Absentminded lapses of attention and a general tendency not to be fully mindful of our present experiences are common threads running through most of our daily lives. Although generally taken to be rather trivial events, they can sometimes have dramatic and even life-threatening consequences (Reason, 1984). Indeed, not paying attention to the critical task of driving the train may have been the primary cause of a major commuter train crash near Chatsworth, California, when an engineer became too focused on sending text messages from his cell phone to either notice or respond to operating signals telling him he should stop the train (National Transportation Safety Board, 2008). Such examples highlight the dramatic consequences that disruptions in attention can have when they happen to occur at particularly inopportune moments, but, given the pervasiveness of attention lapses in everyday life, particularly inopportune moments may not be a necessary component in this process. Indeed, the cumulative effects of inattention may have the potential for less obvious and dramatic, but ultimately more persistent emotional consequences. In our earlier work we presented data consistent with the view that seemingly minor episodes of everyday inattention, and their resulting minor cognitive and behavioral errors, are potentially important factors in the development of depression (Carriere et al., 2008). The present paper extends this earlier argument by testing two follow-up hypotheses that the influences of everyday inattention on depression are mediated, in part, by (1) one's general perceived self-efficacy and (2) one's overall experience of psychological stress.

To examine the potential influence of inattention on depression we (Carriere et al., 2008) previously conducted a large-scale questionnaire study and used structural equation modeling to test and compare two hypotheses: (1) that chronic cognitive failures in attention are an initial cause of depression (the Attention-to-Affect hypothesis), and (2) the reverse hypothesis that depression causes non-specific attention and memory failures (the Affect-to-Attention hypothesis). These hypotheses are both based on the commonly held belief that there is a causal connection between attention and depression, but their corresponding models differ with respect to how this causal link plays out in everyday life. To evaluate these hypotheses each model was compared against the null hypothesis, addressing the question of whether either an Attention-to-Affect or Affect-to-Attention model was capable of explaining the observed relations among the relevant variables. These analyses showed only the Attention-to-Affect model provided good fit for the data, by modeling attention lapses and associated everyday mistakes as a common cause of both memory failures and depression (Carriere et al., 2008). This causal flow is consistent with findings from MacLeod and colleagues (MacLeod et al., 2002), but contrasts with the traditional view of how emotion and attention are linked. Indeed, although not often articulated as an explicit hypothesis (owing, presumably, to the ubiquity of this view), the more common perspective is that emotions influence our attention. Indeed, for example, this appears to be the received view in most research on attentional biases in anxiety and depression (e.g., MacLeod et al., 1986; Gotlib et al., 1988; Dalgleish and Watts, 1990; Mogg et al., 1995; see also a review in Ingram et al., 2008) and on the attentional effects of encountering emotionally salient information (e.g., biases toward negative emotions; Fenske and Eastwood, 2003). With respect to the latter view, however, it is worth noting that attention has also been shown to have reciprocal effects on later emotional evaluation of stimuli (Fenske and Raymond, 2006).

Given the above, we now seek to elaborate on the Attention-to-Affect hypothesis by evaluating another path by which attention lapses could influence our affective state—one's sense of self-efficacy. Self-efficacy is a judgement about our ability to successfully achieve our goals—usually in reference to a specific situation or activity (Bandura, 1977). Our sense of self-efficacy reflects both the expectation that we are capable of performing a task and that most others are not substantially more capable of performing the same task (Davis and Yates, 1982). Such beliefs apply not only to special skills or talents, but also to our ability to perform everyday tasks or cope with novel life challenges. These latter beliefs reflect a generalized sense of self-efficacy (Tipton and Worthington, 1984), which could be undermined in the face of chronic attention lapses interfering with our ability to perform even simple everyday tasks.

