• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Psychol Sci. Author manuscript; available in PMC May 24, 2010.
Published in final edited form as:
PMCID: PMC2875116
NIHMSID: NIHMS198794

Stimulus-Specific Delay Activity in Human Primary Visual Cortex

Abstract

Working memory (WM) involves maintaining information in an on-line state. One emerging view is that information in WM is maintained via sensory recruitment, such that information is stored via sustained activity in the sensory areas that encode the to-be-remembered information. Using functional magnetic resonance imaging, we observed that key sensory regions such as primary visual cortex (V1) showed little evidence of sustained increases in mean activation during a WM delay period, though such amplitude increases have typically been used to determine whether a region is involved in on-line maintenance. However, a multivoxel pattern analysis of delay-period activity revealed a sustained pattern of activation in V1 that represented only the intentionally stored feature of a multifeature object. Moreover, the pattern of delay activity was qualitatively similar to that observed during the discrimination of sensory stimuli, suggesting that WM representations in V1 are reasonable “copies” of those evoked during pure sensory processing.

Working memory (WM) allows the on-line storage of behaviorally relevant information. One emerging view is that WM is supported by the same neural mechanisms that encode the sensory information being remembered (we term this the sensory-recruitment model of WM; see Awh & Jonides, 2001; D’Esposito, 2007; Jonides, Lacey, & Nee, 2005; Postle, 2006). For example, neurons in face-selective regions of inferotemporal cortex show sustained amplitude increases while an observer is holding a face in WM (Chelazzi, Miller, Duncan, & Desimone, 1993; Courtney, Ungerleider, Keil, & Haxby, 1997; Druzgal & D’Esposito, 2001; Lepsien & Nobre, 2007; Miller, Li, & Desimone, 1993; Ranganath, Cohen, Dam, & D’Esposito, 2004). The sensory-recruitment hypothesis assumes that this activity represents the specific stimulus values of the stored items. Here we report a study in which functional magnetic resonance imaging (fMRI) and multivoxel pattern analysis (MVPA) provided direct support for this claim, showing that activation patterns in relevant sensory regions represent the specific stimulus value that is held in WM.

MVPA provides a useful tool for identifying the neural regions that mediate WM by focusing on changes in activation patterns as opposed to simply changes in the mean amplitude of the blood-oxygenation-level-dependent (BOLD) response. For example, Offen, Schluppeck, and Heeger (in press) used fMRI to index activation changes in primary visual cortex (V1), a region known to represent orientation and spatial frequency. Although mean response amplitudes in V1 increased during sustained deployments of spatial attention, activation levels were indistinguishable from a low-level baseline when information about orientation (or spatial frequency) was stored in WM. This finding appears to contradict the sensory-recruitment model. However, as Offen et al. noted, neurons that respond preferentially to the remembered orientation should become more active, whereas neurons tuned away from the remembered orientation should be suppressed (relatively speaking; see, e.g., Martinez-Trujillo & Treue, 2004). A differential pattern of activity across the relevant sensory neurons is thought to represent the encoded orientation (Paradiso, 1988; Pouget, Dayan, & Zemel, 2003; Sanger, 1996), and therefore the sensory-recruitment account holds that this pattern should be maintained during a WM delay period as well. However, if the BOLD response spatially integrates information from neurons that are more active (i.e., those tuned to the remembered orientation) with information from neurons that are less active (Logothetis, Pauls, Augath, Trinath, & Oeltermann, 2001), then a failure to find a sustained amplitude increase in orientation-specific regions of cortex during a delay period does not provide strong evidence against the sensory-recruitment model of WM.

We tested the sensory-recruitment hypothesis by determining if WM is mediated by sustained feature-selective activation patterns in cortical regions that process the relevant sensory information. Using fMRI, MVPA, and a pattern-classification algorithm, we examined feature-specific WM modulations in V1 while subjects remembered either an orientation or a color for 10 s (Haxby et al., 2001; Haynes & Rees, 2005; Kamitani & Tong, 2005, 2006; Norman, Polyn, Detre, & Haxby, 2006; Peelen & Downing, 2007; Serences & Boynton, 2007a, 2007b). The observation of feature-specific activation patterns in V1 suggests that sensory mechanisms are recruited to support the storage of information in WM and furthermore indicates that subjects have top-down control over which features of a multifeature stimulus are stored.

