NeuroImage 89 (2014) 289–296

Contents lists available at ScienceDirect

NeuroImage journal homepage: www.elsevier.com/locate/ynimg

Parietal structure and function explain human variation in working memory biases of visual attention David Soto a,⁎, Pia Rotshtein b, Ryota Kanai c,d a

Imperial College London, Division of Brain Sciences, London, UK University of Birmingham, Behavioural and Brain Sciences Centre, Birmingham, UK University College London, Institute of Cognitive Neuroscience, London, UK d University of Sussex, School of Psychology, UK b c

a r t i c l e

i n f o

Article history: Accepted 17 November 2013 Available online 25 November 2013

a b s t r a c t Recent research indicates that human attention appears inadvertently biased by items that match the contents of working memory (WM). WM-biases can lead to attentional costs when the memory content matches goalirrelevant items and to attentional benefits when it matches the sought target. Here we used functional and structural MRI data to determine the neural basis of human variation in WM biases. We asked whether human variation in WM-benefits and WM-costs merely reflects the process of attentional capture by the contents of WM or whether variation in WM biases may be associated with distinct forms of cognitive control over internal WM signals based on selection goals. Human ability to use WM contents to facilitate selection was positively correlated with gray matter volume in the left superior posterior parietal cortex (PPC), while the ability to overcome interference by WM-matching distracters was associated with the left inferior PPC in the anterior IPS. Functional activity in the left PPC, measured by functional MRI, also predicted the magnitude of WM-costs on selection. Both structure and function of left PPC mediate the expression of WM biases in human visual attention. © 2013 Elsevier Inc. All rights reserved.

Introduction Neurocognitive mechanisms underlying the interplay between working memory (WM) and selective attention are central to our understanding of goal-directed behavior — think of searching for your spouse in the crowd or finding a pair of shoes that match your jacket. According to biased competition theory, visual selection may be guided strategically from the contents of WM (Desimone and Duncan, 1995; Duncan and Humphreys, 1989). In addition to these strategic effects, WM may bias selection independently of attention goals, namely, inadvertent WM guidance (Olivers et al., 2011; Soto et al., 2008) leading to the capture of attention to irrelevant stimuli matching the WM contents. Neuroimaging studies have identified a fronto-thalamic circuit linked to visual cortex that responds to the presence of WMmatching stimuli (Grecucci et al., 2010; Rotshtein et al., 2011; Soto et al., 2007). Involvement of the posterior parietal cortex (PPC) has not been reliably observed in studies assessing the interplay between WM and visual attention (Soto et al., 2007, 2011), though evidence of increased pulvinar–parietal coupling has however been observed in similar paradigms (Rotshtein et al., 2011). Additional recent fMRI evidence by Soto and colleagues indicates that PPC and middle temporal (hippocampal) cortices may be involved in the strategic modulation of ⁎ Corresponding author at: Imperial College London, Department of Medicine, Division of Brain Sciences, Charing Cross Campus, St Dunstan's Rd, W6 8RP London, UK. E-mail address: [email protected] (D. Soto). 1053-8119/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.neuroimage.2013.11.036

WM biases through expectations/foreknowledge about the incoming validity of WM items for visual selection goals, namely, boosting or suppressing WM biases when WM contents predict a target or a distracter (Soto et al., 2012). Behavioral studies have shown that WM biases of visual selection can be subjected to some degree to cognitive control (Carlisle and Woodman, 2011; Han and Kim, 2009; Kiyonaga et al., 2012; Olivers, 2009; Olivers and Eimer, 2011; Olivers et al., 2011; Woodman and Luck, 2007) and the PPC has been consistently associated with strategic attentional control processes, namely, maintaining an attentional set in search for a relevant target feature (Egner et al., 2008; Schenkluhn et al., 2008; Toth and Assad, 2002). We used functional and structural MRI data from four studies assessing WM biasing of perceptual selection (Soto et al., 2007, 2011, 2012a,b) to further our understanding of the mechanisms that shape the expression of WM biases and its cognitive control. The paradigm requires observers to keep an object in WM and to perform concurrently a search task for an unrelated target (see Fig. 1) (Soto et al., 2005). Relative to a neutral baseline, search is slower when the WM cue reappears surrounding a distracter (invalid trials) and faster when it matches the sought target (valid trials). A subsequent recognition test for the initial cue is presented to ensure it was held in WM. Since the different trial types happen with the same probability observers have little incentive to use the cue strategically to find the target. Further, evidence indicates that WM-biases are present even when the WM-cue is always associated with a search distracter (Mannan et al., 2010; Olivers et al., 2006; Pan and Soto, 2010; Soto and Humphreys, 2007; Soto et al.,

290

D. Soto et al. / NeuroImage 89 (2014) 289–296

Fig. 1. Illustration of the experimental protocol to assess WM guidance of visual attention.

