- Email: [email protected]
Amygdala modulation of parahippocampal and frontal regions during emotionally influenced memory storage
NeuroImage 20 (2003) 2091–2099
www.elsevier.com/locate/ynimg
Amygdala modulation of parahippocampal and frontal regions during emotionally influenced memory storage Lisa Kilpatrick and Larry Cahill* Department of Neurobiology and Behavior and Center for the Neurobiology of Learning and Memory, University of California, Irvine, CA 92697-3800, USA Received 8 April 2003; revised 30 July 2003; accepted 4 August 2003
Abstract Considerable evidence from both animal and human subject research supports the hypothesis that the amygdala, when activated by emotional arousal, modulates memory storage processes in other brain regions. By this hypothesis, changes in the functional interactions of the amygdala with other brain regions during emotional conditions should underlie, at least in part, enhanced memory for emotional material. Here we examined the influence of the human amygdala on other brain regions under emotional and nonemotional learning conditions using structural equation modeling (SEqM). Eleven male subjects received two PET scans for regional cerebral glucose metabolism— one scan while viewing a series of emotionally provocative (negative) film clips and a second scan while viewing a series of more emotionally neutral film clips. Enhanced activity in the right amygdala was related to enhanced memory for the emotional films. To identify potential candidate voxels for SEqM, the functional connectivity of the maximally activated voxel within the right amygdala was investigated using partial least squares. A subset of regions identified by this analysis showing differences functional connectivity with the amygdala between the emotional versus neutral film conditions were then submitted to SEqM, which revealed significantly increased amygdala influences on the ipsilateral parahippocampal gyrus and ventrolateral prefrontal cortex during the emotional relative to the neutral film viewing condition. These findings support the view that increased influences from the amygdala, presumably reflecting its memorymodulation function, occur during emotionally arousing learning situations. © 2003 Elsevier Inc. All rights reserved. Keywords: Amygdala; Memory; Emotion; Parahippocampal gyrus; Ventrolateral prefrontal cortex
Converging evidence from studies involving animal and human subjects indicates that the amygdala, when activated by emotional arousal, modulates memory storage processes involving other brain regions (Cahill, 2000; McGaugh et al., 2000). Consistent with this hypothesis, several functional brain imaging studies have shown that amygdala activity during encoding relates to long-term memory for emotionally arousing material but not to memory for more emotionally neutral material (Cahill et al., 1996; Hamann et al., 1999; Canli et al., 2000). Additionally, Cahill et al. (2001) found that enhanced activity of the amygdala while viewing emotional films (relative to more emotionally neutral films)
* Corresponding author. CNLM, Qureshey Lab, University of California, Irvine, CA 92697-00. Fax: ⫹1-949-824-5244. E-mail address: [email protected] (L. Cahill). 1053-8119/$ – see front matter © 2003 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2003.08.006
related significantly to enhanced long-term memory for the emotional films. Although these studies support the “memory-modulation” hypothesis of amygdala function by showing a selective role for the amygdala in enhanced memory for emotionally arousing material, they do not take advantage of the whole-brain coverage available in functional neuroimaging to determine how influences from the amygdala to other brain regions change during emotionally arousing versus neutral learning conditions. A number of studies have begun to address this aspect of the memory modulation hypothesis, showing increased functional connectivity between the amygdala and other brain regions during emotional learning situations. Zald et al. (1998) found increased correlations between the activity of the amygdala and orbitofrontal cortex during exposure to a highly aversive “sulfide cocktail” relative to control con-
2092
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
ditions. Using regression analyses, Morris et al. (1998, 1999) reported increased covariance of amygdala activity with the activity of numerous brain regions during processing of fearful stimuli relative to nonfearful stimuli. Hamann et al. (1999) investigated the intercorrelations among loci in the amygdala, hippocampus, and parahippocampal gyrus maximally correlated with long-term memory for emotional material. Significant correlations between amygdala and hippocampal/parahippocampal activity were found during emotional conditions. Although these results are consistent with a modulatory view of amygdala function, the methods used do not disentangle direct and indirect influences between regions. Furthermore, these results are equally consistent with changes in amygdala afferent, rather than efferent, influences during emotional arousal. Given the memory modulation hypothesis of amygdala function, it follows that the enhancement of memory for the emotional material seen in Cahill et al. (1996) and Cahill et al. (2001) is related to changes in the direct influence of the amygdala on other brain regions. To examine this issue, we used structural equation modeling (SEqM) to investigate changes in the influence of the amygdala on other brain regions in emotional versus neutral conditions within this data set. SEqM can disentangle direct and indirect influences as well as the direction of influence (McIntosh and Gonzalez-Lima, 1994), providing strong new evidence regarding the amygdala modulation view. Regions included in this modeling exercise were derived from an investigation of amygdala functional connectivity using partial least squares, as well as from theoretical interest. We hypothesized that influences from the amygdala to other memoryrelated structures should increase during the emotionally arousing compared with neutral film viewing conditions. In particular, we anticipated an increased amygdala influence on the posterior parahippocampal gyrus, as this structure has been linked to successful memory encoding in both glucose PET (Alkire et al., 1998) and fMRI (Brewer et al., 1998; Wagner et al., 1998) studies and because animal research indicates that the amygdala modulates memory processes in several structures within the hippocampal/parahippocampal complex (Packard et al., 1994; Packard and Teather, 1998; Roesler et al., 2002).
Materials and Methods Experimental procedures Data for the present study were taken from Cahill et al. (2001). Details of the experimental design are given in this study and in Cahill et al. (1996). Briefly, 11 healthy, righthanded males (average age ⫽ 21.5 ⫾ 0.88 years) underwent two PET scans while watching a series of film clips. Exclusionary criteria included any major medical or psychiatric illness, substance abuse, or history of head injury. The PET sessions were separated by 2–7 days. During one scan,
subjects viewed 12 emotionally arousing film clips (eliciting negative emotions such as fear and disgust); during the other scan, subjects viewed 12 similar but relatively emotionally neutral film clips. Subjects were asked to watch each film clip and then rate how emotional they found the clip to be on a scale of 0 (“not emotional at all”) to 10 (“extremely emotional”). Subjects were not told which film set they were watching, nor was any mention of a memory test made. Three weeks after the second PET session, subjects were asked to freely recall as many film clips as possible for both film sessions with no time limit. After the recall session, the subjects were debriefed as to the intent of the study. All subjects indicated that they were unaware that their memory would be tested. All subjects were recruited through campus advertisements and paid $200 for their participation. All gave informed consent in accordance with the University of California, Irvine Institutional Review Board. PET scan procedures Each film set began about 30 s before the subject was injected with 18F-fluoro-2-deoxyglucose (FDG), a glucose analog tracer used to determine regional brain glucose metabolic rate (GMR). Subjects then watched the film clips for approximately 32 min while 80 –90% of the FDG was taken up by the brain. Scanning was performed after FDG uptake with a GE2048 head-dedicated scanner (FWHM resolution about 4.5 mm in-plane). Transmission scans obtained for each subject were used for attenuation correction. Thirty overlapping axial slices parallel to the canthomeatal (CM) line were obtained at 6-mm intervals (15 slices simultaneously). Each subject wore an individually modeled thermoplastic mask to locate the CM line and hold the head still during the scanning. GMR was calculated following Sokoloff et al. (1977) in milligrams of glucose per 100 g of brain tissue per minute. Conjunction analysis The data obtained were reprocessed using the statistical parametric mapping (SPM-99) software from the Welcome Department of Cognitive Neurology, London, UK, implemented in Matlab (Mathworks, Sherborn, MA). All images were realigned, normalized into a stereotactic space, and smoothed using an (8-mm FWHM) isotropic Gaussian kernel. As in Cahill et al. (2001), a design matrix was specified using a conjunction analysis relating the difference in memory performance for the emotional versus neutral sessions to the difference in glucose metabolism for those sessions. The two conjoined components consisted of (1) the brain regions where regional relative glucose metabolism was significantly higher during the encoding of the emotional films compared with the encoding of the neutral films and (2) the brain regions where the increase in memory performance (emotional ⫺ neutral) significantly correlated with activity during encoding of the emotional films. As in the original
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
2093
Table 1 Representative voxels Region
Amygdala PHG BA 20 BA 47 BA 9/46 RSG BA 7 BA 24 BA 8
Coordinates (x, y, z, mm)
20 22 40 36 44 20 24 12 20
⫺4 ⫺54 ⫺76 34 52 ⫺52 ⫺58 32 12
⫺24 ⫺6 ⫺18 ⫺18 18 24 48 24 54
Ratio
— 1.86 3.15 2.90 2.72 ⫺2.64 ⫺2.80 ⫺2.86 ⫺3.12
Amygdala correlation
Memory correlation
Neutral
Emotional
Neutral
Emotional
— 0.03 ⫺0.05 ⫺0.49 ⫺0.71 0.18 ⫺0.10 0.58 0.44
— 0.77 0.77 0.42 ⫺0.19 ⫺0.52 ⫺0.70 ⫺0.02 ⫺0.20
⫺0.43 0.05 0.50 0.18 0.25 ⫺0.27 ⫺0.24 ⫺0.50 ⫺0.37
0.40 0.39 0.50 0.43 ⫺0.29 ⫺0.53 ⫺0.80 0.04 ⫺0.65
Note. Location (MNI coordinates) of the representative voxels associated with each selected region and the ratio of voxel salience (from the second latent variable of the PLS analysis) to standard error are reported. A positive salience indicates a more positive correlation with amygdala activity during the emotional condition relative to the neutral condition. A negative salience indicates a more negative correlation with amygdala activity during the emotional condition relative to the neutral condition. The correlations between voxel activity and memory and between voxel activity and amygdala activity as computed in the first step of the PLS analysis are also reported.
