Journal of Psychosomatic Research 49 (2000) 107 ± 118 Anxiety sensitivity mediates the relationships between childhood learning experiences and elevated hypochondriacal concerns in young adulthood Margo C. Watt*, Sherry H. Stewart Department of Psychology, Dalhousie University, 1355 Oxford Street, Halifax, Nova Scotia, Canada B3H 4JI Received 29 April 1999; accepted 18 January 2000 Abstract Objective: The present study investigated childhood learning experiences potentially associated with the development of elevated hypochondriacal concerns in a non-clinical young adult sample, and examined the possible mediating roles of anxiety sensitivity (i.e., fear of anxiety-related symptoms) and trait anxiety (i.e., frequency of anxiety symptoms) in explaining these relationships. Method: 197 university students participated in a retrospective assessment of their childhood instrumental (i.e., parental reinforcement) and vicarious (i.e., parental modeling) learning experiences with respect to arousal ± reactive (e.g., dizziness) and arousal ± non-reactive (e.g., lumps) bodily symptoms, respectively. Childhood learning experiences were assessed using a revised version of the Learning History Questionnaire (LHQ), anxiety sensitivity levels with the Anxiety Sensitivity Index (ASI), trait anxiety levels with the State-Trait Anxiety Inventory-Trait (STAIT) scale, and degree of hypochondriacal concerns with the Illness Attitudes Scale (IAS)-Total score. Results: Consistent with earlier findings [Watt MC, Stewart SH, Cox BJ. A retrospective study of the learning history origins of anxiety sensitivity. Behav Res Ther 1998;36:505 ± 25.], elevated anxiety sensitivity levels were associated with increased instrumental and vicarious learning experiences related to both arousal ± reactive and arousal ± non-reactive bodily symptoms. Similarly, individuals with elevated hypochondriacal concerns also reported both more instrumental and vicarious learning experiences around bodily symptoms than did students with lower levels of such concerns. However, contrary to the hypothesis, the childhood learning experiences related to hypochondriacal concerns were not specific to arousal ± nonreactive symptoms, but instead involved parental reinforcement and modeling of bodily symptoms in general (arousal ± reactive and ± non-reactive symptoms alike). Anxiety sensitivity, but not trait anxiety, partially mediated the relationships between childhood learning experiences and elevated hypochondriacal concerns in young adulthood. Conclusions: Elevated anxiety sensitivity appears to be a risk factor for the development of hypochondriasis when learning experiences have involved both arousal ± reactive and arousal ± non-reactive bodily symptoms. D 2000 Elsevier Science Inc. All rights reserved. Keywords: Learning history; Anxiety sensitivity; Hypochondriacal concerns Anxiety sensitivity is defined as the fear of anxietyrelated bodily sensations arising from beliefs that these sensations have harmful physical, psychological, and/or social consequences [30,39]. For example, a person with high anxiety sensitivity might fear heart palpitations believing they portend a heart attack; fear dizziness believing it to signify imminent mental breakdown; and/or fear trembling in anticipation of public ridicule. In contrast, a person low in anxiety sensitivity would perceive such sensations to be unpleasant, but transient and harmless, consequences of being in an anxious state. * Corresponding author. Tel.: +1-902-485-8939; fax: +1-902-494-6585. E-mail address: ncwatt@north.nsis.com (M.C. Watt). Anxiety sensitivity has been suggested to be involved in the development and maintenance of anxiety disorders, particularly panic disorder [30,31]. Consistent with this suggestion, anxiety sensitivity levels have been shown to be elevated in panic disorder patients (e.g., [38,45]) and to serve as a pre-morbid vulnerability factor for the development of panic attacks in longitudinal research (e.g., [33]). More recently, interest has turned to the role of anxiety sensitivity in the development/maintenance of hypochondriasis. Cox [9] found elevated anxiety sensitivity levels in a sample of patients with primary hypochondriasis relative to healthy controls. Moreover, anxiety sensitivity levels have been shown to correlate positively with levels of hypochondriacal concerns in patients with panic disorder [25], patients with major depression [24], and non-clinical young adults [41]. 0022-3999/00/$ ± see front matter D 2000 Elsevier Science Inc. All rights reserved. PII: S 0 0 2 2 - 3 9 9 9 ( 0 0 ) 0 0 0 9 7 - 0 108 M.C. Watt, S.H. Stewart / Journal of Psychosomatic Research 49 (2000) 107±118 Reiss and McNally [31] originally suggested that anxiety sensitivity could arise from either genetic factors and/or learning experiences. Stein et al. [37] examined the heritability of anxiety sensitivity in 179 monozygotic and 158 dizygotic twin pairs. They found that anxiety sensitivity has a strong heritable component, accounting for nearly half of the variance in anxiety sensitivity levels (h2 = 0.45). However, the greatest proportion of variance in anxiety sensitivity levels (i.e., 55%) was attributable to environmental influences. Stein et al. [37] suggested that childhood learning experiences may increase anxiety sensitivity and, as a result, the risk for anxiety-related disorders. Learning theory posits at least two mechanisms in the etiology and maintenance of behavior: instrumental learning and vicarious conditioning. With instrumental learning, also referred to as operant conditioning, the individual's behavior is ``instrumental'' in getting something he/she desires (i.e., positive reinforcement) or removing something he/she does not desire (i.e., negative reinforcement). Instrumental learning could contribute to sensitivities to somatic sensations if a child's display of, or complaints about, bodily symptoms (e.g., dizziness, lumps) were rewarded in some way, such as being afforded special attention or being allowed to miss school. Vicarious conditioning, or observational learning, refers to learning by watching and imitating others [3]. Observational learning could account for the development of fear of somatic sensations if a parent modeled, and was rewarded for, fear reactions to their own bodily symptoms in the presence of their child, and/ or verbally transmitted their beliefs about the harmfulness of such symptoms to the child. Several studies have examined learning history contributions to the development of elevated hypochondriacal concerns. Bursky et al. [5] examined the childhood learning histories of a sample of hypochondriasis patients. The hypochondriasis patients reported more frequent illness as children, and being allowed to miss school more often, as compared to a control group of non-hypochondriacal medical outpatients. Similarly, Schwartz et al. [34] assessed the influences of childhood learning experiences in accounting for college students' levels of hypochondriacal concerns. All four social learning variables examined (i.e., familial modeling influences, sympathy when sick, avoidance of responsibility, and reinforcement of illness behavior) were correlated with scores on the Minnesota Multiphasic Personality Inventory-Hypochondriasis scale (MMPI-Hs; [17]). Sympathy when sick and avoidance of responsibility were also correlated with scores on a second measure of hypochondriacal concerns Ðthe Illness Attitudes Scale (IAS; [19]). Parker and Lipscombe [26] examined relations between childhood learning experiences and hypochondriacal concerns in adulthood in general practice outpatients. When experiencing childhood somatic symptoms (e.g., abdominal pains), those scoring higher on the `General Hypochondriasis' subscale of the Illness Behavior Questionnaire [29] rated both parents as eviden- cing more sympathy, and their mothers as more likely to call the doctor, than did remaining subjects. These retrospective studies provide evidence for the potential roles of both instrumental and vicarious conditioning in the development of hypochondriasis. In the first published study of the learning history origins of anxiety sensitivity [47], we found levels of anxiety sensitivity in young