Empirical Status of the Cognitive Model 59
Bradley, 1999a, 2004; Wells & Matthews, 1994; Williams et al., 1997). Because human
attentional capacity is limited, some stimuli will capture attentional resources and oth-
ers will be ignored. The presence of an attentional bias for threat is expected to cause an
increased propensity to experience anxiety (McNally, 1999). Below we organized our
review of the attentional research around three types of experimentation; emotional
Stroop, dot probe detection, and stimulus identification.
Emotional Stroop
In order to experimentally investigate attentional bias in anxiety, clinical researchers
have borrowed and then modified various information- processing tasks from cognitive
experimental psychology. One of the most popular of these experimental paradigms
has been the emotional Stroop task. Based on the classic Stroop color- naming para-
digm (Stroop, 1935), participants are asked to name as quickly as possible the color of
emotionally threatening (e.g., “disease,” “cancer,” “embarrassed” “disaster,” “dirty,”
“inferior”) and nonthreatening (e.g., “upward,” “network,” “leisure,” “secure”) words
printed in blue, yellow, green, or red and to disregard the meaning of the word. Typi-
cally, anxious but not nonanxious individuals take longer to name the printed color of
threat words compared with nonthreat words (e.g., Bradley, Mogg, White, & Millar,
1995; Mathews & Klug, 1993; Mathews & MacLeod, 1985; Mogg, Mathews, & Wein-
man, 1989; Mogg, Bradley, Williams, & Mathews, 1993). This longer color- naming
latency suggests that anxious individuals exhibit preferential allocation of attention to
the threat meaning of the word (Mogg & Bradley, 2004). Thus the extent of interference
in color- naming response by the meaning of the word is assumed to reflect attentional
bias for threat.
The emotional Stroop threat interference effect has been found in all five of the anxi-
ety disorders discussed in this volume: panic disorder (e.g., Buckley, Blanchard, & Hick-
ling, 2002; Lim & Kim, 2005; Lundh, Wikström, Westerlund, & Öst, 1999; McNally,
Riemann, & Kim, 1990); OCD (e.g., Kyrios & Iob, 1998; Lavy, van Oppen, & van den
Hout, 1994); social phobia (e.g., Becker, Rinck, Margraf, & Roth, 2001; Hope, Rapee,
Heimberg, & Dombeck, 1990); PTSD (e.g., J. G. Beck, Freeman, Shipherd, Hamblen,
& Lackner, 2001; Bryant & Harvey, 1995); and GAD (e.g., Bradley et al., 1995; Mogg,
Bradley, Millar, & White, 1995). Moreover, threat interference effects significantly cor-
relate in the low to moderate range with state and symptom anxiety measures (e.g.,
MacLeod & Hagan, 1992; Mathews, Mogg, Kentish, & Eysenck, 1995; Spector, Pec-
knold, & Libman, 2003) and become more apparent as the threat stimulus intensity
increases from mild to severe intensity (Mogg & Bradley, 1998). In addition, the best
discrimination of attentional bias in high trait and nonclinically anxious individuals
versus low anxiety individuals might be with weak to moderately threatening cues in
which the nonanxious person would show no preferential bias for threat (Mathews &
Mackintosh, 1998).
The most consistent and robust interference effects are found with words that are
semantically related to the current emotional concerns of the anxious person (Mathews
& Klug, 1993); this content- specificity seems particularly pronounced in OCD, social
phobia, and PTSD (J. G. Beck et al., 2001; Becker et al., 2001; Buckley et al., 2002;
Foa, Ilai, McCarthy, Shoyer, & Murdock, 1993; Hope et al., 1990; Kyrios & Iob, 1998;
Lavy et al., 1994; Mattia, Heimberg, & Hope, 1993; Spector et al., 2003). However, the