2005) and the size of the WM bias on behavior is not different from situations where the WM contents are equally likely to be valid, neutral or invalid for search (Soto et al., 2005), suggesting that WM-biases are not simply based on strategic decisions. But importantly participants may vary in the extent they employ strategic control decisions for utilizing the WM contents during the attention task. For example, some individuals may show a tendency to drive visual selection based on internal knowledge held in WM while others may be ‘averse’ of being taxed with WM contents which are not directly relevant for the search task hence minimizing costs from WM biases. Based on prior evidence that inadvertent effects of WM on selection may be subjected to cognitive control, we expected individual variation in the ability to suppress search distracters matching the WM contents. We also expected that observers would vary in their ability to use their WM to benefit search when WM contents turn out to match the sought target. Individual variation in WM-cost and WM-benefits of selection might be mediated by a single process associated with the capture of attention from the contents of WM. Alternatively we questioned whether variation in WMcosts and WM-benefits may be associated with distinct processes of cognitive control with fractionated neural substrates. We approached these questions by means of whole-brain analyses of blood-oxygenated level dependent (BOLD) responses and voxelbased morphometry (VBM) analyses of brain structure and function which have proven as useful procedures to relate individual variation in cognitive performance to brain anatomy (Fleming et al., 2010; Kanai and Rees, 2011; Schwarzkopf et al., 2011). Methods Participants Forty-four participants were tested across the four Experiments. The number of participants per Experiment and other related information are noted below. Each participant took part in a single Experiment only. None of the participants were aware of the purpose of the experiment and none had any prior history of neurological or psychiatric disorders. Their vision was normal or corrected-to-normal and all were right handed. Local research ethics committee approval was granted for the experimental procedures. All participants provided written informed consent and received monetary compensation for their participation.

(Psychology Software Tools Inc., Pittsburgh, PA; http://www.pstnet. com/eprime.cfm). Each trial presented observers with a cue in the form of a geometrical shape (e.g. triangle, circle, square, diamond and hexagon) appearing in five different colors (e.g. red, green, blue, yellow, pink). Observers were instructed to hold the cue in WM through the trial. The cue was followed by a delay and then by a search display. This was composed of different colored shapes, each containing a line. The target line was tilted either to the left or to the right; distracter lines were vertical. Participants had to indicate the orientation of the target line by means of a button press, as fast and as accurate as possible. One of the colored shapes in the search display could match the properties of the cue (both color and shape). On valid trials, the matching object surrounded the target line while on invalid trials it surrounded a vertical distracter. On neutral trials, none of the objects in the search display matched the properties of the cue. One third of the trials were valid, 1/3 invalid and 1/3 neutral. A memory test followed after completion of the line orientation task. A memory probe (a single colored shape) appeared and participants were instructed to respond ‘same’ if both dimensions of the cue and probe object matched; otherwise they responded ‘different’. On ‘different’ trials, the cue and probe items could have just their color in common, just their shape in common, or neither attribute in common. Responses in the memory task were made by pressing one key for ‘same’ and one for ‘different’ stimuli. Fig. 1 depicts the sequence of events. Experiment 1 (Soto et al., 2007) This was a 3 (cue validity: valid, neutral and invalid) × 2 (memory requirement: WM, priming) factorial design. The memory requirement varied across blocks, while cue validity was randomly varied across trials. For the purposes of the current analyses, we only considered trials from the WM condition where observers had to keep the color and the shape of the cue in memory for a subsequent recognition test. Participants Ten participants between the ages of 18 and 23 years were recruited. Experimental design Each trial began with a fixation display for 500 ms, followed by a WM-cue that was displayed twice (100 and 500 ms, respectively) with a 100-ms blank interval in between. After a 250-ms blank interval, the search display appeared for 1 s. Memory items were five simple outline colored shapes. In the WM condition both the color and shape had to be remembered throughout the trial. The search display was composed of two unique colored shapes presented to the left and right of fixation, each containing the target and a distracter line respectively. Twenty percent of the trials (catch trials) included a memory test following completion of where participants responded whether a probe was same or different (in both color and shape) relative to the memory cue. Further details can be found in Soto et al. (2007). Experiment 2 (Soto et al., 2012a) This study assessed the influence of cognitive load on WM guidance of selection. The experimental design was a 3 (validity: valid, neutral and invalid) × 3 (WM load: low, high and no load). WM load varied across blocks and the validity factor varied on a trial-by-trial basis. For the purposes of the current analyses, we only considered trials from the low load WM condition where observers had to keep a single item with a unique color and shape in memory.

General experimental task and procedure The basic experimental procedure was similar across the four Experiments. The tasks were programmed and run using E-Prime

Participants Twelve healthy volunteers aged between 22 and 26 years (7 female, all right-handed) were recruited.

D. Soto et al. / NeuroImage 89 (2014) 289–296

Experimental design Each trial began with a fixation cross for 500 ms, followed by a cue display composed of three vertically aligned colored shapes on a gray background. In the high WM load condition, participants had to keep in WM the three colored shapes in the memory cue display. In the low load condition, they were instructed to remember only the central colored shape. For the purposes of the current study, and to keep the experimental conditions as similar as possible, across the 4 experiments, we only included trials from the low load condition. Each stimulus in the cue display was unique in color and shape. The cue display was presented for 100 ms, followed by a brief presentation (100 ms) of a blank screen, and another 500 ms presentation of the cue display. A blank screen followed for 700 ms. A search display composed of two horizontally aligned colored shapes followed. Each item could contain the tilted target or a vertical distractor. The search display remained onscreen for 175 ms and was followed by a further 750 ms response period during which a blank screen was presented. A memory probe test followed completion of the search task. Experiment 3 (Soto et al., 2011) The experiment had a 3 (task: WM only, WM + spatial cueing, priming) by 3 (object cue validity: invalid, neutral, valid) design. The task factor varied across blocks whereas object cue validity varied across trials. For the purposes of the current analyses, we only considered trials from the WM only condition where no spatial cue was given and observers had to keep the color and the shape of the cue in memory for a subsequent recognition test. Participants Twelve healthy volunteers aged between 19 and 26 years (5 females) participated. Experimental design Each trial began with a memory cue displayed two times (100 and 500 ms respectively) with a 100 ms blank interval in between. In the WM condition, observers were asked to keep the shape and color of the cue in memory throughout the trial for a subsequent recognition test. For the purposes of the current study only trials from the WMonly condition where considered. The offset of the cue was followed by a 700 ms blank interval. During WM-only blocks, a cross shape was presented instead for 175 ms (spatial cues never appeared in priming or WM-only blocks). A search display composed of two different colored shapes presented to the right and left of the center of the screen appeared for another 175 ms. Observers were required to identify the orientation of the tilted line within a time window of 925 ms from search onset. A memory test followed after completion of the search task. Experiment 4 (Soto et al., 2012b) The design was a factorial within-subjects design with WM-cue type (Verbal: written word such as “Red Square”; Visual: a visual colored shape) and cue validity (valid, neutral, invalid) as factors. For the purposes of the current study, only trials from the visual condition were included in the analyses. Participants Ten volunteers aged between 22 and 26 years (6 females) participated.

