Behac. Rex Thu. Vol. 28. No. 6. pp. 455168. Printed in Great Britain. AJI rights reserved

1990 Copyright

INVITED

WHY WORRY?

c

OOQ5-7967,90 53.00 + 0.00 1990 Ptrgamon Press plc

ESSAY

THE COGNITIVE ANXIETY

FUNCTION

OF

ANDREW MATHEWS Department of Psychology, Louisiana State University, Baton Rouge, LA 70803-5501, U.S.A. (Received

24 July

1990)

Summary-The phenomenon of worry is considered to arise from cognitive processes involved in anxiety, that serve to maintain high levels of vigilance for personal danger. Rather than rely on self-report alone, the research described here draws on information processing methodology, to investigate this hypothesized cognitive function. Evidence is summarized to show that anxious subjects selectively attend to threatening information, and interpret ambiguous events in a relatively threatening way. However, the evidence on memory suggests that although such information may be easily activated, it is not necessarily more accessible. The allocation of attentional priority to threatening information is seen as a characteristic of anxious (rather than depressed) mood, while the ease with which this processing mode is adopted may underlie trait anxiety and vulnerability to anxiety disorders.

Common-sense has it that worry is pointless. (“Why worry?-it may never happen’*). Although often distressing, it will be argued here that anxiety and worry serve an important cognitive function. It is well established that fear triggers biological reactions that favor rapid flight (or behavioral inhibition; Gray, 1982). That anxiety and worry have related cognitive functions is perhaps less obvious. In fact, anxiety will be shown to involve a characteristic pattern of cognitive processing, having the effect of maintaining high levels of vigilance for possible danger. It could be objected that this is nothing new: vigilance is known to be enhanced in states of arousal (Eysenck, 1982), and elevated vigilance is used as a diagnostic criterion for clinical anxiety states (DSM III-R, American Psychiatric Association, 1987). Furthermore, under a variety of names such as sensitization-repression (Krohne, 1978), or monitoring-blunting (Miller, 1987), as well as vigilance-avoidance (Mathews & Ridgeway, 1981; Suls & Fletcher, 1985); many researchers have described an individual cognitive style in the response to perceived threat. Highly vigilant individuals (sensitizers or monitors) are said to be more likely than are avoiders to seek out or attend to information relating to potential threat. While a connection between anxiety and vigilance is thus an old idea, it will be argued here that the adoption of methods and concepts developed in the study of normal information processing allows vigilance to be defined and measured more objectively. Furthermore, this approach leads to a more precise and testable formulation of cognitive theories of anxiety (e.g. Beck & Emery, 1985) than has previously been possible. Before describing anxiety research using information processing methodology, I will first consider the picture that emerges from self-report studies of cognitive content. Despite the inherent limitations in the scope and validity of this approach, it may be helpful to clarify the nature of reportable content associated with anxiety, and identify some of the experiential phenomena that need to be accounted for in any satisfactory theory. Such phenomena can then be used to develop specific and testable hypotheses accounting for excessive worry. Finally, the remainder of this review focuses on evidence for or against these hypotheses, and on the theoretical view of anxiety that emerges.

Cognitire

content

in anxiety

and worry

Direct questioning of clinically anxious Ss has led to the observation that episodes of anxiety are associated with aversive thoughts about social or physical dangers (Beck, Laude 8: Bohnert, 1974; Hibbert, 1984). However, apart from the general problem of the validity of self report data, these studies did not include control or comparison groups, leaving it unclear whether such thoughts are more common in highly anxious individuals.

456

ASDREW MATHEWS

An alternative approach has been to select normal individuals who are either relatively high or low in the reported frequency of worry, and to compare them (Borkovec, Robinson, Pruzinsky & DePree, 1983). Worry is used here to describe the persistent awareness of possible future danger, which is repeatedly rehearsed without being resolved. One consistent finding, from both interview and self monitoring data, is that thoughts labelled as worrying seem particularly difficult to dismiss from one’s mind. This seems to be true even of those who do not consider themselves to be ‘worriers’, if they are in a negative mood state at the time. Clinically anxious clients, and self-iabelled worriers, report experiencing particular problems in excluding worrisome thoughts (Borkovec, Shadick & Hopkins, 1990). In exploratory work on worry content, we have carried out several cluster analyses of worry frequency ratings with normal popuIation samples, to investigate whether content falls into consistent domains (Tallis, 1989). Some evidence of consistency was obtained across samples, with the most commonly recurring clusters being concerned with close relationships, social confidence, life accomplishments, work competence, financial solvency, and general world problems (e.g. pollution). A comparison of clinically anxious and normal Ss showed that the anxious clients worried more about all personal topics, but not general world problems, where there was no difference. This suggests that excessive worry is mainly concerned with personal and emotional threats to self, rather than with dangers in a more general or abstract sense. Other findings suggest that the conscious content of worry varies according to current life circumstances. Students report worrying most about academic progress and family or other interpersonal problems (Borkovec er al., 1983), while an elderly sample said they worried most about their physical health (Wisocki, 1988). A few other studies have contrasted the reported frequency of worries in those with anxiety disorder and controls, and again found that patients worry more frequently, and about a wider range of topics. Both anxious and control Ss reported on interview that their most common worries were family or interpersonal problems (Borkovec et al., 1990); whereas the use of self-monitoring with anxious Ss suggested that health worries were more frequent (Craske, Rapee, Jacket & Barlow, 1989). Lt’orry and physiological arousal

An objection that can be raised to the use of the term worry is that, in the absence of any agreed definition of either worry or anxiety, the two terms cannot be usefully distinguished. One way of distinguishing the two is to reserve the term worry for the specifically cognitive (or experiential) component of anxiety, without invoking other components, such as physiological arousal. Factor analytic studies of anxiety questionnaires typically reveal two or more underlying factors, one concerned with the awareness of somatic symptoms, and another with more cognitive aspects such as intrusive and unwanted thoughts (Morris, Davis & Hutching, 1981; Lehrer & Woolfolk, 1982). These two factors, somatically based emotionality and worry, appear to have different correlates. For example, worry appears to be the better predictor of deficits in the performance of test anxious Ss (Deffenbacher, 1980). On the other hand, reported episodes of worry are not associated with raised levels of physiological arousal, other than left frontal EEG activation (Carter, Johnson & Borkovec, 1986). Verbally mediated thoughts appear to predominate over visual imagery in worry (Borkovec & Inz, 1990), and induced worry paradoxically inhibited the usual cardiovascular response to a subsequent phobic image (Borkovec & Hu, 1990). If worry is verbally encoded, and is independent of physiological arousal, then the term may capture a more specific and more cognitive process than does the relatively broad construct of anxiety. Worry as preparation for danger It is proposed here that the phenomenon of worry is an important clue to the nature of anxiety; and in particular to its cognitive function. In the same way that fear has been described as a biological alarm system preparing the organism for escape (Barlow, 1988), so worry can be seen a special state of the cognitive system, adapted to anticipate possible future danger. In contrast to other species, humans can readily represent possible future events within the cognitive system, and thus worry may be a uniquely human phenomenon. Our personal experience of worry suggsts that we are rehearsing possible aversive events and outcomes, and at the same time searching for

