and against a background of continuous spoken Arabic presented at 75 dB(A). No effect of unattended speech was observed on either the speed or accuracy of processing. A further study required subjects to decide whether visually presented nonwords were homophonous with real words. Again, performance was not impaired by unattended speech, although a clear effect was observed on an immediate serial memory task. Our results give no support to the perceptual interpretation of the unattended speech effect.

Immediate memory for visually presented sequences of items is impaired when presentation is accompanied by irrelevant speech sounds. The effect is not dependent on the meaning of the unattended material because disruption occurs with material in a language unknown to the subject (Colle & Welsh, 1976; Salamé & Baddeley, 1983) and with nonsense material (Salamé & Baddeley, 1982). The effect does not appear to be sensitive to the intensity of the unattended speech (Colle, 1980; Salamé & Baddeley, 1983) but is greater when the unattended material comprises phonemes similar to the items remembered (Salamé & Baddeley, 1982). The latter study also showed that requiring subjects to suppress articulation abolishes the effect.

Salamé and Baddeley (1982) explained the unattended speech effect in terms of the articulatory loop component of working memory (Baddeley & Hitch, 1974). They suggested that the articulatory loop comprises two components, a phonological store and an articulatory control process. All spoken items are assumed to have obligatory access to the phonological store, whereas visually presented items may be registered by the control process of articulatory subvocalization. Items within the store are assumed to decay over a period of 1–2 s unless rehearsed, with rehearsal again relying on...