Anxiety Disorders

18 (2004) 609627

Attentional bias to threat and emotional

response to biological challenge$
William T. Naya,*,1, Geoffrey L. Thorpea,
Roxann Roberson-Nayb,1, Jeffrey E. Heckera,
Sandra T. Sigmona
a
University of Maine, Orono, ME, USA
b
University of Maryland, College Park, MD, USA

Received 15 October 2002; received in revised form 16 April 2003; accepted 4 August 2003

Abstract

Attentional bias towards threat reliably correlates with clinical anxiety status as well as
elevated trait anxiety. Although such findings have led many to posit a potential causative
or predictive role of threat-biased attentional processes on anxiety problems, little
informative research exists. The present investigation was designed to address the role
of threat-biased attentional processes on emotional/fearful responding. Eighty-seven
participants provided baseline measures of anxiety vulnerability (i.e., anxiety sensitivity;
unmasked/masked emotional Stroop task indices) and then underwent biological challenge
procedures (inhalations of 20% carbon dioxide (CO2)-enriched air). Following challenge,
participants completed measures of emotional response. Regression analyses indicated that
both unmasked and masked attentional bias indices significantly predicted emotional
responding above and beyond anxiety sensitivity. Exploratory analyses also revealed
a gender effect, with prediction of emotional response largely attributable to females.

$
This paper is based on the doctoral dissertation of the first author. Portions of this paper were
presented at the annual meeting of the Association for the Advancement of Behavior Therapy,
Philadelphia, PA, USA, 2001.
*
Corresponding author. Tel.: 1-301-594-9287; fax: 1-301-480-3610.
E-mail address: nayw@intra.nimh.nih.gov (W.T. Nay).
1
Present address: National Institute of Mental Health, Mood and Anxiety Disorders Program,
Building 1, Room 3B-20, Bethesda, MD 20892, USA.

0887-6185/$ see front matter # 2003 Elsevier Inc. All rights reserved.
doi:10.1016/j.janxdis.2003.08.003
610 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627

These findings support attentional bias towards threat as a relatively independent factor
predictive of emotional responding.
# 2003 Elsevier Inc. All rights reserved.

Keywords: Selective attention; Cognitive bias; Anxiety; Fear; Gender differences; Biological challenge

Individuals suffering from anxiety disorders typically recognize that their fears
and consequent reactions to potential threats are out of proportion to reality, yet
they feel helpless in controlling these seemingly automatic responses (McNally,
1995; Ohman & Soares, 1994). Thus, many researchers have directed their efforts
towards consideration of automatic cognitive processes that may be at least partly
responsible for the dissociation between intentioned, verbal-cognitive control,
and fearful responding. Research efforts have been directed towards such
cognitive processes as interpretive, attentional, and memory biases (MacLeod
& Rutherford, 1998; McNally, 1994, 1995). These phenomena have been
investigated for their potential as markers, and as possible causes, of clinical
anxiety dysfunction (e.g., MacLeod, Rutherford, Campbell, Ebsworthy, &
Holker, 2002; Mathews & MacLeod, 2002). The present investigation focuses
specifically on the role of attentional bias towards threat in anxiety disturbance.
A characteristic tendency to disproportionately allocate attentional resources
towards potential threats is believed to contribute to a vicious anxiety cycle
(Williams, Mathews, & MacLeod, 1996). This hypothesis proposes that high trait
vulnerability to anxiety leads to a propensity to shift attention to personally
relevant threat stimuli. Such propensity, in turn, independently causes these stimuli
to be processed more often and to a greater degree of personal emotional
significance. Increased processing of these potential threats globally increases
anxiety. Finally, increased global anxiety may make threat stimuli more salient,
thus increasing the interpreted probability of harm. Thus, cognitive models
assume that attentional bias is not simply a by-product of the emotional disorder
but plays a vital role in its causation and maintenance (Williams et al., 1996, p. 3).
By means of the emotional Stroop task or variants thereof (see, MacLeod, 1991,
for a review), numerous studies have revealed reliable relations between clinical
anxiety disorder status and attentional bias to threat (e.g., GAD: Mathews &
MacLeod, 1985; social phobia: Mattia, Heimberg, & Hope, 1993; specific phobia
of spiders: Watts, McKenna, Sharrock, & Treazise, 1986; panic disorder: Ehlers,
Margraf, Davies, & Roth, 1988; PTSD: Kaspi, McNally, & Amir, 1995; OCD:
Lavy, van Oppen, & van den Hout, 1994). Furthermore, the bias appears to be most
strongly evident with stimuli relevant to the disorder in question (e.g., cata-
strophe in panic) versus generally negative stimuli. Within clinical populations,
attentional bias effects also are observed when the word stimuli are pattern
masked. Pattern masking involves rapidly replacing (usually on the order of
1430 ms) the experimental stimuli with meaningless ones to prevent or reduce
conscious awareness of the critical stimuli. Consistent evidence of attentional
 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627 611

