Intraindividual Coupling of Daily Stressors and Cognitive Interference in Old Age

Abstract

Objectives.

The current study examined emotional and cognitive reactions to daily stress. We examined the psychometric properties of a short cognitive interference measure and how cognitive interference was associated with measures of daily stress and negative affect (NA) between persons and within persons over time.

Methods.

A sample of 87 older adults (M_age = 83, range = 70–97, 28% male) completed measures of daily stress, cognitive interference, and NA on 6 days within a 14-day period.

Results.

The measure yielded a single-factor solution with good reliability both between and within persons. At the between-person level, NA accounted for the effects of daily stress on individual differences in cognitive interference. At the within-person level, NA and daily stress were unique predictors of cognitive interference. Furthermore, the within-person effect of daily stress on cognitive interference decreased significantly with age.

Discussion.

These results support theoretical work regarding associations among stress, NA, and cognitive interference, both across persons and within persons over time.

THE ability to keep unwanted thoughts out of mind is a central construct in different domains of psychology. Individual differences in thought control are related to higher order cognitive functions (for review, see Engle & Kane, 2004) and physiological and emotional reactions to stressful experiences (Brosschot, Gerin, & Thayer, 2006; Moberly & Watkins, 2008a, 2008b). Furthermore, the ability to inhibit intrusive mental content has been hypothesized to decrease with age, and the experience of intrusive thoughts has been proposed as a primary explanation for cognitive aging (Hasher & Zacks, 1988).

Cognitive interference has been used as an index of one's ability to control and censor their mental contents and has been linked to poorer cognitive function (Brewin & Beaton, 2002; Klein & Boals, 2001; Stawski, Sliwinski, & Smyth, 2006), physiological reactivity to stress (Brosschot et al., 2006; Zoccola, Dickerson, & Zaldivar, 2008), anxiety, depression, and poor physical health (Watkins, 2008). Much of the previous research, however, has focused on global assessments and individual differences in cognitive interference (CI), with little empirical attention being devoted to naturally occurring short-term intraindividual variability in CI and its potential contextual correlates. The current study was conducted to accomplish two aims. First, we sought to develop a short measure of CI for use in intensive repeated measures studies, such as daily diary designs. Second, we examined whether daily stressors were associated with CI between persons, as well as within persons across days among older adults who are potentially susceptible to the experience of CI.

CI is defined as the experience of intrusive, off-task thoughts, and the intentional suppression of and failure to suppress such thoughts, which interfere with otherwise normal everyday goal/task-oriented thinking. These thoughts can stem from stressful experiences, recur in the absence of the actual stressor, and impair the ability to concentrate (Klein & Boals, 2001; Sarason, Pierce, & Sarason, 1996; Stawski et al., 2006). CI is similar to repetitive thought, rumination, worry, and perseverative cognition (for review, see Watkins, 2008) and is characterized by thoughts that are relatively brief, sudden, unexpected, and less clearly catalyzed by a particular precipitating event (Clark, 2004). The content of CI can be both related (e.g., thinking about the quality of one's performance) and unrelated (e.g., thinking about an argument one had earlier in the day) to the current task performance (Sarason, Sarason, Keefe, Hayes, & Shearin, 1986), and CI is known to be influenced by both dispositional and situational characteristics (Pierce et al., 1998; Sarason et al., 1986, 1996).

Many existing measures of CI treat the construct as a dispositional characteristic, assessing the types of thoughts people have (e.g., Thought Occurrence Questionnaire, Sarason et al., 1986; White Bear Suppression Inventory; Wegner & Zanakos, 1994), and what they do to control their unwanted intrusive thoughts (e.g., Thought Control Questionnaire; Wells & Davies, 1994). Situational measures of CI do exist but are highly context specific, assessing CI with respect to a specific set of tasks (e.g., Cognitive Interference Questionnaire, Sarason et al., 1986) or a specific experience from one's past (e.g., Impact of Events Scale, Horowitz, Wilner, & Alvarez, 1979). Furthermore, little research exists demonstrating the utility of assessing CI in intensive repeated measures studies or whether CI exhibits considerable within-person variability. Thus, a measure that is not context or situation specific but is sensitive to these forces would be of considerable utility researchers aiming to link CI to variables of theoretical interest within persons over time.

