the influence of trait anxiety on autonomic response and cognitive performance during an...

DEPRESSION AND ANXIETY 23:210–219 (2006)

Research Article

THE INFLUENCE OF TRAIT ANXIETY ON AUTONOMICRESPONSE AND COGNITIVE PERFORMANCE DURING

AN ANTICIPATORY ANXIETY TASK

Jennifer Barrett, Ph.D.� and Jorge L. Armony, Ph.D.

The interaction between emotion and cognition is thought to be intimately involvedin the development and maintenance of anxiety disorders. In a set of studies, weinvestigated whether trait anxiety modulates cognitive performance and autonomicactivity during an anticipatory anxiety task. Participants completed a letter-sizedecision-making task with two alternating 28–32 s background screen color-blocks.One of the colors was associated with the presentation of an aversive noise[unconditioned stimulus (UCS)]. Participants were aware of the background colorthat would (CTX1) and would not (CTX�) be paired with the UCS but did notknow when or how often the UCS would be presented. Two experiments wereconducted. In Experiment 1, the UCS was presented during the decision-makingtask in the CTX1 color-blocks using a partial reinforcement schedule. Differentnoises were presented each time to increase unpredictability and preventhabituation. In Experiment 2, the UCS was never presented during thedecision-making task. Results suggested that only the paradigm used inExperiment 1 was successful in eliciting anticipatory anxiety. In Experiment 1,continuously measured skin conductance response (SCR) data suggested thatanxiety was significantly greater during CTX1 compared to CTX� trials; no SCRdifferences were found between high and low trait-anxious participants. Resultsfurther indicated that high trait-anxious participants responded significantly fasteron the decision-making task during CTX1 compared to CTX� trials, whereas lowtrait-anxious participants displayed the opposite pattern. Our results reveal aninteresting dissociation between the effects of individual differences in trait anxietyon autonomic activity and cognitive performance during an anticipatory anxietytask. Depression and Anxiety 23:210–219, 2006. & 2006 Wiley-Liss, Inc.

Key words: individual differences; anxiety vulnerability; context conditioning;cognition; skin conductance response

INTRODUCTIONIt has been suggested that the mechanisms underlyinganxiety are closely related to the fear response [Grillon,2002; Lang et al., 2000; LeDoux, 1996]. The purposeof the fear response, a biological process that has beenhighly conserved across evolution, is to detect environ-mental threat rapidly and respond appropriately.Response to a feared stimulus involves the recruitmentand coordination of cognitive, motor, autonomic, andendocrine systems [Lang et al., 2000; LeDoux, 1996].However, whereas fear is typically a highly organized,acute response to an explicit threat–object, for example,an angry bear, in anxiety, the response is generally more

Published online 1 March 2006 in Wiley InterScience (www.

interscience.wiley.com).

DOI 10.1002/da.20143

Received for publication 19 July 2005; Revised 23 August 2005;

Accepted 27 September 2005

Contract grant sponsor: Canadian Institutes of Health Research

(operating grant to Jorge Armony).

�Correspondence to: Jorge Armony, Douglas Hospital Research

Centre, 6875 LaSalle Boulevard, Verdun, Quebec, Canada H4H

1R3.

E-mail: [email protected]

Douglas Hospital Research Centre and Department of Psy-

chiatry, McGill University, Montreal, Quebec, Canada

rr 2006 Wiley-Liss, Inc.

diffuse, longer lasting, and often without an explicit cue[Lang et al., 2000; LeDoux, 1996]. In the short-term,acute anxiety may represent a relatively healthyreaction to a stressor or threat, providing input andgenerating bodily signals to guide behavior. In the longterm, nonetheless, chronic anxiety can be debilitatingand result in great impairment in everyday functioning.

Trait anxiety reflects an individual’s predisposition toexperience or display anxiety-relevant thoughts, feel-ings, and behaviors in everyday life; high levels of traitanxiety are often found in individuals with anxietydisorders [Chambers et al., 2004; Kabacoff et al., 1997;Oei et al., 1990; Plehn and Peterson, 2002]. Therefore,much preclinical research has focused on identifyingbehavioral and physiological response patterns asso-ciated with high trait anxiety in the presence of fear-relevant stimuli. In particular, researchers have exten-sively studied the effects of anxiety on attention,employing a variety of paradigms, including the Stroop[e.g., Egloff and Hock, 2001; Mogg et al., 2000],probe-detection [e.g., Ioannou et al., 2004; Mogg et al.,2000], and attentional-cueing tasks [e.g., Fox et al.,2002], to examine the influence of briefly presentedthreat stimuli on processes such as shifting of attention,task interference, and attention disengagement.Whereas this body of work has been highly informativeas to the role of trait anxiety in the cognitive processesrelated to threat detection and the ensuing automaticresponses, less is known about how trait anxiety mayinfluence cognitive and physiological activity duringsustained anticipatory anxiety in response to a threatwhose occurrence cannot be fully predicted. This is ofparticular relevance, because aversive stimuli tend to bemore anxiogenic when they are unpredictable [Minekaand Kihlstrom, 1978], and individuals with anxietydisorders have difficulty coping with aversive eventsthat cannot be predicted [Barlow, 2000].

