[1] 10w Memory-guided attention: bilateral hippocampal volume positively predicts implicit contextual learning

[1] 153w Several studies have begun to demonstrate that contextual memories constitute an important mechanism to guide our attention. Although there is general consensus that the hippocampus is involved in the encoding of contextual memories, it is controversial whether this structure can support implicit forms of contextual memory. Here, we combine automated segmentation of structural MRI with neurobehavioral assessment of implicit contextual memory-guided attention to test the hypothesis that hippocampal volume would predict the magnitude of implicit contextual learning. Forty healthy subjects underwent 3T magnetic resonance imaging brain scanning with subsequent automatic measurement of the total brain and hippocampal (right and left) volumes. Implicit learning of contextual information was measured using the contextual cueing task. We found that both left and right hippocampal volumes positively predicted the magnitude of implicit contextual learning. Larger hippocampal volume was associated with superior implicit contextual memory performance. This study provides compelling evidence that implicit contextual memory-guided attention is hippocampus-dependent.

[1] 121w In addition to its traditional role in episodic memory and spatial navigation, a substantial body of research in humans has shown that the hippocampus is also involved in contextual processing. Specifically, it has been reported that the hippocampus is critical for encoding of object-context associations (Henke et al. 1997;Cohen et al. 1999;Giovanello et al. 2003;Staresina and Davachi 2008), to establish associations between objects and their spatial location (Burgess et al. 2002;Ekstrom et al. 2003;Olson et al. 2006;Piekema et al. 2006) and the acquisition of contextual fear conditioning (Alvarez et al. 2008;Lang et al. 2009;Pohlack et al. 2012). Collectively, these studies suggest that one of the most important functions of the hippocampus consist of binding elements with their context to form memories.

[2] 76w While it is generally agreed that encoding of contextual or relational memories depends critically on the hippocampus, the selective role of the hippocampus in conscious expressions of these memories has been a subject of debate (Chun and Phelps 1999;Manns and Squire 2001;Park et al. 2004;Greene et al. 2007;Preston and Gabrieli 2008;Westerberg et al. 2011). Traditional models of hippocampal function emphasize the role of the hippocampus in the conscious Frauke Nees and Herta Flor joint last authors.

[1] 55w Less than 3% of all trials were omitted from the analyses due to incorrect trials and trials with RT outside of the 0.5-6 s range. Overall accuracy for the new and repeated trials was 99.14 ± 0.9% and 99.34 ± 0.67%, respectively. Accuracy did not differ by array type (t(39) = 1.67, p = 0.102).

[2] 120w Consistent with previous findings (Chun and Jiang 1998;Manelis and Reder 2012) we found a significant main effect of context (novel vs. repeated), F(1,39) = 15.41, p < 0.001), indicating that search RTs were faster for repeated than for novel contexts. There was a significant main effect of epoch, F(1,39) = 169.95, p < 0.001, indicating that search RTs decreased during the task for both types of contexts. In addition, the epoch × context interaction was significant, F(1,39) = 23.25, p < 0.001, indicating that during the task, search RTs for repeated contexts decreased more than for novel contexts (Fig. 2). The magnitude of contextual cueing was significantly greater than zero (0.125 ± 0.129 s, t(39) = 6.09, p < 0.001).

[3] 137w Before the explicit recognition test, the participants were asked if they noticed that some displays were repeated during the experiment. None of the participants reported noticing the repeated displays. For the explicit recognition test, the probability that repeated contexts were correctly recognized (hit rate) was 0.32, and this did not differ from the probability that new contexts were misidentified as repeated contexts (false alarm), 0.26, (t(39) = 1.90, p = 0.064). Furthermore, neither the hit rate nor the corrected hit rate correlated with the magnitude of the contextual cueing effect (r = 0.04, p = 0.760, r = -0.01, p = 0.934, respectively). Likewise, the corrected hit rate did not correlate with either left (r = 0.04, p = 0.760) or right hippocampal volumes. Taken together, there was no clear evidence of explicit recognition of repeated contexts.

