Encoding- and retrieval-related brain activity underlying false recognition

We examined the neural activity associated with true and false recognition during both encoding and retrieval using the Remember/Know procedure to separate recollection (i.e., mental reinstatement of experienced events during which unique details of a memory are recalled) and familiarity (i.e., mental awareness that an event has been experienced previously without the unique details of the event) in recognition memory. Neuroimaging data at retrieval revealed that the right parahippocampal gyrus was activated during recollection-based true recognition compared with familiarity-based true recognition, indicating the item-specific retrieval of visual details. This effect in the right parahippocampal gyrus was not observed for false recognition. Contrary to our expectation, the reactivation effect in early visual cortex was not observed during true recognition, as opposed to false recognition. Neuroimaging data at encoding revealed that the right visual cortex (the right occipitotemporal sulcus) was activated during the encoding of items that yielded recollection-based true recognition compared with familiarity-based true recognition, indicating item-specific visual processing. This effect in the right visual cortex was not observed for false recognition. These results suggest that the subjective feeling of Remember/Know with respect to both veridical and false memories varies with the neural activity during both encoding and retrieval.


Introduction
While memory failure is most commonly an inability to retrieve desired information, individuals may also have memories of events that did not occur. Psychologists have examined this false memory phenomenon for decades, and it is now broadly accepted that human memory is prone to distortions and illusions (Loftus, 2003;Roediger, 1996;Schacter, 1999). Memory distortions, as well as normal forgetting, provide valuable opportunities for researchers to scientifically analyze memory processing, which would be extremely difficult if memory were perfect.
Previous neuroimaging studies focused on false memories during retrieval processes, especially false recognition, which is a process whereby individuals incorrectly claim that they have recently seen or heard a stimulus that they have not actually encountered (Underwood, 1965). Several researchers have attempted to interpret the difference in brain activation between true and false recognition using the sensory reactivation hypothesis (Schacter et al., 2011;Schacter and Slotnick, 2004). This sensory reactivation hypothesis states that the retrieval of true (but not false) memories reactivates regions that were active during the encoding of the information because the retrieval of true memories is accompanied by the retrieval of more perceptual details than the retrieval of false memories.
For example, Slotnick and Schacter (2004) used abstract shapes as stimuli in an old-new recognition memory task to confirm that the difference in brain activity between true and false recognition can be explained by the sensory reactivation hypothesis. In this study, during functional magnetic resonance imaging (fMRI), the participants studied exemplar shapes and later made recognition memory decisions in response to studied shapes (i.e., shapes that were identical to those presented at encoding), nonstudied lures (i.e., shapes that were similar, but not identical, to those presented at encoding), and new shapes (i.e., shapes that were not presented at encoding). Slotnick and Schacter (2004) reported that true recognition of previously studied shapes, compared with false recognition of non-studied lures, was accompanied by increased activity of early visual cortex, an activation pattern that could reflect a sensory signature of true memory. Other neuroimaging studies using different paradigms, such as the Deese-Roediger-McDermott (DRM) task (Deese, 1959;Roediger and McDermott, 1995; for review, see Gallo, 2010), have also provided evidence that true recognition necessitates the reactivation of sensory processing areas that were presumably active at encoding (Abe et al., 2008;Schacter et al., 1996). Thus, both of these previous studies reported that true recognition, relative to false recognition, of previously heard words was associated with increased fMRI signals that implicate auditoryverbal processing (i.e., the left temporoparietal region). In addition to the brain regions that are responsible for sensory processing, the involvement of the medial temporal lobe in distinguishing true from false recognition has also been reported. For example, Cabeza et al. (2001) reported that hippocampal activity was commonly associated with both true and false recognition, whereas activity in the parahippocampal gyrus was specific to true recognition.
