Neural substrates underlying human delay and trace eyeblink conditioning

Transcription

1 Neural substrates underlying human delay and trace eyeblink conditioning Dominic T. Cheng, John F. Disterhoft, John M. Power, Deborah A. Ellis, and John E. Desmond Department of Neurology, Division of Cognitive Neuroscience, Johns Hopkins University School of Medicine, 1620 McElderry Street, Reed Hall East 2, Baltimore, MD 21205; Department of Physiology, Institute for Neuroscience, Northwestern University Feinberg School of Medicine, 303 East Chicago Avenue, Chicago, IL 60611; and Queensland Brain Institute, Building 79, University of Queensland, St. Lucia QLD 4072, Australia Edited by Larry R. Squire, Veterans Affairs Medical Center, San Diego, CA, and approved April 11, 2008 (received for review January 14, 2008) Classical conditioning paradigms, such as trace conditioning, in which a silent period elapses between the offset of the conditioned stimulus (CS) and the delivery of the unconditioned stimulus (US), and delay conditioning, in which the CS and US coterminate, are widely used to study the neural substrates of associative learning. However, there are significant gaps in our knowledge of the neural systems underlying conditioning in humans. For example, evidence from animal and human patient research suggests that the hippocampus plays a critical role during trace eyeblink conditioning, but there is no evidence to date in humans that the hippocampus is active during trace eyeblink conditioning or is differentially responsive to delay and trace paradigms. The present work provides a direct comparison of the neural correlates of human delay and trace eyeblink conditioning by using functional MRI. Behavioral results showed that humans can learn both delay and trace conditioning in parallel. Comparable delay and trace activation was measured in the cerebellum, whereas greater hippocampal activity was detected during trace compared with delay conditioning. These findings further support the position that the cerebellum is involved in both delay and trace eyeblink conditioning whereas the hippocampus is critical for trace eyeblink conditioning. These results also suggest that the neural circuitry supporting delay and trace eyeblink classical conditioning in humans and laboratory animals may be functionally similar. cerebellum functional MRI hippocampus Classical conditioning has been widely used as a model system to study the neural substrates of associative learning. One form of classical conditioning, eyeblink classical conditioning (EBCC), has received considerable attention by the neuroscience community. In EBCC, repeated pairings of a neutral conditioned stimulus (CS), such as a tone, and an unconditioned stimulus (US), such as an airpuff to the eye, eventually result in the CS alone eliciting conditioned responses (CRs), suggesting that an association between the CS and US has been learned. Delay and trace conditioning are two different procedures that vary in timing. In delay conditioning, the CS and US coterminate whereas in trace conditioning, a silent period (called the trace interval) elapses between offset of the CS and delivery of the US. This minor variation in temporal contiguity has significant consequences on behavioral performance in both healthy individuals and medial temporal lobe amnesiacs. For example, hippocampal patients show normal acquisition of delay conditioning but cannot learn trace conditioning (1 4). Considerable evidence from animal and human research suggests that the hippocampus plays a critical role during trace conditioning whereas the cerebellum is involved in both delay and trace conditioning (5 8). Many experimental techniques, such as electrophysiological recordings, lesion procedures, and genetic manipulations, have been used in laboratory animals to test the idea that the hippocampus is essential for successful trace EBCC (9 19). However, the role of the human hippocampus during trace EBCC has been mainly addressed through behavioral studies of healthy individuals and amnesic patients (1, 2, 4). Previous functional neuroimaging investigations (20 26) [positron emission tomography and functional magnetic resonance imaging (fmri)] of eyeblink conditioning have used only delay procedures to explore associative learning-related changes in the brain, whereas trace eyeblink conditioning has been largely ignored. Because the hippocampus has been shown to be critical for trace EBCC (9 19) and significant hippocampal activity has also been reported during delay EBCC (21, 22, 24, 27, 28), the question of whether hippocampal responding during trace EBCC differs significantly from delay EBCC remains. To address this question, the present EBCC work compared fmri brain activation during delay and trace conditioning by using a within-subjects design. Concurrent measurements of eyeblink responses were recorded by using infrared light delivered through a fiber-optic probe (29). Participants received pseudoconditioning and acquisition blocks while watching a silent movie. Pseudoconditioning consisted of four delay-alone, four trace-alone, and four airpuff-alone blocks, whereas acquisition consisted of alternating delay and trace blocks (16 of each). Each block consisted of nine trials and lasted 27 sec. Participants were instructed to attend to the movie and were informed that the study was designed to understand how their knowledge about the movie would be affected by distracting sounds and airpuffs to the left eye. After conditioning, awareness of the CS US relationship and movie knowledge were assessed by a postexperimental questionnaire (1). Behavioral and imaging data were parsed into early and late acquisition periods to assess timedependent changes as a function of learning. Results Behavioral Results. Learning-related changes in behavior were assessed by repeated-measures ANOVA on both % CRs and peak response latencies. Changes in participant s % CRs developed over the course of the experiment as demonstrated by a time main effect for delay (F 1, , P 