analysis, P values less than 0.05 with a small volume correction were accepted as significant. The maximally activated voxel within the amygdala obtained with this analysis ([20, ⫺4, ⫺24]) was very similar to that found in the original analysis (Cahill et al., 2001), suggesting that the finding is not overly sensitive to minute differences in preprocessing. The maximally activated voxel within the amygdala was selected as a seed voxel for use in the partial least squares analyses and SEqM. Partial least squares analysis Seed-partial least squares (Seed-PLS) was employed in an analysis of amygdala functional connectivity with the rest of the brain to identify candidate brain voxels to be submitted to SEqM. As used in neuroimaging, partial least squares is a multivariate technique that can be used to describe the relation between a set of measures (such as design contrasts, memory scores, or amygdala activity) and a set of functional brain images. A more complete mathematical description of PLS is available elsewhere (McIntosh et al., 1996). PLS was performed in three major steps using Matlab (Mathworks, Sherborn, MA). In the first step, the correlation between activity of the maximally activated voxel within the right amygdala and activity values for all other brain voxels was computed across subjects within each film condition, producing one correlation map for each condition. In the second step, singular value decomposition of these correlation maps was performed, resulting in two pairs of latent variables (LVs), or patterns of activation across the whole brain, that were related to amygdala activity. These LVs completely reproduce the cross-correlation matrix computed in the first step. The first element in each LV pair consisted of a matrix of weights that applied to amygdala activity in each condition. The second element consisted of a matrix of weights that applied to all brain voxels that were then remapped into the singular image.
Each singular image generated in this manner indicates the collection of voxels that as a group are most closely related to the corresponding amygdala activity weights (first element). The reliability of voxel saliences in the singular image was assessed by bootstrap estimation of standard errors (2000 iterations). Saliences greater than two times their standard error were considered reliable (Sampson et al., 1989). In the third step, the dot-product of the singular image and the raw images for each subject were computed to produce individual brain scores. These brain scores indicate the extent to which an individual subject’s scan reflects the pattern represented in the singular image. PLS, as applied here, enables us to derive commonalties and differences in amygdala functional connectivity between experimental conditions. To test the “memory-modulation” hypothesis, we focused on regions showing differences in amygdala functional connectivity between the emotional versus neutral film conditions. Regions with statistically reliable peak voxels were considered potential candidates to be submitted to SEqM. However, because SEqM is limited in the number of regions that can be included, only a subset of the candidate regions identified by the PLS analysis could be included in the model. The following rationales guided region selection: (1) a priori theoretical interest in a region (e.g., parahippocampal gyrus) and (2) clear anatomical evidence suggesting connections among regions. Additionally, the correlation between memory for the films and voxel activity was computed across subjects within each film condition (Table 1). Within a selected region, representative voxels with a higher correlation between activity during the emotional condition and memory for emotional material were chosen. Given our theoretical interest in including the posterior parahippocampal gyrus in the SEqM model, a voxel with the highest correlation between activity during the emotional condition and memory for the films within this region was chosen even though its salience was not quite twice the standard
2094
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
Fig. 1. Schematic representing the neuroanatomical model employed in SEqM. Approximate locations of regions are presented on a sagittal view of a glass brain. Arrows between regions represent an anatomical connection between regions. Numbers represent Brodmann’s areas; RSG, retrosplenial gyrus; PHG, parahippocampal gyrus; AMYG, amygdala.
error (1.86). Furthermore, only voxels from the right hemisphere were chosen because our focus is on the connectivity of the right hemisphere amygdala and the overwhelming majority of amygdala connections are ipsilateral (Porrino et al., 1981; Amaral and Price, 1984). Structural equation modeling As used in neuroimaging, SEqM uses two sources of information, the observed interregional correlations and the known neuroanatomical connectivity between regions (McIntosh and Gonzalez-Lima, 1994). It then computes “path coefficients” representing the direct functional influence between anatomically connected regions that best fit the observed interregional correlations. SEqM decomposes the total effects of one region on another region into direct and indirect effects by the information provided in the anatomical model. Furthermore, unlike simple pairwise correlations, SEqM incorporates the directional information provided in the anatomical model and has the ability to solve asymmetrical reciprocal relationships (McIntosh and Gonzalez-Lima, 1994). Brain regions selected to be included in the model are shown in Table 1. Interregional correlations of activity between the selected regions were computed across emotional and neutral scans and combined with a neuroanatomical model (presented in Fig. 1) to compute structural equation models using LISREL 8.3 (Scientific Software International, Inc., Lincolnwood, IL). Connections between brain regions were specified on the basis of neuroanatomical knowledge derived from primate research (Amaral and Price, 1984; Barbas and De Olmos, 1990; Carmichael and Price, 1995; Stefanacci et al., 1996; Morris et al., 1999; Petrides and Pandya, 1999; Cavada et al., 2000; Lavenex et al., 2002; Petrides and Pandya, 2002) and with the assistance of an expert anatomical consultant (J. Fallon). Residual influences were set to 0.30 for all regions. Assessment of significant differences in functional influences between the neutral and emotional conditions was done using the “stacked-model” approach (McIntosh and
Gonzalez-Lima, 1994). This approach involved comparing the fit of an alternative model with that of a null model as determined by a chi-squared difference test. In the null model, the estimate of the path coefficients was constrained to be equal between conditions. In the alternative model, the estimate of the path coefficients was allowed to differ between conditions. A significant difference in the chi-squared goodness-of-fit index allows the conclusion that a significant difference exists in the functional influences between conditions (McIntosh and Gonzalez-Lima, 1994). The statistical significance of changes in the functional influence between specific regions across conditions can be assessed in a similar manner. The null model constrains the estimate of the path coefficient of interest to be equal between conditions, whereas the alternative model allows the estimate to vary between conditions. The significance of the difference in the primary path coefficients at issue (representing amygdala influences) was assessed in this manner allowing statements regarding changes in the direct influence of the amygdala on other brain regions (see Kohler et al., 1998 for a similar rationale and procedure). Given our a priori hypothesis regarding an increased amygdala influence on the parahippocampal gyrus during the emotional condition, a P value less than 0.05 was considered significant. Bonferroni correction for multiple comparisons was employed for all other amygdala influences.
Results Behavioral results Subjects recalled an average of 2.1 (⫾ 0.6) neutral films and 5.73 (⫾ 0.8) emotional films. The average arousal rating for the neutral film clips was 2.8 (⫾ 0.5) and for the emotional film clips was 5.2 (⫾ 0.5). Subjects recalled significantly more emotional films than neutral films [t(10) ⫽ 4.1, P ⬍ .002] and rated the emotional films as significantly more arousing than the neutral films [t(10) ⫽ 3.2, P ⬍ .01]. Functional connectivity of the right amygdala The first LV produced by the seed-PLS (accounting for 81% of the cross-block covariance) revealed brain areas that showed correlations with the amygdala common to both the emotional and neutral conditions. The second LV (accounting for 19% of the cross-block covariance) revealed brain areas that showed strong changes in functional connectivity with the right amygdala between the emotional and neutral conditions (shown in Fig. 2). Results from the second LV revealed many widespread changes in functional connectivity with the right amygdala spanning both hemispheres. In general, ventral regions showed positive saliences with the second LV (i.e., more positive correlation with amygdala activity during the emotional condition relative to the neutral condition), whereas dorsal regions showed negative
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
2095
Fig. 2. Results of the seed-PLS analysis. (A) Singular image identifying regions whose covariation with amygdala activity changed during the emotional condition relative to the neutral condition (LV2). Yellow/red regions are associated with a more positive correlation with the amygdala during the emotional condition; green/blue regions are associated with a more negative correlation with the amygdala during the emotional condition. The image is displayed on a structural MRI conforming to stereotactic space. Horizontal slices start at ⫺48 mm from the anterior commissure and increase in increments of 6 mm to ⫹72 mm. Left is left and the top is anterior. (B) Scatterplot of adjusted amygdala activity with latent variable scores for each subject during the emotional and neutral conditions. The r value in each plot is the correlation between amygdala activity and LV scores.
saliences (i.e., more negative correlation with amygdala activity during the emotional condition relative to the neutral condition). Structural equation modeling To investigate changes in the influence of the right amygdala on other brain regions during the emotional con-
dition relative to the neutral condition, SEqM was performed on a subset of regions identified in the second LV (listed in Table 1). The functional network significantly differed between conditions [2diff(27) ⫽ 89.74, P ⬍ 0.01]. The functional network models from the emotional and neutral scans are depicted in Fig. 3. To investigate changes in amygdala influence on other brain regions differences in individual path coefficients representing amygdala influence
2096
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
Fig. 3. Functional network models from the emotional and neutral scans. Arrow thickness reflects the functional influence of one region onto another through their anatomical connection as indicated in the legend. Significant changes between emotional and neutral conditions in functional influence of the amygdala on other brain regions are in bold.
between conditions were assessed. The influence of the amygdala on the parahippocampal gyrus [2diff(1) ⫽ 3.88, P ⬍ 0.05, uncorrected] showed a significant increase from the neutral to emotional film viewing conditions. The influence of the amygdala on the ventrolateral prefrontal cortex (BA 47) [2diff(1) ⫽ 14.71, P ⬍ 0.05, corrected] showed a significant change from a negative influence in the neutral condition to a positive influence in the emotional condition. The influence of the amygdala on both the cingulate gyrus [2diff(1) ⫽ 8.24, P ⬍ 0.05, corrected] and on the dorsolateral prefrontal cortex (BA 9/46) [2diff(1) ⫽ 11.43, P ⬍ 0.05, corrected] showed a significant decrease to essentially zero in the emotional condition; however, the nature of the decrease differed. For the cingulate gyrus, amygdala influence decreased from a positive influence in the neutral condition. For the dorsolateral prefrontal cortex, amygdala influence decreased from a negative influence in the neutral condition. The influence of the amygdala on other regions did not show significant changes (P ⬎ 0.05, corrected). Path coefficients representing amygdala influences showing a significant difference between emotional and neutral conditions are shown in Table 2.
