em enhanced retention

Emotion Enhanced Retention of Cognitive Skill Learning

Stephan Steidl, Fathima Razik, and Adam K. AndersonUniversity of Toronto and Rotman Research Institute, Toronto, ON, Canada

Ample evidence suggests that emotional arousal enhances declarative/episodic memory. By contrast,there is little evidence that emotional enhancement of memory (EEM) extends to procedural skill basedmemory. We examined remote EEM (1.5-month delay) for cognitive skill learning using the weatherprediction (WP) probabilistic classification task. Participants viewed interleaved emotionally arousing orneutral pictures during WP acquisition. Arousal retarded initial WP acquisition. While participants in theneutral condition showed substantial forgetting of WP learning across the 1.5-month delay interval, thearousal condition showed no evidence of forgetting across the same time period. Thus, arousal duringencoding determined the mnemonic fate of cognitive skill learning. Emotional enhancement of WPretention was independent of verbally stated knowledge of WP learning and EEM for the picture contextsin which learning took place. These results reveal a novel demonstration of EEM for cognitive skilllearning, and suggest that emotional arousal may in parallel enhance the neural systems that supportprocedural learning and its declarative context.

Keywords: emotional enhancement of memory, weather prediction task, explicit memory, striatum,implicit memory

Emotional events are surrounded with a special subjective viv-idness and are thought to persist in memory unlike more mundaneevents. Both human and animal research has consistently impli-cated the importance of sympathetic arousal and the amygdala inthis emotional enhancement of memory (EEM) (for review, seeCahill & McGaugh, 1999). In particular postencoding manipula-tion of noradrenergic activation in the amygdala appears to medi-ate the effect of sympathetic activation on memory consolidation(McIntyre, Power, Roozendaal, & McGaugh, 2003; Roozendaal,Castello, Vedana, Barsegyan, & McGaugh, 2008). In humans, bothpositron emission tomography (PET) and functional MRI (fMRI)studies have shown that amygdala activation during encoding(Cahill et al., 1996; Canli, Zhao, Brewer, Gabrieli, & Cahill, 2000;Dougal, Phelps, & Davich, 2007; Kensinger & Schacter, 2005) andrecall (Sharot, Delgado, & Phelps, 2004) is correlated with greaterexplicit memory for emotionally arousing material. In contrastwith valence specific contributions (i.e., positive or negative),which are more associated with the prefrontal cortices (Kensinger& Corkin, 2005; Phelps, LaBar, & Spencer, 1997), amygdalainfluences on memory formation are most associated with theintensity of emotional experience, or arousal, common to bothpositive and negative events (Cahill & McGaugh, 1999; Hamann,Ely, Grafton, & Kilts, 1999). Examination of neuroimaging data

using structural equation modeling demonstrates greateramygdala-parahippocampal interactions during emotionally arous-ing relative to neutral event encoding, potentially representing theneural bases of EEM for declarative/epidsodic memory (Kilpatrick& Cahill, 2003). Collectively, these studies suggest an interactionbetween sympathetic arousal, the amygdala, and a hippocampal-based declarative memory system that supports conscious recol-lection of past experiences (Kilpatrick & Cahill, 2003; Hamann,2001; LaBar & Cabeza, 2006; Packard & Cahill, 2001).

Although the psychological and neural bases of EEM have beenextensively studied with respect to declarative/episodic memory,storage of past events is accomplished through multiple qualita-tively distinct memory systems in the brain, each acquiring differ-ent types of information (Cohen & Squire, 1980; Knowlton, Man-gels, & Squire, 1996; McDonald & White, 1993; Packard, Hirsh,& White, 1989; Schacter & Tulving, 1994; White & McDonald,2002). Traditionally, a distinction has been made between ahippocampus-based declarative and a striatum-based nondeclara-tive/procedural memory system (Cohen & Squire, 1980). Forexample, different versions of a 8-arm radial maze task, designedto emphasize different forms of learning, have been shown todissociate the contributions of the hippocampus (a task based onspatial cues in the environment) and dorsal striatum (a task basedon stimulus–response associations), suggesting that each brain areaacquires different information about task situations (McDonald &White, 1993). These data are consistent with the notion that thedorsal striatum in rats is associated with stimulus-response habit-type learning.

In rodents, the amygdala has been shown to modulate spatialand cue-based memory tasks in parallel (McGaugh, 2004;McGaugh, McIntyre, & Power, 2002; Packard & Cahill, 2001;Packard, Cahill, & McGaugh, 1994). For example, posttrainingintraamygdala injection of the catecholaminergic agonistd-amphetamine has been shown to enhance retention on both

This article was published Online First November 8, 2010.Stephan Steidl, Fathima Razik, and Adam K. Anderson, Department of

Psychology, University of Toronto, and Baycrest Centre, Rotman ResearchInstitute, Toronto, ON, Canada.

We thank Rifat Alam for help in editing this article. Supported by anNSERC grant to AKA and an NSERC CGS-D to SS.

Correspondence concerning this article should be addressed to Adam K.Anderson, Department of Psychology, University of Toronto, 100 St.George Street, Toronto, ON M5S 3G3 Canada. E-mail: [email protected]

Emotion © 2010 American Psychological Association2011, Vol. 11, No. 1, 12–19 1528-3542/10/$12.00 DOI: 10.1037/a0020288

12

spatial and cued versions of the water maze task (Packard et al.,1994). This work from nonhuman animals suggests that sympa-thetic arousal during emotional states should play a pervasiveneuromodulatory role in enhancing memory formation in puta-tively different memory systems.

Whereas much evidence supports the role of emotional arousaland the amygdala in modulating human declarative memory, thereis little evidence to support a similar role in nondeclarative pro-cedural memory. One way habit learning in humans has previouslybeen assessed is by using the weather prediction task (WP), aprobabilistic classification task in which subjects are required toclassify combinations of visually distinctive cues as predicting oneof two weather outcomes (i.e., sunshine or rain). The cue combi-nations and their outcomes are probabilistically associated soexplicit memorization of their relationship is difficult. Instead,performance of this task is more easily accomplished through theimplicit learning of complex cue-outcome contingencies that arethought to, at least in part, depend on the striatum (Knowlton,Squire, & Gluck, 1994). Early phases of WP task acquisition areintact in medial temporal lobe amnesiacs but disrupted in patientswith striatal dysfunction associated with Parkinson’s disease(Knowlton et al., 1994). Event-related fMRI shows medial tem-poral lobe activation very early on during WP acquisition (Pol-drack et al., 2001), that quickly becomes deactivated, while thestriatum comes to dominate (Poldrack et al., 2001; Poldrack,Prabhakaran, Seger, & Gabrieli, 1999).

