Suspicious minds: The motive to acquire

relationship-threatening information

WILLIAM ICKES,a JEREMY W. DUGOSH,b JEFFRY A. SIMPSON,c AND

CAROL L. WILSONcaUniversity of Texas at Arlington, bAmerican Board of Internal Medicine (Philadelphia)and cTexas A & M University

AbstractIn four studies, we obtained evidence for the reliability and validity of a 21-item scale designed to measure a new

theoretical construct: individual differences in the motivation to acquire relationship-threatening information

(hereafter, MARTI). Study 1 provided evidence for the MARTI scale’s reliability and discriminant validity,

revealing that it was reliable and not significantly correlated with measures of the Big Five personality traits, adult

attachment styles, or more general social orientations. Studies 2 and 3 provided evidence for the scale’s convergent

and discriminant validity, showing that dating partners with higher MARTI scores (i.e., those who were more

motivated to acquire relationship-threatening information) scored lower in relational trust and reported engaging in

more ‘‘suspicion behaviors.’’ Study 4 provided behavioral evidence for the scale’s predictive validity, revealing that

(a) dating partners with higher MARTI scores were more likely to break up within 5 months, and (b) the breakup

rate was most pronounced for dating partners who scored higher on the scale and who also reported being less

close. We discuss how this new construct and measure can be used to study important relationship dynamics.

We can’t go on togetherWith suspicious minds.

—Lines from the 1969 Elvis Presley hit

Was Elvis right? Are the relationships ofpartners who have suspicious minds moresusceptible to break-ups than the relation-ships of partners who don’t? If they are,how do suspicious minds contribute to thedemise of intimate relationships? Using arecently developed measure of individualdifferences in the motive to acquire relation-ship-threatening information (MARTI), we

address these and other questions in theresearch presented below.

Our decision to study this motive wasguided by recent theory and research on theempathic accuracy of couples in relationship-threatening situations (see Ickes & Simpson,1997, 2001; Simpson, Ickes, & Blackstone,1995; Simpson, Ickes, & Grich, 1999).Accordingly, we begin with a brief reviewof this theory and research, which high-lights the importance of assessing the ac-curacy motives of the partners in closerelationships.

Ickes and Simpson’s Empathic

Accuracy Model

In a recent revision (Ickes & Simpson, 2001)of our original empathic accuracy model

Studies 1–3 were supported by National ScienceFoundation grant SBR-9732476 to Jeffry A. Simpsonand William Ickes. Study 4 was supported by a grantfrom the Timberlawn Psychiatric Research Foun-dation to William Ickes. We thank Ira H. Bernsteinfor his advice regarding the exploratory andconfirmatory factor analyses reported in Study 1.

Requests for reprints should be sent to WilliamIckes, Department of Psychology, University ofTexas at Arlington, Arlington, TX 76019-0528;e-mail: ickes@uta.edu.

Personal Relationships, 10 (2003), 131–148. Printed in the United States of America.Copyright # 2003 ISSPR. 1350-4126/02

131

(see Ickes & Simpson, 1997), we attemptedto account for the complex relationsbetween empathic accuracy and relation-ship satisfaction and stability by positing ageneral rule and two major exceptions tothe rule. As a general rule, we proposedthat empathic accuracy tends to be goodfor close relationships. In the majority ofroutine, everyday interactions in whichknowledge of a partner’s thoughts and feel-ings carries little or no potential to threatenor destabilize a relationship, greaterempathic accuracy should promote thekind of mutual understanding that enablespartners to coordinate their individual andshared goals and actions and thereby main-tain a more satisfying and stable relation-ship.

There are, however, two importantexceptions to this general rule. The firstexception involves situations in which oneor both partners, recognizing that greaterempathic accuracy might damage their rela-tionship, display motivated inaccuracy inorder to buffer the relationship from thedissatisfaction and instability that mightotherwise occur. The second exceptioninvolves situations in which, despite realiz-ing that greater empathic accuracy mightdamage the relationship, one or both part-ners become hypervigilant and displaymotiv-ated accuracy with respect to their partner’sunexpressed thoughts and feelings. Thissecond exception to the general rule—whichis pertinent to individuals with suspiciousminds—is the primary focus of the presentresearch.

The general rule: Empathic accuracyhelps relationships

Many studies in the marital adjustment lit-erature have found a positive associationbetween marital adjustment and accuratelyunderstanding the attitudes, role expect-ations, and self-perceptions of one’s spouse.Ickes and Simpson (1997, 2001), for exam-ple, identified more than 20 studies in whichsuch an association was reported, and manyof these studies were cited in an earlierreview by Sillars and Scott (1983).

Though fewer relevant studies exist in themore recent empathic accuracy literature,some of these studies also suggest that part-ners’ accuracy in inferring each other’sthoughts and feelings tends to predict bet-ter relationship outcomes. The connectionbetween empathic accuracy and positiverelationship outcomes appears to be particu-larly strong during the early stages ofrelationship development. During this period,greater empathic accuracy may enablepartners to anticipate and avoid conflicts asthey learn new patterns of behavior that willeventually facilitate mutual accommodationwithin the relationship (cf. Ickes, Stinson,Bissonnette, & Garcia, 1990; Kilpatrick,Rusbult, & Bissonnette, 2002; Thomas,Fletcher, & Lange, 1997). During the laterstages of relationships, heightened empathicaccuracy may be beneficial in a more subtleway by cueing mutually accommodativebehaviors after they have become habitual(Kilpatrick et al., 2002).

The first exception: Sometimes motivatedinaccuracy helps relationships

Ickes and Simpson (2001) proposed twoexceptions to the general rule that empathicaccuracy is beneficial to relationships inmundane, nonthreatening situations. Bothexceptions are presumed to occur in thoserelatively infrequent interaction contexts inwhich relationship partners suspect—orknow for certain—that their partner is har-boring thoughts and feelings that theymight be better off not knowing (i.e.,thoughts or feelings that, if accuratelyinferred, could damage the relationship).

The first exception occurs when oneor both partners are motivated to misinfertheir partner’s potentially threateningthoughts and feelings, thereby sparing them-selves and their relationship the pain andinjury that might otherwise be experienced.In these cases, motivated inaccuracy providesan important exception to the general rulethat partners will benefit from a full andaccurate understanding of each other.

Simpson et al. (1995) reported prelimin-ary evidence that partners may use motivated

132 W. Ickes et al.

inaccuracy to defend against impendingrelationship threats. Heterosexual datingpartners tried to infer each other’sthoughts and feelings during a taskdesigned to threaten their relationships.Specifically, the partners took turns ratingand discussing with each other the physicaland sexual attractiveness of several opposite-sex target persons as potential datingpartners. In the high-threat condition, thepotential dating partners were all highlyattractive individuals; in the low-threatcondition, they were all below average inattractiveness. Following the rating-and-discussion task, each dating partner triedto infer his or her partner’s private thoughtsand feelings from a videotape of their inter-action during the task.

As predicted, the lowest empathic accur-acy was exhibited by partners who werecloser (i.e., more interdependent) and lesssecure about the long-term stability oftheir relationship, and who rated attractive(vs. less attractive) potential dating partnersin each other’s presence. Moreover, therelation between these three variables andempathic accuracy was mediated by the per-ceiver’s degree of self-reported threat duringthe rating-and-discussion task. In fact, thecouples who were the most threatened—those who were closer, less secure abouttheir relationships, and who also ratedhighly attractive opposite sex persons—displayed near-chance levels of empathicaccuracy. Nevertheless, all of these coupleswere still dating four months later, whereasthe remaining couples in the study had asignificantly higher breakup rate (28%).These results suggest that when partnersfind themselves in relationship-threateningsituations that cannot be avoided, they mayuse motivated inaccuracy to protect theirrelationships from the damage that mightresult from more accurately ‘‘reading’’ eachother’s thoughts and feelings.