01) and trace (F 1, , P 32) conditioning (Fig. 1d). Post hoc tests revealed that both early (t , P 03) and late (t , P 01) delay % CRs were greater than % CRs measured during pseudoconditioning. Early trace % CRs (t , P 0.111) did not differ from pseudoconditioning, but participants showed significantly greater late trace % CRs (t , P 32). These behavioral findings suggest that participants successfully learned the CS US relationship for both delay and trace stimuli. As an additional index of conditioning, peak response Author contributions: D.T.C., J.F.D., and J.E.D. designed research; D.T.C., D.A.E., and J.E.D. performed research; D.T.C., J.M.P., and J.E.D. analyzed data; and D.T.C., J.F.D., J.M.P., D.A.E., and J.E.D. wrote the paper. The authors declare no conflict of interest. This article is a PNAS Direct Submission. To whom correspondence should be addressed. dcheng14@jhmi.edu. This article contains supporting information online at /DCSupplemental by The National Academy of Sciences of the USA PNAS June 10, 2008 vol. 105 no cgi doi pnas

2 d % CR a c CS US CS US Delay Conditioning 850 ms 100 ms Delay Block * Early Delay Conditioning Late 250 ms 500 ms Trace Block 100 ms latencies were calculated, and they further supported the idea that subjects acquired both delay and trace conditioning (Fig. 1e). Learning-related changes developed over the course of the experiment as a repeated-measures ANOVA revealed a time main effect for delay (F 1, , P 001) and trace (F 1, , P 003) conditioning. Post hoc tests showed that, relative to pseudoconditioning, both early delay (t , P 002) and late delay (t , P 001) peak responses occurred closer to the time of US presentation. Comparable results were seen for early (t , P 01) and late (t , P 003) trace peak response latencies. These findings indicate that participants produced more appropriately timed CRs during acquisition relative to pseudoconditioning. Direct comparisons between delay and trace conditioning trials showed no significant changes in acquisition between these two forms of learning. Differences in % CRs (t , P 79) and latency measures (t , P 0.895) during early delay and trace trials were not significant. Also, % CRs (t , P 0.156) and latency measures (t , P 83) did not differ between delay and trace trials during late acquisition. These data suggest that participants learned both conditioning paradigms equally quickly. Awareness of the CS US relationship was measured by a true/false postexperimental questionnaire (1). Participants correctly answered the same number of delay and trace questions b CS US Pseudoconditioning Pseudoconditioning Trace Conditioning Early Trace Conditioning * * * * * * Late Fig. 1. Study design and behavioral results. (a c) Temporal relationship between the CS and US in both delay and trace conditioning. (a) Delay CSs lasted 850 ms and coterminated with a 100-ms US presentation. (b) Trace CSs lasted 250 ms and were followed by a 500-ms trace interval before a 100-ms US presentation. (c) During acquisition, participants received 16 delay and 16 trace blocks in an alternating fashion (nine trials per block). (d and e) Behavioral data showing learning-related changes over the course of acquisition. (d) Behavioral performance expressed as % CR. Relative to pseudoconditioning, participants increased their % CRs to delay CSs during early and late acquisition whereas greater % CRs for trace CSs were evident during late acquisition. No significant behavioral differences were seen between delay and trace CSs within each acquisition phase. (e) Behavioral performance expressed as latency of peak response. A significant increase in response latencies was observed for both delay and trace conditioning during early and late acquisition (i.e., during training, the peak response shifted temporally closer to US presentation) e Latency of Peak Response (t , P 0.570). There did not appear to be a significant correlation between delay awareness and delay % CRs (r 0.313, P 0.35). However, there was a positive trend (r 0.567, P 69) between trace awareness and trace % CRs during the final four blocks. Participants answered 9.4 movie questions (of 10) correctly. Imaging Results. To investigate whether neural activity during trace conditioning differs from delay conditioning, we separated the acquisition period into early (first eight blocks) and late (last eight blocks) acquisition and contrasted activation between delay and trace conditioning. Anatomically defined regions of interest based on probability maps defined by a cytoarchitectonic mapping study (30) were used to sample brain activity in medial temporal lobe (MTL) regions on an individual subject level. The development of associative learning-related changes in these structures was assessed by comparing late acquisition with early acquisition for both delay and trace conditioning. Overall responding in these structures during the acquisition of delay and trace conditioning is shown [supporting information (SI) Fig. S1]. Bilateral MTL regions showed similar response profiles, but only the right MTL showed significantly greater trace-related activity compared with delay (t , P 35). Contralateral hippocampal activation to the trained eye is consistent with expectations based on animal work (17). Given the overall significance of the right MTL, we further explored activations restricted to specific subregions of this area. These subregions were determined by a cytoarchitectonic mapping study (30) and differentiated between the cornu ammonis (CA), the subiculum (SUB), the fascia dentata (FD), the hippocampal amygdaloid transition area (HATA), the perirhinal (PRh) and entorhinal cortex (EC) (30). Of all of these subregions comprising the right MTL, only the FD showed significant differential responding between the development of delay and trace conditioning [mean (t , P 48) peak (t , P 49)] (Fig. 2 a and b). Activity during trace conditioning increased over the course of the experiment, whereas a relative decrease was detected as delay conditioning progressed (Fig. 