Not all neuroimaging studies of false memory have focused on the sensory reactivation hypothesis (for a review, see Abe, 2012). Recent neuroimaging studies of false recognition have utilized either the confidence rating (high or low) or the Remember/Know procedure (recollection or familiarity) to further delineate the neural basis of false recognition. For example, Kim and Cabeza (2007b) required participants to read short lists of categorized words and then measured neural activity during the performance of a recognition test. By combining neuroimaging with a confidence rating, Kim and Cabeza (2007b) observed that true recognition with high confidence was associated with increased medial temporal lobe activity, whereas false recognition with high confidence was associated with increased frontoparietal activity. In a recent study by Dennis et al. (2012), the participants were presented with a series of pictures during the encoding phase and were later asked to make a Remember/Know/New judgment in response to target pictures (the same pictures viewed at encoding), related lures (novel pictures similar, but not identical, to those viewed at encoding), or unrelated lures (novel pictures not similar to those viewed at encoding). Dennis et al. (2012) reported that the activity in the hippocampus distinguished between recollection-based true recognition and recollection-based false recognition. These two previous studies indicate that the neural correlates of false recognition can be modulated by subjective feelings on retrieved memory as measured using the confidence rating or the Remember/Know procedure.
Certain previous neuroimaging studies have focused on memory distortion during encoding processes. Encoding-based studies of false memory are designed to identify the patterns of brain activity that predict later memory distortion. For example, Kim and Cabeza (2007a) examined gist-based memory distortion. During scanning, the participants read short lists of categorized words. During a later retrieval phase, the participants made an old/new recognition memory judgment with a confidence rating (for a retrieval-based study, see Kim and Cabeza, 2007b). Encoding-related brain activity was analyzed as a function of whether that activity predicted the subsequent true recognition of the target words or the subsequent false recognition of critical lures. These authors observed that increased activity at encoding in the medial temporal lobe and in the early visual cortex were associated with later true, but not false, recognition.
In another study that focused on memory distortion during encoding processes, Garoff et al. (2005) examined the neural activity that predicts specific memory retrieval for the details of previously encountered objects and a more general memory retrieval for the type of previously encountered objects. During scanning, the participants made size judgments regarding a series of common objects. In a subsequent recognition memory test, the participants were asked to make respective Same, Similar, or New judgments regarding the same objects as viewed at encoding, novel objects that were similar, but not identical, to those viewed at encoding, and novel objects that were dissimilar from those viewed at encoding. Specific recognition was indicated by a Same response to the same object, whereas general and non-specific recognition was indicated by either a Same response to the similar object (false memory) or a Similar response to the same object (partial memory). Garoff et al. (2005) reported that increased activity in the right fusiform gyrus predicted specific recognition memory, indicating the involvement of this region in specific feature encoding. These two studies (Garoff et al., 2005;Kim and Cabeza, 2007a), in addition to other encoding-based studies using the reality monitoring paradigm (Gonsalves et al., 2004;Kensinger and Schacter, 2005) or the post-event misinformation paradigm (Baym and Gonsalves, 2010;Edelson et al., 2011;Okado and Stark, 2005), have suggested that different cognitive and neural processes at encoding are associated with subsequent true and false recognition.
Successful memory requires both successful encoding and retrieval. That is, false memory occurs due to unsuccessful encoding, unsuccessful retrieval, or both. However, given that many retrieval-based and encoding-based neuroimaging experiments have been conducted separately, it is difficult to determine how the encoding and retrieval processes jointly contribute to memory distortion. Although Kim and Cabeza (2007a,b) collected data during both the encoding and retrieval phases, due to study limitations, the findings from the encoding and retrieval phases were based on different numbers of participants. Moreover, during the encoding phase, the participants were presented with multiple stimuli per trial, making it difficult to isolate the brain activations that were associated with true or false memory formation related to a single stimulus.
The present study was designed to determine whether brain activations to true and false memories differed during memory encoding, memory retrieval, or during both memory phases. In the encoding task, the participants were presented with a series of pictures and were asked to identify each stimulus as either "living" or "non-living". In the recognition memory task, the participants were asked to make recognition memory decisions regarding previously presented pictures (Same items), novel pictures that were similar to the previously presented pictures (Similar lures), and novel pictures that were not similar to the previously presented pictures (Dissimilar lures). We used a single-stimulus paradigm as opposed to the typical DRM paradigm (i.e., presentation of multiple stimuli belonging to a specific category for creating false memory) because it is difficult to identify which time point is critical for subsequent false memory phenomena during the encoding of multiple stimuli. The present paradigm allowed us to measure brain activity that was associated with both true and false recognition during the presentation of a single stimulus during both encoding and retrieval. In addition, given the importance of recollection/familiarity distinctions in recognition memory (Diana et al., 2007;Skinner and Fernandes, 2007), we used the Remember/Know procedure to delineate the brain regions associated with the recollection and familiarity of true and false memories. This procedure was included to determine whether the neural correlates of false recognition can be modulated by subjective feelings of remembering and knowing.
