Introduction: The Relationship Between Depression Severity and Cognitive Errors
Cognitive errors (CEs) are a key construct in cognitive therapy (CT), especially when assessing, researching, or treating patients suffering from depression. Early on,
Beck (1976) outlined specific cognitive errors, including arbitrary inference, selective abstraction, overgeneralization, magnification (catastrophizing) and minimization, personalizing, and absolutistic dichotomous thinking.
Burns (1999) defined and outlined the importance of other cognitive errors such as mind-reading, fortune telling, mental filter, all-or-nothing thinking, should statements, discounting the positive, emotional reasoning, and labeling and mislabeling. More recently,
Yurica and DiTomasso (2005) conducted a review and identified 17 cognitive distortions common in the literature, including arbitrary inference/jumping to conclusions, catastrophizing, comparison, dichotomous/black-and-white thinking, disqualifying the positive, emotional reasoning, externalizing self-worth, fortune-telling, labeling, magnification, mind reading, minimization, overgeneralization, perfectionism, personalization, selective abstraction, and “should” statements. Despite the availability of definitions for these individual CEs, and suggestions that individual CEs are important to understand depression, studies examining cognitive errors in depression reported findings in terms of a general level of distortion, rather than for specific CEs (e.g.,
Briere, 2000;
Hamblin, Beutler, Scogin, & Corbishley, 1993;
Monroe, Slavich, Torres, & Gotlib, 2007;
Tang, DeRubeis, Beberman, & Pham, 2005).
Exceptions to this include a study by
Wenzlaff and Grozier (1988). They examined the use of
overgeneralizations in dysphoric and nondysphoric participants who received false negative feedback after a social perceptiveness test. While both groups lowered their self-evaluations of social perceptiveness after receiving the feedback, only the participants with dysphoria lowered their ratings in proficiency judgments. More recently Van den Heuvel, Derksen, Eling, and van der Staak (2012) explored how overgeneralization relates to different mood disorders. The researchers examined tendencies of patients to make overgeneralizations targeted at the self and across situations. Participants diagnosed with major depression and those diagnosed with bipolar disorder were found to exhibit high levels of overgeneralization towards the self. However, only patients with bipolar disorder demonstrated elevated levels of overgeneralizations across situations. This study is of particular importance because it demonstrates subtle differences among patients regarding a specific CE, and it highlights the need to explore specific types of cognitive errors.
Wenzlaff’s and Grozier’s (1988) study was particularly important in that it examined the CE,
magnification/minimization, which is the tendency to evaluate oneself, others, or a situation in a way that magnifies or minimizes the negative or positive aspects (
Drapeau et al., 2008). Consistent with Beck’s theory of depression, dysphoric participants rated social perceptiveness to be more important when they were given failure feedback on the task. Participants who were not dysphoric rated the task as more important when they received successful feedback on the task. This finding is consistent with Ellis’ theory (1980) that individuals who are depressed tend to “awfulize” or magnify the unpleasantness of situations. However, Wenzlaff’s and Grozier’s finding does not necessarily illustrate that participants with dysphoria committed an “error” that was disproportionate to the demands of the situation; it demonstrated that they responded differently to the feedback than did nondepressed controls. Additionally, the participants with dysphoria in Wenzlaff and Grozier’s study did not minimize the importance of their success on social perceptiveness (relative to controls) as Beck’s theory postulates they would. The results, however, do suggest that dysphoric individuals (relative to controls) may be differentially influenced by negatively and positively valenced events. Exploring CE types and valence may have important implications.
Another CE empirically investigated is
fortune telling, which involves “making the assumption that the worst or best possible outcome will occur in a situation” (
Drapeau et al., 2008, p. 18). Fortune telling has been found to relate to identifiable psychotherapy processes and to specific psychological disorders, including depression and anxiety (
Miranda, Fontes, & Marroquin, 2008).
Miranda and Mennin (2007) assessed fortune telling in a group of participants with depression by presenting participants with a future events task. This task assessed the certainty in which participants believed that positive and negative events would occur. There was a positive correlation between greater severity of depression and the likelihood of negative events occurring. In addition, the more severe the depression, the more certain were the participants that positive future events would not occur. Few studies assessed several individual CEs, and often, within studies on individual cognitive errors, the focus was directed to distortion as a whole (
Curry et al., 2006;
Tang, DeRubeis, Beberman, & Pham, 2005;
Burns, Shaw, & Croker, 1987,) or was specific to very few CEs within the typology of the scale used (
Flett, Goldstein, Hewitt, and Wekerle, 2012;
Jager-Hyman et al., 2014;
Leung & Poon, 2001).
