Frontoparietal and Cingulo-opercular Networks Play Dissociable Roles in Control of

George Wallis, Mark Stokes, Helena Cousijn, Mark Woolrich, and Anna Christina Nobre

Abstract ■ We used magnetoencephalography to characterize the spatio- cues this activation was transient and was succeeded by a cingulo- temporal dynamics of cortical activity during top–down control of opercular network activation. We also characterized the time working memory (WM). fMRI studies have previously implicated course of top–down modulation of alpha activity in visual/parietal both the frontoparietal and cingulo-opercular networks in control cortex. This modulation was transient following retrocues, occur- over WM, but their respective contributions are unclear. In our ring in parallel with the activation. We sug- task, spatial cues indicating the relevant item in a WM array oc- gest that the frontoparietal network is responsible for top–down curred either before the memory array or during the maintenance modulation of activity in sensory cortex during both preparatory period, providing a direct comparison between prospective and attention and orienting within memory. In contrast, the cingulo- retrospective control of WM. We found that in both cases a fron- opercular network plays a more downstream role in cognitive toparietal network activated following the cue, but following retro- control, perhaps associated with output gating of memory. ■

INTRODUCTION suggest that whether control is prospective and selects Performance in working memory (WM) tasks is strongly from perceptual input or retrospective, selecting from modulated by selection cues that allow people to priori- within WM results in the following: (1) substantially over- tize the most task-relevant item. Sperling (1960) demon- lapping frontoparietal control regions are recruited (Nee strated that cues picking out the task-relevant items given & Jonides, 2009; Nobre et al., 2004) and (2) activity in before or immediately following the presentation of a sensory cortex is modulated retinotopically (Kuo, Stokes, WM array (when items were still available in the iconic Murray, & Nobre, 2014; Munneke, Belopolsky, & Theeuwes, buffer) improved WM performance. This effect has a ready 2012; Sligte, Scholte, & Lamme, 2009; Mangun, Hopfinger, & interpretation in terms of gating of encoding to a capacity- Buonocore, 2000). This fits well with a model of WM in limitedWMstore,reducingmemoryloadtooneitem. which persistent activity in perceptual cortex is responsible However, more surprisingly, cueing the task-relevant item for maintaining WM representations (Harrison & Tong, during the memory retention interval, termed “retrocue- 2009; Pasternak & Greenlee, 2005): Top–down influences ing” (Sligte, Scholte, & Lamme, 2008; Nobre et al., 2004; bias perceptual activity during WM retention in the same Griffin & Nobre, 2003; Landman, Spekreijse, & Lamme, way as it is biased by perceptual attention. However, this 2003), improves response accuracy almost as much. This model may be too simple. First, recent work suggests that finding challenges the assumption that WM performance not all items in WM are associated with a persistent activity is only limited by storage capacity. Retrieval mechanisms state in sensory cortex (LaRocque, Lewis-Peacock, & Postle, or “output gating” (Chatham, Frank, & Badre, 2014) may 2014). Second, retrocues recruit additional sites in frontal be just as important a determinant of performance as “in- cortex besides the frontoparietal network. In particular, put gating” of WM. Although precues are thought to permit Higo, Mars, Boorman, Buch, and Rushworth (2011) and, input gating of WM, retrocues may facilitate output gating. more recently, Nelissen, Stokes, Nobre, and Rushworth Our study investigated the different patterns of involve- (2013) have suggested that the dorsomedial pFC and the ment of control networks in these different forms of top– anterior insula or frontal operculum [fO] (hereafter we down control of WM. use the term frontal operculum for clarity) may be the crit- One hypothesis holds that a shared top–down atten- ical sites for top–down control over WM representations— tional mechanism acts both on perceptual input and on as opposed to frontoparietal sites. These areas are nodes WM representations (Gazzaley & Nobre, 2012) to medi- of a second “cingulo-opercular” control network, which is ate both precue and retrocue benefits. fMRI experiments distinct from the frontoparietal network classically associ- ated with top–down control (Petersen & Posner, 2012; Dosenbach et al., 2007). The additional recruitment of University of Oxford the cingulo-opercular network by retrocues, as compared

© 2015 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 27:10, pp. 2019–2034 doi:10.1162/jocn_a_00838 to precues, is an intriguing clue to how its functional Figure 1A is a task schematic. Four memory items were role may differ from and complement the frontoparietal presented. Following a delay period of 3310 msec, a sin- network. gle probe item appeared. The probe was identical to one In this study, we used a precision/capacity (Zhang & of the items from the memory array but rotated about Luck, 2008) visual WM task giving precues (Murray, the circular end body, through 5°, 15°, or 45° clockwise Nobre, & Stokes, 2011; Sperling, 1960) and retrocues or anticlockwise (equal probability). Participants re- (Murray, Nobre, Clark, Cravo, & Stokes, 2013; Landman sponded with a right-hand button press; the left button et al., 2003) on different trials to compare prospective (index finger) indicated an anticlockwise rotation, the and retrospective control of WM. We first reviewed prior right button (middle finger) clockwise. On one-third of fMRI work using a meta-analysis technique to confirm the trials, a spatial precue 1540 msec before array onset indi- different spatial patterns of cortical recruitment between cated the relevant memory item. On a separate one- precues and retrocues and obtain a priori spatial ROIs. third of trials, there was a spatial retrocue 1540 msec into We then used magnetoencephalography (MEG), an imag- the retention interval. On remaining trials (“neutral-cue ing modality with high temporal resolution, to character- trials”), no information was given about which item ize the pattern of cortical activity in these areas after both would be probed. Trial types were randomly interleaved. precues and retrocues. Our time-resolved data reveal the Cues were always valid. The cue consisted of a white and temporal relationship between activation in these two a black arrow formed by the sides of a square (Figure 1A). networks and top–down modulation of activity in sensory For half of the participants, the black arrow indicated the cortex. They show that both precues and retrocues sim- cued quadrant, and for half of participants, the white ilarly recruit the frontoparietal network, but retrocues ad- arrow indicated the cued quadrant, controlling for any ditionally recruit the cingulo-opercular network at a later effects driven by the physical properties of the cue. time point. Participants were asked to maintain central fixation, and eye movements were monitored using an infrared binocular eyetracker (Eyelink 1000, SR Research, Ottowa Canada). During the training session, stimuli were pre- METHODS sented on an LCD screen (Samsung 2233Rz, Samsung, Seoul) at a viewing distance of 80 cm. During the MEG Participants and Behavioral Task session, stimuli were back-projected (Panasonic PT Fifty volunteers were recruited (26 women, 24 men; D7700E, Panasonic, Osaka Japan) onto a screen at a view- mean age = 24 years, range = 19–34). All participants ing distance of 85 cm. In both sessions, the WM stimuli were healthy, had normal or corrected-to-normal vision, were presented at an eccentricity 6° visual angle from the and were right-handed. Ethical approval was obtained fixation cross, and each stimulus subtended 1.2°. from the National Health Service South Central Berkshire To reduce noise from trials in which the participants ethics committee (11/SC/0053). One participant per- were settling into the task, the first 50 trials were dis- formed at chance in the behavioral task, and one was un- carded from the training block, as well as the first 10 trials able to complete the MEG session. Forty-eight full data from the MEG session (including these trials does not sets were therefore available for behavioral analysis. Fur- substantively alter any aspect of the results). Any trials ther participants were excluded from the MEG analysis. in which the RT was shorter than 0.1 sec or longer than Four made eye movements larger than 2° horizontal dis- 10 sec (the median RT was 1.1 sec) were discarded from placement during the WM retention intervals (as mea- the analysis, as they likely represented anticipation re- sured on the basis of the eye-tracker signal), potentially sponses or a lapse in task engagement, respectively. contaminating the MEG signal. Six participants were re- For most participants (40 of 48), five or fewer trials were jected because their MEG data were of poor quality: excluded on the basis of RT. The maximum number of For two participants, the spatial coregistration of the trials rejected was 20. MEG forward model with the signal space failed, and A mixture model was fitted for the accuracy data for four participants, the signal was heavily contaminated (Zhang & Luck, 2008) in which responses across trials with electronic artifacts. This left us with MEG data from are assumed to come from one of two distributions: a 38 participants. The behavioral task was programmed in uniform ‘guess’ distribution, representing trials in which Matlab and PsychToolbox (Pelli, 1997). Behavioral and participants had no information about the probed item MEG analyses were performed using custom-written and responded at random, and a Von Mises distribution Matlab software, SPM8, FSL, Fieldtrip, and the in-house (circular analogue of a Gaussian) that represents the OHBA Software Library. fidelity of the WM representation on trials when partici- We used a four-item WM task with predictive cues pants were not guessing. We checked for response bias to- (precues) and retrodictive cues (retrocues). The behav- ward “clockwise” or “anticlockwise” responses by ioral task was run in two separate sessions: The training taking the difference of net clockwise and anticlockwise re- session contained 6 × 36 trial blocks (216 trials), and the sponses across all trials for each cue condition. For neutral MEG session contained 9 blocks (324 trials). trials, there was a tendency for participants to respond