Based on previous findings, we predicted that efficient visual processing during memory encoding would be characterized by increased activity in the visual cortices, whereas efficient reactivation during memory retrieval would be characterized by increased activity in the visual cortices and/or the medial temporal lobe, as these areas jointly contribute to the formation and retrieval of veridical memories. We also predicted that the encoding-and retrieval-related brain activity associated with false memory would depend on whether the memory was based on either recollection or familiarity processes.

Participants
Thirty-three right-handed male volunteers with no history of neurological or psychiatric disease were paid to participate in the present study. The examination of the participants' MRIs revealed no brain pathology. To yield stable activation maps, four participants who did not have at least 20 events in every condition used in imaging contrasts were excluded from the analysis. Thus, the results are based on the remaining 29 participants (age range: 18-25 years, mean age: 21.3 years). All of the participants provided written informed consent in accordance with the Declaration of Helsinki and the guidelines approved by the Ethical Committee of Tohoku University.

Stimuli
A total of 480 pairs of namable, color photographs were used as stimuli. A subset of these stimuli has been used in previous studies (Abe et al., 2011;Hashimoto et al., 2012). Each pair of photographs comprised two objects with the same verbal label, which differed in several perceptual aspects, such as color, shape, orientation, or surface pattern. The photographs were divided into 360 pairs that were presented during both the study and test phases and 120 distractor pairs that were only presented during the test phase. The list of 360 pairs was further subdivided into the following two lists: 120 Same item pairs, for which only one of the objects from a pair was used in both the study and test phases, and 240 Similar item pairs, for which one of the objects was used in the study phase, and the other object was used in the test phase. We used more Similar items than Same items to obtain sufficient target events of false memory for comparison in the neuroimaging data analyses. The assignment of these 360 pairs to Same or Similar categories was counterbalanced across participants. One item from each of the remaining 120 Dissimilar pairs was only presented during the test phase.

Tasks
The experiment consisted of three study phase runs followed by three test phase runs. The fMRI data were acquired during both the study and test phases. Fig. 1 illustrates the experimental design.
During the study phase, the participants were presented with a series of 360 stimuli in a pseudorandom order. The participants were asked to identify each stimulus as either "living" or "nonliving" by pressing a button with their right hand index or middle fingers, respectively. Each stimulus was presented for 2 s, and the intervals between the stimuli, during which cross-fixation was continuously presented, ranged from 3 to 10.5 s to maximize the efficiency of the event-related design (Dale, 1999).
After the study phase, all of the participants were escorted to another room and took a 1-h break. The participants were asked to simply take a break and were not asked to engage in any specific cognitive task. After the break, the participants initiated the test phase. During the test phase, 480 stimuli were sequentially presented (120 Same items, 240 Similar items, and 120 Dissimilar items) in a pseudorandom order. For each stimulus, the participants were asked to press one of the three buttons on the response box according to their memory of the item from the study phase. If the participants were able to recall something specific about the stimulus (e.g., what the stimulus made them think about or how the stimulus appeared on the screen), the participants were asked to give a Remember response with their index finger. If they were not able to recall anything specific regarding the previous presentation of the stimulus but felt a familiarity with the item and believed that the item had been presented during the study phase, the participants were asked to give a Know response with their middle finger. If the participants believed that the item had not been presented during the study phase, they were asked to give a New response with their ring finger. The participants were encouraged to respond as rapidly and as accurately as possible. Each test phase stimulus was presented for 2 s followed by a variable-duration fixation time that ranged from 3 to 15.5 s.

Data acquisition and analysis
Whole-brain imaging was performed using a 1.5-Tesla General Electric Signa MRI scanner. A T1-weighted anatomical image of each participant's brain was obtained for coregistration purposes. A T2*-weighted echo planar imaging (EPI) sequence that was sensitive to blood oxygenation level-dependent (BOLD) contrast with the following parameters was used for functional imaging: repetition time (TR) = 2500 ms, echo time (TE) = 30 ms, flip angle = 90 • , acquisition matrix = 64 × 64, field of view = 260 mm, inplane resolution = 4.06 mm × 4.06 mm, and 25 axial slices, with a slice thickness of 4 mm, and an interslice gap of 1 mm. Firm padding was placed around the head of each participant to restrict head motion. The visual stimuli were projected onto a screen and were viewed through a mirror that was attached to a standard head coil. The participants' responses were collected using an MRIcompatible response box. The first four scans in each run were "dummy" scans and were discarded to allow for T1 equilibration.
