Introduction
In 1963, Aaron Beck first described overgeneralization as a hallmark cognitive distortion in depression and defined it as “patients’ pattern of drawing a general conclusion about their ability, performance or worth on the basis of a single incident” (Beck,
1963; pp. 328–329). In his later writings, it was further described as “drawing a general rule or conclusion on the basis of one or more isolated incidents and applying the concept across the board to related and unrelated situations” (Beck et al.,
1979; p. 14) or simply as “unjustified generalization on the basis of a single incident” (Beck,
1976; p. 94). It is important to note here that Beck focuses on overgeneralizations following
negative events, which we define as
negative overgeneralization.
1 An example would be a person experiencing symptoms consistent with depression who, in response to a minor remark of his boss about a recent presentation, has the thought “you see, I’m a complete failure, I can’t get anything right!”. Another example of overgeneralized thinking by the same person, when his wife asks him later that day to fix the garden fence, would be: “My presentation at work was a disaster, so for sure I won’t be able to repair the fence”.
The evidence so far has largely supported the association between negative overgeneralization and depression.
2 A robust finding in the literature, primarily using the Attitudes Towards Self scale (ATS; Carver & Gannellen,
1983) is that there is a substantial positive association between negative overgeneralization to the self and severity of depressive symptoms, with
rs ranging from 0.30 to 0.60 (e.g., Carver & Gannellen,
1983; Carver et al.,
1988; Eisner et al.,
2008; Ganellen,
1988; MacLeod & Williams,
1990). This cross-sectional association has been observed for individuals with current major depressive disorder (van den Heuvel et al.,
2012) and individuals with a history of major depressive disorder (Eisner et al.,
2008) compared to people who never had experiences consistent with depression. When examined over time, it has been found that negative overgeneralization prospectively predicts higher levels of depression (Carver,
1998; but see Carver et al.,
1988). Further, there is some evidence for the specificity of the association between negative overgeneralization and depression. For example, negative overgeneralization to the self is not significantly related to anxiety (Ganellen,
1988), and is fully accounted for by current depressive symptoms in both bipolar disorder (Eisner et al.,
2008) and borderline personality disorder (Van den Heuvel et al.,
2012).
Other studies, however, have focused on a somewhat different form of overgeneralization and therefore used other measures, such as subscales of the Cognitions Questionnaire (CQ; Fennell & Campbell,
1984), the Cognitive Errors Questionnaire (CEQ; Lefebvre,
1981), or the Children’s Negative Cognitive Error Questionnaire (CNCEQ; Leitenberg et al.,
1986). These instruments measure
negative overgeneralization across situations, which involves (over)generalizing the negative outcome of a situation (e.g., I didn’t get the job I applied for last month) to a different situation (e.g., “I bet the new neighbors won’t like me”) which reflects seeing unpleasant events as typical for one’s life (Fennell & Campbell,
1984), without necessarily making negative inferences about one’s personal overall self-worth. To further distinguish, overgeneralization about the self involves broad negative inferences about oneself from singular circumstances, whereas overgeneralization across situations will involve broad negative inferences about circumstances external to the person, but potentially relevant to them, from singular situations (e.g., that’s math’s test was hard) to a broader range of situations (e.g., tests are generally too stress-inducing). Correlations between this negative overgeneralization across situations and depressive symptom severity have also been observed (MacLeod & Williams,
1990). In one of the rare studies examining overgeneralization in children, Leitenberg et al. (
1986) found that children with elevated depressive symptoms showed higher levels of negative overgeneralization than a control group. Adults recovered from depression showed higher levels of negative overgeneralization across situations relative to people who never had experiences consistent with depression (Fennell & Campbell,
1984). Unlike negative overgeneralization to the self, to the authors’ knowledge there currently are no prospective studies that also examine depressive symptoms.
Despite the fact that these types of overgeneralization, focusing on different aspects of negative cognition, can be distinguished, given that both fit with Beck’s original definition of negative overgeneralization in depression, and they behave comparably in terms of their relation with depression one might expect an association between them. Yet we are aware of only one study to date that has explicitly examined the relation between both types of negative overgeneralization. MacLeod and Williams (
1990) found that negative overgeneralization to the self and across situations were only weakly and non-significantly related,
r(41) = 0.24.
