Abstract

Renewal is defined as the recovery of an extinguished response if extinction and retrieval contexts differ. The context dependency of extinction, as demonstrated by renewal, has important implications for extinction-based therapies. Persons showing renewal (REN) exhibit higher hippocampal activation during extinction in associative learning than those without renewal (NOREN), demonstrating hippocampal context processing, and recruit ventromedial pFC in retrieval. Apart from these findings, brain processes generating renewal remain largely unknown. Conceivably, processing differences in task-relevant brain regions that ultimately lead to renewal may occur already in initial acquisition of associations. Therefore, in two fMRI studies, we investigated overall brain activation and hippocampal activation in REN and NOREN during acquisition of an associative learning task in response to presentation of a context alone or combined with a cue. Results of two studies demonstrated significant activation differences between the groups: In Study 1, a support vector machine classifier correctly assigned participants' brain activation patterns to REN and NOREN groups, respectively. In Study 2, REN and NOREN showed similar hippocampal involvement during context-only presentation, suggesting processing of novelty, whereas overall hippocampal activation to the context–cue compound, suggesting compound encoding, was higher in REN. Positive correlations between hippocampal activation and renewal level indicated more prominent hippocampal processing in REN. Results suggest that hippocampal processing of the context–cue compound rather than of context only during initial learning is related to a subsequent renewal effect. Presumably, REN participants use distinct encoding strategies during acquisition of context-related tasks, which reflect in their brain activation patterns and contribute to a renewal effect.

INTRODUCTION

The renewal effect in extinction describes the reoccurrence of an extinguished response if the context of extinction differs from the context of acquisition and recall (Bouton & Ricker, 1994), which was first demonstrated in an animal study (Bouton & Bolles, 1979). Renewal has since been shown in appetitive conditioning (Bouton & Peck, 1989) and taste aversion learning (Rosas & Bouton, 1997) in rats as well as in fear conditioning in both animals and humans (Vansteenwegen et al., 2005, 2006; Bouton & King, 1983). The most solid renewal effect is elicited by a design in which conditioning and test are performed in the same context, whereas extinction is performed in a different context, a so-called ABA design (Bouton, 2004). By coupling the retrieved association to the context in which it has been acquired, the effect impressively demonstrates the context dependency of extinction learning. Despite potential clinical implications of renewal, such as an unwanted coupling of therapy success to the context of the therapeutic setting (Mystkowski, Craske, Echiverri, & Labus, 2006; Mystkowski, Mineka, Vernon, & Zinbarg, 2003; Mineka, Mystkowski, Hladek, & Rodriguez, 1999), and the behavioral research carried out on the subject, the brain activation correlates of the renewal effect have only recently been investigated.

In a recent fMRI study, we demonstrated that, in healthy participants, the hippocampus has a central role in encoding context information during associative extinction learning and thus mediates the renewal effect, together with ventromedial pFC (vmPFC) during extinction recall (Lissek, Glaubitz, Uengoer, & Tegenthoff, 2013). Participants who showed renewal during extinction recall exhibited stronger hippocampal activation during extinction learning than those who failed to show renewal. Although these findings are in line with fear extinction studies in animals reporting that hippocampal inactivation impaired context encoding (Corcoran, Desmond, Frey, & Maren, 2005; Corcoran & Maren, 2001) and in humans that found hippocampus and vmPFC involved in contextual extinction learning (Milad et al., 2007; Kalisch et al., 2006), they additionally demonstrate that hippocampal processing of context occurs in extinction learning regardless of whether this learning is fear related. Lesion studies in humans also implicated hippocampus in reversal learning, proposing a role for hippocampus in configural learning. Patients with amnestic mild cognitive impairment and hippocampal lesions showed, after intact initial task acquisition, deficits in reversal learning of a configuration of stimuli (Shohamy, Myers, Hopkins, Sage, & Gluck, 2009) and in a reversal task, which required associating a new cue with a new outcome in a context formerly associated with a different cue and outcome (Levy-Gigi, Kelemen, Gluck, & Kéri, 2011).

An ongoing debate concerns the question whether, in context-related associative extinction learning, the context is encoded only during the extinction phase or whether it may be processed already during acquisition of a task. One position based primarily on behavioral animal research claims that extinction of a cue–outcome association is more context specific than initial acquisition (Rosas, Todd, & Bouton, 2013; Bouton & Ricker, 1994), because unlike extinction, acquisition of a cue–outcome relationship is not affected by a change of context and performance is easily transferred between contexts (Bouton, 1993). The surprising change in the contingency between cue and outcome in extinction is supposed to induce the organism to pay attention to the context, whereas during acquisition, attention to context is minimal because context is irrelevant for solving the task (Darby & Pearce, 1995). In contrast, based on behavioral evidence from renewal studies, it has been suggested that, in humans, not only fear extinction but also fear acquisition is controlled by its current context (Effting & Kindt, 2007). The authors observed a more prominent renewal effect in the context of acquisition (ABA) than in a completely new context (ABC), suggesting that the contextual stimuli were already processed during acquisition. These findings have been corroborated by animal research using instrumental conditioning, where ABA and ABC renewal strength was associated with the amount of acquisition training (Todd, Winterbauer, & Bouton, 2012). Transient hippocampal activation to the CS+ context in fear conditioning has also been observed (Baeuchl, Meyer, Hoppstädter, Diener, & Flor, 2015). In line with its role in context processing, hippocampus has been implicated in configural learning for the spatial domain and the nonspatial domain (Ryals, Wang, Polnaszek, & Voss, 2015; Kumaran et al., 2007), particularly in rapid binding between a context and an item during task acquisition (Howard, Kumaran, Ólafsdóttir, & Spiers, 2011). However, whether stimulus compounds are processed elementally or configurally appears to depend in part on past experience or pretraining (Mehta & Russell, 2009; Lober & Lachnit, 2002) and in part on stimulus properties, task demands, and instructions (Melcher, Shanks, & Lachnit, 2008). Because tasks used to research renewal allow participants to process stimuli configurally or nonconfigurally, their solving strategies may differ accordingly. These diverging findings on task processing open up the possibility that context processing during acquisition of an associative learning task may be initiated by strategies the participants apply, such as configural processing of context and cue—strategies that may operate independently of, or are not automatically triggered by, characteristics of the task design.

Along these lines, the phenomenon of renewal, which tends to occur in roughly half of participants in a given study of contextual extinction learning (Lissek et al., 2013), could result from a learning strategy that favors context processing and potentially configural processing throughout all phases of a learning task and that might be reflected in concurrent activation of context-processing regions.

Human imaging data on processes mediating context-related initial learning of an associative task that eventually generate a renewal effect are lacking. Therefore, in this study, we investigated potential processing differences during acquisition of an associative task between persons who later show or do not show renewal. Our first aim was to determine whether systematic differences in overall brain activation patterns exist already in task acquisition between persons who show or do not show renewal. In view of the crucial role of hippocampus for context processing in extinction learning, our second aim was to determine whether there are differences in context-related hippocampal activation already during acquisition between persons who show or do not show renewal, differences that relate particularly to the processing of a context–cue combination compared with processing of a context by itself.

We used a predictive learning task designed to evoke renewal in which participants had to acquire simple cue–outcome associations against the background of particular contexts. These cue–outcome associations were reversed in extinction learning, either in a novel context or in the acquisition context. In extinction recall, stimuli were again presented in the acquisition context. With a different group of participants, an additional phase presenting only the context was introduced before each trial, allowing comparison of the hippocampal responses to the context alone and to the combination of context and cue. Participants were assigned to renewal (REN) or no renewal (NOREN) groups depending on their performance in extinction recall.

In Study 1, we characterized single-participant activation patterns for brain regions processing context and extinction learning and performed a multivariate pattern analysis to determine whether the brain activation patterns of REN and NOREN participants allow reliable classification into distinct groups. In Study 2, we focused on potential differences in the level of hippocampal activation between REN and NOREN participants during the trial phases showing either context only or the context–cue compound to determine whether context-related hippocampal activation during acquisition has a predictive value for later renewal.

For Study 1, we hypothesized that overall brain activation during acquisition should yield distinguishable patterns in REN and NOREN participants.

On the basis of the behavioral findings for context processing during acquisition in context-related conditioning, as well as on our previous findings on prominent hippocampal activation in REN participants during extinction, we hypothesized for Study 2 that participants who show renewal, in contrast to those who do not, will be more likely to process contextual stimuli in combination with the cues already during acquisition of a task, an effort that will reflect in stronger hippocampal involvement.

METHODS

Participants

Twenty-eight healthy right-handed individuals (14 women, 14 men), with mean ± SD age of 25.32 ± 3.878 years (range = 20–30 years; Study 1), and 30 healthy right-handed individuals (15 women, 15 men), with mean ± SD age of 26.9 ± 3.88 years (range = 21–39 years; Study 2), without a history of neurological disorders, participated in this study after giving written informed consent. Before the experiments, participants received handouts informing them about the fMRI procedures. The protocol was approved by the ethics committee of the Ruhr-University Bochum. The study conforms to the Code of Ethics of the World Medical Association (Declaration of Helsinki). The participants received a monetary compensation for their participation (in the amount of €40). Participants were assigned to the REN and NOREN groups according to the procedure described in the Behavioral Data Analysis section.

