Daily Linkages among High and Low Arousal Affect and Subjective Cognitive Complaints

Abstract

Objectives

Subjective cognitive complaints may be an early indicator of Alzheimer’s disease pathology and related dementias that can be detectable prior to objective, performance-based decline. Negative and positive affective states (NA and PA, respectively) are established psychosocial correlates of cognition in older adulthood and have demonstrated capacity for meaningful within-person fluctuations based on person-environment interactions, age, and measurement approach.

Method

We utilized data from a 100-day, microlongitudinal study of 105 community-dwelling older adults (M_age=63.19, SD=7.80, Range=52–88) to explore within- and between-person associations between high and low arousal NA and PA, and memory- and attention-related complaints.

Results

For memory-related complaints, those who reported experiencing greater NA-high arousal had increased forgetfulness (OR=2.23, 95%CI: 1.11–4.49, p<.05). Within persons, reporting more NA-high arousal than usual was associated with increased forgetfulness (OR=1.01, 95%CI: 1.004–1.018, p<.01). For attention-related complaints, those who reported experiencing greater NA-low arousal had increased trouble staying focused (OR=2.34, 95%CI: 1.17–4.66, p<.05). Within persons, reporting more NA-low arousal (OR=1.02, 95%CI: 1.01–1.03, p<.001) and less PA-high arousal (OR=0.96, 95%CI: 0.95–0.97, p<.001) than usual was associated with increased trouble staying focused. Additionally, reporting more PA-low arousal than usual was associated with decreased trouble staying focused among those with higher levels of conscientiousness (OR=0.72, 95%CI: 0.57–0.92, p<.01).

Conclusion

Results from this study offer a means to maximize resource allocation and personalized cognitive health efforts by pinpointing for whom and on which days boosting PA and/or reducing NA may both serve as pathways to benefit daily subjective cognition.

Introduction

Studying early indicators of age-related cognitive decline and pathological impairment is crucial to inform approaches that optimize cognitive health and reduce decline. Subjective cognitive complaints (henceforth referred to as cognitive complaints) may be an early indicator of Alzheimer’s disease pathology (Amariglio et al., 2012) and related dementias (Jonker, Geerlings, & Schmand, 2000) that can be detectable prior to performance-based decline. Importantly, just over 11% of adults 45 years of age and older have reported increased memory loss or confusion, with one-half of these individuals specifying functional limitations due to this subjective decline (CDC, 2018).

Affective states – both negative and positive experiences of emotion – are correlates of subjective cognition (e.g., Drogos et al., 2013), and have meaningful within-person fluctuations based on person-environment interactions, age, and measurement approach (e.g., Röcke, Li, & Smith, 2009; Shifren & Hooker, 1995). In the current study we elucidate ways in which negative and positive affective states (NA and PA, respectively) are associated with reports of forgetting something and trouble staying focused. We do so because we hope to identify factors within- and between-persons that might optimize cognitive health in older adulthood. Informed by the circumplex model of affect (Feldman, 1995), we examine both high and low arousal NA and PA. This model defines a circular ordering of stimuli around two dimensions of valence (negative/positive) and arousal (high/low). Examining all four dimensions of the circumplex (NA-low arousal, NA-high arousal, PA-low arousal, PA-high arousal) enables more complete descriptions of affect (Russell, 1979) and offers clinical relevance (e.g., low arousal ‘depressed’ can be valuable for screening purposes; Lawton, Kleban, Dean, Rajagopal, & Parmelee, 1992).

Affect and Cognition in Older Adulthood

Affect-cognition associations have been established in theoretical (e.g., Eysenck & Calvo, 1992) and empirical work (e.g., Hülür, Hertzog, Pearman, Ram, & Gerstorf; 2014). Particularly relevant to this study are the resource allocation model, processing efficiency theory, and broaden-and-build theory. Resource allocation model (Ellis & Ashbrook, 1988) suggests a low arousal depressed mood state can reduce necessary attentional capacity to appropriately address a task, thereby resulting in poorer task performance. Across 17 years, Hülür et al. (2014) showed that subjective memory was rated better than usual on occasions when individuals reported fewer depressive symptoms than usual. Further, recent work by Hill and colleagues (2019) showed that older adults who reported more depressive symptoms than usual had increased likelihood of perceived ten-year memory decline at the subsequent year of assessment.

Processing efficiency theory (Eysenck & Calvo, 1992) suggests high arousal anxiety impairs one’s ability to process stimuli efficiently because there is considerable distraction from anxious-related affective states that prevent efficient processing and task execution. Across six years, Comijs, Deeg, Dik, and Jonker (2002) showed that memory complaints were linked to anxiety and depressive symptoms, especially among adults with higher control beliefs.

Broaden-and-build theory (Frederickson & Branigan, 2005) posits the utility of PA in broadening the scope of attention and thought-action repertoires. Cross-sectional work showed lower levels of PA and higher levels of NA were associated with more memory complaints (Drogos et al., 2013).

A systematic review by Hill et al. (2016) recognized the need for additional longitudinal assessment to better understand the temporality of the kinds of associations we have just described. Indeed, there is precedent for research to examine affect as an intervention target for promoting cognitive health, and to examine subjective cognition as a preclinical indicator of pathological depressive and anxiety symptomatology (Hill et al., 2016). For example, Mogle et al. (2019) identified within-person associations between memory lapses and affect across seven consecutive days such that older adults reported lower PA and higher NA on days with a prospective memory lapse.

Neuroticism and Conscientiousness in the Context of Cognitive Health and Aging

Neuroticism and conscientiousness are particularly relevant correlates of memory complaints (Pearman & Storandt, 2004; Hill, Mogle, Bhargava, Bell, & Wion, 2019). A systematic review found higher neuroticism and lower conscientiousness was consistently associated with subjective cognitive impairment in older adults without dementia (Koller, Hill, Mogle, & Bhang, 2019). Among community-dwelling older adults, better ratings of subjective memory were associated with higher levels of conscientiousness and lower levels of neuroticism (Hülür, Hertzog, Pearman, & Gerstorf, 2015). Moderation analyses revealed those who reported high levels of conscientiousness had more robust associations between subjective and objective memory (Hülür et al., 2015) than those with low levels of conscientiousness, in support of the potentially adaptive role of conscientiousness (Hertzog & Pearman, 2013). It is important to consider conscientiousness and neuroticism as potential individual differences factors when examining correlates of cognitive complaints.

The Current Study

Affective and subjective cognitive processes vary within-persons and these variations occur in day-to-day life (e.g., Mogle et al., 2019), qualities of psychological processes that Molenaar and Campbell (2009) argued warrants examination of intraindividual variability (IIV). An IIV approach recognizes that data from a single time point, or even multiple time points arrayed over time with long intervals between measurement occasions, are insufficient for examining an individual’s behavior (Nesselroade & Ram, 2004). The microlongitudinal design used in the current study enables us to utilize the IIV approach to examine associations among affect and subjective indicators of cognitive health, both within- and between-persons.