Independent from the Beckian literature on cognitive distortions, there exists a literature in depression research on overgeneralization of memories. Research on overgeneralization of memory has grown ever since a seminal report by Williams and Broadbent (
1986) on a phenomenon called
overgeneral autobiographical memory (OGM). The basic observation is that when asked to describe a specific memory in response to an emotional cue-word, people with depression tend to respond, relatively more than healthy controls, with overgeneralized or so-called categoric memories. OGM is usually tested with the Autobiographical Memory Test (AMT; Williams & Broadbent,
1986), which consists of a set of positive and negative cue-words. Whereas a specific memory refers to a single event that occurred at a particular place and time and lasted less than one day (e.g., “The moment that the instructor informed me that I didn’t pass my driving test”, in response to the cue ‘failed’), an overgeneral categoric memory refers to a class or summary of similar events (e.g., “the times that I failed important tests”; see Williams et al.,
2007, for a review). A recent review has shown that OGM persists in people who have recovered from, or formerly experienced depression (Hallford et al.,
2021a). A recent meta-analysis, including 32 prospective studies, concluded that OGM is a predictor of the course of depression, in that OGM predicts higher follow-up symptoms (Hallford et al.,
2021b). However, it remains unclear whether OGM occurs prior to the very first onset of depression, or is a ‘scar’ of prior episodes that predicts later episodes simply because it is a marker of previous depression.
Although the research tradition on overgenerality in the retrieval of autobiographical events convincingly points to an important role of overgeneralized processing in depression at the level of memory functioning, the association of this memory overgeneralization with the Beckian type of overgeneralized thinking processes in depression (negative overgeneralization to the self and across situations) has never been examined despite strong theoretical arguments that make a relationship between them highly plausible.
According to Beck et al. (
1979), cognitive distortions, such as overgeneralization, are triggered by the activation of a dysfunctional schema, especially in reaction to a stressful or negative experience. These schemas are believed to direct current self-evaluative thinking, as well as influence memory functioning, such as the retrieval or recollection of past personal experiences: “Whether the particular cognitive process [is] recollection [of memories], evaluation of [one’s] current status or attributes, or prediction of the future, the thoughts [bear] the imprint of [the] schema” (Beck et al.,
1979; p. 565). It is a central tenet of information processing accounts in cognitive psychology that schemas give rise to (schema-congruent) biased or selective processing in the domains of attention, interpretation and memory.
The question then arises whether OGM and overgeneralized thinking are actually one and the same thing. Dalgleish et al. (
2003) indicated that they may be. They suggest that cue-words on the AMT may map onto the content of self-schemas, which consist of generic negative aspects of the self. Certain cue words like ‘failed’ and ‘unsuccessful’ may then further activate negative schemas. As they state, “participants’ responses on the cue-word task could amount to a propositional ‘read-off’ of the activated self-schemas. Such a read-off would essentially look like a categorical autobiographical memory” (Dalgleish et al.,
2003, p. 220). Similarly, Ramponi et al. (
2004) provided evidence to show that people who have less differentiated affect-related schematic models retrieve more categoric memories on the AMT. In sum, it is not clear whether these different domains of overgeneralization derive from the same underlying process. More recent theorizing suggests that OGM may be explained by a tendency for people with negativity emotionality/affectivity to process information in a way that is low in sensory-perceptual details, allowing threat-related information in prior categorical beliefs to dominate conscious recall of self-related information (Van den Bergh et al.,
2021). This may be somewhat adaptive in that it reduces the threshold of being aware of self-related aversive experiences in a ‘better safe than sorry’ type of strategy. One important way of clarifying the issue is not only to examine the associations between them, but to see whether they independently predict the course of later depression.
Overview of the Present Study
The present study focused on the three following forms of overgeneralization: negative overgeneralization to the self, negative overgeneralization across situations, and overgeneral autobiographical memory (OGM). These overgeneralization variables were studied in a large community sample with a follow-up assessment six months later. We also measured probable depression at baseline and at follow-up using both dimensional measures of symptom severity, and diagnostic criteria (to assess the probable presence or absence of a major depressive episode).
The study had three main aims. The first aim was to examine the correlations between the different forms of depression-relevant overgeneralization. For negative overgeneralization to the self and across situations we expected a positive, but weak association based on prior findings of MacLeod and Williams (
1990). As for OGM, our study is the first to test its association with the Beckian type of overgeneralization. Based on theoretical arguments from schema-based accounts (Dalgleish et al.,
2003; Ramponi, et al.,
2004), we hypothesized that OGM would be positively associated with both forms of negative overgeneralization.