Why worry? The cognitive function of anxiety

457

ways of avoiding them. This process would appear to have obvious advantages in that we are less likely to be taken by surprise when an anticipated threat does in fact materialize, and may indeed be better prepared to cope with it as a result. However, the more that this process is successful, the less likely are we to label the process as worry, as opposed to preparatory coping, or problem solving. Worry thus resembles problem-solving in some respects: but instead of leading to a satisfactory outcome, it is as if the danger is constantly being rehearsed without a solution ever being found. In extreme cases of excessive worry, entirely imaginary threat scenarios may be constructed, having no reference to the real world at all, and thus clearly incompatible with successful problem solving. Normal

and abnormal

worry

Anxiety and worry are common to many emotional disorders, but in generalized anxiety disorder, worry is the primary diagnostic criterion. Thus, according to DSM III-R (American Psychiatric Association, 1987) to be diagnosed as suffering from generalized anxiety a patient must demonstrate “unrealistic or excessive anxiety and worry (apprehensive anticipation) about two or more life circumstances”. Individuals with generalized anxiety disorders (GAD) would therefore seem to provide a particularly clear example of excessive worry. The question then arises of what distinguishes those who worry excessively in this way, from those who do not, given identical circumstances. One possibility is that individuals may vary in their sensitivity to cues signalling the existence of future threats. Indeed, it is argued later that excessive worry and abnormal anxiety states are associated with an enhanced tendency to detect, and selectively process, threat cues in the environment. People who worry little may simply be those who rarely detect environmental cues signalling potential threat. A second (related) possibility, is that individuals who worry excessively tend to select the more emotionally threatening interpretation of cues relating to possible aversive events. This possibility arises because many of the cues that may signal danger in everyday life are ambiguous in nature. For example. episodes of worry about possible rejection by a friend, or breakup of a relationship, may be triggered by external cues, such as unexpected lack of warmth or affection, whether real or imagined. Such cues may signal a real threat, or they may be attributable to irrelevant factors such as a temporary preoccupation on the part of the other person. A tendency to select the more threatening possibility or interpretation of such cues may be associated with a greater experience of worry. A similar, but more specific bias, favoring the catastrophic interpretation of internal bodily cues has been proposed to account for panic disorder (Clark. Salkovskis, Gelder, Koehler, Martin, Anastasiades, Hackmann, Middleton & Jeavons, 1988). Thus, both internal and external cues may be subject to interpretative bias, perhaps leading to different types of worry or anxiety disorder. The third possibility is that intrusive thoughts about aversive outcomes reflect differences in the way in which information about danger is stored in long term memory. According to Michael Eysenck (1985) individuals high in trait anxiety may have more tightly organized clusters of worry-related information in long term memory, which are responsible for the intrusive, repetitious and uncontrollable nature of excessive worry. This idea is related to Bower’s (1981) network model of mood and memory, in which mood-congruent information is organized in a network around a central emotional node. If this model is extended to take account of individual differences, then perhaps those prone to excessive worry are characterized either by more anxiety congruent material in the network, or by material in memory which is more threatening in content. Whenever anxiety is elicited, activation is assumed to spread automatically through the network of worry-related information, making it more accessible. Thus, if worry-prone (or high trait-anxious) individuals have more such information in memory, they should be more likely to recall threatening events. These three possibilities are neither mutually exclusive, nor are they exhaustive. Several other possibilities exist, including differences in cognitive strategy or behavioral response to threats, that may inadvertently serve to exacerbate rather than ameliorate worry. Distracting oneself from worry-related cues may only maintain their impact, in the same way that avoidance slows the extinction of conditioned fear. Alternatively, if worry is seen as being more like unsuccessful attempts at problem solving, then perhaps worriers are trying to solve insoluble problems.

4%

ASDREW MATHEWS

However, these latter possibilities do not appear to have been investigated using experimental techniques, and the research to be described has focused on the first three alternatives. Selectice processing and interference from emoti5nu~ st~rn~~i One way of attempting to measure the extent to which processing resources are selectively allocated to threatening stimuli, is to test how much the incidental presence of such stimuli interferes with an ongoing task. For example, the Stroop color-naming task requires that Ss name the color that a word is written in while ignoring the word’s content. Differences in color-naming speed, dependent on the irrelevant word content, may be taken as an index of the extent to which word meaning has been selectively processed. Thus, spider phobics are slower in color-naming words related to spiders than are non-phobic controls, and this relative slowing is eliminated by desensitization treatment (Watts, McKenna, Sharrock & Trezise. 1986). Since treatment involves increased exposure to phobic words, the decreased interference after treatment indicates that a high frequency of usage is unlikely to account for the original slowing effect. Rather, it depends on the emotional significance that the word meaning holds for that individual. The version of the test that we have used (Mathews & MacLeod, 1985), contained words related to physical and social threat (e.g. ‘disease’ or ‘criticized’), or were non-threatening control words. When non-anxious controls were required to color-name sets of threatening or non-threatening words, there were no significant differences in coior-naming time as a function of word content. By contrast, Ss suffering from clinical anxiety (GAD) were significantly slower in the color-naming task when the words were threatening in content. In a recent replication (Mogg, Mathews & Weinman, 1989) the degree of interference was found to vary according to reported worries or concerns, as well as mood state. Ss who reported predominantly physical worries, such as illness, were slowed most by words such as death or disease; whereas those who were more concerned with social worries, were slowed more by words such as criticism or rejection. Similar effects have been found in depressed Ss (Gotlib & McCann, 1984; Williams & Nulty, 1986), and in those reporting panic attacks (Ehlers, Margraf, Davies & Roth, 1988) using words that matched the concerns of these individuals. Thus, the slowing effect does not seem to be unique to generalized anxiety disorders, but is apparent whenever words match the content of the S’s worries. It has recently been suggested that interference may be even less specific than this, and depend only on ‘emotionality’: that is, interference may arise from any emotional stimulus, including positive words (Martin, Williams & Clark, 1988). To test this hypothesis, we (Mathews & Klug, in preparation} selected both negative and positive words that were matched for emotionality ratings, but varied in their relevance to the concerns of anxious clients: for example, ‘crazy’ or ‘relaxed’ are considered to be relevant, while ‘quarrel’ or ‘romantic’ were not. Results clearly showed that both positive and negative words can cause interference, but only those that were relevant to worry content, This finding suggests that interference associated with positive words may be caused by close semantic associations between them and their negative opposites, which form the focus of current worry. Such emotional interference effects can be explained by assuming that all words presented to a S will be automatically processed for meaning, but subsequently disregarded if they are irrelevant to the present task. However, when the meaning of a word is in fact highly related to a current concern, rejection as irrelevant is more difficult and therefore slower. Correspondingly, if the task is to detect a word related to that concern, it will be quicker and easier to perform (e.g. Parkinson & Rachman, 1981). Before giving further consideration to the implications of this interpretation, it may first be useful to discuss other evidence concerned with attentional bias in anxiety. Anxiety and attention Although interference in color naming may result from irrelevant words capturing attentional resources (Kahneman & Chajczyk, 19831, this is only one of several possible mechanisms. In fact, the explanation above assumes interference can arise from mechanisms operating later than attentional capture in the processing sequence (Seymour, 1977). Because interference can arise in several possible ways, the Stroop test cannot be used to demonstrate that attention is drawn towards threatening stimuli in anxious Ss. For example, interference could be produced by the subsequent emotional reactions to words, or by a defensive shift in attention away from them.