[4] 403w Linear regression analyses conducted on the entire sample (n = 40) revealed that left hippocampal volume significantly predicted contextual cueing performance (β = 0.51, t(38) = 3.69, p = 0.001, r 2 = 0.26). Right hippocampal volume also significantly predicted contextual cueing performance (β = 0.43, t(38) = 2.96, p = 0.005, r 2 = 0.18), which is illustrated in Fig. 3. As expected, we found not significant effects of either left (β = 0.30, t(38) = 1.97, p = 0.056, r 2 = 0.09) or right (β = -0.00, t(38) = -0.02, p = 0.979, r 2 = 0.00) amygdalar, left (β = -0.07, t(38) = -0.47, p = 0.639, r 2 = 0.01) or right (β = -0.07, t(38) = -0.04, p = 0.629, r 2 = 0.01) caudate nucleus, left (β = 0.12, t(38) = 0.80, p = 0.426, r 2 = 0.02) or right (β = 0.08, t(38) = 0.54, p = 0.589, r 2 = 0.01) putamen, left (β = 0.03, t(38) = 0.20, p = 0.842, r 2 = 0.00) or right (β = -0.14, t(38) = -0.91, p = 0.367, r 2 = 0.02) pallidum, or left (β = 0.29, t(38) = 1.88, p = 0.068, r 2 = 0.09) or right (β = 0.30, t(38) = 1.96, p = 0.057, r 2 = 0.09) cerebellar cortex volumes, in predicting contextual cueing performance (Fig. 3). Similar results were obtained including only those participants who exhibited a positive contextual cueing effect (n = 34). The results revealed that either left (β = 0.38, t(32) = 2.36, p = 0.024, r 2 = 0.15) or right (β = 0.41, t(32) = 2.60, p = 0.014, r 2 = 0.17) hippocampal volumes significantly predicted contextual cueing performance. On the contrary, either left (β = 0.24, t(32) = 1.43, p = 0.161, r 2 = 0.06) or right (β = 0.07, t(32) = 0.40, p = 0.691, r 2 = 0.00) amygdalar, left (β = -0.05, t(32) = -0.28, p = 0.744, r 2 = 0.00) or right (β = -0.02, t(32) = -0.14, p = 0.888, r 2 = 0.00) caudate nucleus, left (β = 0.18, t( 32 With regard to the relationship between cortical volumes and contextual cueing performance, linear regression analyses conducted on the entire sample (n = 40) revealed that formance. Solid lines represent the line of best fit, and dashed lines represent 95% confidence intervals

[5] 228w both left (β = 0.37, t(38) = 2.47, p = 0.018, r 2 = 0.14) or right (β = 0.50, t(38) = 3.53, p = 0.001, r 2 = 0.25) parahippocampal gyrus, left (β = 0.41, t(38) = 2.79, p = 0.008, r 2 = 0.17) entorhinal cortex, left (β = 0.38, t(38) = 2.51, p = 0.016, r 2 = 0.14) supramarginal gyrus, or left (β = 0.37, t(38) = 2.48, p = 0.017, r 2 = 0.14) isthmus of the cingulate cortex volumes, significantly predicted contextual cueing performance, and (Table 1). In contrast, no significant relationships between other cortical regions (Table 1), or total gray matter volumes (β = 0.27, t(38) = 1.72, p = 0.092, r 2 = 0.07) and contextual cueing performance were observed. Finally, we conducted linear regression analyses including only those participants who showed a positive contextual cueing effect (n = 34). revealed that both left (β = 0.46, t(32) = 2.94, p = 0.006, r 2 = 0.21) or right (β = 0.40, t(32) = 2.48, p = 0.018, r 2 = 0.16) parahippocampal gyrus, left (β = 0.38, t(32) = 2.51, p = 0.016, r 2 = 0.14) supramarginal gyrus, or left (β = 0.41, t(32) = 2.54, p = 0.016, r 2 = 0.17) isthmus of the cingulate cortex volumes significantly predicted contextual cueing performance (Supplementary Table 1). Conversely,

