Overperception of Spousal Criticism in Dysphoria and Marital Discord

www.elsevier.com/locate/bt

Available online at www.sciencedirect.com

Behavior Therapy 39 (2008) 300–312

Overperception of Spousal Criticism in Dysphoriaand Marital Discord

David A. Smith, Kristina M. PetersonUniversity of Notre Dame

Depression and marital discord are related to feelingcriticized by others, especially by spouses (e.g., Hooley,J. M., & Teasdale, J. D. 1989). This study evaluated theextent to which criticism was overperceived in relation to“actual” spousal critical comments, with actual criticalcomments being established by independent observers andby criticizing spouses themselves. Using dyadic interactionand questionnaire data from 72 married couples, signaldetection and regression analyses suggested that bothdysphoria and marital discord were associated with ageneral bias towards feeling criticized. Marital discord'sassociation with criticality bias subsumed dysphoria's, butdysphoria's did not subsume marital discord's. Criticalitybias also accounted for a significant proportion of perceivedspousal criticism. A common cognitive process may underlieestablished associations among perceived criticism, dys-phoria, and marital discord.

DEPRESSION AND MARITAL DISCORD are related tofeeling criticized by others, especially spouses (e.g.,Hooley & Teasdale, 1989; Lynch, Robins, &Morse, 2001; Riso, Klein, Anderson, Ouimette, &Lizardi, 1996). It is unclear, however, whether thesereports reflect (a) actual exposure to hypercriticalspouses, (b) the generally negative views taken bypeople who are feeling depressed or maritally dis-

This research was supported by National Institutes of HealthGrant MH-066301-02. We thankMatthew J. Breiding and Dana C.Villines for their help on this project and Chaunce R. Windle for hercomments on an earlier version of this article.

Address correspondence to David A. Smith, Department ofPsychology, University of Notre Dame, 118 Haggar Hall, NotreDame, IN 46556-5636; e-mail: [email protected]/08/300–312$1.00/0© 2008 Association for Behavioral and Cognitive Therapies. Published byElsevier Ltd. All rights reserved.

cordant (e.g., Bradbury & Fincham, 1990; Epstein& Baucom, 2002; Gotlib & Krasnoperova, 1998;Ingram, Miranda, & Segal, 1998; Townsley, Beach,Fincham, & O’Leary, 1991), or (c) some combina-tion of spouse hypercriticality and criticality bias.The purpose of this study is to examine these pos-sibilities in a design that permits observational as-sessment of exposure to interspousal criticismindependent of self-reported perceived criticism.

Criticism and DepressionCriticism has been linked to depression relapse instudies of expressed emotion (EE; Hooley, Orley, &Teasdale, 1986; Hooley & Teasdale, 1989; Vaughn& Leff, 1976a). EE, which is indicated by familymembers' critical, hostile, or emotionally over-involved statements about their relatives, hastraditionally been assessed with the CamberwellFamily Interview (Brown& Rutter, 1966; Rutter &Brown, 1966; Vaughn & Leff, 1976b) or with theFive Minute Speech Sample (Magaña et al., 1986).The importance of EE is underscored by a study

of potential demographic and clinical relapsepredictors in a sample of patients hospitalized forunipolar depression (Hooley et al., 1986). Amongthe candidate demographic and clinical variablestested, only EE predicted relapse to a statisticallysignificant degree. A meta-analysis of this literaturesuggests that patients who are hospitalized for thetreatment of depression have a 2 to 3 times greaterrisk of relapse when they live with “high EE” rela-tives than when they live with “low EE” relatives(Butzlaff & Hooley, 1998).EE has been linked to poor outcomes in a number

of other psychological and medical conditions, suchas schizophrenia, eating disorders, rheumatoid ar-thritis, and diabetes (for a review, see Wearden,Tarrier, Barrowclough, Zastowny, & Rahill, 2000).

mailto:[email protected]

301cr i t i cal i ty b i a s in dysphor ia and mar i tal d i scord

However, the link between EE and relapse maybe especially strong with respect to depression. Ameta-analysis of studies examining EE and severalpsychological disorders revealed that EE was a sig-nificantly better predictor of relapse in mood dis-orders than in schizophrenia (Butzlaff & Hooley,1998). Furthermore, the criticism dimension ofEE may be especially toxic for those with depres-sion. Criticism from others may reinforce depressedpeople's own self-critical tendencies, validating andperpetuating depressive cognitions (Wearden et al.,2000). Relapse in depression is associated withlower levels of criticism than is relapse in schizo-phrenia, suggesting that depressed people may beparticularly sensitive to criticism (Hooley et al.,1986; Vaughn& Leff, 1976a). Additionally, amongEE, marital adjustment, and responses to thequestion, “How critical is your spouse of you?”,the single-item measure of perceived criticism wasthe best predictor of relapse following recoveryfrom depression (Hooley & Teasdale, 1989). Per-ceived criticism has been described as an especiallypotent relapse predictor because, beyond indicatingmere exposure to criticism, it reflects how muchcriticism is “getting through” to patients (Hooley&Teasdale, 1989).

Criticism and Marital DiscordCriticism can be considered a generic stressorbecause it has been linked not only to variousforms of medical and psychiatric pathology but alsoto marital dysfunction (Hayhurst, Cooper, Paykel,Vearnals, & Ramana, 1997; Hooley & Teasdale,1989; Lynch et al., 2001). In both observationaland self-report investigations of the “demanding-ness” dimension of the demand-withdraw pattern,criticism has been shown to correlate with andpredict marital adjustment (Christensen & Shenk,1991; Lynch et al., 2001). Gottman (1994) iden-tifies criticism as one of the “four horsemen of theApocalypse,” factors he believes to be especiallypredictive of divorce. Finally, according to themarital discord model of depression, poor maritaladjustment may lead to depression, in part becauseit undermines spousal support and increases stressand criticism (see Beach, Sandeen, & O’Leary,1990).

Depression and Marital DiscordProcesses such as interspousal criticism that relate toboth marital discord and depressed mood areparticularly interesting because they might accountfor at least part of the substantial relationship be-tween poor marital adjustment and depressivesymptoms (Dehle & Weiss, 1998; O'Leary, Chris-

tian, & Mendell, 1994; Scott & Cordova, 2002; seeBeach, Smith,&Fincham, 1994;Gotlib&Hammen,1992; Whisman, 2001, for reviews). For instance, arecent meta-analysis revealed that approximately16% of the variance in depressive symptoms can beexplained by level of marital adjustment (Whisman,2001). Longitudinally, depressive symptoms pre-dict later declines in marital adjustment (Beach &O'Leary, 1993a), and marital adjustment predictslevels of dysphoria over time, even after controllingfor earlier levels of depressive symptoms (Beach,Katz, Kim, & Brody, 2003; Beach & O'Leary,1993b). First episodes of major depressive disorderare often preceded bymarital and other interpersonalstressors (Whisman & Bruce, 1999).