One construct previously linked to CI is stress. Stressful experiences have important consequences for emotional and physical health and well being (Baum & Posluzny, 1999; Lazarus, 1999; Zautra, 2003). These consequences are thought to be particularly evident during old age (Smith, 2003). In addition to the well-documented emotional and physical consequences of stressful experiences, some theorists have suggested that CI is a normative reaction to stressful events (Horowitz, 1975). Indeed, previous research has shown that stressful life events including illness, combat exposure, natural disaster, and bereavement/loss all elicit CI (for review, see Sundin & Horowitz, 2003). Experimental studies have shown that stressful stimuli such as an emotionally negative film clip (Horowitz, 1975) or public speaking (Stawski, Sliwinski, & Smyth, 2009) elicit CI. In contrast to research on major life events or the utilization of experimentation, less is known about how naturally occurring daily stressors may be related to CI.

Although major life events and traumas are relatively rare, daily stressors occur more frequently but can still carry emotional and physical consequences (Almeida, 2005; Lazarus, 1999; Zautra, 2003). Daily stress research often employs intensive repeated assessments designs allowing for the examination of covariation between variables within persons over time (Bolger, Davis, & Rafaeli, 2003). Such designs are ideal for examining within-person variability in constructs such as CI and determining whether this variability is temporally yoked to daily stressful experiences. Previous research linked the experience of daily stressors to increases in negative affect (NA; e.g., Bolger & Schilling, 1991; Stawski, Sliwinski, Almedia, & Smyth, 2008) and self-reported physical symptoms (e.g., Almeida & Horn, 2004; Grzywacz, Almeida, Neupert, & Ettner, 2004). Research linking daily stress to CI, however, is scant.

Wegner (1988) found that daily overall stress levels were positive related to reports of intrusive thoughts, whereas Moberly and Watkins (2008a, 2008b) observed a positive association between daily NA and ruminative self-focus. Although previous research suggests that reliable within-person variability in CI may exist, it has been linked to subjective stress levels and mood states, not the explicit experience of a stressful event. Thus, it is unclear whether observed day-to-day associations reflect a unique association between stressful experiences and CI or simply the extent to which one is thinking about their NA. For example, it may be that CI has little to do with the experiences of stressors per se and is instead a consequence of ruminations regarding NA (e.g., Nolen-Hoeksema, 1991). It could also be the case that CI is a normative cognitive reaction to the experience of stressors that is in part independent of negative emotions (e.g., Horowitz, 1975).

The ability to inhibit intrusive thoughts and irrelevant information declines with age (Hasher & Zacks, 1988). Additionally, aging is thought to be associated with greater susceptibility to the biological, psychological, and cognitive effects of stress (e.g., Smith, 2003). Previous research on daily stress with respect to cognition has shown older adults report increased memory failures associated with the experience of daily stressors (Neupert, Almeida, Mroczek, & Spiro, 2006), whereas Sliwinski, Smyth, Hofer, and Stawski (2006) showed that performance on attentionally demanding cognitive tasks was worse on stressor days compared with stressor-free days, and these performance decrements were larger in older relative to younger adults. Thus, previous research supports the notion that cognition is susceptible to the effects of daily stressors among older adults, and older adults may be more vulnerable to the effects of daily stressors on cognition. Consistent with this previous research and theory, increased age may be expected to be associated with higher levels of CI and larger daily stressor-related increases in CI, however, little empirical evidence exists to support this.

The current study addressed two aims. First, from a psychometric perspective, we examined the psychometric properties of this measure for use in intensive repeated measures studies (e.g., daily diary designs). Second, from a substantive perspective, we examined whether daily stressors were associated with cognitive interference at two levels of analysis, between persons and within persons across days, and that these associations were independent of physical symptoms and NA. Specifically, we predicted that individuals who reported more frequent daily stressors would also report higher levels of CI, and CI would be higher on days individuals reported experiencing daily stressors compared with stressor-free days. Finally, we explored the hypothesis that increased age would be associated with greater vulnerability to the effects of daily stressors on cognitive interference by testing age as a moderator of the within-person association between daily stressors and CI.

