Patients with damage to the dorsolateral prefrontal cortex are impaired on cognitive tasks such as the Wisconsin Card Sort Test, the Stroop Test and an anti-saccade paradigm, in which sensory-guided habitual responses must be suppressed in favour of conceptually or memory-guided responses. We report here recordings from prefrontal neurons in rhesus monkeys trained to perform a delayed anti-saccade task based on tests that have been used with humans. Activity in the same prefrontal neurons was recorded across conditions when saccades were made toward a remembered target, and also when this prepotent response was suppressed and a saccade in the opposite direction required. Our findings show that most prefrontal neurons code the location of the visual stimulus in working memory, and that this memory can be engaged to suppress as well as prescribe a response. These results establish, in a subset of prefrontal neurons, the iconic nature of the memory code, and suggest a role for visual memory in response suppression.

Single-unit recording studies of posterior parietal neurons have indicated a similarity of neuronal activation to that observed in the dorsolateral prefrontal cortex in relation to performance of delayed saccade tasks. A key issue addressed in the present study is whether the different classes of neuronal activity observed in these tasks are encountered more frequently in one or the other area or otherwise exhibit region-specific properties. The present study is the first to directly compare these patterns of neuronal activity by alternately recording from parietal area 7ip and prefrontal area 8a, under the identical behavioral conditions, within the same hemisphere of two monkeys performing an oculomotor delayed response task. The firing rate of 222 posterior parietal and 235 prefrontal neurons significantly changed during the cue, delay, and/or saccade periods of the task. Neuronal responses in the two areas could be distinguished only by subtle differences in their incidence and timing. Thus neurons responding to the cue appeared earliest and were more frequent among the task-related neurons within parietal cortex, whereas neurons exhibiting delay-period activity accounted for a larger proportion of task-related neurons in prefrontal cortex. Otherwise, the task-related neuronal activities were remarkably similar. Cue period activity in prefrontal and parietal cortex exhibited comparable spatial tuning and temporal duration characteristics, taking the form of phasic, tonic, or combined phasic/tonic excitation in both cortical populations. Neurons in both cortical areas exhibited sustained activity during the delay period with nearly identical spatial tuning. The various patterns of delay-period activity-tonic, increasing or decreasing, alone or in combination with greater activation during cue and/or saccade periods-likewise were distributed to both cortical areas. Finally, similarities in the two populations extended to the proportion and spatial tuning of presaccadic and postsaccadic neuronal activity occurring in relation to the memory-guided saccade. The present findings support and extend evidence for a faithful duplication of receptive field properties and virtually every other dimension of task-related activity observed when parietal and prefrontal cortex are recruited to a common task. This striking similarity attests to the principal that information shared by a prefrontal region and a sensory association area with which it is connected is domain specific and not subject to hierarchical elaboration, as is evident at earlier stages of visuospatial processing.

An important question in neuroscience is whether and how temporal patterns and fluctuations in neuronal spike trains contribute to information processing in the cortex. We have addressed this issue in the memory-related circuits of the prefrontal cortex by analyzing spike trains from a database of 229 neurons recorded in the dorsolateral prefrontal cortex of 4 macaque monkeys during the performance of an oculomotor delayed-response task. For each task epoch, we have estimated their power spectrum together with interspike interval histograms and autocorrelograms. We find that 1). the properties of most (about 60%) neurons approximated the characteristics of a Poisson process. For about 25% of cells, with characteristics typical of interneurons, the power spectrum showed a trough at low frequencies (<20 Hz) and the autocorrelogram a dip near zero time lag. About 15% of neurons had a peak at <20 Hz in the power spectrum, associated with the burstiness of the spike train; 2). a small but significant task dependency of spike-train temporal structure: delay responses to preferred locations were characterized not only by elevated firing, but also by suppressed power at low (<20 Hz) frequencies; and 3). the variability of interspike intervals is typically higher during the mnemonic delay period than during the fixation period, regardless of the remembered cue. The high irregularity of neural persistent activity during the delay period is likely to be a characteristic signature of recurrent prefrontal network dynamics underlying working memory.

Dorsolateral prefrontal and posterior parietal cortex share reciprocal projections. They also share nearly identical patterns of neuronal activation during performance of memory-guided saccades. To test the hypothesis that the reciprocal projections between parietal and prefrontal neurons may entrain their parallel activation, the present experiments have combined cortical cooling in one cortical area with single-unit recording in the other to more precisely determine the physiological interactions between the two during working memory performance. The activity of 105 cortical neurons during the performance of an oculomotor delayed response (ODR) task (43 parietal neurons during prefrontal cooling, 62 prefrontal neurons during parietal cooling) was compared across two blocks of trials collected while the distant cortical area either was maintained at normal body temperature or cooled. The mean firing rates of 71% of the prefrontal neurons during ODR performance changed significantly when parietal cortex was cooled. Prefrontal neurons the activity of which was modulated during the cue, delay, or saccade periods of the task were equally vulnerable to parietal inactivation. Further, both lower and higher firing rates relative to the precool period were seen with comparable frequency. Similar results were obtained from the converse experiment, in which the mean firing rates of 76% of the parietal neurons were significantly different while prefrontal cortex was cooled, specifically in those task epochs when the activity of each neuron was modulated during ODR performance. These effects again were seen equally in all epochs of the ODR task in the form of augmented or suppressed activity. Significant effects on the latency of neuronal activation during cue and saccade periods of the task were absent irrespective of the area cooled. Cooling was associated in some cases with a shift in the best direction of Gaussian tuning functions fit to neuronal activity, and these shifts were on average larger during parietal than prefrontal cooling. In view of the parallel between the similarity in activity patterns previously reported and the largely symmetrical cooling effects presently obtained, the data suggest that prefrontal and parietal neurons achieve matched activation during ODR performance through a symmetrical exchange of neuronal signals between them; in both cortical areas, neurons activated during the cue, delay, and also saccade epochs of the ODR task participate in reciprocal neurotransmission; and the output of each cortical area produces a mixture of excitatory and inhibitory drives within its target.

A key idea in Lashley's formulation of the problem of serial order in behavior is the postulated neural representation of all serial elements before the action begins. We studied this question by recording the activity of individual neurons simultaneously in small ensembles in prefrontal cortex while monkeys copied geometrical shapes shown on a screen. Monkeys drew the shapes as sequences of movement segments, and these segments were associated with distinct patterns of neuronal ensemble activity. Here we show that these patterns were present during the time preceding the actual drawing. The rank of the strength of representation of a segment in the neuronal population during this time, as assessed by discriminant analysis, predicted the serial position of the segment in the motor sequence. An analysis of errors in copying and their neural correlates supplied additional evidence for this code and provided a neural basis for Lashley's hypothesis that errors in motor sequences would be most likely to occur when executing elements that had prior representations of nearly equal strength.