adulthood to be positively related to retrospectively-reported instrumental and vicarious conditioning experiences in childhood. Contrary to predictions, however, high anxiety sensitive individuals reported more such learning experiences related to both anxiety and cold symptoms prior to age 18 than individuals with lower levels of anxiety sensitivity. The lack of specificity to anxiety symptoms in our previous study stands in contrast to previous findings for panic disorder. For example, Ehlers [14] reported significant differences between panickers (i.e., panic disorder patients and infrequent panickers) and normal controls in terms of instrumental and vicarious learning experiences related to personal and parental anxiety symptoms, but no evidence for panic group differences in nonspecific parental encouragement of sick-role behavior in response to non-anxiety-related (i.e., cold) symptoms in childhood. A conceptual replication of Ehlers' study with the panickers vs. non-panickers in our previous study yielded results identical to those reported by Ehlers for panickers vs. normal controls [47]. Based on our previous study results, we suggested that the origins of heightened anxiety sensitivity may lie in learning to catastrophize the meaning of somatic symptoms in general rather than anxiety-related symptoms in particular [47]. As such, the childhood learning experiences of high anxiety sensitive individuals might bear some similarities not only to those individuals with panic disorder, but also to those individuals with hypochondriasis [47]. Given that the theories that the development and maintenance of both panic disorder (e.g., [8]) and hypochondriasis (e.g., [46]) involve catastrophic misinterpretation of somatic sensations (i.e., a construct similar to anxiety sensitivity), and the fact that anxiety sensitivity levels are elevated in both disorders [9,45], we suggested that the two disorders may share anxiety sensitivity as a common pre-morbid vulnerability factor [47]. Despite their similarities, panic disorder and hypochondriasis are distinguishable from one another in several respects including the types of somatic symptoms that are misinterpreted and feared [43,44]. Individuals with panic disorder misinterpret the significance of ``arousal ± reactive'' bodily symptoms (i.e., those which are exacerbated by physiological arousal, such as racing heartbeat and dizziness), whereas individuals with hypochondriasis tend to misinterpret the significance of ``arousal ± non-reactive'' bodily symptoms (i.e., those which are not immediately exacerbated by arousal, such as lumps and tiredness). Thus, we suggested that different learning history factors may be involved in the development of panic disorder vs. hypo- M.C. Watt, S.H. Stewart / Journal of Psychosomatic Research 49 (2000) 107±118 chondriasis [47]. For example, if a child learns to fear arousal ± reactive symptoms specifically through parental reactions to the child's or parent's anxiety symptoms, such learning experiences could place the child at risk for the development of panic attacks and panic disorder in adulthood (cf. Refs. [14,47]). If a child were to learn to fear arousal ± non-reactive bodily symptoms specifically, the child might be placed at risk for the development of general illness-related fears and hypochondriasis in adulthood. The general predisposing characteristic acquired in both of these childhood learning scenarios could be heightened anxiety sensitivity Ð an elevated tendency to catastrophize the meaning of bodily symptoms [25]. There were several limitations to our original study [47]. First, participants' learning experiences with respect to anxiety symptoms were compared with cold symptoms only (cf. Ref. [14]). The use of only one arousal± nonreactive symptom scale as a basis for comparison may limit the generalizability of our previous findings regarding the potential role of learning experiences with respect to somatic symptoms, in the development of elevated anxiety sensitivity. Second, whereas we included separate scales to examine the differential contributions of instrumental vs. vicarious conditioning experiences related to anxiety symptoms, only an instrumental learning scale was included for the experience of cold symptoms (cf. Ref. [14]). Consequently, we were unable to draw conclusions about the potential involvement of vicarious conditioning with respect to arousal ± non-reactive symptoms in the development of heightened anxiety sensitivity levels. Third, although we speculated about the possible roles of learning experiences and anxiety sensitivity in the development of hypochondriasis, no measure of hypochondriacal concerns was included in our previous study. Finally, given the on-going debate (e.g., Ref. [21] vs. Ref. [22]) as to the distinctiveness of anxiety sensitivity vs. trait anxiety (i.e., the tendency to experience anxiety symptoms in stressful circumstances; [36]), a further limitation of our original study was the failure to include a measure of trait anxiety. The present study was designed to replicate and extend the results of our first study on the learning history origins of elevated anxiety sensitivity [47]. We used several design modifications to overcome each of the limitations to our original study noted above. The present study tested three main hypotheses using a retrospective assessment methodology in a large non-clinical sample of young adults. First, we expected to replicate and extend our previous findings that elevated anxiety sensitivity levels would be associated with more learning experiences related to both arousal± reactive and arousal± non-reactive somatic symptoms prior to age 18 than lower levels of anxiety sensitivity, and that these learning experiences would include both instrumental and vicarious forms of conditioning. In order to test this hypothesis, the inquiry regarding the subjects' experiences with arousal± non-reactive symptoms was altered from cold 109 symptoms to include pain, lumps, stomach problems and tiredness. In addition, we incorporated a new scale to assess the role of childhood vicarious conditioning experiences in relation to these arousal± non-reactive bodily symptoms. Second, by incorporating a measure of hypochondriacal concerns, we were able to directly investigate the childhood learning history contributions to elevated hypochondriacal concerns in young adulthood. It was predicted that individuals reporting higher levels of hypochondriacal concerns would report only more instrumental and/or vicarious learning experiences specifically related to arousal±non-reactive symptoms compared to individuals with lower levels of hypochondriacal concerns. Third, consistent with the suggestion that anxiety sensitivity may constitute an underlying vulnerability for the development of elevated hypochondriacal concerns, we expected to find that anxiety sensitivity would serve a ``mediating'' or intervening role in explaining the hypothesized relationships between childhood learning experiences and hypochondriacal concerns in young adulthood (cf. Ref. [47]). By including a measure of trait anxiety, we were able to test the specificity of the hypothesized mediating role of anxiety sensitivity after accounting for the general tendency to experience anxiety when under stress (i.e., trait anxiety). Method Participants A total of 197 (156 female; 41 male) undergraduate students at Dalhousie University volunteered to serve as research participants in exchange for credit points. Students' mean age was 21.9 years (SD = 4.6). Eighty-two parents (73 mothers; 9 fathers) of the student respondents (i.e., 42%) completed a parental version of the Revised Learning History Questionnaire (LHQ-R) as a validity check.1 Measures Demographics measure An author-compiled measure sought basic demographic information (i.e., age and gender) for purposes of sample description. LHQ-R The original LHQ was developed by Ehlers [14] to explore the relationship of childhood experiences and panic. 