biases to threat under masking conditions (in clinically anxious populations) has
supported arguments that these biases emerge automatically, in the sense that they
are unintended, do not require elaborate conscious processing, and yet affect
behavior (e.g., delayed reaction times on threat-related emotional Stroop tasks). It
remains uncertain, however, whether attentional bias effects observed under
masking conditions also are disorder content-specific. A number of studies, for
example, have found that bias effects under masked conditions are attuned
generally to stimuli of negative affective tone, and not specifically to stimuli
relevant to features of particular anxiety disorders (Lundh, Wikstrom, Westerlund,
& Ost, 1999; Mogg, Bradley, Williams, & Mathews, 1993).
Given the vicious cycle hypothesis, one of the most intriguing questions
regarding attentional bias in anxiety is whether the bias precedes, and whether it
possibly contributes to, the development of anxiety dysfunction (MacLeod &
Hagan, 1992; McNally, 1994; McNally, Hornig, Hoffman, & Han, 1999). Using a
conceptual variant of the emotional Stroop taskthe visual dot-probe task,
Broadbent and Broadbent (1988) found an interesting relationship between trait
anxiety and attentional bias towards threat cues among a sample of college
females (males were not assessed). Among those highest in trait anxiety, there
appeared to be a strong, positive, linear association with attentional bias. Lower
scores on a measure of trait anxiety, however, appeared to bear no definite
relationship with attentional bias. In total, the findings suggested that attentional
bias is a continuous anxiety-related factor observable in the general population.
However, this relationship is not apparent in anyone, but rather is observed only in
individuals with high levels of trait anxiety.
Until recently (e.g., MacLeod et al., 2002; Mathews & MacLeod, 2002), little
work had been undertaken to determine whether attentional bias per se has any
unique contributory influence on anxiety or other emotional dysfunction. This
recent laboratory research has provided preliminary evidence of a causal role for
attentional bias in anxiety disturbance (MacLeod et al., 2002; Mathews &
MacLeod, 2002). MacLeod and his coworkers manipulated attentional bias in
normal control populations by reinforcing threat-biased responses to multiple
visual dot-probe training trials (or, conversely, reinforcing neutral-biased
responses in comparison groups). The authors found that, although endorsement
of negative emotional states was comparable across groups after attentional
bias training trials, the groups differed in their emotional response to stress
subsequent to training (i.e., insoluble or difficult anagrams). Participants trained
to respond with a threat bias had greater negative emotional reactions to
subsequent stress. Having directly manipulated attentional biases, the authors
make a strong case for the causal influence of attentional bias on emotional
vulnerability. However, as the authors argued, that effect is not for attentional bias
to directly influence emotion, but rather it affects emotional reactivity to sub-
sequent stress.
To date, MacLeod and his coworkers (MacLeod et al., 2002; Mathews &
MacLeod, 2002) are the only researchers to provide experimentally sound
612 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627

evidence of a possible causal role of attentional bias to threat in emotional

vulnerability. Conceptually, however, there are potential limitations of these
findings in terms of temporal reliability. Specifically, it is possible that effects
of laboratory-based attentional bias training are transient. Second, because
attentional bias in these studies is experimentally induced, the possibility
exists that the laboratory findings are limited in their ecological validity. For
instance, among individuals for whom an attentional bias develops naturally,
the processes may be somehow different. Thus, the present study attempts to
provide a conceptual bridge between: (a) correlational studies demonstrating a
relationship between attentional bias and anxiety in non-clinical populations,
and (b) laboratory paradigms of MacLeod and coworkers demonstrating a
causal role of induced attentional bias on emotional vulnerability. In particular,
this investigation examines naturalistically occurring attentional biases towards
panic-relevant threat stimuli as a predictor of emotional responding to a
biological challenge task (brief inhalations of 20% carbon dioxide (CO2)-
enriched air).
The construct of anxiety sensitivity is central to the current investigation.
Anxiety sensitivity and trait anxiety are considered vulnerabilities in the
development of anxiety dysfunction. Although the exact nature of the relation
between these two constructs has been controversial (Lilienfeld, Jacob, &
Turner, 1989; Lilienfeld, Turner, & Jacob, 1993; McNally, 1996), numerous
studies employing measures of both constructs concurrently have demonstrated
that anxiety sensitivity possesses incremental validity above and beyond trait
anxiety for many criteria relevant to fear-related anxiety dysfunction (Lilien-
feld, 1999). For example, anxiety sensitivity is a significantly stronger pre-
dictor of both, response to biological challenge (Eifert, Zvolensky, Sorrell,
Hopko, & Lejuez, 1999; Holloway & McNally, 1987), as well as to the
naturalistic, longitudinal development of panic attacks (Schmidt, Lerew, &
Jackson, 1997, 1999). Thus, anxiety sensitivity appears to reliably assess
meaningful variance in anxiety dysfunction that is relatively unique, and is
more specifically tied to the fear of anxiety sensations (Eifert et al., 1999;
Holloway & McNally, 1987; Keogh et al., 2001). Finally, although anxiety
sensitivity is associated with emotional response to biological challenge tasks,
evidence indicates that this construct may not correlate with unmasked and
masked Stroop interference indices for threat information, suggesting that each
measure uniquely accounts for variance in emotional responding to a stressor
(Lundh et al., 1999).
Given results of previous correlational work with clinical and non-clinical
samples, we anticipate that measures of attentional bias towards threat will
predict, in a linear fashion, individuals emotional responses to challenge.
Second, although the extant body of literature is considerably smaller on this
matter, we expect that our indices of attentional bias towards threat will account
for unique variance in emotional responding beyond that which can be accounted
for by anxiety sensitivity.
 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627 613

1. Method

1.1. Experimental overview

Unselected, undergraduate volunteers first consented to the study and then

completed baseline questionnaires assessing anxiety sensitivity, current anxiety
symptomatology, and demographic information (see Section 1.3 below). Follow-
ing completion of these measures, participants were interviewed about health
issues that may contraindicate inhalation of 20% CO2-enriched air. Participants
were then consented a second time before participating in the Stroop task.
Although the second consent informed the participant that they may or may
not be exposed to CO2-enriched air, all participants were exposed to CO2. This
procedure was implemented to reduce predictability and sense of control over
CO2 exposure, an important factor in the pathogenesis of anxiety and panic
(Schmidt & Lerew, 2002; Zvolensky, Eifert, Lejuez, & McNeil, 1999).
All participants then took part in computer-assisted, panic-relevant, emotional
Stroop tasks. These tasks included unmasked and masked threat stimuli. After
completion of these tasks, students were informed of the biological challenge
procedure. Participants consenting and permitted to continue then underwent the
biological challenge task involving two 30-s inhalations of 20% CO2-enriched air,
intermixed with a second set of Stroop tasks (findings from the challenge-
concurrent assessment of attentional bias to threat were reported in Nay, Rober-
son-Nay, & Thorpe, 2001). Upon completing the biological challenge task,
students completed questionnaires (i.e., Symptom Checklist, POMS) assessing
emotional response to the challenge task.