2020 Journal of Cognitive Neuroscience Volume 27, Number 10 Figure 1. Task schematic and behavioral data. (A) Precue/ retrocue WM task. The three trial types are randomly interleaved, each making up one-third of trials overall. In this case, the white arrow indicated the cued direction. For half of participants, this was reversed, and the black arrow indicated the cued direction. (B) Accuracy data. The proportion of clockwise responses is plotted for each orientation change of the probe item (x-axis, clockwise positive). Both precues (blue) and retrocues (red) improve performance at all orientation change magnitudes. (C) RT distributions for each trial type, binned into quintiles. Precues and retrocues both speed responses. (D, E) Mixture model analysis. Precues and retrocues substantially reduce the proportion of trials on which participants are guessing (D). Precues substantially increase the precision with which items are represented in memory but retrocues have only a modest effect on precision (E). (F) Misgating analysis. If there were no effect of nontarget items on responses, this plot would show a flat line. Nontarget items influenced responding most strongly in the neutral-cue condition. Error bars represent ±1 SEM.

“anticlockwise” ( p(clockwise) − p(anticlockwise) = ticlockwise orientation changes were collapsed together −0.12, p = .001). This was also significant for precue for modeling, factoring out the response bias. The prob- trials (−0.077, p = .042) and present as a trend for retro- ability that a participant made the correct response on a cue trials (−0.066, p = .08). This suggested that, when given trial depends on k, the precision of the Von Mises dis- guessing, participants were slightly more likely to press tribution (higher values indicate a tighter distribution); the index-finger than middle-finger button. Clockwise/an- pGuess, the probability that participants are guessing on

Wallis et al. 2021 any particular trial; and θ, the size of the orientation change. the probability of responding clockwise was calculated. Note This is captured in Equation 1. that, for any given bin, the probability of the probe item having been clockwise or anticlockwise of the target item ðÞ¼ðÞþ : ðÞ− pcorrect pGuess 0 5 1 pGuess was always the same (.5), so if there were no influence of ðÞ; θ vonmisescdf k (1) nontarget items, the probability of responding “clockwise” vonmisescdf is the cumulative density function of the should be equal across all bins. Von Mises distribution, which can be evaluated numerically. The model was fitted for each participant and each condition Meta-analysis of fMRI Data/Derivation of MEG ROIs separately using a maximum likelihood approach to get values of pGuess and k (Figure 1D, E). We performed meta-analyses of spatial precueing and ret- The mixture model analysis assumes that nontarget rocueing fMRI studies to guide analysis of MEG data items have no effect on responding, but previous work using the freely available GingerAle software package using similar multi-item WM tasks suggests that nontarget (Eickhoff et al., 2009). items can sometimes influence behavior (Bays, Gorgoraptis, Separate meta-analyses were conducted for precue and Wee, Marshall, & Husain, 2011; Bays, Catalao, & Husain, retrocue experiments. Imaging studies were searched 2009). In the current task, the orientation of each mem- using Pubmed and Google Scholar. Retrocue studies ory item was assigned randomly and independently of were included if they involved a late cue (>1 sec after the orientation of the other items. The effect of non- the memory array) to focus on an item already in mem- target items could therefore be evaluated by running a ory, and we used data from contrasts that isolated the similar analysis to that for the target item, examining brain response to this cue. Precue studies were restricted the probability of responding “clockwise” or “anticlock- to studies cueing spatial attention to isolate activations wise,” depending on whether the probe item was rotated associated with spatial orienting. We included only those clockwise or anticlockwise of each nontarget item. Be- studies that could differentiate responses to the cue from cause the target and nontarget orientations were ran- subsequent responses to the cued targets. Seventeen domly and independently assigned, the orientation studies were included in the retrocue meta-analysis, and difference between the probe item and a given nontarget eight studies were included in the precue meta-analysis. item spanned the full circle between −180° and 180°. This These studies are listed in Tables 1 and 2. Local maxima range was divided into eight bins: clockwise/anticlockwise were extracted from the resulting meta-analysis activation 0°–45°, 45°–90°, 90°–135°, and 135°–180°, and for each bin maps and are listed in Tables 3 and 4.

Table 1. Studies Included in the Retrocue Meta-analysis

Study Study No. No. of Participants Rowe, Toni, Josephs, Frackowiak, and Passingham (2000) 1 6 Rowe and Passingham (2001) 2 6 Raye, Johnson, Mitchell, Reeder, and Greene (2002) 3 12 Johnson, Raye, Mitchell, Greene, and Anderson (2003) 4 14 Nobre et al. (2004) 5 10 Johnson, Mitchell, Raye, and Greene (2004) 6 14 Lepsien, Griffin, Devlin, and Nobre (2005) 7 10 Johnson et al. (2005) 8 14 Lepsien and Nobre (2006) 9 14 Yeh, Kuo, and Liu (2007) 10 10 Johnson, Mitchell, Raye, D'Esposito, and Johnson (2007) 11 15 Yi, Turk-Browne, Chun, and Johnson (2008) 12 8 Raye, Mitchell, Reeder, Greene, and Johnson (2008) 13 29 Johnson and Johnson (2009) 14 14 Nee and Jonides (2009) 15 18 Roth, Johnson, Raye, and Constable (2009) 16 22 Higo et al. (2011) 17 21

2022 Journal of Cognitive Neuroscience Volume 27, Number 10 Table 2. Studies Included in the Precue Meta-analysis

Study Study No. No. of Participants Mangun et al. (2000) 1 6 Corbetta, Kincade, Ollinger, McAvoy, and Shulman (2000) 2 13 Giesbrecht, Woldorff, Song, and Mangun (2003) 3 10 Nobre et al. (2004) 4 10 Woldorff et al. (2004) 5 20 Wilson, Woldorff, and Mangun (2005) 6 16 de Haan, Morgan, and Rorden (2008) 7 12 Egner et al. (2008) 8 14

To simplify this spatial pattern into a single set of ROIs averaged activations common to the precue and retrocue for MEG analysis (Table 5), with a spatial sampling appro- meta-analyses (using a distance threshold of 16 mm). priate to the comparatively coarse spatial resolution of This set of combined coordinates was converted into a the method (as compared with fMRI), we converted symmetric set of ROIs by averaging corresponding left these local maxima into a left/right symmetric spatially and right hemisphere ROI locations and mirroring the sparse set of brain locations by averaging local maxima activations that occurred in the left hemisphere only (mid- for each cluster within each meta-analysis map. We then dle temporal gyrus [MTG], TPJ) in the right hemisphere.