Discussion This experiment tested the hypothesis that increased efferent influences from the amygdala to other brain regions, and in particular to the parahippocampal region, will occur in emotionally arousing compared with emotionally neutral learning conditions. Consistent with this hypothesis, results Table 2 Path coefficients representing amygdala influences showing a significant difference between the emotional and neutral conditions Amygdala influence onto
PHG 47 9/46 24
Path coefficient Neutral
Emotional
0.01 ⫺0.30 ⫺0.81 0.86
0.89 0.67 ⫺0.05 ⫺0.12
from an SEqM analysis of PET data revealed a significantly more positive amygdala influence on two brain regions— the parahippocampal gyrus and Brodmann’s area 47— during the emotional relative to the neutral film viewing conditions. The findings constitute the first evidence that increased outflow from the amygdala, presumably reflecting its well-established memory-modulating ability, occurs in the human brain during periods of heightened emotional arousal. In SEqM, a path coefficient represents the expected change in the activity of one region given a unit change in the region influencing it, assuming no other influencing regions show any change (McIntosh and Gonzalez-Lima, 1994). Therefore, the results are consistent with the possibility that (1) an increase in amygdala activity during the neutral condition would directly lead to an increase in cingulate activity, a decrease in dorsolateral prefrontal cortex activity, a smaller decrease in ventrolateral prefrontal cortex, and very little change in parahippocampal gyrus activity; and (2) in contrast, an increase in amygdala activity during the emotional condition would lead to an increase in ventrolateral and parahippocampal activity and very little change in cingulate gyrus and dorsolateral prefrontal cortex activity. SEqM is a powerful analytical technique that goes beyond simply correlating activity between brain regions to illuminating how such correlations may be mediated. To date, there has been only one other research investigation that used SEqM to examine the influence of the amygdala on other brain regions during emotional conditions (Iidaka et al., 2001). However, these authors did not use the “stacked model” approach whereby the patterns of influence in the two conditions can be directly compared; instead, they considered each condition individually. Consequently, although the path coefficients appear to be different in the two conditions in Iidaka et al. (2001), the question of whether they indeed were different was not actually tested. In contrast, the present experiment directly examined this key issue. A further strength of the present experiment in addressing the memory-modulation hypothesis of amygdala function is that memory measures were taken. The analysis in
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
Iidaka et al. (2001) was based on the area of the amygdala maximally activated during their emotional condition. In contrast, the present experiment investigated the changes in influence of the particular region within the amygdala maximally related to enhanced memory for emotional material, an approach more pertinent to the memory-modulation hypothesis. Amygdala influence on activity in the parahippocampal gyrus increased significantly during the emotional relative to neutral film viewing conditions. Several studies have linked activity in the parahippocampal gyrus to subsequent memory in a variety of paradigms (Alkire et al., 1998; Brewer et al., 1998; Wagner et al., 1998). Alkire et al. (1998) found that activity in the posterior parahippocampal gyrus during encoding correlated very highly (r ⫽ 0.98) with long-term free recall of nonemotional words in a longterm incidental memory test. Both Brewer et al. (1998) and Wagner et al. (1998) found that parahippocampal activity distinguished subsequently remembered versus forgotten stimuli. Consistent with these human brain imaging findings, animal research (particularly involving primates) has heavily implicated the parahippocampal region in “declarative” memory formation (Squire and Zola, 1996). Finally, research with rats demonstrates that the amygdala influences memory-related activities in various portions of the hippocampal/parahippocampal complex (Packard et al., 1994; Packard and Teather, 1998; Roesler et al., 2002). Collectively, these findings make the parahippocampal region the prime candidate on theoretical grounds to receive a heightened influence from the amygdala during an emotionally arousing learning situation. The present findings add new support to this view. Amygdala influence on the ventrolateral prefrontal cortex (Brodmann’s area 47) changed significantly from a negative influence in the neutral condition to a positive influence in the emotional film viewing condition. The ventrolateral prefrontal cortex has also been associated with successful subsequent memory. Both Wagner et al. (1998) and Alkire et al. (1998) found that activity in the ventrolateral prefrontal cortex was higher during encoding of words that were subsequently remembered than during encoding of words that were forgotten. Increased activity in the ventrolateral prefrontal cortex has also been found during a deep encoding task relative to a shallow encoding task (Kapur et al., 1994). Amygdala influence on the cingulate gyrus and dorsolateral prefrontal cortex (Brodmann’s area 9/46) decreased significantly in emotional relative to neutral film conditions. The inclusion of these regions was based on the results of the seed-PLS analysis of right amygdala functional connectivity. Although these regions showed strong changes in functional connectivity with the right amygdala between emotional film conditions, their activity did not show a strong correlation with memory in the emotional condition (see Table 1). In fact, these two regions had the smallest correlation between activity during the emotional condition
2097
and memory for the emotional films, as well as the smallest correlation with amygdala activity during the emotional condition of all the regions included in the anatomical model. These observations suggest that not all brain regions will receive a heightened amygdala influence during emotional conditions. It is conceivable that some brain regions receive a tonic influence from the amygdala in neutral conditions that is reduced during periods of emotional arousal. Possibilities such as these deserve more attention in future studies of amygdala effective connectivity. Future investigations will also need to address potential sex-related differences in amygdala influences on other brain regions. Two recent investigations reported a sexrelated hemispheric lateralization of amygdala involvement in emotionally influenced memory (Cahill et al., 2001; Canli et al., 2002). In both, enhanced activity of the right, but not the left, amygdala in men related significantly to enhanced memory for emotional material. Conversely, enhanced activity of the left, but not the right, amygdala in women related significantly to enhanced memory for emotional material. We focused on activity in males for our initial efforts to model amygdala function for two reasons. First, the significance in our previous PET study of the relationship between amygdala activity and memory for the emotional films was higher in males than it was in females; second, the location of the region of maximal correlation with memory for the emotional films was more directly located within the basolateral nucleus in males than it was in females in that study (Cahill et al., 2001). This latter fact is highly important because it is the basolateral nucleus that is most strongly implicated by animal research in memory modulation (Cahill and McGaugh, 1998). However, the question of whether similar changes in amygdala effective connectivity will be found in women (presumably involving the left amygdala) remains, in our view, a highly important question to address. Indeed, we are presently investigating sex-related differences in amygdala function using an eventrelated fMRI paradigm in which the peak voxel within the left amygdala in women and the peak voxel within the right amygdala in men relating to memory for emotional material were previously found to be comparably located within the basolateral nucleus complex (Canli et al., 2002). These data should allow for a more powerful comparison of amygdala effective connectivity in men and women than is possible from our existing PET data set. Further improvements upon the current study afforded by the on-going event-related fMRI investigation include the possibility for a larger sample size than that used in the present study. It is possible that our findings may have been produced without the presence of item-specific correlations between brain regions. However, it is not necessary for there to be item-specific correlations between, for example, the amygdala and parahippocampal gyrus in order for a modulatory function of the amygdala on this region to exist (McIntosh, 1999). The amygdala may modulate parahippocampal activity in a more global rather than item-specific fashion.
2098
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
Indeed, given the relative slowness of consolidation processes (which clearly last many hours at minimum), it is entirely possible that glucose PET may prove more sensitive to at least some aspects of amygdala-based modulatory processes than will event-related fMRI. It should be noted that although enhanced outflow from the amygdala in emotional conditions seems likely to reflect its role in modulating memory consolidation as established by animal research (McGaugh et al., 2000), it may also reflect, at least in part, hypothesized amygdala influences on arousal and/or attentional processes with emotion (e.g., Kapp et al., 1992; Anderson and Phelps, 2001). The present data cannot distinguish these possibilities. Regarding the prefrontal cortex, it is possible that the heightened amygdala influence with emotion reflects a taking “off-line” of this structure by the amygdala during a period of emotional stress (Arnsten, 1998). Recently Keightley et al. (2003) reported results from a seed-PLS analysis of amygdala function during emotional picture viewing conditions, which revealed generally positive correlations with the right hemisphere amygdala and more dorsally located regions during viewing of positive and negative stimuli. In contrast, the present seed-PLS analysis revealed more negative correlations with the right amygdala and dorsally located regions during viewing of emotionally valenced films. The reasons for this seeming disparity between the two seed-PLS analyses of right amygdala function are not clear, but several possibilities can be considered. First, Keightley et al. (2003) combined data from men and women, whereas the present analysis involved only men. Also, Keightley et al. (2003) investigated commonalities in amygdala functional connectivity across positive and negative emotional conditions absent a neutral condition, whereas the present experiment compared amygdala functional connectivity in neutral versus negative emotional conditions, and with reference to long-term memory for the material. Despite their specific differences, both studies indicate that the functional connectivity of the human amygdala is influenced by the emotional state. Concluding remarks Structural equation modeling of PET data revealed heightened amygdala outflow during emotional compared with neutral learning conditions, most notably to the parahippocampal gyrus. These findings provide new support for the view that strong memory for emotionally arousing events results, at least in part, from modulatory influences from the amygdala on memory storage processes occurring in other brain regions.
