Prrson.
individ.
DfJ
Vol. 23, No. 4, pp. 653663, 1997
cc”, 1997Elsevier Science Ltd.
Pergamon PII: SO19143869(97)00071-8
All rights reserved Printed in Great Britain 0191-8869/97 $17.00+0.00
A MEMORY BIAS FOR THREAT IN HIGH-TRAIT ANXIETY John Reidy’* and Anne Richards2 ‘Department of Psychology, University of East London, Romford Road, London El5 4LZ and ‘Department of Psychology, Birkbeck College, University of East London, Romford Road, London El 5 4LZ, England i Received 22 Oclober 1996; receivedfor publication 22 April 1997)
Summary-The suggestion that high-trait anxiety is associated with a memory bias for threatening information has so far received little empirical support. Two studies are reported which were designed to test this prediction and replicate findings recently reported by the current authors. In both studies subjects were required to encode and recall a list containing positive, threatening and non-threatening negative words. Analyses revealed that whereas the low-trait anxious subjects recalled equivalent numbers of threatening and non-threatening words, the high-trait anxious subjects recalled significantly more of the threatening words. Further analyses suggested that the tendency for anxious individuals to recall more threatening than non-threatening words was not simply due to more of these words being endorsed as selfdescriptive by these subjects. Additionally, both high- and low-trait anxious subjects showed a strong memory bias for positive rather than negative words, a finding which is consistent with previous research. The theoretical and methodological implications of these findings are discussed. 0 1997 Elsevier Science Ltd
INTRODUCTION
For many years cognitive psychologists were criticized for failing to consider the relationship between cognition and emotion (Bower, 1981; Eysenck, 1984). These criticisms, fortunately, are no longer valid. In the last decade, largely as a result of the schema theory of Beck and his associates (Beck, 1976; Beck & Emery, 1985) and the semantic network theory of Bower (1981), there has been a remarkable increase in research into the cognitive aspects of emotion. These theories predict that emotional experience is accompanied by increased processing of emotionally congruent information. For example, anxiety should be associated with a memory bias for anxiety-related information and depression associated with a memory bias for depression-related information. There is now extensive evidence supporting the latter of these predictions (see Blaney, 1986, for a review). However, the corresponding research into anxiety has led to somewhat equivocal findings. Consistent with the prediction, several studies have shown that anxiety is associated with superior memory for threatening rather than non-threatening information (Norton, Schaefer, Cox, Dorward & Wozney, 1988; Breck & Smith, 1983; McNally, Foa & Donnell, 1989; Reidy & Richards, 1997). However, other research has suggested no memory bias at all with anxious subjects (Mogg & Mathews, 1990; Richards & French, 1991; Dalgleish, 1994; Nugent & Mineka, 1994; Bradley, Mogg & Williams, 1994). Perhaps more surprising though, is the research which has shown anxiety to be associated with poorer memory for threatening information (Mogg, Mathews & Weinman. 1987; Watts & Dalgleish, 1991; Watts, Trezise & Sharrock, 1986). How can such an inconsistent pattern of results be explained? Sanz (1996) pointed out that the anxiety and memory research has used a variety of methodologies and many different types of anxiety groups. Therefore, such inconsistencies are perhaps not surprising. It may be a mistake, however, to take such a global view of the anxiety research. Zinbarg and Barlow (1996) have argued that whereas all anxiety disorders do have shared characteristics, they also have elements that are specific to each disorder. It might be that differences between the various forms of anxiety are the reason for the contradictory research findings. Concentrating on the research at the level of specific forms of anxiety may, therefore, be more appropriate (Cloitre, Cancienne, Heimberg, Holt & Liebowitz, 1995). For example, there is much research which has shown that panic disorder patients have a bias for anxiety related information (McNally et al., 1989; Norton et al., 1988). The research
*To whom all correspondence should be addressed. 653
654
John Reidy and Anne Richards