The second aim was to further examine the correlations by testing the predictive association between our different overgeneralization variables and the course of depression. Based on the findings of Carver (
1998), and the putative causal role for negative overgeneralization in depression (e.g., Beck et al.,
1979), we hypothesized that both forms of negative overgeneralization would predict levels of depression symptoms prospectively, as well as the probable
recurrence of such episodes in participants with a history of major depressive disorder. For overgeneralization at the level of memory functioning (i.e., OGM), we hypothesized, given the findings of the meta-analysis of Hallford et al., (
2021a,
2021b,
2021c,
2021d), that OGM would also predict higher follow-up symptoms and probable recurrence of a major depressive episode. As mentioned above, we also assessed probable diagnostic status and history of depression. This allowed us to fulfill our third aim, which was to examine for the first time whether OGM predicts the probable first onset of a major depressive episode. Given OGM’s associations with various aspects of depression, we hypothesized that OGM would predict the probable first onset of a depressive episode. Exploratory analyses were also conducted to tests for possible interaction effects between the three overgeneralization variables.
Prediction of Depression at Follow-Up
A hierarchical regression analysis was performed with DASS-D scores at T2 as the criterion variable. DASS-D scores at T1 were included in Step 1, alongside sex, given this is commonly found to predict depression (male coded as 1, female as 2) and probable history of a major depressive episode (no previous episode(s) coded as 0, previous depressive episode(s) coded as 1). In Step 2, the three forms of overgeneralization were included. In Steps 3 and 4, interaction terms created from the centered overgeneralization variables were entered into the model. Table
3 summarizes the results of this analysis. Notably, in the last step, all two-way interactions were being tested as adjusted by the three-way interaction.
Table 3
Summary of hierarchical regression analysis for variables predicting depression symptoms at T2
Step 1 | | | |
Constant | 1.93 (0.53) | | |
Sex | − 0.30 (0.30) | − 0.04 | |
DASS-D T1 | 0.32 (0.03) | 0.37*** | |
Past MDE | 0.79 (0.35) | 0.09* | 0.16*** |
Step 2: ΔR2 = .04 (p < .001) | | | |
Constant | 0.33 (0.63) | | |
Sex | − 0.48 (0.30) | − 0.06 | |
DASS-D T1 | 0.25 (0.04) | 0.28*** | |
Past MDE | 0.54 (0.35) | 0.06 | |
ATS-OG | 0.10 (0.02) | 0.19*** | |
CEQ-R-OG | 0.06 (0.05) | 0.05 | |
AMT-C | 0.20 (0.09) | 0.08* | 0.20*** |
Step 3: ΔR2 = .01 (p = .048) | | | |
Constant | 2.45 (0.54) | | |
Sex | − 0.41 (0.30) | − 0.05 | |
DASS-D T1 | 0.25 (0.04) | 0.29*** | |
Past MDE | 0.71 (0.35) | 0.08* | |
ATS-OG | 0.10 (0.02) | 0.19*** | |
CEQ-R-OG | 0.04 (0.05) | 0.03 | |
AMT-C | 0.19 (0.09) | 0.08* | |
ATS-OG × CEQ-R-OG | − 0.01 (0.01) | − 0.05 | |
AMT-C × ATS-OG | 0.04 (0.02) | 0.10* | |
OGM × CEQ-R-OG | − 0.002 (0.03) | − 0.002 | 0.21*** |
Step 4: ΔR2 = .01 (p = .023) | | | |
Constant | 2.38 (0.54) | | |
Sex | − 0.37 (0.30) | − 0.04 | |
DASS-D T1 | 0.24 (0.04) | 0.28*** | |
Past MDE | 0.60 (0.35) | 0.07 | |
ATS-OG | 0.11 (0.02) | 0.21*** | |
CEQ-R-OG | 0.03 (0.05) | 0.02 | |
AMT-C | 0.16 (0.09) | 0.06 | |
ATS-OG × CEQ-R-OG | − 0.01 (0.01) | − 0.01 | |
AMT-C × ATS-OG | 0.04 (0.01) | 0.10** | |
OGM × CEQ-R-OG | − 0.01 (0.03) | − 0.01 | |
AMT-C × ATS-OG × CEQ-R-OG | 0.01 (0.00) | 0.08* | 0.22*** |