Studies in humans have shown that emotions can facilitate as wellas inhibit the use of stimulus-response learning strategies (Thomas &LaBar, 2008; Schwabe et al., 2007). For example, Thomas and LaBar(2008) showed that emotional arousal during encoding of skill learn-ing, as indexed by a variant of the WP task (Knowlton et al., 1994),resulted in more optimal strategy use during a second Session 24 hrlater. However, it remains unclear whether emotional arousal en-hances long term memory for procedural learning. A prior study usedthe WP task to investigate the effects of emotional arousal duringcognitive skill acquisition on long-term retention (Steidl, Mohi-uddin,& Anderson, 2006). WP trials were interrupted by intermittent emo-tional arousal by showing subjects either highly arousing (e.g., amutilated body) or neutral (e.g., a building) pictures. Subjects returnedto the lab one week later and were asked to complete another WPsession. While neither initial learning nor 1-week retention of WPlearning was significantly affected, there was a robust enhancement ofrecognition memory for the arousing context (i.e., picture recognition)in which WP learning occurred. This suggested that the effects ofemotional arousal on procedural and declarative/episodic memorymay be dissociable. This potential dissociation between emotionalinfluence on declarative/episodic and cognitive skill learning, how-ever, is complicated by two factors: (a) this prior study examinedfurther WP learning as an index of memory, which can obscure howemotional arousal modulates retention and (b) procedural and declar-ative memory may have different time courses of forgetting. Withrespect to the latter, emotional influences on memory consolidationare typically manifested as enhanced retention/resistance to forgettingas delay between study and test increases, with relative maintenanceof memory for emotionally arousing events and more precipitousforgetting for neutral events (Kleinsmith & Kaplan, 1963; LaBar &Phelps, 1998; Sharot & Phelps, 2004). As various forms of implicitmemory are maintained in memory over long periods (Schacter,1987), it may be difficult to visualize EEM for procedural learning

because of limited forgetting over shorter study-test delays. Indeed, inour previous study (Steidl et al., 2006), preliminary data from a smallsubset (n � 5) of participants suggested EEM for skill learning mayonly be present after memory decay following long delays (3 months).

In the present study we followed up this preliminary observationby examining EEM for WP cognitive skill learning over a longretention delay (1.5 months) that permits sufficient forgetting. Asemotionally significant pictures evoke both sympathetic arousaland amygdala activation and have been linked to enhanced mem-ory for surrounding neutral contexts (Canli et al., 2000; Hamann etal., 1999; Anderson, Wais, & Gabrieli, 2006; Maratos, Dolan,Morris, Henson, & Rugg, 2001), we hypothesized that emotionalarousal and associated amygdala recruitment evoked by inter-leaved emotionally arousing pictures should enhance WP memorywhen tested at a long delay. This would provide evidence thatEEM extends to the long-term retention of skill learning.

Materials and Method

Participants

Participants were recruited through advertisements postedthroughout the University of Toronto campus as well as throughintroductory psychology classes. A total of 44 participants (agerange � 17–26) participated in the study which involved twosessions separated by �1.5 months. Participants were randomlyassigned to either the arousal condition (n � 21; 10 men), or theneutral condition (n � 23; 16 men). The average test-retest delaywas 46 � 3 days for the neutral and 46 � 2 days for the arousalcondition. One participant in the neutral condition was removedfrom analysis because of performing �2.5 SDs below the mean atthe end of training (22% accuracy).

The WP Task

All participants performed the weather prediction task on a com-puter (adapted from Knowlton et al., 1996). The WP task was de-signed using Super Lab (Version 2.0, Cedrus Inc.) and viewedthrough a 17 in. flat screen LCD monitor. Participants were presentedwith varying combinations of four types of cues depicting simplegeometric forms: 7 black squares, 10 black triangles, 9 black circles,or 13 black diamonds. On each trial combinations of one, two, or threeof these cues were presented side by side on the computer screen (14possible cue combinations; Table 1) and individuals were asked topredict the weather outcome as either sunshine or rain by pressing oneof two keys indicated on a keyboard. Each of the four cards wasassociated with an outcome of rain or sunshine with a fixed proba-bility. Feedback was given after each response to indicate a correct orincorrect response. In the event of a correct response, participantsheard a high pitch sound and were shown a happy face as well as atext message informing them that they were correct. In the event of anincorrect response participants heard a low pitch buzzer tone and wereshown a sad face as well as a text message informing them that theywere incorrect and indicating the correct answer. A total of 150 trialswere presented to each subject. Within each trial, a particular cardconfiguration was presented for 5 seconds and participants had torespond within this time. After the selection was entered, feedbackwas presented for 2 s. Five self-paced introductory instruction screensappeared at the beginning of the experiment.

13AROUSAL AND COGNITIVE SKILL LEARNING

Emotional arousal was manipulated by showing participantsemotionally arousing or neutral pictures between every third WPtrial (approximately every 20 s). Participants in both arousal andneutral groups were asked to evaluate the intensity of their emo-tional response to a presented picture, that is, the subjective cor-relate of peripheral physiological arousal and amygdala activity(Anderson et al., 2003; Canli et al., 2000; Hamann et al., 1999) inresponse to the question “How intense do you feel?” As arousalcommon to both negative and positive valence most stronglycharacterizes amygdala contributions to EEM (Hamann et al.,1999), we randomized presentations of positive and negative pic-tures, and as such were not interested in valence specific contri-butions to memory formation (Kensinger & Corkin, 2004). Thepictures were selected from the International Affective PictureSystems (IAPS, Lang, Bradley, & Cuthbert, 2005). A total of 50pictures were presented. In the arousal condition, 80% of thepictures were selected from those catalogued in the IAPS as havinghigh valence (20 positive and 20 negative) and high arousal, while20% of the pictures were neutral with low valence and low arousal.Inclusion of negative, positive, and neutral pictures was expectedto limit habituation and thus evoke more consistently intenseemotional responses for arousing pictures. Neutral pictures alsoserved as an additional measure of the influence of evoked emo-tional arousal on memory for the episodic picture context of WPlearning. For the neutral condition, all selected pictures wereneutral in valence and low in arousal. Neutral pictures depictedpeople engaging in everyday situations (e.g., writing, or lookingout of a window), scenery or flora (e.g., plants and trees), orcommon objects (e.g., rolling pin, fan, clothing items). In thispaper, the term “context” will be used to refer to the IAPS picturesintermittently presented during WP acquisition, as it provides adirect index of episodic memory for the events surrounding WPlearning.

