Remembering Objects in Contexts: ERP Correlates of a
Modified Behavioral Pattern Separation Task
Miriam Hollarek
10608753
August 2015
Master Thesis
Graduate School of Psychology Department of Brain and Cognition University of Amsterdam/ Lund University First Supervisor: Jeroen Raaijmaakers Second Assessor: Jaap Murre
Table of Contents
Abstract 3
Introduction 4
Methods 10
Participants 10
Stimuli and Design 11
Procedure 12
EEG Recording and Preprocessing 13
Results 13 Behavioral Results 13 ERPs 15 Discussion 20 References 27 Appendix 34
Abstract
This study examined the so far unknown event-related potentials (ERPs) associated with pattern completion and pattern separation, two proposed processes in episodic memory. In a behavioral pattern separation task with the addition of contexts and their direct modification, participants encoded pictures of objects in contexts and were subsequently asked to identify if stimuli were (1) repetitions, (2) new, or (3) similar, depicting a different version of the same exemplar. If classified as similar, participants had to further judge whether the object, the context, or both was similar. We expected to find that false alarms to similar presentations would resemble ERPs associated with familiarity, and correct rejections of similar presentations would resemble ERPs associated with recollection. Analysis focused on the frontal and parietal old/new effects, previously identified as neural correlates of familiarity and recollection. Results showed a tendency towards a frontal old/new effect during test but no parietal differences between the presented picture types. These results suggest that familiarity mechanisms might be operating during pattern completion, though pattern separation can occur in the absence of recollection.
Remembering Objects in Contexts: ERP Correlates of a Modified Behavioral Pattern Separation Task
Two essential processes for efficient encoding and retrieval in episodic memory are assumed to be pattern separation and pattern completion. These constructs have been used in models of memory to describe specific forms of memory formation and retrieval. The models suggest that pattern separation occurs during encoding when similar, overlapping input is stored orthogonally in separate representations. Pattern completion on the other hand is suggested to occur during retrieval, when exposure to a stimulus, similar to a previously encountered stimulus, results in the retrieval of one common representation (Hunsaker & Kesner, 2013). Thereby, pattern separation minimizes interference of competing memories and thus permits later memory discrimination (Brock Kirwan et al., 2012), while pattern completion allows for reactivation of previously encoded patterns from partial or degraded cues (Yassa et al., 2011).
There is a general consensus about the anatomical structures involved in pattern separation and completion (Kim & Yassa, 2013). After computational models had suggested a key role of sub-regions of the hippocampus, emphasizing the optimal structural properties of the
hippocampus for the formation and retrieval of episodic memories (Kesner, Gilbert, & Wallenstein, 2000; Treves & Rolls, 1994), many researchers set out to test this theory using intracellular recordings in rodents. These studies have identified the CA3 and the dentate gyrus as important hippocampal sub-regions for pattern separation, while the CA1 appears to be involved in pattern completion (Denny et al., 2014; Kesner et al., 2000; Kesner, 2013; Lacy, Yassa, Stark, Muftuler, & Stark, 2011; J. K. Leutgeb, Leutgeb, Moser, & Moser, 2007; S. Leutgeb & Leutgeb, 2007; van Hagen, van Goethem, Lagatta, & Prickaerts, 2015; Vazdarjanova & Guzowski, 2004)
In order to investigate these neuronal substrates in humans, a pattern separation task was designed to resemble tasks of rodent studies (Bakker, Kirwan, Miller, & Stark, 2008; Lacy et al., 2011). During this task participants have been viewing series of objects which are either exact replications of previously seen objects, completely new objects; or belong to a third category displaying similar, but not identical objects. The associated brain structures in humans, as assessed through functional magnetic resonance imaging (fMRI) recordings during viewing of the similar stimuli, closely resemble previous findings in rodents. These studies, together with studies of patients with hippocampal lesions, relate the bilateral CA3 and DG to pattern
separation, and suggest a key role of the CA1 for pattern completion (among others: Azab, Stark, & Stark, 2014; Bakker et al., 2008; Duff et al., 2012; Hunsaker & Kesner, 2013; Kirwan & Stark, 2007; Lacy et al., 2011; Paleja, Girard, Herdman, & Christensen, 2014).
While these experiments compared brain activation during mere viewing of the lure objects (Bakker et al., 2008), later studies included overt responses to be able to obtain a measure of pattern separation abilities. During these experiments participants are asked to label the viewed stimuli as old, new or similar. Lure presentations which are correctly identified as similar are viewed as successful pattern separation, whereas lure presentations which are incorrectly identified as old are treated as examples of pattern completion.
Employing the techniques described above, another line of research proposed a link between adult neurogenesis in the hippocampus and pattern separation abilities. Correlational studies seem to support this view. Structural changes in the DG and CA3 area and worse pattern separation deficits has been found in older adults, a population known to have decreased
neurogenesis (H. M. Holden, Toner, Pirogovsky, Kirwan, & Gilbert, 2013; Heather M. Holden & Gilbert, 2012; Heather M. Holden, Hoebel, Loftis, & Gilbert, 2012). Further, depressed
individuals have been found to have decreased adult neurogenesis as well as decreased pattern separation abilities, and lastly exercising, known to increase adult neurogenesis, also improved pattern separation abilities (Déry et al., 2013).
The importance of intact pattern separation abilities has been highlighted by its clinical relevance as various psychiatric disorders have been found to be associated with decreased task performance. For instance, patients with anxiety disorders and post-traumatic stress disorder (PTSD) are known to overgeneralize contextual cues related to the feared stimulus or the experienced trauma (Hull, 2002; Kheirbek, Klemenhagen, Sahay, & Hen, 2012). Accordingly, studies showed a decrease in hippocampus volume in these patients together with a tendency towards pattern completion (Hull, 2002; Kheirbek et al., 2012). Moreover, impaired pattern separation abilities were identified in people suffering from schizophrenia (Das et al., 2014), and as stated above, in people with depression, including subclinical levels of depressive symptoms (Déry et al., 2013; Fujii, Saito, Yanaka, Kosaka, & Okazawa, 2014). Finally, patients with Alzheimer’s disease (AD) also exhibit impaired pattern separation abilities (Ally et al., 2013).