Arousal and affective states

Previous research using established measures of affect have primarily focused on aggregate indices of affect based on valence (i.e., positive and/or negative), largely ignoring the potential import of arousal. Therefore, a critical gap in this literature is the uncertainty around the impact of potentially high arousal affect (e.g., feeling annoyed or irritated) versus low arousal affect (e.g., feeling sad or depressed) on self-reported indicators of cognitive health.

Multiple domains of cognition

Most research on subjective cognition focuses on memory (Rabin et al., 2015) and examines psychosocial correlates over macro time scales (annual assessments or multiple years separating time points). However, cognition includes more domains than memory (e.g., attention, working memory, processing speed), and the rate of change in these domains varies widely among individuals (Schaie, Willis, & Caskie, 2004). An additional gap, then, is the understanding of associations among psychosocial risk factors and different domains of cognitive complaints.

Research questions

We address these gaps in the understanding of relationships between affect, forgetting something, and trouble staying focused. To further explore individual differences in these relationships, additional analyses examined the moderating roles of personality traits. We ask two research questions:

1. Are arousal-specific NA and PA and cognitive complaints associated both within-persons over time and between-persons? Within-persons, we hypothesized that participants would be more likely to report cognitive complaints on days when NA is higher and PA is lower than usual over the 100 days. Between-persons, individuals reporting higher levels of NA and lower levels of PA on average would be more likely to report cognitive complaints. Further, we examined whether high arousal affect relates to a greater likelihood of cognitive complaints than low arousal affect.

2. Do neuroticism and conscientiousness moderate within-person associations between affect and cognitive complaints? We examined within-person affect by personality trait interactions to evaluate whether the adaptive/maladaptive role of affect strengthens depending on levels of neuroticism and conscientiousness.

Method

Study Design and Procedure

We utilized data from the Personal Understanding of Life and Social Experiences (PULSE) project (Hooker et al., 2013), a 100-day microlongitudinal study of health- and social-related behaviors and self-regulatory processes. After a baseline survey of demographic and psychosocial measures, participants completed 100 daily surveys between June 2010 and October 2010. A quarter of the sample was randomly assigned to a burst group where a total of four 7-day bursts were administered throughout the 100-day period as part of a secondary goal of the PULSE project to investigate methodological questions involving multiple temporal intervals (e.g., Mejía, Hooker, Ram, Pham, & Metoyer, 2014). Burst and daily measurement groups were analyzed together because burst group missingness was set by design. Further, measurement groups did not differ in sociodemographic, health, affect, or cognitive complaint variables, T²=24.76, F(14,87)=1.54, p=.11. Daily reminder emails, with a link to the custom internet-based survey, were sent to each participant in the evening (expiring at 2:00 a.m.).

Sample Demographics

The sample consists of 99 community-dwelling older adults (M_age=63.24, SD=7.94, Range=52–88) drawn from a participant pool (N=450) in Oregon (Table 1). The primarily female (88%), white (97%), married (72%), healthy (<1 daily physical symptom reported on average; 93% self-rated their health as either ‘good’ or ‘excellent’) and highly educated (79% ≥ college degree) participants were partially representative of older adults in Oregon. Each of the demographics and study variables were statistically comparable across the categorical marital status covariate, except for age (participants who identified as widowed were significantly older than other marital statuses; see Appendix A Table 1).

Table 1.

Descriptive Statistics and Bivariate Correlations for Between-Person Variables

Variable	M(SD)	Range	1	2	3	4
1. NA-High Arousal	19.29(15.18)	0.45–60.63	-
2. NA-Low Arousal	28.29(24.96)	0.85–102.37	.83***	-
3. PA-High Arousal	69.81(15.28)	38.09–96.96	−.69***	−70***	-
4. PA-Low Arousal	102.66(23.31)	51.40–145.40	−.72***	−.72***	.94***	-
5. Trouble (N instances, %)	491(7.00)	-	.34***	.48***	−.36***	−.39***
6. Forgot (N instances, %)	332(4.73)	-	.29**	.27**	−.23*	−.20*
7. N	47.94(13.14)	25.64–78.20	.57***	.57***	−.44***	−.53***
8. C	46.51(11.64)	15.33–68.96	−.32**	−.26*	.31**	.29**
9. Age	63.24(7.94)	52–88	−.00	−.02	−.02	−.01
10. Health	2.42(0.62)	1–3	−.19†	−.23*	.28**	.29**
11. Symptoms	0.84(1.04)	0–9	.32**	.33***	−.36***	−.31**
12. Female (%)	87.88	-	.10	.16	−.08	−.07
13. White (%)	96.97	-	−.00	.00	.06	.06
14. College Degree (%)	78.79	-	−.08	−.09	.10	.09
Variable	5	6	7	8	9	10	11	12	13	14
5. Trouble (N instances, %)	-
6. Forgot (N instances, %)	.44***	-
7. N	.30**	.24*	-
8. C	−.19†	−.27**	−.55***	-
9. Age	−.15	.09	−.09	.02	-
10. Health	−.26**	−.18†	−.30**	.25*	.09	-
11. Symptoms	.42***	.40***	.30**	−.26*	.05	−.46***	-
12. Female (%)	.18†	.15	.00	−.12	−.00	−.10	.12	-
13. White (%)	−.18†	−.03	.11	−.18†	.10	.21*	−.06	.12	-
14. College Degree (%)	.03	.04	−.02	.17†	−.19†	.08	−.14	−.12	−.09	-

Note. N=99. NA=negative affect. PA=positive affect. Trouble=trouble staying focused or concentrating. Forgot=forgot something. N=neuroticism T-score. C=conscientiousness T-score. Health=self-rated general health. Symptoms=physical symptoms. College Degree=percent of sample who graduated with at least a 4 year college degree.

^†p<.10

^*p<.05

^**p<.01

^***p<.001.

Measures

Personality traits

Neuroticism and conscientiousness were measured by the NEO-Five Factor Inventory (NEO-FFI; Costa & McCrae, 1992) at baseline. This 60-item inventory included 12 conscientiousness items (e.g., ‘I have a clear set of goals and work toward them in an orderly fashion’) and 12 neuroticism items (e.g., ‘I often get angry at the way people treat me’). Answered on a 5-point Likert-type scale, the NEO-FFI ranges from 0 (strongly disagree) to 4 (strongly agree). Total scores for trait subscales indicate the degree in which the participant exhibits the trait with higher scores meaning greater display of the trait. We utilized T-scores recommended by Costa and McCrae (1992). The NEO-FFI demonstrated adequate internal consistency for neuroticism (α=.92) and conscientiousness (α=.83) subscales.

