Affect Recall Bias: Being Resilient by Distorting Reality

In recent decades, one of the most studied constructs in the psychological field has been represented by emotion regulation (ER) (Fernández-Álvarez et al. 2018), which refers to the process of down-regulating or up-regulating ongoing emotions in order to achieve desirable states (Gross 1998, 2015). Although the previous literature mainly investigated ER in relation to negative emotions, there is now increasing evidence showing the importance of positive affect (PA) in many aspects of our life (Colombo et al., 2020a, b), and highlighting the importance of cultivating positive emotions (Carl et al. 2013; Diener and Seligman 2002; Lyubomirsky et al. 2005; Tugade and Fredrickson 2007). Accordingly, people adopt up to five different categories of strategies to regulate positive emotions (Gross 2015), which can be implemented either before, during or after the emotion-generation process (Quoidbach et al. 2015). More specifically, the use of strategies to increase PA after the occurrence of an event has been extensively supported, and it is now widely accepted that recalling positive memories is an effective strategy to upregulate positive emotions (Mitchell et al. 1997; Wilson and Ross 2003), enhance emotional well-being, and increase happiness (Bryant et al. 2005; Quoidbach et al. 2010).

Nonetheless, the emotions experienced during an event do not necessarily match with the emotions prompted by the associated memory: The intense sadness experienced after losing a job, for instance, might be remembered less intensely (i.e. underestimation) or more intensely (i.e. overestimation) sometime later. This phenomenon, called recall bias, has received increasing attention in recent decades because of its correlates with adjustment to life events (Skoronski 2010; Walker and Skowronski 2009). Hedges et al. (1985) first showed that people tend to overestimate experienced levels of PA and negative affect (NA) when asked to retrospectively recall their general mood (i.e., not in relation to a specific event). In recent years, different studies have confirmed this overestimation bias in recalling past affect (Ben-Zeev et al. 2009; Colombo et al. 2019a, b, c; Kardum and Tićac Daskijević 2001; Thomas and Diener 1990; Wirtz et al. 2003), and they have pointed to different factors affecting this mechanism, including beliefs, event reappraisal, contextual information, and personality traits (Levine et al. 2001; Robinson and Clore 2002; Safer et al. 2002). Whereas the tendency to overestimate negatively-valenced emotions has been interpreted as an adaptive evolutionary mechanism to increase the salience of threatening events, which is essential for survival (Miron-Shatz et al. 2009), it is still not clear whether the presence of a bias in recalling positive emotional experiences may play a role for mental health.

According to the Positive Illusion Theory, enhanced self-evaluations, exaggerated perceptions of control, and unrealistic optimism are adaptive mechanisms that improve people’s happiness, satisfaction in life, and well-being (Brookings and Serratelli 2007; Taylor and Brown 1988, 1994). Indeed, holding positive illusions is likely to increase the perceived abilities to deal with stressors, thus enhancing motivation and enthusiasm while carrying out actions (Taylor and Gollwitzer 1995). This, in turn, would make it easier to reappraise negative events and implement successful coping strategies (Brown 1993). Likewise, holding an optimistic past-oriented disposition (i.e., overestimating past affective experiences) may represent a sort of strategy to maintain and upregulate positive emotions over time (Bryant 1989), which in turn could increase people’s resilience and perceived coping skills: That is, the set of personal qualities that enable an individual to thrive in the face of adversity (Connor and Davidson 2003). The cultivation of positive emotions has been widely found to be an essential coping skill that helps people to deal with daily stressors, with important implications for mental health (Tugade and Fredrickson 2007). Accordingly, the habitual use of savoring strategies has been associated with higher levels of happiness and well-being (Jose et al. 2012; Quoidbach et al. 2010), whereas dampening (i.e. the use of strategies to downregulate positive affective experiences) has been linked to prolonged NA levels (Li et al. 2017; Wood et al. 2003). This could also explain the tendency of depressed patients to retrospectively underestimate PA (Ben-Zeev et al. 2009; Colombo et al. 2019a, b, c), which may arguably reflect the inability of these individuals to recall positive experiences and upregulate positive emotions (Bryant 2003).