The second exception: Sometimes motivatedaccuracy hurts relationships

Individuals enter relationships with priorbeliefs and expectations about themselves

and their partners, and about how relation-ships tend to function (Collins & Read,1994). According to Bowlby (1973), individ-uals with a history of receiving intermittent,unpredictable care and support fromsignificant others develop low self-esteemand are preoccupied with the possibility oflosing or being abandoned by their roman-tic partners. To guard against this possibil-ity, these highly anxious/preoccupiedindividuals become hypervigilant withregard to what their partners are thinkingand feeling, particularly when they sensethat their relationships might be in jeopardy(Cassidy & Berlin, 1994).

One interesting implication of this line ofreasoning is that the highly anxious/preoccu-pied individuals in the Simpson et al. (1995)study should have responded very differentlyfrom the rest of the sample. Instead of dis-playing motivated inaccuracy in the relation-ship-threatening situation, these individuals(identified by high scores on the anxietyattachment dimension) should have displayedhypervigilance and increased accuracy wheninferring their partners’ potentially threaten-ing thoughts and feelings.

To test this intriguing possibility,Simpson, Ickes, and Grich (1999) recodedand reanalyzed the data from the Simpsonet al. (1995) study, using anxious attach-ment scores as a predictor variable. Consist-ent with the hypervigilance prediction, themore anxiously attached participants in thestudy (particularly women) displayedgreater empathic accuracy. That is, highlyanxious individuals displayed enhancedaccuracy rather than motivated inaccuracywhen asked to infer their partners’ poten-tially threatening thoughts and feelings.They also reported feeling more distressed,threatened, and jealous, and displayedother signs of relationship dissatisfactionand instability.

In effect, these highly anxious individualsbehaved as if they had suspicious minds thatcompelled them to know, at the first sign ofthreat, what their partner was thinking andfeeling. Although their suspiciousness mayhave allowed them to gauge the severity ofa threat more accurately by giving them

Suspicious minds: The MARTI scale 133

clearer insights into their partner’s destabil-izing thoughts and feelings, it also carried ahigh price in terms of the correspondingemotional and relational distress it engen-dered. Overall, then, these findings offeredpreliminary support for the second excep-tion to the rule in our theoretical model:Sometimes motivated accuracy can hurtrelationships.

Presumably, the enhanced accuracy ofanxiously attached individuals is motivatedby their desire to gauge the degree of rela-tionship threat, and this motive should beroutinely activated in relationship-threaten-ing situations. But what if a more generalmotive of this type exists—one that reflectsa more chronic and enduring need toacquire and confront relationship-threaten-ing information? Would the relationships ofindividuals who score high in this motivealso exhibit the instability and vulnerabilityobserved in the relationships of dating part-ners who scored high in anxious attach-ment? Conversely, would the relationshipsof dating partners who score low in thismotive reveal the protective, buffering effectof avoiding information that could poten-tially threaten the relationship?

We addressed these questions in the fourstudies reported below. Our first step was todevelop a self-report measure designed toidentify individuals who had suspiciousminds. This measure, which we have labeledthe MARTI scale, assesses the strength ofthe motive to acquire (vs. avoid) relation-ship-threatening information. The items onthis scale (see the Appendix) were written bythe first author and were designed to applyto dating relationships. They were struc-tured so that the most relationship-threat-ening items are located at or near the end ofthe measure. Items having to do with apartner’s (hypothetical) sexual infidelityconcern protected (‘‘safe’’) sex rather thanunprotected (‘‘unsafe’’) sex so that the issueof threats to one’s relationship will not beconfounded with the issue of threats toone’s physical health. Finally, only posi-tively keyed (relationship-threatening)items are included in the scale. Attempts towrite negatively keyed (reverse-scored)

items were not successful because pilotdata revealed that such items contributedlittle or no variance (i.e., virtually allrespondents reported wanting to knowevery positive and reassuring thing theycould about their dating relationship).

In Study 1, we report evidence for thereliability and discriminant validity of theMARTI scale. In Studies 2 and 3, we reportadditional evidence for the scale’s conver-gent and discriminant validity—evidencethat links individuals’ MARTI scores withtheir scores on measures of relational trust(Study 2) and suspicion behaviors (Study 3),but not with their scores on measures ofsocially desirable responding or hostile/paranoid forms of suspiciousness (Study 3).In Study 4, we report behavioral evidenceregarding the predictive validity of theMARTI scale, showing that the relation-ships of couples who are motivated toacquire relationship-threatening informationare, in fact, more likely to break up inresponse to a major destabilizing influence(less subjective closeness).

Study 1

In Study 1, we examined the internal consis-tency of the MARTI scale and confirmed thatit is a unidimensional measure of the strengthof the motive to acquire versus avoid relation-ship-threatening information. We establishedthe uniqueness of the measure by testing itsdiscriminant validity against the Big Five per-sonality traits, adult attachment styles, andmore general dimensions of social relating.

Method

Participants

The participants were 159 undergraduatesat the University of Texas at Arlington whocompleted a questionnaire study on the‘‘characteristics of partners in dating rela-tionships.’’ Each student received partialcourse credit in an introductory psychologyclass for participating in the study. Datawere excluded for two participants becausethey did not follow the instructions.

134 W. Ickes et al.

Sociodemographic characteristics

Of the 157 participants whose data wereused, 84 were women (53.5%) and 73 weremen (46.5%). The racial and ethnic compos-ition of the sample reflected the generalcomposition of the student body at theUniversity of Texas at Arlington. Based onself-reports, the sample was composed of 21African Americans (13.4%), 23 Asians(14.6%), 25 Hispanics (15.9%), 1 NativeAmerican (0.6%), 84 Whites (53.5%), and3 Others (2%). Participants ranged in agefrom 18–36 years, with a mean age of 20.6years, and a standard deviation of 3.54 years.In terms of dating status, 30 participants(19.1%) were not dating anyone at the timeof the study, 26 (16.6%) were dating boththeir current partners and others, and 85(54.1%) were dating their current partnerexclusively. Eight of the remaining partici-pants were engaged (5.1%) and 8 weremarried (5.1%).

Procedure and materials

When participants arrived for the study,they were given a brief introduction andthen signed a consent form. Each partici-pant was then given a survey packet con-taining the following self-report scales:

1. A short demographic questionnairethat assessed the participant’s gen-der, age, race/ethnicity, and currentdating status.

2. The 17-item Adult AttachmentQuestionnaire (AAQ; Simpson,Rholes, & Phillips, 1996), whichcontains two subscales that assessthe two orthogonal attachmentdimensions of avoidance and anxiety.The avoidance subscale meas-ures the tendency to retract fromintimacy in relationships and strivefor psychological and emotionalindependence; the anxiety subscaletaps preoccupied concerns aboutlosing one’s current partner or notbeing fully appreciated by him orher. Each item was answered on a7-point scale, anchored by 1¼ I

strongly disagree and 7¼ I stronglyagree. For evidence of the AAQ’sreliability and validity, see Simpsonet al. (1996).

