- Email: [email protected]
Trait anxiety influences on judgments of frequency and recall
Personality and Individual Dierences 29 (2000) 395±404
www.elsevier.com/locate/paid
Trait anxiety in¯uences on judgments of frequency and recall p Karan S. Kverno* The George Washington University, 2121 I St. NW, Washington, DC 20052, USA Received 26 March 1999; accepted 31 August 1999
Abstract Individuals high in trait anxiety believe that they are more at risk for a variety of threatening events. Subjective probability estimates of risk presumably rely on frequency-of-occurrence information in memory. This study sought to compare individuals diering in the trait of anxiety on frequency estimates and recall. Neutral and threatening words were presented varying numbers of times. An interaction between trait anxiety and word type indicated that individuals with high trait anxiety estimated greater frequencies for threatening words than did individuals with low trait anxiety. There were no dierences between the groups in subsequent recall. Higher false recognitions of threatening words by high trait anxious participants suggested the use of a dierent decision strategy or greater activation and covert rehearsal of threat. 7 2000 Elsevier Science Ltd. All rights reserved. Keywords: Anxiety; Personality; Memory; Probability judgment
1. Introduction ``How likely is it that rely on several sources experiences as a victim, episodes reported by the
you will be assaulted?'' To answer this question, an individual may of information in memory. These sources may include personal accounts from friends or relatives victimized by assault, or graphic news media. The individual might even recall several times that he or
This work was based on a dissertation by the author, submitted to the graduate school at The George Washington University in partial ful®llment of the requirements for the degree of Doctor of Philosophy. * Tel.: +1-410-706-7556; fax: +1-410-706-0253. E-mail address: [email protected] (K.S. Kverno). p
0191-8869/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 1 9 1 - 8 8 6 9 ( 9 9 ) 0 0 2 0 1 - 9
396
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
she imagined being assaulted. All of these memories concern frequency-of-occurrence information, or a count of how often something has occurred. Frequency-of-occurrence information in memory is used to judge the likelihood of a future event (Hasher & Zacks, 1984). Individual dierences in the personality trait of anxiety have been found to bias subjective judgments of risk. High trait anxious individuals are particularly prone to believing that they are at risk for certain negative or harmful events (Butler & Mathews, 1987; Gasper & Clore, 1998). Although the literature shows that high trait anxious individuals overestimate risk, we do not know how they monitor the actual frequency of threatening information in the environment. The purpose of this study was to examine how individuals diering in the trait of anxiety monitor the frequency-of-occurrence of presented threatening and neutral word stimuli. Dierences in the ability to monitor actual frequency might explain at least some of the judgment biases seen in high trait anxious individuals. Several studies have shown that trait anxiety in¯uences interpretations of uncertainty or ambiguity. High trait anxious individuals are more likely than low trait anxious individuals to write down the spelling with the more threatening meaning when presented auditorily with homophones having both threatening (e.g., die) and nonthreatening (e.g., dye) meanings (Byrne & Eysenck, 1993; Dalgleish, 1994; Eysenck, MacLeod & Mathews, 1987). Similarly, when presented with sentences having ambiguous meanings (e.g., ``The doctor examined little Emily's growth.''), high trait anxious individuals are more likely than low trait anxious individuals to interpret the sentences with a threatening meaning (MacLeod & Cohen, 1993). Only two studies have examined the relative contributions of state and trait anxiety to judgments of frequency or probability. Butler and Mathews (1987) compared risk estimates of a group of students anticipating an exam having important consequences with a similar group of students who were not expecting an exam. High trait anxiety was associated with higher perceived risk to oneself for a variety of negative events related to and unrelated to the exam. State anxiety was also associated with increased risk estimates, but related to the exam only. In a similar study, Gasper and Clore (1998, Experiment 1) compared risk estimates of students anticipating an exam and found that individuals with high trait anxiety estimated higher probabilities than individuals with low trait anxiety for a variety of personal and impersonal risks. In addition, they showed that individuals with high trait anxiety relied more on their increased state anxiety as information in making probability judgments. Thus it appears that increases in an anxious mood state may increase the subjective probabilities of risk associated with the perceived cause of the increased anxiety (Butler & Mathews, 1987); or may be a source of information used by individuals when making judgments of risk (Gasper & Clore, 1998). But despite changes in state anxiety, trait anxiety is associated with a pervasive tendency to believe that the probability of risk or harm is greater. Butler and Mathews (1987) propose two possible cognitive mechanisms for the judgment and interpretive biases seen in high trait anxiety: the use of judgmental heuristics and the in¯uences of existing schemata in memory. When people make judgments about the likelihood of certain events they may rely on an availability heuristic, that is, the ease with which past occurrences of similar events can be accessed from memory (Tversky & Kahneman, 1973, 1974). The ease with which past occurrences can be accessed is in turn thought to be a function of the extent or elaboration of activated threat-related schemata in memory. Butler and Mathews suggest
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
397
that the threat-related schemata are more extensive and elaborated in individuals with high trait anxiety compared to individuals with low trait anxiety. Activated threat-related schemata could have the eect of increasing the availability of threat-related information in memory, raising the sensitivity to anxiety related cues in the environment, and increasing the tendency to label or categorize input as threatening Ð all of which could lead to an overestimation of subjective personal risk. The schema-based theories set forth the view that subjective probability estimates involve cognitive processes that take place both at retrieval and at encoding. MacLeod and Campbell (1992) reported several studies suggesting that when asked to judge the probability of future events, people rely on retrieval of individual instances of similar events. Other studies have shown estimates of frequency to be largely independent of recall. Watkins and LeCompte (1991) reported several experiments comparing the relative deviations of recall and frequency estimation from true category frequencies and found that recall was not suciently informative to account for frequency estimates. Manis, Shedler, Jonides and Nelson (1993) reported studies showing that performance on a frequency-of-occurrence estimation task was predicted by true frequency-of-occurrence and not by the recall of speci®c instances. According to Hasher and Zacks (1979, 1984) encoding of frequency is an automatic cognitive process that takes place without intention or eort and is independent of the more eortful or strategic cognitive processes. They presented evidence suggesting that a variety of individual dierences such as age, ability, stress, and arousal that traditionally aect tests of eortful cognitive processing (e.g., recall) have no eect on memory for frequency. Although many studies have questioned automatic encoding of frequency (e.g., Birnbaum, Taylor, Johnson & Raye, 1987; Fisk, 1986), it has often been dicult to rule out other explanations. Strong support for the view that frequency is monitored automatically is gained from a recent study of Alzheimer's patients showing that frequency judgments can remain quite accurate even when recall is severely impaired (Wiggs, Martin & Sunderland, 1997). Because mood congruent memories appear to be more available for various judgments and interpretations, many researchers have attempted to document a preferential bias to recall threat-related information in anxiety on explicit memory tasks. Studies examining moodcongruent memory in anxiety typically have participants study lists consisting of both anxietyrelated and nonanxiety-related words. Participants are then tested on the recall of those words. An explicit memory bias for anxiety-related words in high trait anxious individuals was reported by Reidy and Richards (1997), however upon further analysis the researchers determined that the eect was due to state anxiety dierences. Other studies looking for an explicit memory bias for threat words in trait anxiety have found none (Mathews, Mogg, May & Eysenck, 1989; Richards & French, 1991), even when high trait anxious individuals have shown a processing (interpretive) bias for the same words in a previous task (Dalgleish, 1994). Several current theories suggest that anxiety may aect both automatic and strategic cognitive processes (e.g., Beck & Clark, 1997; Mathews & MacLeod, 1994; Williams, Watts, MacLeod & Mathews, 1988). These theories can account for both the interpretive biases and the lack of memory bias in high trait anxiety. Automatic processing of threat cues may predominate in attentional and interpretive tasks whereas strategic processing (intentional avoidance of elaboration) may predominate in recall tasks. Because frequency-of-occurrence judgments rely (at least to some extent) on automatic cognitive processes and because