Distinguishing “Perceived” and “Actual” CriticismAlthough the picture emerging from this briefsummary of the criticism literature is fairly con-sistent, it is worth remaining mindful of a poten-tially important distinction not being emphasized,namely, the distinction between criticism as it isperceived and criticism as it is expressed. On the onehand, EE ratings in the depression literature arebased on observations of criticism expressed duringa Camberwell Family Interview or Five MinuteSpeech Sample, both undertaken in the absence ofthe patient (e.g., Hooley et al., 1986; Magaña et al.,1986). On the other hand, perceived criticism isassessed via patient self-reports of their global ex-periences (e.g., Riso et al., 1996). In the maritalliterature, the prototypic observational codingstudy of 10-minute videotaped marital problem-solving interactions is concerned with expressedcriticism (Weiss & Summers, 1983). However,owing to their self-report nature, the more typicalmarital assessments are concerned with perceivedcriticism (Christensen & Shenk, 1991; Fiscella &Campbell, 1999; Lynch et al., 2001). Because thesedisparate literatures rely on different assessmentperspectives, synthesizing their results is perilous.The more broadly integrative the effort, the morereliant it would be on the untested assumption thatperceived and expressed criticism are substantiallyrelated to each other.Especially insofar as depression and marital dis-

cord are concerned, it is probably unwise to as-sume expressions and perceptions correspond witheach other. The general negativity bias that is knownto underlie interpersonal perceptions of people whoare depressed or maritally discordant (e.g., Brad-bury & Fincham, 1990; Epstein & Baucom, 2002;Gotlib & Krasnoperova, 1998; Ingram et al., 1998;Townsley, Beach, Fincham,&O’Leary, 1991)mighteven be defined as a lack of correspondence between

302 sm i th & peter son

what others express and what is perceived. Unfortu-nately, correlations between traditional measures ofperceived and expressed criticism do not directlyaddress the correspondence of interest, becauseperceived criticism ratings are notmadewith respectto the behavior from which expressed criticism isscored. That is, patients rating perceived criticismdo not do so based on their observations of theCamberwell Family Interviews or Five-MinuteSpeech Samples. In the present study, perceivedcriticism is assessed in reference to the same be-haviors used to assess expressed criticism. Ex-pressed criticism that participants directly rate forperceived criticism will be referred to as “actual”criticism in order to distinguish it from expressedcriticism that participants do not directly ratefor perceived criticism. We consider it actual cri-ticism because we are able to generate convergingassessments of this behavior—from the criticizersthemselves and from outside observers—that areindependent of any biases on the part of the par-ticipants who rate perceived criticism.

The Present InvestigationNo study has directly compared perceived andactual criticism during an interspousal inter-action. The purpose of this study is to directlycompare perceived and actual (viz. observed andintended) criticism and to clarify the nature of therelationships among marital discord, depressivesymptoms, and perceptions of criticism in marriage.Specifically, we will start by using signal detectionmethods (Macmillan & Creelman, 2005) to char-acterize the amount and variability of bias in per-ceiving criticism from one's spouse; that is, we willdescriptively characterize “criticality bias.” Second,we will examine the correlations among actual andperceived criticism, criticality bias, marital discordand dysphoria. Third, we will test for differencesbetween spouses who overperceive criticism andthose who underperceive criticism. Fourth, we willexamine redundancy in the associations amongcriticality bias, perceived criticism, dysphoria, andmarital discord. And lastly, we will explore whethercriticality bias moderates the association betweendysphoria and marital discord.

Methodparticipants

We recruited 72 married couples from St. JosephCounty, IN, using newspaper advertisements thatoffered $50.00 remuneration for participation ina “marital communication” study. The majorityof participants gave their race/ethnicity as White(82.6%), with 9.7% indicating African American,

2.8% indicating Hispanic, 0% indicating Asian,and 4.2% indicating another race/ethnicity (e.g.,“mixed”). One participant did not indicate a race/ethnicity. Couples had been married for an averageof 9.77 years (SD = 11.37; range = .25 to 53). Mostof the participants were in their first marriages(72.9%). Husbands averaged 38.07 years of age(SD = 12.70) and 14.13 years of education (SD =3.00). Wives averaged 36.46 years of age (SD =12.60) and 14.39 years of education (SD = 2.55).Couples averaged 1.83 children (SD = 1.46; range =0 to 6). Annual family incomes ranged from $0 to$150,000 (M = $46,467, SD = $29,564).

procedure

Couples participated in a 3-hour laboratory visitthat involved completing questionnaires, engagingin a 10-minute videotaped desired-change interac-tion (cf. Pasch & Bradbury, 1998), and rating theirown interaction video. Prior to the desired-changeinteraction, participants also undertook assess-ments that are not pertinent to the present inves-tigation, including a structured diagnostic interview(SCID-I / NP, Feb 2001 revision: First, Gibbon,Spitzer & Williams, 2001; Spitzer, Williams,Gibbon, & First, 1992; Williams et al., 1992) andan audiotaped assessment of EE, the Five MinuteSpeech Sample (Magaña et al., 1986).For the desired-change interaction task, spouses

met together with an experimenter, and a previouslyrandomly selected spouse was designated as the“target” for the upcoming interaction. For expositorypurposes, we will refer to the nontarget spouse as the“source,” though participants were not aware of thisdesignation. Each spouse then independently listedfive things they would like changed about the targetspouse and ranked these topics in order of impor-tance. At the point of generating topics, spousesunderstood that one of the topicswouldbe chosen fora subsequent videotaped discussion. The experimen-ter then chose the highest ranked item appearing onboth lists as the topic for discussion. If there were nooverlapping items, the experimenter asked permis-sion to disclose the target's list to the source spousewho then chose a topic from the target's list. Spousesdid not otherwise see or discuss each other's lists oftopics. Couples were then videotaped engaging in a10-minute discussion of the desired target spousechange. Two identically time-stamped videotapeswere made of this 10-minute interaction for sub-sequent independent rating by each spouse.Following the discussion, spouses were escorted

to separate rooms where they each watched theirown copy of a time-stamped videotape of the desiredchange interaction. Experimenters paused eachspouse's video after identical 30-second intervals.


When the videos were paused, target participantsrated the criticism they recalled perceiving duringthe 30 seconds of video just completed. Sourceparticipants, on the other hand, rated the criti-cism they recalled intending during the 30seconds of video just completed. This way, twoperspectives on the same interaction wereobtained; target spouses, who were the topic ofdiscussion, rated perceived criticism, while sourcespouses, who were not the topic of discussion,rated intended criticism. Dividing each 10-minuteinteraction into 30-second intervals produced 20paired instances of perceived and intended criti-cism for comparison.

observational measures

“Insider” ratings of intended and perceived cri-ticism. When prompted after each 30-second inter-val of their videotaped discussions, target spousesresponded to the question, “How critical was yourpartner being of you?” while source spouses rated“How critical were you intending to be of yourpartner?” during the just-completed 30-second inter-val. Participants responded on 0-to-3 scales (0 = notat all critical, 1 = slightly critical, 2 = moderatelycritical, and 3 = very critical).

“Outside” ratings of observed criticism. At leastfour outside coders independently rated critical com-ments made by the source spouse during the desired-change interaction, using scales developed specifi-cally for this investigation. Sets of outside coderswereapproximately balanced between sexes to control forpossible gender-related perceptual differences. Train-ing ofmost coders beganwith reading primary sourcearticles that provided definitions of criticism as it wasinstantiated via other major observational codingsystems (e.g., Magaña et al., 1986; Vaughn & Leff,1976b;Weiss& Summers, 1983). Specifically, coderswere trained to consider as critical any comment thatexpressed dislike, disapproval, or resentment of thespouse's personality or behavior. Later coders weretrained by previous ones, with periodic reliabilitychecks to alert us to any rater drift. Ratings weremade following each 30-second interval using a 0-to-3 scale (0 = no criticism, 1 = low intensity and shortduration criticism, 2 = high intensity or long durationcriticism, and 3 = high intensity and long durationcriticism, or global criticism of character). The 30-second intervals matched those used by spouses, andthe 0-to-3 rating scale was the same as the spouses' 0-to-3 scales. Prior to coding study tapes, sample tapeswere coded in groups, with coders discussing theirratings out loud in order to expose and resolvemisunderstandings and to clarify definitions. Codersthen rated study tapes independently of the othercoders.

