Implicit associations between anxiety-related symptoms and catastrophic consequences in high anxiety sensitive individuals

Page created by Brian Pope
 
CONTINUE READING
COGNITION AND EMOTION
2006, 20 (2), 295±308

     Implicit associations between anxiety-related
   symptoms and catastrophic consequences in high
              anxiety sensitive individuals
                                 Marie-joseÂe Lefaivre
                Dalhousie University, Halifax, Nova Scotia, Canada
                                     Margo C. Watt
 Dalhousie University, Halifax and St. Francis Xavier University, Antigonish,
                            Nova Scotia, Canada
                   Sherry H. Stewart and Kristi D. Wright
                     Dalhousie University, Nova Scotia, Canada

   Anxiety sensitivity refers to the fear of anxiety-related physical sensations arising
   from beliefs that these sensations have harmful consequences (Reiss & McNally,
   1985). The present study examined whether individuals with high (vs. low) anxiety
   sensitivity show stronger implicit associations in memory between anxiety-related
   symptoms, as opposed to neutral body parts, and harmful, as compared to harm-
   less, consequences. A total of 22 undergraduate students (14 F, 8 M) completed the
   Extrinsic Affective Simon Task (EAST; De Houwer, 2003). Results indicated that
   high anxiety sensitive individuals (n = 10) tended to implicitly associate harmful
   consequences with anxiety-related symptoms. Their performance was significantly
   faster on trials where target words related to anxiety symptoms were mapped on to
   the same response key as harmful consequences. No significant difference in
   performance was found for low anxiety sensitive individuals (n = 12) or when
   target words were body parts unlikely related to diseases. Between-group differ-
   ences persisted after controlling for trait anxiety and history of panic attacks, but
   not when illness-related beliefs were introduced as a covariate. Identifying this
   implicit association bias provides additional empirical support for the concept of
   anxiety sensitivity.

    Correspondence should be addressed to Dr Margo C. Watt, Department of Psychology, St.
Francis Xavier University, P.O. Box 5000, Antigonish, Nova Scotia B2G 2W5, Canada; e-mail:
mwatt@stfx.ca
    The authors would like to thank members of Dr Ray Klein's laboratory at Dalhousie University
(particularly Patti Devlin) for their assistance with the computer programming and Tara MacDonald
at St. Francis Xavier University for her assistance with data collection.

                               # 2006 Psychology Press Ltd
http://www.psypress.com/cogemotion                        DOI:10.1080/02699930500336466
296     LEFAIVRE ET AL.

Anxiety sensitivity is an individual difference factor that refers to the fear of
anxiety-related physical sensations. This fear arises from beliefs that anxiety-
related sensations have harmful somatic, psychological, or social consequences
(Reiss & McNally, 1985). Whereas individuals with low anxiety sensitivity
recognise that bodily sensations associated with anxiety are unpleasant but
transient and harmless, individuals with high anxiety sensitivity fear these
symptoms may have dire consequences (McNally, 1999a). Anxiety sensitivity,
as measured by the Anxiety Sensitivity Index (ASI; Peterson & Reiss, 1992), has
demonstrated incremental validity above and beyond measures of trait anxiety in
discriminating between panic disorder and generalised anxiety disorder (e.g.,
Taylor, Koch, & McNally, 1992), as well as in predicting fear and panic
(Peterson & Reiss, 1992). Anxiety sensitivity has been shown to be an important
risk factor for the development and maintenance of panic disorder (see review
by Cox, Borger, & Enns, 1999), and hypochondriasis (Otto, Demopulos,
McLean, Pollack, & Fava, 1998; Stewart & Watt, 2000).
   Information-processing models consider anxiety disorders to result from
maladaptive schemas in memory misguiding information processing (Teachman
& Woody, 2003; for a review, see Coles & Heimberg, 2002). These models
predict that anxiety-congruent information will be subject to attentional, pro-
cessing, and retrieval biases, which contribute to the development and main-
tenance of anxiety disorders (Foa & Kozak, 1986; for a review, see Coles &
Heimberg, 2002). For example, Beck and Clark (1997) propose a three-stage
model to explain the relative roles of explicit (conscious) and implicit (auto-
matic) information-processing in the etiology of pathological anxiety. Whereas
most research has focused on the third stage of Beck and Clark's model (i.e.,
self-reported cognitions), more recently, researchers have started to explore the
earlier stages; stages not accessible to conscious processing (see McNally,
1995).
   In order to examine the more automatic and implicit stages of information-
processing, researchers generally must employ tasks that measure, in milli-
seconds, automatic actions, or judgements that do not, by definition, require
conscious awareness on the participant's part (see Egloff & Schmukle, 2002;
McNally, 2001). These types of tasks provide complementary information to
explicit research designs (e.g., self-report, free-recall test) interested in the more
reflective and conscious processing stage of Beck and Clark's (1997) model by
providing insight on earlier information-processing stages.
   Studies employing implicit cognition tasks have demonstrated that, as
opposed to controls, clinically diagnosed anxiety disorder patients have a ten-
dency to show selective processing biases favouring information relevant to
their anxiety-related concerns (see reviews in Coles & Heimberg, 2002;
McNally, 1999b; Williams, Mathews, & MacLeod, 1996). Other studies have
examined the relationship between anxiety sensitivity and information-
processing biases before the onset of anxiety problems to determine if anxiety
ANXIETY SENSITIVITY AND IMPLICIT BIAS         297