Unsurprisingly, depression symptoms at baseline significantly predicted depression symptoms at follow-up. Negative overgeneralization to the self and overgeneral memory (AMT categoric memories) each independently significantly predicted depression symptoms at T2. The more negative overgeneralization and the more overgeneral memory at T1, the higher the level of depression symptoms at T2, when baseline symptomatology, sex, and probable history of past depressive episode were used as covariates. Notably, OGM was a significant predictor in the model despite a non-significant zero-order correlation with depressive symptoms at T2 (r = .07, p = .125). This could be explained, in part, because of variation in sampling error given a different sample used in zero-order correlations vs. regression analysis (N = 536 vs. N = 514), although the zero-order correlation was almost identical in the regression sample: r = .065, p = .141). We also assessed for potential statistical suppression. The regression model was run a series of times, each time removing one of the predictor variables while leaving the others in the model. In all of those models, OGM remained a significant predictor (p < .05). Therefore, any statistical suppression likely resulted from some combination of variables, with each of these suppressors having only a trivially small impact on their own. It is also likely that other variables in the model controlled for irrelevant variance, increasing power for this predictor to account for relevant variance. Entering the interaction terms in Step 3 and 4 predicted significantly more variance in depressive symptoms, with two significant interactions at Step 4: OGM and overgeneralisation to the self, and a three-way interaction between overgeneralisation variables. On probing for the conditional effect of OGM on depressive symptoms, simple slopes tests were used, and it was found to be strongest at higher levels of negative overgeneralisation to the self (+ 1 SD: B = 0.41, SE = 0.13, p = .003), relative to average (mean: B = 0.17, SE = 0.09, p = .065) or lower levels (-1 SD: B = − 0.06, SE = 0.14, p = .662). On probing the three-way interaction, again, simple slopes tests were used to indicate that the interaction effect of higher OGM at higher levels of overgeneralisation to the self was larger at higher levels of overgeneralisation to situations (+ 1 SD: B = 0.59, SE = 0.18, p = .001; mean: B = 0.38, SE = 0.13, p = .005; -1 SD: B = 0.17, SE = 0.20, p = .383). In summary, these interactions indicated that OGM more strongly predicted depressive symptoms over time when overgeneralisation to the self and across situations were higher.
We then examined the predictive value of the overgeneralization measures for probable depressive episode during the follow-up period. The presence or absence of a probable major depressive episode at any point during the follow-up was coded with a dummy variable (1 = present, 0 = absent). A logistic regression was conducted in the subgroup of people who probably formerly had experiences consistent with depression or had recovered from depression to investigate the predictive value of our measures for recurrence while covarying sex and baseline depression (Table
4). For prediction of probable recurrence of depressive episode, categoric memories were a significant predictor, with more categoric memories associated with a higher likelihood of a recurrence of depression. Negative overgeneralization to the self, measured by the ATS-OG, was also, independently, a predictor of probable depression recurrence. There were no significant interactions between the overgeneralization variables. Notably, in the fourth step, all two-way interactions were being tested as adjusted by the three-way interaction. To achieve the third aim of the study, the same model of logistic regression was conducted in the subgroup of those who probably never experienced depression. This served to examine the predictive value of our measures for first onset (Table