Reduced self-efficacy involves the belief that one's failures are the result of an inherent lack of ability rather than a consequence of situational factors or simply insufficient effort (Bandura, 1997). The obvious parallels between such beliefs and the counterproductive, negative affect-laden biases that are typical of depression support Bandura's claim that depression is a potential outcome of decreased self-efficacy. In particular, as he noted, if through our sense of inefficacy we come to believe our successes are the result of luck and failures represent our actual ability, depression will result in part because we devalue our accomplishments and overvalue our failures. Ultimately, because failure at a task is seen as more meaningful than success, even when initial failures are followed with later successes, we may still be less willing than others to engage in the same task again. On this view, reduced self-efficacy is thought to create cognitive biases that produce depression, and potentially influence attention by continually directing it away from information critical to successful task completion (Bandura, 1997). Furthermore, given that our sense of self-efficacy is relative to the ability of others, depression, and an unwillingness to even attempt tasks, is an especially likely result when we expect others would not have difficulty completing the same tasks with which we have had trouble (Davis and Yates, 1982).

According to the Attention-to-Affect hypothesis (Carriere et al., 2008) frequent attention lapses lead to attention-related errors that interfere in task success and should precede reductions in self-efficacy for that task. Given that attention lapses occur even in simple, everyday activities—particularly those we know are not problematic for most people—it follows that these continual mistakes could produce a generally diminished sense of self-efficacy and may then lead to disengagement from everyday behavior in general and eventually to depression. The present research thus applied the Attention-to-Affect hypothesis to illuminate more precisely the structure of these relations. In particular, we sought to evaluate the potential causal flow from attentional errors—specifically attention lapses and subsequent attention-related errors—through to decreased self-efficacy and on to a proneness to experience depression.

To examine the hypothesis that perceived self-efficacy mediates the association between failures of basic cognitive mechanisms and depression, we conducted a path analysis using structural equation modeling in AMOS (Arbuckle, 2005) to examine the relations among five self-report questionnaires. Frequency of attention lapses and associated cognitive errors were assessed via the Mindful Attention Awareness Scale–Lapses Only and the Attention-Related Cognitive Errors Scale (MAAS-LO and ARCES: Carriere et al., 2008). We assessed depression via the Beck Depression Inventory—Second Edition (BDI-II: Beck et al., 1996). Self-efficacy was measured via the Generalized Self Efficacy scale (GSE: Schwarzer and Jerusalem, 1995).

Based on previous findings (Carriere et al., 2008) and the present arguments, we predicted the MAAS-LO would explain a significant amount of the variance in the BDI-II, while the ARCES would function as a mediator between the MAAS-LO and GSE, such that the ARCES would not explain a significant amount of the variance in the BDI-II once relations with the MAAS-LO and GSE were accounted for. The ARCES is the key variable in this case, since it allows us to place specific a priori constraints on the model, which are necessary for evaluating causal claims with correlational data, and it is most clearly recognized as an effect of inattention (i.e., it is non-sensical to instead argue that attention-related errors cause the inattention that was necessary for them to have occurred). With these same data we also evaluated the more conventional Affect-to-Attention hypothesis that negative affect creates cognitive biases which influence attention (e.g., through reduction of attentional capacity by rumination) and our perceived self-efficacy.

Consistent with our earlier findings (Cheyne et al., 2006; Carriere et al., 2008), the ARCES and MAAS-LO were found to have good distributional and psychometric properties. All measures showed a good range of scores, no significant deviations from normality in skewness and kurtosis, and demonstrated very satisfactory internal consistency (see Supplementary Table S1, for detailed information).

Pearson Product-Moment correlation coefficients are presented in Table 1. All observed coefficients were moderate to large. As predicted, the ARCES and MAAS-LO were positively correlated, and both were negatively correlated with the GSE. Furthermore, as predicted, the GSE was also negatively associated with the BDI-II. Overall, the correlations between attentional and mood measures replicated our previous findings (Carriere et al., 2008).

Our primary interest was in extending our knowledge about the role that everyday attention lapses play in the onset of depression. Hence, a model was constructed which allowed self-efficacy to mediate the relations from everyday inattention (the MAAS-LO) and attention-related errors (the ARCES) to depression (the BDI-II). To evaluate the conventional Affect-to-Attention hypothesis we simply reversed the causal paths involving depression, producing a model in which depression predicts inattention and self-efficacy.

