Journal of Counseling Psychology1998, Vol. 45, No. 3, 235-246

Copyright 1998 by the American Psychological Association, Inc.0022-0167/98/S3.00

The Differentiation of Self Inventory:Development and Initial ValidationElizabeth A. Skowron and Myrna L. Friedlander

University at Albany, State University of New York

Despite the importance of Bowen theory (M.Bowen, 1976,1978; M. E. Kerr&Bowen, 1988)in the field of family therapy, there have been relatively few studies to date examining itsconstructs or propositions. To fill this gap, a self-report instrument, the Differentiation of SelfInventory (DSI) has been developed. The DSI is a multidimensional measure of differentiationthat focuses specifically on adults (age 25+), their significant relationships, and currentrelations with family of origin. Six-hundred and nine adults participated in a series of 3studies, in which DSI scores—reflecting less emotional reactivity, cutoff, and fusion withothers, and a greater ability to take an "I position"—predicted lower chronic anxiety, betterpsychological adjustment, and greater marital satisfaction. Other results consistent withBowen theory are discussed, along with the potential contribution of the DSI for testingBowen theory, as a clinical assessment tool, and as an indicator of psychotherapeutic outcome.

Bowen theory (Bowen, 1976, 1978) is regarded as one ofthe few comprehensive explanations of psychological devel-opment from a systemic and multigenerational perspective(Gurman, 1991; Nichols & Schwartz, 1998). Indeed, Bowentheory provides a foundation for the field of family therapythat renders it distinct from the multitude of theoreticalapproaches to individual psychotherapy. At present, many ofBowen's (1976, 1978) concepts (e.g., differentiation of self,interlocking triangles, or reactive emotional distancing)pervade the family systems literature. Despite the vastattention Bowen theory has received from clinicians andtheorists alike, there have been, to date, few programmaticattempts to test its validity with respect to personalityfunctioning or quality of interpersonal relations or to changesas a result of psychotherapy. To begin filling this notablegap, we developed the Differentiation of Self Inventory, aself-report instrument for adults (ages 25+). In this article,we present psychometric support for the measure, validationstudies to date, and implications for theory, research, andpractice.

Of the various constructs that compose Bowen theory,differentiation of self is the personality variable most criticalto mature development and the attainment of psychological

Elizabeth A. Skowron and Myrna L. Friedlander, Department ofCounseling Psychology, University at Albany, State University ofNew York.

Portions of this research, based on a doctoral dissertation byElizabeth A. Skowron under the direction of Myrna L, Friedlander,were presented at the 100th and 103rd Annual Conventions of theAmerican Psychological Association. We gratefully acknowledgethe valuable comments and suggestions of Richard F. Haase,Michael P. Nichols, Robert Noone, Collie Connelly, BarbaraWhite, and Douglas Rait.

Correspondence concerning this article should be addressed toElizabeth A. Skowron, who is now at the Department of Educa-tional Psychology, University of Wisconsin, Milwaukee, Wiscon-sin 53201-0413. Electronic mail may be sent to eskowron®soe.uwm.edu.

health. Differentiation of self is defined as the degree towhich one is able to balance (a) emotional and intellectualfunctioning and (b) intimacy and autonomy in relationships(Bowen, 1978). On an intrapsychic level, differentiationrefers to the ability to distinguish thoughts from feelings andto choose between being guided by one's intellect or one'semotions (Bowen, 1976, 1978). Greater differentiation al-lows one to experience strong affect or shift to calm, logicalreasoning when circumstances dictate. Flexible, adaptable,and better able to cope with stress, more differentiatedindividuals operate equally well on both emotional andrational levels while maintaining a measure of autonomywithin their intimate relationships.

In contrast, poorly differentiated persons tend to be moreemotionally reactive (Kerr & Bowen, 1988, p. 320), findingit difficult to remain calm in response to the emotionality ofothers. With intellect and emotions fused, they tend to makedecisions on the basis of what "feels right"; in short, theyare trapped in an emotional world (Bowen, 1976; Kerr,1985).