The data preprocessing and statistical analyses were performed using SPM8 (Wellcome Department of Imaging Neuroscience, London, UK). All of the volumes from each participant were realigned to correct for small movements between scans. This process generated an aligned set of images and a mean image per participant. The realigned images were then corrected for different slice acquisition times. Each participant's T1-weighted structural MRI was coregistered to the mean of the realigned EPI images. The anatomical and functional images were then spatially normalized to a custom template generated from all of the participants' anatomical images using diffeomorphic anatomical registration through an exponentiated lie algebra (DARTEL) algorithm (Ashburner, 2007) and then affine-transformed into the Montreal Neurological Institute (MNI) stereotactic space. The functional images were resampled into 2 mm × 2 mm × 2 mm voxels and spatially smoothed with an 8-mm full-width half-maximum (FWHM) Gaussian kernel.
The fMRI data were analyzed using an event-related model. For each participant, the activity associated with each experimental condition was modeled using a canonical hemodynamic response function. We used subsequent memory analysis for the encoding phase data. Each trial was assigned to a different analysis condition based on the participant's button response for that item. During the encoding phase, the two types of stimuli (Same or Similar items) and the three types of responses (Remember, Know, or New) created a total of six conditions. During the retrieval phase, a total of nine conditions were constructed by combining the three types of stimuli (Same, Similar, or Dissimilar items) and the three types of responses (Remember, Know, or New). Targets for which an incorrect living/non-living response or no response was made were omitted from the analysis. A high-pass filter of 1/128 Hz was used to remove low-frequency noise, and an AR (1) model was employed to correct for temporal autocorrelation.
As we were interested in the different neural mechanisms that underlie true and false recognition, we analyzed the encodingand retrieval-related brain activity resulting from the following four conditions: (a) recollection-based true recognition (Remember During the study phase with the fMRI scanning, the participants were asked to judge whether each photograph represented a living or a non-living thing. (B) During the test phase with fMRI scanning, the participants were asked to make Remember/Know/New judgments in response to each photograph. The aim of this study was to analyze four conditions for both the encoding and retrieval phases: recollection-based true recognition (Remember response to Same items), familiarity-based true recognition (Know response to Same items), recollection-based false recognition (Remember response to Similar items), and familiarity-based false recognition (Know response to Similar items).
response to a Same item), (b) familiarity-based true recognition (Know response to a Same item), (c) recollection-based false recognition (Remember response to a Similar item), and (d) familiarity-based false recognition (Know response to a Similar item). A second-level analysis of the parameter estimates for the voxels in each of these four conditions was then computed as a repeated-measures analysis of variance (ANOVA). In this analysis, each participant was treated as a random effect. We created a three-way 2 × 2 × 2 ANOVA model with the task type (encoding or retrieval), the stimulus type (Same or Similar), and the response type (Remember or Know) as factors. Appropriate corrections were made for non-sphericity and repeated measures (Friston et al., 2002a,b). Comparisons between the conditions were performed using appropriately weighted linear contrasts and were determined on a voxel-by-voxel basis. For the encoding and retrieval data sets, we created two separate 2 × 2 ANOVA models with the stimulus type (Same or Similar) and the response type (Remember or Know) as factors. This procedure allowed for us to perform Ftests to identify interactions between the two factors across brain regions during the encoding and retrieval phases. For the wholebrain analyses, the threshold of significance was set at p < 0.001 (uncorrected for multiple comparisons), with an extent threshold of 10 contiguous voxels. The peak voxels of clusters that exhibited reliable effects are reported in MNI coordinates.
We used the MarsBaR software to extract the percentage signal changes in activated regions and to perform a region-of-interest (ROI) analysis (Brett et al., 2002). Each ROI was defined based on the results of the SPM whole-brain analyses (i.e., functional ROI), and signal changes were averaged across all of the voxels in a given cluster. A repeated-measures two-way ANOVA with Bonferroni's pairwise comparisons (p < 0.05/4) was used to analyze the differences in percentage signal changes in each ROI.