While the neuronal substrates associated with pattern separation and pattern completion have received much attention, electrophysiological correlates of the processes have not yet been directly investigated. Due to the lack of electroencephalogram (EEG) studies and because of the poor temporal resolution of the fMRI signal, little is known about the temporal dynamics of pattern separation and pattern completion. However, a closely linked line of research has
received much attention in the EEG literature, namely recognition memory and with it the notion of the dual process theory.
The dual process theory suggests that recognition can be based on two distinct processes: familiarity and recollection (Mandler, 1980). To empirically study the processes, familiarity was
defined as a sensation of general similarity between a stimulus and a previously encountered stimulus, although participants are unable to retrieve any additional episodic information.
Recollection, on the other hand, refers to the conscious retrieval of a specific prior event that will lead to recognition of the stimuli (Gruber, Tsivilis, Giabbiconi, & Müller, 2008; Mandler, 1980).
One typical design to test the dual process theory involves source memory judgement. Here recognition is thought to occur on the basis of recollection if participants can retrieve additional information about the previously encountered stimuli, such as where or when they first encountered the stimuli; while recognition based on familiarity is thought to take place if they recognize the item but fail to provide any additional information regarding the first encounter (Janowsky, Shimamura, & Squire, 1989). Another widely used design is the Remember/Know paradigm. During this task participants overtly indicate if they recognize an item because they can actively ‘remember’ the event when they first encountered the item, thus depending on recollection, or if they simply ‘know’ they have seen an item before, based on feeling of familiarity (Tulving, 1985). Note, however, that this paradigm has been criticized for not being ‘process pure’, as results can also be explained by high versus low memory strength of
familiarity and recollection processes (Wixted, 2009).
Employing these techniques, ERP studies have identified two separate components believed to reflect familiarity and recollection processes, the frontal old/new effect, or FN400, and the parietal old/new effect, also referred to as late positive component (LPC), respectively (Tim Curran & Cleary, 2003; Rugg & Curran, 2007). For example one study found that the magnitude of the parietal old/new effect co-varied with the number of correct source memory judgements (Wilding, 2000). Another study presented plural word forms of previously shown singular words, and vice versa, as similar stimuli and found that the FN400 old/new effect
differentiated similar (and studied) words from new words, that is familiar words from new words, while the parietal old/new effect differentiated similar (and new) words from studied words, that is similar and new words from exact repetitions that can be recollected (Curran, 2000). Most closely related to the behavioral pattern separation task, the same authors also demonstrated this selectivity of familiarity of the frontal old/new effect and recollection for the parietal old/new effect in pictures of objects and their mirrored image as similar items (Curran & Cleary, 2003).
The task demands posed by the experiments investigating the dual process theory seem closely related to those posed by the behavioral pattern separation task. Successful pattern separation as indicated by correct identification of a similar lure, might occur if the response is based on a rich memory of the original item, thus based on recollection. Contrary, if participants base their response on familiarity a similar lure might be falsely labeled as old, thus reflecting pattern completion. This idea was tested in a behavioral pattern separation task, where
participants additionally provided Remember/Know - and source memory judgments during classification of similar objects (Kim & Yassa, 2013). In contrast to Curran & Cleary (2003), results showed that not only correctly identified similar object (presumed pattern separation), but also similar objects which were falsely remembered as old (presumed pattern completion) were associated with a bias towards remember judgements, as well as a higher number of correct source judgements. Participants thus seemed to be able to recollect additional information from the prior study episode even when their memory for the studied object was imprecise. The authors thus concluded that recollection can occur during both pattern separation and pattern completion and that pattern completion is not simply driven by familiarity (Kim & Yassa, 2013).
In this study we aim to further explore the relationship between pattern separation and the dual process theory by recording EEG signals during a behavioral pattern separation task. This study, as opposed to previous behavioral pattern separation tasks, additionally included the integration of objects in a contexts. With these modified stimuli participants were unable to simply rely on object recognition to succeed in the task, but had to integrate objects and contexts. Successful identification of similar pictures in this task should thus require recollection. Firstly, this extension adds more external validity to the task, since in real life objects are almost never encountered isolated in blank backgrounds. Secondly, object context integration also heavily depends on the identified underlying brain structure for pattern separation, the hippocampus, and therefore the task ensures hippocampus involvement (Eichenbaum & Cohen, 2014; Howard & Eichenbaum, 2014) and thus might more closely reflect the proposed mechanisms of pattern separation and pattern completion. Moreover, the inclusion of contexts in the task also strengthens its clinical relevance, as studies on PTSD showed that the overgeneralization of contexts in particular, contributes heavily to PTSD symptoms (Levy-Gigi, Szabo, Richter-Levin, & Kéri, 2015). Lastly, the modified stimuli not only permit future research to advance the debate about the role of context in object recognition, but it allows to more directly address eventual recognition effects of similar contexts itself, a process which has received very little attention so far.
While we are not aware of any EEG studies directly assessing recognition of similar contexts, there are some studies targeting the role of context in object recognition. Claims of contextual familiarity effects on object recognition have been raised (Tsivilis, Otten, & Rugg, 2001). In contrast others have argued that context only indirectly influences familiarity effects in object recognition, for example via attention and that context might actually elicit a separate
familiarity signal on its own (Ecker, Zimmer, Groh-Bordin, & Mecklinger, 2007). The role of context recognition in familiarity, but even more so effects directly associated with context recognition seem to be not yet fully resolved.
Briefly, by recording EEG signals during our modified behavioral pattern separation task, this study aimed to shed more light on the temporal dynamics of pattern separation and pattern completion and how the processes relate to the ERP components associated with recollection and familiarity, during retrieval. Specifically we hypothesized that during test similar stimuli, falsely classified as old would elicit more positive going ERPs in frontal regions (300 – 500 ms), thus showing a frontal old/new effect and resembling findings of familiarity. And secondly, during test old presentations would elicit more positive going ERPs than correctly identified similar stimuli in parietal regions (500 – 800 ms), thus showing a parietal old/new effect and resembling findings of recollection.
