Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2016 Nov 18.
Published in final edited form as: Neuron. 2015 Nov 8;88(4):832–844. doi: 10.1016/j.neuron.2015.10.001

A source for feature based attention in the prefrontal cortex

Narcisse P Bichot 1,*, Matthew T Heard 1, Ellen M DeGennaro 1, Robert Desimone 1
PMCID: PMC4655197  NIHMSID: NIHMS729299  PMID: 26526392

SUMMARY

In cluttered scenes, we can use feature-based attention to quickly locate a target object. To understand how feature attention is used to find and select objects for action, we focused on the ventral pre-arcuate (VPA) region of prefrontal cortex. In a visual search task, VPA cells responded selectively to search cues, maintained their feature selectivity throughout the delay and subsequent saccades, and discriminated the search target in their receptive fields with a timecourse earlier than in FEF or IT cortex. Inactivation of VPA impaired the animals’ ability to find targets, and simultaneous recordings in FEF revealed that the effects of feature attention were eliminated while leaving the effects of spatial attention in FEF intact. Altogether, the results suggest that VPA neurons compute the locations of objects with the features sought and send this information to FEF to guide eye movements to those relevant stimuli.

INTRODUCTION

In scanning a complex scene, we often know what we are looking for, but not necessarily where it is. The ability to quickly find an object based on a memory of its features is normally attributed to feature-based attention, which shares some properties with memory recall and visual imagery. For simplicity, we will not distinguish here between attention to features of an object versus attention to objects as configurations of multiple non-spatial features. The memory of the searched-for object has been described as the “attentional template” for search (Desimone and Duncan, 1995; Duncan and Humphreys, 1989; Wolfe et al., 1989). FEF, area LIP, and the superior colliculus have all been described as containing “priority maps”, in which responses to a stimulus in a given location in the retinotopic map are scaled according to the similarity of the stimulus to the searched-for target feature (Basso and Wurtz, 1998; Kusunoki et al., 2000; Thompson and Bichot, 2005). For example, if a monkey is searching for a yellow banana in a scene, the locations of all yellow stimuli in the priority maps might be signaled by enhanced neural activity. Cells in those areas respond as though they have received information about the similarity between the stimulus features in their receptive fields (RFs), and the features of the searched-for target, ultimately resulting in the selection of a single stimulus for a saccade target or further visual processing (Findlay and Walker, 1999; Hamker, 2005; Itti and Koch, 2001; Olshausen et al., 1993; Wolfe et al., 1989). However, cells in those structures show little or no selectivity for features such as yellow or activity related to the memory of these features. Thus, it seems unlikely that these areas compute the similarity between the features of the attentional template and the features of a stimulus. How is the match between the feature at a given location and those of the search object computed?

One possibility is that the match is computed in early visual areas, such as V4, where the responses of cells are feature selective and are also influenced by feature attention, i.e. the features of the target the animal is searching for (Chelazzi et al., 2001; Hayden and Gallant, 2005; Martinez-Trujillo and Treue, 2004; McAdams and Maunsell, 2000; Motter, 1994). In particular, we have previously shown that, during free-viewing visual search, the responses of V4 neurons are maximally enhanced when there is a preferred feature in their RF, and that feature matches some or all of the target features, independently of the locus of spatial attention (Bichot et al., 2005; Zhou and Desimone, 2011), as predicted by parallel search models (Desimone and Duncan, 1995; Wolfe et al., 1989).

However, recent studies with paired recordings in FEF and V4 have shown that the onset of feature-based selection in a free-viewing visual search task (Zhou and Desimone, 2011) occurs earlier in FEF than in V4, and the same relative timing difference has been found in a color-cueing spatial attention task (Gregoriou et al., 2009). If the effects of feature and spatial attention occur later in V4 than in FEF, it seems very unlikely that V4 is the source of the selection signals observed in FEF.

Instead, parts of prefrontal cortex (PFC) outside of FEF seem more likely to be a major source of computations for feature-based object selection. PFC has traditionally been associated with executive control (for review, see (Miller and Cohen, 2001)), and working memory for locations and objects (Everling et al., 2006; Funahashi et al., 1989; Fuster and Alexander, 1971; Mendoza-Halliday et al., 2014; Miller et al., 1996; Rainer et al., 1998; Rao et al., 1997). Human imaging studies show that parts of PFC are active during both spatial and feature attention (Bressler et al., 2008; Egner et al., 2008; Gazzaley and Nobre, 2012; Giesbrecht et al., 2003), and a recent human MEG and fMRI study has reported that a particular region in PFC, the inferior frontal junction (IFJ), played an important role in the top-down control of feature-based attention (Baldauf and Desimone, 2014).

In the monkey, we focused on the portion of ventral PFC that extends forward from FEF onto the pre-arcuate gyrus and ventral bank of the principal sulcus. This regions has interconnections with IT, TEO, and possibly V4, on the one hand, and connections with FEF and other parts of PFC on the other (Barbas and Pandya, 1989; Webster et al., 1994). A monkey imaging study has shown that this region, along with FEF and a posterior portion of area 46, is differentially activated during search for a salient target (Wardak et al., 2010). Because the physiological properties of the cells in the ventral bank of the principal sulcus (which we will term “VPS”) and cells on the ventral pre-arcuate gyrus (which we will term “VPA”) appeared to be somewhat different, we have presented the results from the two subregions of PFC separately, using strictly anatomical designations.

We also recorded from the central portion of the inferior temporal (IT) cortex, which plays an important role in object recognition (for review, see (DiCarlo et al., 2012)) to test the alternative possibility that a stage of visual processing later than V4 is the source of feature-based attention, consistent with known feedback of attentional modulation from higher-order to lower-order visual areas (Buffalo et al., 2010). The distinctive properties of cells in VPA and the effects of VPA deactivation on behavior and FEF responses suggest that this region could be the equivalent of the IFJ in humans and thereby play a key role in feature based attention.

RESULTS

Monkeys were trained to perform a free-viewing visual search task as described in previous studies (Bichot et al., 2005; Zhou and Desimone, 2011), but with natural images (including those of faces) rather than simple colored shapes in order to increase selective responses in IT (Desimone et al., 1984; Moeller et al., 2008). Briefly, the animals were presented with a central cue object (serving as the search target) at fixation followed by a delay. The monkeys held the memory of the target during the delay. An array of 8 stimuli then appeared, containing both distracters and a single instance of the search target (Figure 1); the target and distracter items were pseudo-randomly chosen from a fixed set of 8 complex objects on each trial. The monkeys could use free gaze to find the target in the array, and they were rewarded for maintaining fixation on the target for 800 ms continuously. Detection trials, in which the search array contained only the target and no distracters, were randomly interleaved amongst the search trials in order to map neurons’ RFs across the twelve possible stimulus locations, as well as their visual selectivity for the objects used in the experiment.

Figure 1.

Figure 1

Open in a new tab

Schematic representation of behavioral tasks. Dotted circles represent the monkey’s current point of fixation. The initial sequence of events (i.e., fixation, cue, and delay periods) were the same in detection and free-viewing visual search trials. The target (i.e., cued stimulus) was presented alone in detection trials, and along with distractors in search trials. In this example of a search trial, the animal made two saccades (represented by the sequence of black arrows) before finding the target stimulus.

As described above, we found it useful to distinguish cells recorded in the VPA versus VPS regions, and we therefore report their properties separately. Multi-unit activity was recorded simultaneously in IT, VPA, and FEF of two monkeys (monkey B, 15 sessions; monkey R, 13 sessions), using multi-contact electrodes with 16 contacts spaced over 2.25 mm. We will refer to the multi-unit activity at each site simply as “units”. In two other monkeys, we recorded simultaneously from VPS, VPA, and FEF (monkey F, 19 sessions; monkey M, 11 sessions). Penetrations were made through multiple holes in a grid, and surface reconstructions of the grid hole locations are shown in Figure S1. On two penetrations in the most anterior part of VPA, all units were unresponsive, and the data were not included in any analyses. Given the known topographically-organized RF eccentricity representation in FEF (Bruce et al., 1985), recording locations in this area were chosen based on exploratory mapping sessions so that RFs at the recording sites encompassed the fixed stimulus locations used throughout the study. Based on the depths within sulci at which units were recorded at various sites, we sampled a total of approximately 28, 34, 29, and 48 mm2 of cortex in IT, VPA, FEF, and VPS, respectively.

