- Email: [email protected]
The interfering effect of related events on recognition memory discriminability: a functional magnetic resonance imaging study
Cognitive Brain Research 22 (2005) 429 – 437 www.elsevier.com/locate/cogbrainres
Research report
The interfering effect of related events on recognition memory discriminability: a functional magnetic resonance imaging study Martin Lepage*, Franc¸ois Blondin, Ame´lie M. Achim, Matthew Menear, Mathieu Brodeur Brain Imaging Group, Douglas Hospital Research Centre, McGill University, 6875 Boul. LaSalle Verdun, Que´bec, Canada H4H 1R3 Accepted 1 October 2004 Available online 11 November 2004
Abstract Retrieval of information from memory often involves the selection of an event among competing related events, a process that frequently gives rise to interference effects. The present study used a forced-choice recognition test to identify neural correlates of the interfering effect of related events on recognition memory discriminability. Participants encoded landscape pictures divided into three segments. One segment was presented during encoding, and a forced-choice recognition task contrasted a studied and a nonstudied segment for each landscape. For half of the landscapes, the third segment was presented between encoding and recognition tasks to induce associative interference by reducing recognition discriminability. A behavioral study with 40 subjects yielded a significant difference in the correct recognition rate between control and interference trials (76% and 64%, respectively, pb0.001). A subsequent event-related fMRI study with 16 subjects yielded significant activations for correct interference recognition trials relative to control trials in left superior parietal regions, which suggests that these regions play a role in the representation of stimuli and associated information. The opposite contrast yielded significant activations in inferior prefrontal regions bilaterally, right dorsolateral prefrontal cortex and right parahippocampal cortex. Since this contrast was conducted using only correctly recognized trials, these findings could reflect an index of memory discriminability or saliency which could influence conscious recollection. D 2004 Elsevier B.V. All rights reserved. Theme: Neural basis of behavior Topic: Cognition Keywords: Prefrontal cortex; Episodic memory; Parietal cortex; Recognition memory
1. Introduction Retrieval of information from episodic memory often involves the selection of an event among competing related events, a process that frequently gives rise to interference effects. Although research on memory interference effects has a long history [1,20], little is known about its neural basis in humans. Some of the evidence that has been gathered so far points to an involvement of the frontal lobes, as it has been shown that patients with lesions in this area
* Corresponding author. Fax: +1 514 888 4064. E-mail address: [email protected] (M. Lepage). 0926-6410/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.cogbrainres.2004.10.003
demonstrate faulty control of interference during memory tasks [17,25,26]. This notion has been further supported by several functional neuroimaging studies carried out with healthy subjects that have found the prefrontal cortex to be a potential neural correlate of the control of memory interference. Of these studies, two used Positron Emission Tomography (PET) to examine interference during memory encoding [10,14]. In both studies, subjects learned word pairs and then had to learn new associations between already studied words. Increased left prefrontal cortex activation was observed when learning associations, suggesting that this activation could reflect the neural correlate of a form of proactive interference. In a SPECT study by Uhl et al. [30], memory interference during retrieval was
430
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
examined using an AB–AD memory interference cuedrecall task, a form of paired-associate learning task where a cue word (A) is successively paired with two different target words (B, D). A right prefrontal anterior cortical region exhibited greater activity during the retrieval of the AD (interference prone) list relative to the AB (interference free) list. Recently, Henson et al. [16] developed an ingenious fMRI experiment that allowed them to study AB–AD paired-associate recall and to record overt verbal responses while minimizing fMRI signal distortion caused by speechrelated head motion. In this study, neural correlates of the encoding and retrieval components of an AB–AD paired associate cued-recall paradigm were examined. A region in the left inferior frontal cortex and a bilateral frontopolar region showed greater activity for memory interference trials relative to control trials during both encoding and retrieval. These results confirm the important role of the prefrontal cortex in the control of memory interference for both encoding and retrieval. It is well established in cognitive psychology that the interference effects on memory performance that are usually observed on the paired-recall tasks employed by investigators such as Uhl and Henson occur rarely, if ever, when examining recognition memory [1]. One exception is the modified recognition memory test proposed by Chandler and Gargano [6,7] in which the amount of retrieval interference can be manipulated across experimental conditions. During a study episode, nature pictures (targets) are followed by either related pictures (interference items) or by unrelated pictures (control items). A forced-choice recognition test shortly follows in which the studied items are presented concurrently with a new distractor and subjects must indicate which of the two pictures had been presented earlier. In this paradigm, subjects usually remember fewer items when their related counterparts have also been presented, suggesting that related representations can produce retrieval interference. In addition, since this interfering effect is time limited and is not observed following a long (48 h) delay, the original trace could not have been binterferedQ with by the related events, thus pointing to retrieval as the memory stage where interference occurs. The vast majority of the functional neuroimaging studies that have examined memory retrieval, and in particular those that were conducted using fMRI, have used a recognition memory test format. Thus, Chandler’s modified recognition memory test appears well-suited to bridge the gap between the literature on the neural correlates of recognition memory and memory retrieval interference. We therefore used the modified forced-choice recognition memory test developed by Chandler [6] to identify neural correlates of recognition memory interference in an eventrelated fMRI study. Based on the previous findings from the functional neuroimaging studies of memory interference using recall tasks, we hypothesized that anterior prefrontal cortex regions would exhibit greater activity during inter-
ference trials relative to control trials. We first conducted a behavioral study to assess whether our own version of the modified recognition memory test would replicate the interference effect observed in Chandler’s studies. The versions differed in two main ways: (i) we presented subjects with a computerized version of the task, and (ii) we used a different set of stimuli. Following the behavioral study, we performed a functional neuroimaging study using fMRI. Because these two studies share similar behavioral methods, they are described together.
2. Methods 2.1. Subjects Forty subjects (24 females, mean age 33.2, range 19–50 years) participated in the behavioral study and were tested on our adapted version of Chandler’s modified recognition memory test. None had a current or previous history of neurological or psychiatric disorders and all but one of the participants were right-handed. Following the completion of this study, 16 new subjects (8 females; mean age 25.4 range 22–29 years) who were all right-handed according to the Edinburgh Inventory [21] participated in the fMRI study. Once again, none of the participants had a current or previous history of neurological or psychiatric disorders. The study was approved by the Research Ethics Board of the Montreal Neurological Institute and written informed consent was obtained from all subjects. 2.2. Materials For both the behavioral and fMRI studies, the stimulus materials consisted of 72 pictures portraying nature scenes and landscapes. These 72 pictures were assigned, randomly in the behavioral study and pseudo-randomly in the fMRI study, to one of two equal lists of 36 (List A and List B). With respect to the fMRI study, another methodological manipulation was done to increase the number of correctly recognized trials and hence performance under the interference condition (see Results for further details). Each original picture was divided vertically into three segments of equal size, a left segment denoted SL, a central one denoted SC, and a right segment denoted SR. 2.3. Design and procedure The modified forced-choice recognition test consisted of three successive tasks (i) a memory encoding task, (ii) an intervening task, and (iii) a forced-choice memory recognition test. Fig. 1 provides a schematic illustration of the different tasks. During the memory encoding task, one segment from each of the 72 pictures was presented for 3500 ms at a rate of 1 segment/5 s. To encourage effective encoding of these stimuli into memory, subjects were
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
431
Fig. 1. Illustration of the modified recognition memory test. Panel A presents four nature scenes/landscapes that were used as stimuli. Each one was divided into three segments (Sl=left, Sc=center, and Sr=right segments, respectively). Panel B presents a schematic representation of the task administered as a function of time. Panel C, D, and E illustrate the three successive tasks administered, for both trial types. During encoding, one segment from each of the 72 stimuli was presented. The intervening task followed, in which 36 related segments from half the stimuli were presented and represented the interference stimuli. During the forced-choice recognition task, each of the segments presented during encoding were presented concurrently with a segment from the same stimulus that had never been studied before.