Anxiety-related autonomic activity has also beenwidely studied [see Arena and Hobbs, 1995; Carilloet al., 2001; Fuller, 1992; Gonzalez-Bono et al., 2002;Harrison and Turpin, 2003; Hubert and de Jong-Meyer, 1992; Stamps et al., 1979]. Key to many anxietydisorders are changes in the sympathetic branch of theautonomic nervous system (ANS), responsible for the‘‘fight-or-flight’’ response [for reviews, see Berntsonet al., 1998; Charney and Deutch, 1996; Hoehn-Saricand McLeod, 1988]. In fear, stress, and anxiety, theANS involuntarily induces change in the internalenvironment to respond to perceived or real danger,resulting in a cascade of physiological events, includingincreased heart rate and blood pressure, dilation of thepupils, trachea, and bronchi, and regulation of bodytemperature via increased sweat gland output [Bremneret al., 1996; Chrousos, 1998]. Unfortunately, in anxiety,the absence of a real threat or the inappropriateness offight-or-flight behavior renders these physiologicalchanges a source of great discomfort, contributing tofurther anxiety and impairing social and occupationalfunctioning. Although there is little argument that the

experience of anxiety is associated with stereotypicalchanges in ANS activity, what is less clear is the extentto which those who are predisposed to anxiety (e.g.,high trait-anxious) may display an autonomic signaturethat is different from that of persons not predisposed toanxiety either at rest or when anxious. For example,previous studies have indicated that physiologicalresponses in high trait-anxious individuals duringanticipatory anxiety were larger [Gonzalez-Bonoet al., 2002], smaller [Wilken et al., 2000], or similar[Mauss et al., 2003] to those of low trait-anxiousparticipants.

Thus, our aim in this study was to examine how traitanxiety modulates cognitive performance and auto-nomic response to the sustained threat of an unpre-dictable aversive stimulus, using a context conditioningparadigm. Context conditioning has been previouslyused in rats and humans to examine anxiety and anxiousapprehension [for a review, see Grillon, 2002]. A typicalcontext conditioning paradigm is based on the princi-ples of Pavlovian conditioning, in which the presenta-tion of an aversive stimulus (unconditioned stimulus orUCS) is cued by the presentation of a neutral stimulus(conditioned stimulus or CS). Over a number ofpairings, the fear response associated with the UCSgeneralizes to the CS such that the presentation of theCS alone is capable of generating the fear response. Incue-specific conditioning, the CS is typically a brieflypresented stimulus (e.g., an auditory tone) that is highlypredictive of the temporal occurrence of the aversiveevent. In context conditioning, in contrast, the CS islonger lasting and less predictive of the UCS [Grillon,2002]. It is thought that the sustained anxiousapprehension often observed during context condition-ing with unpredictable UCS presentation is a moreaccurate model of anxiety states than explicitly cuedfear conditioning [Grillon, 2002].

In our study, a differential context conditioningparadigm was incorporated into a computer-baseddecision-making task by associating one screen back-ground color (CTX1) with the possible presentationof a loud noise, whereas the other color (CTX�) wasnever paired with the noise [Armony and Dolan, 2001].Cognitive performance was measured through accu-racy and response times in a continuous letter-sizedecision task, while autonomic activity was indexed byelectrodermal (skin conductance) responses. Based onour previous research [Armony and Dolan, 2001], weexpected enhanced skin conductance activity anddecreased performance (either by reduced accuracy orincreased response times) in the context associated withthe UCS compared to the safe context. The criticalquestion we sought to answer was whether there wouldbe an influence of trait anxiety on these measures.

In the second experiment, the same paradigm wasused, except that no UCS was presented during thetask. This experiment acted in part as a control and alsoallowed us to examine the suggestion that simplyinstructing participants of the CTX1/UCS association

211Research Article: Anxiety, Cognitive Performance, and Autonomic Response

Depression and Anxiety DOI 10.1002/da

is sufficient to observe conditioned responses to theCTX1 context [see Olsson and Phelps, 2004].

MATERIALS AND METHODSEXPERIMENT 1

Participants. Twenty-three volunteers participatedin this study (14 women, 9 men). The ages ofparticipants ranged between 19 and 30 years; meanage was 23.5 years (SD 5 3.8 years). All participantshad normal hearing and vision, did not have a currentor past diagnosis of any neurological or psychiatricdisorder, and were not taking any medication (with theexception of oral contraceptives). Volunteers wererecruited through advertisements placed at McGillUniversity and the Douglas Hospital Research Centre,and received $20 as compensation for their time andinconvenience. All experimental procedures wereapproved by the Research Ethics Board of the DouglasHospital Research Centre and were conducted inaccordance with the Code of Ethics of the WorldMedical Association (Declaration of Helsinki).

Decision-Making Task. Upon arrival at the labo-ratory, participants heard a description of the study and

provided written informed consent. The study wasdescribed as an investigation of the effect of anoccasionally presented loud noise on decision-makingand sweat gland activity. Participants were told thatthey would complete a computerized decision-makingtask during which trials would be grouped into twoalternating background screen color-blocks (i.e., thebackground color of the computer screen alternatedbetween green and purple). One of the color-blockswould occasionally be accompanied by a loud noise-burst presented through headphones; participants weretold of the color–noise relationship at the start of thestudy (half of the subjects had green and half hadpurple paired with noise).