sensitivity could be considered a cognitive risk factor for anxiety disorders
(Keogh, Dillon, Georgiou, & Hunt, 2001; Lees, Mogg, & Bradley, 2005;
McCabe, 1999; McNally, Horning, Hoffman, & Han, 1999; Stewart, Conrod,
Gignac, & Pihl, 1998). It remains unclear whether cognitive biases precede or
follow the development of an anxiety disorder given the limited amount of
research on nonclinical samples to date (see McNally et al., 1999).
   The above studies have varied by sample and methodology. Moreover, none
have used association-type tasks thereby preventing the examination of auto-
matic associative processes as a function of anxiety sensitivity levels. By defi-
nition, high anxiety sensitive individuals should be differentiated from low
anxiety sensitive individuals not simply by their fear of anxiety-related sensa-
tions but also by their beliefs about the catastrophic consequences of experi-
encing these sensations. Consequently, the objective of the present study was to
extend previous findings by examining whether high anxiety sensitive indivi-
duals tend to catastrophise the consequences of their physical symptoms. More
precisely, we were interested in whether there is a stronger implicit association
(vs. dissociation) between physical symptoms and harmful outcomes (vs.
harmless outcomes) concepts in memory for high anxiety sensitive individuals
as compared to low anxiety sensitive individuals. Identifying the presence of an
implicit association bias between anxiety-related sensations and negative out-
comes among high anxiety sensitive individuals using a novel research task
could provide additional empirical support for the concept of anxiety sensitivity,
as well as explicating possible cognitive mechanisms involved in high anxiety
sensitive individuals' risk for panic disorder and hypochondriasis.
   The present study employed the Extrinsic Affective Simon Task (EAST)
recently developed by De Houwer (2003). The EAST is a modified version of
the popular Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz,
1998) that measures participants' reaction times (RTs) when categorising stimuli
by pressing either a left or a right computer key. The underlying assumption that
forms the basis of tasks, such as the EAST or IAT, is that a relationship exists
between cognitive structures and automaticity (see Teachman & Woody, 2003).
The EAST is therefore seen as a measure of cognitive structure or automatic
associations in memory based on RTs. The EAST was chosen for the present
study because it circumvents some of the problems associated with the IAT (see
De Houwer, 2003). In the present study, individuals were asked to categorise
body parts (e.g., abdomen, elbow) and anxiety-related sensations (e.g., breath-
less, dizzy) based on their colour, as well as to categorise positive (e.g., benign,
healthy) and negative (e.g., faint, stroke) health outcomes, presented in white,
based on their valence (see details in the Method section). As suggested by
Palfai and Ostafin (2003), we selected words representing body parts unlikely to
be associated with negative physical outcomes (e.g., diseases) as control targets,
because similar evaluative responses would be expected across individuals
therefore providing, hopefully, a stable contrast.
298     LEFAIVRE ET AL.

   It was hypothesised that, if a tendency to catastrophise anxiety-related phy-
sical sensations existed among the high anxiety sensitive participants, their
performance would be superior on trials where they needed to select an
extrinsically negative (vs. positive) response key for coloured target words that
were anxiety-related sensations. It was also expected that this performance
facilitation for targets by pairing with the extrinsically negative response key
would be specific to high anxiety sensitive (as opposed to low anxiety sensitive)
individuals, and to trials involving anxiety-related sensations (as opposed to
body parts) as target items.

                                    METHOD
Participants

The participants were 22 (14 F, 8 M; mean age = 18.5, SD = 0.80) undergraduate
students from St. Francis Xavier University who received course credit for
participating in the study. Participants were randomly selected and assigned to
groups according to their scores on the Anxiety Sensitivity Index (ASI; Peterson
& Reiss, 1992) administered during a mass in-class screening. The high and low
anxiety sensitivity groups were composed of participants scoring at least one
standard deviation above (> 17.9 + 8.7) or below (< 17.978.7) the ASI mass
screening sample mean, respectively (see Peterson & Reiss, 1992; Watt,
Stewart, & Cox, 1998).
    The results of independent-sample t-tests and chi-square analyses showed
that the two groups did not significantly differ in terms of age, gender ratio, level
of education in years, or number of participants who had ever experienced a
spontaneous panic attack but they did differ significantly in trait anxiety as
assessed by the State-Trait Anxiety Inventory-Trait Subscale (Spielberger,
Gorsuch, Lushene, Vagg, & Jacobs, 1983) and health anxiety as assessed by the
Illness Attitudes Scale (Kellner, 1987). The descriptive statistics for both groups
are presented in Table 1.
    Participants had normal or corrected-to-normal vision. Students who were
colour-blind and/or students whose language was not English were excluded
from the study given that the EAST is a cognitive task that requires the rapid
distinction between two similar colours (i.e., green and blue) and the rapid
categorisation of English words based on their semantic meaning.

