Explaining age differences in the memory-experience gap

Abstract

Emotions and symptoms are often overestimated in retrospective ratings, a phenomenon referred to as the “memory-experience gap”. Some research has shown that this gap is less pronounced among older compared to younger adults for self-reported negative affect, but it is not known whether these age differences are evident consistently across domains of well-being and why these age differences emerge. In this study, we examined age differences in the memory-experience gap for emotional (positive and negative affect), social (loneliness), and physical (pain, fatigue) well-being. We also tested four variables that could plausibly explain age differences in the gap: (1) episodic memory and executive functioning, (2) the age-related positivity effect, (3) variability of daily experiences, and (4) socially desirable responding. Adults (n = 477) from three age groups (21–44, 45–64, 65+ years old) participated in a 21-day diary study. Participants completed daily end-of-day ratings and retrospective ratings of the same constructs over different recall periods (3, 7, 14, 21 days). Results showed that, relative to young and middle-aged adults, older adults had a smaller memory-experience gap for negative affect and loneliness. Lower day-to-day variability partly explained why the gap was smaller for older adults. There was no evidence that the magnitude of the memory-experience gap for positive affect, pain or fatigue depended on age. We recommend that future research considers how variability in daily experiences can impact age differences in retrospective self-reports of well-being.

Introduction

The majority of studies on psychological aging rely on retrospective self-report measures that ask respondents to summarize their experiences over extended periods of time (e.g., a week or a month). A well-established finding is that compared to assessments involving momentary or daily diary ratings, recall measures produce higher ratings for emotions and symptoms, a phenomenon referred to as the “memory-experience gap” (e.g., Miron-Shatz, Stone, & Kahneman, 2009). It has been argued that this gap occurs because self-reports of immediate experiences and retrospective evaluations contain different yet complementary information. Whereas the former measure the ebb and flow of the “experiencing self” in daily life, the latter are tied to the “remembering self” that filters and consolidates experiences and shapes decision making and long-term planning (Conner & Barrett, 2012; Kahneman & Riis, 2005; Robinson & Clore, 2002a).

To date, several questions remain about the memory-experience gap as it relates to age. First, whereas studies on negative affect have shown the gap to be less pronounced among older compared to younger adults (e.g., Charles et al., 2016; Neubauer, Scott, Sliwinski, & Smyth, 2020; Ready, Weinberger, & Jones, 2007), findings for positive affect have been mixed (e.g., Ready et al., 2007; Röcke, Hoppmann, & Klumb, 2011), and it is not known if age differences in the memory-experience gap generalize to other domains of wellbeing. Second, it is unclear whether the length of the recall period contributes to age differences in the gap. Third, we do not know the mechanisms that may drive age differences in the memory-experience gap.

The present study aimed to answer these questions by examining age differences in the memory-experience gap for different domains of well-being and for recall periods of varying lengths. We also tested four variables that could plausibly explain age differences in the gap.

The memory-experience gap for positive and negative experiences

Most existing research has examined the memory-experience gap without taking age differences into account. The memory-experience gap has been shown to be especially robust for negative affect, which is rated at higher levels when recalled over longer periods compared to more immediate experiences (Ben-Zeev, Young, & Madsen, 2009; Lay, Gerstorf, Scott, Pauly, & Hoppmann, 2017; Mill, Realo, & Allik, 2016; Neubauer et al., 2020; Röcke et al., 2011). Research on physical symptoms, such as pain and fatigue, has observed the same pattern (Shiffman, Stone, & and Hufford, 2008; Stone, Broderick, Junghaenel, Schneider, & Schwartz, 2016; Walentynowicz, Schneider, & Stone, 2018). Findings for positive affect, however, have been equivocal. Although some research has found recalled positive affect levels to exceed those for more immediate experiences (Parkinson, Briner, Reynolds, & Totterdell, 1995), other studies did not consistently find a gap for positive affect (Lay et al., 2017; Neubauer et al., 2020; Walentynowicz et al., 2018) or found it to be smaller than for negative affect (Miron-Shatz et al., 2009).

Differences in the memory-experience gap by age

Socioemotional selectivity theory posits that due to decreasing time horizons, older adults selectively remember positive information more strongly and negative information less strongly to facilitate well-being (Carstensen, 1992; Carstensen, Fung, & Charles, 2003; Carstensen, Isaacowitz, & Charles, 1999). Accordingly, researchers have hypothesized that the memory-experience gap might decrease with age for negative and increase with higher age for positive experiences.

To date, studies examining age differences in the memory-experience gap have confirmed this hypothesis for negative affect: the gap is less pronounced among older compared to younger adults (e.g., Charles et al., 2016; Lay et al., 2017; Mill et al., 2016; Neubauer et al., 2020; Ready et al., 2007). This effect was found across different modalities (internet and paper-and pencil; Charles et al., 2016; Ready et al., 2007), participant samples (United States Americans and Estonians; Charles et al., 2016; Lay et al., 2017; Mill et al., 2016; Neubauer et al., 2020), and sampling protocols (ecological momentary assessment, daily diaries contrasted with recall ratings; Charles et al., 2016; Lay et al., 2017; Mill et al., 2016; Neubauer et al., 2020).

Empirical findings on age differences in the memory-experience gap for positive affect have been mixed. Whereas Ready et al. (2007) found a memory-experience gap for positive affect that was greater among older compared to younger adults, other studies found limited or no evidence of this age-related pattern (Lay et al., 2017; Neubauer et al., 2020; Röcke et al., 2011). There are several possible explanations for the inconsistent results for positive affect. It has been argued that the gap might not be the same for emotions of different valence, because cognitive heuristics play a larger role in people’s recall of negative compared to positive affect (Ganzach & Yaor, 2019). In addition, a substantial body of research suggests that older age is predominantly characterized by a reduction in negativity compared to an increase in positivity (e.g., Comblain, D’Argembeau, & Van der Linden, 2005; Knight, Maines, & Robinson, 2002; Mather et al., 2004; Rösler et al., 2005). To date, however, robust evidence on age-differences in the memory-experience gap for positive affect is outstanding, and we are not aware of any studies examining age-differences in the gap for physical symptom reports.