Acknowledgments We greatly appreciate the expert and detailed anatomical advice offered by James Fallon in the creation of the neuroanatomical model and our wonderful path analysis con-
sultant Randy McIntosh. Funded by NIMH Grant MH57508 to L.C.
References Alkire, M.T., Haier, R.J., Fallon, J.H., Cahill, L., 1998. Hippocampal, but not amygdala, activity at encoding correlates with long-term, free recall of nonemotional information. Proc. Natl. Acad. Sci. USA 95 (24), 14506 –14510. Amaral, D.G., Price, J.L., 1984. Amygdalo-cortical projections in the monkey (Macaca fascicularis). J. Comp. Neurol. 230 (4), 465– 496. Anderson, A.K., Phelps, E.A., 2001. Lesions of the human amygdala impair enhanced perception of emotionally salient events. Nature 411 (6835), 305–309. Arnsten, A.F.T., 1998. Catecholamine modulation of prefrontal cortical cognitive function. Trends Cogn. Sci. 2 (11), 436 – 447. Barbas, H., De Olmos, J., 1990. Projections from the amygdala to basoventral and mediodorsal prefrontal regions in the rhesus monkey. J. Comp. Neurol. 300, 549 –571. Brewer, J.B., Zhao, Z., Desmond, J.E., Glover, G.H., Gabrieli, J.D., 1998. Making memories: brain activity that predicts how well visual experience will be remembered. Science 281 (5380), 1185–1187. Cahill, L., 2000. Modulation of long-term memory storage in humans by emotional arousal: adrenergic activation and the amygdala, in: Aggleton, J.P. (Ed.), The Amygdala: A Functional Analysis, Oxford University Press, New York, pp. 425– 445. Cahill, L., Haier, R.J., Fallon, J., Alkire, M.T., Tang, C., Keator, D., Wu, J., McGaugh, J.L., 1996. Amygdala activity at encoding correlated with long-term, free recall of emotional information. Proc. Natl. Acad. Sci. USA 93 (15), 8016 – 8021. Cahill, L., Haier, R.J., White, N.S., Fallon, J., Kilpatrick, L., Lawrence, C., Potkin, S.G., Alkire, M.T., 2001. Sex-related difference in amygdala activity during emotionally influenced memory storage. Neurobiol. Learn. Mem. 75 (1), 1–9. Cahill, L., McGaugh, J.L., 1998. Mechanisms of emotional arousal and lasting declarative memory. Trends Neurosci. 21 (7), 294 –299. Canli, T., Desmond, J.E., Zhao, Z., Gabrieli, J.D., 2002. Sex differences in the neural basis of emotional memories. Proc. Natl. Acad. Sci. USA 99 (16), 10789 –10794. Canli, T., Zhao, Z., Brewer, J., Gabrieli, J.D., Cahill, L., 2000. Eventrelated activation in the human amygdala associates with later memory for individual emotional experience. J. Neurosci. 20 (19), RC99. Carmichael, S.T., Price, J.L., 1995. Limbic connections of the orbital and medial prefrontal cortex in macaque monkeys. J. Comp. Neurol. 363, 615– 641. Cavada, C., Company, T., Tejedor, J., Cruz-Rizzolo, R.J., Reinoso-Suarez, F., 2000. The anatomical connections of the macaque monkey orbitofrontal cortex: a review. Cereb. Cortex 10 (3), 220 –242. Hamann, S.B., Ely, T.D., Grafton, S.T., Kilts, C.D., 1999. Amygdala activity related to enhanced memory for pleasant and aversive stimuli. Nat. Neurosci. 2 (3), 289 –293. Iidaka, T., Omori, M., Murata, T., Kosaka, H., Yonekura, Y., Okada, T., Sadato, N., 2001. Neural interaction of the amygdala with the prefrontal and temporal cortices in the processing of facial expressions as revealed by fMRI. J. Cogn. Neurosci. 13 (8), 1035– 47. Kapp, B.S., Whalen, P.J., Supple, W.F., Pascoe, J.P., 1992. Amygdaloid contributions to conditioned arousal and sensory information processing, in: Aggleton, J.P. (Ed.), The Amygdala: Neurobiological Aspects of Emotion, Memory, and Mental Dysfunction, Wiley–Liss, Inc., New York. Kapur, S., Craik, F.I., Tulving, E., Wilson, A.A., Houle, S., Brown, G.M., 1994. Neuroanatomical correlates of encoding in episodic memory: levels of processing effect. Proc. Natl. Acad. Sci. USA 91 (6), 2008 – 2011.
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099 Keightley, M.L., Winocur, G., Graham, S.J., Mayberg, H.S., Hevenor, S.J., Grady, C.L., 2003. An fMRI study investigating cognitive modulation of brain regions associated with emotional processing of visual stimuli. Neuropsychologia 41 (5), 585–596. Kohler, S., McIntosh, A.R., Moscovitch, M., Winocur, G., 1998. Functional interactions between the medial temporal lobes and posterior neocortex related to episodic memory retrieval. Cereb. Cortex 8 (5), 451– 461. Lavenex, P., Suzuki, W.A., Amaral, D.G., 2002. Perirhinal and parahippocampal cortices of the macaque monkey: projections to the neocortex. J. Comp. Neurol. 447 (4), 394 – 420. McGaugh, J.L., Ferry, B., Vazdarjanova, A., Roozendaal, B., 2000. Amygdala: role in modulation of memory storage, in: Aggleton, J.P. (Ed.), The Amygdala: A Functional Analysis, Oxford University Press, New York, pp. 391– 424. McIntosh, A.R., 1999. Mapping cognition to the brain through neural interactions. Memory 7 (5– 6), 523–548. McIntosh, A.R., Bookstein, F.L., Haxby, J.V., Grady, C.L., 1996. Spatial pattern analysis of functional brain images using partial least squares. NeuroImage 3 (31), 143–157. McIntosh, A.R., Gonzalez-Lima, F., 1994. Structural equation modeling and its application to network analysis in functional brain imaging. Hum. Brain Mapp. 2, 2–22. Morris, J.S., Friston, K.J., Buchel, C., Frith, C.D., Young, A.W., Calder, A.J., Dolan, R.J., 1998. A neuromodulatory role for the human amygdala in processing emotional facial expressions. Brain 121 (1), 47–57. Morris, J.S., Ohman, A., Dolan, R.J., 1999. A subcortical pathway to the right amygdala mediating “unseen” fear. Proc. Natl. Acad. Sci. USA 96 (4), 1680 –1685. Morris, R., Petrides, M., Pandya, D.N., 1999. Architecture and connections of retrosplenial area 30 in the rhesus monkey (Macaca mulatta). Eur. J. Neurosci. 11 (7), 2506 –2518. Packard, M.G., Cahill, L., McGaugh, J.L., 1994. Amygdala modulation of hippocampal-dependent and caudate nucleus-dependent memory processes. Proc. Natl. Acad. Sci. USA 91 (18), 8477– 8481. Packard, M.G., Teather, L.A., 1998. Amygdala modulation of multiple memory systems: hippocampus and caudate-putamen. Neurobiol. Learn. Mem. 69 (2), 163–203.
2099
Petrides, M., Pandya, D.N., 1999. Dorsolateral prefrontal cortex: comparative cytoarchitectonic analysis in the human and the macaque brain and corticocortical connection patterns. Eur. J. Neurosci. 11 (3), 1011– 1036. Petrides, M., Pandya, D.N., 2002. Comparative cytoarchitectonic analysis of the human and the macaque ventrolateral prefrontal cortex and corticocortical connection patterns in the monkey. Eur. J. Neurosci. 16 (2), 291–310. Porrino, L.J., Crane, A.M., Goldman-Rakic, P.S., 1981. Direct and indirect pathways from the amygdala to the frontal lobe in rhesus monkeys. J. Comp. Neurol. 198 (1), 121–136. Roesler, R., Roozendaal, B., McGaugh, J.L., 2002. Basolateral amygdala lesions block the memory-enhancing effect of 8-Br-cAMP infused into the entorhinal cortex of rats after training. Eur. J. Neurosci. 15 (5), 905–910. Sampson, P.D., Streissguth, A.P., Barr, H.M., Bookstein, F.L., 1989. Neurobehavioral effects of prenatal alcohol. II. Partial least squares analysis. Neurotoxicol. Teratol. 11 (5), 477– 491. Sokoloff, L., Reivich, M., Kennedy, C., DesRosiers, M., Patlak, D., Pettigrew, K., Sakurada, O., Shinohora, M., 1977. The [14C] deoxyglucose method for the measurement of local cerebral glucose utilization: theory, procedure, and normal values in the conscious and anesthetized albino male rat. J. Neurochem. 28, 897–916. Squire, L.R., Zola, S.M., 1996. Structure and function of declarative and nondeclarative memory systems. Proc. Natl. Acad. Sci. USA 93 (24), 13515–13522. Stefanacci, L., Suzuki, W.A., Amaral, D.G., 1996. Organization of connections between the amygdaloid complex and the perirhinal and parahippocampal cortices in macaque monkeys. J. Comp. Neurol. 375 (4), 552–582. Wagner, A.D., Schacter, D.L., Rotte, M., Koutstaal, W., Maril, A., Dale, A.M., Rosen, B.R., Buckner, R.L., 1998. Building memories: remembering and forgetting of verbal experiences as predicted by brain activity. Science 281 (5380), 1188 –1191. Zald, D.H., Donndelinger, M.J., Pardo, J.V., 1998. Elucidating dynamic brain interactions with across-subjects correlational analyses of positron emission tomographic data: the functional connectivity of the amygdala and orbitofrontal cortex during olfactory tasks. J. Cereb. Blood Flow Metab. 18 (8), 896 –905.
