BEHAVIOR THERAPY 31,745-756,2000

Explicit Memory Bias for Threat Words in Generalized Anxiety Disorder

BRUCE H. FRIEDMAN
Virginia Polytechnic Institute and State University
JULIAN E THAYER

University of Missouri-Columbia
THOMAS D . BORKOVEC

The Pennsylvania State University

Although findings of an implicit memory bias for threat words in generalized anxiety disorder (GAD) are fairly robust, the data regarding an explicit bias in this disorder are less consistent. This issue was investigated in the context of two studies directed primarily at the examination of attentional and physiological underpinnings of GAD. In these experiments, GAD clients and nonanxious control participants (35 and 29 in Study 1, and 22 and 31 in Study 2, respectively) engaged in an S1-$2 conditioning procedure that involved the presentation of a series of neutral stimuli (colored dots) paired with threat and nonthreat words, followed by a free recall test. Instructions were to simply look at the dot and read the word silently. A free recall task was administered at the end of each experimental session. Contrary to previous trends in the literature, a marked bias in the GAD group toward recall of the threat words emerged in both studies. These results are considered in the light of methodological differences with previous research, information processing characteristics of GAD, and the role of physiological assessment in cognitive studies of anxiety.

Information processing approaches to anxiety have yielded numerous insights into the cognitive mechanisms that operate in generalized anxiety disorder (GAD). In particular, examination of memory biases for disordercongruent information suggests that characteristic cognitive styles can distinPreparation of this manuscript was supported in part by National Institute of Mental Health Grant MH-39172 to the third author. The authors would like to thank Mary Borkovec, Holly Hazlett-Stevens, Jim Lyonfields, Katie Peasley, and the late Silvia Molina for assistance in data collection and quantification. These data were collected at the Stress and Anxiety Disorders Institute, Department of Psychology, The Pennsylvania State University. Address correspondence to Bruce H. Friedman, Department of Psychology, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061--0436; e-mail: BHFRIEDM@VT.EDU. 745 005-7894/00/07454)75651.00/0 Copyright2000 by Associationfor Advancementof BehaviorTherapy All rightsfor reproductionin any formreserved.

746

F R I E D M A N ET A L .

guish among various psychopathologies. For example, an implicit memory bias (enhancement of stimulus recognition by prior stimulus exposure or "priming,' without conscious recollection of the exposure; see Squire, 1992) has been typically associated with anxiety, and an explicit memory (available to conscious recall) bias has been found more characteristic of depression (see Williams, Watts, MacLeod, & Mathews, 1997, for review). Implicit memory processes activate stored information, thereby improving its accessibility, though not necessarily affecting its direct retrievability (Graf, 1994). Conversely, explicit memories involve conscious, deliberate recall of information that was elaborated at the time of encoding, and are therefore more available due to embellishment from contextual cues. Anxiety is marked by a nonconscious, automatic, pre-attentive bias toward threat cues (Mathews, 1990) that may prime such information in implicit memory tests. However, this defensive attentional mechanism thwarts further processing of anxiety-provoking stimuli, and so no explicit memory bias appears for such material. In contrast, depression is characterized by conscious elaboration of sad material, which increases its availability for explicit, directed recall. However, these specific memory effects in anxiety may not generalize across diagnostic categories. For example, although the prediction of an implicit but not explicit memory bias for disorder-congruent material has been replicated in GAD (MacLeod & McLaughlin, 1995), others have found both implicit and explicit memory biases for relevant stimuli in panic disorder (Cloitre & Leibowitz, 1991 ). A direct comparison between panic disorder and GAD revealed an explicit memory bias in the former, but not the latter (Becker, Roth, Andrich, & Margraf, 1999). Becker et al. suggested that methodological inconsistencies across studies might account for these discrepant memory effects in the anxiety literature, and advocated the combination of incidental learning and free recall as the optimal paradigm for detecting memory biases in clinical anxiety. They further proposed that this pairing is enhanced with compatible learning and retrieval tasks; that is, both should be either conceptual or perceptual (e.g., Eysenck & Byrne, 1994). An opportunity to meet these standards and thus conceptually replicate Becker et al.'s (1999) investigation of explicit memory bias for threat emerged in two of our ongoing studies that were directed primarily at attentional and physiological elements of GAD. An S1-$2 paradigm, which entails the presentation of a series of paired stimuli separated by a fixed interstimulus interval (Bohlin & Kjelberg, 1979), was used in these two studies to examine higher order conditioning to threat in GAD clients and matched control groups. In the first study, a colored dot (S 1) was consistently paired with either threat or nonthreat words ($2) over multiple trials (see Thayer, Friedman, Borkovec, Johnsen, & Molina, 2000, for details). Individuals were instructed to simply look at the dot and then read the word silently. The second study was an exact replication of the first: the physiological and attentional data from Experiment 2 have not yet been reported (Thayer, Friedman, & Borkovec, 2000). To address the issue of a potential memory bias, a free