Estimating the reliability of these outside observerratings was complex for several reasons. First, therewere different raters and different numbers of ratersduring different periods of the study. The minimumnumber of judges on any interaction was four, whilethe maximum number of judges was six. There werefour different configurations of coders in all, withsome coders appearing in more than one configura-tion. A second complexity is that 20 dependentobservations for each subject were coded, so judgeagreement is inflated owing to a large number ofratings (20) being nested within participants. Andlastly, to correspond with the level at which signaldetection indices were calculated, which is the levelat which substantive data analyses were under-taken, the reliability of individual 30-second inter-vals required estimation, not the reliability of ratingsaggregated across the entire 10-minute interaction.Subsets of couples coded by the same set of raters

were formed using a dummy blocking variable,representing each of the four coder subsets. Each ofthe twenty 30-second rating intervals, ICC(3, k),was calculated (Shrout & Fleiss, 1979) for eachrater configuration subset. These ICCs were thenconverted to Fischer's z' and averaged within ratersubsets. These four averages were then weighted bythe number participants rated by each coder subsetand averaged. After converting this grand averagefrom units of z' to units of ICC, the weightedaverage reliability of criticality expressed during anygiven 30-second interval, for any rater set, and anycouple, was .84. Importantly, there was no maineffect of rater configuration set on criticism ratings.

self-report measures

Depressive symptoms. Self-reported depressivesymptoms were assessed using the Beck DepressionInventory (BDI; Beck, Steer, & Garbin, 1988) or theBDI-II (Beck, Steer, & Brown, 1996). The first 19couples completed the BDI; the remaining 53couples completed the BDI-II. Scores from the BDIwere converted into BDI-II equivalent scores using aconversion table developed by Beck and colleagues(1996). The correlation between the BDI and BDI-IIhas ranged from .84 and .93 (Beck et al., 1996).The BDI and the BDI-II each consist of 21 items

assessing somatic, cognitive, and affective symptomsof depression, each item scored on a 0-to-3 scale, withhigher scores indicating greater levels of depressivesymptomatology. Like the BDI, the BDI-II has soundpsychometric properties, with internal consistencyestimates from a variety of samples in the .90 to .95range (Arnau, Meagher, Norris, & Bramson, 2001;Grothe et al., 2005; Steer, Cavalieri, Leonard, &Beck, 1999). The BDI-II evidences discriminantvalidity with respect to psychiatric symptoms other

Table 1Descriptive Statistics and Intercorrelations of Primary Variables

1 2 3 4 5 6 7

1. Depressive Symptoms __2. Marital Adjustment −.63*** __3. Perceivedglobal .38*** −.54*** __4. Intendedglobal .31* −.45*** .42*** __5. Perceivedlab .29* −.40*** .60*** .19 __6. Intendedlab .10 −.23 .37** .11 .59***, a __7. Observedlab .07 −.20 .29* −.08 .41*** .32** __M 10.03 106.10 2.04 2.64 .73 .72 .30SD 9.52 20.79 1.37 1.25 .67 .66 .32N 70 70 70 70 66 68 69

Note. * pb .05. ** pb .01. *** pb .001.aThe correlation between Perceivedlab and Intendedlab is computed between spouses.

1 Inasmuch as nominal scale agreement statistics, such asCohen's kappa, rely on high levels of exact numerical concordancebetween spouses/raters (i.e., a preponderance of values on the maindiagonal of a husband-by-wife rating scale matrix), we worriedthat the original 10-point Perceived Criticism scale created toomany opportunities for “near misses,” and that distributing even arelatively large number of couples into a 10×10 agreement matrixwould produce many empty cells and cells with frequencies in thesingle digits. Put more psychologically, we feared that a 10-pointscale was calibrated too finely for our participants, and that near-misses in the agreement matrix caused by their attempts to applythis fine-grained scale would markedly reduce kappa-type indices.Once we elected to deviate from the original 1-10 format of thescale, we then decided to invoke a few other preferences in theproduction of a replacement scale. First of all, we felt that thenumber “0” better represented the concept “not at all critical” thandid the number “1,” which is the anchor number of the Hooleyscale (Hooley & Teasdale, 1989). Secondly, we sought to avoid abias toward the middle, or “average” range of the scale, bychoosing an even number of response options (in our case 0-5 givessix response options, which offers participants no center value). Apreponderance of middle values could distort the marginaldistributions, biasing nominal scale agreement and reducingvalidity. Finally, in order to choose among possible even numbersfor the scale, we felt that when asking participants to reflect on thetotality of their experiences with their spouses (as opposed to theirexperiences during the 10-minute interaction), 4 scale points wastoo few, and 8 scale points was not sufficiently fewer than theoriginal 10 to justify the alterations.


than depression (Steer, Ball, Ranieri, & Beck, 1997),and construct and concurrent validity in its associa-tions with other measures of depression (Beck et al.,1996; Dozois, Dobson, & Ahnberg, 1998; Grotheet al., 2005; Steer et al., 1997; Storch, Roberti, &Roth, 2004; Whisman, Perez, & Ramel, 2000).The BDI-II mean and standard deviation for target

spouses are presented in Table 1. Cronbach's alphaestimate of the internal consistency of the BDI-II was.92 for the target spouses in this sample. A cutoff scoreof 18 has correctly classified 92% of patients withmajor depressive disorder (Arnau et al., 2001).According to this criterion, 17.1%of the target spouseswere experiencing potentially clinical levels of depres-sion. Amajority of the target participants (74.3%)wasexperiencing minimal or no depressive symptoms.

Marital adjustment. Marital adjustment was mea-sured with the Dyadic Adjustment Scale (DAS), a 32-item self-report scale assessing marital satisfaction,cohesion, consensus, and affective expression (Spanier,1976). Scores on the DAS range from 0 to 151, withlower scores indicating less marital adjustment (ormore marital discord). The DAS is a widely usedmeasure with excellent internal consistency reliability(N 90) aswell as compelling evidence of convergent anddiscriminant validity (Heyman Sayers, & Bellack,1994; Kazak, Jarmas, & Snitzer, 1988; Spanier,1976). The DAS is highly correlated with othermeasures of relationship conflict, such as the Qualityof Marriage Index (Heyman et al., 1994); it distin-guishes married from divorced couples; and scoreschange with improvements made during maritaltherapy (Spanier, 1976; Whisman& Jacobson, 1992).The DAS mean and standard deviation for target

spouses are presented in Table 1. Cronbach's alphaestimate of the internal consistency of the DAS was.95 for the target spouses in this sample. Using theconventional cutoff of 98, which has been adoptedbecause of evidence that it maximally separates

distressed clinic couples from nondistressed commu-nity couples (Eddy,Heyman,&Weiss, 1991), 27.1%of target participants in this sample were maritallydistressed.

Spousal criticism. Weused amodified versionof thePerceived Criticism scale developed by Hooley andTeasdale (1989) to assess criticism.This scale consistsof two questions: (a) “How critical is your spouse ofyou?” and (b) “How critical are you of yourspouse?” The original scale ranged from 1 (not atall critical) to 10 (very critical indeed). For purposesof the current study, the scale was changed to 0 (notat all critical) to 5 (extremely critical).1


The Perceived Criticism scale has demonstratedgood temporal stability and concurrent validitywith the Camberwell Family Interview (Van Hum-beeck, Van Audenhove, De Hert, Pieters, & Storms,2002). Likewise, it has evidenced discriminantvalidity by its lack of correlation with measures ofdepression or personality traits (Riso et al., 1996).

