www.elsevier.com/locate/bt

Available online at www.sciencedirect.com

Behavior Therapy 39 (2008) 366–374

Engagement in Activities Is Associated With Affective Arousal inAlzheimer's Caregivers: A Preliminary Examination of the

Temporal Relations Between Activity and Affect

Brent T. Mausbach, University of CaliforniaDavid W. Coon, Arizona State University

Thomas L. Patterson, Igor Grant, University of California

The primary purpose of this study was to examine thesynchronous and temporal relations between engagementin activities and the two primary dimensions of affect—namely, positive and negative affect—using an intensivetime-series design called concomitant time series analysis(CTSA). Twenty-four dementia caregivers completed 56diary measures (4 times per day for 2 weeks) assessing theirexperience of positive and negative affect as well as en-gagement in a variety of activities. Total number of ac-tivities was significantly correlated with positive affect(r= .40), but not negative affect (r=-.12). Obtained plea-sure from activities was significantly correlated with bothpositive (r= .42) and negative affect (r=- .17). These resultsmay help further develop behavioral models of depressionby suggesting that behavioral or self-reinforcing activitiesare associated primarily (or more saliently) with one'sexperience of positive affect. Future research examining theeffect of behavioral interventions on both positive andnegative affect is suggested, as is the examination of factorsthat may be more strongly associated with negative affect.

THOSE WHO PROVIDE care for a loved one with de-mentia are at increased risk for suffering both mentaland physical health consequences (Schulz, O'Brien,

This research was supported by awards 23989 and 15301 fromthe National Institute on Aging.

Address correspondence to Brent T.Mausbach, Ph.D., Departmentof Psychiatry (0680), University of California SanDiego, 9500GilmanDrive, La Jolla, CA 92093-0680, USA; e-mail: bmausbach@ucsd.edu.0005–7894/08/366–374$1.00/0© 2008 Association for Behavioral and Cognitive Therapies.Published by Elsevier Ltd. All rights reserved.

Bookwala, & Fleissner, 1995). Perhaps the mostnotable consequence of caregiving is an increased riskfor experiencing depressive symptoms. Indeed, com-pared to noncaregivers, caregivers are at appro-ximately twice the risk for depression (Baumgartenet al., 1992), with studies placing caregivers' risk fordepression between 28% and 33% (Cohen et al.,1990; Williamson & Schulz, 1993). Although re-search examining depressive symptoms among care-givers is quite common, little research has examinedboth positive and negative affect as individualconstructs.Research over the past decades has established two

basic, independent dimensions of affect, termedpositive and negative affect. Where positive affectreflects one's level of excitement, interest, and enthu-siasm, negative affect reflects one's level of distress,fear, and relaxation (Watson, Wiese, Vaidya, &Tellegen, 1999). According to Watson, Clark, andCarey (1988), depressive disorders result from acombination of low positive affect and high negativeaffect. This tripartite model of depression (Clark &Watson, 1991) further identifies high negative affectas a common factor underlying both depression andanxiety, whereas low positive affect is a primary dis-tinguishing feature of depression as compared toother negativemood states (i.e., anxiety). This theoryhas been supported in multiple studies (Brown,Chorpita, & Barlow, 1998; Chorpita, Albano, &Barlow, 1998; Joiner, Catanzaro, & Laurent, 1996;Watson et al., 1995). For example, Watson andcolleagues (1999) demonstrated that positive andnegative affect each contributed unique variance tothe prediction of a diagnosis of major depression,with the number of depressive symptoms highly cor-related with positive affect (r=-.40) and negativeaffect (r=.57). Similarly, others have found that

367act i v i ty and affect

Hamilton Depression Scale scores are highly corre-lated with positive affect (r=-.52) and negative affect(r=.44), respectively, within a nonclinical sample(Crawford & Henry, 2004). Furthermore, thesefindings appear stable into later life. In a sample ofmothers caregiving for adult children with disabil-ities, positive and negative affect have been indepen-dently related to depressive symptoms (Pruchno &Meeks, 2004). Another study of older adults foundpositive affect to be inversely correlated with depres-sive symptoms (i.e., Beck Depression Inventoryscores); greater endorsement of negative affect wasassociated with more depressive symptoms (Cook,Orvaschel, Simco, Hersen, & Joiner, 2004). Giventhat depressive symptoms appear to consist of bothpositive and negative dimensions, researchers andpractitioners should identify behavioral and copingstrategies that simultaneously increase positive anddecrease negative affect to effectively and efficientlycombat depressive symptoms.A key line of research has examined the role of

reduced engagement in self-reinforcing activity(Lewinsohn, 1974, 1975), or activity restriction(Williamson & Shaffer, 2000), in the developmentof emotional distress. These approaches posit thedynamic interplay of engagement in activities andmental health outcomes. For example, the ActivityRestriction Model posits that restriction of normalactivities mediates the relationship between lifestressors and psychological adjustment (Williamson& Shaffer, 2000). To date, research testing thismodelhas been promising, particularly with regard to therelationship between activity restriction and negativeaffect. For example, activity restriction has been as-sociatedwith depressed affect in community-dwellingolder adults (Benyamini & Lomranz, 2004;William-son& Schulz, 1992) and among geriatric outpatientssuffering from cancer (Williamson & Schulz, 1995).Other studies have found that activity restriction isassociated with depressed affect among limb ampu-tees (Williamson, Schulz, Bridges, & Behan, 1994),caregivers of cancer patients (Williamson, Shaffer,& Schulz, 1998), and dementia family caregivers(Thompson et al., 2002).Only a few studies have explored the relations

