Many previous information processing studies have noted that people wi
th severe anxiety selectively attend to threat stimuli. The experiment
reported here tests the hypothesis that, when real threat stimuli are
used instead of semantic stimuli, attention may be divided between th
reat and safety. Spider phobics and controls participated in a reactio
n time experiment, in which the target stimulus (a light) was programm
ed to randomly occur either by the only door to the room or by the wal
l opposite that door. They were asked to press a response button as so
on as they detected the light coming on by the wall or by the door. Ne
xt, for half the participants, a live Zebra Tarantula was placed next
to the stimulus light, either by the door (threat and safety/escape co
incide) or by the wall (threat and safety/escape divided), and the exp
eriment repeated. The rest of the participants repeated the baseline c
ondition to control for practice effects. Response latencies to the li
ght stimuli were analysed. Results suggest that spider phobics (but no
t controls) were faster to respond to the light when it occurred in th
e location where the threat and the escape stimuli coincided, (when th
e spider was by the door), than when it occurred in the location where
the threat and the escape stimuli were divided (when the spider was b
y the wall). These results suggest that phobics may allocate attention
not only to threat but also to safety. It is proposed that such effec
ts may be less detectable, or absent, when the threat stimuli are sema
ntic or symbolic because participants discriminate between threat and
its symbolic representation. (C) 1998 Elsevier Science Ltd. All rights
reserved.