METHODS

Participants

Eighty-seven older adults were recruited from the greater Syracuse area, as well as a senior residence facility. The average age of the sample was 82.59 (SD = 6.06, range = 70–97; 72% female). The sample was highly educated (M = 15.08 years, SD = 2.45, range = 8–22).

Materials

Cognitive interference was assessed using six items, constituting the short cognitive interference measure (SCIM; see Table 1), which were adapted from existing measures of CI: White Bear Suppression Inventory (Wegner & Zanakos, 1994), the Thought Control Questionnaire (Wells & Davies, 1994), Impact of Events Scale (Horowitz et al., 1979), and Thought Occurrence and Cognitive Interference Questionnaires (Sarason et al., 1986). Participants indicated what best described their thoughts “today, before you came to this session.” Responses were made on a 5-point scale (Never = 1, Once = 2, A Few Times = 3, Often = 4, Very Often = 5). Questions were selected and adapted based on two primary criteria. First, we sought questions that would be sensitive to and exhibit variation when administered repeatedly over shorter periods of time (e.g., hours or days). Second, we excluded questions which explicitly asked about thoughts regarding emotional or physical states to avoid potential confounds between thoughts and feelings.

Table 1.

Variance Components Extracted in G-theory Analysis

Variance source	Estimate	% of total variance
Item	0.05	4.04
Day	0.04	3.26
Person	0.32	26.30
Day × person	0.22	18.02
Item × person	0.17	13.98
Item × day	0.00	0.00
Error	0.41	34.41
Total	1.20	100.00

Daily stressors were assessed using the Daily Inventory of Stressful Events (DISE: Almeida, Wethington, & Kessler, 2002). The DISE is a semistructured interview in which participants report whether any of a series of events had occurred within the past 24 hr. This version of the DISE included questions assessing the experience of arguments, health-related problems, and events related to individuals in the respondents social network and has been described in more detail elsewhere (Sliwinski et al., 2006; Stawski et al., 2008). For the purposes of the current study, a dichotomous measure was used indicating whether any stressor had occurred as we were most interested in establishing the yoking of cognitive interference ratings to the reported occurrence of stressors.

Daily NA was assessed using the Philadelphia Geriatric Center Negative Affect Scale of Lawton, Kleban, Dean, Rajagopal, and Parmelee, (1992). Five items (sad, annoyed, worried, irritated, and depressed) measured NA, and participants indicated the extent to which they were experiencing each of the adjectives on a 5-point Likert Scale (not at all, a little, moderately, quite a bit, extremely), right now, at this very moment. A total score was obtained by summing responses across all items with higher scores reflecting greater levels of NA.

Daily physical symptoms were measured using Larsen and Kasimatis (1991) physical symptom checklist. This checklist assessed five constellations of symptoms: aches/pain, gastrointestinal symptoms, symptoms associated with cardiovascular functioning, upper respiratory symptoms, and a category for “other” physical symptoms or discomforts. At each session, participants indicated whether they had experienced each symptom over the past 24 hr. The total number of symptoms reported was used as an index of physical symptoms.

Procedure

Participants came to a research laboratory on six occasions over a 14-day period. Half of the assessments took place in the morning (before 11:00 a.m.) and half in the afternoon (after 1:00 p.m.), and the same research assistant administered the protocol at each session. During each session, participants completed the cognitive interference, NA, and physical symptom measures, followed by the daily stress diary.

Analytic Strategy

The factor structure of the CI items was examined using multilevel confirmatory factor analysis with MPLUS v5 (Muthén & Muthén, 2007). The factor structure of the items is assessed both across time (within-person structure) and across people (between-person structure), yielding factor loadings and fit indices for the factor solutions for characterizing the CI items across people and time.

Estimates of reliability at the between- and within-person levels of analysis were obtained following methods outlined by Cranford and colleagues (2006). Adequate between-person reliability would indicate that the scale was able to differentiate reliable variability between individuals and stable individual differences in CI. Within-person reliability would indicate that the scale captured systematic variance within individuals across days. We used Generalizability theory (G-theory) to partition the variance in items in the daily measure of CI into that related to items, persons, days, and the two-way interactions of each (Table 1). Any variability not due to these sources is considered error variability.