Method
Participants
Thirty-one participants (17 female, age range 19-33) took part in the experiment in exchange for a movie ticket. Before starting the experiment, participants signed an informed consent form. All participants were free of neurological deficits, right handed and had normal or corrected to normal vision and basic understanding of the English language. One participant was excluded from all ERP analyzes due to equipment failure.
Stimuli and Design
Adapted from previous studies (Bakker et al., 2008), the stimuli consisted of combinations of 430 paired original and similar objects, and 430 paired original and similar contexts. The
objects were displayed in the center of the screen, superimposed on the context scenes. To ensure desired rates of correct responses, the stimuli were validated in a separate pilot study. The study consisted of two blocks, each block started with an encoding phase containing 120 study trials, followed by a retrieval phase containing 150 test trials. Study stimuli consisted of randomly selected object- context combinations with 60 exemplars per condition. The paired test stimuli also consisted of 60 exemplars per condition and belonged to one of the following conditions: (1) “new”, a completely new presentation, i.e. a new object in a new context; (2) “old”, an exact repetition of a previously seen stimulus, i.e. the same object in the same context; or (3)
“similar”, a similar presentation. In the similar conditions participants were asked to additionally differentiate between (4) “similar object”, a similar object in an old context; (5) “similar
context”, a similar context with an old object; and (6) “both similar”, a similar object in a
similar context. Example stimuli are displayed in figure 1. All object - context combinations were randomized between subjects and presentation order was randomized within blocks. The response buttons and the presentation order of the blocks were counterbalanced between participants. The experiment was programmed and implemented in E-Prime 2.0 software.
Procedure
Participants sat down in front of a computer screen at a distance of approximately 70 cm. After the application of all EEG electrodes, participants were instructed to relax their facial muscles throughout the experiment and to avoid blinking during stimulus presentation. Task instructions were displayed on the screen before the start of each experimental phase.
During the encoding phase participants viewed the above described stimuli. Presentation of a stimulus always started with a fixation cross (700 ms) displayed in the center of the screen. Subsequently a blank screen was presented (300 ms) followed by the stimulus picture (2000 ms).
Then a response cue was displayed during which participants were asked about their own opinion on whether the object matched the context. This study response was implemented to ensure that participants studied both object and context. Once participants pressed a
corresponding response button the next trial started.
Figure 1: Example of an original study picture (top middle) and all possible paired test
conditions.
After the encoding phase, the test phase started during which participants were asked to identify the different types of stimuli. Before the first test phase started, an example of an original object in its original context and its paired similar object and similar context were displayed side by side to illustrate in what way similar test pictures differed from the original study picture. Each stimulus was again preceded by a fixation cross (700 ms) and blank screen (300 ms), followed by the stimulus (2000 ms). Then a response cue was presented “new, old, or
similar” until participants indicated their response on the corresponding button. Whenever participants responded with “similar”, a second response cue appeared asking the participants to judge whether the object was similar while the context was old, the context was similar while the object was old, or both was similar. Participants could decide to take a short break to rest their eyes between every 20 trials. For 30 seconds between phases and blocks participants were instructed to count backwards in steps of three from a random three digit number.
EEG Recording and Preprocessing
The EEG signal was continuously recorded using 32 Ag/AgCl scalp electrodes, positioned according to the extended 10–20 system (Jasper, 1958). Electro-scalp impedances were kept below 5 Ω. Electrodes were referenced to the left mastoid and re-referenced off-line to the average of both left and right mastoids. To monitor electrooculogram, additional electrodes were placed below the left eye and on the outer canthi of the left and right eye. All voltages were amplified and digitalized at a 22 bit resolution with a 1000 Hz sampling rate.
After downsampling of to the EEG data to 500 Hz, a digital band-pass filter was applied (0.1 - 40 Hz). The data was epoched, starting 300 ms prior to stimulus onset and ending 2000 ms after stimulus onset. The pre-stimulus interval was used for baseline correction. Ocular artifacts were corrected using independent component analysis (Jung et al., 2000). Epochs containing amplitude differences between the most positive and most negative voltage which exceeded 100 µV within a 200 ms interval, were rejected prior to averaging. Consistently bad channels were replaced by their spherical interpolation for seven subjects with a median of one interpolated channel per subject. All preprocessing steps were implemented in ERPLAB and self-written Matlab code (Lopez-Calderon & Luck, 2014).
Results
Behavioral Results
Proportion correct was analyzed separately for the broad picture types (old, new, all similar) and the specific similar conditions (object similar, context similar, both similar), corresponding to the participants’ first and second response, respectively. Greenhouse - Geisser corrections were used in case of violations of sphericity and Bonferroni corrections were used to adjust for multiple testing.
First, differences between broad conditions were assessed in a repeated measures ANOVA with dependent variable proportion correct and independent variable broad condition. The analysis showed significant differences in proportion correct between broad conditions: New (M = 0.84, S.E. = 0.02), and Old (M = 0.80, S.E. = 0.02) > All similar (M = 0.59, S.E. = 0.03) (F (2, 60) = 50.00, p < 0.001, ηp² = 0.47). As can be expected, post hoc pairwise comparisons yielded significantly higher proportions correct for new than similar pictures (p < 0.001) as well as significantly higher proportion correct for old than similar pictures (p < 0.001). Accuracies are displayed in figure 2 (left).
Figure 2: Left: Behavioral Accuracy of broad picture types (all similar presentations collapsed
which were correctly identified as similar). Right: Accuracy of specific similar conditions (proportion of trials first correctly identified as similar and subsequently correctly identified which aspect had changed). Error bars depict standard errors.