2a). Furthermore, a direct comparison of late delay and late trace acquisition revealed significantly greater trace-related activity (t , P 01) (Fig. 2c). Importantly, there were no differences in behavioral CR expression between these two forms of conditioning during the late stages of acquisition, suggesting that increased responding in the FD was not performance-related. Similar analysis procedures were conducted to determine whether responding in the cerebellum differentiates between delay and trace conditioning. Contrasts between delay and trace conditioning blocks vs. a baseline period showed a common region of functional activation in the left cerebellar lobule HVI (Fig. 3). Mean and peak values showed significantly higher responding during delay (mean: t , P 047; peak: t , P 0005) and trace (mean: t , P 43; peak: t , P 002) conditioning compared with baseline. However, delay- and trace-related activations did not differ from each other (mean: t , P 27; peak: t 10 4, P 0.813), suggesting that this region is involved in both delay and trace conditioning but does not discriminate between these two forms of learning. Discussion The neural basis underlying EBCC has been well characterized by using a variety of animals and techniques and continues to be one of the most studied forms of associative learning. The cerebellar cortex and deep nuclei have been linked to an animal s ability to learn successfully both delay and trace EBCC (5 8), whereas additional structures, such as the hippocampus, are recruited during trace conditioning (9 19). The present work extends our current understanding of this neural circuit by using PSYCHOLOGY NEUROSCIENCE Cheng et al. PNAS June 10, 2008 vol. 105 no

3 Right Hippocampus a β -values * b β -values * 0.6 c * β -values Late minus Early (mean) Late minus Early (peak) Late Trace minus Late Delay (peak) Delay Trace L -29 R Fig. 2. Delay and trace activity specific to a subregion of the MTL (center of mass Talairach coordinates: 23, 29, 4). Brain maps illustrate the right fascia dentata (FD) (30) overlayed onto high-resolution T1-weighted MPRAGE images. (a) Similar to overall MTL findings, the right FD showed significant differential responding between delay and trace conditioning. (b) Peak responses within this region also resulted in significantly greater trace-related activity. (c) Although behavioral performance between late delay and late trace conditioning was equivalent (Fig. 1 d and e), a direct comparison of FD responding between these two forms of learning revealed significantly greater FD activity during late trace conditioning, suggesting that differential responding in this area may not be performance-related. fmri to compare directly brain activity during human delay and trace eyeblink conditioning within the same acquisition session. Behavioral results showed that humans were capable of learning both delay and trace conditioning in parallel. Brain imaging data outlined the important role of the human cerebellum during delay and trace EBCC and the critical role of the hippocampus during trace EBCC. Δ β -values Left Cerebellum (Lobule HVI) Delay minus Baseline L * * * * 1.0 mean Δ β -values peak Trace minus Baseline -57 R Fig. 3. Delay and trace activity observed within the left cerebellum (lobule HVI; center of mass Talairach coordinates: 24, 57, 23). Mean and peak response measures show significantly greater activity for both delay and trace conditioning relative to baseline. Unlike hippocampal regions, differences between delay and trace conditioning were not found in this cerebellar ROI, suggesting that this area does not discriminate between these two forms of learning. Behavioral CRs. Participants were able to learn both delay and trace conditioning in parallel, which was demonstrated by calculating two different indices of conditioning: % CR and latency of peak response. A significantly greater number of delay and trace conditioned responses were measured during acquisition relative to pseudoconditioning, and the time of maximal eyelid closure (peak response) shifted closer to US presentation over the course of the experiment (Fig. 1 d and e). Both of these findings indicated that participants produced more adaptive eyeblink CRs during acquisition. Comparable levels of delay and trace conditioned responses were measured throughout acquisition, suggesting that participants learned both conditioning paradigms equally quickly. In the current work, identical interstimulus intervals (ISIs) for both trace and delay trial types were used to equate the stimulus conditions for fmri contrasts, and as a result, distinguishing between delay and trace CRs on the basis of temporal topography of the CR was not possible. However, evidence that separate learning processes occurred include differential patterns of brain activity between these two forms of learning and the positive relationship between awareness and trace CRs. Furthermore, the fact that delay and trace stimuli were sufficiently dissimilar (pure tone vs. white noise) such that all participants reported hearing two different auditory stimuli suggests that CRs were specific to each trial type. Concurrent delay and trace conditioning has also been successfully demonstrated by using fear conditioning procedures (31), but the present fmri study uses eyeblink conditioning procedures. This general approach allowed us to assess common and unique brain responses during the parallel acquisition of both delay and trace conditioning. Although significant learning-related changes were measured, the level of conditioning inside the scanner was less robust than what is typically seen in human behavioral conditioning studies (5). This effect likely reflects the distractions that accompany the MRI environment (e.g., radiofrequency pulse; gradient noise). Other factors such as pseudoconditioning (potentially producing latent inhibition) and a conservative CR time window (250 ms pre-us) may have also contributed to the level of conditioning observed in the current work cgi doi pnas Cheng et al.