[6] 429w Table 1 Summary of linear regression analysis results predicting contextual cueing performance from cortical volumes of Desikan parcellations *P < 0.05, **P < 0.01 Cortical regions Left hemisphere Right hemisphere β t(38) P value r 2 β t(38) P value r 2 Temporal lobe Entorhinal cortex 0.41 2.79 0.008** 0.17 0.13 0.78 0.437 0.02 Parahippocampal gyrus 0.37 2.47 0.018* 0.14 0.50 3.53 0.001** 0.25 Temporal pole 0.15 0.94 0.353 0.02 0.14 0.89 0.379 0.02 Fusiform gyrus -0.13 -0.78 0.440 0.02 0.05 0.33 0.746 0.00 Superior temporal gyrus 0.16 1.01 0.321 0.03 0.15 0.96 0.343 0.02 Middle temporal gyrus -0.01 -0.07 0.944 0.00 0.05 0.30 0.763 0.00 Inferior temporal gyrus 0.22 1.36 0.182 0.05 0.28 1.77 0.085 0.08 Transverse temporal cortex 0.13 0.79 0.437 0.02 0.15 0.95 0.346 0.02 Banks of the superior temporal sulcus 0.16 0.98 0.334 0.02 -0.05 -0.33 0.743 0.00 Frontal lobe Superior frontal gyrus 0.12 0.72 0.473 0.01 -0.08 -0.48 0.632 0.01 Rostral middle frontal gyrus -0.09 -0.53 0.598 0.01 -0.04 -0.22 0.824 0.00 Caudal middle frontal gyrus -0.07 -0.44 0.662 0.01 -0.09 -0.58 0.568 0.01 Inferior frontal gyrus: pars opercularis 0.02 0.10 0.918 0.00 -0.23 -1.46 0.153 0.05 Inferior frontal gyrus: pars triangularis 0.05 0.29 0.774 0.00 0.02 0.09 0.925 0.00 Inferior frontal gyrus: pars orbitalis -0.07 -0.42 0.675 0.00 0.18 1.14 0.261 0.03 Lateral orbitofrontal cortex 0.20 1.27 0.211 0.04 0.16 0.98 0.331 0.02 Medial orbitofrontal cortex 0.09 0.53 0.596 0.01 0.10 0.64 0.527 0.01 Frontal pole 0.27 1.72 0.093 0.07 0.18 1.14 0.263 0.03 Precentral gyrus -0.06 -0.38 0.703 0.00 0.03 0.18 0.858 0.00 Paracentral lobule -0.03 -0.21 0.835 0.00 -0.19 -1.20 0.238 0.04 Parietal lobe Postcentral gyrus 0.15 0.91 0.371 0.02 0.29 1.87 0.069 0.08 Supramarginal gyrus 0.38 2.51 0.016* 0.14 -0.03 -0.16 0.871 0.00 Superior parietal cortex -0.13 -0.81 0.423 0.02 -0.22 -1.38 0.176 0.05 Inferior parietal cortex -0.03 -0.17 0.866 0.00 0.14 0.85 0.403 0.02 Precuneus cortex 0.08 0.47 0.639 0.01 0.09 0.57 0.574 0.01 Occipital lobe Lingual gyrus 0.19 1.19 0.241 0.04 0.27 1.71 0.095 0.07 Pericalcarine cortex -0.07 -0.41 0.683 0.00 0.05 0.34 0.739 0.00 Cuneus cortex -0.01 -0.03 0.972 0.00 0.18 1.13 0.267 0.03 Lateral occipital cortex 0.02 0.14 0.888 0.00 0.01 0.06 0.953 0.00 Rostral anterior cingulate cortex -0.11 -0.68 0.500 0.01 -0.18 -1.13 0.265 0.03 Caudal anterior cingulate cortex -0.21 -1.36 0.183 0.05 -0.07 -0.46 0.650 0.01 Posterior cingulate cortex 0.03 0.21 0.838 0.00 -0.05 -0.32 0.751 0.00 Isthmus cingulate cortex 0.37 2.48 0.017* 0.14 0.22 1.42 0.164 0.05 no significant relationships between other cortical regions (Supplementary Table 1) and contextual cueing performance were found.