between activity and positive affective states, withpreliminary evidence suggesting that increased ac-tivitymaybe associatedwith affectivewell-being. Forexample, a recent study of community-dwellingmiddle-aged and older adults found that those whoengaged in more activities reported greater affectivewell-being and life satisfaction (Warr, Butcher, &Robertson, 2004). Another study found that indivi-duals with major depressive disorder experienced asignificant increase in positive affective states (i.e.,positive well-being and vigor) following 30 minutes

of aerobic physical exercise (Bartholomew, Morri-son, & Ciccolo, 2005). Using a time-series design,Vittengl and Holt (1998) found significant synchro-nous correlations between active social interactionand positive affect. However, these studies aretypically limited to one domain of activity (i.e.,physical exercise and social interaction, respectively)rather than a variety of activities encompassing men-tal, physical, and social engagement.Although separate studies have demonstrated re-

lationships between activity and both positive andnegative affective states, the existing research is notwithout limitations. For example, it remains unclearhow activity relates to both positive and negativeaffect in the same individuals. It could be, forexample, that behavioral activation works entirelyon positive affect, whereas other mechanisms, suchas cognitions, more strongly relate to fluctuations innegative affect. This conceptualization would beconsistent with Beck's “Cognitive Triad” (Beck,Rush, Shaw, & Emery, 1979), whereby individualsestablish negative cognitions about themselves, theirexperiences, and their future, which in turn arebelieved to increase negative affect. One goal of thepresent study is to take a first step toward aligning theindependent research findings on the relationshipbetween activity and affect by examining the intra-individual relationships between activity levels andboth positive and negative affect.Another notable limitation of the existing re-

search is its limited use of different types of researchdesigns. With the exception of the study conductedby Vittengl and Holt (1998), the existing researchhas focused on either cross-sectional or longitudinaldesigns, both of which have their share of weak-nesses. For example, cross-sectional designs do notallow for the examination of intra-individual fluc-tuations in activity and affect over time and do notallow for the examination of temporal relationsbetween variables of interest (Barnett & Gotlib,1988). Alternately, whereas longitudinal designsattempt to take temporal precedence into account,many longitudinal studies assume that the con-structs being measured are traitlike and remainstable over time. Specifically, participants are typi-cally asked to recall their activity levels and affectivestates over an extended period of time prior to eachassessment point (e.g., the previousmonth). Further,many investigations of change assume linear changein variables over time. This type of assessment maybe problematic because individuals might experi-ence daily (or even hourly) fluctuations in their ac-tivity levels and their affect, and these intra-individual fluctuations may be concomitantly andtemporally related to one another. To address theseconcerns, idiographic fluctuations in constructs of

368 mausbach et al .

interest and their covariation over time needthorough investigation.The present study takes a novel approach by using

time series analysis to help address the methodolo-gical issues just raised. Specifically, we examinedidiographic fluctuation in both positive and negativeaffect aswell as several activities over time in a sampleof dementia caregivers. Because they have significantcaregiving responsibilities that frequently (althoughnot completely) interfere with their ability to en-gage in self-reinforcing activities (Ory, Hoffman, Yee,Tennstedt, & Schulz, 1999), dementia caregiversprovide an excellent source inwhich to examine thesefluctuations and their temporal relations.The usefulness of time series analysis for analyzing

behavior has been reported in the literature (Birditt,Fingerman, & Almeida, 2005; Mroczek & Almeida,2004; Tryon, 1982). One approach to examiningtime-series data is concomitant time-series analysis(CTSA). In CTSA, data are collected over extendedperiods of time (e.g., weeks) through the use of aregular and frequent assessment schedule (e.g.,daily). This design has several advantages over tra-ditional research designs (Nasby & Read, 1997;Stader & Hokanson, 1998; West & Hepworth,1991). First, CTSA, because of its longitudinal na-ture, can reveal trends and cycles in data that bythemselves might divulge important information toboth investigators and participants. Second, CTSAreveals details of the variation among variables overtime that most traditional analyses obscure. Third,CTSA can explicitly represent finite causal lag timesand help establish causal inferences that traditionalcross-sectional designs cannot. Fourth, CTSA can beused to examine not only the correlations betweentime-series at the same point in time but also therelations of prior values of one variable with currentvalues of another. Finally, CTSA detects and modelsserial dependencies of repeated observations overtime to ensure valid statistical inference. Therefore,CTSA is an idiographic approach to determining notonly a description of day-to-day fluctuations in con-structs of interest but also the covariation of theseconstructs over time (Stader & Hokanson, 1998).The current study had two aims: (a) to test the

synchronous and temporal relationships between ac-tivity and affective states, and (b) to demonstratethe usefulness of a unique research design (i.e., CTSA)for examining these relationships. In doing so, weemployed an intensive time-series design in which24 individuals completed a 2-week diary assessinggeneral activity, obtained pleasure from these activ-ities, and positive and negative affect. These assess-ments were made at multiple time points each day,allowing us to examine both synchronous (at thesame timepoint) and lagged correlations (correlations

between variables at different points during the day).Based on behavioral theories of depression (Lewin-sohn, 1975; Williamson & Shaffer, 2000), we hypo-thesized that greater engagement in self-reinforcingactivities would be positively associated with positiveaffect and inversely related to negative affect.