Transfer Task

After completion of the WP task during Session 1, we admin-istered a transfer task to assess the subjects’ ability to state thesubjective probability of rain or sunshine (Reber, Knowlton, &Squire, 1996). Subjects were shown either each of the four singlecue cards individually, the six possible combinations of cue cardpairs, or the three possible combinations of cue card triplets (seeTable 1), and asked the question: If just this card/these cards areshowing, what percentage of time will it be sunny?

Retention Task

Participants were asked to return to the laboratory 1.5 monthsafter the initial session. To assess retention of previous WP learn-ing, participants were asked to complete 150 WP trials. Thissession was conducted as described for Session 1, except that therewas no feedback given or arousal manipulation due to interleavedpictures.

Picture Recognition Task

Following WP retention testing, participants were given a rec-ognition task for the pictures encountered during the first session,as an index of contextual memory. Seventy-five pictures wereshown in the recognition task, of which 50 had been previouslyseen in Session 1, and 25 were new foil items. The 25 foil itemswere matched thematically with previously seen items to increasememory discrimination difficulty (10 positive, 10 negative, and 5neutral). Participants in the neutral condition were presented with50 neutral old pictures and 25 neutral foil items. Subjects in bothgroups were asked to indicate whether or not a given picture hadbeen previously encountered with a yes/no response.

Data Acquisition and Analysis

Analyses described below were first conducted with genderincluded as a between-participants factor. However, in no case wasthere any main effect or interaction with sex, so data from men andwomen were collapsed for statistical analysis.

Picture Context Recognition

The picture recognition data was analyzed for accuracy bysubtracting the proportion of responses to new pictures (i.e., falsealarms) from the proportion of responses to old pictures (i.e., hits).

WP Task

Following prior studies, a response was scored as accurate if itwas associated with the most likely outcome. Performance on trialsthat had a 50% chance of being rain or sunshine were not scored,as they did not provide any indication of learning across trials. The150 trials were split into 5 blocks of 30 trials each. Consolidationof implicit learning on the WP task across the 1.5 month periodseparating Session 1 and the retrieval test in Session 2 was as-sessed by comparing the end of Session 1 (block 5) with Session2. Data were collapsed across the 150 trials in Session 2 as nofeedback was provided to allow a more pure measure of retention.

Table 1Weather Prediction Task Probability Structure Taken FromReber, Knowlton, and Squire (1996)

Cue no.Cue

arrangement Frequency Sun Rain P(Sunshine)

1 1100 4 4 0 12 1110 1 1 0 13 1000 7 6 1 .864 1010 4 3 1 .755 0100 5 3 2 .606 1001 2 1 1 .507 0110 2 1 1 .508 1011 2 1 1 .509 1101 2 1 1 .50

10 0010 5 2 3 .4011 0101 4 1 3 .2512 0001 7 1 6 .1413 0111 1 0 1 014 0011 4 0 4 0

Note. The numbers in the cue combination column refer from left to rightto cues 1 to 4: A 1 indicates that a cue is present on a trial, while a 0indicated that a cue is absent on a trial. Frequency indicates how often aparticular combination occurs out of 50 trials. Sunshine and rain indicatedhow often each outcome occurs. The final column indicates the probabilitythat an outcome (sunshine in this case) will occur.

14 STEIDL, RAZIK, AND ANDERSON

Transfer Task

The 14 different cue types were collapsed to represent 5 levelsof cue-outcome predictability (5, 33, 50, 68, and 95%). Transfertask performance was analyzed using a mixed measures ANOVAwith cue-outcome probability (5 to 95%) as the repeated measuresfactors and encoding condition (arousal vs. neutral) as the betweensubjects factor.

Results

Arousal Ratings

Subjects were asked to evaluate the intensity of their emotionalresponses to interleaved pictures during WP learning. An indepen-dent samples t test comparing self-reported arousal ratings in thearousing (collapsed across negative and positive) and neutral con-ditions revealed greater self-reported arousal in the arousal condi-tion (5.7 � 0.3 and 2.9 � 0.31 in arousal and neutral conditions,respectively), t(42) � 6.32, p � .0001, d � 1.95.

Picture Context Recognition

After 1.5 months subjects were given a recognition test forpicture contexts seen during initial WP learning. An independentsamples t test revealed there was a significantly higher number ofhits, t(42) � 5.02, p � .0001, d � 1.44, as well as a trend towardfewer false alarms, t(42) � 2.01, p � .051, in the arousing relativeto the neutral condition. Thus, mean corrected picture recognition(hits–false alarms) was significantly more accurate in the arousing(collapsed across picture types) relative to the neutral condition,t(42) � 5.39, p � .001, d � 1.66. Specifically, arousing picturesin the arousing condition (60% hits and 6.6% false alarms) werebetter remembered than neutral pictures in the neutral condition(44.5% hits and 9.5% false alarms), t(42) � 5.37, p � .001, d �1.66. Moreover, this enhanced memory extended to neutral pic-tures in the arousal condition (52.4% hits and 6.3% false alarms),which were better remembered than neutral pictures in the neutralcondition, t(42) � 3.47, p � .01, d � 1.07 (Figure 1a). Insummary, arousal was associated with enhanced memory for sur-rounding WP picture context.

