Abstract
Considerable attention has been given to accounts of psychological or cognitive diatheses to emotional problems such as depression. Various postulates have been explored in the literature, with different theoretical arguments presented.
In two recent studies we explored a ‘lock and key’ hypothesis of adult depression [1, 2]. Here, we review our findings and the relevant literature on cognitive risk, as well as consider other empirical tests of cognitive vulnerability theories, and with focus on why only few studies have found support for cognitive schema theories.
The lock and key hypothesis posits that early adverse life events or circumstances are capable of establishing vulnerability ‘locks’ which may be later primed when the individual is faced with mirroring life events (‘keys’) in adulthood. The hypothesis assumes that locks are cognitive structures, akin to concepts such as cognitive schemas or core beliefs. Essentially a lock and key hypothesis is in line with other diathesis-stress models of cognitive vulnerability (e.g. Abramson et al. [3]) in that an ‘unlocking’ role of a salient stressor in those with a cognitive diathesis to depression is assumed. This premise holds that early adversity, which may be acute or chronic in nature, creates a cognitive template within the individual which translates into a type of accepted and reifying self-knowledge or more general world knowledge.
Clearly, such a lock and key hypothesis to depression is not novel, however, it emphasizes the salience or specificity of the stressor or key in unlocking the cognitive diathesis. Several theorists have posited various accounts of cognitive vulnerability to psychological disturbance, particularly depression [3–6]. The theoretical and practical conceptualization of cognitive vulnerability to emotional disturbance has an extensive history, with most current accounts of vulnerability (which emphasize the developmental construction of mistaken beliefs about the self and the environment, e.g. ‘I am inferior to others’) being extensions of early work undertaken by Alfred Adler (see Ansbacher and Ansbacher [7]).
Those who prescribe schema models argue for the role of environmental stressors in the activation of such core beliefs [4, 8, 9]. Most related theories of depression view cognitive vulnerability as essentially traceable to some environmental or life event origin (usually in childhood). For instance, Young [6] identified a set of early maladaptive schemas (EMS) which refer to stable and constant themes that emerge during childhood (e.g. ‘I am a failure’ or ‘I know that I am unlovable’). These schemas are said to operate as templates for the processing of later experience. Schemas are considered to be a priori truths that are implicit and readily taken for granted by the individual.
Rather than conditional beliefs like the ‘faulty underlying assumptions’ proposed by Beck [4], EMSs are more concrete and unconditional. According to Young [6], EMSs are more likely to emerge from ongoing patterns of noxious experiences (rather than isolated traumatic events) and usually develop out of dysfunctional relationships with caregivers or significant others during early life. This account of schema development is similar to related theories of vulnerability to depression including Bowlby's [10, 11] attachment theory which provides an account of cognitive vulnerability relevant to interpersonal contexts. He proposed that dysfunctional attachments in childhood are internalized mentally into models which dictate the nature of future interpersonal relationships, and that poor attachments have a continual impact upon the individual's emotional stability or instability throughout life. Another account not unlike Bowlby's is that of object relations theory [12], which describes the intrapsychic representation of interpersonal relationships and the evolution of self–other patterns over time.
All contemporary schema theories define schemas as stable underlying constructs which contribute to the core of the individual's self-concept. They also propose that such constructs operate to create distress via more superficial cognitive biases which result in the dysfunctional assimilation of environmental and intrapersonal data, and that the distortion of information regarding the self and the environment has the potential to give rise to negative automatic thoughts and subjective distress [4, 6,13–15]. Operative schemas are said to also result in errors in logical thinking (e.g. selective abstraction) and are also closely related to cognitive vulnerability models of explanatory or attributional style [16]. Explanatory style, which refers to people's explanations for the causes of events, is closely linked with locus of control theories originally posited by Bandura [17].