www.elsevier.com/locate/ynimg
Amygdala modulation of parahippocampal and frontal regions during emotionally influenced memory storage Lisa Kilpatrick and Larry Cahill* Department of Neurobiology and Behavior and Center for the Neurobiology of Learning and Memory, University of California, Irvine, CA 92697-3800, USA Received 8 April 2003; revised 30 July 2003; accepted 4 August 2003
Abstract Considerable evidence from both animal and human subject research supports the hypothesis that the amygdala, when activated by emotional arousal, modulates memory storage processes in other brain regions. By this hypothesis, changes in the functional interactions of the amygdala with other brain regions during emotional conditions should underlie, at least in part, enhanced memory for emotional material. Here we examined the influence of the human amygdala on other brain regions under emotional and nonemotional learning conditions using structural equation modeling (SEqM). Eleven male subjects received two PET scans for regional cerebral glucose metabolism— one scan while viewing a series of emotionally provocative (negative) film clips and a second scan while viewing a series of more emotionally neutral film clips. Enhanced activity in the right amygdala was related to enhanced memory for the emotional films. To identify potential candidate voxels for SEqM, the functional connectivity of the maximally activated voxel within the right amygdala was investigated using partial least squares. A subset of regions identified by this analysis showing differences functional connectivity with the amygdala between the emotional versus neutral film conditions were then submitted to SEqM, which revealed significantly increased amygdala influences on the ipsilateral parahippocampal gyrus and ventrolateral prefrontal cortex during the emotional relative to the neutral film viewing condition. These findings support the view that increased influences from the amygdala, presumably reflecting its memorymodulation function, occur during emotionally arousing learning situations. © 2003 Elsevier Inc. All rights reserved. Keywords: Amygdala; Memory; Emotion; Parahippocampal gyrus; Ventrolateral prefrontal cortex
Converging evidence from studies involving animal and human subjects indicates that the amygdala, when activated by emotional arousal, modulates memory storage processes involving other brain regions (Cahill, 2000; McGaugh et al., 2000). Consistent with this hypothesis, several functional brain imaging studies have shown that amygdala activity during encoding relates to long-term memory for emotionally arousing material but not to memory for more emotionally neutral material (Cahill et al., 1996; Hamann et al., 1999; Canli et al., 2000). Additionally, Cahill et al. (2001) found that enhanced activity of the amygdala while viewing emotional films (relative to more emotionally neutral films)
* Corresponding author. CNLM, Qureshey Lab, University of California, Irvine, CA 92697-00. Fax: ⫹1-949-824-5244. E-mail address: [email protected] (L. Cahill). 1053-8119/$ – see front matter © 2003 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2003.08.006
related significantly to enhanced long-term memory for the emotional films. Although these studies support the “memory-modulation” hypothesis of amygdala function by showing a selective role for the amygdala in enhanced memory for emotionally arousing material, they do not take advantage of the whole-brain coverage available in functional neuroimaging to determine how influences from the amygdala to other brain regions change during emotionally arousing versus neutral learning conditions. A number of studies have begun to address this aspect of the memory modulation hypothesis, showing increased functional connectivity between the amygdala and other brain regions during emotional learning situations. Zald et al. (1998) found increased correlations between the activity of the amygdala and orbitofrontal cortex during exposure to a highly aversive “sulfide cocktail” relative to control con-
2092
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
ditions. Using regression analyses, Morris et al. (1998, 1999) reported increased covariance of amygdala activity with the activity of numerous brain regions during processing of fearful stimuli relative to nonfearful stimuli. Hamann et al. (1999) investigated the intercorrelations among loci in the amygdala, hippocampus, and parahippocampal gyrus maximally correlated with long-term memory for emotional material. Significant correlations between amygdala and hippocampal/parahippocampal activity were found during emotional conditions. Although these results are consistent with a modulatory view of amygdala function, the methods used do not disentangle direct and indirect influences between regions. Furthermore, these results are equally consistent with changes in amygdala afferent, rather than efferent, influences during emotional arousal. Given the memory modulation hypothesis of amygdala function, it follows that the enhancement of memory for the emotional material seen in Cahill et al. (1996) and Cahill et al. (2001) is related to changes in the direct influence of the amygdala on other brain regions. To examine this issue, we used structural equation modeling (SEqM) to investigate changes in the influence of the amygdala on other brain regions in emotional versus neutral conditions within this data set. SEqM can disentangle direct and indirect influences as well as the direction of influence (McIntosh and Gonzalez-Lima, 1994), providing strong new evidence regarding the amygdala modulation view. Regions included in this modeling exercise were derived from an investigation of amygdala functional connectivity using partial least squares, as well as from theoretical interest. We hypothesized that influences from the amygdala to other memoryrelated structures should increase during the emotionally arousing compared with neutral film viewing conditions. In particular, we anticipated an increased amygdala influence on the posterior parahippocampal gyrus, as this structure has been linked to successful memory encoding in both glucose PET (Alkire et al., 1998) and fMRI (Brewer et al., 1998; Wagner et al., 1998) studies and because animal research indicates that the amygdala modulates memory processes in several structures within the hippocampal/parahippocampal complex (Packard et al., 1994; Packard and Teather, 1998; Roesler et al., 2002).
Materials and Methods Experimental procedures Data for the present study were taken from Cahill et al. (2001). Details of the experimental design are given in this study and in Cahill et al. (1996). Briefly, 11 healthy, righthanded males (average age ⫽ 21.5 ⫾ 0.88 years) underwent two PET scans while watching a series of film clips. Exclusionary criteria included any major medical or psychiatric illness, substance abuse, or history of head injury. The PET sessions were separated by 2–7 days. During one scan,
subjects viewed 12 emotionally arousing film clips (eliciting negative emotions such as fear and disgust); during the other scan, subjects viewed 12 similar but relatively emotionally neutral film clips. Subjects were asked to watch each film clip and then rate how emotional they found the clip to be on a scale of 0 (“not emotional at all”) to 10 (“extremely emotional”). Subjects were not told which film set they were watching, nor was any mention of a memory test made. Three weeks after the second PET session, subjects were asked to freely recall as many film clips as possible for both film sessions with no time limit. After the recall session, the subjects were debriefed as to the intent of the study. All subjects indicated that they were unaware that their memory would be tested. All subjects were recruited through campus advertisements and paid $200 for their participation. All gave informed consent in accordance with the University of California, Irvine Institutional Review Board. PET scan procedures Each film set began about 30 s before the subject was injected with 18F-fluoro-2-deoxyglucose (FDG), a glucose analog tracer used to determine regional brain glucose metabolic rate (GMR). Subjects then watched the film clips for approximately 32 min while 80 –90% of the FDG was taken up by the brain. Scanning was performed after FDG uptake with a GE2048 head-dedicated scanner (FWHM resolution about 4.5 mm in-plane). Transmission scans obtained for each subject were used for attenuation correction. Thirty overlapping axial slices parallel to the canthomeatal (CM) line were obtained at 6-mm intervals (15 slices simultaneously). Each subject wore an individually modeled thermoplastic mask to locate the CM line and hold the head still during the scanning. GMR was calculated following Sokoloff et al. (1977) in milligrams of glucose per 100 g of brain tissue per minute. Conjunction analysis The data obtained were reprocessed using the statistical parametric mapping (SPM-99) software from the Welcome Department of Cognitive Neurology, London, UK, implemented in Matlab (Mathworks, Sherborn, MA). All images were realigned, normalized into a stereotactic space, and smoothed using an (8-mm FWHM) isotropic Gaussian kernel. As in Cahill et al. (2001), a design matrix was specified using a conjunction analysis relating the difference in memory performance for the emotional versus neutral sessions to the difference in glucose metabolism for those sessions. The two conjoined components consisted of (1) the brain regions where regional relative glucose metabolism was significantly higher during the encoding of the emotional films compared with the encoding of the neutral films and (2) the brain regions where the increase in memory performance (emotional ⫺ neutral) significantly correlated with activity during encoding of the emotional films. As in the original
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
2093
Table 1 Representative voxels Region
Amygdala PHG BA 20 BA 47 BA 9/46 RSG BA 7 BA 24 BA 8
Coordinates (x, y, z, mm)
20 22 40 36 44 20 24 12 20
⫺4 ⫺54 ⫺76 34 52 ⫺52 ⫺58 32 12
⫺24 ⫺6 ⫺18 ⫺18 18 24 48 24 54
Ratio
— 1.86 3.15 2.90 2.72 ⫺2.64 ⫺2.80 ⫺2.86 ⫺3.12
Amygdala correlation
Memory correlation
Neutral
Emotional
Neutral
Emotional
— 0.03 ⫺0.05 ⫺0.49 ⫺0.71 0.18 ⫺0.10 0.58 0.44
— 0.77 0.77 0.42 ⫺0.19 ⫺0.52 ⫺0.70 ⫺0.02 ⫺0.20
⫺0.43 0.05 0.50 0.18 0.25 ⫺0.27 ⫺0.24 ⫺0.50 ⫺0.37
0.40 0.39 0.50 0.43 ⫺0.29 ⫺0.53 ⫺0.80 0.04 ⫺0.65
Note. Location (MNI coordinates) of the representative voxels associated with each selected region and the ratio of voxel salience (from the second latent variable of the PLS analysis) to standard error are reported. A positive salience indicates a more positive correlation with amygdala activity during the emotional condition relative to the neutral condition. A negative salience indicates a more negative correlation with amygdala activity during the emotional condition relative to the neutral condition. The correlations between voxel activity and memory and between voxel activity and amygdala activity as computed in the first step of the PLS analysis are also reported.