1 The mean ASI, STAI-T, and IAS-Total scores of students whose parents completed the LHQ-R did not differ from those of the remaining students (ASI Ms (SDs) = 17.2 (8.9) vs. 16.0 (8.4), respectively; t (196) = 0.97, n.s.; STAI-T Ms (SDs) = 41.5 (10.7) vs. 39.7 (10.7), respectively; t (196) = 1.17, n.s.; IAS-Total Ms (SDs) = 30.5 (12.5) vs. 29.3 (13.1), respectively; t (196) = 0.61, n.s.). These results suggest that the 82 parent ± child validation pairs were representative of the larger sample of students. 110 M.C. Watt, S.H. Stewart / Journal of Psychosomatic Research 49 (2000) 107±118 Ehlers [14] modified a questionnaire designed by Whitehead et al. [48] to assess the social learning of sick-role behavior related to menstrual discomfort and cold symptoms in a sample of nursing students. Whitehead et al. [48] reported internal consistencies (Cronbach's alpha) of 0.81, 0.79, and 0.82 for scales entitled: `Encouragement of SickRole/Menstrual Distress,' `Encouragement of Sick-Role/ Colds,' and `Modeling of Sick-Role/Menstrual Distress,' respectively. The validity of the subjects' responses as determined by correlation with the mothers' responses was approximately r = 0.50 for each of the scales [48]. In her study, Ehlers [14] reported internal consistencies (Cronbach's alpha) for anxiety disorder groups of 0.78, 0.84, and 0.78 for the scales `Encouragement of Sick-Role/Panic Symptoms,'' `Modeling of Sick-Role/Panic Symptoms' and `Encouragement of Sick-Role/Colds,' respectively. Using Ehlers' [14] questionnaire, JaÈger [18], reported correlations between the statements of 16 infrequent panickers and their parents to range from r = 0.60 (Frequency of Subjects' Panic Symptoms) to r = 0.78 (Modeling of Sick-Role/ Panic Symptoms). In our previous study [47], we used an expanded version (Ehlers, personal communication, May 1995) of Ehlers' [14] LHQ which included the same subscales, but a larger number of learning history items per subscale. In the Watt et al. [47] study, we found internal consistency estimates (Cronbach's alpha) to range from 0.90 for the `Encouragement of Sick-Role/Colds' scale to 0.92 for both the `Encouragement of Sick-Role/Anxiety' and `Modeling of SickRole/Anxiety' scales. A validity check revealed that the students' LHQ responses were significantly positively correlated with those of their parents for the Experience/ Anxiety (r = 0.26) and Observation/Anxiety (r = 0.35) scales, but not the Experience/Colds (r = 0.12) scale [47]. The version of the LHQ used in our previous study [47] was further revised for use in the present study to assess both instrumental and vicarious learning experiences related to arousal ± reactive and arousal ± non-reactive somatic symptoms, respectively. This version will be referred to as the LHQ-R. The LHQ-R consisted of scales designed to assess the following information. Subjects' sick-role experiences/somatic symptoms. Subjects were first asked if and how often, prior to the age of 18, they experienced arousal± reactive bodily symptoms (e.g., racing heart, dizziness, shortness of breath, strong nausea) or arousal± non-reactive bodily symptoms (e.g., pains, lumps, stomach problems, tiredness). If subjects had experienced any of the listed somatic symptoms, they were asked to specify which of the listed symptoms applied. The specific symptoms endorsed, and the ratings of frequency of symptom occurrence were used in calculation of two symptom frequency scores for each participant: `Personal Arousal ± Reactive' and `Personal Arousal ± Non-Reactive,' respectively. If subjects endorsed either the `Personal Arousal ± Reactive' and/or the `Personal Arousal ±Non-Reactive' symptoms item(s), they were then asked to respond to questions regarding parental encouragement of sick-role behavior in response to their childhood bodily symptoms (`Encourage Symptoms' scale; 22 items). Subjects were asked 10 questions related to negative reinforcement of sick-role behavior such as ``When you had these symptoms prior to age 18 . . . did your parents encourage you to stay home from school? . . . were you excused from school homework?'' and positive-reinforcement of sick-role behavior such as `` . . . did you receive special care? . . . were you allowed to do things which were otherwise not allowed such as watching TV or staying up late?''; eight questions related to parents' verbal transmission of the idea that the child's bodily symptoms were dangerous such as `` . . . did you parents warn you of the possible dangers of your symptoms? . . . did your parents warn you that your symptoms could get worse and run out of control?''; and four questions related to parental discouragement (``punishment'') of sick-role behavior such as `` . . . were you made to feel responsible for having caused your symptoms? . . . did you feel left alone?'' For consistency with instrumental learning theory predictions, the four punishment items were reverse scored (cf. Ref. [47]). The `Personal Arousal ± Reactive' and `Personal Arousal ± Non-Reactive' symptom scores were separately multiplied by the mean `Encourage Symptoms' score to yield two composite scores reflecting the frequency of events in which the subject had, as a child, experienced arousal± reactive or arousal non-reactive bodily symptoms, respectively, and had also received special parental attention or instructions to take care of themselves (cf. Refs. [14,47]). These two composite scores were referred to as the `Experience/Arousal ± Reactive' and `Experience/Arousal ± Non-Reactive' scores, respectively. Observation of parental sick-role/somatic symptoms. Subjects were first asked `` . . . did your father or mother . . . suffer from symptoms such as racing heartbeat, dizziness, shortness of breath, or strong nausea?'' (Parental Arousal ± Reactive score) or ``pains, lumps, stomach problems, or tiredness?'' (Parental Arousal ± Non-Reactive score). Again, subjects were asked to specify which of the listed arousal± reactive and/or arousal ± non-reactive symptoms applied. The specific symptoms endorsed, and the ratings of frequency of symptom occurrence were used in calculation of two parental symptom frequency scores for each participant: `Parental Arousal ±Reactive' and `Parental Arousal ±NonReactive,' respectively. If subjects endorsed either the `Parental Arousal ±Reactive' and/or the `Parental Arousal ± Non-Reactive' symptoms item(s), they were then asked to respond to questions regarding parental modeling of sickrole behavior when parents suffered these symptoms (`Modeling Symptoms' scale; 20 items). These questions resembled the items on the `Encourage Symptoms' scale described above, including eight ``positive and negative reinforcement'' items, eight ``verbal transmission'' items, M.C. Watt, S.H. Stewart / Journal of Psychosomatic Research 49 (2000) 107±118 and four ``punishment'' items. The four punishment items were again reverse scored. `Parental Arousal ±Reactive' and `Parental Arousal ± Non-Reactive' symptom scores were separately multiplied by the mean `Modeling Symptoms' score to yield two composite scores of `Observation/ Arousal ± Reactive' and `Observation/Arousal ± Non-Reactive,' respectively, which reflected the frequency with which parents took special care of themselves or obtained special attention when experiencing arousal ± reactive or arousal± non-reactive bodily symptoms, respectively. Subjects responded to all questions on a relative frequency scale ranging from 0 ±3 where 0 = ``never,'' 1 = ``seldom'' (once or twice a year); 2 = ``occasionally'' (three to six times a year), and 3 = ``often'' (more than six times per year). Thus, scores on the four LHQ-R composite scales could each range from 0 to 9.2 The version of the LHQ-R completed by 82 students' parents was identical to the student version described above, except that all items were reworded to allow parents to retrospectively report on their adult child's frequency of symptoms during childhood, their own (parental) frequency of symptoms, and their responses to their child's and their own symptoms, when their child was growing up. Parents were also asked to indicate their relationship to the student participant (e.g., mother, father). Anxiety sensitivity index (ASI) The ASI [27] is a 16-item self-report inventory in which participants are asked to rate the extent to which they agree or disagree with each item pertaining to beliefs that anxiety symptoms are signs of harmful/aversive consequences, by