398
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
frequency information in memory may play an important role in subjective judgments and beliefs (Hasher & Zacks, 1984), a simple frequency monitoring study was undertaken to look at the eects of trait anxiety on memory for frequency-of-occurrence and recall. In this word frequency estimation paradigm, college students were presented with a series of physically threatening and neutral words. The words were shown either two, ®ve, or eight times at a fairly rapid rate. Following the presentation of words, the participants were asked to estimate the number of times they saw each word. Some of the words on the frequency test were words that did not appear on the study list, and those words served as distractors. It was predicted that high trait anxious participants would estimate relatively greater frequencies for threatening words than would low trait anxious participants. It was also predicted that there would be no dierences between groups on recall. 2. Method 2.1. Participants A total of 159 (65 males and 94 females) volunteer college students participated in the experiment for extra credit. 2.2. Materials Materials included a subject information sheet for the collection of demographic information, the Spielberger State-Trait Anxiety Inventory (STAI) (Spielberger, Gorsuch, Lushene, Vagg & Jacobs, 1983), single word stimuli on slides, the frequency estimation test and recall test. 2.2.1. Word stimuli Slides were prepared using the pool of physically threatening and neutral words reported by Mathews et al. (1989). The stimulus word sets were matched for word frequency (frequency per million) in the English language according to Carroll, Davies, and Richman's (1971) norms, t (46)=ÿ0.44, NS. The Carroll et al. (1971) norms are scaled to provide an index of frequency that parallels those used by Thorndike and Lorge (1944). The stimulus word sets were also equated for word length, t (46)=ÿ1.58, NS. The buer words were high frequency neutral words that were equated with the word sets for word length. There were a total of 24 words in each of the two (neutral and threatening) valence conditions. Six words from each valence condition were randomly assigned to each of three presentation frequencies: two, ®ve, and eight. Six words from each valence condition were not shown and these words served as distractors on the frequency test. Word order was randomly determined with the exception of two restrictions that were made in order to control the variability of spacing between repetitions. First, the words were divided into three presentation units or blocks. Each block consisted of two neutral and two threatening words from each of the three frequency levels (two, ®ve or eight). Second, repeated words always had at least four spaces between them. This spacing rule was based on
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
399
Hintzman's (1969) ®nding that judgments of frequency increase as a function of the spacing of words up to four spaces between repetitions, and then levels o. In summary, there were a total of 180 target slides shown for the non-zero frequencies (six neutral and six threat at frequency levels two, ®ve and eight) plus eight primacy-recency buers. The words appeared in large black type, all in capital letters. 2.2.2. Frequency and recall tests The frequency test was made up of the 36 presented words and the 12 distractors, listed in random order. Each word was followed by a short line on which the participants were to record their estimated frequencies. The following instruction appeared at the top of the page: ``In the blank next to each word, please write the number of times that you saw or studied it. If you are not sure, guess''. The recall test was a blank piece of paper with the following instructions typed at the top of the page: ``Please write down all of the words you remember seeing, in any order''.
2.3. Design and procedure Groupings for the between-subjects variable, trait anxiety, were determined by median split scores on the STAI Trait-Anxiety scale. Two variables were manipulated within subjects: word type (neutral or threatening) and presented word frequency (two, ®ve or eight). The dependent variable for the 2 (group) 2 (word type) 3 (frequency) comparison was the estimated frequency of occurrence of the words. 2.3.1. Acquisition Participants were tested in groups of four to 14. After reading and signing the consent form and ®lling out a demographic questionnaire, participants were instructed that they would be presented with a long list of words that would be shown on slides at a fairly rapid rate. They were told to pay close attention to each of the slides and to look at each word for as long as it appeared on the screen, and that they would be tested on these words afterwards. No mention was made of a frequency test. Words were then presented at a rate of 3 s per word. 2.3.2. Frequency and recall tests Immediately following the acquisition phase, participants were given the frequency test forms. The instructions that appeared on the top of the forms were read aloud. Participants recorded the number of times they thought they had seen each word and forms were collected as soon as each participant ®nished. As soon as all participants ®nished the frequency test, the recall forms were passed out and the instructions that appeared on the top of each form were read aloud. Participants were given a maximum of 6 min to recall as many words as possible in any order. These forms were collected from each participant as soon as he or she indicated completion. 2.3.3. Psychometric measures After all participants completed the frequency test, they were given the STAI and instructed
400
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
to answer all questions as best they could. They were then debriefed and given the opportunity to ask questions. 3. Results 3.1. Participants The participants were ®rst divided into high and low trait anxiety groups on the basis of a median split on the STAI Trait scale. The median STAI Trait score was 38 so all individuals who scored below 38 were included in the low trait anxiety group (n = 74) and all individuals who scored at or above 38 were included into the high trait anxiety group (n = 85). The mean STAI trait scores for the low and high trait anxious groups were as follows: 31.15 (SD=4.17) and 46.60 (SD=7.66). The high trait anxiety group was signi®cantly higher than the low trait anxiety group in state anxiety, t (157)=ÿ 7.76, P < 0.001 and state anxiety was signi®cantly correlated with trait anxiety, r = 0.66, P < 0.001. The mean ages of the participants in the low and high groups were as follows: 19.49 (SD=0.18) and 19.12 (SD=0.12). There were no age dierences between the groups, t (157)=0.10, NS. 3.2. Frequency estimates Each subject's frequency estimates were averaged over each of the actual frequency conditions. The means of these mean frequency estimates were subjected to a repeated measures analysis of variance (ANOVA) that included trait anxiety as the between-group factor and word type (neutral or threatening) and presentation frequency (two, ®ve, or eight) as within-group factors. The means and standard deviations of the frequency estimates can be viewed in Table 1. There were signi®cant main eects of word type, F (1,157)=80.35, P < 0.0001, and presentation frequency, F (2,314)=365.11, P < 0.0001, and a signi®cant interaction Table 1 Mean frequency estimates for presented and distractor words Groupa
Presentation frequencies 0
Low M SD High M SD a
2
5
8
N
T
N
T
N
T
N
T
0.38 0.87
0.52 0.88
2.63 1.17
3.49 1.75
4.42 1.67
4.26 1.66
5.34 1.80
6.38 2.30
0.48 0.80
0.93 1.20
2.70 1.08
4.16 1.80
4.48 1.55
4.78 1.85
5.99 2.20
6.98 2.63
N denotes neutral words, T denotes threatening words.