MEMORY BIAS I N G E N E R A L I Z E D A N X I E T Y D I S O R D E R

747

recall test of the stimulus words was given following the conditioning procedure. Thus, the pairing of the S 1-$2 procedure (an incidental learning task) and free recall (both tasks are also conceptual), utilized in both studies, conforms to the recommendations of Becker et al. (1999). Importantly, however, the conditioning task involved extended exposure (8 seconds) to the threat words, which may be more likely to elicit an explicit memory bias than the brief exposure times that typify most studies (e.g., Stroop paradigms) of this phenomenon (Borkovec, Ray, & Stober, 1998). Furthermore, retention is generally enhanced for emotionally arousing stimuli (McGaugh, 1992), and threat words would presumably be especially arousing to GAD clients. Based on these notions, the following hypotheses were generated and tested in the present studies: (a) GAD clients will recall more threat than nonthreat words, (b) these individuals will recall more threat words when compared to a matched control group, (c) recall for threat words will be greater than nonthreat words in the control group, as well, because of the arousing nature of threat words, and (d) there will be no differences between the groups on recall of nonthreat words.

Experiment 1
Method Participants. Thirty-five treatment-seeking GAD clients and 29 matched
control participants completed the laboratory session. Agency-referred and self-referred clients were first given a 30-minute phone interview by clinical graduate student assessors to determine whether GAD was the likely diagnosis. Subsequently, the assessor conducted a diagnostic interview using the Anxiety Disorder Interview Schedule-Revised (ADIS-R; DiNardo & Barlow, 1988). Clients were admitted to a therapy trial if they met criteria for a principal DSM-III-R (American Psychiatric Association, 1987) diagnosis of GAD with an assessor severity rating of 4 (moderate anxiety) or greater on a 0-to8-point severity scale, and if they did not meet criteria for panic disorder with or without agoraphobia, organic brain syndrome, substance abuse, or medical conditions that may contribute to anxiety. The ADIS-R was conducted a second time by a Ph.D. clinical psychologist. It was required that both diagnostic interviewers agree that the criteria were met. Control subjects were recruited by newspaper advertisements and were assessed by the phone interview and a single administration of the ADIS-R conducted by one of the clinical assessors. These individuals were included in the study if they did not meet diagnostic criteria for any disorder at even mild severity levels and if they reported no prior history of psychopharmacological or psychological treatment of any kind. Physiological data from this study are reported in Thayer et al. (2000). The distribution of gender across groups was 22 women and 13 men for GAD and 18 women and 11 men for control. The distribution of race was 31 Caucasian, 1 African American, 2 Hispanic, and 1 Middle Eastern for the

748

F R I E D M A N ET A L .