1.2. Participants

Undergraduates (n 116) in introductory and upper-level psychology courses

were recruited to participate (see Table 1 for demographic information). Twenty-
nine participants did not complete the biological challenge part of the study
(Partial Completers). Twenty-five of these 29 individuals were either excluded
from (see Health Status Interview) or declined to participate in the biological
challenge phase of the study. The primary reason individuals did not complete all
parts of the study was they were excused from the challenge task due to positive
responses to questions about possible contraindicating conditions on the Health
Status Interview (n 14; 12.1% of total sample). One participant self-elected not
to undergo CO2 challenge (prior to being excused by the experimenter) due to
possible contraindicating conditions. Of the remainder, 8 (6.9%) otherwise
declined the CO2 task, 2 (1.7%) aborted during this phase, and data for 4
(3.4%) were lost due to experimenter error. In total, 87 participants took part
in the biological challenge task and provided complete data.
As presented in Table 1, Partial Completers did not differ significantly from
Full Completers in terms of age or gender. Further between-group t-tests revealed
614 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627

Table 1
Means and standard deviations for demographic and baseline anxiety measures by completer status

Total Partial Full Completers w2 or

sample (n 116) Completers (n 29) (n 87) t-valuea

Demographics
Age 21.4 (5.63) 23.31 (8.58) 20.70 (4.10) 1.58
Gender (% female) 58.6 69.0 55.2 1.71
Baseline measures
ASI 20.57 (9.25) 20.79 (8.27) 20.49 (9.59) 0.15
STAI-T 41.91 (9.92) 43.34 (9.63) 41.44 (10.02) 0.90
STAI-S 37.05 (9.19) 38.38 (9.77) 36.61 (9.01) 0.90
Unmasked threat 14.46 (67.50) 25.53 (79.62) 10.77 (63.04) 1.02
interference index
Masked threat 3.04 (35.65) 10.10 (44.83) 0.69 (31.98) 1.23
interference index
a
All values are non-significant (P > :05).

no significant differences on baseline measures of anxiety and anxiety vulner-

ability. Thus, the final group of participants, from whom dependent measures
were taken, appears representative of the population from which they were drawn.
All participants received extra credit towards their psychology courses for
participation in the study.

1.3. Materials

1.3.1. Questionnaires and rating scales

1.3.1.1. Symptom Checklist. The Symptom Checklist is a 19-item rating scale

listing the 13 DSM-IV (American Psychiatric Association, 1994) panic attack
symptoms, and six additional cognitive and somatic fear-related symptoms
(Appendix). This measure is similar in content and structure to the Diagnostic
Symptom Questionnaire (DSQ; Sanderson, Rapee, & Barlow, 1989) used in prior
biological challenge research. Items on the Symptom Checklist are listed with
5-point Likert-type ratings of severity (0 none to 4 very severe). Thus, the
potential range of scores is 076. This measure served as a primary dependent
variable.

1.3.1.2. Profile of Mood States. The POMS (McNair, Lorr, & Droppleman, 1981)
is a 65-item self-report inventory designed to assess transient, fluctuating mood
states. The items are rated on a 04 Likert-type scale of current intensity, where 0
is not at all and 4 is extremely. The Tension-Anxiety (T/A) subscale, which
is being used in the current investigation, is composed of nine of the sixty-five
items (range of scores 036) and includes items describing somatic tension and
psychomotor agitation. Internal consistency for the T/A scale across two studies
with psychiatric outpatients averaged .91 and test-retest reliability was .70 over an
 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627 615

average of 20 days. Means for the T/A scale in a college population are 13.9
(S:D: 7:4) for females and 12.9 (S:D: 6:8) for males. Scores on the T/A
subscale of the POMS discriminates individuals with panic disorder and non-
clinical panickers (Norton, Dorward, & Cox, 1986) as well as high-anxiety
sensitive college students from low-anxiety sensitive college students (Donnell &
McNally, 1989; Holloway & McNally, 1987), both upon biological challenge.

1.3.1.3. Health Status Interview. A Health Status Interview was devised to probe
for medical issues that may contraindicate brief inhalations of 20% CO2-enriched
air. Because of the relative safety of this procedure, such medical screenings are
uncommon in the extant literature. However, we conservatively elected to dismiss
from the challenge task any participant who endorsed taking medications (other
than contraceptives or aspirin/analgesics), adult asthma or rhinitis (as evidenced
through current use of inhaler or oral medication), history of cardiac disease or
dysfunction, history of high blood pressure treated with medication, history of
regular fainting episodes, history of epilepsy, or that a physician recommended
that they do not exercise.

1.3.1.4. Anxiety Sensitivity Index. The ASI (Peterson & Reiss, 1992) is a 16-item
self-report inventory devised to assess the degree to which a person fears anxiety
sensations. Each item is scored on a 5-point Likert-type scale that ranges from 0
(very little) to 4 (very much), with a total score range of 064. The average
score on the ASI for normal samples is 19.01 (S:D: 9:11; Peterson & Reiss,
1992). Internal reliabilities (Cronbachs alpha) across both normal and patient
populations range between .8 and .9. Also, acceptable test-retest reliabilities have
been demonstrated in college students at 2-week intervals (r :75) and at 3-year
intervals (r :71).

1.3.1.5. State-Trait Anxiety Inventory. The STAI (Spielberger, Gorsuch, Lushene,

Vagg, & Jacobs, 1983) is divided into two independent 20-item tests; a trait
anxiety scale (STAI-T) and a state anxiety scale (STAI-S). The STAI-T is a
commonly used measure of dispositional anxiety in non-clinical attentional bias
studies. Items are scored on a 14 scale, where a score of 4 reflects the presence of
a high level of anxiety for that item (roughly half must be reverse scored before
totaling). Each scale, therefore, has a possible range of scores from 20 to 80.
Normative means (and standard deviations) based on over 800 college students
are 38.8 (S:D: 12) for females and 36.5 (S:D: 10) for males on the STAI-S.
For the STAI-T, they are 40.4 (S:D: 10:2) for females and 38.3 (S:D: 9:2) for
males. Measures of internal consistency for this population are roughly .92 for
STAI-S and .90 for STAI-T, suggesting high coherence for these measures.

1.3.2. Stroop materials

Stroop task stimuli were selected from a previous investigation on attentional
bias and its potential relationship with anxiety vulnerability (McNally et al.,
616 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627

1999). Split-half sets were created from each valence (panic-related and neutral
household words) from the item pool, creating four sets of 14 words each. Words
were matched between sets to equate for length and frequency of English usage
(Francis & Kucera, 1982).
Emotional Stroop tasks were run through a Dell Dimension XPS T500 Pentium
III mini-tower computer. Stimuli were presented on a 19 in. Dell P991 Super
VGA monitor, which was set at its lowest possible resolution for the purpose of
maximizing refresh rate. Vocal reaction times to the Stroop tasks were recorded
by the computer via a serial response box (Psychology Software Tools, Inc., 1999)
and a microphone. Stimulus and mask presentation was controlled by specialized
software.