Methodparticipants

A total of 24 dementia caregivers participated in thisstudy. They were recruited through a variety ofmeans, including community support groups, localsenior centers andmedical clinics, and senior serviceprofessionals. Criteria for inclusion in the studyincluded: (a) being age 21 or older, (b) providing 8ormore hours of care per week to an individual witha physician-diagnosis of Alzheimer's disease (AD),(c) being in generally good health, and (d) reportingthree or more symptoms of depression over the pastmonth per DSM-IV criteria. In addition, patientswith AD must have required assistance with at leasttwo instrumental activities of daily living (IADLs;e.g., shopping, using the telephone) or one activityof daily living (ADL; e.g., dressing, toileting).Individuals were excluded if they (a) did not speakEnglish or (b) reported having a serious medicalcondition (e.g., cancer).

measures

Demographic and baseline characteristics. Atbaseline, all participants were administered demo-graphic questionnaires assessing their age, gender,relationship to their care recipient, ethnicity, years ofcaregiving, yearly income, and hours of care pro-vided per week. Caregivers were also administeredthe Activities of Daily Living Scale (ADL; Katz, Ford,Moscowitz, Jackson, & Jaffee, 1963) and theInstrumental Activities of Daily Living Scale (IADL;Lawton & Brody, 1969), which assess the level ofcare required by the care recipient for performingbasic daily living tasks.Problem behaviors exhibited by the care recipient

(e.g., asking the same question over and over;trouble remembering recent or past events; appear-ing sad or depressed) were assessed using theRevised Memory and Behavior Problem Checklist(RMBPC; Teri et al., 1992). For this scale, care-givers report on whether or not their care recipientexhibited each of 24 problem behaviors over thepast week, with a total score consisting of the sumof the 24 responses (range=0–24). For each pro-blem behavior endorsed, caregivers rate the extentto which they were bothered or upset by theproblem behavior. Responses to these follow-upquestions are on a Likert scale ranging from 1 (not

369act i v i ty and affect

at all) to 5 (extremely). The cross-products (Occur-rence×Upset) for the 24 behaviors are summed tocreate an overall Distress score (range=0–120).

Affective experience. The Positive and NegativeAffect Schedule (PANAS; Watson, Clark, & Telle-gen, 1988) was used to assess affective experience.This scale consists of 20 mood adjectives. Ten itemscomprise the Positive Affect (PA) subscale, includ-ing adjectives such as “interested,” “excited,” and“enthusiastic.” The remaining 10 items comprisethe Negative Affect (NA) subscale, which consistsof adjectives such as “distressed,” “scared,” and“irritable.” The participant responded to each ad-jective by rating how much she felt that way duringthe past 4 hours. Response options were: 1=veryslightly or not at all, 2=a little, 3=moderately,4=quite a bit, 5=extremely.

Activity levels. The Pleasant Events Schedule-AD(PES-AD; Logsdon & Teri, 1997) was used to assesseach caregiver's engagement in activities. The scaleconsists of 20 activities in which individuals engageduring thecourseofa typicalday (e.g., listening tomusic,havingmeals with friends or family, exercising). At eachassessment point, the participant rated the frequencywith which he or she engaged in each activity over thepast 4 hours. Response options were as follows: 0=notat all, 1=1 time, and 2=2 ormore times. A total activityscore (TA) was calculated for each assessment bysumming the 20 items (range=0–40).In addition, for each activity engaged, the

participant was asked to report how enjoyable theactivity was on a scale ranging from 0 (not at all) to2 (a great deal). For each of the 20 activities, a cross-product of the frequency and pleasure scores wascalculated (range=0–4 for each item). The sum ofthese cross products was used to create an obtainedpleasure (OP) score (range=0–80).

procedures

All participants volunteered and provided writtenconsent to participate in the Pleasant Events Project(PEP), which was approved by the InstitutionalReview Board at the University of California, SanDiego. The participant was given a “diary” of re-search forms (mentioned above) and was instructedto complete these measures at the following timeseach day for a period of 2 weeks: (a) early morning,(b) noon, (c) late afternoon, and (d) prior to retiringin the evening. Therefore, for the current study,each participant completed a total of 56 diaryentries for each time-series construct. This schedulewas chosen because the literature suggests a mini-mum of 50 observations when using a time-seriesdesign (West & Hepworth, 1991). The forms werecounterbalanced to eliminate potential bias due tothe order in which forms were completed.