After a brief practice session, participants completeda letter-size decision-making task presented withE-Prime software (Version 1.1; Psychology SoftwareTools, Inc., Pittsburgh, PA) on a desktop computer. Anoverview of the experimental paradigm is depicted inFigure 1. For the decision-making task, participantshad to decide which of two letters were of larger size bypressing either the ‘‘1’’ (i.e., left letter is larger) or ‘‘2’’(right letter is larger) button on the number pad of thekeyboard. For each 2-s trial, two letters were presentedto the left and right of a centrally located white fixation

Figure 1. The experimental design. For the decision-making task, participants had to decide which of two letters were larger by pressingeither the ‘‘1’’ (left letter is larger) or ‘‘2’’ (right letter is larger) button on the number pad of the keyboard. For each 2s trial, two letterswere presented to the left and right of a centrally located white fixation cross for 250 ms; the fixation cross remained on the screen forthe duration of the trial (1,750 ms). Trials were grouped into two different color-block conditions: noise anticipated (CTX1, 15 color-blocks) and noise not anticipated (CTX�, 15 color-blocks). During five CTX1 color-blocks (CTX1 paired), a noise (different in eachcase) was presented. Block length was jittered between 14 and 16 trials (i.e., 28–32 s).

212 Barrett and Armony


cross (14-point Courier font) for 250 ms; the fixationcross remained on the screen for the duration of thetrial (i.e., it was present throughout the experiment).Letters were offset from center on the vertical axis toincrease task difficulty (visual angle: 2.21). In each trial,the same letter was presented: Y, V, H, or T. Bymanipulating font size (15–19 points), letter pairs werecombined to make easy (e.g., 15,18) and difficult (e.g.,16,17) trials. A previous pilot study indicated anaccuracy of 61% for the letter pairs differing by 1 fontpoint and 87% for letter pairs differing by 2–3 fontpoints. Each letter pair (i.e., font pair) was presented infour different configurations according to the side oflarger font (left, right) and the offset (left up, rightdown, or left down, right up).

There were two different color–block conditions:noise anticipated (CTX1, 15 color–blocks) and noisenot anticipated (CTX�, 15 color–blocks). During fiveCTX1 color–blocks (CTX1 paired), a noise (differentin each case) was presented. In addition, a CTX�color–block always followed a CTX1 color–blockcontaining noise. Because we were not interested inthe immediate or lasting effect of the actual noise, datacollected during the CTX1-paired blocks with noise,as well as the CTX� block that immediately followed,were discarded. For CTX1 and CTX� conditions,font-pair trials were combined into 15 blocks. Blocklength was jittered between 14 and 16 trials (i.e., 28–32 s)to reduce the predictability of block changes. Eachblock contained approximately 30% hard trials. Blocksfor all condition types were pseudorandomly presentedacross subjects to ensure that (1) 3/5 CTX1 pairedblocks were presented in the first half of the experi-ment, and (2) no more than two CTX1 or CTX�color-blocks could be presented in succession. A 1-sgray screen and white fixation cross marked the changebetween blocks (i.e., interstimulus interval, or ISI). Forthe analysis, median response time and mean accuracy(percent correct) values were determined for all trials inthe CTX1 unpaired and CTX� conditions (approxi-mately 150 trials per condition; missed trials wereexcluded).

Noise (UCS). A different noise was presentedduring each of the CTX1 paired blocks, 1,000 msinterval after presentation of one letter pair and atvariable points throughout the blocks. Five different200-ms noises were created with CoolEdit Pro (Cool-Edit Pro; Syntrillium Software Corp., Phoenix, AZ)and comprised white noise and white noise mixed withhigh-frequency (e.g., 1 kHz or greater) frequency-modulated tones. The mean sound pressure level of thenoises was 86.4 dBA (SD 5 2.1), and the meanmaximum sound pressure level of the 101.8 dBA(SD 5 3.5). Participants listened to the noise with thehighest mean and maximum sound pressure level priorto the start of the experiment; no one described thenoise as painful or unbearable.

Skin Conductance Response (SCR). Electrodermalactivity, or skin conductance response (SCR), was

continuously measured throughout the procedureusing the Biopac MP100 acquisition system (BiopacMP100; Biopac Systems Inc, Goleta, CA). SCR signalwas acquired using two Ag-AgCl unpolarizable elec-trodes filled with conducting paste and attached to thepalmar surface of the middle phalanx of the left hand;participants were asked to keep this hand as still aspossible throughout the testing procedure. Prior toapplying the electrodes, the skin contact area waslightly abraded and cleaned with alcohol. The signal wasamplified and then stored at a sampling rate of 200 Hzusing Acqknowledge software (Biopac Systems Inc,Goleta, CA). To facilitate analysis of the SCR data, thetiming of the E-Prime presentation and Acqknowledgeacquisition was synchronized. To remove noise and driftfrom the signal, SCR data were mean-corrected andhigh-pass filtered in Acknowledge using a finite impulseresponse filter (window: Blackman �74 dB; cutofffrequency: 0.02 Hz; number of coefficients: 20,000).

We analyzed SCRs using the general linear model toestimate the parameters associated with the followingevent type: (1) the onset of the noise and (2) the onsetof the CTX1 and CTX� conditions. Each condition/event was modeled with 14 finite impulse responsebasis functions, with a 2 s delay. Thus, each analysisproduced 14 estimates (weights) of the amplitude of theSCR for each condition (CTX1, CTX�) and event(noise), covering 28 s.

Questionnaires. As indices of change in affect,150-mm visual analog scales (VASs) measuring valenceof affect and arousal were administered immediatelybefore and after the decision-making task; participantswere instructed to rate the post-task VAS according tohow they felt during the task. Valence of affect andarousal are viewed as being basic dimensions of morecomplex affective states and have been captured usingdifferent scale formats, including pictures [Self-Assess-ment Manikin; Bradley and Lang, 1994] and grids[Affect Grid; Russell et al., 1989]. We chose to use theVAS format to obtain more quantitative affect scores.Following the task, participants also completed theState–Trait Anxiety Inventory—Trait Version [STAI-T;Spielberger, 1983], the Penn State Worry Question-naire [PSWQ; Stober and Bittencourt, 1998], theIntolerance of Uncertainty Scale [IUS; Buhr andDugas, 2002], and a 10-item questionnaire measuringlevels of time management, distractibility and atten-tion, designed to diminish participant awareness of ourinterest in anxiety. In addition, participants completeda debriefing questionnaire that asked questions regard-ing task difficulty, task pace, screen colors, and noise, aswell as general impressions.

