Research Supports Distributed Model of Central Executive in Working Memory

APS sponsors the William James Distinguished Lecture at the meetings of regional psychological associations as part of an overall effort to work more closely with the regional groups. The following report summarizes the presentation by John Jonides at the Western Psychological Association’s 2001 meeting. Reports on APS William James Lectures by Sam Glucksberg, Princeton University, and Robert Cialdini, University of Arizona will appear in upcoming issues of the Observer.

Jonides

It isn’t easy competing with Paradise, but John Jonides managed to do it. Jonides delivered this year’s APS William James Lecture on “Behavior and Neuroimaging Studies of Executive Control” to a capacity crowd at the Western Psychological Association (WPA) annual meeting. Normally, this would be a no-brainer, so to speak. Jonides is a leader in the field of cognitive neuroscience and is widely recognized for his work on perception, attention, and memory. You would expect a big audience to hear about the work of such a distinguished researcher.

But this year’s WPA meeting was in Maui, HI. Given this enticing setting, it is a real testament to Jonides’ work that people put off their time in the sun to hear him. They were rewarded with a fascinating presentation of the latest behavioral and neuroscience perspectives on executive control processes in working memory. Among other things, Jonides’ work calls into question some popular models of working memory.

Jonides is at the University of Michigan where he is a professor of psychology and a professor of neuroscience. He also is the co-director of the university’s functional neuroimaging center. Following is a summary of his APS William James Lecture.

MODELS OF WORKING MEMORY

Two of the primary functions attributed to the central executive are shifting attention and management of multiple tasks. Jonides discussed two types of models describing the role of executive functioning in working memory: Unitary models assert that the focusing and multi- task management functions are two sides of the same coin and are mediated by the same mechanism. In contrast, the distributed models of executive processing posit two distinct mechanisms, one which switches attention from one focus to another and one to allocate attention across multiple tasks.

Through a series of inventive behavioral and imaging experiments, Jonides has found evidence supporting a distributed model of executive functioning. He based his research on the supposition that the unitary models predict interference between tasks requiring switches of attentional focus and tasks requiring the allocation of attention because they depend on the same mechanism. Conversely, because distributed models describe separate mechanisms for switching and allocating attention, they predict no interference between them.

Jonides devised a series of challenging tasks that required subjects to switch among mental mathematical operations, while simultaneously monitoring and updating several numeric values in memory. Here are the basic elements of the experiment:

• Subjects viewed a screen displaying several “counters.” Each counter was assigned an initial numeric value. As the experiment progressed, subjects were required to perform mathematical operations on these values and to mentally update the counters with the results.
• The new values of the counters did not appear on the screen; the subjects had to hold these values in their working memories. For example, if one counter was initially set to zero and the subject was cued to add 3, she would then need to remember that the counter’s value as 3. If she were then cued to multiply the counter by 2, she would need to update the counter’s value to 6.
• The number of counters on the screen and the number of mathematical operations required of subjects varied among experimental conditions. Cues appeared adjacent to the counters to indicate which operation was required. For example, if a subject saw an upward-pointing arrow above a counter, he or she might recognize this as the cue to add seven to that counter’s value. A downward-pointing arrow would call for a different operation.
• Subjects’ performance on these tasks was measured by reaction time; subjects pressed a switch each time they finished a mathematical operation and mentally updated the counter. Each trial could necessitate a change of mathematical operation (from the previous trial) or a change in which counter was operated upon. Some trials required only one change (either operation or counter), some required both and some required neither.

Reaction times were fastest when no changes were required and slowest when both were required, indicating that changing took more effort than not changing. Jonides hypothesized that the task of switching mathematical operations depends on the focusing functions of executive control and the monitoring of multiple counters depends on the multi-task management function of executive control.

By varying the difficulty of each of these tasks, Jonides was able to directly observe whether there are any inhibitory interactions between the tasks. In one version of the experiment, the number of counters varied between two and three, while the number of operations was held constant at two. Results showed that increasing the number of counters increased reaction time when subjects had to change counters, but had no effect on reaction time when subjects had to change operations.

In another version of the experiment, the number of counters was held steady at two but the number of operation varied between two and three. These results showed that increasing the number of operations increases reaction time when changing operations but does not affect reaction time when subjects changed counters. Taken together, these results show these tasks to be separate; making one more difficult does not affect performance on the other. These experiments appear to demonstrate a double dissociation between switching and allocation and support the distributed view of executive functioning.

COUNTER INTELLIGENCE

These results call into question models and experimental paradigms that treat the central executive as a single unified processor.