Optimism and cognitive functioning trajectories in a cohort of aging men

Abstract

Objectives

Robust evidence supports optimism as an asset for good physical and emotional health in aging populations, but its role in cognitive aging remains understudied. This study evaluated whether higher optimism levels would be prospectively associated with higher initial levels and slower decline in cognitive functioning over 26 years in a community-dwelling cohort of aging men.

Methods

Participants included 847 men from the Veterans Affairs Normative Aging Study who completed the Revised Optimism-Pessimism scale of the Minnesota Multiphasic Personality Inventory-2 in 1986 and ≥1 cognitive assessment repeated triennially in 1993–2019. At each assessment, scores from seven cognitive tests were combined into a global composite and three domain-specific composites: verbal memory, executive functioning, and visuospatial ability. Mixed-effects regression models evaluated the associations between optimism and cognitive trajectories.

Results

Higher optimism levels were associated with higher initial levels but not less decline in global cognitive functioning over time (B = 0.04, 95% CI: 0.001–00.07), , adjusted for demographics, practice effects, and lag between optimism assessment and the first cognitive assessment. In domain-specific analyses, optimism was associated with higher initial levels but not magnitude of decline in verbal memory (B = 0.06, 95% CI: 0.01–0.12), and unrelated to executive functioning or visuospatial ability trajectories.

Discussion

This study adds specificity to a nascent literature linking optimism to cognitive aging, indicating an association with initial levels, but not decline—particularly in verbal memory—in older men. Examining these relationships earlier in life may further clarify the etiologic role of optimism in cognitive health across the developmental span.

The global burden of dementia is rising at an unprecedented pace, and identifying modifiable factors that can prevent or delay the onset of cognitive impairment has never been more critical (Zissimopoulos et al., 2018). One such factor in the psychosocial realm is optimism, a habitually positive orientation typically defined either as an attributional style or a disposition (Scheier & Carver, 2018). Robust evidence supports optimism as an asset contributing to capacity to attain and maintain good physical health (Scheier et al., 2021), yet optimism has received little attention in the cognitive aging literature, wherein the quest for modifiable risk factors has primarily focused on education, physical health conditions, health behaviors, and among psychosocial factors, depression, and social isolation (Livingston et al., 2024). The current study addresses this gap by evaluating whether optimism may be associated with healthier trajectories of global and domain-specific cognitive functioning in a cohort of aging men.

Positive psychological well-being as (cognitive) health assets

There has been intense interest in identifying positive attributes or assets that can improve health beyond risk factors that accelerate deterioration. Positive psychological well-being (PPWB) refers to positive affective and cognitive evaluations individuals make regarding their lives (Kubzansky et al., 2015). PPWB factors are prime candidates for consideration as health assets because many are modifiable (Kubzansky et al., 2023) and their overall levels are often largely preserved from midlife to early old age, before age-related vulnerabilities set in. Lifespan theorists proffer that affective well-being improves with age as individuals become increasingly motivated to optimize emotional satisfaction as their time horizon shrinks (Socioemotional Selectivity Theory; Carstensen, 1992) and as individuals gain mastery in socioemotional competency (Strength and Vulnerability Integration; Charles, 2010). Such gains are thought to be maintained at least through early old age, before they may be overwhelmed by cognitive and physical declines and diminishing control over one’s environments (Charles, 2010; Labouvie-Vief et al., 2007). Indeed, across multiple large-scale investigations, optimism was largely stable from midlife into early old-age, with declines evident starting at 70 years and older (Tetzner et al., 2024), paralleling findings of favorable developmental trajectories for life satisfaction and negative affect across adulthood followed by declines in old age (Buecker et al., 2023). Moreover, optimism appears to be malleable in the context of life events and changes in health (Chopik et al., 2020; Purol & Chopik, 2021). Thus, optimism may be a viable health-promoting resource for many adults through early old age.

Among PPWB constructs, optimism has among the strongest evidence as a health asset, demonstrating prospective associations with diverse physical health outcomes, above and beyond the absence of negative psychological well-being (Lee et al., 2019; Rozanski et al., 2019; Scheier et al., 2021). However, few studies have investigated its potential role as a cognitive health asset.

Optimism and cognitive aging

A growing literature has examined PPWB in relation to cognitive aging, with positive affect, meaning in life, and life satisfaction being the most frequently examined PPWB constructs. Findings suggested that PPWB-cognition associations are not universal across well-being constructs (Bell, Singham, Saunders, Buckman et al., 2022; Bell, Singham, Saunders, John et al., 2022). Six longitudinal or prospective studies have examined optimism in relation to objectively measured cognitive performance in population-based samples that include older adults. Of these, three examined associations of optimism with cognitive functioning operationalized by continuous measures of performance as opposed to incident dementia or mild cognitive impairment. Of the three, only one (Oh et al., 2020) had sufficient (i.e., ≥3) occasions to separately evaluate the associations of optimism with initial levels and change in cognitive function. Two studies (Bhattacharyya & Molinari, 2023, 2024) used overlapping samples and considered two-point change over 10 years. Although the three studies examined different cognitive domains, together, they provided preliminary evidence for associations of higher optimism with more favorable levels of verbal memory performance and mental status, slower age-related decline in mental status (Oh et al., 2020), and less 10-year decline in executive function and episodic memory (Bhattacharyya & Molinari, 2023, 2024). There were mixed findings in studies predicting incident cognitive impairment or dementia (Gawronski et al., 2016; Sachs et al., 2022; Sutin et al., 2018), which may be due to between-study differences in gender, baseline age, follow-up duration, assessing optimism as a bipolar versus unipolar construct, and diagnostic thresholds. Thus, the evidence on optimism and the course of cognitive aging remains inconclusive, especially with respect to both initial levels and magnitude of change in cognitive performance over long follow-up periods.