Weather Prediction

WP acquisition. Data were submitted to a two-way ANOVAwith acquisition block (1–5) and encoding condition (arousal vs.neutral) as within and between subject factors, respectively. Amain effect of training block, F(4, 168) � 5.86, p � .001, �2 �0.122 confirmed that overall participants showed significant WPacquisition across trial blocks, increasing linearly from block 1 to5, F(1, 168) � 23.22, p � .001, �2 � 0.104. A main effect ofcondition revealed participants’ overall performance in the arousalcondition was significantly lower compared to the neutral condi-tion, F(1, 42) � 11.41, p � .002, �2 � 0.214 (Figure 1b). Inaddition, there was a marginally significant interaction betweenarousal condition and training block, F(4, 168) � 2.19, p � .07.Separate analysis of the two encoding conditions confirmed suc-cessful WP learning across trial blocks, with performance increas-ing from beginning to end of acquisition (block 1 to 5), in both theneutral, F(1, 88) � 21.31, p � .001, �2 � 0.145 and the arousal

condition, F(1, 80) � 4.16, p � .05, �2 � 0.064, but statisticallymore robust in the former.

Transfer task. To estimate the influence of arousal on trans-ference of declarative knowledge (i.e., those that participants couldverbally state) regarding contingencies learned in the WP task(Reber et al., 1996), participants estimated the percent likelihoodof sunshine for each of the cue card combinations. Transfer taskdata from one subject from the arousal condition was lost. Esti-mation of sunshine depended on cue-outcome predictability, F(4,164) � 25.9, p � .001, �2 � .387. However, there was no maineffect or interaction with encoding condition, Fs � 1.10. Bothencoding conditions tended to overestimate the least predictive andunderestimate the most predictive events, even those with cue-outcome contingencies with �95% diagnosticity, F(1, 41), p �.63, (Percent max sunshine �95%, neutral � 65.28%, SD �25.37; arousal � 61.73%, SD � 22.21) where explicit memorywould likely have been greatest.

Retention. Participants returned 1.5 months later and wereasked to complete 150 WP trials without any feedback. This

A

b1 b2 b3 b4 b540

50

60

70

80

WP

per

cent

cor

rect

± s

em

Training block

Arousal

Neutral

20

30

40

50

60

20

30

40

50

60

Cor

rect

ed r

ecog

nitio

n (%

) ±

sem

Encoding condition

Arous-Neut

Neutral

B

Study Test40

50

60

70

80

Per

cent

cor

rect

Time

C D

Study Test40

50

60

70

80

Per

cent

cor

rect

Arousal

Figure 1. (A) Remote (1.5 months) picture recognition memory for thearousal and neutral encoding conditions. Arousal � arousing pictures in thearousal condition; Arous-Neut � neutral pictures in the arousal condition;Neutral � neutral pictures in the neutral condition. (B) WP acquisitioncurves for neutral and arousal conditions in Session 1. (C) Retention of WPlearning across a 1.5 month delay. Study refers to performance from thefinal training block in Session 1. Test � retention test. (D) Retention of WPlearning across a 1.5 month delay in participants matched for performanceat the end of Session 1. Study refers to performance from the final trainingblock in Session 1. Test � retention test.


allowed assessment of the effects of arousal level during initial WPacquisition on long-term retention. Neither subjects in the arousal,F(4, 80) � 1.88 p � .12, nor neutral conditions, F(4, 88) � 1.61p � .17, showed significant differences across retention trialblocks; thus, data from the retrieval session was collapsed forfurther analyses. To examine retention and its interaction withencoding condition, data were submitted to a two-way ANOVAwith encoding condition (arousal vs. neutral) and delay (block 5 ofDay 1 vs. 1.5 months collapsed across blocks) as between andwithin subject factors, respectively. There was a main effect ofsession, F(1, 42) � 13.77, p � .001, �2 � 0.247, demonstratingthat the 1.5 month delay resulted in significant forgetting of WPlearning. This main effect of session was modified by a significantinteraction between session and encoding condition, F(1, 42) �10.66, p � .003, �2 � 0.202. While participants in the neutralcondition showed substantial forgetting across the 1.5 month delayinterval, F(1, 22) � 22.58, p � .001, �2 � 0.507, the arousalcondition showed no evidence of forgetting across the same timeperiod, F � 1 (Figure 1c).

To examine whether this effect of arousal on retention was anartifact of enhanced performance in the neutral condition duringinitial learning, we removed the top performers in the neutralcondition (n � 12, leaving 11) and the bottom (n � 11, leaving 11)performers in the arousal condition to equate performance at theend of acquisition (Figure 1d). This performance matching resultedin statistically equivalent levels of overall learning at the end ofacquisition (arousal � 62.4, SE � 3.1 vs. neutral � 64.8, SE �2.3; F(1, 32) � 1), as well as matched performance on two indicesof declarative memory: (a) WP acquisition of the most highlydiagnostic cues (�95% diagnosticity), F(1, 32) � 1; and (b)verbally declared knowledge in the transfer task performance forhighly diagnostic cues, F(1, 31) � 1. Despite this statistical match-ing between the arousal and neutral conditions at the end oftraining, there remained a significant effect of delay, F(1, 32) �8.23, p � .008, �2 � 0.204 as well as an interaction with encodingcondition, F(1, 32) � 6.12, p � .02, �2 � 0.160. Participants in theneutral, F(1, 12) � 10.24, p � .008, �2 � 0.460, but not thearousal condition, F � 1, demonstrated robust forgetting 1.5months later (Figure 1d). Thus, greater retention of WP learning inthe arousal condition was not a reflection of the poorer perfor-mance during initial training.

Relation Between WP Acquisition and Retention

We next more closely examined individual data to better under-stand how emotional arousal may potentially alter retention of WPlearning. Inspection of the individual differences in retention revealedthat half of the arousal group showed evidence of WP performanceincreasing over the 1.5 month delay, a much higher level of occur-rence relative to the neutral group (52.4 vs. 8.7% of participants,�2(1) � 10.06, p � .002). Greater incidence of this hypermnesia waspresent even after matching performance between the arousal andneutral groups at the end of acquisition (52.4 vs. 9.1% of participants,�2(1) � 7.02, p � .009). This greater incidence of memory improve-ment with time was thus not a reflection of the poorer performanceduring initial training. Furthermore, performance at the end of acqui-sition in performance-matched subjects was significantly correlatedwith greater retention in the arousal (r � .47, p � .05), but not theneutral group (r � .16, p � .5). This suggests that greater initial

learning was associated with enhanced remote retention in the arousalbut not neutral encoding conditions.