into spider phobia, however, suggests that phobic individuals have poorer memories for spiderrelated information than for neutral information (Watts et al., 1986; Watts & Dalgleish, 1991). Similarly, Eysenck (1992) has argued that generalized anxiety disorder (GAD) patients avoid deep processing of threatening information and as a result have poorer memories for such information. This is best illustrated by the research of Mogg et al. (1987), who found that GAD patients had a greater tendency than normal controls to recall non-threatening rather than threatening words. These findings are consistent with the model proposed by Williams, Watts, MacLeod and Mathews (1988), which asserts that whereas anxious individuals direct attentional resources towards threat, they tend to direct elaborative processing resources away from threat. Mogg et al. suggested that such a vigilance-avoidance pattern of processing may actually help to maintain the high levels of anxiety observed in GAD patients. There are many indications that high-trait anxiety predisposes to generalized anxiety disorder (see Eysenck, 1992). It might therefore be expected that high-trait anxiety is associated with a similar memory bias to that observed with GAD patients. In fact, most of the research fails to uphold this expectation. For example, Dalgleish (1994) tested memory using free-recall and recognition tests and found no evidence of a bias associated with high-trait anxiety. Richards and French (1991) tested high and low anxious subjects using a free-recall task, and also found no bias with the high anxious subjects. Similarly, although Nugent and Mineka (1994) found some supportive evidence in their first experiment, this was not repeated in their second study. They concluded that there was no convincing evidence for such a memory bias. Reidy and Richards (1997) argued that one of the reasons for the contradictory findings in the anxiety and memory literature is the inconsistent choice of stimuli found in such research. Where consistent results have been found, the materials used tended to be related to the form of anxiety under investigation. For example, the research by McNally et al. (1989) found that panic patients had a memory bias only for information that was specifically related to anxiety. It was argued that, as panic patients have a specific fear of anxiety, information related to anxiety is particularly important to them and therefore more memorable. Similar results have been reported by Norton et al. (1988), using non-clinical panickers. The spider phobia research showed that phobics have poorer memories for information that was specifically related to spiders (Watts et al., 1986; Watts & Dalgleish, 1991). However, Watts and Coyle (1993) found that whereas spider phobics’ memory was impaired for anxiety response words (e.g. nervous), it was comparatively enhanced for phobic stimulus words (e.g. fangs). Finally, Zeitlin & McNally (1991) found that, relative to controls, Vietnam veterans suffering from post-traumatic stress disorder (PTSD) had a memory bias for words relating to combat. These findings indicate that extreme care should be taken in the selection of stimuli when investigating memory effects. One problem with stimulus selection for research into high-trait anxiety is the identification of the most relevant type of information. The study by Mogg ef al. (1987) indicated that the distinction between threatening words (e.g. humiliated) and non-threatening words (e.g. bored) is appropriate for research involving GAD patients. Reidy and Richards (1997) suggested that if high-trait anxiety predisposes to GAD, then a similar distinction might be useful for research into high-trait anxiety. Using a procedure almost identical to that used by Mogg et al., and exercising stricter control over word selection and the positioning of words within the word list, Reidy and Richards found that high-trait anxious subjects recalled more threatening than non-threatening words, This finding indicates that the distinction between threatening and nonthreatening words is a useful one with these subject groups. One problem highlighted by the two studies reported by Reidy and Richards (1997) was the tendency for subjects to recall a disproportionate number of words from the final part of the word list. Over 50% of all words recalled had been presented in the final quarter of the word list. They argued that such a large influence may be problematic in experiments where the positioning of words in the word list was not carefully controlled. This was certainly the case with Reidy and Richards (1997, Study 1) and may have been a problem not recognised in the Mogg et al. (1987) study. Therefore, in order to further investigate the recall bias reported by Reidy and Richards (Study 2) and to reduce the influence of the latter words in the word list the following two experiments were conducted. They both employed similar procedures to that used by Mogg et al. (1987) and Reidy and Richards, however, the interval between list presentation and recall was increased to
Anxiety and memory
6.55
reduce the disproportionate recall from the latter part of the world list. Bjork and Whitten (1974) suggested that the superior level of recall of words from the latter part of a word list will by reduced if the interval between list presentation and recall is significantly greater than the inter-word interval. Therefore, in the following experiments, the length of time spent on the filler task was increased from 20 seconds to three minutes. Consistent with the findings of Reidy and Richards, it was predicted that the high-trait anxious subjects would have a recall bias for threatening rather than non-threatening negative words.