291

of an imaginary square centered at fixation. Each object contained a line; three of the lines were vertical while the target one was tilted either to the left or to the right. Participants had to discriminate the orientation of the target during a time window of 1500 ms. A memory test followed the search task. Structural imaging T1-weighted images were used, images were acquired using a sagittal acquisition plane with a 1 × 1 × 1 resolution. Experiments 1 and 4 used a Phillips 3 T Achieva system to acquire 175 sagittal slices with a resolution of 1 × 1 × 1 mm (TR = 8400 ms; TE = 38 ms; flip angle = 8°). Experiments 2 and 3 used a Siemens 3 T fMRI scanner to collect 176 sagittal slices with a resolution of 1 × 1 × 1 mm (TR = 1900 ms, TE = 2.48 ms, flip angle = 9°). Functional imaging Blood Oxygenated Level Dependent contrast-weighted echoplanar images were acquired. Experiment 1: 39 oblique slices, 2 mm thick with a 1-mm gap and in-plane resolution of 3 × 3 mm2, 80° flip angle, 30 ms echo time, and 2110-ms slice repetition time. Experiment 4: 37 oblique slices, 2-mm thickness and 1 mm gap, and in-plane resolution of 3 × 3 mm2, 80° flip angle, 35 ms echo time, and 2020-ms slice repetition time. Images were acquired using an eight-channel phase array coil with a sense factor of 2. Experiments 2 and 3: 38 slices, 3 mm thickness with 0% gap and in-plane resolution was 2.4 × 2.4 mm2, 78° flip angle, 30-ms echo time, and 2200 ms slice repetition time. Voxel based morphometry Image preprocessing The T1-weighted MR images were first segmented for gray matter (GM) and white matter (WM) using the segmentation tools in SPM8 (http://www.fil.ion.ucl.ac.uk/spm). Subsequently, we performed Diffeomorphic Anatomical Registration through Exponentiated Lie Algebra (DARTEL) in SPM8 for inter-subject registration of the tissue segmented images. To ensure that the total gray matter volume was retained before and after spatial transformation, the image intensity was modulated by the Jacobian determinants of the deformation fields. The registered images were then smoothed with a Gaussian kernel (FWHM = 8 mm) and were then affine transformed to Montreal Neurological Institute (MNI) stereotactic space using affine and non-linear spatial normalization implemented in SPM8 for multiple regression analysis. Multiple regression analyses The four experiments were modeled as a factor of no interest to regress out any effects attributable to differences in the MRI scanner and experimental parameters across the studies. Also, the total gray matter volume of the whole brain was included in the design matrix as a covariate of no interest and was thus regressed out in the analysis. In the analysis, both the individual WM-benefits (Neutral RT–Valid RT) and WM-costs (Invalid RT–Neutral RT) of the validity of WM cues were included in the design matrix to separately identify regions related to these measures. We used p b 0.05 corrected for the whole brain volume at a cluster level with non-stationary correction (Hayasaka et al., 2004) as the criterion for statistical significance. For descriptive proposes, we plotted the behavioral measures against the gray matter probability values from the peak voxel identified in the whole brain voxel-based analysis. fMRI analyses

Experimental design The memory cue appeared at fixation for 928 ms. The cue was followed by a blank screen for 188 ms and then the search display for 100–183 ms and a random dot mask for 1500 ms. The search display was composed of four different colored shapes displayed at the corners

First, we estimated the individual effect size for each condition averaged across all the fMRI sessions of a specific experiment. We modeled the onsets of the WM cue on each trial separately for each condition. We also included a regressor for the onset of error trials. Motion

292

D. Soto et al. / NeuroImage 89 (2014) 289–296

realignment parameters and harmonic capturing slow frequencies in the data (1/128 Hz) were also included in the design matrix. Further, for each condition, we added the search RT as a covariate to control for differences in RT. All other experimental conditions were modeled as effects of no interest. These regressors were convolved with the canonical hemodynamic response function. Further details can be found in Soto et al. (2007, 2012a,b). We analyzed the fMRI data for two reasons. First, we wanted to test whether the results we reported in each of the published experiments (often using un-corrected Z-scores, due to the relatively low number of participants) would be observed when all the data is pooled together. For this we used random effect models for the analyses across the 4 Experiments, including only effects based on the conditions that were similar across the experiments (i.e. the 3 validity conditions in the context of a cue held in WM). Here we assessed effects that were consistent across participants. In line with our previous studies we focused on two contrasts: 1) The WM validity contrast: valid [1] N neutral [0] N invalid [−1]; and the WM re-appearance effect: valid [1] + invalid [1] N Neutral [−2]. We report clusters with a voxelwise threshold of Z N 2.33, FDR corrected for multiple comparisons at the cluster level. Second, we aimed to assess how variability in neural-function (indexed by BOLD responses) across participants relates to the observed brain structural variability between participants by using an ROI approach. Two ROIs were defined based on the clusters from VBM analyses associated with the WM benefit and WM cost: left superior and inferior parietal cortices (anterior IPS), respectively (see Results below). From these two ROIs we extracted the BOLD responses of each participant for each of the three conditions. We then computed for each participant and each ROI the neural response associated with WM-cost: Invalid–Neutral; and the WM-benefit: Valid–Neutral. These differential activation levels were correlated with the RT cost and benefit effects measured in behavioral tasks. We used partial correlations, in which the experiment/scanner factor was regressed out to control for any differences in the MRI scanner and experimental parameters across the four experiments.