Table 3. Activation Clusters from the Retrocue Meta-analysis

Local Maxima (MNI Coordinates)

Label Cluster No. xyzCluster Volume (mm3) Right anterior MFG 1 44 46 24 1696 34 50 16 Left anterior MFG 2 −40 36 28 944 Left precentral / Left anterior MFG 3 −50 −2 40 5040 −50 22 28 −54 12 34 Right anterior insula (fO) 4 48 12 −4 2056 Left anterior insula (fO) 5 −40 14 −4 1304 dACC/pre-SMA 6 0 18 48 3272 −23036 Right precentral 7 48 6 42 440 Left MTG 8 −58 −38 0 824 −66 −42 0 Left TPJ 9 −56 −38 32 480 −48 −38 32 Left IPS 10 −38 −48 44 488 Right SPL 11 12 −70 52 1400 18 −62 54 Right IPS 12 34 −58 42 1048 38 −44 44 30 −64 38

Wallis et al. 2023 Table 4. Activation Clusters from the Precue Meta-analysis

Local Maxima (MNI Coordinates)

Label Cluster # xyzCluster Volume (mm3) Left anterior MFG 1 −40 30 24 472 Right FEF 2 32 0 50 1560 Left precentral/left FEF 3 −36 −4 42 464 −42 −452 Left FEF 4 −22 −2 50 512 −22 −858 Left IPS 5 −40 −48 36 544 Right SPL 6 26 −56 58 1112 Left IPS 7 −22 −66 54 3960 −18 −58 54 −22 −68 40 Right SPL 8 22 −72 50 768 28 −68 44 Right IPS0/ V7 9 34 −78 26 584 34 −74 26 Left IPS0/ V7 10 −28 −78 30 400 Left occipital 11 −46 −70 −10 800 Right occipital 12 34 −80 14 800 34 −84 14

MRI Scan head coil was used to obtain a T1-weighted anatomical image with 224 (1-mm) slices. This anatomical image A structural MRI was acquired for each participant using a was used to define the single-shell MEG forward model Siemens 3T scanner (OCMR, Oxford, UK). A 32-channel (Nolte,2003).SPM’s spm_eeg_inv_mesh was used to compute the transformation that mapped a set of canon- Table 5. ROIs Derived from the fMRI Meta-analysis ical meshes for the cortical surface, skull, and scalp to each participant’s individual anatomical MRI, and this MNI Coordinates transformation was used to define a MEG forward model ’ xyztailored to each participant s head shape, computed using Fieldtrip’s forward toolbox (Donders Institute, IPS0 ±34 −76 26 Nijmegen, The Netherlands; shared with SPM). Spatial Mid-IPS ±30 −68 40 coregistration between the forward model and MEG space was by aligning the spaces based on anatomical − Anterior IPS ±39 48 40 landmarks (nasion and left/right preauricular points) for SPL ±12 −68 60 a first estimate and then refining this fit using points FEF ±27 −352recorded from the scalp surface (see MEG scan), which were matched to the scalp surface mesh using an itera- Precentral cortex (or iFEF) ±46 1 43 tive closest point algorithm as implemented in SPM. Anterior MFG ±40 39 23 TPJ ±52 −38 32 MEG Scan − MTG ±62 40 0 MEG data were acquired using an Elekta Neuromag 306- fO ±44 13 −4 channel system (Elekta, Stockholm, Sweden; 204 planar gradiometers, 102 magnetometers). The MEG suite is Pre-SMA 0 24 42 passively shielded. ECG and vertical/horizontal EOG

2024 Journal of Cognitive Neuroscience Volume 27, Number 10 were recorded. Head position was continuously moni- Sensor Space Classification Analysis of tored using emitting coils affixed to the participant’s Alpha Lateralization scalp and a Polhemus 3D tracking system (Polhemus We used a simple correlation-based pattern classification EastTrach 3D, Polhemus, Vermont, Unites States). Ana- approach to test whether the pattern of decrease in alpha tomical landmarks (nasion and left/right preauricular power to an attentional cue resembled the pattern of points) were recorded, as well as ∼100 points spread event-related alpha desynchronization (ERD) in response out over the scalp surface. MEG data were recorded in to a physical stimulus. We averaged the induced re- three blocks of ∼15 min each. sponses to the probe stimulus between 300 and 500 msec (the time window for which within-epoch alpha-band MEG Preprocessing quadrant classification for the probe item was highest), yielding four sensor space patterns for trials in which MEG data were initially inspected to remove any chan- each of the four visual quadrants was probed. These nels severely corrupted by noise and then de-noised probe ERD topographies were then correlated against and corrected for head movements using Elekta’s Maxfilter the trial-wise cue-induced topographies from the precue Signal Space Separation algorithm (Taulu, Kajola, & and retrocue epochs. The trials were classified to a quad- Simola, 2004). Data were epoched and visually inspected rant based on whichever probe stimulus quadrant pat- again using Fieldtrip’s visual artifact rejection tool for stan- tern was most correlated with the pattern of activity on dard artifacts: Contaminated trials and channels were that trial. A leave-one-out approach was used to prevent tagged on the basis of abnormal variance, kurtosis, and cross-temporal correlations within trials from confound- maxima/minima in the time domain data. Eyeblinks were ing the analysis: The to-be-classified trial was always ex- detected from the EOG, and eye-tracker data using a semi- cluded from the averages of probe stimulus activity. The automatic algorithm and data from 200 msec before and classification results were averaged over time–frequency until 300 msec after each blink were excluded from all bins in the alpha band (8–12 Hz) and the interval 0.4– analyses, including estimation of the beamformer weights. 0.8 sec after the cue for plotting (Figure 2C, D).

MEG Analysis Cross-temporal Correlation Analysis Data were cut into three epochs for analysis: precue ep- To establish whether the patterns of brain activity we och, array epoch, and retrocue epoch (see Figure 1A). observed were stable or transient, we correlated cue- Key experimental contrasts compared (1) activity in trials induced brain states across time, as indexed by the in which there was a precue or retrocue with neutral-cue induced response topographies. We randomly subdi- “ ” trials ( cue effects ) and (2) activity in leftward-cued trials vided all of the experimental trials into two groups of “ ”— with activity in rightward-cued trials ( cue laterality to equal size (discarding trials if more trials had survived measure alpha-power lateralization in ). preprocessing in one condition than another). Within each of the two halves, data were averaged within condi- tions, and the cue effects contrast was computed. This Sensor Space Analysis of Alpha yielded two independent estimates of the cue effects Power Lateralization topography. The time domain sensor space signal was transformed to We performed this analysis for the theta (3–7 Hz), al- the frequency domain in 50-msec steps, using a Hanning pha (8–12 Hz), and beta (18–30 Hz) bands separately. As taper/FFT algorithm with a taper spanning four cycles of in previous analyses (Stokes et al., 2013), one half of the the filtered frequency, for frequencies between 3 and data was designated the training data, and the other half 30 Hz in 1-Hz steps. The resulting power spectra were av- the test data. The topographies were extracted from the eraged over trials within each cue condition. The power training data for each time point in the epoch. This to- time series in the planar gradiometer pairs were then pography was correlated with the topography at every combined (Cartesian sum), giving a 102-channel com- time point in the test data, building up a cross-temporal bined planar gradiometer map of sensor space power. correlation matrix (King & Dehaene, 2014). If a state is The cue laterality subtractions [precue left minus precue transient, then it gives rise to high correlation values right] and [retrocue left minus retrocue right] were com- mainly on the diagonal of this matrix. By contrast, tempo- puted per participant for the precue and retrocue rally stable states will give rise to high correlation values epochs, respectively. Sensor space cluster permutation confined to the diagonal of this matrix. The data were statistics (Maris & Oostenveld, 2007) were computed split randomly, so each time this analysis is run, a slightly for these topographies by permuting cue left/cue right different result will be obtained. We therefore performed condition labels (using Fieldtrip’s ft_freqstatistics). Clus- this analysis 20 times and averaged the results; this boot- ters were formed in space (sensor proximity) and time, strapping procedure stabilizes the estimate of the cor- averaging over the alpha (8–12 Hz) band. relation structure. We statistically evaluated the strength

Wallis et al. 2025 Figure 2. Modulation of alpha- band (8–12 Hz) power in visual cortex. (A, B) Sensor space topography of alpha power for the contrast [cue left minus cue right], at 0.6 sec following the precue (200 msec FWHM) (A) and retrocue (B). Sensors belonging to significant clusters (see Results) are circled (black, positive clusters; white, negative clusters). (C, D) Classifying the direction of attention by correlating the cue- induced topography with the topography of the induced response to the probe item. Classification percentages are shown relative to the cued quadrant, averaged over the 8– 12 Hz band, from 0.4 to 0.8 sec postcue. E, F show an alpha lateralization index calculated for the IPS0 virtual electrode. Alpha lateralization is persistent following precues, until presentation of the memory array. Lateralization subsides for most of the maintenance interval but ramps up just before presentation of the probe item. Following retrocues, alpha lateralization is transient, returning to near- baseline level by 1 sec postcue.