Dispositional versus attributional style optimism

All six aforementioned studies on optimism and cognitive aging considered dispositional optimism, defined as having a set of relatively stable expectations that one is likely to achieve favorable outcomes across diverse life domains, and assessed with the Life Orientation Test or its revision (LOT and LOT-R; Scheier & Carver, 1985; Scheier et al., 1994). Attributional style optimism draws from the idea that how we explain past events determines how we predict the future. More optimistic people are thought to attribute past failures to causes that are external to oneself, transient, and situation-specific, and to explain positive events to internal, stable, and global causes (Scheier & Carver, 2018). Attributional style optimism has yet to be investigated in relation to cognitive aging. Considering both allows for triangulation of evidence, which can improve our nascent understanding of how facets of optimism relate to cognitive aging, and enhance the validity and credibility of findings.

Current study

The current study examined prospective associations of optimism with trajectories of cognitive performance in a well-characterized cohort of aging men. To address shortcomings in follow-up duration in the extant literature, we leveraged triennial cognitive assessment data spanning 26 years to characterize the course of cognitive functioning. Our primary hypothesis was that higher levels of attributional style optimism would be associated with higher levels and slower decline in global cognitive functioning over time. To maximize comparability with other studies and evaluate global vs. domain-specific associations, secondary analyses examined cognitive trajectories in three domains: executive functioning, verbal memory, and visuospatial ability. To correct for potential bias in estimating cognitive trajectories given repeated assessments, we evaluated practice effects. We accounted for a range of other relevant covariates suggested in prior work on psychosocial predictors of cognitive aging (Roberts et al., 2022) that could confound and/or mediate the associations of interest. In sensitivity analyses, we repeated the analyses using a measure of dispositional optimism. We posited that the two measures would exhibit similar relationships with cognitive outcomes. However, due to design considerations, including different assessment timing for dispositional and attributional style optimism, different time lag between each optimism assessment and first cognitive assessment, and different sample sizes, comparison of findings must be made cautiously.

Methods

Study design and sample

Participants are from the Normative Aging Study (NAS), a longitudinal study of normative and pathological aging processes founded at the Veterans Affairs (VA) Boston Outpatient Clinic. Between 1961 and 1970, NAS enrolled 2,280 men aged 21–81 who were free of major physical and mental illnesses and showed evidence of geographic stability. Over 95% were military veterans, although it was not an inclusion criterion. Participants underwent regular in-person visits involving biomedical exams and laboratory testing, and completed periodic surveys by mail. Attributional style optimism was assessed via mail survey in 1986 (82.4% response rate among eligible participants at the time), which serves as the baseline for our main analyses. Dispositional optimism was measured beginning in 1994 as part of psychosocial surveys administered concurrently with triennial in-person visits.

Between 1993 and 2019, cognitive assessments were incorporated as part of the triennial in-person visits. Trained study staff conducted testing in a quiet room in the research wing of VA Boston. From 935 men who completed the attributional style optimism assessment in 1986 and at least one occasion of cognitive testing, we excluded 17 men who were missing more than 5% of optimism data and 71 men who did not have a valid cognitive testing occasion (defined as having valid data on at least five of seven cognitive tests), yielding an analytic sample of 847. Over half (54.4%) of this sample participated in more than three cognitive testing occasions; 13% participated in more than five occasions. Because few men had more than six valid cognitive testing occasions (n = 45–53 across outcomes) and unstable trajectory estimates could result from sparse data coverage at the extremes of the temporal axis, for each outcome, we used data from up to the first five valid cognitive testing occasions per man. This resulted in an analytic sample comprising 2,456 cognitive observations across the 847 men, with an average cognitive follow-up duration of 7.7 years [standard deviation (SD) = 6.3, range = 0–25.6) (see Supplementary Figure S1 for a flow chart). The 88 excluded men were older and had a lower body mass index (BMI) than the analytic sample, but they did not differ on other demographics, health conditions, or health behaviors. In the subset of 77 men excluded due to missing cognitive data, relative to the main analytic sample, their optimism levels did not differ (Supplementary Table S1). Compared to the 1,433 NAS men who did not meet inclusion criteria for this study, our analytic sample (n = 847) was born slightly later (M_birthyear (SD): 1926 (7) vs. 1921 (10)), had more years of NAS follow-up (M_follow-up (SD): 39.1 (7.7) vs. 18.6 (13.0)), and had lower mortality rates (71% vs. 89% by December 31, 2020). The VA Boston Healthcare System Institutional Review Board approved the NAS protocol. Participants provided written informed consent.

Measures

Optimism

Attributional style optimism (henceforth “optimism”) was assessed in 1986 with the Revised Optimism-Pessimism scale (PSM-R; Malinchoc et al., 1995) of the Minnesota Multiphasic Personality Inventory-2 (MMPI-2; Butcher et al., 1989). Malinchoc et al. (1995) applied the Content Analysis of Verbatim Explanations technique to MMPI-2 items and identified 263 that were weighted to yield a bipolar score on a continuum ranging from optimistic to pessimistic. Following prior work using this data (Lee et al., 2019), we created maximum likelihood estimates of missing items based on available responses and set the total score to missing if more than 5% of items were missing (<2% of respondents). The PSM-R scores are originally on a T-score metric (i.e., M = 50, SD = 10) based on the original MMPI normative sample (Malinchoc et al., 1995), with T < 50 representing an optimistic explanatory style. For ease of interpretation, we reverse-coded and z-standardized PSM-R scores against the mean and SD of all NAS 1986 MMPI-2 respondents, such that each additional unit corresponds to one SD higher in optimism level. The optimism z-scores were further categorized into quartiles for descriptive purposes and to evaluate possible nonlinear associations with cognitive trajectories. The PSM-R had good internal consistency (Kuder–Richardson-20 = 0.86) and temporal stability over 5 years (r = .87, p < .0001) in NAS (Lee et al., 2019).