In a second repeated measures ANOVA we also included the participants’ study response - matching or not matching object context combinations - as an independent variable to see whether their study response affected later accuracy during the test phase. Since the participants’ study response is nonexistent for the picture type ‘new’, this variable was not included in the analysis of the broad conditions. We thus conducted a specific conditions (object similar, context similar, both similar) x study response (matching, nonmatching) repeated measures ANOVA. The analysis showed a main effect of specific conditions: Object similar (M = 0.40, S.E. = 0.03), and context similar (M = 0.43, S.E. = 0.03) > both similar (M = 0.28, S.E. = 0.03) (F (2, 60) = 21.19, p < 0.001, ηp² = 0.41). Post hoc pairwise comparisons showed significantly higher proportion correct for object similar than both similar (p < 0.001) and significantly higher proportion correct for context similar than both similar (p < 0.001). Accuracies are displayed in figure 2 (right).
Moreover a main effect of study response was found: Matching (M = 0.40, S.E. = 0.16) > nonmatching (M = 0.35, S.E. = 0.15) (F (2, 29) = 8.23, p = 0.007, ηp² = 0.22). However this difference cannot reliably be attributed to the matching or nonmatching nature of the stimuli since the frequency of the responses differed greatly; 70% of combinations were rated nonmatching compared to 30% matching. As the study was not designed to investigate this difference and since the different frequencies introduce a confounding variable, this distinction was not further explored. There was no significant interaction effect between study response and specific conditions (F (2, 29) = 2.63, p = 0.08, ηp² = 0.08).
ERP Results
First, ERPs were analyzed during retrieval for the following broad picture types: (1) ‘old hits’, i.e. correctly identified old presentations; (2) ‘new correct rejections’, i.e. correctly
identified new presentations; (3) ‘similar correct rejections’, i.e. all similar presentations collapsed which were correctly identified as similar; (4) ‘similar false alarms’, i.e. all similar presentations falsely identified as old. For these analyses we analyzed mean amplitudes for the frontal and parietal old/new effect time windows (350 - 500 ms and 500 - 800 ms), as well as for a third late time window (900 - 1200 ms) (Johansson, Mecklinger, & Treese, 2004). Regions consisted of pooled electrodes for left frontal (FP1, F3, F7, FC5), mid frontal (FZ, FCZ, FC1, FC2), right frontal (FP2, F4, F8, FC6), left parietal (P7, P3, CP5, PO9), mid parietal (PZ, CP1, CP2), and right parietal regions (P4, P8, CP6, PO10). These six pooled regions were analyzed for each time window separately, resulting in three repeated measures ANOVAs with independent variables picture type (old hit, new correct rejection, similar correct rejection, similar false alarm) X AP (anterior, posterior) X hemisphere (left, mid, right). Following previous studies, subsequent ANOVAs comparing picture types in priori selected regions of interests (ROIs) were
performed for the frontal and parietal old/new effects, reflecting sites where these effects were found to be maximal in previous studies. These electrodes were F3, FZ, F4 for the frontal old/new effect, and P3, PZ, P4 for the parietal old/new effect (Curran & Cleary, 2003; Curran & Friedman, 2004).
Results for the old/new effects during test and subsequent performance effects during study start with a description of the general grand average wave forms and scalp topographies before reporting statistical results. For all analyses only significant effects involving the
experimental manipulation of picture type are reported, tables presenting complete ANOVAs can be found in the appendix.
Old/new effects during test.
Twenty-six participants with a minimum of 15 trials per condition were included in the analysis. Mean trial numbers were 39, 41, 87, and 42, for old hits, new correct rejections, similar correct rejections, and similar false alarms, respectively.
Visual inspection of the waveforms for similar false alarms and new correct rejections from the frontal electrodes F3, FZ, and F4 showed the expected old/new pattern during the corresponding time window for the frontal old/new effect (350 - 500 ms): Perceived old presentations, i.e. similar false alarms were more positive than new correct rejections (figure 2A). The Scalp topography demonstrates that this difference is widely spread in frontal, up to central areas and is most prominent in midline regions (figure 2C, left). ERPs from the partial electrodes P3, PZ, and P4 do not seem to differ for correctly recognized old presentations, and similar correct rejections during the corresponding time window for the parietal old/new effect (500 - 800 ms) (figure 2B). The scalp topography indicates that instead of the expected pattern of
more positive voltages for old stimuli, similar correct rejections elicited higher voltages in parietal regions as compared to old hits (figure 2C, right).
Figure 3: (A) Grand average waves of frontal electrodes F3, FZ, and F4. Boxes indicate the
frontal old/new effect time window (350 – 500 ms). (B) Grand average waves of frontal electrodes P3, PZ, and P4. Boxes indicate the parietal old/new effect time window (500 – 800 ms). (C) Scalp topographies depicting the average amplitude difference of similar false alarms minus new correct rejections from 350 - 500 ms (left), and old hits minus similar correct rejections from 500 – 800 ms (right)
A
C
B
Frontal old/new effect. The statistical analysis of the frontal old/new effect (350 - 500
ms), employing the six regions analysis, as described above, revealed no significant interactions or main effects involving picture type (appendix A1). Nevertheless, as many previous studies only employed ROI analysis using electrodes where the effect had been observed to be maximal, this analysis was followed by a ROI analysis. The ROI repeated measures ANOVA showed a significant main effect of picture type with means of old hits and similar false alarms showing more positive going ERPs than new correct rejections (F (3, 78) = 2.95, p = 0.038, ηp² = 0.10). An overview of all means is given in table 1. However pairwise comparisons employing Bonferroni multiple comparison corrections showed no significant differences (old hits against new correct rejections, p = 0.14; similar false alarms against new correct rejections, p = 0.3), challenging the effect’s reliability.
Table 1: Mean voltages and their standard deviations of ROI analysis for the frontal and parietal
windows and parietal regions for the late window.