Per Cranford and colleagues (2006), between-person reliability was calculated as the percentage of variability due to persons and the person by item interaction (indicating that individuals may respond differently to different items) divided by those terms plus the estimated variability related to error (Equation 1). (Note that these reliability estimates assume that individuals were measured on the same fixed day, and items were treated as fixed effects across days.)

(1)

Within-person reliability was calculated as the person by day interaction divided by this same term plus the error variance (Equation 2).

(2)

This coefficient quantifies whether there is reliable interindividual variability in intraindividual change across days. In the above equations, i refers to the number of items.

For examining the effects of daily stressors on cognitive interference, we employed multilevel models (Snidjers & Bosker, 1999) using SAS PROC MIXED, as this methodology allows us to examine associations between daily stress and CI at the within- and between-person levels (Levels 1 and 2, respectively). The following model was used:

(3)

where, at Level 1, b_0j is the cognitive interference score for session i and person j. Parameters b_1j and b_2j are linear and quadratic growth parameters accounting for changes in cognitive interference across the study sessions which could potentially bias the effects of time-varying/within-person covariates (cf. Sliwinski et al., 2006). The parameter b_3j reflects the within-person daily stress effect. Because daily stressors were coded dichotomously (0 = no stressors reported, 1 = any stressors reported), b_3j indicates the difference in cognitive interference on stressor days compared with stressor-free days and is the critical estimate of the within-person association between daily stress and cognitive interference. β₀₀ and β₃₀ are the sample average within-person intercept and daily stress effects (i.e., fixed effects), and β₁₀ and β₂₀ are the sample average linear and quadratic session trends and are assumed to be invariant across individuals. β₀₁ and β₀₂ are Level 2 effects and reflect age and sex differences in the average levels cognitive interference, respectively. β₀₃ is the context effect of daily stress, reflecting each individual's average proportion of stressor days. This estimate captures the difference between the within- and between-person daily stressor effects, and adding these together provides the between-person daily stressor effect and is the critical test of whether individual differences in the frequency of daily stressors are associated with cognitive interference. u_0j and u_3j are random effects allowing for individual differences in the average level of cognitive interference and within-person daily stressor effect, respectively. e_ij is the Level 1, or residual variance.

RESULTS

Structure of the SCIM

Table 1 presents descriptive statistics for the six items comprising the SCIM. Average values for the individual items were low indicating that participants reported these types of events to be relatively infrequent occurrences. The between- and within-person standard deviations, however, show that there was considerable variation both across individuals and across days for a given individual. The intraclass correlation coefficient, an index of the total variability attributable to individual differences, for the items ranged from 0.32 to 0.48 indicating that between 32% and 48% of the variability for a given item reflected differences across individuals. Thus, a majority of the variability for scores on each item (52%–68%) was observed within persons across days.

Next, we submitted the items to a multilevel confirmatory factor analysis to test a single-factor structure of the items at the between- and within-person levels. This initial model had an acceptable fit, χ²(19) = 89.08, p < .01, comparative fit index (CFI) = 0.89, root mean square error of approximation (RMSEA) = 0.08, standardized root mean residual (SRMR)_within = 0.05, SRMR_between = 0.10 (for factor loadings, see Table 1). The solution accounted for 83% and 42% of the variance at the between- and within-person levels, respectively. Modification indices suggested that allowing the residual variances between Items 3 and 4 at the within-person level to be correlated would significantly improve model fit. Adding this parameter did improve model fit, χ²(19) = 36.96, p = .01, CFI = 0.97, RMSEA = 0.05, SRMR_within = 0.03, SRMR_between = 0.07 but had no effect on the factor structure as the single-factor solution at each levels of analysis remained, and the factor loadings were virtually identical. A more restrictive model constraining the factor loadings to be equivalent across items and levels (i.e., tau-equivalent model; Raykov, 1997) provided good absolute fit to the data, χ²(26) = 73.57, p < .01, CFI = 0.92, RMSEA = 0.06, SRMR_within = 0.05, SRMR_between = 0.09, however, a chi-square difference test indicated that constraining the loadings led to a significant reduction in model fit, ${\text{χ}}_{\text{diff}}^2$(7) = 36.61, p < .01.

