Assessment Volume 16 Number 2

June 2009 145-158 2009 SAGE Publications

10.1177/1073191108324440http://asmnt.sagepub.com

hosted athttp://online.sagepub.com

145

Factor Structure and MeasurementInvariance of the Cognitive FailuresQuestionnaire Across the Adult Life SpanPhilippe RastDaniel ZimprichUniversity of ZurichMartin Van BoxtelJellemer JollesUniversity of Maastricht

The Cognitive Failures Questionnaire (CFQ) is designed to assess a persons proneness to committing cognitive slipsand errors in the completion of everyday tasks. Although the CFQ is a widely used instrument, its factor structureremains an issue of scientific debate. The present study used data of a representative sample (N = 1,303, 24-83 yearsof age) from the Maastricht Aging Study (MAAS) to test and compare factor solutions for the CFQ previouslyreported in the literature by means of confirmatory factor analysis of ordered categorical variables. A three-factormodel of the CFQ from an exploratory factor analysis was tested for increasing levels of measurement invarianceacross six age groups. Factor (co-)variances remained stable across the age groups, mean differences were observedfor the factor Forgetfulness with higher means for older participants, and the factor Distractibility where participantsolder than 60 years of age had lower means.

Keywords: cognitive failures; measurement invariance; factor structure; ordered categorical; life span

Cognitive slips and errors attract attention notonly in everyday lifesometimes as comicalmistakes such as putting flour in ones own coffee oras more serious lapses such as turning the wrongway in a one-way streetbut also in psychology(Broadbent, Cooper, FitzGerald, & Parkes, 1982;Reason, 1988; Wallace, 2004). A number of researchershave examined the mechanisms underlying everydayslips and failures. Norman (1981) subdivided cogni-tively based slips and failures into three categories ofmistakes: errors in the formation of intentions, faultyactivation of schemas, and false triggering, that is,interrupted processing of sequences of actions. Incontrast, Reason (1990) attributed failures to inatten-tion and overattention (see Hasher & Zacks, 1979).Inattention leads to capture slips in which subjects

cannot complete actions as wished, omissions follow-ing interruptions, reduced intentionality, perceptualconfusions, and interference errors. Overattentionresults in omissions, repetitions, and reversals.

Another prominent account of everyday slipsand errors is proposed by Broadbent et al. (1982). Acognitive failure may involve perceptual failures,failures of memory, or misdirected physical actions.The common element is that there is a departure fromthe normal smooth flow of function, and events donot proceed in accordance with intention (p. 1). Theassumption underlying cognitive failures is that vari-ous perceptual, action, and memory failures are influ-enced by a general and rather enduring factor. Thisfactor might be described as a general proneness orliability to cognitive failures which should be rela-tively independent of traditional personality andintelligence measures (see Klumb, 2001).

The Cognitive Failures Questionnaire (CFQ)To assess the frequency of everyday cognitive fail-

ures, Broadbent et al. (1982) developed the Cognitive

Authors Note: Please address correspondence to Philippe Rast,Gerontopsychology, Department of Psychology, University ofZurich, Binzmhlestr. 14/24, CH-8050 Zurich; e-mail: p.rast@psychologie.uzh.ch. Philippe Rast and Daniel Zimprich con-tributed equally to this article.

at HOWARD UNIV UNDERGRAD LIBRARY on February 19, 2015asm.sagepub.comDownloaded from

Failures Questionnaire (CFQ), which comprises 25items derived from three areas of slips and errors: per-ception slips, memory slips, and slips in motor func-tioning. These items were assembled from failureepisodes which the majority of people would accept asoccurring to them at least occasionally. Respondentsare offered examples such as Do you fail to noticesignposts on the road?, Do you read something andfind you havent been thinking about it and must readit again?, Do you bump into people? and are askedto report the frequency of these incidents in the past 6months on a 5-point Likert-type scale.