Overall, monkeys performed similarly, finding the search target on >95% of trials after an average of 2.9 (+/− 0.2 SEM) saccades with an average saccadic latency of 203.8 ms (+/− 3.8 ms SEM) over those recording sessions. These performance measures show that the animals used object information to efficiently guide their search as they were significantly smaller than would be expected if the animals had chosen to search the display strictly serially or randomly (i.e., compared to averages of 4.5 saccades or 800 ms fixation durations; one-sample T-tests, t = 6.75 and 160.37, respectively, P < 10−8 for both comparisons). Data from the animals have been combined because they were qualitatively similar (one-way ANOVA; number of saccades, F = 1.12, P = 0.35; saccade latency, F = 0.43, P = 0.73).

Stimulus selectivity

In our sample, we found significant stimulus selectivity in VPA, VPS, and IT in 35%, 27%, and 48% of the units, respectively, based on an ANOVA (evaluated at P < 0.05) computed on the responses to the set of stimuli in the detection trials. Figure 2A shows the ordered responses from best to worst stimulus for those cells. The locations of stimulus-selective units in PFC are shown in Figure S2. In contrast, no units in FEF showed stimulus selectivity based on the same ANOVA, consistent with previous studies of this area (Bichot and Schall, 1999; Bichot et al., 1996; Mohler et al., 1973; Schall et al., 1995). Thus, in terms of feature selectivity, cells in VPA were more similar to the other two areas than to FEF. The time courses of feature selective responses for the cue presented at the fovea and the cued target presented alone in the detection trials in VPA, IT, and VPS are shown in Figures 3A and B.

Figure 2.

Figure 2

Open in a new tab

Selectivity and spatial tuning in VPA, FEF, IT, and VPS. A. Selectivity tuning showing ordered average responses from best to worst stimulus in areas where neurons with stimulus selective responses were found (i.e., VPA, IT, and VPS). Selectivity tuning in FEF is shown for comparison purposes since no significant selectivity was found in the area. Responses are normalized by the response to the best stimulus. B. Receptive field (RF) spatial tuning in VPA (solid line), FEF (dotted line), IT (dashed line), and VPS (dashed-dotted line). The ratio of the average response at each location relative to the average response at the center of the RF (i.e., location eliciting the largest average response) is shown as a function of the distance between that location and the center of the RF. Error bars represent SEM. See also Figures S2 and S3.

Figure 3.

Figure 3

Open in a new tab

Neural correlates of working-memory in IT, VPA, and VPS during free-viewing visual search. Normalized firing rates averaged across the population of recorded neurons are shown when the search target was the neurons’ preferred stimulus (red lines) compared to when the search target was the neurons’ non-preferred stimulus (blue lines). SEM (±) at each time point is indicated by shading over the lines. Plotted are normalized population responses to the centrally presented cue (A), responses to the target presented alone during detection trials (B), responses during search prior to the first saccade made to the target or to a distractor (C and D, respectively), and responses during visual search on the second and subsequent saccades when they were made to the target or to a distractor (E and F, respectively) with activity aligned to the end of the previous saccade at time zero. Only activity from correct trials and before saccade initiation (i.e., first saccade for B, C, and D, and subsequent saccade for E and F) was used in the analyses. SEM (±) at each time point is indicated by shading over the lines. See also Figure S4 and Table S1.

Spatial selectivity

We tested for significant spatial selectivity (RFs), using an ANOVA (P < 0.05) computed on the responses to extrafoveal stimuli in the detection trials. In VPA and VPS, about two-thirds (104/154 and 64/108, respectively) of stimulus-selective neurons also had well-defined extrafoveal RFs determined by significant differences in average responses across extrafoveal stimulus locations (Figure S2), while only about half of IT stimulus-selective neurons (61/121) exhibited such extrafoveal spatial selectivity. The remaining neurons in all these regions usually had very large receptive fields responding to all stimulus locations equally (i.e., no statistical difference), including locations in the ipsilateral visual field. As shown in Figure 2B, the RFs of the units with significant spatial tuning were, in our sample, largest on average in VPS, followed by IT cortex, and then VPA and FEF, which were similar to each other. While no neurons in VPA had RF centers in the ipsilateral hemifield, many of the RFs (40/104) extended into the ipsilateral hemifield. It is possible that with longer presentation times, more of the PFC units would have had larger, more bilateral RFs (see (Zaksas and Pasternak, 2006)) as Kadohisa et al (2015) have shown that large PFC fields develop slowly over time. Thus, both VPA and VPS have spatial and feature selectivity, consistent with previous studies of PFC (Everling et al., 2006; Rainer et al., 1998; Rao et al., 1997), although the spatial selectivity in VPA is more similar to FEF.

Many units in IT and VPS with spatially-selective extrafoveal responses also responded significantly to the cue presented foveally (46% and 49%, respectively), whereas this was less frequent in VPA and FEF (37% and 18%, respectively). The median RF center eccentricity of the spatially selective units was 6 degrees (dva) in all areas (Figure S3) and was not significantly different across areas (Kruskal-Wallis one-way ANOVA, χ2 = 1.81, P = 0.61).

Persistent stimulus selective activity

Given that VPA, VPS and IT cortex all showed stimulus-selective cue responses, we asked whether cue-related information persisted throughout the trial. Figure 3 shows the population responses in the three stimulus-selective areas during several phases of the search trials, separately for trials when the preferred versus non-preferred stimulus was the search cue. It was not possible to perform this analysis for FEF as the units did not have preferred stimuli. Population responses leading up to the first saccade were analyzed separately from later saccades as they contain the visually-evoked response to array onset (Bichot et al., 2005; Zhou and Desimone, 2011).

Cells in all three areas showed stimulus selective responses to the search cues and the target presented alone in detection trials, as shown in the population average histograms for the preferred and non-preferred stimulus for each cell in Figure 3A and B, respectively. However, cells in VPA differed from cells in the other two areas in that the population activity remained higher throughout the search trial when their preferred stimulus was the cue (i.e. when the animal was searching for the preferred stimulus as the target) than when the non-preferred stimulus was the cue (Figures 3C–F and S4, Table S1), and this higher activity persisted through the memory delay and through the response intervals for targets and distracters, on the first saccade and subsequent saccades. Units in VPS had higher activity during the memory delay following the preferred stimulus as the cue, but, unlike in VPA, this difference was only marginally significant on the first saccade and did not persist for the following saccades to targets or distracters. Thus, unlike VPS, VPA retained information about the sought-after target identity during the major decision times during the trial, and the difference between VPA and VPS was highly significant (T-test, difference in normalized activity between preferred search and non-preferred search; 100–200 ms after array onset [before first saccade], saccade to target: t = 6.14, P < 10−8, saccade to distractor: t = 5.99, P < 10−8; 100–200 ms from previous fixation [before subsequent saccades], saccade to target: t = 4.83, P < 10−5, saccade to distractor: t = 5.30, P < 10−6). Units in IT cortex gave somewhat higher responses to the preferred stimulus as the target on the first and subsequent saccades, but this difference did not persist for saccades to distracters. The IT response modulation might have been due to spatial attention to the target stimulus.

Feature selection/attention

Although VPA had distinctive stimulus-selective activity throughout the trial, a key question was whether the cells communicated information about the relationship between the attended target features and the features of the stimulus in the RF, independent of spatial attention. To separate out the effects of feature-based and spatial-based attention, we used a strategy that has been used in previous studies of FEF and V4 (Bichot and Schall, 1999; Gregoriou et al., 2009; Zhou and Desimone, 2011). For feature attention, we examined responses to the stimulus in the RF at times during the trial when the animal was preparing a saccade to a stimulus outside the RF. With spatial attention directed outside the RF, we asked whether the response to the RF varied according to whether the RF stimulus matched the features of the searched-for target (red lines in Figures 4A and S5A) or did not match (i.e., a distractor was in the RF; blue lines in Figures 4A and S5A). For spatial attention, we examined responses to the target stimulus in the RF when the animal was preparing to make a saccade to it (green lines in Figures 4A and S5A) or to a stimulus outside the RF (red lines in Figures 4A and S5A). For VPA, IT, and VPS, we analyzed activity on trials in which the animals searched for the preferred target of cells; for FEF, all target conditions were combined since the neurons did not show selectivity for the different stimuli.