Together, there is evidence suggesting that people’s past-oriented disposition and, more specifically, the way people recall experienced PA, may in itself be a tool to regulate current emotions. Consequently, the presence of a recall bias may have direct consequences for present affective states and, therefore, individuals’ mental health status and well-being. In the present study, we explored the potential effects of PA recall bias (i.e. the tendency to retrospectively overestimate or underestimate positive affective experiences) and NA recall bias (i.e. the tendency to retrospectively overestimate or underestimate negative affective experiences) on mental health outcomes (depressive symptoms, anxiety symptoms, and psychological well-being), and we investigated resilience as a potential mechanism (i.e., mediator) explaining this relationship. To do so, we asked participants to self-report their mood during a two-week Ecological Momentary Assessment (EMA) study, and we subsequently asked them to retrospectively recall the experienced PA and NA levels. We anticipate that PA overestimation will positively contribute to well-being and will be associated with increased resilience. More specifically, we hypothesize PA overestimation to positively impact mental health through the enhancement of an individual’s resources to cope with daily stressors. Conversely, and consistently with the previous literature (Miron-Shatz et al. 2009), we expect NA overestimation to be an evolutionary rather than coping mechanism and, therefore, not be associated with well-being.

In the present study, we explored affect recall bias by comparing daily to retrospective assessments of affect. Consistent with the previous literature, participants tended to overestimate negative affective experiences. Nevertheless, we did not find the general PA retrospective overestimation that was observed in previous studies (Ben-Zeev et al. 2009; Hedges et al. 1985). Specifically, in our sample, half of the participants were likely to overestimate past PA, but the other half were likely to underestimate it. This divergent result may be due to the novel assessment approach used in the present study (i.e., EMA using a smartphone app) (Colombo et al. 2019a, b, c), but also to the characteristics of the recruited sample, which included mildly depressed/anxious students. In relation to the latter, despite reporting a similar NA overestimation pattern, depressed individuals have been shown to overestimate PA to a lesser extent (Ben-Zeev et al. 2009) or even to underestimate it (Colombo et al. 2019a, b, c). Accordingly, it is possible that the presence of participants with mild symptoms influenced the results observed in this study, increasing the variability of PA bias distribution and leading to an equal number of overestimators and underestimators. This hypothesis cannot be confirmed at this stage and will require further investigation.

An interesting finding in the present study was that PA bias correlated and predicted mental health outcomes. More specifically, PA underestimation was associated with higher depression and anxiety symptoms, which confirms a recent meta-analysis that pointed out no significant differences in the effect sizes of reduced PA levels in depressed and anxious patients (Khazanov and Ruscio 2016). Additionally, PA overestimation correlated with and predicted higher levels of psychological well-being (self-acceptance, environmental mastery, purpose in life and positive relations), thus supporting the hypothesis that overestimating past positive affective experiences is likely to be an adaptive strategy that positively impacts many dimensions of mental health. Notably, an indirect effect of PA bias on psychological well-being through resilience was observed: In other words, our results suggest that the tendency to overestimate PA does not directly affect mental health, but instead adds up to the set of personal qualities that people use to face adversities (Connor and Davidson 2003) and indirectly enhances mental health. Consistent with the broaden-and-build theory (Fredrickson and Joiner 2002; Tugade and Fredrickson 2007), these results confirm the idea that cultivating positive emotions is a fundamental tool for the enhancement of resilience (Herrero et al. 2019).