The Attention-to-Affect model assessed the simultaneous effects of the MAAS-LO and GSE on the BDI-II, using the MAAS-LO as an exogenous variable predicting the ARCES, GSE and BDI-II. The GSE was also entered as a mediator of the effect of the MAAS-LO on the BDI-II, and the ARCES was entered as a mediator of the effect of the MAAS-LO on the GSE. This model provided reasonably good fit indices for the data, χ² (1, N = 188) = 2.05, p = 0.152, CFI = 0.995, TLI = 0.972, RMSEA = 0.075, consistent with the Attention-to-Affect hypothesis, via changes in self-efficacy. As anticipated based on previous findings (Carriere et al., 2008) the direct path between the ARCES and BDI-II was found to be non-significant (b^* = 0.105, p = 0.156) and omitted. The only notable shortcoming of this model was that the path from the MAAS-LO to the GSE was not significant (b^* = −0.027, p = 0.772). This is because almost the entirety of the MAAS-LO's relation with the GSE was mediated by the ARCES in this sample—consistent with the hypothesis that one's everyday inattention leads to a decreased sense of self-efficacy primarily through the attention-related cognitive errors that also result from attention lapses. Accordingly, a more parsimonious model, removing the non-significant path from the MAAS-LO to the GSE, is shown in Figure 1A and provides very good fit indices for the data, χ² (2, N = 188) = 2.15, p = 0.341, CFI = 0.999, TLI = 0.998, RMSEA = 0.020. For all direct and indirect effects see Supplementary Table S2.

The traditional Affect-to-Attention model was tested next. This model treats the BDI-II as an exogenous variable predicting both the MAAS-LO and GSE, and with the MAAS-LO mediating the relation of the BDI-II and ARCES. The Affect-to-Attention model, which implicitly directs most research, does not specifically predict a direct influence of self-efficacy on attention-related errors—that is, without influencing attention lapses first—so in the Affect-to-Attention model that we tested the direction of the path between the ARCES and GSE remained consistent with the model shown in Figure 1A. Accordingly, the Affect-to-Attention model shown in Figure 1B matches Figure 1A, except that all paths directly connecting with the BDI-II are reversed. This model provided much poorer fit indices for the data, χ² (1, N = 188) = 6.61, p = 0.010, CFI = 0.976, TLI = 0.854, RMSEA = 0.173. For all direct and indirect effects see Supplementary Table S3. Importantly, this model, which most closely matches the conventional Affect-to-Attention hypothesis, clearly does a poor job of representing the obtained pattern of relations and thus the implication is that conventional theories may be inadequate. Furthermore, this model requires a direct connection from the MAAS-LO to the GSE in order to better explain the observed covariance matrix, so is also potentially less parsimonious than the Attention-to-Affect model. A large source of variance left unaccounted for in the traditional Affect-to-Attention model involved the relation of the BDI-II with the ARCES, suggesting that, ultimately, the Affect-to-Attention model is not as effective in predicting the occurrence of attention-related errors as the Attention-to-Affect model is in predicting depression.

Given the present findings, the Attention-to-Affect model, shown in Figure 1A, best and most parsimoniously fits the data. Within this model, the MAAS-LO predicts all cognitive and affective variables, with the GSE mediating relations between the MAAS-LO and BDI-II. Moreover, in conjunction with attention-related cognitive errors (e.g., going to the fridge to get some milk, and instead taking out the juice), such attention lapses may be viewed as influencing our affective wellbeing via their impact on our sense of self-efficacy. Thus, the present findings once again highlight that maintaining an awareness of our actions is an important contributor not only to the outcomes of everyday activities, but potentially to our long-term emotional wellbeing.

The Attention-to-Affect model naturally focuses more on cognitive factors (i.e., attention lapses, attention-related errors, and self-efficacy) rather than affective factors (depression). Although we have thus far stressed these cognitive routes, we also acknowledge the existence of other affective mechanisms on the route from attention lapses to depression. One such potential linking mechanism is stress, particularly given that attention lapses can interfere with our ability to perform our normal, everyday activities and through such interference we may become stressed.

