Vulnerability to Anxiety 121
rather rapidly. The final experiment demonstrated that induced bias for threat will lead
to an increase in anxiety only when it is activated by generating personally threatening
meanings. The authors conclude that their results provide direct experimental evidence
that activation of threat interpretation bias plays a causal role in anxiety.
In a more recent study Wilson et al. (2006) used the homograph interpretative bias
induction of Grey and Mathews (2000) and randomly assigned 48 nonanxious students
to a threat or nonthreat training condition. Analysis revealed the expected differential
interpretation bias with training but no direct effect on depressed or anxious mood.
However, interpretation bias did have a significant impact on emotional reactivity to
four stressful video clips with the threat- trained group showing an elevation in state
anxiety in response to the stressor. The authors concluded that threat interpretation bias
can make “a causal contribution to anxiety reactivity” (Wilson et al., 2006, p. 109).
Yiend, Mackintosh, and Mathews (2005) used the text-based ambiguous social
scenarios from Mathews and Mackintosh (2000) to demonstrate that the induction of
a threat interpretation bias can endure over at least 24 hours but, like previous studies,
there was no significant direct effect on state anxiety. In another study Mackintosh,
Mathews, Yiend, Ridgeway, and Cook (2006) again found that induced interpretation
bias endured over a 24-hour time period and survived changes in environmental context
between training and testing. This enduring effect of induction training was replicated
in a second experiment using text-based scenarios involving potential physical threat.
Furthermore, individuals with the negative interpretation training showed the largest
increases in state anxiety after viewing stressful accident video clips a day after training.
However, a replication study of Mathews and Mackintosh (2000) failed to find that the
effects of interpretative bias training generalized to indices of interpretative process-
ing that differed from the training task, although they did find that negatively trained
individuals had significant increases in state anxiety (Salemink et al., 2007a). A second
experiment, however, produced negative results, with positive and negative interpreta-
tive bias training having no significant effect on state anxiety or emotional reactivity
to stress (Salemink et al., 2007b). Together these results indicate that interpretative
training effects can endure over time and across differing environmental and possibly
stimulus contexts, and that changes in emotional reactivity due to training may also
have some measure of durability.
In a special issue of the Journal of Abnormal Psychology a series of studies based
on cognitive bias training demonstrated that significant therapeutic benefits could be
achieved from directly training anxious individuals to generate benign or positive inter-
pretations to emotionally ambiguous material, or to selectively attend to nonthreatening
stimuli; the procedures were labeled cognitive bias modification (for a discussion see
MacLeod, Koster & Fox, 2009). Four studies are of particular relevance in demonstrat-
ing the causal status of threat bias. In the first study nonclinical students who were
trained over several days to selectively avoid emotionally negative or threatening words
using a home-based dot probe program had significantly lower trait anxiety scores and
weakened stress reactivity to a naturalistic stressor encountered 48 hours after training
than a no-train control group (MacLeod & Bridle, 2009).
In a second study high worriers trained to access benign meanings to threat-related
homographs and emotionally ambiguous scenarios had significantly fewer negative
thought intrusions and less anxiety during a focused breathing task than the no- training
control group (Hirsch, Hayes, & Mathews, 2009). In two final studies involving atten-