Figure 4.

Figure 4

Open in a new tab

Time course of feature-based and spatial selection. A. From top to bottom, normalized responses in FEF, VPA, IT, and VPS, aligned to the onset of the search array when the first saccade was made to the target in the RF (green lines), when the target was in the RF but the saccade was made to a distractor outside the RF (red lines), and when the target was outside the RF (and a distractor was in the RF) and the saccade was made to a distractor outside the RF (blue lines). Responses in VPA, IT, and VPS were from correct trials and when the target was the preferred stimulus. Red vertical lines represent the onset of feature-based selection (difference between red and blue lines), and green vertical lines represent the onset of spatial selection (difference between green and red lines). SEM (±) at each time point is indicated by shading over the lines. Only spikes occurring prior to saccade initiation were used in the analyses. Because sample sizes were different across regions, we computed the timecourse of regions with more units by subsampling their population with the lowest number of units found in any region (i.e., IT) and obtaining an average over 10,000 iterations; shown response SEM for these regions is the average of the SEM calculated for the subsamples. B. Cumulative distribution of feature-based (left) and spatial (right) attentional effect latencies, computed from individual recording sites. There were more available trials for analysis in FEF (due to the lack of stimulus selectivity) than the other regions which all had similar numbers of contributing trials. Therefore, we subsampled the available trials in FEF with the average number of trials used in VPA, VPS, and IT (i.e., 22 trials for the target in RF / saccade to target condition, 37 trials for the target in RF / saccade outside RF condition, and 65 trials for the target outside RF / saccade outside RF condition) and averaged results over 10,000 iterations. Units that contributed less than five trials to any of the conditions were excluded from all analyses. For feature selection, all units had at least 10 trials contributing to each compared condition. For spatial selection, only 11/397 units had less than 7 trials in the saccade to the target in the RF condition which, as described above, yields on average the least number of trials. See also Figures S2 and S5.

Population responses in VPA and FEF showed substantial effects of feature based attention (100–200 ms after array onset, T-test, VPA: t = 9.42, P < 10−14; FEF: t = 8.96, P < 10−15), with an increase in response of 21.8% and 8.1% with feature attention in VPA and FEF, respectively. IT cortex and VPS showed smaller effects of feature attention (4.2% and 1.1% increase, respectively), and these effects were not significant during the same time period (IT: t = 1.06, P = 0.29; VPS: t = 0.52, P = 0.60).

We compared the latencies of feature selective effects in two different ways. We first computed the earliest detectable effect of attention in the population response histograms. The population histograms might reveal very early differences that are not significant at the level of individual units, although very few units may contribute to early effects. The latency for the effects of feature-based attentional selection in VPA (90 ms) was somewhat earlier than in FEF (100 ms), although the difference was not statistically significant (two-sided permutation test, P = 0.62). In contrast to both VPA and FEF, the effects of feature attention in VPS did not meet the criteria for the determination of a feature attention latency (i.e., difference in activity significant at the 0.05 level for at least 10 ms). The effects of feature attention in IT were smaller than in VPA and FEF, and the time of earliest feature selection in IT (189 ms) was significantly later than in both VPA (P = 0.015) and FEF (P = 0.024).

We next compared the areas by measuring the latency of feature attention effects for each recorded unit in each area, and then comparing the cumulative distributions of latencies, as shown in Figure 4B. Small proportions of cells in VPA and FEF showed early effects of feature attention (below 100 ms), consistent with the analysis of population histograms, but the cumulative distribution in VPA rose more steeply (earlier) than in FEF and the other two areas. At a cumulative distribution of 10 % of units, VPA led FEF by 20 ms, and this difference grew to 58 ms by a cumulative distribution of 35%. Overall, VPA had the largest proportion of units exhibiting feature-based selection (Chi-square test, vs. FEF, χ2 = 6.15, P = 0.013; vs. IT, χ2 = 12.51, P < 10−3; vs. VPS, χ2 = 11.03, P < 10−3) with overall earlier onset times (T-test, vs. FEF, t = 2.45, P = 0.016; vs. IT, t = 3.98, P < 10−3; vs. VPS, t = 3.06, P < 0.01), followed by FEF, while IT and VPS had the lowest proportions of units exhibiting such discrimination along with overall later times (see also Figure S2).

A signal-detection analysis also showed that VPA exhibited greater feature-based selection than any other region we sampled, as shown in Figure 5A. For each cell, the magnitude of feature-based selection was quantified by calculating the area under the receiver operating characteristic curve (AUROC) comparing activity (100–200 ms after array onset) when the target was in the RF and monkeys made a saccade to a distractor outside the RF to activity when the target was outside the RF and monkeys made a saccade to a distractor outside the RF. This measure of feature-based selection was largest in VPA (one-way ANOVA, F = 145.22, P < 10−62; T-tests comparing VPA to each of the other regions, P < 10−27 for all comparisons). Overall, it appears that feature-based selection in VPA occurs early enough and with a magnitude large enough to influence or be the source of feature-based selection in FEF. Feature-based selection is not a prominent property in VPS, clearly distinguishing it from VPA and FEF.

Figure 5.

Figure 5

Open in a new tab

Magnitude of feature-based (A) and spatial-based (B) selection. Error bars represent SEM. The number of units contributing to each area is shown in Figure 4.

VPA units also showed feature enhancement for their non-preferred target in the RF (Figure S5B), albeit with weaker and later effects than when animals searched for the neurons’ preferred target. No feature enhancement was found in either IT or VPS for the non-preferred target.

Spatial selection/attention

The time course of spatial selection revealed a nearly opposite trend compared to feature attention. We first examined the earliest evidence of spatial selection in the population response histograms. In contrast to feature-based selection, spatial selection occurred earlier in the FEF population response than in VPA (105 ms vs. 138 ms), although again the VPA-FEF difference was not significant (P = 0.35). The time of spatial selection in VPS (140 ms) was similar to that in VPA (see also Figure S2), while spatial selection did not meet the criteria to determine an onset of discrimination in IT.

As was found with feature-based selection, the analysis of cumulative distributions of spatial selection latencies revealed clear differences among the areas (Figure 4B). Small proportions of cells in VPA and FEF showed early effects of spatial selection, consistent with the analysis of population histograms, but the cumulative distribution in FEF rose more steeply (earlier) than in VPA and the other two areas. At a cumulative distribution of 10 % of units, FEF led VPA by 16 ms, and this difference grew to 39 ms by a cumulative distribution of 35%. Overall, FEF had the largest proportion of units (Chi-square test, vs. VPA, χ2 = 4.93, P = 0.026; vs. IT, χ2 = 20.17, P < 10−5; vs. VPS, χ2 = 8.29, P < 0.01) showing spatial-based selection with the earliest onset times (T-test, vs. VPA, t = 2.91, P < 0.01; vs. IT, t = 2.99, P < 0.01; vs. VPS, t = 2.26, P = 0.025).

A signal-detection analysis also showed that FEF exhibited greater spatial-based selection than any other region we sampled, as shown in Figure 5B. For each cell, the magnitude of spatial-based selection was quantified by calculating the AUROC comparing activity (100–200 ms after array onset) when the target was in the RF and monkeys made a saccade to it to activity when the target was in the RF and monkeys made a saccade outside the RF. This measure of spatial-based selection was largest in FEF (one-way ANOVA, F = 33.56, P < 10−18; T-tests comparing FEF to each of the other regions, P < 10−5 for all comparisons). These results suggest that while VPA may be the source of feature-based selection in FEF, the decision to make a saccade to a potential target likely originates in FEF and/or other related oculomotor structures and may be passed on to VPA. VPS is similar to VPA in terms of spatial selection, but clearly differs in feature-based selection.

Deactivation studies

To test for a causal role of VPA in feature-based selection, we tested the effects of VPA deactivation on both behavioral performance and selection in FEF during the random-design visual search (i.e., target changed randomly from trial to trial). We limited injections to the central portion of the VPA recording region, to avoid spread of muscimol into FEF and VPS. We nonetheless inactivated a substantial portion of this central region, using muscimol injections in six sessions (three each in monkeys F and M). In each session, three injections spaced 700 μm apart in depth were made with cannulas at each of two locations (Figure S1). Because the cannulas were inserted at an angle to the cortex, several square mm of cortex relative to the surface were likely affected.