Behavioral data
The proportion and reaction times of the participants' responses to Same, Similar, and Dissimilar items are shown in Fig. 2. The data for recollection-based true recognition (Remember response to Same item), familiarity-based true recognition (Know response to Same item), recollection-based false recognition (Remember response to Similar item), and familiarity-based false recognition (Know response to Similar item) were analyzed using a repeatedmeasures two-way ANOVA with the stimulus type (Same or Similar) and response type (Remember or Know) as factors.
With regard to the proportion of responses at retrieval, we observed no significant main effect of the response type (F[1,28] = 0.119, p = 0.732); however, we observed a significant main effect of the stimulus type (F[1,28] = 111.989, p < 0.001) and a significant interaction between the two factors (F[1,28] = 93.338, p < 0.001). Post hoc t-tests with Bonferroni's correction (p < 0.05/4) revealed that the proportion of Remember responses was higher than the Know responses to the Same items (t[28] = 4.120, p < 0.001), whereas the proportion of Remember responses was lower than Know responses to Similar items (t[28] = −5.056, p < 0.001). The proportion of Remember responses to Same items was higher than that for the Similar items (t[28] = 12.634, p < 0.001), whereas the proportion of Know responses to the Same items was lower than that for the Similar items (t[28] = −3.534, p < 0.005).
With regard to the reaction time data at retrieval, we observed no significant main effect of the stimulus type (F[1,28] = 2.390, p = 0.133); however, we observed a significant main effect of the response type (F[1,28] = 98.374, p < 0.001) and a significant interaction between the two factors (F[1,28] = 10.797, p < 0.005). Post hoc t-tests with Bonferroni's correction (p < 0.05/4) revealed that the reaction time of Remember responses was shorter than the Know responses to the Same (t[28] = 9.085, p < 0.001) and Similar items (t[28] = 9.345, p < 0.001). The reaction time for Remember responses to Same items was shorter than for the Similar items (t[28] = 4.196, p < 0.001), and there was no difference in the reaction times for the Know responses to the Same items and the Know responses to the Similar items (t[28] = −1.081, p = 0.289).
We also analyzed the mean reaction time during the encoding phase for items that yielded recollection-based true recognition, familiarity-based true recognition, recollection-based false recognition, and familiarity-based false recognition. We observed no significant main effects of the stimulus type (F[1,28] = 0.282, p = 0.600) or the response type (F[1,28] = 0.735, p = 0.399), nor was any interaction observed between the two factors (F[1,28] = 2.569, p = 0.120).

Brain activation at retrieval
The results of the retrieval phase SPM whole-brain analyses are summarized in Table 1. We first compared neural activity during true recognition (i.e., recollection-based and familiarity-based true recognition) versus false recognition (i.e., recollection-based and familiarity-based false recognition). We observed significant activations in the left insula, the left and the right thalamus, the right insula (subcortical area), the right postcentral gyrus (subcortical area), and in the right fusiform gyrus (subcortical area). However, no activations were observed in the early visual cortex or in the medial temporal lobe. The opposite comparison showed no significant activations.
We then compared neural activity during the retrieval of items associated with Remember responses to those associated with Know responses. We observed significant activations in the left superior frontal gyrus, the left medial frontal cortex, the left inferior frontal gyrus, the left posterior cingulate gyrus/precuneus, the left angular and middle occipital gyrus, the left and the right anterior cingulate gyrus, the right middle frontal gyrus, the right inferior frontal gyrus, and the right postcentral gyrus. The opposite comparison indicated significant activations in the left supplementary motor area, the left superior frontal gyrus, the right superior frontal gyrus, the right superior temporal gyrus (subcortical white matter), and in the right middle temporal gyrus. The patterns of neural activity that were observed in the regions identified by the main effect of response type are not discussed because this comparison is not directly relevant to examining the neural correlates of false memory; we were primarily interested in the regions that exhibited different activation in true and false recognition and in regions that exhibited an interaction between these two factors (i.e., stimulus type and response type).