METHOD

Observers

Ten neurologically intact observers participated in a single 2-hr scanning session. All observers gave written informed consent in accord with the requirements of the institutional review board at the University of Oregon. Data from 3 observers were discarded because of technical problems or voluntary withdrawal from the study. Each observer was trained in the experimental task for approximately 1.5 hr prior to scanning, to set sample-test disparities to threshold (see Staircase Procedure).

Behavioral Task

Stimuli were rendered on a light-gray background and displayed via a rear-mounted projector (see Fig. 1). Observers were instructed to maintain fixation on a central square (subtending 1° visual angle from a viewing distance of 58 cm) that was present for the duration of each scan. At the beginning of each trial, observers were shown the sample, a Gabor stimulus (radius = 13°) with a small circular aperture (2°) cut around the fixation square. The sample stimulus was rendered in one of two orientations (45° or 135°, plus or minus an additional offset randomly selected from a range of ±10°) and in one of two colors (middle red or middle green, plus or minus an offset randomly selected from a range of ±10% saturation). The addition of random jitter around the canonical features discouraged the use of verbal labels, which would not have been precise enough to support accurate performance in this task. For some observers (n = 2), the spatial frequency of the stimulus varied randomly across trials (0.75–1.25 cycles/deg); for the others, the spatial frequency was fixed at 1 cycle/deg. The sample stimulus flickered on and off at 5 Hz and was presented for a total of 1 s, followed by a 10-s blank retention interval. Next, observers were shown a test stimulus (again flickering at 5 Hz for a total of 1 s) that either was identical to or mismatched the sample along one feature dimension (i.e., either the orientation or the color was slightly different). Observers then used a custom-made button box to make a two-alternative forced-choice response regarding whether the sample and test stimuli matched. The sample and test stimuli differed on 50% of trials. A 10-s intertrial interval followed the offset of each test stimulus.

Fig. 1
The behavioral paradigm. On each trial, a sample stimulus that flickered at 5 Hz was presented for 1 s; observers were instructed to remember either the exact orientation or the exact color of this sample over the following 10-s delay period. After the ...

At the beginning of each block of trials (or scan), observers were informed that the sample and test stimuli would differ along only one feature dimension (color or orientation). Thus, remember-orientation and remember-color trials were run in separate scans. Each scan contained eight trials, and each observer completed seven or eight scans in each memory condition. Feedback (percentage correct) was given after each scan.

Staircase Procedure

To encourage a narrow focus of attention on only the relevant feature, and to discourage verbal-labeling strategies, we titrated task difficulty for each observer in a separate behavioral testing session (7–8 blocks of each feature condition). The task was identical to that just described, and the sample-test disparity was adjusted independently for each feature dimension until a criterion level of performance (approximately 75% correct) was reached. The resulting orientation and color disparities were used to determine the sample-test disparity for each feature during scanning.

fMRI Data Acquisition and Analysis

Scanning was performed using a 3-T Siemens Allegra system at the Robert and Beverly Lewis Center for Neuroimaging at the University of Oregon. Anatomical images were acquired using a spoiled-gradient-recalled T1-weighted sequence that yielded images with a 1-mm3 resolution. Whole-brain echo-planar images (EPIs) were acquired in 33 transverse slices (2,000-ms repetition time, 30-ms echo time, 90° flip angle, 64 × 64 matrix, 192-mm field of view, 3.5-mm slice thickness, no gap).

Data analysis was performed using BrainVoyagerQX (Version 1.86; Brain Innovations, Maastricht, The Netherlands) and custom time-series analysis and pattern-classification routines written in MATLAB (Version 7.2; Mathworks, Natick, MA). Either seven (n = 2) or eight (n = 5) scans of the remember-orientation and remember-color tasks were collected per subject (14–16 scans per subject), with each scan lasting 186 s. EPI images were slice-time-corrected, motion-corrected (within and between scans), and high-pass-filtered (three cycles per run).