[1] 321w Contextual memories are an important mechanism to generate predictions (Bar and Ullman 1996;Hyman et al. 2012), to guide our attention (Chun and Jiang 1998;Auckland et al. 2007;Turk-Browne et al. 2008;Hannula and Ranganath 2009;Goldfarb et al. 2016), to modulate our emotion (Rudy et al. 2004;Brewin et al. 2007;Lang et al. 2009;Acheson et al. 2012;Maren et al. 2013) and ultimately to construct meaning. Since contextual memories play a pivotal role in perception and cognition, there has been an intense interest in understanding the neural mechanisms that support such relational memories. Neuropsychological, electrophysiological, and neuroimaging studies have suggested that the hippocampus is a key structure for contextual memory (Burgess et al. 2002;Ekstrom et al. 2003;Manelis et al. 2011) and that the hippocampus can support implicit forms of contextual memory (Chun and Phelps 1999;Turk-Browne et al. 2008, 2010;Hannula and Ranganath 2009). Here we investigated the role of hippocampal volume in predicting contextual learning in healthy humans. Consistent with our predictions, we found that hippocampal, but not, amygdalar, caudate, putamen, pallidum, or cerebellar cortex volumes predicted the magnitude of contextual learning. Importantly, left and right parahippocampal gyrus left supramarginal gyrus and left isthmus of the cingulate cortex volumes also predicted contextual cueing performance. The results reported here supported convergent evidence that the implicit contextual learning is hippocampus-dependent. The present data were largely consistent with previous neuropsychological and pharmacological studies on the hippocampus involvement in contextual cueing effect (Chun and Phelps 1999;Park et al. 2004). Our results also complement previous structural MRI findings that shows a specific link between hippocampal volume and contextual memory performance (Rajah et al. 2010;Horner et al. 2012;Pohlack et al. 2012). Similarly, these results also are congruent with previous fMRI and PET studies that has shown increased hippocampal activity during the processing of contextual associations (Henke et al. 1997;Cohen et al. 1999;Greene et al. 2007;Alvarez et al. 2008;Staresina and Davachi 2008;Lang et al. 2009;Manelis and Reder 2012;Giesbrecht et al. 2013;Goldfarb et al. 2016).

[2] 92w In agreement with previous reports, our results indicate that the facilitation for repeated context did not depend on explicit recognition. These results are consistent with previous studies showing that the hippocampus participates in some kinds of implicit learning (Hannula and Ranganath 2009;Turk-Browne et al. 2008, 2010) and support the view that the hippocampus is not only responsible for integrating contextual information, but also that these processes maybe done in the absence of explicit recognition (Cohen et al. 1997;Konkel and Cohen 2009;Reder et al. 2009: Henke 2010;Hannula and Ranganath 2009;Hannula and Greene 2012).

[3] 101w Considered together, these results are consistent with relational memory theory (Eichenbaum et al. 1994), which suggests that the hippocampus is critical for forming associations among items of events. Contrary to the idea that the hippocampus plays a selective role in conscious memory (Squire 2004), an extension of the relational memory theory (Cohen et al. 1997;Ryan and Cohen 2003;Moses and Ryan 2006;Moscovitch 2008) suggests that the hippocampus can support relational memory in the absence of consciousness and that the conscious awareness of encoded and retrieved content is determined by the interaction of the hippocampus with other structures (e.g., prefrontal and parietal regions).