We next explored brain regions that exhibited a significant interaction between the two factors using the F-contrast in a twoway ANOVA in SPM. This analysis revealed that the left pallidum, the left putamen, the left caudate nucleus, the left cerebellum, the left supramarginal gyrus (cortex and subcortical area), the right anterior cingulate gyrus (subcortical area), the right putamen, and most importantly, the right parahippocampal gyrus, exhibited significant interaction effects (Fig. 3). To clarify the nature of the interaction that was observed in the right parahippocampal gyrus, we performed a ROI analysis. The results of the ROI analysis confirmed a significant interaction (F[1,28] = 13.576, p < 0.001) without a significant main effects of stimulus type (F[1,28] = 0.127, p = 0.725) or response type (F[1,28] = 0.184, p = 0.671). Post hoc t-tests with Bonferroni's correction (p < 0.05/4) revealed that there were significant differences in neural activity between recollection-based true recognition and familiarity-based true recognition (t[28] = 2.679, p = 0.012) and between familiarity-based true recognition and familiarity-based false recognition (t[28] = −3.677, p < 0.001). There were marginally significant differences between recollectionbased true recognition and recollection-based false recognition (t[28] = 2.529, p = 0.017) and between recollection-based false recognition and familiarity-based false recognition (t[28] = −2.493, p = 0.019).

Brain activation at encoding
The results of the SPM whole-brain analyses of the encoding phase imaging data are summarized in Table 2. We first compared neural activity during the encoding of items that were correctly retrieved (i.e., recollection-and familiarity-based true recognition) versus those that were falsely retrieved (i.e., recollection-and familiarity-based false recognition). We observed significant activations in the right supplementary motor area, the right precentral gyrus (subcortical area), the right middle temporal gyrus (subcortical white matter), and the right angular gyrus. The opposite comparison revealed no significant activations. Although the activation of the angular gyrus was not predicted, in the light of the importance of this region in the memory literature (for review, see Uncapher and Wagner, 2009), the pattern of neural activity in this region is illustrated and discussed below (Fig. 4A). RT, recollection-based true recognition; FT, familiarity-based true recognition; RF, recollection-based false recognition; FF, familiarity-based false recognition.  . 3. Retrieval-related activity in the right parahippocampal gyrus. The activity in the right parahippocampal gyrus was significantly increased during recollection-based true recognition relative to familiarity-based true recognition; however, this effect was not observed when comparing recollection-based false recognition with familiarity-based false recognition. The error bars represent the standard error of the mean.
We then compared neural activity during the encoding of items that yielded Remember responses versus those that yielded Know responses. We observed significant activations in the left inferior frontal gyrus, the right amygdala, and in the bilateral inferior temporal gyri. Using the opposite comparison, we observed significant activations in the right middle frontal gyrus, the right superior frontal gyrus, and in the bilateral posterior parietal cortices. Patterns of neural activity in the regions identified in the main effect of response type are not discussed because this comparison is not directly relevant to examining the neural correlates of false memory; we were primarily interested in the regions that exhibited different brain activity during true and false recognition and in the regions that exhibited interactions between the stimulus type and response type factors.
The regions that exhibited a significant interaction between the two factors were explored using F-contrast in a two-way ANOVA in SPM. This analysis revealed that the right occipitotemporal sulcus exhibited a significant interaction between the stimulus type and response type (Fig. 4B). We performed ROI analysis to clarify the nature of this interaction. The results of the ROI analysis confirmed a significant interaction (F[1,28] = 15.170, p < 0.001) and revealed a significant main effect of response type (F[1,28] = 13.063, p < 0.005) without a main effect of stimulus type (F[1,28] = 1.029, p = 0.319). Post hoc t-tests with Bonferroni's correction (p < 0.05/4) revealed that there were significant differences in neural activity between recollection-based true recognition and familiarity-based true recognition (t[28] = 4.615, p < 0.001) and between familiaritybased true recognition and familiarity-based false recognition (t[28] = −3.678, p < 0.001). No significant difference was observed Fig. 4. Encoding-related activity in (A) the right angular gyrus and (B) the right occipitotemporal sulcus. The activity in the right angular gyrus was increased for items that were correctly retrieved (i.e., both recollection-and familiarity-based true recognition) relative to those that were falsely retrieved (i.e., both recollection-and familiaritybased false recognition). The activity in the right occipitotemporal sulcus was increased for items that were associated with recollection-based true recognition relative to familiarity-based true recognition; however, this effect was not observed when comparing the items that were associated with recollection-based false recognition with those that were associated with familiarity-based false recognition. The error bars represent the standard error of the mean. between recollection-based true recognition and recollectionbased false recognition (t[28] = 2.154, p = 0.040) or between recollection-based false recognition and familiarity-based false recognition (t[28] = −0.118, p = 0.907).