Besides, NA bias correlated with and predicted the presence of mild depressive symptoms. Our results are coherent with a long tradition of research showing the presence of a negative bias in depressed patients, which involves increased elaboration of negative information, recall of more negative memories than positive ones, and difficulties in disengaging from negative information (Gaddy and Ingram 2014; Gotlib and Joormann 2010). Accordingly, the tendency to overestimate negative emotional experiences may reflect the negative bias that it is usually reported by depressed people (Craske and Pontillo 2001), and it may be already observable in individuals with mild symptoms. However, the direction of the association between depression and the negative bias remains an open question, and the correlational nature of our results does not allow to disentangle whether the presence of mild depressive symptoms provokes a negative bias or the other way around. Notably, NA bias only barely correlated with one of the Ryff’s well-being dimensions and it was not significantly associated with resilience. In addition, almost all the participants in our sample tended to overestimate past NA, and only a few people underestimated it. In other words, NA bias does not seem to have regulatory or coping functions, but it may rather be a “normal” tendency of individuals. Thus, whereas people appear to overestimate past PA to cope with daily stressors, they are not likely to use NA underestimation to feel better and deal with negative events.

In the present study, we also conducted an exploratory cluster analysis to investigate whether the combination of biases could lead to different associations with the outcome measures. Interestingly, participants in the “double bias” cluster (i.e., PA underestimation and NA overestimation) reported the poorest mental health status (in terms of depression, anxiety and psychological well-being). These scores were clearly worse than participants in the “positive bias” group (i.e., PA overestimation), but also slightly, yet not significantly worse than participants in the “negative bias” group (i.e., NA overestimation, PA accuracy). Although the latter differences were not statistically significant, these results and the marked differences with the “positive bias” group suggest that the concurrent overestimation of NA and underestimation of PA might potentially represent an important vulnerability factor for mental health. Due to the small sample size of each cluster, future studies are needed to confirm this hypothesis, which might open new avenues for research, prevention and treatment purposes. Another interesting finding was that people who tended to overestimate past positive emotions benefited from higher psychological well-being (self-acceptance and environmental mastery), even when compared to individuals who were almost accurate in retrieving their affect (“accurate” cluster). Despite representing a cognitive bias, we therefore suggest that the overestimation of past positive affective experiences can be considered an adaptive distortion of reality.

For many decades, accurate predictions have been considered a marker of mental health: Cognitive distortions or non-accurate representations of reality were considered vulnerability factors for psychopathology (Jahoda 1953). Consistent with the information processing perspective, humans were regarded as scientists gathering and elaborating information from the environment with the aim of building realistic and accurate pictures of the world (Fischhoff 1975). The Positive Illusion Theory offered a divergent perspective, suggesting that cognitive biases are adaptive in many circumstances (Taylor and Brown 1988), and our results are consistent with this latter idea.

According to Taylor and Brown (1988), people are likely to report three positive biases: Unrealistically positive views of the self, exaggerated perceptions of control, and unrealistic optimism. Rather than representing processing errors, these positive biases have been shown to be protective factors for mental health (i.e., useful resources to maintain and promote well-being and happiness). Consistent with this perspective, a growing body of studies has focused on the importance of a future-oriented disposition (Colombo et al. 2020a, b) and, more specifically, the repercussions that future perception has on mental health (Mikus et al. 2017; Weinstein 1980). Accordingly, a new construct called “openness to the future” has been proposed, which refers to the “positive expectations about what life may bring, a sense of competence and ability to cope with events, the anticipation, planning and perseverance to reach an outcome even in the face of adversity, and the acceptance of what cannot be resolved or predicted”. Similar to Taylor’s positive biases (1988), openness to the future has been shown to be a protective factor for well-being, and it has been associated with higher PA levels, psychological well-being, and self-esteem, as well as with reduced depressive symptoms, anxiety, and worry (Botella et al. 2018). Here, we propose that also people’s past-oriented disposition may represent an important protective factor for mental health. Building on our results, we showed that the way we perceive and recall our past experiences is associated with many mental health related dimensions. More specifically, overestimating past affective experiences is likely to be a protective factor associated with greater well-being. Conversely, the tendency to underestimate PA and strongly overestimate NA might potentially represent a risk factor for mental health, although the direction of this association cannot be clarified due to the observational nature of the design.