3. The 18-item Social Orientation Scale(SOS; Ickes & Hutchison, 2000),which has two subscales that assessthe orthogonal dimensions of socialabsorption and social individuation.Social absorption is the tendency tobecome highly involved and interde-pendent with others (e.g., It’s easyfor me to get ‘‘in sync’’ with otherpeople and to ‘‘merge’’ with themduring the time we’re together).Social individuation is the tendencyto distinguish self from otherswith respect to attitudes, values,intentions, and motives (e.g., In myinteractions with others, I have aclear and definite sense of thedifference between my perspectiveand theirs). The two-factor structureof the SOS has been supportedby confirmatory factor analysesin four samples (see Ickes &Hutchison, 2000). Each item wasanswered on a 4-point scale,anchored by 1¼ very uncharacteris-tic of me and 4¼ very characteristicof me. For evidence of the reliabilityand validity of the SOS subscales,see Ickes, Hutchison, and Mashek(in press).

4. A 15-item measure of the Big Fivepersonality traits (Ickes, 1997),based on prototypical markers ofeach of the Big Five dimensionsreported by John (1990). Sampleitem dimensions are: quiet versustalkative and shy versus sociable(extraversion), friendly versus un-friendly and cold versus warm(agreeableness), responsible versusirresponsible and careful versuscareless (conscientiousness), tenseversus relaxed and calm versusanxious (neuroticism), and imagin-ative versus unimaginative andclosed-minded versus open-minded(openness). Each item dimension

Suspicious minds: The MARTI scale 135

was answered on a 1 to 7 scale, with4 as the neutral midpoint betweeneach adjective pair.

5. The 21-item Motive to AcquireRelationship-Threatening Information(MARTI) scale (see the Appendix).

After they had completed the surveypacket, all participants were thanked anddebriefed.

Results

Reliability of the measures

Cronbach alpha coefficients were computedto assess the internal consistency of eachscale. The alpha values for the subscales ofthe Adult Attachment Questionnaire were.71 for avoidance and .74 for anxiety. Simi-larly, the alpha values for the two subscalesof the Social Orientation Scale were .76 forsocial individuation and .74 for socialabsorption. The internal consistencies ofthe Big Five scales were more variable:extraversion (.84), neuroticism (.71), con-scientiousness (.68), openness (.61), andagreeableness (.52). Because the MARTIscale was intended to measure a single,unidimensional construct, we calculated itsinternal consistency as a single 21-itemscale. The alpha value was high (.85).

Factor analyses of the MARTI scale

One major goal of Study 1 was to testwhether the MARTI scale measured a sin-gle latent variable that was distinguishablefrom the Big Five personality traits, attach-ment styles, and other major social/rela-tional orientations. To achieve this goal,we conducted a series of confirmatory factor

analyses (CFAs) using all 21 items of theMARTI scale. We tested four CFA models:(i) the hypothesized single-factor (centroid)model, (ii) a model based on item means,(iii) a model based on item variances, and(iv) a model based on item skewness.1 Theresults indicated that the single-factor (cen-troid) model accounted for substantiallymore variance (49.4%, GFI fit index¼ .95)than did the models based on item means(13.6%, GFI¼ .63), item variances (16.4%,GFI¼ .70), or item skewness (10.3%,GFI¼ .57).

We next conducted an exploratory prin-cipal components analysis (PCA) that pro-duced five factors with eigenvalues greaterthan 1.0. To determine whether thisexploratory five-factor model provided abetter fit to the data than did the single-factor model, we performed a second CFAin which the single-factor (centroid) modelwas tested against the five-factor model.The five-factor model accounted for only16.5% of the total variance, much lessthan the 49.4% accounted for by the one-factor model.

These results clearly demonstrate thatthe MARTI scale is a single-factor measure.This conclusion is also supported by thescale’s face validity, the fact that the scaleitems were developed to tap a single psycho-logical construct, and the scale’s high inter-nal consistency.

Descriptive statistics

The average MARTI score for the 157 par-ticipants in the Study 1 sample was 25.0,with a standard deviation of 3.93, a skew-ness index of 1.12, and a kurtosis index of.78. The average MARTI score for the menin the sample (M¼ 25.4) did not differ sig-nificantly from the average score for thewomen (M¼ 24.6), t¼ 1.28, ns.21. Artifactual factors (historically known as ‘‘difficulty

factors’’) can occasionally emerge when sets of itemson a scale intercorrelate because they have similar(and typically extreme) means, variances, or skews(Bernstein & Teng, 1989; Carroll, 1945; Ferguson,1941). To ensure that the factors underlying theMARTI scale were content-based rather thanartifactual, we compared the one-factor (centroid)model solution against the other three.

2. There was also no significant difference between themen’s and the women’s MARTI scores in Studies 2,3, and 4. And because the participants’ gender didnot significantly moderate any of the findingsreported in these four studies, we will not discussthe gender variable further.

136 W. Ickes et al.

Discriminant validity: Correlations betweenthe MARTI scale and other measures

Given the CFA results, we next correlatedparticipants’ MARTI scores with theirscores on the other self-report measures.These analyses were performed to testwhether the motive to acquire relationship-threatening information is empirically dis-tinct from general personality traits andbasic social/relational orientations.

The results consistently supported thediscriminant validity of the MARTI scale.The participants’ MARTI scores were notsignificantly correlated with their scores onthe social orientation dimensions of socialabsorption (r¼ .05) or social individuation(r¼ .02). They were also not significantlycorrelated with Big Five scores on extra-version (r¼�.04), agreeableness (r¼ .01),conscientiousness (r¼�.13), neuroticism(r¼�.11), or openness to experience(r¼ .00). As a measure of the motive toacquire relationship-threatening informa-tion, therefore, the MARTI scale appearsto assess a construct that is both theoret-ically and empirically distinct from person-ality traits and general social orientations.

Ickes and Simpson (2001) speculatedthat the motive to acquire relationship-threatening information should be rela-tively independent of adult attachmentstyles. The rationale for this prediction isthat information motives are presumed tobe conscious and relatively consistent acrosssituations, whereas attachment motivesmay be largely unconscious and activatedonly in certain situations (i.e., those inwhich the threat of relationship loss seemsimminent; see Bowlby, 1973). Ickes andSimpson’s speculation was confirmed inthe present sample: MARTI scores wereessentially unrelated to participants’ scoreson both the avoidance (r¼ .05) and theanxious (r¼�.03) attachment subscales.These data provide further support for theconceptual distinctiveness of the MARTIscale as a measure of the motive to acquirerelationship-threatening information.

In summary, the results of Study 1revealed that the 21-item MARTI scale is a

highly reliable measure of a single con-struct. They further revealed that this con-struct is empirically distinct from the BigFive personality traits, the social orienta-tion dimensions of social absorption andsocial individuation, and the relationship-relevant dimensions of avoidant andanxious attachment styles. What was stilllacking, however, was evidence for the con-vergent and predictive validity of theMARTI scale. We sought to obtain suchevidence in the next three studies weconducted.

Study 2

In Study 2, we explored the convergentvalidity of the MARTI scale in terms of itsassociation with relationship trust. Peoplewho are strongly motivated to acquire rela-tionship-threatening information may seekto do so because they distrust their part-ners. Any acquired evidence that supportstheir initial distrust might only strengthenthis motive. On the assumption that a reci-procal relationship exists between distrustand the motive to acquire relationship-threatening information, we predicted thatcouples whose members were motivated toacquire relationship-threatening informa-tion (high MARTI scorers) would reportlower relationship trust than coupleswhose members were motivated to avoidrelationship-threatening information (lowMARTI scorers).