Common brain activation at retrieval and encoding
The results of the whole-brain SPM analyses are summarized in Table 3. Extrapolating across the two memory phases, we calculated the main effect of true memory (i.e., recollection-based true recognition + familiarity-based true recognition vs. recollectionbased false recognition + familiarity-based false recognition). We observed significant activations in the left caudate nucleus, left thalamus, right middle frontal gyrus, right postcentral gyrus (subcortical area), and right fusiform gyrus (subcortical area); however, as in the retrieval phase, no activations were observed in the early visual cortex or in the medial temporal lobe.

Discussion
We used an event-related fMRI design to examine the neural activity associated with true and false recognition during both encoding and retrieval. We also used the Remember/Know procedure to identify brain regions associated with recollectionand familiarity-based true recognition and recollection-and familiarity-based false recognition. During retrieval, increased activity in the right parahippocampal gyrus was associated with recollection-based true recognition relative to familiarity-based true recognition. In contrast, this difference was not observed between recollection-based false recognition and familiarity-based false recognition. Contrary to our expectation, the reactivation effect in early visual cortex was not observed during true recognition, as opposed to false recognition. During encoding, the right visual cortex (specifically, the right occipitotemporal sulcus) was activated during the encoding of items that yielded recollection-based true recognition compared with familiaritybased true recognition. This effect in the right visual cortex was not observed between items that yielded recollection-based and familiarity-based false recognition. The present results indicate that the subjective feeling of Remember/Know with respect to both veridical and false memories varies with the neural activity during both encoding and retrieval.
In the present study, we expected to find that the visual cortex would be differentially activated between true and false recognition during the retrieval phase. However, we did not detect the activation in the early visual cortex. In addition, the main effect of true versus false memory across encoding and retrieval phases did not reveal an effect described by the sensory reactivation hypothesis. A possible explanation for this null finding is that the conditions in the present study did not encourage the encoding of sufficient sensory information to increase differences in brain activations between true and false recognition (see Cabeza et al., 2001). For example, if the participants were asked to judge certain perceptual aspects rather than the semantic aspects of the presented stimuli during encoding, the sensory signature of the stimuli may be stored in the early visual cortex. Our results raise the possibility that, at least in the present experimental paradigm, the reactivation of sensory brain regions does not necessarily distinguish between true and false recognition (see also Kahn et al., 2004).
We observed an interaction effect in the parahippocampal gyrus whereby increased parahippocampal activity during recollection-based true recognition was greater than that observed in familiarity-based true recognition, which is thought to be a neural correlate of item-specific recollection. However, we did not observe increased parahippocampal activity during the retrieval of false memories. Our results suggest that this effect is specific to recollection-based true recognition. The increased medial temporal-lobe activity for recollection-based true recognition relative to familiarity-based true recognition is consistent with previous neuroimaging studies that used the Remember/Know procedure (Eldridge et al., 2000;Johnson and Rugg, 2007;Vilberg and Rugg, 2007;Woodruff et al., 2005). This result is also consistent with a recent neuropsychological study of patients with Alzheimer's disease (AD). AD causes severe atrophy of the medial temporal lobe, including the parahippocampal gyrus. Patients with AD exhibit difficulty in using item-specific recollection processes to reduce false memories during retrieval (Abe et al., 2011).
In a recent study by Dennis et al. (2012), greater activation of the parahippocampal gyrus was observed during recollection-based true recognition compared with familiarity-based true recognition; these previous results are in strong agreement with those of the present study. However, these authors also observed greater activation of the parahippocampal gyrus during recollection-based false recognition compared with familiarity-based false recognition, which is inconsistent with the results of the present study. One possible reason for these contrasting results are differences in the experimental paradigms between the present study and Dennis et al. (2012), such as the number of stimuli and the interval between encoding and retrieval. Further studies are required to determine whether the present results can be replicated.
