Age moderates within-person associations between momentary affect and cognitive performance in daily life

Abstract

Objectives:

This paper examined how momentary negative affect (NA) and positive affect (PA) are linked to ambulatory cognitive performance and whether these associations vary by age. Although prior research has identified between-person associations linking affect and cognitive performance, few ecological momentary assessment studies have investigated within-person relationships across adulthood.

Methods:

Data were drawn from the Effects of Stress on Cognitive Aging, Physiology, and Emotion (ESCAPE) study. Participants were 260 racially and economically diverse adults aged 25–65 (Mage = 46.5) who completed five daily assessments over 14 days. At each assessment, participants reported their current affect and completed brief, standardized cognitive tasks via smartphone. Multilevel models examined within-person associations between NA, PA and ambulatory cognitive performance (processing speed, working memory, spatial memory) and tested age as a moderator.

Results:

Across all ages, moments of heightened NA (compared to individuals’ averages) were related to worse working memory (γ_NA = −0.0004, SE = 0.0002, p = .02). Moments of heightened NA were related to decreases in processing speed (γ_NA = −0.0001, SE = 0.0000, p = .02) but only among adults around the sample mean age (46 years) and older. In contrast, moments of heightened PA were associated with decreases in working memory (γ_PA = −0.0003, SE = 0.0002, p = .04) only among adults under the sample mean age. No significant associations were observed between momentary affect and spatial memory.

Conclusion:

These results indicate that within-person associations between affect and cognition in daily life vary as a function of age-related contextual and behavioral factors. These findings suggest potential clinical implications for cognition, particularly during periods of elevated NA in later midlife.

Research consistently shows that individuals who report high levels of negative affect (NA) perform worse on laboratory-based cognitive tasks, including slower processing speed and poorer working memory (Forgas, 2008; Mogle et al., 2019; Payne and Schnapp, 2014; Verhagen et al., 2019). Findings for positive affect (PA) are less consistent: some studies link PA to better cognitive performance (Hittner et al., 2020; Yang et al., 2013), whereas others report null effects (Garrison and Schmeichel, 2022; Nealis et al., 2016) or even impaired working memory (Martin and Kerns, 2011). Importantly, these studies compare individuals at the between-subjects level and rely on laboratory-based settings, which cannot capture how affect and cognition fluctuate among individuals over time within social and environmental contexts. For example, the connection between affect and cognitive performance at any given time might vary based on situational context and the specific type of cognitive task.

Ecological momentary assessment (EMA) addresses these limitations by capturing real-time within-person processes in daily life (Luo et al., 2023; Scott et al., 2015; von Stumm, 2016; Wang et al., 2019). This approach reveals how daily fluctuations in affect shape cognition in natural settings (Cerino et al., 2021; Röcke and Brose, 2013), insights often missed by trait-level surveys or controlled lab tasks. Daily affect and cognition dynamics may become increasingly consequential across adulthood as age-related changes in cognitive resources and emotion regulation processes emerge, potentially altering how affective experiences relate to cognitive performance (Charles, 2010; Charles and Luong, 2013). The present study addresses this gap by testing within-person links between momentary NA, PA, and ambulatory cognitive performance across multiple tasks, and by evaluating whether these links differ by age.

1. Framework for affective wellbeing and cognitive performance

Self-reported affect and laboratory-based cognitive performance share a well-established connection, such that affective states (i.e., NA and PA) are interdependent with cognitive processes (Hogan et al., 2013; LeMoult and Gotlib, 2019; Matthews and Campbell, 2010; Storbeck and Clore, 2007). One framework that describes these connections is the cognitive model of depression, which posits a cyclical relationship between depressive symptoms (e.g., NA) and cognitive interference (LeMoult and Gotlib, 2019). In this model, negative psychological states influence self-referential thoughts, attention, and memory, thereby impairing cognitive control and diminishing the capacity to process new information effectively. Based on this theoretical framework, moments characterized by higher-than-typical NA for a given individual could interfere with attention and cognitive control during the completion of ambulatory cognitive tasks, resulting in poorer performance.

The relationship between PA and cognitive performance is less well understood. The broaden-and-build theory of positive emotions suggests that PA may support broader attentional scope, flexible thinking, and adaptive cognitive processing (Fredrickson, 2004). From this perspective, individuals experiencing higher-than-usual PA may have greater cognitive resources available for task engagement, which could support better performance (Kiken and Fredrickson, 2017). However, other findings linking PA to cognitive function suggest that when heightened PA is coupled with greater arousal (e.g., excited), this could increase cognitive flexibility but also reduce attentional control (Dreisbach and Goschke, 2004). Experimental work manipulating PA and arousal found that poorer cognitive outcomes depended on factors including arousal level and the specific cognitive domain assessed (Nealis et al., 2016). Together, these studies indicate that the associations between PA and cognitive performance may vary across contexts and types of cognitive domains, highlighting the need to examine these relationships in daily life.

2. Age differences in affect and cognitive performance

There are age-related patterns in both affect and cognition. Beginning around midlife, PA tends to remain stable or increase, whereas NA generally declines (Charles and Carstensen, 2010). The strength and vulnerability integration (SAVI) model proposes that as individuals reach older adulthood, they become increasingly physiologically and cognitively vulnerable in the face of heightened or sustained negative emotional experiences (Charles and Piazza, 2024). At the same time, later midlife and older adulthood is often accompanied by greater emotional stability and an increased tendency to prioritize emotionally meaningful experiences (Charles and Luong, 2013). Although average levels of NA tend to decline in older adulthood, when older adults experience heightened NA they may experience a greater demand on their regulatory and cognitive resources. For example, when people had to inhibit irrelevant information in an emotional Stroop task, cognitive performance was more disrupted by high-arousal stimuli among older adults (M = 74 years) than younger adults (M = 22 years; (Wurm et al., 2004). In another study testing whether older age was related to greater distractibility, older adults (ages 60-86; M = 65) performed worse than young adults (ages 19-32; M = 25) on a concentration task (n-back) when the content included emotionally arousing material compared to neutral material (Gallant et al., 2020). In contrast, low-arousal PA (such as feeling calm or content) may reflect lower emotional demands, allowing an individual to maintain regulatory and motivational resources to support engagement in everyday cognitive tasks (Gasper and Hackenbracht, 2014). Together, these mixed findings suggest that high-arousal affective experiences may not carry the same cognitive implications across adulthood. These perspectives also provide a theoretical basis and motivation to examine age-related variation in within-person affect-cognition associations from young adulthood through later midlife.

A growing body of literature suggests that cognitive performance is not static across day or moments but rather fluctuates alongside changes in individuals’ everyday psychological experiences (Sliwinski et al., 2006). Furthermore, the nature of these within-person associations has been shown to differ across age groups. For example, laboratory studies have shown that induced affective arousal differentially impacts cognitive performance in younger versus older adults (Chow et al., 2007; Pupillo et al., 2021). Moving beyond the laboratory, daily diary and ecological momentary assessment (EMA) studies have revealed that daily stressors and psychological arousal can negatively impact cognitive performance, often with the strength of these associations varying by age (Joshi et al., 2023; Riediger et al., 2014).