Differences in the memory-experience gap by length of the recall period

The length of the recall period may influence the magnitude of the memory-experience gap, which is practically important because reporting periods vary widely across common self-report measures used in aging research. Prior research has evidenced that the mean levels of self-report ratings near monotonically increase with longer reporting periods. This has been consistently demonstrated for positive and negative affect (Robinson & Clore, 2002b; Walentynowicz et al., 2018; Watson, Clark, & Tellegen, 1988) and symptom reports (Broderick et al., 2008; Houtveen & Oei, 2007; Stone, Broderick, Schwartz, & Schwarz, 2008).

Age differences in the memory-experience gap by recall period

Very little is known about whether age differences in the memory-experience gap depend on the recall period. Charles et al. (2016) examined age differences in negative affect across a month, a week, and a day as part of two measurement bursts. The results showed that age differences in the memory-experience gap increased with longer reporting periods (weekly compared to monthly recall). A similar pattern as found by Ready et al. (Ready et al., 2007, Study 1). These findings are in line with the model of strength and vulnerability integration (SAVI) (Charles, 2010), which posits that older adults are more likely to infuse their recollection of past negative experiences with positive experiences compared to younger adults, especially over longer time periods (Charles, 2010). However, more complex patterns are also theoretically plausible. Robinson and Clore (2002a) postulate that longer reporting periods lower the amount of detail that is accessible making respondents switch from an episodic retrieval strategy to a semantic one. If such a switch in retrieval strategies occurs “sooner” (i.e., involving relatively shorter recall periods) for older compared to younger people because of episodic memory limitations at older ages, this could yield age differences in the memory-experience gap that are already pronounced for relatively short reporting periods.

Variables that could explain age differences in the memory-experience gap

To date, we do not know the mechanisms through which these age differences emerge. In the present study, we examined four mediators that could potentially explain this relationship: (1) episodic memory and executive functioning, (2) age-related positivity effect (selective information processing), (3) the pattern (i.e., variability) of daily experiences, and (4) socially desirable responding.

Episodic memory and executive functioning

Cognitive decline is well-documented as a normal process of aging (McArdle, Fisher, & Kadlec, 2007; Salthouse & Ferrer-Caja, 2003; Wilson et al., 2002). Age-related reductions in episodic memory and in executive functioning may both impact age differences in the memory-experience gap, though for slightly different reasons. Completing recall ratings is a cognitively complex task that involves searching one’s memory for relevant information and integrating the retrieved details into a single summary judgment (Tourangeau, 1984). As memory functioning declines in older age, access to episodic details becomes more limited (Schwarz, 2006; Schwarz & Knäuper, 2012), such that older people may rely more strongly on their general beliefs or self-schemata in recall. Similarly, as executive functioning declines in older age, the task of generating accurate retrospective self-reports may become increasingly challenging and effortful. Individuals with difficulties in executive functioning (e.g., lower working memory capacity and cognitive flexibility) may instead choose a less demanding retrieval and response strategy (a phenomenon Krosnick (1991) called “weak survey satisficing”) and provide answers consistent with their general beliefs or views about themselves. Given that people’s self-schemata are often overly favorable (Taylor & Brown, 1988), an increasing reliance on these overly favorable self-schemata in recall could reduce the memory-experience gap at older ages for negative experiences and possibly enhance it for positive experiences.

Age-related positivity effect

People’s memories become more selective with age. As people grow older they show a preference for positive over negative information. This may be because they aim to maintain a positively-valenced emotional equilibrium (Carstensen & Mikels, 2005; Mather & Carstensen, 2005), because processing negative information is more cognitively challenging (Labouvie-Vief, Grühn, & Studer, 2010), or because of age-related neural degeneration that lessens emotional reactions to negative information (Cacioppo, Berntson, Bechara, Tranel, & Hawkley, 2011). This “aging-related positivity effect” has been well-documented to affect retrieval from memory (Reed, Chan, & Mikels, 2014). If older adults recall positive information and experiences more easily than negative experiences compared to younger adults, this could reduce the memory-experience gap for negative experiences and enhance it for positive experiences in older ages.

Variability of experiences

Another explanation is that changes in the patterns of daily experiences (rather than memory processes) are responsible for age differences in the memory-experience gap. The variability of a person’s experiences systematically relates to the degree to which people overestimate their experiences in recall; that is, the memory-experience gap is wider for people with more variable experiences (Kikuchi et al., 2006; Stone, Schwartz, Broderick, & Shiffman, 2005). This phenomenon has been viewed as an extension of the peak effect in recall (which posits that people disproportionately attend to peaks or exacerbations in retrospective summary ratings) in that lower intraindividual variability implies fewer and/or less distinctive peak experiences (Stone et al., 2005). Prior research has documented that compared to younger people, older adults are less variable in their day-to-day positive and negative affect (Carstensen, Pasupathi, Mayr, & Nesselroade, 2000; 2011; Röcke, Li, & Smith, 2009). Thus, declines in day-to-day variability of experiences (e.g., more emotional stability) from younger to older adulthood might account for a lower memory-experience gap for negative as well as positive experiences.

Social desirability

Finally, age differences in the memory-experience gap may be due to differences in people’s general willingness to report negative or undesirable thoughts and feelings (social desirability). Evidence indicates that individuals’ willingness to report poor psychosocial functioning is lower at older ages, and this effect may be exacerbated when participants feel they could be negatively evaluated as a person (Carstensen & Cone, 1983; Luong, Charles, Rook, Reynolds, & Gatz, 2015; Soubelet & Salthouse, 2011). Schwarz (2007) argued that social desirability may play a more pronounced role in retrospective summary ratings compared to reports of limited episodes, such that admitting that one had a “bad” day may be easier and less damaging to one’s general self-regard than admitting negative feelings that lasted over a longer period of time. Prior research has argued that this process might be especially important for older adults to counter common aging stereotypes of mental and physical deterioration (Fastame & Penna, 2012). This may, therefore, make them especially motivated to present themselves favorably in recall reports, which in turn may reduce the memory-experience gap for negative experiences and enhance it for positive experiences relative to younger people.