[4] 273w In addition, this study provides evidence that three cortical regions volumes also predicted contextual cueing performance. These areas include the parahippocampal gyrus, the supramarginal gyrus, as well as the isthmus of the cingulate cortex. The involvement of parahippocampal gyrus in implicit contextual memory is consistent with previous observations that parahippocampal cortex responds preferentially to the global spatial configuration of the scenes independent of individual items in the scenes (Epstein and Kanwisher 1998;Epstein and Ward 2010). Likewise, our finding is in agreement with previous reports of parahippocampal cortex engagement in the encoding of spatial contextual information (Bar et al. 2008;Aminoff et al. 2013). Importantly, it has been suggested that the parahippocampal cortex supports the visual encoding of contexts, independently of the fact that this learning is accessible or not accessible to consciousness (Aminoff et al. 2013). Similarly, our results extend previous findings that have implicated the isthmus of the cingulate cortex (retrosplenial cortex) in the processing of contextual associations (Aminoff et al. 2008;Peters et al. 2009;Kveraga et al. 2011), suggesting that the retrosplenial cortex plays a significant role in translating a particular context into a more abstract representation of that context. In addition to the supramarginal gyrus involvement in phonological short-term memory (Buchsbaum and D'Esposito 2009) and long term memory retrieval (Rosen et al. 2017), recent evidence indicates that the ventral posterior parietal cortex (including the supramarginal gyrus) activity also occurs for implicit retrieval (Elman and Shimamura 2011). Taken together, these findings suggest that the hippocampus, parahippocampal cortex, supramarginal gyrus, and the isthmus of the cingulate cortex are part of a brain network that supports both implicit and explicit contextual memory formation and retrieval.

[5] 110w Although the aim of the current study was to examine if hippocampal volume would positively predict contextual cueing performance, the hippocampal system does not tend to work in isolation. Visual search facilitation in repeated contexts could involve reciprocal interactions between medial temporal lobe structures and both dorsal and ventral attentional networks (Hutchinson and Turk-Browne 2012) and the striatum (Goldfarb et al. 2016). Consequently, voluntary control of attention; stimulus-response associations, and exogenous attentional load can interfere with the encoding and expression of implicit memories (Lee et al. 2013;Turk-Browne et al. 2013;Goldfarb et al. 2016). Future research should examine the influence of both dorsal and ventral attentional networks in implicit contextual learning.

[1] 151w The online version of this article (https ://doi.org/10.1007/s0042 9-019-01887 -9) contains supplementary material, which is available to authorized users. retrieval of past events (Tulving and Schacter 1990;Moscovitch 1992;Squire 2004). According to those theories, the hippocampus exclusively support the conscious recollection of event and facts, while other forms of learning that does not require conscious or deliberative access, depend on different brain systems such as the amygdala, caudate, putamen, globus pallidus and cerebellum (Tulving and Schacter 1990;Moscovitch et al. 2006). However, there is evidence indicating that the hippocampus is also involved in implicit forms of learning (Turk-Browne et al. 2008, 2010;Hannula and Ranganath 2009), including contextual facilitation, suggesting that the hippocampus is not only a brain region in which contextual representations may be formed but that it is also important when this process does not correspond to a conscious experience (Cohen et al. 1997;Konkel and Cohen 2009;Henke 2010;Ranganath 2010;Hannula and Greene 2012).

[2] 293w One important source of evidence for the role of the hippocampus in implicit contextual memory comes from the use of the contextual cueing paradigm developed by Chun and Jiang (1998), which has been shown to be sensitive to hippocampal function. In this paradigm, subjects are encouraged to search a target item (e.g. a letter T) embedded in a spatial array of distractors items (e.g. a set of the letters L). When the target is embedded in a repeated configuration of arrays, subjects spend less time in finding the target compared to a novel configuration of arrays. In the first case, the distractor items configure a context that predicts the location of the target facilitating the visual search. Interestingly, subjects are not able to explicitly recognize memory traces for repeated configurations of arrays. Using this paradigm, Chun and Phelps (1999) observed that subjects with hippocampal damage did not show the contextual facilitation in search times compared to controls subjects. Similar results were reported by Park et al. (2004), who injected the GABA-A agonist midazolam, a benzodiazepine that induces anterograde amnesia, in healthy subjects. In addition, Greene et al. (2007), using functional magnetic resonance imaging (fMRI), found that faster search times for repeated contexts are related to greater activation in the hippocampus. Other studies, however, have suggested that hippocampal activity is correlated with explicit recognition of repeated contexts (Preston and Gabrieli 2008;Westerberg et al. 2011). Indeed, it has been suggested that patients with circumscribed hippocampal damage performed normally in the contextual cueing task (Manns and Squire 2001). The source of this discrepancy remains under debate (Greene et al. 2007;Preston and Gabrieli 2008;Westerberg et al. 2011;Geyer et al. 2012). Taken together, there is increasing evidence that the hippocampus is critically involved in implicit contextual memory-guided attention.