4). For prediction of first onset, the group of three generalization measures did not significantly improve the predictive value of the model, either as independent predictors or as interaction terms.
Table 4
Logistic regressions, (a) prediction of probable depression recurrence (Total N = 97, recurrence n = 12) and (b) prediction of probable depression onset (Total N = 381, first onset n = 15)
(a) | | | |
Step 1: χ2 (2) = 9.16* | | | |
Constant | − 3.40 (0.66) | | |
Sex | 1.3 (0.73) | 4.03 | 0.95, 17.05 |
DASS-D T1 | 0.19 (0.07) | 1.21*** | 1.04, 1.41 |
Step 2: Δχ2 (3) = 10.6* | | | |
Constant | − 8.23 (2.31) | | |
Sex | 2.1 (0.91) | 8.29* | 1.36, 50.27 |
DASS-D T1 | 0.13 (0.09) | 1.14 | 0.94, 1.37 |
ATS-OG | 0.15 (0.06) | 1.17* | 1.03, 1.32 |
CEQ-R-OG | 0.13 (0.13) | 1.14 | 0.88, 1.48 |
AMT-C | 0.53 (0.24) | 1.70* | 1.05, 2.77 |
Step 3: Δχ2 (3) = 4.72, ns | | | |
Constant | − 4.49 (1.04) | | |
Sex | 2.45 (1.01) | 11.66*** | 1.59, 85.44 |
DASS-D T1 | 0.21 (0.11) | 1.23 | 0.99, 1.53 |
ATS-OG | 0.14 (0.06) | 1.15* | 1.01, 1.32 |
CEQ-R-OG | 0.04 (0.16) | 1.04 | 0.75, 1.45 |
AMT-C | 0.45 (0.37) | 1.58 | 0.75, 3.31 |
ATS-OG × CEQ-R-OG | 0.01 (0.01) | 1.01 | 0.96, 1.05 |
AMT-C × ATS-OG | 0.10 (0.06) | 1.10 | 0.97, 1.25 |
AMT-C × CEQ-R-OG | 0.20 (0.14) | 1.22 | 0.91, 1.63 |
Step 4: Δχ2 (4) = 0.64, ns | | | |
Constant | − 4.72 (1.14) | | |
Sex | 2.69 (1.0) | 14.79*** | 1.77, 123.52 |
DASS-D T1 | 0.21 (0.11) | 1.23 | 0.99, 1.53 |
ATS-OG | 0.16 (0.07) | 1.17* | 1.01, 1.36 |
CEQ-R-OG | 0.01 (0.18) | 1.01 | 0.71, 1.44 |
AMT-C | 0.39 (0.38) | 1.49 | 0.69, 3.18 |
ATS-OG × CEQ-R-OG | 0.01 (0.02) | 1.01 | 0.97, 1.05 |
AMT-C × ATS-OG | 0.10 (0.06) | 1.11 | 0.97, 1.27 |
AMT-C × CEQ-R-OG | 0.22 (0.15) | 1.24 | 0.92, 1.67 |
AMT-C × ATS- OG × CEQ-R-OG | − 0.01 (0.02) | 0.98 | 0.94, 1.02 |
(b) | | | |
Step 1: χ2 (2) = 10.06** | | | |
Constant | − 4.13 (0.49) | | |
Sex | 3.6 (0.54) | 1.43 | 0.49, 4.13 |
DASS-D T1 | 0.19 (0.05) | 1.21*** | 1.08, 1.35 |
Step 2: Δχ2 (3) = 6.48, ns | | | |
Constant | − 5.54 (0.54) | | |
Sex | 0.51 (0.57) | 1.67 | 0.54, 5.17 |
DASS-D T1 | 0.16 (0.06) | 1.18** | 1.04, 1.33 |
ATS-OG | 0.03 (0.04) | 1.03 | 0.94, 1.13 |
CEQ-R-OG | 0.14 (0.08) | 1.15 | 0.97, 1.37 |
AMT-C | − 0.41 (0.30) | 0.66 | 0.36, 1.21 |
Step 3: Δχ2 (3) = 1.60, ns | | | |
Constant | − 4.40 (0.61) | | |
Sex | 0.38 (0.59) | 1.47 | 0.46, 4.69 |
DASS-D T1 | 0.17 (0.06) | 1.19** | 1.04, 1.34 |
ATS-OG | 0.05 (0.06) | 1.05 | 0.94, 1.18 |
CEQ-R-OG | 0.21 (0.11) | 1.23 | 0.99, 1.53 |
AMT-C | − 0.57 (0.42) | 0.56 | 0.24, 1.28 |
ATS-OG × CEQ-R-OG | 0.01 (0.01) | 0.99 | 0.96, 1.01 |
AMT-C × ATS-OG | 0.01 (0.05) | 1.01 | 0.91, 1.11 |
AMT-C × CEQ-R-OG | 0.10 (0.11) | 1.11 | 0.89, 1.38 |
Step 4: Δχ2 (4) = 1.75, ns | | | |
Constant | − 5.30 (0.68) | | |
Sex | 0.24 (0.60) | 1.27 | 0.38, 4.20 |
DASS-D T1 | 0.17 (0.06) | 1.19** | 1.05, 1.35 |
ATS-OG | − 0.01 (0.08) | 0.98 | 0.83, 1.16 |
CEQ-R-OG | 0.24 (0.13) | 1.27 | 0.97, 1.66 |
AMT-C | − 0.78 (0.51) | 0.45 | 0.16, 1.24 |
ATS-OG × CEQ-R-OG | 0.01 (0.02) | 1.01 | 0.96, 1.05 |
AMT-C × ATS-OG | − 0.07 (0.08) | 0.93 | 0.79, 1.09 |
AMT-C × CEQ-R-OG | 0.13 (0.12) | 1.14 | 0.89, 1.47 |
OGM × ATS- OG × CEQ-R-OG | 0.02 (0.01) | 1.02 | 0.98, 1.06 |
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.