of the correlations at the group level by forming clusters whole brain-induced response analyses for the theta in the time/time correlation space and then tested these band (3–7Hz),alphaband(8–12 Hz), and beta band against a permutation distribution of cluster size. (18–30 Hz). The 4-D spatiotemporal map for each analy- sis epoch (precue, retrocue) was averaged over succes- sive 300-msec windows. Cluster permutation statistics Source Space Analyses for the 3-D maps were computed for the informative cue versus neutral cue contrasts with a cluster-forming To characterize the time course of activation in each ROI, threshold of 3 (t statistic). a virtual electrode was created for each ROI coordinate using a linearly constrained minimum variance beamfor- mer (Woolrich, Hunt, Groves, & Barnes, 2011). The RESULTS time–frequency representation of the data was then com- puted at each virtual electrode, between 3 and 30 Hz. The Behavioral Data time–frequency data were averaged across task conditions Group level accuracy data are shown in Figure 1B: Both pre- within participants. The condition averages (cue effects cues and retrocues improved response accuracy, increasing and cue laterality) were then subtracted within partici- the proportion of correct responses across all magnitudes pants. These contrasts were averaged at the group level of orientation change. Group level parameters for the mix- to create time–frequency maps of cue-related activity for ture model analysis described in Methods are plotted in each ROI. Significance testing was performed by forming Figure 1D, E. Precues and retrocues both reduced guess clusters in the time/frequency space for each ROI, for pos- rate, but only precues substantially increased precision. A itive and negative deviations in power, and then testing repeated-measures ANOVA with factor Cue condition against a permutation distribution of cluster size. (3 levels: neutral, precue, retrocue) found evidence for a To visualize spatial patterns of activation over the main effect of Cue condition upon guess rate (F(1.74, whole brain and verify that these were appropriately sam- 82.3) = 203.1, p < .0005). Paired sample t tests against the pled by the ROIs derived from the meta-analysis, we ran neutral condition confirmed that guess rate was significantly

2026 Journal of Cognitive Neuroscience Volume 27, Number 10 reduced in the precue condition (t(47) = 16.95, p <.0005) Behavioral analyses performed over the subset of 38 and retrocue condition (t(47) = 14.55, p <.0005). participants for whom we were able to analyze the For the precision parameter, there was also a main ef- MEG data were qualitatively the same as for the set of fect of Cue condition (F(2, 94) = 17.2, p < .0005). Paired- 48 participants who successfully completed the behav- sample t tests indicated that the precision in the precue ioral task, except that the marginally significant effect of condition was significantly higher than in the neutral con- retrocues upon precision (kappa) did not reach signifi- dition (t(47) = 5.63, p < .0005), but the increase in pre- cance (t(37) = −1.58, p = .12). cision in the retrocue condition was only marginally significant (t(47) = 2.04, p = .047). Precision in the pre- Alpha Power Modulation in Perceptual and Parietal cue condition was significantly higher than precision in Cortex Indexes the Allocation of Attention the retrocue condition (t(47) = 3.56, p = .001). We also tested whether nontarget items affected be- Alpha power in perceptual and parietal cortex is robustly havior. This has previously been termed “misbinding” modulated by preparatory attention (van Ede, Köster, & (Bays et al., 2011) but could also be characterized as “mis- Maris, 2012; Haegens, Händel, & Jensen, 2011; Siegel, gating,” if as argued here, the effect is attributable to se- Donner, Oostenveld, Fries, & Engel, 2008; Worden, Foxe, lecting the wrong item in memory to guide behavior. We Wang, & Simpson, 2000). Typically, when attention is di- performed a repeated-measures ANOVA with factors Ori- rected to one side of space, there is a relative increase in entation bin (8 levels corresponding to the orientation alpha power in the ipsilateral cortex and decrease in alpha bins described in Methods) and Cue condition (3 levels: power in the contralateral cortex, compared with when neutral, precue and retrocue). There was a main effect of attention is directed to the other hemifield. We hypoth- Orientation bin (F(7, 329) = 13.65, p < .0005) confirm- esized that the pattern of alpha activity in visual and pa- ing that the nontarget items affected responding, and rietal cortex would be similarly modulated by precues there was an interaction between Orientation bin and and retrocues, as both cue types may recruit a common Cue condition (F(14, 658) = 4.11, p < .0005) indicating top–down mechanism. We first performed a sensor space the degree of misgating differed between the cue condi- analysis subtracting the pattern of activation in trials in tions. These data are shown in Figure 1F. which a quadrant in the right hemifield was cued from To quantify the differences between cue conditions the pattern of activation when a quadrant in the left hemi- contributing to this interaction, we ran repeated-measures field was cued. The results (for a representative time point ANOVAs comparing pairs of conditions (i.e., as described 0.6 sec postcue) are plotted in Figure 2A (precues) and B above, but with cue condition levels (1) precue, retrocue; (retrocues). Both precues and retrocues robustly lateralized (2) precue, neutral cue; (3) retrocue, neutral cue). There alpha activity. was no evidence for a Cue condition × Orientation bin Cluster permutation tests revealed a significant cluster interaction for (1) precue versus retrocue (F(7, 329) = of sensors with increased power over the left occipito- 0.598, p = .757), but there was a significant Cue condi- parietal sensors for the contrast precue left minus precue tion × Orientation bin interaction for (2) precue versus right ( p = .0005). The cluster of sensors with decreased neutral cue (F(7, 329) = 5.71, p < .0005) and for (3) ret- power over the right hemisphere sensors did not reach rocue and neutral cue (F(7, 329) = 5.87, p < .0005). significance ( p = .10). For the retrocue left minus retro- Therefore, both precues and retrocues significantly re- cue right contrast, there was both a significant cluster of duced the propensity to respond on the basis of a non- sensors with increased powercenteredovertheleft target item. hemisphere occipito-parietal sensors ( p = .0065) and a Relative to the neutral condition, precues and retro- significant cluster of sensors with decreased power over cues also reduced RTs. In Figure 1C, RT distributions the right hemisphere sensors ( p = .001). are expressed in terms of quintile means: Precues and The source space time course for the laterality contrast retrocues were associated with a strikingly similar RT dis- was extracted at occipito-parietal ROIs and converted in- tribution, with a reduction in RT relative to the neutral to a lateralization index by flipping the sign of the de- condition that was present across all quintiles (mean re- crease in alpha power in the right hemisphere and duction in RT for precues, 422 msec, SEM =34msec; adding it to the increase in alpha power in the left. Later- retrocues, 432 msec, SEM = 36 msec). We compared me- alization was strongest in the IPS0 ROI, for which precue dian RTs between the cue conditions using a repeated- and retrocue time courses are shown in Figure 2E and F. measures ANOVA with factor Cue condition (3 levels). Precues gave rise to a more sustained alpha lateralization There was a significant main effect of Cue condition (F(2, lasting from ∼0.5 sec postcue until the presentation of 94) = 136.56, p < .0005). We then compared the condi- the memory array, whereas retrocues gave rise to a more tions using paired-sample t tests. There was a significant transient lateralization of alpha power between 0.5 and difference between median RT for neutral and precue tri- 1 sec postcue that returned to baseline before the probe − als ( p =5.7×10 7) and neutral and retrocue trials ( p = stimulus appeared. Note the very similar time course of 2.9 × 10−7), but no significant difference between precue decrease in alpha power in IPS for the cue effects con- and retrocue trials ( p = .90). trast (Figure 5A).