4 Cerebellum. Significantly greater activation in the left (ipsilateral to the trained eye) cerebellum (lobule HVI) was detected during both delay and trace conditioning (Fig. 3). The present results are consistent with findings from previous functional neuroimaging studies on delay conditioning, which reported learningrelated changes in the cerebellum/lobule HVI (20 26). Evidence from animal and human studies has shown that the cerebellar cortex and deep nuclei are involved in both delay and trace eyeblink conditioning (6 8). An event-related fmri study of single-cue delay conditioning in the rabbit reported decreased blood-oxygenation-level-dependent (BOLD) responses in lobule HVI of the cerebellar cortex during early and asymptotic learning, with no changes visible during the unpaired pseudoconditioning trials that preceded learning, although the BOLD response reduced in area and became more localized as learning proceeded (26). The decreased BOLD response was interpreted as consistent with the fact that reduced drive from cerebellar cortical Purkinje cells onto cerebellar deep nuclei would be required to initiate CR performance controlled from the cerebellar dentate interpositus as training proceeded. In the present work, we compared the BOLD response during late compared with early training trials and observed no significant differences. Although the early vs. late training contrast was not significant for the cerebellar cortex in the present work, we did find significantly greater cerebellar activity relative to baseline for both delay and trace conditioning. However, differential responding between delay and trace conditioning was not detected, suggesting that this region does not discriminate between these two forms of learning. Hippocampus. Contrasting early and late acquisition periods allowed us to investigate the development of hippocampal activity as a function of delay and trace learning. Greater hippocampal responding was detected in the later stages of trace conditioning (Fig. 2a), suggesting that hippocampal involvement gradually increased over the course of trace learning. This response pattern, which closely parallels the development of subject s behavioral trace % CRs, is in agreement with data showing that greater activity in rabbit hippocampal cell recordings was detected on the same day that significant trace CRs began to develop (16). The fact that the largest BOLD response was seen in the hippocampus contralateral to the trained eye late in acquisition is consistent with the pattern of changes in single-unit responses seen in rabbit hippocampus (17), where the larger firing rate increases were seen in the hippocampus contralateral compared with ipsilateral to the trained eye as conditioning became well established. These findings are also consistent with a rodent study (18) that reports hippocampal single-unit activity during the late stages of trace acquisition. In contrast to the late developing increase in trace-related hippocampal activity, a late decrease in hippocampal responding during delay conditioning was observed (Fig. 2a). This temporal profile suggests relatively greater hippocampal activity (mean values: 8 4) during early delay conditioning, which also corresponds to the development of delay CRs. The delay-related hippocampal activity also appears to be in line with animal electrophysiological evidence. These studies report that hippocampal cells respond very rapidly during the early stages of delay learning, correspond strongly with behavioral CRs, and attenuate with continued training (18, 27, 28, 32). The pattern of hippocampal activation in the present work is consistent with the interpretation that trace conditioning recruits additional forebrain structures, such as the hippocampus, to bridge the temporal gap between the CS and US, whereas delay eyeblink conditioning relies only on circuitry within the cerebellum. Initial hippocampal responding during delay conditioning suggests an early involvement of the hippocampus that eventually subsides because the cerebellum is sufficient to support this form of learning. Although both delay and trace conditioning rely on cerebellar deep nuclei to produce CRs, trace conditioning depends on an intact hippocampus to provide inputs to the deep nuclei via pontine afferents (8). Animal neurophysiology studies showed trace-related activity in recordings made from the CA region (14 18), whereas the current work observed significant responding in the human FD. Although the anatomical regions of interest (ROIs) were based on precise measurements from a cytoarchitectonic mapping study (30), the spatial resolution of our functional voxels and the close proximity between the FD and CA make it difficult to rule out any involvement of the CA region. Higher spatial resolution functional images of the human hippocampus may provide a better method to distinguish between structures within the hippocampus proper. By simply considering the temporal profile of hippocampal activity during delay and trace conditioning, it is tempting to conclude that activity in this region is simply related to the production of CRs. However, a direct comparison between late delay and trace learning revealed greater hippocampal activity during trace learning (Fig. 2c). If trace-related activity in the hippocampus was strictly a function of CR expression, differential responding in the FD should not have been observed in this contrast because delay and trace CRs were comparable during this stage of acquisition (Fig. 1 d and e). These findings suggest that increased activity in the hippocampus during