Predictive Learning Task

The predictive learning task that we used in this study was adapted for use in an fMRI setting from a task originally devised by Üngör and Lachnit (2006), which constitutes an established paradigm to study the renewal effect. By means of the task design in which acquisition, extinction in various contexts, and test phase are run back-to-back without breaks in between, a renewal effect can be reliably evoked, as demonstrated in a number of behavioral (Lucke, Lachnit, Koenig, & Uengoer, 2013; Lachnit, Ko, Üngör, & Melchers, 2008; Nelson & Callejas-Aguilera, 2007; Rosas & Callejas-Aguilera, 2006; Üngör & Lachnit, 2006) as well as imaging (Lissek, Glaubitz, Güntürkün, & Tegenthoff, 2015; Lissek et al., 2013) studies. The task design allows for learning of associations between cues and outcomes with or without encoding of the context, because regarding or ignoring the context does not impact the ability to learn the task.

In the predictive learning task, participants were asked to put themselves in the position of a physician and predict whether various articles of food served in different restaurants would lead to the aversive consequence of a stomachache in their patient. The learning process consisted of the three successive phases of (a) acquisition of associations, (b) extinction, and (c) test phase (see Figure 1).

Figure 1. 

Predictive learning task. (A) Example of a trial of the task version used in Study 2, containing a phase of context-only presentation. (B) Example of a trial of the task version used in Study 1, where the context is always shown in combination with the cue. (C) Design of the predictive learning task. In Condition AAA, extinction is performed in the same context as acquisition. In Condition ABA, extinction is performed in a different context. In both conditions, the final test for renewal is performed in the context of acquisition. (D) Food images used as stimuli.

Figure 1. 

Predictive learning task. (A) Example of a trial of the task version used in Study 2, containing a phase of context-only presentation. (B) Example of a trial of the task version used in Study 1, where the context is always shown in combination with the cue. (C) Design of the predictive learning task. In Condition AAA, extinction is performed in the same context as acquisition. In Condition ABA, extinction is performed in a different context. In both conditions, the final test for renewal is performed in the context of acquisition. (D) Food images used as stimuli.

During the acquisition phase (80 trials), participants learned to associate an article of food with a specific consequence. In each trial, one of eight stimuli (vegetables or fruits) was presented to the participant in one of two different contexts (indicated by the restaurant names “Zum Krug” and “Altes Stiftshaus” and a frame in either red or blue color). The stimulus in its context was first presented alone for 3 sec, and then a question asking whether the patient will develop a stomachache was superimposed, with the response options “yes” or “no.” Response time was 4 sec; participants responded by pressing the respective button on an fMRI-ready keyboard (Lumitouch; Photon Control, Inc., BC, Canada). After the response, or in case of a missing response after expiration of the response time, a feedback with the correct answer was displayed for 2 sec, namely, “The patient has a stomachache” or “The patient does not have a stomachache.” The actual response of the participant was not commented on. In the second study, an additional phase containing only the context was displayed on the screen for 4 sec at the beginning of each trial. The food stimuli were presented in randomized order; each stimulus was presented 10 times. Four stimuli were presented per context. Stimuli were counterbalanced with regard to their causing the aversive consequence of a stomachache, with two stimuli per context causing stomachache during acquisition, whereas the other two did not.

During the extinction phase (80 trials), half of the stimuli were presented in the same context as during acquisition (Condition AAA: no context change, 40 trials); and the other half, in the other context (Condition ABA: context change, designed to induce a renewal effect, 40 trials), in randomized order. In addition, stimuli were subdivided into two types: For actual “extinction stimuli,” the consequence changed, and the new consequence had to be learned; and for “distractor stimuli,” which were introduced to make overall learning more difficult, the consequence during extinction remained unchanged (during acquisition, stimuli that later were subdivided into extinction and distractor stimuli were still indistinguishable). Per context, we used two extinction stimuli and two distractor stimuli. In all other respects, trials were identical to those during acquisition.

During the test phase (24 trials), all stimuli were presented once again in the context of acquisition (three presentations per stimulus). With the exception that, during the test phase, no feedback with the correct response was given, trials were identical to those during acquisition.

Imaging Data Acquisition

Functional and structural brain scans were acquired using a whole-body 3-T scanner (Philips Achieva 3.0 T X-Series; Philips, Amsterdam, The Netherlands) with a 32-channel SENSE head coil. BOLD contrast images were obtained with a dynamic T2*-weighted gradient echo EPI sequence using SENSE (repetition time = 3200 msec, echo time = 35 msec, flip angle = 90°, field of view = 224 mm, slice thickness = 3.0 mm, voxel size = 2.0 × 2.0 × 3.0 mm). We acquired 45 transaxial slices parallel to the AC–PC line, which covered the whole brain. High-resolution structural brain scans of each participant were acquired using an isotropic T1 Turbo Field Echo sequence (field of view = 240 mm, slice thickness = 1.0 mm, voxel size = 1 × 1 × 1 mm) with 220 transversally oriented slices covering the whole brain. The task was presented to the participants via fMRI-ready LCD goggles (Visuastim Digital; Resonance Technology Inc., Northridge, CA) connected to a laptop, which ran specific software programmed in MATLAB (The Mathworks, Natick, MA). Responses were given by means of an fMRI-ready keyboard (Lumitouch response pad; Photon Control Inc., BC, Canada).

Imaging Data Analysis

For preprocessing and statistical analysis of fMRI data, we used the software SPM, version 8 (Wellcome Department of Cognitive Neurology, London, UK), implemented in MATLAB R2008a. Three dummy scans, during which the BOLD signal reached steady state, preceded the actual data acquisition of each session; thus, preprocessing started with the first acquired volume. Preprocessing on single-participant level consisted of the following steps: slice timing correction to account for time differences because of multislice image acquisition; realignment of all volumes to the first volume for motion correction; spatial normalization into standard stereotactic coordinates with 2 × 2 × 2 mm3 using an EPI template of the Montreal Neurological Institute (MNI) provided by SPM; and smoothing with a 6-mm FWHM kernel, in accordance with the standard SPM procedure. The acceptable limit for head motion was 2 mm for translational movements and 0.5° for rotational movements.

In a first-level single-participant analysis, we calculated activation during acquisition of the stimulus–outcome associations, contrasted against baseline. To evaluate the brain activation patterns of the groups, for Study 1, we used an anatomically defined mask that was constructed using the software MARINA (BION Bender Institute of Neuroimaging, University of Giessen, Germany; Walter et al., 2003), which is based on the automated anatomical labeling atlas (Tzourio-Mazoyer et al., 2002). The mask contained, as a priori ROIs, pFC (superior, middle, and inferior frontal gyri), hippocampus, amygdala, and insula. All data contained in this combined mask were analyzed together in a single analysis. To evaluate differential hippocampal activation to the context-only and context–cue phases, for Study 2, we used an anatomical ROI of bilateral hippocampus defined by means of the MARINA software, based on the automated anatomical labeling atlas (Tzourio-Mazoyer et al., 2002). These different masks used for the two studies reflect our two foci of interest: In Study 1, we investigated whether REN and NOREN participants can be distinguished on the basis of their overall brain activation pattern during acquisition, and in Study 2, we investigated whether hippocampal processing in acquisition differs between REN and NOREN.

For both studies, we used an event-related design, modeling the events of each trial (onset of context-only, context–cue, questions, feedback) using distinct stick functions convolved with the canonical hemodynamic response function in SPM, with our analysis based on the onsets of context–cue phases of each trial in Study 1. For Study 2, we modeled two regressors for the context-only and context–cue phases of each trial, by convolving a series of stick functions (onsets of the respective trial phases) with the canonical hemodynamic response function in SPM. These two phases were presented with an SOA of 4 sec. In addition, we performed analyses to account for the potential problem of collinearity between these regressors, to ensure that brain activity to the context-only phase could be reliably distinguished from activity to the context–cue phase.

We examined the “design orthogonality” matrix in SPM for each participant regarding the value of the cosine of the angle between the regressors for the context-only condition and the context–cue condition. The group mean for the absolute value of angles between the regressors was cos (θ) = −0.004 (range = −0.07 to 0.07), indicating no substantial collinearity between regressors, that is, the regressors do not explain the same variance in our signal.

The contrast images from the single-participant analyses were entered into second-level random effects analyses to calculate in one-sample tests the activation patterns of the two subgroups (REN: renewal effect, NOREN: no renewal effect) for the acquisition phase. We used an FWE-corrected threshold of p < .05 voxel level for the complete sample analysis of Study 1. For second-level analyses of within-group activation during the context-only and context–cue phases, we applied an FWE-corrected threshold of p < .05 cluster level, unless otherwise specified.