Wallis et al. 2027 To establish whether the alpha-power changes were separately for the theta (3–7 Hz), alpha (8–12 Hz), and quadrant-specific and resembled an ERD to a physical beta (18–30 Hz) bands (Figure 3B). stimulus, we classified the cue-induced topographies The cross-temporal correlation matrix shows that, after based on the ERD to the probe item. The results are precues, a stable state emerges from ∼0.6 sec postcue shown in Figure 2C and D, in which the classification pat- until the presentation of the memory array (Figure 3B, tern is shown averaged over time–frequency bins in the “square” in upper right for precues). Retrocues gave rise alpha band (8–12 Hz) between 0.4 and 0.8 sec following to a different pattern. Cue-induced activity following a precues and retrocues, coded relative to the cued quad- retrocue continues to evolve throughout the analysis ep- rant. We tested the quadrant specificity of this effect sta- och. This overview of the temporal structure of the cue- tistically by separately comparing up–down classification induced responses can be compared with the source and left–right classification and testing resulting time– space analysis of induced responses shown in Figure 5. frequency clusters of above-chance classification against a permutation distribution of cluster sizes under the null fMRI Meta-analysis hypothesis. Quadrant classification was significant for both precues (up/down cluster, p = .029; left/right clus- The results of the meta-analyses are given in Tables 3 and 4, ter, p = .0002) and retrocues (up/down cluster, p < and the ROIs derived from the meta-analysis results are .0002, left/right cluster, p < .0002). shown in Figure 4. There was overlap between precue- We also tested for an induced-response analogue to and retrocue-associated activations in the anterior middle contralateral delay activity (Vogel & Machizawa, 2004) frontal gyrus (MFG; or dlPFC) and (IPS). in which there is a sustained ERP lateralization during Both cue types also activated regions in the precentral gy- the retention interval following lateralized encoding by rus. Comparing the meta-analysis results, there was a disso- classifying cued quadrant in the WM retention interval ciation between cue types, in that precue studies reported in precue trials. For comparison, Sauseng et al. (2009) activity in the FEF whereas retrocue studies reported reported alpha lateralizationinthedelayintervalofa activity in a more inferior region we call iFEF (adopting task similar to that used by Vogel and colleagues. We Derrfuss’ terminology; Derrfuss, Vogt, Fiebach, von found a transient effect in the alpha band (8–12 Hz) Cramon, & Tittgemeyer, 2012). This is consistent with soon after the memory array, significant only for the the previous study by Nee and Jonides (2009), which iden- left/right decoding ( p = .0062) between 0.5 and 0.7 sec tified this dissociation. iFEF is in proximity to IFJ, and the following the array onset (compare with the timing of retrocue meta-analysis activation cluster encompassed peak lateralization following retrocues). We also found the coordinates of both regions. Kastner and colleagues a stronger effect (Figure 2E) that emerged in the run-up (2007) have also identified these two dissociable precen- (<1 sec before) to the probe stimulus (up/down decod- tral regions. ing, p = .0022; left/right decoding, p = .0002). Our re- Retrocues additionally activated the bilateral fO and sults therefore partly replicate those of Sauseng et al. the pre-SMA. The left posterior MTG and the left inferior (2009), but the longer retention interval in our experi- parietal lobule were also activated following retrocues. ment (3000 msec as compared to 900 msec in the pre- The meta-analysis results were consistent with previ- vious study) revealed that alpha lateralization in the ously described frontoparietal and cingulo-opercular con- retention interval manifested over short intervals when trol networks (Petersen & Posner, 2012; Dosenbach attention was most likely to be lateralized, consistent et al., 2007). Precues and retrocues both activated the with the transient lateralization observed following frontoparietal network, and retrocues additionally acti- retrocues. vated the cingulo-opercular network.

Cross-temporal Classification Analysis Induced Responses in Control Networks To picture the overall temporal dynamics of brain activity The ROIs derived from the fMRI meta-analysis were used to in response to a precue or retrocue without first spatially extract source space-induced responses from frontal and selecting the data, we randomly split the trials in each parietal control regions. The contrast [informative cue mi- condition into two halves and correlated the induced re- nus neutral cue] between 3 and 30 Hz is shown for fronto- sponse topographies of contrasts performed on each half parietal and cingulo-opercular ROIs in Figure 5A. The separately, across time (i.e., correlating the topography at remaining parietal ROIs (IPS, SPL) (not shown) showed a time 1 with the topography at time 2—for all combina- similar pattern of responses to the mid-IPS. tions of t1 and t2). As illustrated in Figure 3A, on-diagonal The spatial pattern of induced responses in the control correlations reflect the reproducibility of topographies region ROIs replicated the spatial pattern of BOLD activa- across the independent data sets, and the off-diagonal tions captured in the meta-analysis. Both precues and retro- correlations capture the temporal persistence of these cues activated the anterior MFG and mid-IPS, with precues cue-induced brain states (see King & Dehaene, 2014, additionally activating FEF and retrocues activating iFEF. In for an in-depth discussion). The analysis was performed mid-IPS, an increase in theta power was accompanied by a

2028 Journal of Cognitive Neuroscience Volume 27, Number 10 Figure 3. Temporal dynamics of cue-induced brain states. (A) The expected correlation patterns for unstable and stable brain states. A continually changing brain state will not be correlated when t1 ≠ t2, that is, on the off- diagonal. By contrast a stable state will lead to off-diagonal correlations. (B) The correlation structure for the [precue – neutral] and [retrocue – neutral] topographies, that is, the stability of cue-induced brain states as indexed using sensor space induced responses in the theta (3–7Hz),alpha(8–12 Hz), and beta (18–30 Hz) bands, over all sensors. Group level t statistics are shown. The correlation structure was tested using a permutation distribution of cluster size and is thresholded at t = 2. Significant correlations are in full color saturation; nonsignificant correlations are unsaturated. The response to precues remains significantly correlated between 0.4 sec and the end of the precue epoch (presentation of the memory array) indicating a persistent, stable state. By contrast, the response to retrocues is less stable, evolving throughout the analysis epoch.

decrease in alpha/beta power that was sustained for pre- cues and transient for retrocues. This decrease was stron- gest in the left IPS. These frontoparietal network power increases occurred at a similar latency after the cue fol- lowing both precues and retrocues. Only retrocues coactivated the fO and pre-SMA (cingulo- opercular network), again replicating the pattern observed in the fMRI meta-analysis. In contrast to the frontoparietal network, these regions coactivated late in the retrocue epoch, after the decrease in power in mid-IPS had returned to baseline ∼1 sec postcue. The fO power increases were in the theta/alpha band, and the pre-SMA in the beta band. The pre-SMA also responded in the theta band immedi- ately following both precues and retrocues. These power increases were not present in the [precue minus neutral cue] analysis for the same retrocue epoch time period and are therefore unlikely to correspond to preparatory activity for the probe stimulus. We verified that the ROIs appropriately sampled the pat- tern of activity in source space by computing whole-brain maps of activity over successive 300-msec windows. The whole brain-induced response maps reproduced the pat- Figure 4. Meta-analysis of prospective and retrospective cueing tasks. terns expected on the basis of the fMRI meta-analysis. This MEG ROIs derived from meta-analysis activation results. The ROIs are is illustrated in Figure 5B, which shows the time period colored by network membership (following Dosenbach et al., 2007). – As ROIs were symmetric across hemispheres, we show the left 300 600 msec postcue, and in Figure 5C, which shows hemisphere only. the time period 1200–1500 msec postcue.

Wallis et al. 2029 Figure 5. Cue effects contrasts. (A) Induced responses in frontoparietal and cingulo- opercular ROIs reveals biphasic time course of control network activation. The time–frequency (TF) plots are for the contrast: informative cue minus neutral cue. TF data are averaged over left and right hemisphere ROIs, as most effects were bilateral, with the exception of the two plots marked with asterisks (anterior MFG and mid IPS following precues) for which effects were left-lateralized. Significant activation clusters are shown in full color saturation. Precues give rise to a sustained alpha/beta desynchronization in the left mid-IPS ( p < .0002) lasting until the presentation of the memory array, matching the time course of alpha lateralization in occipital cortex, consistent with a role for left IPS as a proximal control region for this attentional effect. Right anterior MFG ( p = .0004, early cluster; p = .029, late cluster) and bilateral FEF ( p = .0016) are activated in the theta band early following the cue. Cingulo- opercular nodes are not activated with the exception of a short-lived activation in the pre-SMA immediately following the cue ( p = .006). Retrocues gave rise to an alpha/beta desynchronization in the mid- IPS ( p < .0002), which lasted until ∼1 sec postcue, matching the time course of alpha lateralization. Activations in the theta and alpha/beta band in the anterior MFG ( p = .0054, p = .0004) preceded the parieto-occipital effects. Consistent with the pattern in fMRI, retrocues did not activate the FEF but did give rise to a bilateral activation in the more ventral precentral ROI ( p = .0036). Retrocues also gave rise to activations in the cingulo-opercular nodes. The pre-SMA was activated immediately following the retrocue in the theta band ( p = .0006) and also later in the epoch in the beta-band (∼1.2 sec postcue; p = .006). At this later time point, there was also a bilateral activation in the anterior insula/fO in the theta/alpha band ( p < .0002). (B) Whole-brain activation maps for the contrast [informative cue – neutral cue] projected onto the CARET cortical surface, averaged over the 300–600 msec time period following both precues and retrocues. Only activation belonging to a statistically significant activation cluster is shown. (C) Whole-brain activation maps for the contrast [informative cue – neutral cue] projected onto the CARET cortical surface, averaged over the 1200–1500 msec time period following both precues and retrocues. Only activation belonging to a statistically significant activation cluster is shown.