trace learning could not simply be performance-related. Rather, the hippocampus may be recruited during the production of delay and trace CRs and may be critically relied on during the associative processes mediating trace conditioning. The current findings are also consistent with conclusions drawn from animal lesion and human amnesic studies. The acquisition of delay conditioning does not appear to be affected by lesions to the hippocampus (33 35), suggesting that this structure is not a part of the minimal essential circuitry required for delay learning (36, 37). However, trace conditioning is severely impaired by lesions to the hippocampus (9 12), highlighting this region s importance during this form of learning. Measuring greater hippocampal activity during trace than delay conditioning in the current work further emphasizes the critical role of the hippocampus during trace conditioning. A limited number of studies have used functional neuroimaging to study the role of the hippocampus during trace conditioning. A magnetencephalogram (MEG) study reported greater hippocampal activity related to trace but not delay eyeblink conditioning (38). However, subcortical localization with MEG is difficult because it lacks the anatomical resolution that MRI provides. Two studies used fear conditioning procedures and fmri to report trace-related hippocampal activity (31, 39). In particular, results from a similar within-subjects fear conditioning design used online US expectancy ratings to show that participants rapidly acquired the trace CS US relationship. Interestingly, this early explicit learning was accompanied by differences in hippocampal response magnitude related to the accuracy of predicting US presentation during trace conditioning. These fear conditioning results and the current findings suggest that the hippocampus is active around the time of successful trace conditioning. Several theories regarding the specific role of the hippocampus during trace conditioning have been proposed. For example, the hippocampus may be needed to overcome stimulus discontiguity, to time CRs accurately, or to distinguish between the interstimulus interval and trace interval (10, 17, 40). Another theory is that trace conditioning is an associative learning task dependent on awareness or declarative memory, which requires the hippocampus (1, 41 44). One other possibility is that the hippocampus is more active during complicated and difficult forms of classical conditioning (e.g., reversal and trace) (13). PSYCHOLOGY NEUROSCIENCE Cheng et al. PNAS June 10, 2008 vol. 105 no

5 Unfortunately, the design of the current work limits our ability to test several of these theories. We were not able to rule out certain theories (10, 17) because the short ISIs and rapid presentations of conditioning trials in a block design prevented us from disentangling the hemodynamic response to the CS, trace interval, and US. Longer trace intervals may require a greater dependence on an intact hippocampus, but shorter trace intervals have also been shown to have a deleterious effect on patients with hippocampal damage. Using a 500-ms trace interval, one study (4) showed that MTL amnesiacs were impaired relative to control subjects, suggesting that the MTL is involved during trace conditioning with a 500-ms trace period. Given these findings and the numerous studies that emphasize the importance of declarative knowledge/awareness during trace conditioning (1, 2, 41 44), it is reasonable to predict hippocampal activity even during shorter trace intervals. Awareness. The idea that awareness of CS US relationships is critical for successful trace learning has been debated (45 47). All participants were able to differentiate between delay and trace tones, allowing us to present separate postexperimental questionnaires for each stimulus. Performance on these questionnaires was not significantly correlated with their behavioral CRs for either delay or trace conditioning. A closer examination showed a positive trend (P 69) between the number of correctly answered questions on the trace questionnaire and eyeblink CRs recorded at the end of acquisition. This finding is in agreement with previous work (48) showing that awareness was significantly correlated with single-cue trace (r 9) but not single-cue delay conditioning. In fact, the relationship between awareness and trace conditioning may even be stronger in the present work (r 0.60) than in the previous study (48) (r 9) but simply did not reach statistical significance because of a small sample size (n 11). The present behavioral findings are in agreement with several studies reporting that awareness is necessary for trace learning (1, 2, 41 44). Although a participant s hippocampal activity was not correlated with awareness, it should be noted that this finding does not discount the idea that awareness is necessary for trace conditioning. There are several procedural differences between the current article and previous reports supporting the hippocampus awareness trace conditioning theory. Perhaps most importantly, participants in prior studies (1, 41 43) received only trace conditioning, whereas participants in the current work received both delay and trace trial types. It has been shown that learning delay conditioning can affect the development of trace conditioning (13, 49), which leaves open the possibility that awareness of one trial type could influence awareness of the other. Comparisons between the current results and previous awareness trace conditioning papers should be made with these procedural differences in mind. Conclusions