[3] 45w While several fMRI studies have identified associations between implicit contextual learning and hippocampal BOLD activity, there are no studies on structural brain correlates of implicit contextual learning. Here, we use structural MRI to test the hypothesis that hippocampal volume would positively predict contextual cueing performance.

[4] 138w Forty healthy subjects (28 males; mean age 27.23, SD 2.12, range 21-32 years) participated in the study. They were recruited in schools for ambulance rescue workers located in Southern Germany as part of a longitudinal study investigating predictors of post-traumatic stress disorder (Pohlack et al. 2011(Pohlack et al. , 2012(Pohlack et al. , 2014;;Cacciaglia et al. 2015;Winkelmann et al. 2017). None of the participants had to be excluded due to mental disorders such as major depressive disorder, current or chronic substance abuse, schizophrenia or borderline personality disorder, as assessed with the German version of the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders-IV (Wittchen et al. 1997). The Ethics Committee of the Medical Faculty Mannheim of the University of Heidelberg approved the study, and we obtained written informed consent from all persons before participation.

[5] 173w Each experimental session consisted of a 25-min test phase in which participants completed a version of the Contextual Cueing task (Chun and Jiang 1998) implemented and executed in Matlab (Natick, MA, USA). The participants were encouraged to search a target item embedded in a spatial array of distractor items. The target was a T stimulus rotated 90° to the right or to the left. The distractor stimulus was an L shape presented randomly in one of four orientations (0°, 90°, 180°, 270°) with a 10 pixel offset at the line junctions. Each display consisted of 12 items (a single target and 11 distractors) randomly positioned in an invisible 8 × 6 matrix (37.2° × 28.3°). For repeated display, the target appeared in the same location within an invariant configuration of distractors across blocks. For each new display, the distractor configurations were newly generated in each block. Visual stimuli were presented on a gray background in a 17-inch Viewsonic VG710b LCD monitor. The participants were seated 50 cm in front of the computer monitor.

[6] 112w Each trial consisted of the presentation of a 500 ms fixation cross, a display presentation (for a maximum of 6 s) and a variable duration inter-trial-interval (ITI; range 500-1000 ms). During the display presentation, the participants pressed one of the two buttons ("C" or "M") on a computer keyboard, corresponding to whether the bottom of the T was pointed to the right or to the left. An example of the trial sequence is shown in Fig. 1. Following one practice block, each subject performed 20 blocks of the search task. Each block contained 24 intermixed trials of 12 repeated and 12 new displays. There were rest periods of 10 s between bocks.

[7] 121w After block 20 of the contextual cueing task, the participants were asked if they noticed that certain configurations of the stimuli were repeated from block to block (Chun and Jiang 1998). Then, an explicit recognition test was carried out to assess awareness of the contextual displays. Importantly, the participants were not informed at the beginning of the experimental session about the recognition test. In this test, the 12 repeated displays used in the experimental session and 12 novel displays were presented in random order (Chun and Jiang 1998). The participants were informed that certain display configurations were repeated during the experiment and were instructed to indicate whether each display presented was "repeated" or "new" by responding with 1 of 2 keys.