1.3.3. CO2 delivery system

The 20% CO2-enriched air delivery system was designed in general accor-
dance with Lejuez, Forsyth, and Eiferts (1998) recommendations for manual
delivery systems. An intake port of a bi-valve respiratory mask (New Rudolph
Mouth Breathing Face Mask, Two-Way Non-Rebreathing with a T-shape valve,
Hans Rudolph, Inc.) was connected via flexible tubing (Hytrel Tubing, Hans
Rudolph, Inc.) to a gas-impermeable three-way valve (Manual Directional
Control Valve, Three-Way T-Shape Stopcock Type, Series 2120, Hans Rudolph,
Inc.) located in the researchers observation space. One port of the three-way
valve was open to room air. The remaining port (with one already fitted by tubing
to the face mask intake valve) was fitted with a 30-l meteorological balloon. This
balloon acted as a reservoir for the 20% CO2-enriched air (20.1% CO2, 20.4%
oxygen, 59.5% nitrogen; SpecAir Specialty Gases, Auburn, ME, USA). Use of a
balloon reservoir made it possible to deliver the 20% CO2-enriched air in a
manner dictated by a participants natural breathing rate and depth.

1.4. Procedures

Following consent, participants completed baseline questionnaires and were

interviewed about health status. The first informed consent procedure described
only the baseline questionnaires, Health Status Interview, and the emotional
Stroop task. Designing the study with two consent procedures was necessary to
control for anticipatory anxiety (a potential confound) that may arise in parti-
cipants if they are concerned about the biological challenge task. Participants
completed the emotional Stroop procedures regardless of responses to health
probes.
Following the Health Status Interview, participants were moved to a room
outfitted for the Stroop task and were seated in a chair so that the participants face
was roughly 18 from the center of a 19 in. color computer monitor. The
investigator then provided practice instructions for the emotional Stroop task.
Practice trials consisted of 10 neutral words that were not included in the
experimental trials (each word randomly presented in any of the four possible
 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627 617

color backgrounds, allowing presentation of up to 20 unmasked trials followed by

up to 20 masked trials). Once the participant demonstrated understanding of the
procedure and completed at least 10 trials of each presentation type, practice was
terminated.
Following practice trials, participants were told that instructions appearing on
the monitor would initialize the experimental Stroop task. Computer automated
procedures began by notifying participants of a rest period in which they were to
relax comfortably. Experimental Stroop procedures were initiated after 120 s of
rest and 15 s of on-screen instructions reminding participants of the Stroop
procedures.
For the unmasked Stroop task, software was programmed to start trials with a
fixation cross (13 mm high  30 mm wide) for 750 ms in the center of the screen
(1250 ms blank screen prior to each fixation box, creating a 2-s Inter-Trial Interval
(ITI)). This was immediately followed by Stroop stimulus words presented in
white uppercase letters 8 mm high. These words were presented on a patch of
background color (approximately 10 mm  38 mm) randomly selected among
red, blue, green, and yellow, with the restriction that no color was presented twice
in a row. The word and background colors were displayed for up to 3 s (in the
circumstance that no response was detected by the microphone). Controlling
software automatically returned the screen to a blank, white display upon receipt
of a response from the participant or after 3 s. Upon receipt of a response or after
3 s had passed, the next 2-s ITI was initiated.
The same general procedure was used for the masked Stroop tasks. The
exception is the presentation of masks following word presentation. An array of
Xs equal in number of characters to the preceding word appeared 25 ms after onset
of the original word and background color. As with the experimental stimuli in the
unmasked trials, the array of Xs remained on display until the participant provided
a voice response or 3 s had passed.
Words were presented individually within blocked sets (e.g., Neutral One
block), with the order of blocks randomized across participants. Participants
completed one unmasked set, then one masked set (of the same valence set), the
opposite valence unmasked, and then masked set. Words within the appropriate
set were presented randomly, with constraints that no word or background color
presented twice in succession. Each individual stimulus word was presented four
times (once with each background color, never in consecutive presentations), for a
total of 224 trials. For example, one participant may have received the Neutral
(1) set unmasked, then Neutral (1) masked, then a rest period followed by
Threat (2) unmasked, then Threat (2) masked. The constraint that unmasked
presentations came first was placed on the design due to previous research
suggesting that the effects of automatic attentional bias may need to be primed
for effects to be observed (Fox, 1996). Undoubtedly, this is an important
conceptual issue, but one that was not addressed in this particular examination.
Participants had communication abilities with the experimenter throughout the
experiment.
618 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627

Once the participant had completed the Stroop assessment, the researcher
re-entered the room and provided the participant with mid-point self-report
assessments to be used for manipulation checks (POMS-T/A scale and
Symptom Checklist). After completion of the questionnaires, the second
consent procedure was undertaken. The second consent procedure explained
that participants might be administered dosages of CO2-enriched air. Further-
more, consent procedures informed participants that it would not be possible
for them to know when the CO2 could be administered. All participants,
regardless of their responses to the health status questionnaire, were given the
opportunity to accept or decline. Once the participant made his or her election,
the experimenter excused those who answered affirmatively to any of the
contraindicating conditions and/or who declined. Asking participants for their
assent without regard to eligibility was implemented so that it may be possible
later to statistically look at factors that may predict refusal of CO2 admin-
istration.
For those participants eligible and assenting to continue with the biological
challenge, the experimenter then fitted the participant with a respiratory mask.
After 60 s of rest/acclimation, the researcher opened the three-way valve to the
20% CO2-enriched air for 30 s. A 30-s period of breathing room air followed, and
then additional Stroop tasks were initiated. Half-way through the additional tasks
(matching the rest period in the middle of the initial Stroop procedures), the three-
way valve was once again opened to the 20% CO2-enriched air for 30 s. Following
completion of the experimental session, the participant was asked to complete a
second POMS-T/A subscale, a second Symptom Checklist, and a second STAI-S.
Upon completion of the questionnaires, each participant was debriefed and given
research credit.