ResultsCriticism during the marital interaction task wasassessed from three different perspectives. Forpurposes of describing these three perspectives,“Perceivedlab” denotes target spouse reports ofperceptions of criticism received from their sourcepartners during the laboratory task. “Intendedlab”denotes source spouse reports of their intendedcriticism towards their target partners during thelaboratory task. “Observedlab” denotes outsideobserver ratings of criticism expressed by the sourcespouse toward the target spouse during thelaboratory task. In addition to these three micro-analytic indicators of specific criticisms expressedin 30-second intervals during the laboratory task,we assessed perceived and intended criticism glo-bally via the Perceived Criticism scale. The per-ceived criticism item “How critical is your spouseof you?” will be referred to as “Perceivedglobal”and the intended criticism item “How criticalare you of your spouse?” will be referred to as“Intendedglobal.”

criticism analyses

The means, standard deviations, and intercorrela-tions among the laboratory interaction measures(Perceivedlab, Intendedlab, and Observedlab), thegeneral criticism measures (Perceivedglobal andIntendedglobal), depressive symptoms (BDI-II), andmarital adjustment (DAS) are presented in Table 1.2

Inspection of the means and standard deviationsreveals that target participants were experiencingon average only mild depressive symptoms andwere generally not maritally discordant. Globally,target spouses reported that both they and theirpartners were moderately critical. On the labora-tory task, participants perceived and intended lessthan one mild critical comment each 30-secondinterval during the interaction. Observed criticism

2 It is important to bear inmind that Perceivedlab and Intendedlabare comparable, in that even though they come from differentpeople they are reports on the same person's criticality (viz. that ofthe source spouse). Conversely, Perceivedglobal and Intendedglobalcome from the same person reporting on two different people (viz.one's spouse and one's self). Hence, Intendedglobal and Intendedlabare not comparable because they not only come from differentreporters but also are in reference to different people.

occurred at a rate of less than one half criticalcomment per 30-second interval.A within-spouses dependent t-test revealed that

target spouses rated themselves as more globallycritical (Intendedglobal) than they viewed their par-tners to be (Perceivedglobal), t(69) = − 3.54, p b .001.Despite this mean difference, Perceivedglobal andIntendedglobal, were significantly correlated, r(70) =.42, p b .001, suggesting that while one's own globalintended criticism varies systematically with one'sglobal perceptions of criticism from one's spouse,intended criticism occurs at a recognizably higherlevel.These initial effects concerned perceived criticism

from one's spouse and intended criticism towardsone's spouse, both reported by the same person buttargeting two different people's criticism (one's ownand one's spouse's). More informative with respectto criticality bias are effects relating one's ownperceived criticism to one's spouse's intendedcriticism, inasmuch as these between-spouse effectsconcern two different people's reports on thecriticality of the same person. A between-spousedependent t-test revealed that target participantsperceived less criticism (Perceivedglobal) than theirpartners reported intending (Intendedglobal), t(69) =−3.93, p b .001. Despite this mean difference, one'sown Perceivedglobal was correlated with one'spartner's Intendedglobal, r(70) = .54, p b .001,suggesting a moderate level of correspondencebetween what is globally intended and what one'spartner perceives, though again, levels of intendedcriticism exceeded levels perceived.In the laboratory ratings, within-spouse ana-

lyses are not possible because only target spousesrated perceived criticism and only source spousesrated intended criticism. A between-spouse depen-dent t-test revealed no differences between per-ceived and intended criticism, with Perceivedlabversus Intendedlab, t(65) = −.26, ns.However, meanvalues for the outside ratings of observed criticismduring the laboratory task (Observedlab) weresignificantly lower than both Perceivedlab, t(64) =5.43, p b .001, and Intendedlab, t(66) = 5.28,p b .001. Intercorrelations between global andlaboratory ratings were also significant. Specifi-cally, Perceivedglobal and Perceivedlab were stronglyrelated, r(66) = .60, p b .001, as were Intendedglobaland Intendedlab, r(66) = .45, p b .001, suggestingnot only that people's global predispositions toperceive and intend criticism manifest themselvesduring this specific laboratory task, but also thatthe laboratory and global measures of criticism arevalid.Insofar as depressive symptoms and marital ad-

justment are concerned, Table 1 shows that as

Table 2Signal Detection Index Means (Standard Deviations)

P-I P-O

Hit Rate (H) .56 (.28) .56 (.28)False Alarm Rate (F) .37 (.28) .36 (.29)Sensitivity (d') .58 (.74) .67 (.66)Bias (c) .10 (.83) .12 (.87)

Note. P-I treats intended criticism as the criterion against whichperceived criticism is indexed. P-O treats observed criticism as thecriterion against which perceived criticism is indexed.

3 Obscured by the binary present/absent recoding is thepossibility that some spouses systematically express criticism inmore obvious ways than do others, making the task easier for theirpartners. If data are converted to binary format, and if maritallydiscordant or depressed participants are systematically exposed tomore obviously critical comments, then these participants mightreport more criticism, not because they are exposed to more orperceive more, but simply because their task is easier. Therefore,before committing to the binary representation, we correlated thenumbers of ratings of “1,” “2,” and “3” with scores on the BDI-IIand DAS. Results indicated that neither dysphoria nor maritaladjustment was correlated with the numbers of intended criticismratings of “1,” “2,” or “3.” Dysphoria and marital adjustment alsofailed to correlate significantly with the numbers of observedcriticism ratings of “1” or “2.” Only two participants receivedobserved criticism ratings greater that “2.”


expected, depressive symptoms (BDI-II) wereinversely associated with marital adjustment(DAS), r(70) = − .63, p b .001. Furthermore, globalperceived criticism (Perceivedglobal) and laboratoryperceived criticism (Perceivedlab) were significantlyassociated with both depressive symptoms, r(70) =.38, p b .001 and r(66) = .29, p b .05, respectively,and marital adjustment r(70) = − .54, p b .001 andr(66) = − .40, p b .001, respectively.We also sought to examine the previously un-

tested assumption that criticism coded by outsideobservers is strongly related to self-reported per-ceived criticism. The correlation between observedand perceived criticism during the laboratory taskwas r = .41, p b .001, suggesting perceived criticismcaptures a reliable proportion of what might beobtained by using outside observers. Similarly, thetraditional global perceived criticism item corre-lated significantlywith observed criticism during thelaboratory task, r = .29, p b .05, providing somemutual validation of the measure and of theobservational ratings. It is important to note, how-ever, that levels of observed criticism were signifi-cantly lower than both perceived and intendedcriticism during the laboratory interaction, suggest-ing either that spouses were biased in their reports ofcriticism or that the outside coders were insensitiveto criticisms that were apparent to spouses them-selves, perhaps owing to spouses' broader aware-ness of their own relationship's idiosyncrasies.

signaldetectionanalysesof interactiondata

To further clarify the nature of the relationshipsamong perceived, intended, and observed criticism,the laboratory data were subjected to signal detec-tion analysis to estimate sensitivity and bias inparticipants' perceptions of criticism. Two separatesignal detection analyses were conducted to com-pare (a) perceived and intended criticism (P-I) and(b) perceived and observed criticism (P-O).Recalling that each 10-minute interaction was

rated in 30-second intervals, to calculate signaldetection scores, each pair of scores (P-I, P-O) wascompared at each of the 20 time intervals to

determine hit (H) and false alarm (F) rates of eachtarget participant. Because of our interest in bias andsensitivity of perceived criticism, for P-I and P-Ocomparisons, I and O served as “signals” to bedetected or not. If couples had missing data for morethan 5 out of the 20 intervals, or if they had either nocriticism or else criticism on every interval, they wereexcluded from the signal detection analyses becausefalse alarm or hit rates could not be calculated orwould be based on insufficient numbers of observa-tions. As a result, for the final analyses, there were 54and 57 participants with P-I and P-O scores,respectively. Couples who were excluded did notdiffer systematically on any study variables from thosewho were retained. As shown in Table 2, on average,the hit rate (H) was just over .55, and the false alarmrate (F) was just over .35. The hit rate suggestsparticipants were missing almost half (45%) of thecriticisms that were either intended by their spouses orobserved by others. The false alarm rate shows arelatively pronounced tendency to report criticismsthat were neither intended nor observed by others.The tendency to miss criticisms at a rate in excess ofthe rate of inappropriately reporting them gives rise toa general positive bias (described below), which isperhaps not surprising in this nonclinical sample.After hit and false alarm rates were computed,

bias and sensitivity were calculated. Ordinarilywith a graded response (0-3) measure one woulduse additive ROC procedures (Macmillan & Creel-man, 2005, pp. 126-130), but in our data very fewcouples used all four response options (usuallyneglecting the rating of “3”). In order to preservethe sample size, therefore, we elected to collapseresponses into the more familiar binary form,comparing zero (0) to nonzero (1, 2, and 3) ratings.This recoding then represents the simple absenceversus presence of criticism during each 30-secondinterval.3

Table 3Correlates of Perceived Criticality Bias

P-Ibias P-Obias

Depressive symptoms −.22 −.30*Marital adjustment .40** .44***Perceivedglobal −.52*** −.53***Intendedglobal −.05 −.21

Note. * pb .05. ** pb .01. *** pb .001.