Measures

   Demographic information. Age, gender, and level of education of the
participants were collected with a researcher-compiled demographic information
questionnaire.
ANXIETY SENSITIVITY AND IMPLICIT BIAS                 299
                                    TABLE 1
       Means (and standard deviations) comparing the two anxiety sensitivity
                                     groups

                                                      LAS                 HAS
                                                    (n = 12)            (n = 10)

      Age                                          18.42 (0.9)         18.6 (0.7)
      Gender                                         5 M; 7 F           3 M; 7 F
      Level of education                            14.5 (1.0)         14.2 (0.4)
      ASI-Total                                     7.0 (1.8)         32.9 (2.8)***
      STAI-T                                       36.5 (8.6)         46.0 (10.7)*
      IAS-Total                                    28.9 (12.6)        43.3 (15.4)*
      PAQ (% spontaneous panic reporters)            2 (17%)            2 (20%)

         LAS, low anxiety sensitive; HAS, high anxiety sensitive; STAI-T, State Trait
      Anxiety Inventory-Trait Scale; ASI, Anxiety Sensitivity Index; PAQ, Panic Attack
      Questionnaire; IAS, Illness Attitudes Scale. * p < .05; ** p < .01; *** p < .001.

   Anxiety Sensitivity Index (ASI; Peterson & Reiss, 1992). Using 5-point
Likert-type scales, the ASI assesses the respondent's level of agreement/
disagreement with 16 statements relating to the beliefs that anxiety sensations
are associated with physical, social, or psychological consequences. Psycho-
metric information concerning the excellent reliability and validity of the
measure for clinical and nonclinical populations is available in Peterson and
Reiss (1992).

   Illness Attitudes Scale (IAS; Kellner, 1987). This 29-item self-report scale
is divided into nine subscales that assess fears, behaviours, beliefs, and effects
associated with hypochondriasis (see Stewart & Watt, 2000), as well as
psychopathology associated with abnormal illness behaviours. The IAS has been
shown to have good test-retest reliability (see Fava, Kellner, Zielzny, & Grandi,
1988; Kellner, 1987), acceptable internal reliability (see Stewart & Watt, 2000;
Wise & Sheridan, 2001), good concurrent validity (Kellner, Abbot, Winslow, &
Pathak, 1987), and convergent validity (Speckens, Spinhoven, Sloekers, Bolk, &
van Hemert, 1996).

   Panic Attack Questionnaire±Revised (PAQ-R; Cox, Norton, & Swinson,
1992). The PAQ-R provides a description of a panic attack as defined in the
Diagnostic and Statistical Manual of Mental Disorders, 3rd edition-Revised
(American Psychiatric Association, 1987) and asks participants to indicate
whether they have ever experienced such an attack. The experience of at least
one spontaneous (``out of the blue'') panic attack was used to assess panic
history.
300     LEFAIVRE ET AL.

   State-Trait Anxiety Inventory-Trait Subscale (STAI-T; Spielberger et al.,
1983). This self-report subscale assesses participants' general tendency to
react anxiously to potentially threatening stimuli. Participants are asked to rate
how they generally feel about 20 statements using 4-point Likert-type scales.
The STAI manual (Spielberger et al., 1983) provides information about the good
psychometric proprieties of this scale.

   EAST Stimulus Materials. Five anxiety-related symptoms (i.e., breathless,
dizzy, numbness, palpitation, shaky) were selected from word lists previously
used in anxiety sensitivity studies (McCabe, 1999; McNally et al., 1999). Five
negative health outcomes matching these particular symptoms (i.e., suffocate,
faint, stroke, heart attack, seizure) were then selected in order to reflect the
anticipated negative consequences of experiencing such symptoms. Subse-
quently, five body parts identified as unlikely related to diseases (i.e., abdomen,
ankle, elbow, finger, nose) and five positive health outcomes (e.g., benign,
healthy, innocuous, resilient, thriving) were also selected to complete the four
stimuli categories for the EAST. The results of two one-way analyses of
variance (ANOVA) indicated that the four word groups did not differ in average
word length, F(3, 16) = 1.661, ns, and average word frequency, F(3, 16) = 0.828,
ns. The average word frequency was calculated using a popular Internet search
engine (www.google.com; see Blair, Urland, & Ma, 2003).