Behavior before and after inactivation of VPA revealed significant post-inactivation deficits when the target was in the contralateral hemifield to the injection hemisphere, and to a lesser extent, when it was on the midline (Figures 6A and S6A, Table S2). The number of saccades to find a contralateral target increased while the opposite was true for an ipsilateral target. The total time to find the target, saccadic reaction times, and the percentage of trials in which the animals did not find the target all increased for both contralateral and midline targets. There were no effects on behavioral performance as a function of search block sequence or time during a session as assessed in training sessions a day prior to injection sessions (Figure S6B).

Figure 6.

Figure 6

Open in a new tab

Effects of VPA inactivation on behavioral performance and target selection in FEF. A. Effects of VPA inactivation on behavioral performance during search trials as a function of target location relative to the hemisphere in which VPA was inactivated. Data from the random and blocked visual search sessions are shown in orange and blue, respectively. Across session averages of behavioral measures are shown before (hashed bars) and after (solid bars) VPA inactivation. Midline locations (on the vertical meridian) were neither ipsilateral nor contralateral to the hemisphere of inactivation. Asterisks (*) mark significant effects of inactivation. B. Effects of VPA inactivation on behavioral performance during detection trials. Behavioral measures are shown before (hashed bars) and after (solid bars) VPA inactivation. Data from random and blocked design sessions were combined as search cue frequency had no effect on saccades to targets presented alone. For all analyses, trials in which monkeys broke fixation prior to the presentation of the target alone (detection) or with distractors (search) were not included. C, D. Effects of VPA inactivation on selection in FEF during random and blocked visual search, respectively. Population normalized responses in FEF during detection trials (top panels) and search trials (bottom panels) are shown before (left panels) and after (right panels) VPA inactivation. For detection trials, activity is shown when the target was inside (solid lines) or outside (dashed lines) the RF. For search trials, conventions as in Fig. 4. Only activity from correct trials and before saccade initiation was used in the analyses. See also Figures S6 and S7, and Table S2.

We also found a significant increase in saccades to the target with a following saccade away from it in the contralateral hemifield (pre: 7.5%, post: 10.5%; T-test, t = 4.56, P < 0.01), and a decrease of such behavior in the ipsilateral hemifield (pre: 7.9%, post: 3.8%; t = 7.63, P < 10−3). Furthermore, the pattern of distractor fixations in the contralateral hemifield was significantly affected by inactivation compared to the ipsilateral hemifield (correlation between pre- and post-inactivation distractor fixation patterns; mean Fisher z-transform: contralateral, 0.65, ipsilateral, 0.82; T-test, t = 8.11, P < 10−3). In sum, the monkeys had difficulty matching stimuli to the cue in the contralateral hemifield following inactivation of VPA.

The injection sessions were treated as independent across days, to account for day-to-day variations in performance but they were not independent across locations in VPA because of the large size of the injections, as described above. As a conservative test of the deactivation effects on behavior, we summed the trials across all deactivation sessions and simply compared proportions of saccade errors before and during the deactivations using a chi-square test. This test also showed a significant increase in errors post-inactivation for targets in the contralateral hemifield and on the midline (χ2 = 124.11, P < 10−28; and χ2 = 37.77, P < 10−9, respectively), but not for targets in the ipsilateral hemifield (χ2 = 3.19, P = 0.07).

We recorded the activity of 42 FEF units during visual search before and after VPA inactivations (Figures 6C and S7). While neural activity during detection trials (100–200 ms following stimulus onset) was not affected by VPA inactivation (repeated-measures two-way ANOVA; target in RF vs. out RF, F = 2057.1, P < 10−15; pre- vs. post inactivation, F = 3.83, P = 0.06; interaction, F = 0.06, P = 0.80), activity during search was significantly altered (two-way ANOVA; target in RF and saccade to target vs. target in RF and saccade outside RF vs. target outside RF and saccade outside RF, F = 85.27, P < 10−15; pre- vs. post inactivation, F = 8.55, P < 0.01; interaction, F = 12.45, P < 10−4). Most strikingly, feature selection in FEF (difference between red and blue lines) was completely abolished post-inactivation (T-test, pre-inactivation, t = 6.27, P < 10−6; post-inactivation, t = 0.64, P = 0.53). By contrast, even though neural activity when the saccade was made to the target in the RF was modestly lower post-inactivation, the effect of spatial attention (difference between green and red lines) was still present (t = 6.05, P < 10−6), and it was not significantly affected by the inactivation (pre- vs. post-inactivation, t = 1.29, P = 0.21). The differential effect of VPA inactivation on feature and spatial attention was confirmed by a signal-detection analysis. AUROC computed on a cell-by-cell basis in the 100–200 ms period after array onset showed that feature attention information significantly decreased (pre: 0.633, post: 0.508; T-test, t = 13.27, P < 10−21), while spatial attention information was not significantly altered (pre: 0.717, post: 0.698; T-test, t = 1.25, P = 0.21). Thus, VPA inactivation eliminated the effects of feature selection in FEF but not spatial selection.

As a control for the possibility that the effects of feature selection in FEF would normally decline in the second half of the recording session even without VPA inactivation, we compared feature selection in FEF of the same monkeys in the first versus second half of the session on days without VPA deactivation (these sessions had twice as many search trials as those during the pre- or post-inactivation blocks in the deactivation sessions). The results of recordings from 58 neurons (Figure S5A) showed no difference in the magnitude of feature selection between the two halves of the sessions (T-test, t = 0.30, P = 0.77).

Previous studies showed that when attention to a cue or stimulus is repeated for many trials, the effects of PFC lesions on attention are greatly reduced (Pasternak et al., 2015; Rossi et al., 2007). We therefore repeated the inactivation of VPA in six additional sessions (three in each monkey) using a blocked design, in which the target cue remained the same in blocks of 20 consecutive trials. We found that while deficits after inactivation were somewhat mitigated (Figure 6A, Table S2), monkeys still made more saccades, took longer to find the target, and made more errors when searching for a target in the contralateral hemifield. However, there were no longer significant effects on saccade latencies, or for targets at midline locations in general.

We also recorded from 38 units in FEF during these blocked sessions (Figures 6D and S7). Consistent with the somewhat reduced behavioral deficits in the blocked-design search, the effects of feature attention in FEF were still significant post-inactivation but smaller compared to pre-inactivation (T-test, t = 8.43, pre/feature, P < 10−9; post/feature, t = 2.45, P = 0.02; pre vs. post, t = 4.54, P < 10−4), while spatial enhancement was again unchanged after inactivation compared to before (pre/spatial, t = 3.08, P < 0.01; post/spatial, t = 2.99, P < 0.01; pre vs. post, t = 0.41, P = 0.68). The differential effect of VPA inactivation on feature and spatial attention was again confirmed by a signal-detection analysis. AUROC computed on a cell-by-cell basis in the 100–200 ms period after array onset showed that feature attention information significantly decreased (pre: 0.672, post: 0.559; T-test, t = 12.22, P < 10−18), while spatial attention information was not significantly altered (pre: 0.605, post: 0.583; T-test, t = 1.79, P = 0.08). Also, consistent with the lack of changes in spatial selection in FEF and neural activity during detection trials, neither accuracy nor saccade latencies during the detections trials that were interleaved with either variant of the search task were affected by VPA inactivation (Figure 6B, Table S2). In sum, both behavior in the task and the effects on FEF responses are more sensitive to the loss of VPA inputs when the cue changes frequently, but VPA seems important for feature-attention even with repeated cues.

Finally, we compared the behavioral effects of VPA inactivation to those from inactivating a nearby portion of VPS (Figure 7A, Table S3). We inactivated VPS with muscimol in six sessions (three each in monkeys F and M); in each session, three injections spaced 1 mm apart in depth were made at each of two different locations (Figure S1). Similar to the effects of VPA inactivation, inactivation of VPS caused large behavioral deficits for targets in the contralateral hemifield or on the midline during random-design visual search. Following inactivation, there were increases in the number of saccades made to find the target, the total time to find the target, and error rates for both contralateral and midline target locations, and increases in saccade latencies for all target locations. However, unlike with VPA inactivation, there were no significant behavioral deficits during blocked-design search. Performance during detection trials was not affected by VPS inactivation in either type of session (Figure 7B, Table S3). Thus, unlike VPA, VPS seems to play an important role in feature based selection only when attention switches frequently.