STUDY
1
Method Subjects. Thirty-two subjects took part in the experiment. The majority were students recruited from the university precinct who were paid for their participation. Sixteen subjects (ten males) with a mean age of 23.60 (SD=6.80) were classified as high-trait anxious (scoring above 40 on the Spielberger State and Trait Anxiety Inventory, STAI; Spielberger, Gorsuch, Lushene, Vagg & Jacobs, 1983) and 16 subjects (nine males) with a mean age of 23.60 (SD= 5.01) were classified as low-trait anxious (scoring 40 or below on the STAI). Materials. The materials and list structure were the same as those used by Reidy and Richards (1997, Study 2). There were 20 threatening words (e.g. Mutilated), 20 non-threatening words (e.g. Bored) and 40 positive words (e.g. Happy). These 80 words were divided into two lists, each containing 20 positive, 10 threatening and 10 non-threatening words. One list would be presented with self-referencing instructions, the other with instructions referring to Terry Wogan (a popular TV presenter from Ireland who had a prime-time chat show at the time the study was conducted). The total set of 80 words was also divided into four subsets of 20 words, each containing ten positive, five threatening and five non-threatening words. The words in each of these sub-lists were randomly arranged and four different sub-list orders created, so that each sub-list appeared in every position of the word list overall. Thus, in the four list orders created, list A would appear in positions 1, 2, 3 and 4, as would sub-lists B, C and D. These four lists were recorded onto audiotape for presentation to subjects (one word every ten seconds) and were carefully rotated across subject groups. Procedure. The procedure was essentially the same as that used by Mogg et al. (1987) and Reidy and Richards (1997). Subjects were tested individually and told that the study was designed to investigate the way people make judgements about words. They were informed that they would be presented with a number of words on audiotape and that they should decide whether or not the words described the person indicated by the instructions following each word (‘You’ or ‘TV’, where TV referred to Terry Wogan). They were told to inform the experimenter of their decision for each word by pointing to the relevant ‘Yes’ or ‘No’ response cards set up in front of them. The subjects were then given a short practice task, consisting of eight neutral words (e.g. tall), which was followed by the presentation of the 80 experimental words. The subjects’ responses were recorded by the experimenter. Upon completion of the encoding task the subjects were given a three minute filler task consisting of cancelling the vowels from a sheet of random letters. They were then provided with a blank sheet of paper and given six minutes to recall as many of the experimental words as possible. Finally, the subjects completed the STAI, the Beck Depression Inventory (BDI: Beck, Ward, Mendelson, Mock & Erbaugh, 1961); and the Mill Hill Vocabulary Scale Synonym Selection test (MHV: Raven, Court & Raven, 1969). Results Subject characteristics. Table 1 shows the means and standard deviations of the subjects’ questionnaire scores. This shows that there was a significant difference between the two anxiety groups in their state anxiety scores (t(30) = 1.93, p < 0.05), trait anxiety scores (t(30) = 9.88, p
656
John Reidy and Anne Richards Table 1. Means and standard deviations of questionnaire scores for high- and low-trait anxious subiects in Studv I Low-trait
STAI: Trait’ State BDI’ MHV Age
High-trait
Mean
SD
Mean
SD
35.81 34.3 I 4.16 19.13 23.60
3.17 5.91 3.02 4.37 5.01
47.94 39.19 Il.78 20.75 23.60
3.75 8.17 5.56 4.82 6.80
‘Groups differ at ptO.OO1.
to a reduction in recall of words from the latter part of the word list. The proportion of words recalled from each block were: 0.19,0.20,0.26, and 0.35. Thus, 35% of all words recalled were from the final block of 20 words in the list. Although this is higher than would be expected by chance, it is considerably lower than reported by Reidy and Richards (1997), where 50% of all words recalled came from this final block of 20. Thus, the current experiment has been at least partially successful in reducing the influence of the words in this latter part of the word list. Recall analysis. Analysis of the words used in the current study revealed that the threat and nonthreat words differed significantly in terms of their emotionality ratings (as indicated by the norms of Reidy, 1994; t(38)= 5.90, p