For further detailed second-level analyses of hippocampal activation within and between the subgroups, in response to early and late context-only and context–cue phases in Study 2, as well as for the extraction of mean signal intensities (MSIs) for correlational analyses, we used a functional ROI consisting of bilateral hippocampal clusters (volume = 927 voxels) found activated in all participants of Study 1 during the acquisition phase, that is, during the combined context–cue phase, thresholded at p < .05 FWE-corrected voxel level. This functional ROI from Study 1 was chosen to secure independence of ROI selection and analysis in Study 2. Only activated clusters within this functional ROI were included in the results. To further evaluate smaller differences between the groups, we adapted the correction level to a small volume correction with a sphere of 10 mm around the peak voxel of the activated cluster within the ROI.

Correlational Analyses

For all participants of Study 2, we extracted MSIs of activation during the context-only and context–cue phases from the functional ROIs within hippocampus described above, using the Marsbar Toolbox (Brett et al., 2002). The extracted MSIs were entered into correlational analyses using the IBM SPSS Statistics for Windows software package, version 22.0 (IBM Corp., Armonk, NY), to analyze the relation between hippocampal activation during the context-only and context–cue phases of acquisition with the level of the renewal effect (i.e., the percentage of renewal responses shown during the ABA test phase, relative to all ABA test trials) as well as with learning performance (i.e., the error rate in percent during acquisition and extinction learning).

Multivariate Pattern Analysis

Furthermore, for Study 1, we performed a multivariate analysis to determine whether, based on brain activation patterns observed in REN and NOREN participants during acquisition of the predictive learning task, we could train a classifier, a linear support vector machine (SVM), to differentiate between the two groups and to identify patterns that were specific to one group or the other. For the multivariate analysis, we used the “Pattern Recognition for Neuroimaging Toolbox” (PRoNTo) software implemented in MATLAB and SPM8 (Schrouff et al., 2013) with a gray matter mask to train a binary SVM on the data set of Study 1 using a leave-one-subject-out cross-validation approach. The β-images of the general linear model analysis of the original data acquired during acquisition of the predictive learning task, which represent the partial correlations between predictor and the observed data (i.e., the unique contribution of each predictor to explain the observed data), were entered into the multivariate analysis. Classification was performed by training a linear SVM (f(x) = w0 + wTx, where the weights w are the model parameters learned in the training phase and represent the relative contribution of each feature to the predictive task (Schrouff et al., 2013). A default regularization parameter C = 1 was applied. Results were analyzed with regard to overall accuracy, class accuracy, and balanced accuracy. To determine if the balanced accuracy of classification significantly exceeded chance levels (50%), we used a permutation test with 1000 runs. Permutation testing in this context is preferable to parametric testing, as it does not require assumptions regarding the distribution. In each run of the permutation test, group labels were randomly assigned to each image, and the entire cross-validation procedure was repeated. The number of runs in which the permuted test accuracy exceeded the one obtained for the true groups was divided by the total number of tests to yield a p value for the classification. In a second step, we projected the average weight maps onto a brain template to determine brain regions that contributed to the classification as an REN or NOREN participant.

Behavioral Data Analysis

For all three learning phases, log files were written that contained information on response latency, response type, and correctness of response, from which we calculated error rates. For calculation of the renewal effect, only responses to stimuli with consequence change (extinction stimuli) during the test phase were analyzed. The behavioral renewal effect in the predictive learning task is supposed to occur only in the Condition ABA because of the context change during extinction learning. In case of renewal, associations learned during acquisition in Context A will reappear in the test phase, which is again performed in Context A, whereas extinction was performed in Context B.

In contrast, the AAA condition constitutes a control condition for extinction learning, because here all learning phases are performed in an identical context. If extinction learning is successful, responses during the test phase will reflect the associations learned during extinction. Only if extinction learning is impaired, responses in the AAA test phase will reflect associations learned during acquisition.

Errors in acquisition and extinction learning were defined as responses stating the incorrect association between the context–cue compound and the consequence. During the test phase, a response that referred to the association that was correct during acquisition constituted an error in the AAA condition and a renewal response in the ABA condition.

Studies that previously used the predictive learning task outside an fMRI setting have found that a considerable portion of their participants (on average, about 40%) did not exhibit the renewal effect (see also Lissek et al., 2013). After evaluation of their behavioral data, each of our participants therefore was assigned to one of the two groups, REN or NOREN. Participants were assigned to the REN and NOREN groups, respectively, based on their results in the ABA trials with consequence change during the test phase, that is, those trials that evoked the renewal effect. All participants who never showed the renewal effect (i.e., 0% renewal responses) were assigned to the NOREN group, and all participants who showed the renewal effect (minimum of 30% renewal responses) were assigned to the REN group. In two-tailed independent t tests, we compared the learning performance of REN and NOREN participants in acquisition and extinction. Statistical analyses were performed using the IBM SPSS Statistics for Windows software package, Version 20.0.

RESULTS

Assignment to Groups Based on Performance in the Recall Phase

Study 1

Seventeen participants showed a renewal effect and were assigned to the REN group. Eleven participants did not show a renewal effect and were assigned to the NOREN group. (One participant each from the REN and NOREN group had to be excluded from further analysis because of corrupt imaging data.) The REN group showed a renewal score of 69.66% in ABA trials (±3.74 SEM, range = 33.33–100%).

Study 2

Twenty participants showed a renewal effect and were assigned to the REN group. Ten participants did not show a renewal effect and were assigned to the NOREN group. The REN group showed a renewal score of 80.5% in ABA trials (±4.67 SEM, range = 30–100%). The NOREN group by definition did not show renewal (0.00% in ABA trials).

Ratio of ABA Renewal Responses to AAA Errors

To evaluate whether the observed renewal effect did actually indicate renewal, and not weak memory for the contingencies learned during the extinction phase, we calculated the ratio between those ABA and AAA responses given during the test phase that reflected the original contingency of the acquisition procedure (ABA − AAA/ABA + AAA). Here, a ratio of 1 indicates that responses reflected the acquisition contingency only in ABA trials (i.e., they constituted renewal responses), whereas a ratio of −1 indicates that this was the case only in AAA trials (i.e., errors). If the responses reflecting the acquisition contingency occur only or predominantly during ABA trials with a ratio close to 1, this indicates that, with a high probability, the behavior does not occur because of weak memory for the extinction contingencies but reflects a differentiation between the ABA and AAA conditions. In Study 1, 11 of 16 REN participants showed a ratio of 1; the remaining four showed ratios of 0.2, 0.3, 0.5, and 0.6. In Study 2, 15 of 20 REN participants showed a ratio of 1, three showed ratios of 0.8, and the remaining two showed ratios of 0.27 and 0.53. These results indicate that the renewal effect on which group assignment was based was genuine.

Behavioral Results in Acquisition and Extinction

Study 1

The REN and NOREN groups did not differ with regard to their learning performance in the predictive learning task. We observed no significant differences in acquisition of associations between the REN and NOREN groups (t(24) = 0.542, p = .593 two-tailed; percentage of errors: REN = 14.06% ± 1.47 SEM, NOREN = 12.62% ± 2.39 SEM). Also in extinction learning, groups did not differ with regard to percentage of errors (t(24) = 1.328, p = .197 two-tailed; percentage of errors: REN = 13.75% ± 1.51 SEM, NOREN = 10.75% ± 1.50 SEM). Groups did not differ with regard to response times in acquisition (t(47) = 0.086, p = .932 two-tailed; REN = 0.9356 sec, ±0.0704 SEM; NOREN = 0.9252 sec, ±0.1021 SEM).

Study 2

The REN and NOREN groups differed with regard to their learning performance during acquisition (t(28) = 2.678, p = .012; percentage of errors: REN = 13.06% ± 1.81 SEM [similar to Study 1] and NOREN = 6.13% ± 0.35 SEM [lower than in Study 1]). Again, during extinction, error rates did not differ significantly between groups (t(28) = 1.488, p = .148 two-tailed; percentage of errors: REN = 13.94% ± 1.59 SEM, NOREN = 10.25% ± 1.39 SEM).

Imaging

Activation within the REN and NOREN Groups

The REN and NOREN groups of Study 1 showed differential patterns of activation during acquisition of the predictive learning task. Both groups showed activation in dorsolateral pFC (dlPFC), OFC, and insula (right-hemispheric in REN; dlPFC: BA 9, OFC: BA 47), bilateral and more extended in NOREN (dlPFC: BAs 9, 44, 45, 46; OFC: BAs 47 and 10). In addition, the REN group exhibited activation in bilateral hippocampus and lingual gyrus, which was absent in the NOREN group. The NOREN group, in contrast, showed activation in right amygdala and parahippocampal gyrus (BA 35) and left fusiform gyrus (BA 37), which was absent in the REN group (see Figure 2 and Table 1).

Figure 2. 

Study 1: Activation patterns during acquisition of the predictive learning task in participants who later show renewal (REN) depicted in red and those that do not (NOREN) in blue (p < .05 FWE-corrected on cluster level, minimum cluster size = 20 voxels). Whereas REN participants predominantly activate regions in bilateral hippocampus, cingulate, insula, and a small area in dlPFC, NOREN participants predominantly activate dlPFC, OFC, and insula.