Retinotopic Mapping

Retinotopic maps were obtained using a rotating checkerboard stimulus and standard presentation and analysis techniques (Engel et al., 1994; Sereno et al., 1995). This procedure was used to identify V1, a region known to respond to both color and orientation (e.g., Johnson, Hawken,& Shapley, 2001; Leventhal, Thompson, Liu, Zhou, & Ault, 1995; Sincich & Horton, 2005; Solomon & Lennie, 2007), as well as ventral (V2, V3, hV4) and dorsal (V2, V3, V3a) visual areas.

Voxel Selection

Independent functional localizer scans were used to identify regions of occipital visual cortex that responded to the spatial position occupied by the stimulus aperture in the main experiment. Colored Gabor stimuli identical to those used in the WM task were cycled on and off at 5 Hz for 10 s on each trial and followed by a 10-s passive fixation epoch. At the beginning of each localizer run, observers were instructed to attend to either the color or the orientation of the stimulus for the duration of that run. Color and orientation runs were presented in alternation. During each trial, two or three target events (a change in the value of the relevant feature dimension for 100 ms) occurred, and observers pressed a button whenever they detected a target event. Each localizer run contained 12 trials, and observers completed either one (n = 4) or two (n = 3) runs per attended feature dimension. We then used a general linear model (GLM) to identify voxels within each visual area that responded more strongly during epochs of stimulation than during epochs of passive fixation. The single regressor in the GLM was created by convolving a boxcar model of the stimulus protocol with a gamma function (Boynton, Engel, Glover, & Heeger, 1996). In each visual area, all voxels that passed a statistical threshold of p < .05 (corrected for multiple comparisons using the false-discovery-rate algorithm in BrainVoyager) were retained for further analysis.

Multivoxel Pattern Analysis

For the MVPA, we first extracted the raw time series from each voxel within each region of interest during a time period extending from 4 s to 10 s after the presentation of each sample stimulus. These time series were then normalized using a z transform on a scan-by-scan basis. Temporal epochs from all but one scan were extracted to form a “training” data set for the classification analysis; data from the remaining scan were defined as the “test” set (we use the term scan to refer to an entire 186-s data-collection sequence, so the training and test data sets were always independent). We then trained a Support Vector Machine (SVM; specifically, the OSU-SVM implementation, http://sourceforge.net/projects/svm/; see also Kamitani & Tong, 2005, 2006) using only the training data and then used the SVM to classify the orientation or the color of the sample stimulus on each trial from the test scan (classification of color and classification of orientation were carried out separately, so chance for all comparisons was 50%).

This procedure was repeated using a hold-one-scan-out cross-validation approach, so that data from every scan were used as a test set in turn. The SVM’s overall classification accuracy for each observer was then defined as the average classification accuracy across all seven or eight permutations of holding one scan out for use as a test set (depending on the number of scans the subject completed). Classification accuracy was averaged across corresponding regions of interest in the left and right hemispheres because no significant differences were observed between left and right visual areas.

RESULTS

Behavioral Performance

Behavioral performance is summarized in Table 1. During scanning, discrimination thresholds did not differ between the two subtypes of either feature dimension (45° vs. 135°, red vs. green), and overall accuracy was maintained at the level predetermined by the staircase procedure.

TABLE 1
Mean Performance on the Behavioral Task

Analysis of WM-Related Activation in Visual Cortex

Our data figures (Figs. 2, ,3,3, and and4)4) are based on the 62 most responsive voxels within each visual area because this was the minimum number of voxels with significant activity across subjects and visual areas; however, our main conclusions are robust even when more or fewer voxels are included in the analysis.

Fig. 2
Mean amplitude of the blood-oxygenation-level-dependent (BOLD) response in primary visual cortex (V1) across the working memory (WM) delay period and intertrial interval (ITI). All time series were computed against a baseline of the activation level at ...
Fig. 3
Feature-selective working memory (WM) modulations revealed by multivoxel pattern analysis. The graphs show classification accuracy as a function of the stimulus feature (color or orientation) being classified and whether the subject was instructed to ...
Fig. 4
Comparison of classification accuracy (collapsed across feature dimensions) in two time bins: 4 through 8 s poststimulus and 4 through 10 s poststimulus. The graph in (a) shows classification accuracy for remembered features (i.e., for orientation when ...