Negative words
Non-emotional
Threatening
Non-threatening
Self
Other
Self
Other
0.96
I .63 1.15
0.69 0.79
1.50 I .03
0.63 0.89
I.81 I .47
2.69 I .45
1.50 1.56
I .75
0.8 I I .05
Self
Other
Self
Other
Low-trait: Mean SD
3.25 2.30
I .25 1.00
2.44 I .26
High-trait: Mean SD
3.56 1.59
1.31 I .45
3.44 I .3?
I.44
1.61
Anxiety and memory
657
Finally, the analysis revealed a significant interaction between anxiety, valence and emotionality (F( 1,30) = 4.17, p =O.OS). Table 2 suggests that whereas the anxious and non-anxious subjects have similar patterns of recall of emotional and non-emotional positive words, there is a difference in the pattern of recall for negative words. This pattern of negative word recall is illustrated in Fig. 1. Simple interaction analyses showed that the interaction between anxiety and emotionality for (1.30) = 0.72, ns), however, the same interaction for positive words was clearly non-significant (FComp negative words just failed to reach significance (Fcomp(1,30) = 3.14, p = 0.087). Planned simple effects analysis on this latter interaction revealed that for the low-trait anxious subjects there was no difference in the number of threatening and non-threatening words recalled (Fcomp(1,30) = 0.11, ns; means of 2.32 and 2.13 respectively for threatening and non-threatening words), whereas the high-trait anxious subjects recalled significantly more threatening than non-threatening words (F,,,,(l,30) = 8.01, ~~0.01; means of 4.19 and 2.56 respectively). It appears, therefore, that the high-trait anxious subjects have a memory bias for threatening rather than non-threatening words. Bias scores. For each subject, bias scores were calculated which take account of the total number of words recalled (see Mogg et al., 1987). The number of negative words recalled by each subject was subtracted from the number of positive words recalled and this divided by the total number of words recalled. The resulting ratio, ranging from + 1 to - 1, gives an indication of the degree of bias in recall, a positive score indicates a bias for positive words and a negative score indicates a bias for negative words. There was no significant difference between the high and low anxious subjects in terms of this bias score (t(30) = 0.68, ns; means of 0.29 and 0.23 respectively for low and high anxious subjects). A similar bias score was calculated for the threatening and non-threatening words. The number of threatening words recalled was subtracted from the number of non-threatening words recalled and then divided by the total number of negative words recalled. A positive score suggests a bias for non-threatening words, whereas, a negative score suggests a bias for threatening words. There was a significant difference between the high and low anxious subjects in terms of this threat bias score (t(30) = 1.91, p < 0.05, one-tailed; means of 0.05 and - 0.32 respectively for low and high anxious subjects). Correlations. As the anxiety groups differed in terms of state anxiety and depression as well as trait anxiety, Pearson correlation coefficients were calculated between the two bias scores described
6r
* K l
\ +------* l Low anxious 0 High anxious
Threat
Non-threat
Experiment
1
Threat
Non-threat
Experiment
2
Type of word recalled in Experiments 1 and 2. Fig. 1. Mean number of threat and non-threat words recalled by high- and low-trait anxious subjects in Experiments 1 and 2.
658
John Reidy and Anne Richards
above and the state and trait anxiety (STAI) and the depression (BDI) scores. These revealed only one significant correlation and that was between trait anxiety scores and the threat bias scores (r = - 0.3 1, N = 32, p < 0.05 one-tailed). The correlations between threat bias scores and state anxiety (r = -0.15, N= 32, ns) and depression score (r = -0.21, N= 32, ns) were not significant. This suggests that trait anxiety rather than state anxiety or depression was the most important variable underlying the different patterns of recall. MacLeod (1990) has argued that in studies using self-referencing there is the possibility that the superior recall of anxiety-related words by anxious subjects reflects the fact that these individuals rated more of these words as self-descriptive than the low anxious subjects. To investigate this possibility correlation coefficients were calculated for the number of each type of word endorsed as self-descriptive in the encoding task (‘Yes’ responses) and the number of these words recalled. There was one significant correlation and that was for the non-emotional positive words (r =0.24, N = 32, p < 0.05). The other correlations were non-significant (r = 0.11, 0.20, 0.08 respectively for emotional positive, threatening and non-threatening words; N=32, ns). These suggest that the superior recall of threatening words by the high anxious subjects was not simply the result of more of these words being self-referenced. Bias scores (similar to those calculated above for the number of words recalled) were also calculated which took account of the number of negative words endorsed as self-descriptive. Here a positive ratio indicates a bias towards endorsing non-threatening words as self-descriptive and a negative ratio indicates a bias towards endorsing threatening words. A r-test revealed that there was no significant difference between the high and low anxious subjects in terms of this bias score (t = 0.19, df = 30, ns; means of 0.60 and 0.57 respectively for the low and high anxious subjects). These bias scores show that both groups have a bias towards endorsing nonthreatening words as self-descriptive. Discussion The findings from the current experiment provide support for the prediction that high-trait anxiety is associated with a memory bias for threatening words. However, given that the interaction between anxiety and emotionality for the negative words just failed to reach significance such support for the prediction should be treated with caution. Taken in conjunction with the findings of Reidy and Richards (1997) the current results are promising. A similar pattern of recall emerged in these studies, which supports the proposal of Eysenck (1992) that high-trait anxiety is associated with a recall bias for threatening information. These findings also demonstrate that the distinction between threatening and non-threatening negative words is a useful one for research with these anxiety groups. Furthermore, the finding that all subjects had a recall bias for positive rather than negative words adds credibility to the suggestion by Reidy and Richards that such a pattern of recall may mask genuine anxiety related differences, if these differences are exclusively related to negative and not to positive information. STUDY
