PET Evidence for an Amodal Verbal Working Memory System ERIC H. SCHUMACHER ,* ,1 ERICK LAUBER ,* ,2 EDWARD A WH,* JOHN JONIDES,* EDWARD E. SMITH,* AND ROBERT A. KOEPPE† *Department of Psychology and †Division of Nuclear Medicine, University of Michigan, Ann Arbor, Michigan 48109 Received October 2, 1995 Current models of verbal working memory assume that modality-specific representations are translated into phonological representations before entering the working memory system. We report an experiment that tests this assumption. Positron emission tomogra- phy measures were taken while subjects performed a verbal working memory task. Stimuli were presented either visually or aurally, and a visual or auditory search task, respectively, was used as a control. Re- sults revealed an almost complete overlap between the active memory areas regardless of input modality. These areas included dorsolateral frontal, Broca’s area, SMA, and premotor cortex in the left hemisphere; bilateral superior and posterior parietal cortices and anterior cingulate; and right cerebellum. These results correspond well with previous research and suggest that verbal working memory is modality independent and is mediated by a circuit involving frontal, parietal, and cerebellar mechanisms. r 1996 Academic Press, Inc. INTRODUCTION Current models of working memory postulate two processing components for verbal material: a phonologi- cal store and a phonological rehearsal mechanism (Baddeley, 1992). These mechanisms are assumed to process amodal mental representations (i.e., represen- tations that are independent of input modality). Before storage and rehearsal ensue, modality-specific represen- tations of verbal material are presumably translated into amodal phonological codes. Researchers have begun to use neuroimaging tech- niques to investigate the neural correlates of the verbal working memory system. The results from these stud- ies converge on a model that involves left inferior frontal gyrus (typically described as Broca’s area) and other speech-related areas for rehearsal, and superior and posterior parietal cortices for storage. Petrides et al. (1993) found significant bilateral acti- vation in mid-dorsolateral frontal, premotor, and poste- rior parietal cortices when activation related to a task in which subjects repeated the numbers from 1 to 10 in order was subtracted from activation related to a task in which subjects randomly generated aloud the num- bers from 1 to 10, without repetition. The latter task required subjects to hold in working memory the digits that they already said while new digits were generated. The former task did not require much use of working memory. The active areas are therefore presumably involved in the storage and maintenance of verbal material. Paulesu et al. (1993) reported evidence that frontal areas mediate the rehearsal of verbal material, whereas parietal areas mediate the storage of such material. Their experiment required subjects either to remember a set of English letters (memory task) or to make rhyming judgments about English letters (rhyming task). When the control task, in which subjects made nonverbal decisions about unfamiliar Korean letters, was subtracted from the memory task, significant activation remained in bilateral SMA, Broca’s area, and left posterior parietal cortex, among other areas. The posterior parietal activation was associated with the memory but not the rhyming task. This suggests that posterior parietal cortex mediates working memory storage. By contrast, Paulesu et al. (1993) found that activation in Broca’s area was associated with both the memory and the rhyming tasks, which suggests that this area mediates working memory rehearsal. Awh et al. (1996) showed directly that Broca’s area is part of the circuit for phonological rehearsal and that bilateral superior and posterior parietal cortices medi- ate working memory storage. Their experiment in- cluded one experimental and two control tasks. The experimental task was a continuous-performance, two- back task. Subjects saw a sequence of letters, one at a time, and responded positively if the current letter was identical to the one seen two previously in the sequence and negatively otherwise. This task requires subjects to hold at least two letters in working memory at all times and to keep track of the temporal order of the 1 To whom correspondence and reprint requests should be ad- dressed. E-mail: eschu@umich.edu. 2 Present address: Department of Psychology, University of Georgia. NEUROIMAGE 3, 79–88 (1996) ARTICLE NO. 0009 79 1053-8119/96 $18.00 Copyright r 1996 by Academic Press, Inc. All rights of reproduction in any form reserved.