RESULTSEXPERIMENT 1

Questionnaires. Mean and median STAI-T scoreswere 41.1 (SD 5 7.3) and 39.0, respectively. Based on a



median-split of these scores, participants were assignedto low (n 5 10, M 5 34.7, SD 5 3.0) and high (n 5 10,M 5 48.1, SD 5 4.6) Trait Anxiety groups; three sub-jects scoring 39 were excluded. Our median STAI-Tscores and assignment strategy are consistent with thatof previous investigations [e.g., Ioannou et al., 2004].

Debrief ratings for task difficulty (1 5 extremely easy,7 5 extremely difficult), task pace (1 5 extremely slow,7 5 extremely fast), pleasantness of the noise (1 5extremely pleasant, 7 5 extremely unpleasant), expectancyof noise during CTX1 (never, 50%, 75%, 100%), andexpectancy of noise during CTX� suggest that mostparticipants found the letter-size decision-making taskto be of moderate pace and difficulty. In addition,participants rated the noise as being unpleasant andexpected it to be presented in more than half of thetrials (see Table 1a). Analysis of variance (ANOVA)with Trait Anxiety as a between-subjects factor (high,low) conducted on debrief ratings revealed no sig-nificant effects (all F-valueso1.3, all P-values 4 .25).

To examine differences in pre- and during-task VASaffect ratings, an ANOVA with Affect (valence, arousal)and Time (pretask, during-task) as within-subjectfactors and Trait Anxiety (low, high) as a between-subjects factor was performed. Results revealed asignificant Affect by Time interaction only, withparticipants feeling less pleasant (pre: M 5 95.0,SD 5 27.4; during: M 5 77.3, SD 5 27.2) and havinghigher arousal (pre: M 5 64.7, SD 5 34.1, during:M 5 83.9, SD 5 33.2) during the experiment comparedto before the experiment [F (1, 18) 5 8.44, Po.01].Finally, an ANOVA with Trait Anxiety as a between-subjects factor was conducted to examine differences inPSWQ and IUS scores. Results revealed that hightrait-anxious participants reported more worry com-pared to low trait-anxious participants [high: M 5 51.8,SD 5 13.9; low: M 5 38.8, SD 5 8.8; F (1, 19) 5 6.23,Po.05].

Decision-Making Task. For the letter-size deci-sion-making task, an ANOVA with Condition (CTX1,CTX�) as a within-subjects factor and Trait Anxiety asa between-subject factor was performed on responsetime (RT) and accuracy values. RT results revealed atrend toward a main effect of Trait Anxiety, with hightrait-anxious participants displaying faster overallresponse times compared to low trait-anxious partici-pants [high: M 5 524.3, SD 5 76.5; low: M 5 611.9,

SD 5 76.5; F (1, 18) 5 3.16, P 5.09], and a Conditionby Trait Anxiety interaction with high trait-anxiousparticipants displaying faster response times toCTX1 compared to CTX� trials, and low trait-anxious participants displaying the opposite pattern[F (1, 18) 5 4.71, Po.05; see Table 2a]. For accuracy,there were no significant main effects or interactions(all F-valueso0.1, all P-values 4.5; for values seeTable 2a).

We performed Pearson correlation analyses toexamine further the relationship between Trait Anxiety,Worry, and the difference in RT to CTX� comparedto CTX1 trials (RT Difference). Results revealedsignificant correlations between Trait Anxiety andWorry (r 5 .53, Po.01, one-tailed), Trait Anxiety andRT Difference (r 5 0.44, Po.05, one-tailed) and Worryand RT Difference (corrected for Trait Anxiety:r 5 0.42, Po.05, one-tailed). Correlations for TraitAnxiety/RT Difference are plotted in Figure 2.

Skin Conductance Response. The parameter esti-mates (i.e., weights) of the SCR to the onset of thenoise are plotted in Figure 3, which indicates that thenoise elicited a response characteristic of the SCR [e.g.,Lim et al., 1997] and that peaks between 6 and 8 sfollowed event onset. Parameter estimates for the noisewere analyzed in an ANOVA with Time as a within-subjects factor (fourteen 2-s time points or bins) and

TABLE 1. Debrief questionnaire responses

Task difficulty Task pace Noise rating Noise expected during CTX1 Noise expected during CTX�

a) Experiment 1M 4.45 5.55 6.35 58.15% 0%SD 1.15 0.51 0.67 19.52 0

b) Experiment 2M 4.92 5.07 4.72 55.36% 0%SD 1.14 1.07 1.01 40.64 0

TABLE 2. Response time and accuracy during thedecision-making task

CTX1 CTX�

Trait anxiety RT Accuracy RT Accuracy

a) Experiment 1Low M 620.7 76.58 603.2 76.37

SD 144.6 5.94 127.48 7.68

High M 518.85 78.15 529.75 77.86SD 79.65 3.42 77.06 4.1

b) Experiment 2Low M 525.71 77.03 527.14 75.71

SD 64.84 6.67 54.68 6.0

High M 558.28 79.18 561.5 78.12SD 74.48 3.24 75.53 5.23



Trait Anxiety as a between-subjects factor. Resultsrevealed a main effect of Time only: F (13, 234) 5 8.49,Po.01 (with Greenhouse–Geisser correction).