   EAST apparatus. The EAST (De Houwer, 2003) was conducted using the
SuperLab software programmed on an iMac computer running in OS 9.1.
Participants were seated in front of the computer at a distance of approximately
40 cm from the 14-inch screen. The words' letters were around 7 mm high and 5
mm wide and were presented on a black background. The descriptors were
presented in white using the default Adobe Photoshop values. The target words
were presented in either blue or green. The red, blue, and green settings used to
create the blue (0,117,97) and the green (0,97,117) produced two colours that
were quite similar. Participants used either the ``A'' key or ``L'' key of the
(QWERTY) keyboard to categorise the stimuli. Participants' response times,
defined as the time between the onset of a word and the first key press, were
measured using the highly accurate (< 1 ms) Timer-1 included in the SuperLab
software.

Procedure
After reading and signing the Consent Form, the participants completed the
EAST and the questionnaires individually. The EAST included several trials
during which participants were presented with words one by one, and were
asked to classify these words based on either their meaning or colour. Partici-
pants had to categorise white words (i.e., descriptors) according to the word's
ANXIETY SENSITIVITY AND IMPLICIT BIAS          301

valence (i.e., positive or negative). Participants were told to press the GOOD key
when words referred to something positive (e.g., healthy, innocuous) or the BAD
key if the word was something negative (e.g., stroke, faint). Participants also had
to categorise coloured words (i.e., targets) according to their colour (i.e., blue vs.
green) regardless of the valence. The target words were either anxiety-related
symptoms or body parts unlikely to be related to disease. On the trials of the test
blocks, one descriptor and one colour were mapped on to one response key (e.g.,
A = positive and blue) and the other descriptor and the other colour were
mapped on to the other response key (e.g., L = negative and green). Conse-
quently, the targets become extrinsically related to the descriptors' valence
because of the task instructions (see De Houwer, 2003; see also Appendix). The
key mapping assignments for the descriptors and the targets were manipulated in
an orthogonal way to control for possible location-effect between participants.
The participants were told to respond as quickly as possible without making any
errors. A red cross appearing on the screen informed participants when they had
made an incorrect response.
   The EAST took about 15 minutes to complete and included two practice
blocks of 20 trials followed by four test blocks of 30 trials. Participants were
presented the same stimuli during the practice and test blocks (see De Houwer,
2003). The first practice block consisted of white words only during which each
10 stimuli were presented twice in a random order. The random presentation of
the 10 coloured words, once in each colour, was the second practice block. The
four test blocks of 30 trials followed during which each of the 10 target words
were presented once in each colour and each of the 10 descriptor words were
presented once in white. The words were all presented in random order with the
restriction that the same word could not be presented on two or more con-
secutive trials and that the required response could not be the same on four or
more consecutive trials. At each trial, the presentation of the word was preceded
by a white fixation symbol of 350 ms. The word remained on the screen until the
participants pressed a response key. The intertrial interval was 250 ms. Prior to
starting each block, participants were presented with instructions reminding
them what key to press for which type of stimuli.

                                    RESULTS
Results of the evaluation of assumptions indicated a violation of normality,
which required log-transformation of data, but satisfactory homogeneity of
variance-covariance matrices. The analyses were performed using the mean log-
transformed RTs of the test trials where a correct response was made for
coloured words only. The RTs of trials with an incorrect response were excluded
from the analyses as were trials with a RT below 300 ms or above 3000 ms (see
Brendl, Markman, & Messner, 2001). The mean log-transformed RT was cal-
culated for the four types of trials: (1) body parts extrinsically mapped on to a
302     LEFAIVRE ET AL.

                                   TABLE 2
      Log-transformed means (and standards deviations) of the trial types for
                        both anxiety sensitivity groups

      Trial type                                        LAS                  HAS
                                                      (n = 12)             (n = 10)

      Body part + Positive Outcome                  2.829   (0.112)     2.847   (0.060)
      Body part + Negative Outcome                  2.924   (0.091)     2.934   (0.088)
      Symptom + Positive Outcome                    2.827   (0.102)     2.887   (0.108)a
      Symptom + Negative Outcome                    2.839   (0.080)     2.857   (0.129)a

      LAS, low anxiety sensitive; HAS, high anxiety sensitive.
      a
        Significant difference between these two trial types for the HAS group at p < .05.