Here we adopt Selye's (1984) belief that the term stress should be reserved to describe physiological or emotional outcomes while the term stressors should be used to describe those events that cause stress. This view of stress as an emotional outcome is compatible with the work of Lazarus (1993), who argued that stress can be seen as a cognitive-emotional trait which persists across situations and remains present even in the absence of external stressors. On this view, there may sometimes be no identifiable situational cause for an individual's stress; that is, stress is simply an abundance of psychological tension, arousal, and other non-specific emotional distress resulting from no particularly salient cause. A similar view was put forward by Selye (1984), who argued that stress is a potential consequence of almost all physical, cognitive, or emotional activity—that essentially everything we do or feel has the potential to produce a stress response in the body, even if only to a very minimal extent. Accordingly, to say someone is feeling stressed reflects the individual's recognition of an abundance of the stress response rather than a truly categorical change in his or her present state. Expanding on the views of Lazarus (1993), we view stress more generally as a host of unpleasant affective traits, such as general irritability, that attention lapses may help to create, and which then potentially produce further emotional distress in the form of depressive affect. The challenge, accordingly, is to determine how stress fits into the Attention-to-Affect model.

Once again, all measures were found to have good distributional and psychometric properties, with only the GSE reflecting a larger degree of kurtosis (though still acceptable, for full details see Supplementary Table S4). Pearson Product-Moment correlation coefficients are presented in Table 2. All coefficients were significant, and the majority was moderate to large. As predicted by the theory discussed earlier, both the MAAS-LO and ARCES show strong relations with DASS-Stress. Furthermore, consistent with the findings of Study 1 and the present theory, the GSE showed strong relations with the DASS-Stress, and DASS-Depression.

Starting from our earlier model (Carriere et al., 2008), which postulated attention lapses as a common cause of a sequence of cognitive, behavioral, and affective outcomes, the present studies were designed to investigate additional potential intermediates between attention lapses and affective dysfunction. The addition of self-efficacy as a collection of beliefs about our ability to perform everyday tasks (Study 1) and stress as an individual tendency to experience an unpleasant abundance of tension, arousal and general emotional distress (Study 2) allowed the expansion of our causal model from initial attention failures to affective outcomes. Once again, the present findings provided good support for the view of attention lapses as a cause of depression. The present findings are also consistent with a view of attention lapses as potential stressors in their own right—capable of producing stress without the consequent major life events more typically identified as causes of stress. There remains the distinct possibility, however, that everyday inattention also increases the likelihood that we encounter difficult life situations, and so stressful life events are still an important component of the attention–stress relation. As we did not specifically inquire about these participants' experiences of stressful life events, the present findings cannot directly address this question.

It is worthwhile highlighting once again that relation of the MAAS-LO with the GSE was fully explained by the ARCES in both Study 1 and Study 2. Accordingly, we conclude that reduced self-efficacy results from attention lapses that lead us to make attention-related errors when completing everyday tasks, although stress may play an important role in this process as well. That is, it is only when we either experience stress or notice the negative effect inattention has on completing tasks, that our sense of self-efficacy is diminished. Thus, future studies looking to model the effect of inattention on self-efficacy should incorporate at least attention-related errors as a mediator of this relation.

A limitation of the present models is that they rely solely on relations observed among self-report questionnaire assessments of attention and emotion. This limitation is a consequence of the present research focusing on trait-level tendencies of the individual. That is, we investigated general tendencies to be inattentive, stressed, or depressed in everyday life regardless of specific situations one might encounter. Taking these general tendencies to the level of specific situations may not be easily accomplished because any given situation will inevitably introduce its own complexities. Such complexities are often unexpected and difficult to account for, and thus correct interpretation of the results is made more difficult. For example, it is probable that, although attention lapses may serve an important role in the etiology of depression, depressed affect will be associated with self-focused depressive rumination that provides additional attentional load and is associated with mind-wandering away from important characteristics of the task at hand (Smallwood et al., 2003, 2007)—and this process is likely to play out over both long and short timescales. Indeed, real-life scenarios are likely to involve reciprocal relations between trait inattention, trait and situational stress, and depression where a general tendency toward inattention increases the likelihood of stressful events occurring (such as in the example of the distracted Chatsworth train engineer from the introduction) and stressful events then lead to depression, both of which create new cognitive loads that make maintaining attention even more difficult, and the cycle repeats. Likewise, stressful life events could be responsible for the creation of short-term attentional difficulties and depression, and once these events are resolved the other effects will not necessarily persist. Notably, the studies discussed here, being focused solely on trait-level cognitive and emotional processes, do not preclude any such possibilities. What they do suggest is that everyday attentiveness should serve as a protective factor, minimizing the potential impact of traumatic life events, and that with careful control and interpretation, future studies might benefit from including measures of attention lapses and an attempt to better understand the long-term cognitive, emotional and behavioral interactions that play an important role in the onset and maintenance of depression. In the end, that several different studies have produced similar findings, even while involving different mediators of the link between attention lapses and affective distress, should make their findings more compelling. Accordingly, the results obtained in the present studies should be interpreted as a stimulus for additional research on the potential involvement of seemingly harmless lapses of attention in everyday life in the development of highly problematic emotional states.