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
401
between word type and frequency, F (2,314)=28.55, P < 0.0001. Simple eects analyses indicated that larger frequency estimates were given to threat words than neutral words at presentation frequencies two, F (1,157)=87.29, P < 0.0001 and eight, F (1,157)=55.17, P < 0.0001, but not at presentation frequency ®ve, F < 1.00. The central prediction of an interaction between word type and anxiety group was signi®cant, F (2,314)=3.88, P = 0.05. Although all participants estimated relatively higher frequencies for threatening than neutral words, the eect was more pronounced for anxious participants. Simple eects analyses indicated that the eect of trait anxiety collapsed across presentation frequencies was signi®cant for threatening words, F (1,157)=4.68, P < 0.05, but not for neutral words, F (1,157)=1.43, NS. 3.3. Recall The mean numbers of neutral words recalled were 8.74 and 8.32 by the low and high trait anxiety groups respectively. The mean numbers of threatening words recalled were 9.78 and 9.00 by the low and high trait anxiety groups respectively. A two-way repeated measures ANOVA was performed with trait anxiety as the between-group factor and word type as the within-subjects factor. There were no recall dierences between the groups, F (1,157)=1.63, NS. Overall, threatening words were recalled better by both groups, F (1,157)=13.41, P < 0.001. There was no interaction between trait anxiety and word type, F < 1.00. 3.4. Recognition The zero frequency level data were analyzed separately to examine the possibility that the high trait anxious participants were exhibiting a response bias, that is whether they were more likely than the low trait anxiety participants to give higher estimates of threatening words regardless of information stored in memory. Recognition was assessed by asking the participants to rate the frequency of the presented words (old) and the distractor words (new). Since there was no true frequency of one, an estimate of the probability of detecting a threat word vs a neutral word was made by de®ning hits as the proportion of words from presentation level two correctly recognized as having occurred (i.e., subject gave a frequency estimate greater than zero). False alarms were de®ned as the proportion of distractor words which were given frequency estimates greater than zero. The group means were then subjected Table 2 Mean proportion of hits, false alarms and corrected recognition by trait anxiety groups Groupa
Low High a
Hits
False alarms
Corrected
N
T
N
T
N
T
0.92 0.90
0.92 0.94
0.18 0.22
0.21 0.32
0.74 0.67
0.71 0.62
N denotes neutral words, T denotes threatening words.
402
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
to a two-way repeated measures ANOVA with trait anxiety as the between-group factor and word type as the repeated measure. The resulting means are displayed in Table 2. An analysis of hit rates resulted in a main eect of word type, F (1,157)=4.09, P < 0.05, with threat words being given higher estimates than neutral words. For false alarms, the interaction between group and word type was signi®cant, F (1,157)=4.85, P < 0.05. High trait anxiety participants made more false alarms than low trait anxiety participants on threatening words, F (1,157)=4.29, P < 0.05, but not on neutral words, F < 1.00. Subtracting the proportion of false alarms from the proportion of hits gave an estimate of corrected recognition for the presented words. There were no recognition dierences between the groups, F (1,157)=2.54, NS, nor any group by word type interactions, F < 1.00. Both groups recognized more neutral words than threatening words, F (1,157)=4.83, P < 0.05.
4. Discussion As predicted, individuals with high trait anxiety estimated greater frequencies for threatening words than individuals with low trait anxiety, but similar frequencies for neutral words. There were no dierences between the groups on recall. These ®ndings support a distinction between high and low trait anxious individuals in memory for frequency-of-occurrence information related to threat. The distinction appears to be one of degree rather than kind. Both groups estimated higher frequencies for threatening words than for neutral words and both groups recalled more threatening than neutral words. The similarities between the groups deserve comment. Based on the pattern of ®ndings for frequency estimates and recall, it is reasonable to suggest that all participants were using an availability heuristic to some extent. The threat words were more accessible than the neutral words for frequency judgments and recall. Certain word characteristics were not controlled (e.g., categorization and imagability) and may have made the threatening words easier to recall. For example, the threat words were all from one category (physical threat) and the neutral words were from dierent categories. In addition, it is unknown what the carry-over eects were from the estimate to the recall task. Exposure to the words during the frequency estimation task may have eliminated potential recall dierences between groups, although the failure to ®nd an explicit memory eect in trait anxiety is in line with previous research (e.g., Mathews et al., 1989; Richards & French, 1991). The dierences between groups were evident only on frequency estimates of threatening words: words that were presented and words that never occurred. Neither an automatic frequency counting mechanism at encoding nor the use of an availability heuristic at retrieval can easily explain the group dierences on distractor estimates. Compared to the low trait anxious participants, the high trait anxious participants appeared to be using a more lenient criterion for judging the occurrence of threatening words. They were more likely to report seeing a threatening word, whether or not it had occurred. Whether or not they actually believed they had seen the threat distractors or whether they were simply guessing when uncertain is not clear from this study. Alternatively, if high trait anxious individuals do have more extensive and elaborated threatrelated schemata in memory which are activated by the presence of threatening events (in this
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
403