A number of studies have shown that cognitive fail-ures, as measured by the CFQ, are related to absent-mindedness (Reason & Lucas, 1984), slowperformance on focused attention tasks (Meiran,Israeli, Levi, & Grafi, 1994), automobile accidents andwork accidents (Larson & Merritt, 1991; Wallace &Vodanovich, 2003), dissociative experiences(Merckelbach, Muris, & Rassin, 1999), daytime sleepi-ness and boredom proneness (Wallace & Vodanovich,2003), computing losses because of forgetting to saveones data in humancomputer interaction (Jones &Martin, 2003), and reduced cognitive inhibition(Bloem & Schmuck, 1999). The CFQ has goodtestretest reliability, with stability coefficients atapproximately rtt = .80 across 6 to 65 weeks, indicatinga high degree of stability of individual differences(Broadbent et al., 1982; Merckelbach, Muris, Nijman,& de Jong, 1996; Vom Hofe, Mainemarre, & Vannier,1998). The same authors provided coefficient alphameasures for the CFQ at approximately .90 althoughMerckelbach et al. reported somewhat lower alpha val-ues in 3 samples, ranging from .75 to .81, implyingadequate internal consistency for research purposes.Wallace (2004) examined the convergent and discrim-inant validity of the CFQ. He examined the associationbetween CFQ total scores and measures of comparable(e.g., absentmindedness) as well as opposite constructs(e.g., conscientiousness and everyday memory). In asample of 386 undergraduate students he found that thefrequency of self-reported cognitive failures correlatedpositively (rs = .50 to .53) with similar constructs,whereas the associations with opposite constructs werenegative (rs = .13 to .41). The broad acceptance andusefulness of the CFQ are also reflected by the fact thatthe CFQ has been translated into several languages, forexample, Dutch (Merckelbach et al., 1996), German(Klumb, 1995), Hebrew (Meiran et al., 1994),and Spanish (Garca Martnez & Snchez-Cnovas,1994). In summary, the CFQ is a commonly used

questionnaire which has proved to be a useful instru-ment to identify individuals prone to cognitive failures.

Factor Structure of the CFQIn most applied studies, the sum score across all

CFQ items is used as a measure of proneness toeveryday slips and errors based on the assumptionthat the CFQ captures a general liability to cognitivefailure. In accordance with this assumption, Broadbentet al. (1982) conducted a number of factor analyses indifferent samples and concluded that a single, generalfactor adequately captured the structure of the CFQ.The authors found that apart from the obvious generalfactor (p. 5) results were rather variable. Subsequently,several investigators reexamined the factor structure ofthe CFQ and their results question the notion of onlyone factor (Larson, Alderton, Neideffer, & Underhill,1997; Pollina, Greene, Tunick, & Puckett, 1992;Wagle, Berrios, & Ho, 1999; Wallace, 2004; Wallace,Kass, & Stanny, 2002). Details regarding these modelscan be retrieved from Wallace (2004) where in theappendix a tabular comparison of models is given. HenceBroadbent and colleagues assumption that a generalfactor adequately describes the CFQ is contrasted bylater findings that propose more than one factor.

Almost all researchers used principal componentsanalysis (PCA) to identify the dimensional structureof the CFQ. PCA is a data reduction procedure and isnot considered the best approach to identify latentfactors (Floyd & Widaman, 1995). Matthews, Coyle,and Kraig (1990) administered the CFQ to a sampleof 475 college students. They found two components,a general component and an additional componentrelating to memory for names, although the latterfactor constituted only two items. Larson et al. (1997)examined the structure of the CFQ in a sample of2,379 American Navy recruits. By their own asser-tion, the two components appeared to incorporate ahodgepodge of different types of items (p. 31) andthus were not meaningfully interpretable. In conclu-sion, the authors argued for a general component interms of Broadbent et al. (1982) and a memory fornames component. In a recent study with 335 par-ticipants (223 undergraduate students and 112 U.S.Navy personnel), Wallace and colleagues (2002)reported a four-factor solution that emerged from aPCA followed by varimax rotation: Memory,Distractibility, Blunders, and Names. In a subsequentconfirmatory factor analysis (CFA) in a sample of709 university students, this solution was replicated