The link between causal attributions, control-related beliefs and depression is well illustrated by the hopelessness theory of depression [3, 18] which developed from the earlier learned helplessness theory [19]. According to the hopelessness theory, symptoms of hopelessness manifest largely from one's faulty expectations; namely that highly aversive events or outcomes will occur and/or conversely, that highly desired outcomes will not occur. Crucial to the notion of ‘hopelessness’ is the belief that the individual has no response in their repertoire that will change the likelihood of occurrence of these outcomes. Thus, there is a strong argument to suggest that cognitive schemas (or core beliefs with a life-event origin) can play an active role in determining depressogenic expectations of future events, or further, that schemas become expressed through expectations.
Weinstein [20] has proposed that an individual's ‘perceived vulnerability’ to harm or negative experiences relates to their former history of such experiences. He [20–24] established that individuals without aversive background events tended toward optimism in evaluating their own vulnerability to negative life events. In other words, in order to progress through life and deal with day-to-day routine, most people assume a certain ‘invulnerability’ or ‘rosy glow’. This invulnerability is likely to serve a ‘protective’ purpose. Thus, by giving little (if any) attention to the possibility of a random life crisis, the individual can remain uncompromised by anxiety and pessimism [25]. Personal experience of early aversive events may work to undermine any such optimistic bias, and may manifest in a vulnerability producing a lower general threshold to life stressors, or a susceptibility to specific ones. It follows then, that for people with early experience of a negative event (e.g. loss, abuse, threat to personal safety), the bias toward optimism may be reduced and their vulnerability to decompensate in response to a future (real or perceived) negative event is increased, either generally or with specificity to only salient events. The extent to which ‘general’ versus ‘specific’ vulnerability is created is an issue of some importance, both theoretically and in designing therapeutic paradigms. A lock and key hypothesis of depression argues for specific or selective vulnerability.
Hammen and Goodman-Brown [26] studied the relationship between children's self-schema and the onset of depressive symptoms following exposure to lifeevent stressors which were relevant to the self-schema. They found that the extent of ‘matching’ between negative life events and type of pre-existing self-schema was significantly related to depression outcome over a 6-month period, particularly for children of depressed mothers. Studies comparing subjects with experience of serious and/or threatening events and those without such experience, have found that individuals with a history of an aversive set of circumstances (e.g. victimization) have significantly greater ‘perceived vulnerability’ to psychological distress when exposed to like circumstances in the future (e.g. victimization) than individuals without such past experiences [27, 28], hence supporting the role of specificity in vulnerability.
Mireault and Bond [29], for example, found that in a sample of college students, bereaved subjects perceived themselves to be more likely to be exposed to subsequent loss than did a group of non-bereaved controls. Their study, together with others that have shown early loss to be significantly predictive of higher degrees of ‘generalized worrying’ [30, 31], supports the view that individuals are less likely to perceive themselves to be as vulnerable to threatening events if they lack any earlier experience of such events [20]. Mireault and Bond [29] found that such vulnerability (i.e. worry, the expectation of future loss and a heightened vigilance to loss cues) was a stronger predictor of anxiety and depression than was the actual loss itself.
Some models of cognitive risk suggest that depressogenic cognitions as vulnerability factors are more relevant for recurrent depressions than first episodes due to the influence of information processing during past episodes [32]. In support of this thesis, Lewinsohn et al. [33] found that the link between dysfunctional thinking and depressive symptoms, was greater for adolescents with a previous history of major depression, than for those without a history of depression.
Brown et al. [34] studied a sample of women with depression, and proposed that vulnerability factors play a crucial role because they compromise an individual's ability to develop and maintain an optimistic view about controlling their environment. With such a vulnerability already wired, it becomes clearer how a further stressful event (especially if salient) can be actively involved in the onset of some depressions. This concept of susceptibility to depression, is further understood as a lowered threshold for experiencing feelings of hopelessness when exposed to stressors [35]. A previous episode of depression and hopelessness, or a previous negative event (e.g. a loss), may advance some form of induced sensitivity to subsequent stressors, however minor. Thus, minor but relevant events, may well possess the ability to ‘bring home’ the implications of a major problem. For example, the threat of abuse for a person with an endangerment history, may act to bring home the implication of future abuse and could reinforce their expectation that future abuse is certain. Triggers which are minor, but which have specificity to the existing vulnerability (here a schema in the form of an expectation), have the potential to feed into one's expectation of a worst possible scenario.