Method

Participants

Participants were 96 heterosexual datingcouples who completed a relationship ques-tionnaire before participating in a labora-tory study unrelated to the presentinvestigation. At least one member of eachcouple was enrolled in an introductory psy-chology class at the University of Texas atArlington and received partial course creditfor his or her participation. If the otherpartner was not also an introductory

Suspicious minds: The MARTI scale 137

psychology student, she or he was paid $10for participating.

Sociodemographic characteristics

Once again, the racial and ethnic compos-ition of the sample was similar to the generalcomposition of the student body at theUniversity of Texas at Arlington. Based onself-reports, the sample was composed of 28African Americans (14.5%), 14 Asians(7%), 31 Hispanics (16%), 1 NativeAmerican (0.5%), 111 Whites (58%), and 7Others (4%). Participants ranged in agefrom 18–50 years; their mean age was 21.6years, with a standard deviation of 3.50years. In terms of dating status, 14 weredating both their current partners andothers (7.3%) and 150 were dating theircurrent partner exclusively (78.1%). Of theremaining participants, 24 were engaged(12.5%) and 4 were married (2.1%).

Procedure and materials

As in Study 1, when the participants arrivedfor the study, they were given a brief intro-duction and then signed a consent form.Each participant was then given a surveypacket similar to the one used in Study 1. Thepacket contained the MARTI scale, the AdultAttachment Questionnaire, and some add-itional measures. These additional measuresincluded (i) a short demographic question-naire that assessed each participant’s gender,age, race/ethnicity, and current dating status,and (ii) the Trust Scale (Rempel, Holmes, &Zanna, 1985). Participants completed theirpackets in separate cubicles to ensure thattheir dating partner could not influencethem to censor or distort their responses.

Results

Cronbach’s alpha coefficients were com-puted to assess the internal consistency ofeach scale. The alpha values for the sub-scales of the AAQ were .74 for avoidanceand .72 for anxiety. The coefficient alphasfor the MARTI scale and the Trust Scalewere .90 and .79, respectively.

To examine the discriminant and con-vergent validity of the MARTI scale,correlations were computed between theparticipants’ MARTI scores and theirscores on the other measures. As in Study1, the zero-order correlations between theMARTI scale and the subscales of the AAQwere not significant (r¼ .03 for the avoid-ance subscale, r¼ .05 for the anxiety sub-scale). On the other hand, the predictednegative correlation between the MARTIscale and the Trust Scale was statisticallysignificant, r¼�.19, p< .01. Specifically,partners who had stronger motives toacquire relationship-threatening informa-tion reported lower levels of trust in theircurrent relationships.

The dating partners’ scores on the TrustScale were interdependent, as evidenced byan intraclass (interpartner) correlation of.58, F(95,96)¼ 1.83, p< .05. For this rea-son, we recomputed the correlation betweenthe MARTI and the Trust Scale at both theindividual and dyad levels of analysis, usingprocedures recommended by Gonzalez andGriffin (2000). The interdependence-adjusted individual-level correlationbetween MARTI and trust was about thesame as before (r¼�.20, p< .05); thelatent-variables estimate of the dyadic cor-relation was slightly larger (r¼�.24). Theadjusted individual-level correlation indi-cates that as the motive to acquire relation-ship-threatening information increases,individuals are less trusting (or vice versa).Similarly, the dyad-level correlation indi-cates that in dating couples where bothmembers have higher MARTI scores, bothmembers also tend to trust each other less.

Because mistrust should be conceptuallyrelated to the motive to acquire relation-ship-threatening information, the Study 2data offer important support for the con-vergent validity of the MARTI scale. On theother hand, although the negative correla-tion between MARTI and trust was signifi-cant, as we predicted, it was also lowenough (�.20 to �.25) to justify making aconceptual distinction between MARTIand relational trust.

138 W. Ickes et al.

Study 3

Study 3 was designed to provide furtherconstruct validity evidence for the MARTIscale. With respect to convergent validity,we predicted that if individuals are stronglymotivated to acquire relationship-threaten-ing information, they should be moreinclined to engage in ‘‘suspicion behaviors’’driven by concerns about the possiblelack of commitment or the infidelity of pastor current romantic partners. Accordingly,individuals who score higher on theMARTI scale should be more likely toreport having eavesdropped on their past orcurrent partners’ phone conversations,checked up on their partners’ whereabouts,secretly followed or spied on their partners,monitored their partners’ behavior for signsof infidelity, or set up test situations todetermine whether their partners werelying (see Ickes & Simpson, 1997, p. 238).

With respect to discriminant validity, wepredicted that scores on the MARTI scaleshould not correlate significantly withscores on a measure of socially desirableresponding (i.e., impression-management;Paulhus, 1984) or on a measure of hostile/paranoid (i.e., more clinical) forms of suspi-cion (Buss & Durkee, 1957). Because theMARTI was developed to assess individualdifferences in the desire to acquire relation-ship-threatening information in romanticrelationships (rather than the desire tomake good social impressions or the ten-dency to harbor paranoid concerns aboutpeople in general), it should not share sub-stantial variance with measures of people’stendencies to present themselves in anoverly positive light or to harbor hostileand paranoid worries about the malevo-lence of all people.

As a final goal of Study 3, we wanted toconfirm that the MARTI scale (which has adichotomous response format) correlatesstrongly with a version of the scale inwhich all 21 items are answered using acontinuous, Likert-type response format.Accordingly, both versions of the scalewere administered to the Study 3participants.

Method

Participants

Seventy-four men and 114 women, allenrolled in an introductory psychology classat Texas A & M University, participated forcourse credit. To be eligible for the study,participants had to have dated someone atsome point in the past. At the time of thestudy, 44.1% percent of the participantswere not currently dating anyone, 49.5%were dating one partner exclusively, 4.3%were currently dating more than one partner,and 2.1% were engaged. The participants’average age was 18.90 years, with a standarddeviation of .87 years.Based on self-reports,79.8% of the participants classified them-selves as White, 14.4% as Latino/Hispanic,and the remaining 5.8% as AfricanAmerican, Asian, or Other.

Procedures

Participants completed a short set of self-report questionnaires in small groups. Inaddition to completing a general demo-graphic questionnaire and the MARTIscale, they also responded to each of thefour following measures.

1. Suspicion Behavior Checklist (SBC).This 11-item scale, developed for thepresent study, asks individuals howoften they have engaged in variousacts/behaviors designed to check,monitor, or test for a romantic part-ner’s potential disloyalty or infidel-ity (either their current partner orprevious ones). On a 4-point scale(where 1¼ never, 2¼ 1 or 2 times,3¼ 3 or 4 times, and 4¼ 5 or moretimes), the participants indicate howoften they have eavesdropped on apartner’s private phone conversa-tions, called to see if a partner waswhere he or she was supposed to be,secretly followed or spied on a part-ner, set up test situations to see if apartner would lie about the relation-ship, had someone check to see who

Suspicious minds: The MARTI scale 139

a partner was with, confronted apartner about suspicions of beingromantically involved with otherpeople, closely monitored a part-ner’s daily behavior for signs of infi-delity, unexpectedly visited a partnerto see whether he or she wasdoing something wrong, or asked apartner details about his or her pastlovers.

This checklist proved to be reliable(alpha¼ .85), but because the beha-viors it assessed were relatively rareand socially undesirable, the distribu-tion of scores was markedly (and posi-tively) skewed. To help normalize thedistribution, total scores on the SBCwere transformed using a log10 trans-formation to reduce the skew. Higherscores indicated engaging in suspicionbehaviors more frequently.