The plotted mean-corrected parameter estimates ofthe SCR to the onset of the CTX1 and CTX�conditions are shown in Figure 4. As in the decision-making task, data from CTX1 paired blocks, as well asthe CTX� blocks immediately following, were dis-carded. Again, both conditions elicited responses thatwere similar in size and shape, and relatively character-istic of the SCR. However, the SCR to the onset of theCTX� block peaked between 2 s and 4 s and the SCRto the onset of the CTX1 block peaked between 4 sand 7 s. In addition, there were notable differences inthe shape of the SCR between the conditions, with thepeak SCR to the CTX� condition decreasing morerapidly compared to that observed for the CTX1condition. Parameter estimates were analyzed togetherin an ANOVA with Condition (CTX1, CTX�) andTime (fourteen 2-s time points or bins) as a within-subjects factors and Trait Anxiety as a between-subjects

factor. Results revealed a main effect of Condition,with greater overall SCR to the CTX1 condition thanto the CTX� condition [CTX1: M 5 0.01, SD 5 0.07;CTX�: M 5�0.01, SD 5 0.05; F (1, 18) 5 7.22,Po.05] and a main effect of Time [F (13, 234) 5 5.21,Po.05 (with Greenhouse–Geisser correction)]. Therewas also a Condition by Time interaction [F (13,234) 5 2.18, Po.05], but this effect was no longersignificant following Greenhouse–Geisser correction(P 5.11).

EXPERIMENT 2

The design, procedure, and analysis strategy ofExperiment 2 was identical to that of Experiment 1,except that no noise was actually presented during thedecision-making task. Fifteen volunteers participated inthis study (10 women, 5 men). The age of participantsranged between 18 and 31 years; mean age was 20.8years (SD 5 3.18 years).

EXPERIMENT 2

Questionnaires. Mean and median STAI-T scoreswere 38.6 (SD 5 9.0) and 37.0, respectively. Based on amedian-split of these scores, participants were assignedto low (n 5 7, M 5 31.3, SD 5 3.4) and high (n 5 7,M 5 46.1, SD 5 7.1) Trait Anxiety groups; one subjectscoring 37 was excluded.

Similar to Experiment 1, debrief ratings for taskdifficulty, task pace, pleasantness of the noise (noiseheard prior to beginning the task), expectancy of noiseduring CTX1, and expectancy of noise during CTX�suggest that most participants found the letter-sizedecision-making task to be of moderate pace anddifficulty. In addition, participants rated the noise asbeing somewhat unpleasant and expected it to bepresented in more than half of the trials (see Table 1b).An ANOVA with Trait Anxiety as a between-subjects

-100-80-60-40-200

204060

25 35 45 55 65

Trait Anxiety

RT

Dif

fere

nce

Figure 2. In Experiment 1, the significant correlation betweenthe RT Difference (response time during CTX� context minusresponse time during CTX1 context) and Trait Anxiety (r 5 .44,Po.05, one-tailed).

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

2 4 6 8 10 12 14 16 18 20 22 24 26 28

Time (seconds)

SC

R

Figure 3. The mean-corrected parameter estimates of the SCR28 s after the onset of the UCS (noise) in Experiment 1. Resultsindicated a main effect of Time [F (13, 234) 5 8.49, Po.01,Greenhouse–Geisser corrected].

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

2 4 6 8 10 12 14 16 18 20 22 24 26 28

Time (seconds)

SC

R

CTX-

CTX+

Figure 4. The mean-corrected parameter estimates of the SCR28 s after the onset of the CTX1 (dashed line) and CTX� (solidline) contexts in Experiment 1. Results indicated a main effectof Condition [F (1, 18) 5 7.22, Po.05] and a main effect of Time[F (13, 234) 5 5.21, Po.05, Greenhouse–Geisser corrected].



factor (high, low) conducted on debrief ratings revealedno significant effects (all F-valueso1.7, all P-values 4 .2).

To examine differences in pre- and during-task VASaffect-ratings, an ANOVA was performed, with Affect(valence, arousal) and Time (pretask, during-task) aswithin-subject factors and Trait Anxiety (low, high) as abetween-subjects factor. Results revealed a main effectof Session [pre: M 5 84.1, SD 5 27.5; during:M 5 73.1, SD 5 32.6; F (1, 12) 5 4.80, Po.05] and atrend toward an Affect by Time interaction, withparticipants feeling less pleasant (pre: M 5 99.8,SD 5 16.7; during: M 5 77.5, SD 5 25.7) and reportingno change in arousal (pre: M 5 68.2, SD 5 27.4,during: M 5 68.7, SD 5 38.8) during the experimentcompared to before the experiment [F (1,12) 5 4.04,P 5.067]. Finally, we conducted an ANOVA with TraitAnxiety as a between-subjects factor to examinedifferences in PSWQ and IUS scores. Results revealedthat compared to low trait-anxious participants, hightrait-anxious participants reported more worry [high:M 5 58.4, SD 5 18.1; low: M 5 36.1, SD 5 17.9; F (1, 13)5 5.38, Po.05] and higher intolerance of uncertainty[high: M 5 69.9, SD 5 12.7; low: M 5 43.1, SD 5 8.1; F(1, 13) 5 22.04, Po.01].

