70 COGNITIVE THEORY AND RESEARCH ON ANXIETY
Studies employing word stem completion have produced mixed results that can only
be interpreted as weak evidence of implicit memory in anxiety. In some studies clinically
anxious patients or high trait- anxious individuals have generated more threat word
completions, which suggests an implicit memory for threat (e.g., Cloitre, Shear, Can-
cienne, & Zeitlin, 1994; Eysenck & Byrne, 1994; Mathews, Mogg, May, & Eysenck,
1989; Richards & French, 1991). However, other studies have failed to find an implicit
threat bias (e.g., Baños, Medina, & Pascual, 2001; Lundh & Öst, 1997; Rapee, McCal-
lum, Melville, Ravenscroft, & Rodney, 1994). McNally (1995) considers word stem
completion a poor test of implicit memory in anxiety because it is strongly affected by
the physical attributes of words rather than by their meaning.
Lexical Decision Tasks
In lexical decision tasks individuals are shown a list of mixed- valence words in which
some may be anxiety- relevant, some depression- relevant, and others neutral. After a
filler task individuals are shown a second list of words that will contain some of the
“old” words, some “new” words, and also some nonword distractors (e.g., eupine,
mard, flidge). Participants are told to indicate as quickly as possible whether the stimu-
lus is a “word” or a “nonword.” Quicker lexical decision for previously presented words
suggests an implicit memory priming effect. In anxiety we would predict quicker lexi-
cal decision for previously presented threat than nonthreat words. In this experimental
paradigm priming effects can be investigated subliminally or supraliminally depending
on whether the first exposure occurs above or below the threshold of awareness.
In two lexical decision experiments, Bradley and colleagues (Bradley, Mogg, &
Williams, 1994, 1995) failed to find evidence of an anxiety- congruent implicit memory
bias in either subliminal or supraliminal priming conditions (see also Foa, Amir, Ger-
shuny, Molnar, & Kozak, 1997, for negative results). Amir and colleagues utilized a
more sensitive measure of automatic encoding of the meaning of information by requir-
ing perceptual rather than word judgments to more complex stimuli. In two studies
socially anxious individuals exhibited a significant auditory or visual preferential rating
for previously presented threat stimuli that was interpreted as indicating an implicit
memory- priming effect for social threat stimuli (Amir, Bower, Briks, & Freshman,
2003; Amir, Foa, & Coles, 2000). However, Rinck and Becker (2005) failed to find an
implicit memory bias for socially threatening words in an anagram task (i.e., identify
the word from scrambled letters). Thus findings from standard lexical decision experi-
ments or more recent perceptually oriented priming studies have not been particularly
supportive of implicit (automatic) threat evaluation in anxiety.
Primed Stimulus Identification Tasks
A number of studies have investigated implicit memory bias by determining if anx-
ious individuals show more accurate detection of briefly presented threatening words
(stimuli) as a result of prior exposure to threat and nonthreat stimuli. MacLeod and
McLaughlin (1995) found an implicit memory bias for threat in GAD patients com-
pared to nonanxious controls based on a tachistoscopic word identification task. The
GAD group exhibited better detection of old threat than nonthreat words, whereas