Although there are several parametric binarysensitivity and bias indices to choose from, wechose d-prime,

dV¼ zðHÞ ¼ zðFÞ;

where z is the standard normal deviate, H is the hitrate, and F is the false alarm rate (Macmillan &Creelman, 2005). We indexed bias using c, where

c ¼ �:5½zðHÞ þ zðFÞ�;

and z, H, and F are defined as above. Bias andsensitivity calculations produce infinite values forperfect hit and false alarm rates. Therefore, assuggested by Macmillan & Creelman (2005, p. 8),we converted perfect proportions of 0 and 1 to 1/(2N) and 1-1/(2N), where N = the number of trialsin the proportion.Means and standard deviations for criticality

sensitivity and bias are presented in Table 2. Onaverage, participants did not very successfullydistinguish the presence from the absence ofcriticism (sensitivity). Out of a maximum possiblescore of 3.29, participants produced mean scores of0.58 and 0.67 for P-I and P-O, respectively. Out ofpossible criticality bias scores ranging from − 1.64(indicating there was criticism) to + 1.64 (indicatingthere was not criticism), with zero indicating nobias, participants produced mean bias scores of+ 0.10 and + 0.12 for P-I and P-O, respectively. Thepositive sign on these means reflects a generalpositive bias, or the tendency to report there was nocriticism. The magnitude of this reporting tendencycan be approximated by considering these means asa percentage of the maximum bias (+ 1.64),suggesting overall bias in the +6-7% range. Whilethe practical importance of effects of this size mightbe questioned, when it is remembered that themarital interaction studied likely occurs many timesover the course of one's marriage, the cumulativeeffect of such a frequently occurring, albeit small,effect could be substantial.

correlates of sensitivity and bias

t-tests revealed no gender differences on any of thesensitivity or bias measures. Similarly, moderatedmultiple regression analyses revealed no genderinteractions with either sensitivity or bias in relationto any other study variables. Therefore, analyses ofbias and sensitivity will not be broken down bygender. As well, sensitivity was not significantlyrelated to depressive symptoms, marital adjustment,or the global tendency to report one's spouse as cri-tical. Apparently, the ability ormotivation to carefully

discriminate between the presence and absence ofcriticism is either not a stable individual difference, orif it is a stable individual difference, it is one that is notreliably related to the other personal and relationshipphenomena we investigated. Bias, on the other hand,evidenced significant correlates.As expected, depressive symptoms were signifi-

cantly associated with a bias towards feeling criticized(Table 3) when target ratings of criticism were com-pared to outside observer ratings of the same cri-ticisms (P-O), r(57) = −.30, p b .05. Bias in targetcriticism ratings relative to spouse's intended criticismratings was also related to depression symptoms,though this effectwas not statistically significant, (P-I),r(54) = − .22, p N .05. Decreased marital adjustmentwas also associated significantly with criticalitybias, r(57) = .44, p b .001 and r(54) = .40, p b .01 forP-O and P-I, respectively. Furthermore, as predicted,bias was associated with a global tendency to viewone's spouse as critical, r(54) = − .52, p b .001 andr(57) = − .53, p b .001, for P-I and P-O respectively.To say that bias correlates with depressive symp-

toms or with marital adjustment is not to say thatrelatively depressed or maritally discordant partici-pants are “biased,” either positively or negatively, inan absolute sense. Table 2 shows an overall positivebias, which raises the possibility that our “criticality”bias measure reflects relatively more or less positivebias rather than involving negative bias, per se. Toexamine actual criticality bias, we tested whetherparticipants obtaining negative bias scores actuallydiffered from those obtaining positive bias scores.Independent sample t-tests comparing partici-

pants with positive P-O bias to those with nega-tive P-O bias were statistically significant for theBDI-II, t(55) = 2.17, p b .05, the DAS, t(55) = − 3.00,p b .01, and Perceivedglobal,, t(55) = 4.22, p b .001.Hence, participants who overperceive criticism dur-ing the laboratory task are relativelymore depressed,maritally discordant, and predisposed to reportingglobal criticism than are participants who under-perceive criticism. Interestingly, after classifyingparticipants into those with negative bias scoresand those with positive bias scores, we found that themeans of criticism overperceivers (those with nega-tive bias scores) on both the BDI-II (13.36) and on the

FIGURE 1 Moderation of the association between depressivesymptomatology (BDI-II) andmarital adjustment (DAS) by criticality bias.


DAS (96.92) were on the distressed sides of tradi-tional clinical thresholds (13 and 98, respectively).Means of criticism underperceivers (those withpositive bias scores) were on the nondistressedsides of these thresholds (8.06 on the BDI-II and113.31 on the DAS). In relation to intended criti-cism, P-I overperceivers reported relatively moremarital discord, t(52) = − 2.76, p b .01, and morePerceivedglobal, t(52) = 4.72, p b .001, but did notdiffer on depressive symptomatology, t(52) = 1.44,ns, from P-I underperceivers.

explication of bias, marital adjustment,depressive symptoms, and criticism

Because marital adjustment, depressive symptoma-tology, and global perceived criticality were sig-nificantly intercorrelated, we were concerned aboutredundancy among these correlations. That is, wefelt it was important to determine whether the sixcorrelations among these four constructs repre-sented six independent main effects or variations ona smaller set of more fundamental effects that gaverise to the observed associations. Therefore, weundertook a series of multiple regression analyses toestimate the relative contributions of each of thefour constructs to the pair-wise associations amongthe others. We also tested whether the associationsbetween various pairs of variables were altered byother variables, using moderated multiple regres-sion (Aiken & West, 1991). That is, we askedwhether some of the obtained effects were especiallystrong (or weak) for people at varying levels of theothers. Each regression was computed separatelyfor both P-I bias (cP-I) and P-O bias (cP-O), producingtwo regression results for each of the constructs ofinterest (marital discord, depressive symptomatol-ogy, and criticality). All effects described in thissection are significant at the p b .05 level, unlessotherwise noted.

Marital discord. To help discern whether maritaldiscord might be associated with depressive symp-tomatology because of their respective associationswith criticality bias, we cast marital discord in therole of dependent variable, regressing it on bias,depressive symptomatology, and their interaction.This analysis showed that the association betweendepressive symptoms and marital adjustment wasexacerbated by increases in bias and, equivalently,the association between bias and marital adjust-ment was strengthened as depressive symptomsincreased. As an aid to interpretation of theseeffects, Fig. 1 provides a simple slopes plot of theregression of the BDI-II on the DAS at negative,zero, and positive bias values (cP-O). The plot for cP-Iis not given because it is nearly identical. Theseresults suggest that while both bias and depressive

symptomatology contribute independent incre-ments to the prediction of marital adjustment, theeffect of depressive symptoms is especially strongfor spouses who are more negatively biased (and,equivalently, the effect of criticality bias is strength-ened by depressive symptoms).