1.5. Data reduction strategy

Typical data reduction procedures used in the attentional bias literature were
employed (e.g., MacLeod & Hagan, 1992; Mogg et al., 1993). Mean scores were
calculated for each participant across the four critical Stroop epochs: unmasked
threat, unmasked neutral, masked threat, and masked neutral. Prior to calculating
these means, however, instances of no response or response artifacts were
eliminated. Instances of no response were evident as 3-s response times. Also,
responses shorter than 100 ms were eliminated as likely artifacts. At this point,
within-individual means and standard deviations per valence set (neutral words
vs. threat words) and per presentation type (unmasked vs. masked) were calcu-
lated. Data points beyond three standard deviations of these specific within-
individual means were eliminated.
For analytic purposes, data were reduced further. Attentional bias to threat
index scores were calculated. To accomplish this, individual mean scores for each
neutral set were subtracted from the individual mean score from the appropriate
threat set (Lundh et al., 1999; MacLeod & Hagan, 1992; Mogg et al., 1993). A
 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627 619

positive index score reflects some degree of bias towards threat. In all, each
participant provided one unmasked and one masked threat index.

2. Results

2.1. Manipulation check

The questions to be addressed in this study are dependent on the assumption that
inhalation of CO2-enriched air will increase symptoms of fearful/anxious respond-
ing. Therefore, t-tests were conducted on baseline STAI-S compared to post-CO2
STAI-S anxiety, mid-point Symptom Checklist compared to post-CO2 Symptom
Checklist, and mid-point POMS-T/A subscale versus post-CO2 POMS-T/A sub-
scale. Use of the mid-point measures, where available, was considered a more
conservative test of symptom increase specifically due to CO2 than baseline
assessment of symptoms would have been. Specifically, it is arguable that the (first
set of) Stroop tasks may have increased symptoms compared to baseline due to the
stress of the procedure and/or the presentation of threat stimuli. Therefore, a more
direct argument can be made that any change in symptoms after the initial Stroop
tasks to post-CO2 should be due primarily to the effects of inhaling CO2-enriched air.
As predicted, Symptom Checklist and POMS scores significantly increased from
post-Stroop to post-CO2 inhalation (M 5:5 vs. M 11:7, t 8:69, P < :05;
M 8:2 vs. M 12:2, t 6:52, P < :05, respectively). Significant increases in
state anxiety also were observed from baseline to post-CO2 (M 36:6 vs.
M 47:0, t 8:07, P < :05). Results suggest that the two 30-s inhalations of
20% CO2-enriched air successfully increased fearful/anxious symptoms.

2.2. Strength of relations between anxiety sensitivity, trait anxiety, and

symptomatic response

The overall zero-order correlation matrix for the predictors and dependent
variables can be found in Table 2. Inspection of the matrix reveals that all
predictors were significantly correlated with post-CO2 symptoms (i.e., Symptom
Checklist scores). Similarly, the post-CO2 POMS T/A subscale was significantly
correlated with all measures, except for the masked threat index. The Symptom
Checklist and POMS T/A subscale were highly correlated (r :75, P < :001),
suggesting that these measures assess largely the same construct. Therefore, to
reduce the likelihood of Type I error, only the Symptom Checklist was retained as
a primary dependent measure. This measure contains more panic-relevant items
and maps directly onto DSM-IV (American Psychiatric Association, 1994) panic
attack criteria.
Contrary to a priori expectations, association between the ASI and Symptom
Checklist was not statistically greater than for the STAI-T and the Symptom
Checklist (r :40 vs. r :32, respectively; Zdiff 0:60; P > :05). As the
620 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627

Table 2
Zero-order correlation matrix for predictors/covariates and dependent variables

Measure 1 2 3 4 5

Predictor variables
ASI
Trait anxiety .58**
Unmasked index .12 .23*
Masked index .05 .24* .31**
Criterion variables
Symptoms Checklist .40** .32** .50** .24*
POMS-T/A .36** .37** .33** .18 .75**
n 87; n 86 for POMS T/A.
*
P < :01.
**
P < :001.

STAI-T and ASI were moderately correlated (r :58; P < :01), Symptom Check-
list scores were regressed onto ASI and STAI-T scores using stepwise procedures.
This analysis resulted in a final prediction model including only the ASI
(F 16:36; P < :01). Our finding suggests that any non-overlapping variance
accounted for by the STAI-T does not significantly add predictive power above and
beyond that accounted for by the ASI, despite the moderately strong zero-order
correlations held by both with respect to symptomatic response to CO2 inhalation.

2.3. Predicting symptomatic response to CO2 challenge

Multiple regression procedures were used to predict individual differences in

fearful responding to CO2-enriched air inhalation. The dependent or criterion
variable for this series of regression models was the total score from the Symptom
Checklist assessed after the CO2 inhalation. In all, two prediction models were
examined. For Model 1, predictor variables were entered into a multiple regres-
sion procedure in the following order: ASI total score (force entered), unmasked
threat interference index, and the centered ASI  unmasked threat interference
index interaction. The latter two predictors were entered as a block using forward
stepping criteria within the block. Model 2 replicated Model 1, except that the
masked threat interference index and centered ASI  masked threat interference
index interaction were entered as predictors, again using forward stepping
procedures within the block.
Results of the two multiple regression models are described in Table 3.
Consistent with previous findings, the ASI was a moderately strong predictor
of Symptom Checklist scores. Also, as predicted, the unmasked threat inter-
ference index was a significant predictor of self-reported panic symptoms, even
with variance due to anxiety sensitivity accounted for beforehand. Furthermore,
the unmasked index was actually a stronger predictor of symptomatic response
than anxiety sensitivity.
 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627 621

Table 3
Summary of multiple regression analyses: unmasked and masked index in the prediction of Symptom
Checklist scores

Predictors B S.E.B. b R2 DR2

Model 1a
Step 1
(Hierarchical) .16**
**
ASI .32 .08 .35
Step 2
(Forward stepping within block) .37** .21**
Unmasked index .00 .01 .46**

Model 2b
Step 1
(Hierarchical) .16**
**
ASI .32 .08 .35
Step 2
(Forward stepping within block) .21** .05*
Masked index .00 .03 .22*
a
Centered ASI  unmasked index interaction did not enter significantly (P > :05).
b
Centered ASI  masked index interaction did not enter significantly (P > :05).
*
P < :01.
**
P < :001.