It should be noted that prior research examining age differences in affect and cognitive performance has largely relied on comparisons between younger and older adults. Consequently, less is known about how associations between momentary affect and cognitive performance vary across midlife. Further, much of the existing ambulatory research has relied on subjective self-reports of perceived cognitive performance (e.g., Joshi et al., 2023) or focused primarily on reactivity to specific daily stressors rather than general fluctuations in daily affect. By modeling across a wide range of ages (25–65), the present study will examine whether these associations change gradually rather than assuming discrete age-group differences. Further, the use of high-frequency, validated ambulatory cognitive assessments enables greater ecological validity for objective cognitive measurements in naturalistic settings (Sliwinski et al., 2018). More broadly, this study moves beyond asking whether affect is associated with cognitive performance to examining when, in whom, and across which cognitive domains these associations emerge in everyday life.

3. Daily affect and ambulatory cognitive performance

Within-person affect can be captured through EMA methods such as daily diary or smartphone prompts (i.e., “how happy are you feeling right now?”) with relative ease and low participant burden (Mogle et al., 2019). Daily variations in affect can be important indicators of overall wellbeing, which may also extend into cognitive health (Rickenbach et al., 2019; Stawski et al., 2019). However, utilizing EMA to assess both affect and cognitive performance is a less common but emerging area of research (Scott et al., 2015). Ambulatory cognitive assessments, delivered on smart-phones multiple times per day across days, are a relatively new type of approach within EMA methods that provide researchers with a more accessible assessment of cognitive performance (Hakun et al., 2024; Thompson et al., 2022). This type of ambulatory assessment can capture several different cognitive domains using tasks of working memory, visual/spatial memory, processing speed, and attention (Sliwinski et al., 2018). Assessing cognition at multiple time points across the day can provide a more reliable and consistent measure of objective cognition (Goga et al., 2025; Hakun et al., 2023). Completing a cognitive assessment in an individual's own contextual setting provides researchers with a more ecologically valid result compared to a lab-based or traditional paper-and-pencil cognitive battery (Crawford et al., 2022). Importantly, fluctuations in processing speed, working memory, and spatial memory may have implications for everyday functioning, including decision-making, problem-solving, navigation, and the completion of work or household tasks (Weizenbaum et al., 2020).

Recent work using daily diary and EMA designs demonstrate that affective states fluctuate significantly within persons over time and are tied to cognitive performance in daily life (Brose et al., 2014; Stawski et al., 2019). Prior literature assessing the coupling of affect and cognition often spans a wide range of ages, which provides useful insights into differences across age cohorts. However, the potential moderating effects of age within the daily dynamics of affect and cognition are not known (Stawski et al., 2011; von Stumm, 2016). Evidence connecting NA and the cognitive domain of working memory is fairly robust, but the association has not always been consistent. For example, in a study of adults aged 20 to 31, participants showed poorer working memory performance on days when reported NA was higher than their average level, as assessed by a daily in-person measure (Brose et al., 2012). Among college-aged students, lower working memory capacity was likewise associated with reporting more moments of NA than their average level which may interfere with higher-order cognitive functions such as problem-solving (Garrison and Schmeichel, 2022).

Yet, when day-to-day associations were examined between NA and self-perceived memory functioning among older adults (aged 60+), there was no significant relationship (Stawski et al., 2013). In another sample of older adults (mean age of 80 years) that included a daily diary NA assessment and a response time inconsistency task (a proxy measure for cognitive efficiency), cognitive efficiency was significantly poorer on days when NA was higher than individuals’ usual levels (Cerino et al., 2021). Differences across findings may be linked to the inconsistency across studies in their methods and cognitive domain of focus (i.e., subjective versus objective), but also to sociodemographic factors like age. These studies underscore the importance of examining sources of within-person variability, as affective states shift throughout the day and may dynamically influence cognitive performance (Röcke and Brose, 2013; Rocke et al., 2009). Collectively, these findings suggest that NA is often associated with poorer cognitive performance, but that the strength and direction of this association may depend on measurement approach, cognitive domain, and characteristics of the sample, including age. Together, these findings identify a gap in research that examines affect-cognition associations across adulthood using objective and ambulatory cognitive measures and explicitly considers potential age-related differences in these relationships.

Research investigating within-person PA and cognitive performance has yielded mixed results. Some daily diary and EMA studies suggest that PA may offer short-term cognitive benefits, consistent with the broaden-and-build framework (Fredrickson, 2004). In another daily diary study by Brose and colleagues (Brose et al., 2014), young adults between 20 and 31 years were found to have better working memory performance on days when they experienced PA above their own average. Such findings show that PA may enhance cognitive capacity, thereby supporting improved cognitive performance. However, other studies using daily diary assessments found no associations between reports of PA and cognitive performance among younger adults (Guevarra et al., 2024). In contrast, examinations of day-to-day affect and memory among older adults with an average age of 67 found no associations between PA and tasks assessing working memory, short-term memory, or processing speed (von Stumm, 2016). These inconsistent findings demonstrate the need for further investigation into daily PA dynamics. Null results may reflect age-related differences in the daily salience of affect on cognitive processes, or PA may relate to cognitive performance only through cumulative or indirect effects that are not captured by within-person frameworks.

Overall, the literature based on objective ambulatory cognitive assessments suggests that within-person affect-cognition associations are plausible but heterogeneous, underscoring the need for further investigation using such measures. Emerging evidence further suggests that psychosocial and affective experiences may not relate uniformly across cognitive domains, with some ambulatory cognitive tasks appearing more sensitive than others (Kang et al., 2024, 2025). These findings highlight the importance of examining processing speed, working memory, and spatial memory separately and assessing whether associations between affect and cognition vary across adulthood.

4. Current investigation

The present paper examined how momentary reports of NA and PA relate to performance on ambulatory cognitive tasks of processing speed, spatial memory, and working memory in a sample of adults aged 25 to 65. Given evidence that affective experiences may differentially relate to specific cognitive domains, these outcomes were examined separately. Our first aim was to assess whether within-person fluctuations in affective states throughout the day correspond to changes in concurrent ambulatory cognitive performance. We hypothesized that moments of heightened NA would be associated with worse performance on ambulatory processing speed, working memory, and spatial memory tasks. In contrast, we expected moments of heightened PA were expected to be associated with better performance on these same tasks, consistent with the broaden-and-build theory and prior work linking positive affect to adaptive cognitive functioning. Our second aim was to assess whether age moderates within-person associations. Drawing on developmental theory and prior findings suggesting that affective experiences may have different cognitive implications across adulthood, we hypothesized that increasing age would strengthen the negative association between NA and cognitive performance. Based on the mixed results across prior literature, we tested whether the within-person association between PA and cognitive performance varied across adulthood without specifying a directional hypothesis. As an exploratory analysis, we examined whether within-person associations between PA and cognitive performance were non-linear.