Figure 7.

Figure 7

Open in a new tab

Effects of VPS inactivation on behavioral performance. Conventions as in Fig. 6. See also Table S3.

DISCUSSION

Although much is known about the sources of top-down signals for visual spatial attention in monkey cortex, much less has been known about the sources of signals important for feature attention. A previous study found that feature-based target selection in area V4 occurs later than in FEF (Zhou and Desimone, 2011), suggesting that the earliest site of feature-based selection may be outside of visual cortex. Here we found that neurons in the VPA region of prefrontal cortex exhibit feature-based attentional modulation with a time course early enough to be a major cause of feature-based selection in FEF and possibly all other ventral stream areas. Combining our results with the earlier study of V4 and FEF by Zhou and Desimone (2011), feature-based selection also occurs earlier in VPA than in area V4. Spatially-selective VPA units also had RFs similar to those in FEF but, unlike FEF, many also showed selectivity for the objects used in our task. This selectivity could reflect an underlying selectivity for the component features of the objects, or selectivity for the objects acquired through learning to search for them (i.e., based on task demands) (e.g. (Freedman et al., 2001; Kadohisa et al., 2015; McKee et al., 2014)). Thus, VPA units seem to combine information about object features with their spatial location (see also (Kadohisa et al., 2015; Rainer et al., 1998; Rao et al., 1997)), and may provide information about both the identity and location of targets with behavioral relevance in the visual field.

We recorded from cells in IT cortex because it seemed possible that IT cortex might contain early feature-based signals for target selection even though area V4 does not. However, we found relatively late selection signals in IT, consistent with the findings of earlier studies of IT responses during search tasks (Chelazzi et al., 1998; Monosov et al., 2010; Sheinberg and Logothetis, 2001). Our results extend those previous findings by providing the first direct comparison of the time course of feature-based attention in IT and FEF dissociable from the effects of spatial selection in a task with an attentional template.

VPA as a source of feature-based selection is supported by our finding that feature-based, but not spatial selection in FEF is impaired by VPA inactivation. Following VPA inactivation, FEF cells respond as though they no longer have access to information about the location of objects with target features. We do not yet know whether VPA also sends direct feedback to other visual areas to support feature-based attention in these areas. Consistent with the differential effects of VPA inactivation on feature and spatial attention in FEF, our analysis of the time course of attentional modulation suggests that, within PFC, spatial selection originates in FEF, and feedback from FEF is likely a major (but not sole) source of feedback to visual cortex during spatial attention (Gregoriou et al., 2014; Moore and Armstrong, 2003; Moore and Fallah, 2001).

How do VPA cells compute the similarity between the features of the stimulus in their RF and the features of the target that the animal is searching for, or what has been referred to as the attentional template? One clue is that VPA seems to be unique among the regions we studied in having an explicit representation of the attentional template (the “cue”) throughout the delay and the search trial, even persisting across saccades. VPA cells have higher firing rates throughout the trial when their preferred stimulus is the cue/target, compared to non-preferred stimuli. This persistence of the attentional template in VPA may be used to directly compute stimulus similarity during search.

The combined feature and spatial information we observed in VPA is consistent with previous recordings in overlapping parts of PFC. However, because we recorded multi-unit activity, we cannot be certain that feature and spatial selectivity was combined at the level of individual VPA cells. Other studies have shown that individual PFC neurons can encode a working memory of both objects and locations during the delay period (Kadohisa et al., 2015; Rainer et al., 1998; Rao et al., 1997). Furthermore, the sustained representation of the attentional template we found is similar to the robust memory trace observed in PFC, but not IT, during a non-spatial match-to-sample task (Miller et al., 1996). Similarly, the discrimination of target objects (even when non-preferred) in VPA is consistent with the selective representation of task-relevant objects at preferred locations previously found in PFC (Everling et al., 2006).

We have also shown that such feature-based modulation of neural activity throughout the search trial is not ubiquitous in PFC, with nearby neurons in VPS exhibiting little to no such effects. The time of feature selection effects in VPS for units that showed any such effect was also significantly later than those in VPA and FEF. Furthermore, while the effects of VPS inactivation during visual search were mitigated by repetition of the target cue, deficits persisted with cue repetition after VPA inactivation, suggesting that VPS may be more important for attention switching or working memory while VPA may be more important for feature attention across the board. It is possible that while neurons in VPS are more involved in encoding the cue or the ability to adapt to changes in the cue, neurons in VPA process the stimuli of the search display as potential matches to the cue (i.e., a spatial “match-to-sample”). Neither region appears to play a role in saccade production per se; their inactivation does not cause any deficits in making a visually-guided saccade to a target presented alone, unlike the impairments observed following FEF inactivation (Dias and Segraves, 1999).

Our goal was to determine whether activity in PFC beyond FEF can be the source of feature-based selection signals found in FEF, and we have found units consistent with this hypothesis in VPA. Anatomical studies have shown that this region has connections with TEO, IT cortex and possibly area V4 (Barbas and Pandya, 1989; Webster et al., 1994). Although our recordings in VPA showed clear differences with cells recorded in adjacent areas VPS and FEF, we do not claim that VPA is a functionally defined area with clear boundaries. We did not study all of VPS and other parts of PFC to be sure whether there are other regions with properties similar to those in VPA. Several other studies have reported substantial regional overlap for coding of different types of information in dorsolateral PFC (e.g. (Kadohisa et al., 2015; Wallis et al., 2001; Watanabe, 1986; White and Wise, 1999)). One possible explanation could be that many complex neuronal properties are shared across PFC subregions but signals for top-down feature based attention are more concentrated in VPA. The adjacency of VPA to FEF suggests it could have a special relationship to this area. Another possible explanation could be that many studies of PFC tested across different subregions for the presence or absence of various types of information at any time in the trial (e.g., the delay period following the sample during a match-to-sample paradigm). We also found feature-based attentional effects on responses in V4, IT, FEF, and VPA at some time point during the trial, and would likely find them throughout the visual cortex, PFC, and regions of the parietal cortex through feedforward and feedback connectivity. However, the critical question for this study was where the feature selection effects emerged the earliest, and that appears to be VPA. Consistent with our findings, a recent study in which monkeys reported the color or motion of foveally presented stimuli found that choice signals developed in lateral prefrontal cortex and parietal regions and were fed back to FEF and sensory cortex (Siegel et al., 2015).

We have referred to our recording region as VPA simply as a description of its anatomical location. Our recording sites likely encompass multiple cytoarchitectonic areas such as Areas 45A and 12, and even possibly Area 46v. In future studies, it will be necessary to functionally map much more of the PFC, including more dorsal and anterior portions, to determine whether VPA is unique, or whether it might even be considered a separate, functionally-defined ”region”. An imaging study in monkeys searching for a salient target found activation only within a restricted portion of PFC, including the region we termed VPA, FEF and a posterior part of Area 46 (Wardak et al., 2010).

We did not find evidence for the early selection of targets defined by feature in IT cortex, consistent with the results of other studies in IT during visual search (Chelazzi et al., 1998; Monosov et al., 2010; Sheinberg and Logothetis, 2001). However, we did not record throughout the entire IT region, and therefore we cannot be sure that some IT cells with properties similar to those in VPA do not exist. Likewise, there could be other cortical sources for signals important for feature attention outside of PFC, including the parietal cortex, for example. At this stage, we can only be confident that VPA has the necessary signals at an early enough time to support feature based selection, and that VPA deactivation leads to behavioral impairments and a loss of feature-based selection in FEF.

Altogether, our results suggest a prefrontal, rather than visual cortical, source of feature-based attention, culminating in the priority maps in FEF from which a target is chosen for overt or covert orienting. FEF may, in turn, send feedback to topographically organized visual areas, enhancing activity at locations in the visual field representations containing stimuli that share target features. In that case, some of the effects of feature-based attention found in extrastriate areas (Bichot et al., 2005; Chelazzi et al., 2001; Hayden and Gallant, 2005; Martinez-Trujillo and Treue, 2004; McAdams and Maunsell, 2000; Motter, 1994) may have been caused by FEF feedback targeted strictly to the visual field locations, rather than the representation of stimulus features, of potential targets.