Figure 2. 

Study 1: Activation patterns during acquisition of the predictive learning task in participants who later show renewal (REN) depicted in red and those that do not (NOREN) in blue (p < .05 FWE-corrected on cluster level, minimum cluster size = 20 voxels). Whereas REN participants predominantly activate regions in bilateral hippocampus, cingulate, insula, and a small area in dlPFC, NOREN participants predominantly activate dlPFC, OFC, and insula.

Table 1. 

Activation in REN and NOREN Groups (Study 1) during Acquisition of the Predictive Learning Task (FWE-corrected p < .05 Cluster Level, Minimum of 20 Contiguous Voxels)

RegionBAHemRENNOREN
MNI CoordtVoxelMNI CoordtVoxel
dlPFC 54, 14, 36 6.53 58 54, 18, 34 10.45 53 
    −34, 6, 26 11.22 47 
45    58, 20, 14 14.77 80 
46    52, 40, 14 11.19 57 
    −44, 38, 18 8.72 37 
44    −48, 12, 22 10.46 23 
OFC 47 46, 20, 4 6.81 74 52, 22, −2 9.12 22 
    −36, 24, −22 11.57 81 
     −24, 18, −16 13.09 76 
     −52, 22, −10 8.21 93 
10    −28, 52, 24 11.91 37 
Cingulate gyrus (anterior/mid) 32 4, 14, 40 5.89 111    
  4, 24, 26 5.57     
Insula  38, 26, 0 7.29 197 38, 26, −6 8.25 21 
    −24, 20, −16 8.94 37 
     −38, −18, −2 9.43 20 
     −44, 4, 4 13.46 26 
Hippocampus  22, −24, −12 6.63 101    
 −24, −24, −8 9.79 118    
Lingual gyrus  14, −32, −4 14.98 117    
 −14, −48, −4 6.01 81    
Parahippocampal gyrus 35    22, −34, −8 8.36 20 
Fusiform gyrus 37    −28, −44, −18 7.77 20 
Amygdala     26, 2, −18 9.71 30 
Thalamus/pulvinar  22, −30, 4 10.73 105    
 −22, −32, 8 8.99 141 −22, −30, 8 8.51 62 
RegionBAHemRENNOREN
MNI CoordtVoxelMNI CoordtVoxel
dlPFC 54, 14, 36 6.53 58 54, 18, 34 10.45 53 
    −34, 6, 26 11.22 47 
45    58, 20, 14 14.77 80 
46    52, 40, 14 11.19 57 
    −44, 38, 18 8.72 37 
44    −48, 12, 22 10.46 23 
OFC 47 46, 20, 4 6.81 74 52, 22, −2 9.12 22 
    −36, 24, −22 11.57 81 
     −24, 18, −16 13.09 76 
     −52, 22, −10 8.21 93 
10    −28, 52, 24 11.91 37 
Cingulate gyrus (anterior/mid) 32 4, 14, 40 5.89 111    
  4, 24, 26 5.57     
Insula  38, 26, 0 7.29 197 38, 26, −6 8.25 21 
    −24, 20, −16 8.94 37 
     −38, −18, −2 9.43 20 
     −44, 4, 4 13.46 26 
Hippocampus  22, −24, −12 6.63 101    
 −24, −24, −8 9.79 118    
Lingual gyrus  14, −32, −4 14.98 117    
 −14, −48, −4 6.01 81    
Parahippocampal gyrus 35    22, −34, −8 8.36 20 
Fusiform gyrus 37    −28, −44, −18 7.77 20 
Amygdala     26, 2, −18 9.71 30 
Thalamus/pulvinar  22, −30, 4 10.73 105    
 −22, −32, 8 8.99 141 −22, −30, 8 8.51 62 

Multivariate Pattern Analysis

The results of the multivariate pattern analysis showed that REN and NOREN participants could be distinguished on the basis of their brain activation pattern during acquisition of the predictive learning task: The overall accuracy of the SVM classifier was 80.8% (i.e., 21 of 26 participants of Study 1 were classified correctly with regard to their group membership). Within groups, class accuracy for the REN group was 87.5% (14 of 16 participants classified correctly), whereas class accuracy for the NOREN group was 70% (7 of 10 participants classified correctly). Balanced accuracy was 78.8%, significant at a p value of <.001 according to the permutation test.

Visual inspection of the weight vector w, used by the SVM to distinguish between groups, indicated that the multivariate analysis relied on regions similar to the ones found by univariate testing. In particular, activation in bilateral hippocampus, right-hemispheric insula, right inferior frontal gyrus, dlPFC (BA 9), and right lingual gyrus as well as right SMA contributed to classification as an REN participant, whereas activation in left-hemispheric insula, left dlPFC (BA 9/46) and OFC (BA 47), right parahippocampal gyrus, and thalamus/pulvinar contributed to classification as an NOREN participant.

Hippocampal Activation in Context and Context + Cue Phases

In a first analysis, we evaluated hippocampal BOLD responses to the context-only and context–cue compound phases in Study 2. During presentation of the context alone, we observed bilateral middle and posterior hippocampal activation in the REN group as well as posterior hippocampus activation in the NOREN group. During the presentation of the context–cue compound, we observed bilateral posterior hippocampal activation in the REN group. However, there was no significant hippocampal activation in response to the context–cue compound in the NOREN group alone (see Figure 3A and Table 2).

Figure 3. 

(A) Study 2: Posterior hippocampal activation in REN (red) and NOREN (green) participants during context-only presentation (left) and presentation of the context–cue compound (right; p < .05 FWE-corrected on cluster level, minimum cluster size = 10 voxels). During context presentation, both REN and NOREN groups show a hippocampal response, suggesting a novelty response. During the subsequent presentation of the context–cue compound, only the REN group maintains hippocampal activation, suggesting enhanced processing of the combination of stimulus and context. (B) Higher hippocampal activation in the REN group compared with the NOREN group during presentation of the context–cue compound (two-sample t test, SVC p < .05 FWE-corrected on voxel level).

Figure 3. 

(A) Study 2: Posterior hippocampal activation in REN (red) and NOREN (green) participants during context-only presentation (left) and presentation of the context–cue compound (right; p < .05 FWE-corrected on cluster level, minimum cluster size = 10 voxels). During context presentation, both REN and NOREN groups show a hippocampal response, suggesting a novelty response. During the subsequent presentation of the context–cue compound, only the REN group maintains hippocampal activation, suggesting enhanced processing of the combination of stimulus and context. (B) Higher hippocampal activation in the REN group compared with the NOREN group during presentation of the context–cue compound (two-sample t test, SVC p < .05 FWE-corrected on voxel level).

Table 2. 

Hippocampal Activation (Study 2) during the Context-only and Context–Cue Phases (One-sample t Tests, p < .05 FWE-corrected Cluster Level, Minimum of 10 Contiguous Voxels)

GroupContext-only PhaseContext–Cue Phase
MNI CoordVoxeltMNI CoordVoxelt
REN 20, −32, −4 66 5.97 14, −38, 10 14 5.64* 
32, −20, −10 63 4.55    
−18, −28, −10 72 6.41 −18, −40, 8 45 8.12 
NOREN 26, −32, −4 23 7.66*    
−16, −30, −6 34 8.46    
GroupContext-only PhaseContext–Cue Phase
MNI CoordVoxeltMNI CoordVoxelt
REN 20, −32, −4 66 5.97 14, −38, 10 14 5.64* 
32, −20, −10 63 4.55    
−18, −28, −10 72 6.41 −18, −40, 8 45 8.12 
NOREN 26, −32, −4 23 7.66*    
−16, −30, −6 34 8.46    
*

p < .05 FDR-corrected cluster level.

Two-sample comparisons of REN and NOREN participants showed higher BOLD activation in bilateral hippocampus during the context–cue compound phase (peak MNI coordinates: 22, −24, −12; −20, −40, 6). There was no region in hippocampus in which NOREN showed higher activation than REN (see Figure 3B and Table 3).

Table 3. 

Comparison of Hippocampal Activation (Study 2) during Context-only and Context–Cue Phases (Two-sample t Tests, SVC p < .05 FWE-corrected Voxel Level, Minimum of 10 Contiguous Voxels)

ContrastContext-only PhaseContext–Cue Phase
MNI CoordVoxeltMNI CoordVoxelt
REN > NOREN    22, −24, −12 20 3.89 
   −20, −40, 6 52 3.99 
NOREN > REN       
ContrastContext-only PhaseContext–Cue Phase
MNI CoordVoxeltMNI CoordVoxelt
REN > NOREN    22, −24, −12 20 3.89 
   −20, −40, 6 52 3.99 
NOREN > REN       

Hippocampal Activation in the First and Last 20 Trials of the Context-only and Context–Cue Phases: Comparison of REN and NOREN Groups

To further assess the time course of hippocampal activation during the context-only phase, we separately analyzed the activation data for the first and last 20 trials of the acquisition session in the REN and NOREN groups. One-sample activation data for the REN and NOREN groups are listed in Table 4.