Relation Between Picture Context Memoryand WP Retention

We next assessed the relationship between WP retention and mem-ory for WP picture contexts. If enhanced WP retention in the arousalcondition was supported by episodic/declarative memory, then WPretention may correlate with enhanced memory for the picture con-text. We found that arousal-enhanced WP retention was independentfrom picture context memory, r � .165, F(1, 19) � 1. Further, thisindependence characterized memory for WP retention and picturecontext memory not only for intrinsically arousing pictures, r � .14,F(1, 19) � 1, but also for enhanced memory of neutral picturesstudied in an arousing context, r � .19, F(1, 19) � 1. This dissoci-ation did not reflect insufficient variability on either WP retention(SD � 17.8%, min–max spread � 64.5%) or picture context recog-nition (SD � 11.6%, min–max spread � 40%). Thus, evoked emo-tions during encoding were associated with a parallel but largelyindependent enhancement of remote memory for WP learning andepisodic/declarative memory for its associated picture context.

Discussion

Presentation of emotionally arousing pictures at encodingslowed initial acquisition of probabilistic classification learning inthe WP tasks, a form of cognitive skill learning. By contrast, whileparticipants in the neutral encoding condition showed substantialforgetting over a 1.5 month period, participants in the arousalcondition maintained WP performance, suggesting that emotionalarousal enhanced memory retention, decreasing the rate of mem-ory decay. This decreased rate of memory decay was found evenafter matching initial acquisition in the arousal and neutral condi-tions. Emotional arousal experienced during WP learning was alsoassociated with enhanced remote memory for the picture contextsin which the WP task was acquired. These results provide evidencethat the modulatory influences of emotional arousal extend tocognitive skill learning.

The current results are consistent with animal studies showingamygdala modulation of different mnemonic tasks that putativelyoriginate from different neural systems (McGaugh, 2004; McGaughet al., 2002; Packard & Cahill, 2001; Packard et al., 1994). However,more recent work suggests the distinction between a hippocampal/medial temporal lobe-based declarative and a striatal-based proce-dural memory system is less clear. Evidence suggests that thedorsomedial portion of the rat striatum, corresponding to thecaudate in primates, is involved in flexible place learning (Devan,McDonald, & White, 1999; Devan & White, 1999; Yin & Knowl-ton, 2004; Yin, Knowlton, & Balleine, 2005; Yin, Ostlund, Knowl-ton, & Balleine, 2005). For example, lesions confined to thedorsomedial portion of the striatum produce a preference forcue-based responding in a water maze task (Devan et al., 1999).Thus, it is unclear to what extent different forms of memory can beanatomically separated strictly in terms of medial temporal lobeand dorsal striatal systems. Because the cues and their associatedoutcomes in the weather prediction task are probabilistically re-lated, it is difficult to explicitly encode the relation between theindividual stimuli and their associated outcomes. Rather, the sub-


ject must learn a series of stimulus-response association acrossmany trials. As such, the task emphasizes nondeclarative memoryprocesses. However, patients with mild Parkinson’s disease whocan perform the task normally relative to controls, show medialtemporal lobe cortical activation unlike controls (Moody,Bookheimer, Vanek, & Knowlton, 2004), suggesting that bothstriatal and declarative-based memory systems can contribute tolearning the WP task. Manipulations designed to disrupt implicitlearning processes have also been shown to have no effect on WPlearning (Price, 2009). In neurologically intact controls, the twobrain areas may in fact compete with one another during taskacquisition. There is also evidence of a change in the underlyingmemory systems used with progressive WP training. Patient stud-ies show that early on during training learning depends on an intactstriatum, while later on, learning depends on an intact hippocam-pus (Knowlton et al., 1994). Functional imaging studies supportthis switch in memory systems, showing very early activation inthe medial temporal lobe quickly replaced by engagement of thestriatum (Poldrack et al., 2001) followed by later activation of thehippocampus (Poldrack et al., 1999; Iaria, Petrides, Dagher, Pike,& Bohbot, 2003). In the current study, the number of trials usedmay have recruited both memory systems across the course oftraining. Thus, for these reasons it is possible that the enhancedlong-term retention observed in the arousal condition in the currentstudy reflects arousal enhancement of both nondeclarative anddeclarative memory mechanisms. However, there are a fewsources of evidence to suggest the arousal enhancement of WPretention reflects, at least in part, a reliance on a nondeclarativeprocedural learning mechanism.

The precise timing of a switch between declarative and proce-dural systems is unclear in the present study. Recent data show thatattentional distraction from a secondary task impairs overall prob-abilistic classification performance and declarative knowledge ofcue-outcome associations (Foerde, Poldrack, & Knowlton, 2007)and further blocks the shift from procedural/striatal to declarative/medial temporal lobe memory systems (Foerde, Knowlton, &Poldrack, 2006). This is consistent with data showing that stressduring learning facilitates use of procedural relative to declarativelearning strategies in both humans (Schwabe et al., 2007) andanimals (Packard & Wingard, 2004; Wingard & Packard, 2008). Inthe present study, the intermittent presentations of picture contextsand ratings between every few trials of WP training may haveserved a similar role. Indeed, verbal statements in the transfer taskat the end of Session 1 regarding the outcomes of even the mostdiagnostic cues (�95%), where declarative memory should bemost evident, revealed substantial underestimation in both theneutral and arousal conditions. Further, we assessed the relation-ship between WP retention and memory for surrounding picturecontexts. Individual differences in emotion enhanced memory forpicture contexts were not associated with greater WP retention.This suggests retention of probabilistic classification learning wasindependent of retrieving information from the original studycontext, similar to implicit learning in medial temporal lobe am-nesiacs (Knowlton et al., 1996; Reber et al., 1996).

Arousal during initial encoding clearly had a disruptive effect onWP acquisition, consistent with our previous study (Steidl et al.,2006). Similarly, Thomas and LaBar (2008) have shown thatarousal during encoding specifically leads to suboptimal strategyuse, particularly in those participants that found the emotional

stimuli most anxiety provoking. Studies of explicit memory some-times show an initial disrupting effect of emotional arousal onmemory for neutral events at immediate test delays (Hurlemann etal., 2007; Kleinsmith & Kaplan, 1963; Strange, Hurlemann, &Dolan, 2003), but enhancement at longer delays (Anderson et al.,2006; Kleinsmith & Kaplan, 1963; Maratos et al., 2001). As theWP task involves simultaneous acquisition and retrieval, this maypartly contribute to the observed initial deficit in the arousalcondition. Our data suggest that the effects of emotional arousal onWP learning may be biphasic, characterized by an initial disruptionin acquisition that is contrasted with a later retention enhancement.