The goal of this study was to use fMRI and MVPA to determine whether, as predicted by the sensory-recruitment hypothesis, there are stimulus-specific modulations in early regions of visual cortex while an observer is remembering a specific orientation or color. Although we examined the response properties of several visual areas in occipital cortex (V1, V2v, V3v, hV4, V2d, V3d, V3a), we focused on V1 because this region contains neurons that are selective for both orientation and color (Johnson et al., 2001; Leventhal et al., 1995; Sincich & Horton, 2005; Solomon & Lennie, 2007), making it the ideal region to test for orientation-and color-selective modulations during the storage of information in WM. Independent functional localizer scans were used to identify the 62 V1 voxels in each observer that were most selective for the retinotopic position occupied by the stimulus aperture in the WM task (see the Method section for our voxel-selection logic).

Before performing the MVPA, we compared the mean amplitude of the BOLD response in V1 (collapsed across all 62 voxels) during the delay period of the WM task with the mean amplitude of the BOLD response in a corresponding temporal epoch following the test stimulus (see Figs. 2a and 2b). This comparison controlled for low-level sensory factors, as the sample and test stimuli were essentially identical. Even though response amplitudes were slightly higher 10 s postsample than 10 s posttest, overall activation levels during the WM delay period were statistically indistinguishable from activation levels during the corresponding epoch following the test stimulus on both remember-orientation trials and remember-color trials. Two-way repeated measures analyses of variance (ANOVAs) with delay type (WM vs. ITI) and time point (four levels, from 4 s through 10 s) as factors showed no reliable main effect of delay type, F(1, 6) = 0.4, p = .55, η2 = .06, and F(1, 6) = 0.27, p = .62, η2 = .04, respectively. We examined activation 4 s through 10 s poststimulus because this epoch should reflect activation associated with the retention of information in WM. No other retinotopically organized region showed sustained amplitude increases related to storing information in WM.1

Because comparing activation during the WM delay period with activation during a “passive” ITI following the presentation of the test stimulus is potentially problematic, we also plotted the mean time series for 24 s following the presentation of each sample stimulus (as in Offen et al., in press). Activation levels fell back to baseline approximately 8 s after the onset of the sample stimulus (see Figs. 2c and 2d), in contrast to the sustained amplitude increases often observed in parietal and frontal cortex across the delay period in a WM task (e.g., see Fig. 1 in D’Esposito, 2007). Thus, at least in our study, V1 did not exhibit a robust sustained amplitude increase that is often associated with WM maintenance.

It is important to note that the lack of amplitude changes in V1 does not rule out the possibility that stimulus-specific patterns of activation are involved in maintaining information in WM. Therefore, we used MVPA to determine if V1 exhibits a stimulus-specific activation pattern during the delay period, consistent with the predictions of the sensory-recruitment hypothesis. The MVPA approach is based on the assumption that some subregions of visual cortex contain submillimeter columns of neurons that are selective for different stimulus features, such as orientation. In comparison, fMRI voxels are large (~3 mm3); however, if slightly more neurons within a voxel prefer a particular orientation than prefer other orientations, then that voxel may exhibit a weak but detectable response bias (Kamitani & Tong, 2005). By examining the distributed voxel-by-voxel activation pattern across a visual area such as V1, one can make inferences about changes in the underlying population response profile, and pattern-classification algorithms can be used to predict the specific feature that an observer is viewing, attending, or (in our case) remembering (Haynes & Rees, 2005; Kamitani & Tong, 2005, 2006; Norman et al., 2006; Peelen & Downing, 2007; Serences & Boynton, 2007a, 2007b).