Daily-level cognitive complaints

Two cognitive complaints were measured daily using the Self-Rated Health, Pain, and Symptoms Checklist (Winter, Lawton, Ruckdesche, & Sando, 2007). Daily attention-related complaints were measured with the single item ‘trouble staying focused or concentrating.’ Daily memory-related complaints were measured with the single item ‘forgot something.’ Both items were dichotomous (0=No, 1=Yes). There were 491 (7%) instances of participants reporting trouble staying focused and 332 (5%) instances of participants reporting they forgot something (Table 1). Trouble staying focused was positively associated with forgetting something both between-persons (r=.44, p<.001) and within-persons (r=.11, p<.001).

Daily-level affect

Affect was measured daily using the 10-item Philadelphia Geriatric Center Positive and Negative Affect Scales (Lawton et al., 1992). High and low arousal subscales were computed for both NA and PA by summing responses for specific items based on alignment with the circumplex model of affect (Feldman, 1995). NA-high arousal included the sum of responses for ‘annoyed’ and ‘irritated’, and NA-low arousal included the sum of responses for ‘sad’, ‘depressed’, and ‘worried.’ PA-high arousal included the sum of responses for ‘energetic’ and ‘interested’, and PA-low arousal included the sum of responses for ‘happy’, ‘content’, and ‘warm hearted.’ Participants were asked to ‘Use the slide indicator to describe how well these words describe your feelings and emotions today’ on a continuous scale ranging from 0 (not at all) to 49 (extremely), although participants were unable to see the numbers along the scale to promote independent assessments each day (e.g., Hooker et al., 2013). Higher scores reflected greater affective experience for each domain. Following Cranford and colleagues (2006), we computed within- and between-person reliability estimates for NA-high arousal (within-person=.86; between-person=.99), NA-low arousal (within-person=.76; between-person=.99), PA-high arousal (within-person=.62; between-person=.99), and PA-low arousal (within-person=.82; between-person=.99).

Covariates

Higher prevalence of cognitive complaints is linked to older age, female gender, and functional limitations (Jonker et al., 2000; Hülür et al., 2014). Poorer mental health is associated with lower physical health (World Health Organization, 2014) and higher proportions of widowed women (Kessler, Mickelson, Walters, Zhao, & Hamilton, 2004). Therefore, age, gender, marital status, daily physical symptoms, and self-rated health were included as covariates. Self-rated health was measured at baseline with the item, “In general, would you say your health is poor, fair, good, or excellent” on a scale ranging from 0 (poor) to 3 (excellent). Physical symptoms were measured daily and defined as the sum of the following symptoms: shortness of breath, mobility trouble, allergies, poor appetite, dizziness, heart pounding, nausea/upset stomach, chest tightness, constipation/diarrhea, stiffness/muscle soreness.

Analytic Strategy

We utilized logistic generalized linear mixed models (binary outcome distribution and logit link function in PROC GLIMMIX; SAS Institute, 2013) to address the research questions. Intraclass correlation coefficients from unconditional mixed linear models (PROC MIXED; SAS Institute, 2013) were used to determine within- and between-person variation in primary study variables. Level-1 (within-person) affect variables and daily physical symptoms were computed by subtracting an individual’s average level of affect and physical symptoms (across 100 days) from their daily scores (Hoffman & Stawski, 2009). Level-2 (between-person) variables included person-mean centered affect and physical symptoms variables (‘mn’ appended to variable names in equation below), grand-mean centered conscientiousness, neuroticism, age, and self-rated health (centered at the sample averages to facilitate meaningful interpretations of intercept and slope estimates), as well as gender, marital status, and affect by personality trait interaction terms. Maximum likelihood estimation was used due to missing data and attrition across days. Separate models were estimated to predict trouble staying focused and forgetting something. Each model included the four arousal-specific affect terms. The equation below, however, specifies only NA and PA to condense the information necessary for describing the modeling approach.

Research questions examining associations between affect and cognitive complaints, and potential moderation from conscientiousness and neuroticism were assessed with 2-level logistic generalized MLMs as follows:

Logistic Generalized Linear Mixed Models for Cognitive Complaints

Level-1:	CognitiveComplaintdi = β0i(Intercept) + β1i(Dayi) + β2i(NAdi–NAmni) + β3i(PAdi–PAmni) + β4i(Symptomsdi–Symptomsmni)
Level-2:	β0i = γ00 + γ01(Agei) + γ02(Genderi) + γ03(MaritalStatusi) + γ04(Healthi) + γ05(Symptomsmni) + γ06(NAmni) + γ07(PAmni) + V0i
	β1i = γ10
	β2i = γ20 + γ21(Conscientiousnessmn) + γ22(Neuroticismmn)
	β3i = γ30 + γ31(Conscientiousnessmn) + γ32(Neuroticismmn)
	β4i = γ40

Research Question 1 examined main effects of affect, where the odds of reporting a cognitive complaint (e.g., forgetting something) for person i at day d is a function of level-1 intercept (β_0i), linear trend across days (β_1i), within-person NA effect (β_2i), within-person PA effect (β_3i), within-person physical symptoms effect (β_4i), level-2 between-person differences in age (γ₀₁), gender (γ₀₂), marital status (γ₀₃), health (γ₀₄), physical symptoms (γ₀₅), NA (γ₀₆), PA (γ₀₇), and a random effect for the intercept (V_0i) to allow for variation across persons. Question 2 required additional affect by personality trait interaction terms. Thus, the within-person NA [PA] effect on a cognitive complaint is a function of the within-person NA [PA] effect at the sample average age (γ₂₀) [(γ₃₀)] and its cross-level interactions with conscientiousness (γ₂₁) [(γ₃₁)] and neuroticism (γ₂₂) [(γ₃₂)].

We report odds ratios (OR) with lower limit and upper limit of 95% confidence intervals (CI) to indicate the odds of reporting a cognitive complaint given a 1-unit increase in the within-person predictor variable. Between-person associations were rescaled to reflect the odds of reporting a cognitive complaint with a 1SD increase in the predictor variable. To aid in the interpretability of interaction effects, we calculated the predicted probability ($\frac{Odds}{\left(1+Odds\right)}$) of reporting a cognitive complaint given a 1SD increase and 1SD decrease in the within-person predictor variable (affect) at +1SD and −1SD of the moderator variable (personality). Within-person SDs were calculated by obtaining the square root of the level-1 variance (i.e., residual) in unconditional mixed linear models for relevant within-person predictor variables (affect).

Results

Tables 1 and 2 provide descriptive statistics and bivariate correlations for between-person and within-person variables, respectively. Unconditional generalized MLMs showed significant variation in NA-high arousal, NA-low arousal, PA-high arousal, PA-low arousal, trouble staying focused, and forgetting something (Figure 1).