146 Assessment

at HOWARD UNIV UNDERGRAD LIBRARY on February 19, 2015asm.sagepub.comDownloaded from

(Wallace, 2004). Pollina et al. (1992) examined thestructure of the CFQ in a sample of 387 collegestudents. A PCA yielded five components: dis-tractibility, misdirected actions, spatial or kinaes-thetic memory, interpersonal intelligence, andmemory for names. Only three components, however,were considered reliable of which distractibilityalone accounted for 27% of the variance.

To summarize, findings have been mixed. Thestructures of the presented solutions differed acrossauthors both with respect to their content and com-plexity. One to five component solutions have beenreported, but only a few were replicable in indepen-dent samples. In fact, only the solution by Wallaceet al. (2002) was retested and confirmed by means ofCFA (Wallace, 2004). This heterogeneity in resultsmay stem in part from the approach used to extractthe alleged factors. By relying on PCA, the possibil-ity of different solutions is maximized by the inclu-sion of error variance for each time. The common useof varimax rotation is also problematic as factors areforced to be independent, which is unlikely to accu-rately characterize the relationships between cogni-tive failure factors (Preacher & MacCallum, 2003).Another complication of the previous literature is thatinvestigations of the factor structure of the CFQ havebeen mainly based on young, adult populations. Thusit is unclear whether any of the previously presentedsolutions can be generalizedboth in terms of thegeneral structure and with respect to measurementpropertiesto other populations.

Cognitive Failures Across the Life SpanAn underrepresented aspect in previous research

on cognitive failures is whether the frequency of self-reported slips, errors, and lapses changes across thelife span in cognitively intact adults (but see Boomsma,1998). There are, however, reasons to expect that theself-reported frequency of some cognitive failuresincreases into old age. Lay impressions hold thatolder adults are more forgetful, absentminded, andclumsy than younger adults (Heckhausen, Dixon, &Baltes, 1989) all of which are attributes that form partof cognitive failures. More generally, it was foundthat attributes carrying negative connotations, such asbeing indicative of memory failures or cognitive failures,are believed to be more pronounced in older personsboth by younger and older adults (Lineweaver &Hertzog, 1998). Consistent with these lay impressions,

if adults are asked to judge their own cognitive ormemory functioning, usually a negative relation betweenage and self-reported cognitive or memory perfor-mance emerges (Bolla, Lindgren, Bonaccorsy, &Bleecker, 1991; Derouesn, Lacomblez, Thibault, &LePoncin, 1999; Hertzog, Hultsch, & Dixon, 1998).At the same time, individual differences in subjectiveassessments of ones own cognitive or memory function-ing are only weakly related to individual differencesin ones actual cognitive and memory performance asmeasured by psychological tests (Hertzog & Hultsch,2000; Ponds, van Boxtel, & Jolles, 2000; Zimprich,Martin, & Kliegel, 2003), implying that subjectivejudgments of cognitive functioning are only partlybased on objective performance. An explanatoryaccount for these findings was offered by McDonald-Miszczak, Hertzog, and Hultsch (1995), who pro-posed a social cognition framework, which posits thatimplicit knowledge about a general decline of cogni-tive functioning in old age might bias judgments ofolder persons about their own cognitive functioningtoward the general expectation of decline. Similarly,Cavanaugh, Feldman, and Hertzog (1998) pointedout that memory failures may be seen as part of acommon self-theory of aging: When asked about per-sonal memory beliefs, older adults are more likely toaccess memory-failure concepts and to make disposi-tional evaluations relative to young adults or relativeto ones own past.