The present study

To achieve the study goals, participants from three age groups completed daily end-of-day ratings of their emotional (positive and negative affect), social (loneliness), and physical (pain, fatigue) well-being and completed retrospective ratings of the same constructs over a 3-, 7-, 14-, and 21-day period that overlapped with their daily diary ratings. Our hypotheses were: (1) Consistent with prior research on the memory-experience gap, we hypothesized that negative emotional, social, and physical aspects of well-being will be rated at higher levels in retrospective ratings compared to daily diaries across all age groups. (2) Older age will be associated with lower ratings of negative affect, loneliness, and physical symptoms and higher ratings of positive affect across all reporting periods. (3) The memory-experience gap will be smaller for older adults compared to younger ones for negative experiences. For positive affect, we expected that the gap should become larger for older adults based on the framework of socioemotional selectivity theory. (4) The magnitude of the memory-experience gap will differ depending on the length of the recall period, such that longer reporting periods yield a greater gap. (5) The magnitude of age differences in the memory-experience gap will differ depending on the length of the recall period. Given the limited prior evidence and lack of a clear theoretical expectation for the directionality of this effect, this hypothesis was exploratory. (6) Finally, for each of the putative mediators, we hypothesize that they will account for observed age differences in the memory-experience gap.

Methods

Recruitment

The study was approved by the University of Southern California’s Institutional Review Board (IRB). Participants were recruited through the national online opt-in internet panel Dynata. The panel consists of members agreeing to periodically participate in online surveys for minimal rewards. Dynata notified panelists of this research opportunity and, if interested, they were directed to a Qualtrics link that included information about the study’s purpose, the anticipated level of commitment, and eligibility questions. Panelists were eligible to participate if they (1) were 21 years or older; (2) were in the Central or Eastern time zone (because our online data collection system had limited programming capabilities for defining acceptable “windows” for the completion of the end-of-day daily diaries); (3) were fluent in English; (4) had access to a computer with high-speed internet at home; (5) had no difficulty reading a computer screen; (6) had no upcoming major events that would interfere with their daily assessments; (7) did not work a night shift; (8) and were willing to participate for 25 consecutive days. Interested and eligible panelists were asked to provide their contact information to the research team. Recruitment was stratified by age groups (21–44, 45–64, and 65 + years of age) and gender (50% male). Study invitations were sent to panelists in batches until approximately comparable sample sizes across strata were reached.

Procedures

Participants engaged in a 25-day daily diary study protocol consisting of baseline questionnaires, followed by an online aging-related positivity effect assessment, 21 end-of-day daily online questionnaires, recall assessments with intervals of 3, 7, 14, and 21 days of the daily assessment period, and a telephone-based cognitive (memory) assessment at the end of the study. All panelists provided electronic informed consent before engaging in the study. Communication with participants occurred online and over the telephone. During an introductory phone call, participants were trained to complete their assessments through Assessment Center (AC) (http://www.assessmentcenter.net/), a secure online data collection platform supported by the National Institutes of Health (NIH). After the training, participants completed the baseline questionnaire on the AC website followed by an aging-related positivity effect assessment via Qualtrics (participants also completed other measures not reported here).

Beginning the following day, participants completed daily diary questions each evening, between the hours of 6pm-12am midnight in the participants’ own time zone, as close to their bedtime as possible, for the next 25 consecutive evenings. The first four days were considered training days and were not included in the data analyses. Participants also completed one 3-, 7-, 14-, and 21-day recall assessment throughout the daily diary study period. Participants were randomly assigned to one of two arms that differed only in the day and order in which the 3-, 7-, and 14-day recall assessments were presented. Recall assessments were presented on days that allowed for a full coverage of corresponding daily diaries (e.g., the 14-day recall assessment was conducted mid-way through the study to be able to match it with 14 corresponding days of daily diaries).

Research staff checked participant compliance daily and contacted participants to troubleshoot if necessary. Participants could opt into daily reminders (emails and/or text messages) to facilitate assessment completion. Within a week of completing their final daily and recall assessment, participants completed a telephone-based episodic memory and executive functioning assessment.

Participants were compensated up to $150 with an Amazon gift card for completion of the study. Compensation was proportional to overall participation and timely completion of the daily assessments. Participants who fully completed all tasks were entered into a lottery in which one of every nine compliant participants were compensated with an additional $50 Amazon gift card.

Measures

Demographics

Participants completed questions about their age, gender, marital status, living situation, employment status, educational attainment, family income, ethnicity, and race.

End-of-day daily diaries and recall assessments

Daily diaries consisted of questions about participants’ emotional, social, and physical well-being experienced “since waking up today”. The recall assessments were 3-, 7-, 14-, and 21-day versions of the same daily items. On days with a recall assessment, participants completed the recall first and then the daily diary.

To generate a set of summary scores from daily diaries for comparison with the four recall reports, the daily diaries pertaining to the reporting period of each recall assessment (i.e., 3, 7, 14, and 21 days, respectively) were averaged for each participant and each outcome variable. Thus, there was a set of aggregated daily scores for the 3 days that corresponded to a participant’s 3-day recall period, and so on. In few cases, the recall assessment was completed a day late, in which case those diaries were aggregated that fell into the time period covered by when the recall assessment was actually completed.

Emotional well-being - positive and negative affect.

Positive and negative affect items were based on the circumplex model of affect (Russell, 1980). The positive affect measure consisted of six items, (“Since waking up today/In the past 3/7/14/21 days, I felt happy, excited, enthusiastic, calm, relaxed, and content). The response options ranged from 1 (Not at all) to 7 (Extremely). Cronbach alphas ranged from .96 to .97 for the aggregated daily scores and were all .94 for recall scores for each of the four reporting periods.

The negative affect measure consisted of six items (“Since waking up today/In the past 3/7/14/21 days, I felt distressed, frustrated, tense, bored, discontent, and dissatisfied”). The response options ranged from 1 (Not at all) to 7 (Extremely). Alphas ranged from .93 to .95 for aggregated daily scores and from .90 to .92 for recall scores across the four reporting periods.

Physical well-being – symptoms.

Fatigue and pain were assessed with measures from the NIH Patient-Reported Outcomes Measurement Information System (PROMIS®; Cella et al., 2010).

Fatigue.

The fatigue measure consisted of five items (“Since waking up today/In the past 3/7/14/21 days, how often did you find yourself getting tired easily, how often were you sluggish, how often were you energetic, how tired did you feel on average, and how energetic were you on average”) with five response options from “Never” to “Always” or “Not at all” to “Very much.” The Cronbach alphas were .92 for aggregated daily scores and ranged from .87 to .89 for recall scores.

Pain.

The pain measure consisted of a single item, as is common practice in clinical pain assessment (Jensen, Hu, Potts, & Gould, 2013). The item asked: “Since waking up today/In the past 3/7/14/21 days, how would you rate your pain on average?” and used response options on a scale of 0–10 with greater numbers indicating more pain.

Social well-being: loneliness.