Depressive symptomatology. To help discernwhether depressive symptomatology was associatedwith marital discord because they were eachassociated with bias, we cast depressive symptoma-tology in the role of dependent variable, regressing iton bias, marital discord, and the bias-by-discordinteraction. The marital discord association withdepressive symptomatology was not moderated bybias, and, equivalently, the association betweenmarital discord and bias was not moderated bydepressive symptomatology. This analysis furthershowed that marital adjustment contributed sig-nificantly to the prediction of depressive sympto-matology above and beyond bias, but bias did notsignificantly increment the predictability of depres-sive symptomatology beyond that predictable frommarital adjustment alone.

Global perceived criticism. Finally, we returnedto global perceived criticism and examined theextent to which the association between depressivesymptoms and perceived criticism was attributableto criticality bias. For this analysis, we regressedPerceivedglobal on the BDI-II and either cP-I or cP-O.When bias was indexed relative to partner-intendedcriticism, depressive symptomatology did not pro-vide additional predictive validity, though themagnitude of the depressive symptomatology effectwas similar (unstandardized B-coefficient of 0.03)to that associated with cP-O (B=0.04). When biaswas indexed relative to outside observers, depres-sive symptoms contributed significantly to theprediction of global perceived criticism, suggestingthat depressive symptoms were associated withglobal perceived criticism beyond the extent towhich their criticism ratings are biased.


Discussion

Weexamined the relationbetweenperceived criticismfrom one's spouse and “actual” spousal criticalcomments, with actual critical comments establishedboth by independent observers and by the criticizingspouses themselves. Although there was a smallpositive bias overall (+6-7%), spouses with negativebias scores differed from those with positive biasscores on both marital discord and depressive symp-tomatology, with negative bias spouses averaging inthe clinical range on measures of both discord anddysphoria. These findings are consistent with theoperation of self-and relationship-destructive cogni-tive biases that increase the perceived criticality ofone's partner relative to that partner's actualcriticality. Conversely, it would also be correct tosuggest that we found self-and relationship-enhan-cing biases that reduce the perceived criticality ofone's partner, inasmuch as criticality bias scoresextended into the positive range in this sample.This pattern of findings is reminiscent of Beck's

(1967) early suggestion that the negative cognitivebiases observed in depressed patients are counterpartsto positive cognitive biases observed in nondepressedpeople. He argued that the normal capacity to distortin order to enhance the self can be distressing if it getsdeployed against the self. Social psychologists haveextended these ideas to interpersonal relationshipswith their work on the relationship-enhancing effectsof perceptual distortions (e.g., Murray, Holmes, &Griffin, 1996a; Simpson, Gangestad, & Lerma,2004). The present study demonstrates not only thiskind of positive bias in close relationships, but it alsodemonstrates negative bias in relationships. Our find-ings are also consistent with more general cognitivetheories of both marital discord and depression.Results suggest independent contributions of both theenvironment (actual criticality) and perceptions of theenvironment (criticality bias) in personal and relation-shipdysfunction.Data such as these could be adducedin support of clinical directives to attend both tocognitive and to behavioral aspects of depression anddiscord (e.g., Beck, Rush, Shaw and Emery, 1979;Epstein & Baucom, 2002).Had we not indexed perceived criticality to actual

criticality using signal detectionmethods, it would beambiguous whether perceived criticality reflectedcognitive biases or genuine partner criticality. Whilethese data establish bias as a component of perceivedcriticality, it isworth emphasizing that bias andactualcriticality are both important, at least insofar asdysphoria is concerned. While the relation betweenbias and perceived criticality was significant, so toowas the association between perceived criticality anddysphoria, even after controlling for bias. Therefore,

it would be inappropriate to attribute reports ofperceived criticism solely to bias. Instead, our datasuggest that, at least in this instance, there areseparable and independent effects of actual criticismand criticality bias. As suggested by Hooley andTeasdale (1989), reports of perceived criticismappearto reflect how much criticism is “getting through” tospouses. As such, not only do reports of perceivedcriticism reflect the actual criticism directed at theperceiver, but they also reflect the perceiver'scognitive bias towards over-or underperceiving cri-ticism. It is perhaps because global perceived criticismreflects both criticality bias and actual criticism, thatit is such a strong predictor of response to treatment(Chambless & Steketee, 1999) and relapse fromdepression (Hooley & Teasdale, 1989), relative to“actual” observed criticism as measured, say, by theCamberwell Family Interview or Five Minute SpeechSample, which do not capture criticality bias. If so,criticality bias might be a helpful additional focus ofinterventions for depression and marital discord,beyond the more traditional goal of reducing actualcriticism through communication and problem-sol-ving training (e.g., Beach et al., 1990; Jacobson et al.,1991; O'Leary & Beach, 1990).Neither the causal sequence nor the causal mech-

anisms connecting criticality bias and actual criticismare presently known. It is fairly easy to imagine howprior experiencewith a critical spouse could heightenthe expectation of future criticality, producing asubsequent criticality bias (Gotlib&Whiffen, 1989).It is less easy to imagine the reverse causal direction,that is, to imagine how prior criticality bias mayproduce later actual partner criticality. Nevertheless,there is evidence that “positive illusions” are asso-ciated with later reduced conflict in some couples(Murray, Holmes, & Griffin, 1996b), and this mightbe a complement process to the one we investigated,which was a kind of “negative illusion.” Therefore,causal paths between criticality bias and actualcriticality are possible in both directions, and it willrequire longitudinal and experimental studies todetermine the relative contributions of each path.Only marital discord made a unique contribution

when dysphoria and discordwere considered togetherin relation to criticality bias. This is consistent withprevious data suggesting that marital cognitions areespecially closely related to marital distress and de-pressive cognitions are especially closely related todepression (e.g., Townlsey et al., 1991). Our measureof bias was gleaned from amarital interaction task, soit is perhaps not surprising that itwasmore closely tiedto broader relationship functioning than it was todepressive symptomatology. Similarly, in a sample ofdepressed outpatients, perceived criticism was notcorrelated with any maladaptive personality traits,


individual functioning, or with social relationshipfunctioning in general, but perceived criticism wasrelated to marital adjustment (Riso et al., 1996). Al-though perceived criticism's performance as a psy-chopathology prognostic indicator is prominent inthis literature, it is worth considering whetherperceived criticism is as much or more of an indicatorof relationship functioning than of personal function-ing. Our results suggest this may be especially true ofthe criticality bias component.Apart from its performance as a specific indicator

of marital cognition, criticality bias may also be anindicator of a more general latent cognitive style—aglobal cognitive style that is also involved in de-pression (e.g., pessimism). In the case of depression,the focus of the latent cognitive style would be moreon the self than on the spouse or the marriage. Itmay be the common association of criticality biasand depressive symptoms with the latent globalcognitive style that produces the association be-tween them. Delineation of this latent cognitivestyle would be of great theoretical and practicalimportance as it could help integrate a number ofotherwise disparate areas of inquiry.The final purpose of this study was to measure

and compare both global reports of perceived andintended criticism with laboratory coded criticismduring a structured interaction task. Previous stud-ies have compared perceived and intended criticismwithout knowledge of actual criticism (Hooley &Teasdale, 1989) or have inferred actual levels ofspousal criticism from interviews about patients inthe absence of the patient, assuming behavior in theinterview reflects the way relatives actually interactwith the patient (Hooley, Orley, & Teasdale, 1986;Vaughn & Leff, 1976a). Our results support thevalidity of global reports of perceived criticism asindicators of actual criticism coded by outside ob-servers, supporting the previously untested assump-tion that criticism coded by outside observers (e.g.,the Camberwell Family Interview and Five MinuteSpeech Sample) is strongly related to perceivedcriticism.4

4 It might be imagined that asking participants to code theirspouses' Camberwell Family Interviews or Five Minute SpeechSamples would provide a direct test of the assumed overlapbetween expressed and perceived criticism, but that would notnecessarily be an improvement over the estimates made in thepresent investigation, primarily because the Camberwell FamilyInterview and the Five Minute Speech Sample are both conductedin the absence of the participant. Hence, the behavior coded by theparticipant would be behavior to which they had not been exposed.Spouses are likely to speak about the participants differently thanthey speak to the participants.