Based on these arguments, one might hypothesizethat the self-reported frequency of cognitive failuresincreases with advanced age. This may be especiallytrue for failures associated with memory problems.

Aims of the Present StudyThe present study had three aims. First, we set out

to find an adequate factorial representation of theCFQ in a large, representative sample covering thewhole adult life span. To do so, we investigated pre-viously reported factor solutions by means of confir-matory factor analysis and compared these models toa solution derived from an exploratory factor analysisof the present data. Second, measurement invarianceof the most viable factor structure was tested acrosssix age groups to examine whether the CFQ is unbi-ased with respect to age. After establishing strictmeasurement invariance of the CFQ across agegroups, the third aim was to investigate age differ-ences in factor covariances, variances, and means.

Rast et al. / Measurement Invariance of the CFQ Across Age 147

at HOWARD UNIV UNDERGRAD LIBRARY on February 19, 2015asm.sagepub.comDownloaded from

Method

ParticipantsThe sample for this study comprised individuals

from the Maastricht Aging Study (MAAS), a longitu-dinal study on the biological determinants and cogni-tive consequences of normal aging stratified by age,sex, and occupational achievement. In an early phaseof MAAS, the sample was obtained through the reg-istration network of family practices (RNH) super-vised by the Department of General Practice of theUniversity of Limburg. All participants had low riskfor dementia; individuals with documented centralnervous system (CNS) pathology or mini-mentalstate examination (MMSE) scores below 24 wereexcluded (for a detailed description of inclusion cri-teria and sampling methodology refer to Jolles, Houx,van Boxtel, & Ponds, 1995).1 The main study ofMAAS consisted of four cross-sectional panels, A1 toA4, sharing the same methodology with respect tosample frame, subject inclusion, stratification crite-ria, and basic measurement protocol. In the first waveof the MAAS study, the CFQ was part of the assess-ment in three panels, A2 to A4, summing to 1,354participants. In the present study, participants rangingin age from 24 to 83 years (M = 51.2, SD = 16.2) whohad complete data records with respect to the CFQwere included. Fifty-one participants (3.8% of thetotal sample) were excluded from further analyses asthey did not provide complete data records concern-ing the CFQ, constituting a sample size of N = 1,303participants, 49% of them female. Missingness ofCFQ data was unrelated to age, gender, and educa-tional level. The sample was split into six age groups,which, in the remainder of this study, will be referredto as Group 1 (age: 24-33 years, M = 27.9, SD = 2.9)the reference group, Group 2 (age: 34-43 years, M =38.1, SD = 2.7), Group 3 (age: 44-53 years, M = 47.7,SD = 2.6), Group 4 (age: 54-63 years, M = 57.9,SD = 2.7), Group 5 (age: 64-73 years, M = 67.8, SD =2.7), and Group 6 (age: 74-83 years, M = 76.3, SD =2.1; for descriptive statistics see Table 1). Across thesix age groups, there were no differences in the pro-portion of female participants (2 = 0.88, df = 5, p >.97). Age groups, however, differed significantly inlevel of formal education (F = 35.83, df = 5, 1297, p /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages false /GrayImageMinResolution 150 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages false /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /FlateEncode /AutoFilterGrayImages false /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages false /MonoImageMinResolution 300 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages false /MonoImageDownsampleType /Bicubic /MonoImageResolution 1200 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /FlateEncode /MonoImageDict > /AllowPSXObjects true /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile () /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName (http://www.color.org) /PDFXTrapped /False

/CreateJDFFile false /Description > /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ > > /FormElements true /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles true /MarksOffset 6 /MarksWeight 0.250000 /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /DocumentCMYK /PageMarksFile /RomanDefault /PreserveEditing true /UntaggedCMYKHandling /UseDocumentProfile /UntaggedRGBHandling /UseDocumentProfile /UseDocumentBleed false >> ]>> setdistillerparams> setpagedevice