Condition Frontal (300 – 500 ms) Parietal (500 - 800 ms) Parietal (900 - 1200 ms) Old Mean (SD) -2.43 (1.96) 2.76 (4.08) 0.51 (0.45) New Mean (SD) -2.87 (2.29) 3.73 (3.27) 1.65 (0.42) Similar Hit Mean
(SD) -2.51 (1.82) 3.13 (3.70) 0.32 (0.40) Similar Miss Mean
(SD) -2.40 (2.05) 3.04 (4.04) 0.11 (0.48)
Parietal old/new effect. The six regions analysis of the parietal old/new effect (500 - 800
ms) showed no significant main effects, but a significant interaction effect involving picture type and AP (F (3, 72) = 4.76, p = 0.008, ηp² = 0.165) (see appendix A2). Follow up simple main effects of parietal regions, however, showed no significant differences between picture types. Moreover, examination of the means (Table 1) demonstrate that the observed tendencies exhibit an opposite direction than would be expected for the parietal old/new effect; ERPs of new correct rejections were more positive going than all other conditions. Therefore the observed tendencies do not seem to reflect parietal old/new differences of recollection. Confirming this view, the following ROI analysis showed no main effect of picture type (F (3, 78) = 2.53, p = 0.077, ηp²= 0.89).
Late window. A third analysis using the six regions analysis was conducted for a late time window (900 - 1200 ms). A significant main effect of picture type was found (F (3, 72) = 7.65, p < 0.001, ηp² = 0.242), with pairwise comparisons showing that new > similar hit (p = 0.018) and new > similar miss (p = 0.002). Further, a significant interaction between broad conditions and hemisphere was observed (F (3, 72) = 7.42, p < 0.001, ηp²= 0.238). Follow up analysis of simple main effects showed significant differences of picture type for the parietal regions (F (3, 72) = 9.42, p < 0.001, ηp² = 0.282). Pairwise comparisons showed that new > similar hit (p = 0.004) and new > similar miss (p < 0.001). As we had no a priori selected ROIs or predictions for this effects, no ROI analysis was conducted for the late window.
Discussion
Behavioral performance on this modified pattern separation task demonstrated that participants are equally able to identify test presentations as similar, regardless if the object or
the context was slightly changed. Performance for ‘both similar’ conditions was lowest, as the frequency of false identification as new increased. ERPs during the test phase of the task showed tendencies towards a frontal old/new effect for similar false alarms and old hits. In other words, all presentations perceived as old elicited more positive going ERPs than new correct rejections. However these tendencies were not reliable after correction for multiple comparisons. Moreover no reliable parietal old/new effects were found for any of the picture types. Only during the late window reliable differences were observed, demonstrating a reversed late parietal old/new effect during which new correct rejections showed more positive going ERPs than both similar correct rejections and similar false alarms. In the following we discuss what our findings suggest about the relationship between the two processes of pattern completion and pattern separation and the dual process theory, and how the addition of context in this task might contribute to further our understanding of the role of context in recognition memory.
During test, a tendency towards a frontal old/new effect was found, distinguishing similar false alarms from new correct rejections. No such old/new tendencies were observed
distinguishing similar correct rejections from new correct rejections. These results are in line with Curran and Cleary (2003), who found frontal old/new effects which distinguished between correctly rejected new objects and their similar false alarms (using the object’s mirrored image as their similar presentation), but not between new correct rejections and similar correct rejections. Previous research has identified the frontal old/new effect as a neural correlate of familiarity (reviewed in: Rugg & Curran, 2007). In our task, similar false alarms reflect a failure to separate the similar presentation from the original study image, thus representing pattern completion. Similar false alarms and new correct rejections differences were evident in the frontal old/new effect, and may therefore be due to different levels of familiarity associated with them. Assuming
that the observed tendencies reflect genuine reliable frontal old/new effects, this would suggest that higher engagement of familiarity processes is associated with a bias towards pattern
completion. However, before any strong conclusions can be drawn, planned pairwise comparison of similar false alarms and new correct rejections in a separate data set need to validate the effect, as the observed frontal old/new effect was not reliable after correction for multiple comparison.
Analysis concerning the parietal old/new effect revealed no differences between any of the picture types. There are several reasons which might explain the lack of this neural correlate of recollection in our study.
Firstly, the utilization of recollection mechanisms might differ between participants. The large performance variance for similar correct rejections could influence the ERPs associated with recollection. We did not provide any explicit instruction on which strategy participants should use to reach a decision and strategies.Moreover, abilities probably differ greatly between participants. Hence, one possibility is that only good performers show a parietal old/new effect. Such a distinction has previously been reported in Curran and Cleary (2003).
Secondly, one could still argue that recollection is not required to successfully identify the broad picture type ‘similar’. Before participants are asked to specifically identify which aspect, object or context, has changed in the similar presentation, participants do not need to recollect specific details about the study episode, but could employ a strategy where they reject pictures with low levels of familiarity as new (similar correct rejections), and retain pictures with high levels of familiarity as old (similar false alarms). The lack of evidence for the need of recollection in pattern separation is in line with Kim and Yassa (2013) who argue that pattern separation can occur both in the presence and in the absence of recollection, based on their
findings of remember/know and source memory judgements in a behavioral pattern separation task. Still, recollection effects could possibly be evident for similar correct rejections where the second question, which aspect was similar, had also been answered correctly. However, since participants are asked to make a stepwise decision this might only be reflected during a later stage.
Regardless of the above described assumptions, pattern separation did take place, because similar correct rejection rates were high, reflecting participants’ ability to separate similar
presentations from its original study image. Neural correlates of recollection, however were absent and explanations for this lack are readily available. Taken together these results thus suggests that at least with previous and present task demands, pattern separation can occur without recollection.
Lastly, our results regarding the parietal (and frontal) old/new effect might also differ due to the nature of the stimuli used in this modified task. The addition of contexts in this task and related possible impacts on our results and their potential interpretations are discussed next. In contrast to previously employed behavioral pattern separation tasks, this study added and also manipulated contexts in which the objects were presented. Regarding familiarity effects, previous research has proposed opposing theories of the role of context. Tsivilis et al. (2001) have argued that context can also affect recognition based on familiarity. This suggestion was based on their observation that objects presented in a new context elicited a less pronounced frontal old/new effect (Tsivilis et al., 2001). Ecker et al. (2007) on the other hand argue that contextual modulations of the frontal old/new effect are indirect and might be mediated by other factors, such as attention. In their study they showed that the contextual effects on the frontal old/new effect diminish when participants are cued to direct their attention towards the object
(Ecker et al., 2007). Moreover the authors raised the issue to what extend objects and contexts can be regarded as an item (object) and its background (context), instead of two separate and independent entities. The authors suggested that salient contexts might elicit their own
familiarity or recollection effects. Hence, the observed tendency towards a frontal old/new effect in this study could either stem from a relative familiarity of a united representation of object and context, or from two separate familiarity effects of objects and contexts. Further examination of the differences between ERPs elicited by similar objects and similar contexts might help to disentangle the underlying processes.
