Self-Reported Cognitive Failures in Everyday Life: A Closer Look at Their Relation to Personality and Cognitive Performance

Participants

Study 1 included N = 158 healthy adults (54% female) between 20 and 86 years of age (M = 51.7, SD = 15.5) recruited in a city in Southern Germany. Their native language was mostly German (94.9%), or German combined with another language (1.9%), or a language other than German (3.2%). All were fluid in German. The education level was as follows: basic school graduation: 12.0%; finished vocational training: 18.4%, high school graduation: 30.4%; bachelor’s degree: 6.3%; master’s degree: 27.8%; PhD: 4.4% (no information: 0.6%). The inclusion criteria were (1) German as native language or fluent in German, (2) no medical problems potentially hampering cognition (e.g., meningitis) and no pervasive developmental or learning disorders (e.g., autism), and (3) normal or corrected-to-normal vision and hearing.

The sample size was based on power considerations. Using G*Power 3 (Faul, Erdfelder, Lang, & Buchner, 2007), we estimated that in a linear regression model with possibly eight predictor variables (Big Five, processing speed, memory, and inhibition), with multiple correlation coefficient ρ² = 0.1 (small effects), an α error probability = .05 (for “suggestive evidence,” Benjamin et al., 2017), and a power = .80, one would need N = 146 participants. We recruited more, in case some participants dropped out between the sessions or had to be excluded (which did not happen). After data collection, we decided to analyze the data with structural equation models (SEM), which are a combination of linear regression and factor models. Thus, Supplementary Table S5 (available in the online version of the article) includes the findings of Study 1 based on the originally planned linear regression models. The pattern of findings and general interpretation of the data did not differ from SEM.

Procedure

Study 1 consisted of two sessions in the laboratory, lasting ~90 minutes. The first session included the measures of demographics, cognitive failures, personality, memory, processing speed, and life satisfaction (a control variable). The second session included the inhibition tasks and was, on average, 4 days after the first session (M = 3.5, SD = 3.0). A psychologist or a trained research assistant instructed the assessment and answered questions. All participants provided written informed consent. They were paid 25 euros. There was no dropout.

Measures

Data Analyses

The analyses were based on structural equation modeling. All models were calculated with Mplus 7.4, using full information maximum likelihood estimation (FIML) with robust standard error correction (MLR). Only 0.4% of the data were missing, mostly because participants skipped an item occasionally. Because of FIML estimation, all observed data points were used in the analyses. In line with Beauducel and Wittmann (2005), model-fit was evaluated with the χ² test, the comparative fit index (CFI), the root mean square error of approximation (RMSEA), and the standardized root mean square residual (SRMR). For model identification, the first loading of each factor was fixed to 1. All latent factors were allowed to correlate. In line with the current discussion on redefining statistical significance, we followed the suggestion of Benjamin et al. (2017) and considered all p values less than .005 to be significant and all p values less than .05 to be suggestive evidence.

Descriptive Statistics and Reliability

Table S1 of the online supplementary material includes the descriptive statistics and reliability estimates of all measures of Study 1. The internal consistencies of the CFQ and cognitive performance were good to acceptable. The personality domains demonstrated acceptable internal consistencies for a short scale, which were similar to other findings on this scale (BFI-S, e.g., Hahn et al., 2012).

The Relations of Cognitive Failures, Personality, and Cognitive Performance

The main analyses consisted of three steps: (1) testing the pattern of relations between self-reported cognitive failures, personality, and cognitive performance; (2) considering relevant control variables such as the participants’ age; and (3) investigating whether differential effects for cognitive failure types exist. All three steps were analyzed with SEM.

First, we tested the correlations between cognitive failures, personality, and cognitive performance (Figure 1). The model fitted the data well and showed that, in line with our expectations, cognitive failures (as indicated by an overall CFQ score) were positively correlated with neuroticism and there was also suggestive evidence for a negative relation with conscientiousness. ³ Thus, on average, participants who were more neurotic and less conscientious, scored higher on the CFQ. Furthermore, in line with our expectations, cognitive failures were unrelated with agreeableness and openness. Against our expectations, there was suggestive evidence for a negative relation with extraversion, albeit this correlation was rather small. As the descriptively strongest effects were found for conscientiousness and neuroticism, our findings largely support previous findings on cognitive failures and personality (e.g., Wallace, 2004). Against our expectations, cognitive failures were unrelated with processing speed and inhibition. In line with our expectations, they were also unrelated with memory. Thus, on the level of latent cognitive performance, there are no relations between cognitive failures and cognitive performance. All other correlations, which were not relevant to our hypotheses, are displayed in Table S2 (available in the online version of the article; e.g., the correlations among the personality domains and cognitive performance).