On an interpersonal level, differentiation of self refers tothe ability to experience intimacy with and independencefrom others. More differentiated persons are capable oftaking an / Position in relationships: maintaining a clearlydefined sense of self and thoughtfully adhering to personalconvictions when pressured by others to do otherwise(Bowen, 1978, p. 252). Differentiation allows for flexibleboundaries that permit emotional intimacy and physicalunion with another without a fear of merger (Bowen, 1978;Kerr, 1988).

When overwhelmed by emotionality in their familyrelationships, poorly differentiated individuals tend to en-gage infusion or emotional cutoff (Ken & Bowen, 1988).According to Bowen theory, highly fused individuals remainemotionally "stuck" in the position they occupied in theirfamilies of origin, have few firmly held convictions andbeliefs, are either dogmatic or compliant, and seek accep-tance and approval above all other goals (Bowen, 1976,

235

236 SKOWRON AND FRIEDLANDER

1978). Emotional cutoff is personified by the reactiveemotional distancer, who appears aloof and isolated fromothers, tends to deny the importance of family, often boastsof his or her emancipation from parents, and displays anexaggerated facade of independence (Nichols & Schwartz,1998). Whereas the fused person tends to experience separa-tion as overwhelming, the emotionally cutoff person findsintimacy profoundly threatening. Yet both individuals arepoorly differentiated, basing self-esteem largely on theapproval of others and generally conforming to those aroundthem.

Theoretically, one's level of differentiation has a numberof important consequences for an individual. Foremost,Bowen (1978) proposed that less differentiated individualsexperience greater chronic anxiety: "The average level ofchronic anxiety of a person and of a . . . family parallels thebasic level of differentiation of that individual and family[and] the lower the level of basic differentiation, the higherthe average level of chronic anxiety" (Ken" & Bowen, 1988,p. 115). According to Bowen (1976, 1978; Kerr & Bowen,1988), less differentiated individuals also become dysfunc-tional under stress more easily and thus suffer more psycho-logical and physical symptoms (e.g., anxiety, somatization,depression, alcoholism, and psychoticism).

Conversely, highly differentiated individuals are thoughtto demonstrate better psychological adjustment. Some evi-dence has emerged in support of these notions. Greene,Hamilton, and Rolling (1986) discovered that inpatient andoutpatient participants, regardless of diagnosis, reportedsignificantly lower levels of differentiation than did those ina nonclinical control group. Likewise, adults who report lessfusion in their significant relationships have been shown toexperience fewer self-reported health problems (Bray, Har-vey, & Williamson, 1987).

More highly differentiated individuals are also expectedto remain in satisfying contact with their families of origin,establish more satisfying marriages, and be effective prob-lem solvers (Bowen, 1976, 1978). At present, only indirectsupport exists for the theoretical link between differentiationand marital satisfaction. Jacobson and his colleagues (Jacob-son, Follette, & McDonald, 1982; Jacobson, Waldron, &Moore, 1980) found that behavioral reactivity, defined as thetendency for spouses to react at the affective level to someimmediate stimulus from the partner, was associated withmarital distress. Couples who reported greater maritalsatisfaction showed less emotional reactivity in their ex-changes, whereas interactions of distressed couples werecharacterized by heightened emotional reactivity to immedi-ate positive and negative events in their relationships(Jacobson, Follette, & McDonald, 1982; Jacobson, Waldron,& Moore, 1980). Harvey, Curry, and Bray (1991) observedthat greater fusion and less intimacy with one's parentspredicted deficits in intimacy and greater emotional reactiv-ity with one's spouse.

Concern has been expressed about the paucity of empiri-cal research on the basic principles or constructs in Bowentheory (Gurman, 1978,1991). If Bowen theory is to continueto contribute significantly to the field, empirical means areneeded to test (and potentially modify) its basic assump-

tions. Thus, we undertook development of the Differentia-tion of Self Inventory (DSI) to create a self-report instru-ment for adults, age 25+, capable of (a) testing theoreticalassumptions, (b) assessing individual differences in adultfunctioning, and (c) evaluating psychotherapeutic outcomesfrom a systemic perspective. By defining adulthood with alower limit of 25 years of age, we sought to ensure that thesamples obtained consisted of those individuals who, from afamily life cycle perspective (Carter & McGoldrick, 1988),could be considered adults (i.e., postcollege or working,living apart from the parental home, and largely financiallyindependent).

