Evaluating models of working memory: FMRI and behavioral evidence on the effects of concurrent irrelevant information

FMRI and behavioral methods were used to examine working memory impairments resulting from articulatory suppression, irrelevant speech, and irrelevant nonspeech. While the deleterious effects of these three irrelevant information types are well established in the behavioral literature, theoretical models provide conflicting accounts of the origins of these effects. To adjudicate between these accounts, two experiments were conducted. Experiment 1 examined fMRI signal changes in a delayed probed recall task with articulatory suppression, irrelevant speech, or irrelevant nonspeech imposed during the encoding and delay periods. Within the principally frontal and left-lateralized network of brain regions engaged by the task, articulatory suppression caused a relative increase in activity early in the trial, while both irrelevant speech and nonspeech conditions caused relative reductions in regional activity later in the trial. In a subsequent behavioral experiment (Experiment 2), the specific timing of interference was manipulated to further explore apparent differences in the temporal specificity of the effects. Subjects performed a delayed serial recall task while irrelevant information was imposed during specific trial stages: encoding, delay, or recall. Articulatory suppression was found to be most effectual when it coincided with item encoding, while both irrelevant speech and irrelevant nonspeech were most effectual when presented during the post-presentation delay. Taken together, these experiments provide convergent evidence for a dissociation of articulatory suppression from the two irrelevant sound conditions, but suggest that the effects of irrelevant speech and irrelevant nonspeech are functionally equivalent. This pattern of dissociation is predicted by the Embedded-Processes model (Cowan, 1995), but proves challenging to explain in the context of alternative theories.