TABLE 1 COMORBID DIAGNOSES FOR G A D CLIENTS Experiment I Social phobia Specific phobia Agoraphobia Major depression (recurrent) Major depression Dysthymia Major depression (partially remitted) Obsessive-compulsive disorder Posttraumatic stress disorder Seasonal affective disorder 19 12 3 3 2 1 I 1 1 1 Experiment 2 15 3 2 0 0 2 1 0 0 0

GAD group, and 26 Caucasian, 1 African-American, 1 Hispanic, and 1 Middle Eastern for the control group. Average ages of the clients and controls were 36.9 (SD = 11.8) and 33.8 (SD = 10.9) years, respectively, and average years of education were 15.7 (SD = 2.6) and 16.1 (SD = 2.3), respectively. The groups did not differ significantly on any of these variables. The average duration of the clients' disorder was 12.5 years (SD = 12.6). Seventy-seven percent had at least one comorbid diagnosis (see Table 1) and 5 were taking psychotropic medication ~at the time of the session. At the end of the diagnostic interview using the ADIS-R, the Hamilton Anxiety Rating Scale (HARS; Hamilton, 1960) and the Hamilton Rating Scale for Depression (HRSD; Hamilton, 1959) were also administered. In a separate self-report session, clients (before therapy began) and controls completed the State-Trait Anxiety Inventory-Trait (STAI; Spielberger, Gorsuch, Lushene, Vagg, & Jacobs, 1983); the Reactions to Relaxation and Arousal Questionnaire, a factor analytically derived measure of fear of relaxation (RRAQ; Heide & Borkovec, 1983); the Penn State Worry Questionnaire (PSWQ; Meyer, Miller, Metzger, & Borkovec, 1990), a measure of worry that distinguishes GAD from all other anxiety disorders (Brown, Antony, & Barlow, 1992); and the Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961). This research was approved by the Institutional Review Board and informed consent was obtained from each participant prior to the start of the experiment. Procedure. Full details on the S1-$2 task and physiological recording procedure can be found in Thayer et al. (in press); included here is an overview of the S 1-$2 procedure and details of the free recall test. Each participant received a tour of the laboratory and a description of the recording methOne was taking Prozac, 2 were taking hypnotic-sedatives, 1 was taking Lisinopril (for high blood pressure), and 1 was taking Synthroid (a thyroid medication).

MEMORY BIAS IN GENERALIZED ANXIETY DISORDER

749

ods and tasks prior to the actual session. At the experimental session and after electrode attachment, the participant was seated in a recliner chair in a small sound-attenuated room adjacent to the equipment room and left alone with door closed; the experimenter communicated with the participant via intercom. The initial phase of the experiment involved physiological recording during periods of quiet resting, self-relaxation, and worrying (see Thayer, Friedman, & Borkovec, 1996). After a final 5-minute baseline, the S1-$2 task was initiated, which involved the presentation of a colored dot (S 1) displayed at the center of a computer monitor. The color of the dot (yellow or green) was consistently paired with either threat or nonthreat words ($2); this association was counterbalanced across subjects within both GAD and control groups. This contingency was not made explicit, which allowed for the assessment of higher order conditioning. The instructions simply required that each word be read silently as it appeared on the screen. The S 1 and $2 stimuli were sequentially presented for 8 seconds each, with an intertrial interval of 30 seconds. A total of 20 trials (10 threat, 10 nonthreat) was presented. A single order was used with all participants; the order of the words had been randomly determined before the experiment, with the constraint that no more than 2 consecutive trials contain words of one type. The 20 words were taken from Mathews and MacLeod (1986). Following the S 1-$2 procedure, subjects were asked to write down as many words from the previous task that they could remember. The number of correctly recalled threat and nonthreat words served as the dependent variable.

Experiment 2
Method Participants and procedure. Participants were selected along the same criteria as described in Experiment 1. The distribution of gender across groups was 15 women and 7 men for GAD and 22 women and 9 men for the control group. The distribution of race was 18 Caucasian, 1 African American, 2 Hispanic, and 2 Indian for the GAD group, and 26 Caucasian, 1 African American, 2 Hispanic, 1 Indian, and 1 Asian for the control group. Average ages of the clients and controls were 35.81 (SD = 12.1) and 39.60 (SD = 13.2) years, respectively, and average years of education were 15.9 (SD = 2.3) and 16.80 (SD = 2.2), respectively. The groups did not differ significantly on any of these variables. The average duration of the clients' disorder was 10.9 years (SD = 9.8). Eighty-six percent had at least one comorbid diagnosis (see Table 1) and 2 were taking psychotropic medication 2 at the time of the session. The GAD clients and control participants completed the same selfreport measures of anxiety and depression that were administered in Experiment 1 (see Table 2 for mean scores). The initial procedure, which involved physiological recording during peri2 One was taking Xanax and 1 was taking Buspar.