positive health outcome response key; (2) body parts extrinsically mapped on to
a negative health outcome response key; (3) anxiety-related symptoms extrin-
sically mapped on to a positive health outcome response key; (4) anxiety-related
symptoms extrinsically mapped on to a negative health outcome response key.
The overall untransformed mean of these fours trial types for both anxiety
sensitivity groups combined was 760.4 (SD = 200.2). The log-transformed
means of the four trial types for each anxiety sensitivity group are presented in
Table 2.
    An independent-samples t-test was conducted to evaluate the usefulness of
the selected control categories (i.e., body parts identified as unlikely to be
related to diseases and positive health outcomes). Results indicated no sig-
nificant difference in RT between anxiety sensitivity groups when target
words referring to body parts were mapped on to the same response key as
the white words referring to positive health outcome, t(20) = 70.47, ns. The
absence of a significant difference suggested that these control categories pro-
vided a stable contrast from which a comparison could be made with the
trials of main interest (i.e., trials wherein the participants needed to select
extrinsically negative responses for coloured words that were anxiety-related
symptoms).
    The error rates and deleted trials represented, on average, 5.8% and 1.3% of
all trials, respectively. Two 2 6 2 6 2 (anxiety sensitivity group 6 target
words 6 health outcome responses) ANOVA with repeated measures was
conducted on the error rate and deleted trials data. No significant effects were
revealed. The internal consistency of the EAST was assessed using 40 RT
differences for targets paired with negative vs. positive outcomes. Error rates
and deleted trials were treated as missing values and were replaced prior to alpha
calculation. Cronbach's alpha indicated adequate reliability of .63.
    A 2 6 2 6 2 ANOVA with repeated measures was conducted to evaluate
the effects of the anxiety sensitivity groups (high vs. low), target words (body
ANXIETY SENSITIVITY AND IMPLICIT BIAS         303

parts vs. anxiety symptoms), and extrinsic health outcome responses (positive
vs. negative) on RTs. Results of the ANOVA indicated a significant three-
way interaction, F(1, 20) = 10.07, p < .005, partial Z2 = .34. Consequently,
two 2 6 2 ANOVAs were conducted to evaluate the effects of target words
and extrinsic health outcome responses on RT for each anxiety sensitivity
group. Results of the ANOVA for the low anxiety sensitivity group found no
significant main effects for target words, F(1, 11) = 0.38, ns, or health out-
comes, F(1, 11) = 0.03, ns, nor a significant interaction, F(1, 11) = 1.69, ns.
These results suggested that participants in the low anxiety sensitivity group
did not differ significantly in their RTs whether the target words (i.e., body
parts vs. anxiety symptoms) were extrinsically mapped on to the same
response key as positive or negative health outcomes. The results of the
ANOVA for the high anxiety sensitivity group found no significant main
effects for target words, F(1, 9) = 1.55, ns, or health outcomes, F(1, 9) = 1.07,
ns. However, the results indicated a significant target words 6 health out-
come interaction, F(1, 9) = 20.50, p < .001, partial Z2 = .70.
   To further examine this significant two-way interaction for the high anxiety
sensitivity group, post hoc analyses were performed. Given that the specific
comparison of interest was the performance of high anxiety sensitive indivi-
duals on trials where they needed to select an extrinsically negative (vs. posi-
tive) response for words that were anxiety-related (vs. body parts), paired t-
tests with Dunnett's correction was used (Tabachnick & Fidell, 2001). This
approach revealed that high anxiety sensitive participants responded sig-
nificantly faster when anxiety-related symptoms were extrinsically mapped
onto the same response key as negative health outcomes vs. positive outcomes,
t(9) = 2.28, p < .05. There were no significant differences in participants'
responding to the pairings of body parts with either negative or positive out-
comes. These findings support the contention that there exists a stronger
implicit association (vs. dissociation) between anxiety-related physical symp-
toms and harmful (vs. harmless) concepts among high anxiety sensitive
individuals.
   The three-way interaction, F(1, 19) = 4.89, p < .05, and the contrast of
interest, t(9) = 1.86, p < .05 remained significant when trait anxiety as assessed
by the STAI-T (Spielberger et al., 1983) was entered as a covariate. When
history of panic attacks (no spontaneous panic attack vs. at least one sponta-
neous panic attack) was entered as a covariate, the three-way interaction,
F(1, 19) = 9.53, p < .01 and the contrast of interest, t(9) = 2.22, p < .05, also
remained significant. The three-way interaction remains significant when
hypochondriacal concerns, as assessed by the IAS (Kellner, 1987), was entered
as a covariate, F(1, 19) = 7.09, p < .05, but the contrast of interest becomes only
marginally significant, t(9) = 1.78, p < .10. The three-way interaction, however,
is only marginally significant when the illness-related beliefs factor is intro-
duced by itself as the covariate, F(1, 19) = 4.248, p = .053.
304     LEFAIVRE ET AL.