While it is generally agreed that the retrieval of contextual specifics requires recollection (Rugg & Yonelinas, 2003), context additions in this study might have contributed to the
observed lack of a parietal old/new effect. If objects and contexts were treated as two separate entities, each would be able to elicit their own familiarity or recollection signal. Since the context for ‘object similar’ and the object for ‘context similar’ conditions were always old, every
presentation could be based on familiarity to some extent. This logic does not apply to the ‘both similar’ condition, however behavioral results showed that these presentations were often falsely identified as new, and were therefore probably perceived as too different to evoke recollection processes.
Even if objects and contexts are processed as two separate and independent entities, integration processes are still likely to occur during encoding and retrieval of this task.
Therefore, the context addition seems to more closely resemble task demands for real life pattern separation where multiple item integration is common. Furthermore, though future analysis using this task we might gain more insights into the possibly separate processing of objects and
recollection, in the following we discuss the observed differences during our third time window of interest.
A reliable effect showing more positive going ERPs for new correct rejections as compared to similar false alarms and similar correct rejections was found in parietal regions during a late window. This effect might depict a late posterior negative slow wave (LPN) which has previously been reported in studies requiring the binding of items with contextual
information and is typically more negative for previously studied items (among others: Curran, 2000; Cycowicz, Friedman, & Snodgrass, 2001). The observed differences have been proposed to reflect reconstruction of the prior study episode during response conflict, when relevant specifics are not readily available and thus need prolonged evaluation (Johansson & Mecklinger, 2003). Our findings neatly fit these interpretations, as reconstruction of the study episode can assist to identify the more ambiguous similar presentations which ask for extended evaluation, and accordingly elicited a more negative going LPN. Reconstruction of the study episode on the other hand is not possible for new presentations, and accordingly new presentations showed a more positive going LPN.
A possible limitation of the context addition in this task is a potential limited
distinctiveness of the context stimuli pairs. Even though we used a wide variety of indoor and outdoor scenes ranging from bedrooms and supermarkets to beaches and mountains, we cannot rule out the possibility that some of the 300 presented contexts evoke familiarity not only for their similar pair, but for example for all contexts depicting the outside perspective of a
monument. Even though ERPs could have been affected hereby, comparable behavioral accuracy for correctly identified similar objects and correctly identified similar contexts suggest that participants at least consciously distinguish the contexts.
Future investigations are crucial in order to address a number of issues. Importantly, pattern separation has been defined to occur during encoding. Future studies should thus include ERP analysis during the study phase to see whether successful pattern separation can be
predicted by ERPs during encoding. The separate contributions of objects and contexts were not the main concern of this study and therefore not analyzed. These differences should be directly addressed. This is in principle possible using in the current data set, given that we have enough artifact free trials for each specific condition. Further, future studies should try to direct
tendencies towards pattern separation. Successful manipulation, possibly by directing attention towards a specific aspects of a presentation could bias participants towards pattern separation and might in the long run help to decrease symptoms of PTSD, anxiety disorders and depression.
In sum, the present study does not support the notion that the dual process theory can fully account for the processes of pattern completion and pattern separation. Our results suggest that while familiarity processing seems to be elevated during pattern completion, pattern
separation can also occur in the absence of recollection processes. Still, the possibility remains that specific task requirements of the behavioral pattern separation task might alter the
contributions of familiarity and recollection and could introduce a necessity of recollection for successful pattern separation.
References
Ally, B. a., Hussey, E. P., Ko, P. C., & Molitor, R. J. (2013). Pattern separation and pattern completion in Alzheimer’s disease: Evidence of rapid forgetting in amnestic mild cognitive impairment. Hippocampus, 23, 1246–1258. http://doi.org/10.1002/hipo.22162
Azab, M., Stark, S. M., & Stark, C. E. L. (2014). Contributions of human hippocampal subfields to spatial and temporal pattern separation. Hippocampus, 24, 293–302.
http://doi.org/10.1002/hipo.22223
Bakker, a., Kirwan, C. B., Miller, M., & Stark, C. E. L. (2008). Pattern Separation in the Human Hippocampal CA3 and Dentate Gyrus. Science, 319(March), 1640–1642.
http://doi.org/10.1126/science.1152882
Brock Kirwan, C., Hartshorn, A., Stark, S. M., Goodrich-Hunsaker, N. J., Hopkins, R. O., & Stark, C. E. L. (2012). Pattern separation deficits following damage to the hippocampus.
Neuropsychologia, 50(10), 2408–2414. http://doi.org/10.1016/j.neuropsychologia.2012.06.011
Curran, T. (2000). Brain potentials of recollection and familiarity. Memory & Cognition, 28(6), 923– 938. http://doi.org/10.3758/BF03209340
Curran, T., & Cleary, A. M. (2003). Using ERPs to dissociate recollection from familiarity in picture recognition. Cognitive Brain Research, 15, 191–205.
http://doi.org/10.1016/S0926-6410(02)00192-1
Curran, T., & Friedman, W. J. (2004). ERP old/new effects at different retention intervals in recency discrimination tasks. Cognitive Brain Research, 18(2), 107–120.