In sensitivity analyses, we considered associations of dispositional optimism measured by the LOT (Scheier & Carver, 1985) with cognitive trajectories. The LOT consists of two 4-item subscales, which assess unipolar optimism (e.g., “In uncertain times, I usually expect the best”) and unipolar pessimism (e.g., “If something can go wrong for me, it will”), with item scores ranging from 1 to 5. We computed the bipolar LOT total score by reverse-scoring pessimism items before summing across all items, such that higher scores represent greater optimism. We also scored LOT unipolar optimism and pessimism subscales by summing items in each subscale, such that higher scores represented higher optimism and pessimism, respectively. We used mean substitution for missing items, permitting up to one missing item each for unipolar optimism and pessimism subscales, and two missing items for the total (bipolar) score.

In a subsample (n = 486) of our analytic sample with both optimism measures administered within a 5-year period (PSM-R in 1991, LOT total score between 1994–1996), the correlation between the attributional style and dispositional optimism was r = 0.46, p < .0001.

Cognition

Cognitive functioning was assessed via seven cognitive tests: (a) the Mini Mental State Examination with the “county” item excluded (Folstein et al., 1975); (b) Immediate and (c) Delayed Recall from the Consortium to Establish a Registry for Alzheimer’s Disease 10-item word list learning (Morris et al., 1989); (d) WAIS-R Digit Span Backwards (Wechsler, 1955); (e) Animal Naming Test (Tombaugh et al., 1999); (f) Figure Copying items from Consortium to Establish a Registry for Alzheimer’s Disease and the Developmental Test of Visual Motor Integration (Beery, 1989) and (g) Pattern Comparison from the Neurobehavioral Evaluation System, Version 2 (Letz, 1991). Individual scores on each test were z-standardized against the analytic sample mean at each man’s first occasion.

Our primary outcome was global cognitive functioning, assessed by all seven aforementioned tests (Farooqui et al., 2017). Secondary analyses considered performance in three domains: verbal memory assessed by Immediate and Delayed Recall, executive functioning assessed by Digit Span Backwards and Animal Naming Test, and visuospatial ability assessed by Figure Copying and Pattern Comparison. Individual tests were z-standardized against the analytic sample mean at each man’s first testing occasion, and z-standardized test scores were averaged to create the global cognitive functioning composite and three domain-specific composites (see Supplementary Table S2, for details). For each occasion, the global composite was scored for men with valid data on ≥5 tests, and each domain-specific composite was scored for men with valid data on ≥1 test for that domain.

Covariates

Demographics were assessed by questionnaires and included fathers’ occupation, age (centered at a mean of 69.3 years at the first cognitive occasion), education (mean-centered at 14.2 years), and being married versus not. Father’s occupation was reported by respondents at the NAS entry. For age, marital status, educational attainment, health behaviors, and depressive symptoms, we used exam, chart-review, and questionnaire-based data from the triennial examination closest in time and within 1.5 years of the first cognitive testing occasion. Major chronic conditions were measured via medical chart review as a count score of cancer, cardiovascular disease, chronic obstructive pulmonary disease, and diabetes. Health behaviors, queried by staff and via questionnaires, included smoking status (never/former/current), alcohol consumption (none to moderate, former, or heavy/problematic), and BMI (kg/m²) calculated from height and weight measured by study staff. Past-month depressive symptoms were measured as continuous scores of the Brief Symptom Inventory-Depression subscale (Derogatis & Melisaratos, 1983).

Statistical analysis

We generated descriptive statistics and examined distributions of covariates by optimism quartiles using one-way analysis of variance and chi-square tests. Given the clustering of repeated cognitive assessments within participants, we used multilevel regression to test the primary hypothesis that higher levels of optimism would be associated with higher levels and less decline in global cognitive functioning over time. Secondary analyses considered each domain-specific composite as outcomes. All models were constructed with time since first cognitive assessment in years as the temporal metric and adjusted for age at first cognitive assessment. Intercepts reflect expected levels of cognitive functioning at age 69, the sample mean at the first cognitive assessment. Linear slopes represent the magnitude of cognitive change per year. Quadratic slopes represent nonlinear cognitive changes and are interpreted as the rate of cognitive change per year (acceleration/deceleration) at the intercept. Fixed effects reflect the sample average for a given parameter; random effects quantify individual differences in the parameter.

Analyses were conducted in four steps. First, we determined the best-fitting model of cognitive trajectories over age (henceforth: “best-fitting trajectory model”) by visually inspecting plots of raw cognitive scores, calculating intraclass correlation coefficients, and evaluating different forms of cognitive change over age: no change, linear change, and quadratic change. We estimated fixed and random effects of intercepts and slopes in all models. Please see Supplementary Methods and Supplementary Table S3 for further details.

The best-fitting trajectory model for each cognitive outcome was used as the basis for hypothesis testing. In a series of up to three models per outcome, we tested associations of optimism with cognitive functioning, adjusted for the time lag (years) between optimism in 1986 and each man’s first cognitive assessment. The first model specified a main effect of optimism indicating its association with initial levels of cognitive functioning (i.e., at first cognitive assessment for each man). The second model added the interaction of optimism*time, indicating its association with magnitude of change in cognitive functioning. If the best-fitting trajectory model included quadratic change, we tested a third model containing optimism × time². Beginning with a model with all significant interactions and their respective lower-order terms (Model 1), we added covariates incrementally, including demographics (Model 2), chronic health conditions (Model 3 [core model]), health behaviors (Model 4), and depression (Model 5). Covariate missingness was assumed to be missing at random and handled via multiple imputation with 30 imputed datasets.