2. Impression Management Scale.Paulhus’s (1984) 10-item ImpressionManagement Scale is answered on a7-point scale (anchored by 1¼No, notat all and 7¼Yes, a great deal).Sample items include: Once in a while,I laugh at a dirty joke (reverse-keyed);I am always courteous, even to peoplewho are disagreeable; and Sometimesat elections I vote for candidates I knowlittle about (reverse-keyed). The scalehad a lower-than-expected reliability(alpha¼ .53), which did not improvewhen potentially problematic itemswere considered for exclusion. Highscores on this measure reflect a ten-dency to present oneself in an overlypositive light.

3. Alternative MARTI Scale (continu-ous response format). The parti-cipants then responded to the 21items of the MARTI scale again,but this time using a Likert-typerating scale (anchored by 1¼ I wouldprefer to NOT know this informa-tion and 7¼ I would prefer TOKNOW this information). This con-tinuous-scale version of the MARTIwas highly reliable (alpha¼ .92), withhigher scores reflecting a stronger

desire to acquire relationship-threatening information.

4. Hostile/Paranoid Suspicion Scale.This 10-item measure (Buss &Durkee, 1957) assesses hostile/para-noid forms of distrust and suspicionabout people in general (includingstrangers). It is answered in a true/false format. Sample items are:I sometimes have the feeling that othersare laughing at me; There are a numberof people who seem to be jealous of me;and I know that people tend to talkabout me behind my back. The scalewas reasonably reliable (alpha¼ .63)given its brevity and dichotomousresponse format. Higher scores indi-cated greater paranoid suspicion anddistrust.

Results

As expected, the dichotomous and continu-ous versions of the MARTI scale proved tobe highly correlated, both in the full sample,r¼ .77, p< .001, and within each sex, r¼ .79for men and r¼ .74 for women. When theunreliability of each scale was corrected forattenuation (Cohen & Cohen, 1983), thedisattenuated correlation was .91. Thisresult indicates that the two versions of theMARTI scale (dichotomous and continuousresponse formats) yield total scale scores forwhich respondents are rank-ordered almostidentically, suggesting that the two versionscan be used interchangeably in research.

To further examine the factor structureof the MARTI scale, we conducted twoprincipal components analyses: one onthe 21 MARTI items answered in the origi-nal (dichotomous) response format, andone on the 21 MARTI items answered in aLikert-type response format. The scree testfor the analysis involving the 21 itemsanswered in the dichotomous format indi-cated one component (factor) thatexplained 23.15% of the variance. Withthe exception of one item (item 13), allitems loaded .30 or greater on this factor.The scree test for the analysis involving

140 W. Ickes et al.

MARTI items answered on the Likert-typescales also revealed one component (factor)that accounted for 39.50% of the variance.All 21 items loaded .35 or greater on thisfactor. Thus, replicating results reportedin Study 1, both forms of the MARTI scaleappear to measure a single underlying factor.

Despite the low frequency (and the highlyskewed distribution) of the suspicion beha-viors reported in this sample, the partici-pants’ scores on the MARTI scalecorrelated significantly with their scores onthe Suspicion Behaviors Checklist, r¼ .17,p< .02 in the full sample (r¼ .14 for womenand r¼ .19 for men). When the unreliabilityof each scale was corrected for attenuation,the disattenuated correlation was .20 in thefull sample (.17 for women and .23 for men).That is, individuals who want to acquiremore relationship-threatening informationreport having engaged more frequently in‘‘suspicion behaviors’’ involving past orcurrent romantic partners.

As predicted, two nonsignificant discrim-inant correlations were found. Specifically,the MARTI did not correlate with eitherPaulhus’s (1984) Impression ManagementScale, r¼�.06, ns, in the full sample(r¼�.02 for men and r¼�.14 for women,both ns) or Buss and Durkee’s (1957)Hostile/Paranoid Suspicion Scale, r¼�.05,ns (r¼�.05 for men and r¼�.08, both ns).

In sum, this pattern of convergent anddiscriminant correlations provides furthersupport for the construct validity of theMARTI scale. The significant convergentcorrelation involving the MARTI scaleand the Suspicion Behaviors Checklist wasonly modest, however, analogous to theweak-but-significant correlations typicallyobserved in other domains (i.e., researchon criminal behavior or on coronary heartdisease) in which the criterion outcomeoccurs infrequently and has a positivelyskewed response distribution.

Study 4

The goal of Study 4 was to marshal addi-tional evidence for the construct validity ofthe MARTI scale by replicating the scale’s

discriminant validity properties and testingthe scale’s predictive validity. Ickes andSimpson (1997, 2001) proposed that indivi-duals who find it difficult to avoid relation-ship-threatening information should bevulnerable to experiencing relationship dis-solution over relatively short periods oftime, especially if they feel less close to(i.e., less personally connected with) theirpartners. People who feel less subjectivelyclose to their partners should find relation-ship-threatening thoughts and feelings moredamaging because their levels of relation-ship commitment and trust are not highenough to allow them to ignore, downplay,or discount threatening information.

This line of reasoning suggests that datingpartners’ MARTI scores should predict theprobability of relationship dissolution overtime. Specifically, couples who score higheron the MARTI scale (i.e., couples with moresuspicions) should be significantly more likelyto break up than couples who score lower(i.e., couples with fewer or less intense suspi-cions), and this instability should be mostapparent in couples who also report lowerlevels of perceived (subjective) closeness.

Method

Participants

The participants were 85 heterosexual dat-ing couples who volunteered for a study onthe ‘‘characteristics of partners in datingrelationships.’’ With the aid of posters andfliers, couples were recruited from theUniversity of Texas at Arlington campusand the surrounding community. Each par-tner was paid $10 for his or her participation.Data from four couples were excluded fromthe analyses (in two cases because of tech-nical difficulties, and in two others becauseof a failure to follow instructions). Datafrom the remaining 81 couples were used inthe analyses reported below.

Sociodemographic characteristics

Similar to Study 1, the racial and ethic com-position of the Study 4 sample reflected the

Suspicious minds: The MARTI scale 141

general composition of the student body atthe University of Texas at Arlington. Therewere 15 Hispanic participants (9.3%), 24African Americans (14.8%), 23 Asians(14.2%), 94 Whites (58.0%), and 6 Others(3.7%). Participants ranged in age from18–43 years, with a mean age of 22.1 years.The average length of their relationshipswas 22.5 months, with a standard deviationof 1.3 months. With regard to dating status,only 4 participants were dating their currentpartner and others (2.5%), 126 people weredating only their current partner (77.7%),and 32 were engaged (19.8%).

Procedure

When each couple arrived for the study, theexperimenter escorted the partners to separ-ate cubicles where they completed a packetof self-report measures in private. Thepacket contained each of the scales used inStudy 1 (except the Big Five scales) plus anadditional one (the Inclusion of Other in theSelf scale). When both partners had fin-ished, they were reunited. They then partici-pated in a laboratory task not relevant tothe present study. When this task was over,each couple was given $20 ($10 per person)for their participation, after which theywere thanked and debriefed.

Measures

The Study 4 self-report scales included thebrief demographic questionnaire, the AdultAttachment Questionnaire (Simpson et al.,1996), the Social Orientation Scale (Ickes &Hutchison, 2000), and the MARTI scale.Participants also completed the Inclusionof Other in the Self scale (IOS; Aron,Aron, & Smollan, 1992), which is a well-validated measure of perceived relationshipcloseness. The IOS consists of 7 Venn dia-grams that depict self and partner as circleswith different degrees of overlap (perceivedcloseness). The participants choose the dia-gram (1 through 7) for which the degree ofoverlap best represents their perceptionof the degree of closeness in their currentrelationship. Higher scores reflect greatersubjective closeness.

