Empirical Status of the Cognitive Model 71
nonanxious controls had better identification on nonthreat than threat stimuli. How-
ever, others have failed to find speeded detection of previously presented threat versus
nonthreat words in panic disorder or PTSD (Lim & Kim, 2005; Lundh et al., 1999;
McNally & Amir, 1996). There is little evidence, then, for an implicit memory bias for
threat from stimulus identification priming studies.
Other Tests of Automatic Threat Evaluation
Amir et al. (1998a) employed a homograph paradigm to investigate activation and inhi-
bition of threat- relevant information in individuals with generalized social phobia (GSP)
and healthy controls. Individuals read short sentences that were followed by a single
word that either did or did not fit the meaning of the sentence. Individuals had to decide
whether or not the cue word matched the meaning of the sentence. As predicted, only
the GSP group showed a slower response to cue words that followed homographs with
a possible social threat meaning. This effect was only present at short sentence prim-
ing intervals, which suggests that GSP individuals were able to suppress or inhibit an
automatic evaluation of the sentence’s threat meaning when more effortful processing
was allowed.
Employing a memory task called release of proactive interference (RPI) that taps
into the semantic organization of memory, Heinrichs and Hofmann (2004) failed to find
the predicted memory effects of socially threatening information for high socially anx-
ious students. In fact, the opposite effect was found with the low social anxiety group
demonstrating a RPI effect for socially threatening words. In a study involving analysis
of eye movement to angry, happy, and neutral faces, Rohner (2004) was able to show
that individuals learned to avert their attention away from angry faces. In this experi-
ment, then, anxiety was related to an implicit memory for threat avoidance.
Finally, an experimental paradigm called the Implicit Association Test (IAT) has
been used to examine automatic memory-based associations between two concepts
(Greenwald, McGhee, & Schwartz, 1998). It is considered an index of implicit atti-
tudes because it is relatively uninfluenced by conscious controlled processes (Teachman
& Woody, 2004). In a study involving individuals highly fearful of snakes or spiders,
Teachman, Gregg, and Woody (2001) found significant differences in implicit negative
associations for snake versus spider attitudes across several semantic categories that
matched individuals’ fear concerns (Teachman & Woody, 2003; see de Jong, van den
Hout, Rietbrock, & Huijding, 2003, for negative findings of implicit associations for
spider cues in a group with high fear of spiders). Moreover, fear- related implicit associa-
tions were shown to change over the course of a three- session group exposure treatment
for phobias (Teachman & Woody, 2003).
Two studies have compared high and low socially anxious individuals on the IAT.
Tanner, Stopa, and de Houwer (2006) found that both high and low socially anxious
groups had positive implicit self- esteem as indicated by their reaction times to IAT word
classification. However, implicit self- esteem was significantly less positive in the high
social anxiety group, suggesting that a self- favoring effect was weaker in those with
high self- reported social anxiety. De Jong (2002) also concluded that high socially anx-
ious individuals have a weaker self- favoring bias, but his results suggested this was due
to significantly higher esteem associations for others. Although only a few IAT stud-