instructed to judge the visual complexity of the scenes depicted. On each trial, the subject pressed one button if the scene was complex (e.g., picture including trees, mountains, waterfalls) or another button if the scene was judged to be simple (e.g., picture including only desert sand). Such task instructions are known to produce effective encoding and
good retention of the studied information. An intervening task using the same presentation parameters followed 2 min after completion of the encoding task. In this task, 36 picture segments were presented twice in a random order, and the subjects assessed the pleasantness of the picture (pleasant vs. unpleasant). These 36 segments were related (i.e., taken
432
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
from the same picture) to one of the segments presented during the encoding task. For instance, if the subject saw segment SL during the encoding task, he/she would then be presented with the SC or SR picture segment. A forcedchoice recognition test immediately followed and consisted of the successive presentation of each and every picture segment presented during encoding concurrently with a related segment that had never been presented before in the experiment. For example, if the subject studied segment SL during encoding, and saw the related segment SC during the intervening task, then he/she would be presented simultaneously with segments SL and SR during recognition and would have to indicate which one had been studied before. If, on the other hand, a segment from another picture presented at encoding, denoted SRV, was not followed by a related segment during the intervening task, then the recognition trial would consist of the presentation of SRV and either SCV or SLV. Again, the subject had to indicate which of the segments had been presented during encoding. It should be noted that none of the picture segments presented during the intervening task were presented during recognition. The forced-choice recognition test always involved a stimulus that had been presented during the encoding list and a related segment that had never been presented before. Hence, there were two kinds of trials during recognition: (i) interference trials; that is, trials for those items that were followed by the presentation of a related item, and (ii) control or interference-free trials; that is, trials for those items that were not followed by a related item. Recognition trials were presented at a rate of one every 6.5 s, with each pair of picture segments appearing for 2500 ms. Fig. 1 illustrates these different tasks. For the fMRI experiment, all three tasks were carried out in the scanner but only the forced-choice recognition test was actually scanned. During recognition, an abstract image consisting of a visually distorted stimulus that had the same layout as the other recognition stimuli served as a baseline event and subjects were instructed to simply ignore this stimulus. Thirty-six occurrences of this baseline event were interspersed throughout scanning.
Functional scans were acquired parallel to the anterior– posterior commissural plane. fMRI images were analyzed with fmristat [31]. The T2* images were first realigned to the fifth image in their respective run and then spatially smoothed with a 6 mm (fwhm) isotropic Gaussian kernel. The statistical analysis of the fMRI data was based on a linear model with correlated errors. The design matrix of the linear model was convolved with a hemodynamic response function modeled as a difference of two gamma functions timed to coincide with the acquisition of each slice. Low-frequency drifts were removed by including polynomial covariates, up to degree 3, in the design matrix. The correlation structure was modeled as an autoregressive process of degree 1. At each voxel, the autocorrelation parameter was estimated from the least squares residuals using the Yule–Walker equations. The autocorrelation parameter was first regularized by spatial smoothing with a 10-mm fwhm Gaussian filter, and then used to whiten the data and the design matrix. The linear model was then re-estimated using least squares on the whitened data to produce estimates of effects and their standard errors. For each subject, the effects and standard deviations from the two recognition runs were then averaged and the resulting images were normalized to standard space using the MNI template [8] as a reference. In a second step, subjects were combined using a mixed-effect linear model using the effects taken from the previous analysis. This was fitted using restricted maximum likelihood (ReML) estimates implemented by the maximization (EM) algorithm. A random effects analysis was performed by first estimating the ratio of the random effects variance to the fixed effects variance, then regularizing this ratio by spatial smoothing with an 8-mm fwhm Gaussian filter. The variance of the effect was then estimated by the smoothed ratio multiplied by the fixed effects variance to achieve higher degrees of freedom. Statistical maps were thresholded at pb0.001, uncorrected for multiple comparisons (t(101)=3.15). Only regions consisting of at least five contiguous voxels above the threshold are reported.
2.4. fMRI data acquisition and analyses
3. Results
Scanning was carried out at the Montreal Neurological Institute (MNI) on a 1.5-T Siemens Sonata system, using gradient echo EPI sequences. Stimuli were generated by a Pentium class PC Laptop computer running E-Prime (http:// www.pstnet.com) and projected via a LCD projector and mirror system. A three key response mouse connected to the computer recorded subjects’ responses. The scanner triggered the start of stimulus presentation. A scanning session began with a high-resolution T1-weighted acquisition for anatomical localization followed by two functional runs, each consisting of 100 T2*-weighted images acquired with blood oxygenation level-dependent (BOLD) contrast (TR=3100 ms, TE=50 ms, 25 slices, 225 mm voxels).
3.1. Behavioral study on the modified recognition memory test Data from one subject was discarded because of a technical problem that prevented this individual’s behavioral responses from being recorded. The behavioral study yielded a mean hit rate of 76% (s.d.=10) for control (interference-free) trials and 64% (s.d.=15) for interference trials. The difference in mean hit rate was significant (t(38)=4.77, pb0.001). Since we had planned to use an event-related design for our fMRI study and we wanted specifically to compare correct interference trials to correct control (interference-free) trials, we subsequently per-
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
formed an item analysis on the stimuli in the interference condition. The goal was to identify the nine stimuli that were the worst remembered and replace them by new, beasierQ ones, as assessed informally by one of us (M.B.). This new version of the task was then used in the fMRI study. It should also be noted that the stimuli used for interference and no-interference trials were not counterbalanced across subjects.
433
3.2. Behavioral and fMRI results of the main study Data from one subject was discarded due to performance being below chance levels. With regards to the forcedchoice recognition test carried out in the scanner, we noticed that a fair proportion (22%) of the errors made by subjects consisted of errors of omission. These omission errors where not systematically associated with a specific type of
Fig. 2. (A) Three-dimensional rendering of an MRI brain, which indicates the localization on the z axis of the Talairach coordinate system for each of the slices presented below. (B) Event-related imaging results ( pb0.001, uncorrected) for the contrast between recognition interference trials relative to recognition control trials on horizontal MR slices from the average brain computed from the structural images of our participants. The numbers in the upper left corner of each slice denote the distance in mm from the anterior commissural point on the z axis of the Talairach coordinate system. Significant activity in left parietal regions and right cerebellum can be observed. (C) Event-related imaging results ( pb0.001, uncorrected) for the contrast between recognition control trials relative to recognition interference trials on horizontal MR slices from the average brain computed from the structural images of our participants. The numbers in the upper left corner of each slice denote the distance in mm from the anterior commissural point on the z axis of the Talairach coordinate system. Significant activity in left prefrontal cortex, right parahippocampal cortex, and retrosplenial cortex bilaterally can be observed.
434
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
trial, as a comparison for the ratio of omission errors over the total number of errors did not yield a significant statistical difference between interference and no-interference recognition trials (t(14)=1.4, p=0.18). Because the recognition stimulus disappeared after 2500 ms, this could have inadvertently biased the subject to not respond beyond that point. Mean response time was 1839 ms (s.d.=181 ms) for interference trials and 1801 ms (s.d.=219 ms) for control trials (t(14)=0.98, p=0.35). Thus, omissions were excluded from further analyses. We then computed a percentage based on the number of correct responses over the total number of responses (correct responses+commissions errors). This analysis yielded a mean hit rate of 72% (s.d.=11) for control (interference-free) trials and of 76% (s.d.=7) for interference trials. The difference in mean hit rate was not significant (t(14)= 1.75, p=0.10). Two main contrasts for the fMRI data were examined. A first contrast compared correct interference trials to control trials and identified significant activations in left parietal regions including precuneus and inferior parietal lobule, left Table 1 Brain regions showing significant activation ( pb0.001, uncorrected) for the contrast between interference versus control (interference-free) correct recognition and for the opposite contrast (control versus interference) Region
BA
MNI coordinates X
InterferenceNcontrol Left Midbrain (red nucleus) Precuneus Angular gyrus Right Cingulate sulcus Postcentral gyrus Cerebellum Cerebellum ControlNinterference Left Middle frontal gyrus Inferior frontal gyrus Superior temporal gyrus Precentral gyrus Middle temporal gyrus Cerebellum Retrosplenial Cerebellum Right Middle frontal gyrus Inferior frontal gyrus Anterior cingulate Superior frontal gyrus Caudate Superior frontal gyrus Parahippocampal gyrus Fusiform gyrus Retrosplenial Cerebellum
Y
t value Z
6 14 32
22 58 62
18 52 32
4.24 4.47 7.37
6 5
22 14 24 46
10 34 50 74
40 60 38 34
5.46 6.06 5.67 4.12
10 13 22 6 21
32 42 50 16 52 32 10 14
46 28 10 20 20 42 56 56
4 6 6 70 8 40 12 38
4.21 5.75 4.35 4.5 5.44 5.26 4.5 5.19
26 40 8 24 12 12 26 38 18 32
48 40 38 34 0 12 36 44 48 80
20 4 36 32 26 64 14 20 6 32
3.68 4.32 4.79 4.02 3.97 4.23 3.72 4.14 3.99 4.8
7 39
30
11 47 32 9 6 36 37 30
BA=Brodmann area; XYZ coordinates of local maxima are listed according to the Talairach system [28].