5. Method and materials

5.1. Sample

The present study used data from the baseline wave of The Effects of Stress on Cognitive Aging, Physiology, and Emotion (ESCAPE) Project. ESCAPE was approved by the ethical review board of the Albert Einstein College of Medicine of Yeshiva University (Scott et al., 2015). Systematic probability sampling was used for recruitment from Co-Op City, a Bronx, New York housing co-operative of approximately 60,000 residents, resulting in a racially and economically diverse population that is representative of the greater sociodemographics of the Bronx, New York. Participant eligibility for recruitment into the ESCAPE study also included: 25-65 years of age, English-speaking, ambulatory, and able to operate the study smartphone (e.g., free of visual impairments that would interfere with participation). For additional details on recruitment and methodology of the broader ESCAPE study, see Scott and colleagues (2015). The sample for this paper consisted of 260 adults with an average age of 46.49 years (SD = 11.09), 65% were female (n = 172), a majority identifying as Black or African American (n = 167, 63%), and 70% having at least some college education. Participants in this sample reported an annual income of $40,000 (Range: $20,000 - $60,000). Half of the sample were employed (n = 132, 50%) and the majority of the sample were not partnered (n = 178, 67%).

5.2. Procedures

Participants first attended a baseline lab visit, during which they provided information on demographics, physical health, and mental health. Following a training phase, they began a 14-day EMA burst using study-assigned smartphones (see Scott et al., 2015 for full ESCAPE methods). During this period, participants received five quasi-random beeps per day across their waking hours, prompting the completion of real-time EMA surveys assessing affective states, situations, and experiences. The quasi-random timing helped prevent anticipation of beeps. Immediately following each psychosocial EMA survey, participants completed ambulatory cognitive tasks. The median completion rate for prompted smartphone surveys and cognitive tasks was 83%. Across participants, up to 17,024 cognitive task observations were possible. Participants completed 15,696 processing speed assessments, 15,750 spatial memory assessments, and 14,127 working memory assessments, corresponding to completion rates of 92.2%, 92.5%, and 83.0%, respectively. Missingness was higher for working memory than for the other cognitive tasks, which was administered last in the fixed cognitive task sequence. Participants were eligible for inclusion in EMA analyses if they provided at least one valid EMA assessment (affective or cognitive) per study day.

5.3. Measures

5.3.1. Demographics

Demographic information collected included participants’ age, gender (male/female), race and ethnicity, education, income, partner status, and work status. For analytical purposes, racial and ethnic identity were collapsed into three categories: 1) non-Hispanic Black (63.46%), 2) Other/Minoritized (including those identifying as Hispanic-White, Hispanic-Black, Asian, and Other; 28.08%), and 3) non-Hispanic White (8.46%). Education was assessed on a five-level scale from “less than grade school” to a “graduate or professional degree”. Those reporting grade school or less (0.5%) were combined with those reporting some high school (5.4%). Over half of the participants (56.5%) had at least some college education and 26% earned a college degree. Income (average annual) was reported based on eight categories (<$5000; $5000-19,999; $20,000-39,999; $40,000-59,999; $60,000-79,999; $80,000-99,999; $100,000-149,999; ≥ $150,000). Responses marked “decline to answer” (n = 22, 8.5%) were coded as missing. Work status was dichotomized as “working” (employed, 50.51%) or “not working” (retired, 14.09%; unemployed and looking, 27.18%; unemployed and not looking, 8.21%). Marital status was recoded into a dichotomous variable called partner status including: “partnered” (married, remarried, or cohabiting; 40.71%) and “not partnered” (divorced, separated, never married, widowed, or other; 59.29%).

5.3.2. Affect

Momentary affect was captured via smartphone, prior to cognitive assessments, in each EMA survey. Five times throughout the day, participants rated momentary NA and PA using alternating positive and negative affect descriptors selected from established affect literature (Diener and Emmons, 1984) and adapted for the parent study ESCAPE (see Scott et al., 2015). NA items included tense, depressed, angry, unhappy, and frustrated. PA items included happy, enjoyment/fun, joyful, and pleased. Participants were prompted on their screens to rate how well each of the various affective descriptors described how they felt “right now” using a sliding bar on the screen ranging from 0 (not at all) to 100 (extremely). Momentary scores were calculated by averaging the individual NA and PA items.

We report the reliability estimates for these measures published by Kang and colleagues (2026), who used the same affect items and scoring from the parent ESCAPE dataset (Scott et al., 2015). Using the reliability-for-change approach described by Cranford et al. (2006), between-person reliability was excellent for both PA (0.99) and NA (0.99), and within-person reliability was also strong (PA = 0.89; NA = 0.84), indicating that the measures reliably distinguished individuals from one another and captured meaningful within-person fluctuations across time. Although PA and NA were assessed using four and five items, respectively, both scales demonstrated strong within-person reliability, suggesting that differences in scale length were unlikely to account for the observed pattern of findings.

5.3.3. Ambulatory cognitive assessment

At each beep, cognitive performance was assessed using the study smartphone. Figs. 1-3 provide screenshots from each ambulatory cognitive task. Participants completed three ambulatory cognitive tasks assessing processing speed, spatial memory, and working memory in a fixed order (Scott et al., 2015). Together, these tasks capture distinct aspects of daily cognitive functioning. The ambulatory cognitive tasks were adapted from established cognitive paradigms and have demonstrated reliability, construct validity, and criterion validity in prior mobile cognitive assessment research (e.g., Sliwinski et al., 2018; Cerino et al., 2021a,b,c; Hakun et al., 2024). These tasks were specifically developed for repeated administration in naturalistic settings.

Fig. 1.

Processing Speed Task: Symbol Search Task

Note. An example of a trial from the symbol search task. The task measures processing speed by asking participants to decide as quickly as possible which of the two pairs at the bottom of the screen matches a symbol pair at the top of the screen.

Fig. 3.

Working Memory Task: Modified N-Back Paradigm

Note. An example of the N-back task completed by participants. In the practice phase (a) participants decide if the two cards match while both are facing up. Next, cards are shifted from right to left and the new target cards now face down (b). Participants then go through retrieval and updating (c) where they determine if the new target card (facing up) matches the facedown card that was presented two phases prior.

Processing Speed.

Each assessment included a total of 16 trials in which participants examine three symbol pairs on the top of the smartphone screen and compare them to the two symbol pairs located at the bottom of the screen (see Fig. 1). Processing speed reflects the efficiency with which individuals rapidly perceive, evaluate, and respond to information. In this task, participants decide as quickly as possible which pair of symbols at the bottom match up with any of the three symbol pairs on the top of the screen. Processing speed is scored using ‘throughput’, a composite response measure calculated by dividing the accuracy rate (correct response per minute) by average reaction time (Thorne, 2006). Higher scores represent better processing speed performance.