Results Behavioral data WM effects on search performance were strongly significant across the four studies datasets included here (N = 44) (Soto et al., 2007, 2012a,b). Validity effects (i.e. Invalid RT–Valid RT = 50.3 ± 7.3 ms, t(43) = 6.89, p b 10− 7) were reliable both in terms of cost effects (Invalid RT–Neutral RT = 34.4 ± 6.7 ms, t(43) = 5.17, p b 10−5); and also in terms of WM benefit effects (Neutral RT–Valid RT = 15.9 ± 4.7 ms, t(43) = 3.38, p b 0.01). The distributions of these effects across participants however indicate that there is substantial variability in WM-validity effects (Fig. 2). Next we tested whether individual differences of the WM-costs correlated with those of WM-benefits. We

observed that the WM-costs and WM-benefits on selection did not correlate (Pearson correlation: −0.21, p = 0.18). This suggests that the sources of the variability in these two measures (costs and benefits) may be different. Consistent effects across participants Based on the fMRI data, we computed WM effects on neural responses across participants pooling together the data from the four studies above. The data were derived from similar experimental conditions of these previously published studies (see Fig. 1 for examples of the different WM validity conditions). We first examined validity effects (valid N neutral N invalid). Consistent with the reported results of these individual studies (Soto et al., 2007, 2012a,b), we observed increased responses to valid relative to invalid trials in anterior and dorsal frontal cortex, bilateral inferior temporal and left lateral occipital and superior temporal cortices (Fig. 3A). We further observed thalamic involvement in WM guidance in keeping with reports from our previous studies; there were validity effects in anterior thalamic nuclei (MNI = − 3, − 3, 6, p b 0.01, corrected) and also in the vicinity of the right pulvinar/posterior hippocampus (MNI = 25, − 35, 2, p b 0.01, corrected). We also found WM-validity related activation in the left PPC around supramarginal and the angular gyri. This finding is interesting because PPC activity across each of the prior studies considered individually did not consistently discriminate the presence of WM biases. The re-appearance effects (valid + invalid N neutral) were observed in bilateral occipital, superior frontal and left anterior temporal cortices (Fig. 3B). Re-appearance WM-effects were also observed in the left pulvinar/hippocampus (MNI = − 20, − 34, 6, p b 0.01, corrected) and also in left posterior pulvinar (MNI = − 6, − 34, 6, p b 0.01, corrected). VBM analysis In order to assess the relationship between brain structure and the strength of the WM biases, we performed a whole-brain correlation analyses between the size of the WM effect on search RTs and gray matter (GM) integrity. Note that the VBM model included regressors representing the WM-benefit on search (Neutral RT–Valid RT), the WM cost effects (Invalid RT–Neutral RT) and nuisance regressors for the different experiments and scanners. We found two functionally fractionated subregions within the left PPC (Fig. 4A). The size of the WM-costs positively correlated with the volume of the gray matter (GM) in the anterior part of the intraparietal sulcus (aIPS) (Figs. 4B; r = 0.65, t(37) = 5.28, x = − 42, y = − 43, z = 40, p b 0.05, corrected, Z = 4.53). This shows that higher GM volume of the left anterior IPS made individuals more prone to be distracted and/or less prone to disengage from invalid WM distracters in the search. Note that WM-benefit correlation with gray matter in this region was obtained even when we controlled for WM-cost in the model. No region showed the opposite correlation.

Fig. 2. The frequency histograms of WM effects are shown as a function of effect size. (A) The distribution of WM validity effects across 44 participants from four studies. (B) The distribution of WM-cost effects. (C) The distribution of WM-benefit effects. The dotted green lines indicate the mean across participants. These were all significantly higher than zero (see the main text).

D. Soto et al. / NeuroImage 89 (2014) 289–296

293

WM-benefit cluster around the SPL did not correlate with search RT benefits (r = −0.0765, p = 0.311) but interestingly SPL structure did correlate with overall memory performance in the subsequent recognition test (see below). Relationship to overall memory performance

Fig. 3. WM effects on neural responses. Left and right sagittal views of (A) WM validity effects (B) WM-reappearance effects. The activation is displayed at a voxelwise threshold Z N 2.33, corrected for multiple comparisons.