We therefore examined activation patterns in V1 during the delay period of the WM task (4–10 s following sample onset) to determine if information about the remembered feature was being actively represented. As predicted by the sensory-recruitment hypothesis, when observers were remembering the orientation of the sample stimulus, activation patterns in V1 discriminated stimulus orientation, but not stimulus color (see Fig. 3a). The complementary pattern was observed when observers were instructed to remember the color of the stimulus. A two-way repeated measures ANOVA with memory instruction (remember orientation vs. remember color) and stimulus feature (classify orientation vs. classify color) as factors yielded a significant interaction, F(1, 6) = 21.4, p < .005, η2 = .78. In contrast, activation patterns associated with a corresponding temporal epoch following the test stimulus—which was physically identical and required a challenging discrimination without storage—did not support above-chance classification accuracy (see Fig. 3b), F(1, 6) = 0.14, p = .72, η2 = .02, so that there was a three-way interaction of delay type (WM vs. ITI), memory instruction, and stimulus feature, F(1, 6) = 6.9, p < .05, η2 = .54. Given that the test stimulus evoked a BOLD response whose amplitude was statistically indistinguishable from that evoked by the sample stimulus (see Fig. 2), these data suggest that the active discrimination of the test stimulus for 1 s was not sufficient to drive above-chance classification accuracy. Thus, we conclude that the stimulus-specific pattern of delay activity in V1 was a direct consequence of active maintenance in WM. Activation patterns in other retinotopically organized visual areas did not consistently discriminate the remembered feature of the sample stimulus (see Table 2 for classification accuracies).

TABLE 2
Mean Classification Accuracy in Extrastriate Regions

The three-way interaction illustrated in Figure 3 was also significant when 80 V1 voxels were used to perform the classification, F(1, 6) = 6.9, p < .05, η2 = .53. Thus, a similar pattern of classification accuracy is observed even when more than 62 voxels are considered in the analysis. Qualitatively similar results were obtained for pattern sizes ranging from 40 to 100 voxels as well. When a linear discriminant classifier based on the Mahalanobis distance between activation patterns was used to compute classification accuracy, the three-way interaction was again significant, F(1, 6) = 10.8, p < .025, η2 = .64. Thus, the results were not idiosyncratically dependent on the use of an SVM. At first glance, it is striking that classification accuracy for orientation was below chance when participants were remembering stimulus color (see Fig. 3a). However, this effect was not robust across all activation pattern sizes, and we never observed below-chance color classification accuracy when participants were remembering orientation.

Together, these analyses suggest that sustained stimulus-specific patterns in V1 reflect active storage in WM, and are not a passive consequence of the attentive encoding of the sample stimulus. Although both the sample and the test stimuli required attentive processing, significant classification accuracy was obtained only during the WM delay period following the sample stimulus. These differences in classification accuracy cannot easily be explained by differences in general arousal or effort related to task demands, as the overall amplitude of the evoked BOLD response was roughly equivalent for the test and the sample stimuli. However, to provide additional support for sustained feature-selective modulations during the WM delay period, we repeated the classification analyses with and without data from the last time point in the delay period (i.e., 10 s after the onset of the sample; see Figs. 2a and 2b). If there were a sustained WM-related activation pattern, then data from the last time point in the delay period would contribute to classification accuracy. By contrast, if feature-selective activation patterns were not sustained across the entire delay period, then adding data from the last time point would not improve classification accuracy (and might even impair classification accuracy if the activation patterns associated with the last time point were dominated by noise).

As Figure 4 shows, classification of the remembered stimulus was significantly better when activation patterns associated with the last time point in the delay period were included. However, the addition of this last time point in the delay period did not alter classification accuracy for the nonremembered feature (see Fig. 4a). Thus, the mere addition of more data did not necessarily improve the performance of the pattern classifier. A two-way ANOVAwith memory instruction (orientation vs. color) and time bin (4 s through 8 s vs. 4 s through 10 s) as factors confirmed that adding information from the last time point in the delay period (10 s poststimulus) selectively enhanced classification accuracy for the remembered feature, F(1, 6) = 6.7, p < .05, η2 = .53. Furthermore, although including data from 10 s postsample clearly improved classification accuracy for the remembered stimulus feature, including data from 10 s posttest (at the end of the ITI) had little effect on classification of the test stimulus (see Fig. 4b). The two-way interaction of delay type and time bin was significant, F(1, 6) = 6.2, p < .05, η2 = .51.