Table 4. 

Hippocampal Activation (Study 2) during the Context-only and Context–Cue Phases, Blocks of First and Last 20 Trials of Acquisition (One-sample t Tests, SVC p < .05 FWE-corrected Voxel Level, Minimum of 10 Contiguous Voxels)

GroupContext-only PhaseContext–Cue Phase
First 20 TrialsLast 20 TrialsFirst 20 TrialsLast 20 Trials
MNI CoordVoxeltMNI CoordVoxeltMNI CoordVoxeltMNI CoordVoxelt
REN 20, −30, −4 95 6.61 26, −30, −4 54 4.77 40, −32, −12 25 5.08    
−26, −32, −2 131 5.42 −18, −32, −4 54 5.14 −32, −40, −2 121 5.89    
NOREN 34, −18, −12 23 9.03          
−18, −24, −10 125 7.21 −22, −28, −6 53 7.75       
GroupContext-only PhaseContext–Cue Phase
First 20 TrialsLast 20 TrialsFirst 20 TrialsLast 20 Trials
MNI CoordVoxeltMNI CoordVoxeltMNI CoordVoxeltMNI CoordVoxelt
REN 20, −30, −4 95 6.61 26, −30, −4 54 4.77 40, −32, −12 25 5.08    
−26, −32, −2 131 5.42 −18, −32, −4 54 5.14 −32, −40, −2 121 5.89    
NOREN 34, −18, −12 23 9.03          
−18, −24, −10 125 7.21 −22, −28, −6 53 7.75       

Two-sample t tests revealed significant differences in hippocampal recruitment between the REN and NOREN groups. During the first presentations of the context–cue compound, REN participants showed higher activation in left-hemispheric hippocampus, a difference that was based on a strong deactivation in NOREN participants compared with a slighter deactivation in REN (MSI: −3.05 in NOREN and −0.33 in REN). In contrast, during the last trials, NOREN participants showed higher left-hemispheric hippocampal responses, because of a deactivation in this region in REN participants (MSI: 0.86 in NOREN and −1.63 in REN). This time course in the REN group suggests a hippocampal process of coupling context and cue during the early trials in which learning still occurs, whereas in the final trials, when associations have been established and no more errors occur, hippocampus is no longer active.

With small volume correction with a sphere of 10 mm around the peak voxel of the activated cluster, we found additional differential activation during the context-only phase between the groups: REN showed higher activation in right-hemispheric middle hippocampus during both first and last 20 trials, whereas NOREN showed higher activation in right anterior hippocampus, only during the first 20 trials. These results are consistent with the notion of a novelty response to the context.

Moreover, additional regions in left-hemispheric mid and posterior hippocampus were recruited in REN participants during the last context-only presentations and the first context–cue presentations.

In summary, REN and NOREN participants exhibit slight differences in hippocampal activation to the presentation of context-only and context–cue compounds in early and late learning, which may be related to their differences in processing an association between context and cue (see Table 5).

Table 5. 

Comparison of Hippocampal Activation (Study 2) during Context-only and Context–Cue Phases (Two-sample t Tests, SVC p < .05 FWE-corrected Voxel Level, Minimum of 10 Contiguous Voxels)

ContrastContext-only PhaseContext–Cue Phase
First 20 TrialsLast 20 TrialsFirst 20 TrialsLast 20 Trials
MNI CoordVoxeltMNI CoordVoxeltMNI CoordVoxeltMNI CoordVoxelt
REN > NOREN 36, −28, −6 11 3.63a 24, −18, −12 20 3.46a       
   −30, −14, −12 10 3.24a −16, −30, −8 23 3.90    
      −32, −40, −2 13 3.80a    
NOREN > REN 34, −4, −24 19 3.73a          
         −28, −32, −8 66 4.62 
ContrastContext-only PhaseContext–Cue Phase
First 20 TrialsLast 20 TrialsFirst 20 TrialsLast 20 Trials
MNI CoordVoxeltMNI CoordVoxeltMNI CoordVoxeltMNI CoordVoxelt
REN > NOREN 36, −28, −6 11 3.63a 24, −18, −12 20 3.46a       
   −30, −14, −12 10 3.24a −16, −30, −8 23 3.90    
      −32, −40, −2 13 3.80a    
NOREN > REN 34, −4, −24 19 3.73a          
         −28, −32, −8 66 4.62 
a

Thresholded SVC sphere of 10-mm diameter around peak voxel of activated cluster, volume ca. = 40 voxels.

MSI in Bilateral Hippocampus

An ANOVA with repeated measures showed a significant main effect of group (F(3) = 7.637, p = .011) with regard to hippocampal MSI, without significant effects of the repeated factor measuring site/time (F(3) = 0.871, p = .460) and of the interaction (F(3) = 0.155, p = .926), suggesting an overall higher level of hippocampal activation in the REN group.

Correlations of Hippocampal Activation and Renewal Level

According to our hypotheses, hippocampal activation in terms of BOLD MSI showed significant correlations with the renewal level across participants not only during the context–cue phase but also during the context-only phase: Activation in hippocampus correlated positively with renewal level during the context-only phase (Pearson's r = .436, p = .013, left; r = .415, p = .0175, right) as well as during the context–cue compound phase (Pearson's r = .429, p = .0145, left; r = .389, p = .025, right), indicating that a higher activation in hippocampus was associated with a higher renewal rate during the retrieval phase.

Correlations of Hippocampal Activation and Learning Performance

MSI in left hippocampus during both context-only and context–cue phases also was found positively correlated to acquisition error rate (Pearson's r = .419, p = .033, context-only; Pearson's r = .417, p = .034, context + cue); furthermore, there was a trend toward a significant correlation with the extinction error rate (Pearson's r = .354, p = .076, context-only; Pearson's r = .340, p = .089, context + cue).

DISCUSSION

REN and NOREN Show Distinct Brain Activation Patterns in Acquisition of an Associative Learning Task

The patterns of activation observed in the REN and NOREN groups, respectively, show a characteristic focus on hippocampal and cingulate regions in REN participants and on prefrontal regions in NOREN participants. Although both groups activated dlPFC and OFC, the NOREN's group activation in these regions was more pronounced and widespread, suggesting increased prefrontal involvement in stimulus processing. Insula activation was also present in both groups. In addition to these commonly activated regions, REN participants showed bilateral hippocampal and right-hemispheric cingulate activation that was completely absent in the NOREN group. Regions of activation exclusive to the NOREN group were right-hemispheric amygdala and bilateral parahippocampal/fusiform areas. The pattern of results partially corresponds to that found in extinction learning and recall phase in REN and NOREN participants (Lissek et al., 2013). Most importantly, results of a multivariate analysis using an SVM classifier yielded a significantly high accuracy rate distinguishing between the REN and NOREN groups based on their brain activation during acquisition. This finding demonstrates that activation patterns differ significantly between the groups, suggesting that, already during initial exposition to the stimuli, participants who later will or will not show renewal use distinct approaches to processing the demands of the predictive learning task.

Stronger Hippocampal Activation in REN Suggests Superior Encoding of Context Information Already during Acquisition

On the basis of behavioral and imaging studies, it has been suggested that, in contextual extinction learning, context information will only be encoded in response to the unexpected change in cue–outcome contingencies during extinction (Darby & Pearce, 1995), whereas initial acquisition of the cue–outcome contingencies is assumed to proceed without attention to the context (Rosas et al., 2013; Bouton & Ricker, 1994). However, diverging findings suggested that context may already be processed during acquisition of a task (Todd et al., 2012; Effting & Kindt, 2007).

Human hippocampus has previously been demonstrated to be involved in processing context information during extinction learning with (Milad et al., 2007; Kalisch et al., 2006) and without (Lissek et al., 2013) a fear component and also in other types of learning (Smith & Mizumori, 2006; Good, de Hoz, & Morris, 1998). Findings from a previous study (Lissek et al., 2013) showed that REN and NOREN participants differed significantly with regard to hippocampal activation in extinction learning, which was substantially higher in REN participants, in particular when the change in the contingency between cue and outcome had to be associated with a novel context. The comprehensive functions of the hippocampus in novelty processing (Yamaguchi, Hale, D'Esposito, & Knight, 2004; Knight, 1996) on the one hand and in sustaining neural representations that code for associative information such as object–object or object–place information (Kumaran & Maguire, 2009; Eichenbaum, 2004) on the other are presumably crucial for integrating relevant context information into a representation of an association between a stimulus and outcome.

Our findings demonstrate that hippocampal activation in response to a context–cue compound is not restricted to extinction learning but may occur already during acquisition of an associative learning task, suggesting early processing of context information—a result consistent with the notion of attention to context already during conditioning, which has been demonstrated in a fear extinction study (Effting & Kindt, 2007). In our experiment, this hippocampal activation distinguishes REN participants from NOREN participants already in a task phase during which the context has no evident relevance, suggesting that REN participants use an encoding strategy that favors a representation of the complete context–cue compound. Conceivably, such a representation will promote a decision strategy in the test phase that takes context into consideration and thus yields a renewal effect.