Studies examining the relationship between PFC and visual cortex during working memory for motion signals in match-to-sample tasks have found that, while robust template encoding is indeed present in PFC, MST may be the source of the delay activity seen in PFC (Mendoza-Halliday et al., 2014). MT may also play an important role in the comparison between sample and test stimuli (Zaksas and Pasternak, 2006). Our analyses have focused on the search period to determine the source of feature attention and thus it is difficult to make direct comparisons with match-to-sample tasks in which distracting information is not present with the target. It is also possible that synchrony measures (Gregoriou et al., 2009) or dynamic population coding (Mante et al., 2013; Stokes et al., 2013) beyond the scope of this study will reveal more complex interactions between different subregions of PFC and visual cortex in different phases of the search task.

Nonetheless, our findings in VPA are consistent with a recent study showing the prefrontal gating of object-based attention in humans (Baldauf and Desimone, 2014), and VPA may be the nonhuman primate homolog of the inferior frontal junction (IFJ) described in that report as a source of feedback in object-feature based attention (also see (Neubert et al., 2014)). Given the similarity between the “attentional template” that seems to be represented in VPA, and the object representations thought to be actively maintained during visual working memory and recall, VPA may have a very general role in covertly maintaining and manipulating visual object information.

EXPERIMENTAL PROCEDURES

Subjects and surgical procedures

Four adult male rhesus monkeys weighing 8–10 kg were used. Under aseptic conditions, monkeys were implanted with a headpost and chambers that allowed access to brain regions for neural recording and inactivation. All procedures and animal care were in accordance with NIH guidelines.

Behavioral tasks

The experiments were under the control of a PC computer using MonkeyLogic software (University of Chicago, IL), which presented the stimuli, monitored eye movements, and triggered the delivery of the reward. Monkeys were seated in an enclosed chair and eye position was monitored using an EyeLink II infrared system (SR Research Ltd., Ontario, Canada). Stimuli were presented on a video monitor viewed binocularly at a distance of 57 cm in a dark isolation box.

The stimuli were a fixed set of eight natural object images that were matched for the number of pixels different from the gray background, and subtended an area of approximately 1.5×1.5 dva. After fixating a small, white central fixation point for 800 ms, the monkeys were presented with a central cue that informed it of the stimulus selected as the detection or search target for that trial. In the search condition, the remaining seven stimuli became distractors for that trial. The cue stimulus stayed on for 1000 ms, after which time it was extinguished and replaced by the fixation spot for another 800 ms. The monkeys were required to hold fixation at the center of the screen during this delay period. At the end of the delay, the fixation spot was extinguished and, simultaneously, either the target was presented alone (detection trials) or presented among distractors (search trials). The monkeys were required to fixate the target stimulus for 800 ms continuously to receive a reward. For search trials, the animals had 8 s from search array onset to find the target, and no constraints were placed on their search behavior in order to allow them to conduct the search naturally. Even though the animals could fixate distractors as long as they wanted within a trial, only 3.5% of distractor fixations lasted 800 ms or longer. A search trial was considered an error only if an animal never fixated the search target continuously for 800 ms within the 8 s search duration. For detection trials, the animals had 50 ms to enter the target window and keep fixation at the target location until reward (i.e., multiple saccades were not allowed during detection trials in order to accurately map the properties of the RF). The target location was selected pseudo-randomly such that, within an experimental block, there were fifteen repetitions (five detection and ten search trials) of each stimulus presented as the target at each of twelve possible stimulus locations. The target locations, like the object identities, were fixed throughout the experiment. Once the location for the target stimulus was selected, the remaining seven distractors occupied locations selected randomly from the remaining eleven. The target identity, location, and trial type (i.e., detection vs. search) changed pseudo-randomly each trial, and all eight stimuli became the target on an equal number of total correct trials within an experimental block (1440 trials total for both detection and search trials). All neurophysiological recording data presented (except those from inactivation sessions – see below) came from sessions in which monkeys successfully completed all conditions in an experimental block.

For inactivation sessions, given the added duration from the injection procedure and resulting increased difficulty in maintaining stable recordings, the experimental block length was reduced by implementing only three repetitions of each target stimulus at each target location for detection trials, and five repetitions for search trials. In the random design, target identity, location, and trial type changed pseudo-randomly each trial as described above for regular recording sessions. In the blocked design, the search target remained the same in blocks of 20 correct trials. The sequence of targets between blocks was pseudo-random such that all conditions were included within an experimental block.

Neural recordings

Recordings began only after the monkeys were fully proficient in the search task and performance was stable. Recordings were conducted with multi-contact laminar electrodes (Plexon Inc., Dallas, TX) with 16 contacts spaced at 150 μm intervals, using the Omniplex system (Plexon Inc.). Due to the long duration of sessions, it was difficult to keep isolation on a single neuron; thus, the majority of the data are from small clusters of cells, or multi-unit activity, and are presented as such. To address the possibility that overlapping neural activity was recorded on adjacent contacts, we compared the zero-shift cross-correlation during the fixation period of signals on adjacent contacts to those at least three contacts away. There was only a very small increase of 1.2% of coincident spikes on adjacent contacts (2.9% vs. 4.1%), which may be partly due to an increased probability of common input connectivity of units on nearby contacts.

A grid system with holes 1 mm apart was used inside all the recording chambers to guide electrode penetrations and localize them relative to structural MRI images (see Figure S1 for recording sites). Penetration locations were confirmed with gray to white matter transition depths. FEF recording sites were in the rostral bank of the arcuate sulcus. VPS recording sites were in the ventral bank of the principal sulcus. VPA recording sites were on the pre-arcuate gyrus, anterior to the arcuate sulcus and ventral to the principal sulcus, and the penetrations did not enter either the arcuate sulcus or the principal sulcus (i.e., white matter was reached by the expected depth).

Neural inactivation

Muscimol (5 μg/μl) was injected in either VPA or VPS. The locations and depths were chosen based on the basis of exploratory recordings (Figure S1). In a given session, we made injections of 1 μl at three different depths and two locations within the selected area. The injections started at the deepest location where neurons were found, and subsequent injections were made by retracting the cannulas by steps of 700 um in VPA and 1 mm in VPS. The injections were made at a rate of 0.05 μl/min with a 5-minute wait between injections, and data collection began 35 minutes after the last injection. When concomitant recordings were made in FEF, the electrode was not moved or adjusted after the injection relative to its location before the injection.

Data analysis

Spike density functions were generated by convolving spikes with an asymmetric, forward-only filter designed to represent the postsynaptic consequences of cell activity (Thompson et al., 1996). The spike density function of each neuron was normalized by its maximum firing rate. The object and spatial selectivity of each site was determined using a two-way ANOVA with stimulus object and stimulus location during detection trials as the two main effects. If significant effects of object or location were found, post-hoc contrasts (T-tests) were used to determine preferred and non-preferred stimuli or locations inside and outside the RF of the units, respectively. Just as neurons can have RFs encompassing more than one stimulus location, they can also respond preferentially to more than one stimulus. The use of post-hoc contrasts to identify the preferred and non-preferred stimuli or locations, rather than just using best and worst ones was necessary in order to maximize the number of useable trials for the analyses. Object selectivity at the fovea was determined separately with a one-way ANOVA of responses to the different objects presented as the cue. Overall, a median of two stimuli were selected as preferred in VPA, VPS and IT; medians of four, three and five stimuli were selected as non-preferred in VPA, VPS and IT, respectively.

The time courses of feature-based and spatial selection were determined with a T-test at each millisecond following the time of search array presentation. The onset of selection was defined as the first millisecond when the difference between conditions became significant (P < 0.05) and remained significant for the next 10 ms.

Supplementary Material

1
2

Highlights.

  • Prefrontal cortex plays a key role in finding objects based on visual features

  • Neurons in the VPA region of PFC exhibit the earliest times of feature selection

  • Deactivation of VPA impairs the ability to find objects based on their features

  • VPA appears to be the source of feature selection in FEF, but not spatial selection

Acknowledgments

This work was funded by US National Eye Institute grant EY017921 and NSF grant CCF 1317348 (to R.D.).