To adequately measure differentiation, we included boththe intrapsychic and interpersonal components, that is, thethinking-feeling and separateness-togetherness dimensions.Historically, transgenerational theorists (e.g., Boszormenyi-Nagy & Ulrich, 1981; Framo, 1992) have described indi-vidual and family functioning solely in terms of interper-sonal and intergenerational family processes. Self-reportinstruments developed within this tradition include Kear's(1978) Differentiation of Self Scale, the Emotional CutoffScale (McCollum, 1991), the Family-of-Origin Scale (Hov-estadt, Anderson, Piercy, Cochran, & Fine, 1985), and thePersonal Authority in the Family System Questionnaire(Bray, Williamson, & Malone, 1984). Although each repre-sents an important contribution to the field, none attempts tooperationalize the range of interpersonal components ofdifferentiation (i.e., fusion to emotional cutoff), and nonefocuses on the intrapsychic aspects of differentiation (seeBowen, 1978; Kerr & Bowen, 1988).

For example, Kear's (1978) Differentiation of Self Scaleconsists of three factors: Separation of Thinking and Feel-ing, Emotional Maturity, and Emotional Autonomy; yetitems reflect only interpersonal components of differentia-tion and ignore quality of relations with spouse or partner.The Differentiation of Self Scale suffers also from signifi-cant methodological limitations. For instance, a factoranalysis used to create its subscales was conducted on 72initial items using only 50 participants (see Nunnally, 1978).

McCollum's (1986, 1991) Emotional Cutoff Scale is anexcellent measure of the degree to which respondentsmanage their emotional attachment to each parent throughcutoff. Yet its limited focus on relations with parents ignoresthe presence of emotional cutoff in current significantrelationships as well as other aspects of differentiation. Torespond to the Family of Origin Scale (Hovestadt et al.,1985), adults provide retrospective perceptions of theirfamily of origin relations, whereas adolescents are asked togive their current perceptions of relations with family (e.g.,Niedermeier, Handal, Brown, Searight, & Manley, 1992).The retrospective ratings emphasize the past and ignore therespondent's current relations with family members. Andalthough the Personal Authority in the Family SystemQuestionnaire (Bray et al., 1984) includes items aboutcurrent relationships, it neglects the concept of emotionalcutoff as well as the intrapsychic aspects of Bowen's (1976,1978) concept of differentiation.

There also exist several self-report measures of separation-individuation based on object relations theory (e.g., Hoff-

DIFFERENTIATION OF SELF 237

man, 1984; Levine, Green, & Millon, 1986; Olver, Aries, &Batgos, 1990). These separation-individuation measureswere designed for use with late adolescents rather thanadults, and none contain items that deal with maritalrelations or that reflect problems in achieving a balancebetween intimacy and autonomy. The concept of differentia-tion, as defined by Bowen (1976, 1978), is often misinter-preted in the family therapy literature and equated withindividuation or autonomy. Although similar in some re-spects, separation-individuation is not equivalent to differen-tiation of self. Individuation, from an object relationsperspective (e.g., Bios, 1975; Mahler, Pine, & Bergman,1975), involves the achievement of independence and aunique sense of identity. Differentiation of self is thecapacity to maintain autonomous thinking and achieve aclear, coherent sense of self in the context of emotionalrelationships with important others.

To create the DSI, a series of studies was undertakenbased on three different samples. The purpose of thesestudies was to develop and validate the DSI using a constructapproach to test construction (e.g., Jackson, 1970; Jackson& Messick, 1958; Loevinger, 1957; NunnalLy, 1978). Jack-son's recommendations for personality scale developmentwere used to construct items that would adequately reflectthe domain (i.e., differentiation of self), be clear andunambiguous, be relatively free of social desirability biasand other content biases, have high discriminatory power,and, as a set, sufficiently represent the underlying constructof differentiation (Jackson, 1970).

