60 COGNITIVE THEORY AND RESEARCH ON ANXIETY
attentional bias in GAD and, to a lesser extent, panic may be more emotionally oriented
and thus elicited by any negative emotional stimuli, and in some cases, even positive
information (e.g., Becker et al., 2001; Bradley, Mogg, White, & Millar, 1995; Buckley
et al., 2002; Lim & Kim, 2005; Lundh et al., 1999; Martin, Williams, & Clark, 1991;
McNally et al., 1994; Mogg et al., 1993; Mogg, Bradley, Millar, & White, 1995).
To investigate the automaticity of attentional threat bias, researchers modified the
emotional Stroop task to include subliminal (below conscious awareness) and supral-
iminal (above conscious awareness) conditions. In these studies individual threat and
nonthreat words are presented very briefly (20 milliseconds or less ) followed by a mask,
which usually involves a string of random letters presented in the same location as the
word. In some studies participants are asked to name the color of the word whereas in
other studies they are asked to name the color of the background of the word. In the
supraliminal condition the words remain unmasked on the screen until a color- naming
response is made. Figure 3.1 provides an illustration of the modified emotional Stroop
task.
In a number of studies anxious patients exhibited significantly slower color- naming
latencies to subliminal threat words, suggesting that selective attention to threat occurs
at the automatic preconscious level (e.g., Bradley et al., 1995; Kyrios & Iob, 1998; Lundh
et al., 1999; Mogg et al., 1993). Since this threat interference effect was found on both
subliminal and supraliminal trials within the same study, it suggests that attentional
bias for threat involves both automatic and elaborative cognitive processes (e.g., Bradley
et al., 1995; Lundh et al., 1999; Mogg et al., 1993).
Another important issue addressed in the emotional Stroop research is the relation
of attentional threat bias to state and trait anxiety. MacLeod and Rutherford (1992)
reported that automatic attentional threat bias is most influenced by an interaction
between state and trait anxiety. They compared nonclinical high and low trait- anxious
students on a modified emotional Stroop task and found that the high trait- anxious
students under stress (tested 1 week before exams) showed greater subliminal Stroop
interference for threat, whereas stress did not enhance threat interference for the low
trait- anxious students. In the supraliminal condition both high and low trait- anxious
students showed intentional avoidance of threat words. Other studies have also found
that increased stress and arousal are associated with greater attentional bias, especially
in high trait or fearful individuals (Chen, Lewin, & Craske, 1996; Mogg, Mathews,
Bird, & MacGregor- Morris, 1990; Richards, French, Johnson, Naparstek, & Williams,
1992; see McNally, Riemann, Louro, Lukach, & Kim, 1992, for contrary findings).
However, the effects of state and trait anxiety on attentional bias may be more compli-
cated than first thought. High trait anxious individuals exhibit a preconscious, auto-
matic attentional bias for threat, but unlike clinical samples, this attentional bias may be
sensitive to negative valence more generally rather than to specific threat content (e.g.,
Fox, 1993; Mogg & Marden, 1990). In addition, elevated state anxiety may lead to
greater automatic threat bias in high trait anxiety individuals (interaction effect), but at
the more elaborative, strategic level, stress may have independent effects on attentional
threat bias. MacLeod and Hagan (1992) suggested that nonclinical individuals may be
able to strategically modify their automatic threat bias, thereby eliminating any differ-
ential interference effects in the supraliminal condition. Anxious patients, on the other
hand, may fail to strategically modify their preconscious attentional threat bias so that
threat differences continue to emerge at the elaborative stage of information processing.