Similar to the previous analysis, symptomatic response was regressed onto

ASI, masked interference index, and the centered interaction of ASI  masked
bias index. As predicted, the masked index score was a significant predictor of
physical symptoms resulting from CO2 beyond the variance that could be
accounted for by anxiety sensitivity.

2.4. Gender differences

Exploratory reanalysis was undertaken to help determine if gender mediates

the interrelationship among anxiety sensitivity, attentional bias to threat, and
symptomatic response to challenge. This reanalysis is driven by recent findings
suggesting that respiratory-related symptoms occur with greater frequency in
female versus male panic populations (Sheikh, Leskin, & Klein, 2002). Basic
research into the factors that can help account for these types of gender
differences, therefore, is of utmost importance.
In the current sample, females did not report significantly more panic
symptoms than males in response to the CO2 (t 1:06, P > :05; M 12:60
vs. M 10:56, respectively). Neither ASI nor any attentional bias measure
differed significantly by gender. To be consistent with the statistical methods
used for a priori data analysis, zero-order correlations were conducted separately
for males and females on all independent and dependent variables (see Table 4).
622 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627

Table 4
Zero-order correlation matrix for predictors and dependent variables: males (n 39)/females
(n 48)

1 2 3

Predictor variables
ASI
Unmasked index .05/.15
Masked index .01/.07 .35*/.29*
Criterion variable
Symptom Checklist .23/.48* .31/.59** .14/.29*
*
P < :01.
**
P < :001.

On inspection of Table 4, it is apparent that gender differentially affects which

predictors are significantly related to post-CO2 Symptom Checklist severity
scores. In particular, every predictor correlated with post-CO2 symptoms for
females. Conversely, there were no significant zero-order correlations for males
on the dependent variable. An initial question arises from these findings about
whether gender was differentially related to the degree of symptom increase from
post-Stroop to post-CO2. A 2  2 mixed-model MANOVA was conducted to
examine this question. Time (post-Stroop and post-CO2) served as the within-
subjects factor and Gender served as the between-subjects factor. As expected,
given the overall a priori manipulation checks, there was a significant main
effect for Time on Symptom Checklist severity scores (Wilks lambda :54,
F 73:54, P < :001). There was, however, no significant Gender main effect
or Time  Gender interaction (MS 155:20, F 1:59, P > 0:05; Wilks
lamda 1:00, F :038, P > 0:05, respectively). Therefore, the differential
relations between predictors and symptoms according to gender cannot be
attributed to greater symptom increase by either males or females.
The differential zero-order correlations of Table 4 suggest that power in this
studys central regression models may have been suppressed by male noise. For
this reason, regression models were performed with females only. As with the
initial analyses, two models were examined wherein the ASI was force entered in
the first block followed by the interference indices and their interactions with the
ASI in the second block using forward stepping procedures. Because multiple
statistical comparisons were made using data of female participants, an alpha
adjustment (Bonferroni correction) was undertaken to reduce the possibility of
Type I error. Thus, a corrected alpha of .025 (.05/2) was required to meet
statistical significance. Resultant data are located in Table 5.
The pattern of results found for females was much the same as the initial
regression models using the full sample. The magnitude of effects (i.e., full model
R2s) across all three regression analyses, however, were larger when only females
were included in the analyses.
 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627 623

Table 5
Summary of multiple regression analyses: unmasked and masked index in the prediction of Symptom
Checklist scores (females only)

Predictors B S.E.B. b R2 DR2

Model 1a
Step 1
Hierarchical .23**
**
ASI .34 .09 .40
Step 2
Forward stepping within block .50** .27**
Unmasked index .06 .01 .53**

Model 2b
Step 1
Hierarchical .23**
**
ASI .34 .09 .40
Step 2
Forward stepping within block .30** .07*
Masked index .06 .03 .26*
a
Centered ASI  unmasked index interaction did not enter significantly (P > :025).
b
Centered ASI  masked index interaction did not enter significantly (P > :025).
*
P < :01.
**
P < :001.

3. Discussion

Overall findings support the basic premise of this investigation: individual

differences in the tendency to disproportionately allocate attentional resources to
low-level threats are uniquely predictive of emotional response to challenge.
These findings are largely supportive of the vicious cycle hypothesis (Williams
et al., 1996). Specifically, this study supports prior research in which dispositional
anxiety (i.e., anxiety sensitivity) is directly related to measures of attentional bias
(cf. Broadbent & Broadbent, 1988; MacLeod & Rutherford, 1998). A particular
contribution of this study, however, is support of the second major aspect of the
vicious anxiety cycle hypothesis. That is, once developed, stable encoding biases
for threatening information represent a unique and independent individual
difference factor in predicting anxious and fearful responding to a stressor.
Regression analysis substituting the masked threat interference index for the
unmasked threat interference index produced similar findings. One apparent
difference between the two analyses, however, was the smaller magnitude of
unique effect for the masked index than for the unmasked index (DR2 :06 vs.
.21). The differences in size of effect for these two indices supports the notion that
varying the level of stimuli awareness results in, at least, quantitative differences.
Previous finding of quantitative, and at times qualitative, differences among
unmasked and masked measures of attentional bias in relation to anxiety may
624 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627