To the best of our knowledge, no neuroimaging study has examined the patterns of encoding-related brain activity that is associated with recollection-based and familiarity-based false recognition (for a retrieval-based study, see Dennis et al., 2012). In the present study, we found a notable interaction effect in the right occipitotemporal sulcus. The activity of the right occipitotemporal sulcus that was observed during the encoding phase was specific to the items that were subsequently correctly remembered (i.e., recollection-based true recognition) compared with the items that were later reported as familiar (i.e., familiarity-based true recognition). This activity difference was not observed for items that yielded later false recognition. The increased activity in this region is consistent with our prediction and may reflect the item-specific visual processing of perceptual details during encoding.
The occipitotemporal sulcus is adjacent to the fusiform gyrus, which is a region that is essential for object perception in the ventral form-processing stream (Haxby et al., 2001). Using a similar set of Same, Similar, and Dissimilar stimuli, Garoff et al. (2005) reported that increased activity in the right fusiform gyrus predicts specific recognition memory compared with general recognition memory and argued that this region is associated with the encoding of specific stimulus features. Taken together, the results of these previous studies and the present study indicate that activity in the higherorder visual cortices is essential for the formation of vivid memories that promote subsequent recollection-based true recognition. Inefficient activity of these brain regions in the encoding process may be responsible for false memory at retrieval.
Although not predicted, we observed a significant difference in the activity of the right posterior inferior parietal cortex (i.e., the angular gyrus) during the encoding of correctly recognized items versus falsely recognized items. One possible explanation is that the right posterior parietal cortex participates in enhancing attention during encoding (Corbetta et al., 2008;Raz and Buhle, 2006; for a review, see Kim, 2011), thereby enabling the subsequent retrieval of veridical memory. However, our results indicate parietal "deactivation" in all of the four experimental conditions, which is non-specific to the task. Given that this region is associated with the default mode network (Watanabe, 2011), where task-induced deactivation is observed, negative BOLD responses might reflect the degree of attentional resources directed toward the task, not the cognitive processing associated with memory formation. In a meta-analysis, Uncapher and Wagner (2009) reported that the results of memory studies regarding the inferior parietal lobule activity have been mixed. Certain studies have reported negative relationships between brain activity in this area and subsequent memory effects (e.g., Daselaar et al., 2004;, whereas other studies have reported positive effects (e.g., Buckner et al., 2001;Cansino et al., 2002;Sommer et al., 2006). Future research will be required to determine the critical factors, including the form of retrieval, such as recollection-based or familiarity-based recognition, for parietal encoding activity in episodic memory.
It is necessary to mention the limitations of the present study. First, the differences between the neural correlates of recollectionand familiarity-based recognition in true and false memories are derived from the subjective measurement of the Remember/Know distinction. In future studies, it would be informative to apply more objective measures. For example, the source-monitoring paradigm, in which participants are required to judge in which the encoding contexts (sources) of the stimulus was studied (Johnson et al., 1993), would be suitable to investigate recollection-and familiarity-based false recognition. Second, we recruited only male participants to avoid potentially confounding gender differences; this approach may prevent us from generalizing our results. Third, at the time of encoding, the participants had no knowledge of whether they would be tested using Same items or Similar lures. Even if a participant exhibited recollection-or familiarity-based false recognition to a Similar lure at retrieval, this participant may still be able to correctly recognize the item if they were simultaneously presented with a Same item (see Guerin et al., 2012). Finally, in the present study, the Same items predominantly elicited Remember responses, whereas Similar lures elicited more Know responses, raising the possibility that this bias affected the present results. Therefore, although the results of the present study represent an important step toward clarifying the neural mechanisms that underlie the phenomenon of false memory, further studies are required before firm conclusions can be drawn.

Conclusions
The present study revealed that different brain activations are associated with true and false memories during both encoding and retrieval. At retrieval, the parahippocampal activity was associated with recollection-based recognition (relative to familiarity-based recognition) only for true memory, indicating the item-specific retrieval of visual details. At encoding, item-specific visual processing in the higher-order visual cortex is thought to be crucial for subsequent recollection-based true recognition. The present findings underscore the importance of the recollection/familiarity distinction in recognition memory for future studies of neural correlates of false memory; the subjective feeling of Remember/Know with respect to both veridical and false memories varies with neural activity during both encoding and retrieval.