Spatial memory.

Spatial memory reflects the ability to encode, retain, and retrieve visuospatial information over a brief period of time. To assess spatial memory, a location dot memory task has participants memorize the location for three red dots (see Fig. 3). The three red dots are presented on a 5x5 grid for 3 s, then a visual distractor exercise is completed for 8 s. Afterwards, participants are asked to recall the location of the three red dots. This task was performed across two trials each assessment. To score the dot memory task, Euclidean distance is calculated based on the distance of the location of the incorrect dot compared to the correct grid location. Higher scores indicate less accurate placement of the dot and worse spatial memory performance.

Working Memory.

Working memory reflects the ability to actively maintain and update information over short periods while performing ongoing cognitive operations. Participants completed a modified version of the N-back paradigm using a 2-back condition to assess working memory (see Fig. 2; Scott et al., 2015; Verhagen et al., 2019). Three standard playing cards in boxes first appear face-up on the phone screen. Participants determine if the first and third cards match, and then the cards are turned face-down. The leftmost card then disappears, as the other two cards move one box to the left. A new card is then presented face-up in the right most spot. Finally, participants decide whether the cards in the first and third boxes match. A total of 12 trials are completed for each assessment. Scores for the N-back tasks were calculated using throughput, a composite measure of the proportion of correct responses per minute with average reaction time (Thorne, 2006). Higher scores represent better working memory performance.

Fig. 2.

Spatial Memory Task: Dot Memory Task

Note. An example of the dot memory task for spatial memory. Participants are shown three red dots for 3 s (a), then engage in a distraction phase (b), and the retrieval phase (c) then asks participants to recall where the three dots were located.

6. Analytic plan

To test our hypotheses, we used 3-level multilevel models (beeps nested within days nested within participants) in SAS version 9.6 (PROC MIXED) with restricted maximum likelihood (REML) estimation. Separate models were analyzed for each of the three cognitive task outcomes (processing speed, working memory, and spatial memory), with NA and PA included simultaneously in each model. Statistical significance was determined using a two-tailed test with an α of 0.05. Level 1 (beep; within-person) affect predictors and other time-varying continuous covariates were person-mean centered to capture momentary deviations from an individual's typical affective level. To separate within- and between-person effects, person-level mean NA and PA scores (calculated across the EMA burst) were included as Level 3 (between-person) and grand-mean centered. Age was modeled as a continuous Level 3 moderator through cross-level interaction terms to examine whether within-person associations between affect and cognitive performance varied across adulthood. Each model adjusted for the linear effects of beep number and study day, as well as between-person differences in average NA, average PA, gender, race and ethnicity, income, education level, work status, partner status, and age. Models were estimated using REML, which incorporates all available observations and allows participants with intermittent missing EMA or cognitive task assessments to contribute all available data to model estimation.

$$\begin{gathered} Level1(beep):{\text{CognitivePerformance}}_{ijk}={\pi }_{0ji}(\text{Intercept}) \\ +{\delta }_{1ji}\left({\text{Beep}}_{ijk}\right){\delta }_{2ji}(NA-{\text{NAmn}}_{ijk})+{\delta }_{3ji}(\text{PA}-{\text{PAmn}}_{ijk})+i_{ik} \\ Level2(day):{\pi }_{0ji}={\beta }_{00\mathrm{i}}+{\beta }_{01\mathrm{i}}\left({\text{Day}}_{ji}\right)+{\mathrm{r}}_{0ji} \\ {\delta }_{2ji}={\beta }_{20\mathrm{i}} \\ {\delta }_{3ji}={\beta }_{30\mathrm{i}} \\ Level3(person):{\beta }_{00\mathrm{i}}={\gamma }_{000}+{\gamma }_{001}({\text{Gender}}_i)+{\gamma }_{002}(\text{Race}\&{\text{Ethnicity}}_i) \\ +{\gamma }_{003}({\text{Income}}_i)+{\gamma }_{004}({\text{Education}}_i)+{\gamma }_{005}({\text{WorkStatus}}_i) \\ +{\gamma }_{006}({\text{PartnerStatus}}_i)+{\gamma }_{007}({\text{Age}}_i)+{\gamma }_{008}({\text{Namn}}_i) \\ +{\gamma }_{009}({\text{PAmn}}_i)+{\mathrm{u}}_{00i} \\ {\beta }_{20\mathrm{i}}={\gamma }_{200}+{\gamma }_{201}({\text{Age}}_i) \\ {\beta }_{30\mathrm{i}}={\gamma }_{300}+{\gamma }_{301}({\text{Age}}_i) \end{gathered}$$

Cognitive performance for person i at day j on beep k was regressed on level-1 linear trend for beeps within study days (δ1ji), momentary NA (δ2ji), momentary PA (δ3ji), residual variance (i_lk), level-2 linear trend for days across assessment period (β_01i), and level-3 between-person differences in gender (γ₀₀₁), race and ethnicity (γ₀₀₂), income (γ₀₀₃), education (γ₀₀₄), work status (γ₀₀₅), partner status (γ₀₀₆), age (γ₀₀₇), average NA (γ₀₀₈), and average PA (γ₀₀₉). To test for age moderation of momentary affect-cognitive performance associations, age*momentary NA (γ₂₀₁) and age*momentary PA (γ₃₀₁) interaction terms were also included. As an exploratory analysis, we also examined whether associations between momentary PA and cognitive performance were better characterized by nonlinear effects by including a quadratic within-person PA term in each model.

7. Results

7.1. Descriptive statistics and correlation

Descriptive statistics and between-person bivariate correlations among study variables are presented in Table 1. Higher average NA was significantly correlated with poorer working memory (r = −0.14, p < .001) and spatial memory performance (r = −0.21, p < .001). Higher average PA was significantly correlated with poorer processing speed performance (r = −0.14, p < .001). In our sample, age was not significantly correlated with average NA (r = −0.05) but was positively correlated with PA (r = 0.14, p < .001).

Table 1.

Descriptives and correlations.