This work combines fmri and a within-subjects design of delay/trace eyeblink conditioning to show that participants can learn both delay and trace conditioning in parallel. Significant cerebellar activity to both delay and trace conditioning was observed, but differential responding between these two trial types was not. These findings support the view that this region is involved in both delay and trace conditioning but does not discriminate between the two. Our hippocampal results confirm and extend previous animal lesion, human amnesic, and electrophysiology data by showing greater hippocampal activity during trace than delay conditioning in humans. The timing of this activity appeared to coincide with the development of delay and trace CRs. However, differential hippocampal responding was measured when CR expression was comparable (trace vs. delay contrast), suggesting that this activity was not strictly performance-related. These results suggest that the hippocampus may be critically relied on during the associative processes mediating trace conditioning. Materials and Methods Participants. Eleven healthy, right-handed volunteers ( years; six females, five males) were compensated $20 per hour for their participation. All procedures were approved by the Institutional Review Boards for human subject research at the Johns Hopkins University School of Medicine. Behavioral Apparatus. Stimulus presentation and behavioral data acquisition were controlled and analyzed by using a laptop computer interfaced to DT9834 data acquisition module (Data Translations) running custom software developed under LabView version 7.1 (National Instruments). Presentation of auditory stimuli was achieved by using pneumatic headphones (MRA, Inc.). A movie (Charlie Chaplin s The Gold Rush) was projected onto a backilluminated screen and viewed through an adjustable mirror attached to the head coil. Custom modifications to standard laboratory safety goggles were made so that the end of a polyethylene tube could be attached (Nalgene) for airpuff delivery and also to accommodate a MRI-compatible infrared sensor for recording eyeblinks. A fiber-optic probe (RoMack, Inc.) measured the reflectance of infrared light from the left eye (29), and airpuff delivery was controlled by a solenoid valve (Asco). Imaging Apparatus. Whole-brain functional imaging was performed on a 3-tesla MRI scanner (Phillips) by using a T2*-weighted gradient echo planar imaging (EPI) pulse sequence. Contiguous 7-mm oblique (perpendicular to the axis of the hippocampus) slices were collected (TR, 2,000 ms; TE,3 ms; FOV, 24 cm; flip angle, 70 ) in a series of 810 sequential images (for a total of 1,620 sec). Structural images were acquired by using a T1-weighted magnetizationprepared rapid acquisition gradient echo (MPRAGE) pulse sequence. Stimuli. Delay and trace CSs were counterbalanced between a binaural 1,200-Hz tone and white noise (90 db). The delay stimulus lasted 850 ms and coterminated with a 100-ms corneal airpuff (4 psi) to the left eye (Fig. 1a). The trace stimulus lasted 250 ms and was followed by a 500-ms trace period before airpuff presentation (Fig. 1b). Trials were grouped into blocks: nine trials per block, 2 sec per trial, 3-sec intertrial interval, such that each block lasted 27 sec (Fig. 1c). Pseudoconditioning was seamlessly followed by acquisition blocks. Pseudoconditioning consisted of alternating four delay-alone, four tracealone, and four airpuff-alone blocks, whereas acquisition consisted of alternating 16 delay and 16 trace blocks. Presentation order was counterbalanced among subjects. Procedure. Once participants were fitted with the safety goggles and placed in the magnet, they were instructed to watch and pay attention to the silent movie while distracting sounds and airpuffs were presented. They were told that their movie knowledge would be tested after the experiment. Further, they were informed that this study investigated the effects that distracting sounds and airpuffs have on their ability to remember details about the movie. After conditioning, a movie quiz and postexperimental questionnaire assessing awareness of the CS US contingencies were administered. Behavioral Eyeblink Analysis. The criterion for a response to be considered a CR is that the difference between the maximum and minimum responses in a 250 ms pre-us time window must exceed four times the standard deviation of the mean of the baseline period (250 ms before CS presentation). A time window of 250 ms pre-us presentation was selected to minimize the inclusion of voluntary and responses as CRs (50). Performance was expressed as % CR. Peak response latencies relative to CS onset during this 250-ms time window were also calculated as an additional behavioral measure of conditioning. Repeated-measures ANOVA were performed on these two variables. Imaging Analysis. Structural and functional imaging preprocessing and statistical analyses were performed with Statistical Parametric Mapping (SPM2) software (Wellcome Department of Cognitive Neurology, London, U.K.). EPI functional images were realigned and resliced correcting for minor motion artifacts and MPRAGE structural images were coregistered to the mean motion-corrected functional image for each subject. Both functional and structural images were transformed into standard stereotaxic space according to the Montreal Neurological Institute (MNI) protocol, and the functional images were smoothed with a Gaussian filter (full-width half-maximum 5 mm). The general linear model was used to estimate individual subject activations, and cgi doi pnas Cheng et al.