Why worry?

The cognitive

function

of anxiety

459

However, there is now quite compelling evidence, from another experimental paradigm, that attentional shifts towards threatening stimuli are indeed characteristic of anxiety. In an initial unpublished experiment by a student of Michael Eysenck (C. Haliopoulos), high trait anxious Ss were found to detect tones in a dichotic tape more rapidly if these followed a threatening word. Converting this to the visual modality, we found a similar effect in clinically anxious (GAD) Ss, (MacLeod, Mathews & Tata, 1986). Word pairs were displayed on a computer screen, and Ss required to detect the occasional appearance of a small dot, in a location just vacated by one of the two words. Anxious Ss were faster to detect the dot if it replaced a threatening word, while non-anxious controls were faster when the dot replaced a non-threatening word. Thus it appears that the anxious Ss were more likely to have their attention captured by the more threatening word in a pair, while controls seem to show a reverse shift, towards the non-threatening word in a pair. On the other hand, when depressed Ss were tested with the same method, they did not appear to distinguish between threatening or control words. This apparent attentional indifference in depressed Ss was confirmed in a subsequent experiment using words chosen to be relevant to depression (Gotlib, McLachlan & Katz, 1988). Assuming that these results are reliable, their implications are quite intriguing. Interference in color-naming is common to both anxiety and depression, but absent in controls. In contrast, the attentional vigilance effect is present in anxiety, but not in depression, and may even be reversed in non-anxious controls. In real life, we could suppose that anxious (but not depressed) individuals would have their attention caught by any fleeting or inconsequential cues associated with threat, while the same cues would be actively neglected by those who are less anxious. An interesting question then arises. Is attentional vigilance a consequence of anxious mood (a state effect); or is it a more enduring characteristic of those who are prone to anxiety (a trait effect)? If it is an enduring cognitive characteristic, then attentional vigilance may be a cause, rather than only a consequence, of worry and anxiety. Relation of vigilance to trait and state anxiety Because trait and state measures tend to be closely correlated, disentangling their effects can be difficult. In fact, measures of color-naming interference tend to correlate with both, sometimes with state anxiety being more closely associated (e.g. Mathews & MacLeod, 1985) and sometimes trait (Mogg et al., 1989; Richards & Millwood, 1989). Transient induction of depressed mood produced little effect (Gotlib & McCann, 1984), and the effect that was found with depressive disorders disappeared on remission (Gotlib & Cane, 1987). Thus, although the pattern of results is not yet entirely clear in anxiety, it seems that a combination of state and trait factors is the best predictor of color-naming interference in depression. Using the dot-detection measure of attention with non-clinical Ss, Broadbent and Broadbent (1988) found evidence of a small but significant vigilance effect that was associated with trait anxiety levels, with state anxiety levels having a less reliable effect. Most interesting was the finding that the relation with trait levels was non-linear, so that attentional vigilance was unaffected by trait scores in the low to moderate range, but increased exponentially at the very highest levels. State anxiety only predicted vigilance when the trait score was also high, again suggesting an interactive effect, but one that only operates at the upper end of the distribution of trait anxiety scores. Another way of testing this idea is to assess the vigilance of high and low trait Ss, when their state levels are also high or low. In a study of students assessed some time before, and then immediately prior to an important examination, high trait Ss showed more evidence of attentional vigilance for examination-related words as the examination approached (MacLeod & Mathews, 1988). Interestingly however, the low trait students did not show this effect, and if anything tended to shift attention away from examination-related words as the examination approached. This finding strongly supports the hypothesis of an interaction between trait and state, such that high trait anxious subjects become increasingly vigilant under stress, perhaps further enhancing their anxiety level. In contrast to this positive feedback loop, low trait anxious Ss may show a defensive response under stress, serving to restrain further anxiety increases via a negative feedback loop. Unfortunately, other results suggest a further complication. Acute stress, in the form of a contrived failure on an anagram task, seems to produce vigilant reactions to all threat cues, for both high and low trait students (Mogg, Mathews, Bird & MacGregor-Morris, 1990). Perhaps such transient

460

ANDREW %‘fATHEWS

increases in vigilance are initially typical of both groups, but while high trait Ss maintain or increase vigilance levels over time, low trait Ss can reverse this effect through the use of adaptive coping strategies. A final piece of evidence in favor of both state and trait effects comes from a study of recovery in clinically anxious Ss (Mathews, May, Mogg & Eysenck, 1990). Ss were required to ignore various types of distracters, and search instead for a neutral target word. As expected, the presence of threat distracters slowed detection of the target by anxious Ss, relative to controls. However, all distracters, irrespective of emotional content, disrupted the performance of the currently anxious group, whereas only threat distracters slowed the recovered group. Since these effects were only apparent when the position of the target was unknown, we assume that distraction resulted from attentional capture during search. If so, then attentional capture by threat would seem to be a function of anxiety proneness, while state anxiety is additionally associated with a general distractability effect. To summarize, there is some evidence indicating a link between persistent vigilance for threat and trait anxiety level, with more tentative indications that stress may potentiate this effect. While the causal relationships involved remain uncertain, it is plausible that the ease with which individuals adopt a vigilant attentional style under stress, underlies proneness to anxiety. Since this style will lead to the intake of more information concerned with potential danger, it could well play a part in triggering or maintaining episodes of worry and anxiety. Selective attention or pre-attentive bias?