Decision-Making Task. For the letter-size deci-sion-making task, an ANOVA with Condition (CTX1,CTX�) as a within-subject factor and Trait Anxiety asa between-subjects factor was performed on responsetime and accuracy values; results revealed no significantmain effects or interactions (all F-valueso1.4, allP-values 4 0.25; see Table 2b).

We performed correlation analyses to examinefurther the relationship between Trait Anxiety, Worry,and the difference in response time to CTX�compared to CTX1 trials (RT Difference). Resultsrevealed a significant correlation between Trait Anxietyand Worry only (r 5 0.79, Po.01, one-tailed).

Skin Conductance Response. The plotted para-meter estimates of the SCR to the onset of the CTX1 and CTX� conditions are shown in Figure 5. As inExperiment 1, both conditions elicited responses thatare similar in size and shape, and relatively character-istic of the SCR, peaking between 4 s and 6 s aftercondition onset. Parameter estimates were analyzedtogether in an ANOVA with Condition (CTX1,CTX�) and Time (fourteen 2-s time points or bins)as a within-subjects factors and Trait Anxiety as abetween-subjects factor. Results revealed a main effectof Time only: F (13, 156) 5 3.85, P 5.051 (withGreenhouse–Geisser correction).

DISCUSSIONOur aim in this study was to use an anticipatory

anxiety paradigm to examine how trait anxiety mod-ulates cognitive performance and autonomic responseto the presence of the sustained threat of unpredictableaversive stimuli. To elicit anxiety, we used a contextconditioning paradigm where, during the decision-

making task, one context (i.e., the background color ofthe computer screen) was associated with the unpre-dictable presentation of aversive noise (UCS) and theother context was neutral or safe (i.e., not associatedwith aversive noise). Two experiments were conducted.In Experiment 1, the UCS was presented during theCTX1 color-blocks using a partial reinforcementschedule. In Experiment 2, the UCS was neverpresented.

EXPERIMENT 1

In Experiment 1, participants were grouped into highand low trait-anxious groups based on a median split ofthe STAI-T scores. All participants described the UCSas unpleasant and expected it to be presented in over50% of the CTX1 blocks and in none of the CTX�blocks. In general, the observed SCR to the onset ofthe UCS, as well as the questionnaire data, wereconsistent with the UCS being aversive. Affectmeasurements revealed that all participants felt sig-nificantly less pleasant and experienced more arousalduring the task compared to before the task.

The SCR data demonstrates that our paradigm wassuccessful in eliciting anticipatory anxiety. Participantsdisplayed a greater SCR to the onset of CTX1 color-blocks (without UCS presentation) compared toCTX� color-blocks. This finding suggests that parti-cipants experienced greater or more sustained anxietyin response to the onset of color-blocks where the UCSwas anticipated. Interestingly, no significant differencesin SCRs were found between high and low trait-anxious individuals, suggesting that individual differ-ences in trait anxiety did not influence the amount ofanxiety and/or arousal elicited by our paradigm. Thisobservation is consistent with previous studies ofanticipatory anxiety in high and low trait-anxietygroups [Mauss et al., 2003].

-0.08

-0.06

-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

0.12

2 8 10 12 14 16 18 20 22 24 26 28

Time (seconds)

SC

R

CTX-

CTX+

4 6

Figure 5. The mean-corrected parameter estimates of the SCR28 s after the onset of CTX1 (dashed line) and CTX� (solidline) contexts in Experiment 2. Results indicated a main effectof Time [F (13, 156) 5 3.85, P 5.051, Greenhouse–Geissercorrected].



In contrast, individual differences in trait anxietywere found to influence performance on the decision-making task. When we compared reaction timesbetween CTX1 and CTX� color-blocks, high trait-anxious participants responded more quickly to letterpairs presented during CTX1 compared to CTX�color-blocks, whereas low trait-anxious participantsdisplayed the opposite pattern; that is, during color-blocks, when the UCS was anticipated compared tosafe color-blocks, high trait-anxious participants werefaster to respond to the letter pairs, and low trait-anxious participants responded more slowly to theletter pairs. The speed of performance appeared to beunrelated to how well participants discriminatedbetween the letter pairs, because no significantbetween-group differences were found for accuracy.Although one could speculate that a lack of between-group differences in accuracy may reflect a ceilingeffect, the observed performance (between 75% and80%, see Table 2) does not support this suggestion.Our results further revealed that the RT difference (RTduring CTX� color-blocks�RT during CTX1) wasmodulated by both level of trait anxiety and worry. Therelationship between the RT difference and worry wasespecially intriguing as worry is a prominent cognitivesymptom of many anxiety disorders.

EXPERIMENT 2

In Experiment 2, participants were again groupedinto high and low trait-anxious groups based on amedian split of the STAI-T scores. All participantsdescribed the UCS as somewhat unpleasant andexpected it to be presented in over 50% of the CTX1blocks and in none of the CTX� blocks. It appears thatthe UCS was perceived as being less aversive than inExperiment 1. This was not surprising given that, inExperiment 2, the UCS was only presented to theparticipants once (prior to the start of the task). Affectmeasurements revealed that all participants felt sig-nificantly less pleasant during the task comparedto before the task; no significant changes in arousalwere observed.

In contrast to Experiment 1, no differences wereobserved for the SCR to the onset of CTX1 color-blocks compared to CTX�color blocks; thus, thecontext SCR findings did not strongly suggest thatthe paradigm was successful in eliciting anticipatoryanxiety. It is possible that the observed SCRs for theCTX1 and CTX� color-blocks reflect that partici-pants experienced the same level of anxiety or arousalin response to both conditions. Once again, individualdifferences in trait anxiety were not found to influencethese autonomic responses. Finally, for the decision-making task, no between- or within-group differenceswere found for either accuracy or response time.