Given the single-factor solutions at each level of analysis, we then estimated the reliability of the scale using methods recommended by Cranford and colleagues (2006; see Equations 1 and 2 and variance estimates in Table 1), distinguishing reliability across individuals (between person) and across time (or days; within person). The reliabilities were 0.83 and 0.76 at the between- and within-person levels, respectively, indicating that the scale is useful for discriminating average levels of CI between persons, as well as discriminating between levels of CI among assessment occasions within a given person over time.

Daily Stressors and Cognitive Interference

Next, we used multilevel models to examine the effects of daily stressors on CI, using sum scores of the items. Although the results we present examining associations between daily stress and cognitive interference were obtained using traditional linear multilevel models (SAS PROC MIXED) with the sum score of the CI items as our dependent variable, we also estimated all reported models within a multilevel structural equation framework (MPLUS v5), whereby latent factors reflecting the common variance among the CI items at both levels of analysis served as the dependent variable. This later approach allowed us to consider our substantive objectives within the same modeling framework used to demonstrate the psychometric properties of the cognitive interference scale. Although there were differences in the absolute values of the estimates obtained from the two analytic approaches, all of the substantive results, significant effects, and the conclusions that we draw from them were the same. As such, we have opted to only report the results using the traditional multilevel modeling approach, but the results utilizing the multilevel structural equation modeling approach are available from the authors. Model 1 (Table 3) tested an empty model which decomposed the variability in daily CI. Based on this model, approximately 51% of the variability in CI reflects individual differences. Next, we tested whether CI varied as a function of daily stressors at both the between- and within-person levels. These models included age and sex as covariates, as well as linear and quadratic trends to control for systematic trends across study days, as suggested by Sliwinski and colleagues (2006). During preliminary analyses, we included the within-person daily stressor effect as a random effect to test for individual differences in the magnitude of the daily stressor–CI relationship, however, the effect was not significant (estimate = 1.41, SE = 1.55, n.s.) and was excluded from the final models. A likelihood ratio test of nested models confirmed that excluding this random slope and covariance between the random intercept and slope did not substantially alter model fit, ${\text{χ}}_{\text{diff}}^2$(2) = 1.20, p = .55. The results of this model can be seen in Table 2 (Model 2).

Table 3.

Estimates for Multilevel Models Examining the Effects of Daily Stressors on Cognitive Interference

	Model 1	Model 2	Model 3	Model 4
	Estimate (SE)	Estimate (SE)	Estimate (SE)	Estimate (SE)
Fixed effects
Intercept	10.80 (0.40)**	12.26 (0.54)**	12.37 (0.48)**	12.32 (0.53)**
Session	—	−1.89 (0.29)**	−1.78 (0.27)**	−1.90 (0.29)**
Session × session	—	0.28 (0.06)**	0.27 (0.05)**	0.28 (0.06)**
Age	—	0.11 (0.06)+	0.05 (0.06)	0.08 (0.06)
Sex	—	−0.16 (0.83)	−0.04 (0.73)	−0.36 (0.81)
Daily stress (WP)	—	1.57 (0.36)**	0.87 (0.35)**	1.61 (0.36)**
Daily stress (BP)	—	5.30 (1.17)**	1.79 (1.24)	4.34 (1.23)**
Negative mood (WP)	—	—	0.54 (0.07)**	—
Negative mood (BP)	—	—	0.92 (0.18)**	—
Physical symptoms (WP)	—	—	—	−0.21 (0.17)
Physical symptoms (BP)	—	—	—	0.48 (0.23)*
Random effects
Intercept	12.08 (2.15)**	9.62 (1.73)**	7.24 (1.33)**	9.07 (1.64)**
Residual	11.67 (0.81)**	9.78 (0.68)**	8.52 (0.59)**	9.75 (0.67)**

Note: +p < .10; *p < .05; **p < .01. BP = between person, WP = within person.

Table 2.