In the tasks described so far, Ss could easily become aware of the nature of the emotional words used. For this reason it is impossible to know whether interference or attentional vigilance arise from automatic processes that are not dependent on awareness, or whether voluntary control processes are involved. For reasons to be discussed below, we believe that at least some of the processes involved are automatic, so that they may occur involuntarily and without Ss being aware of their nature. Initial attempts to investigate this question used the dichotic listening technique, in which sensitivity to unattended stimuli in one audio channel is measured, while Ss actively attend to another channel. For example, unattended sexually explicit words are particularly likely to intrude involuntarily into awareness, and especially so in Ss with high manifest anxiety scores (Nielsen & Sarason, 1981). Obsessive-compulsive Ss are better at detecting words associated with dirt or contamination than controls, and this effect disappears after successful treatment (Foa & McNally, 1986). Using a simultaneous secondary reaction time task, Bargh (1982) has shown that performance can be slowed due to the presence of personally relevant words, even when Ss are unable to report on the nature of these unattended stimuli. We have found similar effects arising from unattended threatening words in clinically anxious, but not control Ss (Mathews & MacLeod, 1986), again in the apparent absence of awareness for any of the critical words. Anxious Ss, but not controls, were significantly slower on a simple reaction time task in trials involving unattended threat words. Since Ss were unable to report on the content of these words, we assumed that threat cues may be selectively processed, and cause emotional interference, prior to reaching awareness. The problems of inferring lack of awareness in dichotic listening tasks have been extensively discussed by Holender (1986). It is impossible to be sure that effects do not arise because Ss rapidly shift their attention between channels, and become transiently aware of the critical content, but then forget it equally rapidly. While none of the existing paradigms can totally rule out such transient processes, it is possible to severely limit subjective awareness using very brief word presentations followed by a pattern mask. In one such study, clinical and control Ss were presented with threatening or neutral words very briefly (16 msec) in color, with the display continued in white for half of the trials only, until the Ss responded by naming the color of the original display (Mathews, MacLeod & Tata, cited in Mathews, 1988). After some practice, the Ss could see and identify the flash of color, but reported that they could not read the words on the brief presentation trials. Although small in magnitude, there were significant interference effects from threat words in both anxious and depressed clients, relative to controls, and these were not significantly different for brief versus prolonged presentation trials. Even if momentary awareness cannot be completely ruled out, these results demonstrate that a drastically curtailed ability to report on the nature of

Why worry? The cognitive function of anxiety

461

stimuli does not necessarily lead to a corresponding reduction in their capacity to produce interference effects. In the clinical context, this would suggest that when individuals cannot say why they began to worry or feel anxious, it remains possible that environmental threat cues triggered the worry episode. It thus appears that anxiety and worry are associated with an automatic processing bias, initiated prior to awareness, but serving to attract attention to environmental threat cues, and thus facilitating the acquisition of threatening information. We speculate that anxious individuals may come to have the impression of being in some personal danger as a result. Moreover, since this attentional capture effect is presumed to be automatic, anxious Ss may not always be able to report on the reasons for their sense of danger. Depressed individuals also seem to process emotionally threatening information automatically, although apparently without equivalent consequences for attention.

emotional

The inflation of subjective risk in anxiety The second hypothesis put forward about excessive worry was that ambiguous events might be more likely to be interpreted as threatening by those prone to anxiety. Apart from the clinical observations supporting this idea, there is strong evidence that, under condition of uncertainty, anxious individuals will estimate the risk of danger as being higher than will non-anxious controls. In an initial study (Butler & Mathews, 1983), anxious (GAD) and depressed Ss picked more threatening interpretations of ambiguous scenarios, and tended to estimate the risk of future aversive events as being higher than did controls; particularly when these risk estimates were being made for oneself as opposed to other people. Similar effects following experimental mood inductions have been reported by Bower (1983) and Johnson and Tversky (1983). Others have found that agoraphobic and panic patients also estimate risks more highly than normals, particularly when they concern body sensations (Clark et al., 1988); and these estimates normalize after recovery (McNally & Foa, 1987). These data support the conclusion that an inflation of subjective risk is the outcome of negative mood states, rather than being a characteristic of certain individuals. Furthermore, the effect of mood seems to be similar for different negative emotions (at least for anxiety and depression), and across a wide range of events (e.g. Johnson & Tversky, 1983). There may be qualifications to these conclusions, however. The inflation that accompanies stress appears to be fairly global in high trait anxious Ss, but more specific to the event causing an increase in anxiety in low trait Ss (Butler & Mathews, 1987). Also, there is evidence that inflations are greater in areas of personal concern to the individual concerned (e.g. Clark et al., 1988). In unpublished Ph.D. work, Gillian Butler has found fairly high correlations between self-reported worry about particular topics and the inflation of subjective risk for those topics. There thus seems to be an association between the extent that individuals worry about an event, and how much they see themselves at risk from that event. It remains unclear at this stage whether there are any direct causal links between worry and subjective risk. It seems likely that an inflated perception of risk would lead to concern and worry, but there is also evidence that worry can serve to further inflate subjective risk, so that each may influence the other. For example, cognitive manipulations that increase availability of reasons why an event could occur, even if totally imaginary in nature, tend to increase the subjective probability of that event (Tversky & Kahneman, 1973; Gregory, Cialdini & Carpenter, 1982). The interpretation of ambiguity An alternative hypothesis is that both worry and subjective risk are driven by an interpretative bias in which ambiguous cues are more likely to be interpreted as threatening. Some theories of semantic access assume that ‘all possible meanings of an ambiguous word are processed in parallel, until one is selected and reaches awareness (Simpson, 1984). A selection bias favoring the more threatening meaning of a stimulus could thus operate in the same way that attention can be captured by the more threatening of two environmental cues. Suppose, for example, that you were to notice a small lump on your skin which had not been apparent before. Depending on interpretation, this could be seen as an inconsequential blemish, or as the first warning of skin cancer, and thus a source of considerable concern. To the extent that everyday life is full of such

162

ANDREW,hfATHEWS

ambiguous signals, a bias in the interpretation of threat cues could be a relatively common precipitant of worry episodes. Equally, when making judgements of the risk of uncertain outcomes, the more threatening of several alternatives may be seen as more probable than the rest. In an experiment on the interpretation of ambiguous cues (Eysenck, MacLeod & Mathews, 1987), Ss listened to lists of words having two distinct spellings, one corresponding to a threatening and one to a non-threatening meaning (homophones such as die/dye, pain/pane etc). The number of threatening spellings used when writing the words was positivley correlated with trait anxiety scores. In a second study of clinical anxiety (GAD), a currently anxious and a recovered group again listened to a tape-recorded list of words, and wrote down what they heard (Mathews, Richards & Eysenck, 1989). Currently anxious Ss used spellings which implied a threatening interpretation more often than controls, with the recovered group being intermediate. Finally, in another two studies of clinical anxiety, Ss listened to sentences that could be interpreted in different ways, such as “The two men watched as the chest was opened”. The interpretation chosen was determined by performance on a subsequent recognition test using sentences corresponding to one possible disambiguation (e.g. “The two men watched as the patient’s chest was cut open”). Recovered and control Ss endorsed more of the non-threatening sentences, while currently anxious Ss endorsed (relatively) more of the threatening versions. A signal detection analysis confirmed that this was more likely to reflect the interpretation made than a simple response bias (Eysenck, Mogg, May, Richards & Mathews, 1990). Other disorders may reveal a similar, but perhaps more specific, bias in the interpretation of ambiguous material. For example, Clark et al. (1988) exposed panic patients to partial sentences, such as “My heart pounding could mean that I was. . .“; followed by a lexical decision task. Words such as ‘dying’, that imply a catastrophic interpretation of symptoms, were primed more in the panic group than in controls, but other types of threatening sentence did not result in differential priming. It seems plausible that interpretative bias will be most obvious when there is a match between pre-exisiting concerns and one meaning of the ambiguous material. However, a negative bias may be apparent only when worries are currently active, since some evidence suggests that recovered patients resemble normals in showing a reverse effect, favoring positive interpretations. This positive bias in normals, that was also noted in the attentional data, may be part of a protective mechanism whereby anxious mood is usually avoided. Non-depressed Ss have similarly been observed to demonstrate a positive self-serving bias in the recall of personal events (Alloy 8~ Abramson, 1979). However, rather than concluding that normals are biased and anxious or depressed individuals are not, we assume that the processing of emotionally or personally relevant information is almost always biased, albeit in different ways. At an early stage of processing, pre-attentive bias can direct attention towards or away from threat cues, with both the direction and sensitivity of this process being associated with anxiety level. When such cues match a current concern or worry, the decision to ignore or process the information further must be made, and this may cause interference with other tasks, such as color-naming. These interference effects seem to be less specific to anxiety than attentional capture, and occur in a variety of mood states. Finally, when stimuli are ambiguous with respect to threat value, selection of meaning is required, and this decision process is also influenced by mood state and current concerns. The structure of worry in memory