Using a sample of depressed adults, we undertook two studies to explore the issue of specificity within cognitive vulnerability to depression, examining the role of early and later life events and the significance of held core beliefs or schemas.
Lock-key links in a clinical sample
In our initial study [1] we sought to explore some of these assumptions by investigating the existence of lockkey links (and lock-key specificity) within a sample of 270 clinically depressed patients. We examined for associations between early life events (i.e. incidents or experiences possibly responsible for vulnerability locks) and precipitating ‘keys’ (i.e. events triggering depression). Patients were interviewed to ascertain both any early adverse experiences and the life events which precipitated their depressive episode. First, patients were required to rate their degree of exposure to a list of childhood environment characteristics (e.g. violence, left alone a lot, ongoing criticism) with options being ‘not at all’, ‘slight’, ‘moderate’ or ‘severe’ and coded 0–3. Later in the interview, patients were asked to consider the life-event precipitants to their current depressive episode (‘keys’) and to judge which one most ‘affected’ them or let them down the most. Patients were then asked to rate the extent to which that particular stressor corresponded with a list of response descriptors (i.e. the extent to which the stressor led to feelings of being ‘violently or physically abused’, ‘abandoned’, ‘criticized’, for example). Options again were ‘not at all’, ‘slightly’, ‘moderately’ or ‘severely’ and coded 0–3. See Parker et al. [1] for a detailed coverage of methods.
We then examined for associations between these two lists (locks and keys). Correlational analyses indicated several significant associations between similar ‘early’ stressors (locks) and themes or responses evoked by ‘precipitant’ stressors (keys). Three molar locks (‘abuse’, ‘insecurity’ and ‘loss’) and three molar keys (‘rejection’, ‘unsafe/danger’ and ‘control/abuse’) (i.e. lock and key domains) were created from these original lists of lock and key stressors. We explored associations between these lock and key domains and found three separate patterns of associations. First, there was evidence of a general vulnerability effect, whereby exposure to an ‘insecure’ emotional base in childhood was linked similarly to our three molar keys (or precipitant classes): ‘rejection’, ‘unsafe/danger’ and ‘control/abuse’. Second, there was some evidence of specificity whereby early ‘abusive’ stressors were more closely associated with abusive and dangerous life-event precipitants. Third, absence of any link in some analyses. Here, early ‘loss’ stressors (including parental death or separation) appeared unrelated to any of the molar life-event keys, presumably because we failed to include ‘death’ or ‘separation’ losses in our original list of life-event precipitants. Thus, these exploratory analyses suggested that early adverse experiences might, at times, establish both specific and non-specific patterns of vulnerability to having depression triggered by exposure to relevant life-event stressors.
Patients were also questioned about their ‘interpretation’ of the significant life event preceding their depression (e.g. ‘My bypass operation was like a physical assault’), and about other cognitions linked to that meaning in order to identify possible schemas or core beliefs held currently by the depressed patient (e.g. ‘I am powerless’).
Our qualitative analyses, which first involved categorization of patients' most significant early ‘stressor’ and most significant ‘precipitant’ event (documented during clinical interview), suggested the existence of an adverse life event in childhood (before 16 years) and a mirroring or matching recent triggering life event in one-third of our sample. In these cases, the presence of salient depressogenic cognitions was also apparent. For example, one patient's early adversity involved the death of both parents at 11, followed by separation from siblings and placement in foster homes. Her recent key stressor was loss of a job shortly after loss of a relationship. The patient interpreted this stressor as reflecting a loss of a secure base and a feeling of being generally let down. Assessment of possible schemas revealed that she carried the core belief of ‘I am always left abandoned’.