750

FRIEDMAN ET AL. TABLE 2 G A D CLIENT AND CONTROL SCORES ON SELF-REPORT MEASURES OF ANXIETY AND DEPRESSION Experiment 1 GAD Mean SD 8.03 6.31 7.32 7.12 7.56 7.72 Control Mean 1.21 2.53 1.47 12.92 27.69 28.07 SD 1.70 1.98 1.27 3.99 4.89 6.31 GAD Mean 16.24 23.51 10.26 28.29 55.82 67.05 SD 7.76 6.31 4.89 6.54 8.23 8.36 Experiment 2 Control Mean 1.84 2.82 1.29 13.10 25.45 28.55 SD 4.19 2.23 1.21 4.47 5.43 9.41

BDI HARS HRSD RRAQ STAI PSWQ Note.

18.12 24.35 14.10 28.15 59.03 68.58

BDI - Beck Depression Inventory; HARS = Hamilton Anxiety Rating Scale; HRSD = Hamilton Rating Scale for Depression; RRAQ = Reactions to Relaxation and Arousal Questionnaire; STAI = State-Trait Anxiety Inventory-Trait: PSWQ = Penn State Worry Questionnaire.

ods of quiet rest, self-relaxation, and worry, and subsequent S l - S 2 and free recall procedures, were identical to Experiment 1.

Results Statistical Analyses

As a manipulation check, differences between the clients and the controls on the self-report measures of anxiety and depression were examined using t tests. In both experiments the clients scored higher than the controls on all measures (all ps < 0.00001). In addition, the two client groups and the two control groups did not differ significantly on any measures after control of Type I error rates. Planned t tests were conducted, based on the predictions stated in the introduction: mean numbers of recalled words (out of 10 threat and 10 nonthreat words) are reported in Table 3. In addition, effect sizes (point-biserial correlations) are reported. A modified Bonferroni procedure (Simes, 1986) was used to control hypothesis-wise Type I error rates. In both experiments, the GAD group recalled significantly more threat words than nonthreat words [Experiment 1: t(62) = 6.29,p < .001, rpb = 0.62; Experiment 2: t(51) = 3.69,p < .004, one-tailed, /'i0b 0.46]. Furthermore, the GAD group recalled significantly more threat words than the control group did in both studies [Experiment 1: t(62) = 3.69,p < .001, rpb ---- 0.42; Experiment 2: t(51) = 2.00,p < .03, one-tailed, rpb ---- 0.27]. There was a marginally significant trend in the control group in Experiment 1 for recall of more threat than nonthreat words, t(62) = 1.46, p < .08, one-tailed, rpb ---- 0.18, that reached significance in Experiment 2, t(51) = 1.91,p < .03, one-tailed, rpb ----0.26, but did not sur=

MEMORY BIAS IN GENERALIZED ANXIETY DISORDER TABLE 3 MEANS AND STANDARD DEVIATIONS FOR NUMBER OF THREAT AND NONTHREAT WORDS RECALLED Word Type Group Experiment 1 GAD Control Experiment 2 GAD Control Threat

751

Nonthreat

4.6 (1.5) 3.3 (1.3) 4.7 (1.7) 3.7 (1.8)

3.0 (1.7) 2.9 (1.2) 3.5 (1.7) 3.0 (1.3)

vive error correction. However, the effect sizes indicated that the effect was nontrivial. The difference between GAD and control groups for recall of nonthreat words did not approach significance in either experiment (Experiment 1: rpb = 0 . 0 2 ; Experiment 2: rpb = 0.16). To examine possible differential recall of social threat versus physical threat words, analysis of variance using group (GAD vs. Control) as a between-subjects factor and word type (social vs. physical threat) as a within-subjects factor was conducted for each experiment. There were no significant interactions between group and word type [Experiment l: F(1, 62) = 0.066, p = 0.79; Experiment 2: F(1, 51) = 2 . 3 5 , p -- 0.13].