selecting one of five points on a Likert scale. The scale ranges from ``very little'' (zero point) to ``very much'' (four points). The total ASI score, which can range from 0 to 64, is obtained by summing the item scores. Considerable evidence supports the good psychometric properties of the ASI, including evidence of its high internal consistency, high test ± retest reliability, criterion-related validity in distinguishing anxiety disordered patients from controls, and construct validity as a measure of fear of anxiety as distinct from trait anxiety [27]. State-trait anxiety inventory-trait scale (STAI-T) The STAI-T [36] is a 20-item self-report measure on which participants rate how anxious they ``generally'' feel on four-point Likert scales that range from ``almost never'' (one point) to ``almost always'' (four points). The STAI-T contains nine reverse-scored, anxiety-absent items. Total STAI-T scores can range from 20 to 80. Considerable 2 In order to compare all subjects according to their childhood learning experiences, we chose to assign item scores of 0 for the `Encourage Symptoms' and `Modeling Symptoms' scales (prior to the calculation of the composite LHQ-R scales), in cases where subjects reported not having experienced any of the listed somatic symptoms, and thus were not exposed to the learning experiences described (cf. Ref. [47]). 111 evidence supports the good psychometric properties of the STAI-T including high reliability, and good construct validity in predicting degree of anxiety in response to stressors (see the review in Ref. [36]). The STAI-T was included in the present study as a measure of trait anxiety Ða personality construct which is conceptually and empirically distinct from anxiety sensitivity [22]. IAS The IAS [19] is a 29-item self-report instrument designed to measure fears, attitudes, and beliefs associated with hypochondriasis and abnormal illness behavior. Items are self-rated on five-point Likert scales, with endpoints of ``no'' (zero point) and ``most of the time'' (four points). Two of the items are open-ended questions asking about the nature of illness and nature of treatment and are not included in scoring. The IAS-Total score (sum across all 27 scorable items) was used in the present study as an index of degree of hypochondriacal concerns (cf. Ref. [34]), given factor analytic findings that the IAS possesses a hierarchical structure with a single, global factor of `general hypochondriacal concerns' at the higher-order level ([16,41]; also, see the review in Ref. [42]). The IAS possesses good psychometric properties. Test ± retest reliability is high [15,19]. Good concurrent validity has been demonstrated for the IAS in terms of its ability to discriminate hypochondriasis patients from family practice patients, non-hypochondriacal psychiatric patients, and non-patient employees [20]. Finally, convergent validity of the IAS has been demonstrated in Refs [34,35]: IAS scores are significantly correlated with the scores on alternative hypochondriasis measures including the Whitley Index [28], the Somatosensory Amplification Scale [6], and the MMPI-Hs scale [17]. Procedure The purpose of the study as explained to the participants was to investigate the relationships between childhood learning experiences and attitudes toward health/ illness in young adulthood. After providing written informed consent, the volunteer student participants completed the self-report measures anonymously during class time in the order listed above. All participants were invited to have one of their parents (i.e., the parent to whom they had referenced their responses to the relevant LHQ-R items) complete a comparable version of the LHQ-R. A parental package was distributed to willing students, which contained an informed consent form in which the purposes of the study were described, a parental version of the LHQ-R, and a stamped envelope in which the parent could return the completed questionnaires. Parental questionnaire packages were mailed to the students' parents in cases where the students were living at a distance from their parents. A numerical coding scheme was used to match the student and parent questionnaire responses, to preserve the respondents' anonymity. 112 M.C. Watt, S.H. Stewart / Journal of Psychosomatic Research 49 (2000) 107±118 Results Summary data On average, the sample scored 16.5 (SD = 8.6) on the ASI, 29.8 (SD = 12.9) on the IAS-Total, and 40.5 (SD = 10.7) on the STAI-T. These means compare well with published norms for university students on these measures (e.g., [27,34,36]). Bivariate correlations computed between scores on the ASI, IAS-Total, and STAI-T revealed that ASI scores were positively correlated with both IAS-Total and STAI-T scores (rs = 0.61 and 0.60, respectively; ps < 0.001; one-tailed tests). IAS-Total scores were also positively correlated with STAI-T scores (r = 0.37; p < 0.001; onetailed test). Additional information on the pattern of relationships between anxiety sensitivity, trait anxiety, and specific components of IAS hypochondriacal concerns can be found in Ref. [41]. The sample means, medians, and SDs on the four composite LHQ-R scales in the present study were: Experience/ Arousal ±Non-Reactive (M = 1.18; Mdn = 0.91; SD = 1.45); Experience/Arousal ± Reactive (M = 0.96; Mdn = 0.00; SD = 1.31); Observation/Arousal ±Non-Reactive (M = 0.94; Mdn = 0.00; SD = 1.64); and Observation/Arousal ± Reactive ( M = 0.61; Mdn = 0.00; SD = 1.23). These values compare well with the sample means, medians, and SDs on the three original LHQ scales in the study in Ref. [47]: Experience/ Colds (M = 1.32; Mdn = 1.05; SD = 0.99); Experience/ Anxiety (M = 0.61; Mdn = 0.00; SD = 1.10); and Observation/Anxiety (M = 0.54; Mdn = 0.00; SD = 1.17). Across the two studies, student participants tended to score higher on learning history scales tapping parental responses to arousal± non-reactive symptoms than on comparable scales tapping parental responses to arousal± reactive symptoms, and higher on learning history scales tapping instrumental learning vs. vicarious learning experiences. Prior to further analyses, square root transformations were conducted on the four LHQ-R scale scores since raw scores revealed significant positive skew (cf. Refs. [14,47]). Bivariate correlations were computed between square root transformed scores on the four LHQ-R composite scales. All scales were significantly positively inter-correlated with one another with shared variance ranging from 8% (Experience/Arousal ± Reactive with Observation/Arousal ± Non-Reactive) to 38% (Observation/Arousal ± Reactive with Observation/Arousal ± Non-Reactive) (all ps < 0.001; onetailed tests). The average inter-scale correlation was r = 0.40 (p < 0.001; one-tailed test) indicating an average shared variance between LHQ-R scales of 16%. Reliability and validity data for LHQ-R Internal consistencies (Cronbach's alphas) for the two multi-item LHQ-R scales were calculated based on the responses of the total student sample (N = 197). Coefficient alphas were 0.91 for the `Encourage Symptoms' scale and 0.96 for the `Modeling Symptoms' scale, indicating excellent internal consistency [23]. As a validity check, correlations were performed between the four square root transformed composite LHQ-R scales of the 82 student ± parent pairs. One-tailed tests indicated that parents' and adult children's responses were significantly positively correlated with one another in each case: r = 0.34, p < 0.001 for Observation/Arousal ± Non-Reactive; r = 0.31, p < 0.005 for Experience/ Arousal ±Non-Reactive; r = 0.23, p < 0.05 for Observation/Arousal ± Reactive; and r = 0.21, p < 0.05 for Experience/Arousal ±Reactive, respectively. These results support the validity of the students' LHQ-R scores (cf. Refs. [18,47]). Hypothesis testing To address the first of our hypotheses (i.e., that levels of anxiety sensitivity would be positively associated with learning experiences related to both arousal±reactive and arousal ± non-reactive somatic symptoms and that these learning experiences would include both instrumental and vicarious forms of conditioning), correlations between childhood learning experiences and anxiety-related personality variables (i.e., anxiety sensitivity and trait anxiety) were examined. Bivariate correlations were computed between scores on the four square root transformed composite scores of the LHQ-R and scores on the ASI. These four correlations are displayed in the upper left of Fig. 1a ± d, respectively. Consistent with the predictions and the findings of our earlier study [47], ASI scores were significantly positively correlated with the scores on all four LHQ-R scales (see Fig. 1a ±d).3 This result indicates that elevated anxiety sensitivity is associated with greater instrumental and vicarious childhood learning experiences with respect to both arousal± reactive and arousal ±non-reactive symptoms. A similar pattern was observed in the case of correlations between STAI-T scores and LHQ-R scale scores (lower left, Fig. 1a ± d),4 save that STAI-T scores were unrelated to the scores on the `Observation/Arousal ± Reactive' scale of the LHQ-R (see Fig. 1c). Similarly, we tested the second of our hypotheses (i.e., that levels of hypochondriacal concerns would be positively related to instrumental and/or vicarious learning experiences specifically related to arousal± non-reactive symptoms only) by examining bivariate correlations between childhood 3 We compared the magnitude of the correlations between ASI scores and each of the four LHQ-R variables using a set of dependent sample ztests to determine whether anxiety sensitivity was more strongly related to any particular childhood learning experience(s). No significant differences were found between correlations (largest z = 1.2, n.s.). 4 We compared the magnitude of the correlations between STAI-T scores and each of the four LHQ-R variables using a set of dependent sample z-tests to determine whether trait anxiety was more strongly related to any particular childhood learning experience(s). No significant differences were found between correlations (largest z = 1.1, n.s.). M.C. Watt, S.H. Stewart / Journal of Psychosomatic Research 49 (2000) 107±118 113 Fig. 1. Summary of regression analyses evaluating the potential mediating roles of anxiety sensitivity (ASI scores) and trait anxiety (STAI-T scores) in explaining relations between childhood learning experiences (LHQ-R scale scores) and hypochondriacal concerns in young Adulthood (IAS-Total scores): (a) instrumental learning experiences for arousal ± reactive symptoms (LHQ-R Experience/Arousal ± Reactive scores); (b) instrumental learning experiences for arousal ± non-reactive symptoms (LHQ-R Experience/Arousal ± Non-Reactive scores); (c) vicarious learning experiences for arousal ± reactive symptoms (LHQ-R Observation/Arousal ± Reactive scores); and (d) vicarious learning experiences for arousal ± non-reactive symptoms (LHQ-R Observation/Arousal ± Non-Reactive scores). ****p < 0.001; ***p < 0.005; **p < 0.01; *p < 0.05 (one-tailed tests). Partial correlation coefficients are displayed in parentheses. Partial correlations between LHQ-R variables and IAS-Total scores control for ASI and STAI-T scores. Partial correlations between ASI and IAS-Total scores control for the LHQ-R variable in question and STAI-T scores. Partial correlations between STAI-T scores and IAS-Total scores control for the LHQ-R variable in question and ASI scores. ASI = Anxiety Sensitivity Index [27]; STAI-T = State-Trait Anxiety Inventory-Trait scale [36]; IAS-Total = Illness Attitudes Scale [19] Total score; LHQ-R = Revised version of Ehlers' [14] Learning History Questionnaire. LHQ-R scale scores square root transformed to reduce skew. learning experiences and hypochondriacal concerns (see bivariate correlations, middle of Fig. 1a ± d, respectively). Consistent with our hypothesis, IAS-Total scores were significantly positively correlated with the scores on the `Experience/Arousal ± Non-Reactive' and `Observation/ Arousal ±Non-Reactive' LHQ-R scales (see Fig. 1b and d, respectively), indicating that greater levels of hypochondriacal concerns were associated with greater instrumental and vicarious learning experiences for childhood arousal± non-reactive symptoms. Contrary to expectation, IAS-Total scores were also significantly positively correlated with the scores on the `Experience/Arousal ± Reactive' and `Observa- tion/Arousal± Reactive' scales of the LHQ-R (see Fig. 1a and c, respectively), suggesting a lack of symptom-specificity in the childhood learning histories of young adults with high hypochondriacal concerns.5 To test the third hypothesis of the study (i.e., that anxiety sensitivity mediates the relationship between childhood 5 We compared the magnitude of the correlations between IAS-Total scores and each of the four LHQ-R variables using a set of dependent sample z-tests to determine whether hypochondriacal concerns were more strongly related to any particular childhood learning experience(s). No significant differences were found between correlations (largest z = 1.1, n.s.). 114 M.C. Watt, S.H. Stewart / Journal of Psychosomatic Research 49 (2000) 107±118 learning experiences and hypochondriacal concerns), four sets of mediator regression analyses were performed using square root transformed scores on each of the four LHQ-R composite scores as predictor variables, respectively. The results of these analyses are schematically depicted in Fig. 1a ±d, respectively. For the analyses involving square root transformed Experience/Arousal ± Reactive LHQ-R scores (Fig. 1a), in the first step, Experience/Arousal ± Reactive scores were found to have a significant effect on the ASI scores (t = 4.10, p < 0.001) and a significant effect on the STAI-T scores (t = 2.74, p < 0.01) suggesting that either personality variable could serve as a potential mediator. In the second step, Experience/Arousal ± Reactive scores were found to have a significant effect on the levels of hypochondriacal concerns (i.e., IAS-Total scores) (t = 4.64, p < 0.001). In the third step, IAS-Total scores were simultaneously regressed on the ASI scores, STAI-T scores, and Experience/Arousal ±Reactive scores. ASI scores (t = 7.89, p < 0.001) and Experience/ Arousal ±Reactive scores (t = 2.67, p < 0.01), but not STAIT scores (t = 0.06, n.s.), were significant univariate predictors of levels of hypochondriacal concerns. Although Experience/Arousal ± Reactive scores continued to predict IAS-Total scores, the magnitude of the relation was substantially reduced relative to the second step, after accounting for the effects of the personality variables (see Fig. 1a). For the analyses involving square root transformed Experience/Arousal ± Non-Reactive LHQ-R scores (Fig. 1b), in the first step, Experience/Arousal ± Non-Reactive scores were found to have a significant effect on the ASI scores (t = 2.37, p < 0.05) and a significant effect on the STAI-T scores (t = 3.07, p < 0.005) suggesting that either personality variable could serve as a potential mediator. In the second step, Experience/Arousal ± Non-Reactive scores were found to have a significant effect on the levels of hypochondriacal concerns (i.e., IAS-Total scores) (t = 3.21, p < 0.005). In the third step, IAS-Total scores were simultaneously regressed on the ASI scores, STAI-T scores, and Experience/Arousal ± Non-Reactive scores. ASI scores (t = 8.48, p < 0.001) and Experience/Arousal ±Non-Reactive scores (t = 2.20, p < 0.05), but not STAI-T scores (t = 0.17, n.s.), were significant univariate predictors of the levels of hypochondriacal concerns. Although Experience/Arousal ± Non-Reactive scores continued to predict IAS-Total scores, the magnitude of the relation was substantially reduced relative to the second step, after accounting for the effects of personality variables (see Fig. 1b). For the analyses involving square root transformed Observation/Arousal ±Reactive LHQ-R scores (Fig. 1c), in the first step, Observation/Arousal ±Reactive scores were found to have a significant effect on the ASI scores (t = 2.45, p < 0.05) but no significant effect on the STAI-T scores (t = 1.15, n.s.) suggesting that only anxiety sensitivity could serve as a potential mediator. In the second step, Observation/Arousal ±Reactive scores were found to have a significant effect on the levels of hypochondriacal concerns (i.e., IAS-Total scores) (t = 2.88, p < 0.005). In the third