Empirical Status of the Cognitive Model 63
ized by a significant avoidance of emotional faces (Chen, Ehlers, Clark, & Mansell,
2002; Mansell, Clark, Ehlers, & Chen, 1999). One possibility is that social phobia
involves an initial attentional vigilance for social evaluation followed by an avoidance
of social threat stimuli once more elaborative processing occurs (Chen et al., 2002; see
findings by Mogg et al., 2004).
Dot probe experiments have been used to investigate cognitive vulnerability to
anxiety by determining if high trait anxiety is characterized by speeded detection of
threat stimuli. The most consistent finding is that high trait- anxious individuals exhibit
quicker probe detection to threatening words or faces compared to low-trait anxious
individuals, especially at shorter exposure intervals (Bradley, Mogg, Falla, & Hamilton,
1998; Mogg & Bradley, 1999b; Mogg, Bradley, Miles, & Dixon, 2004; Mogg et al.,
2000, Experiment 2). Other studies, however, have reported entirely negative findings
for trait anxiety, concluding that hypervigilance for threat was due to state anxiety (or
immediate stress) either alone or in interaction with trait anxiety (e.g., Bradley, Mogg,
& Millar, 2000; Mogg et al., 1990).
It is likely that these inconsistent findings occur because attentional bias in anxiety
involves both hypervigilance and avoidance of threat stimuli (Mathews & Mackintosh,
1998; Mogg & Bradley, 1998). Generally hypervigilance for threat has been more appar-
ent during brief exposures when preconscious automatic processes predominate and at
higher levels of threat intensity. Avoidance of threat stimuli more likely occurs at longer
exposure intervals when more elaborative processing comes into effect and with mildly
threatening stimuli. This vigilance- avoidance pattern may be particularly evident in
specific fears, with high trait anxiety characterized by initial vigilance for threat with-
out subsequent avoidance (Mogg et al., 2004; see Rohner, 2002, for contrary findings).
However, Rohner (2002) did not confirm this distinction between anxiety and fear.
In a study that directly examined the effects of varying levels of threat intensity,
Wilson and MacLeod (2003) compared probe detection times of high and low trait-
anxious students to very low, low, moderate, high, and very high anger facial expres-
sions paired with a neutral face. All participants failed to show attentional bias to the
very low threat stimuli, attentional avoidance of mildly threatening faces, and atten-
tional vigilance at the most intensely threatening stimuli. Interestingly, group differ-
ences in attentional deployment were only apparent with the moderately threatening
faces where only high trait- anxious group showed quicker detection of threatening than
neutral faces. Others have also found that attentional bias for threat increases with stim-
ulus threat value (Mogg et al., 2004; Mogg et al., 2000). In a more recent study high
trait- anxious individuals showed clear evidence of facilitated attention and impaired
disengagement from high threat at 100 milliseconds but attentional avoidance at 200
or 500 milliseconds (Koster, Crombez, Verschuere, Van Damme, & Wiersema, 2006).
Finally, in an attentional training experiment by MacLeod, Rutherford, Campbell, Ebs-
worthy, and Holker (2002), students given training to attend away from negative words
had reduced emotional response to a stress induction compared to students trained to
attend to negative probes. This indicates that attentional bias can have a causal impact
on emotional response.
In summary both semantic (words) and visual (faces) dot probe detection research
provides the strongest experimental evidence for an automatic, preconscious hypervigi-
lance for threat. Hypervigilance for threat is more likely when conscious elaborative
processing is restricted (shorter exposures with reduced awareness), when threat stimuli