As indicated previously, the theoretical approaches of Gordon Bower (1981), and Michael Eysenck (1983), emphasize links in memory among mood-congruent representations. Applied to worry and anxiety, Bower’s model predicts that activation spreads through a network of worry-related information, linked via a central anxiety node. When activated, either by the induction of anxious mood, or by priming worry representations in memory, the network should make all threat related information more accessible, both perceptually and from memory. This model has been successful in predicting negative recall bias (in depression), attentional vigilance (in anxiety), interference from emotional words, and global inflation of subjective risk (in both). Given these successes, one might confidently expect that excessive worry would be associated with a bias favoring the recall of threatening information. In fact, the findings in this area have been complex and confusing.

Why worry? The cognitive function of anxiety

463

Early claims that phobic Ss preferentially recalled words related to their fears (Nunn, Stevenson & Whalen, 1984) can be criticized methodologically, and have not been replicated (Pickles & van den Broek, 1988). One subsequent study provided evidence of mood congruent recall in panic disorder (McNally, Foa & Donnell, 1989). Results with normal Ss have also been mixed, with some indicating a bias favoring negative self-descriptors in high anxious groups (Breck & Smith, 1983) and others failing to find significant effects (Foa, McNally & Murdock, 1989). In generally anxious Ss, not only have we failed to find clear indications of a mood congruent bias in memory, but in one experiment there was a significant effect in the reverse direction, with normal controls recalling proportionally more threatening words (Mogg, Mathews & Weinman, 1987). Failure to find consistent evidence of a bias in recall associated with anxiety thus represents a potential problem for Bower’s model. Failures

to find cognitive bias in anxiety

Negative findings with recall of threatening words is not our only failure to find mood congruent effects in anxiety. For example, in unpublished pilot work, we asked anxious Ss to identify visually degraded pictures that were either pleasant (such as a wedding) or unpleasant in nature (such as a funeral). Anxious individuals showed no differences in inspection time, or relative advantage in identi~cation speed or accuracy for the two types of picture. Similarly we have found no difference in the speed of identifying threat or no-threat words, either when presented tachistoscopicatly, or in a lexical decision task. Other failures with the lexical decision task have been reported in depression (see Clark, Teasdale, Broadbent & Martin, 1983; MacLeod, Tata & Mathews, 1987). One explanation of these negative findings is that mood states in general, and anxiety in particular, do not affect all cognitive processes equally. Rather, we suppose that anxiety is associated primarily with an attentional bias, that accords priority to threatening information. When the task involves only one stimulus, as in the lexical decision task, there are unlikely to be group differences, because in the absence of competition, attentional priority becomes irrelevant. Competition for attentional resources may thus be a necessary, although not sufficient, condition for the presence of bias in anxiety. In other unpublished studies investigating this issue, we have found evidence of more rapid lexical decisions for threatening words, but only when two stimuli are presented Simultaneo~sIy (MacLeod & Mathews, described in Mathews, 1988). A subsequent replication carried out by Karin Mogg has provided essentially similar findings, consistent with the hypothesis that lexical decisions are speeded only when there is opportunity for attention to be allocated differently. These findings suggest the need to revise models that predict similar effects of mood across all aspects of emotional information processing. Anxiety and cognitive avoidance

Despite this argument, it still seems difficult to understand why the results of expe~ments on recall of threat by anxious Ss should be so variable and inconclusive, in striking contrast to the consistent findings in depression. If anxious individuals attend selectively to threat cues, then threatening information should be preferentially registered in memory. Even if the to-be-remembered words are presented without competing stimuli, priority could still be given to threat representations during retrieval If, however, initial registration of threat is followed by cognitive avoidance responses, recall could be relatively impaired. We have speculated that anxiety, unlike depression, is associated with a tendency to avoid elaboration on the associates or attributes of threatening information (Williams, Watts, MacLeod & Mathews, 1988). Elaboration is used here to describe the various voluntary processes by which Ss link an activated representation with other information in memory (Mandler, 1980). Thus we suppose that highiy charged emotional events may be encoded primarily in terms of their immediate emotional significance (e.g. as aversive and/or dangerous), while other aspects are ignored or even inhibited. This could explain why traumatic or phobic experiences lead to reports of intrusive memories, but paradoxically poor performance on recognition tasks that require detailed discrimation among aversive stimuli (Watts, Trezise & Sharrock, 1986; Christiansen & Loftus, 1987). Similarly, failure of highly anxious Ss to iink threat words with other information in memory, would handicap later voluntary retrieval of these specific words (Mogg et al., 1987). Despite this

464

ANDREWMATHEWS

apparent difficulty in voluntary retrieval, the quality of worrisome thoughts (for example, their persistent and repetitious nature) suggests that they are easily activated automatically. Implicit memory for threat in anxiety To investigate this issue, we adapted the word completion method employed by Graf and Mandler (1984). Ss completing a stem with the first word that comes to mind, are more likely to use a word they have seen recently, even wher. they claim not to recognize it (implicit memory). However, completion is relatively unaffected by encoding strategies, and for this reason Graf and Mandler argue that it reflects automatic activation of the memory trace, rather than voluntary elaborative processes. Following presentation of threatening and control words, we required anxious and non-anxious Ss to complete three-letter stems with the first word that came into their mind, or to use the same three-letter stems as a recall cue (Mathews, May, Mogg & Eysenck, 1989). The pattern of cued recall did not differ significantly across groups, but anxious Ss produced proportionally more threat word completions. This effect was significant only for words that had been presented in the list, and not for new three-letter stems. It would thus appear that the threat word representations were more strongly primed in anxious Ss, although recall of threat was not clearly facilitated. The implication would seem to be that threatening information is stored in such a way as to be more easily activated by cues in anxious Ss, but that its voluntary retrieval may be handicapped by cognitive avoidance. If so, then perhaps this could explain how intrusive thoughts and worries could be automatically triggered by cues that only partly resemble danger representations in memory. Furthermore, perhaps this combination of easy activation and restricted elaboration, accounts for the persistent nature of worry, since repeated intrusions followed by (short-term) avoidance could prevent either permanent storage in less emotive form (Rachman, 1980) or the incorporation of corrective information (Foa & Kozack, 1986). Anxiety as a mode of the cognitive system