midbrain, right cingulate gyrus, right postcentral gyrus and right cerebellum. Table 1 provides the coordinates for the significant foci of activations and Fig. 2 illustrates the parietal activations overlaid onto the average MRI template. The opposite contrast comparing control trials to interference trials yielded significant activation in anterior prefrontal regions bilaterally, left prefrontal opercular region, left premotor cortex, right caudate, right parahippocampal gyrus and bilaterally in medial prefrontal cortex, retrosplenial cortex and cerebellum. Once again, the coordinates for the significant foci of activations are provided in Table 1, and Fig. 2 illustrates the prefrontal and parahippocampal activations overlaid onto the average MRI template.
4. Discussion This study aimed to identify and compare the neural correlates of forced-choice recognition memory, in which the presence or absence of related (associative) information in memory was manipulated. The presence of related information in memory is hypothesized to induce itemspecific retroactive interference [6,7], a result corroborated by our behavioral study. The main fMRI findings are that correct memory recognition under interference yielded significant cortical activity mainly in left parietal regions whereas control (interference-free) recognition trials yielded greater activity in bilateral prefrontal and retrosplenial cortices and in right parahippocampal regions. Taken together, these results suggest that the presence or absence of related events in memory can influence the discriminability of temporally and contextually similar memory traces and that this effect can be observed in the pattern of brain activation. Considering that only correct forced-choice recognition trials were compared, thus holding retrieval success constant, the finding of greater prefrontal/parahippocampal activity during control recognition trials suggests that these activations could reflect neural correlates of effective discriminability of events in memory, which could represent a channel of information in the service of conscious recollection. At the behavioral level, it should be noted, however, that we did not observe the expected decrease in recognition performance for the interference condition relative to the control condition. We nonetheless assume that our manipulation at the item level was effective in affecting the discriminability of memory traces. 4.1. The discriminability/competition of representations in memory and parietal cortex Relative to control recognition trials, interference trials were associated with significant cortical activity mainly in left parietal regions, including the precuneus and angular gyrus. Previous functional neuroimaging studies have proposed that the left parietal cortex plays a role in retrieval
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
435
success (see Ref. [23] for a review). Since only correct trials for both kinds of recognition memory were examined, this finding cannot be explained in terms of differences in retrieval success. Functional neuroimaging studies on response selection point to a role of the parietal cortex in the active representation of a currently presented stimulus and the information that is strongly associated with it [3,24]. A recent study by Sohn et al. [27] systematically examined the effect of manipulating the number of associations in person–location pairs during a recognition memory task and found that both prefrontal and parietal regions were activated when a region-of-interest approach (restricted to small prefrontal, parietal and sensorimotor regions) was used. Sohn et al. then performed a confirmatory analysis and found that only the prefrontal cortex responded when the amount of competing information was manipulated. However, in a second exploratory analysis of the whole brain, a significant effect of competition was observed for a right medial parietal area in the vicinity of the precuneus. Thus, although the results of Sohn’ confirmatory analysis conflict with those from the present study, the results from their exploratory analysis suggests that the medial parietal region indeed does play a role and its activity may be modulated as a function of the competition/discriminability of related items in memory. These different results clearly present us with a controversial but interesting problem that must be addressed if we are to expand our understanding of the role of the left parietal cortex in recognition memory beyond what is already known concerning its involvement in retrieval success [23]. Another brain region revealed by the comparison of interference trials with control trials was a cortical area in the vicinity of the cingulate sulcus. Previous studies have reported a role of anterior cingulate cortex in response competition or at least in the detection of a state such as response competition [5,22]. Finally, two regions of the right cerebellum also exhibited significant activations. This is interesting because the cerebellum’s role in memory retrieval is garnering an increasing amount of attention, with several functional neuroimaging studies reporting significant cerebellar activations when subjects consciously retrieve information from memory [2,9] and several other neuropsychological studies pointing to significant impairments in simple associative learning (on simple associative learning tasks) in patients with cerebellar lesions [4,12].
success in terms of the locations of significant peaks of activity. Indeed, several studies have reported anterior prefrontal activation during the recognition of old items relative to new ones. Retrieval success, which includes judgments based on both conscious recollection and familiarity assessment, is associated with activation in the left anterior prefrontal cortex and bilateral medial and lateral parietal regions (see Ref. [23] for a review). For instance, Donaldson et al. [11] compared recognition of old and new items and observed increased activity in the anterior left prefrontal cortex and bilateral medial parietal and lateral parietal regions for old items. Similarly, McDermott et al. [19] reported increased activation in the anterior middle frontal gyrus bilaterally when comparing hits with the correct rejection of unrelated lures. By examining only hits, the pattern of brain activation can contrast those items that are recognized on the basis of conscious recollection from those based on familiarity assessment using the remember/know paradigm [29]. Briefly, brememberQ judgments refer to instances on a recognition memory test where the participant can consciously remember the initial episode, such as the occurrence of a word as part of a stimulus list. bKnowQ judgments refer instead to those instances where the participant only has a sense of familiarity for the prior occurrence of a given stimulus. Using such a paradigm, Henson et al. [15] observed activation in the left anterior prefrontal cortex, left inferior and superior parietal gyrus and posterior cingulate when remember judgments were compared to know judgments. Similarly, Eldridge et al. [13], in an fMRI study using the remember/know paradigm restricted to items already judged boldQ, directly compared remember judgments to know judgments. This contrast revealed significant activation in the left hippocampus, right parahippocampal gyrus, left middle and right inferior frontal gyrus, as well as the right posterior cingulate and bilateral inferior parietal gyrus. It is interesting to note that this right parahippocampal region has been shown to be involved in the recognition of visual scenes [18]. Thus, being able to better discriminate information in memory could represent a sufficient condition for conscious recollection. It would be of interest to combine the remember/know paradigm with Chandler’s modified recognition memory test. One could predict that we should observe fewer remember responses for interference trials relative to interference-free trials.
4.2. Prefrontal–parahippocampal activity and interference-free memory recognition
4.3. Limitations of the present study
Relative to interference trials, the control recognition trials were associated with increased activity in left inferior and right inferior and superior prefrontal activity. In addition, a right parahippocampal cortex region and the retrosplenial cortex bilaterally also exhibited significant activity. Overall, this pattern of neural activity bears a strong resemblance to previous descriptions of the patterns induced by retrieval
It is important to note that several limitations in the current should be taken into consideration. As we mentioned before, we did not find a behavioral interference effect during the fMRI study. There is nonetheless one very important difference between interference trials and nointerference trials, which is the presence or absence of related (associative) information. It is important to consider that the concept of interference refers to a behavioral phenomenon
436
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
that is defined as a reduction of performance attributable to a previous experience and does not refer to a specific neural mechanism that may be responsible for such an effect. In the present study, interference trials and no-interference trials are operationally defined as a function of the presence or absence of related information in memory. Thus, even in the absence of a behavioral interference effect, it follows that whatever differences we may have observed in fMRI signal between these two trial types can safely be assumed to be related to our task manipulation and not to a random factor. We work on the assumption that this manipulation may for an unknown proportion of trials decrease recognition discriminability, which ultimately induces an interference effect. We also assume that the process or processes that decreased recognition memory accuracy are systematically at play whenever related information has been presented. Another limitation is the fact that we did not counterbalance our interference and no-interference stimuli across subjects, thus opening up the possibility that some material-specific effects may act as a confound.
