Irrelevant background speech disrupts immediate recall of visually presented items. Salamé and Baddeley (1982) found that increasing the phonological similarity between the irrelevant speech and the visual items greatly increased this disruption. In contrast, Jones and Macken (1995) found little evidence for such an increase. The present experiments directly manipulated the phonological similarity of the irrelevant speech background and the to-be-remembered visual items. Experiments 1-4 compared background speech that shared virtually no phonemes with the visual stimuli with background speech that shared all of the phonemes of the visual stimuli. No effect of phonological similarity was found. Experiment 5 replicated the method of Salamé and Baddeley's critical experiment but not their results. With regard to the two primary explanations of the irrelevant speech effect, these data present a strong challenge to the phonological store hypothesis while offering some support to the changing state hypothesis.