Finally, we repeated the analysis after removing data from the peak of the stimulus-evoked BOLD response (4s; see Fig. 2) and used only data collected 6 s to 10 s poststimulus to classify the remembered feature attribute. Even when data from the peak were excluded, classification accuracy for the remembered feature was significantly higher than classification accuracy for the nonremembered feature, collapsed across remember-orientation and remember-color trials (.584 vs. .474), t(6) = 2.9, prep > .87. These control analyses support our conclusion that the interaction depicted in Figure 3a reflects the on-line maintenance of information in WM, rather than the aftereffects of a phasic sensory response.

Although our data suggest that maintaining information in WM gives rise to sustained feature-selective activation patterns in V1, a stronger prediction of the sensory-recruitment hypothesis is that the pattern of activation during the delay period will literally mimic the pattern of activation evoked during sensory processing of the same stimulus. To test this prediction, we trained a classification algorithm using data from the functional localizer scans that were initially used to identify visually responsive voxels in each subregion of occipital cortex (see Method). The stimuli used in the localizer scans were identical to those used in the WM study, except that they were presented continuously for 10 s instead of only 1 s, and WM was not required during the localizer tasks. Each subject completed two to four of these localizer scans (in half of the scans, they attended orientation and ignored color; in the other half, they attended color and ignored orientation). One SVM was trained using data from attend-orientation localizer scans, and another was trained using data from attend-color localizer scans. These SVMs were then used to predict the orientation or color that subjects were remembering on each trial during the main WM task. Collapsed across remember-orientation and remember-color trials, classification accuracy was .607 (SEM = .42), t(6) = 2.5, prep > .87. This above-chance classification accuracy demonstrates that the V1 activation pattern that is sustained during WM resembles the sensory-evoked response that is observed during sensory processing alone.

DISCUSSION

These results demonstrate that the maintenance of information in visual WM elicits stimulus-specific activation patterns in the same regions of visual cortex that encode the to-be-remembered sensory information. These activation patterns were specifically tied to the delay period, when active rehearsal in WM was required. Classification accuracy was not above chance following the test stimulus, which involved identical bottom-up stimulation and discrimination of the same stimulus dimension, but no WM load. In addition, the sustained activation patterns observed during the delay period were similar to patterns evoked by the continuous presentation of identical sensory stimuli, which suggests that early feature-selective visual areas are recruited to maintain a “copy” of remembered stimulus attributes, as opposed to a more abstract or categorical representation. In addition to providing these empirical results, this study demonstrates that MVPA is a valuable tool for answering questions about the neural mechanisms that mediate the storage of specific stimulus values in WM.

Finally, these findings are also relevant to the claim that the capacity of visual WM is determined by the number of individuated objects that have to be stored, rather than the total amount of visual detail contained within those items (Awh, Barton, & Vogel, 2007; Irwin, 1992; Luck & Vogel, 1997; Woodman & Vogel, 2008; Xu & Chun, 2006; Zhang & Luck, 2008). For example, Luck and Vogel (1997) showed that capacity estimates for objects defined by a single feature (e.g., color or orientation) were equivalent to capacity estimates for multifeatured objects (e.g., colored oriented lines). This suggests that capacity is determined by the number of objects that are stored, rather than by the total information load. Alternatively, other researchers have proposed that an obligatory set of core features, including attributes such as color and orientation, is maintained regardless of the observer’s intentions (Alvarez & Cavanagh, 2004). This hypothesis might explain Luck and Vogel’s observation of equivalent capacity estimates for single and multifeature objects if all the possible features were obligatorily stored even when only a single feature was relevant. However, our results (Fig. 3) reveal that early sensory areas selectively represent only behaviorally relevant features during a WM delay period and suggest that observers have top-down control over which features are stored (see also Olivers, Meijer, & Theeuwes, 2006; Woodman & Vogel, 2008).