The lack of robust SCR results in Experiment 2indicate that simply instructing participants of therelationship between the CTX1 context and the UCS

is not sufficient to observe a context-specific anxietyresponse. Although it is also possible that the expecta-tion of noise in the absence of CTX1/UCS pairings inExperiment 2 resulted in participants experiencing anincrease in anxiety across both contexts, self-reportaffect data do not strongly support this suggestion.This findings contrasts with a study by Phelps andcolleagues [2001] using a somewhat similar paradigm,and in which significant amygdala activation to a CS1was observed. Further studies are required to establishthe reasons for this discrepancy.

CONCLUSIONOur findings in this set of experiments revealed that

context-specific increases in anxiety can be observedusing context conditioning. Importantly, when antici-patory anxiety is increased, trait anxiety appears to playa role in how participants perform, but not how theyfeel. Although differences in autonomic response havebeen observed in anxiety disorders and in high versuslow trait-anxious individuals [e.g., Arena and Hobbs,1995; Carillo et al., 2001; Fuller, 1992; Gonzalez-Bonoet al., 2002; Harrison and Turpin, 2003; Hubert and deJong-Meyer, 1992; Stamps et al., 1979], there remainsuncertainty regarding whether individual differences intrait anxiety modulate autonomic response whenanxiety is triggered [Gonzalez-Bono et al., 2002; Mausset al., 2003; Wilken et al., 2000]. The results of ourstudies clearly suggest that individual differences intrait anxiety do not influence autonomic responseassociated with anticipatory anxiety. Thus, our findingssuggest that autonomic responses elicited by theanticipation of a stressor may not be a reliable markerof trait anxiety.

Although individual differences in trait anxiety werenot found to influence context-specific increases inSCRs, both trait anxiety and worry were found tomodulate response time during the decision-makingtask. The fact that the effects of trait anxiety and worryon cognitive performance were only observed inExperiment 1, in which significant autonomic re-sponses to the CTX1 were elicited, strongly indicatesthat a certain level of anticipatory (or state) anxietymust be reached for these anxiety-relevant traits toimpact performance. During trials in which anxiety wasincreased compared to ‘‘safe’’ trials, high trait-anxiousparticipants made letter-size decisions more quickly,and low trait-anxious individuals made letter-sizedecisions more slowly. In our letter-size task, increasedor decreased response time could reflect alteredprocessing at many different stages in the decision-making process, including attention, perception, andmotor output, as well as other factors such asdistractibility or auditory monitoring. Due to thesimple nature of our decision-making task, it isimpossible to know the precise mechanism responsiblefor our trait anxiety/RT findings. However, theobserved relationship between trait anxiety/worry and



cognitive performance are nonetheless intriguing andsuggest that individual differences in anxiety-relevanttraits, including those specific to prominent cognitivesymptoms such as worry, may manifest in basic aspectsof cognition and information processing.

Given our focus on anxiety-related personality traits,it is pertinent to discuss briefly the measurement oftrait anxiety. Because our anxiety-related question-naires were administered following the experiments,and state and trait anxiety are known to be correlated, itis possible that the task-induced increase in stateanxiety influenced the trait anxiety scores. Psycho-metric data suggests, however, that trait anxiety asmeasured by the STAI-T is a relatively stablepersonality trait [Barnes et al., 2002]. Thus, we areconfident that are our trait anxiety scores are moreindicative of individual differences in trait, not state,anxiety.

To conclude, the results and implications of our setof experiments have raised a number of interestingquestions for future research. Our finding of adissociation between the effect of trait anxiety oncognitive performance versus autonomic response inparticular may hold important clues as to the nature ofanxiety vulnerability. Further studies of cognition inanxiety and anxiety disorders will aid our under-standing of how rather subtle individual differences ininformation processing may have a significant impacton emotional functioning.

Acknowledgments. Our thanks to the contribu-tions of anonymous reviewers in revising this manu-script, and to Rina Zelmann for her help in the analysisof the skin conductance data.

REFERENCESArena JG, Hobbs SH. 1995. Reliability of psychophysiological

responding as a function of trait anxiety. Biofeedback Self Regul20:19–37.

Armony JL, Dolan RJ. 2001. Modulation of auditory neuralresponses by a visual context in human fear conditioning.Neuroreport 12:3407–3411.

Barlow DH. 2000. Unraveling the mysteries of anxiety and itsdisorders from the perspective of emotion theory. Am Psychol 55:1247–1263.

Barnes LLB, Harp D, Jung WS. 2002. Reliability generalization ofscores on the Spielberger State–Trait Anxiety Inventory. EducPsychol Meas 62:603–618.

Berntson GG, Sarter M, Cacioppo JT. 1998. Anxiety and cardiovas-cular reactivity: The basal forebrain cholinergic link. Behav BrainRes 94:225–248.

Bradley MM, Lang PJ. 1994. Measuring emotion: The Self-Assessment Manikin and the semantic differential. J Behav TherExp Psychiat 25:49–59.

Bremner JD, Krystal JH, Southwick SM, Charney DS. 1996.Noradrenergic mechanisms in stress and anxiety: I. Preclinicalstudies. Synapse 23:28–38.

Buhr K, Dugas MJ. 2002. The Intolerance of Uncertainty Scale:Psychometric properties of the English version. Behav Res Ther40:931–945.