data analysis

For the current study, four time-series were ex-amined: (a) positive affect (PA), (b) negative affect(NA), (c) total activity level (TA), and obtainedpleasure from activities (OP). Prior to conductingcross-correlations, CTSA requires that each vari-able be described separately and that threats to thevalidity of a time series be removed. The most con-servative approach to removing threats involvesprewhitening each time series one participant at atime, a process that reduces each variable to whitenoise residuals. Prewhitening first entails detecting,estimating, and removing long-term trends as wellas cycles from each series (Nasby & Read, 1997;West & Hepworth, 1991).An assumption of time series analysis is statio-

narity, which requires, among other criteria, thatthe mean of the repeated measures not vary sys-tematically over time (Nasby & Read, 1997; West&Hepworth, 1991). Sources of nonstationarity aremodeled and removed according to the followingdecomposition model:

Y tð Þ¼ T tð ÞþC tð ÞþD tð Þ

Where T(t), C(t), and D(t) refer to the componentsof long-term trend, cyclic patterns (superimposedabout the long-term trend), and daily variations(superimposed about the long-term trend and cycles),respectively (Nasby&Read, 1997). The componentsT(t) and C(t) delineate the variation of the mean of Yover time, and require estimation and removal toisolate the daily component and prevent distortion ofthe cross-correlations between “daily” variations intwo series (Nasby&Read, 1997;West&Hepworth,1991). Once trends and cycles are removed, the dailycomponent of the series is isolated.Another assumption of CTSA is that each ob-

servation is independent of the others. Serial de-pendencies, which may exist in the daily component,violate this assumption. That is, if the correlation ofthe residuals fromadjacent time points is “non-zero,”the result would be that one would conduct regres-sion analyses that violate the assumption that ob-servations are independent of one another. West andHepworth (1991) indicate that the result of conduct-ing analyses with correlated residual error termswould be to “bias the standard error of the regressioncoefficients so that inferential statistical tests based onthe data will be incorrect” (p. 619). These analysesbecome particularly problematic when non-zerocorrelations of the residuals exist in two time seriesto be cross-correlated. Serial dependencies musttherefore be modeled using the Autoregressive In-tegrated Moving Average (ARIMA) technique (Box& Jenkins, 1976). This was accomplished using

Table 1Characteristics of the Sample

Variable

Caregiver Age, M (SD) 62.71 (11.33)Gender, n (%)Male 1 (4)Female 23 (96)

Relationship, n (%)Spouse 15 (63)Non-spouse 9 (37)

Ethnicity, n (%)Caucasian 20 (83)Latina/Hispanic 4 (17)

Years Caregiving, M (SD) 3.88 (2.59)Income, n (%)b$50,000 12 (52)≥$50,000 11 (48)

Hours of Care/Week, M (SD) 58.61 (52.04)Patient Age, M (SD) 76.71 (8.26)Activities of Daily Living, M (SD) 2.71 (2.37)Instrumental Activities of Daily Living,M (SD) 7.05 (1.60)

RMBPC Total Behaviors, M (SD) 11.06 (3.47)RMBPC Distress, M (SD) 1.55 (0.74)

Note. One participant did not provide income data. RMBPC=Re-vised Memory and Behavior Problem Checklist.

370 mausbach et al .

autocorrelograms. After serial dependencies weremodeled for each series, the Box-Ljung Q statisticindicated that all serieswere equivalent towhite noiseresiduals. For each participant, all four time series inthis study were reduced to their white noise residualsusing this prewhitening process.Once prewhitening was completed and each time

series had been decomposed to its white noise resi-duals, cross-correlations were calculated betweenvariables of interest. In the present study, cross-correlations consisted of both synchronous andlagged correlations. Synchronous correlations revealthe relations between variables during the same timeframe (“lag” 0). In addition to synchronous correla-tions, each variable was alternately placed as the leadvariable (e.g., TA preceding PA by one or more timeintervals) and the lag variable (e.g., PA preceding TA

Table 2Models Used to Prewhiten Diary Variables

Transformation Trend

Variable Sqrt Log10 (+) (-) (

Frequency 3 1 4 4 5Obtained Pleasure 0 0 4 2 5Positive Affect 1 1 6 5 5Negative Affect 1 1 4 4 4

Note. Numbers inside cells correspond to the number of participanprewhitening, with a maximum cell value of 24. Sqrt=Square root; (+) and(Par) corresponds to a parabolic trend; Cycles of 2 and 4 correspond to 2maximum of 24 participants was possible except for obtained pleasure,

by one or more time intervals), and the pair of var-iables was examined for significant lagged correla-tions. In the present analyses, each variable wasplaced as the lead variable by four time points (i.e.,“lag” 4), as this would indicate one full day of datacollection. It is these lagged correlations that helpestablish temporal inferences between the variablesof interest. In all, we conducted cross-correlationsbetween the following variables: (a) TA and PA,(b) TA and NA, (c) OP and PA, and (d) OP and NA.Using the methods described by Vittengl andHolt

(1998), we calculated the number of participantswith significant correlations (pb .05, two-tailed) asan indication of the prevalence of the effect. Asreported by Vittengl and Holt, a binomial distribu-tion was used to calculate the minimum number ofsignificant effects needed to deviate from chancegiven a two-tailed alpha of .05. Using a Bonferronicorrection, we determined that the probability ofsignificance in each direction was .0007 (p-value of.025 divided by 36 total cross-correlations). In thiscase, 5 or more participants with significant effectsindicated a departure from chance. Finally, effectsizes were calculated using Fisher's z transforma-tion and weighting participants' values by the num-ber of diaries (out of 56 total) they completed. Asper Cohen's (1988) guidelines, mean effect sizes forcorrelational data of .10, .30, and .50 were con-sidered small, medium, and large, respectively.