indicate that these varying methods tap different levels of cognitive analysis (i.e.,
strategic versus automatic; post-attentive versus pre-attentive; see, Williams et al.,
1996, for a review). The current findings can be interpreted in at least two ways. One
interpretation is that, in fact, the two different procedures do tap different levels of
cognitive analysis. Another interpretation would be that the level of cognitive
processing involved is the same, but the effect is weakened by degraded or impeded
awareness. Unfortunately, the methods involved in the current investigation do not
allow a means to clarify which of these two hypotheses better accounts for the
findings. Regardless, the findings of the current investigation suggest that the
proportion of attentional resources allocated to even a procedurally masked threat
stimulus has a direct, linear effect on ones affective reactions.
Reanalysis of the regression models according to gender group status revealed
unexpected findings, given that gender differences have not been noted frequently
in the literature on attentional biases. In particular, it appears that the prediction of
symptomatic response by anxiety sensitivity and interference indices can largely
be attributed to effects observed among female participants. One caveat is offered
in considering the generalizability of the gender difference findings. Specifically,
the magnitude of symptom increase, although roughly the same for males and
females, was small in absolute terms. It is possible that true but less robust effects
in males were translated into non-significant findings. Further research on the
issue of gender differences in attentional bias is clearly warranted.
Given the possibility that attentional bias to threat represents a vulnerability
factor in anxiety disturbance, the speculation naturally arises that perhaps it is
possible to re-train or re-condition vulnerable individuals so that they no longer
disproportionately attend to threat. Recent studies hint at the possibility of
implementing laboratory procedures designed to train individuals in neutral or
positively skewed attentional allocation (MacLeod et al., 2002; Mathews &
MacLeod, 2002). Other practical applications of the current findings also are
evident, including the use of attentional bias paradigms as a method of anxiety
vulnerability assessment. Through the development of standardized procedures,
measures of attentional bias to threat may become valuable clinical tools in multi-
method assessment. Future research is needed, however, in addressing the
temporal stability and reliability of attentional bias assessments.
The current study contains a number of methodological limitations that can be
improved upon in future research. One such limitation is the lack of a subjective
awareness check on masked stimuli. This check was not conducted in the current
study for a number of reasons. First, the chosen interval between stimulus
presentation and mask presentation (25 ms) falls well in the range of stimulus
presentation duration found effective for blocking subjective awareness in
numerous previous studies (Esteves & Ohman, 1993; Lundh et al., 1999; Mogg
et al., 1993; Mogg, Kentish, & Bradley, 1993; Whalen et al., 1998). Second, a
logistical complication arises regarding where in the protocol to inject awareness
checks. If done anywhere in the middle of the protocol, these checks may affect
later trials. If awareness checks are conducted at the end of the trials, one has to
 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627 625

consider the influence of both practice effects (or spreading cognitive activation)
and of having unmasked presentations of the same stimuli beforehand. Certainly,
many of these difficulties can be overcome or simply accepted as a complication
of conducting awareness checks. However, given the preliminary nature of the
current study and the additional burden the awareness checks would have put on
research participants, it was decided to limit discussion to the effects of pro-
cedural masking.
To maximize emotional responding, predictability, and control were important
considerations in the present experimental design. Nonetheless, future investiga-
tions could experimentally manipulate these factors in an attempt to examine the
ability of attentional biases to predict emotional responding when these factors
vary. One final consideration for the present study is that the pattern of results
observed may not apply to clinical samples. Future investigations could more
directly assess the relations among these variables within clinical samples to
determine whether similar patterns of association emerge.

Acknowledgments

The research was supported, in part, by a 1999 University of Maine Faculty

Research Grant awarded to Dr. Geoffrey L. Thorpe. The authors would like to
acknowledge Dr. Mark Jackson, Director, Cutler Health Center, for his consulta-
tion on this project and his assistance with the development of the Health Status
Interview. Additionally, we extend our appreciation to Dr. Colin Martindale and
Dr. Christine Fink for their reviews of previous versions of this manuscript.

Appendix A. Additional symptoms included in the Symptom Checklist

(1) Desire to escape or run out

(2) Difficulty keeping calm
(3) Visual difficulties (blurred vision, tunnel vision)
(4) Hearing difficulties (ringing in the ears, difficulty hearing)
(5) Feelings of embarrassment
(6) Difficulty concentrating

References

American Psychiatric Association. (1994). Diagnostic and Statistical Manual of Mental Disorders
(4th Ed.). Washington, DC: Author.
Broadbent, D., & Broadbent, M. (1988). Anxiety and attentional bias: state and trait. Cognition &
Emotion, 2, 165183.
Donnell, C. D., & McNally, R. J. (1989). Anxiety sensitivity and history of panic as predictors of
response to hyperventilation. Behaviour Research & Therapy, 27, 325332.
626 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627

Ehlers, A., Margraf, J., Davies, S., & Roth, W. T. (1988). Selective attention of threat cues in subjects
with panic attacks. Cognition & Emotion, 2, 201219.
Eifert, G. H., Zvolensky, M. J., Sorrell, J. T., Hopko, D. R., & Lejuez, C. W. (1999). Predictors of
self-reported anxiety and panic symptoms: an evaluation of anxiety sensitivity, suffocation fear,
heart-focused anxiety, and breath-holding duration. Journal of Psychopathology and Behavioral
Assessment, 21, 293305.
Esteves, F., & Ohman, A. (1993). Masking the face: recognition of emotional facial expressions as a
function of the parameters of backward masking. Scandinavian Journal of Psychology, 34, 118.
Fox, E. (1996). Selective processing of threatening words in anxiety: the role of awareness. Cognition
& Emotion, 10, 449480.
Francis, W. N., & Kucera, H. (1982). Frequency analysis of English usage. Boston: Houghton Mifflin.
Holloway, W., & McNally, R. J. (1987). Effects of anxiety sensitivity on the response to
hyperventilation. Journal of Abnormal Psychology, 96, 330334.
Kaspi, S. P., McNally, R. J., & Amir, N. (1995). Cognitive processing of emotional information in
posttraumatic stress disorder. Cognitive Therapy & Research, 19, 433444.
Keogh, E., Dillon, C., Georgiou, G., & Hunt, C. (2001). Selective attentional biases for physical
threat in physical anxiety sensitivity. Journal of Anxiety Disorders, 15, 299315.
Lavy, E. H., van Oppen, P., & van den Hout, M. A. (1994). Selective processing of emotional
information in obsessive compulsive disorder. Behaviour Research & Therapy, 32, 243246.
Lejuez, C. W., Forsyth, J. P., & Eifert, G. H. (1998). Devices and methods for administering carbon
dioxide-enriched air in experimental and clinical settings. Journal of Behavior Therapy &
Experimental Psychiatry, 29, 239248.
Lilienfeld, S. O. (1999). Anxiety sensitivity and the structure of personality. In: S. Taylor (Ed.),
Anxiety sensitivity: theory, research, & treatment of the fear of anxiety (pp. 149180). Mahwah,
NJ: Lawrence Erlbaum Associates.
Lilienfeld, S. O., Jacob, R. G., & Turner, S. M. (1989). Comment on Holloway and McNallys (1987)
Effects of anxiety sensitivity on the response to hyperventilation. Journal of Abnormal
Psychology, 98, 100102.
Lilienfeld, S. O., Turner, S. M., & Jacob, R. G. (1993). Anxiety sensitivity: an examination of
theoretical and methodological issues. Advances in Behavior Research & Therapy, 15, 147187.
Lundh, L., Wikstrom, J., Westerlund, J., & Ost, L. (1999). Preattentive bias for emotional information
in panic disorder with agoraphobia. Journal of Abnormal Psychology, 108, 222232.
MacLeod, C. M. (1991). Half a century of research on the Stroop effect: an integrative review.
Psychological Bulletin, 109, 163203.
MacLeod, C., & Hagan, R. (1992). Individual differences in the selective processing of threatening
information, and emotional responses to a stressful life event. Behaviour Research & Therapy, 30,
151161.
MacLeod, C., & Rutherford, E. M. (1998). Automatic and strategic cognitive biases in anxiety and
depression. In: K. Kirsner, C. Speelman, M. Maybery, A. OBrien-Malone, M. Anderson, & C.
MacLeod (Eds.), Implicit and explicit mental processes (pp. 233254). Mahwah, NJ: Lawrence
Erlbaum Associates.
MacLeod, C., Rutherford, E., Campbell, L., Ebsworthy, G., & Holker, L. (2002). Selective attention
and emotional vulnerability: assessing the causal basis of their association through the
experimental manipulation of attentional bias. Journal of Abnormal Psychology, 111, 107123.
Mathews, A., & MacLeod, C. (1985). Selective processing of threat cues in anxiety states. Behaviour
Research & Therapy, 23, 563569.
Mathews, A., & MacLeod, C. (2002). Induced processing biases have causal effects on anxiety.
Cognition & Emotion, 16, 331354.
Mattia, J. L., Heimberg, R. G., & Hope, D. A. (1993). The revised Stroop color-naming task in social
phobics. Behaviour Research & Therapy, 31, 305315.
McNair, D. M., Lorr, M., & Droppleman, L. F. (1981). EdITS manual: Profile of Mood States. San
Diego, CA: Educational and Industrial Testing Service.
 W.T. Nay et al. / Anxiety Disorders 18 (2004) 609627 627