Interestingly, we found that these lock-key links were more prevalent in non-melancholic depressions, particularly those classed by our clinicians as having a reactive depressive disorder. Reactive or situational depression has been conventionally understood as a depressive episode which follows a severe life stressor. However, for our subsample, it appeared more to be a reaction to a specifically salient or congruent stressor: one which possessed some relevant meaning for the patient. Therefore, the reactive depressive response may have a foundation vulnerability component locked in by some earlier significant life-event adversity, and be more likely than the other depressive subtypes to be activated by mirroring stressors.
Assessment of cognitive schema
In a second study [2] we assessed the extent to which those identified associations between early adverse events and precipitating stressors (lock-key links) might be mediated by cognitive structures like schemas. We found evidence of depressogenic cognitive schemas in 60% of our sample. We categorized these schemas and identified nine domains: (i) ‘threat, danger and lack of safety’; (ii) ‘lack of secure emotional base’; (iii) ‘loss or separation’; (iv) ‘rejection’; (v) ‘victimization and/or abuse’; (vi) ‘worthlessness and/or inadequacy’; (vii) ‘abandonment’; (viii) ‘powerlessness and lack of control’; and (ix) ‘sense of failure’. Patients were given a score on each of these schema domains. Patients' schemas were more strongly associated with their early life events than with events which were said to precipitate their depressions. This finding gives some support for the thesis that experiences of early adversity (whether acute or chronic) leave behind some degree of cognitive residue in the form of a core belief about the self (future or world), which in essence becomes a cognitive vulnerability to depression following exposure to life stress (consistent with a diathesisstress model). However, we failed to find consistent evidence to support vulnerability specificity: whereby a lock or a key was most clearly associated with a cognitive schema which reflected the same type of adversity theme. Rather, evidence pointed towards more general, less specific associations. Despite this lack of evidence for wide specificity, interestingly, the strongest association between a schema theme and lock or key class was for the ‘victimization and/or abuse’ schema being associated most significantly with the ‘unsafe/danger’ class of lifeevent precipitants (keys).
We correlated our three molar locks (‘abuse’, ‘insecurity’, and ‘loss’) with our three molar keys (‘rejection’, ‘unsafe/danger’ and ‘control/abuse’) and then re-examined the coefficients after partialling out any influence of the nine schema domain scores. If cognitive schemas are higher-order ‘locking in’ variables driving any such links between ‘mirroring’ locks and keys, we would have expected that associations should have disappeared when salient cognitive schemas were partialled out. However, we found no evidence for this central hypothesis. When we controlled for salient schema scores (e.g. partialling out ‘lack of secure emotional base’ scores for associations including the ‘insecurity’ lock), associations remained basically unchanged. Such a finding appears to challenge the role of cognitive schemas and their locking in (of vulnerability) status. Therefore, if not due to methodological limitations, one parsimonious possibility is that cognitive schemas are not, as proposed, latent structures, but instead products of mood-state changes. That is, at some level of depression, individuals might interpret their recent stressors and earlier environment according to the cognitive impact of the depression (e.g. ‘I feel worthless’), and not according to any ongoing schema highlighted by a salient stressor.
Thus, in our first study, we did find some support for a lock and key vulnerability hypothesis for our sample of patients, however, in our second study, it appeared that that support might constitute only secondary or artifactual findings. We consider this issue in some detail below.
Discussion
If as most cognitive theorists claim, schema are latent constructs intrinsic to those who develop depression and activated by mirroring life-event stressors, certain consequences should follow. Thus, depressed subjects when euthymic should be more likely than ‘never depressed’ subjects to show some evidence of cognitive vulnerability. However, numerous studies have illustrated that although various types of cognitive vulnerability are recognizable during depression (e.g. depressogenic attitudes and casual attributions), they remain undetectable during periods of recovery [36].
Self-report scores on the measures used in many of these studies are almost always elevated during depression, but more often than not, return to the normal range when those depression-prone subjects recover. Evidence like this could suggest that cognitions thought to be stable depressogenic schemas, may instead be state-like manifestations of a depressed mood. In a recent review of numerous relevant studies, Ingram et al. [37] observed that ‘an inescapable conclusion from the majority of the studies is that depressive cognition is largely state dependent’.