The foregoing account suggests that a network model, such as that described by Bower (198 I), is insufficient to explain cognitive processes in anxiety. Rather, our present view is more closely related to the theoretical account of Oatley and Johnson-Laird (1987), which proposes a limited number of basic emotions, each having evolved to meet a specific need, and each corresponding to a unique organizational mode of the cognitive system. In any specific emotional state, the cognitive system is organized in a manner appropriate for dealing with the new set of priorities arising from particular types of event. For example, anxiety is said to occur when a background self-preservation goal is perceived to be at risk. The related cognitive re-organization serves the function of maintaining vigilance until the threat has been avoided or no longer exists. This theoretical framework is quite consistent with our finding that the cognitive biases seen in anxiety and depression seem to involve quite different processing operations. We suppose that anxiety is associated with the assignment of high priority to those processes involved in detecting and acting on information related to danger, and a corresponding reduction in the cognitive resources available for other tasks. By contrast, sadness is said to be provoked by failures in major plans, or loss of active goals (Oatley & Johnson-Laird, 1987), serving to motivate a search for new plans and goals. The processes given priority are thus more likely to be those concerned with elaboration, facilitating the retrieval of related events from memory. Although it is difficult to see clinical depression as having survival value, perhaps it represents an extreme (and maladaptive) form of normal sadness; and therefore shares the same mode of cognitive organization. With these considerations in mind, we can now return to the interaction between trait and state that was discussed earlier. It is proposed that anxious mood can be regarded as the experience corresponding to the cognitive system entering the vigilant mode, in which priority is given to those cogntive operations necessary to detect and interpret potential threats. However, there is the problem that, under some circumstances, low trait individuals seem to respond to stress by shifting attention away from threat cues. To account for this we might suppose that potentially threatening information is being monitored to some extent all the time, but that until a certain threshold of probable danger has been reached, the system does not switch into its vigilant mode. Below this threshold, attentional resources may be directed away from threat cues to allow other tasks to be

Why worry? The cognitive function of anxiety

165

pursued without distraction. Once a threshold level of threat has been reached, however, the system switches into vigilant mode, external and/or internal threat cues are more likely to enter awareness, and individuals may report episodes of anxiety or worry. In the case of low trait individuals who report that they worry relatively little, we assume that the threshold is usually set relatively high, and the system typically remains in ‘defensive’ mode. Of course, in order that this does not have maladaptive consequences, such as failing to avoid a real danger, once a sufficient magnitude of threat is detected, the switch to vigilant mode must occur even in low trait anxious individuals. In contrast to such adaptive responding, those reporting excessive worry and high levels of trait anxiety are assumed to have thresholds set very low, causing them to enter a vigilant mode relatively easily. As a result, they will attend selectively to threat cues, interpret ambiguous events in a threatening way, and experience repeated intrusive thoughts about potential threats. Prolonged experience of this vigilant mode may have the additional effect of accumulating more information about potential dangers, and thus increasing the intensity of the anxiety experienced. In extreme cases, this excessive worry could lead to the breakdown of normal adaptive behavior described as an anxiety disorder. CONCLUSIONS

AND

PROSPECT

Like most research endeavors, the work described here has raised as many questions as it has answered. Many of the suggestions made are likely to prove incomplete, or incorrect. Much depends on the idea that anxious individuals selectively attend to threat cues, although it is not yet clear how specific the effect is. In the color-naming task, interference seems to be greatest when the material used matches current concerns, although no such evidence of specificity has been found on the attentional deployment task. This could be explained if it is assumed that attention is initially drawn towards any potential threat (or other object of interest), thus allowing for better stimulus identification, while interference effects depend on a match with current concerns. Although the case has been argued for a stronger attentional bias in anxiety than depression, and vice-versa in the case of mood-congruent recall, the confused state of the evidence does not justify any unequivocal conclusions. Even if the argument should prove valid, it is likely to be a matter of degree, rather than an absolute distinction. In any case, the frequent coexistence of anxiety and depression would make such an absolute distinction difficult to detect. It may be more fruitful to focus on the extent to which specific automatic or controlled processes underlie the cognitive biases that have been documented so far. For example, if negative elaboration underlies the recall bias seen in depression; then it should be possible to modify the bias by teaching Ss to encode material more positively. To investigate whether emotional trait variables influence how information is structured in memory, it might be more productive to use measures of automatic processing. Implicit measures of memory (word completion), suggest that threat representations are stored in a form that makes them easier to activate in highly anxious Ss, but that links with other types of information may be restricted. Effective treatments for anxiety states, such as exposure, may work by linking threat representations to other more positive representations in memory, as suggested by Foa and Kozack (1986). Rather than relying on explicit measures such as recall to test these ideas, it would seem more appropriate to use measures of automatic processes, such as priming. Another unresolved issue is that of whether interference tasks (such as color-naming, and dichotic listening), measure the same processes as tasks designed to index the deployment of attention. Because interference can arise in many different ways, it has been argued that these former tasks cannot always be taken to reflect attention alone. Rather, it is proposed that emotional states can cause interference due to difficulty in rejecting congruent information as irrelevant, at subsequent stages of processing. Further, it is assumed that this process is associated with greater attentional deployment towards the congruent information in anxiety, but with greater elaboration on negative implications for oneself in depression. At this stage of research however, this account can only be speculative. Perhaps the most important unanswered question is of whether the biases that have been found have a causal role in worry and anxiety, or whether they are merely secondary consequences of mood state. The nature of attentional and interpretive biases makes it plausible that they could

466

ASDRE~ MATHEWS

at least maintain, and possibly augment a pre-existing anxious mood state. In vulnerable (high trait) individuals, attentional vigilance provoked by stress does not seem to decline over time as it does in low trait Ss. As a result, the former should be more likely than the latter to remain aware of possible dangers, and thus maintain anxious mood at an elevated level. Because most evidence of negative bias is eliminated following clinical recovery, it seems unlikely that this anxiety elevating process is dictated by enduring trait variables alone. Rather, it seems more likely that readiness to adopt a vigilant cognitive mode (the postulated mechanism underlying trait anxiety), interacts with stressful life events to determine anxious mood state. Further studies of stress effects in high and low trait Ss, and clinical studies of the processes involved in recovery and relapse, are now needed to test this hypothesis. To conclude; does anxiety serve a useful cognitive function? Anxiety and worry are adaptive, in preparing us for future dangers that often require representation in imaginal form. But anxiety seems to be a rather inflexible mode of the information processing system; it accords priority to some processes and inhibits others in a relatively stereotyped way. Thus, if handling a threat requires creative thought, or even worse, when there is no actual problem or danger in the real world, then worry will probably cause more problems than it resolves. In anxiety disorders, worry results in unnecessary distress, and disrupts the ability to concentrate on anything else. Use of information processing methodology to study the cognitive function of worry, may be an important step towards improving our ability to understand and treat such distressing clinical conditions. Ackno~ledgemenrs-Much of the research on which this article is based was supported by grants from the Medical Research Council, U.K. and from the Wellcome Trust. The invaluable contribution of ideas and help from many colleagues is gratefully acknowledged: especially Tom Borkovec, Gillian Butler, Michael Eysenck, Cohn MacLeod, Jon May, Karin Mogg. Anne Richards, Fraser Watts, Mark Williams and Phillip Tata, as well as many others; although final responsibility for errors and deficiencies remains with the author. Thanks are due to Jim Geer and Valerie Ridgeway for helpful comments on the manuscript.