[8] 79w Magnetic resonance imaging was performed at the Central Institute of Mental Health in Mannheim with a 3T MAG-NETOM Trio whole-body scanner (Siemens Medical Solutions, Erlangen, Germany) equipped with a standard 12-channel head coil. T1-weighted high-resolution images were acquired with a magnetization-prepared rapid gradient echo (MPRAGE) sequence [TR 2300 ms, TE 2.98 ms, field of view 240 × 256 mm 2 , 160 sagittal slices, voxel size 1.0 × 1.0 × 1.1 mm 3 , parallel imaging (GRAPPA) factor 2].

[9] 289w The segmentation of the hippocampus, amygdala, caudate, putamen, pallidum, cerebellar cortex, and cortical regions (Desikan et al. 2006) were performed using the Freesurfer 5.0 image analysis suite, which is documented and freely available for download online (http://surfe r.nmr.mgh.harva rd.edu/). Technical details of these procedures are described in prior publications (Dale et al. 1999;Fischl et al. 1999;Fischl and Dale 2000). The processing involves motion correction, removal of non-brain tissue using a hybrid watershed/surface deformation procedure, automated Talairach transformation, segmentation of the subcortical white and deep gray matter volumetric structures, intensity normalization, tessellation of the gray/white matter boundary, automated topology correction, and surface deformation following intensity gradients to optimally place the gray/ white matter and gray matter/cerebrospinal fluid borders at the location where the greatest shift in intensity defines the transition to the other tissue class. Processing was performed on an 8-core UNIX workstation with the Graphics-Processing Unit option enabled. Freesurfer morphometric procedures have been validated against manual segmentation and VBM8 (Grimm et al. 2015) and demonstrate good test-retest reliability across scanner manufacturers and across field strengths (Han et al. 2006). Finally, right and left hippocampal and amygdalar, caudate, putamen, pallidum, cerebellar cortex, cortical regions, and total gray matter volumes were calculated in cubic centimeters and then normalized to the Each block contained 24 randomly interleaved trials of 12 repeated and 12 new contexts. On each trial, subjects fixated a central cross for 500 ms. After that, a search array was presented for a maximum of 6 s or until a response was made (subjects indicated whether the bottom of the T was pointed to the right or to the left). A variable duration ITI (500-1000 ms) separated subsequent trials estimated total intracranial volume of each individual, also provided by FreeSurfer.

[10] 309w For analysis purposes, search reaction times (RTs) of the 20 blocks were grouped into sets of five yielding four epochs and analyzed using two-way repeated measures ANOVA, based on trials performed correctly within [0.5, 6] s. The independent variables were (1) Context (novel vs. repeated), and (2) Epoch (1 vs. 4). The contextual cueing effect was calculated as the difference in RTs between Old and New arrays collapsed across epochs 3-4 (Chun and Phelps 1999). Differences in the magnitude of contextual cueing effect were evaluated with Student's t test. To examine whether participants had explicit recognition of repeated contexts, a paired-samples t test on hits and false alarms for the recognition test was conducted (Manginelli et al. 2013). To further address the possibility that explicit recognition of repeated contexts could influence visual search facilitation in the contextual cueing task, the correlations between hit rate and the magnitude of the contextual cueing, corrected hit rate and the magnitude of the contextual cueing effect, and corrected hit rate and left or right hippocampal volumes were calculated (Preston and Gabrieli 2008;Manginelli et al. 2013). To explore relationships between volumetric and behavioral data, linear regression analyses were conducted on the entire sample (n = 40), separately for amygdalar, hippocampal, caudate, putamen, pallidum, cerebellar cortex, cortical regions, and total gray matter volumes. For each regression analysis, contextual cueing performance was entered as dependent variable. An additional linear regression analyses was carried out including only those participants who showed a positive (above-zero) contextual cueing effect (n = 34). To control for possible influences of demographic variables, partial correlations between volumetric and behavioral data were performed, including age, sex and years of education as confounds. For each of the analyses, the alpha level was set to 0.05. Data are reported as mean ± SD unless otherwise stated. All analyses and statistical tests were performed in Matlab.