5. Conclusion We have used Chandler’s modified recognition memory test to probe recognition memory interference and its neural correlates. The retroactive interference induced by this task is hypothesized to be related to a specific problem in discriminating between temporally and contextually similar memory traces formed during a specific contextual episode. Contrary to our expectations, we did not observe greater neural activity in any regions of the prefrontal cortex. This finding is at variance with the results of functional neuroimaging studies that have examined interference in a pairedlearning task [16,30] and in a recognition memory study of person–location pairs [27], both of which reported greater prefrontal activity during interference conditions. It is likely that the process responsible for the interference in our study is quite different from that observed in paired-associate learning. The interference induced during the retrieval of paired-associates (but also during the encoding of such pairs) may be more related to the generation and selection of appropriate responses whereas in the current study interference is closely linked to the representation of information of memory. Thus, taken together, the present study and others point to the existence of different processes contributing to memory interference and these processes likely have different neural correlates.
Acknowledgments This research was supported by the Natural Sciences and Engineering Research Council of Canada (grant #165762) and by the Fonds de Recherche sur la Nature et les Technologies (grant #ER-80024). We thank B. Pike, M.
Ferreira, and K. Worsley from the Montreal Neurological Institute for assistance with the implementation and analysis of this study, and A. Cormier and the staff of the Brain Imaging Centre for their technical expertise.
References [1] M.C. Anderson, J.H. Neely, Interference and inhibition in memory retrieval, in: E.L. Bjork, R.J. Bjork (Eds.), Memory, Academic Press, San Diego, 1996, pp. 237 – 313. [2] N.C. Andreasen, D.S. O’Leary, S. Paradiso, T. Cizadlo, S. Arndt, G.L. Watkins, L.L. Ponto, R.D. Hichwa, The cerebellum plays a role in conscious episodic memory retrieval, Hum. Brain Mapp. 8 (1999) 226 – 234. [3] S.A. Bunge, E. Hazeltine, M.D. Scanlon, A.C. Rosen, J.D. Gabrieli, Dissociable contributions of prefrontal and parietal cortices to response selection, NeuroImage 17 (2002) 1562 – 1571. [4] A.G.M. Canavan, R.E. Passingham, C.D. Marsden, N. Quinn, M. Wyke, C.E. Polkey, Sequence ability in parkinsonians, patients with frontal lobe lesions and patients who have undergone unilateral temporal lobectomies, Neuropsychologia 27 (1989) 787 – 798. [5] C.S. Carter, A.M. Macdonald, M. Botvinick, L.L. Ross, V.A. Stenger, D. Noll, J.D. Cohen, Parsing executive processes: strategic vs. evaluative functions of the anterior cingulate cortex, Proc. Natl. Acad. Sci. U. S. A. 97 (2000) 1944 – 1948. [6] C.C. Chandler, How memory for an event is influenced by related events: interference in modified recognition tests, J. Exper. Psychol., Learn., Mem., Cogn. 17 (1991) 115 – 125. [7] C.C. Chandler, G.J. Gargano, Retrieval processes that produce interference in modified forced-choice recognition tests, Mem. Cogn. 26 (1998) 220 – 231. [8] C.A. Cocosco, V. Kollokian, R.K.S. Kwan, A.C. Evans, Brainweb: online interface to a 3D MRI simulated brain database, NeuroImage 5 (1997) S425. [9] J.E. Desmond, J.A. Fiez, Neuroimaging studies of the cerebellum: language, learning and memory, Trends Cogn. Sci. 2 (1998) 355 – 362. [10] R.J. Dolan, P.C. Fletcher, Dissociating prefrontal and hippocampal function in episodic memory encoding, Nature 388 (1997) 582 – 585. [11] D.I. Donaldson, S.E. Petersen, J.M. Ollinger, R.L. Buckner, Dissociating state and item components of recognition memory using fMRI, NeuroImage 13 (2001) 129 – 142. [12] J. Drepper, D. Timmann, F.P. Kolb, H.C. Diener, Non-motor associative learning in patients with isolated degenerative cerebellar disease, Brain 122 (Pt. 1) (1999) 87 – 97. [13] L.L. Eldridge, B.J. Knowlton, C.S. Furmanski, S.Y. Bookheimer, S.A. Engel, Remembering episodes: a selective role for the hippocampus during retrieval, Nat. Neurosci. 3 (2000) 1149 – 1152. [14] P. Fletcher, T. Shallice, R.J. Dolan, Sculpting the response space-an account of left prefrontal activation during encoding, NeuroImage 12 (2000) 404 – 417. [15] R.N.A. Henson, M.D. Rugg, T. Shallice, O. Josephs, R.J. Dolan, Recollection and familiarity in recognition memory: an event-related functional magnetic resonance imaging study, J. Neurosci. 19 (1999) 3962 – 3972. [16] R.N.A. Henson, T. Shallice, O. Josephs, R.J. Dolan, Functional magnetic resonance imaging of proactive interference durign spoken cued recall, NeuroImage 17 (2002) 543 – 558. [17] A. Incisa della Rocchetta, B. Milner, Strategic search and retrieval inhibition: the role of the frontal lobes, Neuropsychologia 31 (1993) 503 – 524. [18] S. Kfhler, J. Crane, B. Milner, Differential contributions of the parahippocampal place area and the anterior hippocampus to human memory for scenes, Hippocampus 12 (2002) 718 – 723.
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437 [19] K.B. McDermott, T.C. Jones, S.E. Petersen, S.K. Lageman, H.L. Roediger, Retrieval success is accompanied by enhanced activation in anterior prefrontal cortex during recognition memory: an event-related fMRI study, J. Cogn. Neurosci. 12 (2000) 965 – 976. [20] J.A. McGeoch, The Psychology of Human Learning, Longmans, Green, NY, 1942. [21] R.C. Oldfield, The assessment and analysis of handedness: the Edinburgh inventory, Neuropsychologia 9 (1971) 97 – 113. [22] C.C. Ruff, T.S. Woodward, K.R. Laurens, P.F. Liddle, The role of the anterior cingulate cortex in conflict processing: evidence from reverse Stroop interference, NeuroImage 14 (2001) 1150 – 1158. [23] M.D. Rugg, R.N.A. Henson, Episodic memory retrieval: an (eventrelated) functional neuroimaging perspective, in: A. Parker, E.L. Wilding, T.J. Bussey (Eds.), The Cognitive Neuroscience of Memory: Encoding and Retrieval, Psychology Press, New York, 2002, pp. 3 – 37. [24] E.H. Schumacher, M. D’Esposito, Neural implementation of response selection in humans as revealed by localized effects of stimulusresponse compatibility on brain activation, Hum. Brain Mapp. 17 (2002) 193 – 201.
437
[25] A.P. Shimamura, P.J. Jurica, J.A. Mangels, F.B. Gershberg, R.T. Knight, Susceptibility to memory interference effects following frontal lobe damage: findings from tests of paired-associate learning, J. Cogn. Neurosci. 7 (1995) 144 – 152. [26] M.L. Smith, G. Leonard, J. Crane, B. Milner, The effects of frontal- or temporal-lobe lesions on susceptibility to interference in spatial memory, Neuropsychologia 33 (1995) 275 – 285. [27] M.H. Sohn, A. Goode, V.A. Stenger, C.S. Carter, J.R. Anderson, Competition and representation during memory retrieval: roles of the prefrontal cortex and the posterior parietal cortex, Proc. Natl. Acad. Sci. U. S. A. 100 (2003) 7412 – 7417. [28] J. Talairach, P. Tournoux, Co-Planar Stereotaxic Atlas of the Human Brain, Thieme, New York, 1988. [29] E. Tulving, Memory and consciousness, Can. Psychol. 26 (1985) 1 – 12. [30] F. Uhl, I. Podreka, L. Deecke, Anterior frontal cortex and the effect of proactive interference in word pair learning-results of brain-SPECT, Neuropsychologia 32 (1994) 241 – 247. [31] K.J. Worsley, C.H. Liao, J. Aston, V. Petre, G.H. Duncan, F. Morales, A.C. Evans, A general statistical analysis for fMRI data, NeuroImage 15 (2002) 1 – 15.