6 a random-effects analysis was conducted for all subjects. Estimated mean and peak weights were used as a measure of brain activation. A Talairach transformation was performed, and these coordinates are reported in the figures and text (51, 52). Overwhelming evidence from laboratory animal studies and the human lesion literature identify regions in the MTL as critical for learning trace conditioning. Structural ROIs based on probabilistic maps of the MTL region reported in a cytoarchitectonic mapping study (30) and made available as an SPM toolbox were used to sample neural activity on an individual subject level. The MTL region consisted of the cornu ammonis, the subiculum, the fascia dentata, the hippocampal amygdaloid transition area, the perirhinal, and entorhinal cortex (30). Contrasts were designed to explore the nature of MTL activity during both delay and trace conditioning. The development of MTL responding was investigated by separating the acquisition period into early (first eight blocks) and late (last eight blocks) acquisition and contrasting these two periods during delay and trace conditioning. Specific subregions within the MTL were analyzed separately, and a direct comparison between late delay acquisition and late trace acquisition was conducted to test the idea that the hippocampus is critically important for trace conditioning. Similar imaging analyses were conducted to determine whether responding in the cerebellum differentiates between delay and trace conditioning. Because of the known involvement of cerebellar lobule HVI in EBCC from animal and human literature (6 8), a union of activation from both averaged trace and delay conditioning SPM {Z} maps (based on a delay/trace minus pre-cs baseline contrast), thresholded at P 25, within the anatomical boundaries of left HVI (53) was made to create an appropriate cerebellar ROI. This ROI was used to sample cerebellar activity on an individual subject level to investigate the idea that this region responds equally to both delay and trace conditioning. ACKNOWLEDGMENTS. We thank M. M. Oh and C. Weiss for technical assistance. This work was supported by National Institutes of Health Grant R01 AG (to J.E.D.). 1. Clark RE, Squire LR (1998) Classical conditioning and brain systems: The role of awareness. Science 280: Knuttinen M-G, Power JM, Preston AR, Disterhoft JF (2001) Awareness in classical differential eyeblink conditioning in young and aging humans. Behav Neurosci 115: Gabrieli JDE, et al. (1995) Intact delay-eyeblink classical conditioning in amnesia. Behav Neurosci 109: McGlinchey-Berroth R, Carrillo MC, Gabrieli JDE, Brawn CM, Disterhoft JF (1997) Impaired trace eyeblink conditioning in bilateral, medial temporal lobe amnesia. Behav Neurosci 111: Disterhoft JF, et al. (2003) in Neuropsychology of Memory, eds Squire LR, Schachter D (Guilford, New York), pp Christian KM, Thompson RF (2003) Neural substrates of eyeblink conditioning: and retention. Learn Mem 11: Gerwig M, Kolb FP, Timmann D (2007) The involvement of the human cerebellum in eyeblink conditioning. Cerebellum 6: Woodruff-Pak DS, Disterhoft JF (2008) Where is the trace in trace conditioning? Trends Neurosci 31: Solomon PR, Vander Schaaf ER, Thompson RF, Weisz DJ (1986) Hippocampus and trace conditioning of the rabbit s classically conditioned nictitating membrane response. Behav Neurosci 100: James GO, Hardiman MJ, Yeo CH (1987) Hippocampal lesions and trace conditioning in the rabbit. Behav Brain Res 23: Moyer JR, Deyo RA, Disterhoft JF (1990) Hippocampectomy disrupts trace eye-blink conditioning in rabbits. Behav Neurosci 104: Weiss C, Bouwmeester H, Power JM, Disterhoft JF (1999) Hippocampal lesions prevent trace eyeblink conditioning in the freely moving rat. Behav Brain Res 99: Beylin AV, et al. (2001) The role of the hippocampus in trace conditioning: Temporal discontinuity or task difficulty? Neurobiol Learn Mem 76: Moyer JR, Thompson LT, Disterhoft JF (1996) Trace eyeblink conditioning increases CA1 excitability in a transient and learning-specific manner. J Neurosci 16: Weiss C, Kronforst-Collins MA, Disterhoft JF (1996) Activity of hippocampal pyramidal neurons during trace eyeblink conditioning. Hippocampus 6: McEchron MD, Disterhoft JF (1997) Sequence of single neuron changes in CA1 hippocampus of rabbits during acquisition of trace eyeblink conditioned responses. J Neurophysiol 78: Weible AP, O Reilly J-A, Weiss C, Disterhoft JF (2006) Comparisons of dorsal and ventral hippocampus cornu ammonis region 1 pyramidal neuron activity during trace eyeblink conditioning in the rabbit. Neuroscience 141: Green JT, Arenos JD (2007) Hippocampal and cerebellar single-unit activity during delay and trace eyeblink conditioning in the rat. Neurobiol Learn Mem 87: Tseng W, Guan R, Disterhoft JF, Weiss C (2004) Trace eyeblink conditioning is hippocampally dependent in mice. Hippocampus 14: Molchan SE, Sunderland T, McIntosh AR, Herscovitch P, Schreurs BG (1994) A functional anatomical study of associative learning in humans. Proc Natl Acad Sci