Attentional distraction during WP acquisition may have beengreater in the arousal condition (e.g., Anderson, 2005; Most, Chun,Widders, & Zald, 2005; Schimmack & Derryberry, 2005), as well asdifferential engagement of declarative and nondeclarative learningmechanisms between neutral and arousal groups during encoding(Foerde et al., 2006, 2007), accounting for significantly poorerinitial performance. Because declarative memory may be moresusceptible to forgetting than skill learning (Steidl et al., 2006),this could result in differential patterns of forgetting as shownin the present results. To address this, we performance matchedthe arousal and neutral groups at the end of acquisition. Perfor-mance was also matched on highly diagnostic cue types wheredeclarative knowledge would be greatest, as well as verbally statedknowledge regarding cue-outcome contingencies. As such, therelative contributions of declarative and procedural learning mech-anisms should have been similar across the two groups. Despiteequivalent WP performance at the end of training, these matchedgroups demonstrated markedly distinct mnemonic fates—nearcomplete retention for the arousal group (Figure 1d), but noevidence of retention in the neutral condition after the 1.5 monthdelay.

Similarly, arousal levels experienced during delayed testing mayhave facilitated differential engagement of procedural and declar-ative strategies between neutral and arousing groups (Schwabe etal., 2007). Because retention testing took place in the same envi-ronment in which subjects were exposed to arousing picture stim-uli during acquisition, it is possible that reexposure to this envi-ronment during retention testing may have differentially increasedarousal levels in the arousal group. Studies of pure contextual fearconditioning in humans are very rare. One study has shown thathumans acquire contextual fear conditioning (assessed by bothGSR and subjective anxiety ratings) to a distinct virtual realityenvironmental context paired with unpredictable footshock withinan experimental session (Alvarez, Biggs, Chen, Pine, and Grillon,2008). However, whether contextual fear conditioning is evidentupon delayed exposure to the shock-paired context was not as-sessed. In the absence of physiological measures of arousal (i.e.,galvanic skin response) in the current studies, it is difficult todetermine the extent to which such conditioned arousal induced byreexposure to the training environment may have differentiallycontributed to WP performance in the arousal group. Undoubtedly,future studies would benefit from including measures of physio-logical arousal during both retention and acquisition.

Whether the neuroanatomical basis of the memory observed in thecurrent study is of pure striatal or mixed striatal/medial temporalorigin, the present results do appear to reflect a form of memorydistinct from previous demonstrations of EEM (Cahill et al., 1996;Cahill, Prins, Weber, & McGaugh, 1994; Canli et al., 2000; Dolcos,


LaBar, & Cabeza, 2004). Whereas studies of EEM largely focus onshorter retention intervals and recollection of the episodic context inwhich emotions are experienced (Cahill et al., 1994, 1996; Dolcos,LaBar, & Cabeza, 2004; Kensinger & Corkin, 2004) the present studydemonstrates that emotional arousal enhances remote retention of acognitive skill, putatively a nonepisodic form of memory. Emotionsthus not only modulate the subjective experience of remembering butpotentially the neural mechanisms that support a form of memory thatis free from the episodic context in which it was acquired (Bechara etal., 1995; Tulving, 1987).

In accordance with rat models of EEM (McGaugh, 2004), wehypothesize that the amygdala plays a central role in the orchestrationof arousal enhancement of multiple forms of memory, as shown herein the EEM for cognitive skill learning and its episodic picturecontext. In rats the dorsal striatum and the hippocampus both receiveanatomical projections from the basolateral amygdala (Finch et al.,1986; Kelley, Domensick, & Nauta, 1982; Pikkarainen, Rönkkö,Savander, Insausti, & Pitkanen, 1999; Price, 2003; Russchen, Bakst,Amaral, & Price, 1985; Russchen & Price, 1984) via the stria termi-nalis and numerous studies converge on the appreciation that theamygdala plays an important role in the emotional modulation ofmemory retention (e.g., McGaugh, 2004). In humans, structural equa-tion modeling of neuroimaging data has demonstrated the neuralbases of EEM for hippocampal-declarative memory (Kilpatrick &Cahill, 2003), supported by anatomical connectivity between theamygdala and surrounding anterior temporal lobe (Powell et al.,2004). By contrast, knowledge of amygdala-striatal functional andanatomical connectivity in humans is limited. The present findingssuggest modulatory influences on long-term memory for a probabi-listic classification task, a putatively striatal based form of cognitiveskill learning. Future work will need to establish whether the presentdemonstration of enhanced remote retention of cognitive skill learn-ing does indeed reflect amygdala modulatory influences on striatalmemory systems.

References

Alvarez, R. P., Biggs, A., Chen, G., Pine, D. S., & Grillon, C. (2008).Contextual fear conditioning in humans: Cortical-hippocampal andamygdala contributions. Journal of Neuroscience, 28, 6211–6219.

Anderson, A. K. (2005). Affective influences on the attentional dynamicssupporting awareness. Journal of Experimental Psychology: General,134, 258–281.

Anderson, A. K., Christoff, K., Stappen, I., Panitz, D., Ghahremani, D. G.,Glover, G., . . . Sobel, N. (2003). Dissociated neural representations ofintensity and valence in human olfaction. Nature Neuroscience, 6, 196–202.

Anderson, A. K., Wais, P. E., & Gabrieli, J. D. (2006). Emotion enhancesremembrance of neutral events past. Proceedings of the National Acad-emy of Sciences, USA, 103, 1599–1604.

Bechara, A., Tranel, D., Damasio, H., Adolphs, R., Rockland, C., &Damasio, A. (1995). Double dissociation of conditioning and declarativeknowledge relative to the amygdala and hippocampus in humans. Sci-ence, 269, 1115–1118.

Cahill, L., Haier, R. J., Fallon, J., Alkire, M. T., Tang, C., Keator, D., . . .McGaugh, J. L. (1996). Amygdala activity at encoding correlated withlong-term, free recall of emotional information. Proceedings of theNational Academy of Sciences, USA, 93, 8016–8021.