There arguably remains a paucity of research about the nature of specific cognitive errors in depression, despite the claimed importance of cognitive errors in theory and practice. This may be understandable, in part, because measuring cognitive errors has always been a challenge. For example,
Hammen and Krantz (1976) designed a self-report questionnaire to measure specific CEs using case vignettes. When developing the scale items, independent raters could not agree on the type of CE represented; therefore, items were categorized as being distorted or not distorted.
Lefebvre (1981) also attempted to assess CEs using a self-report measure of seven different CEs, but because of difficulties with reliability, these seven were condensed down to the four errors of catastrophizing, overgeneralizing, personalization, and selective abstraction. It was no surprise that most researchers have relied on experimental design in a laboratory setting, or have used self-report questionnaires. The former is certainly important and can offer invaluable information, but remain costly to implement and the ecological validity of the findings provided may be questionable. Despite utility of the self-report questionnaire, the latter is not without limitation. For example, a person’s mood can influence how he/she responds to items on a questionnaire (Moussavi, et al., 2007), especially in the case of depression where a negative reporting bias is likely (
Owsley, & McGwin, 2004; Shean, & Baldwin, 2010). Further, participants may think in ways not captured in the format in which the questionnaire was written, and participants may not be able to report accurately on the frequency of their thoughts. Difficulty concentrating, which is a symptom of major depression, may inhibit the act of completeing a questionnaire (e.g.,
Gotlib, & Joorman, 2010).
Another limitation is that questionnaires require content for the items. For example, questions may include items relating to particular academic or social events (e.g., the Negative and Positive Cognitive Error Questionnaire by McKenna, 1987), and participants may neglect to endorse an item because they cannot relate to the content. Thus, their actual level of distortion is under represented (
Floyd & Scogin, 1998). Additionally, unique thoughts and beliefs may not be captured by paper and pencil measures (
Segal, 1984), and self-report measures rely on the assumption that maladaptive attitudes and beliefs reflect how people think during stressful situations (
Gunthert, Cohen, Butler, & Beck, 2005). Gunthert and colleagues argued dispositional self-report measures may be too far removed from what happens in stressful situations, and that “a more direct strategy would be to assess participants’
actual thoughts in response to naturally occurring stress rather than to request their more
general reports of underlying dysfunctional attitudes” (p. 78). One implication of these measurement limitations was articulated by Jacobson and colleagues (1996) who underlined the possible “inadequacy of currently available measuring instruments” (p. 303) in CT. Concurrent with this, reseachers recommended (
Dozois & Dobson, 2004;
Oei & Free, 1995) assessing key constructs by using a
variety of assessment modalities (e.g., self-report and observer-rated).
This study is a first step in addressing these challenges. It aims to examine cognitive errors in depression, as assessed by trained observers. It thus provides a perspective complementary to what has been obtained through experimental studies and the use of questionnaires. This study also aims to increase our understanding of the association between many possible CEs and depression. Indeed, most CEs have not been researched, possibly because of a lack of methods to assess them. For example, cognitive errors are not explicitly identified on the Cognitive Bias Questionnaire (
Krantz & Hammen, 1979), or the Cognitive Distortions Scale (CDS,
Briere, 2000), and only four are represented on
Lefebvre’s (1981) Cognitive Error Questionnaire.
Henriques and Leitenberg (2002) have stated that “negative cognitive errors such as personalization, overgeneralization, selective abstraction, and catastrophization may play a role in the etiology and maintenance of dysphoric mood and are therefore deserving of more research separate from dysfunctional attitudes and automatic thoughts” (p. 258). It would be useful to examine more closely cognitive errors independent of the content domains used on existing scales, since these may not address the experiences of those using these questionnaires.
Therefore, the first aim of this study was to describe the type and frequency of specific cognitive errors in depressed participants at therapy intake. It also aimed to determine if depressed participants displayed more negative than positive CEs and whether or not positive and negative CEs were related to one another. A second aim was to examine the relationships between cognitive errors and severity of depression. Indeed, while research supports the idea that CEs are different in individuals with and without depression, little is known about the relationship that may exist between CEs and levels or severity of depression.