ResultsBaseline characteristics of the sample are presentedin Table 1. On average, participants were approxi-mately 62 years of age and had been caregiving for3.8 years. Most caregivers were spouses (63%) andof Caucasian ethnicity (83%).Participants were highly compliant in completing

their diaries. Specifically, participants completed anaverage of 51.6 consecutive PES-AD forms and 51.0consecutive PANAS forms. The total number of diaryobservations across participants was approximately1,200 for eachof our time series studied (i.e., PA,NA,

Cycle Autoregressive parameter

Par) 2 4 1 2 3 4 5

2 13 5 1 4 0 11 10 5 1 3 0 00 8 7 1 0 1 10 3 8 5 0 0 0

ts' models that required the specific time-series parameter for(-) correspond to positive and negative linear trends, respectively;(half-day) and 4 (1 day) diary periods, respectively. For each cell, ain which a maximum of 23 was possible.

371act i v i ty and affect

TA, and OP). One participant did not adequatelycomplete the pleasure scale of the PES-AD. There-fore, cross-correlations between obtained pleasureand affective outcomes were conducted for 23 of the24 participants. Table 2 presents a summary of thetrends, cycles, and autoregressive parameters used toprewhiten diary data. As can be seen, daily cycleswere relatively common, most notably for thefrequency and obtained pleasure variables. For PAand NA, an autoregressive parameter was fairlycommon, indicating that a given diary record forthese constructs was correlated with adjacent diaryrecords.

cross-correlation results

Results for selected cross-correlations are presentedin Table 3. As can be seen, Total Activity (TA) waspositively related with positive affect (PA), demon-strating a medium to large effect (r=.40) at lag 0(synchronous correlation). Of 24 participants, 18demonstrated a significant positive correlation be-tween TA and PA, which exceeded the numberexpected by chance. When TA followed PA by onediary period (lag 1), there was a small effect (meanr=.09). However, only 2 participants demonstratedsignificant correlations, which did not reach ourpredefined number of participants needed to de-viate from chance (i.e., 5 participants).Similarly, obtained pleasure (OP) was positively

associated with PA, also demonstrating a medium tolarge effect (r=.42) at lag 0. Of 23 participants, 18demonstrated a significant correlation between theseconstructs (pb .05). When PA preceded OP by one

Table 3Cross-Correlations between Affect and Activity Variables

Minimum Maximum S (-) S (+) M

Lag 0 Cross-CorrelationsFrequency with

Positive Affect .12 .67 0 18 .40Negative Affect - .43 .10 4 0 -.12

Obtained Pleasure withPositive Affect .17 .81 0 18a .42Negative Affect - .36 .21 7 0 -.17

Lag 1 Cross-CorrelationPositive Affect with

Frequency -.19 .38 0 2 .09Obtained Pleasure -.21 .44 0 3 .11

Negative Affect withFrequency -.33 .11 0 0 -.07Obtained Pleasure -.21 .09 0 0 -.07

Note. N=24; a N=23 for Obtained Pleasure. S(-)=Number ofparticipants with significant negative correlations. S(+)=Numberof participants with significant positive correlations. M=Meancorrelation.

diary period, there was a small effect (r=.11), with 3of 23 participants reaching statistical significance.The relationship between TA and negative affect

(NA) appeared weaker than that between TA andPA. Specifically, a small effect was observed betweenTA and NA (r=�.12), with 4 participants demon-strating a significant negative correlation. When TApreceded NA by one diary period, the average effectwas -.07, with no significant correlations.Finally, the number of participants with signifi-

cant negative correlations between obtained plea-sure (OP) and negative affect (NA) at lag 0 exceededthat expected by chance (n=7). The mean effect forthis relationship was small to medium (r=-.17).

DiscussionThe purpose of this study was to examine the syn-chronous and lagged correlations between activityand affect in a sample of dementia family caregivers.Our results suggest that alterations in caregivers'experience of positive and negative affect wereassociated with their level of activity, particularlyengagement in activities deemed to be pleasurable.That is, greater experience of positive affect andreduced experience of negative affect were associatedwith obtained pleasure from activities. These results,which encompass over 1,200 observations, provide afirst look at the relations between behavioral acti-vation, particularly self-reinforcing activity, andone's positive and negative affective states over time.Although previous research has focused on the

relations between activity and depressive symptoms,this study expands upon this research by demonstrat-ing that engagement in self-reinforcing activities (i.e.,obtained pleasure) is associated with each of the twocentral domains of affect (i.e., positive and negativeaffect) known to be associated with depressivesymptoms (Crawford & Henry, 2004; Watson,Clark, & Carey, 1988). Interestingly, self-reinforcingactivities appeared to be much more strongly relatedto positive affect than negative affect, an effect thatbehavioral researchers may wish to investigate morethoroughly in future studies. Indeed, these results arepreliminary and correlational, and caution should beexercised in determining the causality of thesefindings. Although CTSA is a unique method ofestablishing concomitant and temporal correlationsamong longitudinal data, it remains a correlationalanalysis that is open to alternative explanations (e.g.,the impact of important confounding variables notincluded in the analyses). Further, among correla-tional data such as presented in this study, causalinferences are developed through thoughtful delib-erations on the nature of the statistical association(e.g., if patterns were predicted a priori, if potentially