analysis, P values less than 0.05 with a small volume correction were accepted as significant. The maximally activated voxel within the amygdala obtained with this analysis ([20, ⫺4, ⫺24]) was very similar to that found in the original analysis (Cahill et al., 2001), suggesting that the finding is not overly sensitive to minute differences in preprocessing. The maximally activated voxel within the amygdala was selected as a seed voxel for use in the partial least squares analyses and SEqM. Partial least squares analysis Seed-partial least squares (Seed-PLS) was employed in an analysis of amygdala functional connectivity with the rest of the brain to identify candidate brain voxels to be submitted to SEqM. As used in neuroimaging, partial least squares is a multivariate technique that can be used to describe the relation between a set of measures (such as design contrasts, memory scores, or amygdala activity) and a set of functional brain images. A more complete mathematical description of PLS is available elsewhere (McIntosh et al., 1996). PLS was performed in three major steps using Matlab (Mathworks, Sherborn, MA). In the first step, the correlation between activity of the maximally activated voxel within the right amygdala and activity values for all other brain voxels was computed across subjects within each film condition, producing one correlation map for each condition. In the second step, singular value decomposition of these correlation maps was performed, resulting in two pairs of latent variables (LVs), or patterns of activation across the whole brain, that were related to amygdala activity. These LVs completely reproduce the cross-correlation matrix computed in the first step. The first element in each LV pair consisted of a matrix of weights that applied to amygdala activity in each condition. The second element consisted of a matrix of weights that applied to all brain voxels that were then remapped into the singular image.
Each singular image generated in this manner indicates the collection of voxels that as a group are most closely related to the corresponding amygdala activity weights (first element). The reliability of voxel saliences in the singular image was assessed by bootstrap estimation of standard errors (2000 iterations). Saliences greater than two times their standard error were considered reliable (Sampson et al., 1989). In the third step, the dot-product of the singular image and the raw images for each subject were computed to produce individual brain scores. These brain scores indicate the extent to which an individual subject’s scan reflects the pattern represented in the singular image. PLS, as applied here, enables us to derive commonalties and differences in amygdala functional connectivity between experimental conditions. To test the “memory-modulation” hypothesis, we focused on regions showing differences in amygdala functional connectivity between the emotional versus neutral film conditions. Regions with statistically reliable peak voxels were considered potential candidates to be submitted to SEqM. However, because SEqM is limited in the number of regions that can be included, only a subset of the candidate regions identified by the PLS analysis could be included in the model. The following rationales guided region selection: (1) a priori theoretical interest in a region (e.g., parahippocampal gyrus) and (2) clear anatomical evidence suggesting connections among regions. Additionally, the correlation between memory for the films and voxel activity was computed across subjects within each film condition (Table 1). Within a selected region, representative voxels with a higher correlation between activity during the emotional condition and memory for emotional material were chosen. Given our theoretical interest in including the posterior parahippocampal gyrus in the SEqM model, a voxel with the highest correlation between activity during the emotional condition and memory for the films within this region was chosen even though its salience was not quite twice the standard
2094
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
Fig. 1. Schematic representing the neuroanatomical model employed in SEqM. Approximate locations of regions are presented on a sagittal view of a glass brain. Arrows between regions represent an anatomical connection between regions. Numbers represent Brodmann’s areas; RSG, retrosplenial gyrus; PHG, parahippocampal gyrus; AMYG, amygdala.
error (1.86). Furthermore, only voxels from the right hemisphere were chosen because our focus is on the connectivity of the right hemisphere amygdala and the overwhelming majority of amygdala connections are ipsilateral (Porrino et al., 1981; Amaral and Price, 1984). Structural equation modeling As used in neuroimaging, SEqM uses two sources of information, the observed interregional correlations and the known neuroanatomical connectivity between regions (McIntosh and Gonzalez-Lima, 1994). It then computes “path coefficients” representing the direct functional influence between anatomically connected regions that best fit the observed interregional correlations. SEqM decomposes the total effects of one region on another region into direct and indirect effects by the information provided in the anatomical model. Furthermore, unlike simple pairwise correlations, SEqM incorporates the directional information provided in the anatomical model and has the ability to solve asymmetrical reciprocal relationships (McIntosh and Gonzalez-Lima, 1994). Brain regions selected to be included in the model are shown in Table 1. Interregional correlations of activity between the selected regions were computed across emotional and neutral scans and combined with a neuroanatomical model (presented in Fig. 1) to compute structural equation models using LISREL 8.3 (Scientific Software International, Inc., Lincolnwood, IL). Connections between brain regions were specified on the basis of neuroanatomical knowledge derived from primate research (Amaral and Price, 1984; Barbas and De Olmos, 1990; Carmichael and Price, 1995; Stefanacci et al., 1996; Morris et al., 1999; Petrides and Pandya, 1999; Cavada et al., 2000; Lavenex et al., 2002; Petrides and Pandya, 2002) and with the assistance of an expert anatomical consultant (J. Fallon). Residual influences were set to 0.30 for all regions. Assessment of significant differences in functional influences between the neutral and emotional conditions was done using the “stacked-model” approach (McIntosh and
Gonzalez-Lima, 1994). This approach involved comparing the fit of an alternative model with that of a null model as determined by a chi-squared difference test. In the null model, the estimate of the path coefficients was constrained to be equal between conditions. In the alternative model, the estimate of the path coefficients was allowed to differ between conditions. A significant difference in the chi-squared goodness-of-fit index allows the conclusion that a significant difference exists in the functional influences between conditions (McIntosh and Gonzalez-Lima, 1994). The statistical significance of changes in the functional influence between specific regions across conditions can be assessed in a similar manner. The null model constrains the estimate of the path coefficient of interest to be equal between conditions, whereas the alternative model allows the estimate to vary between conditions. The significance of the difference in the primary path coefficients at issue (representing amygdala influences) was assessed in this manner allowing statements regarding changes in the direct influence of the amygdala on other brain regions (see Kohler et al., 1998 for a similar rationale and procedure). Given our a priori hypothesis regarding an increased amygdala influence on the parahippocampal gyrus during the emotional condition, a P value less than 0.05 was considered significant. Bonferroni correction for multiple comparisons was employed for all other amygdala influences.
Results Behavioral results Subjects recalled an average of 2.1 (⫾ 0.6) neutral films and 5.73 (⫾ 0.8) emotional films. The average arousal rating for the neutral film clips was 2.8 (⫾ 0.5) and for the emotional film clips was 5.2 (⫾ 0.5). Subjects recalled significantly more emotional films than neutral films [t(10) ⫽ 4.1, P ⬍ .002] and rated the emotional films as significantly more arousing than the neutral films [t(10) ⫽ 3.2, P ⬍ .01]. Functional connectivity of the right amygdala The first LV produced by the seed-PLS (accounting for 81% of the cross-block covariance) revealed brain areas that showed correlations with the amygdala common to both the emotional and neutral conditions. The second LV (accounting for 19% of the cross-block covariance) revealed brain areas that showed strong changes in functional connectivity with the right amygdala between the emotional and neutral conditions (shown in Fig. 2). Results from the second LV revealed many widespread changes in functional connectivity with the right amygdala spanning both hemispheres. In general, ventral regions showed positive saliences with the second LV (i.e., more positive correlation with amygdala activity during the emotional condition relative to the neutral condition), whereas dorsal regions showed negative
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
2095
Fig. 2. Results of the seed-PLS analysis. (A) Singular image identifying regions whose covariation with amygdala activity changed during the emotional condition relative to the neutral condition (LV2). Yellow/red regions are associated with a more positive correlation with the amygdala during the emotional condition; green/blue regions are associated with a more negative correlation with the amygdala during the emotional condition. The image is displayed on a structural MRI conforming to stereotactic space. Horizontal slices start at ⫺48 mm from the anterior commissure and increase in increments of 6 mm to ⫹72 mm. Left is left and the top is anterior. (B) Scatterplot of adjusted amygdala activity with latent variable scores for each subject during the emotional and neutral conditions. The r value in each plot is the correlation between amygdala activity and LV scores.
saliences (i.e., more negative correlation with amygdala activity during the emotional condition relative to the neutral condition). Structural equation modeling To investigate changes in the influence of the right amygdala on other brain regions during the emotional con-
dition relative to the neutral condition, SEqM was performed on a subset of regions identified in the second LV (listed in Table 1). The functional network significantly differed between conditions [2diff(27) ⫽ 89.74, P ⬍ 0.01]. The functional network models from the emotional and neutral scans are depicted in Fig. 3. To investigate changes in amygdala influence on other brain regions differences in individual path coefficients representing amygdala influence
2096
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
Fig. 3. Functional network models from the emotional and neutral scans. Arrow thickness reflects the functional influence of one region onto another through their anatomical connection as indicated in the legend. Significant changes between emotional and neutral conditions in functional influence of the amygdala on other brain regions are in bold.
between conditions were assessed. The influence of the amygdala on the parahippocampal gyrus [2diff(1) ⫽ 3.88, P ⬍ 0.05, uncorrected] showed a significant increase from the neutral to emotional film viewing conditions. The influence of the amygdala on the ventrolateral prefrontal cortex (BA 47) [2diff(1) ⫽ 14.71, P ⬍ 0.05, corrected] showed a significant change from a negative influence in the neutral condition to a positive influence in the emotional condition. The influence of the amygdala on both the cingulate gyrus [2diff(1) ⫽ 8.24, P ⬍ 0.05, corrected] and on the dorsolateral prefrontal cortex (BA 9/46) [2diff(1) ⫽ 11.43, P ⬍ 0.05, corrected] showed a significant decrease to essentially zero in the emotional condition; however, the nature of the decrease differed. For the cingulate gyrus, amygdala influence decreased from a positive influence in the neutral condition. For the dorsolateral prefrontal cortex, amygdala influence decreased from a negative influence in the neutral condition. The influence of the amygdala on other regions did not show significant changes (P ⬎ 0.05, corrected). Path coefficients representing amygdala influences showing a significant difference between emotional and neutral conditions are shown in Table 2.