Hippocampal Processing over the Course of Acquisition Learning Differs between REN and NOREN

The assumption of a decision strategy in REN that considers context is supported by the findings of the second study. Here, we investigated hippocampal processing in response to presentation of the context by itself or as part of a context–cue compound.

Results show comparable bilateral hippocampal activation of REN and NOREN to presentation of the context-only phase over the complete learning session and, in particular, during the early learning phase (i.e., the first 20 trials). Presumably, this activation is a response pertaining to unexpected stimuli and thus relates to the hippocampal function of processing novelty, which has been reported in a number of studies (Kumaran & Maguire, 2009; Köhler, Danckert, Gati, & Menon, 2005; Yamaguchi et al., 2004; Knight, 1996), and that REN and NOREN participants have in common.

In contrast, the continued bilateral hippocampal activation in the REN group, as well as the left-hemispheric activation in NOREN participants, to the context-only phase during the last 20 trials suggests further processing of the context beyond a mere novelty response.

During the context–cue compound phase of each trial, overall bilateral hippocampal activation over the complete learning session is sustained in REN participants but lacking in NOREN participants.

Here, as in the first study, it can be assumed that the hippocampal activation in REN is associated with early encoding of the context–cue compound. In particular, the REN hippocampal activation appears to reflect the learning process, as it is prominent during the phase of high error rates in the first 20 trials when the new associations are being established but absent during the final 20 trials when associations have already been established so that errors occur only very rarely. Thus, REN hippocampal activation during the context + cue phase of acquisition presumably represents a process of early association of context and stimulus along the lines of the “binding of item and context” model (Diana, Yonelinas, & Ranganath, 2007; Eichenbaum, Yonelinas, & Ranganath, 2007), which in a broader sense ascribes to hippocampus a role in supporting recollection by associating learned items (words, pictures) with the overall context of learning in a framework of episodic memory.

In contrast, only in the final 20 trials, we observed a cluster of unilateral hippocampal activation in the NOREN group in a direct comparison with REN, which was not in evidence in any of the one-sample t tests nor the two-sample comparison of activation to the context–cue compound over the complete session and which might reflect associating of context and cue in late learning. This activation, however, had no bearing on later renewal.

Hippocampal Activation during Processing of Context-only and of the Context–Cue Compound Is Correlated with Later Renewal Behavior

Additional findings from Study 2 further demonstrate a relation between activation in hippocampus and the renewal effect. For the phases of both context-only and context–cue presentation, higher bilateral hippocampal activation is correlated with a higher renewal level across groups, corroborating the notion that hippocampal processing of context information already in the initial stage of learning is particularly prominent in participants who later consider the context in their responses.

An additional relation was found with regard to learning performance—higher error rates are associated with higher activity in left-hemispheric hippocampus during acquisition trials. It can be speculated that more errors may trigger a more thorough examination of the stimulus material, which is again reflected in hippocampal activity and results in more intense processing of context information. Such a view would be consistent with the explanation proposed by Darby and Pearce (1995) for extinction learning, who suggest that errors contribute to directing attention to context. In this framework, however, better encoding of context or the context–cue compound does not necessitate better learning in terms of fewer errors resulting from faster associations between (context–)cue and outcome.

REN Participants May Use a Different Strategy than NORENs in Processing Task Demands

In view of the findings reported above, we can assume that REN participants use a different encoding strategy for this associative learning task already from the outset. Such a strategy may be related to differential allocation of attention by way of executive attention that is used when there is a conflict between several attention cues (Petersen & Posner, 2012) or to a more holistic or configural processing of the visual information presented (Pearce, 1994). Executive attentional processing has been associated with anterior cingulate activation (Petersen & Posner, 2012; Fan, McCandliss, Fossella, Flombaum, & Posner, 2005), which is in line with our findings of cingulate activation in REN participants during task acquisition. Similar higher anterior cingulate activation was also found with stimulation of the NA system by atomoxetine compared with placebo during extinction of the predictive learning task, a manipulation that led to faster extinction (Lissek, Glaubitz, Güntürkün, et al., 2015).

In imaging studies, holistic processing of visual information has been investigated predominantly in the realm of face processing, assigning a crucial role to fusiform cortex, in particular, the fusiform face area (Arcurio, Gold, & James, 2012; Gold, Mundy, & Tjan, 2012; Zhang, Li, Song, & Liu, 2012). Moreover, the fusiform face area may also have a role in processing other configurations of multipart stimuli (Bilalić, Langner, Ulrich, & Grodd, 2011). However, in our study, fusiform cortex activation during display of the context–cue compound was rather found in NOREN participants.

The role of hippocampus in configural processing in general has been demonstrated in a number of studies. Several studies relate hippocampal activity to configural processing (Ryals et al., 2015; Davachi & Wagner, 2002; O'Reilly & Rudy, 2001) or describe deficits in patients with hippocampal lesions in both spatial and nonspatial configural learning (Kumaran et al., 2007); however, one fMRI study implicated regions in anterior medial temporal lobe, but not hippocampus, in configural learning (Preston & Gabrieli, 2008). Findings from our study suggest that hippocampal activation during presentation of the context–cue compound is related to configural processing in terms of rapid item–context binding (Howard et al., 2011). However, we did not use additional measures to investigate potential processing strategies in our subjects. Future studies on extinction and renewal in predictive learning should therefore employ additional tests suitable to explore holistic/configural processing in REN and NOREN participants.

Behavioral Findings

In Study 1, as expected, learning performance did not differ between REN and NOREN participants, but in Study 2, a difference in error rates was observed: Whereas the REN participants' error rate was similar to those found in Study 1 as well as in our previous studies (Lissek, Glaubitz, Güntürkün, et al., 2015; Lissek, Glaubitz, Wolf, & Tegenthoff, 2015), the NOREN group made significantly fewer errors. Thus, the introduction of an additional context-only phase did not affect REN participants' learning progress but facilitated learning for the NOREN group. It is conceivable that presenting the context on its own without any consequence made it appear even more irrelevant to participants less inclined to configural processing. A recent study showed that irrelevant contexts receive less attention and favor context-independent processing (Lucke, Lachnit, Stüttgen, & Uengoer, 2014). Yet, how such perceived irrelevance might accelerate learning remains unclear at present. Further studies are needed to identify task-related factors that might differentially affect participants who differ in their propensity for configural processing.

Transferability of Findings

As an example of associative learning, the predictive learning task constitutes a simple association task in which associations between a cue and an outcome are learned against the background of a particular context. Although predictive learning tasks differ in various procedural aspects from classical conditioning tasks often used in (fear) extinction research, for example, in the biological significance of the stimuli, there are fundamental similarities between these learning scenarios. In either case, participants predict the occurrence of an event based on the presence or absence of specific stimuli. Moreover, several phenomena observed in classical conditioning can be found in predictive learning too; for example, factors known to influence the rate of conditioning, such as contingency and cue competition also affect predictive learning (Shanks, Holyoak, & Medin, 1996). Because of these parallels between predictive learning and classical conditioning, many researchers assume that both types of learning are governed by similar mechanisms (Lachnit, Schultheis, Konig, Üngör, & Melchers, 2008; Miller & Matute, 1996; Allan, 1993; Gluck & Bower, 1988; Alloy & Tabachnik, 1984). Attesting to this assumption, the use of the predictive learning task in a previous study demonstrated the crucial role of hippocampus for context encoding in extinction learning and of vmPFC during contextual retrieval (Lissek et al., 2013), thus delivering results that correspond to findings of hippocampal and vmPFC involvement in human contextual fear extinction learning (Milad et al., 2007; Kalisch et al., 2006). As a consequence, we can assume that our present results on differential processing of context information during acquisition will also be valid for other forms of associative learning that are susceptible to renewal, such as contextual fear extinction learning.

Conclusions

In our study, we found hippocampal activity in processing context-related information during acquisition of an associative learning task linked to a later renewal effect. These results complement our previous findings on hippocampal context processing during extinction learning predicting later renewal.

Our comparative analysis of healthy human participants who showed or did not show a renewal effect (i.e., considered context during extinction recall) demonstrates that these groups can be distinguished as early as this initial learning phase, based on the differential patterns of their overall brain activation and of their hippocampal activation in particular.

Already during acquisition, when context was still irrelevant, participants who would later show renewal exhibited prominent activation of hippocampus. This activation was specifically related to the processing of the context–cue compound and not to mere novelty of the display: Although both REN and NOREN participants showed a hippocampal response to a presentation of the context by itself, REN participants maintained overall hippocampal processing during the following presentation of context and cue together. Moreover, brain activation patterns exhibited during acquisition were distinct enough to predict with high accuracy whether a given person would show renewal.

These findings further indicate that hippocampal processing of context information is not necessarily triggered by an unexpected change in cue–outcome contingencies during extinction but can also occur based on a, presumably, configural perception of the context–cue compound during initial learning. Furthermore, they suggest that healthy human participants may implement distinct strategies already in acquisition of an associative learning task, strategies that are reflected in distinct brain activation patterns, and that, in extinction recall, are associated with the display or lack of a renewal effect.