372 mausbach et al .

confounding covariates have been modeled and ruledout). Nonetheless, this study lays early groundwork forusing traditional experimental methods of testingwhether behavioral activation produces affectivechange, particularly change in positive affect. Themost salient means of establishing this would be toconduct a randomized trial comparing a behavioraland control intervention for producing change inpositive and negative affect. Although others (Coon,Thompson, Steffen, Sorocco, &Gallagher-Thompson,2003; Gallagher-Thompson et al., 2000) have con-ducted randomized clinical trials that successfullydemonstrated the benefit of behaviorally based inter-ventions in the reduction of depressive symptomsamong dementia caregivers, these authors did notexamine the impact of these strategies on positiveaffect or the mediating role of pleasurable activitieson these outcomes. It may be particularly interestingto utilize daily diary approaches during the course ofa behavior therapy trial, thereby allowing researchersto incorporate analyses such as CTSA to tease outdaily time-lag effects while simultaneously examin-ing the causal effects of the intervention via the RCT.It should be noted that the participants in this study

were not necessarily clinically depressed, which mayhave limited the correlation between participants'activity ratings and negative affect. Future researchamong participants experiencing significant symp-toms of depression may reveal a unique pattern fromthe one presented in this study. Nonetheless, activitylevels, particularly engagement in self-reinforcingactivities, were significantly associatedwith both posi-tive and negative affect in this sample, thus providingpreliminary evidence to be replicated in future studies.Interestingly, there was a lack of evidence sup-

porting lagged correlations between constructsassessed in this study, thereby making it unclear asto whether activity or affect preceded change in theother construct. No significant lagged correlationswere found when activities (TA) preceded positiveaffect. These data suggest a weak effect of positiveaffect on one's engagement in self-reinforcing acti-vities, whereby individuals engaged in slightly moreactivities during diary periods immediately follow-ing elevations in PA, or fewer activities during diaryperiods immediately following reductions in PA.As an example, it seems highly plausible that ifindividuals are feeling fatigued, disinterested, orunenthusiastic during the afternoon, they wouldforego engaging in activities during the eveninghours. Although we did not test this effect, the lackof significant lagged correlations when activity pre-ceded affect might also be explained by a “burnout”period following periods of “exceptionally” highlevels of activity. That is, individuals who engaged inhigh levels of activity during the afternoonmay have

become “fatigued” by the stimulation and elected torest or conserve energy during the evening, whichwould be reflected as lower levels of positive affect.Similarly, from a caregiving context, a caregiverwho engaged in activities apart from her care reci-pient may experience a concomitant increase inpositive affect, but a return to caregiving duties laterthat day, in turn, reduces both activity and positiveaffect. However, these conclusions are speculativeand beyond the scope of the current study. Verifica-tion is therefore needed using greater diary data and/or a more advanced analytic framework than usedin this study.It should be noted that an important feature of the

design used in this study was that our correlationswere obtained after removal of trends, cycles, andserial dependencies. This “prewhitening” processserved to reduce otherwise spuriously high correla-tions between activity and affect. Indeed, cycles of2 and 4 diary periods were common in these data, aswere autocorrelations within the time series. Speci-fically, cycles of 4 diary periods appeared quite fre-quently for our total activity and obtained pleasuredomains, which may suggest that for some indivi-duals, activities are scripted, or planned, for specifictimes of the day (e.g., a walk in the evening or lateafternoon, a favorite television show before bed,etc.). This may be particularly true for dementiacaregivers who need to keep a specific schedule tohelp manage their caregiving responsibilities andhave less flexibility in their schedules. Also, auto-regressive parameters were common for positive andnegative affect, indicating that reports of theseaffective states were correlated at nearby diary points(e.g., those reporting high positive affect at noon alsoreported high levels of positive affect at dinner). Ifbehavioral activation does indeed produce changesin positive affect, as Bartholemew et al. indicated(2005), these increases in positive affectmay not onlycarry over to future time periods, but they may cycleback to one's engagement in activities. Specifically,an individual may engage in pleasurable activitiesand experience an immediate increase in positiveaffect, which in turn increases one's engagement infuture activities, and so on, over time. Indeed, thiscan be conceptualized as a reversal of the “down-ward spiral” or “vicious cycle” effect wherein a lackof engagement leads to a cycle of depressive symp-toms reinforcing disengagement and greater mooddisturbance. This familiar cycle is a noted feature ofbehavioral models of depression, although its effectcan be difficult to observe statistically given the com-plex intercorrelations occurring over time. None-theless, we do observe in this study that affect andpleasurable activities appear correlated during thesame time period, and that affective experiences