We conducted three sets of sensitivity analyses. First, given observations of higher optimism among more highly educated and earlier-born cohorts (Boehm et al., 2015; Chopik et al., 2020), we explored whether education or birth cohort moderated the associations of interest. We specified interactions of optimism with birth cohort operationalized as three groups [<age 65 (reference), 65–70, ≥71 at first cognitive assessment] and with education operationalized as two groups [less than high school (reference) vs. high school or more].

Second, to allow more direct comparison with prior studies, we substituted the PSM-R with the LOT (total, optimism subscale, and pessimism subscale scores) and repeated the analyses. We then compared the overall pattern of results between the two sets of analyses. The LOT analyses used a subset of the main analytic sample who completed the LOT prior to their first cognitive assessment used in the main analysis (thus bypassing the need to remodel the cognitive trajectories as both sets of analyses used identical cognitive data for the included men). As the LOT administration began in 1994, one year after the initiation of cognitive assessments in NAS (in 1993), the LOT sample was smaller and excluded men whose first cognitive occasion preceded LOT administration. Within the main analytic sample, men included the LOT subsample (n = 566) compared to those excluded (n = 281) were older at the first cognitive occasion (M_age: 70.1 vs. 67.7) and had slightly higher number of chronic conditions (M_conditions: 1.2 vs. 1.0); the two groups did not differ on other demographics nor depression.

Finally, we considered nonlinear effects of attributional style optimism in the main analyses by substituting the continuous PSM-R z-score with dummy codes for quartiles.

Results

Sample characteristics

Table 1 includes characteristics of the overall sample and by optimism quartiles. At the first cognitive assessment, men were on average 69 years old (SD = 7.3, range = 51–98), had some college education, and had one chronic health condition. Most (76%) were married. More optimistic men had higher educational attainment, fewer chronic conditions and depressive symptoms, and less alcohol consumption.

Table 1.

Descriptive statistics of the analytic sample by optimism quartiles.

Variable	Full sample n = 847	Q1 n = 212	Q2 n = 212	Q3 n = 211	Q4 n = 212	F/χ2	df  a
	Mean (SD) or %	Mean (SD) or %	Mean (SD) or %	Mean (SD) or %	Mean (SD) or %
Demographics
Age (years)	69.3 (7.3)	69.02 (7.5)	69.34 (7.4)	69.29 (7.2)	69.49 (7.3)	0.2	(3,846)
Education (years)	14.2 (2.7)	13.80 (2.6)	14.26 (2.6)	14.16 (2.7)	14.70 (2.7)	4.1	(3,846)
Marital status (% married)	75.7%	72.6%	75.0%	78.7 %	76.4%	2.5	3
Paternal occupation						19.9	12
Unskilled	15.5%	18.9%	16.5%	17.1%	9.4%
Semi-skilled	24.1%	23.6%	20.3%	26.1%	26.4%
Skilled & foreman	34.2%	35.4%	32.6%	35.1%	34.0%
Semi-professional and white collar	14.9%	14.6%	18.4%	11.9%	14.6%
Professional	10.5%	6.6%	11.8%	9.0%	14.6%
Missing	0.8%	0.9%	0.5%	0.9%	0.9%
Health status
Major chronic conditions	1.2 (0.9)	1.28 (0.9)	1.17 (0.9)	1.16 (1.0)	1.02 (0.9)	3.1	(3,846)
Depressive symptoms	0.2 (0.4)	0.55 (0.6)	0.20 (0.4)	0.07 (0.2)	0.08 (0.2)	60.5	(3,764)
Health behaviors
Smoking status						5.6	6
Current smoker	11.8%	12.7%	11.3%	12.3%	10.9%
Former smoker	65.1%	65.1%	64.6%	69.2%	61.3%
Never smoker	23.1%	22.2%	24.1%	18.4%	27.8%
Alcohol consumption						23.6	6
Heavy or problematic	4.3%	4.7%	5.7%	3.3%	3.3%
Former	13.5%	22.6%	9.9%	10.9%	10.4%
None to moderate	71.1%	61.3%	73.6%	73.5%	75.9%
Missing	11.2%	11.3%	10.9%	12.3%	10.4%
Body mass index	27.9 (3.9)	28.51 (4.1)	27.83 (3.9)	27.60 (3.9)	27.61 (3.6)	2.6	(3,824)

Note. Age, education, marital status, health status, and health behaviors were assessed at first cognitive assessment using values closest in time and within 1.5 years. Paternal occupation (1 = unskilled, 2 = semi-skilled, 3 = skilled and foreman, 4 = white collar or semi-professional, 5 = professional, managerial, proprietary) was assessed at study enrollment. Major chronic conditions were assessed as a count score of cancer, cardiovascular disease, chronic obstructive pulmonary disease, and diabetes. Past-month depressive symptoms were measured as continuous scores of the 6-item Brief Symptom Inventory (possible scores range 0–4, observed scores ranged 0–3). Distributions shown are of non-imputed covariates and thus include missing as a response category, when applicable. In analyses, missingness in covariates was handled via multiple imputation.

^a df = degrees of freedom, shown as (df_Between, df_Within).

Optimism and global cognitive functioning trajectories

Of the variance in global cognitive functioning, 65% was attributable to between-person differences and 35% due to within-person variation, with the latter comprising systematic change over time and occasion-specific fluctuation, including measurement and random error (Supplementary Figure S2). Among candidate models of global cognitive functioning trajectories, a random linear model provided the best fit to the data (Supplementary Table S3; Supplementary Figure S3).

