Cognitive Therapy of Anxiety Disorders

Empirical Status of the Cognitive Model 67

that individuals with generalized social phobia had difficulty learning nonthreat inter-
pretations of ambiguous social information. Also, a psychophysiological study found
that combat veterans with PTSD were less expressive to emotionally positive standard
pictorial stimuli (i.e., lower zygomatic facial EMG response) after viewing a 10-minute
trauma videotape (Litz, Orsillo, Kaloupek, & Weathers, 2000; see Miller & Litz, 2004,
for failure to replicate). These findings suggest that diminished processing of nonthreat
or safety information may be evident in anxiety but this may only occur at the later
stage of strategic processing (see Derryberry & Reed, 2002). Also, the provision of
safety cues may have difficulty overriding the strong information- processing bias for
threat (i.e., Hayward, Ahmad, & Wardle, 1994) and there is even evidence that indi-
viduals with panic may show a recognition bias for “safe” facial expressions (Lundh,
Thulin, Czyzykow, & Öst, 1998).
At this point too few studies have investigated the processing of safety cues in
anxiety and so the empirical status of Hypothesis 2 cannot be determined. Clearly, stud-
ies are needed that directly compare the automatic and strategic processing of threat-
relevant and safety- relevant information in clinically anxious and nonanxious controls.
In addition it would be important to establish a relationship between diminished safety
cue processing as a mediator of safety- seeking behavior.

Low Anxiety: Enhanced Safety Signal Processing

Two outcomes are possible when investigating safety signal processing in the absence of
anxiety. It is possible that attention is drawn toward positive stimuli or safety cues so
that a positivity bias is evident in nonanxious states. An alternative outcome is that no
attentional bias occurs in low anxiety so that an evenhanded processing of threatening
and safety cues prevails.
At this point we know very little about the processing of safety- relevant informa-
tion in low anxiety states. In the original dot probe experiment MacLeod et al. (1986)
found that the nonanxious control group tended to shift their attention away from
threat words (see also Mogg & Bradley, 2002). However, this effect has not been rep-
licated in most subsequent studies (e.g., Mogg, Mathews, & Eysenck, 1992; Mogg,
Bradley, et al., 2004; Mogg et al., 2000). On the other hand, MacLeod and Rutherford
(1992) found that low trait- anxious students evidenced a significant reduction in color-
naming interference for threat words as their state anxiety level increased in a high stress

Early Processing Later Processing
HIGH aNXIETY Heightened Attention to Threat
Diminished Attention to Safety

X Delayed Threat Disengagement

Inadequate Safety Cue Processing



Increased Safety-Seeking Behavior

LOW aNXIETY Reduced Attention to Threat
Enhanced Attention to Safety

X Low Threat Engagement

Appropriate Safety Cue Processing



Safety-Seeking Behavior absent

figure 3.3. Proposed relation of threat and safety processing biases in high and low anxiety.