Five-month follow-up of dating status

Five months after the study, each of thedating partners was contacted by telephone.Each partner was asked the following ques-tion: ‘‘When you participated in the studyon [date of participation], you were dating aperson named [name of dating partner]. Areyou still dating this person?’’ (answeredeither ‘‘Yes’’ or ‘‘No’’). Both partners hadto agree about their current dating statusfor their relationship status data to be usedin the analyses. Of the 81 couples who par-ticipated in Study 4, concordant relation-ship status information was obtained from75 couples (92%).

Results

Discriminant validity: Correlations betweenthe MARTI scale and other measures

As in Study 1, scores on the MARTI scalewere not significantly correlated with scoreson any of the other self-report scalesassessed in Study 4. In particular, MARTIscores were not correlated with scores oneither the avoidance (r¼ .10) or the anxious(r¼�.08) subscales of the AAQ, and theyalso were unrelated to both the socialabsorption (r¼�.06) and social individu-ation (r¼ .03) subscales of the Social Orient-ation Scale. Similar to the findings of Study1, these null results support the discriminantvalidity of the motive to acquire relation-ship-threatening information.

Evidence for the predictive validity of theMARTI scale

As a behavioral test of predictive validity,we calculated the dyad-level scores for theMARTI scale and the IOS measure (foreach scale, we averaged the scores of themale partner and the female partner withineach dyad) and used these two scores topredict couples’ relationship status (brokenup vs. still together) at the 5-month follow-up. Because information would have beenlost by reducing the predictor variables toclass variables (needed to run in anANOVA model), we used a moderated

142 W. Ickes et al.

multiple regression approach to test theunique effects of each predictor variable—dyad-level MARTI and perceived closenessscores—and the interaction of these twovariables (for information about thistechnique, see Arnold & Evans, 1979;Cohen, 1978; McClelland & Judd, 1993;Nunnally & Bernstein, 1994; Saunders, 1956).

As predicted, couples who scored higheron the MARTI scale were more likely tohave broken up 5 months later than werecouples who scored lower. When, for illus-trative purposes, median splits were used todivide the couples into subgroups, the per-centages of couples who were still dating at5-month follow-up were 54% of those withhigh MARTI scores and 78% of those withlow MARTI scores, t(74)¼�1.93, p¼ .06.A similar trend was found for couples whowere classified by median split as higherversus lower in perceived closeness: Ahigher percentage of more-close couples(77%) than less-close couples (65%) werestill dating 5 months later, t(74)¼�1.74,p< .08.

These main effects, however, were quali-fied in the moderated multiple regressionanalysis by a significant MARTI�Perceived Closeness interaction, F(1,74)¼ 4.47, p< .05. This predicted interaction(shown in Figure 1) revealed that, amongcouples who had a stronger motive toacquire relationship-threatening informa-tion, those who were less close were lesslikely to stay together (47%) than those

who were closer (67%). In contrast, amongcouples with a weaker motive to acquirerelationship-threatening information, therewas no such difference. In fact, less-closecouples were just as likely to stay togetheras more-close couples in this case (78% and76%, respectively).

This interaction provides evidence thatthe avoidance of relationship-threateninginformation can actually buffer dating part-ners from risk factors that might otherwisedestabilize their relationships. Lower levelsof perceived closeness appear to be a riskfactor for relationship dissolution, but onlyin combination with a strong motive toacquire relationship-threatening informa-tion. When this motive is weak (i.e., whencouples avoid such information), the rela-tionships of less-close couples are no moreunstable than those of more-close couples,at least within the time frame examined inStudy 4.

General Discussion

The findings of the present studies supportthe reliability, distinctiveness, and constructvalidity of the MARTI scale—a measure ofthe tendency to approach or avoid relation-ship-threatening information. Results ofStudy 1 reveal that the MARTI scale hasgood internal consistency and that it meas-ures a single underlying factor. The resultsof Study 1 also provide discriminant valid-ity evidence for the MARTI scale, demon-strating that it is empirically unrelated tomeasures of conceptually distinct constructssuch as the Big Five personality traits, adultattachment styles, and general socialorientations.

Data from Studies 2 and 3 provide con-vergent and discriminant validity evidencefor the MARTI scale. Specifically, results ofthese studies reveal that dating partnerswith higher MARTI scores tend to trusttheir partners less (Study 2) and reportengaging in more ‘‘suspicion behaviors’’designed to ascertain their partners’ degreeof fidelity or commitment (Study 3). Inaddition, scores on the MARTI scale donot correlate with the tendency to convey

78%

47%

76%67%

0%

20%

40%

60%

80%

100%

Low MARTI High MARTI

Low CloseHigh Close

Figure 1. Percentage of couples stilltogether at the time of the 5-monthfollow-up in Study 4.

Suspicious minds: The MARTI scale 143

socially desirable impressions or withhostile/paranoid forms of suspiciousness.

The data from Study 4 replicate the dis-criminant validity findings from Study 1and provide important predictive validityevidence for the MARTI scale. As expected,dating couples who are more motivated toacquire relationship-threatening informa-tion are more likely to break up across a5-month period, particularly if their rela-tionship is low in subjective closeness. Con-versely, couples who are motivated to avoidrelationship-threatening information areless likely to break up, even if their relation-ship is less close. These results suggest that astrong motive to avoid relationship-threat-ening information might buffer relation-ships from destabilizing influences (such aslow closeness) that could undermine them ifleft unchecked.

Considered together, the results of thesestudies indicate that the MARTI scaleassesses a novel and empirically distinctpsychological dimension that may play animportant role in the long-term functioningand well-being of romantic relationships. Inwhat follows, we suggest ways in which themotive to acquire relationship-threateninginformation might have both deleteriousand (in certain cases) beneficial effects onrelationships. We then propose some viabledirections for future research.

When is seeking relationship-threateninginformation bad versus good forrelationships?

As a general rule, relationships should suf-fer if one or both partners possess a strongmotive to acquire relationship-threateninginformation, particularly if one or bothpartners ruminate about or embellishworst-case scenarios when confronted withsuch information. Strong motives toacquire threatening information should beparticularly damaging to relationshipswhen the knowledge of potential threatsprolongs, revives, or intensifies existingrelationship problems. This should be trueeven in situations where problems seemminor (i.e., are not central to the relation-

ship) or are transient (see Ickes & Simpson,2001; Simpson et al., 1995). The damageshould be greatest, however, in situationswhere problems are major, enduring, diffi-cult, or impossible to resolve.

On the other hand, there may beinstances in which it might be adaptive forpartners to possess at least moderate levelsof the motive to acquire relationship-threatening information. Moderate levels ofthis motive could potentially benefit relation-ships when the resulting awareness of legit-imate relationship threats enables partners tomake necessary corrections or changes intheir relationship. This process should bemost evident in situations where the prob-lems associated with the threat are criticalto the long-term well-being of the relation-ship, are likely to persist over time, couldrecur frequently, or could be resolved ifdealt with directly. Conversely, moderatelevels of the motivation to acquire relation-ship-threatening information may be harmfulif couples continually encounter situationswhere their differences seem irreconcilable ortheir relationship problems are viewed asirremediable.