Study 1

The purpose of this study was to create the DSI. First,definitions, descriptions, and examples from Bowen (1976,1978; Anonymous, 1972) and his successors (Kerr, 1985;Kerr & Bowen, 1988; Nichols, 1984; Nichols & Schwartz,1998; Papero, 1990) were used to generate a pool of itemsthat exemplify differentiation of self. Items (N = 96) gener-ated by our research team reflected the ability to distinguishand balance (a) thinking and feeling and (b) the capacity forintimacy with and autonomy from others in current impor-tant relationships as well as with parents and siblings.Differentiation was operationalized in a multidimensionalfashion, given that Bowen (1976, 1978) described manycomponents of differentiation in his writings. Further, Gur-man (1978) argued that differentiation, like any complexpsychological construct, is inherently multidimensional. Weused a principal-components analysis to identify the DSFsdimensionality and determine final item selection. Theoreti-cal relations between differentiation and chronic anxietywere tested to assess the initial construct validity of the DSI(i.e., Bowen's proposition that poorly differentiated individu-als also experience more chronic anxiety).

Method

Participants. Participants were adults (A =313) living in NewYork, Ohio, and California, including (a) randomly selected facultyand staff at a large state university, (b) parents of children on

a suburban athletic team, (c) graduate students in counselingpsychology, clinical psychology, and social work, and (d) availablefriends and acquaintances of research team members. Completedquestionnaires were returned by 213 women and 98 men (2 genderunspecified), 75% of whom were married, 49% with children. Onaverage, participants were 36.8 years of age (SD = 9.69,range = 25-65). In terms of ethnicity, 5.1% of the sample wereAfrican American, 4.5% Asian American, 2.2% Latino-Latina,1.9% Native American, 82.7% White, and 3.2% other.

Instruments. Participants completed the 96-item DSI describedabove. The Trait version of the State-Trait Anxiety Inventory(STAI-T; Spielberger, Gorsuch, & Lushene, 1970) is a well-established 20-item self-report measure of relatively stable indi-vidual differences in anxiety proneness. Internal consistency esti-mates for the STATT-T have ranged from .86 to .92, and atest-retest reliability correlation over a 3-month interval wasreported to be .75. In contrast to large changes in STAI-State scoresproduced by stress conditions, STAI-T scores of chronic anxietyremain stable and unaffected by experimentally induced stresses(Spielberger et al., 1970).

Procedure. We contacted participants by form letter and askedthem to take part in a research project that focused on adults'interpersonal relationships and their relationships with their fami-lies of origin. Questionnaire packets consisted of the DSI, ademographic sheet, and the STAI-T. Each packet included a coverletter stating the purpose of the study and explaining the voluntaryand anonymous nature of the research. Postage-paid envelopeswere provided.

Results and Discussion

Subscales were developed on the basis of the responses of313 adults. A principal-components analysis was conductedusing an orthogonal rotation. We used a principal-compo-nents analysis because we were interested in identifying afew coherent dimensions that best reflected the variousaspects of the differentiation. Bowen's theory has manyconstructs that are not mutually exclusive but that relate todifferentiation of self. To have created subscales basedsolely on our own biases as to the relative importance ofthese theoretical constructs seemed less rigorous (cf. Jack-son, 1970) than allowing the respondents' ratings to helpdetermine the salient dimensions of the measure. Thus,although we created an initial pool of 96 items representa-tive of the substantive domain of differentiation, the finalbasis of item selection was empirical.

Four factors were identified with eigenvalues greater than3.0, ranging from 11.43 to 3.34. Results of Cattail's screeplot of the factor variances showed a substantial break afterfour factors; these four factors accounted for 26.2% of thevariance. To interpret the factors and construct scales, weconsidered only those items loading at least .40 on a singlefactor (n = 43). The following factors were identified:Factor 1, with 12 items, was defined as Emotional Reactiv-ity; Factor 2, with 10 items, was defined as taking anI Position; Factor 3, with 13 items, was defined as ReactiveDistancing; and Factor 4, with 9 items, was defined asFusion With Parents, (A table listing items and their factorloadings is available from Elizabeth A. Skowron.)

We conducted subsequent analyses, using the foursubscale scores and a total DSI score. Scores were reversedon the items constituting Emotional Reactivity, Reactive

238 SKOWRON AND FRIEDLANDER

Distancing, and Fusion With Parents subscales to signifyless differentiation; thus, higher scores on each subscalereflected greater levels of differentiation. To compute theDSI full scale, all 44 items were summed so that higherscores reflected greater differentiation of self.