For descriptive purposes we only plotted the behavioral measures against the gray matter volume of the peak correlation observed in Fig. 4C. We also noted participants' data point in each study using different symbols and colors. This figure illustrates that the same trends were reproduced for individual studies separately, indicating that the observed correlation is not due to differences in the scanner manufacturers or differences in experimental parameters across the studies. The performance benefits from WM cueing correlated with the gray matter volume of a more superior part of the left PPC (SPL) (r = 0.58, t(37) = 4.37, x = − 32, y = − 55, z = 52, p b 0.05, corrected, Z = 3.90. cluster size = 988.9 mm3) (see Figs. 4D and E). This demonstrates that individuals with lower GM volume in the left SPL were less able to take advantage of valid WM contents to facilitate selection. No region showed the opposite effect. In line with the absence of correlation between the WM-benefits and WM-costs in search performance, there was no significant correlation in GM density between the cost peak and benefit peak of the VBM clusters (r = 0.237, t(42) = 1.58, p = 0.12). Taken together, this pattern of results suggests that the above dissociation between WM cost in the anterior IPS and the WM-benefit in posterior SPL may reflect functionally distinct operations in fractionated neural regions, rather than reflecting the operation of a single process associated with attentional capture by the contents of WM. We address this point further in the discussion.

Individual difference based on fMRI ROI data We assessed any correspondence between PPC structures and the functional (BOLD) activation during the task with regard to individual variation in WM-costs and WM-benefits on search. Specifically, it was predicted that functional activity in the clusters derived from the structural VBM analyses ought to be correlated with the size of the WM-costs and WM-benefits on search behavior. We therefore computed the correlation between WM-costs and WM-benefits in RT and the functional activations for the cost contrast (BOLD Invalid N BOLD Neutral) and the benefit contrast (BOLD Valid N BOLD Neutral) within the VBM cluster associated with WM-costs and WM-benefits respectively. Note that the region of interest was based on the structural data (which is independent from the functional data) with the aim of assessing the correlation of behavior with the functional activations. Variation in the cost-related activation in the anterior IPS was significantly correlated with the RT cost effect of individual subjects (r = 0.296 b 0.025, one-tailed) (Fig. 5). However, activations in the

Memory performance was very high (mean correct = 92.4%) but not perfect. Given the increasing evidence for functional overlap for memory and visual attention processes (i.e. Olivers et al., 2011; Soto et al., 2008; Theeuwes et al., 2009) we assessed whether the VBM clusters associated with the size of the WM bias on search behavior could be also predictive of performance in the delayed recognition test. Interestingly we found significant correlations between the accuracy in the recognition test and PPC structure. We performed ROI based analyses based on the VBM clusters associated with the WM-benefit, around the left SPL and WM-cost, around the anterior IPS. Small volume correction analyses at left SPL (x = − 32, y = −55, z = 52) with a sphere (10 mm radius) showed positive correlation with memory performance (P(corr) = 0.001, r = 0.62, t(38) = 4.82. peak coordinate x = − 33, y = − 58, z = 52) (see Fig. 6). This finding shows that high GM volume in the left SPL is associated with both enhanced WM benefits on selection and strengthening of the WM representation. There was no correlation between gray matter volume around anterior IPS (associated with the WM-cost) and memory performance. Thus only the structure of the VBM cluster associated with WM benefits on search performance was correlated with memory performance. We note however that inspection of Fig. 6 may suggest that the above correlation between SPL structure and memory performance might have been driven by some of the participants with the lower memory accuracy. Because the memory-task was relatively easy, it is possible that those participants with lower memory performance may be less motivated or task-compliant, and this should be considered when interpreting the generality of this particular result. Discussion The main aim of the current study was to investigate how individual variability in brain structure and function may explain the magnitude of WM biases in visual search. VBM analyses revealed that anterior IPS was associated with variation in WM-costs by the presence of irrelevant WM items in search. Increased GM volume in this region was associated with higher WM-costs and hence a reduced ability to inhibit search interference from invalid WM contents. Increased BOLD responses in the anterior IPS to the presence of WM distracters in search (relative to a neutral baseline) were also associated with larger WM costs in search. Gray matter volume in left SPL on the other hand was positively correlated with the WM-benefits on search from valid WM cues, with higher gray GM associated with larger WM-benefits. Further, SPL volume correlated with memory performance in the delayed recognition test. On the other hand, BOLD responses in SPL to the presence of valid WM items in search did not correlate with the WM-benefits in search RTs, however. This null BOLD result does not match the VBM association between SPL structure and WM-benefits on search behavior, however, caution must be taken in making inferences from this null finding because, to the best of our knowledge, there is no evidence that VBM and BOLD may necessarily reflect the operation of similar neural mechanisms. Indeed few studies to date in addition to the present one have directly compared VBM and BOLD data within the same experimental protocol (see further discussion on this point below). Together these findings indicate two fractionated regions in the left PPC that appear specifically associated with individual variation in WM-costs on selection (aIPS) on the one hand and with interindividual variation in WM-benefits on the other (SPL). This argument is also consonant with the lack of correlation between the behavioral manifestation of WM-costs and WM-benefits on search and with the

294

D. Soto et al. / NeuroImage 89 (2014) 289–296

Fig. 4. Structural correlates of WM-cost and WM-benefit effects on search RTs. (A) The significant clusters for WM-cost effects (red) and WM benefit effects (green) are rendered together on a standard brain. (B) The significant cluster for WM-cost effects is displayed on a standard brain image. The left inferior parietal lobe correlated with individual differences in WM-cost effects. (C) The gray matter density at the peak voxel for the WM-cost cluster is plotted against WM-cost effect. The notations are identical to panel B. (D) The significant cluster for WM benefit effects is displayed on a standard brain image. The left superior parietal lobe correlated with individual differences in WM benefit effects. The whole brain gray matter volume and the experiment factors were regressed out and the residuals are displayed in arbitrary unit (a.u.). Data from each of the four experiments are shown in different colored markers. Regression lines are shown by the dotted lines separately for each experiment to illustrate the consistent trends across all the experiments. The solid black line is the regression line for the data points collapsed across all the experiments. (E) The gray matter density at the peak voxel for the WM benefit cluster is plotted against WM benefit effect. The notations are identical to panel B.