Discussion
An explicit memory bias for threat words was replicated across two separate GAD samples under conditions recommended as ideal for detection of memory biases in studies of anxiety: namely, incidental learning and free recall (Becker et al., 1999). Indeed, the overall total number of correctly recalled words was in accord with this research. However, the present data conflict with a body of literature that has failed to find an explicit memory bias in GAD (Becker et al.; Bradley, Mogg, & Williams, 1995; MacLeod & McLaughlin, 1995; Mathews, Mogg, May, & Eysenck, 1989), though exceptions to this trend can be found (Mogg & Mathews, 1990; Otto, McNally, Pollack, C h e n , & Rosenbaum, 1994). Mogg et al. attributed their finding to a response bias in GAD toward mood-congruent words, rather than a selective memory effect. However, participants in this study were aware that they would receive a recall test, and so the task does not qualify as incidental learning. This difference renders the findings less directly comparable to either Becket et al. or the present study. Otto et al. found an explicit memory bias only in left-hemisphere-dominant GAD subjects; this effect is considered in greater detail below. The hypothesized memory bias for threat in the control group was less

752

F R I E D M A N ET AL.

robustly supported, as evidenced by smaller effect sizes and the marginal significance of the effect. One likely explanation for this difference is that the threat words were less arousing and significant for the control group. This possibility is supported by the clear bias for the threat words by the GAD patients in both studies when compared to the control groups. Moreover, as predicted, there were no group differences in recall for nonthreat words, which further supports the salience of the threat words to the GAD group. In general, the present results highlight the points emphasized by Becker et al. (1999; i.e., methodological inconsistencies and the role of physiological arousal in anxiety) in their attempt to account for inconsistent memory bias findings across anxiety studies. One methodological variance that may account for differences in findings involves the nature of the tasks. Many studies of memory bias in GAD have used very brief stimulus exposure times, to which an active response is required (Borkovec et ai., 1998). In contrast, the protracted, passive viewing of the S1-$2 task in the present study may have overridden the avoidant aspects of worry, and so resulted in enhanced retention of threat words. Another distinction can be made regarding the nature of the stimuli. A commonly used procedure involves the generation of self-referent images to threat and nonthreat words (e.g., Becker et al., 1999; Mathews et al., 1989). In contrast, instructions in the present task were to simply read the words silently. It is possible that the process of imagining oneself in an anxietyprovoking scenario elicits greater cognitive avoidance and therefore less strategic elaboration than merely reading a threatening word. The latter task may in fact be better matched to the characteristic tendency toward worry and rumination in GAD, which is marked by the dominance of verbal-linguistic processing over imagery (Borkovec & Inz, 1990; Borkovec et al., 1998; Lyonfields, Borkovec, & Thayer, 1995). Furthermore, the use of imagery in general may yield no particular advantage in terms of acquisition and retention (Postman, 1974). In sum, the S 1-$2 task used in the present study may be singularly congruent with GAD in terms of detecting an explicit memory bias. On the other hand, although imagery tasks may be conceptually compatible with free recall (Becker et al.), they may not be well matched to GAD. Indeed, the low capacity for generation of fear imagery in GAD has been invoked to explain therapy outcome differences between GAD and other anxiety disorders such as specific phobias, in which vivid fear images are more readily produced (Lang, Levin, Miller, & Kozak, 1983). The fact that the effect sizes reported here are two to three times greater than those reported in Becker et al. lends further support to these notions. This reasoning is consistent with the finding of an explicit memory bias in GAD individuals who display a left-hemispheric bias (Otto et al., 1994). The left cerebral hemisphere is generally dominant for language and verbal memory, and the right hemisphere for imagery and nonverbal memory (Bradshaw, 1989; Bryden & Ley, 1983; Kolb & Whishaw, 1990). Hence, a memory bias involving verbal material is more likely to appear in left-dominant individu-