It is worth noting that, although progressing down a somewhat different causal path, the present line of research is perfectly compatible with the findings of Killingsworth and Gilbert (2010). Through an experience sampling study these researchers have shown a significant relation between mind wandering in everyday situations and future happiness, with happiness reportedly decreasing following reports of mind wandering. The effect of mind wandering on happiness was independent of, and larger than, the emotional effect of the activity performed at the time. Furthermore, mind wandering was associated with decreased happiness when the content of that thought was either neutral or negative, relative to the same activity performed without mind wandering. Interestingly, mind wandering also did not lead to an increase in happiness when the topic was pleasant. Together these findings suggest mind wandering has a general negative effect on our emotional state, which may be only partially counteracted by thinking about pleasant topics. Of course, other slightly more recent work similarly examining the connection between mind wandering and negative mood produced contradictory results, where sadness predicted mind wandering but mind wandering did not predict sadness on short timescales (Poerio et al., 2013). Again, though similar in some respects to the present studies, these findings are based on reports of mind wandering rather than attention lapses or attention-related errors, and momentary happiness rather than depression over longer timescales. Nevertheless, there are obvious links between inattention and mind wandering, and between depression and momentary happiness, and so it is good to see some potential convergence on the same general conclusion in these separate lines of research.

On the whole, the present results provide good support for the hypothesis that chronic attention lapses may set in motion processes that ultimately lead to depression, including those working via their influence on our stress level. This process is likely to be, in part, also mediated by the effects of failures of attention on our general sense of our ability to accomplish both mundane and novel tasks. These models redefine the role of attention lapses in our everyday emotional experiences and are all the more compelling as a result of the consistency with which attempts to address the various causal relations underlying these experiences produce similarly well-fitting models. As a result, the present findings suggest that direct attentional training exercises could be an important, and presently underutilized, method of preventing onset or relapse of depressive episodes. Of course, the present research is only an initial foray in this domain, and it is worth noting again that the present models contrast the most simplified views of the interrelations among attention, self-efficacy, stress, and depression. We do not believe there is any debate over the validity of these interrelations, nor the belief that causality is involved in these relations. The Affect-to-Attention models we tested therefore represent what we believe are the most typical theoretical perspectives taken when discussing causation among these particular variables; those models failed to explain the data. The Attention-to-Affect models likewise represent our preferred alternative theoretical perspectives on the causal roles, after recognizing that seemingly minor inattentiveness is certainly capable of producing a wide range of major life consequences and that at least one study has already suggested it is possible to treat major depressive disorder with only attention training (Papageorgiou and Wells, 2000). But it is undoubtedly the case that in the real world the interrelations are more complex than any of our models allow. Indeed, the primary causal flow may be from attention-to-affect in general, but from affect-to-attention (particularly from stressors to inattention) in many real-world scenarios. It is clear there is a great deal of additional research to do with regard to testing even more potential pathways from either attention-to-affect or affect-to-attention (e.g., anxiety is likely to be another important variable) and we will not fully understand the connection between attention and affect without considering more complex alternatives from all plausible causal perspectives.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by University of Waterloo Research Ethics Board. The patients/participants provided their written informed consent to participate in this study.

JSAC is the main author, responsible for most data analysis, and writing. AN was responsible for data collection in Study 2 and assisting in developing the theoretical models. JAC and DS was a supervisor and assisted with proofreading and conceptual development of the models. All authors contributed to the article and approved the submitted version.