Vulnerability to Anxiety 119
characterize the anxiety disorders, but this research is still in its infancy and many
fundamental questions about the nature of schematic vulnerability in anxiety have not
been addressed.
INDUCED THREAT INTERPRETATIoN bIAS
It is now well established that a tendency to endorse threatening interpretations of ambig-
uous information is an important feature of the selective processing bias for threat that
characterizes anxiety (Mathews, 2006). However, demonstrating that threat- processing
bias, and its underlying schematic threat activation by extension, has causal influence
is more difficult because most of the research has been correlational or involved cross-
sectional research designs. Mathews and MacLeod (2002) note that evidence of dif-
ferential bias in anxious and nonanxious groups, reduction of threat bias with treat-
ment, or differential activation of bias in high and low trait anxious individuals after
a stressful event can not rule out a noncausal explanation such as the influence of a
third unidentified variable. Thus research showing that experimental manipulation of
interpretative bias through deliberate training conditions has a considerable impact on
emotion is strong empirical evidence for causality in evaluative processing of threat.
Furthermore, this research is important for cognitive vulnerability because it provides
evidence for a basic precondition of vulnerability: that biased information processing
has a causal effect on emotion.
The basic aim of induction procedures is to train volunteers to engage in selective
processing of new anxiety- relevant information and assess changes in subsequent anxi-
ety. Two effects are necessary to demonstrate. First, that training in differential process-
ing bias has been successful and generalizes to the processing of new information. And
second, an increase or decrease in threat- processing bias results in changes in level of
anxiety. A third question often addressed is whether there are individual differences
in susceptibility to threat-bias training that might suggest heightened vulnerability to
a n x iet y.
MacLeod and colleagues conducted a series of experiments on induced attentional
bias for threat in student volunteers. In the typical experiment individuals were randomly
assigned to an attentional threat training condition or the avoidance of threat in favor of
emotionally neutral cues (Mathews & MacLeod, 2002). In a series of unpublished pilot
experiments (see discussion in Mathews & MacLeod, 2002), MacLeod and colleagues
adapted the dot probe detection paradigm so that participants were randomly assigned
to 576 training trials in which the dot always appeared in the location of threatening or
neutral words. Analysis of 128 test trials revealed a significant training effect in which
participants trained to detect threat words were significantly faster at probe detection
after a threat word and slower to detect probes after a neutral word. This training effect
was replicated in another pilot experiment using happy and angry faces.
In their first major published study, MacLeod et al. (2002) reported on two studies
involving experimental manipulation of attentional bias. In the first experiment 64 non-
vulnerable students (trait anxiety scores in the middle range) were randomly assigned
to an “attend negative” training condition or an “attend neutral” condition. Training
involved 576 trials in which 50% of the word pair presentations were at a short exposure
interval (i.e., 20 milliseconds) and the other 50% were at a longer exposure duration
(i.e., 480 milliseconds). Ninety-six test trials were distributed throughout the training