Internal consistency estimates using Cronbach's alphasuggested high reliabilities for the DSI total scale and eachof the four subscales: DSI = .88; Emotional Reactivity =.83; Reactive Distancing = .80; Fusion With Parents = .82;and I Position = .80. Subscale correlations with the DSI fullscale were moderate to high: .59 (Fusion With Parents), .65(I Position), .75 (Reactive Distancing), and .80 (EmotionalReactivity). Correlations among the four subscales weresmall to moderate: .37 (Emotional Reactivity and I Posi-tion), .45 (Emotional Reactivity and Reactive Distancing),.31 (Emotional Reactivity and Fusion With Parents), .34 (IPosition and Reactive Distancing), .17 (I Position andFusion With Parents), and .18 (Reactive Distancing andFusion With Parents). In support of the DSI's constructvalidity, level of differentiation, as measured by the DSI,correlated highly with a measure of chronic anxiety. DSIfull-scale scores significantly predicted Trait Anxiety, mea-sured by the STAI-T (r = .64, p < .0001). Correlationsbetween Trait Anxiety and the four subscales ranged from.16 (p < .01, Fusion With Parents) to .51 (I Position), .55(Reactive Distancing), and .58 (Emotional Reactivity), allremaining ps < .0001.

Study 2

The purpose of Study 2 was to revise the theoretical focusand item content of the original DSI because of theconsiderable amount of variance left unaccounted for in theprevious factor analysis. In this study, the DSI subscalesunderwent conceptual revisions, and its psychometric prop-erties were strengthened on the basis of item analyses and acritical examination of social desirability bias. Once again, aconstruct approach to personality scale construction (e.g.,Cronbach & Meehl, 1955; Jackson, 1970; Jackson &Messick, 1958; Loevinger, 1957; Nunnally, 1978) was used.

The factor structure of the original DSI was retained in thepresent revision. First, the four or five items with the highestitem-total correlations within each subscale were identified.The content of these items guided our decisions aboutretaining or modifying each subscale name or definition. TheEmotional Reactivity and I Position subscales appeared tobest represent the constructs as in the literature and thusunderwent only modest revisions. Because the ReactiveDistancing and Fusion With Parents subscales had emergedas conceptually weaker, we refined their conceptualizationsand renamed them Emotional Cutoff and Fusion WithOthers, respectively.

Revisiting definitions and descriptions of differentiationbased on Bowen theory (Anonymous, 1972; Bowen, 1976,1978; Kerr & Bowen, 1988), we generated a pool of 78items, which was submitted along with subscale definitionsto two experts on Bowen theory, who suggested the revisionof some items and subscale definitions. Next, on the basis ofthe responses of a second adult sample, item analyses were

conducted to minimize social desirability bias and select thebest items for each subscale. Descriptive statistics werecomputed, along with internal consistency reliabilities.

MethodParticipants, Adults (n = 169, 111 women and 58 men), age

25+, who were employed at a large northeastern state agency, tookpart in the research. Participants averaged 42.34 years of age(SD = 8.59). The majority were married (70.2%; M - 15.04years), 13.7% were single, 6.3% were unmarried and living with apartner, and 9.5% were separated or divorced. In terms of ethnicity,90.4% were White, 5.4% African American, 0.6% Asian American,0.6% Latino-Latina, 0.6% Native American, and 2.4% other.Approximately 15% of participants were currentLy in therapy; 45%had sought therapy in the past.

Instruments. The DSI used in Study 2 contained 78 itemsconstituting four subscales: Emotional Reactivity, I Position,Emotional Cutoff, and Fusion With Others. To rate each item,respondents used a 6-point Likert-type scale, ranging from not atall true of me (1) to very true of me (6). Crowne and Marlowe's(1964) Social Desirability Scale (SDS), a 33-item true-falseself-report measure, was used to estimate the tendency to describeoneself in favorable terms. Internal consistency reliability has beenestimated at .88, with test-retest correlations at .88 and .89(Crowne & Marlowe, 1960,1964; Robinson & Shaver, 1973).