case words), it may be that they are also more likely to automatically activate the related distractor concepts in memory. In addition, frequency judgments increase when events are highly salient or vivid such as when they receive extra attention in the media (Lichtenstein, Slovic, Fischho, Layman & Combs, 1978; Tversky & Kahneman, 1973, 1974). If high trait anxious individuals thought about or imagined more threat-related events during the study or if their imaginings of threat were more vivid, they may have had greater diculty distinguishing between the two sources in memory (internal vs external) and had more uncertainty at test. False recognition of distractors may re¯ect the inability to distinguish between real and imagined events in memory (Johnson & Raye, 1981). To test this possibility, in two subsequent studies an attempt was made to hold the frequency of internal and external generations constant between the high and low trait anxious individuals (Kverno, 1994). Once constrained, the dierences in threat frequency estimates between the high and low anxiety groups disappeared. In addition, compared to the low trait anxious individuals, the high trait anxious individuals, no longer exhibited a greater speci®c response bias on threat distractors. 5. Implications The ®ndings have implications for theories concerning the global nature of subjective probabilities related to threat observed in individuals with anxiety (Butler & Mathews, 1987; Constans & Mathews, 1993; Johnson & Tversky, 1983). Not only do high trait anxious individuals estimate higher frequencies for threatening future events, they estimate higher frequencies for threatening events that have previously occurred (a speci®c eect) and for related threatening events that have never occurred (a global eect). High trait anxious individuals believe that they are more at risk for harmful events (e.g., being assaulted or harmed). The results of this study suggest that when high trait anxious individuals are asked to monitor or estimate the frequencies of personally or clinically relevant threatening events (e.g., panic attacks, phobic fears, pain, binges, depressive episodes), the estimates might also be in¯ated. An important area for future research will be to further examine the mechanisms by which anxiety in¯uences frequency monitoring. References Beck, A. T., & Clark, D. A. (1997). An information processing model of anxiety: automatic and strategic processes. Behaviour Research and Therapy, 35(1), 49±58. Birnbaum, I. M., Taylor, T. H., Johnson, M. K., & Raye, C. L. (1987). Is event frequency encoded automatically? The case of alcohol intoxication. Journal of Experimental Psychology: Learning, Memory and Cognition, 13(2), 251±258. Butler, G., & Mathews, A. (1987). Anticipatory anxiety and risk perception. Cognitive Therapy and Research, 11(5), 551±565. Byrne, A., & Eysenck, M. W. (1993). Inidividual dierences in positive and negative interpretive biases. Personality and Individual Dierences, 14(6), 849±851. Carroll, J. B., Davies, P., & Richman, B. (1971). Word frequency book. New York: American Heritage. Constans, J. I., & Mathews, A. M. (1993). Mood and the subjective risk of future events. Cognition and Emotion, 7(6), 545±560.
404
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
Dalgleish, T. (1994). The relationship between anxiety and memory biases for material that has been selectively processed in a prior task. Behaviour Research and Therapy, 2, 227±231. Eysenck, M. W., MacLeod, C., & Mathews, A. (1987). Cognitive functioning and anxiety. Psychological Research, 49, 189±195. Fisk, A. D. (1986). Frequency encoding is not inevitable and is not automatic: a reply to Hasher and Zacks. American Psychologist, (February), 215±216. Gasper, K., & Clore, G. L. (1998). The persistent use of negative aect by anxious individuals to estimate risk. Journal of Personality and Social Psychology, 74(5), 1350±1363. Hasher, L., & Zacks, R. T. (1979). Automatic and eortful processes in memory. Journal of Experimental Psychology: General, 108(3), 356±388. Hasher, L., & Zacks, R. T. (1984). Automatic processing of fundamental information. The case of frequency of occurrence. American Psychologist, 39(12), 1372±1388. Hintzman, D. L. (1969). Apparent frequency as a function of frequency and the spacing of repetitions. Journal of Experimental Psychology, 80(1), 139±145. Johnson, E. J., & Tversky, A. (1983). Aect, generalization, and the perception of risk. Journal of Personality and Social Psychology, 45(1), 20±31. Johnson, M. K., & Raye, C. L. (1981). Reality monitoring. Psychological Review, 88(1), 67±85. Kverno, K. (1994). Frequency monitoring and anxiety. Unpublished doctoral dissertation, The George Washington University, Washington, DC. Lichtenstein, S., Slovic, P., Fischho, B., Layman, M., & Combs, B. (1978). Judged frequency of lethal events. Journal of Experimental Psychology: Human Learning and Memory, 4(6), 551±581. MacLeod, C., & Campbell, L. (1992). Memory accessibility and probability judgments: an experimental evaluation of the availability heuristic. Journal of Personality and Social Psychology, 63(6), 890±902. MacLeod, C., & Cohen, I. L. (1993). Anxiety and the interpretation of ambiguity: a text comprehension study. Journal of Abnormal Psychology, 102(2), 238±247. Manis, M., Shedler, J., Jonides, J., & Nelson, T. E. (1993). Availability heuristic in judgements of set size and frequency of occurrence. Journal of Personality & Social Psychology, 65(3), 448±457. Mathews, A., & MacLeod, C. (1994). Cognitive approaches to emotion and emotional disorders. Annual Review of Psychology, 45, 25±50. Mathews, A., Mogg, K., May, J., & Eysenck, M. (1989). Implicit and explicit memory bias in anxiety. Journal of Abnormal Psychology, 98(3), 236±240. Reidy, J., & Richards, A. (1997). Anxiety and memory: a recall bias for threatening words in high anxiety. Behaviour Research and Therapy, 35(6), 531±542. Richards, A., & French, C. C. (1991). Eects of encoding and anxiety on implicit and explicit memory performance. Personality and Individual Dierences, 12(2), 131±139. Spielberger, C. D., Gorsuch, R. L., Lushene, R., Vagg, P. R., & Jacobs, G. A. (1983). Manual for the State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press, Inc. Thorndike, E. L., & Lorge, I. (1944). The teachers' word book of 30,000 words. New York: Columbia University. Tversky, A., & Kahneman, D. (1973). Availability: a heuristic for judging frequency and probability. Cognitive Psychology, 5, 207±232. Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: heuristics and biases. Science, 185, 1124±1131. Watkins, M. J., & LeCompte, D. C. (1991). Inadequacy of recall as a basis for frequency knowledge. Journal of Experimental Psychology: Learning, Memory, and Cognition, 17(6), 1161±1176. Wiggs, C. L., Martin, A., & Sunderland, T. (1997). Monitoring frequency of occurrence without awareness: evidence from patients with Alzheimer's Disease. Journal of Clinical and Experimental Neuropsychology, 19(2), 235±244. Williams, J. M. G., Watts, F. N., MacLeod, C., & Mathews, A. (1988). Cognitive psychology and emotional disorders. NY: John Wiley & Sons.