373act i v i ty and affect

appear to carry over to future time periods, which inturn appear positively correlated with activity levels.Overall, this study provides a preliminary exam-

ination of the synchronous and temporal relationsbetween activity and affect in a sample of Alzhei-mer's caregivers. We find that although both posi-tive and negative affect are associated withengagement in pleasurable activities, the associa-tion between positive affect and pleasurable activ-ities is stronger than that of negative affect. Notedstrengths of this study are (a) our examination ofthe two core aspects of depressed affect (i.e.,positive and negative affect) and their uniquecorrelations with activity, and (b) utilization of apowerful longitudinal design (i.e., concomitanttime-series analysis), which allowed for the exam-ination of fluctuations in activity and affect andanalysis of the synchronous and temporal relationsbetween these constructs over time. Future investi-gations should examine: (a) the impact of beha-vioral activation interventions on participants'engagement in activities and both positive andnegative affect and (b) whether mechanisms otherthan engagement in activities (e.g., changes incognitions) more strongly influence negative affect.

References

Barnett, P. A., & Gotlib, I. H. (1988). Psychosocial functioningand depression: Distinguishing among antecedents, conco-mitants, and consequences. Psychological Bulletin, 104,97–126.

Bartholomew, J. B., Morrison, D., & Ciccolo, J. T. (2005).Effects of acute exercise on mood and well-being in patientswith major depressive disorder. Medicine and Science inSports and Exercise, 37, 2032–2037.

Baumgarten, M., Battista, R. N., Infante-Rivard, C., Hanley, J.A., Becker, R., & Gauthier, S. (1992). The psychological andphysical health of family members caring for an elderlyperson with dementia. Journal of Clinical Epidemiology, 45,61–70.

Beck, A. T., Rush, A. J., Shaw, B. F., & Emery, G. (1979).Cognitive therapy of depression. New York: GuilfordPress.

Benyamini, Y., & Lomranz, J. (2004). The relationship of activityrestriction and replacement with depressive symptoms amongolder adults. Psychology and Aging, 19, 362–366.

Birditt, K. S., Fingerman, K. L., & Almeida, D. M. (2005). Agedifferences in exposure and reactions to interpersonal tensions:A daily diary study. Psychology and Aging, 20, 330–340.

Box, G. E. P., & Jenkins, G. M. (1976). Time series analysis:Forecasting and control. San Francisco: Holden-Day.

Brown, T. A., Chorpita, B. F., & Barlow, D. H. (1998).Structural relationships among dimensions of the DSM-IVanxiety and mood disorders and dimensions of negativeaffect, positive affect, and autonomic arousal. Journal ofAbnormal Psychology, 107, 179–192.

Chorpita, B. F., Albano, A. M., & Barlow, D. H. (1998). Thestructure of negative emotions in a clinical sample ofchildren and adolescents. Journal of Abnormal Psychology,107, 74–85.

Clark, L. A., & Watson, D. (1991). Tripartite model ofanxiety and depression: Psychometric evidence and taxo-nomic implications. Journal of Abnormal Psychology, 100,316–336.

Cohen, J. (1988). Statistical power analysis for the behavioralsciences. Hillsdale, NJ: Lawrence Erlbaum Associates.

Cohen,D., Luchins,D., Eisdorfer, C., Paveza,G. J., Ashford, J.W.,Gorelick, P., et al. (1990). Caring for relatives with Alzheimer'sdisease: The mental health risks to spouses, adult children,and other family caregivers. Behavior, Health, and Aging, 1,171–182.

Cook, J. M., Orvaschel, H., Simco, E., Hersen, M., & Joiner, T.(2004). A test of the tripartite model of depression andanxiety in older adult psychiatric outpatients. Psychologyand Aging, 19, 444–451.

Coon, D. W., Thompson, L., Steffen, A., Sorocco, K., &Gallagher-Thompson, D. (2003). Anger and depressionmanagement: Psychoeducational skill training interventionsfor women caregivers of a relative with dementia. TheGerontologist, 43, 678–689.

Crawford, J. R., & Henry, J. D. (2004). The Positive andNegative Affect Schedule (PANAS): Construct validity,measurement properties and normative data in a largenon-clinical sample. British Journal of Clinical Psychology,43, 245–265.

Gallagher-Thompson, D., Lovett, S., Rose, J., McKibbin, C.,Coon, D., Futterman, A., et al. (2000). Impact of psychoe-ducational interventions on distressed family caregivers.Journal of Clinical Geropsychology, 6, 91–110.

Joiner Jr., T. E., Catanzaro, S. J., & Laurent, J. (1996).Tripartite structure of positive and negative affect, depres-sion, and anxiety in child and adolescent psychiatricinpatients. Journal of Abnormal Psychology, 105, 401–409.

Katz, S., Ford, A. B., Moscowitz, R. W., Jackson, B. A., &Jaffee, M. W. (1963). The index of ADL: A standardizedmeasure of biological and psychosocial function. Journal ofthe American Medical Association, 185, 914–919.

Lawton, M. P., & Brody, E. M. (1969). Assessment of olderpeople: self-maintaining and instrumental activities of dailyliving. The Gerontologist, 9, 179–186.

Lewinsohn, P. M. (1974). A behavioral approach to depression.In R. J. Friedman & M.M. Katz (Eds.), The psychology ofdepression: Contemporary theory and research. New York:John Wiley & Sons.