Table 2.

Bivariate Correlations for Within-Person Variables

Variable	1	2	3	4	5	6	7
1. NA-High Arousal	-
2. NA-Low Arousal	.51***	-
3. PA-High Arousal	−.32***	−.44***	-
4. PA-Low Arousal	−.49***	−.58***	.65***	-
5. Trouble (N instances, %)	.14***	.21***	−.22***	−.20***	-
6. Forgot (N instances, %)	.08***	.05***	−.02†	−.04***	.11***	-
7. Symptoms	.08***	.14***	−.16***	−.15***	.06***	−.01	-

Note. N=99. NA=negative affect. PA=positive affect. Trouble=trouble staying focused or concentrating. Forgot=forgot something. N=neuroticism T-score. C=conscientiousness T-score. Health=self-rated general health. Symptoms=daily physical symptoms.

^†p<.10

^*p<.05

^**p<.01

^***p<.001.

Figure 1.

Variance Decompositions for Primary Study Variables

Note. Values depicted reflect proportion of variation across persons and days.

Affect Associated with Odds of Trouble Staying Focused or Concentrating

Main effects

Examination of main effects (Table 3, Models 1–2; Figure 2, Left Panels) revealed significant within- and between-person associations between NA-low arousal, PA-high arousal, and trouble staying focused. On days when participants reported higher levels of NA-low arousal than usual, there were significantly higher odds of reporting trouble staying focused (OR=1.02, 95%CI: 1.01 to 1.03, p<.001). On days when participants reported higher levels of PA-high arousal than usual, there were significantly lower odds of reporting trouble staying focused (OR=0.96, 95%CI: 0.95 to 0.97, p<.001). Between-persons, participants who reported experiencing greater NA-low arousal on average (+1SD) also had significantly higher odds of reporting trouble staying focused (OR=2.34, 95%CI: 1.17 to 4.66, p=.03).

Table 3.

Logistic Generalized MLMs for High and Low Arousal Affect Associated with Odds of Trouble Staying Focused

	Model 1: Unadjusted	Model 2: Adjusted	Model 3: Personality Moderation
Fixed Effects	OR [95% CI]	OR [95% CI]	OR [95% CI]
Day1	0.99 [0.99, 1.00]***	0.99 [0.99, 1.00]***	0.99 [0.99, 0.99]***
Age (centered at mean 63 years) SD		0.68 [0.46, 1.02]†	0.72 [0.49, 1.07]
Gender (0=Male, 1=Female)		1.87 [0.50, 6.97]	2.49 [0.68, 9.13]
Marital Status		1.02 [0.70, 1.49]	0.80 [0.54, 1.19]
Self-rated Health SD		1.15 [0.74, 1.80]	1.13 [0.73, 1.77]
WP Physical Symptoms		0.96 [0.84, 1.11]	0.96 [0.84, 1.10]
BP Physical Symptoms SD		2.17 [1.40, 3.35]***	1.59 [1.03, 2.46]*
WP NA-High Arousal	1.00 [1.00, 1.01]	1.00 [1.00, 1.01]	1.00 [0.99, 1.01]
BP NA-High Arousal SD	1.06 [0.49, 2.29]	0.90 [0.45, 1.78]	0.81 [0.40, 1.63]
WP NA-Low Arousal	1.02 [1.01, 1.03]***	1.02 [1.01, 1.03]***	1.02 [1.01, 1.02]***
BP NA-Low Arousal SD	2.67 [1.22, 5.86]*	2.34 [1.17, 4.66]*	2.46 [1.22, 4.97]*
WP PA-High Arousal	0.96 [0.95, 0.97]***	0.96 [0.95, 0.97]***	0.96 [0.95, 0.97]***
BP PA-High Arousal SD	1.00 [0.30, 3.36]	1.45 [0.49, 4.27]	1.36 [0.44, 4.16]
WP PA-Low Arousal	0.99 [0.99, 1.00]	1.00 [0.99, 1.00]	0.99 [0.98, 1.00]*
BP PA-Low Arousal SD	0.72 [0.20, 2.63]	0.54 [0.17, 1.70]	0.55 [0.17, 1.77]
N SD			0.94 [0.53, 1.69]
C SD			0.68 [0.43, 1.09]
WP NA-High Arousal*N			1.00 [1.00, 1.00]
WP NA-High Arousal*C			1.00 [1.00, 1.00]
WP NA-Low Arousal*N			1.00 [1.00, 1.00]
WP NA-Low Arousal*C			1.00 [1.00, 1.00]
WP PA-High Arousal*N			1.00 [1.00, 1.00]
WP PA-High Arousal*C			1.00 [1.00, 1.00]
WP PA-Low Arousal*N			1.00 [1.00, 1.00]
WP PA-Low Arousal*C			0.999 [0.9989, 1.0001]†
Random Effect
Intercept Estimate (SE)	3.39 (0.74)	2.44 (0.53)	2.28 (0.49)
−2LL	2259.10	2241.72	2222.62
Residual ICC	0.51	0.43	0.41

Note. N = 99. OR = odds ratio. WP = Within-person. BP = Between-person. SD = standard deviation difference in between-person effect. Day1 = Linear trend. NA = negative affect. PA = positive affect. N = neuroticism T-score. C = conscientiousness T-score. Health = self-rated general health. Symptoms = physical symptoms. Residual ICC = Residual intraclass correlation coefficient indicating the proportion of variance unexplained by variables in the model that can be attributed to individual differences.

^†p<.10

^*p<.05

^**p<.01

^***p<.001.

Figure 2.

Within- and Between-Person Associations among Affect and Cognitive Complaints

Note. ln(Odds Ratio)=coefficients (betas) from logistic generalized linear mixed models. êbeta=odds ratio. WP=within-person. BP=between-person. NA=negative affect. PA=positive affect. *p<.05, **p<.01, ***p<.001.

Personality differences

There was marginal evidence of moderation by conscientiousness for the within-person association between PA-low arousal and trouble staying focused (OR=0.99, 95%CI: 0.998 to 1.000, p=.08; Table 3, Model 3; Figure 3). On days when levels of PA-low arousal were +1SD (within-person SD=19.39) higher than usual among participants with relatively higher levels of conscientiousness, there were significantly lower odds of reporting trouble staying focused (+1SD; OR=0.72, 95%CI: 0.57 to 0.92, p<.01). This relationship did not emerge among those with relatively lower levels of conscientiousness (−1SD; OR=0.92, 95%CI: 0.75 to 1.12, p=.38). In terms of predicted probability per SD difference in PA-low arousal, the probability of reporting trouble staying focused on days when PA-low arousal was +1SD higher [−1SD lower] than usual was .42 [.58] for those with relatively higher levels of conscientiousness – compared to .48 [.52] for those with relatively lower levels of conscientiousness. Neuroticism did not moderate associations between arousal-specific affect and reporting trouble staying focused.