Discussion This experiment tested the hypothesis that increased efferent influences from the amygdala to other brain regions, and in particular to the parahippocampal region, will occur in emotionally arousing compared with emotionally neutral learning conditions. Consistent with this hypothesis, results Table 2 Path coefficients representing amygdala influences showing a significant difference between the emotional and neutral conditions Amygdala influence onto
PHG 47 9/46 24
Path coefficient Neutral
Emotional
0.01 ⫺0.30 ⫺0.81 0.86
0.89 0.67 ⫺0.05 ⫺0.12
from an SEqM analysis of PET data revealed a significantly more positive amygdala influence on two brain regions— the parahippocampal gyrus and Brodmann’s area 47— during the emotional relative to the neutral film viewing conditions. The findings constitute the first evidence that increased outflow from the amygdala, presumably reflecting its well-established memory-modulating ability, occurs in the human brain during periods of heightened emotional arousal. In SEqM, a path coefficient represents the expected change in the activity of one region given a unit change in the region influencing it, assuming no other influencing regions show any change (McIntosh and Gonzalez-Lima, 1994). Therefore, the results are consistent with the possibility that (1) an increase in amygdala activity during the neutral condition would directly lead to an increase in cingulate activity, a decrease in dorsolateral prefrontal cortex activity, a smaller decrease in ventrolateral prefrontal cortex, and very little change in parahippocampal gyrus activity; and (2) in contrast, an increase in amygdala activity during the emotional condition would lead to an increase in ventrolateral and parahippocampal activity and very little change in cingulate gyrus and dorsolateral prefrontal cortex activity. SEqM is a powerful analytical technique that goes beyond simply correlating activity between brain regions to illuminating how such correlations may be mediated. To date, there has been only one other research investigation that used SEqM to examine the influence of the amygdala on other brain regions during emotional conditions (Iidaka et al., 2001). However, these authors did not use the “stacked model” approach whereby the patterns of influence in the two conditions can be directly compared; instead, they considered each condition individually. Consequently, although the path coefficients appear to be different in the two conditions in Iidaka et al. (2001), the question of whether they indeed were different was not actually tested. In contrast, the present experiment directly examined this key issue. A further strength of the present experiment in addressing the memory-modulation hypothesis of amygdala function is that memory measures were taken. The analysis in
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
Iidaka et al. (2001) was based on the area of the amygdala maximally activated during their emotional condition. In contrast, the present experiment investigated the changes in influence of the particular region within the amygdala maximally related to enhanced memory for emotional material, an approach more pertinent to the memory-modulation hypothesis. Amygdala influence on activity in the parahippocampal gyrus increased significantly during the emotional relative to neutral film viewing conditions. Several studies have linked activity in the parahippocampal gyrus to subsequent memory in a variety of paradigms (Alkire et al., 1998; Brewer et al., 1998; Wagner et al., 1998). Alkire et al. (1998) found that activity in the posterior parahippocampal gyrus during encoding correlated very highly (r ⫽ 0.98) with long-term free recall of nonemotional words in a longterm incidental memory test. Both Brewer et al. (1998) and Wagner et al. (1998) found that parahippocampal activity distinguished subsequently remembered versus forgotten stimuli. Consistent with these human brain imaging findings, animal research (particularly involving primates) has heavily implicated the parahippocampal region in “declarative” memory formation (Squire and Zola, 1996). Finally, research with rats demonstrates that the amygdala influences memory-related activities in various portions of the hippocampal/parahippocampal complex (Packard et al., 1994; Packard and Teather, 1998; Roesler et al., 2002). Collectively, these findings make the parahippocampal region the prime candidate on theoretical grounds to receive a heightened influence from the amygdala during an emotionally arousing learning situation. The present findings add new support to this view. Amygdala influence on the ventrolateral prefrontal cortex (Brodmann’s area 47) changed significantly from a negative influence in the neutral condition to a positive influence in the emotional film viewing condition. The ventrolateral prefrontal cortex has also been associated with successful subsequent memory. Both Wagner et al. (1998) and Alkire et al. (1998) found that activity in the ventrolateral prefrontal cortex was higher during encoding of words that were subsequently remembered than during encoding of words that were forgotten. Increased activity in the ventrolateral prefrontal cortex has also been found during a deep encoding task relative to a shallow encoding task (Kapur et al., 1994). Amygdala influence on the cingulate gyrus and dorsolateral prefrontal cortex (Brodmann’s area 9/46) decreased significantly in emotional relative to neutral film conditions. The inclusion of these regions was based on the results of the seed-PLS analysis of right amygdala functional connectivity. Although these regions showed strong changes in functional connectivity with the right amygdala between emotional film conditions, their activity did not show a strong correlation with memory in the emotional condition (see Table 1). In fact, these two regions had the smallest correlation between activity during the emotional condition
2097
and memory for the emotional films, as well as the smallest correlation with amygdala activity during the emotional condition of all the regions included in the anatomical model. These observations suggest that not all brain regions will receive a heightened amygdala influence during emotional conditions. It is conceivable that some brain regions receive a tonic influence from the amygdala in neutral conditions that is reduced during periods of emotional arousal. Possibilities such as these deserve more attention in future studies of amygdala effective connectivity. Future investigations will also need to address potential sex-related differences in amygdala influences on other brain regions. Two recent investigations reported a sexrelated hemispheric lateralization of amygdala involvement in emotionally influenced memory (Cahill et al., 2001; Canli et al., 2002). In both, enhanced activity of the right, but not the left, amygdala in men related significantly to enhanced memory for emotional material. Conversely, enhanced activity of the left, but not the right, amygdala in women related significantly to enhanced memory for emotional material. We focused on activity in males for our initial efforts to model amygdala function for two reasons. First, the significance in our previous PET study of the relationship between amygdala activity and memory for the emotional films was higher in males than it was in females; second, the location of the region of maximal correlation with memory for the emotional films was more directly located within the basolateral nucleus in males than it was in females in that study (Cahill et al., 2001). This latter fact is highly important because it is the basolateral nucleus that is most strongly implicated by animal research in memory modulation (Cahill and McGaugh, 1998). However, the question of whether similar changes in amygdala effective connectivity will be found in women (presumably involving the left amygdala) remains, in our view, a highly important question to address. Indeed, we are presently investigating sex-related differences in amygdala function using an eventrelated fMRI paradigm in which the peak voxel within the left amygdala in women and the peak voxel within the right amygdala in men relating to memory for emotional material were previously found to be comparably located within the basolateral nucleus complex (Canli et al., 2002). These data should allow for a more powerful comparison of amygdala effective connectivity in men and women than is possible from our existing PET data set. Further improvements upon the current study afforded by the on-going event-related fMRI investigation include the possibility for a larger sample size than that used in the present study. It is possible that our findings may have been produced without the presence of item-specific correlations between brain regions. However, it is not necessary for there to be item-specific correlations between, for example, the amygdala and parahippocampal gyrus in order for a modulatory function of the amygdala on this region to exist (McIntosh, 1999). The amygdala may modulate parahippocampal activity in a more global rather than item-specific fashion.
2098
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099
Indeed, given the relative slowness of consolidation processes (which clearly last many hours at minimum), it is entirely possible that glucose PET may prove more sensitive to at least some aspects of amygdala-based modulatory processes than will event-related fMRI. It should be noted that although enhanced outflow from the amygdala in emotional conditions seems likely to reflect its role in modulating memory consolidation as established by animal research (McGaugh et al., 2000), it may also reflect, at least in part, hypothesized amygdala influences on arousal and/or attentional processes with emotion (e.g., Kapp et al., 1992; Anderson and Phelps, 2001). The present data cannot distinguish these possibilities. Regarding the prefrontal cortex, it is possible that the heightened amygdala influence with emotion reflects a taking “off-line” of this structure by the amygdala during a period of emotional stress (Arnsten, 1998). Recently Keightley et al. (2003) reported results from a seed-PLS analysis of amygdala function during emotional picture viewing conditions, which revealed generally positive correlations with the right hemisphere amygdala and more dorsally located regions during viewing of positive and negative stimuli. In contrast, the present seed-PLS analysis revealed more negative correlations with the right amygdala and dorsally located regions during viewing of emotionally valenced films. The reasons for this seeming disparity between the two seed-PLS analyses of right amygdala function are not clear, but several possibilities can be considered. First, Keightley et al. (2003) combined data from men and women, whereas the present analysis involved only men. Also, Keightley et al. (2003) investigated commonalities in amygdala functional connectivity across positive and negative emotional conditions absent a neutral condition, whereas the present experiment compared amygdala functional connectivity in neutral versus negative emotional conditions, and with reference to long-term memory for the material. Despite their specific differences, both studies indicate that the functional connectivity of the human amygdala is influenced by the emotional state. Concluding remarks Structural equation modeling of PET data revealed heightened amygdala outflow during emotional compared with neutral learning conditions, most notably to the parahippocampal gyrus. These findings provide new support for the view that strong memory for emotionally arousing events results, at least in part, from modulatory influences from the amygdala on memory storage processes occurring in other brain regions.
Acknowledgments We greatly appreciate the expert and detailed anatomical advice offered by James Fallon in the creation of the neuroanatomical model and our wonderful path analysis con-
sultant Randy McIntosh. Funded by NIMH Grant MH57508 to L.C.