Lewinsohn, P. M. (1975). Engagement in pleasant activities anddepression level. Journal of Abnormal Psychology, 84,729–731.

Logsdon, R. G., & Teri, L. (1997). The Pleasant EventsSchedule-AD: Psychometric properties and relationship todepression and cognition in Alzheimer's disease patients.The Gerontologist, 37, 40–45.

Mroczek, D. K., & Almeida, D. M. (2004). The effect of dailystress, personality, and age on daily negative affect. Journalof Personality, 72, 355–378.

Nasby, W., & Read, N. W. (1997). The life voyage of a solocircumnavigator: Integrating theoretical and methodologi-cal perspectives. Journal of Personality, 65, 785–1068.

Ory, M. G., Hoffman III, R. R., Yee, J. L., Tennstedt, S., &Schulz, R. (1999). Prevalence and impact of caregiving: Adetailed comparison between dementia and nondementiacaregivers. The Gerontologist, 39, 177–185.

Pruchno, R. A., & Meeks, S. (2004). Health-related stress,affect, and depressive symptoms experienced by caregivingmothers of adults with a developmental disability. Psychol-ogy and Aging, 19, 394–401.

Schulz, R., O'Brien, A. T., Bookwala, J., & Fleissner, K. (1995).Psychiatric and physical morbidity effects of dementia

374 mausbach et al .

caregiving: Prevalence, correlates, and causes. The Geron-tologist, 35, 771–791.

Stader, S. R., & Hokanson, J. E. (1998). Psychosocialantecedents of depressive symptoms: An evaluation usingdaily experiences methodology. Journal of AbnormalPsychology, 107, 17–26.

Teri, L., Truax, P., Logsdon, R., Uomoto, J., Zarit, S., &Vitaliano, P. P. (1992). Assessment of behavioral problems indementia: The revised memory and behavior problemschecklist. Psychology and Aging, 7, 622–631.

Thompson, L. W., Solano, N., Kinoshita, L., Coon, D. W.,Mausbach, B., & Gallagher-Thompson, D. (2002). Pleasur-able activities and mood: Differences between Latina andCaucasian dementia family caregivers. Journal of MentalHealth and Aging, 8, 211–224.

Tryon, W. W. (1982). A simplified time-series analysis forevaluating treatment interventions. Journal of AppliedBehavior Analysis, 15, 423–429.

Vittengl, J. R., & Holt, C. S. (1998). A time-series diary study ofmood and social interaction. Motivation and Emotion, 22,255–275.

Warr, P., Butcher, V., & Robertson, I. (2004). Activity andpsychological well-being in older people. Aging and MentalHealth, 8, 172–183.

Watson,D., Clark, L. A.,&Carey, G. (1988). Positive and negativeaffectivity and their relation to anxiety and depressive disorders.Journal of Abnormal Psychology, 97, 346–353.

Watson, D., Clark, L. A., & Tellegen, A. (1988). Developmentand validation of brief measures of positive and negativeaffect: The PANAS scales. Journal of Personality and SocialPsychology, 54, 1063–1070.

Watson, D., Weber, K., Assenheimer, J. S., Clark, L. A., Strauss,M. E., & McCormick, R. A. (1995). Testing a tripartitemodel: I. Evaluating the convergent and discriminantvalidity of anxiety and depression symptom scales. Journalof Abnormal Psychology, 104, 3–14.

Watson, D., Wiese, D., Vaidya, J., & Tellegen, A. (1999). Thetwo general activation systems of affect: Structural findings,evolutionary considerations, and psychobiological evidence.Journal of Personality and Social Psychology, 76, 820–838.

West, S. G., & Hepworth, J. T. (1991). Statistical issues in thestudy of temporal data: Daily experiences. Journal ofPersonality, 59, 609–662.

Williamson,G.M.,& Schulz, R. (1992). Pain, activity restriction,and symptoms of depression among community-residingelderly adults. Journal of Gerontology, 47, 367–372.

Williamson, G. M., & Schulz, R. (1993). Coping with specificstressors in Alzheimer's disease caregiving. The Gerontolo-gist, 33, 747–755.

Williamson, G. M., & Schulz, R. (1995). Activity restrictionmediates the association between pain and depressed affect:A study of younger and older adult cancer patients.Psychology and Aging, 10, 369–378.

Williamson, G. M., & Shaffer, D. R. (2000). The activityrestriction model of depressed affect: Antecedents andconsequences of restricted normal activities. In G. M.Williamson, D. R. Shaffer, & P. A. Parmelee (Eds.), Physicalillness and depression in older adults: A handbook of theory,research, and practice (pp. 173–200). New York: KluwerAcademic/Plenum.

Williamson, G. M., Schulz, R., Bridges, M. W., & Behan, A. M.(1994). Social and psychological factors in adjustment tolimb amputation. Journal of Social Behavior and Person-ality, 9, 249–268.

Williamson, G. M., Shaffer, D. R., & Schulz, R. (1998). Activityrestriction and prior relationship history as contributors tomental health outcomes among middle-aged and olderspousal caregivers. Health Psychology, 17, 152–162.

RECEIVED: January 9, 2007ACCEPTED: October 16, 2007Available online 20 April 2008