Given the dearth of research on the link between depression severity and cognitive errors, we tested whether depression severity is positively correlated with negative CEs and negatively correlated with positive CEs. We also examined if individuals with high and low distortions differ from one another on levels of depression. We questioned if subdividing the sample of depressed patients based on levels of distortion might maximize variance in the variable of interest, namely cognitive errors, and allow for a more precise observation of processes in subgroups of patients (see
Peris et al., 2010).
Results
Mean BDI scores were 30.10 (SD = 6.51). As the sample contained fewer than 50 participants, the Shapiro-Wilk test was used to determine the normality of the variables. Results indicated that all CE variables were non-normally distributed; therefore non-parametric tests were used for all analyses.
Profile of Cognitive Errors
On average, participants spoke 3365.31 words (
SD = 1488.08) during a 50-minute session of early therapy. These sessions contained an average of 3.50 cognitive errors (CEs) per 1000 words (
SD = 2.23), totaling 10.91 CEs per session (
SD = 6.46). Of these CEs, 10.27 (94%) were negative in valence, with only .64 (6%) being positive in valence. The frequency of each cognitive error per 50-minute session, and expressed as a proportion of total CEs can be found in
Table 1 (see
Drapeau, Perry & Dunkley, 2008 for specific information on data conversion procedures).
Prevalence of Negative vs. Positive CEs
A two-tailed Wilcoxon signed-rank test indicated that participants had more negative than positive CEs for total scores, and for each of the four CE clusters: fortune telling, overgeneralization, selective abstraction, and personalizing (see
Table 2).
Correlations Between Negative and Positive CEs
Spearman two-tailed correlations were computed to determine if positive and negative cognitive errors are related to one another. No significant correlation was found for positive and negative CEs (r =—.20, p = .19). The four CE clusters of fortune telling, overgeneralization, selective abstraction, and personalizing were correlated with each other. Fortune telling negative and selective abstraction negative were significantly correlated (r = .31, p = .04), and overgeneralization negative and selective abstraction negative were significantly correlated (r = .31, p = .04). We found no other combination; no positive CE cluster was related to a negative CE cluster.
The above correlations were recomputed for high and low distorters. The split of the sample into two such groups offers another way to examine the data, one that considers the possibility that processes may be different for subgroups of depressed individuals. Participants were categorized as high or low distorters based on their median split scores for early therapy total cognitive errors (e.g.,
Dozois, Covin, & Brinker, 2003). A two-tailed Mann-Whitney test confirmed that the group that had been classified as high distorters endorsed significantly more total cognitive errors (
Mdn = 4.74, range = 2.75–10.26), than did the low distorters (
Mdn = 1.94, range = .00–2.57;
U = .00;
p < .001). These differences were largely due to negative cognitive errors, which significantly differed between the groups (high distorters:
Mdn = 4.33, range = 2.59–10.26; low distorters:
Mdn = 1.70, range = .00–2.40;
U = .00;
p <.001), while positive CEs did not (high distorters:
Mdn = .00, range = .00–1.15; low distorters:
Mdn = .17, range = .00–.86;
U = 233.00;
p = .62).
Two-tailed Spearman correlations indicated that total positive CEs and total negative CEs were not correlated for high distorters (r = –.35, p = .11). However, there were significant correlations observed for the following CE clusters: fortune telling positive and selective abstraction positive (r = .70, p <.001); fortune telling negative and selective abstraction positive (r = –.43,p = .04); selective abstraction negative and personalizing positive (r = –.51, p = .01). Among the low distorters, again total positive CEs and total negative CEs were not significantly correlated (r = –21, p = .34); however, the following CE clusters were significantly correlated: Selective abstraction negative and selective abstraction positive (r = .43, p = .04), personalizing positive and overgeneralization negative (r = –.43, p = .04).
CEs and Level of Depressive Symptoms
For the complete sample, no significant correlations were found among total, total positive, or total negative CEs and level of depression on the Beck Depression Inventory (BDI) at early therapy based on the use of Spearman one-tailed correlations. Similarly, no positive or negative CE Cluster was significantly correlated with depressive symptoms at early therapy on the BDI. In terms of the 30 specific CEs, no positive CEs were significantly correlated with level of depression on the BDI. Significant correlations for negative CEs were found for magnification (r = .30, p = .02) and for should and must statements (r = – .27, p = .04). As more correlations were hypothesized, high versus low distorters were examined separately.