Procedure. Adults, age 25+, who were employed at a largenortheastern state agency, took part in the research. Cluster-sampling procedures were used to randomly select 2 departmentsfrom a large northeastern state agency, out of 32 total from which tosolicit participants. Three hundred potential participants werecontacted by interagency mail. Each packet included a cover letterexplaining the voluntary and anonymous nature of the research, thetwo counterbalanced questionnaires, and a demographic sheet. Thestudy was described as "focusing on adults' interpersonal relation-ships, relationships with family members, and (their) generalattitudes." Participants returned completed packets by mail insealed envelopes. One hundred sixty-nine participants returnedcompleted questionnaires, for a 56% return rate.

ResultsItem analyses. Statistical analyses were performed at

the item level to discern the DSI's inherent factor and toensure that each subscale was homogeneous and distinctfrom the other three subscales (Campbell & Fiske, 1959;Jackson, 1970). All items met a priori criteria for responsedistribution (i.e., items with skewness and kurtosis valuesbetween —1.5 and 1.5 and SDs >: 1 were retained). Thirty-five items were eliminated due to low item-scale correla-tions (i.e., items with item-subscale correlations <.45 wereeliminated). Seven items were rekeyed because they loadedhighly on another subscale and demonstrated good discrimi-nation between subscales and because the Bowen experts weconsulted suggested that those items corresponded morehighly to that respective subscale.

No additional items were eliminated on the basis ofcriteria for evaluating social desirability bias. Correlationsbetween DSI items and social desirability scores rangedfrom -.15 to .49. None of the remaining 43 items werefound to lower the internal consistency reliability of their

DIFFERENTIATION OF SELF 239

respective subscales; thus, no additional items were elimi-nated. (Tables illustrating item response distributions, skew-ness and kurtosis values, item correlations with the SDS,item-subscale correlations, and item reliability analyses areavailable from Elizabeth A. Skowron.)

Description of the DSL The resulting 43-item DSI (seeAppendix) contains four subscales: Emotional Reactivity, IPosition, Emotional Cutoff, and Fusion With Others. The11-item Emotional Reactivity subscale reflects the degree towhich a person responds to environmental stimuli withemotional flooding, emotional lability, or hypersensitivity.The I Position subscale contains 11 items that reflect aclearly defined sense of self and the ability to thoughtfullyadhere to one's convictions when pressured to do otherwise.The 12-item Emotional Cutoff subscale reflects feelingthreatened by intimacy and feeling excessive vulnerabilityin relations with others. Items reflect fears of engulfment andbehavioral defenses like overfunctioning, distancing, ordenial. Finally, the 9-item Fusion With Others subscalereflects emotional overinvolvement with others, includingtriangulation and overidentification with parents.

To compute the DSI full-scale score, raw scores on allitems in the Emotional Reactivity, Emotional Cutoff, andFusion With Others subscales and on one item in the IPosition subscale (#35) are reversed, so that higher scoressignify greater differentiation. Scores on all items are thensummed and divided by the total number of items, so that thefull-scale score ranges from 1 (low differentiation) to 6 {highdifferentiation). To facilitate comparison of full-scale andsubscale scores, each subscale is also computed by reversingrespective items, summing item scores, and then dividing bythe number of items in the subscale (Emotional Reactiv-ity = 11, I Position = 11, Emotional Cutoff = 12, FusionWith Others = 9). Scores on each subscale thus range from1 to 6, with higher scores reflecting greater differentiation.

Descriptive statistics. All scores were normally distrib-uted; subscale means ranged from 2.07 to 4.34 (full-scaleM = 3.73, SD = 0.58; see Table 1). Subscale-full-scalecorrelations were moderate to high, ranging from .43(Fusion With Others) to .80 (Emotional Reactivity), all ps <.001. Intercorrelations among the subscales were low tomoderate, ranging from .08 (Fusion With Others and I

Position) to .53 (Fusion With Others and Emotional Reactiv-ity; see Table 2). Correlations between DSI subscales andSDS scores were negligible to moderate {r — .42 for Emo-tional Reactivity, r = .49 for I Position, r = .34 for Emo-tional Cutoff, and r = - .02 for Fusion With Others).Cronbach's alpha was used to estimate internal consistencyreliabilities for the DSI full scale and each of the foursubscales (DSI a = .88, Emotional Reactivity a = .84; IPosition a = .83, Emotional Cutoff a = .82; Fusion WithOthers a = .74).