Footnotes

AUTHOR CONTRIBUTIONS

N.P.B. and R.D. designed the experiments, analyzed the data and wrote the paper. N.P.B., E.M.D. and M.T.H conducted the experiments.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Baldauf D, Desimone R. Neural mechanisms of object-based attention. Science. 2014;344:424–427. doi: 10.1126/science.1247003. [DOI] [PubMed] [Google Scholar]
  2. Barbas H, Pandya DN. Architecture and intrinsic connections of the prefrontal cortex in the rhesus monkey. J Comp Neurol. 1989;286:353–375. doi: 10.1002/cne.902860306. [DOI] [PubMed] [Google Scholar]
  3. Basso MA, Wurtz RH. Modulation of neuronal activity in superior colliculus by changes in target probability. J Neurosci. 1998;18:7519–7534. doi: 10.1523/JNEUROSCI.18-18-07519.1998. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bichot NP, Rossi AF, Desimone R. Parallel and serial neural mechanisms for visual search in macaque area V4. Science. 2005;308:529–534. doi: 10.1126/science.1109676. [DOI] [PubMed] [Google Scholar]
  5. Bichot NP, Schall JD. Effects of similarity and history on neural mechanisms of visual selection. Nat Neurosci. 1999;2:549–554. doi: 10.1038/9205. [DOI] [PubMed] [Google Scholar]
  6. Bichot NP, Schall JD, Thompson KG. Visual feature selectivity in frontal eye fields induced by experience in mature macaques. Nature. 1996;381:697–699. doi: 10.1038/381697a0. [DOI] [PubMed] [Google Scholar]
  7. Bressler SL, Tang W, Sylvester CM, Shulman GL, Corbetta M. Top-down control of human visual cortex by frontal and parietal cortex in anticipatory visual spatial attention. J Neurosci. 2008;28:10056–10061. doi: 10.1523/JNEUROSCI.1776-08.2008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Bruce CJ, Goldberg ME, Bushnell MC, Stanton GB. Primate frontal eye fields. II. Physiological and anatomical correlates of electrically evoked eye movements. J Neurophysiol. 1985;54:714–734. doi: 10.1152/jn.1985.54.3.714. [DOI] [PubMed] [Google Scholar]
  9. Buffalo EA, Fries P, Landman R, Liang H, Desimone R. A backward progression of attentional effects in the ventral stream. Proc Natl Acad Sci U S A. 2010;107:361–365. doi: 10.1073/pnas.0907658106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Chelazzi L, Duncan J, Miller EK, Desimone R. Responses of neurons in inferior temporal cortex during memory-guided visual search. J Neurophysiol. 1998;80:2918–2940. doi: 10.1152/jn.1998.80.6.2918. [DOI] [PubMed] [Google Scholar]
  11. Chelazzi L, Miller EK, Duncan J, Desimone R. Responses of neurons in macaque area V4 during memory-guided visual search. Cereb Cortex. 2001;11:761–772. doi: 10.1093/cercor/11.8.761. [DOI] [PubMed] [Google Scholar]
  12. Desimone R, Albright TD, Gross CG, Bruce C. Stimulus-selective properties of inferior temporal neurons in the macaque. J Neurosci. 1984;4:2051–2062. doi: 10.1523/JNEUROSCI.04-08-02051.1984. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Desimone R, Duncan J. Neural mechanisms of selective visual attention. Annu Rev Neurosci. 1995;18:193–222. doi: 10.1146/annurev.ne.18.030195.001205. [DOI] [PubMed] [Google Scholar]
  14. Dias EC, Segraves MA. Muscimol-induced inactivation of monkey frontal eye field: effects on visually and memory-guided saccades. J Neurophysiol. 1999;81:2191–2214. doi: 10.1152/jn.1999.81.5.2191. [DOI] [PubMed] [Google Scholar]
  15. DiCarlo JJ, Zoccolan D, Rust NC. How does the brain solve visual object recognition? Neuron. 2012;73:415–434. doi: 10.1016/j.neuron.2012.01.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Duncan J, Humphreys GW. Visual search and stimulus similarity. Psychol Rev. 1989;96:433–458. doi: 10.1037/0033-295x.96.3.433. [DOI] [PubMed] [Google Scholar]
  17. Egner T, Monti JM, Trittschuh EH, Wieneke CA, Hirsch J, Mesulam MM. Neural integration of top-down spatial and feature-based information in visual search. J Neurosci. 2008;28:6141–6151. doi: 10.1523/JNEUROSCI.1262-08.2008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Everling S, Tinsley CJ, Gaffan D, Duncan J. Selective representation of task-relevant objects and locations in the monkey prefrontal cortex. Eur J Neurosci. 2006;23:2197–2214. doi: 10.1111/j.1460-9568.2006.04736.x. [DOI] [PubMed] [Google Scholar]
  19. Findlay JM, Walker R. A model of saccade generation based on parallel processing and competitive inhibition. Behav Brain Sci. 1999;22:661–674. doi: 10.1017/s0140525x99002150. discussion 674–721. [DOI] [PubMed] [Google Scholar]
  20. Freedman DJ, Riesenhuber M, Poggio T, Miller EK. Categorical representation of visual stimuli in the primate prefrontal cortex. Science. 2001;291:312–316. doi: 10.1126/science.291.5502.312. [DOI] [PubMed] [Google Scholar]
  21. Funahashi S, Bruce CJ, Goldman-Rakic PS. Mnemonic coding of visual space in the monkey’s dorsolateral prefrontal cortex. J Neurophysiol. 1989;61:331–349. doi: 10.1152/jn.1989.61.2.331. [DOI] [PubMed] [Google Scholar]
  22. Fuster JM, Alexander GE. Neuron activity related to short-term memory. Science. 1971;173:652–654. doi: 10.1126/science.173.3997.652. [DOI] [PubMed] [Google Scholar]
  23. Gazzaley A, Nobre AC. Top-down modulation: bridging selective attention and working memory. Trends Cogn Sci. 2012;16:129–135. doi: 10.1016/j.tics.2011.11.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Giesbrecht B, Woldorff MG, Song AW, Mangun GR. Neural mechanisms of top-down control during spatial and feature attention. Neuroimage. 2003;19:496–512. doi: 10.1016/s1053-8119(03)00162-9. [DOI] [PubMed] [Google Scholar]
  25. Gregoriou GG, Gotts SJ, Zhou H, Desimone R. High-frequency, long-range coupling between prefrontal and visual cortex during attention. Science. 2009;324:1207–1210. doi: 10.1126/science.1171402. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Gregoriou GG, Rossi AF, Ungerleider LG, Desimone R. Lesions of prefrontal cortex reduce attentional modulation of neuronal responses and synchrony in V4. Nat Neurosci. 2014;17:1003–1011. doi: 10.1038/nn.3742. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Hamker FH. The reentry hypothesis: the putative interaction of the frontal eye field, ventrolateral prefrontal cortex, and areas V4, IT for attention and eye movement. Cereb Cortex. 2005;15:431–447. doi: 10.1093/cercor/bhh146. [DOI] [PubMed] [Google Scholar]
  28. Hayden BY, Gallant JL. Time course of attention reveals different mechanisms for spatial and feature-based attention in area V4. Neuron. 2005;47:637–643. doi: 10.1016/j.neuron.2005.07.020. [DOI] [PubMed] [Google Scholar]
  29. Itti L, Koch C. Computational modelling of visual attention. Nat Rev Neurosci. 2001;2:194–203. doi: 10.1038/35058500. [DOI] [PubMed] [Google Scholar]
  30. Kadohisa M, Kusunoki M, Petrov P, Sigala N, Buckley MJ, Gaffan D, Duncan J. Spatial and temporal distribution of visual information coding in lateral prefrontal cortex. Eur J Neurosci. 2015;41:89–96. doi: 10.1111/ejn.12754. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Kusunoki M, Gottlieb J, Goldberg ME. The lateral intraparietal area as a salience map: the representation of abrupt onset, stimulus motion, and task relevance. Vision Res. 2000;40:1459–1468. doi: 10.1016/s0042-6989(99)00212-6. [DOI] [PubMed] [Google Scholar]
  32. Mante V, Sussillo D, Shenoy KV, Newsome WT. Context-dependent computation by recurrent dynamics in prefrontal cortex. Nature. 2013;503:78–84. doi: 10.1038/nature12742. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Martinez-Trujillo JC, Treue S. Feature-based attention increases the selectivity of population responses in primate visual cortex. Curr Biol. 2004;14:744–751. doi: 10.1016/j.cub.2004.04.028. [DOI] [PubMed] [Google Scholar]