In relation to the previous, there is evidence suggesting that, while experiencing both positive and negative emotional states is adaptive and essential for survival, dysregulated NA levels along with reduced PA can be associated with the onset of many emotional disorders (Hofmann et al. 2012). From an evolutionary perspective, NA entails the implementation of avoidance or withdrawal behaviors in the face of challenging and dangerous situations, whereas PA fosters approach behaviors and the exploration of novel situations (Cacioppo and Berntson 1999). It is therefore possible to hypothesize that owning distorted emotional representations could affect an individual’s behaviors and attitude towards the external world. Accordingly, the concurrent overestimation of NA and underestimation of PA may lead to avoid new experiences and to focus more on negative rather than positive stimuli, which in turn may discourage positive attitudes such as exploration and curiosity.

From a clinical point of view, these results might have some important implications. Reduced PA is indeed a core component of anhedonia, which in turn is a vulnerability factor for mental health. Furthermore, altered levels of both NA and PA has been found to account for the onset and maintenance of many emotional disorders (Brown 2007; Brown and Barlow 2009). However, the primary focus of most of the available interventions has been placed on the reduction of symptoms and on the alleviation of negative emotional states. Considering the growing literature showing the short- and long-terms benefits of PA on health and mental health (Pressman et al. 2018), more efforts should be made in order to create interventions that focus on increasing PA levels and learning strategies to regulate not only negative emotions, but also positive ones. An example in this direction is the Positive Affect Treatment, which focuses specifically on the enhancement of PA. In a recent randomized controlled trial, the authors revealed better clinical outcomes in terms of depression, affect, anxiety, stress, and suicidal ideation in the group of patients receiving Positive Affect Treatment, compared to a Negative Affect Treatment (Craske et al. 2019). Additionally, a possible future line of research could lie in the self-monitoring of daily affect and ER deployment in relation to positive emotions by means of a smartphone based EMA similar to the one used in the present investigation. Symptom self-monitoring has already been shown to be an efficacious tool to increase awareness and self-empowerment in depressed patients, leading to decreased symptoms and fewer maladaptive behaviors (Simons et al. 2015; Snippe et al. 2016). Consistently, it is possible that monitoring daily affect and related emotion regulation strategies increases people’s awareness of experienced positive and negative emotions. Along this line, Sharot (2011) showed that individuals tend to update their predictions when provided with a piece of information that is positive rather than negative (i.e., when the information provides a more optimistic perspective on the prediction made). This updating process appears to produce optimism that is resistant to change. Similarly, providing a feedback about daily experienced emotions may lead habitual PA underestimators to update their predictions to more closely match the real affect experienced, thus reducing the PA bias.

Although this study deepens our knowledge about affect bias and its relationship with mental health, we acknowledge several limitations that could be addressed in future research. An important aspect that needs to be considered is content validity. Although the single items we used to assess PA and NA via EMA significantly correlated with and predicted affect scores, it would be necessary to replicate this finding in a larger population in order to guarantee content validity. In addition, the PANAS consists of 10 items to assess PA and 10 items to evaluate NA. Conversely, we used single items to collect momentary measures of PA and NA because of the daily, repeated nature of the ecological assessments. It is possible that a complex construct like affect might be grasped differently when using 10 items as opposed to one item. While acknowledging this, it is important to note that it would have been too demanding for users to complete the 20 items on the PANAS three times a day for 2 weeks. Similar to other studies (Suso-Ribera et al. 2018), we decided to use single items in order to manage the difficulties in obtaining high adherence rates when dealing with EMA research (Desirée Colombo et al. 2018). Additionally, the indirect effects of recall bias on mental well-being observed in this study should be considered with caution, considering the cross-sectional nature of data (Maxwell and Cole 2007). Further studies are needed to confirm the causal role of PA bias on Ryff's measures through resilience. Finally, further limitations are related to the nature of our sample, which was mainly female and composed of undergraduate students, and that also included individuals with mild depressive symptoms. An important goal for future research would include investigating whether affect recall is affected by sex and whether different patterns can be observed in relation to age. According to the “positivity effect”, indeed, old individuals as compared to younger ones are likely to recall more positive than negative information (Carstensen and DeLiema 2018; Reed and Carstensen 2012). It would be therefore important to explore whether the benefits of holding biased representations of past emotional experiences entail the same implications in different populations.