For depression scores, a two-tailed Mann-Whitney test indicated that high and low distorters did not differ from one another on their BDI scores (U = 248.00, p = .91). Among the low distorters (n = 22), the only significant correlation between CEs and depressive symptoms was an inverse relationship between positive fortune telling and depression scores on the BDI (r = –.44, p = .02). For high distorters (n = 23), significant correlations were found for depression and total CEs (r = .47, p = .01), negative CEs (r = .38, p = .04), and negative selective abstraction (r = .35, p = .05). Depressive symptoms were also inversely related to positive CEs (r = – .14, p = .03), and positive overgeneralization (r = –.41, p = .03).
As individuals with high distortions had associations between positive and negative CEs and depression, while those with low distortions had an association only between positive fortune telling and depression, correlations were conducted to determine if those with high distortions had a general tendency to distort. Tentative support for this hypothesis was found using one-tailed Spearman correlations which indicated that positive and negative CEs approached a significant correlation for high distorters (r = –.35, p = .05), but not for low distorters (r = –.21, p = .17).
Discussion
All cognitive errors found were in the spontaneous speech of depressed participants. The selective abstraction cluster was most prevalent, followed by overgeneralization, personalizing, and fortune telling. When individual CEs were separated from the clusters, the top six CEs were labeling negative, followed by should and must negative, jumping to conclusions negative, mind-reading negative, overgeneralizing negative, and fortune telling negative. These findings were consistent with theoretical tenets, such as Beck and colleagues’ concept of the negative cognitive triad, including negative thoughts about the self (labeling), future (fortune telling) and world (overgeneralizing) (e.g.,
Beck et al., 1979). It was also reflective of Ellis’ emphasis on depressed clients’ tendencies to “awfulize” (use magnification) and use should statements (
Ellis, 1980).
The findings supported the hypothesis that participants would have more negative than positive CEs. This result was consistent with theory and previous research indicating that thinking is distorted negatively in patients with depression (
Beck et al., 1979;
Hamilton & Abramson, 1983;
Hawley, Zuroff, Brozina, & Dobson, 2014;
Pothier, Dobson, & Drapeau, 2012;
Schwartzman, et al., 2012;). These results were also consistent with those of Kramer, Bodenmann, and Drapeau (
2009), who found that patients with bipolar disorder had a higher ratio of negative to cognitive positive errors when in a depressed mood state. When in a manic state, they demonstrated more positive CEs.
The current study also found that there was no relationship between positive and negative CEs for the entire sample. However, there were several instances of positive and negative CEs being significantly correlated among high and low distorters. Previous research was inconsistent in this regard, as negative and positive distortions had been found to be positively correlated (e.g.,
Henriques & Leitenberg, 2002), negatively correlated (e.g.,
Lauren & Black, 2011), and uncorrelated (e.g.,
Mazur, Wolchik, & Sandler, 1992).
Previous research suggested depressive symptoms were significantly correlated with cognitive variables, such as automatic thoughts (e.g.,
Dobson & Breiter, 1983;
Dozois, et al., 2009;
Hjemdal, Stiles, & Wells, 2013;
Oei, Bullbeck, & Campbell, 2006), dysfunctional attitudes (e.g., Oei, et al.,
Segal, et al., 2006; Sherrer & Dobson, 2009), and cognitive errors (e.g.,
Henriques & Leitenberg, 2002;
Pothier, Dobson, & Drapeau, 2012; Sato, 2004; Schwartzman, et al., 2012). The current study found that in most instances, level of depressive symptoms was not related to cognitive errors. The exceptions were significant positive correlation between magnification negative and depressive symptoms and the inverse relationship between should and must statements negative and depressive symptoms.
To further investigate these counterintuitive findings, we examined high and low distortions separately. There were no differences among individuals with high and low distortions on their levels of depression; there were some significant correlations between depression and cognitive errors within the two groups. Among those with low distortions, a greater amount of positive fortune telling was associated with lower levels of depression on the BDI. Among those with high distortions, greater levels of depression were associated with having greater total CEs, negative CEs, and negative selective abstraction CEs. Individuals with high distortions were also found to have lower depression scores associated with higher levels of positive CEs and positive overgeneralization. Distorting in the positive direction may function as a successful short-term mood-boosting strategy for those with high distortions, given it was associated with lower levels of depression, the strategy of distorting information may work against these individuals when negatively valenced information is present, given that negative CEs were associated with greater levels of depression.