Optimism was unrelated to the magnitude of change in global cognition over time (optimism × time: B = −0.002, 95% CI −0.006 to 0.002). Thus, we used a main effect model as the basis for further covariate adjustment. Each SD higher in optimism levels was associated with 0.04 SD (95% CI 0.001–0.07; Table 2, Model 2) higher initial levels of global cognitive functioning, adjusted for age and time since optimism assessment (both at the first cognitive assessment), and demographics. Further adjusting for health conditions attenuated the effect slightly and rendered it marginally significant [B = 0.03, 95% CI −0.003 to 0.07; Model 3 (core model)]. The point estimate was unchanged but became less precise with a wider confidence band after further adjusting for health behaviors (B = 0.03, 95%CI −0.007 to 0.06, Model 4) and depressive symptoms (B = 0.03, 95% CI −0.01 to 0.07, Model 5). Figure 1 displays the mean global cognitive functioning trajectory over age by optimism levels (at the mean and +/− 1 SD). Supplementary Table S9 contains parameter estimates for covariates (excluded from the main table for brevity).

Table 2.

Associations of optimism with global cognitive functioning (n = 847, Obs = 2,456).

Variable	Model 1: Base model		Model 2: + demographics		Model 3: + health conditions		Model 4: + health behaviors		Model 5: + depression
	B	95% CI	B	95% CI	B	95% CI	B	95% CI	B	95% CI
Fixed effects
Intercept	0.12	−0.04, 0.27	0.01	−0.18, 0.20	0.05	−0.14, 0.24	0.04	−0.15, 0.24	0.05	−0.15, 0.24
Optimism	0.05	0.02, 0.09	0.04	0.001, 0.07	0.03	−0.003, 0.07	0.03	−0.01, 0.06	0.03	−0.01, 0.07
Linear time (yrs)	−0.04	−0.05, −0.04	−0.04	−0.05, −0.04	-0.04	−0.05, −0.04	−0.04	−0.05, −0.04	−0.04	−0.05, −0.04
Practice effect	−0.10	−0.14, −0.06	−0.10	−0.14, −0.06	-0.10	−0.14, −0.06	−0.10	−0.14, −0.06	−0.10	−0.14, −0.06
Age	−0.04	−0.05, −0.04	−0.04	−0.05, −0.04	-0.04	−0.05, −0.04	−0.04	−0.05, −0.04	−0.04	−0.05, −0.04
Random effects
Intercept	0.21	0.19, 0.24	0.19	0.17, 0.22	0.19	0.17, 0.22	0.19	0.16, 0.22	0.19	0.16, 0.22
Linear time (yrs)	0.001	0.0004, 0.001	0.001	0.0003, 0.001	0.001	0.0003, 0.001	0.001	0.0003, 0.001	0.001	0.0003, 0.001
Int-time cov	0.001	−0.001, 0.003	0.001	−0.001, 0.003	0.001	−0.001, 0.003	0.0003	−0.002, 0.002	0.0003	−0.002, 0.002
Residual	0.10	0.09, 0.11	0.10	0.09, 0.11	0.10	0.09, 0.11	0.10	0.09, 0.11	0.10	0.09, 0.11

Note. Int-time cov = covariance of intercept with linear time (Yrs); Obs = number of observations; yrs = years, since first cognitive testing occasion. Parameter estimates for covariates are shown in Supplementary Table S9. Model 1 adjusted for age (sample mean-centered at first cognitive assessment) and time lag in years between optimism assessment and first cognitive assessment. Model 2 added demographics; Model 3 added health conditions; Model 4 added health behaviors; Model 5 added depression. Bold: p < .05; Italics: .05 ≤ p < .10.

Figure 1.

Estimated mean cognitive trajectories over age, by optimism level, for global cognitive functioning and verbal memory. The plotted trajectories are based on Model 1 (see Tables 2 and 3 for estimates), and include main effects of optimism, time polynomials, age at first cognitive assessment (sample mean-centered at 69.3 years), a practice effect, and the time lag (years) between Malinchoc optimism assessment and first cognitive assessment. Plots reflect individuals with the mean-sample age at first cognitive assessment and the sample-mean time lag in years between their Malinchoc optimism and first cognitive assessments (9.76 years).

Domain-specific cognitive trajectories and associations with optimism

Nearly two-thirds (64%) of the variance in verbal memory was attributable to individual differences, whereas variance in visuospatial ability and executive functioning was split roughly equally between- and within-person (Supplementary Figure S2). A random linear model with practice effect best characterized trajectories of verbal memory and executive functioning, while a fixed quadratic model with random linear decline and no practice effect provided the best fit for visuospatial ability (Supplementary Table S4; Supplementary Figure S3).

Optimism was unrelated to the magnitude of change across domain-specific outcomes (i.e., optimism × time interactions ns, results not shown for parsimony); thus, as with global cognitive functioning, we present results for the main effects of optimism only. Higher optimism was associated with higher initial levels 0.06 SD (95% CI 0.01–0.12, Table 3, Model 3), but not the magnitude of decline, in verbal memory in the core model. This association was robust to further adjustment for health behaviors (Model 4) and depression (Model 5). Optimism was unrelated to trajectories of executive function and visuospatial ability (Table 3).

Table 3.

Associations of optimism with verbal memory (n = 847, Obs = 2,542), executive functioning (n = 847, Obs = 2,452), and visuospatial ability (n = 847, Obs = 2,403).