The loneliness measure was based on the R-UCLA Loneliness Scale (Russell, Peplau, & Cutrona, 1980) and consisted of three items (“Since waking up today/In the past 3/7/14/21 days, how often did you feel that you lack companionship, did you feel left out, did you feel isolated from others”). The four response options ranged from “Never” to “Often.” Alphas ranged from .93 to .96 for aggregated daily scores and from .93 to .94 for recall scores.

Supplemental assessments of psychological distress.

Daily and recall measures of the psychological distress domains anxiety, anger, and depression were also assessed with PROMIS measures. Please see the online Supplemental Material for details and results for these measures.

Mediator assessments

Socially desirable responding.

The 16-item version of the Balanced Inventory of Desirable Responding (BIDR-16) was completed at baseline to measure social desirability (Hart, Ritchie, Hepper, & Gebauer, 2015). The BIDR-16 includes items, such as “I am a completely rational person”. Response options range from 1 (Not true) to 7 (Very true). Cronbach alpha was .84.

Aging-related positivity effect assessment.

Within four days after the baseline assessment, participants completed an online picture task through Qualtrics in which they were presented with picture stimuli taken from the International Affective Picture System (IAPS) (Lang, Bradley, & Cuthbert, 1997). The procedure for this task was closely modeled after the procedure from Experiment 2 in Barber et al. (2016).

Briefly, participants were presented with seven positively-valenced and seven negatively-valenced low-arousal pictures in random order, and were then asked to recall the content of as many pictures as they were able. A picture was scored as “recalled” if the rater was able to confidently match the picture with a description provided in the recall task. The relative positivity of participants’ recall was assessed as the number of net positive pictures recalled divided by the total number of pictures recalled, where the number of net positive pictures was the number of positive minus the number of negative images recalled (Barber et al., 2016). The scoring of the recall task was performed by two independent raters with an interrater reliability of kappa=0.95.

Variability in daily experiences.

Variability was assessed by calculating the intraindividual standard deviation of each participant’s daily diary outcome measures. The intraindividual standard deviation captures the magnitude (or amplitude) of deviations of a person’s daily scores around the person’s average experience level, and it has been commonly used to measure variability of emotions (Carstensen et al., 2000; Carstensen et al., 2011; Eid & Diener, 1999; Röcke et al., 2009) and physical symptom experiences such as pain (Kikuchi et al., 2006; Stone et al., 2005).

Telephone-based episodic memory and executive functioning assessment.

After participants completed their days of daily diary assessments, they participated in the Brief Test of Adult Cognition by Telephone (BTACT) (Lachman, Agrigoroaei, Tun, & Weaver, 2014), administered by trained research assistants. The BTACT consists of seven cognitive tests and takes about 25 minutes. Participants were asked to have as few distractions as possible and to not use paper or writing utensils. The test was audio-recorded with the permission of the participants. Participants’ audio recordings were reviewed during scoring to double check the responses. Following prior research, two summary measures were created from the component tests: (a) a measure of episodic memory (comprised of both immediate and delayed recall of 15 words; alpha = .89) and (b) a measure of executive functioning (comprised of number series [completing a number pattern], category verbal fluency [number of animals recalled within 60 seconds], digit span backward [highest string of numbers recalled in reverse order], and backward counting from 100 in 30 seconds; alpha = .56). Each of the two summary measures (episodic memory and executive functioning) was computed by first standardizing each test and then averaging those within each of the two domains (Karlamangla et al., 2014). The correlation between the two summary measures was r = .38, consistent with prior research (Lachman et al., 2014).

Analytic Strategy

Analyses to examine the memory-experience gap

To test the specific hypotheses about the memory experience-gap, factorial mixed model ANOVAs with two within-subject factors (random effects) and one between-subject factor (fixed effect) were estimated. The two within-subject factors were assessment type (recall versus corresponding aggregated daily diaries, 2 levels) and reporting period (3, 7, 14, and 21 days, 4 levels), and age group (3 levels) served as between-person factor. A significant main effect for assessment type indicates that the recall ratings differ from the aggregated daily diaries (i.e., supporting Hypothesis 1, the existence of a memory-experience gap). A significant main effect of age indicates that the age groups differ from each other in their average ratings for an outcome variable across the different types of assessment (Hypothesis 2). The age × assessment type interaction tests whether the memory-experience gap differs in magnitude across the three age groups (Hypothesis 3). A significant assessment type × reporting period interaction indicates that the memory-experience gap on average differs between assessments covering shorter versus longer time periods (Hypothesis 4). A significant three-way interaction (assessment type × age × reporting period) indicates that age differences in the memory-experience gap depend on the length of the reporting period (Hypothesis 5). ANOVA models were estimated separately for each outcome variable with multilevel models, with hypotheses tested in three steps: Step 1 entered the three main effects; Step 2 added all two-way interactions; and Step 3 added the three-way interaction. In each step, an omnibus Wald-test of overall group differences was conducted first; significant omnibus tests were followed up by post hoc pairwise comparisons to examine the nature of significant main and interaction effects. All models were estimated using Mplus version 8.4 (Muthén & Muthén, 2017) using maximum likelihood parameter estimation and a sandwich estimator (MLR) that yields standard errors and chi-square statistics robust to non-normality. Because multiple tests were conducted, to control for Type I error inflation, statistical significance was evaluated at a Bonferroni-corrected level of p < .01, adjusted for 5 outcome measures (p =.05/5).

Mediator analyses

The second set of analyses examined whether any of the putative mediators (episodic memory and executive functioning, variability in daily experiences, positivity effect, and social desirability) account for potential age differences in the memory-experience gap (Hypothesis 6). Before testing the mediation models, two sets of preliminary analyses were conducted (a) to examine whether the age groups differed in the proposed mediators (using one-way ANOVAs), and (b) to examine the correlations between the putative mediator variables and individual differences in the memory-experience gap. Even though significant effects in these bivariate analyses are not a prerequisite of significant indirect (mediated) effects, these analyses can provide insights into the magnitude of relationships between the independent variable (age), proposed mediators, and outcomes (memory-experience gap) (Shrout & Bolger, 2002).