                                 DISCUSSION

The present experiment was designed to investigate whether high anxiety sen-
sitive individuals tend to catastrophise the consequences of anxiety-related
symptoms based on cognitive implicit associations. This was examined with the
EAST (De Houwer, 2003), an implicit measure of cognitive structure or auto-
matic associations in memory based on RTs. Participants in the present study
were asked to categorise white descriptor words (i.e., harmless and harmful
health outcomes) based on their valence (i.e., positive vs. negative), as well as to
categorise coloured target words (i.e., body parts unlikely related to diseases and
anxiety-related symptoms) based on their colour (i.e., blue or green) while
ignoring their valence. Consistent with information-processing models of
anxiety (e.g., Beck & Clark, 1997), results indicated that high (vs. low) anxiety
sensitive individuals tend to implicitly associate anxiety-related symptoms with
negative consequences. High anxiety sensitive individuals responded sig-
nificantly faster on trials where anxiety-related target words were mapped on to
the same response key as harmful (vs. harmless) health outcomes. High anxiety
sensitive individuals did not show the same tendency for associating body parts
with negative outcomes on the EAST and none of the RTs for the different
pairings of the target words and the health outcomes were significantly faster for
the low anxiety sensitivity group.
    The between-group differences persisted after controlling for trait anxiety and
history of panic attacks; however, results suggested that illness-related beliefs
may be contributing to this implicit bias of associating anxiety-related symptoms
with negative outcomes in high anxiety sensitive individuals. This could be due
to the fact that illness-related beliefs factor of the IAS (Kellner, 1987) is the
more closely linked to the DSM-IV's (American Psychiatric Association, 2001)
concept of hypochondriasis (see Stewart & Watt, 2000). This factor is also better
at discriminating between parents with hypochondriasis diagnosis vs. controls
than other IAS factors (see Kellner et al., 1987).
    Based on the underlying assumption of the EAST, these results suggest that
the performance of high AS individuals on trials where they needed to make an
negative extrinsic response choice for anxiety-related symptoms (i.e., anxiety-
related target words were mapped on to the same response key as harmful
outcomes) was superior as compared to other trials because this response choice
corresponded more closely to their associations in memory. Thus, anxiety
sensitivity could be a cognitive risk factor for anxiety disorders (see Keogh et
al., 2001; McCabe, 1999; McNally et al., 1999; Stewart et al., 1998).
    Identifying this implicit association bias provides additional empirical sup-
port for the concept of anxiety sensitivity. Indeed, the use of the EAST in the
present study extends the work of previous studies investigating cognitive biases
of threat in nonclinical populations (e.g., Keogh et al., 2001; McCabe, 1999;
McNally et al., 1999; Stewart et al., 1998) by distinguishing between the
ANXIETY SENSITIVITY AND IMPLICIT BIAS         305

anxiety-related symptoms (e.g., breathless) and their possible consequences
(e.g., suffocate). This distinction is important because the results of the present
study suggest that, consistent with the anxiety sensitivity definition, high anxiety
sensitive individuals can be differentiated from low anxiety sensitive persons not
simply by their fear of anxiety-related symptoms but also their beliefs about the
consequences of experiencing such symptoms. This study, however, only
investigated the relationship between anxiety-related symptoms and their pos-
sible physical consequences. It did not examine whether this tendency to cata-
strophise the consequences of anxiety-related sensations would be observed if
participants had also been presented with possible psychological and/or social
consequences of experiencing those symptoms. Consequently, the results might
only be specific to physical concerns associated with high anxiety sensitivity and
should not be generalised to psychological and social concerns associated with
anxiety-related symptoms. It is also possible that the significant results are
simply an artefact of the small sample, although this seems unlikely given that a
predicted three-way interaction emerging by chance is highly unlikely. In any
event, these results require replication with an independent sample to ensure that
they are reliable.
   The present study is considered a necessary first step in addressing a related
research question as to whether it is possible to differentiate between high
anxiety sensitivity individuals and individuals with strong hyponchondriacal
tendencies on the basis of their implicit associations between arousal-reactive
(vs. arousal-nonreactive) symptoms and immediate (vs. delayed) consequences
(e.g., Stewart & Watt, 2000). In this case, it could be hypothesised that stronger
implicit associations between arousal-reactive symptoms (e.g., heart palpita-
tions) and immediate consequences (e.g., heart attack) would be found in high
anxiety sensitive individuals, whereas individuals with more hyponchondriacal
tendencies would also show stronger implicit associations between arousal-
nonreactive symptoms (e.g., lump) and delayed consequences (e.g., cancer).
   The sample size of the present study did not allow for the direct comparison
of gender differences in anxiety-related implicit associations. However, previous
studies suggest that females, compared to males, tend to score higher on the
physical concerns factor of the ASI (e.g., Stewart, Taylor, & Baker, 1997).
Another study by Stewart et al. (1998) found that, when completing a modified-
Stroop colour-naming task, high anxiety sensitive women demonstrated more
interference with physical threat cues as compared to their low anxiety sensitive
counterparts, whereas high anxiety sensitive men displayed more interference
for psychological/social threat as compared to low anxiety sensitive men.
Moreover, samples employed in previous studies (e.g., Keogh et al., 2001;
McNally et al., 1999) have been predominantly female. Consequently, future
research should consider investigating whether high anxiety sensitive women
would show a stronger implicit association between arousal-reactive symptoms
and negative health outcomes than high anxiety sensitive men. Future research
306       LEFAIVRE ET AL.