absence of correlation between GM volume of aIPS and the volume of the SPL clusters associated with behavioral costs and benefits respectively. We propose that WM-costs and WM-benefits on selection do not appear to be the consequence of the operation of a single process (i.e. the capture of attention by stimuli matching the WM contents). The data suggest that left SPL and inferior PPC around aIPS may be associated with different forms of cognitive control applied to WM signals based on current selection goals, which are determined by the validity status of the memory content on a given trial. We propose that left aIPS may be important for the re-deployment of attention following capture by invalid WM distracters, in keeping with prior evidence that the process of ignoring salient distracters is dependent on the left PPC (Kanai et al., 2011; Mevorach et al., 2006). On the other hand, the SPL may support automatic selection of items that appear congruent with the search goals (i.e. valid memory matches), in keeping with a neural architecture of the PPC along a dorsal-superior/ventral-inferior axis

whereby more superior regions determine the voluntary allocation of attention based goal-relevant cues (Corbetta and Shulman, 2002; Egner et al., 2008). It might be argued that larger GM volume ought to be associated with improved function; accordingly, observers with larger aIPS ought to have displayed improved ability to ignore WM-distracters and reorienting attention to the sought target (i.e. reduced WM-costs). This was not the case. Similar findings to ours have been observed however; for example, Kanai et al. (2011) reported that larger SPL made individuals more distractible by bottom-up salient distracters; see also Hyde et al. (2007, 2010). These findings are usually interpreted as reflecting a higher processing efficiency in lower GM regions (Kanai et al., 2011), which is consonant with the course of brain development and with the fact that more mature and efficient brains undergo a reduction in GM in attention areas in the parietal cortex (Gogtay et al., 2004). Reproducing our previous results (Grecucci et al., 2010; Rotshtein et al., 2011; Soto et al., 2007, 2012a,b), the whole-brain analyses of the

D. Soto et al. / NeuroImage 89 (2014) 289–296

295

Fig. 5. Relationship between functional activation and WM-costs and WM-benefits on search performance. (A) Functional activation (beta) for cost effects (Invalid N Neutral) extracted within the inferior PPC–aIPS cluster from the VBM results is plotted against the costs in search RTs; (B) functional activation (beta) for benefit effects (Valid N Neutral) extracted within the SPL benefit-cluster from the VBM results is plotted against the WM-benefits in RTs.

fMRI data revealed a network of prefrontal and superior frontal regions along with temporo-occipital and thalamic nuclei involved in the WM biasing of visual selection. We note that the whole-brain analyses of the fMRI data revealed a cluster in left PPC associated with the WMvalidity effect on selection. This contrasts with the absence of PPC engagement across each of the previous studies assessing WM biases of selection (Soto et al., 2007, 2012b,c), which may be attributed to large inter-individual variability in PPC recruitment during the interaction between WM and attention. The VBM-clusters in the left PPC associated with inter-individual variation in the expression of WM biases were distinct however from the PPC region showing more consistent WMvalidity effect across subjects in the functional activations. While it is common to find correlation between performance and gray matter volume in regions that exhibit BOLD activations in related cognitive tasks (Kanai and Rees, 2011), only a few studies in addition to the present one have directly compared structural differences with differences in functional activation within the same group of participants. A recent study (Ilg et al., 2008) examined how training on reading mirrored words changes gray matter volume and functional activation within the same participants. Importantly, the locus at which gray matter volume associated with the mirror-reading training matched the locus where functional activation increased after training. Functional activation seems to be also linked with underlying white matter connectivity, which can be measured by fractional anisotropy (FA) derived from diffusion-tensor imaging (DTI). In a WM task, the individual differences in FA values in frontoparietal regions correlated with the level of BOLD activation in closely located superior frontal sulcus (Olesen et al.,

Fig. 6. Structural correlates of WM-performance in the left SPL.

2003). These studies suggest that inter-individual variability in the level of functional activation might reflect differences in gray matter and white matter structure mediating the functional activation but the nature of correspondence between structural and functional measures is not yet well specified and awaits further investigation. It is also interesting to compare the current findings and the prior findings on the neural basis of WM-attention interplay. While the functional MRI results clearly indicate the involvement of fronto-thalamic and temporal-occipital regions in WM guidance as previous demonstrated in the prior studies (Grecucci et al., 2010; Rotshtein et al., 2011; Soto et al., 2007, 2012a,b), their structural variability did not account for inter-individual variation in the effect of WM on selection. We propose that WM biases of selection are characterized by increased neural responses in superior frontal and posterior visual regions that are sensitive to the re-appearance of a WM item in the search array. A further network involving more anterior pre-frontal regions and thalamic nuclei may react to the validity of the WM contents for the selection goal (Grecucci et al., 2010; Rotshtein et al., 2011; Soto et al., 2007) modulating further responses in earlier visual regions (Soto et al., 2012a). The present findings suggest however that these two neural mechanisms may be further modulated by the left PPC. Additional work is needed however to elucidate the interplay between the PPC and frontothalamic and temporal-occipital regions for the control of WM biases. A recent study by Soto and colleagues has provided evidence that PPC, as part of a parieto-middle temporal pathway, may support the strategic cognitive control of WM biases based on foreknowledge about the validity of the WM content for attention (i.e. to enhance/inhibit biases when the WM content is known to cue a target or a distracter in search) (Soto et al., 2012). Notably, PPC involvement in the above study was associated with the implementation of preparatory control processes linked to the expectation that the WM cue will be valid or invalid for search. In the present study, the WM cues were equally likely to be valid, neutral or invalid, hence the influence of expectancy-driven top-down control must be greatly reduced in the present study by comparison. Because of this, caution needs to be exerted in making direct comparisons between the two studies. Notwithstanding, the fact that PPC involvement was found at all in these two studies may be informative about the type of cognitive control processes that are supported by the PPC in the context of WM biases of visual attention and which arguably may reflect (i) the utilization of target foreknowledge about WM validity to modulate the implementation of an appropriate attentional set prior to search and (ii) determine inter-individual variation in the extent to which this foreknowledge is implemented in the attention set or (iii) the weighting given to irrelevant/ accessory WM contents and hence the degree with which those accessory memory signals are capable of (inadvertently) biasing attention.