Mediated effects were tested with structural equation (path analysis) models using Bayesian parameter estimation in Mplus. In the mediator models, the memory-experience gap was operationalized with discrepancy scores computed for each individual (calculated as recall scores minus averaged daily diaries, averaged across the four reporting periods), which served as continuous outcome variable. The three age groups served as categorical independent variable. Following recommendations (Hayes & Preacher, 2014), the age groups were coded using sequential coding, which yields indirect effects contrasting (a) the middle-aged versus the young adult group and (b) the older versus the middle-aged adult group, respectively. An omnibus Wald-test (Asparouhov & Muthén) with two degrees of freedom was employed first to examine evidence for overall significant indirect effects across the three age groups, followed by tests for specific indirect effects for pairwise comparisons of age groups. Statistical significance of specific indirect effects was determined using the product of coefficients method, with 95% credible intervals estimated from a minimum of 10,000 iterations in Bayesian analysis using non-informative priors. Like bootstrapping, this procedure appropriately takes the non-normal distribution of indirect effects into account (Muthén & Asparouhov, 2012). Indirect effects were considered significant at the corrected level of p < .01.

Secondary analyses examining correlations between recall and aggregated daily diary scores

An extra set of analyses examined whether the age groups differed in the rank-order correspondence (i.e., correlation) between recalled experiences and aggregated daily diary scores. Apart from differences in expected values (i.e., mean levels), how strongly ratings are correlated represents an additional measure of how closely recalled ratings reflect daily experiences. For each outcome variable, we computed the correlation between recall and aggregated daily diary ratings (pooled across reporting periods) in each age group. The (Fisher z-transformed) correlations were then compared between age groups using multigroup structural equation models (Cheung & Chan, 2005); omnibus Wald-tests were estimated first (in Mplus) to examine evidence for differences in the correlations between the 3 age groups, followed by pairwise comparisons of correlations. The R package MplusAutomation (Hallquist & Wiley, 2018) was used to facilitate and streamline the setup of Mplus models.

Results

Participants

A total of 495 participants completed the study. Compliance was high: participants completed an average of 20.00 (95.24%) out of 21 daily diaries (SD = 1.88 days, range = 2 to 21 days). Only 17 (0.009%) of 1,980 recall assessments were missed. Individuals who did not complete all four recall (n = 11) surveys were excluded from data analysis. In addition, we required completion of at least 66% of the diaries for each reporting period (i.e., at least 2 for the 3-day, 5 for the 7-day, 10 for the 14-day, and 14 for the 21-day reporting period, respectively). This resulted in a final analytic sample size of n = 477.

The demographics of the analytic sample are shown in Table 1. Participants’ mean age was 34.52 (SD = 5.13) in the young, 53.29 (SD = 5.49) in the middle-aged, and 70.68 (SD = 4.36) in the older age group. Participants were predominantly White (84.1%), 54.4% were women, 65.4% had a college degree, and the median family income was $50,000 to $74,999. The proportion of White participants was significantly smaller in young adults (χ² (2, n = 477) = 41.40, p <.05), and the proportion of female participants was significantly smaller in older adults (χ² (2, n = 476) = 17.65, p < .05), compared to the other groups. The age groups did not significantly differ in education and family income.

Table 1.

Demographic information of the overall sample and by age groups

	Overall	Young Adults (ages 21–44)	Middle-Aged Adults (ages 45–64)	Older Adults (age ≥65)
N	477	190	140	147
Age Mean (SD)	51.17 (15.96)	34.52 (5.13)	53.29 (5.49)	70.68 (4.36)
Gender (females)	54.41%	59.79%	62.14%	40.14%
Race
American Indian or Alaska Native	0.21%	0.53%	0.00%	0.00%
Asian	4.19%	9.47%	1.43%	0.00%
Black or African American	7.13%	12.11%	5.00%	2.72%
Native Hawaiian or Other Pacific Islander	0.21%	0.00%	0.71%	0.00%
White	84.07%	71.05%	90.00%	95.24%
Other or Mixed	4.19%	6.84%	2.86%	2.04%
Education (% with college degree)	65.41%	70.00%	57.86%	66.67%
Family Income (median)	$50,000-$74,999	$50,000-$74,999	$50,000-$74,999	$50,000-$74,999

Results of analyses examining the memory-experience gap

A summary of results from the factorial ANOVA models showing the omnibus tests for each outcome variable is provided in Table 2. Descriptive characteristics (means, SDs) of the outcome variables in each age group along with intercorrelations between the outcomes (for each recall and aggregated daily measure) are shown in the online supplement.

Table 2.

Results from omnibus Wald-tests in factorial ANOVA models for each outcome variable.

	Factors	DF	Omnibus Wald-test χ2
			Negative affect	Positive affect	Fatigue	Pain	Loneliness
Step 1	Assessment type	1	266.39***	1.97	362.69***	121.48***	159.90***
	Age group	2	62.11***	19.26***	29.46***	5.17	16.11***
	Reporting period	3	3.97	19.46***	0.75	2.71	2.34
Step 2	Assessment type × Age group	2	23.19***	3.59	1.37	4.39	18.89***
	Assessment type × Reporting period	3	16.99***	7.79	27.95***	7.17	25.08***
	Age group × Reporting period	6	5.75	5.68	8.49	6.54	5.42
Step 3	Assessment type × Age group × Reporting period	6	5.37	5.46	5.68	4.61	5.69

Note.

^***p < .001.

Hypothesis 1: Negative emotional, social, and physical aspects of well-being are rated at higher levels in retrospective ratings compared to daily diaries (main effects of assessment type).

Main effects of assessment type were evident for all outcome variables (ps < .01), except for positive affect (p = .16). Supporting Hypothesis 1, follow-up contrasts for significant outcomes showed that the average levels of the recall scores were significantly higher than those of the daily diary scores, consistent with a memory-experience gap. Standardized effect sizes (Cohen’s d) for the difference between recall scores and daily diaries were 0.26 (negative affect), 0.02 (positive affect, which was not significant), 0.24 (fatigue), 0.12 (pain intensity), and 0.22 (loneliness).

Hypothesis 2: Older age is associated with lower ratings of negative affect, fatigue, pain, and loneliness and higher ratings of positive affect (main effects of age).

Significant main effects of age were found for all outcomes (ps < .01), except for pain intensity (p = .075). Supporting Hypothesis 2, results from pairwise comparisons of age groups showed that older adults reported significantly lower levels of negative affect, fatigue, and loneliness, as well as higher levels of positive affect, compared to both younger and middle-aged adults, with effect sizes ranging from d = 0.33 to d = 0.67. Younger and middle-aged adults did not significantly differ from each other in their reported levels on any of the outcomes.

Hypothesis 3: The memory-experience gap will decrease with age for negative experiences and increase with age for positive affect (interaction effects between age group and assessment type).