MEMORY BIAS I N G E N E R A L I Z E D A N X I E T Y D I S O R D E R

753

als. In fact, the characteristic worry of GAD is marked by left frontal hemispheric activation (Borkovec et al., 1998). Thus, both the central nervous system and information processing characteristics of GAD underscore the potential inappropriateness of imagery tasks for the detection of memory biases in this disorder. Physiological arousal has been invoked loosely in causal speculations of memory differences between panic and GAD (Becker et al., 1999). However, a model that can more specifically relate information processing to physiological responding may be more useful than global, nonspecific depictions of physiological arousal. The former approach was taken in our investigation (from which the present data were obtained) of phasic heart rate (HR) responses to repeated presentation of threat words in GAD (Thayer et al., in press). In that report, the GAD group developed a conditioned HR deceleration in anticipation of threat words, a physiological change that suggests facilitation of stimulus intake (Bohlin & Kjelberg, 1979). Furthermore, these individuals also exhibited a HR acceleration in response to the threat words, which, in accord with models that link increased adrenergic activity with enhanced memory, might have improved stimulus retention (McGaugh, 1992). The adaptive value of this mechanism presumably entails the selective recall of salient emotional events, such as a threatening encounter. In GAD, defensive cognitive operations occur at multiple stages of information processing: There is a pre-attentive bias toward detection of threatening stimuli and excessive rumination of stored threat information. This combination may account for the convergent findings of an implicit memory bias in GAD (see Becker et al., 1999; Williams et al., 1997, for reviews), as well as positive findings of an explicit memory bias in GAD under certain conditions. It has also been suggested that the processing of emotional stimuli is primarily moderated by pre-intentional, automatic mechanisms (Libkuman, Nichols-Whitehead, Griffith, & Thomas, 1999). Stimulus orienting, with accompanying HR deceleration, occurs at this early stage of information processing. This response is indicative of enhanced attention to external stimuli, which may in turn elicit greater intake of peripheral cues, a notion that is consistent with the hypervigilance and characteristic scanning of the environment for threat cues that is associated with anxiety (Mathews, 1990). Viewed as such, the enhanced retention of threat information in anxiety is essentially an encoding, rather than rehearsal or elaboration, phenomenon (Libkuman et al.). The HR acceleration in response to threat words, observed in GAD subjects in Thayer et al. (in press), likely reflects a defensive response, rather than cognitive elaboration, that further improves retention via emotional arousal. In sum, although the data on explicit recall biases in anxiety remain "difficult to characterise" (Williams et al., 1997, p. 286), it seems premature at this juncture to dismiss out of hand such a bias. Rather, such tendencies can be observed in certain situations (e.g., extended stimulus exposure times) and in individuals with particular distinguishing features (e.g., left-hemisphere dominance). The parameters of these tendencies should be pursued further

754

FRIEDMAN ET AL.

through employment of various types and combinations of learning and memory tasks rather than with fixed procedures that may yield consistency across studies but occlude a comprehensive depiction of GAD. It also may be rash to challenge broadly influential cognitive models of memory in anxiety (e.g., Beck, Emery, & Greenberg, 1985; Bower, 1981; Williams et al.; cf. Becker et al., 1999, p. 162) based on a narrowly defined memory methodology. Additionally, the concurrent assessment of physiological data with specific theoretical links to information processing, such as phasic cardiac responding (e.g., Thayer et al., 2000) or HR variability (Friedman & Thayer, 1998; Lyonfields et al., 1995; Thayer et al., 1996) may allow for the precise mapping of somatic activity to cognitive phenomena. This approach is likely to be more fruitful than inclusion of nonspecific, global measures of physiological arousal, a dated paradigm that has been consistently refuted in psychophysiological research (Lacey, 1967; Venables, 1984; Wolf, 1995). Ultimately, methodological diversity coupled with an appreciation of the psychophysiological complexities of GAD will help advance cognitive models that can account for a wide range of information processing phenomena in this and other anxiety disorders.