Cycowicz, Y. M., Friedman, D., & Snodgrass, J. G. (2001). Remembering the color of objects: an ERP investigation of source memory. Cerebral Cortex (New York, N.Y. : 1991), 11(4), 322–334. http://doi.org/10.1093/cercor/11.4.322
Das, T., Ivleva, E. I., Wagner, A. D., Stark, C. E. L., & Tamminga, C. a. (2014). Loss of pattern separation performance in schizophrenia suggests dentate gyrus dysfunction. Schizophrenia
Research, 159(1), 193–197. http://doi.org/10.1016/j.schres.2014.05.006
Denny, C. a., Kheirbek, M. A., Alba, E. L., Tanaka, K. F., Brachman, R. a., Laughman, K. B., … Hen, R. (2014). Hippocampal memory traces are differentially modulated by experience, time, and adult neurogenesis. Neuron, 83(1), 189–201. http://doi.org/10.1016/j.neuron.2014.05.018 Déry, N., Pilgrim, M., Gibala, M., Gillen, J., Martin Wojtowicz, J., MacQueen, G., & Becker, S.
(2013). Adult hippocampal neurogenesis reduces memory interference in humans: Opposing effects of aerobic exercise and depression. Frontiers in Neuroscience, 7(April), 1–15.
http://doi.org/10.3389/fnins.2013.00066
Duarte, A., Ranganath, C., Winward, L., Hayward, D., & Knight, R. T. (2004). Dissociable neural correlates for familiarity and recollection during the encoding and retrieval of pictures. Cognitive
Brain Research, 18(3), 255–272. http://doi.org/10.1016/j.cogbrainres.2003.10.010
Duff, M. C., Warren, D. E., Gupta, R., Vidal, J. P. B., Tranel, D., & Cohen, N. J. (2012). Teasing apart tangrams: Testing hippocampal pattern separation with a collaborative referencing paradigm. Hippocampus, 22, 1087–1091. http://doi.org/10.1002/hipo.20967
Ecker, U. K. H., Zimmer, H. D., Groh-Bordin, C., & Mecklinger, A. (2007). Context effects on familiarity are familiarity effects of context - An electrophysiological study. International
Eichenbaum, H., & Cohen, N. J. (2014). Can We Reconcile the Declarative Memory and Spatial Navigation Views on Hippocampal Function? Neuron, 83(4), 764–770.
http://doi.org/10.1016/j.neuron.2014.07.032
Friedman, D., & Johnson, R. (2000). Event-related potential (ERP) studies of memory encoding and retrieval. Microscopy Research and Technique, 51(January), 6–28. Retrieved from
http://onlinelibrary.wiley.com/doi/10.1002/1097-0029(20001001)51:1<6::AID-JEMT2>3.0.CO;2-R/pdf
Fujii, T., Saito, D. N., Yanaka, H. T., Kosaka, H., & Okazawa, H. (2014). Depressive mood
modulates the anterior lateral CA1 and DG/CA3 during a pattern separation task in cognitively intact individuals: A functional MRI study. Hippocampus, 24(October 2013), 214–224.
http://doi.org/10.1002/hipo.22216
Gruber, T., Tsivilis, D., Giabbiconi, C.-M., & Müller, M. M. (2008). Induced electroencephalogram oscillations during source memory: familiarity is reflected in the gamma band, recollection in the theta band. Journal of Cognitive Neuroscience, 20, 1043–1053.
http://doi.org/10.1162/jocn.2008.20068
Holden, H. M., & Gilbert, P. E. (2012). Less efficient pattern separation may contribute to age-related spatial memory deficits. Frontiers in Aging Neuroscience, 4(May), 1–6.
http://doi.org/10.3389/fnagi.2012.00009
Holden, H. M., Hoebel, C., Loftis, K., & Gilbert, P. E. (2012). Spatial pattern separation in cognitively normal young and older adults. Hippocampus, 22, 1826–1832.
Holden, H. M., Toner, C., Pirogovsky, E., Kirwan, C. B., & Gilbert, P. E. (2013). Visual object pattern separation varies in older adults. Learning & Memory, 20(7), 358–362.
http://doi.org/10.1101/lm.030171.112
Howard, M., & Eichenbaum, H. (2014). Time and space in the hippocampus. Brain Research, 1–14. http://doi.org/10.1016/j.brainres.2014.10.069
Hull, A. M. (2002). Neuroimaging findings in post-traumatic stress disorder. Systematic review. The
British Journal of Psychiatry : The Journal of Mental Science, 181(August), 102–110.
http://doi.org/10.1192/bjp.181.2.102
Hunsaker, M. R., & Kesner, R. P. (2013). The operation of pattern separation and pattern completion processes associated with different attributes or domains of memory. Neuroscience and
Biobehavioral Reviews, 37(1), 36–58. http://doi.org/10.1016/j.neubiorev.2012.09.014
Janowsky, J. S., Shimamura, A. P., & Squire, L. R. (1989). Source memory impairment in patients with frontal lobe lesions. Neuropsychologia, 27(8), 1043–1056. http://doi.org/10.1016/0028-3932(89)90184-X
Jasper, H. H. (1958). The ten twenty electrode system of the international federation.
Electroencephalography and Clinical Neurophysiology, 10, 371–375. Retrieved from
http://ci.nii.ac.jp/naid/10020218106/en/
Johansson, M., & Mecklinger, A. (2003). The late posterior negativity in ERP studies of episodic memory: Action monitoring and retrieval of attribute conjunctions. Biological Psychology. http://doi.org/10.1016/S0301-0511(03)00104-2
Johansson, M., Mecklinger, A., & Treese, A.-C. (2004). Recognition memory for emotional and neutral faces: an event-related potential study. In Journal of cognitive neuroscience (Vol. 16, pp. 1840–1853). http://doi.org/10.1162/0898929042947883
Kesner, R. P. (2013). Role of the hippocampus in mediating interference as measured by pattern separation processes. Behavioural Processes, 93, 148–154.
http://doi.org/10.1016/j.beproc.2012.09.018
Kesner, R. P., Gilbert, P. E., & Wallenstein, G. V. (2000). Testing neural network models of memory with behavioral experiments. Current Opinion in Neurobiology, 10, 260–265.