The interaction between age group and assessment type was significant for 2 of the 5 outcomes: negative affect and loneliness (ps < .01). Consistent with Hypothesis 3, follow-up contrast analyses indicated that the difference between recall scores and daily diaries (i.e., the memory-experience gap) was significantly smaller for older adults than for younger and middle-aged adults on these outcome variables. The memory-experience gap did not significantly differ between younger and middle-aged adults. No significant age by assessment type interactions were found for positive affect and for the physical health outcomes fatigue and pain. Figure 1 shows the mean scores in each age group by assessment type (collapsed across reporting periods) for each outcome variable to illustrate the pattern of results.

Figure 1.

Memory-experience gap by age group across outcomes.

Hypothesis 4: The magnitude of the memory-experience gap differs depending on the length of the recall period (interaction effects between assessment type and reporting period).

As shown in Table 2, the assessment type by reporting period interactions were significant for negative affect, fatigue, and loneliness (ps < .01) but not for positive affect (p = .051) and pain (p = .07). Follow-up contrast analyses showed that the memory-experience gap in 3-day recall was significantly smaller than in all other reporting periods for negative affect, fatigue, and loneliness. The memory-experience gaps between 7-day, 14-day, and 21-day reporting periods did not significantly differ from each other for any outcome variable.

Hypothesis 5: The magnitude of age differences in the memory-experience gap differs depending on the length of the recall period (interaction effects between assessment type, age, and reporting periods).

Three-way interactions (assessment type × age group × reporting period) were not significant for any of the outcomes, suggesting that age differences in the memory-experience gap did not differ between reporting periods.

Given that the age groups differed in demographic characteristics (gender and race), supplementary ANCOVA analyses were conducted controlling for these demographic variables; the results for Hypotheses 1 through 5 did not change.

Hypothesis 6: Age differences in the memory-experience gap are mediated by episodic memory and executive functioning, the age-related positivity effect, variability in daily experience, and/or social desirability.

Age differences in the proposed mediators.

Table 3 shows descriptive statistics (means, SDs) for the putative mediators in each age group. Results from one-way ANOVAs indicated that the age groups significantly differed in executive function, social desirability, and in the magnitude of within-person variability on all outcomes (ps < .01). Specifically, older adults had lower executive functioning than middle-aged (d = .35) and young adults (d = .39), and scored significantly higher on social desirability than middle-aged (d = .42) and young adults (d = .57). Furthermore, older adults had lower within-person variability in the daily experiences on all outcome variables than middle-aged (ds ranging from .36 to .68) and young adults (ds from .47 to .75). Age differences in episodic memory and positivity effect test scores were not significant.

Table 3.

Means of mediators by age groups and ANOVA test results.

	Means			F Test
	Young Adults	Middle-Aged Adults	Older Adults	NumDF	DenDF	F
Episodic Memory	0.10a	0.08a	−0.17a	2	472	3.86
Executive Functioning	0.09a	0.05a	−0.18b	2	473	7.54***
Positivity Effect	0.07a	0.02a	0.07a	2	459	0.93
NA Variability	0.71a	0.69a	0.45b	2	474	21.87***
PA Variability	0.76a	0.69a	0.56b	2	474	15.07***
Fatigue Variability	6.21a	6.02a	4.97b	2	474	11.48***
Pain Variability	1.07a	0.97a	0.76b	2	474	9.41***
Loneliness Variability	0.31a	0.28a	0.14b	2	474	20.03***
Social Desirability	4.58a	4.71a	5.08b	2	474	13.48***

Note.

^**p < .01,

^***p < .001.

Means with different superscripts are significantly different from each other at p < .01.

Associations between the proposed mediators and the memory-experience gap.

Correlations between each proposed mediator and the memory-experience gap (operationalized as the difference between recall and aggregated daily scores, averaged across the reporting periods)¹ are shown in Table 4 for each outcome. Greater within-person variability of experiences was associated with a greater memory-experience gap for all outcomes (rs between .19 and .55, ps < .01), except for positive affect (r = .03, p = 51). Higher social desirability was associated with a smaller memory-experience gap for negative affect (r = −.20, p <.01) and loneliness (r = −.24, p < .01). Episodic memory, executive functioning, and positivity effect scores were not associated with the memory-experience gap for any outcome.

Table 4.

Correlations between putative mediators and memory-experience gap for each outcome variable

Memory-experience gap for well-being outcomes	Episodic memory	Executive functioning	Positivity effect	Variability	Social desirability
Negative affect	.11	.02	−.01	.35***	−.20***
Positive affect	−.04	−.07	−.01	.03	−.03
Fatigue	.01	−.01	−.03	.19***	−.02
Pain	.06	−.06	.07	.30***	.03
Loneliness	.04	.06	.06	.55***	−.24***

Note.

^***p < .001.

Mediated (indirect) effects.

As shown in Table 5, episodic memory, executive functioning, and positivity effect scores did not show significant indirect effects for any outcome variable in the mediation models. However, omnibus tests of indirect effects indicated that the within-person variability in daily experiences significantly mediated age differences in the memory-experience gap for all outcomes (ps <.01), except for positive affect (p = .98).² Tests of specific indirect effects for pairwise comparisons between age groups showed that variability in daily experiences mediated the difference in the memory-experience gap between the older and middle-aged groups for each outcome (see Table 5), explaining between 50.8% (negative affect) and 77.1% (loneliness) of the total effect of these age groups on the memory-experience gap. In addition, social desirability significantly mediated age-differences in the gap for loneliness (omnibus indirect effects test p < .01), with a significant specific indirect effect for the difference between the older and middle aged-groups that explained 23.2% of the total effect of these age groups.