Figure 3.

Conscientiousness Moderates Within-Person Positive Affect-Trouble Staying Focused Associations

Note. ln(Odds Ratio)=coefficients (betas) from logistic generalized linear mixed model. êbeta=odds ratio. C=conscientiousness. Bars reflect odds of trouble staying focused on days when within-person positive affect-low arousal is +1SD higher than usual. **p<.01.

Affect Associated with Odds of Forgetting Something

Main effects

An examination of main effects (Table 4, Models 1–2, Figure 2, Right Panels) revealed significant within- and between-person associations between NA-high arousal and reporting forgetting something. On days when participants reported higher levels of NA-high arousal than usual, there were significantly higher odds of forgetting something (OR=1.01, 95%CI: 1.004 to 1.018, p<.01). Between-persons, participants who reported experiencing greater NA-high arousal on average (+1SD) also had significantly higher odds of forgetting something (OR=2.23, 95%CI: 1.11 to 4.49, p=.03).

Table 4.

Logistic Generalized MLMs for High and Low Arousal Affect Associated with Odds of Forgetting Something

	Model 1: Unadjusted	Model 2: Adjusted	Model 3: Personality Moderation
Fixed Effects	OR [95% CI]	OR [95% CI]	OR [95% CI]
Day1	0.99 [0.98, 0.99]***	0.99 [0.98, 0.99]***	0.98 [0.98, 0.99]***
Age (centered at mean 63 years) SD		1.14 [0.79, 1.63]	1.21 [0.90, 1.61]
Gender (0=Male, 1=Female)		3.02 [0.78, 11.64]	4.40 [1.36, 14.22]*
Marital Status		1.12 [0.79, 1.58]	0.96 [0.73, 1.28]
Self-rated Health SD		0.91 [0.59, 1.40]	0.93 [0.66, 1.33]
WP Physical Symptoms		0.90 [0.75, 1.07]	0.91 [0.77, 1.09]
BP Physical Symptoms SD		1.59 [1.05, 2.41]*	1.28 [0.91, 1.80]
WP NA-High Arousal	1.01 [1.004, 1.019]**	1.01 [1.004, 1.018]**	1.01 [1.001, 1.017]*
BP NA-High Arousal SD	2.39 [1.11, 5.14]*	2.23 [1.11, 4.49]*	1.74 [0.96, 3.16]*
WP NA-Low Arousal	1.00 [1.00, 1.01]	1.00 [1.00, 1.01]	1.00 [1.00, 1.01]
BP NA-Low Arousal SD	0.83 [0.38, 1.79]	0.73 [0.36, 1.48]	0.70 [0.38, 1.29]
WP PA-High Arousal	1.00 [0.99, 1.02]	1.00 [0.99, 1.02]	1.00 [0.99, 1.01]
BP PA-High Arousal SD	0.52 [0.17, 1.58]	0.69 [0.25, 1.90]	0.61 [0.26, 1.44]
WP PA-Low Arousal	1.00 [0.99, 1.01]	1.00 [0.99, 1.01]	1.00 [0.99, 1.01]
BP PA-Low Arousal SD	1.31 [0.41, 4.22]	1.08 [0.37, 3.17]	1.20 [0.48, 2.99]
N SD			0.96 [0.66, 1.40]
C SD			1.23 [0.81, 1.86]
WP NA-High Arousal*N			1.00 [1.00, 1.00]
WP NA-High Arousal *C			1.00 [1.00, 1.00]
WP NA-Low Arousal*N			1.00 [1.00, 1.00]
WP NA-Low Arousal*C			1.00 [1.00, 1.00]
WP PA-High Arousal*N			1.00 [1.00, 1.00]
WP PA-High Arousal*C			1.00 [1.00, 1.00]
WP PA-Low Arousal*N			1.00 [1.00, 1.00]
WP PA-Low Arousal*C			1.00 [1.00, 1.00]
Random Effect
Intercept Estimate (SE)	2.86 (0.65)	2.18 (0.51)	1.27 (0.24)
−2LL	2127.20	2112.04	2111.48
Residual ICC	0.46	0.40	0.28

Note. N = 99. OR = odds ratio. WP = Within-person. BP = Between-person. SD = standard deviation difference in between-person effect. Day1 = Linear trend. NA = negative affect. PA = positive affect. N = neuroticism T-score. C = conscientiousness T-score. Health = self-rated general health. Symptoms = physical symptoms. Residual ICC = Residual intraclass correlation coefficient indicating the proportion of variance unexplained by variables in the model that can be attributed to individual differences.

^†p<.10

^*p<.05

^**p<.01

^***p<.001.

Neuroticism and conscientiousness did not moderate associations between arousal-specific affect and forgetting something.

Assessing Collinearity

Collinearity is a concern for the simultaneous prediction models distinguishing by affective arousal given moderate–strong correlations between NA-high and NA-low arousal and PA-high and PA-low arousal. Appendix B provides a detailed description of sensitivity analyses and confirmatory factor analyses conducted to assess the extent to which collinearity influenced weak, non-significant odds ratios to change direction. In general, non-significant between-person estimates appeared unstable and changed direction more than within-person estimates, broadly consistent with other research that found collinearity to be of comparatively greater concern for between-person findings than within-person findings (Stawski et al., 2019). Theoretical precedent (Feldman, 1995), good model fit (Appendix B, Table 3), and results that were robust to sensitivity analyses (Appendix B, Tables 1–2), however, suggest arousal-specific within-person and between-person affect variables can be simultaneously included in predictive models of trouble staying focused and forgetting something in this sample.

Discussion

In a sample of community-dwelling older adults, high and low arousal NA and PA were associated with cognitive complaints both within-persons over time and between-persons. Patterns of associations offered partial support for a priori hypotheses, as relationships differed depending on affective arousal, individual differences in conscientiousness, and domain of cognitive complaints. The differential findings by arousal underscore the value of using affect scales with all dimensions of the circumplex model of affect (NA-high arousal, NA-low arousal, PA-high arousal, PA-low arousal; Feldman, 1995) represented in order to obtain more thorough understanding of affect in older adulthood (Russell, 1979) and to elucidate which affective states are especially relevant correlates of various domains of subjective cognition. Below, we discuss implications and potential reasons why NA-low arousal and PA may be more relevant for attention-related complaints and NA-high arousal may be comparatively more relevant for memory-related complaints.