OPEN IN VIEWER

Figure 1.

The relations of cognitive failures, personality, and cognitive performance (Study 1).

Notably, all effects were tested with correlations first, which is a more liberal testing procedure than regression models (in which the effects are controlled for the other predictors) and allowed us to detect all correlates of the CFQ. In the next step, we used regression coefficients instead of correlations to understand which correlates of CFQ were also significant predictors over and above the other predictors. When predicting the CFQ score ⁴ with the five personality domains and the three cognitive performance factors, conscientiousness (β = −.26) and neuroticism (β = .36) were the only predictors, indicating that they explained unique variance in cognitive failures (in contrast to extraversion). Together, they explained about 26% of the variance (Pseudo-R²_model = 29%, Δ-Pseudo-R²_{conscientiousness and neuroticism} = 26%).

Second, we included sex, age, and satisfaction with life as control variables (i.e., predictors) in our model (Supplementary Figure S1, available in the online version of the article). This modified model fitted the data well. The CFQ score was negatively related to satisfaction with life, and there was suggestive evidence for a relation with sex as well (with females reporting more cognitive failures). Age (20-86 years) was unrelated to the CFQ score, which is in line with the idea of cognitive failures as a facet of personality rather than cognitive performance (e.g., Wright & Osborne, 2005). The pattern of results was robust regarding these control variables. Neuroticism was significantly correlated to cognitive failures, and there was suggestive evidence for a correlation with conscientiousness and openness (Figure S1). Only conscientiousness (β = −.25) and neuroticism (β = .36) were predictors over and above the control variables and other predictors. Together, they still explained about 18% of the variance (Pseudo-R²_model = 31%, Δ-Pseudo-R²_{conscientiousness and neuroticism} = 18%). The reduction of explained variance is likely due to a construct overlap between neuroticism and satisfaction with life (r = −.29). However, there is a unique contribution of conscientiousness and neuroticism to the CFQ score over and above the control variables.

Third, we investigated whether there were differential effects for the factors clumsiness, retrieval, and intention forgotten (Wilhelm et al., 2010), which characterize different types of cognitive failures. Thus, we included three latent factors instead of one CFQ score (Figure 2). The model fitted the data well. Clumsiness, retrieval, and intention forgotten were correlated (.46-.79) but distinct factors. In line with our assumptions, neuroticism correlated positively with retrieval and intention forgotten, and there was further suggestive evidence for a positive relation to clumsiness (Figure 2). In contrast to our assumptions, conscientiousness was only related to retrieval based on suggestive evidence. Clumsiness (e.g., dropping things) and failures in maintaining intentions, however, were unrelated to conscientiousness.

OPEN IN VIEWER

Figure 2.

Different types of cognitive failures and their relation to personality and cognitive performance (Study 1).

Participants

Study 2 included N = 176 healthy adults (63% female) between 19 and 39 years of age (M = 24.8 years, SD = 4.4) recruited in two cities in southern Germany. Their native language was mostly German (90.9%) or German combined with another language (1.8%), or a language other than German (6.2%). All were fluent in German. The education level was as follows: basic school graduation: 0.6%; finished vocational training: 2.8%, high school graduation: 67.0%; bachelor’s degree: 18.2%; master’s degree: 9.1%; PhD: 0.6% (no information: 1.7%). The inclusion criteria were the same as in Study 1. Notably, the samples of both studies were independent and based on power considerations. Using G*Power 3 (Faul et al., 2007), we estimated that in a linear regression model with possibly 12 predictor variables (6 subfacets each of conscientiousness and neuroticism), with multiple correlation coefficient ρ² = 0.1 (small effect), an α error probability = .05 (for “suggestive evidence,” Benjamin et al., 2017), and a power = .80, one would need N = 170 participants for Study 2. We recruited more, in case some participants had to be excluded (which did not happen).

Procedure

Study 2 lasted ~90 minutes (including a 10-minute break), in which the participants answered one questionnaire in the laboratory (on demographics, cognitive failures, and personality). A psychologist or a trained research assistant instructed the assessment and answered questions. All participants provided written informed consent. They participated voluntarily without compensation. There was no dropout.

Measures

Data Analyses

The analyses were based on correlations and linear regression models. All models were calculated with Mplus 7.4 using FIML with robust standard error correction (MLR). One additional analysis was based on a factor model (Figure 3), which was estimated and evaluated as the models in Study 1. Only 1.1% of the data were missing, because participants skipped an item occasionally. Due to FIML estimation, all observed data points were used in the analyses. We followed the suggestion of Benjamin et al. (2017) and considered all p values less than .005 to be significant and all p values less than .05 to be suggestive evidence.