These hypothesized effects are likely tobe moderated by any or all of several fac-tors, including (a) the type of relationalthreat a person finds most troubling,(b) the person’s own dispositional characteris-tics (e.g., level of anxious attachment or neu-roticism), and (c) the person’s partner’sdispositional characteristics. For example,in certain watershed situations (e.g., whenrelational stability is highly threatened),high scores on both the MARTI andanxious attachment may bode poorly forthe long-term well-being of a relationship.Highly anxious persons tend to be pre-occupied with issues of loss and abandon-ment, and they typically infer the worst inthese relationship-threatening contexts(Cassidy & Berlin, 1994). In contrast,high scores on the MARTI scale in com-bination with greater attachment securityshould be less damaging because moresecure people do not worry about aban-donment and, for that reason, should beless likely to prematurely catastrophize

144 W. Ickes et al.

problems when relationship threats becomesalient.

The characteristics of one’s partner alsomight affect long-term relationship out-comes. Individuals who are involved withpartners scoring high on both the MARTIand anxious attachment, for instance, mighthave to persistently calm their partners andremind them of their steadfast love anddevotion in order to sustain and stabilizetheir relationships. Needless to say, thisrole could become increasing difficult andburdensome over time.

In sum, the meaning of high scores onthe MARTI scale must be interpreted in thecontext of additional qualifying informa-tion about the relationship partners. Astrong desire to acquire relationship-threatening information should havedifferent long-term effects on relationships,depending on how each partner interpretsand processes potentially threateninginformation. If partners use information-processing strategies in which negativeaffect distorts their thinking (Forgas, 1995)or if they rely on emotion-focused copingstrategies when distressed (Lazarus &Folkman, 1984), the motive to acquirerelationship-threatening information mayhave deleterious effects. However, ifpartners do not let negative affect infusetheir information processing or if they useproblem-focused coping strategies whenupset, this motive should be less damagingto their relationships.

Directions for future research

The new MARTI construct and its asso-ciated scale suggest several new avenuesfor research on relationship dynamics.First, future research should determinewhether other relevant individual differencemeasures (e.g., attachment styles or neur-oticism) or dyad-based measures (e.g., rela-tionship trust) interact with the MARTIscale to predict long-term relationship out-comes such as changes in satisfaction orrelationship instability. It would be interest-ing to determine, for example, whetherspecific configurations of personal attributes

(e.g., high scores on the MARTI, highscores on anxious attachment, and lowscores on trust) forecast the most negativerelationship outcomes across time. Predic-tions of this type could be tested in marriedcouples as well as in dating couples. Bysimply changing the scale instructions andcertain of the items, the MARTI can easilybe adapted for use with married samples.

Second, future research should examinewhether different levels of the motiveassessed by the MARTI scale predict long-term marital well-being in the specific situ-ations in which high versus low levels of thismotive theoretically should have beneficialversus damaging effects. It is possible, forexample, that a strong motive to avoidrelationship-threatening information onthe part of one or both spouses may protectmarriages from short-term, minor, or tem-porary downturns, but make them moresusceptible to long-term declines in stabilityand satisfaction when more serious orintractable problems arise. In other words,future studies should identify the specificcircumstances in which the motive to avoidrelationship-threatening information stopsbeing adaptive and becomes maladaptive.

Third, we need to understand how rela-tionship dynamics might differ when part-ners have similar versus different motives toapproach or avoid relationship-threateninginformation. For example, if both partnersare predisposed to avoid relationship-threatening information, avoidance maybecome normative in their relationship andsuch couples may become ‘‘conflict-avoi-dant’’ (Gottman, 1994). If both partnersare predisposed to approach relationship-threatening information, they may developconfrontative relationships similar to thosethat Gottman has characterized as ‘‘vola-tile.’’ Finally, if one partner is predisposedto avoid relationship-threatening informa-tion while the other is predisposed toacquire it, such couples may develop anddisplay a ‘‘demand/withdrawal’’ pattern(Christensen & Heavey, 1990). Ourinformational-motives perspective suggeststhat which partner ‘‘demands’’ and whichone ‘‘withdraws’’ might have more to do with

Suspicious minds: The MARTI scale 145

each partner’s MARTI score than with hisor her gender. If this proves to be true, itwould qualify the conventional wisdom thatwomen usually demand and men usuallywithdraw in romantic relationships.

Fourth, future research should investi-gate the different possible ‘‘origins’’ of themotivation to acquire versus avoid relation-ship-threatening information. Some individ-uals may be motivated to gatherthreatening information because theychronically worry about their partners’activities and whereabouts, even if theirpartners have always been honest and faith-ful. Other individuals may seek suchinformation because painful personalexperiences have taught them that they can-not trust their partners and must remainvigilant to signs of trouble. Scores on theMARTI, therefore, may contain some vari-ance associated with the general dispositionto acquire relationship-threatening infor-mation and some that reflects the dynamicsof the current relationship. Understandingwhy certain individuals are strongly motiv-ated to acquire versus avoid relationship-threatening information is likely to provideimportant clues about the dynamics of suspi-cion and denial in close relationships.

Fifth, additional research should clarifywhere the MARTI scale fits within thelarger nomological network of potentiallyrelated constructs and measures. The cur-rent studies demonstrate that the MARTI

scale displays its convergent validity in rela-tion to conceptually similar constructs suchas dispositional trust and the tendency toengage in suspicion behaviors, but shows itsdiscriminant validity in relation to concep-tually distinct or irrelevant constructs suchas personality traits, response biases, andgeneral social orientations. Because scalevalidation is a continuing process andbecause no initial set of studies can beexpected to rule out the influence of allpossible covariates (uncertainty orientation,need for closure, jealousy, etc.), futureresearch should continue to explore the con-struct validity of the MARTI scale.

Finally, future research should explore ingreater detail the behavioral tactics thatindividuals use to acquire or avoid poten-tially threatening information. Do peopletypically seek such information directly byconfronting their partners? Or do they seekinformation indirectly by monitoring theirpartner’s personal correspondence, check-ing their partner’s claims against verifiablerecords, or calling to confirm their partner’swhereabouts and current activities? Arethere certain types of information or situ-ations in which direct versus indirect tacticsare more likely to be employed? A betterunderstanding of what tactics people rou-tinely use, when they use them, and why theyuse them may provide deeper insights intohow, when, and why these informationalmotives operate.

References

Arnold, H. J., & Evans, M. G. (1979). Testing multiplica-tive models does not require ratio scales. Organiza-tional Behavior and Human Performance, 24, 41–59.

Aron, A., Aron, E. N., & Smollan, D. (1992). Inclu-sion of Other in the Self Scale and the structure ofinterpersonal closeness. Journal of Personality andSocial Psychology, 63, 596–612.

Bernstein, I. H., & Teng, G. (1989). Factoring itemsand factoring scales are different: Spurious evi-dence for multidimensionality due to item categor-ization. Psychological Bulletin, 105, 467–477.

Bowlby, J. (1973).Attachment and loss: Volume 2. Separa-tion: Anxiety and anger. New York: Basic Books.

Buss, A. H., & Durkee, A. (1957). An inventory forassessing different kinds of hostility. Journal ofConsulting Psychology, 21, 343–349.

Carroll, J. B. (1945). The effect of difficulty andchance success on correlations between items andbetween tests. Psychometrika, 10, 1–19.

Cassidy, J., & Berlin, L. J. (1994). The insecure/ambivalent pattern of attachment: Theory andresearch. Child Development, 65, 971–991.

Christensen, A., & Heavey, C. L. (1990). Gender andsocial structure in the demand/withdrawal patternof marital interaction. Journal of Personality andSocial Psychology, 59, 73–81.