Variable	Model 1		Model 2		Model 3		Model 4		Model 5
	B	95% CI	B	95% CI	B	95% CI	B	95% CI	B	95% CI
Verbal memory
Fixed effects
Intercept	0.26	0.02, 0.49	0.27	−0.02, 0.56	0.28	−0.02, 0.58	0.23	−0.08, 0.54	0.21	−0.10, 0.52
Optimism	0.08	0.02, 0.13	0.06	0.01, 0.12	0.06	0.01, 0.12	0.06	0.01, 0.11	0.08	0.02, 0.14
Linear time (yrs)	−0.05	−0.05, −0.04	−0.05	−0.05, −0.04	−0.05	−0.05, −0.04	−0.05	−0.05, −0.04	−0.05	−0.05, −0.04
Practice effect	−0.07	−0.14, −0.004	−0.07	−0.14, −0.004	−0.07	−0.14, −0.004	−0.07	−0.14, −0.001	−0.07	−0.14, −0.001
Age	−0.05	−0.06, −0.04	−0.05	−0.06, −0.04	−0.05	−0.06, −0.04	−0.05	−0.06, −0.04	−0.05	−0.06, −0.04
Random effects
Intercept	0.24	0.19, 0.29	0.46	0.40, 0.53	0.46	0.40, 0.53	0.46	0.39, 0.52	0.45	0.39, 0.52
Linear time (yrs)	0.001	0.001, 0.002	0.001	0.001, 0.002	0.001	0.001, 0.002	0.001	0.001, 0.002	0.001	−0.002, 0.01
Int-time cov	0.01	0.001, 0.01	0.004	−0.001, 0.01	0.004	−0.001, 0.01	0.003	−0.002, 0.01	0.003	0.001, 0.002
Residual	0.32	0.30, 0.35	0.26	0.24, 0.28	0.26	0.24, 0.28	0.26	0.24, 0.28	0.26	0.24, 0.28
Executive functioning
Fixed effects
Intercept	0.09	−0.11, 0.28	0.02	−0.22, 0.25	0.06	−0.17, 0.30	0.01	−0.23, 0.26	0.02	−0.23, 0.27
Optimism	0.04	−0.01, 0.08	0.02	−0.02, 0.06	0.01	−0.03, 0.06	0.01	−0.03, 0.06	0.01	−0.04, 0.06
Linear time (yrs)	−0.03	−0.04, −0.03	−0.03	−0.04, −0.03	−0.03	−0.04, −0.03	−0.03	−0.04, −0.03	−0.03	−0.04, −0.03
Practice effect	−0.07	−0.13, −0.004	−0.07	−0.13, −0.01	−0.07	−0.13, −0.01	−0.07	−0.13, −0.005	−0.07	−0.13, −0.005
Age	−0.03	−0.04, −0.03	−0.03	−0.04, −0.03	−0.03	−0.04, −0.03	−0.03	−0.04, −0.03	−0.03	−0.04, −0.03
Random effects
Intercept	0.34	0.29, 0.39	0.31	0.26, 0.36	0.31	0.26, 0.36	0.31	0.26, 0.36	0.31	0.26, 0.36
Linear time (yrs)	0.001	0.0004, 0.001	0.001	0.0003, 0.001	0.001	0.0003, 0.001	0.001	0.0003, 0.001	0.001	0.0003, 0.001
Int-time cov	-0.01	−0.01, −0.003	−0.01	−0.01, −0.003	−0.01	−0.01, −0.003	−0.01	−0.01, −0.003	−0.01	−0.01, −0.003
Residual	0.24	0.22, 0.26	0.24	0.22, 0.26	0.24	0.22, 0.26	0.24	0.22, 0.26	0.24	0.22, 0.26
Visuospatial ability
Fixed effects
Intercept	0.11	−0.10, 0.31	−0.13	−0.37, 0.12	−0.06	−0.31, 0.19	0.02	−0.24, 0.28	0.04	−0.22, 0.30
Optimism	0.03	−0.01, 0.08	0.02	−0.03, 0.06	0.01	−0.04, 0.06	0.004	−0.04, 0.05	−0.01	−0.07, 0.04
Linear time (yr)	−0.02	−0.03, −0.003	−0.02	−0.03, 0.00	−0.02	−0.03, −0.003	−0.02	−0.04, −0.004	−0.02	−0.04, −0.004
Quad time (yrs2)	−0.002	−0.003, −0.001	−0.002	−0.003, −0.001	−0.002	−0.003, −0.001	−0.002	−0.003, −0.000	−0.002	−0.003, −0.0005
Age	−0.04	−0.05, −0.04	−0.04	−0.05, −0.04	−0.04	−0.05, −0.03	−0.04	−0.05, −0.03	−0.04	−0.05, −0.03
Random effects
Intercept	0.25	0.20, 0.30	0.25	0.20, 0.30	0.24	0.19, 0.29	0.24	0.19, 0.29	0.24	0.19, 0.29
Linear time (yrs)	0.001	0.001, 0.002	0.001	0.001, 0.002	0.001	0.001, 0.002	0.001	0.001, 0.002	0.001	0.001, 0.002
Int-time cov	0.01	0.001, 0.01	0.01	0.001, 0.01	0.01	0.001, 0.01	0.01	0.001, 0.01	0.01	0.001, 0.01
Residual	0.32	0.30, 0.35	0.32	0.30, 0.35	0.32	0.30, 0.35	0.32	0.30, 0.35	0.32	0.30, 0.35

Note. Int-time cov = covariance of intercept with linear time (yrs); Obs = number of observations; yrs = years, since first cognitive testing occasion; yrs² = years squared, since first cognitive testing occasion. Model 1 adjusted for time between optimism assessment and first cognitive assessment and mean-centered age at first cognitive assessment. Model 2 added demographics (mean-centered education, marital status, father’s occupation) to Model 1. Model 3 (core model) added health conditions to Model 2. Model 4 added health behaviors (smoking status, drinking status, BMI) to Model 3. Model 5 added depressive symptoms to Model 4. Bold: p < .05; Italics: .05 ≤ p 10.