http://doi.org/10.1016/S0959-4388(00)00067-2
Kheirbek, M. A, Klemenhagen, K. C., Sahay, A., & Hen, R. (2012). Neurogenesis and generalization: a new approach to stratify and treat anxiety disorders. Nature Neuroscience, 15(12), 1613–20. http://doi.org/10.1038/nn.3262
Kim, J., & Yassa, M. A. (2013). Assessing recollection and familiarity of similar lures in a behavioral pattern separation task. Hippocampus, 23(4), 287–294. http://doi.org/10.1002/hipo.22087
Kirwan, C. B., & Stark, C. E. L. (2007). Overcoming interference: an fMRI investigation of pattern separation in the medial temporal lobe. Learning & Memory (Cold Spring Harbor, N.Y.), 14, 625–633. http://doi.org/10.1101/lm.663507
Lacy, J. W., Yassa, M. A, Stark, S. M., Muftuler, L. T., & Stark, C. E. L. (2011). Distinct pattern separation related transfer functions in human CA3/dentate and CA1 revealed using high-resolution fMRI and variable mnemonic similarity. Learning & Memory (Cold Spring Harbor,
N.Y.), 18(Ward 2001), 15–18. http://doi.org/10.1101/lm.1971111
Leutgeb, J. K., Leutgeb, S., Moser, M.-B., & Moser, E. I. (2007). Pattern separation in the dentate gyrus and CA3 of the hippocampus. Science (New York, N.Y.), 315(5814), 961–966.
Leutgeb, S., & Leutgeb, J. K. (2007). Pattern separation, pattern completion, and new neuronal codes within a continuous CA3 map. Learning & Memory (Cold Spring Harbor, N.Y.), 14, 745–757. http://doi.org/10.1101/lm.703907
Levy-Gigi, E., Szabo, C., Richter-Levin, G., & Kéri, S. (2015). Reduced hippocampal volume is associated with overgeneralization of negative context in individuals with PTSD.
Neuropsychology, 29(1), 151–161. http://doi.org/10.1037/neu0000131
Lopez-Calderon, J., & Luck, S. J. (2014). ERPLAB: an open-source toolbox for the analysis of event-related potentials. Frontiers in Human Neuroscience, 8(April), 1–14.
http://doi.org/10.3389/fnhum.2014.00213
Mandler, G. (1980). Recognizing: The judgment of previous occurrence. Psychological Review,
87(3), 252–271. http://doi.org/10.1037/0033-295X.87.3.252
Paleja, M., Girard, T. a., Herdman, K. a., & Christensen, B. K. (2014). Two distinct neural networks functionally connected to the human hippocampus during pattern separation tasks. Brain and
Cognition, 92, 101–111. http://doi.org/10.1016/j.bandc.2014.10.009
Rugg, M. D., & Curran, T. (2007). Event-related potentials and recognition memory. Trends in
Cognitive Sciences, 11(6), 251–257. http://doi.org/10.1016/j.tics.2007.04.004
Rugg, M. D., & Yonelinas, A. P. (2003). Human recognition memory: A cognitive neuroscience perspective. Trends in Cognitive Sciences. http://doi.org/10.1016/S1364-6613(03)00131-1 Treves, A., & Rolls, E. t. (1994). Computational analysis of the role of the hippocampus in memory.
Hippocampus, 4(3), 374–391. http://doi.org/10.1002/hipo.450040319
Tsivilis, D., Otten, L. J., & Rugg, M. D. (2001). Context effects on the neural correlates of recognition memory: an electrophysiological study. Neuron, 31(3), 497–505. http://doi.org/10.1016/S0896-6273(01)00376-2
Tulving, E. (1985). Memory and consciousness. Canadian Psychology/Psychologie Canadienne,
26(1), 1–12. http://doi.org/10.1037/h0080017
Van Hagen, B. T. J., van Goethem, N. P., Lagatta, D. C., & Prickaerts, J. (2015). The object pattern separation (OPS) task: A behavioral paradigm derived from the object recognition task.
Behavioural Brain Research, 285, 44–52. http://doi.org/10.1016/j.bbr.2014.10.041
Vazdarjanova, A., & Guzowski, J. F. (2004). Differences in hippocampal neuronal population responses to modifications of an environmental context: evidence for distinct, yet
complementary, functions of CA3 and CA1 ensembles. The Journal of Neuroscience : The
Official Journal of the Society for Neuroscience, 24(29), 6489–6496.
http://doi.org/10.1523/JNEUROSCI.0350-04.2004
Wilding, E. L. (2000). In what way does the parietal ERP old/new effect index recollection?
International Journal of Psychophysiology, 35(1), 81–87.
http://doi.org/10.1016/S0167-8760(99)00095-1
Wixted, J. T. (2009). Remember/Know judgments in cognitive neuroscience: An illustration of the underrepresented point of view. Learning & Memory (Cold Spring Harbor, N.Y.), 16(7), 406– 412. http://doi.org/10.1101/lm.1312809
Yassa, M. A., Lacy, J. W., Stark, S. M., Albert, M. S., Gallagher, M., & Stark, C. E. L. (2011). Pattern separation deficits associated with increased hippocampal CA3 and dentate gyrus activity in nondemented older adults. Hippocampus, 21, 968–979. http://doi.org/10.1002/hipo.20808
Appendix
A1 : Repeated measures ANOVA for six regions analysis of the frontal old/new effect.
A2 : Repeated measures ANONA for six regions analysis of the parietal old/new effect.
A3: Repeated measures ANONA for six regions analysis of the third late time window.
DF F p
Picture Type 3, 72 1.52 0.22
Picture Type x AP 3, 72 4.76 0.01
Picture Type x Hemisphere 3.7, 88.1 1.68 0.67
Picture Type x AP x Hemisphere 6, 144 1.14 0.34
DF F p
Picture Type 3, 72 0.85 0.47
Picture Type x AP 3, 72 1.81 0.15
Picture Type x Hemisphere 6, 144 1.12 0.35
Picture Type x AP x Hemisphere 3.7, 89.8 1.47 0.222
DF F p
Picture Type 3, 78 7.53 < 0.001
Picture Type x AP 3, 78 7.21 < 0.001
Picture Type x Hemisphere 3.7, 97.3 1.97 0.16