References
American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd ed., rev.). Washington, DC: Author. Beck, A. T., Emery, G., & Greenberg, R. L. (11985). Anxiety disorders and phobias. New York: Basic Books. Beck, A. T., Ward, C.H., Mendelson, M., Mock, J., & Erlbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatcv. 4, 53-61. Becker, E. S., Roth, W. T., Andrich, M., & Margraf, J. (1999). Explicit memory bias in anxiety disorders. Journal of Abnormal Psychology, 108, 153-163. Bohlin, G., & Kjelberg, A. (1979). Orienting activity in two-stimulus paradigms as reflected in heart rate. In H.D. Kimmel, E.H. van Olst, & J. F. Orlebeke (Eds.), The orienting reflex in humans (pp. 169-197). Hitlsdale, N J: Lawrence Erlbaum Associates. Borkovec, T. D., & Inz, J. (1990). The nature of worry in generalized anxiety disorder: A predominance of thought activity. Behavior Research and Therapy. 28, 153-158. Borkovec, T. D., Ray, W. J., & Stober, J. (1998). Worry: A cognitive phenomenon intimately linked to affective, physiological, and interpersonal behavioral processes. Cognitive Therapy and Research, 6, 561-576. Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129-148. Bradley, B. P., Mogg, K., & Williams, R. (1995). Implicit and explicit memory bias for emotion-congruent information in clinical depression and anxiety. Behaviour Research and Therapy, 33,755-770. Bradshaw, J. L. (1989). Hemispheric specialization and psychological imction. New York: Wiley. Brown, T. A., Antony, M. M., & Barlow, D. H. (1992). Psychometric properties of the Penn State Worry Questionnaire in a clinical anxiety disorders sample. Behaviour Research and Therapy, 30, 33-38. Bryden, M. R, & Ley, R. G. (1983). Right hemisphere involvement in imagery and affect, in E. Perecman (Ed.), Cognitive processing in the right hemL~phere (pp. 111-123). New York: Academic Press.