This study employed novel measures of criticismandnovelmethodsof data analysis, so some caution isappropriate when interpreting its findings. It isreassuring to note, however, that the laboratory inter-action measures of perceived and intended criticismcorrelated quite highly with their global counterparts,rs=.60 and .45, respectively, psb .001. Global andlaboratory perceived criticism also converged in theirassociations with depressive symptomatology andmarital adjustment. Intended laboratory criticism,however, correlated with depressive symptomatologyand marital adjustment differently than did globalintended criticism, suggesting possibly less convergentvalidity for intended criticism or possibly greaterdetermination of intended criticism by the topic beingdiscussed in the lab. Nevertheless, that laboratoryintended criticism was significantly associated withlaboratory perceived criticism reported by one'sspouse further supports the reliability and validity ofthese measures, as does their significant associationswith observed criticism.Inasmuch as this study was conducted with a

nonclinical sample, the implications of our findingsformajor depressive disorder andmarital discord arepresently unclear. However, we did observe an in-verse association between subclinical depressivesymptomatology and marital adjustment that wasof a magnitude (r=−.63) generally observed inclinical samples (Whisman, 2001). Unless the samephenomenon emerges for different reasons in clinicaland nonclinical samples, replication of this founda-tional association gives some basis upon which tohope that the other interpersonal aspects of depressionwe found may also be evident in the clinical range. Forinstance, extrapolation of the moderated associationdepicted in Fig. 1 suggests stronger criticality biaseffects in clinical samples, even if perceived criticismeffects getweaker. Finally, it is interesting to note that inthis community sample, perceived criticismwas relatedto depressive symptoms while previous studies of de-pressed or remitted depressed patients led us to expectthat perceived criticism would be unrelated to depres-sive symptoms (Hooley & Teasdale, 1989; Riso et al.,1996). It is unclear whether this constitutes a failure toreplicate a nonfinding or the emergence of an effectbecause of the broader community sampleweused, butbecause some findings replicated quite well and othersdid not, the overall generalizability of our findings toclinical samples remains an empirical issue for furtherstudy.

ReferencesAiken, L. S., & West, S. G. (1991). Multiple regression: Testing

and interpreting interactions. Newbury Park, CA: Sage.Arnau, R. C., Meagher, M. W., Norris, M. P., & Bramson, R.

(2001). Psychometric evaluation of the Beck Depression


Inventory-II with primary care medical patients. HealthPsychology, 20, 112–119.

Beach, S. R. H., Katz, J., Kim, S., & Brody, G. H. (2003). Pros-pective effects ofmarital satisfaction on depressive symptomsin established marriages: A dyadic model. Journal of Socialand Personal Relationships, 20, 355–371.

Beach, S. R. H., & O'Leary, K. D. (1993a). Dysphoria andmarital discord: Are dysphoric individuals at risk for maritalmaladjustment? Journal of Marital and Family Therapy, 19,355–368.

Beach, S. R. H., & O'Leary, K. D. (1993b). Marital discord anddysphoria: For whom does the marital relationship predictdepressive symptomatology? Journal of Social and PersonalRelationships, 10, 405–420.

Beach, S. R. H., Sandeen, E. E., & O'Leary, K. D. (1990).Depression in marriage: A model for etiology and treatment.New York: Guilford Press.

Beach, S. R. H., Smith, D. A., & Fincham, F. D. (1994). Maritalinterventions for depression: Empirical foundation andfuture prospects. Applied and Preventive Psychology, 3,233–250.

Beck, A. T. (1967). Depression. New York: Harper and Row.Beck, A. T., Rush, A. J., Shaw, B. F., & Emery, G. (1979).

Cognitive therapy of depression. New York: Guilford Press.Beck, A. T., Steer, R. A., & Garbin, M. G. (1988). Psychometric

properties of the Beck Depression Inventory: Twenty-fiveyears of evaluation. Clinical Psychology Review, 8, 77–100.

Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Manual forBeck Depression Inventory – II. San Antonio, TX: Psycho-logical Corporation.

Bradbury, T. N., & Fincham, F. D. (1990). Attributions inmarriage: Review and critique. Psychological Bulletin, 107,3–33.

Brown, G.W., &Rutter, M. (1966). The measurement of familyactivities and relationships: A methodological study.HumanRelations, 19, 241–263.

Butzlaff, R. L., & Hooley, J. M. (1998). Expressed emotion andpsychiatric relapse. Archives of General Psychiatry, 55,547–552.

Chambless, D. L., & Steketee, G. (1999). Expressed emotionand behavior therapy outcome: A prospective study withobsessive-compulsive and agoraphobic outpatients. Journalof Consulting and Clinical Psychology, 67, 658–665.

Christensen, A., & Shenk, J. L. (1991). Communication,conflict, and psychological distance in nondistressed, clinic,and divorcing couples. Journal of Consulting and ClinicalPsychology, 59, 458–463.

Dehle, C., & Weiss, R. L. (1998). Sex differences in prospectiveassociations between marital quality and depressed mood.Journal of Marriage and the Family, 60, 1002–1011.

Dozois, D. J. A., Dobson, K. S., & Ahnberg, J. L. (1998). Apsychometric evaluation of the Beck Depression Inventory-II. Psychological Assessment, 10, 83–89.

Eddy, J. M., Heyman, R. E., & Weiss, R. L. (1991). Anempirical evaluation of the Dyadic Adjustment Scale:Exploring the differences between marital “satisfaction”and “adjustment”. Behavioral Assessment, 13, 199–220.

Epstein, N. B., & Baucom, D. H. (2002). Enhanced cognitive-behavioral therapy for couples: A contextual approach.Washington, DC: American Psychological Association.

First, M. B., Gibbon, M., Spitzer, R. L., & Williams, J. B. W.(2001). SCID-I-TR: Structured Clinical Interview for DSM-IV-TR Disorders (February 2001 Revision). BiometricsResearch, New York.

Fiscella, K., & Campbell, T. L. (1999). Association of perceivedfamily criticism with health behaviors. Journal of FamilyPractice, 48, 128–134.

Gotlib, I. H., & Hammen, C. L. (1992). Psychological aspectsof depression: Toward a cognitive-interpersonal integration.Oxford, England: John Wiley and Sons.

Gotlib, I. H., & Krasnoperova, E. (1998). Biased informationprocessing as a vulnerability factor for depression. BehaviorTherapy, 29, 603–617.

Gotlib, I. H., & Whiffen, V. E. (1989). Stress, coping, andmarital satisfaction in couples with a depressed wife.Canadian Journal of Behavioural Science, 21, 401–418.

Gottman, J. M. (1994).What predicts divorce? The relationshipbetween marital processes and marital outcomes. Hillsdale,NJ: Lawrence Erlbaum Associates.

Grothe, K. B., Dutton, G. R., Jones, G. N., Bodenlos, J.,Ancona, M., & Brantley, P. J. (2005). Validation of the BeckDepression Inventory-II in a low-income African Americansample of medical outpatients. Psychological Assessment,17, 110–114.