Acknowledgments

This study was supported by start-up funds to J.T.S. from the University of California, Irvine; by a National Institutes of Health fellowship to E.F.E.; by National Science Foundation Grant 0617681 to E.K.V.; and by National Institutes of Health Grant R01MH077105-01A2 to E.A. J.T.S. and E.A. conceived the experiment; J.T.S., E.A., E.F.E., and E.K.V. designed the experimental paradigm; E.F.E. collected the data; J.T.S. and E.F.E. analyzed the data; and J.T.S., E.A., E.F.E., and E.K.V. wrote the article.

Footnotes

1No extrastriate area that we identified showed a heightened response while observers were remembering orientation or color (over a temporal window extending from 4 s through 10 s poststimulus, all ps > .2 for the main effect of WM). We do not rule out the possibility that some modest delay-period activity might be observed if enough subjects were scanned. However, our data do demonstrate that sustained WM-related modulations are relatively weak in primary and extrastriate visual areas (see also Offen et al., in press).

REFERENCES

  • Alvarez GA, Cavanagh P. The capacity of visual short-term memory is set both by visual information load and by number of objects. Psychological Science. 2004;15:106–111. [PubMed]
  • Awh E, Barton B, Vogel EK. Visual working memory represents a fixed number of items regardless of complexity. Psychological Science. 2007;18:622–628. [PubMed]
  • Awh E, Jonides J. Overlapping mechanisms of attention and spatial working memory. Trends in Cognitive Sciences. 2001;5:119–126. [PubMed]
  • Boynton GM, Engel SA, Glover GH, Heeger DJ. Linear systems analysis of functional magnetic resonance imaging in human V1. Journal of Neuroscience. 1996;16:4207–4221. [PubMed]
  • Chelazzi L, Miller EK, Duncan J, Desimone R. A neural basis for visual search in inferior temporal cortex. Nature. 1993;363:345–347. [PubMed]
  • Courtney SM, Ungerleider LG, Keil K, Haxby JV. Transient and sustained activity in a distributed neural system for human working memory. Nature. 1997;386:608–611. [PubMed]
  • D’Esposito M. From cognitive to neural models of working memory. Philosophical Transactions of the Royal Society B: Biological Sciences. 2007;362:761–772. [PMC free article] [PubMed]
  • Druzgal TJ, D’Esposito M. Activity in fusiform face area modulated as a function of working memory load. Cognitive Brain Research. 2001;10:355–364. [PubMed]
  • Engel SA, Rumelhart DE, Wandell BA, Lee AT, Glover GH, Chichilnisky EJ, Shadlen MN. fMRI of human visual cortex. Nature. 1994;369:525. [PubMed]
  • Haxby JV, Gobbini MI, Furey ML, Ishai A, Schouten JL, Pietrini P. Distributed and overlapping representation of faces and objects in ventral temporal cortex. Science. 2001;293:2425–2430. [PubMed]
  • Haynes JD, Rees G. Predicting the orientation of invisible stimuli from activity in human primary visual cortex. Nature Neuroscience. 2005;8:686–691. [PubMed]
  • Irwin DE. Memory for position and identity across eye movements. Journal of Experimental Psychology: Learning, Memory, and Cognition. 1992;18:307–317.
  • Johnson EN, Hawken MJ, Shapley R. The spatial transformation of color in the primary visual cortex of the macaque monkey. Nature Neuroscience. 2001;4:409–416. [PubMed]
  • Jonides J, Lacey SC, Nee DE. Processes of working memory in mind and brain. Current Directions in Psychological Science. 2005;14:2–5.
  • Kamitani Y, Tong F. Decoding the visual and subjective contents of the human brain. Nature Neuroscience. 2005;8:679–685. [PMC free article] [PubMed]
  • Kamitani Y, Tong F. Decoding seen and attended motion directions from activity in the human visual cortex. Current Biology. 2006;16:1096–1102. [PMC free article] [PubMed]
  • Lepsien J, Nobre AC. Attentional modulation of object representations in working memory. Cerebral Cortex. 2007;17:2072–2083. [PubMed]