Limitations
There are some limitations involved with using an older dataset. For instance, the
Diagnostic and Statistical Manual of Mental Disorders has been updated. Applying older diagnostic criteria has the potential to diagnose inaccurately psychological disorders. However, the
Diagnostic and Statistical Manual of Mental Disorders, 5
th edition;
DSM-5; American Psychiatric Association, 2013) includes similar criteria for major depression compared to earlier editions (
Paris, 2013). In addition, the Beck Depression Inventory (BDI-IA;
Beck et al., 1979) was applied during the initial study and the most current edition of this scale is the Beck Depression Inventory II (BDI-II;
Beck, Steer, Brown, 1996). These two versions of the depression inventory have, however, been found to diagnose similar levels of depression (Beck, Steer, Ball, & Ranieri, 1996) and are highly correlated in terms of measuring depression (
r = .93,
p < .001).
Although the present study relied on data drawn from an older study, the latter is considered to be a highly important random controlled trial, which is evidenced by the republishing of the original work (
Jacobson et al., 1996,
2000). In addition, other studies have recently used this same dataset for research purposes. For example,
Tang and Derubeis (2005) applied measures of total cognitive change to the dataset through the use of a self-report measure (Patient Cognitive Change Scale; [PCCS];
Tang and DeRubeis, 1999). However, the scale they used was a self-report measure, which may not be the most accurate reporting of cognitive errors for patients with depression. For instance, depressed patients are more likely to be affected by a negative reporting bias (
Owsley, & McGwin, 2004; Shean, & Baldwin, 2010). Using the observer-rated CERS on this dataset allowed for a more objective examination of CEs using a specific cognitive distortion typology.
Conclusion
Several general trends and conclusions may be drawn from this study. The theoretical tenets of cognitive therapy (e.g.,
Beck et al., 1979;
Ellis, 1980) were supported in that all types of cognitive errors were found in the spontaneous speech of participants with depression. The distortions of labeling negative, should and must statements negative, jumping to conclusions negative, mind-reading negative, overgeneralizing negative, and fortune telling negative, were most prevalent. It remains to be seen to what extent this may differ in non-depressed individuals.
The hypothesis that participants have more negative than positive CEs was also supported. No relationship between positive and negative CEs was found at early therapy for the entire sample. However, there were a few significant correlations among individuals with high and low distortions, suggesting that for the most part, depressed participants did not have a general tendency to distort information, but that distortion was more likely to occur in the negative, rather than in the positive direction. There were also no significant correlations between cognitive errors and levels of depression for the entire sample. However, when individuals with low and high distortions were separated, several correlations were found. In particular, among the high distorters, higher levels of negative CEs and lower levels of positive CEs were related to higher levels of depressive symptoms.
This study presents a number of limitations. First, a relatively small sample size may have limited the power to detect additional findings. Also, while the CERS allows for an assessment of the quantity of CEs, the qualitative properties, such as the meaning of CEs to the participant, could not be assessed. Furthermore, while it may be reasonable to assume that priming is not necessary to elicit cognitive errors, as these can be found at multiple levels of cognition (e.g.,
Beck, 1995), the low prevalence of CEs in this sample of depressed patients raises questions about the number of errors displayed in a therapy session that is not particularly stressful. Finally, the methodology did not allow for latent CEs to be assessed.
The strengths of this research is the use of detailed session coding as a methodology to examine the relationship between types and levels of distortions and severity of depression. This window into individual sessions may allow therapists to better understand how CEs function within sessions, and the CERS could be a beneficial tool for researchers and clinicians. Indeed, the majority of the research has used self-report assessment of automatic thoughts and dysfunctional attitudes. Further, much of the research in cognitive therapy for depression has focused on the role of negative distortions. The current study is one of only a few to examine positive cognitive errors. An observer-rated method contributes information that should be considered together with data derived from studies that have used laboratory settings and self-report instruments. This methodology also allowed cognitive errors to be assessed as they naturally occurred in the participants’ speech, rather than having participants recall the frequency of past thoughts, or estimate how they might be likely to distort in a hypothetical situation. Finally, the current study provided a detailed profile of the frequency of cognitive errors in a sample of depressed participants.
The current study contributes to existing knowledge by offering a more detailed account of the role of both positive and negative cognitive errors in depression, and from a novel perspective, using an observer-rated method. Further, individual differences were highlighted so that group means would not obscure unique findings between groups that may be of clinical utility to practitioners. One clinical implication of this study is that positive and negative CEs may signal different psychological processes for different people. Furthermore, some people may have a tendency to distort, while others may selectively distort in the negative direction. Practitioners may wish to pay greater attention to how positive and negative CEs impact the moods of their depressed patients, as theoretical writings have largely emphasized the role that negative CEs have on mood.