www.elsevier.com/locate/paid
Trait anxiety in¯uences on judgments of frequency and recall p Karan S. Kverno* The George Washington University, 2121 I St. NW, Washington, DC 20052, USA Received 26 March 1999; accepted 31 August 1999
Abstract Individuals high in trait anxiety believe that they are more at risk for a variety of threatening events. Subjective probability estimates of risk presumably rely on frequency-of-occurrence information in memory. This study sought to compare individuals diering in the trait of anxiety on frequency estimates and recall. Neutral and threatening words were presented varying numbers of times. An interaction between trait anxiety and word type indicated that individuals with high trait anxiety estimated greater frequencies for threatening words than did individuals with low trait anxiety. There were no dierences between the groups in subsequent recall. Higher false recognitions of threatening words by high trait anxious participants suggested the use of a dierent decision strategy or greater activation and covert rehearsal of threat. 7 2000 Elsevier Science Ltd. All rights reserved. Keywords: Anxiety; Personality; Memory; Probability judgment
1. Introduction ``How likely is it that rely on several sources experiences as a victim, episodes reported by the
you will be assaulted?'' To answer this question, an individual may of information in memory. These sources may include personal accounts from friends or relatives victimized by assault, or graphic news media. The individual might even recall several times that he or
This work was based on a dissertation by the author, submitted to the graduate school at The George Washington University in partial ful®llment of the requirements for the degree of Doctor of Philosophy. * Tel.: +1-410-706-7556; fax: +1-410-706-0253. E-mail address: [email protected] (K.S. Kverno). p
0191-8869/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 1 9 1 - 8 8 6 9 ( 9 9 ) 0 0 2 0 1 - 9
396
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
she imagined being assaulted. All of these memories concern frequency-of-occurrence information, or a count of how often something has occurred. Frequency-of-occurrence information in memory is used to judge the likelihood of a future event (Hasher & Zacks, 1984). Individual dierences in the personality trait of anxiety have been found to bias subjective judgments of risk. High trait anxious individuals are particularly prone to believing that they are at risk for certain negative or harmful events (Butler & Mathews, 1987; Gasper & Clore, 1998). Although the literature shows that high trait anxious individuals overestimate risk, we do not know how they monitor the actual frequency of threatening information in the environment. The purpose of this study was to examine how individuals diering in the trait of anxiety monitor the frequency-of-occurrence of presented threatening and neutral word stimuli. Dierences in the ability to monitor actual frequency might explain at least some of the judgment biases seen in high trait anxious individuals. Several studies have shown that trait anxiety in¯uences interpretations of uncertainty or ambiguity. High trait anxious individuals are more likely than low trait anxious individuals to write down the spelling with the more threatening meaning when presented auditorily with homophones having both threatening (e.g., die) and nonthreatening (e.g., dye) meanings (Byrne & Eysenck, 1993; Dalgleish, 1994; Eysenck, MacLeod & Mathews, 1987). Similarly, when presented with sentences having ambiguous meanings (e.g., ``The doctor examined little Emily's growth.''), high trait anxious individuals are more likely than low trait anxious individuals to interpret the sentences with a threatening meaning (MacLeod & Cohen, 1993). Only two studies have examined the relative contributions of state and trait anxiety to judgments of frequency or probability. Butler and Mathews (1987) compared risk estimates of a group of students anticipating an exam having important consequences with a similar group of students who were not expecting an exam. High trait anxiety was associated with higher perceived risk to oneself for a variety of negative events related to and unrelated to the exam. State anxiety was also associated with increased risk estimates, but related to the exam only. In a similar study, Gasper and Clore (1998, Experiment 1) compared risk estimates of students anticipating an exam and found that individuals with high trait anxiety estimated higher probabilities than individuals with low trait anxiety for a variety of personal and impersonal risks. In addition, they showed that individuals with high trait anxiety relied more on their increased state anxiety as information in making probability judgments. Thus it appears that increases in an anxious mood state may increase the subjective probabilities of risk associated with the perceived cause of the increased anxiety (Butler & Mathews, 1987); or may be a source of information used by individuals when making judgments of risk (Gasper & Clore, 1998). But despite changes in state anxiety, trait anxiety is associated with a pervasive tendency to believe that the probability of risk or harm is greater. Butler and Mathews (1987) propose two possible cognitive mechanisms for the judgment and interpretive biases seen in high trait anxiety: the use of judgmental heuristics and the in¯uences of existing schemata in memory. When people make judgments about the likelihood of certain events they may rely on an availability heuristic, that is, the ease with which past occurrences of similar events can be accessed from memory (Tversky & Kahneman, 1973, 1974). The ease with which past occurrences can be accessed is in turn thought to be a function of the extent or elaboration of activated threat-related schemata in memory. Butler and Mathews suggest
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
397
that the threat-related schemata are more extensive and elaborated in individuals with high trait anxiety compared to individuals with low trait anxiety. Activated threat-related schemata could have the eect of increasing the availability of threat-related information in memory, raising the sensitivity to anxiety related cues in the environment, and increasing the tendency to label or categorize input as threatening Ð all of which could lead to an overestimation of subjective personal risk. The schema-based theories set forth the view that subjective probability estimates involve cognitive processes that take place both at retrieval and at encoding. MacLeod and Campbell (1992) reported several studies suggesting that when asked to judge the probability of future events, people rely on retrieval of individual instances of similar events. Other studies have shown estimates of frequency to be largely independent of recall. Watkins and LeCompte (1991) reported several experiments comparing the relative deviations of recall and frequency estimation from true category frequencies and found that recall was not suciently informative to account for frequency estimates. Manis, Shedler, Jonides and Nelson (1993) reported studies showing that performance on a frequency-of-occurrence estimation task was predicted by true frequency-of-occurrence and not by the recall of speci®c instances. According to Hasher and Zacks (1979, 1984) encoding of frequency is an automatic cognitive process that takes place without intention or eort and is independent of the more eortful or strategic cognitive processes. They presented evidence suggesting that a variety of individual dierences such as age, ability, stress, and arousal that traditionally aect tests of eortful cognitive processing (e.g., recall) have no eect on memory for frequency. Although many studies have questioned automatic encoding of frequency (e.g., Birnbaum, Taylor, Johnson & Raye, 1987; Fisk, 1986), it has often been dicult to rule out other explanations. Strong support for the view that frequency is monitored automatically is gained from a recent study of Alzheimer's patients showing that frequency judgments can remain quite accurate even when recall is severely impaired (Wiggs, Martin & Sunderland, 1997). Because mood congruent memories appear to be more available for various judgments and interpretations, many researchers have attempted to document a preferential bias to recall threat-related information in anxiety on explicit memory tasks. Studies examining moodcongruent memory in anxiety typically have participants study lists consisting of both anxietyrelated and nonanxiety-related words. Participants are then tested on the recall of those words. An explicit memory bias for anxiety-related words in high trait anxious individuals was reported by Reidy and Richards (1997), however upon further analysis the researchers determined that the eect was due to state anxiety dierences. Other studies looking for an explicit memory bias for threat words in trait anxiety have found none (Mathews, Mogg, May & Eysenck, 1989; Richards & French, 1991), even when high trait anxious individuals have shown a processing (interpretive) bias for the same words in a previous task (Dalgleish, 1994). Several current theories suggest that anxiety may aect both automatic and strategic cognitive processes (e.g., Beck & Clark, 1997; Mathews & MacLeod, 1994; Williams, Watts, MacLeod & Mathews, 1988). These theories can account for both the interpretive biases and the lack of memory bias in high trait anxiety. Automatic processing of threat cues may predominate in attentional and interpretive tasks whereas strategic processing (intentional avoidance of elaboration) may predominate in recall tasks. Because frequency-of-occurrence judgments rely (at least to some extent) on automatic cognitive processes and because