Cahill, L., & McGaugh, J. L. (1999). Mechanisms of emotional arousal andlasting declarative memory. Trends in Neuroscience, 21, 294–299.

Cahill, L., Prins, B., Weber, M., & McGaugh, J. L. (1994). �-Adrenergicactivation and memory for emotional events. Nature, 371, 702–704.

Canli, T., Zhao, Z., Brewer, J., Gabrieli, J. D., & Cahill, L. (2000).Event-related activation in the human amygdala associates with latermemory for individual emotional experience. Journal of Neuroscience,20, RC99.

Cohen, N. J., & Squire, L. R. (1980). Preserved learning and retentionpattern-analyzing skill in amnesia: Dissociation of knowing how andknowing that. Science, 210, 207–210.

Devan, B. D., McDonald, R. J., & White, N. M. (1999). Effects of medialand lateral caudate-putamen lesions on place- and cue-guided behaviorsin the water maze: Relation to thigmotaxis. Behavioral Brain Research,100, 5–14.

Devan, B. D., & White, N. M. (1999). Parallel information processing inthe dorsal striatum: Relation to hippocampal function. Journal of Neu-roscience, 19, 2789–2798.

Dolcos, F., LaBar, K. S., & Cabeza, R. (2004). Interaction between theamygdala and medial temporal lobe system predicts better memory foremotional events. Neuron, 42, 855–863.

Dougal, S., Phelps, E. A., & Davachi, L. (2007). The role of medialtemporal lobe in item recognition and source recollection of emotionalstimuli. Cognitive Affective and Behavioral Neuroscience, 7, 233–242.

Finch, D. M., Wong, E. E., Derian, E. L., Chen, X. H., Nowlin-Finch,N. L., & Brothers, L. A. (1986). Neurophysiology of limbic systempathways in the rat: Projections from the amygdala to the entorhinalcortex. Brain Research, 370, 273–284.

Foerde, K., Knowlton, B. J., & Poldrack, R. A. (2006). Modulation ofcompeting memory systems by distraction. Proceedings of the NationalAcademy of Sciences USA, 103, 11778–11783.

Foerde, K., Poldrack, R. A., & Knowlton, B. J. (2007). Secondary-taskeffects on classification learning. Memory & Cognition, 35, 864–874.

Hamann, S. (2001). Cognitive and neural mechanisms of emotional mem-ory. Trends in Cognitive Science, 5, 394–400.

Hamann, S. B., Ely, T. D., Grafton, S. T., & Kilts, C. D. (1999). Amygdalaactivity related to enhanced memory for pleasant and aversive stimuli.Nature Neuroscience, 2, 289–293.

Hurlemann, R., Matusch, A., Hawellek, B., Klingmuller, D., Kolsch, H.,Maier, W., & Dolan, R. J. (2007). Emotion-induced retrograde amnesiavaries as a function of noradrenergic-glucocorticoid activity. Psychop-harmacology, 194, 261–269.

Iaria, G., Petrides, M., Dagher, A., Pike, B., & Bohbot, V. D. (2003).Cognitive strategies dependent on the hippocampus and caudate nucleusin human navigation: Variability and change with practice. Journal ofNeuroscience, 23, 5945–5952.

Kelley, A. E., Domensick, V. B., & Nauta, W. J. (1982). The amygdalos-triatal projection in the rat: An anatomical study by anterograde andretrograde tracing methods. Neuroscience, 7, 615–630.

Kensinger, E. A., & Corkin, S. (2004). Two routes to emotional memory:Distinct neural processes for valence and arousal. Proceedings of theNational Academy of Sciences, USA, 101, 3310–3315.

Kensinger, E. A., & Schacter, D. L. (2005). Retrieving accurate anddistorted memories: Neuroimaging evidence for effects of emotion.Neuroimage, 27, 167–177.

Kilpatrick, L., & Cahill, L. (2003). Amygdala modulation of parahip-pocampal and frontal regions during emotionally influenced memorystorage. Neuroimage, 20, 2091–2099.

Kleinsmith, L. J., & Kaplan, S. (1963). Paired-associate learning as afunction of arousal and interpolated interval. Journal of ExperimentalPsychology, 65, 190–193.

Knowlton, B. J., Mangels, J. A., & Squire, L. R. (1996). A neostriatal habitlearning system in humans. Science, 273, 1399–1402.

Knowlton, B. J., Squire, L. R., & Gluck, M. A. (1994). Probabilisticclassification learning in amnesia. Learning and Memory, 1, 106–120.

LaBar, K. S., & Cabeza, R. (2006). Cognitive Neuroscience of emotionalmemory. Nature Reviews Neuroscience, 7, 54–64.

LaBar, K. S., & Phelps, E. A. (1998). Arousal-mediated memory consol-


idation: Role of the medial temporal lobe in humans. PsychologicalScience, 9, 490–493.

Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (2005). Internationalaffective picture system (IAPS): Affective ratings of pictures and in-struction manual. Technical Report No. A-6, University of Florida,Gainesville, FL.

Maratos, E. J., Dolan, R. J., Morris, J. S., Henson, R. N., & Rugg, M. D.(2001). Neural activity associated with episodic memory for emotionalcontext. Neuropsychologia, 39, 910–920.

McDonald, R. J., & White, N. M. (1993). A triple dissociation of memorysystems: Hippocampus, amygdala and dorsal striatum. Behavioral Neu-roscience, 107, 3–22.

McGaugh, J. L. (2004). The amygdala modulates the consolidation ofmemories of emotionally arousing experiences. Annual Review of Neu-roscience, 27, 1–28.

McGaugh, J. L., McIntyre, C. K., & Power, A. E. (2002). Amygdalamodulation of memory consolidation: Interaction with other brain sys-tems. Neurobiology of Learning and Memory, 87, 539–552.

McIntyre, C. K., Power, A. E., Roozendaal, B., & McGaugh, J. L. (2003).Role of the basolateral amygdala in memory consolidation. Annals of theNew York Academy of Sciences, 985, 273–293.

Moody, T. D., Bookheimer, S. Y., Vanek, Z., & Knowlton, B. J. (2004). Animplicit learning task activates medial temporal lobe in patients withParkinson’s disease. Behavioral Neuroscience, 118, 438–442.