References Alkire, M.T., Haier, R.J., Fallon, J.H., Cahill, L., 1998. Hippocampal, but not amygdala, activity at encoding correlates with long-term, free recall of nonemotional information. Proc. Natl. Acad. Sci. USA 95 (24), 14506 –14510. Amaral, D.G., Price, J.L., 1984. Amygdalo-cortical projections in the monkey (Macaca fascicularis). J. Comp. Neurol. 230 (4), 465– 496. Anderson, A.K., Phelps, E.A., 2001. Lesions of the human amygdala impair enhanced perception of emotionally salient events. Nature 411 (6835), 305–309. Arnsten, A.F.T., 1998. Catecholamine modulation of prefrontal cortical cognitive function. Trends Cogn. Sci. 2 (11), 436 – 447. Barbas, H., De Olmos, J., 1990. Projections from the amygdala to basoventral and mediodorsal prefrontal regions in the rhesus monkey. J. Comp. Neurol. 300, 549 –571. Brewer, J.B., Zhao, Z., Desmond, J.E., Glover, G.H., Gabrieli, J.D., 1998. Making memories: brain activity that predicts how well visual experience will be remembered. Science 281 (5380), 1185–1187. Cahill, L., 2000. Modulation of long-term memory storage in humans by emotional arousal: adrenergic activation and the amygdala, in: Aggleton, J.P. (Ed.), The Amygdala: A Functional Analysis, Oxford University Press, New York, pp. 425– 445. Cahill, L., Haier, R.J., Fallon, J., Alkire, M.T., Tang, C., Keator, D., Wu, J., McGaugh, J.L., 1996. Amygdala activity at encoding correlated with long-term, free recall of emotional information. Proc. Natl. Acad. Sci. USA 93 (15), 8016 – 8021. Cahill, L., Haier, R.J., White, N.S., Fallon, J., Kilpatrick, L., Lawrence, C., Potkin, S.G., Alkire, M.T., 2001. Sex-related difference in amygdala activity during emotionally influenced memory storage. Neurobiol. Learn. Mem. 75 (1), 1–9. Cahill, L., McGaugh, J.L., 1998. Mechanisms of emotional arousal and lasting declarative memory. Trends Neurosci. 21 (7), 294 –299. Canli, T., Desmond, J.E., Zhao, Z., Gabrieli, J.D., 2002. Sex differences in the neural basis of emotional memories. Proc. Natl. Acad. Sci. USA 99 (16), 10789 –10794. Canli, T., Zhao, Z., Brewer, J., Gabrieli, J.D., Cahill, L., 2000. Eventrelated activation in the human amygdala associates with later memory for individual emotional experience. J. Neurosci. 20 (19), RC99. Carmichael, S.T., Price, J.L., 1995. Limbic connections of the orbital and medial prefrontal cortex in macaque monkeys. J. Comp. Neurol. 363, 615– 641. Cavada, C., Company, T., Tejedor, J., Cruz-Rizzolo, R.J., Reinoso-Suarez, F., 2000. The anatomical connections of the macaque monkey orbitofrontal cortex: a review. Cereb. Cortex 10 (3), 220 –242. Hamann, S.B., Ely, T.D., Grafton, S.T., Kilts, C.D., 1999. Amygdala activity related to enhanced memory for pleasant and aversive stimuli. Nat. Neurosci. 2 (3), 289 –293. Iidaka, T., Omori, M., Murata, T., Kosaka, H., Yonekura, Y., Okada, T., Sadato, N., 2001. Neural interaction of the amygdala with the prefrontal and temporal cortices in the processing of facial expressions as revealed by fMRI. J. Cogn. Neurosci. 13 (8), 1035– 47. Kapp, B.S., Whalen, P.J., Supple, W.F., Pascoe, J.P., 1992. Amygdaloid contributions to conditioned arousal and sensory information processing, in: Aggleton, J.P. (Ed.), The Amygdala: Neurobiological Aspects of Emotion, Memory, and Mental Dysfunction, Wiley–Liss, Inc., New York. Kapur, S., Craik, F.I., Tulving, E., Wilson, A.A., Houle, S., Brown, G.M., 1994. Neuroanatomical correlates of encoding in episodic memory: levels of processing effect. Proc. Natl. Acad. Sci. USA 91 (6), 2008 – 2011.
L. Kilpatrick, L. Cahill / NeuroImage 20 (2003) 2091–2099 Keightley, M.L., Winocur, G., Graham, S.J., Mayberg, H.S., Hevenor, S.J., Grady, C.L., 2003. An fMRI study investigating cognitive modulation of brain regions associated with emotional processing of visual stimuli. Neuropsychologia 41 (5), 585–596. Kohler, S., McIntosh, A.R., Moscovitch, M., Winocur, G., 1998. Functional interactions between the medial temporal lobes and posterior neocortex related to episodic memory retrieval. Cereb. Cortex 8 (5), 451– 461. Lavenex, P., Suzuki, W.A., Amaral, D.G., 2002. Perirhinal and parahippocampal cortices of the macaque monkey: projections to the neocortex. J. Comp. Neurol. 447 (4), 394 – 420. McGaugh, J.L., Ferry, B., Vazdarjanova, A., Roozendaal, B., 2000. Amygdala: role in modulation of memory storage, in: Aggleton, J.P. (Ed.), The Amygdala: A Functional Analysis, Oxford University Press, New York, pp. 391– 424. McIntosh, A.R., 1999. Mapping cognition to the brain through neural interactions. Memory 7 (5– 6), 523–548. McIntosh, A.R., Bookstein, F.L., Haxby, J.V., Grady, C.L., 1996. Spatial pattern analysis of functional brain images using partial least squares. NeuroImage 3 (31), 143–157. McIntosh, A.R., Gonzalez-Lima, F., 1994. Structural equation modeling and its application to network analysis in functional brain imaging. Hum. Brain Mapp. 2, 2–22. Morris, J.S., Friston, K.J., Buchel, C., Frith, C.D., Young, A.W., Calder, A.J., Dolan, R.J., 1998. A neuromodulatory role for the human amygdala in processing emotional facial expressions. Brain 121 (1), 47–57. Morris, J.S., Ohman, A., Dolan, R.J., 1999. A subcortical pathway to the right amygdala mediating “unseen” fear. Proc. Natl. Acad. Sci. USA 96 (4), 1680 –1685. Morris, R., Petrides, M., Pandya, D.N., 1999. Architecture and connections of retrosplenial area 30 in the rhesus monkey (Macaca mulatta). Eur. J. Neurosci. 11 (7), 2506 –2518. Packard, M.G., Cahill, L., McGaugh, J.L., 1994. Amygdala modulation of hippocampal-dependent and caudate nucleus-dependent memory processes. Proc. Natl. Acad. Sci. USA 91 (18), 8477– 8481. Packard, M.G., Teather, L.A., 1998. Amygdala modulation of multiple memory systems: hippocampus and caudate-putamen. Neurobiol. Learn. Mem. 69 (2), 163–203.
2099
Petrides, M., Pandya, D.N., 1999. Dorsolateral prefrontal cortex: comparative cytoarchitectonic analysis in the human and the macaque brain and corticocortical connection patterns. Eur. J. Neurosci. 11 (3), 1011– 1036. Petrides, M., Pandya, D.N., 2002. Comparative cytoarchitectonic analysis of the human and the macaque ventrolateral prefrontal cortex and corticocortical connection patterns in the monkey. Eur. J. Neurosci. 16 (2), 291–310. Porrino, L.J., Crane, A.M., Goldman-Rakic, P.S., 1981. Direct and indirect pathways from the amygdala to the frontal lobe in rhesus monkeys. J. Comp. Neurol. 198 (1), 121–136. Roesler, R., Roozendaal, B., McGaugh, J.L., 2002. Basolateral amygdala lesions block the memory-enhancing effect of 8-Br-cAMP infused into the entorhinal cortex of rats after training. Eur. J. Neurosci. 15 (5), 905–910. Sampson, P.D., Streissguth, A.P., Barr, H.M., Bookstein, F.L., 1989. Neurobehavioral effects of prenatal alcohol. II. Partial least squares analysis. Neurotoxicol. Teratol. 11 (5), 477– 491. Sokoloff, L., Reivich, M., Kennedy, C., DesRosiers, M., Patlak, D., Pettigrew, K., Sakurada, O., Shinohora, M., 1977. The [14C] deoxyglucose method for the measurement of local cerebral glucose utilization: theory, procedure, and normal values in the conscious and anesthetized albino male rat. J. Neurochem. 28, 897–916. Squire, L.R., Zola, S.M., 1996. Structure and function of declarative and nondeclarative memory systems. Proc. Natl. Acad. Sci. USA 93 (24), 13515–13522. Stefanacci, L., Suzuki, W.A., Amaral, D.G., 1996. Organization of connections between the amygdaloid complex and the perirhinal and parahippocampal cortices in macaque monkeys. J. Comp. Neurol. 375 (4), 552–582. Wagner, A.D., Schacter, D.L., Rotte, M., Koutstaal, W., Maril, A., Dale, A.M., Rosen, B.R., Buckner, R.L., 1998. Building memories: remembering and forgetting of verbal experiences as predicted by brain activity. Science 281 (5380), 1188 –1191. Zald, D.H., Donndelinger, M.J., Pardo, J.V., 1998. Elucidating dynamic brain interactions with across-subjects correlational analyses of positron emission tomographic data: the functional connectivity of the amygdala and orbitofrontal cortex during olfactory tasks. J. Cereb. Blood Flow Metab. 18 (8), 896 –905.