Variable	n	M (SD) % or [range]	1	2	3	4	5	6	7	8	9	10	11	12
1. Age	260	46.49 (11.09) [25 - 65]	1
2. Sex	260		0.02	1
Male	91	35
Female	169	65
3. Race & Ethnicitya	260		−0.12	−0.08	1
Black	165	63.46%
White	22	8.46%
Other/Minoritized	73	28.08%
4. Incomeb	236	3.83 (1.69) [1 - 8]	0.08	0.08	−0.11	1
5. Educationc	260	3.3 (1.14) [1 – 5]	0.09	−0.01	−0.15	0.34	1
6. Work Status	158		−0.17	−0.08	0.00	0.28	0.15	1
Working	130	50.4%
Not Working	128	49.6%
7. Partner Status	259		0.18	0.13	0.08	0.45	0.12	0.01	1
Yes	81	32.1%
No	178	67.2%
8. Average Negative Affectd	260	23.07 (15.85) [0 – 100]	−0.05	−0.04	0.10	−0.19	0.04	−0.02	0.00	1
9. Average Positive Affectd	260	61.18 (18.40) [0 – 100]	0.14	0.08	−0.03	0.04	−0.18	−0.07	−0.02	−0.50	1
10. Average Processing Speed (Symbol Search)d,e	259	0.23 (0.05) [−2.93 – 4.32]	−0.40	−0.03	0.11	0.16	0.16	0.20	−0.06	−0.10	−0.13	1
11. Average Working Memory (N-Back)d,e	260	0.7 (0.25) [−1.88 – 4.39]	−0.50	0.08	0.04	0.05	0.03	0.11	−0.07	−0.14	−0.06	0.63	1
12. Average Spatial Memory (Dot Memory)d,f	260	−1.84 (0.91) [−7.21 – 1.29]	−0.15	0.17	−0.05	0.26	0.27	0.04	0.08	−0.21	−0.03	0.45	0.37	1

Note. n represents the number of respondents for each variable, % represents the percentage of the sample, M and SD denote the mean and standard deviation. Correlation values are based on Pearson and point-biserial correlations, as appropriate. Correlation values represent between-person associations. Repeatedly assessed affect and cognitive variables reflect participants' mean values averaged across the 14-day EMA burst. Significant values are shown in bold (p < .01).

^aBlack represents non-Hispanic Black, White represents non-Hispanic White/Caucasian, and “Other/Minoritized” includes Hispanic-White (n = 45), Hispanic-Black (n = 18), Asian (n = 1), and “Other” (n = 9). For descriptive correlations, race and ethnicity was treated as a binary variable (0 = White, 1 = Black and other/minoritized).

^bIncome was treated as a continuous variable and was reported using eight categories ranging from < $5000 to ≥ $150,000.

^cEducation was treated as a continuous variable and was reported using five categories ranging from completion of grade school or less to a graduate or professional degree.

^dValues represent participants' averages across the 14-day EMA burst.

^eThroughput scores for processing speed and working memory were calculated by dividing the number of correct responses by the time required to complete the task.

^fSpatial memory scores reflect accuracy, higher scores represent better performance.

Intraclass correlation coefficients indicated substantial within-person variability across both affect and ambulatory cognitive performance. Specifically, within-person fluctuations accounted for 48.30% of the variance in PA, 48.31% in NA, 42.90% in processing speed, 60.20% in spatial memory, and 50.70% in working memory. The remaining variance reflected stable between-person differences, supporting the use of 3-level multilevel models to examine within-person associations between momentary affect and cognitive performance.

We examined both momentary and averaged (across the 14 days) affect as predictors and tested moderation by age for each cognitive task. Results for processing speed (symbol search task) are shown in Table 2, working memory (N-back task) in Table 3, and spatial memory (dot memory task) in Table 4.

Table 2.

Main & interactive effects of momentary affect and age on processing speed (symbol search task) with covariates.

Parameters	γ	SE	t	p	95% CI
					Lower	Upper
Intercept	0.2128	0.0131	16.21	<0.001	0.1870	0.2387
Beep	−0.0011	0.0003	−4.00	<0.001	−0.0017	−0.0006
Day	0.0018	0.0001	16.32	<0.001	0.0016	0.0020
Age	−0.0020	0.0003	−6.28	<0.001	−0.0026	−0.0014
Gendera	−0.0001	0.0070	−0.01	0.990	−0.0138	0.0136
Race & Ethnicityb	−0.0187	0.0131	−1.43	0.154	−0.0445	0.0071
Educationc	−0.0338	0.0157	−2.15	0.033	−0.0648	−0.0028
Work Statusd	0.0029	0.0071	0.41	0.683	−0.0111	0.0169
Partner Statuse	−0.0119	0.0081	−1.47	0.144	−0.0278	0.0041
Income	0.0061	0.0024	2.56	0.011	0.0014	0.0108
Main Effects	γ	SE	t	p	95% CI
					Lower	Upper
Negative Affect (NA)f	−0.0006	0.0002	−2.23	0.027	−0.0009	−0.0001
Positive Affect (PA)f	−0.0007	0.0002	−2.68	0.008	−0.0012	−0.0002
Momentary NA	0.0000	0.0000	1.47	0.141	−0.0000	0.0001
Momentary PA	−0.0001	0.0000	−1.63	0.104	−0.0001	0.0000
Interaction Effects	γ	SE	t	p	95% CI
					Lower	Upper
Momentary NA *Age	−7.22E-6	3.125E-6	−2.31	0.021	−0.0000	−1.1E-6
Momentary PA * Age	−1.93E-6	2.567E-6	−0.75	0.452	−6.96E-6	3.101E-6

Note. Day and beep are included in the models to account for practice effects.

^aReference group = “female”.

^bReference group = “Black (non-Hispanic).

^cEducation was treated as a continuous variable for analyses.

^dReference group = “working”.

^eReference group = “partnered”.

^fPerson-level average of momentary negative and positive affect across 14 EMA days.

Table 3.

Main & interactive effects of momentary affect and age on working memory (N-back task) with covariates.

Parameters	γ	SE	t	p	95% CI
					Lower	Upper
Intercept	0.6112	0.0572	10.68	< 0.001	0.4984	0.7239
Beep	−0.0058	0.0015	−3.75	< 0.001	−0.0088	−0.0028
Day	0.0160	0.0006	27.91	< 0.001	0.0149	0.0172
Age	−0.0112	0.0014	−8.18	< 0.001	−0.0139	−0.0085
Gendera	0.0484	0.0303	1.60	0.112	−0.0113	0.1081
Race & Ethnicityb	−0.0302	0.0570	−0.53	0.597	−0.1424	0.0821
Educationc	−0.1653	0.0685	−2.41	0.017	−0.3002	−0.0303
Work Statusd	−0.0054	0.0308	−0.18	0.861	−0.0662	0.0554
Partner Statuse	−0.0352	0.0352	−1.00	0.319	−0.1045	0.0342
Income	0.0123	0.0103	1.19	0.235	−0.0080	0.0326
Main Effects	γ	SE	t	p	95% CI
					Lower	Upper
Negative Affect (NA)f	−0.0035	0.0011	−3.23	0.001	−0.0056	−0.0014
Positive Affect (PA)f	−0.0014	0.0009	−1.54	0.125	−0.0032	0.0004
Momentary NA	−0.0004	0.0002	−2.43	0.015	−0.0008	−0.0001
Momentary PA	−0.0003	0.0002	−2.03	0.042	−0.0006	−0.0000
Interaction Effects	γ	SE	t	p	95% CI
					Lower	Upper
Momentary NA * Age	0.0000	0.0000	0.70	0.485	−0.0000	0.0000
Momentary PA * Age	0.00004	0.0000	2.57	0.010	8.487E- 6	0.0000

Note. Day and beep are included in the models to account for practice effects.