Most, S. B., Chun, M. M., Widders, Z. M., & Zald, D. H. (2005). Attentionalrubbernecking: Cognitive control and personality in emotion-induced blind-ness. Psychonomic Bulletin and Review, 12, 654–661.

Packard, M. G., & Cahill, L. (2001). Affective modulation of multiplememory systems. Current Opinion in Neurobiology, 11, 752–756.

Packard, M. G., Cahill, L., & McGaugh, J. L. (1994). Amygdala modula-tion of hippocampal-dependent and caudate nucleus-dependent memoryprocesses. Proceedings of the National Academy of Sciences, USA, 91,8477–8481.

Packard, M. G., Hirsh, R., & White, N. M. (1989). Differential effects offornix and caudate nucleus lesions on two radial maze tasks: Evidencefor multiple memory systems. Journal of Neuroscience, 9, 1465–1472.

Packard, M. G., & Wingard, J. C. (2004). Amygdala and emotionalmodulation of the relative use of multiple memory systems. Neurobiol-ogy of Learning and Memory, 82, 243–252.

Phelps, E. A., LaBar, K. S., & Phelps, E. A. (1997). Memory for emotionalwords following unilateral temporal lobectomy. Brain and Cognition,35, 85–109.

Pikkarainen, M., Rönkkö, S., Savander, V., Insausti, R., & Pitkanen, A.(1999). Projections from the lateral, basal, and accessory basal nuclei ofthe amygdala to the hippocampal formation in rat. Journal of Compar-ative Neurology, 403, 229–260.

Poldrack, R. A., Clark, J., Pare-Blagoev, E. J., Shohamy, D., CresoMoyano, J., Myers, C., & Gluck, M. A. (2001). Interactive memorysystems in the human brain. Nature, 546–550.

Poldrack, R. A., Prabhakaran, V., Seger, C. A., & Gabrieli, J. D. (1999).Striatal activation during acquisition of a cognitive skill. Neuropsychol-ogy, 13, 564–574.

Powell, H. W., Guye, M., Parker, G. J., Symms, M. R., Boulby, P., Koepp,M. J., . . . Duncan, J. S. (2004). Noninvasive in vivo demonstration ofthe connections of the human parahippocampal gyrus. Neuroimage, 22,740–747.

Price, A. L. (2009). Distinguishing the contributions of implicit and ex-plicit processes to performance of the weather prediction task. Memory& Cognition, 37, 210–222.

Price, J. L. (2003). Comparative aspects of amygdala connectivity. Annalsof the New York Academy of Sciences, 985, 50–58.

Reber, P. J., Knowlton, B. J., & Squire, L. R. (1996). Dissociable proper-ties of memory systems: Differences in the flexibility of declarative andnondeclarative knowledge. Behavioral Neuroscience, 110, 861–871.

Roozendaal, B., Castello, N. A., Vedana, G., Barsegyan, A., & McGaugh,J. L. (2008). Noradrenergic activation of the basolateral amygdala mod-ulates consolidation of object recognition memory. Neurobiology ofLearning and Memory, 90, 576–579.

Russchen, F. T., Bakst, I., Amaral, D. G., & Price, J. L. (1985). Theamygdalostriatal projections in the monkey. An anterograde tracingstudy. Brain Research, 329, 241–257.

Russchen, F. T., & Price, J. L. (1984). Amygdalostriatal projections in therat. Topographical organization and fiber morphology using the lectinPHA-L as an anterograde tracer. Neuroscience Letters, 47, 15–22.

Schacter, D. L. (1987). Implicit memory: History and current status.Journal of Experimental Psychology, 13, 412–423.

Schacter, D. L., & Tulving, E. (1994). Memory systems. Cambridge: MITPress.

Schimmack, U., & Derryberry, D. (2005). Attentional interference effectsof emotional pictures: Threat, negativity, or arousal? Emotion, 5, 55–66.

Schwabe, L., Oitzl, M. S., Philippsen, C., Richter, S., Bohringer, A.,Wippich, W., & Schachinger, H. (2007). Stress modulates the use ofspatial versus stimulus-response learning strategies in humans. Learningand Memory, 14, 109–116.

Sharot, T., Delgado, M. R., & Phelps, E. A. (2004). How emotion enhancesthe feeling of remembering. Nature Neuroscience, 7, 1376–1380.

Sharot, T., & Phelps, E. A. (2004). How arousal modulates memory:Disentangling the effects of attention and retention. Cognitive Affectiveand Behavioral Neuroscience, 4, 294–306.

Steidl, S., Mohi-uddin, S., & Anderson, A. K. (2006). Effects of emotionalarousal on multiple memory systems: Evidence from declarative andprocedural learning. Learning and Memory, 13, 650–658.

Strange, B. A., Hurlemann, R., & Dolan, R. J. (2003). An emotion-inducedretrograde amnesia in humans is amygdala- and beta-adrenergic-dependent. Proceedings of the National Academy of Sciences, USA, 100,13626–13631.

Thomas, L. A., & LaBar, K. S. (2008). Fear relevancy, strategy use, andprobabilistic learning of cue-outcome associations. Learning and Mem-ory, 15, 777–784.

Tulving, E. (1987). Multiple memory systems and consciousness. HumanNeurobiology, 6, 67–80.

White, N. M., & McDonald, R. J. (2002). Multiple parallel memorysystems in the brain of the rat. Neurobiology of Learning and Memory,77, 125–184.

Wingard, J. C., & Packard, M. G. (2008). The amygdala and emotionalmodulation of competition between cognitive and habit memory. Be-havioral Brain Research, 193, 126–131.

Yin, H. H., & Knowlton, B. J. (2004). Contributions of striatal subregionsto place and response learning. Learning and Memory, 11, 459–463.

Yin, H. H., Knowlton, B. J., & Balleine, B. W. (2005). Blockade of NMDAreceptors in the dorsomedial striatum prevents action-outcome learningin instrumental conditioning. European Journal of Neuroscience, 22,505–512.

Yin, H. H., Ostlund, S. B., Knowlton, B. J., & Balleine, B. W. (2005). Therole of the dorsomedial striatum in instrumental conditioning. EuropeanJournal of Neuroscience, 22, 513–523.

Received September 14, 2009Revision received March 12, 2010

Accepted May 5, 2010 �


em enhanced retention

Documents