Study 3After the DSI subscale revisions in Study 2, a third sample

was obtained to evaluate the DSI's factor structure usingconfirmatory factor analyses and to test theoretically pre-dicted relations between differentiation of self, psychologi-cal symptoms, and marital satisfaction. It was hypothesizedthat (a) significant inverse relationships between symptom-atology and the DSI subscales would support Bowen's(1976,1978) assumption that highly differentiated individu-als are more free of symptoms and generally better adjustedand (b) significant positive relationships between maritalsatisfaction and the DSI subscales would support Bowen's(1976,1978; Kerr & Bowen, 1988) proposition that individu-als with higher levels of differentiation establish moresatisfying marriages.

Method

Participants. A total of 127 adults (118 employees and 9spouses) participated, with only 91 married adults completing theDyadic Adjustment Scale (DAS). Participants were 53 men and 73women (1 gender unspecified), with an average age of 42.23 years(SD — 10.22, range 25-72 years). Married respondents constituted59.5% of the sample (M = 12.72 years married, SD = 9.74). Of theremainder, 15.9% were single, 9.4% were unmarried and livingwith a partner, 13.5% were separated or divorced, and 1,6% werewidowed. More than half of the participants (61.8%) were parents(M = 2.13 children, SD = 0.94). In terms of ethnicity, 90.5% of thesample were White, 4.0% African American, 2.4% Asian Ameri-can, 1.6% Latino-Latina, and 0.8% Native American. Ten percent

Table 1Means and Standard Deviations on the DSI

Study 3

ScaleERIPECFODSI

StudyM

3.354.014.342.973.73

2SD

0.900.830.870.880.58

M3.374.084.532.923.74

TotalSD

0.940.850.790.850.60

M3.69a4.244.443.053.87

MenSD

0.880.900.770.890.55

WomenM

3.18b3.974.612.823.64

SD0.920.810.810.820.61

Note. For Study 2, n = 169; for Study 3, n = 127 (53 men, 73 women, 1 unspecified). DSI =Differentiation of Self Inventory; ER = Emotional Reactivity; IP = I Position; EC = EmotionalCutoff; FO = Fusion With Others. Scores range from 1 to 6. Higher scores on all scales reflect greaterdifferentiation of self. Means in the same row that do not share subscripts differ at p < .01.

240 SKOWRON AND FRIEDLANDER

Table 2Intercorrelations Among Subscalesand Subscale-Full-Scale Correlations

Scale 1Study 2

1. DSI2. ER3. IP4. EC5. FO

1. DSI2. ER3. IP4. EC5. FO

—.80**.61**.64**.43**

—.84**.69**.58**.52**

—.46**.27*.53**

Study 3

—.53**.25.48**

—.31*.08

—.28*.12

-.12 —

-.04 —Note. DSI = Differentiation of Self Inventory; ER = EmotionalReactivity; IP = I Position; EC - Emotional Cutoff; FO = FusionWith Others. Scores on the DSI and subscales range from 1 to 6.Higher scores on all scales represent greater differentiation of self.*p < .001, two-tailed. **p < .0001, two-tailed.

of participants were currently in therapy; 52.70% had past experi-ence in psychotherapy, predominantly (59.40%) in individualtreatment.

Instruments. The four subscales in the 43-item DSI were usedas predictor variables. In the Study 3 sample, internal consistencyreliabilities were moderate to high and similar to those obtained inprevious studies (DSI a = .88, Emotional Reactivity a = .88; 1Position a = .85; Emotional Cutoff a = .79; Fusion With Othersa = .70).

The Hopkins Symptom Checklist (Derogatis, Lipman, Rickels,Uhlenhuth, & Covi, 1974) is a well-known self-report measureassessing psychological symptomatology on five dimensions: So-matization, Obsessive-Compulsive, Interpersonal Sensitivity, De-pression, and Anxiety. Items are rated on a 4-point Likert-typescale, ranging from 1 (not at all) to 4 (extremely), reflecting thedegree of distress experienced within the past 7 days. Its GeneralSeverity Index (GSI) reflects intensity of distress independent ofthe number of symptoms endorsed (Derogatis, Yevzeroff, &Wittelsberger, 1975). The GSI is sensitive to symptom changesover the course of psychotherapy (Rickels et al., 1971) and is usedmost often to provide a summary measure of symptomatology(Derogatis et al., 1974). The GSI is computed by summing the fiveraw symptom subscales and dividing by 58; scores range from 1 to4, with higher scores indicating greater symptomatology. Internalconsistency reliabilities range from .84 to .87, and test-retestcoefficients range from .75 to .84 (Derogatis et al., 1974). GSIscores in the present sample ranged from 1.0 to 2.6 (M = 1.5,SD = 0.33).