2030 Journal of Cognitive Neuroscience Volume 27, Number 10 DISCUSSION basis of temporal scale of control operations: Although the frontoparietal network is involved in moment-by- Selection mechanisms are effective in improving WM accu- moment adjustment of top–down control based on evolv- racy whether cues are prospective, selecting what “gets ing task requirements, the cingulo-opercular network is a into” WM, or retrospective, selecting from within WM. parallel system maintaining task set over longer periods. Although much of the previous literature has suggested However, the finding that the cingulo-opercular network that both effects may be mediated by a common top– is transiently recruited following retrocues suggests a more down mechanism modulating activity in sensory cortex, dynamic role in ongoing cognitive control. Our results can mediated by a frontoparietal network (Gazzaley & Nobre, be compared with an fMRI study by Ploran and colleagues 2012), other investigators have suggested that a separate (2007), in which images were slowly revealed in noise, until cingulo-opercular network is specifically involved in exert- participants were able to make a discrimination response. ing control over the contents of WM (Nelissen et al., 2013; Activity in the frontoparietal network slowly increased as ev- Higo et al., 2011). We addressed this question by record- idence was accumulated, but the cingulo-opercular network ing MEG data while participants processed either a precue was activated only at the moment participants made their or a retrocue. The spatial resolution of the source space response. These data suggested that the cingulo-opercular MEG recordings was sufficient to replicate the spatial pat- network has a more “downstream” role, acting on evidence tern of network activations seen in a meta-analysis of pre- integrated by the frontoparietal network in an interaction vious fMRI studies, and the high temporal resolution of with sensory cortex. Analogously, we suggest that in our task the method allowed us to dissociate the activation time the frontoparietal sites act to retrieve perceptual informa- course of the two networks. tion about the cued item, and the cingulo-opercular net- The spatiotemporal pattern of induced responses to work underpins a downstream stage. This is broadly precues and retrocues implicated the frontoparietal and consistent with a “cascade” account of executive function cingulo-opercular networks in different aspects of cogni- (Banich, 2009), but further work is needed to characterize tive control. A cross-temporal correlation analysis of the the nature of this secondary role for the cingulo-opercular sensor data showed that the brain response to precues network. Plausible functions include prioritization of the had a sustained component, whereas retrocues gave rise retrieved information to drive the response to the probe to a response that continued to evolve over the analysis item or inhibition of interfering information (uncued epoch. A source space analysis showed that the fronto- memoranda). The latter hypothesis might explain why parietal sites associated with precueing were activated we did not observe a similar power increase in the cingu- in the early phase of the response following retrocues, lo-opercular network before the presentation of the probe but that the cingulo-opercular sites activated later. Com- item in the precue condition: If there is only one item in paring these network dynamics with the time course of memory, there is are no other items that might interfere alpha modulation in perceptual cortex, they are consis- with the cognitive operation performed on the relevant tent with the suggestion that the frontoparietal network item (in this case, orientation comparison with the probe is responsible for top–down control over sensory cortex stimulus). (as reviewed in Gazzaley & Nobre, 2012). We propose Our results are also consistent with a conservative view that, consistent with prior hypotheses (Gazzaley & Nobre, of the role of attention in memory maintenance. Modula- 2012; Nobre et al., 2004), a common frontoparietal net- tions of alpha power in perceptual and parietal cortex are work is responsible for endogenously modulating activity a reliable marker for preparatory attention (van Ede, de in sensory cortex, whether this is to bias sensory process- Lange, Jensen, & Maris, 2011; Rihs, Michel, & Thut, 2007). ing during preparatory attention or to reactivate a sensory We found both the expected quadrant-specific modulation representation during memory retrieval. However, we of alpha power following precues, but also a very similar stress that, because we did not use a connectivity analysis, pattern following retrocues, consistent with recent reports we did not obtain direct evidence that the frontoparietal by Poch, Campo, and Barnes (2014) and Myers, Walther, network was responsible for top–down control. Wallis, Stokes, and Nobre (2015). Alpha lateralization was Following retrocues, the cingulo-opercular power in- sustained following precues, until the presentation of the creases occurred later in the trial, once the frontoparietal memory array, but transient following retrocues, peaking and sensory power changes had subsided. This disparity between 0.5 and 1 sec postcue and then returning to base- in activation timing is inconsistent with the hypothesis line. We also observed alpha lateralization in the retention that these cingulo-opercular sites directly modulate activ- interval as previously reported (Sauseng et al., 2009), but ity in sensory cortex during top–down control (Nelissen interestingly this was also not sustained, as might have et al., 2013; Higo et al., 2011) but suggests that they are been expected if memory maintenance consisted in sus- important for a separate operation specifically associated tained top–down activation of representations in parietal with retrocueing, as discussed below. or sensory cortex (Kiyonaga & Egner, 2012; Awh, Vogel, The “dual network” account proposed by Dosenbach, & Oh, 2006). Instead, over the relatively long retention in- Fair, Cohen, Schlaggar, and Petersen (2008) dissociates terval in our experiment, there were two periods of signif- the frontoparietal and cingulo-opercular networks on the icant alpha lateralization: one close after the presentation

Wallis et al. 2031 of the memory array (significant between 0.5 and 0.7 trieval errors in which the wrong item is selected to guide sec postcue in the classification analysis) and then behavior—that is, they mitigate against errors in output another that ramped up toward the presentation of the gating. These aspects of the behavioral data are all consis- memoryprobe(seeFigure2E).Intriguingly,alphalater- tent with retrocues facilitating output gating, as opposed to alization did not also ramp up in readiness for the probe optimizing memory maintenance. item following retrocues. We speculate that this may be In summary, although selection from WM following a because, in the 1.5 sec between retrocue and probe retrocue involves a similar top–down modulation of sen- item, participants are occupied retrieving and preparing sory and parietal cortex and a similar pattern of frontopa- to use the retrocued item, and there is not enough time rietal network activation as does preparatory attention, our to make a preparatory switch in attention back toward data suggest that control over WM is not identical with top– the probe item. down attention acting to bias memory maintenance activity Taken together, these results are consistent with the in sensory and parietal cortex. Sensory reactivation was idea that alpha lateralization in parietal/perceptual cortex transient, and there was no evidence for sustained biasing during memory maintenance tracks top–down activation, of maintenance activity following cues. Instead we would but that this may be dissociable from maintenance pro- suggest that the frontoparietal network mediates top– cesses (LaRocque et al., 2014; Lewis-Peacock, Drysdale, down control over sensory cortex, which can be recruited Oberauer, & Postle, 2012). Rather than interpreting the either to bias perception (attention) or to retrieve per- retinotopic reactivation representing the cued item fol- ceptual content associated with WM. We found, in line lowing a retrocue as biasing of ongoing maintenance ac- with previous studies (Nelissen et al., 2013; Higo et al., tivity, an alternative interpretation is that this reactivation 2011), that the second cingulo-opercular network was reflects a brief “access” event in which the sensory prop- specifically recruited by retrocues, but not precues. erties of the cued item are reactivated in order that it can However, the activation timing suggested it was not di- be prioritized in memory. Various contemporary models rectly involved in control over sensory representations, of WM have adopted a multilevel framework (Oberauer & as prior studies have suggested. The precise role of the Hein, 2012; Olivers, Peters, Houtkamp, & Roelfsema, cingulo-opercular network in cognitive control remains 2011), in which, of the set of items in memory, a single to be elucidated. Broadly, precues facilitate input gating item can be elevated to a special prioritized state. Sen- whereas retrocues may facilitate output gating of mem- sory reactivation following a retrocue may be involved ory (Hazy, Frank, & O’Reilly,2007).Thefrontoparietal in moving items into this prioritized state. network has a role in both input and output gating, but Our behavioral data also support this interpretation, as the cingulo-opercular network may be specifically asso- they were more consistent with retrocues modulating the ciated with output gating. retrievability of an item in memory (“output gating”)than with retrocues changing whether (or with what fidelity) the item was maintained in memory. Both precues and Acknowledgments retrocues reduced guess rate and the rate of responding This work was funded by Wellcome Trust studentships to G. W. about uncued items and also caused a leftward shift in and H. C., MRC fellowship MR/J009024/1 to M. S., a Wellcome the entire RT distribution compared with neutral-cue trials, Trust Equipment Grant to A. C. N. to support OHBA, the Na- consistent with faster retrieval of cued items. However, tional Institute for Health Research (NIHR) Oxford Biomedical Research Centre based at Oxford University Hospitals Trust precues had a substantial effect on the precision with Oxford University, and an MRC UK MEG Partnership grant which the cued items were represented, whereas precision (MR/K005464/1). The authors would like to thank Nils Kolling, was only marginally modulated by retrocues. This is not Nicholas Myers, and Franz Neubert for helpful discussions and consistent with the proposal that retrocues act primarily Sven Braeutigam, Henry Luckhoo, Diego Viduarre, and Adam by protecting cued items from gradual decay (Pertzov, Baker for their assistance with the MEG analysis. Bays, Joseph, & Husain, 2012), as were this the case we Reprint requests should be sent to George Wallis, Oxford Cen- would expect to have seen a more substantial precision ad- tre for Human Brain Activity, Warneford Hospital, Oxford, OX3 vantage following retrocues. An alternative explanation in 7JX, UK, or via e-mail: [email protected]. terms of maintenance processes is that retrocues protect items from “sudden death” during the retention interval, which might explain the difference in guess rate. However, REFERENCES in Murray et al. (2013) retrocues boosted performance Awh, E., Vogel, E. K., & Oh, S. H. (2006). Interactions between even if compared against a condition in which the probe attention and working memory. Neuroscience, 139, item was presented early, at the same time as the retrocue, 201–208. implying that the retrocue benefit could not depend only Banich, M. T. (2009). Executive function: The search for an on protecting items from forgetting during the remainder integrated account. Current Directions in Psychological Science, 18, 89–94. of the retention interval. Finally, in the current study, retro- Bays, P. M., Catalao, R. F. G., & Husain, M. (2009). The precision cues almost completely abolished the effect of nontarget of visual working memory is set by allocation of a shared items on behavior, implying that retrocues can prevent re- resource. Journal of Vision, 9, 7.1–711.