2
In view of the caution expressed above a further study was conducted which used essentially the same procedure, but had one important modification. Instead of using both self- and otherreferenced stimuli, all the words were referenced to the self. The reasons for this change are twofold: first, there was no interaction between reference condition and anxiety group in either the first study or in the Reidy and Richards (1997) studies; and second, to try to encourage higher levels of recall, particularly of the negative words. Much of the previous research using self-referencing procedures has shown that self-referenced words are better recalled than other-referenced words (e.g. Martin, Ward & Clark, 1983; Mogg et al., 1987). Therefore, to avoid the low levels of recall exhibited by some subjects in Study 1, all the words were self-referenced. As in Study 1, it was predicted that high-trait anxiety would be associated with a recall bias for threatening rather than non-threatening negative words. Method Subjects. Thirty-two subjects took part in the study; the majority were students recruited from the university precinct who were paid for their participation. As the allocation of subjects to list
Anxiety and memory
659
Table 3. Means and standard deviations of questionnaire scores for high- and low-trait anxious subjects in Study 2 Low-trait
STAI: Trait” State’ BDI” MHV Age
High-trait
Mean
SD
Mean
SD
33.08 30.5 5.58 20.83 26.17
6.19 Il.21 3.46 5.65 16.99
49.75 40.13 I I .47 19.44 21.40
5.22 8.23 5.83 4.56 3.66
‘Groups differ at p < 0.05; “Groups differ at p < 0.0 I.
number had to be carried out before level of trait anxiety was known, there were unequal numbers of high-trait anxious subjects in each list condition. Therefore, four of the high-trait subjects tested were excluded from the analyses to ensure that within each anxiety group, the word lists were fully balanced. One of the high-trait anxious subjects was excluded because they were an outlier on all memory indices, the remaining three were randomly selected for exclusion. Thus, sixteen subjects (8 males) with a mean age of 21.4 (SD= 3.66) were classified as high-trait anxious and 12 subjects (four males) with a mean age of 26.17 (SD= 16.99) were classified as low-trait anxious. Materials and procedure. The words used in this study were the same as those used for Study 1 and were structured in the same way. However, instead of half of the words having instructions referring to Terry Wogan, all words were self-referenced. The procedure was also the same as for Study 1, except that the subjects were not presented with referencing instructions on the audiotape. They were simply told, before the list was presented, to reference all of the words to themselves. Results Subject characteristics. The means and standard deviations of the high- and low-trait anxiety groups for their anxiety (state and trait), depression, and MHV scores and age are shown in Table 3. This shows that the two groups differed significantly in terms of their state anxiety (t(26) = 2.63, p < 0.05) trait anxiety (t(26) = 7.72, p < 0.001) and depression scores (t(26) = 3.33, p < 0.01). They did not, however, differ in terms of age (t(26) = 0.95, ns) or their MHV scores (t(26) =0.72, ns). Pattern of recall. The change in procedure to using only self-referencing instructions has, as predicted, encouraged greater levels of recall. The mean number of words recalled by each subject in the current study was 19.07, which was considerably greater than those recalled in the first study (14.85). The proportions of words recalled from each of the four blocks of 20 words in the word list were 0.21, 0.22, 0.24, and 0.33. As in Study 1, the current experiment has reduced the proportion of words recalled from the final block of twenty in the list to only 33%. Thus, increasing the interval between encoding and recall appears to be an effective means of reducing the influence of words presented in the final part of the word list. RecaN analysis. Consistent with Study 1, the emotionality of the words was included as a factor in the analyses. Thus, the positive words matched with the threatening words were classed as emotional and those matched with the non-threatening words classed as non-emotional. The mean number of words recalled by the high- and low-trait anxious subjects are presented in Table 4. Table 4. Means and standard deviations for the number of emotional and non-emotional oositive words and threatenine and non-threatenine words recalled in Studv 2 High-trait
Low-trait Mean
SD
MeU
SD
Positive: Emotional Non-emotional
7.92 5.67
2.54 2.10
6.44 5.44
2.56 2.19
Negative: Threatening Non-threatening
3.67 3.83
1.61 I .40
4.00 2.06
I .93 1.12
Word type
660
John Reidy and Anne Richards
An analysis of variance with one between subjects factor of anxiety (high vs low-trait) and two within-subjects factors of valence (positive vs negative) and emotionality (emotional vs nonemotional) was carried out on the raw recall scores. This revealed a significant main effect of valence (F( 1,26) = 77.86, p
Study2 has produced a remarkably similar pattern of results to both Study 1 and Reidy and Richards (1997), showing that the high-trait anxious group had a significantly greater bias towards the recall of threatening words than the low-trait group. Additionally, recall of words from the final portion of the word list was consistent with that seen in Study 1, in that only 33% of all words recalled came from this part of the word list. This compares favourably with both of the studies reported by Reidy and Richards (1997), where 50% of words recalled came from the final part of
Anxiety and memory
661
the list. Increasing the interval between list presentation and recall is, therefore, an effective means of reducing the importance of these final words.