296

D. Soto et al. / NeuroImage 89 (2014) 289–296

Beyond the PPC role in attentional control (i.e. in the memory domain), superior regions of the PPC are thought to support processes taking place prior to retrieval of memory contents such as voluntary directing of attention towards information that may be relevant for retrieval and memory decision-making (Ciaramelli et al., 2010). Here we found that gray matter and the functional activation of the SPL region correlated with improved attention and memory performance in this study. Further, the locus of the WM benefit effect in the SPL (MNI x = −32, y = −55, z = 52) closely matches the locus of PPC activation found in prior studies of recognition memory; MNI x = −34, y = −56, z = 42 (Ciaramelli et al., 2008); MNI x = −36, y = − 57, z = 42 (Ciaramelli et al., 2010); MNI: x = −48, y = −48, z = 51 (Vilberg and Rugg, 2009). Further work is needed however to better understand whether and how individual variability in specific memory substrates (i.e. hippocampus) and specific memory operations may be associated with the deployment of visual attention by information held in memory. We conclude that left PPC structure and function account for human variation in WM guidance by determining the strength that feature memory signals are assigned to bias visual selection. Remarkably, distinct regions within the left PPC determine the efficiency of cognitive control applied to the contents of WM based on current attention goals. Acknowledgments We thank Glyn W. Humphreys for his continued support. This work has been funded by a grant from the Medical Research Council (UK, 89631). References Carlisle, N.B., Woodman, G.F., 2011. Automatic and strategic effects in the guidance of attention by working memory representations. Acta Psychol. (Amst.) 137, 217–225. Ciaramelli, E., Grady, C.L., Moscovitch, M., 2008. Top-down and bottom-up attention to memory: a hypothesis (AtoM) on the role of the posterior parietal cortex in memory retrieval. Neuropsychologia 46, 1828–1851. Ciaramelli, E., Grady, C., Levine, B., Ween, J., Moscovitch, M., 2010. Top-down and bottomup attention to memory are dissociated in posterior parietal cortex: neuroimaging and neuropsychological evidence. J. Neurosci. 30, 4943–4956. Corbetta, M., Shulman, G.L., 2002. Control of goal-directed and stimulus-driven attention in the brain. Nat. Rev. Neurosci. 3, 201–215. Desimone, R., Duncan, J., 1995. Neural mechanisms of selective visual attention. Annu. Rev. Neurosci. 18, 193–222. Duncan, J., Humphreys, G.W., 1989. Visual search and stimulus similarity. Psychol. Rev. 96, 433–458. Egner, T., Monti, J.M., Trittschuh, E.H., Wieneke, C.A., Hirsch, J., Mesulam, M.M., 2008. Neural integration of top-down spatial and feature-based information in visual search. J. Neurosci. 28, 6141–6151. Fleming, S.M., Weil, R.S., Nagy, Z., Dolan, R.J., Rees, G., 2010. Relating introspective accuracy to individual differences in brain structure. Science 329, 1541–1543. Gogtay, N., Giedd, J.N., Lusk, L., Hayashi, K.M., Greenstein, D., Vaituzis, A.C., Nugent III, T.F., Herman, D.H., Clasen, L.S., Toga, A.W., Rapoport, J.L., Thompson, P.M., 2004. Dynamic mapping of human cortical development during childhood through early adulthood. Proc. Natl. Acad. Sci. U. S. A. 101, 8174–8179. Grecucci, A., Soto, D., Rumiati, R.I., Humphreys, G.W., Rotshtein, P., 2010. The interrelations between verbal working memory and visual selection of emotional faces. J. Cogn. Neurosci. 22, 1189–1200. Han, S.W., Kim, M.S., 2009. Do the contents of working memory capture attention? Yes, but cognitive control matters. J. Exp. Psychol. Hum. Percept. Perform. 35, 1292–1302. Hayasaka, S., Phan, K.L., Liberzon, I., Worsley, K.J., Nichols, T.E., 2004. Nonstationary cluster-size inference with random field and permutation methods. NeuroImage 22, 676–687. Hyde, K.L., Lerch, J.P., Zatorre, R.J., Griffiths, T.D., Evans, A.C., Peretz, I., 2007. Cortical thickness in congenital amusia: when less is better than more. J. Neurosci. 27, 13028–13032.