MEMORY BIAS IN GENERALIZED ANXIETY DISORDER

755

Cloitre, M., & Leibowitz, M. R. (1991). Memory bias in panic disorder: An investigation of the cognitive avoidance hypothesis. Cognitive Therapy and Research, 15, 371-386. DiNardo, P. A., & Barlow, D. H. (1988). Attrie~ Disorders bzterviews Schedule-Revised (ADIS-R). Albany: State University of New York at Albany. Eysenck, M. W., & Byrne, A. (1994). Implicit memory, explicit memory bias, and anxiety. Cognition and Emotion, 8, 415-431. Friedman, B. H., & Thayer, J. F. (1998). Anxiety and autonomic flexibility: A cardiovascular approach. Biological Psychology, 49,303-323. Graf, P., (1994). Explicit and implicit memory: A decade of research. In C. Umilta & M. Moscovitch (Eds.), Attention and performance, Vol. XV: Conscious and nonconscious information processing (pp. 681-696). Cambridge, MA: Bradford. Hamilton, M. (1959). The measurement of anxiety states by rating. British Journal of Medical Psychology, 32, 50-55. Hamilton, M. (1960). A rating scale for depression. Journal of Neurology, Neurosurgery, and Psychiatry, 23, 56-62. Heide, F. J., & Borkovec, T. D. (1983). Relaxation-induced anxiety: Paradoxical anxiety enhancement due to relaxation training. Journal of Consulting and Clinical Psychology, 51, 171-182. Kolb, B., & Whishaw, I. Q. (1990). Fundamentals of human neuropsychology (3rd ed.). New York: Freeman. Lacey, J. I. (1967). Somatic response patterning and stress: Some revisions of activation theory. In M. H. Appley & R. Trumbull (Eds.), Psychological stress: Issues in research (pp. 1442). New York: Appleton-Century-Crofts. Lang, P. J., Levin, D. N., Miller, G. A., & Kozak, M. J. (1983). Fear behavior, fear imagery, and the psychophysiology of emotion: The problem of affective response integration. Journal of Abnormal Psychology. 92,276-306. Libkuman, T. M., Nichols-Whitehead, P., Griffith, J., & Thomas, R. (1999). Sources of arousal and memory for detail. Memory & Cognition. 27, 166-190. Lyonfields, J. D., Borkovec, T. D., & Thayer, J. F. (1995). Vagal tone in generalized anxiety disorder and the effects of aversive imagery and worrisome thinking. Behavior Therapy. 26, 457-466. MacLeod, C., & McLaughlin, K. (1995). Implicit and explicit memory bias in anxiety: A conceptual replication. Behaviour Research and Therapy. 33, 1-14. Mathews, A. (1990). Why worry? The cognitive function of anxiety. Behaviour Research and Therapy, 28,455-468. Mathews, A., & MacLeod, C. (1986). Selective processing of threat in anxiety states. Journal of Abnormal Psychology, 95,563-569. Mathews, A., Mogg, K., May, J., & Eysenck, M. (1989). Implicit and explicit memory bias in anxiety. Journal of Abnormal Psychology, 98,236-240. McGaugh, J. L. (1992). Affect, neuromodulatory systems, and memory storage. In S.-A. Christianson (Ed.), Handbook of emotion and memory: Research and theory. (pp. 245-268). Hillsdale, NJ: Erlbaum. Meyer, T. J., Miller, M. L., Metzger, R. L., & Borkovec, T. D. (1990]. Development and validation of the Penn State Worry Questionnaire. Behaviour Research and Therapy, 28, 487-495. Mogg, K., & Mathews, A. (1990). Is there a self-referent mood-congruent recall bias in anxiety? Behaviour Research and Therapy, 28, 91-92. Mogg, K., Mathews, A., & Weinman, J. (1989). Selective processing of threat cues in anxiety states: A conceptual replication. Behaviour Research and Therapy, 27, 317-323. Otto, M. W., McNally, R. J., Pollack, M. H., Chen, E., & Rosenbaum, J. F. (1994). Hemispheric laterality and memory bias for threat in anxiety disorder. Journal of Abnormal Psychology, 103,828-831. Postman, L. (1974). Does imagery enhance long-term retention? Bulletin of the Psychonomic Socie~, 3,385-387.

756

FRIEDMAN ET AL.

Simes, R. J. (1986). An improved Bonferroni procedure for multiple tests of significance. Biometrika, 73, 751-754. Spielberger, C. D., Gorsuch, R. L., Lushene, R., Vagg, P. R., & Jacobs, G. A. (1983). Manual jor the State-Trait An.rietv lm,entorv. Palo Alto, CA: Consulting Psychologists Press. Squire, L.R. (1992). Memory and the hippocampus: A synthesis from findings with rats, monkeys, and humans. Psychological Review, 99, 195-231. Thayer, J. F., Friedman, B. H., & Borkovec, T. D. (1996). Autonomic characteristics of generalized anxiety disorder and worry. Biological Psychiatry. 39,255 266. Thayer, J. F., Friedman, B. H., Borkovec, T. D., (2000). Physiological and attentional characteristics (,?fgeneralized anxiety disorder: A replicruion. Unpublished raw data. Thayer, J. F., Friedman, B. H., Borkovec, T. D., Johnsen, B. H., & Molina, S. (2000). Phasic heart period reactions to cued threat and non-threat stimuli in generalized anxiety disorder. Psychophysiology, 37, 361 368. Venables, R H. (1984). Arousal: An examination of its status as a concept. In M.G.H. Coles, J.R. Jennings, & J.P. Stern (Eds.), Psychophysiological perspectives (pp. 134-142). New York: Van Nostrand. Williams, J. M. G., Watts, F. N., MacLeod, C., & Mathews, A. (1997). Cognitive p,chology and emotional disorders (2nd ed.). Chichester, England: Wiley. Wolf, S. (1995). Dogmas that have hindered understanding, hltegrative Physiological and

Behavioral Science, 30, 3~4.

RECEIVED: January 3, 2000 ACCEPTED: May 1,2000