GENERAL
DISCUSSION
The results from the two studies presented here are consistent with those reported by Reidy and Richards (1997) and represent the only consistent evidence of a memory bias found with high-trait anxious subjects. Whereas these findings support the theories of Beck and Emery (1985) and Bower (1981), they fail to support the model outlined by Williams et al. (1988). Williams et al. proposed that anxious individuals avoid deep processing of threat and as a result have a memory bias for non-threatening rather than threatening information. The current research suggests that this is not the case. It is possible that the avoidance strategy suggested by Williams et al. is only applicable to GAD patients and not to high-trait anxious individuals, an argument which has recently been proposed by Eysenck (1992). Perhaps future research could directly compare the memory performance of GAD patients and non-patients with high levels of trait anxiety. The current findings also support the suggestion by Reidy and Richards (1997) that the distinction between threatening and non-threatening words is relevant for high-trait anxiety as well as for generalized anxiety disorder. However, the question remains, why should the distinction between threatening and non-threatening negative words be so important? One could argue that the main difference between these types of words is that the threatening words are simply more emotional than the non-threatening words and are thus more memorable. Indeed, Martin, Williams and Clark (1991) have argued that emotionality is an important confound in much of the cognition and emotion literature (see also Russo, Patterson, Robertson, Stevenson & Upward, 1996). Such a possibility was examined in the current studies by including emotionality as a factor in the analyses. Given that the pattern of recall of positive words was not the same as that for the negative words, it is unlikely that the superior recall of threatening words was the result of these simply being more emotional than the non-threatening words. Alternatively, it could be that the recall of words is simply related to the number of times such words are endorsed as self-descriptive in the encoding task. This has been argued as an important confound in this type of research by MacLeod (1990). Although, the relationship between selfreferencing and recall is not entirely straightforward, this possibility was examined in the current research by correlational analysis and by calculating bias scores which took account of the number of threatening and non-threatening words endorsed as self-descriptive. The absence of significant positive correlations in Studies 1 and 2 and the failure to find significant differences between high and low anxious subjects in terms of their bias scores suggests that this cannot be held as an explanation for the superior recall of threatening words by the high trait-anxious subjects in these studies. Perhaps a more plausible explanation for the superior recall of threatening words by the anxious individuals is that such words might be related to their personal worries. It has become clear to researchers that worry is an important component of anxiety, indeed, it has been argued that it is the principal component of anxiety (Eysenck, 1992; Mathews, 1990). The importance of worry to trait anxiety is further emphasised by research which has shown that high-trait anxious individuals tend to worry more than low-trait individuals (Tallis, Eysenck & Mathews, 1992; Borkovec, Robinson, Pruzinsky & DePree, 1983; Davey, 1993). Borkovec (1994) has argued that worry primarily involves verbal-linguistic processing and one of the consequences of such mental rehearsal of worry themes is a strengthening of threat-related information in memory. The memory bias observed in the current studies might therefore be the result of the higher levels of worry reported by high-trait anxious individuals. The implication of this suggestion is that a memory bias is most likely to be observed for information relating to an individual’s current concerns. This possibility is currently under investigation, In addition to the bias for threat shown by high-trait anxious individuals, the current research has shown that both high- and low-trait anxious subjects appear to have a recall bias for positive rather than negative words. In both studies the subjects recalled greater numbers of positive words, a finding that is consistent with numerous other studies (Mogg et al.. 1987; Kennedy & Craighead,
John Reidy and Anne Richards
662