Research report
The interfering effect of related events on recognition memory discriminability: a functional magnetic resonance imaging study Martin Lepage*, Franc¸ois Blondin, Ame´lie M. Achim, Matthew Menear, Mathieu Brodeur Brain Imaging Group, Douglas Hospital Research Centre, McGill University, 6875 Boul. LaSalle Verdun, Que´bec, Canada H4H 1R3 Accepted 1 October 2004 Available online 11 November 2004
Abstract Retrieval of information from memory often involves the selection of an event among competing related events, a process that frequently gives rise to interference effects. The present study used a forced-choice recognition test to identify neural correlates of the interfering effect of related events on recognition memory discriminability. Participants encoded landscape pictures divided into three segments. One segment was presented during encoding, and a forced-choice recognition task contrasted a studied and a nonstudied segment for each landscape. For half of the landscapes, the third segment was presented between encoding and recognition tasks to induce associative interference by reducing recognition discriminability. A behavioral study with 40 subjects yielded a significant difference in the correct recognition rate between control and interference trials (76% and 64%, respectively, pb0.001). A subsequent event-related fMRI study with 16 subjects yielded significant activations for correct interference recognition trials relative to control trials in left superior parietal regions, which suggests that these regions play a role in the representation of stimuli and associated information. The opposite contrast yielded significant activations in inferior prefrontal regions bilaterally, right dorsolateral prefrontal cortex and right parahippocampal cortex. Since this contrast was conducted using only correctly recognized trials, these findings could reflect an index of memory discriminability or saliency which could influence conscious recollection. D 2004 Elsevier B.V. All rights reserved. Theme: Neural basis of behavior Topic: Cognition Keywords: Prefrontal cortex; Episodic memory; Parietal cortex; Recognition memory
1. Introduction Retrieval of information from episodic memory often involves the selection of an event among competing related events, a process that frequently gives rise to interference effects. Although research on memory interference effects has a long history [1,20], little is known about its neural basis in humans. Some of the evidence that has been gathered so far points to an involvement of the frontal lobes, as it has been shown that patients with lesions in this area
* Corresponding author. Fax: +1 514 888 4064. E-mail address: [email protected] (M. Lepage). 0926-6410/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.cogbrainres.2004.10.003
demonstrate faulty control of interference during memory tasks [17,25,26]. This notion has been further supported by several functional neuroimaging studies carried out with healthy subjects that have found the prefrontal cortex to be a potential neural correlate of the control of memory interference. Of these studies, two used Positron Emission Tomography (PET) to examine interference during memory encoding [10,14]. In both studies, subjects learned word pairs and then had to learn new associations between already studied words. Increased left prefrontal cortex activation was observed when learning associations, suggesting that this activation could reflect the neural correlate of a form of proactive interference. In a SPECT study by Uhl et al. [30], memory interference during retrieval was
430
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
examined using an AB–AD memory interference cuedrecall task, a form of paired-associate learning task where a cue word (A) is successively paired with two different target words (B, D). A right prefrontal anterior cortical region exhibited greater activity during the retrieval of the AD (interference prone) list relative to the AB (interference free) list. Recently, Henson et al. [16] developed an ingenious fMRI experiment that allowed them to study AB–AD paired-associate recall and to record overt verbal responses while minimizing fMRI signal distortion caused by speechrelated head motion. In this study, neural correlates of the encoding and retrieval components of an AB–AD paired associate cued-recall paradigm were examined. A region in the left inferior frontal cortex and a bilateral frontopolar region showed greater activity for memory interference trials relative to control trials during both encoding and retrieval. These results confirm the important role of the prefrontal cortex in the control of memory interference for both encoding and retrieval. It is well established in cognitive psychology that the interference effects on memory performance that are usually observed on the paired-recall tasks employed by investigators such as Uhl and Henson occur rarely, if ever, when examining recognition memory [1]. One exception is the modified recognition memory test proposed by Chandler and Gargano [6,7] in which the amount of retrieval interference can be manipulated across experimental conditions. During a study episode, nature pictures (targets) are followed by either related pictures (interference items) or by unrelated pictures (control items). A forced-choice recognition test shortly follows in which the studied items are presented concurrently with a new distractor and subjects must indicate which of the two pictures had been presented earlier. In this paradigm, subjects usually remember fewer items when their related counterparts have also been presented, suggesting that related representations can produce retrieval interference. In addition, since this interfering effect is time limited and is not observed following a long (48 h) delay, the original trace could not have been binterferedQ with by the related events, thus pointing to retrieval as the memory stage where interference occurs. The vast majority of the functional neuroimaging studies that have examined memory retrieval, and in particular those that were conducted using fMRI, have used a recognition memory test format. Thus, Chandler’s modified recognition memory test appears well-suited to bridge the gap between the literature on the neural correlates of recognition memory and memory retrieval interference. We therefore used the modified forced-choice recognition memory test developed by Chandler [6] to identify neural correlates of recognition memory interference in an eventrelated fMRI study. Based on the previous findings from the functional neuroimaging studies of memory interference using recall tasks, we hypothesized that anterior prefrontal cortex regions would exhibit greater activity during inter-
ference trials relative to control trials. We first conducted a behavioral study to assess whether our own version of the modified recognition memory test would replicate the interference effect observed in Chandler’s studies. The versions differed in two main ways: (i) we presented subjects with a computerized version of the task, and (ii) we used a different set of stimuli. Following the behavioral study, we performed a functional neuroimaging study using fMRI. Because these two studies share similar behavioral methods, they are described together.
2. Methods 2.1. Subjects Forty subjects (24 females, mean age 33.2, range 19–50 years) participated in the behavioral study and were tested on our adapted version of Chandler’s modified recognition memory test. None had a current or previous history of neurological or psychiatric disorders and all but one of the participants were right-handed. Following the completion of this study, 16 new subjects (8 females; mean age 25.4 range 22–29 years) who were all right-handed according to the Edinburgh Inventory [21] participated in the fMRI study. Once again, none of the participants had a current or previous history of neurological or psychiatric disorders. The study was approved by the Research Ethics Board of the Montreal Neurological Institute and written informed consent was obtained from all subjects. 2.2. Materials For both the behavioral and fMRI studies, the stimulus materials consisted of 72 pictures portraying nature scenes and landscapes. These 72 pictures were assigned, randomly in the behavioral study and pseudo-randomly in the fMRI study, to one of two equal lists of 36 (List A and List B). With respect to the fMRI study, another methodological manipulation was done to increase the number of correctly recognized trials and hence performance under the interference condition (see Results for further details). Each original picture was divided vertically into three segments of equal size, a left segment denoted SL, a central one denoted SC, and a right segment denoted SR. 2.3. Design and procedure The modified forced-choice recognition test consisted of three successive tasks (i) a memory encoding task, (ii) an intervening task, and (iii) a forced-choice memory recognition test. Fig. 1 provides a schematic illustration of the different tasks. During the memory encoding task, one segment from each of the 72 pictures was presented for 3500 ms at a rate of 1 segment/5 s. To encourage effective encoding of these stimuli into memory, subjects were
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
431
Fig. 1. Illustration of the modified recognition memory test. Panel A presents four nature scenes/landscapes that were used as stimuli. Each one was divided into three segments (Sl=left, Sc=center, and Sr=right segments, respectively). Panel B presents a schematic representation of the task administered as a function of time. Panel C, D, and E illustrate the three successive tasks administered, for both trial types. During encoding, one segment from each of the 72 stimuli was presented. The intervening task followed, in which 36 related segments from half the stimuli were presented and represented the interference stimuli. During the forced-choice recognition task, each of the segments presented during encoding were presented concurrently with a segment from the same stimulus that had never been studied before.