Informing Theoretical Accounts of Affect-Cognition Associations

It is possible that feeling more annoyed or irritated than usual, markers of NA-high arousal, leads to inefficient processing and task execution, thereby impairing the encoding process and in turn increasing forgetfulness. The processing efficiency theory (Eysenck & Calvo, 1992) focuses on the role of anxiety in impairing the ability to process stimuli efficiently. The daily linkages between annoyed and irritated affective states and forgetting something may serve as a potential extension of this theory – that is, forgetfulness becomes a consequence of potentially inefficient processing. Additional work should explicitly test the ways in which high arousal anxiety-related affective states influence memory failures through the mechanistic pathway of inefficient processing. This is necessary to fully evaluate the utility of links between NA-high arousal and forgetting something within the tenets of processing efficiency theory. It was surprising that NA-low arousal was not associated with forgetting something. It is possible that daily influences of feeling depressed on memory could not be observed with concurrent reports, and may not emerge until the next day or subsequent days.

Further, feeling more sad or depressed than usual, markers of NA-low arousal, may usurp attentional resources that individuals need throughout the day, thereby leading to compromised focus or concentration. The resource allocation model (Ellis & Ashbrook, 1988) suggests that a low arousal depressed mood state can affect the allocation of attentional processing resources, which can then affect the encoding process of memory. Perhaps the days when sad or depressed affective states are worse than normal are when memory functioning is at greatest risk due, in part, to less focus or concentration and a potentially disrupted encoding process. This is consistent with well-known accounts of the explanatory roles of components of executive function in cognitive aging outcomes (e.g., Park et al., 1996), but what is unique in this study is that this relationship holds on the intraindividual level as well as the between-person level.

It was surprising that NA-high arousal was not associated with trouble staying focused. Perhaps the marker of attention-related complaints did not adequately capture anxiety-provoked inefficient processing purported by Eysenck and Calvo (1992). Further, annoyance and irritation may be less relevant high-arousal states relative to other NA-high arousal items that are more consistent with anxiety-related symptomatology (e.g., anxious, panic) but were not examined in this study.

Both PA-high arousal and PA-low arousal were associated with a lower likelihood of reporting trouble staying focused within-persons over time. Consistent with broaden-and-build theory (Frederickson & Branigan, 2005), it is possible that feeling more interested or excited than usual motivates the learning process and augments intellectual resources, thus lowering the likelihood of attentional lapses. Perhaps the days when participants are particularly interested and excited are when attentional lapses can be minimized due in part to heightened focus or concentration on something they find appealing.

The personality moderation suggests that feeling happier and more content and warm hearted than usual can be an adaptive correlate of attention-related complaints for those with relatively higher levels of conscientiousness. Perhaps these individuals engage in effective strategies to stave off attentional lapses because previous research suggests conscientiousness may be linked to decreased cognitive difficulties through the engagement of behavioral strategies (Hertzog & Pearman, 2013).

Differential Findings for Memory-Related and Attention-Related Complaints

The majority of subjective cognition research has focused on memory. Rabin et al.’s (2015) systematic review of 640 cognitive self-report items from 34 measures found 59% of the items were memory-related (short-term, episodic, semantic, prospective memory). The current study’s use of the ‘forgot something’ complaint contributes to this burgeoning literature and supports prior work that links subjective memory with general NA and PA (Drogos et al., 2013), anxiety and depressive symptoms (Comijs et al., 2002; Hülür et al., 2014), and exposure to daily stressors (Stawski, Mogle, & Sliwinski, 2013). Depressive symptoms have been identified as a correlate of subjective memory within-persons across nine occasions over up to 17 years (Hülür et al., 2014) and across three occasions over six years (Comijs et al., 2002). Both of these studies utilized the Center for Epidemiologic Studies Depression Scale (CES-D; Radloff, 1977) to assess depressive symptomatology experienced in the past week. However, associations between feeling annoyed or irritated and reports of forgetting something across 100 days in the current study indicate that these mental health-subjective memory linkages operate on a comparatively micro timescale as well.

Comparatively less research has targeted subjective attention, as Rabin et al. (2015) reported that only 11% of the 640 cognitive self-report items assessed from the literature were attention-related (e.g., sustained attention, focused attention, concentration, working memory, and processing speed). To the best of our knowledge, our study is the first empirical examination of daily linkages between arousal-specific affect and an attention-related complaint. The complex pattern of associations with NA and PA suggest three promising areas for future inquiry: 1) elucidating days that confer greatest protection from or risk of trouble staying focused (e.g., when PA-low arousal and NA-low arousal are higher than normal, respectively); 2) identifying individuals at greatest risk of trouble staying focused (e.g., individuals high in NA-low arousal); and 3) clarifying the moderating characteristics that amplify daily benefits of PA-low arousal (e.g., highly conscientious individuals).

Value of Microlongitudinal Design for Personalized Approaches to Cognitive Health

Utilizing an IIV approach and leveraging the microlongitudinal design of PULSE allowed for assessment of both within- and between-person associations of affect and cognitive complaints across 100 days. The interval-contingent data collection every evening for up to 100 days was an optimal design for assessing continuous phenomena like affect (Cerino & Hooker, 2019). The multilevel framework and moderation analyses suggests that the likelihood of reporting memory or attentional lapses depends on for whom and when affect is higher (or lower). For any individual, daily ebbs and flows in experiences of emotion can compromise and facilitate daily subjective cognitive health. Days when one feels sadder or more worried than usual may compromise the ability to concentrate. Similarly, days when one feels more annoyed or irritated than usual may compromise the ability to remember something. In addition to deleterious influences of negative experiences of emotions, days of greater interest or energy than usual may promote better concentration. Regardless of average levels of affect, being aware of how one feels throughout the day, and how those feelings might covary with cognition, may help individuals identify ways to experience cognitively sharper days and work toward a lifestyle that prioritizes cognitive health.

Examining environmental and lifestyle characteristics (e.g., social interactions, volunteerism, work) is an important next step in determining potentially modifiable factors in day-to-day life that can promote adaptive influences of affect. Leveraging daily variability in affect is integral to promoting subjective cognitive health. For example, identifying what it is that individuals find particularly interesting or annoying on a given day could prompt ways to heighten or suppress those experiences to potentially reduce cognitive complaints. It may also identify opportunities to promote coping strategies (e.g., passive appraisals that convert conflicts into lessons; Diehl, Coyle, & Labouvie-Vief, 1996) that could potentially mitigate the detrimental influences of NA and maximize facilitative influences of PA for subjective cognition.

Our results may also have important implications for personalized approaches to optimize cognitive health in older adulthood (Hamburg & Collins, 2010; Hill & Payne, 2017). Efforts toward mitigating forgetfulness and trouble staying focused by reducing levels of NA may benefit not only from targeting those with higher average levels of NA, but all older individuals, when they feel more annoyed or irritated, and sadder or more depressed, respectively. Further, resources dedicated toward improving daily feelings of happiness, contentment, and warm heartedness may be best allocated to those relatively high in conscientiousness. The current study helps offer a means to maximize resource allocation and personalized health efforts by pinpointing for whom and on which days boosting PA and/or reducing NA may both serve as pathways to mitigate daily cognitive complaints. Future research examining the temporality of these linkages is needed, however, to adequately and more broadly inform personalized interventions (Hill & Payne, 2017) that can be designed and implemented early on, when cognitive intervention efforts are most viable (Prince et al., 2011).

