Attention includes processes that evaluate stimuli relevance, select the most relevant stimulus against less relevant stimuli, and bias choice behavior toward the selected information. It is not clear how these processes interact. Here, we captured these processes in a reinforcement learning framework applied to a feature-based attention task that required macaques to learn and update the value of stimulus features while ignoring nonrelevant sensory features, locations, and action plans. We found that value-based reinforcement learning mechanisms could account for feature-based attentional selection and choice behavior but required a value-independent stickiness selection process to explain selection errors while at asymptotic behavior. By comparing different reinforcement learning schemes, we found that trial-by-trial selections were best predicted by a model that only represents expected values for the task-relevant feature dimension, with nonrelevant stimulus features and action plans having only a marginal influence on covert selections. These findings show that attentional control subprocesses can be described by (1) the reinforcement learning of feature values within a restricted feature space that excludes irrelevant feature dimensions, (2) a stochastic selection process on feature-specific value representations, and (3) value-independent stickiness toward previous feature selections akin to perseveration in the motor domain. We speculate that these three mechanisms are implemented by distinct but interacting brain circuits and that the proposed formal account of feature-based stimulus selection will be important to understand how attentional subprocesses are implemented in primate brain networks.

The abilities of switching between and maintaining task rules are fundamental aspects of goal-oriented behavior. The PFC is thought to implement the cognitive processes underling such rule-based behavior, but the specific contributions of the several cytoarchitecturally distinct subfields of PFC remain poorly understood. Here, we used bilateral cryogenic deactivation to investigate the relative contributions of two regions of the dorsolateral PFC (dlPFC)—the inferior dlPFC (idlPFC) area, consisting of the cortex lining the caudal principal sulcus, and the dorsally adjacent superior dlPFC (sdlPFC)—to different aspects of rule-based behavior. Macaque monkeys performed two variants of a task that required them to alternate unpredictably between eye movements toward (prosaccade) or away from (antisaccade) a visual stimulus. In one version of the task, the current rule was overtly cued. In the second, the task rule was uncued, and successful performance required the animals to detect rule changes on the basis of reward outcome and subsequently maintain the current task rule within working memory. Deactivation of the idlPFC impaired the monkeys' ability to perform pro- and antisaccades in the uncued task only. In contrast, deactivation of the sdlPFC had no effect on performance in either task. Combined deactivation of idlPFC and sdlPFC impaired performance on antisaccade, but not prosaccade, trials in both task variants. These results suggest that the idlPFC is required for mnemonic processes involved in maintenance of task rules, whereas both idlPFC and sdlPFC together are necessary for the deployment of the cognitive control required to perform antisaccades. Together, these data support the concept of a functional specialization of subregions within the dlPFC for rule-guided behavior.

Failures in monitoring of self-generated actions are thought to underlie the positive symptoms in schizophrenia. It has been hypothesized that these deficits may be caused by a dysfunction of N -methyl- d -aspartate receptors (NMDARs). Here we recorded the activity of prefrontal neurons in monkeys performing an antisaccade task, while we administered a subanesthetic dose of the noncompetitive NMDAR antagonist ketamine. Many neurons discriminated between correct antisaccades and response errors in their postresponse activity. Ketamine increased the activity for the neurons' nonpreferred response, thereby decreasing the neurons' performance selectivity. Ketamine also affected the monkeys' behavior after an error, consistent with a deficit in error detection. The results show that NMDARs play an important role in action monitoring in primates. The decrease in performance selectivity of prefrontal neurons after ketamine can help to explain the deficits in action monitoring found in humans after ketamine administration and provides support for the hypothesis that an NMDAR dysfunction underlies self-monitoring deficits and psychotic symptoms in schizophrenia.

Numerous studies have established a role for the ACC in cognitive control. Current theories are at odds as to whether ACC itself directly engages or alternatively recruits other frontal cortical areas that implement control. The antisaccade task, in which subjects are required to make a saccade to the location opposite a suddenly appearing visual stimulus, is a simple oculomotor paradigm that has been used extensively to investigate flexible oculomotor control. Here, we tested a causal role of the dorsal ACC in cognitive control by applying electrical microstimulation during a preparatory period while monkeys performed alternating blocks of pro- and antisaccade trials. Microstimulation induced significant changes in saccadic RTs (SRTs) in both tasks. On prosaccade trials, SRTs were increased for saccades contralateral to and decreased for saccades ipsilateral to the stimulated hemisphere. In contrast, SRTs were decreased for both ipsi- and contralaterally directed antisaccades. These data show that microstimulation administered during response preparation facilitated the performance of antisaccades and are suggestive of a direct role of ACC in the implementation of cognitive control.