Sensitivity analyses

In the first set of sensitivity analyses, across cognitive outcomes, there was no evidence that birth cohort (Supplementary Table S5) or education (Supplementary Table S6) modified the association of optimism with levels or change in cognitive functioning. For example, the association of optimism with verbal memory was not significantly different for men with high school or more years of education relative to those with lower educational attainment (B = −0.10, 95% CI −0.21 to 0.02).

When considering dispositional optimism operationalized separately as a bipolar dimension (LOT total score) and as unipolar constructs (LOT optimism and pessimism subscales), the overall pattern of findings was consistent with those reported earlier for attributional style optimism. For example, as expected, higher LOT total score was associated with higher initial levels of verbal memory (B = 0.07, 95% CI: 0.0004–0.13) but not its magnitude of change. The effect size estimates for the fixed effect of optimism were highly similar between the dispositional (LOT total score; Supplementary Table S7) and attributional style (PSM-R; Tables 2 and 3) definitions for all cognitive outcomes, although they were less precise (i.e., wider confidence intervals) for the LOT.

Finally, we evaluated potential nonlinear associations between optimism and cognitive trajectories by substituting continuous optimism scores with quartiles in the main analyses (Supplementary Table S8). For global cognitive functioning, associations with optimism quartiles were in the expected direction but not statistically significant in the core model. For verbal memory, being in the highest versus lowest optimism quartile was associated with better performance (B = −0.03, 95% CI −0.04 to −0.03). However, the overlapping confidence intervals across optimism quartiles did not allow us to differentiate between dose–response versus threshold effects. Optimism quartiles were not associated with the other cognitive domains. All optimism quartiles*time interactions were nonsignificant across outcomes.

Discussion

In a prospective cohort of community-dwelling older men, higher levels of dispositional optimism were associated with higher initial levels, but not decline, in global cognitive functioning over 26 years of follow-up. Domain-specific models indicated that the observed association was primarily driven by performance in verbal memory. Optimism was unrelated to executive functioning and visuospatial ability trajectories. The observed associations of optimism with verbal memory levels replicated across the attributional style and dispositional conceptualizations of optimism, suggesting each measure captures a relevant aspect of optimism for cognitive aging processes. Moreover, these associations were robust to adjustment for depression, consistent with a growing literature suggesting that the protective associations of PPWB with cognition are independent from negative aspects of psychological functioning (Willroth et al., 2023). Our findings contribute specificity regarding aspects of PPWB and their relations, or the lack thereof, with domain-specific levels and age-related change in cognitive functioning. Our findings have implications for the potential etiologic timing regarding when optimism might influence cognitive aging. Causal effects, if any, might operate in midlife or earlier (prior to our sample’s mean age at first cognitive assessment).

That optimism was associated with higher initial levels of memory performance is partially consistent with previously reported longitudinal findings on optimism and cognitive functioning in older adults. Our findings agree with Oh et al. (2020)’s observations of higher dispositional optimism with initial levels but not the magnitude of decline in verbal memory performance among older couples in the Health and Retirement Study. Our findings are inconsistent with findings showing associations of optimism with less residualized decline in episodic (verbal and visual) memory and executive function in the Midlife in the United States Study (Bhattacharyya & Molinari, 2023, 2024). However, this study assessed change across two assessments taken at a 10-year interval, and the sample was on average 14 years younger than ours at cognitive baseline, rendering their results less comparable to our trajectory-based outcomes.

We did not observe any association between optimism and cognitive decline across the cognitive domains assessed. Thus, the most consistent finding in this limited literature is a prospective association of higher optimism with higher levels, but not change, in verbal memory in older adults. These findings suggest that any protective effects of optimism on cognitive health may be domain-specific. An additional consideration is the timing of when and how optimism may influence cognitive trajectories. It may be that optimism influences cognitive development or reserve in earlier life by setting trajectories apart in midlife or earlier, before the beginning of our cognitive follow-up, with differences maintained over time following the initial separation of levels. Lending preliminary support to this idea, a Finnish national study observed associations of dispositional pessimism with visual memory as early as midlife (among 46-year-olds), though not earlier (among a separate sample of 26-year-olds) (Karhu et al., 2022). Additionally, as people age, processes contributing to cognitive and physical health declines may increasingly overpower the influence of positive characteristics, such as optimism, that confer benefits earlier in life (Purol & Chopik, 2021). If so, the association of optimism with cognitive decline may be weaker in studies with lengthy follow-up into old age, such as ours. Future research adopting a lifespan design is needed to clarify the developmental timing and mechanisms by which optimism may influence cognitive aging.

The association of optimism with better verbal memory in later life may offer clues about its potential benefits for cognitive aging. Adjustment for health conditions and health behaviors at the first cognitive assessment minimally attenuated the effect sizes, suggesting other pathways may underlie these associations. Considering psychosocial pathways, greater optimism has been linked to more frequent and intense experiences of positive emotions (Lee et al., 2022). Broaden and build theory posits that positive emotions can bolster cognitive and behavioral flexibility in our daily lives, which may in turn promote health and cognition through mechanisms such as preserving mental health, facilitating stress resilience, fostering social connections, and promoting a healthy lifestyle (Fredrickson et al., 2020). Neurocognitive studies have found that the positive expectancies and mood that accompany optimism enhance in-depth information processing, which ultimately facilitates short- and long-term recall (Tyng et al., 2017). Additional research assessing within-person associations of optimism, mood, and cognitive performance over different timescales will likely be fruitful.