Hayhurst, H., Cooper, Z., Paykel, E. S., Vearnals, S., &Ramana, R. (1997). Depressed emotion and depression.British Journal of Psychiatry, 171, 439–443.

Heyman, R. E., Sayers, S. L., & Bellack, A. S. (1994). Globalmarital satisfaction versus marital adjustment: An empiricalcomparison of three measures. Journal of Family Psychol-ogy, 8, 432–446.

Hooley, J. M., Orley, J., & Teasdale, J. D. (1986). Levels ofexpressed emotion and relapse in depressed patients. BritishJournal of Psychiatry, 148, 642–647.

Hooley, J. M., & Teasdale, J. D. (1989). Predictors of relapse inunipolar depressives: Expressed emotion, marital distress,and perceived criticism. Journal of Abnormal Psychology,98, 229–235.

Ingram, R. E., Miranda, J., & Segal, Z. V. (1998). Cognitivevulnerability to depression. New York: Guilford Press.

Jacobson, N. S., Dobson, K., Fruzzetti, A. E., Schmaling, K. B.,& Salusky, S. (1991). Marital therapy as a treatment fordepression. Journal of Consulting and Clinical Psychology,59, 547–557.

Kazak, A. E., Jarmas, A., & Snitzer, L. (1988). The assessmentof marital satisfaction: An evaluation of the DyadicAdjustment Scale. Journal of Family Psychology, 2, 82–91.

Lynch, T. R., Robins, C. J., & Morse, J. Q. (2001). Couplefunctioning in depression: The roles of sociotropy andautonomy. Journal of Clinical Psychology, 59, 1349–1359.

Macmillan, N. A., & Creelman, C. (2005). Detection theory: Auser's guide (2nd ed.).NewYork: CambridgeUniversity Press.

Magaña, A. B., Goldstein, M. J., Karno, M., Miklowtiz, D. J.,Jenkins, J., & Falloon, I. R. H. (1986). A brief method forassessing expressed emotion in relatives of psychiatricpatients. Psychiatry Research, 17, 203–212.

Murray, S. L., Holmes, J. G., & Griffin, D. W. (1996a). Thebenefits of positive illusions: Idealization and the construc-tion of satisfaction in close relationships. Journal ofPersonality and Social Psychology, 70, 79–98.

Murray, S. L., Holmes, J. G., &Griffin, D.W. (1996b). The self-fulfilling nature of positive illusions in romantic relation-ship: Love is not blind, but prescient. Journal of Personalityand Social Psychology, 71, 1155–1180.

O'Leary, K. D., & Beach, S. R. H. (1990). Marital therapy: Aviable treatment for depression and marital discord. Amer-ican Journal of Psychiatry, 147, 183–186.

O'Leary, K. D., Christian, J. L., & Mendell, N. R. (1994). Acloser look at the link between marital discord anddepressive symptomatology. Journal of Social and ClinicalPsychology, 13, 33–41.

Pasch, L. A., & Bradbury, T. N. (1998). Social support, conflict,and the development of marital dysfunction. Journal ofConsulting and Clinical Psychology, 66, 219–230.


Riso, L. P., Klein, D. N., Anderson, R. L., Ouimette, P. C., &Lizardi, H. (1996). Convergent and discriminant validity ofperceived criticism from spouses and family members.Behavior Therapy, 27, 129–137.

Rutter, M., & Brown, G. W. (1966). The reliability andvalidity of measures of family life and relationships infamilies containing a psychiatric patient. Social Psychiatry,1, 38–53.

Scott, R. L., & Cordova, J. V. (2002). The influence of adultattachment styles on the association between maritaladjustment and depressive symptoms. Journal of FamilyPsychology, 16, 199–208.

Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations: Usesin assessing rater reliability. Psychological Bulletin, 86,420–428.

Simpson, J. A., Gangestad, S. W., & Lerma, M. (2004).Perception of physical attractiveness: Mechanisms involvedin the maintenance of romantic relationships. In H. T. Reis,& C. E. Rusbult (Eds.), Close relationships: Key readings(pp. 305–316). Philadelphia: Taylor and Francis.

Spanier, G. B. (1976). Measuring dyadic adjustment: Newscales for assessing the quality of marriage and similardyads. Journal of Marriage and the Family, 38, 15–28.

Spitzer, R. L., Williams, J. B., Gibbon, M., & First, M. B.(1992). The Structured Clinical Interview for DSM-III-R(SCID): I. History, rationale, and description. Archives ofGeneral Psychiatry, 49, 624–629.

Steer, R. A., Ball, R., Ranieri, W. F., & Beck, A. T. (1997).Dimensions of the Beck Depression Inventory-II in clinicallydepressed outpatients. Journal of Clinical Psychology, 55,117–128.

Steer, R. A., Cavalieri, T. A., Leonard, D. M., & Beck, A. T.(1999). Use of the Beck Depression Inventory for primarycare to screen for major depression disorders. GeneralHospital Psychiatry, 21, 106–111.

Storch, E. A., Roberti, J. W., & Roth, D. A. (2004). Factorstructure, concurrent validity, and internal consistency of theBeck Depression Inventory–Second Edition in a sample ofcollege students. Depression and Anxiety, 19, 187–189.

Townsley, R. M., Beach, S. R. H., Fincham, F. D., & O'Leary,K. D. (1991). Cognitive specificity for marital discord anddepression: What types of cognition influence discord?Behavior Therapy, 22, 519–530.

Van Humbeeck, G., Van Audenhove, C., De Hert, M., Pieters,G., & Storms, G. (2002). Expressed emotions: A review ofassessment instruments. Clinical Psychology Review, 22,321–341.

Vaughn, C. E., & Leff, J. P. (1976a). The influence of family andsocial factors on the course of psychiatric illness: Acomparison of schizophrenic and depressed neuroticpatients. British Journal of Psychiatry, 129, 125–137.

Vaughn, C., & Leff, J. (1976b). The measurement of expressedemotion in the families of psychiatric patients. BritishJournal of Social and Clinical Psychology, 15, 157–165.

Wearden, A. J., Tarrier, N., Barrowclough, C., Zastowny, T. R.,& Rahill, A. A. (2000). A review of expressed emotionresearch in health care. Clinical Psychology Review, 20,633–666.

Weiss, R. L., & Summers, K. J. (1983). Marital Interaction CodingSystem–III. In E. E. Filsinger (Ed.), Marriage and familyassessment: A sourcebook for family therap (pp. 85–115).Beverly Hills, CA: Sage.

Whisman, M. A. (2001). The association between depressionand marital dissatisfaction. In S. R. H. Beach (Ed.), Maritaland family processes in depression: A scientific foundationfor clinical practice (pp. 3–24). Washington, DC: AmericanPsychological Association.

Whisman, M. A., & Jacobson, N. S. (1992). Change in maritaladjustment following marital therapy: A comparison of twooutcome measures. Psychological Assessment, 4, 219–223.

Whisman, M. A., & Bruce, M. L. (1999). Marital dissatisfac-tion and incidence of major depressive episode in acommunity sample. Journal of Abnormal Psychology, 108,674–678.

Whisman, M. A., Perez, J. E., & Ramel, W. (2000). Factorstructure of the Beck Depression Inventory–Second Edition(BDI-II) in a student sample. Journal of Clinical Psychology,56, 545–551.

Williams, J. B. W., Gibbon, M., First, M. B., Spitzer, R. L.,Davies, M., Borus, J., et al. (1992). The Structured ClinicalInterview for DSM-III-R (SCID): 2. Multisite test-retestreliability. Archives of General Psychiatry, 49, 630–636.

RECEIVED: September 1, 2006ACCEPTED: September 4, 2007Available online 20 April 2008

Documents

Overperception of Spousal Criticism in Dysphoria and Marital Discord