Carrillo E, Moya-Albiol L, Gonzalez-Bono E, Salvador A, Ricarte J,Gomez-Amor J. 2001. Gender differences in cardiovascular andelectrodermal responses to public speaking task: The role ofanxiety and mood states. Int J Psychophysiol 42:253–264.

Chambers JA, Power KG, Durham RC. 2004. The relationshipbetween trait vulnerability and anxiety and depressive diagnoses atlong-term follow-up of generalized anxiety disorder. J AnxietyDisord 18:587–607.

Charney DS, Deutch A. 1996. A functional neuroanatomy of anxietyand fear: Implications for the pathophysiology and treatment ofanxiety disorders. Crit Rev Neurobiol 10:419–446.

Chrousos GP. 1998. Stressors, stress, and neuroendocrine integrationof the adaptive response: The 1997 Hans Selye Memorial Lecture.Ann NY Acad Sci 851:311–335.

Egloff B, Hock M. 2001. Interactive effects of state anxiety and traitanxiety on emotional Stroop interference. Person Indiv Diff 31:875–882.

Fox E, Russo R, Dutton K. 2002. Attentional bias for threat:Evidence for delayed disengagement from emotional faces. CognEmotion 16:355–379.

Fuller BF. 1992. The effects of stress–anxiety and coping styles onheart rate variability. Int J Psychophysiol 12:81–86.

Gonzalez-Bono E, Moya-Albiol L, Salvador A, Carrillo E, Ricarte J,Gomez-Amor J. 2002. Anticipatory autonomic response to a publicspeaking task in women—the role of trait anxiety. Biol Psychol60:37–49.

Grillon C. 2002. Startle reactivity and anxiety disorders: Aversiveconditioning, context, and neurobiology. Biol Psychiatry 52:958–975.

Harrison LK, Turpin G. 2003. Implicit memory bias and traitanxiety: A psychophysiological analysis. Biol Psychol 62:97–114.

Hoehn-Saric R, McLeod DR. 1988. The peripheral sympatheticnervous system. Its role in normal and pathologic anxiety. PsychiatrClin North Am 11:375–386.

Hubert W, de Jong-Meyer R. 1992. Saliva cortisol responses tounpleasant film stimuli differ between high and low trait anxioussubjects. Neuropsychobiology 25:115–120.

Ioannou MC, Mogg K, Bradley BP. 2004. Vigilance for threat: Effectsof anxiety and defensiveness. Pers Indiv Diff 36:1879–1891.

Kabacoff RI, Segal DL, Hersen M, Van Hasselt VB. 1997.Psychometric properties and diagnostic utility of the Beck AnxietyInventory and the State–Trait Anxiety Inventory with older adultpsychiatric outpatients. J Anxiety Disord 11:33–47.

Lang PJ, Davis M, Ohman A. 2000. Fear and anxiety: Animal modelsand human cognitive psychophysiology. J Affect Disord 61:137–159.

LeDoux J. 1996. The emotional brain. NewYork: Simon and Schuster.Lim CL, Rennie C, Barry RJ, Bahramali H, Lazzaro I, Manor B,

Gordon E. 1997. Decomposing skin conductance into tonic andphasic components. Int J Psychophysiol 25:97–109.

Mauss IB, Wilhelm FH, Gross JJ. 2003. Autonomic recovery andhabituation in social anxiety. Psychophysiology 40:648–653.

Mineka S, Kihlstrom JF. 1978. Unpredictable and uncontrollableevents: a new perspective on experimental neurosis. J AbnormPsychol 87:256–271.

Mogg K, Bradley BP, Dixon C, Fisher S, Twelftree H, McWilliams A.2000. Trait anxiety, defensiveness and selective processing ofthreat: An investigation using two measures of attentional bias.Pers Indiv Diff 28:1063–1077.

Oei TP, Evans L, Crook GM. 1990. Utility and validity of the STAIwith anxiety disorder patients. Br J Clin Psychol 29:429–432.



Olsson A, Phelps EA. 2004. Learned fear of ‘‘unseen’’ faces afterPavlovian, observational, and instructed fear. Psychol Sci 15:822–828.

Phelps EA, O’Connor KJ, Gatenby JC, Gore JC, Grillon C, Davis M.2001. Activation of the left amygdala to a cognitive representationof fear. Nat Neurosci 4:437–441.

Plehn K, Peterson RA. 2002. Anxiety sensitivity as a predictor of thedevelopment of panic symptoms, panic attacks, and panic disorder:A prospective study. J Anxiety Disord 16:455–474.

Russell JA, Weiss A, Mendelsohn GA. 1989. Affect Grid: Asingle-item scale of pleasure and arousal. J Pers Soc Psychol 57:493–502.

Spielberger CD. 1983. Manual for the State–Trait Anxiety Inventory.Mountain View, CA: Consulting Psychologists Press.

Stamps LE, Fehr LA, Lewis RA. 1979. Differential effects of stateand trait anxiety on heart rate responses and reaction time. BiolPsychol 8:265–272.

Stober J, Bittencourt J. 1998. Weekly assessment of worry: Anadaptation of the Penn State Worry Questionnaire for monitoringchanges during treatment. Behav Res Ther 36:645–656.

Wilken JA, Smith BD, Tola K, Mann M. 2000. Trait anxiety and priorexposure to non-stressful stimuli: Effects on psychophysiologicalarousal and anxiety. Int J Psychophysiol 37:233–242.



the influence of trait anxiety on autonomic response and cognitive performance during an...

Documents