120 COGNITIVE THEORY AND RESEARCH ON ANXIETY
trials. Thus half of the participants were trained to attend to the negative information
and the other half were trained to attend away from negative stimuli (attend to neutral
words). After the dot probe training all participants completed a stressful anagram
task. Analysis revealed that students in the negative training condition exhibited faster
dot probe detection to negative words in the test trails, whereas participants trained to
attend away from negative words exhibited a speeding effect to dot probes following
the neutral words. However, this training effect was only evident at the longer exposure
trials, indicating that differential bias was not preconscious. Furthermore, attentional
training had no immediate effect on mood, although after the anagram stress students
trained to attend away from negative information showed significantly lower elevations
in negative mood. The authors concluded that attentional threat avoidance training
may reduce vulnerability for negative emotional response to stress.
In a second replication study all training trials were conducted at a longer exposure
interval and emotional reactivity to stress was assessed before and after attentional
training (MacLeod et al., 2002). Analysis revealed that a differential training effect was
again achieved and that attentional training away from negative stimuli resulted in no
negative emotional response to the anagram stressor, whereas the group that had nega-
tive attentional training showed a pronounced negative emotional response to the stres-
sor. These differential effects were due to training because at baseline the groups did
not differ in showing elevations in negative mood to a preinduction baseline anagram
task. The authors concluded that attentional training modified the degree of emotional
response to a subsequent stressor. Thus the training had its greatest impact not on mood
directly but rather on affecting emotional vulnerability to stress.
Of greatest relevance to Hypothesis 12 are a series of published studies on inter-
pretative bias training. Grey and Mathews (2000) first investigated whether interpre-
tative bias for threat could be trained in volunteers with normal trait anxiety scores.
Individuals were randomly assigned to a threatening or a nonthreatening homograph
training condition in which volunteers were trained to complete a word fragment with a
threatening or nonthreatening homograph. In the first experiment, Grey and Mathews
(2000) found that threat training resulted in faster response for generating threat solu-
tions on 20 critical test items, and the biasing effect of threat training was found to
generalize to a lexical decision task in two further experiments. In a final study that
included an untrained control group, individuals exposed to homograph threat training
showed faster lexical decision for threat than the baseline group. These studies, then,
demonstrated that an interpretative threat bias for ambiguous stimuli can be trained in
nonvulnerable individuals.
Mathews and Mackintosh (2000) conducted five experiments in which interpreta-
tive bias training involved making a negative (threatening) or positive (nonthreaten-
ing) interpretation to a short description of an ambiguous social situation. Sixty-four
descriptions were presented with each one followed by a word fragment that matched
a threatening or nonthreatening interpretation. In the first experiment, volunteers ran-
domly assigned to interpretative threat training were faster at completing negative probe
word fragments and gave higher recognition ratings to threatening interpretations of the
ambiguous descriptions. Furthermore, there was a direct effect on mood, with the threat
group reporting an increase in anxiety after training, although this mood effect was
not replicated in the second experiment. In the fourth experiment interpretative threat
training did result in an increase in state anxiety but its effects were shown to dissipate