However, it is important to note that the majority of these studies rely on the three most widely used selfreport instruments aimed at measuring cognitive vulnerability. They are the Dysfunctional Attitudes Scale (DAS [38]), an instrument comprised mostly of conditional assumption statements; the Attributional Style Questionnaire (ASQ [39]), which detects a negative bias in respondents' causal explanation for events; and the Automatic Thoughts Questionnaire (ATQ [40]), an instrument which measures the existence of negative automatic thoughts, which in theory are present only during a depressive episode. Further, Ingram et al. [37] adds that many of these studies which appear to pose a challenge to the thesis of cognitive vulnerability (because they fail to detect such ‘vulnerability’ in the absence of depressed mood) rely on measures of cognitive products or content which are actually defined as unlikely indicators of vulnerability by the theory they are suppose to validate. Thus, while cognitive theories do predict changes in cognitive products (e.g. negative automatic thoughts) with remission, they specify that enduring depressogenic cognitions exist at the deeper level of structure and process [14].
Therefore it seems likely that ‘depressive thinking’ as measured by instruments such as the DAS, ASQ and ATQ is largely state dependent. However, it may not necessarily follow that cognitive vulnerability in the form of schemas or core beliefs (either global self evaluations: ‘I am unlovable’, or expectations: ‘I will be abandoned in the end’) do not exist as agents of cognitive vulnerability. Either these vulnerabilities are unconditional (and constant) belief themes which can be accessed at any time given the right methodology, or latent beliefs, only detectable once fuelled by life-event stress. If such schemas are latent, then the diathesis-stress model applies and exposure to a stressor is required for the schema to become operative and thus measurable. Therefore, we should not expect to see evidence of such vulnerability in the form of latent beliefs in depressionprone subjects when they are euthymic, without the necessary exposure to a relevant life stressor.
Beck's dysfunctional assumptions or attitudes are said to become operative as a function of changes in circumstances. They are conditional assumptions (e.g. ‘If my partner leaves me, I am unlovable’) rather than absolute beliefs about the self. Therefore instruments such as the DAS are unlikely to indicate cognitive vulnerability in depression-prone individuals when they are not depressed, and when not exposed to congruent stressors. Young's [6] schema theory argues for the unconditional nature of early maladaptive schemas and suggests that schemas are more hypervalent compared with underlying assumptions which require that certain conditions are present. Young has recently developed a Schema Questionnaire for use in schema-focused therapy for personality disorders [6]. The instrument was designed to assess the 16 EMS, several of which translate into depressogenic core beliefs and also overlap with the nine cognitive schema domains identified in our sample. This questionnaire may be of further use for studies investigating the important hypothesis that depressed subjects, when euthymic, should be more likely than never depressed subjects to show some evidence of cognitive vulnerability.
If schemas do exist as enduring components of one's core self-knowledge (or world knowledge), then they must be able to be captured or measured at some level. Therefore, questions remain concerning their identification and assessment. Our findings suggest that it may be in fact impossible to accurately measure depressogenic cognitive schemas in individuals when they are currently depressed.
It appears that any true assessment of the operational role of cognitive vulnerability in depression might be via prospective investigations. For example, a group of young, never-depressed individuals with or without apparent ongoing depressogenic schema (e.g. using the Schema Questionnaire [6]) could be followed up over time allowing for the assessment of any first-episode depressions. Here, we would expect that the group with vulnerability would be more likely to become depressed or become depressed earlier and more frequently than the group with no such vulnerability. In order to include a test of the diathesis-stress model, such a study should investigate whether depression occurred as a consequence of life events which matched the depressogenic schema.