OPEN IN VIEWER

Figure 3.

Neuroticism and cognitive failures (Study 2).

Descriptive Statistics and Reliability

Supplementary Table S3 (available in the online version of the article) includes the descriptive statistics and internal consistencies of all measures. Taken together, the internal consistencies of the CFQ and the personality domains were good (Cronbach’s α = .83-.92), and the subscales of neuroticism and conscientiousness had acceptable internal consistencies (α = .47-.87), which largely overlap with other findings on this personality scale (NEO-PI-R, e.g., Berth & Goldschmidt, 2004).

The Relations of Cognitive Failures With Personality

The main analyses consisted of two steps: (1) conceptually replicating the pattern of relations between self-reported cognitive failures and personality found in Study 1 and (2) testing whether these relations are driven by the theoretically expected subfacets of conscientiousness and neuroticism. Both steps were analyzed with correlations and linear regression models.

First, we tested the correlations between cognitive failures and the Big Five of personality. In line with our expectations and with Study 1, cognitive failures (as indicated by an overall CFQ score) were negatively correlated with conscientiousness (−.43) and positively with neuroticism (.60). Furthermore, in line with our expectations and with Study 1, cognitive failures were unrelated with openness. Against our expectations, but in line with Study 1, cognitive failures were negatively correlated to extraversion, albeit this correlation was rather small (−.21). Against our expectations, and in contrast to Study 1, cognitive failures were negatively correlated with agreeableness (−.23). Thus, in four out of five cases, we conceptually replicated the findings of Study 1. The intercorrelations between the Big Five were mostly moderate and are displayed in Supplementary Table S4 (available in the online version of the article). Again, the strongest effects were found for conscientiousness and neuroticism. Notably, all effects were tested with correlations first, which is a more liberal testing procedure than regression models (in which the effects are controlled for the other predictors) and allowed us to detect all correlates of the CFQ. In a next step, we used regression coefficients instead of correlations. When predicting cognitive failures with the Big Five, conscientiousness (β = −.19) and neuroticism (β = .52) were the only significant predictors, indicating that they explained unique variance in cognitive failures (in contrast to extraversion and agreeableness). Together, they explained about 32% of the variance (pseudo-R²_model = 41%, Δ-pseudo-R²_{conscientiousness and neuroticism} = 32%). This is the same pattern as found in Study 1, and it remained stable after controlling for sex, age, and satisfaction with life (as in Study 1). Together, conscientiousness (β = −.20) and neuroticism (β = .51) still explained about 24% of the variance (pseudo-R²_model = 42%, Δ-pseudo-R²_{conscientiousness and neuroticism} = 24%). This reduction of explained variance is likely due to a construct overlap between neuroticism and satisfaction with life (r = −.50). However, there is a unique contribution of conscientiousness and neuroticism to the CFQ score over and above the control variables.

The high correlation between the CFQ and neuroticism on a manifest scale level raised the question of whether they are even stronger related on a latent level, because they might assess the same underlying construct. To test this question, we estimated the model in Figure 3, which fitted the data well. Of course, neuroticism and cognitive failures were also strongly related on a latent level. Testing their correlation against 1, however, resulted in a severe drop of model fit, Δχ²(df = 1) = 26.01, p < .001; ΔCFI = .06, indicating that two factors represent the data better than one and that neuroticism and cognitive failures are related but separable psychological constructs.

Second, we tested the correlations between cognitive failures and the subfacets of conscientiousness and neuroticism. In line with our assumptions, cognitive failures were negatively related to the subfacets self-discipline (−.42) and order (−.33) of conscientiousness and positively related to the subfacets self-consciousness (.35), depression (.46), and anxiety (.55) of neuroticism. We had no directed hypotheses concerning the other subfacets. Cognitive failures were negatively related to the subfacets competence (−.36) and dutifulness (−.38) of conscientiousness, and there was suggestive evidence for a relation to the subfacet achievement striving (−.17) as well. Cognitive failures were positively related to the subfacets angry hostility (.48), impulsiveness (.37), and vulnerability (.55) of neuroticism. The subfacet deliberation of conscientiousness was the only one unrelated to cognitive failures. Taken together, the relations of cognitive failures with conscientiousness and neuroticism appeared to be based on almost all subfacets of conscientiousness and neuroticism. At this point, we refrained from testing regression models because the subfacets of conscientiousness and neuroticism were too strongly correlated (up to .74) to be included as parallel predictors.