2032 Journal of Cognitive Neuroscience Volume 27, Number 10 Bays, P. M., Gorgoraptis, N., Wee, N., Marshall, L., & Husain, M. in extrastriate visual areas: Top–down effects of refreshing (2011). Temporal dynamics of encoding, storage, and just-seen visual stimuli. Neuroimage, 37, 290–299. reallocation of visual working memory. Journal of Vision, Kastner, S., DeSimone, K., Konen, C. S., Szczepanski, S. M., 11, 1–15. Weiner, K. S., & Schneider, K. A. (2007). Topographic maps Chatham, C. H., Frank, M. J., & Badre, D. (2014). Corticostriatal in human frontal cortex revealed in memory-guided saccade output gating during selection from working memory. and spatial working-memory tasks. Journal of Neuron, 81, 930–942. Neurophysiology, 97, 3494–3507. Corbetta, M., Kincade, J. M., Ollinger, J. M., McAvoy, M. P., & King, J.-R., & Dehaene, S. (2014). Characterizing the dynamics Shulman, G. L. (2000). Voluntary orienting is dissociated of mental representations: The temporal generalization from target detection in human posterior parietal cortex. method. Trends in Cognitive Sciences, 18, 203–210. Nature Neuroscience, 3, 292–297. Kiyonaga, A., & Egner, T. (2012). Working memory as internal de Haan, B., Morgan, P. S., & Rorden, C. (2008). Covert attention: Toward an integrative account of internal and orienting of attention and overt eye movements activate external selection processes. Psychonomic Bulletin & identical brain regions. Brain Research, 1204, 102–111. Review, 20, 228–242. Derrfuss, J., Vogt, V. L., Fiebach, C. J., von Cramon, D. Y., & Kuo, B.-C., Stokes, M. G., Murray, A. M., & Nobre, A. C. (2014). Tittgemeyer, M. (2012). Functional organization of the left Attention biases visual activity in visual short-term memory. inferior precentral sulcus: Dissociating the inferior frontal eye Journal of Cognitive Neuroscience, 26, 1377–1389. field and the inferior frontal junction. Neuroimage, 59, Landman, R., Spekreijse, H., & Lamme, V. A. F. (2003). Large 3829–3837. capacity storage of integrated objects before change Dosenbach, N. U. F., Fair, D. A., Cohen, A. L., Schlaggar, B. L., & blindness. Vision Research, 43, 149–164. Petersen, S. E. (2008). A dual-networks architecture of LaRocque, J. J., Lewis-Peacock, J. A., & Postle, B. R. (2014). top–down control. Trends in Cognitive Sciences, 12, Multiple neural states of representation in short-term 99–105. memory? It’s a matter of attention. Frontiers in Human Dosenbach, N. U. F., Fair, D. A., Miezin, F. M., Cohen, A. L., Neuroscience, 8, 5. Wenger, K. K., Dosenbach, R. A. T., et al. (2007). Distinct Lepsien, J., Griffin, I. C., Devlin, J. T., & Nobre, A. C. (2005). brain networks for adaptive and stable task control in Directing spatial attention in mental representations: humans. Proceedings of the National Academy of Sciences, Interactions between attentional orienting and working- U.S.A., 104, 11073–11078. memory load. Neuroimage, 26, 733–743. Egner, T., Monti, J. M. P., Trittschuh, E. H., Wieneke, C. A., Lepsien, J., & Nobre, A. C. (2006). Attentional modulation of Hirsch, J., & Mesulam, M. M. (2008). Neural integration of object representations in working memory. Cerebral Cortex top–down spatial and feature-based information in visual (New York, N.Y.: 1991), 17, 2072–2083. search. Journal of Neuroscience, 28, 6141–6151. Lewis-Peacock, J. A., Drysdale, A. T., Oberauer, K., & Postle, Eickhoff, S. B., Laird, A. R., Grefkes, C., Wang, L. E., Zilles, K., & B. R. (2012). Neural evidence for a distinction between short- Fox, P. T. (2009). Coordinate-based activation likelihood term memory and the focus of attention. Journal of estimation meta-analysis of data: A random- Cognitive Neuroscience, 24, 61–79. effects approach based on empirical estimates of spatial Mangun, G. R., Hopfinger, J. B., & Buonocore, M. H. (2000). uncertainty. Human , 30, 2907–2926. The neural mechanisms of top–down attentional control. Gazzaley, A., & Nobre, A. C. (2012). Top–down modulation: Nature Neuroscience, 3, 284–291. Bridging selective attention and working memory. Trends in Maris, E., & Oostenveld, R. (2007). Nonparametric statistical Cognitive Sciences, 16, 129–135. testing of EEG- and MEG-data. Journal of Neuroscience Giesbrecht, B., Woldorff, M. G., Song, A. W., & Mangun, G. R. Methods, 164, 177–190. (2003). Neural mechanisms of top–down control during Munneke, J., Belopolsky, A. V., & Theeuwes, J. (2012). Shifting spatial and feature attention. Neuroimage, 19, 496–512. attention within memory representations involves early visual Griffin, I. C., & Nobre, A. C. (2003). Orienting attention to areas. PLoS One, 7, e35528. locations in internal representations. Journal of Cognitive Murray, A. M., Nobre, A. C., Clark, I. A., Cravo, A. M., & Stokes, Neuroscience, 15, 1176–1194. M. G. (2013). Attention restores discrete items to visual short- Haegens, S., Händel, B. F., & Jensen, O. (2011). Top–down term memory. Psychological Science, 24, 550–556. controlled alpha band activity in somatosensory areas Murray, A. M., Nobre, A. C., & Stokes, M. G. (2011). Markers of determines behavioral performance in a discrimination task. preparatory attention predict visual short-term memory Journal of Neuroscience, 31, 5197–5204. performance. Neuropsychologia, 49, 1458–1465. Harrison, S. A., & Tong, F. (2009). Decoding reveals the Myers, N. E., Walther, L., Wallis, G., Stokes, M. G., & Nobre, A. contents of visual working memory in early visual areas. C. (2015). Temporal dynamics of attention during encoding Nature, 458, 632–635. versus maintenance of working memory: Complementary Hazy, T. E., Frank, M. J., & O’Reilly, R. C. (2007). Towards an views from event-related potentials and alpha-band executive without a homunculus: Computational models of oscillations. Journal of Cognitive Neuroscience, 27, the prefrontal cortex/basal ganglia system. Philosophical 492–508. Transactions of the Royal Society of London, Series B, Nee, D. E., & Jonides, J. (2009). Common and distinct neural Biological Sciences, 362, 1601–1613. correlates of perceptual and memorial selection. Higo, T., Mars, R. B., Boorman, E. D., Buch, E. R., & Rushworth, Neuroimage, 45, 963–975. M. F. S. (2011). Distributed and causal influence of frontal Nelissen, N., Stokes, M., Nobre, A. C., & Rushworth, M. F. S. operculum in task control. Proceedings of the National (2013). Frontal and parietal cortical interactions with Academy of Sciences, U.S.A., 108, 4230–4235. distributed visual representations during selective attention Johnson, M. R., & Johnson, M. K. (2009). Top–down and action selection. Journal of Neuroscience, 33, enhancement and suppression of activity in category- 16443–16458. selective extrastriate cortex from an act of reflective attention. Nobre, A. C., Coull, J. T., Maquet, P., Frith, C. D., Vandenberghe, Journal of Cognitive Neuroscience, 21, 2320–2327. R., & Mesulam, M. M. (2004). Orienting attention to locations Johnson, M. R., Mitchell, K. J., Raye, C. L., D’Esposito, M., & in perceptual versus mental representations. Journal of Johnson, M. K. (2007). A brief thought can modulate activity Cognitive Neuroscience, 16, 363–373.