1988; Taylor & Brown, 1988; Sanz, 1996; Reidy & Richards, 1997). This finding further highlights the importance of word selection for this type of research, particularly when utilizing non-clinical anxiety groups. The tendency to recall positive rather than negative information may mask genuine differences between anxious and non-anxious subjects, particularly if such differences are limited to threatening and non-threatening negative words. The current research has also shown that the superior recall of words positioned in the latter part of the word list can be reduced if the interval between encoding and recall is increased to three minutes. Although this is a relatively minor modification to the procedure it does highlight the need for carefully designed studies in this area of research. Future research into mood congruent memory biases needs to ensure that the possible confounding effects of words in the recency portion of word lists are minimized. One problem with the current research is the lack of consistency shown in the correlational analyses. Study I found that trait anxiety significantly correlated with the threat bias scores, whereas there were no significant correlations involving threat bias scores reported in Study 2. One possibility is that the inconsistency shown with these correlations is due to the psychometric weakness of the questionnaires used to measure anxiety and depression. There is a growing awareness that many self-report measures of these constructs tend to be highly correlated (Gotlib, 1984; Creamer, Foran & Bell, 1995). Endler, Cox, Parker and Bagby (1992) have argued that the covariation between many measures of anxiety and depression may reflect the psychometric weaknesses of these scales. It is perhaps these weaknesses that have led to the inconsistent correlational analyses reported in this paper. An alternative possibility is that the relatively small sample sizes employed in the current research have reduced the power of the experiments to detect consistent relationships among the variables investigated. Future research needs to address both of these possibilities. In conclusion, the studies presented in this paper have supported the prediction that high-trait anxiety is associated with a memory bias for threatening rather than non-threatening words. They have also confirmed that non-clinical subjects have a strong tendency to recall positive rather than negative words and suggest that care should be taken in the selection of stimuli and the analysis of such stimuli in future memory research. In addition, these studies have shown that increasing the interval between list presentation and list recall reduces the influence of the final words in the word list. However, given that the present studies and Study2 reported by Reidy and Richards (1997) used the same words, the findings need to be replicated using different materials. Acknowledgements-This
research was supported by a Medical Research Council studentship.
REFERENCES Beck, A. T. (1976). Cognitive theory and fhe emorional disorders. New York: International Universities Press. Beck, A. T. & Emery, G.( 1985). Anxiety disorders andphobias: a cognitiveperspective, New York: Basic Books. Beck, A. T., Ward, C. H., Mendelson, M., Mock, J. E. & Erbaugh, J. K. (1961). An inventory for measuring depression. Archives of General Psychiatry, 4, 561-571.
Bjork, R. A. & Whitten, W. B. (1974). Recency-sensitive retrieval processes in long-term free recall. Cognifive Psychology, 6, 173-189.
Blaney, R. H. (1986). Affect and memory: a review. Psychological Bulletin, 99, 229-246. Borkovec, T. D. (1994). The nature, functions, and origins of worry. In G. C. L. Davey & F. Tallis (Eds), Worrying: perspectives on theory, assessment and treatment (pp. 5-29). New York: Wiley. Borkovec, T. D., Robinson, E., Pruzinsky, T. & DePree, J. A. (1983). Preliminary exploration of worry: some characteristics and processes. Behaviour Research and Therapy, 21. 9-16. Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129-148. Bradley, B. P., Mogg, K. & Williams, R. (1994). Implicit and explicit memory for emotional information in non-clinical subjects. Behaviour Research and Therapy, 32,65-78. Breck. B. E. & Smith, S. H. (1983). Selective recall of self-descriptive traits by socially anxious and nonanxious females. Social Behavior and Personality, I I, 7 1-76.
Cloitre, M., Cancienne, J., Heimberg, R. G., Holt, C. S. & Liebowitz, M. (1995). Memory bias does not generalize across anxiety disorders. Behaviour, Research and Therapy, 33, 305-307. Creamer, M., Foran, J. & Bell, R. (1995). The Beck Anxiety Inventory in a non-clinical sample. Behaviour, Research and Therapy, 33.477485.
Dalgleish, T. (1994). The relationship between anxiety and memory biases for material that has been selectively processed in a vrior task. Behaviour Research and Therapy, 32.227-23 1. Dave;, G. C. L. (1993). A comparison of three worry questionnaires. Behaviour Research and Therapy, 31, 51-56. Endler, N. S., Cox, B. J., Parker, J. D. A. & Bagby, R. M. (1992). Self-reports of depression and state-trait anxiety: evidence for differential assessment. Journal of Personality and Social Ps.vchology, 63, 832-838.