^aReference group = “female”.

^bReference group = “Black (non-Hispanic).

^cEducation was treated as a continuous variable for analyses.

^dReference group = “working”.

^eReference group = “partnered”.

^fPerson-level average of momentary negative and positive affect across 14 EMA days.

Table 4.

Main & Interactive Effects of Momentary Affect and Age on Spatial memory (Dot Memory Task) with Covariates.

Parameters	γ	SE	t	p	95% CI
					Lower	Upper
Intercept	−2.1341	0.2229	−9.57	< 0.001	−2.574	−1.6948
Beep	−0.0434	0.0066	−6.55	< 0.001	−0.0563	−0.0304
Day	0.0161	0.002	6.58	< 0.001	0.0113	0.0209
Age	−0.0166	0.0054	−3.11	0.003	−0.0272	−0.0061
Gendera	0.3651	0.1179	3.10	0.002	0.1328	0.5974
Race & Ethnicityb	0.0218	0.0526	−0.03	0.975	−0.4153	0.4589
Educationc	−0.6059	0.2667	−2.27	0.024	−1.1314	−0.0804
Work Statusd	−0.1506	0.1201	−1.25	0.228	−0.3872	0.0860
Partner Statuse	−0.0250	0.1370	−0.18	0.954	−0.2948	0.2449
Income	0.1106	0.0402	2.75	0.039	0.03143	0.1899
Main Effects	γ	SE	t	p	95% CI
					Lower	Upper
Negative Affect (NA)f	−0.0128	0.0042	−3.06	0.003	−0.0210	−0.0045
Positive Affect (PA)f	−0.0057	0.0036	−1.59	0.103	−0.0127	0.0014
Momentary NA	−0.0013	0.0008	−1.67	0.116	−0.0028	0.0002
Momentary PA	−0.0004	0.0007	−0.56	0.556	−0.0017	0.0009
Interaction Effects	γ	SE	t	p	95% CI
					Lower	Upper
Momentary NA *Age	−0.0000	0.0001	−0.47	0.636	−0.0002	0.0001
Momentary PA * Age	0.0001	0.0001	1.06	0.288	−0.0001	0.0002

Note. Day and beep are included in the models to account for practice effects.

^aReference group = “female”.

^bReference group = “Black (non-Hispanic).

^cEducation was treated as a continuous variable for analyses.

^dReference group = “working”.

^eReference group = “partnered”.

^fPerson-level average of momentary negative and positive affect across 14 EMA days.

7.2. Processing speed (symbol search task)

Higher scores on the symbol search task signify better performance (see Table 2). At the within-person level, neither momentary NA (γ = 0.000, t(220) = 1.47, p = .141) nor momentary PA (γ = −0.0001, t(220) = −1.63, p = .104) were significantly associated with processing speed. At the between-person level, higher average NA and PA were each significantly related to worse processing speed (γ_NA = −0.0006, t(220) = −2.23, p = .027; γ_PA = −0.0007, t(220) = −2.68, p = .008).

7.2.1. Moderation by age

Although momentary NA was not directly associated with processing speed, the relationship was significantly moderated by age (γ = −0.00001, t(221) = −2.31, p = .021). To interpret this interaction, we examined simple slopes at 1 SD above the mean age (~57 years), at the mean age (~46 years), and 1 SD below the mean age (~35 years; M = 46.49, SD = 11.02). Moments characterized by higher than typical NA were associated with significantly worse performance at older ages within the sample (b = −0.0001, t(259) = −3.85, p = < 0.001) and at the mean age (b = −0.0001, t(259) = −2.82, p = .005), but not at younger ages within the sample (b = −0.0000, t(259) = −0.27, p = .787). This interaction is illustrated in Fig. 4. Age did not moderate the association between momentary PA and processing speed (γ = −0.0000, t(220) = −0.75, p = .452).

Fig. 4.

Interaction between Negative Affect and Age on Processing Speed (Symbol Search Task)

Note. Predicted processing speed as a function of negative affect, plotted for age (grand-mean centered) probed at −1 SD (blue), 0 (red), and +1 SD (green) from grand-mean-centered age. Higher values represent better cognitive performance. Shaded areas represent 95% confidence intervals. **0 SD significant slope at p < .01. ***+1 SD significant slope at p < .001.

7.3. Working memory (N-back task)

Results for working memory are presented in Table 3. Higher scores on the N-back task represent better performance. At the within-person level, moments characterized by higher-than-usual NA were significantly associated with poorer working memory, as indicated by lower N-back task performance (γ = −0.0004, t(221) = −2.43, p = .015). Similarly, participants performed significantly worse on the N-back task during moments of higher-than-usual PA (γ = −0.0003, t(221) = −2.03, p = .042). At the between-person level, higher average NA was related to poorer N-back performance (γ = −0.0035, t(221) = −3.23, p = .001), whereas average PA was not significantly associated with N-back performance (γ = −0.0014, t(221) = −1.54, p = .125).

7.3.1. Moderation by age

There was no significant moderation effect of age on the association between momentary NA and working memory task performance (γ = 0.0000, t(221) = 0.70, p = .485). However, momentary PA significantly interacted with age in relation to working memory (γ = 0.00004, t(221) = 2.57, p = .010). To interpret this interaction, we examined simple slopes at 1 SD below the sample mean age (~35 years), at the mean age (~46 years), and 1 SD above the sample mean age (~57 years; M = 46.49, SD = 11.02). As shown in Fig. 5, moments characterized by higher-than-usual PA were associated with significantly poorer working memory performance at younger ages (b = −0.0007, t(259) = −3.25, p = .001) and at the mean age (b = −0.0003, t(259) = −2.08, p = .037), but not at older ages within the sample (b = 0.0001, t(259) = −0.32, p =.748).

Fig. 5.

Interaction between Positive Affect and Age on Working Memory (N-back Task)

Note. Predicted working memory as a function of positive affect, plotted for age (grand-mean centered) probed at −1 SD (blue), 0 (red), and +1 SD (green) from grand-mean-centered age. Higher values represent better cognitive performance. Shaded areas represent 95% confidence intervals. *0 SD significant slope at p = .037. **−1 SD significant slope at p = .001.

7.4. Spatial memory (dot memory task)

Results for spatial memory are presented in Table 4. Higher scores on the dot memory task demonstrate better performance. At the within-person level, neither momentary NA (γ = −0.0013, t(221) = −1.67, p = .116) nor momentary PA (γ = −0.0004, t(221) = −0.56, p = .556) were significantly associated with spatial memory. At the between-person level, higher average NA was significantly associated with poorer spatial memory (γ = −0.0128, t(221) = −3.06, p = .003) but average PA was not related to spatial memory (γ = −0.0057, t(221) = −1.59, p = .103).

7.4.1. Moderation by age

Age did not significantly moderate the relationships between momentary NA or PA and spatial memory (γ_NA = −0.0000, t(221) = −0.47, p = .636; γ_PA = 0.0001, t(221) = 1.06, p = .288).