Considering biological pathways, the “heart brain axis”—a network of physiological interactions between the heart and brain (Saeed et al., 2023)—may be a nexus between optimism and cognitive resilience. In addition to the well-established connections between cardiovascular and cognitive health (Livingston et al., 2024), optimism has been consistently associated with lower risk of incident cardiovascular events (Rozanski et al., 2019) and their upstream biobehavioral determinants (Boehm et al., 2018; Kubzansky et al., 2020).

Common causes of both optimism and cognitive functioning represent an alternative explanation for our findings. Although we adjusted for various sociodemographic, physical, and mental health factors, residual confounding from factors such as genetics and rearing environment is possible. Growing up in a nurturing and mentally enriching environment provides a strong foundation for cognitive development and learning (Rakesh et al., 2024) and is linked to better later-life cognitive functioning (Bell et al., 2024). Resource-rich environments may help cultivate an optimistic outlook through ample opportunities that promote effective problem-solving and self-efficacy, and providing safety nets (e.g., financial buffers) in the face of life stressors.

The association of optimism with higher levels of verbal memory was largely replicated across the attributional and dispositional definitions of optimism, despite differences in sample size, the time lag between each optimism measurement and first cognitive assessment, and follow-up duration in the two sets of analyses. These consistent findings across the two optimism measures are not unexpected given their moderate correlation (r = .46). The LOT-PSM-R correlation is lower than the cross-time correlation of the PSM-R (r = 0.87) over a similar 5-year period, supporting the view that they reflect shared but distinct facets of a broader optimism construct. We note that the effect size estimates of the LOT total score and the PSM-R were highly similar for intercepts of global cognition and verbal memory, but the confidence bands were wider for the LOT, likely due to a smaller analytic sample and fewer cognitive assessment occasions offering less precision. To further place the effect sizes into perspective: the associations of 1 − SD difference in attributional style (B_PSM-R = 0.06) and dispositional optimism (B_{LOT Total} = 0.07) with verbal memory levels was comparable, but stronger, than the associations of an additional year of education (B_edu = 0.05) or one fewer chronic condition (B_cond = −0.05; Supplementary Table S9, core models) with global cognitive functioning levels.

Some limitations should be considered. First, because our sample was on average 69 years old at the start of cognitive follow-up, we were unable to evaluate the optimism–cognition association in midlife or earlier. Second, we were unable to assess bidirectional relationships of optimism and cognition. It is possible that better cognitive performance lends confidence and a more optimistic outlook, especially during a life stage when concerns about cognitive decline are common. Future research may consider examining the temporal dynamics among optimism, objective, and subjective cognition over narrower timescales. Third, NAS men were selected during the 1960s for good health and men in our analytic sample further survived until at least the first cognitive assessment. To the extent that the more cognitively impaired men were less likely to enter the analytic sample and have greater attrition, the estimated association of optimism with initial levels and change in cognitive functioning would be biased downward. Fourth, although we had repeated assessments over a long follow-up, larger samples may be needed to capture associations of optimism with systematic differences in cognitive decline over age. Fifth, though our study is strengthened by our ability to consider both attributional style and dispositional optimism, we were unable to directly evaluate their unique and joint contributions to cognitive trajectories given differences in administration timing and samples. Factor analytic work to clarify their shared and unique facets will likely be fruitful in elucidating which aspects of optimism are most strongly related to cognitive aging. Finally, our findings have limited generalizability to women and non-White individuals. Replications across populations with more diverse sociodemographic characteristics and cultural backgrounds would enrich our knowledge of how and for whom optimism may benefit cognitive aging. Of note, our analyses suggested that our findings were consistent for men across the observed ranges of birth cohorts and education.

These limitations notwithstanding, this is only the second study to examine optimism in relation to longitudinal trajectories of cognitive functioning. We built upon Oh et al.’s (2020) study by considering change in cognitive functioning both globally and in three domains over a lengthier 26-year follow-up, while adjusting for several potential confounds suggested in prior work on psychosocial predictors of cognitive aging. Examining two conceptualizations of optimism allowed us to contribute new evidence on how different aspects of optimism relate to cognitive aging.

Optimism has shown substantial promise as an asset for cardiovascular health (Rozanski et al., 2019) and healthy aging (James et al., 2019). The current study suggests the health benefits of optimism may extend to cognitive aging, specifically in the verbal memory domain. The pattern of findings suggests negligible association of optimism with cognitive decline in late life. However, optimism may be setting individuals onto different cognitive trajectories earlier this life, although additional lifespan data are needed to assess this notion more rigorously. Our study answers a call by Willroth et al. (2023) for greater specificity in the growing literature on PPWB and cognitive health, by considering facets of optimism and their associations with domain-specific cognitive trajectories. Studies aimed at investigating the etiologic role of optimism in both short- and long-term cognitive change, including understanding when in the life span optimism may be most influential, will inform a growing arsenal of strategies that can effectively delay the onset of dementia.

Supplementary material

Supplementary material is available at The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences online.

Data availability

This study was not preregistered. Requests to access the study materials and VA-owned data can be submitted to A. Spiro (avron.spiro@va.gov) and will be considered on a case-by-case basis. Analysis scripts and output files will be made available upon request to the corresponding author.

Funding

This work was supported by grants from the National Institutes of Health (RF1-AG064006; K08-048221; R01-AA008941; R01-AG002287; and R01-AG018436); and a US Department of Veterans Affairs (VA) Merit Review and Research Career Scientist Award. The VA Normative Aging Study is a research component of the Massachusetts Veterans Epidemiology Research and Information Center (MAVERIC) and is supported by the VA Cooperative Studies Program/Epidemiological Research Centers. The views expressed in this article are those of the authors and do not necessarily represent the views or policies of the affiliated institutions.

Conflict of interest

None declared.