Visuospatial working memory is one of the most extensively investigated functions of the dorsolateral prefrontal cortex (DLPFC). Theories of prefrontal cortical function have suggested that this area exerts cognitive control by modulating the activity of structures to which it is connected. Here, we used the oculomotor system as a model in which to characterize the output signals sent from the DLPFC to a target structure during a classical spatial working memory task. We recorded the activity of identified DLPFC–superior colliculus (SC) projection neurons while monkeys performed a memory-guided saccade task in which they were required to generate saccades toward remembered stimulus locations. DLPFC neurons sent signals related to all aspects of the task to the SC, some of which were spatially tuned. These data provide the first direct evidence that the DLPFC sends task-relevant information to the SC during a spatial working memory task, and further support a role for the DLPFC in the direct modulation of other brain areas.

Complex behavior often requires the formation of associations between environmental stimuli and motor responses appropriate to those stimuli. Moreover, the appropriate response to a given stimulus may vary depending on environmental context. Stimulus-response associations that are adaptive in one situation may not be in another. The prefrontal cortex (PFC) has been shown to be critical for stimulus-response mapping and the implementation of task context. To investigate the neural representation of sensory-motor associations and task context in the PFC, we recorded the activity of prefrontal neurons in two monkeys while they performed two tasks. The first task was a delayed-match-to-sample task in which monkeys were presented with a sample picture and rewarded for making a saccade to the test picture that matched the sample picture following a delay period. The second task was a conditional visuomotor task in which identical sample pictures were presented. In this task, animals were rewarded for performing either prosaccades or antisaccades following the delay period depending on sample picture identity. PFC neurons showed task selectivity, object selectivity, and combinations of task and object selectivity. These modulations of activity took the form of a reduction in stimulus and delay-related activity, and a pro/anti instruction-based grouping of delay activity in the conditional visuomotor task. These data show that activity in PFC neurons is modulated by experimental context, and that this activity represents the formal demands of the task currently being performed.

Everyday life typically requires behavior that involves far more than simple stimulus-response associations. Environmental cues are often ambiguous and require different actions depending on the situation. The prefrontal cortex (PFC) is thought to be crucial for this flexible control of behavior. An important task that probes this ability is the antisaccade task in which subjects have to suppress a glance towards a suddenly presented peripheral stimulus and instead look away from the stimulus to its mirror location. Here we recorded the activity of PFC neurons in monkeys trained to alternate between blocks of prosaccade and antisaccade trials with no external instruction cues. We found that the activity of many neurons was different between the two tasks during the fixation period before the peripheral stimulus was presented. These differences were already present on the first correct trials after a task switch. The activity of these neurons also discriminated between correct responses and errors. We hypothesize that the PFC provides bias signals to saccade-related areas that are necessary to preset the oculomotor system for different tasks.

The phenomenon of inhibition of return (IOR) has generated considerable interest in cognitive neuroscience because of its putative functional role in visual search, that of placing inhibitory tags on objects that have been recently inspected so as to direct further search to novel items. Many behavioral parameters of this phenomenon have been clearly delineated, and based on indirect but converging evidence, the widely held consensus is that the midbrain superior colliculus (SC) is involved in the generation of IOR. We had previously trained monkeys on a saccadic IOR task and showed that they displayed IOR in a manner similar to that observed in humans. Here we recorded the activity of single neurons in the superficial and intermediate layers of the SC while the monkeys performed this IOR task. We found that when the target was presented at a previously cued location, the stimulus-related response was attenuated and the magnitude of this response was correlated with subsequent saccadic reaction times. Surprisingly, this observed attenuation of activity during IOR was not caused by active inhibition of these neurons because (a) they were, in fact, more active following the presentation of the cue in their response field, and (b) when we repeated the same experiment while using the saccadic response time induced by electrical micro-stimulation of the SC to judge the level of excitability of the SC circuitry during the IOR task, we found faster saccades were elicited from the cued location. Our findings demonstrate that the primate SC participates in the expression of IOR; however, the SC is not the site of the inhibition. Instead, the reduced activity in the SC reflects a signal reduction that has taken place upstream.