398
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
frequency information in memory may play an important role in subjective judgments and beliefs (Hasher & Zacks, 1984), a simple frequency monitoring study was undertaken to look at the eects of trait anxiety on memory for frequency-of-occurrence and recall. In this word frequency estimation paradigm, college students were presented with a series of physically threatening and neutral words. The words were shown either two, ®ve, or eight times at a fairly rapid rate. Following the presentation of words, the participants were asked to estimate the number of times they saw each word. Some of the words on the frequency test were words that did not appear on the study list, and those words served as distractors. It was predicted that high trait anxious participants would estimate relatively greater frequencies for threatening words than would low trait anxious participants. It was also predicted that there would be no dierences between groups on recall. 2. Method 2.1. Participants A total of 159 (65 males and 94 females) volunteer college students participated in the experiment for extra credit. 2.2. Materials Materials included a subject information sheet for the collection of demographic information, the Spielberger State-Trait Anxiety Inventory (STAI) (Spielberger, Gorsuch, Lushene, Vagg & Jacobs, 1983), single word stimuli on slides, the frequency estimation test and recall test. 2.2.1. Word stimuli Slides were prepared using the pool of physically threatening and neutral words reported by Mathews et al. (1989). The stimulus word sets were matched for word frequency (frequency per million) in the English language according to Carroll, Davies, and Richman's (1971) norms, t (46)=ÿ0.44, NS. The Carroll et al. (1971) norms are scaled to provide an index of frequency that parallels those used by Thorndike and Lorge (1944). The stimulus word sets were also equated for word length, t (46)=ÿ1.58, NS. The buer words were high frequency neutral words that were equated with the word sets for word length. There were a total of 24 words in each of the two (neutral and threatening) valence conditions. Six words from each valence condition were randomly assigned to each of three presentation frequencies: two, ®ve, and eight. Six words from each valence condition were not shown and these words served as distractors on the frequency test. Word order was randomly determined with the exception of two restrictions that were made in order to control the variability of spacing between repetitions. First, the words were divided into three presentation units or blocks. Each block consisted of two neutral and two threatening words from each of the three frequency levels (two, ®ve or eight). Second, repeated words always had at least four spaces between them. This spacing rule was based on
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
399
Hintzman's (1969) ®nding that judgments of frequency increase as a function of the spacing of words up to four spaces between repetitions, and then levels o. In summary, there were a total of 180 target slides shown for the non-zero frequencies (six neutral and six threat at frequency levels two, ®ve and eight) plus eight primacy-recency buers. The words appeared in large black type, all in capital letters. 2.2.2. Frequency and recall tests The frequency test was made up of the 36 presented words and the 12 distractors, listed in random order. Each word was followed by a short line on which the participants were to record their estimated frequencies. The following instruction appeared at the top of the page: ``In the blank next to each word, please write the number of times that you saw or studied it. If you are not sure, guess''. The recall test was a blank piece of paper with the following instructions typed at the top of the page: ``Please write down all of the words you remember seeing, in any order''.
2.3. Design and procedure Groupings for the between-subjects variable, trait anxiety, were determined by median split scores on the STAI Trait-Anxiety scale. Two variables were manipulated within subjects: word type (neutral or threatening) and presented word frequency (two, ®ve or eight). The dependent variable for the 2 (group) 2 (word type) 3 (frequency) comparison was the estimated frequency of occurrence of the words. 2.3.1. Acquisition Participants were tested in groups of four to 14. After reading and signing the consent form and ®lling out a demographic questionnaire, participants were instructed that they would be presented with a long list of words that would be shown on slides at a fairly rapid rate. They were told to pay close attention to each of the slides and to look at each word for as long as it appeared on the screen, and that they would be tested on these words afterwards. No mention was made of a frequency test. Words were then presented at a rate of 3 s per word. 2.3.2. Frequency and recall tests Immediately following the acquisition phase, participants were given the frequency test forms. The instructions that appeared on the top of the forms were read aloud. Participants recorded the number of times they thought they had seen each word and forms were collected as soon as each participant ®nished. As soon as all participants ®nished the frequency test, the recall forms were passed out and the instructions that appeared on the top of each form were read aloud. Participants were given a maximum of 6 min to recall as many words as possible in any order. These forms were collected from each participant as soon as he or she indicated completion. 2.3.3. Psychometric measures After all participants completed the frequency test, they were given the STAI and instructed
400
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
to answer all questions as best they could. They were then debriefed and given the opportunity to ask questions. 3. Results 3.1. Participants The participants were ®rst divided into high and low trait anxiety groups on the basis of a median split on the STAI Trait scale. The median STAI Trait score was 38 so all individuals who scored below 38 were included in the low trait anxiety group (n = 74) and all individuals who scored at or above 38 were included into the high trait anxiety group (n = 85). The mean STAI trait scores for the low and high trait anxious groups were as follows: 31.15 (SD=4.17) and 46.60 (SD=7.66). The high trait anxiety group was signi®cantly higher than the low trait anxiety group in state anxiety, t (157)=ÿ 7.76, P < 0.001 and state anxiety was signi®cantly correlated with trait anxiety, r = 0.66, P < 0.001. The mean ages of the participants in the low and high groups were as follows: 19.49 (SD=0.18) and 19.12 (SD=0.12). There were no age dierences between the groups, t (157)=0.10, NS. 3.2. Frequency estimates Each subject's frequency estimates were averaged over each of the actual frequency conditions. The means of these mean frequency estimates were subjected to a repeated measures analysis of variance (ANOVA) that included trait anxiety as the between-group factor and word type (neutral or threatening) and presentation frequency (two, ®ve, or eight) as within-group factors. The means and standard deviations of the frequency estimates can be viewed in Table 1. There were signi®cant main eects of word type, F (1,157)=80.35, P < 0.0001, and presentation frequency, F (2,314)=365.11, P < 0.0001, and a signi®cant interaction Table 1 Mean frequency estimates for presented and distractor words Groupa
Presentation frequencies 0
Low M SD High M SD a
2
5
8
N
T
N
T
N
T
N
T
0.38 0.87
0.52 0.88
2.63 1.17
3.49 1.75
4.42 1.67
4.26 1.66
5.34 1.80
6.38 2.30
0.48 0.80
0.93 1.20
2.70 1.08
4.16 1.80
4.48 1.55
4.78 1.85
5.99 2.20
6.98 2.63
N denotes neutral words, T denotes threatening words.