instructed to judge the visual complexity of the scenes depicted. On each trial, the subject pressed one button if the scene was complex (e.g., picture including trees, mountains, waterfalls) or another button if the scene was judged to be simple (e.g., picture including only desert sand). Such task instructions are known to produce effective encoding and
good retention of the studied information. An intervening task using the same presentation parameters followed 2 min after completion of the encoding task. In this task, 36 picture segments were presented twice in a random order, and the subjects assessed the pleasantness of the picture (pleasant vs. unpleasant). These 36 segments were related (i.e., taken
432
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
from the same picture) to one of the segments presented during the encoding task. For instance, if the subject saw segment SL during the encoding task, he/she would then be presented with the SC or SR picture segment. A forcedchoice recognition test immediately followed and consisted of the successive presentation of each and every picture segment presented during encoding concurrently with a related segment that had never been presented before in the experiment. For example, if the subject studied segment SL during encoding, and saw the related segment SC during the intervening task, then he/she would be presented simultaneously with segments SL and SR during recognition and would have to indicate which one had been studied before. If, on the other hand, a segment from another picture presented at encoding, denoted SRV, was not followed by a related segment during the intervening task, then the recognition trial would consist of the presentation of SRV and either SCV or SLV. Again, the subject had to indicate which of the segments had been presented during encoding. It should be noted that none of the picture segments presented during the intervening task were presented during recognition. The forced-choice recognition test always involved a stimulus that had been presented during the encoding list and a related segment that had never been presented before. Hence, there were two kinds of trials during recognition: (i) interference trials; that is, trials for those items that were followed by the presentation of a related item, and (ii) control or interference-free trials; that is, trials for those items that were not followed by a related item. Recognition trials were presented at a rate of one every 6.5 s, with each pair of picture segments appearing for 2500 ms. Fig. 1 illustrates these different tasks. For the fMRI experiment, all three tasks were carried out in the scanner but only the forced-choice recognition test was actually scanned. During recognition, an abstract image consisting of a visually distorted stimulus that had the same layout as the other recognition stimuli served as a baseline event and subjects were instructed to simply ignore this stimulus. Thirty-six occurrences of this baseline event were interspersed throughout scanning.
Functional scans were acquired parallel to the anterior– posterior commissural plane. fMRI images were analyzed with fmristat [31]. The T2* images were first realigned to the fifth image in their respective run and then spatially smoothed with a 6 mm (fwhm) isotropic Gaussian kernel. The statistical analysis of the fMRI data was based on a linear model with correlated errors. The design matrix of the linear model was convolved with a hemodynamic response function modeled as a difference of two gamma functions timed to coincide with the acquisition of each slice. Low-frequency drifts were removed by including polynomial covariates, up to degree 3, in the design matrix. The correlation structure was modeled as an autoregressive process of degree 1. At each voxel, the autocorrelation parameter was estimated from the least squares residuals using the Yule–Walker equations. The autocorrelation parameter was first regularized by spatial smoothing with a 10-mm fwhm Gaussian filter, and then used to whiten the data and the design matrix. The linear model was then re-estimated using least squares on the whitened data to produce estimates of effects and their standard errors. For each subject, the effects and standard deviations from the two recognition runs were then averaged and the resulting images were normalized to standard space using the MNI template [8] as a reference. In a second step, subjects were combined using a mixed-effect linear model using the effects taken from the previous analysis. This was fitted using restricted maximum likelihood (ReML) estimates implemented by the maximization (EM) algorithm. A random effects analysis was performed by first estimating the ratio of the random effects variance to the fixed effects variance, then regularizing this ratio by spatial smoothing with an 8-mm fwhm Gaussian filter. The variance of the effect was then estimated by the smoothed ratio multiplied by the fixed effects variance to achieve higher degrees of freedom. Statistical maps were thresholded at pb0.001, uncorrected for multiple comparisons (t(101)=3.15). Only regions consisting of at least five contiguous voxels above the threshold are reported.
2.4. fMRI data acquisition and analyses
3. Results
Scanning was carried out at the Montreal Neurological Institute (MNI) on a 1.5-T Siemens Sonata system, using gradient echo EPI sequences. Stimuli were generated by a Pentium class PC Laptop computer running E-Prime (http:// www.pstnet.com) and projected via a LCD projector and mirror system. A three key response mouse connected to the computer recorded subjects’ responses. The scanner triggered the start of stimulus presentation. A scanning session began with a high-resolution T1-weighted acquisition for anatomical localization followed by two functional runs, each consisting of 100 T2*-weighted images acquired with blood oxygenation level-dependent (BOLD) contrast (TR=3100 ms, TE=50 ms, 25 slices, 225 mm voxels).
3.1. Behavioral study on the modified recognition memory test Data from one subject was discarded because of a technical problem that prevented this individual’s behavioral responses from being recorded. The behavioral study yielded a mean hit rate of 76% (s.d.=10) for control (interference-free) trials and 64% (s.d.=15) for interference trials. The difference in mean hit rate was significant (t(38)=4.77, pb0.001). Since we had planned to use an event-related design for our fMRI study and we wanted specifically to compare correct interference trials to correct control (interference-free) trials, we subsequently per-
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
formed an item analysis on the stimuli in the interference condition. The goal was to identify the nine stimuli that were the worst remembered and replace them by new, beasierQ ones, as assessed informally by one of us (M.B.). This new version of the task was then used in the fMRI study. It should also be noted that the stimuli used for interference and no-interference trials were not counterbalanced across subjects.
433
3.2. Behavioral and fMRI results of the main study Data from one subject was discarded due to performance being below chance levels. With regards to the forcedchoice recognition test carried out in the scanner, we noticed that a fair proportion (22%) of the errors made by subjects consisted of errors of omission. These omission errors where not systematically associated with a specific type of
Fig. 2. (A) Three-dimensional rendering of an MRI brain, which indicates the localization on the z axis of the Talairach coordinate system for each of the slices presented below. (B) Event-related imaging results ( pb0.001, uncorrected) for the contrast between recognition interference trials relative to recognition control trials on horizontal MR slices from the average brain computed from the structural images of our participants. The numbers in the upper left corner of each slice denote the distance in mm from the anterior commissural point on the z axis of the Talairach coordinate system. Significant activity in left parietal regions and right cerebellum can be observed. (C) Event-related imaging results ( pb0.001, uncorrected) for the contrast between recognition control trials relative to recognition interference trials on horizontal MR slices from the average brain computed from the structural images of our participants. The numbers in the upper left corner of each slice denote the distance in mm from the anterior commissural point on the z axis of the Talairach coordinate system. Significant activity in left prefrontal cortex, right parahippocampal cortex, and retrosplenial cortex bilaterally can be observed.
434
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
trial, as a comparison for the ratio of omission errors over the total number of errors did not yield a significant statistical difference between interference and no-interference recognition trials (t(14)=1.4, p=0.18). Because the recognition stimulus disappeared after 2500 ms, this could have inadvertently biased the subject to not respond beyond that point. Mean response time was 1839 ms (s.d.=181 ms) for interference trials and 1801 ms (s.d.=219 ms) for control trials (t(14)=0.98, p=0.35). Thus, omissions were excluded from further analyses. We then computed a percentage based on the number of correct responses over the total number of responses (correct responses+commissions errors). This analysis yielded a mean hit rate of 72% (s.d.=11) for control (interference-free) trials and of 76% (s.d.=7) for interference trials. The difference in mean hit rate was not significant (t(14)= 1.75, p=0.10). Two main contrasts for the fMRI data were examined. A first contrast compared correct interference trials to control trials and identified significant activations in left parietal regions including precuneus and inferior parietal lobule, left Table 1 Brain regions showing significant activation ( pb0.001, uncorrected) for the contrast between interference versus control (interference-free) correct recognition and for the opposite contrast (control versus interference) Region
BA
MNI coordinates X
InterferenceNcontrol Left Midbrain (red nucleus) Precuneus Angular gyrus Right Cingulate sulcus Postcentral gyrus Cerebellum Cerebellum ControlNinterference Left Middle frontal gyrus Inferior frontal gyrus Superior temporal gyrus Precentral gyrus Middle temporal gyrus Cerebellum Retrosplenial Cerebellum Right Middle frontal gyrus Inferior frontal gyrus Anterior cingulate Superior frontal gyrus Caudate Superior frontal gyrus Parahippocampal gyrus Fusiform gyrus Retrosplenial Cerebellum
Y
t value Z
6 14 32
22 58 62
18 52 32
4.24 4.47 7.37
6 5
22 14 24 46
10 34 50 74
40 60 38 34
5.46 6.06 5.67 4.12
10 13 22 6 21
32 42 50 16 52 32 10 14
46 28 10 20 20 42 56 56
4 6 6 70 8 40 12 38
4.21 5.75 4.35 4.5 5.44 5.26 4.5 5.19
26 40 8 24 12 12 26 38 18 32
48 40 38 34 0 12 36 44 48 80
20 4 36 32 26 64 14 20 6 32
3.68 4.32 4.79 4.02 3.97 4.23 3.72 4.14 3.99 4.8
7 39
30
11 47 32 9 6 36 37 30
BA=Brodmann area; XYZ coordinates of local maxima are listed according to the Talairach system [28].