REFERENCES Alloy, L. B. & Abramson, L. Y. (1979). Judgement of contingency in depressed and non-depressed students: Sadder but wiser? Journal of Experimental Psychology, IO8, 441485. American Psychiatric Association (1987). Diagnostic and statislical manual of mental disorders (3rd edn, rev.). Washington, D.C.: APA. Bargh, J. A. (1982). Attention and automaticity in the processing of self-relevant information. Journal of Personality & Social Psychology, 43, 425-436.

Barlow, D. H. (1988). Anxiefy and ifs disorders. New York: Guilford. Beck, A. T. & Emery, G. (1985). Anxiety disorders and phobias: A cognitive perspective. New York: Basic Books. Beck, A. T., Laude, R. & Bohnert, B. (1974). Ideational components of anxiety neurosis. Archives of General Psychiatry, 31, 319-325.

Borkovec, T. D. & Hu, S. (1990). The effect of worry on cardiovascular response to phobic imager)l. Beha-ciour Research and Therapy, 38, 69-73.

Borkovec, T. D. & Inz, J. (1990). The nature of worry in generalized anxiety disorder: A predominance of thought activity. Behaviour Research and Therapy, 28, 153-158.

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. Borkovec, T. D., Shadick, R. & Hopkins, M. (1990). The nature of normal and pathological worry. In Barlow, D. & Raper, R. (Eds), Chronic anxiery and Generalized Anxiety Disorder. New York: Guilford Press. Bower, G. H. (1981). Mood and memory. American Psychologist, 36, 129-148. Bower, G. H. (1983). AfTect and cognition. Philosophical Transactions of Ihe Royal Society of London, 8302, 38742. Breck, B. E. & Smith, S. H. (1983). Selective recall of self-descriptive traits by socially anxious and nonanxious females. Social Behavior & Personality, II, 71-76.

Broadbent, D. & Broadbent, M. (1988). Anxiety and attentional bias: State and trait. Cognifion % Emorion, 2, 165-183. Butler, G. & Mathews, A. (1983). Cognitive processes in anxiety. Advances in Behaviour and Therapy, 5, 51-62. Butler, B. & Mathews, A. (1987). Anticipatory anxiety and risk perception. Cognitive Therapy and Research, 91, 551-565. Carter, W. R., Johnson, M. C. & Borkovec, 5. D. (1986). Worry: An electrocortical analysis. Aaiances in Behaviour Research and TheraDv, 8, 193-204.

Christiansen, S. & Loft& E. F. (1987). Memory for traumatic events. Applied Cognitive Psychology, I, 225-239. Clark, D. M., Teasdale, J. D., Broadbent, D. E. & Martin, M. (1983). Effect of mood on lexical decisions. Bulletin of the Psychonomic Society, 21, 175-178.

Clark, D. M., Salkovskis, P. M., Gelder, M., Koehler, C., Martin, M., Anastasiades, P., Hackmann, A.. Middleton, H. & Jeavons. A. (1988). Tests of a coanitive theory of panic. In Hand, I. & Wittchen, H. U. (Eds). Panic and phobias (Vol. 2, pp. 149-158). Berlin: Springer. . Craske, M. G., Rapee, R. M., Jackel, L. & Barlow, D. H. (1989). Qualitative dimensions of worry in DSM-IIIR generalized anxiety disorder subjects and non-anxious controls. Behaviour Research and Therapy, 27, 397-402. Deffenbacher, J. L. (1980). Worry and emotionality. In Sarason, I. G. (Ed.), Tesr anxiety: Theory, research and applications. Hillsdale, N.J.: Erlbaum.

Why worry? The cognitive function of anxiety

467

Ehlers, A., Margraf, J.. Davies, S. & Roth, W. T. (1988). Selective processing of threat cues in subjects with panic attacks. Cognition and Emotion, 2, 201-219.

Eysenck, M. W. (1982). Attention and arousal: Cognition and performance. Berlin: Springer. Eysenck, M. W. (1985). Anxiety and the worry process. Bulletin of Psychonomic Society, 22, 545-548. Eysenck, M. W., MacLeod, C. & Mathews, A. (1987). Cognitive functioning in anxiety. Psychological Research, 49, 189-195. Eysenck, M. W., Mogg, K., May, J., Richards, A. & Mathews, A. (1990). Bias in interpretation of ambiguous sentences related to threat in anxiety. ~olournulof Abnormal Psychology. In press. Foa, E. B. & Kozak, M. J. (1986). Emotional processing of fear: Exposure to corrective information. Psycho~og~cafBulletin, 99, 20-35.

Foa, E. B. & McNally, R. J. (1986). Sensitivity to feared stimuli in obsessive-compulsives: Cognitive Therapy % Research,

A dichotic listening analysis.

10, 411486.

Foa, E. B., McNally, R. J. & Murdock, T. B. (1989). Anxious mood and memory. Behaviour Research and Therapy, 27, 141-147.

Go&b, f. H. & Cane, D. B. (1987). Construct accessibility and clinical depression: a longitudinal investigation. Journal of Abnormal Psychology, 96, 199-204. Gotlib, I. H. & McCann, C. D. (1984). Construct accessibility and depression: An examination of cognitive and affective factors. Journal of Personality & Social Psychology, 47, 427-439. Gotlib, I. H., McLachlan, A. L. % Katz, A. N. (1988). Biases in visual attention in depressed and nondepressed individuals. Cognition and Emotion, 2, 185-200.

Graf, P. & Mandler, G. (1984). Activation makes words more accessible, but not necessarily more retrievable. Journal of Verbal Learning and Verbal Behavior, 23, 553-568.

Gray, J. A. (1982). The neuropsychology of anxiety. Oxford: Clarendon Press. Gregory, W. L., Cialdini, R. B. & Carpenter, K. M. (1982). Self-relevant scenarios as mediators of likelihood estimates and compliance: Does imagining make it so? Journal of Perso~u~jty & Social Psychology, 43, 89-99. Hibbert, G. A. (1984). Ideational components of anxiety: their origin and content. British Journal of Psychiatry, 144, 618-624.