Descriptive Statistics and Factor Loadings for the items of the Short Cognitive Interference Measure

	Mean	SD (BP)	SD (WP)	ICC	Factor loadings
					BP	WP
Q1. Did you think about personal worries?	2.09	0.70	0.85	0.40	0.71	0.56
Q2. Did you have trouble concentrating?	1.64	0.55	0.81	0.32	0.86	0.50
Q3. Did you try to avoid certain thoughts?	1.66	0.65	0.83	0.38	0.99	0.61
Q4. Did you try to put problems out of your mind?	1.92	0.82	0.85	0.48	0.97	0.58
Q5. Did you think about something you didn’t mean to?	1.51	0.55	0.70	0.38	0.92	0.45
Q6. Did you have thoughts that kept jumping into your head?	1.98	0.81	0.86	0.47	0.81	0.61

Note: BP = between person; WP = within person; ICC = Intraclass Correlation Coefficient. Skew for all items were all below 1.65.

CI scores significantly decreased across the sessions (estimate = −1.89, SE = 0.29, p < .01), however, the rate at which they decreased did slow down across sessions (estimate = 0.28, SE = 0.06, p < .01). Additionally, older adults reported higher levels of CI (estimate = 0.11, SE = 0.06, p = .06). Importantly, the within- and between-person daily stressor effects were significant. Individuals reported higher levels of CI on days when they reported experiencing a stressor compared with stressor-free days (estimate = 1.57, SE = 0.36, p < .01). Similarly, individuals who reported experiencing stressors more frequently reported higher levels of CI (estimate = 5.30, SE = 1.17, p < .01).

Previous daily stress research has shown that NA (Stawski et al., 2008) and physical symptoms reports (Hoffman & Stawski, 2009) are higher on stressor days compared with non-stressor days. As such, we wanted to establish that the observed effect of daily stressors on cognitive interference was not attributable to variability in NA or physical symptoms. We estimated two additional models examining the magnitude of the daily stress effect on CI, after covarying for NA and physical symptoms at both levels of analysis. Time-varying and time-invariant indices of daily NA and physical symptoms were calculated and implemented using methods person-centering as outlined by Hoffman and Stawski. The between-person effect was obtained by calculating each individual's average NA score across the study days (Negative Affect_.j). The within-person effect was obtained by subtracting each individual's NA score on a given day from their average level of NA (Negative Affect_ij − Negative Affect_.j) and reflects solely within-person variation as the stable between-person variability has been subtracted out. Between- and within-person effects for physical symptoms were calculated the same way. Here, it is worth noting that when using person-centering, as we are doing with NA and physical symptoms, the person-mean directly reflects the between-person effect. In contrast, the person-mean estimate for daily stress, using grand-mean centering as we have done, reflects a context effect or the incremental influence of the person over and above the time-varying (Level 1) effect of daily stress.

Table 2 shows the results after covarying for NA (Model 3) and physical symptoms (Model 4) at the between- and within-person levels. We included the within-person NA and physical symptoms predictors as random effects, but preliminary analyses revealed that neither random effect was statistically significant and their exclusion did not affect model fit, thus, the final models reported only include fixed effects.

For the model adding NA (Model 3), the between-person effect was significant, indicating that individuals reporting higher levels of NA also report higher levels of CI (estimate = 0.92, SE = 0.18, p < .01), and the between-person daily stress effect was no longer significant. Similarly, the within-person effect was significant indicating that on days individuals reported their NA to be higher than usual, their CI was higher as well (estimate = 0.54, SE = 0.07, p < .01). The within-person effect of daily stress remained significant, but its estimate was reduced (estimate = 0.87, SE = 0.35, p < .01).

For the model adding physical symptoms (Model 4), the between-person effect of physical symptoms was significant, indicating that individuals reporting more physical symptoms also report higher levels of CI (estimate = 0.48, SE = 0.23, p < .05). The within-person association between physical symptoms and CI, however, was not statistically significant. The within-person daily stress effect remained significant and virtually unchanged (estimate = 1.61, SE = 0.36, p < .01), whereas the between-person daily stress effect was reduced but still significant (estimate = 4.34, SE = 1.23, p < .01). It is worth noting that the same pattern of results was obtained when estimating a model covarying for NA and physical symptoms simultaneously.