also should consider including a measure of depression given the relationship
between anxiety sensitivity and depression (see Taylor, Koch, & Woody, 1996)
and the evidence that depressed individuals tend to show a memory bias for
negative material (Ruiz-Caballero & GonzaÂlez, 1997).
   In conclusion, the findings of the present study might eventually lead to
important clinical implications. Because the cognitive biases in high anxiety
sensitive individuals appear to be operating at an automatic level (i.e., not under
intentional control), treatment approaches could benefit from consulting studies
in the field of social cognition (e.g., Dasgupta & Greenwald, 2001) and the
clinical realm (e.g., Teachman & Woody, 2003) that have demonstrated that
automatic associations are sensitive to intervention. Similarly, it has been sug-
gested that measures of automatic associations could be useful intervention tools
(Teachman & Woody, 2004). Finally measures, such as the EAST, might also be
useful as outcome assessment tools to assess the efficacy of treatment in altering
implicit associations in memory (e.g., Teachman & Woody, 2003).

                                                             Manuscript received 1 December 2004
                                                       Revised manuscript received 17 August 2005

                                        REFERENCES
American Psychiatric Association. (1987). Diagnostic and Statistical manual of mental disorders:
   DSM-III-R (3rd rev. ed.). Washington, DC: Author.
Beck, A. T., & Clark, D. A. (1997). An information processing model of anxiety: Automatic and
   strategic processes. Behaviour Research and Therapy, 35, 49±58.
Blair, I. V., Urland, G. R., & Ma, J. E. (2003). Using internet search engines to estimate word
   frequency. Behavior Research Methods, Instruments, and Computers, 34, 286±290.
Brendl, C. M., Markman, A. B., & Messner, C. (2001). How do indirect measures of evaluation
   work? Evaluating the influence of prejudice in the Implicit Association Test. Journal of Per-
   sonality and Social Psychology, 81, 760±773.
Coles, M. E., & Heimberg, R. G. (2002). Memory biases in the anxiety disorders: Current status.
   Clinical Psychology Review, 22, 587±627.
Cox, B. J., Borger, S. C., & Enns, M. W. (1999). Anxiety sensitivity and emotional disorders:
   Psychometric studies and their theoretical implications. In S. Taylor (Ed.), Anxiety sensitivity:
   Theory, research and treatment of the fear of anxiety (pp.115±148). Mahwah, NJ: Erlbaum.
Cox, B. J., Norton, G. R, & Swinton, R.P. (1992). The panic attack questionnaireÐRevised.
   Unpublished scale, Centre for Addiction and Mental Health, Toronto, Canada.
Dasgupta, N., & Greenwald, A. G. (2001). On the malleability of automatic attitudes: Combating
   automatic prejudice with images of admired and disliked individuals. Journal of Personality and
   Social Psychology, 81, 800±814.
De Houwer, J. (2003). The extrinsic affective simon task. Experimental Psychology, 50, 77±85.
Egloff, B., & Schmukle, S. C. (2002). Predictive validity of an implicit association test for assessing
   anxiety. Journal of Personality and Social Psychology, 83, 1441±1455.
Fava, G. A., Kellner, R., Zielezny, M., & Grandi, S. (1988). Hypochondriacal fears and beliefs in
   agoraphobia. Journal of Affective Disorders, 14, 239±244.
Foa, E. B., & Kozak, M. J. (1986). Emotional processing of fear: Exposure to corrective information.
   Psychology Bulletin, 99, 20±35.
ANXIETY SENSITIVITY AND IMPLICIT BIAS                   307
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual differences in
    implicit cognition: The implicit association test. Journal of Personality and Social Psychology,
    74, 1464±1480.
Kellner, R. (1987). Abridged manual of the illness attitudes scale. Department of Psychiatry, School
    of Medicine, University of New Mexico, USA.
Kellner, R., Abbott, P., Winslow, W. W., & Pathak, D. (1987). Fears, beliefs, and attitudes in DSM-
    III hypochondriasis. Journal of Nervous and Mental Disease, 175, 20±25.
Keogh, E., Dillon, C., Georgiou, G., & Hunt, C. (2001). Selective attentional biases for physical
    threat in physical anxiety sensitivity. Anxiety Disorders, 15, 299±315.
Lees, A., Mogg, K., & Bradley, B.P. (2005). Health anxiety, anxiety sensitivity and attentional biases
    for pictorial and linguistic health-threat cues. Cognition and Emotion, 19, 453±462.
McCabe, R. E. (1999). Implicit and explicit memory for threat words in high- and low-anxiety-
    sensitive participants. Cognitive Therapy and Research, 23, 21±38.
McNally, R. J. (1995). Automaticity and the anxiety disorders. Behaviour Research and Therapy, 33,
    747±754.
McNally, R. J. (1999a). Theoretical approaches to the fear of anxiety. In S. Taylor (Ed.), Anxiety
    sensitivity: Theory, research and treatment of the fear of anxiety (pp. 3±16). Mahwah, NJ:
    Erlbaum.
McNally, R. J. (1999b). Anxiety sensitivity and information-processing biases for threat. In S. Taylor
    (Ed.), Anxiety sensitivity: Theory, research and treatment of the fear of anxiety (pp. 183±197).
    Mahwah, NJ: Erlbaum.
McNally, R. J. (2001). On the scientific status of cognitive appraisal models of anxiety disorder.
    Behaviour Research and Therapy, 39, 513±512.
McNally, R. J., Horning, C. D., Hoffman, E. C., & Han, E. M. (1999). Anxiety sensitivity and
    cognitive biases of threat. Behavior Therapy, 30, 51±61.
Otto, M. W., Demopulos, C. M., McLean, N. E., Pollack, M. H., & Fava, M. (1998). Additional
    findings on the association between anxiety sensitivity and hypochondriacal concerns: Exam-
    ination of patients with major depression. Journal of Anxiety Disorders, 12, 225±232.
Palfai, T. P., & Ostafin, B. D. (2003). Alcohol-related motivational tendencies in hazardous drinkers:
    Assessing implicit response tendencies using the modified-IAT. Behaviour Research and
    Therapy, 41, 1149±1162.
Peterson, R. A., & Reiss, S. (1992). Anxiety sensitivity index manual (2nd ed.). Worthington, OH:
    International Diagnostic Systems.
Reiss, S., & McNally, R. J. (1985). The expectancy model of fear. In S. Reiss & R. R. Bootzin (Eds.),
    Theoretical issues in behavior therapy (pp. 107±122). New York: Academic Press.
Ruiz-Caballero, J. A., & GonzaÂlez, P. (1997). Effects of level of processing on implicit and explicit
    memory in depressed mood. Motivation and Emotion, 21, 195±209.
Speckens, A. E., Spinhoven, P., Sloekers, P. P. A., Bolk, J. H., & van Hemert, A. M. (1996). A
    validation study of the Whitley index, the Illness Attitude Scales, and the Somatosensory
    Amplification Scale in general medical and general practice patients. Journal of Psychosomatic
    Research, 40, 95±104.
Spielberger, C., Gorsuch, R., Lushene, R., Vagg, P., & Jacobs, G. (1983). Manual for the State-Trait
    Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press.
Stewart, S. H., Conrod, P. J., Gignac, M. L., & Pihl, R. O. (1998). Selective processing biases in
    anxiety-sensitive men and women. Cognition and Emotion, 12, 105±133.
Stewart, S. H., Taylor, S., & Baker, J. M. (1997). Gender differences in dimensions of anxiety
    sensitivity. Journal of Anxiety Disorders, 11, 179±200.
Stewart, S. H., & Watt, M. C. (2000). Illness attitudes scale dimensions and their associations with
    anxiety-related constructs in a nonclinical sample. Behaviour Research and Therapy, 38, 83±99.
Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). Needham Heights,
    MA: Allyn & Bacon.
308        LEFAIVRE ET AL.