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
401
between word type and frequency, F (2,314)=28.55, P < 0.0001. Simple eects analyses indicated that larger frequency estimates were given to threat words than neutral words at presentation frequencies two, F (1,157)=87.29, P < 0.0001 and eight, F (1,157)=55.17, P < 0.0001, but not at presentation frequency ®ve, F < 1.00. The central prediction of an interaction between word type and anxiety group was signi®cant, F (2,314)=3.88, P = 0.05. Although all participants estimated relatively higher frequencies for threatening than neutral words, the eect was more pronounced for anxious participants. Simple eects analyses indicated that the eect of trait anxiety collapsed across presentation frequencies was signi®cant for threatening words, F (1,157)=4.68, P < 0.05, but not for neutral words, F (1,157)=1.43, NS. 3.3. Recall The mean numbers of neutral words recalled were 8.74 and 8.32 by the low and high trait anxiety groups respectively. The mean numbers of threatening words recalled were 9.78 and 9.00 by the low and high trait anxiety groups respectively. A two-way repeated measures ANOVA was performed with trait anxiety as the between-group factor and word type as the within-subjects factor. There were no recall dierences between the groups, F (1,157)=1.63, NS. Overall, threatening words were recalled better by both groups, F (1,157)=13.41, P < 0.001. There was no interaction between trait anxiety and word type, F < 1.00. 3.4. Recognition The zero frequency level data were analyzed separately to examine the possibility that the high trait anxious participants were exhibiting a response bias, that is whether they were more likely than the low trait anxiety participants to give higher estimates of threatening words regardless of information stored in memory. Recognition was assessed by asking the participants to rate the frequency of the presented words (old) and the distractor words (new). Since there was no true frequency of one, an estimate of the probability of detecting a threat word vs a neutral word was made by de®ning hits as the proportion of words from presentation level two correctly recognized as having occurred (i.e., subject gave a frequency estimate greater than zero). False alarms were de®ned as the proportion of distractor words which were given frequency estimates greater than zero. The group means were then subjected Table 2 Mean proportion of hits, false alarms and corrected recognition by trait anxiety groups Groupa
Low High a
Hits
False alarms
Corrected
N
T
N
T
N
T
0.92 0.90
0.92 0.94
0.18 0.22
0.21 0.32
0.74 0.67
0.71 0.62
N denotes neutral words, T denotes threatening words.
402
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
to a two-way repeated measures ANOVA with trait anxiety as the between-group factor and word type as the repeated measure. The resulting means are displayed in Table 2. An analysis of hit rates resulted in a main eect of word type, F (1,157)=4.09, P < 0.05, with threat words being given higher estimates than neutral words. For false alarms, the interaction between group and word type was signi®cant, F (1,157)=4.85, P < 0.05. High trait anxiety participants made more false alarms than low trait anxiety participants on threatening words, F (1,157)=4.29, P < 0.05, but not on neutral words, F < 1.00. Subtracting the proportion of false alarms from the proportion of hits gave an estimate of corrected recognition for the presented words. There were no recognition dierences between the groups, F (1,157)=2.54, NS, nor any group by word type interactions, F < 1.00. Both groups recognized more neutral words than threatening words, F (1,157)=4.83, P < 0.05.
4. Discussion As predicted, individuals with high trait anxiety estimated greater frequencies for threatening words than individuals with low trait anxiety, but similar frequencies for neutral words. There were no dierences between the groups on recall. These ®ndings support a distinction between high and low trait anxious individuals in memory for frequency-of-occurrence information related to threat. The distinction appears to be one of degree rather than kind. Both groups estimated higher frequencies for threatening words than for neutral words and both groups recalled more threatening than neutral words. The similarities between the groups deserve comment. Based on the pattern of ®ndings for frequency estimates and recall, it is reasonable to suggest that all participants were using an availability heuristic to some extent. The threat words were more accessible than the neutral words for frequency judgments and recall. Certain word characteristics were not controlled (e.g., categorization and imagability) and may have made the threatening words easier to recall. For example, the threat words were all from one category (physical threat) and the neutral words were from dierent categories. In addition, it is unknown what the carry-over eects were from the estimate to the recall task. Exposure to the words during the frequency estimation task may have eliminated potential recall dierences between groups, although the failure to ®nd an explicit memory eect in trait anxiety is in line with previous research (e.g., Mathews et al., 1989; Richards & French, 1991). The dierences between groups were evident only on frequency estimates of threatening words: words that were presented and words that never occurred. Neither an automatic frequency counting mechanism at encoding nor the use of an availability heuristic at retrieval can easily explain the group dierences on distractor estimates. Compared to the low trait anxious participants, the high trait anxious participants appeared to be using a more lenient criterion for judging the occurrence of threatening words. They were more likely to report seeing a threatening word, whether or not it had occurred. Whether or not they actually believed they had seen the threat distractors or whether they were simply guessing when uncertain is not clear from this study. Alternatively, if high trait anxious individuals do have more extensive and elaborated threatrelated schemata in memory which are activated by the presence of threatening events (in this
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
403