Age as a Moderator of the Daily Stress–Cognitive Interference Association

To test for age differences in the within-person association between daily stressors and CI, we added the age by within-person daily stress cross-level interaction to Equation 3. This interaction was significant (estimate = −0.12, SE = 0.06, p = .05) and indicated that the within-person daily stress effect decreased with age (Figure 1). The daily stress effect was larger and statically significant 1 SD below the mean for age (∼76 years old; estimate = 2.29, SE = 0.51, p < .01) and smaller and not significant 1 SD above the mean for age (∼89 years old; estimate = 0.85, SE = 0.51, p = .09). We also tested for age differences in the association between daily stress and cognitive interference at the between-person level, however, the interaction was not significant (p = .13).

Figure 1.

Within-person association between daily stress and cognitive interference as a function of age.

DISCUSSION

The results of the current study produced a number of findings. First, our short measure of CI displayed a single-factor structure and exhibited adequate reliability and significant variability between persons and within persons over time. Second, daily stress was significantly associated with CI both between persons and within persons over time; individuals reporting more frequent occurrence of daily stressors reported higher levels of CI and CI was significantly higher on stressor days compared with non-stressor days. Third, individual differences in daily NA were associated with higher levels of CI and accounted for the between-person effect of daily stress. At the within-person level, NA and daily stress were each uniquely associated with higher levels of CI. Covarying for daily physical symptoms did little to alter the association between daily stress and CI at either level of analysis. Finally, age significantly moderated the within-person effect of daily stress on CI such that increased age was associated with decreased reactivity.

Our findings regarding the psychometric properties of this short measure of CI are important for a number of reasons. Research employing intensive repeated measures designs (e.g., daily diary, experience sampling, ecological momentary assessment) can often rely on measures developed for assessing stable individual differences or use a single item for assessing a construct of interest. Less attention is given to whether the measure exhibits sufficient within-person variability or whether the measure displays similar factor structures across time as it does across people (cf. Molenaar & Campbell, 2009). Our short measure of CI exhibited a single-factor structure at both levels, suggesting that the items converged on similar structural constructs and that constraining the factor loadings to be the same across items and levels (i.e., tau-equivalent model) did provide a good but worse fit to the data than the model freely estimating the factor loadings. Additionally, using methods described by Cranford and colleagues (2006), our measure of CI exhibited adequate reliability (>0.70) across both levels of analysis. These results suggest that this short measure of CI is suitable for use in an intensive repeated measures/daily diary-type design.

We also provided substantive validation of the measure by linking CI to daily stress, both between persons and within persons over time. CI scores exhibited significant variability across individuals and days (approximately 50% at each level). Furthermore, individuals who reported experiencing daily stressors also reported higher levels of CI, and CI levels were higher on stressor days compared with non-stressor days. These results are consistent with previous findings linking stressful experiences to CI (Stawski et al., 2009; Sundin & Horowitz, 2003; Wegner, 1988), and theoretical perspectives suggesting that increases in CI is a normative component of stress reactivity (Horowitz, 1975).

These results replicate and extend previous research in a number of ways. Although previous research has linked CI with negative mood states (Moberly & Watkins, 2008a, 2008b) and subjective feelings of stress (Wegner, 1988) within persons over time, our findings indicate that stressors are temporally linked to daily reports of CI and that this is independent of one's daily NA or physical symptom reports. Additionally, the significant within-person association between NA and CI replicated with work of Moberly and Watkins (2008a, 2008b) in a sample of older adults. Thus, within persons over time, there is something unique about stressful experiences that increases CI that cannot be accounted for by their NA, the possibility of a reciprocal relationships notwithstanding (e.g., Moberly & Watkins, 2008a). This pattern of results is consistent with theoretical work by both Nolen-Hoeksema (1991) and Horowitz (1975) suggesting that transient fluctuations in CI can be linked to both mood states and stressful experiences. At the between-person level, the association between individual differences in the frequency of daily stressors and CI was accounted for by NA. This pattern is less consistent with Horowitz’s position that CI is a normative consequence of stressful experiences and more consistent Nolen-Hoeksema's (1991) work suggesting that individual differences in CI may reflect depressive rumination, or at the very least, thoughts about one's mood.