Taylor, S., Koch, W. J., & McNally, R. J. (1992). How does anxiety sensitivity vary across the
   anxiety disorders? Journal of Anxiety Disorders, 6, 249±259.
Taylor, S., Koch, W. J., & Woody, S. (1996). Anxiety sensitivity and depression: How are they
   related? Journal of Abnormal Psychology, 105, 474±479.
Teachman, B. A., & Woody, S. R. (2003). Automatic processing in spider phobia: Implicit fear
   associations over the course of treatment. Journal of Abnormal Psychology, 112, 100±109.
Teachman, B. A., & Woody, S. R. (2004). Staying tuned to research in implicit cognition: Relevance
   for clinical practice with anxiety disorders. Cognitive and Behavioral Practice, 11, 149±159.
Watt, M. C., Stewart, S. H., & Cox, B. J. (1998). A retrospective study of the learning history origins
   of anxiety sensitivity. Behaviour Research and Therapy, 36, 505±525.
Williams, J. M. G., Mathews, A., & MacLeod, C. (1996). The emotional Stroop task and psycho-
   pathology. Psychological Bulletin, 120, 3±24.
Wise, T. & Sheridan, M. J. (2001). Psychometric properties of the Illness Attitudes Scale in psy-
   chiatric patients. Psychological Reports, 89, 73±76.

                                          APPENDIX
                        Example of instructions for a test-trial block:

This is a test phase with white and coloured words.
For   white words with a positive meaning: Press the GOOD key (A)
For   white words with a negative meaning: Press the BAD key (L)
For   words in a blue-ish colour: Press the GOOD key (A)
For   words in a green-ish colour: Press the BAD key (L)
You can also read