USA 91: Logan CG, Grafton ST (1995) Functional anatomy of human eyeblink conditioning determined with cerebral glucose metabolism and positron-emission tomography. Proc Natl Acad Sci USA 92: Blaxton TA, et al. (1996) Functional mapping of human learning: A positron emission tomography activation study of eyeblink conditioning. J Neurosci 16: Schreurs BG, et al. (1997) Lateralization and behavioral correlation of changes in regional cerebral blood flow with classical conditioning of the human eyeblink response. J Neurophysiol 77: Ramnani N, Toni I, Josephs O, Ashburner J, Passingham RE (2000) Learning- and expectation-related changes in the human brain during motor learning. J Neurophysiol 84: Knuttinen M-G, et al. (2002) Electromyography as a recording system for eyeblink conditioning with functional magnetic resonance imaging. Neuroimage 17: Miller MJ, et al. (2003) fmri of the conscious rabbit during unilateral classical eyeblink conditioning reveals bilateral cerebellar activation. J Neurosci 23: Berger TW, Alger B, Thompson RF (1976) Neural substrate of classical conditioning in the hippocampus. Science 192: Berger TW, Thompson RF (1978) Neuronal plasticity in the limbic system during classical conditioning of the rabbit nictitating membrane response. I. The hippocampus. Brain Res 145: Miller MJ, Li L, Weiss C, Disterhoft JF, Wyrwicz AM (2005) A fiber optic-based system for behavioral eyeblink measurement in a MRI environment. J Neurosci Methods 141: Amunts K, et al. (2005) Cytoarchitectonic mapping of the human amygdala, hippocampal region and entorhinal cortex: Intersubject variability and probability maps. Anat Embryol 210: Knight DC, Cheng DT, Smith CN, Stein EA, Helmstetter FJ (2004) Neural substrates mediating human delay and trace fear conditioning. J Neurosci 24: Sears LL, Steinmetz JE (1990) of classically conditioned-related activity in the hippocampus is affected by lesions of the cerebellar interpositus nucleus. Behav Neurosci 104: Schmaltz LW, Theios J (1972) and extinction of a classically conditioned response in hippocampectomized rabbits (Oryctolagus cuniculus). J Comp Physiol Psychol 79: Solomon PR, Moore JW (1975) Latent inhibition and stimulus generalization of the classically conditioned nictitating membrane response in rabbits (Oryctolagus cuniculus) after dorsal hippocampal ablation. J Comp Physiol Psychol 89: Akase E, Alkon DL, Disterhoft JF (1989) Hippocampal lesions impair memory of shortdelay conditioned eye blink in rabbits. Behav Neurosci 103: Desmond JE, Moore JW (1982) A brain stem region essential for the classically conditioned but not unconditioned nictitating membrane response. Physiol Behav 28: Mauk MD, Thompson RF (1987) Retention of classically conditioned eyelid responses following acute decerebration. Brain Res 403: Kirsch P, et al. (2003) Cerebellar and hippocampal activation during eyeblink conditioning depends on the experimental paradigm: A MEG study. Neural Plast 10: Büchel C, Dolan RJ, Armony JL, Friston KJ (1999) Amygdala hippocampal involvement in human aversive trace conditioning revealed through event-related functional magnetic resonance imaging. J Neurosci 19: Bolles RC, Collier AC, Bouton ME, Marlin NA (1978) Some tricks for ameliorating the trace-conditioning deficit. Bull Psychonom Soc 11: Clark RE, Squire LR (1999) Human eyeblink classical conditioning: Effects of manipulating awareness of the stimulus contingencies. Psychol Sci 10: Manns JR, Clark RE, Squire LR (2000) Parallel acquisition of awareness and trace eyeblink classical conditioning. Learn Mem 7: Manns JR, Clark RE, Squire LR (2000) Awareness predicts the magnitude of single-cue trace eyeblink conditioning. Hippocampus 10: Knight DC, Nguyen HT, Bandettini PA (2006) The role of awareness in delay and trace fear conditioning in humans. Cogn Affect Behav Neurosci 6: LaBar KS, Disterhoft JF (1998) Conditioning, awareness, and the hippocampus. Hippocampus 8: Lovibond PF, Shanks DR (2002) The role of awareness in Pavlovian conditioning: Empirical evidence and theoretical implications. J Exp Psychol Anim Behav Process 28: Manns JR, Clark RE, Squire LR (2002) Standard delay eyeblink conditioning is independent of awareness. J Exp Psychol Anim Behav Process 28: Manns JR, Clark RE, Squire LR (2001) Single-cue delay eyeblink conditioning is unrelated to awareness. Cogn Affect Behav Neurosci 1: Woodruff-Pak DS (1993) Eyeblink classical conditioning in HM: Delay and trace paradigms. Behav Neurosci 107: Gormezano I (1966) in Experimental Methods and Instrumentation in Psychology, ed Sidowski JB (McGraw Hill, New York), pp Talairach J, Tournoux PA (1988) Co-Planar Stereotaxic Atlas of the Human Brain: 3-Dimensional Proportional System: An Approach to Cerebral Imaging (Thieme, New York). 52. Lancaster JL, et al. (2007) Bias between MNI and Talairach coordinates analyzed using the ICBM-152 brain template. Hum Brain Map 28: Tzouri-Mazoyer N, et al. (2002) Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 15: PSYCHOLOGY NEUROSCIENCE Cheng et al. PNAS June 10, 2008 vol. 105 no