step, IAS-Total scores were simultaneously regressed on the ASI scores, STAI-T scores, and Observation/Arousal ± Reactive scores.6 ASI scores (t = 8.19, p < 0.001), but not STAI-T scores (t = 0.19, n.s.) or Observation/ Arousal ±Reactive scores (t = 1.74, n.s.), were a significant univariate predictor of levels of hypochondriacal concerns (see Fig. 1c). Thus, Observation/Arousal ± Reactive scores were no longer related to levels of hypochondriacal concerns, after accounting for the effects of personality variables. For the analyses involving square root transformed Observation/Arousal ±Non-Reactive LHQ-R scores (Fig. 1d), in the first step, Observation/Arousal ± Non-Reactive scores were found to have a significant effect on the ASI scores (t = 3.34, p < 0.001) and a significant effect on the STAI-T scores (t = 2.34, p < 0.05) suggesting that either personality variable could serve as a potential mediator. In the second step, Observation/Arousal ± Non-Reactive scores were found to have a significant effect on the levels of hypochondriacal concerns (i.e., IAS-Total scores) (t = 3.64, p < 0.001). In the third step, IAS-Total scores were simultaneously regressed on the ASI scores, STAI-T scores, and Observation/ Arousal ±Non-Reactive scores. ASI scores (t = 8.12, p < 0.001) and Observation/Arousal ± Non-Reactive scores (t = 2.01, p < 0.05), but not STAI-T scores (t = 0.08, n.s.), were significant univariate predictors of levels of hypochondriacal concerns. Although Observation/Arousal ±Non-Reactive scores continued to predict IAS-Total scores, the magnitude of the relation was substantially reduced relative to the second step, after accounting for the effects of the personality variables (see Fig. 1d). Discussion The present study was designed to investigate the roles of instrumental and vicarious childhood learning experiences in the development of heightened anxiety sensitivity and elevated hypochondriacal concerns in young adulthood. Watt et al. [47] proposed that hypochondriasis may develop as a result of childhood learning experiences, and that elevated anxiety sensitivity may serve a mediating or intervening role in explaining these childhood experience Ð adult outcome associations. Specifically, we proposed that childhood exposure to parental reinforcement and/or parental modeling of arousal± non-reactive somatic symptoms would lead to a heightened tendency to catastrophize the meaning of bodily symptoms (i.e., elevated anxiety sensi6 Although learning experiences failed to predict STAI-T scores in the first step of the regression analyses using `Observation/Arousal ± Reactive' scores, STAI-T scores were entered simultaneously with ASI scores and `Observation/Arousal ± Reactive' scores in the third step in order to assess the potential mediating role of ASI scores after controlling for the influences of STAI-T scores. M.C. Watt, S.H. Stewart / Journal of Psychosomatic Research 49 (2000) 107±118 tivity), which in turn would predispose the child toward the development of general illness-related fears and hypochondriasis in young adulthood. Although the findings of the present study do confirm a mediating role for anxiety sensitivity in the development of hypochondriacal concerns, the learning experiences involved in the origins of elevated hypochondriacal concerns do not appear to be specific to arousal± non-reactive somatic symptoms. Consistent with the first hypothesis of the present study, elevated anxiety sensitivity levels were found to be related to retrospectively-reported learning experiences in childhood and adolescence. In keeping with the findings of our earlier study [47], the childhood experiences predicting higher levels of anxiety sensitivity were not specific to anxiety-related (arousal ±reactive) symptoms, but involved parental responses to somatic symptoms in general-arousal ± reactive and arousal ±non-reactive symptoms alike. In addition, these learning experiences included both instrumental (i.e., parental reinforcement) and vicarious (i.e., parental modeling) forms of conditioning (cf. Ref. [47]). The present findings thus extend our previous findings linking anxiety sensitivity with a greater intensity of instrumental learning symptoms with respect to cold symptoms, to other arousal± non-reactive symptoms such as lumps, tiredness, pain and stomach problems. Further, through the inclusion of the `Observation/Arousal ± Non-Reactive' scale, the present findings also extend to the vicarious learning modality, our previous results linking anxiety sensitivity with a greater intensity of instrumental learning experiences for arousal± non-reactive symptoms [47]. Thus, the findings of the present study are consistent with our previous suggestion that the origins of heightened anxiety sensitivity may lie in learning to catastrophize the meaning of somatic symptoms in general rather than arousal± reactive (anxiety-related) symptoms in particular. In fact, several recent studies (e.g., [1,2,12]) suggest that anxiety sensitivity may represent a construct broader than fear of anxiety-related sensations, per se. For example, in Ref. [1], it was found that chronic back pain patients with high anxiety sensitivity exhibit greater anxiety in response to pain, and greater fear of potential negative consequences of pain, than low anxiety sensitivity chronic back pain patients. Cox [10] speculated that anxiety sensitivity may be part of a broader set of beliefs about the harmfulness of internal sensations in general (cf. Ref. [25]). A second study hypothesis was that certain childhood instrumental and vicarious learning experiences would be significantly associated with elevated hypochondriacal concerns in young adulthood (cf. Refs. [5,26,34]). Specifically, we hypothesized that the childhood learning experiences predicting elevated hypochondriacal concerns in young adulthood would be specific to arousal± non-reactive symptoms (e.g., lumps, tiredness) as opposed to arousal ±reactive somatic symptoms (e.g., racing heartbeat, dizziness), since a tendency to catastrophize the meaning of internal bodily sensations which are not immediately exacerbated by anxi- 115 ety has been said to be characteristic of patients with clinical hypochondriasis [43,44]. As predicted, students reporting higher levels of hypochondriacal concerns on the IAS indicated a greater intensity of parental reinforcement and parental modeling of arousal±non-reactive symptoms when they were growing up, relative to students with lower levels of hypochondriacal concerns. However, contrary to prediction, elevated hypochondriacal concerns were also associated with a greater intensity of parental reinforcement and parental modeling of arousal±reactive somatic symptoms as well. There are several potential explanations for our lack of support for the predicted symptom-specificity in the learning histories of those with elevated hypochondriacal concerns. The first explanation pertains to measurement artifact associated with the LHQ-R. Since the LHQ-R (in contrast with the version of the LHQ used in Ref. [47]) had respondents' rate parental behavior in response to arousal ± reactive and arousal± non-reactive symptoms on the same scale, redundancy between the `Experience/Arousal ± Reactive' and `Experience/Arousal ± Non-Reactive' scores and between the `Observation/Arousal ±Reactive' and `Observation/Arousal ± Non-Reactive' scores may be at play. However, correlations between scores on the four composite LHQ-R scales revealed that the scales shared an average of only 16% common variance providing little evidence of measurement redundancy across scales. Another possible explanation pertaining to measurement artifact relates to our choice of instrument for operationalizing level of hypochondriacal concerns. The IAS [19] has been criticized as a measure of hypochondriasis due to its general failure to specify fears and concerns about arousal± non-reactive as opposed to arousal ±reactive symptoms ([43,44]; also, see the review in Ref. [42]). Future research should attempt to replicate the current findings after controlling for these potential measurement artifacts. Assuming the present results can be replicated, our findings suggest that elevated hypochondriacal concerns (like high anxiety sensitivity) may originate by way of a child learning to fear bodily symptoms in general (arousal ± nonreactive and arousal± reactive symptoms alike) rather than arousal ± non-reactive symptoms in particular. This result stands in contrast to previous findings that panic attacks and panic disorder are specifically related to instrumental and vicarious learning experiences pertaining to anxiety-related, arousal ± reactive symptoms such as racing heartbeat and dizziness [14,40,47]. Taken together, the results of studies employing the LHQ and LHQ-R are quite consistent, however, with Salkovskis and Clark's [32] contention that ``whilst the panic-type response is likely to be confined to sensations subject to rapid increase when anxious, both anxiety-related and anxiety-unrelated symptoms may play a part in the concerns of hypochondriacal patients'' (p. 29). A third hypothesis of the present study was that anxiety sensitivity (but not the correlated personality variable of trait anxiety) would serve a mediating or intervening role in 116 M.C. Watt, S.H. Stewart / Journal of Psychosomatic Research 49 (2000) 107±118 explaining the relations between childhood learning experiences and elevated hypochondriacal concerns in young adulthood. Indeed, compelling evidence was found for the role of anxiety sensitivity as a ``partial mediator'' [4] in the development of elevated hypochondriacal concerns. In initial bivariate analyses, both learning experiences and anxiety sensitivity were significantly related to the degree of hypochondriacal concerns, but after controlling for anxiety sensitivity, learning experiences had much weaker effects on the level of hypochondriacal concerns. In the case of `Observation/Arousal ±Reactive' scores, learning experiences no longer had any significant effect on the level of hypochondriacal concerns after accounting for anxiety sensitivity levels. The specificity of the mediational effect of anxiety sensitivity was examined by including a correlated but conceptually distinct measure of trait anxiety. In each set of regression analyses, anxiety sensitivity (but not trait anxiety) was found to be a significant mediator. This finding thus adds to the literature about the many ways in which anxiety sensitivity and trait anxiety are empirically distinct constructs (cf. Ref. [22]). Our finding that anxiety sensitivity served a mediating role in the development of elevated hypochondriacal concerns adds to recent findings that anxiety sensitivity partially mediates the relation between childhood learning experiences for arousal± reactive symptoms and the development of panic attacks in young adulthood [40]. Several potential limitations of the current study suggest important avenues for future research. First, since our crosssectional, correlational design cannot establish causality or direction of causality, longitudinal research should further investigate the pathways to the development of hypochondriasis suggested by the present mediational analyses. Specifically, researchers could follow children over time to determine whether instrumental and vicarious learning experiences lead to heightened anxiety sensitivity which in turn increases risk for elevated hypochondriacal concerns in young adulthood. Second, given that concerns have been expressed that the ASI may contain a few items which are conceptually-redundant with hypochondriasis measures [11,41], it would be useful for future research to attempt to replicate the present mediational results after controlling for item redundancy between the ASI and IAS (see Ref. [13], for sample methodology). Third, our study employed a retrospective design. Retrospective reports are often criticized as being subject to biases and distortions due to selective memory of past events; the influences of current attitudes, behaviors, and experiences; and demand characteristics of the study itself. For example, scores on the LHQ-R scales may reflect the same over-concern with bodily sensations as tapped by the measures of present anxious and hypochondriacal symptoms, rather than reflecting actual early experiences. Thus, in keeping with our earlier study [47], we attempted to validate the students' retrospective reports of their childhood experiences by inviting their parents to complete comparable versions of the LHQ-R. As in earlier studies (i.e., [18,47]), parental responses were found to provide statistically significant corroboration of the responses of their adult children. This finding is consistent with the conclusion of Brewin et al. [7] that ``provided that individuals are questioned about the occurrence of specific events or facts that they were sufficiently old and well placed to know about, the central features of their accounts are likely to be reasonably accurate'' (p. 94). Although the small sample size of parents (N = 82) precluded such analyses in the present study, it would nonetheless be useful for future research to correlate parental LHQ-R scores with their adult children's levels of anxiety sensitivity and hypochondriacal concerns, particularly since correlations between parents' and adult children's LHQ-R scale scores were generally small in magnitude (cf. Ref. [47]). If parent ratings showed similar associations with these criterion variables as seen in the present study with adult child ratings, this would strengthen the argument that actual childhood experiences played a significant role in the development of individual differences on these variables. Other potential limitations include our use of a nonclinical sample, and our use of questionnaires vs. structured interviews in the assessment of hypochondriacal concerns, which may preclude the generalization of results to the clinical condition of hypochondriasis. Although our results are consistent with those of Barsky et al. [5] regarding associations between early learning experiences and clinical hypochondriasis, the role of anxiety sensitivity in mediating the relationship between childhood learning experiences and the development of hypochondriasis, still requires testing in a clinical sample. A further potential limitation pertains to the content of the LHQ and LHQ-R used in the present and previous studies. These questionnaires were designed to address issues of symptom specificity (i.e., arousal ±reactive vs. arousal± non-reactive symptoms) and mechanism specificity (i.e., instrumental vs. vicarious conditioning) in the learning histories of individuals with high anxiety sensitivity, panic, or elevated hypochondriacal concerns. Items were not, however, constructed to assess the particular feared consequences of bodily symptoms, which are acquired through these learning pathways. Specifically, a child might learn to fear that bodily symptoms are signs of immediate negative consequences (e.g., learning to misinterpret a racing heartbeat as a sign of an impending heart attack) or signs of more delayed negative consequences (e.g., learning to misinterpret lumps as a sign of an ongoing disease process such as cancer). Salkovskis and Clark [32] have suggested that the former type of catastrophic misinterpretation is characteristic of panic disorder, whereas the latter type is characteristic of hypochondriasis. Future research on the learning history factors associated with the development of these two forms of behavioral pathology might benefit from an examination of this issue of ``feared consequence specificity'' in the conditioning histories of each clinical group. M.C. Watt, S.H. Stewart / Journal of Psychosomatic Research 49 (2000) 107±118 Finally, whereas the present study focused on environmental factors in accounting for anxiety sensitivity levels, the importance of biological/genetic factors must not be overlooked [37]. However, accepting an inherited basis for anxiety sensitivity is not incompatible with the present conception of anxiety sensitivity as arising from learned sensitivities to and fears of somatic sensations. In fact, interactions between genetic and learning factors are likely. For example, parents may respond more to a child with an inherited propensity to fear bodily sensations, and vicarious and instrumental learning factors may serve to ``activate'' or