The neurophysiology of working memory : functional mapping of the human brain with positron emission tomography

Working memory (WM) refers to the retention, or keeping on-line, of information over short periods of time. WM is thought to be important for a variety of cognitive functions, including problem solving, learning and reading. Previous studies in nonhuman primates have mainly implicated sustained neural activity in the prefrontal cortex as the neural correlate of WM. In the present series of experiments, positron emission tomography was used to measure regional cerebral blood flow, a marker of regional metabolic activity, in the brain of human subjects during performance of WM tasks.

In paper I, we found that WM for visual, auditory and somatosensory stimuli activated cortical areas with WM specific activity, independent of the sensory modality of the stimulus. These areas were located in the prefrontal, inferior parietal, cingulate and frontal opercular cortex. We hypothesized that activation of the same part of cortex by two different tasks would be associated with interference when these tasks are performed simultaneously. This hypothesis was supported by the results of a series of psychological experiments (paper II). In paper V, both single and dual-task performance of WM tasks was studied. We could show that no additional areas were activated during dual-task performance compared to single task performance. This also implies that there was no area with activation specifically related to divided attention or task-coordination requirements, which has previously been suggested.

In paper III, WM for abstract visual stimuli was studied in two delayed matching to sample tasks. In the second matching task the WM load was increased by introducing additional contingencies in the instruction. Both matching tasks activated prefrontal and inferior parietal cortex in the left hemisphere, and the activity in these areas was correlated. The cingulate cortex was activated in the same location as in paper I and V. This cingulate activity is likely due to the planning of a motor response in delayed response tasks such as WM tasks. The increasing WM load in the second matching task induced additional activation in the right prefrontal and inferior parietal cortex. Schizophrenic patients performing this task show impaired performance, and reduced frontal metabolism, compared to control subjects. In paper IV we studied encoding and retrieval from long-term memory, the results suggesting that WM may be necessary to perform tasks involving encoding into long-term memory.

In conclusion: a consistent finding in the WM tasks was the coactivation of dorsolateral prefrontal cortex and the middle part of the inferior parietal cortex. This activity was significantly higher during WM tasks than during control tasks involving selective attention. While the prefrontal activation was expected from studies in non human primates, the inferior parietal involvement in WM is a relatively new finding. This area probably has no functionally corresponding area in the non-human primate. Our results also show that two different WM tasks, with memoranda from different sensory modalities, activate overlapping parts of the cortex. This means that information processing during WM tasks is not entirely parallel, but depends on the activation of parts of the cortex which has no modality specificity. Furthermore, we have associated such overlap in activity between tasks with limitations of simultaneous information processing in the brain.