Covarying for physical symptoms did little to alter the associations between daily stress and CI. At the between-person level, both physical symptoms and daily stress were both significantly associated with higher levels of CI, whereas, at the within-person level, physical symptoms and CI were not significantly associated within persons over time, and the daily stress effect was virtually unchanged. These results suggest that the association between daily stressors and cognitive interference is not an artifact of daily stressors being associated with worse health. For the purposes of the current study, we covaried for physical symptoms when exploring the daily stress–CI relationship, other work in this area has suggested that perseverative cognition is important for linking the effects of stress on health (Brosschot et al., 2006).

According to Brosschot, the disposition to experience perseverative CI is associated with prolonged stress-related emotional and physiological reactions which increase vulnerability to disease. Our results are consistent with this position as we observed significant associations among individual differences in daily stress, CI, and physical symptoms but cannot speak to the direction of such effects. Although Brosschot's position focuses on CI and its correlates in understanding individual differences in stress and long-term health and disease processes, our results suggest that more transient daily stressor–related increases in CI are independent of one's current physical state. The association between CI and health might be established over a much longer period of time whereby perseveration and CI have cumulative effects making people more susceptible to disease-related processes, whereas more transient fluctuations in CI are less likely to have more proximal influences on one's symptoms or health at a specific point in time.

The differing patterns of results that emerges linking daily stress to CI when covarying for NA and physical symptoms at the between- and within-person levels of analysis also merits consideration. Although the goal is to predict variability in CI at both levels of analysis, these differing patterns of results underscore the fact that variability at different levels of analysis may reflect different constructs (e.g., Hoffman & Stawski, 2009; Martin & Hofer, 2004). Even though CI was the construct of interest, the reason that one's CI is higher from one day to the next may be different from why one person's CI is higher than another person’s. Indeed, on a given day, an individual's CI was associated with both their stressful experiences and NA on that day, whereas, for a given person, their level of CI is associated with their general level of NA but not their stressful experiences. Assuming that associations at one level of analysis (e.g., between persons) are informative about associations at a different level of analysis (e.g., within persons over time) may be incorrect (Borsboom, Mellenbergh, & van Heerden, 2003; Molenaar, 2004). Even in the context of the current study, the average within-person association may not completely describe all individuals, meriting consideration of more person-specific examinations.

Contrary to prediction, increased age was associated with decreased reactivity to daily stressors. Although increased age was associated with marginally higher levels of CI overall and is consistent with theories of cognitive aging (Hasher & Zacks, 1988), the decreased reactivity with increased age runs counter to theoretical accounts of age-related vulnerability to stress (e.g., Smith, 2003). One explanation is that the older adults in our sample are positively selected and thus less affected by the experience of daily stressors. This, however, does not explain their overall higher levels of CI. An alternative explanation is that, given the older age of the sample (mean age = ∼83), the oldest portion of the sample experiences more CI in general due to their inability to stave it off, whereas their younger counterparts are more capable of regulating their mental contents both in general and under conditions of stress. Exploring these associations across the broader adult life span would provide useful for understanding the relationship between age, CI, and stress.

Although the current study utilized an intensive repeated measures daily diary-type design, it cannot provide strong evidence of directionality, causality, or reciprocal relationships. Future research designed to better understand the temporal sequencing reciprocal relationships among stress, NA, symptoms, and CI is needed, as well as the impact objective stressors have on CI to complement our self-report methods. Again, CI is similar to other forms of maladaptive cognition (e.g., repetitive thought, rumination, worry, and perseverative cognition; Watkins, 2008), however, disagreement about the precise definitions of CI and its related constructs remains (e.g., Blumberg, 2000; Muris, Merckelbach, & Horselenberg , 1996). More thorough assessments of these constructions and their relationships to each other and daily stress over time would be a useful contribution.

Despite these limitations, the current study provides evidence of the utility of a short measure of CI for use in intensive repeated measures studies. CI exhibited considerable variability between persons as well as within persons over time, and we replicated and extended previous research to show that day-to-day fluctuations in CI are uniquely associated with NA and daily stressors among older adults. The results underscore the importance of considering the measurement properties of constructs for use in intensive repeated measures studies and distinguishing between associations across different levels of analysis for testing hypotheses about psychological phenomena.

FUNDING

This research was supported by grants (R01 AG12448 and AG026728) from the National Institutes of Health/National Institute on Aging.