Table 5:

Indirect effects of age groups on memory–experience gap via the putative mediator variables

		Omnibus test of indirect effects	Specific indirect effects
Dependent variable	Mediator variable	χ2 (df=2)	Middle-aged versus Young age groups Estimate (99% CI)	Older versus Middle-aged age groups Estimate (99% CI)
Negative affect
	Episodic memory	2.204	.000 (−.015; .012)	−.008 (−.031; .003)
	Executive functioning	0.063	.000 (−.007; .009)	.001 (−.017; .021)
	Variability	23.461***	−.009 (−.046; .026)	−.073 (−.125; −.032)
	Positivity	0.000	.000 (−.012; .012)	.000 (−.011; .012)
	Social desirability	8.728	−.009 (−.034; .009)	−.025 (−.060; −.004)
Positive affect
	Episodic memory	0.881	.000 (−.007; .010)	.004 (−.007; .020)
	Executive functioning	2.517	.001 (−.009; .014)	.009 (−.005; .030)
	Variability	0.037	.000 (−.013; .011)	−.001 (−.019; .016)
	Positivity	0.019	.000 (−.009; .011)	.000 (−.011; .010)
	Social desirability	0.110	.000 (−.012; .008)	−.002 (−.023; .016)
Fatigue
	Episodic memory	0.064	.000 (−.056; .053)	−.006 (−.129; .093)
	Executive functioning	0.059	.001 (−.050; .065)	.009 (−.123; .150)
	Variability	9.449**	−.038 (−.208; .114)	−.212 (−.471; −.043)
	Positivity	0.217	.007 (−.059; .107)	−.006 (−.108; .060)
	Social desirability	0.103	−.003 (−.097 .063)	−.015 (−.186; .136)
Pain
	Episodic memory	1.009	.000 (−.017; .014)	−.007 (−.037; .011)
	Executive functioning	2.302	.002 (−.014; .023)	.014 (−.009; .050)
	Variability	12.809**	−.023 (−.074; .023)	−.052 (−.113; −.003)
	Positivity	0.907	−.005 (−.030; .008)	.004 (−.009; .032)
	Social desirability	0.854	.003 (−.010; .023)	.009 (−.020; .045)
Loneliness
	Episodic memory	0.174	.000 (−.006; .005)	−.001 (−.013; .008)
	Executive functioning	0.501	.000 (−.008; .005)	−.003 (−.018; .008)
	Variability	32.422***	−.019 (−.061; .022)	−.076 (−.126; −.031)
	Positivity	0.736	−.002 (−.014; .004)	.002 (−.004; .014)
	Social desirability	11.833**	−.008 (−.028; .008)	−.023 (−.050 −.005)

Note:

^**p<.01;

^***p<.001. CI = Credible interval.

When within-person variability and social desirability were considered simultaneously in multiple mediator models to evaluate their unique indirect effects, variability in daily experiences remained a significant mediator for all outcomes (ps <.01; except positive affect, p = .98), whereas the indirect effect of social desirability for loneliness became non-significant.

The mediation results (significant indirect effects) did not change when age was treated as a continuous rather than a categorical predictor variable, and when gender and race differences between the age groups were statistically controlled.³

Secondary analyses: age differences in correlations between recall and aggregated daily scores.

An additional set of analyses compared the correspondence (i.e., correlations) between individuals’ recall and aggregated daily ratings across age groups. As shown in Table 6, all correlations were above .80, and the majority of correlations were above .85. Significant age differences in the correlations were evident for fatigue and pain; for both outcomes, the correlations between recalled and aggregated daily experiences were lower in the younger age group compared to the middle-aged and older adult groups (ps < .01).

Table 6:

Age-group differences in correlations between recall ratings and corresponding aggregated daily diaries (pooled across reporting periods)

	Correlations between recall ratings and aggregated diaries			Wald test for differences between correlations
	Young	Middle-aged	Older	χ2 (df=2)
Negative affect	.85 a	.88 a	.89 a	2.767
Positive affect	.90 a	.93 a	.93 a	4.805
Fatigue	.81 a	.89 b	.92b	19.739***
Pain	.90 a	.95 b	.96b	25.888***
Loneliness	.86 a	.89 a	.92 a	4.431

Note:

^***p<.001.

Means with different superscripts are significantly different from each other at p < .01.

Discussion

This study examined if the magnitude of the memory-experience gap differed by age, domain of well-being, and the length of the recall period (for the retrospective measures). We also examined potential mediators of age differences in the gap. Our first hypothesis that negative aspects of emotional, social, and physical well-being are rated at higher levels in retrospective ratings compared to daily diaries was largely confirmed. Consistent with prior research (Shiffman et al., 2008; Stone et al., 2016; Walentynowicz et al., 2018), we detected memory-experience gaps for all outcomes, except for positive affect. These results are consistent with the notion that negative experiences leave a stronger imprint in memory than positive ones, leading to their overestimation in recall (Miron-Shatz et al., 2009; Rozin & Royzman, 2001).

The second hypothesis that older age is associated with lower levels of negative experiences and higher levels of positive affect across all reporting periods was also supported. Ample research has documented that older adults frequently report lower levels of negative affect than their younger counterparts and that positive affect is maintained at high levels despite physical and cognitive declines that might suggest deterioration of well-being (for review see Carstensen et al., 2000; 2011; Charles, Reynolds, & Gatz, 2001; Stone, Schwartz, Broderick, & Deaton, 2010). We found that this “well-being paradox of aging” (Carstensen et al., 2011; Charles & Carstensen, 2010) was evident not only for affect but for loneliness, and fatigue. The exception was that there was no main effect of age on pain intensity. Instead, the pattern of results was curvilinear and showed that pain levels were low in the younger and also low in the older age group, which is contrary to what one would expect with increasing physical ailments in older adults.

The third and central hypothesis was that the memory-experience gap will decrease with age for negative experiences and increase with age for positive affect. This was partly confirmed. We found that the gap was significantly larger for younger and middle-aged adults compared with older adults for negative socio-emotional experiences (i.e., negative affect and loneliness). However, we did not find an effect for physical symptoms, which raises questions about the generalizability of age differences in the gap. It could be that sensory-based experiences, such as pain and fatigue, do not evoke age differences. Experiences that involve concrete, physical sensations might be more easily remembered than less tactile socio-emotional experiences. We encourage future research to examine age differences in the gap in other domains that are central to healthy aging, such as sleep, diet, and physical activity. For positive affect, we expected a larger gap for older adults based on socioemotional selectivity theory (Carstensen, 1992; Carstensen et al., 1999). This hypothesis was not confirmed in line with prior research that has either failed to observe this effect or produced inconsistent results. Our non-significant finding suggests that reconstructive processes responsible for the memory-experience gap may not impact people’s recall of positive affective experiences, and that this remains relatively stable across the adult life span.