case words), it may be that they are also more likely to automatically activate the related distractor concepts in memory. In addition, frequency judgments increase when events are highly salient or vivid such as when they receive extra attention in the media (Lichtenstein, Slovic, Fischho, Layman & Combs, 1978; Tversky & Kahneman, 1973, 1974). If high trait anxious individuals thought about or imagined more threat-related events during the study or if their imaginings of threat were more vivid, they may have had greater diculty distinguishing between the two sources in memory (internal vs external) and had more uncertainty at test. False recognition of distractors may re¯ect the inability to distinguish between real and imagined events in memory (Johnson & Raye, 1981). To test this possibility, in two subsequent studies an attempt was made to hold the frequency of internal and external generations constant between the high and low trait anxious individuals (Kverno, 1994). Once constrained, the dierences in threat frequency estimates between the high and low anxiety groups disappeared. In addition, compared to the low trait anxious individuals, the high trait anxious individuals, no longer exhibited a greater speci®c response bias on threat distractors. 5. Implications The ®ndings have implications for theories concerning the global nature of subjective probabilities related to threat observed in individuals with anxiety (Butler & Mathews, 1987; Constans & Mathews, 1993; Johnson & Tversky, 1983). Not only do high trait anxious individuals estimate higher frequencies for threatening future events, they estimate higher frequencies for threatening events that have previously occurred (a speci®c eect) and for related threatening events that have never occurred (a global eect). High trait anxious individuals believe that they are more at risk for harmful events (e.g., being assaulted or harmed). The results of this study suggest that when high trait anxious individuals are asked to monitor or estimate the frequencies of personally or clinically relevant threatening events (e.g., panic attacks, phobic fears, pain, binges, depressive episodes), the estimates might also be in¯ated. An important area for future research will be to further examine the mechanisms by which anxiety in¯uences frequency monitoring. References Beck, A. T., & Clark, D. A. (1997). An information processing model of anxiety: automatic and strategic processes. Behaviour Research and Therapy, 35(1), 49±58. Birnbaum, I. M., Taylor, T. H., Johnson, M. K., & Raye, C. L. (1987). Is event frequency encoded automatically? The case of alcohol intoxication. Journal of Experimental Psychology: Learning, Memory and Cognition, 13(2), 251±258. Butler, G., & Mathews, A. (1987). Anticipatory anxiety and risk perception. Cognitive Therapy and Research, 11(5), 551±565. Byrne, A., & Eysenck, M. W. (1993). Inidividual dierences in positive and negative interpretive biases. Personality and Individual Dierences, 14(6), 849±851. Carroll, J. B., Davies, P., & Richman, B. (1971). Word frequency book. New York: American Heritage. Constans, J. I., & Mathews, A. M. (1993). Mood and the subjective risk of future events. Cognition and Emotion, 7(6), 545±560.
404
K.S. Kverno / Personality and Individual Dierences 29 (2000) 395±404
Dalgleish, T. (1994). The relationship between anxiety and memory biases for material that has been selectively processed in a prior task. Behaviour Research and Therapy, 2, 227±231. Eysenck, M. W., MacLeod, C., & Mathews, A. (1987). Cognitive functioning and anxiety. Psychological Research, 49, 189±195. Fisk, A. D. (1986). Frequency encoding is not inevitable and is not automatic: a reply to Hasher and Zacks. American Psychologist, (February), 215±216. Gasper, K., & Clore, G. L. (1998). The persistent use of negative aect by anxious individuals to estimate risk. Journal of Personality and Social Psychology, 74(5), 1350±1363. Hasher, L., & Zacks, R. T. (1979). Automatic and eortful processes in memory. Journal of Experimental Psychology: General, 108(3), 356±388. Hasher, L., & Zacks, R. T. (1984). Automatic processing of fundamental information. The case of frequency of occurrence. American Psychologist, 39(12), 1372±1388. Hintzman, D. L. (1969). Apparent frequency as a function of frequency and the spacing of repetitions. Journal of Experimental Psychology, 80(1), 139±145. Johnson, E. J., & Tversky, A. (1983). Aect, generalization, and the perception of risk. Journal of Personality and Social Psychology, 45(1), 20±31. Johnson, M. K., & Raye, C. L. (1981). Reality monitoring. Psychological Review, 88(1), 67±85. Kverno, K. (1994). Frequency monitoring and anxiety. Unpublished doctoral dissertation, The George Washington University, Washington, DC. Lichtenstein, S., Slovic, P., Fischho, B., Layman, M., & Combs, B. (1978). Judged frequency of lethal events. Journal of Experimental Psychology: Human Learning and Memory, 4(6), 551±581. MacLeod, C., & Campbell, L. (1992). Memory accessibility and probability judgments: an experimental evaluation of the availability heuristic. Journal of Personality and Social Psychology, 63(6), 890±902. MacLeod, C., & Cohen, I. L. (1993). Anxiety and the interpretation of ambiguity: a text comprehension study. Journal of Abnormal Psychology, 102(2), 238±247. Manis, M., Shedler, J., Jonides, J., & Nelson, T. E. (1993). Availability heuristic in judgements of set size and frequency of occurrence. Journal of Personality & Social Psychology, 65(3), 448±457. Mathews, A., & MacLeod, C. (1994). Cognitive approaches to emotion and emotional disorders. Annual Review of Psychology, 45, 25±50. Mathews, A., Mogg, K., May, J., & Eysenck, M. (1989). Implicit and explicit memory bias in anxiety. Journal of Abnormal Psychology, 98(3), 236±240. Reidy, J., & Richards, A. (1997). Anxiety and memory: a recall bias for threatening words in high anxiety. Behaviour Research and Therapy, 35(6), 531±542. Richards, A., & French, C. C. (1991). Eects of encoding and anxiety on implicit and explicit memory performance. Personality and Individual Dierences, 12(2), 131±139. Spielberger, C. D., Gorsuch, R. L., Lushene, R., Vagg, P. R., & Jacobs, G. A. (1983). Manual for the State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press, Inc. Thorndike, E. L., & Lorge, I. (1944). The teachers' word book of 30,000 words. New York: Columbia University. Tversky, A., & Kahneman, D. (1973). Availability: a heuristic for judging frequency and probability. Cognitive Psychology, 5, 207±232. Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: heuristics and biases. Science, 185, 1124±1131. Watkins, M. J., & LeCompte, D. C. (1991). Inadequacy of recall as a basis for frequency knowledge. Journal of Experimental Psychology: Learning, Memory, and Cognition, 17(6), 1161±1176. Wiggs, C. L., Martin, A., & Sunderland, T. (1997). Monitoring frequency of occurrence without awareness: evidence from patients with Alzheimer's Disease. Journal of Clinical and Experimental Neuropsychology, 19(2), 235±244. Williams, J. M. G., Watts, F. N., MacLeod, C., & Mathews, A. (1988). Cognitive psychology and emotional disorders. NY: John Wiley & Sons.