midbrain, right cingulate gyrus, right postcentral gyrus and right cerebellum. Table 1 provides the coordinates for the significant foci of activations and Fig. 2 illustrates the parietal activations overlaid onto the average MRI template. The opposite contrast comparing control trials to interference trials yielded significant activation in anterior prefrontal regions bilaterally, left prefrontal opercular region, left premotor cortex, right caudate, right parahippocampal gyrus and bilaterally in medial prefrontal cortex, retrosplenial cortex and cerebellum. Once again, the coordinates for the significant foci of activations are provided in Table 1, and Fig. 2 illustrates the prefrontal and parahippocampal activations overlaid onto the average MRI template.
4. Discussion This study aimed to identify and compare the neural correlates of forced-choice recognition memory, in which the presence or absence of related (associative) information in memory was manipulated. The presence of related information in memory is hypothesized to induce itemspecific retroactive interference [6,7], a result corroborated by our behavioral study. The main fMRI findings are that correct memory recognition under interference yielded significant cortical activity mainly in left parietal regions whereas control (interference-free) recognition trials yielded greater activity in bilateral prefrontal and retrosplenial cortices and in right parahippocampal regions. Taken together, these results suggest that the presence or absence of related events in memory can influence the discriminability of temporally and contextually similar memory traces and that this effect can be observed in the pattern of brain activation. Considering that only correct forced-choice recognition trials were compared, thus holding retrieval success constant, the finding of greater prefrontal/parahippocampal activity during control recognition trials suggests that these activations could reflect neural correlates of effective discriminability of events in memory, which could represent a channel of information in the service of conscious recollection. At the behavioral level, it should be noted, however, that we did not observe the expected decrease in recognition performance for the interference condition relative to the control condition. We nonetheless assume that our manipulation at the item level was effective in affecting the discriminability of memory traces. 4.1. The discriminability/competition of representations in memory and parietal cortex Relative to control recognition trials, interference trials were associated with significant cortical activity mainly in left parietal regions, including the precuneus and angular gyrus. Previous functional neuroimaging studies have proposed that the left parietal cortex plays a role in retrieval
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
435
success (see Ref. [23] for a review). Since only correct trials for both kinds of recognition memory were examined, this finding cannot be explained in terms of differences in retrieval success. Functional neuroimaging studies on response selection point to a role of the parietal cortex in the active representation of a currently presented stimulus and the information that is strongly associated with it [3,24]. A recent study by Sohn et al. [27] systematically examined the effect of manipulating the number of associations in person–location pairs during a recognition memory task and found that both prefrontal and parietal regions were activated when a region-of-interest approach (restricted to small prefrontal, parietal and sensorimotor regions) was used. Sohn et al. then performed a confirmatory analysis and found that only the prefrontal cortex responded when the amount of competing information was manipulated. However, in a second exploratory analysis of the whole brain, a significant effect of competition was observed for a right medial parietal area in the vicinity of the precuneus. Thus, although the results of Sohn’ confirmatory analysis conflict with those from the present study, the results from their exploratory analysis suggests that the medial parietal region indeed does play a role and its activity may be modulated as a function of the competition/discriminability of related items in memory. These different results clearly present us with a controversial but interesting problem that must be addressed if we are to expand our understanding of the role of the left parietal cortex in recognition memory beyond what is already known concerning its involvement in retrieval success [23]. Another brain region revealed by the comparison of interference trials with control trials was a cortical area in the vicinity of the cingulate sulcus. Previous studies have reported a role of anterior cingulate cortex in response competition or at least in the detection of a state such as response competition [5,22]. Finally, two regions of the right cerebellum also exhibited significant activations. This is interesting because the cerebellum’s role in memory retrieval is garnering an increasing amount of attention, with several functional neuroimaging studies reporting significant cerebellar activations when subjects consciously retrieve information from memory [2,9] and several other neuropsychological studies pointing to significant impairments in simple associative learning (on simple associative learning tasks) in patients with cerebellar lesions [4,12].
success in terms of the locations of significant peaks of activity. Indeed, several studies have reported anterior prefrontal activation during the recognition of old items relative to new ones. Retrieval success, which includes judgments based on both conscious recollection and familiarity assessment, is associated with activation in the left anterior prefrontal cortex and bilateral medial and lateral parietal regions (see Ref. [23] for a review). For instance, Donaldson et al. [11] compared recognition of old and new items and observed increased activity in the anterior left prefrontal cortex and bilateral medial parietal and lateral parietal regions for old items. Similarly, McDermott et al. [19] reported increased activation in the anterior middle frontal gyrus bilaterally when comparing hits with the correct rejection of unrelated lures. By examining only hits, the pattern of brain activation can contrast those items that are recognized on the basis of conscious recollection from those based on familiarity assessment using the remember/know paradigm [29]. Briefly, brememberQ judgments refer to instances on a recognition memory test where the participant can consciously remember the initial episode, such as the occurrence of a word as part of a stimulus list. bKnowQ judgments refer instead to those instances where the participant only has a sense of familiarity for the prior occurrence of a given stimulus. Using such a paradigm, Henson et al. [15] observed activation in the left anterior prefrontal cortex, left inferior and superior parietal gyrus and posterior cingulate when remember judgments were compared to know judgments. Similarly, Eldridge et al. [13], in an fMRI study using the remember/know paradigm restricted to items already judged boldQ, directly compared remember judgments to know judgments. This contrast revealed significant activation in the left hippocampus, right parahippocampal gyrus, left middle and right inferior frontal gyrus, as well as the right posterior cingulate and bilateral inferior parietal gyrus. It is interesting to note that this right parahippocampal region has been shown to be involved in the recognition of visual scenes [18]. Thus, being able to better discriminate information in memory could represent a sufficient condition for conscious recollection. It would be of interest to combine the remember/know paradigm with Chandler’s modified recognition memory test. One could predict that we should observe fewer remember responses for interference trials relative to interference-free trials.
4.2. Prefrontal–parahippocampal activity and interference-free memory recognition
4.3. Limitations of the present study
Relative to interference trials, the control recognition trials were associated with increased activity in left inferior and right inferior and superior prefrontal activity. In addition, a right parahippocampal cortex region and the retrosplenial cortex bilaterally also exhibited significant activity. Overall, this pattern of neural activity bears a strong resemblance to previous descriptions of the patterns induced by retrieval
It is important to note that several limitations in the current should be taken into consideration. As we mentioned before, we did not find a behavioral interference effect during the fMRI study. There is nonetheless one very important difference between interference trials and nointerference trials, which is the presence or absence of related (associative) information. It is important to consider that the concept of interference refers to a behavioral phenomenon
436
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437
that is defined as a reduction of performance attributable to a previous experience and does not refer to a specific neural mechanism that may be responsible for such an effect. In the present study, interference trials and no-interference trials are operationally defined as a function of the presence or absence of related information in memory. Thus, even in the absence of a behavioral interference effect, it follows that whatever differences we may have observed in fMRI signal between these two trial types can safely be assumed to be related to our task manipulation and not to a random factor. We work on the assumption that this manipulation may for an unknown proportion of trials decrease recognition discriminability, which ultimately induces an interference effect. We also assume that the process or processes that decreased recognition memory accuracy are systematically at play whenever related information has been presented. Another limitation is the fact that we did not counterbalance our interference and no-interference stimuli across subjects, thus opening up the possibility that some material-specific effects may act as a confound.