A few studies using similar designs have been conducted. In contrast to the negative findings presented in the review by Ingram et al. [37], some studies have used a ‘behavioural high-risk design’ and have obtained evidence in support of cognitive risk theories. For example, Alloy et al. [41] found support for the hopelessness theory of depression when ‘attributionally vulnerable’ students were more likely to exhibit past episodes and more severe episodes of depression compared with ‘attributionally invulnerable’ students. More recent studies using prospective designs of this nature have found further support in reporting that individuals who evidenced negative cognitive styles were more likely to go on to develop depressive symptoms when faced with negative life events than individuals who had not shown any negative cognitive style [42, 43]. Further, Alloy et al. [44] reported on a continuing longitudinal study. Students were given a number of questionnaires to measure vulnerability including the Cognitive Styles Questionnaire, a new measure designed to capture ongoing core beliefs. From this procedure, they were able to identify generic ‘high’ and ‘low’ (cognitive risk for depression) student groups who had no history of depressive disorders and were currently asymptomatic. They assessed subjects prospectively over 5 years. Their report provides some compelling evidence that depressogenic cognitive styles do infer a significant risk for developing episodes of major depression. They also found that any ‘risk’ was specific to depression (rather than any other axis I disorder), and that cognitive styles appear to remain extremely stable over time [44].
Pursuing a different approach, Williams [45] has argued that those who are vulnerable to depression possess a stable tendency to encode and retrieve events at a categorical or general level, making recall of specific memories (positive or negative) difficult. This tendency known as overgeneral autobiographical memory is considered to be a long-term cognitive style which persists despite changes in mood. Overgeneral memory has been found to be predictive of future recurrent depression in recovered patients [46] and may therefore prove useful as a more widespread assessment strategy or marker of cognitive vulnerability in relation to depression relapse and recurrence.
Mood challenge strategies have also been used experimentally in the detection of depressogenic schemas in formerly depressed but recovered subjects. Cognitions are usually tested before and after a mood challenge (e.g. ‘sad’ music known to induce dysphoric mood). For example, Segal et al. [47] used a mood induction procedure and found that changes in formerly depressed subjects' cognitions that accompanied changes in mood (i.e. induced dysphoria) were predictive of later recurrence of depression. They also found that negative responses to the mood challenge were reduced for those subjects who had recovered after previous cognitive–behavioural therapy (CBT), as opposed to pharmacotherapy, presumably because the CBT had had an effect upon existing cognitive structures. Because prospective studies using this technique have not yet been pursued (i.e. using ‘never-depressed’ but otherwise ‘at risk’ subjects), it is unclear whether the effects observed (i.e. cognitive reactivity to mood challenge) resulted from possible cognitive ‘scarring’ caused by previous episodes or from any intrinsic pre-existing diathesis.
Also significant, as alluded to earlier and in line with work done on self versus other judgements of ‘perceived vulnerability’ to adversity, is one's expectations of future events. Perhaps another way to access ongoing depressogenic schemas is through the assessment of people's expectations of future events (e.g. ‘I expect to lose those around me’). For example, if a significant expectation of ‘loss’ can be identified in a non-depressed individual and if, in the same individual, a future ‘loss-type’ stressor acts as a precipitant to depression, we can conclude something about that individual's vulnerability. Further, if that expectation can be traced back to a ‘loss’ event, then we may acquire evidence regarding the life-event origin of that core belief. Expectations may therefore inform us about schema as enduring cognitive themes over time.
Conclusion
Our preliminary investigations of lock and key depressions highlight the need for a closer look at the methodology involved in identifying depressogenic core beliefs or schemas, as well as the difficulties involved in the identification of schemas in depressed (as opposed to euthymic) individuals. In order to make conclusions about schemas as either state-dependent artifacts of depression or enduring vulnerability in the form of core beliefs, further studies need to move beyond the conventional measures of cognitive products and content. As noted, there has been a move toward prospective longitudinal investigations of cognitive vulnerability theories, with positive findings. Studies like these, which involve long-term follow up, are of considerable importance, not only to confirm the potential of cognitive risk in some depressive disorders, but in reinforcing the relevance of cognitive-based interventions and preventative strategies, particularly for recurrent depressive disorders.