Wallis et al. 2033 Nolte, G. (2003). The magnetic lead field theorem in the quasi- alpha activity. European Journal of Neuroscience, 22, static approximation and its use for magnetoencephalography 2917–2926. forward calculation in realistic volume conductors. Physics Siegel, M., Donner, T. H., Oostenveld, R., Fries, P., & Engel, A. K. in Medicine and Biology, 48, 3637–3652. (2008). Neuronal synchronization along the dorsal visual Oberauer, K., & Hein, L. (2012). Attention to information in pathway reflects the focus of spatial attention. Neuron, 60, working memory. Current Directions in Psychological 709–719. Science, 21, 164–169. Sligte, I. G., Scholte, H. S., & Lamme, V. A. F. (2008). Are there Olivers, C. N., Peters, J., Houtkamp, R., & Roelfsema, P. R. multiple visual short-term memory stores? PLoS One, 3, e1699. (2011). Different states in visual working memory: When it Sligte, I. G., Scholte, H. S., & Lamme, V. A. F. (2009). guides attention and when it does not. Trends in Cognitive V4 activity predicts the strength of visual short-term Sciences, 15, 327–334. memory representations. Journal of Neuroscience, 29, Pasternak, T., & Greenlee, M. W. (2005). Working memory in 7432–7438. primate sensory systems. Nature Reviews Neuroscience, 6, Sperling, G. (1960). The information available in brief visual 97–107. presentations. Psychological Monographs: General and Pelli, D. G. (1997). The VideoToolbox software for visual Applied, 74, 1. psychophysics: Transforming numbers into movies. Spatial Stokes, M. G., Kusunoki, M., Sigala, N., Nili, H., Gaffan, D., & Vision, 10, 437–442. Duncan, J. (2013). Dynamic coding for cognitive control in Pertzov, Y., Bays, P. M., Joseph, S., & Husain, M. (2012). Rapid prefrontal cortex. Neuron, 78, 364–375. forgetting prevented by retrospective attention cues. Taulu, S., Kajola, M., & Simola, J. (2004). Suppression of Journal of Experimental Psychology: Human Perception interference and artifacts by the signal space separation and Performance, 39, 1224. method. Brain Topography, 16, 269–275. Petersen, S. E., & Posner, M. I. (2012). The attention system of van Ede, F., de Lange, F., Jensen, O., & Maris, E. (2011). the human brain: 20 Years after. Annual Review of Orienting attention to an upcoming tactile event involves a Neuroscience, 35, 73–89. spatially and temporally specific modulation of sensorimotor Ploran, E. J., Nelson, S. M., Velanova, K., Donaldson, D. I., alpha- and beta-band oscillations. Journal of Neuroscience, Petersen, S. E., & Wheeler, M. E. (2007). Evidence 31, 2016–2024. accumulation and the moment of recognition: Dissociating van Ede, F., Köster, M., & Maris, E. (2012). Beyond establishing perceptual recognition processes using fMRI. Journal of involvement: Quantifying the contribution of anticipatory Neuroscience, 27, 11912–11924. α- and β-band suppression to perceptual improvement Poch, C., Campo, P., & Barnes, G. R. (2014). Modulation of with attention. Journal of Neurophysiology, 108, 2352–2362. alpha and gamma oscillations related to retrospectively Vogel, E. K., & Machizawa, M. G. (2004). Neural activity predicts orienting attention within working memory. European individual differences in visual working memory capacity. Journal of Neuroscience, 40, 2399–2405. Nature, 428, 748–751. Raye, C. L., Johnson, M. K., Mitchell, K. J., Reeder, J. A., & Wilson, K. D., Woldorff, M. G., & Mangun, G. R. (2005). Control Greene, E. J. (2002). Neuroimaging a single thought: networks and hemispheric asymmetries in parietal cortex Dorsolateral PFC activity associated with refreshing just- during attentional orienting in different spatial reference activated information. Neuroimage, 15, 447–453. frames. Neuroimage, 25, 668–683. Raye, C. L., Mitchell, K. J., Reeder, J. A., Greene, E. J., & Woldorff, M. G., Hazlett, C. J., Fichtenholtz, H. M., Weissman, D. Johnson, M. K. (2008). Refreshing one of several active H., Dale, A. M., & Song, A. W. (2004). Functional parcellation representations: Behavioral and functional magnetic of attentional control regions of the brain. Journal of resonance imaging differences between young and older Cognitive Neuroscience, 16, 149–165. adults. Journal of Cognitive Neuroscience, 20, 852–862. Woolrich, M., Hunt, L., Groves, A., & Barnes, G. (2011). MEG Rihs, T. A., Michel, C. M., & Thut, G. (2007). Mechanisms of beamforming using Bayesian PCA for adaptive data covariance selective inhibition in visual spatial attention are indexed by matrix regularization. Neuroimage, 57, 1466–1479. alpha-band EEG synchronization. European Journal of Worden, M. S., Foxe, J. J., Wang, N., & Simpson, G. V. (2000). Neuroscience, 25, 603–610. Anticipatory biasing of visuospatial attention indexed by Roth, J. K., Johnson, M. K., Raye, C. L., & Constable, R. T. (2009). retinotopically specific-band electroencephalography Similar and dissociable mechanisms for attention to internal increases over occipital cortex. Journal of Neuroscience, versus external information. Neuroimage, 48, 601–608. 20, 1–6. Rowe, J. B., & Passingham, R. E. (2001). Working memory for Yeh, Y.-Y., Kuo, B.-C., & Liu, H.-L. (2007). The neural correlates location and time: Activity in prefrontal area 46 relates to of attention orienting in visuospatial working memory for selection rather than maintenance in memory. Neuroimage, detecting feature and conjunction changes. Brain Research, 14,77–86. 1130, 146–157. Rowe,J.B.,Toni,I.,Josephs,O.,Frackowiak,R.S.,&Passingham, Yi, D.-J., Turk-Browne, N. B., Chun, M. M., & Johnson, M. K. R. E. (2000). The prefrontal cortex: Response selection or (2008). When a thought equals a look: Refreshing enhances maintenance within working memory? Science, 288, perceptual memory. Journal of Cognitive Neuroscience, 20, 1656–1660. 1371–1380. Sauseng, P., Klimesch, W., Stadler, W., Schabus, M., Zhang, W., & Luck, S. J. (2008). Discrete fixed-resolution Doppelmayr, M., Hanslmayr, S., et al. (2009). A shift of visual representations in visual working memory. Nature, spatial attention is selectively associated with human EEG 453, 233–235.

2034 Journal of Cognitive Neuroscience Volume 27, Number 10