  34. McAdams CJ, Maunsell JH. Attention to both space and feature modulates neuronal responses in macaque area V4. J Neurophysiol. 2000;83:1751–1755. doi: 10.1152/jn.2000.83.3.1751. [DOI] [PubMed] [Google Scholar]
  35. McKee JL, Riesenhuber M, Miller EK, Freedman DJ. Task dependence of visual and category representations in prefrontal and inferior temporal cortices. J Neurosci. 2014;34:16065–16075. doi: 10.1523/JNEUROSCI.1660-14.2014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Mendoza-Halliday D, Torres S, Martinez-Trujillo JC. Sharp emergence of feature-selective sustained activity along the dorsal visual pathway. Nat Neurosci. 2014;17:1255–1262. doi: 10.1038/nn.3785. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Miller EK, Cohen JD. An integrative theory of prefrontal cortex function. Annu Rev Neurosci. 2001;24:167–202. doi: 10.1146/annurev.neuro.24.1.167. [DOI] [PubMed] [Google Scholar]
  38. Miller EK, Erickson CA, Desimone R. Neural mechanisms of visual working memory in prefrontal cortex of the macaque. J Neurosci. 1996;16:5154–5167. doi: 10.1523/JNEUROSCI.16-16-05154.1996. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Moeller S, Freiwald WA, Tsao DY. Patches with links: a unified system for processing faces in the macaque temporal lobe. Science. 2008;320:1355–1359. doi: 10.1126/science.1157436. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Mohler CW, Goldberg ME, Wurtz RH. Visual receptive fields of frontal eye field neurons. Brain Res. 1973;61:385–389. doi: 10.1016/0006-8993(73)90543-x. [DOI] [PubMed] [Google Scholar]
  41. Monosov IE, Sheinberg DL, Thompson KG. Paired neuron recordings in the prefrontal and inferotemporal cortices reveal that spatial selection precedes object identification during visual search. Proc Natl Acad Sci U S A. 2010;107:13105–13110. doi: 10.1073/pnas.1002870107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Moore T, Armstrong KM. Selective gating of visual signals by microstimulation of frontal cortex. Nature. 2003;421:370–373. doi: 10.1038/nature01341. [DOI] [PubMed] [Google Scholar]
  43. Moore T, Fallah M. Control of eye movements and spatial attention. Proc Natl Acad Sci U S A. 2001;98:1273–1276. doi: 10.1073/pnas.021549498. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Motter BC. Neural correlates of attentive selection for color or luminance in extrastriate area V4. J Neurosci. 1994;14:2178–2189. doi: 10.1523/JNEUROSCI.14-04-02178.1994. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Neubert FX, Mars RB, Thomas AG, Sallet J, Rushworth MF. Comparison of human ventral frontal cortex areas for cognitive control and language with areas in monkey frontal cortex. Neuron. 2014;81:700–713. doi: 10.1016/j.neuron.2013.11.012. [DOI] [PubMed] [Google Scholar]
  46. Olshausen BA, Anderson CH, Van Essen DC. A neurobiological model of visual attention and invariant pattern recognition based on dynamic routing of information. J Neurosci. 1993;13:4700–4719. doi: 10.1523/JNEUROSCI.13-11-04700.1993. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Pasternak T, Lui LL, Spinelli PM. Unilateral prefrontal lesions impair memory-guided comparisons of contralateral visual motion. J Neurosci. 2015;35:7095–7105. doi: 10.1523/JNEUROSCI.5265-14.2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Rainer G, Asaad WF, Miller EK. Memory fields of neurons in the primate prefrontal cortex. Proc Natl Acad Sci U S A. 1998;95:15008–15013. doi: 10.1073/pnas.95.25.15008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Rao SC, Rainer G, Miller EK. Integration of what and where in the primate prefrontal cortex. Science. 1997;276:821–824. doi: 10.1126/science.276.5313.821. [DOI] [PubMed] [Google Scholar]
  50. Rossi AF, Bichot NP, Desimone R, Ungerleider LG. Top down attentional deficits in macaques with lesions of lateral prefrontal cortex. J Neurosci. 2007;27:11306–11314. doi: 10.1523/JNEUROSCI.2939-07.2007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Schall JD, Hanes DP, Thompson KG, King DJ. Saccade target selection in frontal eye field of macaque. I. Visual and premovement activation. J Neurosci. 1995;15:6905–6918. doi: 10.1523/JNEUROSCI.15-10-06905.1995. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Sheinberg DL, Logothetis NK. Noticing familiar objects in real world scenes: the role of temporal cortical neurons in natural vision. J Neurosci. 2001;21:1340–1350. doi: 10.1523/JNEUROSCI.21-04-01340.2001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Siegel M, Buschman TJ, Miller EK. Cortical information flow during flexible sensorimotor decisions. Science. 2015;348:1352–1355. doi: 10.1126/science.aab0551. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Stokes MG, Kusunoki M, Sigala N, Nili H, Gaffan D, Duncan J. Dynamic coding for cognitive control in prefrontal cortex. Neuron. 2013;78:364–375. doi: 10.1016/j.neuron.2013.01.039. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Thompson KG, Bichot NP. A visual salience map in the primate frontal eye field. Prog Brain Res. 2005;147:251–262. doi: 10.1016/S0079-6123(04)47019-8. [DOI] [PubMed] [Google Scholar]
  56. Thompson KG, Hanes DP, Bichot NP, Schall JD. Perceptual and motor processing stages identified in the activity of macaque frontal eye field neurons during visual search. J Neurophysiol. 1996;76:4040–4055. doi: 10.1152/jn.1996.76.6.4040. [DOI] [PubMed] [Google Scholar]
  57. Wallis JD, Anderson KC, Miller EK. Single neurons in prefrontal cortex encode abstract rules. Nature. 2001;411:953–956. doi: 10.1038/35082081. [DOI] [PubMed] [Google Scholar]
  58. Wardak C, Vanduffel W, Orban GA. Searching for a salient target involves frontal regions. Cereb Cortex. 2010;20:2464–2477. doi: 10.1093/cercor/bhp315. [DOI] [PubMed] [Google Scholar]
  59. Watanabe M. Prefrontal unit activity during delayed conditional Go/No-Go discrimination in the monkey. I. Relation to the stimulus. Brain Res. 1986;382:1–14. doi: 10.1016/0006-8993(86)90104-6. [DOI] [PubMed] [Google Scholar]
  60. Webster MJ, Bachevalier J, Ungerleider LG. Connections of inferior temporal areas TEO and TE with parietal and frontal cortex in macaque monkeys. Cereb Cortex. 1994;4:470–483. doi: 10.1093/cercor/4.5.470. [DOI] [PubMed] [Google Scholar]
  61. White IM, Wise SP. Rule-dependent neuronal activity in the prefrontal cortex. Exp Brain Res. 1999;126:315–335. doi: 10.1007/s002210050740. [DOI] [PubMed] [Google Scholar]
  62. Wolfe JM, Cave KR, Franzel SL. Guided search: an alternative to the feature integration model for visual search. J Exp Psychol Hum Percept Perform. 1989;15:419–433. doi: 10.1037//0096-1523.15.3.419. [DOI] [PubMed] [Google Scholar]
  63. Zaksas D, Pasternak T. Directional signals in the prefrontal cortex and in area MT during a working memory for visual motion task. J Neurosci. 2006;26:11726–11742. doi: 10.1523/JNEUROSCI.3420-06.2006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Zhou H, Desimone R. Feature-based attention in the frontal eye field and area V4 during visual search. Neuron. 2011;70:1205–1217. doi: 10.1016/j.neuron.2011.04.032. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS729299-supplement-1.pdf (411.1KB, pdf)
2
NIHMS729299-supplement-2.docx (173.8KB, docx)

RESOURCES