Anxiety and memory
663
Eysenck. M. W. (1984). A handbook of cognitive psychology. London: Lawrence Erlbaum Associates. Eysenck, M. W. (1992). Anxiety: the cognitive perspecfive. Hove: Lawrence Erlbaum Associates. Gotlib, 1. H. (1984). Depression and general psychopathology in university students. Journal of Abnormal
Psychology,
93,
19-30.
Kennedy, R. E. & Craighead. W. E. (1988). Differential efl’ects of depression and anxiety on recall of feedback in a Teaming task. Behariour Therapy, 19, 431454. MacLeod, C. (1990). Mood disorders and cognition. In: M. W. Eysenck (Ed.), Cognitive psychology: an international reoiew. New York: Wiley. Martin, M., Ward, J. C. & Clark. D. M. (1983). Neuroticism and the recall of positive and negative personality information, Behaoiour
Research and Therapy,
21.495503.
Martin, M., Williams, R. M. & Clark. D. M. (1991). Does anxiety lead to selective processing of threat-related information? Behaviour
Research and Therapy,
29, 147-160.
Mathews, A. (1990). Why worry? The cognitive function of anxiety. Behauiour Research and Therapy, 28.455468. McNally, R. J., Foa, E. B. & Donnell, C. D. (1989). Memory bias for anxiety information in patients with panic disorder. Cognition
and Emotion,
3, 2144.
Mogg, K. 8t Mathews, A. (1990). Is there a self-referent recall bias in anxiety?. Behaciour Research and Therapy, 28. 91-92. Mogg. K., Mathews, A. & Weinman, J. (1987). Memory bias in clinical anxiety. Journal of Abornal Psychology. 96,94-98. Norton, G. R., Schaefer, E., Cox, B. J., Dorward. J. & Wozney, K. (1988). Selective memory effects in nonclinical panickers. Journal
of Anxiety
Disorders,
2, 1699177.
Nugent. K. & Mineka, S. (1994). The effect of high and low trait anxiety on implicit and explicit memory tasks. and Emotion,
Cognition
8, 147-163.
Raven, J. C., Court, J. H. & Raven, J. (1969). Manualfor Raven’sprogressice matrices and cocabulary scales. London: H.K. Lewis. Reidy, J. (1994). Anxiqv and explicif memoryfor emotionally toned words. Unpublished Ph.D. thesis, Reidy. J. 8~Richards, A. (1997). Anxiety and memory: a recall bias for threatening words in high anxiety. Behaviour Research and Therapy,
35, 531-542.
Richards, A. & French. C. C. (1991). Effects of encoding and anxiety on implicit and explicit memory performance. Personality
and Individual
Differences,
12, 131-l 39.
Russo. R., Patterson, N., Robertson, D., Stevenson, N. & Upward, J. (1996). Emotional value of information and its relevance in the interpretation of homophones in anxiety. Cognition and Emotion, IO, 213-220. Sanz, J. (1996). Memory biases in social anxiety and depression. Cognition and Emotion, IO, 87-105. Spielberger, C. D., Gorsuch, R. L., Lushene, R.. Vagg, P. R. & Jacobs, G. A. (1983). Manual for the state-trait anrietj inventory (form Y). Palo Alto, CA: Consulting Psychologists Press. Taylor. S. E. & Brown, J. D. (1988). Illusion and well-being: a social psychological perspective on mental health. Psychological Bulletin.
103, 193-210.
Tallis, F., Eysenck, M. W. & Mathews, A. (1992). A questionnaire for the measurement of nonpathological worry. Watts, F. N.
Differences, 13, 161-168. & Coyle, K. (1993). Phobics show
Personality
and Indiridual
poor recall of anxiety words.
British
Journal
of Medical
Psychology,
66, 3733
Cognition
and Emotion.
S. 313-
382. Watts, F. N. & 329.
Dalgleish, T. (1991). Memory for phobia-related words in spider phobics.
Watts. F. N., Trezise. L. & Sharrock, R. (1986). Processing of phobic stimuli.
British
Journal
of Clinical
Psychology,
25,
253-261.
Williams, J. M. G.. Watts, F. N., MacLeod, C. & Mathews, A. (1988). Cognitice psychology and emotional disorders. Chichester: Wiley, 1988. Zeitlin, S. B. & McNally, R. J. (1991). Implicit and explicit memory bias for threat in post-traumatic stress disorder. Behaoiour,
Research and Therapy,
29,451457.
Zinbarg, R. E. & Barlow. D. H. (1996). Structure of anxiety and the anxiety disorders: a hierarchical model. Abnormal Psychology. 105. 181-193.
Journal
of