Acknowledgments

This work was supported by a grant from the DFG, (Deutsche Forschungsgemeinschaft; FOR 1581 Extinction Learning). We thank Tobias Otto for programming the stimulus presentation software. We appreciate the continued scientific support of Philips, Germany, including MR acquisition tools used in this study.

Reprint requests should be sent to Silke Lissek, Department of Neurology, BG University Hospital Bergmannsheil, Ruhr-University Bochum, Buerkle-de-la-Camp-Platz 1, 44789 Bochum, Germany, or via e-mail: silke.lissek@ruhr-uni-bochum.de.

REFERENCES

REFERENCES
Allan
,
L. G.
(
1993
).
Human contingency judgments: Rule based or associative?
Psychological Bulletin
,
114
,
435
448
.
Alloy
,
L. B.
, &
Tabachnik
,
N.
(
1984
).
Assessment of covariation by humans and animals: The joint influence of prior expectations and current situational information
.
Psychological Review
,
91
,
112
149
.
Arcurio
,
L. R.
,
Gold
,
J. M.
, &
James
,
T. W.
(
2012
).
The response of face-selective cortex with single face parts and part combinations
.
Neuropsychologia
,
50
,
2454
2459
.
Baeuchl
,
C.
,
Meyer
,
P.
,
Hoppstädter
,
M.
,
Diener
,
C.
, &
Flor
,
H.
(
2015
).
Contextual fear conditioning in humans using feature-identical contexts
.
Neurobiology of Learning and Memory
,
121
,
1
11
.
Bilalić
,
M.
,
Langner
,
R.
,
Ulrich
,
R.
, &
Grodd
,
W.
(
2011
).
Many faces of expertise: Fusiform face area in chess experts and novices
.
Journal of Neuroscience
,
31
,
10206
10214
.
Bouton
,
M. E.
(
1993
).
Context, time, and memory retrieval in the interference paradigms of Pavlovian learning
.
Psychological Bulletin
,
114
,
80
99
.
Bouton
,
M. E.
(
2004
).
Context and behavioral processes in extinction context and behavioral processes in extinction
.
Learning and Memory
,
11
,
485
494
.
Bouton
,
M. E.
, &
Bolles
,
R. C.
(
1979
).
Role of conditioned contextual stimuli in reinstatement of extinguished fear
.
Journal of Experimental Psychology: Animal Behavior Processes
,
5
,
368
378
.
Bouton
,
M. E.
, &
King
,
D. A.
(
1983
).
Contextual control of the extinction of conditioned fear: Tests for the associative value of the context
.
Journal of Experimental Psychology: Animal Behavior Processes
,
9
,
248
265
.
Bouton
,
M. E.
, &
Peck
,
C. A.
(
1989
).
Context effects on conditioning, extinction and reinstatement in an appetitive conditioning preparation
.
Animal Learning and Behavior
,
17
,
188
198
.
Bouton
,
M. E.
, &
Ricker
,
S. T.
(
1994
).
Renewal of extinguished responding in a second context
.
Animal Learning and Behavior
,
20
,
313
321
.
Brett
,
M.
,
Anton
,
J.-L.
,
Valabregue
,
R.
, &
Poline
,
J.-B.
(
2002
).
Region of interest analysis using an SPM toolbox
.
Paper presented at the 8th Internationa Conference on Functional Mapping of the Human Brain, June 2–6, 2002, Sendai, Japan
.
Corcoran
,
K. A.
,
Desmond
,
T. J.
,
Frey
,
K. A.
, &
Maren
,
S.
(
2005
).
Hippocampal inactivation disrupts the acquisition and contextual encoding of fear extinction
.
Journal of Neuroscience
,
25
,
8978
8987
.
Corcoran
,
K. A.
, &
Maren
,
S.
(
2001
).
Hippocampal inactivation disrupts contextual retrieval of fear memory after extinction
.
Journal of Neuroscience
,
21
,
1720
1726
.
Darby
,
R. J.
, &
Pearce
,
J. M.
(
1995
).
Effects of context on responding during a compound stimulus
.
Journal of Experimental Psychology: Animal Behavior Processes
,
21
,
143
154
.
Davachi
,
L.
, &
Wagner
,
A. D.
(
2002
).
Hippocampal contributions to episodic encoding: Insights from relational and item-based learning
.
Journal of Neurophysiology
,
88
,
982
990
.
Diana
,
R. A.
,
Yonelinas
,
A. P.
, &
Ranganath
,
C.
(
2007
).
Imaging recollection and familiarity in the medial temporal lobe: A three-component model
.
Trends in Cognitive Sciences
,
11
,
379
386
.
Effting
,
M.
, &
Kindt
,
M.
(
2007
).
Contextual control of human fear associations in a renewal paradigm
.
Behaviour Research and Therapy
,
45
,
2002
2018
.
Eichenbaum
,
H.
(
2004
).
Hippocampus: Cognitive processes and neural representations that underlie declarative memory
.
Neuron
,
44
,
109
120
.
Eichenbaum
,
H.
,
Yonelinas
,
A. P.
, &
Ranganath
,
C.
(
2007
).
The medial temporal lobe and recognition memory
.
Annual Review of Neuroscience
,
30
,
123
152
.
Fan
,
J.
,
McCandliss
,
B. D.
,
Fossella
,
J.
,
Flombaum
,
J. I.
, &
Posner
,
M. I.
(
2005
).
The activation of attentional networks
.
Neuroimage
,
26
,
471
479
.
Gluck
,
M. A.
, &
Bower
,
G. H.
(
1988
).
From conditioning to category learning: An adaptive network model
.
Journal of Experimental Psychology: General
,
117
,
227
247
.
Gold
,
J. M.
,
Mundy
,
P. J.
, &
Tjan
,
B. S.
(
2012
).
The perception of a face is no more than the sum of its parts
.
Psychological Science
,
23
,
427
434
.
Good
,
M.
,
de Hoz
,
L.
, &
Morris
,
R. G.
(
1998
).
Contingent versus incidental context processing during conditioning: Dissociation after excitotoxic hippocampal plus dentate gyrus lesions
.
Hippocampus
,
8
,
147
159
.
Howard
,
L. R.
,
Kumaran
,
D.
,
Ólafsdóttir
,
H. F.
, &
Spiers
,
H. J.
(
2011
).
Double dissociation between hippocampal and parahippocampal responses to object-background context and scene novelty
.
Journal of Neuroscience
,
31
,
5253
5261
.
Kalisch
,
R.
,
Korenfeld
,
E.
,
Stephan
,
K. E.
,
Weiskopf
,
N.
,
Seymour
,
B.
, &
Dolan
,
R. J.
(
2006
).
Context-dependent human extinction memory is mediated by a ventromedial prefrontal and hippocampal network
.
Journal of Neuroscience
,
26
,
9503
9511
.
Knight
,
R.
(
1996
).
Contribution of human hippocampal region to novelty detection
.
Nature
,
383
,
256
259
.
Köhler
,
S.
,
Danckert
,
S.
,
Gati
,
J. S.
, &
Menon
,
R. S.
(
2005
).
Novelty responses to relational and non-relational information in the hippocampus and the parahippocampal region: A comparison based on event-related fMRI
.
Hippocampus
,
15
,
763
774
.
Kumaran
,
D.
,
Hassabis
,
D.
,
Spiers
,
H. J.
,
Vann
,
S. D.
,
Vargha-Khadem
,
F.
, &
Maguire
,
E. A.
(
2007
).
Impaired spatial and non-spatial configural learning in patients with hippocampal pathology
.
Neuropsychologia
,
45
,
2699
2711
.
Kumaran
,
D.
, &
Maguire
,
E. A.
(
2009
).
Novelty signals: A window into hippocampal information processing
.
Trends in Cognitive Sciences
,
13
,
47
54
.
Lachnit
,
H.
,
Ko
,
S.
,
Üngör
,
M.
, &
Melchers
,
K. G.
(
2008
).
Comparing elemental and configural associative theories in human causal learning: A case for attention
.
Journal of Experimental Psychology: Animal Behavior Processes
,
34
,
303
313
.
Lachnit
,
H.
,
Schultheis
,
H.
,
Konig
,
S.
,
Üngör
,
M.
, &
Melchers
,
K.
(
2008
).
Comparing elemental and configural associative theories in human causal learning: A case for attention
.
Journal of Experimental Psychology: Animal Behavior Processes
,
34
,
303
313
.
Levy-Gigi
,
E.
,
Kelemen
,
O.
,
Gluck
,
M. A.
, &
Kéri
,
S.
(
2011
).
Impaired context reversal learning, but not cue reversal learning, in patients with amnestic mild cognitive impairment
.
Neuropsychologia
,
49
,
3320
3326
.
Lissek
,
S.
,
Glaubitz
,
B.
,
Güntürkün
,
O.
, &
Tegenthoff
,
M.
(
2015
).