intensify inherited susceptibilities toward anxiety sensitivity. Heightened anxiety sensitivity may in turn increase the likelihood of development of both panic attacks and elevated hypochondriacal concerns, with panic being the most likely outcome if learning experiences have been specific to arousal ± reactive symptoms [14,47] and hypochondriasis being the most likely outcome if learning experiences have been more general to arousal± reactive and arousal± nonreactive bodily symptoms alike (cf. Ref. [32]). [10] [11] [12] [13] [14] [15] [16] [17] [18] Acknowledgments This research was supported by a grant from the Dalhousie University Research Development Fund for the Arts (RDFA) awarded to the second author. The first author was supported by a Natural Sciences and Engineering Research Council of Canada Doctoral Scholarship and an Honorary Killam Scholarship at the time this research was conducted. The authors would like to acknowledge Dr. Anke Ehlers for providing the copy of the LHQ from which the present LHQ-R was derived. References [1] Asmundson GJG, Norton GR. Anxiety sensitivity in patients with physically unexplained chronic back pain: a preliminary report. Behav Res Ther 1995;33:771 ± 7. [2] Asmundson GJG, Taylor S. Role of anxiety sensitivity in pain-related fear and avoidance. J Behav Med 1996;19:573 ± 82. [3] Bandura A. Social Foundations of Thought and Action: A Social Cognitive Theory. Upper Saddle River, NJ: Prentice-Hall, 1986. [4] Baron RM, Kenny DA. The moderator ± mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol 1986;51:1173 ± 82. [5] Barsky AJ, Wool C, Barnett MC, Cleary PD. Histories of childhood trauma in adult hypochondriacal patients. Am J Psychiatry 1994;151:397 ± 401. [6] Barsky AJ, Wyshak G. Hypochondriasis and somatosensory amplification. Br J Psychiatry 1990;157:404 ± 9. [7] Brewin CR, Andrews B, Gotlib IH. Psychopathology and early experience: a reappraisal of retrospective reports. Psychol Bull 1993; 113:82 ± 98. [8] Clark DM. A cognitive approach to panic. Behav Res Ther 1986;24:461 ± 70. [9] Cox BJ. The Role of Anxiety Sensitivity in Panic and other Disorders [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] 117 [Summary]. Program Abstracts of the 19th National Conference of the Anxiety Disorders Association of America 1999;24:86. Cox BJ. The nature and assessment of catastrophic thoughts in panic disorder. Behav Res Ther 1996;34:363 ± 74. Cox BJ, Borger SC, Enns MW. Anxiety sensitivity and emotional disorders: psychometric properties and their theoretical implications. In: Taylor S, editor. Anxiety Sensitivity: Theory, Research, and Treatment of the Fear of Anxiety. Mahwah, NJ: Erlbaum, 1999. pp. 115 ± 48. Cox BJ, Kuch K, Parker JDA, Shulman ID, Evans RJ. Alexithymia in somatoform disorder patients with chronic pain. J Psychosom Res 1994;38:523 ± 7. Devine H, Stewart SH, Watt MC. Relations between anxiety sensitivity and dimensions of alexithymia. J Psychosom Res 1999;47: 145 ± 58. Ehlers A. Somatic symptoms and panic attacks: a retrospective study of learning experiences. Behav Res Ther 1993;31:269 ± 78. Fava GA, Kellner R, Zielezny M, Grandi S. Hypochondriacal fears and beliefs in agoraphobia. J Affective Disord 1988;14:239 ± 44. Hadjistavropoulos HD, Frombach IK, Asmundson GJG. Exploratory and confirmatory factor analytic investigations of the illness attitudes scale in a nonclinical sample. Behav Res Ther 1999;37:671 ± 84. Hathaway SR, McKinley JC. The Minnesota Multiphasic Personality Inventory Manual. New York: The Psychological Corporation, 1951. JaÈger I. Soziales Lernen and Panikanfalle. Unpublished thesis. University of Tubingen, Germany, 1988. Kellner R. Abridged Manual of the Illness Attitudes Scale. Department of Psychiatry, School of Medicine, University of New Mexico, 1987. Kellner R, Abbott P, Winslow WW, Pathak D. Fears, beliefs, and attitudes in DSM-III hypochondriasis. J Nerv Ment Dis 1987;175:20 ± 5. Lilienfeld SO. Anxiety sensitivity is not distinct from trait anxiety. In: Rapee RM, editor. Current Controversies in the Anxiety Disorders. New York: Guilford Press, 1996. pp. 228 ± 44. McNally RJ. Anxiety sensitivity is distinguishable from trait anxiety. In: Rapee RM, editor. Current Controversies in the Anxiety Disorders. New York: Guilford Press, 1996. pp. 214 ± 27. Nunnally JC. Psychometric Theory. New York: McGraw-Hill, 1978. Otto MW, Demopulos CM, McLean NE, Pollack MH, Fava M. Additional findings on the association between anxiety sensitivity and hypochondriacal concerns: examination of patients with major depression. J Anxiety Disord 1998;12:225 ± 32. Otto MW, Pollack MH, Sachs GS, Rosenbaum JF. Hypochondriacal concerns, anxiety sensitivity, and panic disorder. J Anxiety Disord 1992;6:93 ± 104. Parker G, Lipscombe P. The relevance of early parental experiences to adult dependency, hypochondriasis and utilization of primary physicians. Br J Med Psychol 1980;53:355 ± 63. Peterson RA, Reiss S. Anxiety Sensitivity Index Manual, 2nd edn. Worthington, OH: International Diagnostic Services, 1992 Revised. Pilowsky I. Dimensions of hypochondriasis. Br J Psychiatry 1967; 113:89 ± 93. Pilowsky I, Spence ND. Patterns of illness behavior in patients with intractable pain. J Psychosom Res 1976;19:279 ± 87. Reiss S. Expectancy model of fear, anxiety, and panic. Clin Psychol Rev 1991;11:141 ± 53. Reiss S, McNally RJ. Expectancy model of fear. In: Reiss S, Bootzin RR, editors. Theoretical Issues in Behavior Therapy. New York: Academic Press, 1985. pp. 107 ± 21. Salkovskis PM, Clark DM. Panic disorder and hypochondriasis. Adv Behav Res Ther 1993;15:23 ± 48. Schmidt NB, Lerew DR, Jackson RJ. The role of anxiety sensitivity in the pathogenesis of panic: prospective evaluation of spontaneous panic attacks during acute stress. J Abnorm Psychol 1997;106:355 ± 64. Schwartz SM, Gramling SE, Mancini T. The influence of life stress, 118 [35] [36] [37] [38] [39] [40] [41] M.C. Watt, S.H. Stewart / Journal of Psychosomatic Research 49 (2000) 107±118 personality, and learning history on illness behavior. J Behav Ther Exp Psychiatry 1994;25:135 ± 42. Speckens AE, Spinhoven P, Sloekers PPA, Bolk JH, van Hemert AM. A validation study of the Whitley index, the Illness Attitude Scales, and the Somatosensory Amplification Scale in general medical and general practice patients. J Psychosom Res 1996;40: 95 ± 104. Spielberger C, Gorsuch R, Lushene R, Vagg P, Jacobs G. Manual for the State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press, 1983. Stein MB, Jang KL, Livesley WJ. Heritability of anxiety sensitivity: a twin study. Am J Psychiatry 1999;156:246 ± 51. Stewart SH, Knize K, Pihl RO. Anxiety sensitivity and dependency in clinical and non-clinical panickers and controls. J Anxiety Disord 1992;6:119 ± 31. Stewart SH, Taylor S, Baker JM. Gender differences in dimensions of anxiety sensitivity. J Anxiety Disord 1997;11:179 ± 200. Stewart SH, Taylor S, Jang KL, Cox BJ, Watt MC, Fedoroff IC, Borger SC. Causal modeling of relations among learning history, anxiety sensitivity, and panic attacks. Behav Res Ther (in press). Stewart SH, Watt MC. Illness attitude scale dimensions and their [42] [43] [44] [45] [46] [47] [48] associations with anxiety-related constructs in a non-clinical sample. Behav Res Ther (in press). Stewart SH, Watt MC. Assessment of health anxiety. In: Asmundson GJG, Taylor S, Cox BJ, editors. Health Anxiety: Clinical and Research Perspectives on Hypochondriasis and Related Disorders. Sussex, UK: Wiley, Sussex, UK (in press). Taylor S. Comment on Otto et al. (1992): hypochondriacal concerns, anxiety sensitivity, and panic disorder. J Anxiety Disord 1994;8:97 ± 9. Taylor S. Panic disorder and hypochondriacal concerns: reply to Otto and Pollack (1994). J Anxiety Disord 1995;9:87 ± 8. Taylor S, Koch WJ, McNally RJ. How does anxiety sensitivity vary across the anxiety disorders? J Anxiety Disord 1992;7:249 ± 59. Warwick HM, Salkovskis PM. Hypochondriasis. Behav Res Ther 1990;28:105 ± 17. Watt MC, Stewart SH, Cox BJ. A retrospective study of the learning history origins of anxiety sensitivity. Behav Res Ther 1998;36: 505 ± 25. Whitehead WE, Busch CM, Heller BR, Costa PT. Social learning influences on menstrual symptoms and illness behavior. Health Psychol 1986;5:13 ± 23.