7.5. Exploratory analyses

We conducted post-hoc analyses to test the moderating effect of age on average (between-person) NA and PA. We found no evidence of moderation by age between average levels of NA or PA and across any of the three cognitive tasks.

7.5.1. Tests of curvilinear PA effects

To examine the possibility of non-linear associations, additional models were estimated including a quadratic within-person PA term (constructed from standardized person-mean-centered PA values). Across outcomes, quadratic PA terms were not statistically significant for processing speed (β = 0.000097, SE = 0.000234, p = .68), working memory (β = −0.00116, SE = 0.00127, p = .36), or spatial memory (β = 0.00124, SE = 0.00544, p = .82). We also tested whether quadratic PA effects varied as a function of age; interactions between PA² and age were not significant across all outcomes (ps > 0.68). Thus, within the observed range, associations between momentary PA and cognitive performance were adequately characterized as linear.

8. Discussion

This paper leveraged EMA to examine how momentary shifts in NA and PA relate to ambulatory cognitive performance and how age moderates within-person associations. Within-person increases in NA (relative to one's own average) were associated with momentary decreases in processing speed and working memory, but not with spatial memory. For processing speed, this association was moderated by age, such that the negative associations were evident at the sample mean age and older ages within the sample. Momentary increases in PA (relative to one's own average) were linked to poorer working memory, with no relationships to processing speed or spatial memory. Age moderated the association between PA and working memory, such that the negative association was evident at younger ages and at the sample mean age, but not at older ages within the sample.

Although our primary focus was on within-person associations, between-person associations were examined for comparison and to facilitate the interpretation of within-person findings. Within-person associations describe how individuals performed during moments when they experienced more affect than usual, whereas between-person associations describe differences among individuals who report higher average affect than others. Examining both levels of analysis allowed us to assess whether associations observed within individuals across time were consistent with, or differed from, associations observed between individuals who differed in their average levels of affect. Higher average NA (between-persons) was linked to worse performance across all three cognitive tasks – processing speed, working memory, and spatial memory. Higher average PA (between-persons) was only associated with worse processing speed, with no associations observed for working memory or spatial memory.

8.1. Between-person affect and cognitive performance

The finding that individuals with greater levels of NA (averaged across the 2 weeks of EMA) performed more poorly across all three ambulatory cognitive tasks is consistent with prior research on NA and daily cognitive performance (Stawski et al., 2019; Tieks et al., 2022), as well as past work linking higher NA to poorer cognitive functioning more broadly (Gillet et al., 2012; Hulur et al., 2015). Higher than average NA could be related to poorer cognitive performance through factors such as lifestyle or health habits (e.g., smoking or less physical exercise) that ultimately can lead to cognitive impairment (Ferrer and Mendes, 2018; Zhang et al., 2019).

Our findings demonstrate that daily and momentary ratings of PA may relate differently to ambulatory cognitive performance than to daily or single laboratory measures of cognition. Unexpectedly, higher average PA (between-person) was associated with poorer processing speed, but not with working memory or spatial memory. Importantly, these results should not be interpreted as indicating that PA is broadly detrimental to cognition. Rather, they suggest that under some circumstances in daily life, moments of elevated PA may coincide with contexts involving competing attentional demands or increased engagement that temporarily reduces working memory performance. This relationship between within-person PA and cognition contrasts prior research, which has linked higher trait-like PA with better cognitive performance in laboratory settings (Dreisbach, 2006; Pupillo et al., 2022), including day to day changes in laboratory-based processing speed (Verhagen et al., 2019), or which has shown null results linking trait-like PA with day-to-day changes in laboratory-based cognitive performance (von Stumm, 2016).

Disparate findings in the current body of affect-cognition literature could be due to measurement differences in PA; we used an average of two weeks of affect ratings (with participants using a smartphone slider-bar to respond to how they felt “right now” multiple times a day), as compared to a one-time multiple-choice assessment involving retrospective recall over weeks or months, such as the Positive and Negative Affect Schedule (PANAS). The varied findings may suggest that associations between PA and cognitive performance differ depending on whether PA is assessed as a momentary, repeatedly measured state, or as a trait-like (i.e., between-person) retrospective construct.

8.2. Within-person affect and cognitive performance

8.2.1. Main effects

The primary purpose of the present study was to examine how momentary affect relates to ambulatory cognitive performance across adulthood (ages 25 to 65 years). At the within-person level, higher-than-usual levels of both NA and PA (relative to one's own average) were associated with poorer working memory performance, but not with processing speed or spatial memory. Existing EMA research suggests that days or moments characterized by higher NA (but not PA) are generally associated with poorer working memory across younger and older adults (Brose et al., 2012; Stawski et al., 2013). Although the finding of a negative relationship between within-person PA and working memory performance may seem counterintuitive, this association can be better understood when considering the significant role of age.

8.2.2. Moderation by age

Examination for moderation by age found heightened NA was associated with worse processing speed performance at the sample mean age and at older ages within the sample. In addition, moments of heightened PA were associated with worse working memory performance at younger ages and at the sample mean age, but not at older ages. Interpreting these age effects requires an understanding of both the social and environmental context and sources of affective change, which may differ substantially across adulthood but also in daily life.

Strength and vulnerability integration (SAVI) theory provides a useful framework for interpreting age-related differences in our findings. SAVI posits that although older age is often associated with lower average levels of NA, these lower levels are often the result of daily circumstances and proactive efforts to avoid situations that elicit distress (Charles, 2010). However, when people experience heightened NA, older adulthood is associated with less effective down-regulation of distress. Moreover, age-related changes in physiology may make this regulation more difficult (Charles and Luong, 2013). Consistent with this framework, our analyses indicated that heightened NA was associated with poorer processing speed only among individuals at the average age or older. This pattern suggests that age-related changes in affective-physiological coupling may make processing speed particularly sensitive to momentary fluctuations in NA later in adulthood.

Whereas we focused on moments with higher NA than average, cognitive performance may also be more connected to within-person variation in affect-related physiological arousal (Cerino et al., 2021; Le Vigouroux et al., 2022). In keeping with SAVI, older adults usually report lower levels of high-arousal NA (e.g., anger) than younger counterparts, whereas low-arousal NA (e.g., sad) typically remains consistent across ages (Kessler and Staudinger, 2009; Scheibe and Blanchard-Fields, 2009; Scheibe and Moghimi, 2021). However, because the present study could not distinguish between high- and low-arousal forms of NA the mechanisms underlying these age differences remain unclear. Future research incorporating more nuanced assessments of affective experiences and physiological processes may help clarify why momentary increases in NA appear more strongly related to processing speed in later midlife.