Our fourth and fifth hypotheses were that the magnitude of the memory-experience gap and age differences in the gap differ by the length of the recall period. These hypotheses were confirmed in a limited way. The gap was smallest for the 3-day compared to 7-, 14-, and 21-day recall, but the gap in the latter three recall periods did not differ significantly from one another. This finding is in line with our prior research comparing ecological momentary assessments (EMA) with retrospective ratings over 1, 3, 7, and 28 days; differences between aggregated EMA ratings and recall ratings did not systematically increase with longer reporting periods beyond 3 days (Broderick et al., 2008). We furthermore did not find that age differences in the gap varied across the reporting periods. This result contrasts with prior work by Charles et al. (2016) who found larger age differences in the gap for monthly compared to weekly recall. One possible explanation for this discrepancy lies in methodological differences between the two studies. In our study, retrospective and daily diary ratings covered the same time periods. In contrast, participants in Charles et al.’s (2016) study completed a week of daily diaries that did not match with the days covered in the monthly recall.

Finally, we examined possible mediators of age differences in the gap. The first set of possible mediators, episodic memory and executive functioning, did not significantly mediate age differences in the gap. Our measure of executive functioning had a low internal consistency reliability, which may have reduced statistical power to detect effects that were truly present. Perhaps the mechanisms at work here are also more complex. While lower memory functioning may limit access to episodic details and reduce tendencies to overestimate experiences in retrospective reports (Schwarz, 2006; Schwarz & Knäuper, 2012), lower memory functioning may at the same time increase the use of cognitive heuristics that are associated with a larger memory-experience gap (Stone et al., 2005), and these two opposing effects might cancel each other out. One can also speculate whether our memory assessment, the BTACT which measures episodic memory functioning over the span of several minutes, contributed to this finding. Our longest recall assessment spanned 21 days, which is considerably longer. Other memory measures assessing longer-term recall, such as the Autobiographical Memory Test (Van Vreeswijk & De Wilde, 2004; Williams & Broadbent, 1986), might yield different results.

Our second mediator, the age-related positivity effect, was neither related to age nor to the memory-experience gap. This was surprising given that research has shown that older adults demonstrate a preference for positive over negative information across a broad range of experimental paradigms, laboratory tasks, and stimuli (for review see Reed et al., 2014). Prior studies have argued that experimental or environmental constraints that disrupt older adults’ real-world information processing can greatly reduce the positivity effect (Reed et al., 2014). Older adults in our study might have experienced extraneous cognitive constraints during the task (e.g., completing the assessment while being distracted) that impacted their selective information processing. However, we cannot be sure of this given the remote administration of the task. Future studies utilizing web-based and non-laboratory designs might want to assess the presence of potentially competing demands at the time of the assessment.

The third mediator, variability in daily experiences, showed the predicted associations with age (less variability at older ages) and with the memory-experience gap (less variability was associated with a smaller gap); it also significantly mediated their relationship. A well-established finding is that older adults experience more stability and less day-to-day variability in affect compared to younger people. This has been attributed to developmental processes involving more stable environments, better emotion regulation, and shifts in one’s outlook on life (Carstensen et al., 2000; 2011; Röcke et al., 2009). Our findings suggest that these factors may contribute to age differences in the gap for negative experiences, which in turn can make age differences in recalled negative affect appear larger than they actually are.

Our mediation results on within-person variability in daily experiences have actionable implications. Although our analyses were cross-sectional and cannot provide strong support for causal mechanisms (Lindenberger, Von Oertzen, Ghisletta, & Hertzog, 2011), the results provide preliminary evidence that reducing the impact of variability on recall reports could result in a more complete picture of psychological aging across the adult life span. Strategies could be implemented to mitigate the impact of differences in variability on recall, for example, through the use of techniques that help respondents reinstantiate the full reporting period and minimize reliance on memory heuristics (as in the event history calendar method; Belli, Shay, & Stafford, 2001).

The fourth mediator, social desirability, showed significant associations with age and the memory-experience gap for negative affect and loneliness, and it was a significant mediator of their relationship for loneliness. It has previously been argued that social desirability may play a stronger role in retrospective ratings than in reports of brief (e.g., daily) episodes, especially for negative or undesirable feelings and experiences (Schwarz, 2007). Our finding that the greater tendency of older people to present themselves favorably may be especially pronounced in their recall of loneliness is in line with this speculation. Loneliness is often viewed as stigmatizing, especially in older ages (Hauge & Kirkevold, 2010). Our mediation results for social desirability should, however, be viewed with caution. The indirect effect was only evident for one of the five outcomes it became non-significant after controlling for within-person variability.

This study has several limitations. First, we recruited participants from the U.S. general population, yet the final sample consisted of predominantly White and well-educated respondents. This is often the case for samples drawn from Internet survey panels (e.g., Hays, Liu, & Kapteyn, 2015). Second, given the gray digital divide in older adults (Morris, 2007) the participants in our older age group may not have been representative of older adults in the general population who are not part of Internet survey panels. We encourage future studies to examine whether our findings generalize to more demographically diverse samples. Third, the study did not implement the “gold standard” of assessing real-time experience, ecological momentary assessment. Instead, we contrasted daily diaries with retrospective recall. Daily diaries involve recall over a day and could thus still be impacted by biases and heuristics. Completing an end-of-day diary is not equivalent to rating one’s experiences in the moment (Neubauer et al., 2020) and researchers should be mindful of this. At the same time, while they should not be viewed as substitutes for one another, prior research has shown that momentary and end-of-day assessments demonstrate moderate-to-high correspondence, even for the assessment of several complex emotion dynamics (Broderick, Schwartz, Schneider, & Stone, 2009; Schneider, Junghaenel, Gutsche, Mak, & Stone, 2020). Fourth, as noted by Lindenberger et al. (2011) and by Maxwell and Cole (2007) cross-sectional models can offer only limited insight into developmental mechanisms underlying age-related changes. Our cross-sectional research design precludes statements about changes of well-being with age over time and associated changes in the memory-experience gap. Finally, we did not consider the full range of potential mediators of age differences in the gap. For instance, younger adults might tend to exaggerate their negative experiences in retrospect because they are less able to downregulate their negative emotions. Alternatively, negative experiences might be particularly useful for learning, in line with the developmental goals of younger adulthood. Examining these mechanisms might be fruitful avenues for future research.

In sum, the present study demonstrated age differences in the memory-experience gap for several self-report well-being outcomes. We found that the effect was limited to negative socio-emotional experiences and that it was not present for positive affect and physical symptoms. As such, our results suggest that future research should consider not only the valence but also the type of wellbeing domain for the study of individual and age-related differences in the gap. Furthermore, we recommend that future research considers how variability in daily experience can impact age differences in retrospective self-reports of well-being.