Limitations and Future Directions

Several limitations of this study should be considered. Analyses were correlational, which prevent any causal or directional conclusions concerning affect-cognitive complaint linkages. Future work should assess lagged effects of affect subscales to determine if increases in affect on day one lead to increased or decreased reports of cognitive complaints the next day or subsequent days. The potential bidirectionality of these associations should be assessed to determine if complaints influence affect (e.g., becoming more annoyed as a result of forgetting something). Indeed, Mogle et al. (2019) reported in supplementary analyses that a prospective memory lapse on one day was related to an increase in NA the next day. Examining lagged effects could also clarify the ways in which trouble staying focused and forgetting something work together in affect-subjective cognition associations. Might the daily link between affect and trouble staying focused precede a report of forgetting something? If so, it may be important to achieve greater understanding of how subjective reports of attentional and memory lapses align with potentially disrupted encoding processes.

There is mixed evidence on the relationships between self-reported and performance-based cognition, with some support for small-to-moderate associations within- and between-persons (Hülür et al., 2014). The lack of objective measures in the PULSE protocol, however, prevented explicit comparisons and underscores the need for future research to include both self-reported and performance-based measures. Finally, the interval-contingent data collection that occurred every evening for 100 days provided important information on the participant’s reflection of their affect and cognitive complaints throughout the day and offered an appropriate amount of occasions to obtain reliable estimates of dynamic affect (Mejía et al., 2014). The design does not, however, elucidate the instantaneous affective reaction or influence on cognitive complaints at a certain point in time. With the current study as a foundation, future research should extend these daily linkages toward an ecological momentary assessment space with immediate, momentary reports of how one feels at an exact point in time to avoid potential recall bias, improve ecological validity, and harness greater moment-to-moment precision of associations among affect and cognitive complaints (Naragon-Gainey, McMahon, & Park, 2018).

Conclusion

Using an IIV approach, arousal-specific NA and PA were associated with attention-related and memory-related complaints both within-persons over time and between-persons for a sample of community-dwelling older adults. Associations differed depending on affective arousal, individual differences in conscientiousness, and domain of cognitive complaints. The results suggest efforts focused on maximizing resource allocation and personalizing cognitive health should consider for whom and on which days boosting PA and/or reducing NA may both serve as pathways to benefit daily subjective cognition.

Appendix A: Supplementary Information on Categorical Covariate Marital Status

Appendix A Table 1:

Descriptive Statistics across Marital Statuses

Variable M(SD)	Married (71.72%)	Divorced/Separated (10.10%)	Single (4.04%)	Widowed (12.12%)	Other (2.02%)	p
NA-High Arousal	20.10(15.65)	17.77(14.36)	10.16(4.34)	17.88(13.85)	25.00(29.97)	.72
NA-Low Arousal	29.67(25.34)	27.75(29.61)	13.34(14.47)	24.42(18.84)	35.24(48.63)	.73
PA-High Arousal	69.51(15.63)	72.67(15.74)	73.71(18.56)	68.44(13.67)	66.49(16.13)	.94
PA-Low Arousal	101.95(24.05)	107.88(24.37)	105.82(29.44)	101.66(18.22)	101.38(28.68)	.96
Trouble	0.10(0.17)	0.06(0.06)	0.14(0.18)	0.07(0.15)	0.00(0.00)	.79
Forgot	0.06(0.13)	0.04(0.04)	0.10(0.13)	0.13(0.25)	0.01(0.01)	.48
N	48.31(13.89)	50.47(11.11)	53.46 (16.76)	43.38(7.75)	38.12(12.08)	.45
C	47.14(11.55)	43.70(13.82)	42.07(9.89)	47.05(12.48)	43.88(3.67)	.83
Age	61.89(6.98)	63.30(5.17)	59.50(6.03)	73.50(8.97)	57.00(2.83)	<.001
Health	2.38(0.66)	2.40(0.52)	2.50(0.58)	2.67(0.49)	2.50(0.71)	.70
Symptoms	0.83(0.95)	0.94(0.67)	0.78(0.58)	0.93(0.78)	0.51(0.69)	.97
Female (%)	84.51	100.00	75.00	100.00	100.00	.32
White (%)	98.59	90.00	75.00	100.00	100.00	.06
College Degree (%)	83.10	70.00	75.00	58.33	100.00	.31

Note. N=99. NA=negative affect. PA=positive affect. Trouble=trouble staying focused or concentrating. Forgot=forgot something. N=neuroticism. C=conscientiousness. Health=self-rated general health. Symptoms=physical symptoms. College Degree=percent of sample who graduated with at least a 4 year college degree. Mean age for participants who identified as widowed was significantly older than those who identified as married, divorced/separated, single, or other (all ps<.05).

Appendix B: Assessing Collinearity

Collinearity is a concern for the simultaneous prediction models distinguishing by affective arousal given moderate to strong correlations between NA-high and NA-low arousal and PA-high and PA-low arousal at the level of individual differences (rs ranged from .83–.94, ps<.001; see Table 1) and within-persons over time (rs ranged from .51–.65, ps<.001; see Table 2).

Appendix B Tables 1 and 2 provide matrices of generalized MLMs predicting cognitive complaints with both independent and simultaneous inclusion of affect variables to assess the extent to which collinearity may have led to unstable odds ratios. Albeit non-significant, the odds ratio for between-person NA-high arousal predicting trouble staying focused changed from >1.00 to <1.00 when NA-high and NA-low arousal variables and any PA variables were simultaneously included in the model (Appendix B Table 1, Models 2B, 2G, 2H). The non-significant odds ratio for between-person PA-high arousal predicting trouble staying focused changed from <1.00 to >1.00 when PA-high and PA-low arousal variables were simultaneously included in the model (Appendix Table 1, Models 2B, 2D). The non-significant odds ratio for within-person PA-low arousal predicting trouble staying focused changed from <1.00 to >1.00 only when NA-high and NA-low arousal and PA-high and PA-low arousal variables were simultaneously included in the model (Appendix B Table 1, Model 2B).

Albeit non-significant, the odds ratio for between-person NA-low arousal predicting forgot something changed from >1.00 to <1.00 when NA-high and NA-low arousal variables were simultaneously included in the model (Appendix B Table 2, Models 2B, 2C). Further, the non-significant odds ratio for within-person PA-high arousal predicting forgot something changed from <1.00 to >1.00 when PA-high and PA-low arousal variables were simultaneously included in the model (Appendix B Table 2, Models 2B, 2D, 2H). The non-significant odds ratio for within-person PA-low arousal predicting forgot something changed from <1.00 to >1.00 only when within-person PA-high and PA-low arousal variables and NA-high and NA-low arousal variables were simultaneously included in the model (Appendix B Table 2, Model 2B, 2H).