Cohen, J. (1978). Partialed products are interactions;partialed powers are curve components. Psycho-logical Bulletin, 85, 858–866.

Cohen, J. & Cohen, P. (1983). Applied multipleregression/correlation analysis for the behavioralsciences. Hillsdale, NJ: Erlbaum.

Collins, N. L., & Read, S. J. (1994). Cognitiverepresentations of attachment: The structureand function of working models. In D.Perlman & K. Bartholomew (Eds.), Attachmentprocesses in adulthood (pp. 53–90). London:Kingsley.

146 W. Ickes et al.

Ferguson, G. A. (1941). The factorial interpretation oftest difficulty. Psychometrika, 6, 323–329.

Forgas, J. (1995). Mood and judgment: The affectinfusion model (AIM). Psychological Bulletin,116, 39–66.

Gonzalez, R., & Griffin, D. (2000). On the statistics ofinterdependence: Treating dyadic data withrespect. In W. Ickes & S. Duck (Eds.), The socialpsychology of personal relationships (pp. 181–213).Chichester, UK: Wiley.

Gottman, J. M. (1994). What predicts divorce? Therelationship between marital processes and maritaloutcomes. Hillsdale, NJ: Erlbaum.

Ickes, W. (1997). A brief measure of the Big Five per-sonality dimensions. Unpublished scale developedat the University of Texas at Arlington.

Ickes, W., & Hutchison, J. (2000, February). Socialabsorption and social individuation: Independentdimensions of social relating. Invited talk given atthe preconference on relationships, first annualmeeting of the Society for Personality and SocialPsychology, Nashville, Tennessee.

Ickes, W., Hutchison, J., & Mashek, D. (in press).Closeness as intersubjectivity: Social absorptionand social individuation. To appear in D. Mashekand A. Aron (Eds.), The handbook of closeness andintimacy. Mahwah, NJ: Erlbaum.

Ickes, W., & Simpson, J. A. (1997). Managingempathic accuracy in close relationships. InW. Ickes (Ed.), Empathic accuracy (pp. 218–250).New York: Guilford Press.

Ickes, W., & Simpson, J. A. (2001). Motivationalaspects of empathic accuracy. In G. J. O. Fletcher& M. S. Clark (Eds.), Blackwell handbook of socialpsychology: Interpersonal processes (pp. 229–249).Oxford, UK: Blackwell.

Ickes, W., Stinson, L., Bissonnette, V., & Garcia, S.(1990). Naturalistic social cognition: Empathicaccuracy in mixed-sex dyads. Journal of Personalityand Social Psychology, 59, 730–742.

John, O. P. (1990). The ‘‘Big Five’’ factor taxonomy:Dimensions of personality in the natural language

and in questionnaires. In L. A. Pervin (Ed.),Handbook of personality: Theory and research(pp. 66–100). New York: Guilford Press.

Kilpatrick, S. D., Rusbult, C.E., & Bissonnette, V. L.(2002). Empathic accuracy among newly marriedcouples. Personal Relationships, 9, 369–393.

Lazarus, R. S., & Folkman, S. (1984). Stress, apprai-sal, and coping. New York: Springer.

McClelland, G. H., & Judd, C. M. (1993). Statisticaldifficulties of detecting interactions and moderatoreffects. Psychological Bulletin, 114, 376–390.

Nunnally, J. C., & Bernstein, I. H. (1994). Psycho-metric theory. New York: McGraw-Hill.

Paulhus, D. L. (1984). Two-component models ofsocially desirable responding. Journal of Personal-ity and Social Psychology, 46, 598–609.

Rempel, J. K., Holmes, J. G., & Zanna, M. P. (1985).Trust in close relationships. Journal of Personalityand Social Psychology, 49, 95–112.

Saunders, D. R. (1956). Moderator variables in pre-diction. Educational and Psychological Measure-ment, 16, 209–222.

Sillars, A. L., & Scott, M. D. (1983). Interpersonalperception between intimates: An integrative review.Human Communication Research, 10, 153–176.

Simpson, J. A., Ickes, W., & Blackstone, T. (1995).When the head protects the heart: Empathic accu-racy in dating relationships. Journal of Personalityand Social Psychology, 69, 629–641.

Simpson, J. A., Ickes, W., & Grich, J. (1999). Whenaccuracy hurts: Reactions of anxiously-attacheddating partners to a relationship-threateningsituation. Journal of Personality and SocialPsychology, 76, 754–769.

Simpson, J. A., Rholes, W. S., & Phillips, D. (1996).Conflict in close relationships: An attachmentperspective. Journal of Personality and SocialPsychology, 71, 899–914.

Thomas, G., Fletcher, G. J. O., & Lange, C. (1997).On-line empathic accuracy in marital interaction.Journal of Personality and Social Psychology, 72,839–850.

Appendix: The Motivation to Acquire Relationship-Threatening Information

(MARTI) Scale

In dating relationships, there are things that partners might or might not want to knowabout each other. For each of the following hypothetical items of information, indicatewhether you would prefer to know or not know the information by circling the appropriatealternative in each case. There are no right or wrong answers, but it is important that you beas honest and accurate as possible in responding to each item.

All of the items of information listed below are hypothetical. We have not obtainedany of this information from your dating partner and will not do so. We will keep all ofthe responses to this survey strictly confidential and use the resulting data for statisticalpurposes only. Your dating partner will never see any of your responses at any time.

If you could, which of the following hypothetical items of information about yourdating partner would you prefer to know or not know? (Circle one of the two responsesin each case.)

(continued on next page)

Suspicious minds: The MARTI scale 147

Appendix: (continued)

I would prefer to:

1. The most critical thing my partner has said about meto one of his/her same-sex friends. know not know

2. The most unkind or unfair thing my partner has saidabout me to one of his/her same-sex friends. know not know

3. The most critical thing my partner has said about meto one of my same-sex friends. know not know

4. The most unkind or unfair thing my partner has saidabout me to one of my same-sex friends. know not know

5. The most critical thing my partner has said about meto one of his/her current or former dating partners. know not know

6. The most unkind or unfair thing my partner has said aboutme to one of his/her current or former dating partners. know not know

7. The most intimate detail of my past history that mypartner has revealed to one of his/her same-sex friends. know not know

8. The most intimate detail of my past history that mypartner has revealed to one of my same-sex friends. know not know

9. The most intimate detail of my past history that mypartner has revealed to one of his/her current orformer dating partners. know not know

10. The most intimate detail of our relationship that mypartner has revealed to one of his/her same-sex friends. know not know

11. The most intimate detail of our relationship that mypartner has revealed to one of my same-sex friends. know not know

12. The most intimate detail of our relationship that mypartner has revealed to one of his/her currentor former dating partners. know not know

13. Which of my same-sex friends my partner issecretly the most strongly attracted to. know not know

14. Whether my partner has ever flirted with oneof my same-sex friends without telling me. know not know

15. Whether my partner has, without telling me,ever flirted with any other members of theopposite sex while we have been dating. know not know

16. Whether my partner has ever dated one of mysame-sex friends without telling me. know not know

17. Whether my partner has, without telling me,ever dated any other members of the opposite sexwhile we have been dating. know not know

18. Whether my partner has ever made out with oneof my same-sex friends without telling me. know not know

19. Whether my partner has, without telling me,ever made out with any other members of theopposite sex while we have been dating. know not know

20. Whether my partner has ever had protected sexwith one of my same-sex friends without telling me know not know

21. Whether my partner has, without telling me,ever had protected sex with any other membersof the opposite sex while we have been dating. know not know

148 W. Ickes et al.