McNally, R. J. (1994). Panic disorder: a critical analysis. New York: Guilford.

McNally, R. J. (1995). Automaticity and the anxiety disorders. Behaviour Research & Therapy, 33,
747754.
McNally, R. J. (1996). Cognitive bias in the anxiety disorders. In: D.A. Hope (Ed.), Perspectives on
anxiety, panic, & fear (pp. 211250). Lincoln, NE: University of Nebraska Press.
McNally, R. J., Hornig, C. D., Hoffman, E. C., & Han, E. M. (1999). Anxiety sensitivity and
cognitive biases for threat. Behavior Therapy, 30, 5161.
Mogg, K., Bradley, B. P., Williams, R., & Mathews, A. (1993). Subliminal processing of emotional
information in anxiety and depression. Journal of Abnormal Psychology, 102, 304311.
Mogg, K., Kentish, J., & Bradley, B. P. (1993). Effects of anxiety and awareness on colour-
identification latencies for emotional words. Behaviour Research & Therapy, 31, 559567.
Nay, W. T., Roberson-Nay, R., & Thorpe, G. L. (2001, November). The role of attentional bias to
threat in fear responding. Poster presented at the 35th annual meeting of the Association for the
advancement of Behavior Therapy, Philadelphia, PA.
Norton, G. R., Dorward, J., & Cox, B. J. (1986). Factors associated with panic attacks in nonclinical
subjects. Behavior Therapy, 17, 239252.
Ohman, A., & Soares, J. J. F. (1994). Unconscious anxiety: phobic responses to masked stimuli.
Journal of Abnormal Psychology, 103, 231240.
Peterson, R. A., & Reiss, S. (1992). Anxiety Sensitivity Index revised test manual. Worthington, OH:
International Diagnostic Services.
Sanderson, W. C., Rapee, R. M., & Barlow, D. H. (1989). The influence of an illusion of control on
panic attacks induced via inhalation of 5.5% CO2-enriched air. Archives of General Psychiatry,
46, 157162.
Schmidt, N. B., & Lerew, D. R. (2002). Prospective evaluation of perceived control, predictability,
and anxiety sensitivity in the pathogenesis of panic. Journal of Psychopathology and Behavioral
Assessment, 24, 207214.
Schmidt, N. B., Lerew, D. R., & Jackson, R. J. (1997). The role of anxiety sensitivity in the
pathogenesis of panic: prospective evaluation of spontaneous panic attacks during acute stress.
Journal of Abnormal Psychology, 106, 355364.
Schmidt, N. B., Lerew, D. R., & Jackson, R. J. (1999). Prospective evaluation of anxiety sensitivity in
the pathogenesis of panic: replication and extension. Journal of Abnormal Psychology, 108, 532
537.
Sheikh, J. I., Leskin, G. A., & Klein, D. E. (2002). Gender differences in panic disorder: findings
from the National Comorbidity Survey. American Journal of Psychiatry, 159, 5558.
Spielberger, C. D., Gorsuch, R. L., Lushene, R., Vagg, P. R., & Jacobs, G. A. (1983). Manual for the
State-Trait Anxiety InventorySTAI (Form Y). Palo Alto, CA: Consulting Psychologists Press.
Watts, F. N., McKenna, F. P., Sharrock, R., & Treazise, L. (1986). Colour naming of phobia-related
words. British Journal of Psychology, 77, 97108.
Whalen, P. J., Rauch, S. L., Etcoff, N. L., McInerney, S. C., Lee, M. B., & Jenike, M. A. (1998).
Masked presentations of emotional facial expressions modulate amygdala activity without
explicit knowledge. Journal of Neuroscience, 18, 411418.
Williams, J. M. G., Mathews, A., & MacLeod, C. (1996). The emotional Stroop task and
psychopathology. Psychological Bulletin, 120, 324.
Zvolensky, M. J., Eifert, G. H., Lejuez, C. W., & McNeil, D. W. (1999). The effects of offset control
over 20% carbon-dioxide-enriched air on anxious responding. Journal of Abnormal Psychology,
108, 624632.