CFAs assessing theoretical distinction by arousal

Appendix B Table 3 provides global fit indices from a series of CFAs to formally evaluate the capacity for high and low arousal affect factors in this sample. For NA, a 2-factor within-person (high arousal, low arousal) and 2-factor between-person (high arousal, low arousal) factor structure demonstrated best model fit, χ² (8) = 63.00, p < .001; RMSEA = .03, CFI = .98; TLI = .96. For PA, a similar 2-factor within-person and 2-factor between-person factor structure demonstrated good model fit, χ² (8) = 53.21, p < .001; RMSEA = .03, CFI = .99; TLI = .97. However, a slightly more parsimonious 2-factor within-person and 1-factor between-person factor structure also demonstrated good model fit, χ² (9) = 57.64, p < .001; RMSEA = .03, CFI = .99; TLI = .97. A likelihood ratio test comparing the PA 2-factor within-person / 2-factor between-person model with the PA 2-factor within-person / 1-factor between-person model revealed a significant difference in model fit reducing the between-person factor structure to 1, Δχ2 (1) = 4.44, p < .05, ΔCFI < .001.

Sensitivity analyses using the parsimonious model (Appendix B Table 1, Model 2H; Appendix B Table 2, Model 2H) showed that results remained unchanged when the between-person PA variable is included in the model as a single factor or two factors distinguished by arousal. Taken together with the magnitude of correlations between NA-high and NA-low arousal and PA-high and PA-low arousal, there appears to be some degree of collinearity resulting in weak, non-significant odds ratios to change direction. However, theoretical precedent, good model fit, and results that were robust to sensitivity analyses suggest 2-factor within-person and 2-factor between-person affect variables can be simultaneously included in predictive models of trouble staying focused and forgetting something in this sample.

Appendix B Table 1.

Matrices of Logistic Generalized MLMs for High and Low Arousal Affect Predicting Odds of Trouble Staying Focused

	Independent Inclusion						Simultaneous Inclusion
Variables	1A	1B	1C	1D	1E	1F	2A	2B	2C	2D	2E	2F	2G	2H
WP NA	1.02***						1.01***
BP NA	2.52***						2.20*
WP NA-High Arousal		1.02***						1.00	1.01**		1.01***		1.00	1.00
BP NA-High Arousal		2.17***						0.90	1.00		1.74*		0.92	0.93
WP NA-Low Arousal			1.03***					1.02***	1.03***			1.02***	1.02***	1.02***
BP NA-Low Arousal			2.49***					2.34*	2.50**			2.11**	2.32*	2.34***
WP PA				0.98***			0.98***						0.98***
BP PA				0.43***			0.80						0.79	0.80
WP PA-High Arousal					0.95***			0.96*		0.96***	0.95***			0.96***
BP PA-High Arousal					0.46***			1.45		1.21	0.67
WP PA-Low Arousal						0.96***		1.00		0.98***		0.98***		0.99
BP PA-Low Arousal						0.43***		0.54		0.35†		0.77

Note. WP=within-person. BP=between-person. NA=negative affect. PA=positive affect. Estimates are odds ratios from models adjusting for linear trend, age, gender, marital status, self-rated health, and daily physical symptoms.

^†p<.10

^*p<.05.

Appendix B Table 2.

Matrices of Generalized MLMs for High and Low Arousal Affect Predicting Odds of Forgetting Something

	Independent Inclusion						Simultaneous Inclusion
Variables	1A	1B	1C	1D	1E	1F	2A	2B	2C	2D	2E	2F	2G	2H
WP NA	1.01***						1.01***
BP NA	1.87**						1.51
WP NA-High Arousal		1.01***						1.01**	1.01**		1.01***		1.01**	1.01**
BP NA-High Arousal		2.04***						2.23*	2.39*		1.78*		2.19*	2.19*
WP NA-Low Arousal			1.01***					1.00	1.00			1.01†	1.00	1.00
BP NA-Low Arousal			1.37*					0.73	0.82			1.14	0.73	0.73
WP PA				0.99**			1.00						1.00
BP PA				0.53**			0.73						0.75	0.74
WP PA-High Arousal					0.99			1.00		1.00	1.00			1.00
BP PA-High Arousal					0.55**			0.70		0.74	0.80
WP PA-Low Arousal						0.99**		1.00		0.99**		0.99		1.00
BP PA-Low Arousal						0.66**		1.08		0.69		0.74

Note. WP=within-person. BP=between-person. NA=negative affect. PA=positive affect. Estimates are odds ratios from models adjusting for linear trend, age, gender, marital status, self-rated health, and daily physical symptoms.

^†p<.10

^*p<.05.

Appendix B Table 3.

Global Fit Indices from Confirmatory Factor Analyses Assessing Factor Structure of Affect

Affect	Factors	χ2	χ2/df	CFI	TLI	RMSEA	AIC	BIC
NA	1 WP / 1 BP	2638.90***	263.89	0.16	−0.67	0.19	239446.91	239618.48
	2 WP / 1 BP	310.78***	34.53	0.90	0.79	0.07	236862.17	237040.60
	1 WP / 2 BP	1373.61***	152.62	0.57	0.04	0.15	239174.64	239353.07
	2 WP / 2 BP	63.00***	7.88	0.98	0.96	0.03	236596.30	236781.59
PA	1 WP / 1 BP	101.16***	10.12	0.98	0.95	0.04	230396.14	230567.71
	2 WP / 1 BP	57.64***	6.40	0.99	0.97	0.03	230294.27	230472.70
	1 WP / 2 BP	96.73***	10.75	0.98	0.95	0.04	230385.69	230564.12
	2 WP / 2 BP	53.21***	6.65	0.99	0.97	0.03	230284.64	230469.93
NA 2 WP / 2 BP PA 2 WP / 2 BP		304.41***	5.25	0.98	0.97	0.03	463204.46	463629.95
NA 2 WP / 2 BP PA 2 WP / 1 BP		314.90***	5.16	0.98	0.97	0.02	463211.08	463615.97
LRT		Δχ2 (1) = 10.49, p < .05, ΔCFI = .001.

Note. WP=within-person. BP=between-person. NA=negative affect. PA=positive affect. LRT=likelihood ratio test. A likelihood ratio test comparing the PA 2 WP / 2 BP model with the PA 2 WP / 1 BP model revealed a significant difference in model fit reducing the between-person factor structure to 1, Δχ² (1) = 4.44, p < .05, ΔCFI < .001.