Noradrenergic stimulation modulates activation of extinction-related brain regions and enhances contextual extinction learning without affecting renewal
.
Frontiers in Behavioral Neuroscience
,
9
,
34
.
Lissek
,
S.
,
Glaubitz
,
B.
,
Uengoer
,
M.
, &
Tegenthoff
,
M.
(
2013
).
Hippocampal activation during extinction learning predicts occurrence of the renewal effect in extinction recall
.
Neuroimage
,
81
,
131
143
.
Lissek
,
S.
,
Glaubitz
,
B.
,
Wolf
,
O. T.
, &
Tegenthoff
,
M.
(
2015
).
The DA antagonist tiapride impairs context-related extinction learning in a novel context without affecting renewal
.
Frontiers in Behavioral Neuroscience
,
9
,
1
13
.
Lober
,
K.
, &
Lachnit
,
H.
(
2002
).
Configural learning in human Pavlovian conditioning: Acquisition of a biconditional discrimination
.
Biological Psychology
,
59
,
163
168
.
Lucke
,
S.
,
Lachnit
,
H.
,
Koenig
,
S.
, &
Uengoer
,
M.
(
2013
).
The informational value of contexts affects context-dependent learning
.
Learning & Behavior
,
41
,
285
297
.
Lucke
,
S.
,
Lachnit
,
H.
,
Stüttgen
,
M. C.
, &
Uengoer
,
M.
(
2014
).
The impact of context relevance during extinction learning
.
Learning & Behavior
,
42
,
256
269
.
Mehta
,
R.
, &
Russell
,
E.
(
2009
).
Effects of pretraining on acquisition of novel configural discriminations in human predictive learning
.
Learning & Behavior
,
37
,
311
324
.
Melchers
,
K. G.
,
Shanks
,
D. R.
, &
Lachnit
,
H.
(
2008
).
Stimulus coding in human associative learning: Flexible representations of parts and wholes
.
Behavioural Processes
,
77
,
413
427
.
Milad
,
M. R.
,
Wright
,
C. I.
,
Orr
,
S. P.
,
Pitman
,
R. K.
,
Quirk
,
G. J.
, &
Rauch
,
S. L.
(
2007
).
Recall of fear extinction in humans activates the ventromedial prefrontal cortex and hippocampus in concert
.
Biological Psychiatry
,
62
,
446
454
.
Miller
,
R. R.
, &
Matute
,
H.
(
1996
).
Biological significance in forward and backward blocking: Resolution of a discrepancy between animal conditioning and human causal judgment
.
Journal of Experimental Psychology: General
,
125
,
370
386
.
Mineka
,
S.
,
Mystkowski
,
J. L.
,
Hladek
,
D.
, &
Rodriguez
,
B. I.
(
1999
).
The effects of changing contexts on return of fear following exposure therapy for spider fear
.
Journal of Consulting and Clinical Psychology
,
67
,
599
604
.
Mystkowski
,
J. L.
,
Craske
,
M. G.
,
Echiverri
,
A. M.
, &
Labus
,
J. S.
(
2006
).
Mental reinstatement of context and return of fear in spider-fearful participants
.
Behavior Therapy
,
37
,
49
60
.
Mystkowski
,
J. L.
,
Mineka
,
S.
,
Vernon
,
L. L.
, &
Zinbarg
,
R. E.
(
2003
).
Changes in caffeine states enhance return of fear in spider phobia
.
Journal of Consulting and Clinical Psychology
,
71
,
243
250
.
Nelson
,
J. B.
, &
Callejas-Aguilera
,
J. E.
(
2007
).
The role of interference produced by conflicting associations in contextual control
.
Journal of Experimental Psychology: Animal Behavior Processes
,
33
,
314
326
.
O'Reilly
,
R. C.
, &
Rudy
,
J. W.
(
2001
).
Conjunctive representations in learning and memory: Principles of cortical and hippocampal function
.
Psychological Review
,
108
,
311
345
.
Pearce
,
J. M.
(
1994
).
Similarity and discrimination: A selective review and a connectionist model
.
Psychological Review
,
101
,
587
607
.
Petersen
,
S. E.
, &
Posner
,
M. I.
(
2012
).
The attention system of the human brain: 20 Years after
.
Annual Review of Neuroscience
,
35
,
73
89
.
Preston
,
A. R.
, &
Gabrieli
,
J. D. E.
(
2008
).
Dissociation between explicit memory and configural memory in the human medial temporal lobe
.
Cerebral Cortex
,
18
,
2192
2207
.
Rosas
,
J. M.
, &
Bouton
,
M. E.
(
1997
).
Additivity of the effects of retention interval and context change on latent inhibition: Toward resolution of the context forgetting paradox
.
Journal of Experimental Psychology: Animal Behavior Processes
,
23
,
283
294
.
Rosas
,
J. M.
, &
Callejas-Aguilera
,
J. E.
(
2006
).
Context switch effects on acquisition and extinction in human predictive learning
.
Journal of Experimental Psychology: Learning, Memory, and Cognition
,
32
,
461
474
.
Rosas
,
J. M.
,
Todd
,
T. P.
, &
Bouton
,
M. E.
(
2013
).
Context change and associative learning
.
Wiley Interdisciplinary Reviews: Cognitive Science
,
4
,
237
244
.
Ryals
,
A. J.
,
Wang
,
J. X.
,
Polnaszek
,
K. L.
, &
Voss
,
J. L.
(
2015
).
Hippocampal contribution to implicit configuration memory expressed via eye movements during scene exploration
.
Hippocampus
,
25
,
1028
1041
.
Schrouff
,
J.
,
Rosa
,
M. J.
,
Rondina
,
J. M.
,
Marquand
,
A. F.
,
Chu
,
C.
,
Ashburner
,
J.
, et al
(
2013
).
PRoNTo: Pattern recognition for neuroimaging toolbox
.
Neuroinformatics
,
11
,
319
337
.
Shanks
,
D. R.
,
Holyoak
,
K.
, &
Medin
,
D. J.
(
1996
).
Causal learning
.
San Diego, CA
:
Academic Press
.
Shohamy
,
D.
,
Myers
,
C. E.
,
Hopkins
,
R. O.
,
Sage
,
J.
, &
Gluck
,
M. A.
(
2009
).
Distinct hippocampal and basal ganglia contributions to probabilistic learning and reversal
.
Journal of Cognitive Neuroscience
,
21
,
1821
1833
.
Smith
,
D. M.
, &
Mizumori
,
S. J. Y.
(
2006
).
Learning-related development of context-specific neuronal responses to places and events: The hippocampal role in context processing
.
Journal of Neuroscience
,
26
,
3154
3163
.
Todd
,
T. P.
,
Winterbauer
,
N. E.
, &
Bouton
,
M. E.
(
2012
).
Effects of the amount of acquisition and contextual generalization on the renewal of instrumental behavior after extinction
.
Learning & Behavior
,
40
,
145
157
.
Tzourio-Mazoyer
,
N.
,
Landeau
,
B.
,
Papathanassiou
,
D.
,
Crivello
,
F.
,
Etard
,
O.
,
Delcroix
,
N.
, et al
(
2002
).
Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain
.
Neuroimage
,
15
,
273
289
.
Üngör
,
M.
, &
Lachnit
,
H.
(
2006
).
Contextual control in discrimination reversal learning
.
Journal of Experimental Psychology: Animal Behavior Processes
,
32
,
441
453
.
Vansteenwegen
,
D.
,
Hermans
,
D.
,
Vervliet
,
B.
,
Francken
,
G.
,
Beckers
,
T.
,
Baeyens
,
F.
, et al
(
2005
).
Return of fear in a human differential conditioning paradigm caused by a return to the original acquistion context
.
Behaviour Research and Therapy
,
43
,
323
336
.
Vansteenwegen
,
D.
,
Vervliet
,
B.
,
Hermans
,
D.
,
Beckers
,
T.
,
Baeyens
,
F.
, &
Eelen
,
P.
(
2006
).
Stronger renewal in human fear conditioning when tested with an acquisition retrieval cue than with an extinction retrieval cue
.
Behaviour Research and Therapy
,
44
,
1717
1725
.
Walter
,
B.
,
Blecker
,
C.
,
Kirsch
,
P.
,
Sammer
,
G.
,
Schienle
,
A.
,
Stark
,
R.
, et al
(
2003
).
MARINA: An easy to use tool for the creation of MAsks for Region of INterest Analyses
.
Paper presented at the 9th International Conference on Functional Mapping of the Human Brain, June 19–22, 2003, New York, NY
.
Yamaguchi
,
S.
,
Hale
,
L. A.
,
D'Esposito
,
M.
, &
Knight
,
R. T.
(
2004
).
Rapid prefrontal-hippocampal habituation to novel events
.
Journal of Neuroscience
,
24
,
5356
5363
.
Zhang
,
J.
,
Li
,
X.
,
Song
,
Y.
, &
Liu
,
J.
(
2012
).
The fusiform face area is engaged in holistic, not parts-based, representation of faces
.
PLoS One
,
7
,
e40390
.