  • Leventhal AG, Thompson KG, Liu D, Zhou Y, Ault SJ. Concomitant sensitivity to orientation, direction, and color of cells in layers 2, 3, and 4 of monkey striate cortex. Journal of Neuroscience. 1995;15:1808–1818. [PubMed]
  • Logothetis NK, Pauls J, Augath M, Trinath T, Oeltermann A. Neurophysiological investigation of the basis of the fMRI signal. Nature. 2001;412:150–157. [PubMed]
  • Luck SJ, Vogel EK. The capacity of visual working memory for features and conjunctions. Nature. 1997;390:279–281. [PubMed]
  • Martinez-Trujillo J, Treue S. Feature-based attention increases the selectivity of population responses in primate visual cortex. Current Biology. 2004;14:744–751. [PubMed]
  • Miller EK, Li L, Desimone R. Activity of neurons in anterior inferior temporal cortex during a short-term memory task. Journal of Neuroscience. 1993;13:1460–1478. [PubMed]
  • Norman KA, Polyn SM, Detre GJ, Haxby JV. Beyond mind-reading: Multi-voxel pattern analysis of fMRI data. Trends in Cognitive Sciences. 2006;10:424–430. [PubMed]
  • Offen S, Schluppeck D, Heeger DJ. The role of early visual cortex in visual short-term memory and visual attention. Vision Research. (in press) [PMC free article] [PubMed]
  • Olivers CNL, Meijer F, Theeuwes J. Feature-based memory-driven attentional capture: Visual working memory content affects visual attention. Journal of Experimental Psychology: Human Perception and Performance. 2006;32:1243–1265. [PubMed]
  • Paradiso MA. A theory for the use of visual orientation information which exploits the columnar structure of striate cortex. Biological Cybernetics. 1988;58:35–49. [PubMed]
  • Peelen MV, Downing PE. Using multi-voxel pattern analysis of fMRI data to interpret overlapping functional activations. Trends in Cognitive Sciences. 2007;11:4–5. [PubMed]
  • Postle BR. Working memory as an emergent property of the mind and brain. Neuroscience. 2006;139:23–38. [PMC free article] [PubMed]
  • Pouget A, Dayan P, Zemel RS. Inference and computation with population codes. Annual Review of Neuroscience. 2003;26:381–410. [PubMed]
  • Ranganath C, Cohen MX, Dam C, D’Esposito M. Inferior temporal, prefrontal, and hippocampal contributions to visual working memory maintenance and associative memory retrieval. Journal of Neuroscience. 2004;24:3917–3925. [PubMed]
  • Sanger TD. Probability density estimation for the interpretation of neural population codes. Journal of Neurophysiology. 1996;76:2790–2793. [PubMed]
  • Serences JT, Boynton GM. Feature-based attentional modulations in the absence of direct visual stimulation. Neuron. 2007a;55:301–312. [PubMed]
  • Serences JT, Boynton GM. The representation of behavioral choice for motion in human visual cortex. Journal of Neuroscience. 2007b;27:12893–12899. [PubMed]
  • Sereno MI, Dale AM, Reppas JB, Kwong KK, Belliveau JW, Brady TJ, et al. Borders of multiple visual areas in humans revealed by functional magnetic resonance imaging. Science. 1995;268:889–893. [PubMed]
  • Sincich LC, Horton JC. The circuitry of V1 and V2: Integration of color, form, and motion. Annual Review of Neuroscience. 2005;28:303–326. [PubMed]
  • Solomon SG, Lennie P. The machinery of colour vision. Nature Reviews Neuroscience. 2007;8:276–286. [PubMed]
  • Woodman GF, Vogel EK. Selective storage and maintenance of an object’s features in working memory. Psychological Bulletin & Review. 2008;15:223–229. [PubMed]
  • Xu Y, Chun MM. Dissociable neural mechanisms supporting visual short-term memory for objects. Nature. 2006;440:91–95. [PubMed]
  • Zhang W, Luck SJ. Discrete fixed-resolution representations in visual working memory. Nature. 2008;453:233–235. [PMC free article] [PubMed]
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

  • PubMed
    PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...