Holender, D. (1986). Semantic activation without conscious identification. Behavioral Brain Science 9. l-66. Johnson, E. J. & Tversky, A. (1983). Affect, generalisation, and the perception of risk. Journal of Personality and Social Psychology, 45, 20-3 1. Kahneman, D. & Chajzzyk, D, (1983). Tests of the automaticity of reading: Dilution of Stroop effects by Color-irrelevant stimuli. Journal of Experimental Psychology: Human Perception and Performance, 9, 497-509. Krohne, H. W. (1978). Individual differences in coping with stress and anxiety. In Spielberger, C. D. & Sarason, I. G. (Eds), Stress and anxiety (Vol. 5, pp. 233-260). Washington, D.C.: Hemisphere. Lehrer, P. M. & Woolfolk, R. L. (1982). Self-report assessment of anxiety: Somatic, cognitive and behavioral modalities. Behaoiorol Assessment, 4, 167-177.

MacLeod, C. & Mathews, A. (1988). Anxiety and the allocation of attention to threat. Quarterly Journal of Experimental Psychology, 40, 653-670.

MacLeod, C., Mathews, A. % Tata, P. (1986). Attentional bias in emotional disorders. Journal of Abnormal Psychology, 95, 15-20.

MacLeod. C., Tata, P. & Mathews, A. (1987). Perception of emotionally valenced information in depression. British Journal of Clinicnl Psychology, 26, 67-68.

McNally. R. J. & Foa, E. B. (1987). Cognition and agoraphobia: & Research, If, 567-581.

Bias in the interpretation

of threat. Cognitiue Therapy

McNally, R. J., Foa. E. B. & Donnell, C. D. (1989). Memory bias for anxiety information in patients with panic disorder. Cognition & Emotion, 3, 27-M.

Mandl&, G. (1980). Recognizing: The judgement of previous occurrence. Psychological Rmiew, 87, 252-271. Martin. M.. Williams. R. & Clark. D. M. (1988). Does anxietv lead to selective urocessinn of threat-related information? Paper presented at the World Congress of Behaviour Therapy, Edinburgh, Scotland.Mathews, A. (1988). Anxiety and the processing of threatening information. In Hamilton, V., Bower, G. H. & Frijda, N. H. (Eds). Cognitive perspectives on emotion and motioution (pp. 265-284). London: Kluwer Academic. Mathews, A. & Klug. F. Paper in preparation. Mathews. A. M. & MacLeod, C. (1986). Discrimination of threat cues without awareness in anxiety states. Journal of Abnormal Psychology, 95, 131-138.

Mathews. A. & MacLeod, C. (1985). Selective processing of threat cues in anxiety states. Behaoiour Research und Therapy. 23, 563-569.

Mathews, A. & Ridgeway, V. (1981). Personality and surgical recovery: A review. Briiish Journal of Chnical Psychology, 20, 243-260.

Mathews, A., May, J., Mogg, K. & Eysenck, M. (1990). Attentional bias in anxiety: Selective search or defective filtering? four& of Abnormal Psychology, 99, 166-I 73. Mathews, A., Mogg, K., May, J. & Eysenck, M. (1989). Implicit and Explicit Memory Bias in Anxiety. Journaf of Abnormal Psychology, 98, 236-240.

Miller, S. M. (1987). Monitoring and blunting: Validation of a questionnaire to assess styles of information-seeking under threat. Journal of Personality & Social Psychology, 52, 345-353. Mogg, K., Mathews, A. & Weinman. J. (1987). Memory bias in clincial anxiety. Journafof Abnormaf Psychofogy, 96,94-98. Mogg, K., Mathews. A. & Weinman. J. (1989). Selective processing of threat cues in anxiety states: A replication. Behaciour Research and Therapy, 27, 317-323.

Mogg, K.. Mathews, A., Bird, C. & MacGregor-Morris (1990). Effects of stress and anxiety on the processing of threat stimuli. Journal of Personality and Social Psychology. In press. Morris, L. W., Davis, M. A. & Hutchings, C. H. (1981). Cognitive and emotional components of anxiety: Literature review and a revised worry~motionality scale. Journal of Educarional Psychology, 73, 541-555. Nielsen, S. L. & Sarason, 1. G. (1981). Emotion, personality, and selective attention. Journal of Personality and Social Psychology, 41, 945-960.

468

ANDREW M.CHEWS

Nunn, J. D., Stevenson, R. & Whalan, G. (1984). Selective memory effects in agaraphobic patients. British Journal of Clinical Psychology, 23, 195-201. Oatley, K. & Johnson-Laird, P. (1987). Towards a cognitive theory of emotions. Cognition and Emotion, I, 29-50. Parkinson, L. & Rachman, S. (1981). Intrusive thoughts: The effects of an uncontrived stress. Advances in Behuuiour Research and Therapy, 3, Ill-

118.

Pickles, A. J. & van der Broek, M. D. (1988). Failure to replicate evidence for phobic schemata in agoraphobic patients. British Journal of Clinical Psychology, 27, 271-321.

Rachman, S. (1980). Emotional processing. Behuuiour Research and Therapy, 18, 51-60. Richards, A. & Millwood, B. (1989). Colour-identification of differentially valenced words in anxiety. Cognition and Emotion, 29, 171-176. Richards, A. & Whittaker, T. M. (1990). Effects of anxiety and mood manipulation in autobiographical memory. British Journal of Clinical Psychology, 29, 145-154.

Seymour, P. H. K. (1977). Conceptual

encoding and locus of the Stroop effect. Quarterly Journul of Experimenrul

Psychology, 29, 245-265.

Simpson, G. B. (1984). Lexical ambiguity and its role in models of word recognition. Psychological Bulletin, 96, 316340. Suls, J. & Fletcher, B. (1985). The relative efficacy of avoidant and non-avoidant coping strategies: A meta-analysis. Health Psychology, 4, 249-288.

Tallis, F. (1989). Experimental investigations of worry. Unpublished Ph.D. thesis, University of London, England. Tversky, A. & Kahneman, D. (1973). Availability: A heuristic for judging frequency and probability. Cognitive Psychology, 5. 207-232.

Watts, F. N., Trezise, L. & Sharrock, R. (1986). Processing of phobic stimuli. Brirish Journal of Clinical Psychology, 2J, 253-261.

Watts, F. N., McKenna, F. P., Sharrock, R. & Trezise, L. (1986). Colour naming of phobia related words. British Journul of Psychology, 77, 97-108.

Williams, J. M. G. & Nulty, D. D. (1986). Construct accessibility, depression and the emotional Stroop task: Transient mood or stable structure? Personality and Individual Differences, 7, 485-49 I. Williams, M.. Watts. F., MacLeod, C. & Mathews, A. (1988). Cognitive Psychology and Emotional Disorders. New York: Wiley. Wisocki. P. A. (1988). Worry as a phenomenon relevant to the elderly. Behavior Therapy, 19, 369-379.

Why worry? The cognitive function of anxiety.

The phenomenon of worry is considered to arise from cognitive processes involved in anxiety, that serve to maintain high levels of vigilance for perso...
2MB Sizes 0 Downloads 0 Views