Hyde, K.L., Samson, F., Evans, A.C., Mottron, L., 2010. Neuroanatomical differences in brain areas implicated in perceptual and other core features of autism revealed by cortical thickness analysis and voxel-based morphometry. Hum. Brain Mapp. 31, 556–566. Ilg, R., Wohlschlager, A.M., Gaser, C., Liebau, Y., Dauner, R., Woller, A., Zimmer, C., Zihl, J., Muhlau, M., 2008. Gray matter increase induced by practice correlates with taskspecific activation: a combined functional and morphometric magnetic resonance imaging study. J. Neurosci. 28, 4210–4215. Kanai, R., Rees, G., 2011. The structural basis of inter-individual differences in human behaviour and cognition. Nat. Rev. Neurosci. 12, 231–242. Kanai, R., Dong, M.Y., Bahrami, B., Rees, G., 2011. Distractibility in daily life is reflected in the structure and function of human parietal cortex. J. Neurosci. 31, 6620–6626. Kiyonaga, A., Egner, T., Soto, D., 2012. Cognitive control over working memory biases of selection. Psychon. Bull. Rev. 19, 639–646. Mannan, S.K., Kennard, C., Potter, D., Pan, Y., Soto, D., 2010. Early oculomotor capture by new onsets driven by the contents of working memory. Vision Res. 50, 1590–1597. Mevorach, C., Humphreys, G.W., Shalev, L., 2006. Opposite biases in salience-based selection for the left and right posterior parietal cortex. Nat. Neurosci. 9, 740–742. Olesen, P.J., Nagy, Z., Westerberg, H., Klingberg, T., 2003. Combined analysis of DTI and fMRI data reveals a joint maturation of white and grey matter in a fronto-parietal network. Brain Res. Cogn. Brain Res. 18, 48–57. Olivers, C.N., 2009. What drives memory-driven attentional capture? The effects of memory type, display type, and search type. J. Exp. Psychol. Hum. Percept. Perform. 35, 1275–1291. Olivers, C.N., Eimer, M., 2011. On the difference between working memory and attentional set. Neuropsychologia 49, 1553–1558. Olivers, C.N., Meijer, F., Theeuwes, J., 2006. Feature-based memory-driven attentional capture: visual working memory content affects visual attention. J. Exp. Psychol. Hum. Percept. Perform. 32, 1243–1265. Olivers, C.N., Peters, J., Houtkamp, R., Roelfsema, P.R., 2011. Different states in visual working memory: when it guides attention and when it does not. Trends Cogn. Sci. 15, 327–334. Pan, Y., Soto, D., 2010. The modulation of perceptual selection by working memory is dependent on the focus of spatial attention. Vision Res. 50, 1437–1444. Rotshtein, P., Soto, D., Grecucci, A., Geng, J.J., Humphreys, G.W., 2011. The role of the pulvinar in resolving competition between memory and visual selection: a functional connectivity study. Neuropsychologia 49, 1544–1552. Schenkluhn, B., Ruff, C.C., Heinen, K., Chambers, C.D., 2008. Parietal stimulation decouples spatial and feature-based attention. J. Neurosci. 28, 11106–11110. Schwarzkopf, D.S., Song, C., Rees, G., 2011. The surface area of human V1 predicts the subjective experience of object size. Nat. Neurosci. 14, 28–30. Soto, D., Humphreys, G.W., 2007. Automatic guidance of visual attention from verbal working memory. J. Exp. Psychol. Hum. Percept. Perform. 33, 730–737. Soto, D., Heinke, D., Humphreys, G.W., Blanco, M.J., 2005. Early, involuntary top-down guidance of attention from working memory. J. Exp. Psychol. Hum. Percept. Perform. 31, 248–261. Soto, D., Humphreys, G.W., Rotshtein, P., 2007. Dissociating the neural mechanisms of memory-based guidance of visual selection. Proc. Natl. Acad. Sci. U. S. A. 104, 17186–17191. Soto, D., Hodsoll, J., Rotshtein, P., Humphreys, G.W., 2008. Automatic guidance of attention from working memory. Trends Cogn. Sci. 12, 342–348. Soto, D., Mok, A.Y., McRobbie, D., Quest, R., Waldman, A., Rotshtein, P., 2011. Biasing visual selection: functional neuroimaging of the interplay between spatial cueing and feature memory guidance. Neuropsychologia 49, 1537–1543. Soto, D., Greene, C.M., Chaudhary, A., Rotshtein, P., 2012a. Competition in working memory reduces frontal guidance of visual selection. Cereb. Cortex 22, 1159–1169. Soto, D., Rotshtein, P., Hodsoll, J., Mevorach, C., Humphreys, G.W., 2012b. Common and distinct neural regions for the guidance of selection by visuoverbal information held in memory: converging evidence from fMRI and rTMS. Hum. Brain Mapp. 33, 105–120. Soto, D., Greene, C.M., Kiyonaga, A., Rosenthal, C.R., Egner, T., 2012c. A parieto-medial temporal pathway for the strategic control over working memory biases in human visual attention. J. Neurosci. 32, 17563–17571. Theeuwes, J., Belopolsky, A., Olivers, C.N., 2009. Interactions between working memory, attention and eye movements. Acta Psychol. (Amst.) 132, 106–114. Toth, L.J., Assad, J.A., 2002. Dynamic coding of behaviourally relevant stimuli in parietal cortex. Nature 415, 165–168. Vilberg, K.L., Rugg, M.D., 2009. Functional significance of retrieval-related activity in lateral parietal cortex: evidence from fMRI and ERPs. Hum. Brain Mapp. 30, 1490–1501. Woodman, G.F., Luck, S.J., 2007. Do the contents of visual working memory automatically influence attentional selection during visual search? J. Exp. Psychol. Hum. Percept. Perform. 33, 363–377.

Parietal structure and function explain human variation in working memory biases of visual attention.

Recent research indicates that human attention appears inadvertently biased by items that match the contents of working memory (WM). WM-biases can lea...
910KB Sizes 0 Downloads 0 Views