5. Conclusion We have used Chandler’s modified recognition memory test to probe recognition memory interference and its neural correlates. The retroactive interference induced by this task is hypothesized to be related to a specific problem in discriminating between temporally and contextually similar memory traces formed during a specific contextual episode. Contrary to our expectations, we did not observe greater neural activity in any regions of the prefrontal cortex. This finding is at variance with the results of functional neuroimaging studies that have examined interference in a pairedlearning task [16,30] and in a recognition memory study of person–location pairs [27], both of which reported greater prefrontal activity during interference conditions. It is likely that the process responsible for the interference in our study is quite different from that observed in paired-associate learning. The interference induced during the retrieval of paired-associates (but also during the encoding of such pairs) may be more related to the generation and selection of appropriate responses whereas in the current study interference is closely linked to the representation of information of memory. Thus, taken together, the present study and others point to the existence of different processes contributing to memory interference and these processes likely have different neural correlates.
Acknowledgments This research was supported by the Natural Sciences and Engineering Research Council of Canada (grant #165762) and by the Fonds de Recherche sur la Nature et les Technologies (grant #ER-80024). We thank B. Pike, M.
Ferreira, and K. Worsley from the Montreal Neurological Institute for assistance with the implementation and analysis of this study, and A. Cormier and the staff of the Brain Imaging Centre for their technical expertise.
References [1] M.C. Anderson, J.H. Neely, Interference and inhibition in memory retrieval, in: E.L. Bjork, R.J. Bjork (Eds.), Memory, Academic Press, San Diego, 1996, pp. 237 – 313. [2] N.C. Andreasen, D.S. O’Leary, S. Paradiso, T. Cizadlo, S. Arndt, G.L. Watkins, L.L. Ponto, R.D. Hichwa, The cerebellum plays a role in conscious episodic memory retrieval, Hum. Brain Mapp. 8 (1999) 226 – 234. [3] S.A. Bunge, E. Hazeltine, M.D. Scanlon, A.C. Rosen, J.D. Gabrieli, Dissociable contributions of prefrontal and parietal cortices to response selection, NeuroImage 17 (2002) 1562 – 1571. [4] A.G.M. Canavan, R.E. Passingham, C.D. Marsden, N. Quinn, M. Wyke, C.E. Polkey, Sequence ability in parkinsonians, patients with frontal lobe lesions and patients who have undergone unilateral temporal lobectomies, Neuropsychologia 27 (1989) 787 – 798. [5] C.S. Carter, A.M. Macdonald, M. Botvinick, L.L. Ross, V.A. Stenger, D. Noll, J.D. Cohen, Parsing executive processes: strategic vs. evaluative functions of the anterior cingulate cortex, Proc. Natl. Acad. Sci. U. S. A. 97 (2000) 1944 – 1948. [6] C.C. Chandler, How memory for an event is influenced by related events: interference in modified recognition tests, J. Exper. Psychol., Learn., Mem., Cogn. 17 (1991) 115 – 125. [7] C.C. Chandler, G.J. Gargano, Retrieval processes that produce interference in modified forced-choice recognition tests, Mem. Cogn. 26 (1998) 220 – 231. [8] C.A. Cocosco, V. Kollokian, R.K.S. Kwan, A.C. Evans, Brainweb: online interface to a 3D MRI simulated brain database, NeuroImage 5 (1997) S425. [9] J.E. Desmond, J.A. Fiez, Neuroimaging studies of the cerebellum: language, learning and memory, Trends Cogn. Sci. 2 (1998) 355 – 362. [10] R.J. Dolan, P.C. Fletcher, Dissociating prefrontal and hippocampal function in episodic memory encoding, Nature 388 (1997) 582 – 585. [11] D.I. Donaldson, S.E. Petersen, J.M. Ollinger, R.L. Buckner, Dissociating state and item components of recognition memory using fMRI, NeuroImage 13 (2001) 129 – 142. [12] J. Drepper, D. Timmann, F.P. Kolb, H.C. Diener, Non-motor associative learning in patients with isolated degenerative cerebellar disease, Brain 122 (Pt. 1) (1999) 87 – 97. [13] L.L. Eldridge, B.J. Knowlton, C.S. Furmanski, S.Y. Bookheimer, S.A. Engel, Remembering episodes: a selective role for the hippocampus during retrieval, Nat. Neurosci. 3 (2000) 1149 – 1152. [14] P. Fletcher, T. Shallice, R.J. Dolan, Sculpting the response space-an account of left prefrontal activation during encoding, NeuroImage 12 (2000) 404 – 417. [15] R.N.A. Henson, M.D. Rugg, T. Shallice, O. Josephs, R.J. Dolan, Recollection and familiarity in recognition memory: an event-related functional magnetic resonance imaging study, J. Neurosci. 19 (1999) 3962 – 3972. [16] R.N.A. Henson, T. Shallice, O. Josephs, R.J. Dolan, Functional magnetic resonance imaging of proactive interference durign spoken cued recall, NeuroImage 17 (2002) 543 – 558. [17] A. Incisa della Rocchetta, B. Milner, Strategic search and retrieval inhibition: the role of the frontal lobes, Neuropsychologia 31 (1993) 503 – 524. [18] S. Kfhler, J. Crane, B. Milner, Differential contributions of the parahippocampal place area and the anterior hippocampus to human memory for scenes, Hippocampus 12 (2002) 718 – 723.
M. Lepage et al. / Cognitive Brain Research 22 (2005) 429–437 [19] K.B. McDermott, T.C. Jones, S.E. Petersen, S.K. Lageman, H.L. Roediger, Retrieval success is accompanied by enhanced activation in anterior prefrontal cortex during recognition memory: an event-related fMRI study, J. Cogn. Neurosci. 12 (2000) 965 – 976. [20] J.A. McGeoch, The Psychology of Human Learning, Longmans, Green, NY, 1942. [21] R.C. Oldfield, The assessment and analysis of handedness: the Edinburgh inventory, Neuropsychologia 9 (1971) 97 – 113. [22] C.C. Ruff, T.S. Woodward, K.R. Laurens, P.F. Liddle, The role of the anterior cingulate cortex in conflict processing: evidence from reverse Stroop interference, NeuroImage 14 (2001) 1150 – 1158. [23] M.D. Rugg, R.N.A. Henson, Episodic memory retrieval: an (eventrelated) functional neuroimaging perspective, in: A. Parker, E.L. Wilding, T.J. Bussey (Eds.), The Cognitive Neuroscience of Memory: Encoding and Retrieval, Psychology Press, New York, 2002, pp. 3 – 37. [24] E.H. Schumacher, M. D’Esposito, Neural implementation of response selection in humans as revealed by localized effects of stimulusresponse compatibility on brain activation, Hum. Brain Mapp. 17 (2002) 193 – 201.
437
[25] A.P. Shimamura, P.J. Jurica, J.A. Mangels, F.B. Gershberg, R.T. Knight, Susceptibility to memory interference effects following frontal lobe damage: findings from tests of paired-associate learning, J. Cogn. Neurosci. 7 (1995) 144 – 152. [26] M.L. Smith, G. Leonard, J. Crane, B. Milner, The effects of frontal- or temporal-lobe lesions on susceptibility to interference in spatial memory, Neuropsychologia 33 (1995) 275 – 285. [27] M.H. Sohn, A. Goode, V.A. Stenger, C.S. Carter, J.R. Anderson, Competition and representation during memory retrieval: roles of the prefrontal cortex and the posterior parietal cortex, Proc. Natl. Acad. Sci. U. S. A. 100 (2003) 7412 – 7417. [28] J. Talairach, P. Tournoux, Co-Planar Stereotaxic Atlas of the Human Brain, Thieme, New York, 1988. [29] E. Tulving, Memory and consciousness, Can. Psychol. 26 (1985) 1 – 12. [30] F. Uhl, I. Podreka, L. Deecke, Anterior frontal cortex and the effect of proactive interference in word pair learning-results of brain-SPECT, Neuropsychologia 32 (1994) 241 – 247. [31] K.J. Worsley, C.H. Liao, J. Aston, V. Petre, G.H. Duncan, F. Morales, A.C. Evans, A general statistical analysis for fMRI data, NeuroImage 15 (2002) 1 – 15.