The age-related pattern of findings may reflect differences in the contexts in which positive emotions are experienced across adulthood. Older adults generally report higher levels of daily PA, consistent with age-related increases in motivations to maintain emotional wellbeing and a greater tendency to savor positive experiences (Growney et al., 2025; Smith and Hollinger-Smith, 2015). In contrast, moments of elevated PA among individuals at younger ages may be more likely to occur during goal-directed activities, social engagement, or experiences related to achievement and accomplishment. Supporting this possibility, greater diversity of daily activities has been linked to greater diversity of positive emotional experiences, with this association being strongest at relatively younger ages (Lee et al., 2022). Such contexts may involve competing environmental stimulation or increased attentional demands, potentially reducing the cognitive resources available for working memory performance (Garrison and Schmeichel, 2022; Plass and Kalyuga, 2019). This interpretation may help explain why higher-than-usual PA was associated with poorer working memory performance at younger ages and up through the mean age, but not at older ages within the present sample.

Although this finding contrasts with perspectives suggesting that positive emotions generally benefit cognition, prior research indicates that associations between PA and cognitive performance may depend on the cognitive domain assessed and the context in which positive emotions occur. For example, positive affect has been associated with greater cognitive flexibility but also increased distractibility, suggesting that positive emotional states may sometimes involve tradeoffs between attentional control and broader engagement with the environment (Dreisbach, 2006; Liu and Wang, 2014). Because the present study did not assess the source, context, or arousal characteristics of positive emotional experiences, these explanations remain speculative and warrant further investigation. Additionally, exploratory analyses examining quadratic associations between momentary PA and cognitive performance did not support curvilinear effects, suggesting that the observed associations were adequately characterized as linear within the range of PA observed in this sample.

Overall, these age-specific findings underscore the importance of considering affective context and age-related dynamics when interpreting EMA-based cognitive performance in everyday life. PA may become less strongly linked to cognitive vulnerability with age, whereas heightened NA appears to remain consequential for cognitive performance, especially as we age (Charles and Piazza, 2024). From a SAVI perspective, heightened NA may be more physiologically taxing or distressing as we age across midlife, particularly when it involves high-arousal states, thereby limiting available cognitive resources and constraining performance on tasks such as processing speed. In contrast, momentary increases in PA, and particularly those associated with lower arousal, may be less cognitively disruptive later in adulthood, depending on their source and context.

Importantly, associations between momentary affect and cognition were not uniform across cognitive domains. Significant associations emerged for processing speed and working memory, whereas spatial memory showed comparatively fewer associations. Rather than assuming a single underlying cognitive construct, examining ambulatory cognitive tasks separately may help identify domain-specific patterns that would otherwise be obscured in composite measures.

8.3. Limitations and future directions

Although this study provides valuable insight into how cognitive performance varies with intra-individual differences in affect and age, several limitations should be acknowledged. Despite the intensive repeated measures, these analyses were cross-sectional and non-experimental, limiting our ability to draw causal temporal conclusions. Further, our EMA approach involved responses that were hours apart and across a two-week period, limiting our ability to examine moment-to-moment changes in affect and cognition or generalizing across broader periods of time. However, although it was beyond the scope of the current paper, future studies could use these data to examine day-to-day lagged effects, working to better understand the temporal nature of the relationships. The bidirectional nature of affect and cognition warrants additional study. Prior work suggests that within-person variability in cognition itself can shape affective experiences, underscoring the importance of examining the psychological, social, and physiological mechanisms linking momentary affect and cognitive performance.

Future longitudinal research, particularly on low- and high-arousal emotions in a set of adults ranging beyond age 65, may clarify the ages at which within-person affective experiences relate to differences in mobile cognitive performance. Because our sample had a maximum age of 65, participants toward the top of this range represent relatively “young” older adults, and findings may not generalize to older or more cognitively vulnerable populations. Our results may only generalize to the larger Bronx, NY, population that it represented. For example, the present sample was predominately Black (63%) and Black individuals are more likely to experience racial discrimination, which has been linked to poorer spatial working memory (Harrington et al., 2025). Future work focusing on Black samples and relevant contextual factors may improve understanding of the associations between affect and cognition. Replication in samples that are more diverse with attention to variation in age, racial and ethnic identity, cognitive impairment statuses (e.g., mild cognitive impairment, early dementia), is also warranted.

8.4. Summary and implications

By distinguishing within-person from between-person processes and evaluating age-related differences across multiple objective ambulatory cognitive domains, this work aimed to provide a more nuanced conceptual understanding of affect-cognition dynamics across adulthood. We found that higher average NA (between-persons) was related to poorer processing speed, working memory, and spatial memory performance. Having higher average PA (between-persons) was only associated with slower processing speed and was not associated with working or spatial memory. Heightened momentary NA (within-person) was linked to poorer processing speed and working memory, whereas momentary increases in PA were associated with worse working memory performance. These effects varied by age but did not extend to spatial memory, suggesting that elevated affective states – regardless of valence – may interfere with specific cognitive processes in daily life.

Theoretical and practical implications emerge from these findings. Differences between within- and between-person effects, as well as across types of cognitive tasks, underscore the complexity of associations between momentary affect and cognition. Future research using EMA could further unpack how person-level factors interact with social and environmental contexts to shape these dynamics. Age proved to be an important moderator, highlighting the need to consider age as a source of adaptability in how affective experiences influence cognitive performance across the lifespan. Notably, increasing age across early adulthood and into later midlife may be associated with greater vulnerability to the cognitive costs of NA, potentially due to declines in executive function and a reduced capacity to buffer against emotional disruption.

8.4.1. Clinical and policy implications

The present study suggests that affective experiences may represent an important contextual factor associated with cognitive performance in everyday life. In particular, higher-than-usual levels of NA were associated with slower processing speed at older ages within the present sample, indicating that NA may have greater implications for momentary cognitive efficiency as adults move through midlife. This paper highlights the importance of considering both affective experiences and age when examining cognitive functioning in daily life.

From a clinical perspective, these findings suggest that periods of higher-then-usual NA may coincide with reduced efficiency in processing information and adults in later midlife may experience a greater cognitive burden between NA and cognition. As a result, communication strategies such as reducing information density, presenting information in multiple formats, and providing opportunities for clarification may be particularly beneficial when conveying complex information during emotionally demanding situations. More broadly, the findings underscore the importance of recognizing that cognitive functioning may fluctuate alongside affective experiences across adulthood.

At the policy level, the present findings support efforts to promote emotional wellbeing across adulthood and to recognize the dynamic nature of cognitive functioning in everyday contexts. Understanding how affective experiences relate to cognitive performance in daily life may be particularly important given growing evidence that emotional wellbeing is linked to long-term cognitive health and aging outcomes (Graham-Engeland et al., 2024).

8.5. Conclusion

By capturing affective experiences and cognitive performance as they unfold in daily life, EMA approaches may provide valuable insight into when individuals are most vulnerable to affect-related changes in cognitive functioning. Overall, the present findings highlight the importance of conceptualizing affect and cognition as dynamic, interconnected processes that vary across everyday contexts and across adulthood, particularly as adults move through midlife.

Data availability

The authors do not have permission to share data.