Spanier's (1976) DAS assesses relationship discord and overallmarital satisfaction. The DAS yields a total score ranging from 0 to151, with higher scores reflecting better marital adjustment.Internal consistency reliability of the DAS full-scale score has beenreported at .96 (Spanier, 1976). Construct validity is supported bysignificant correlations with other well-known measures of maritaladjustment and by results showing that divorced couples scoresignificantly lower than married couples (Spanier, 1976, 1988).Scores in the present sample ranged from 50 to 150 (M = 104,SD = 17.9). On the basis of the accepted DAS cutoff score of 98(Eddy, Heyman, & Weiss, 1991; Jacobson et al., 1984), 30 marriedparticipants (30.90%) in Study 3 were classified as maritallydistressed.

Procedure. Three hundred adult participants were randomlyselected from an available sample of staff (e.g., administrative,building maintenance, and clerical), faculty, and their spouses at anortheastern state university. Participants were contacted by letterthrough the campus mail and asked to complete a packet consistingof the three randomly ordered questionnaires and a demographicsheet. Unmarried participants were instructed to disregard theDAS, and married volunteers were invited to request an additionalpacket for their spouse. (The original plan was to compare couples'scores, but too few spouses returned questionnaires.) Each packetincluded a cover letter that described the study as "focusing onadults1 interpersonal relationships, relationships with family mem-bers, and (their) general attitudes" and explained its voluntary andanonymous nature. The return rate was 42.3%.

ResultsPreliminary analyses. Means and standard deviations

on the four DSI subscale and full-scale scores were highlysimilar to those obtained in Study 2 (see Table 1). As inStudy 2, scores were normally distributed, subscale intercor-relations were moderate (Emotional Reactivity and I Posi-tion r — .53; Emotional Reactivity and Fusion With Othersr = .48) to negligible (e.g., Fusion With Others and Emo-tional Cutoff r= - .04; see Table 2).

Factor analyses. We conducted a confirmatory factoranalysis, using Lisrel 7, to evaluate the four-factor structureof the DSI. Given the limitations inherent in using Lisrelprocedures when fitting models with large numbers ofsingle-item indicators such as the DSI, an item-clusteringprocedure was used to increase the stability of the indicators(c.f. Anderson & Gerbing, 1988; Joreskog & Sdrbom, 1986;MacCallum, 1986). For each of the 4 DSI subscales, singleitems were randomly summed into meta-items, comprising 3or 4 items each, resulting in 3 indicators per subscale, or 12indicators in all. Table 3 contains the correlation matrixamong the 12 confirmatory factor analysis indicator variables.

Fit indices used to evaluate the model included thegoodness-of-fit index (GFI), adjusted goodness-of-fit index(adjusted GFI), a chi-square to degrees of freedom ratio(X2'4O, and the root mean squared of the residuals (RMS).GFI values greater than .90, adjusted GFI values greater than.80, a xWrat io less than 2.0, and RMS values less than . 10indicate a well-fitting model (cf, Cole, 1987; Joreskog &Sorbom, 1986; Marsh & Hocevar, 1985).

The four-factor model of differentiation, corresponding tothe four DSI subscales, was tested. Indicators expected toload on each factor were stated in advance (Nunnally, 1978).Each of the indicators was permitted to load freely on itsrespective factor and was constrained to 0 on the otherfactors. Each latent variable was scaled to the first indicatorby fixing its value to 1.00. A maximum-likelihood solutionwas used to fit the model to the data. The fit of thisfour-factor model was good, x2(48, N = 137) = 89.35,p < .0001, GFI = .91, adjusted GFI = .85, ^idf = 1.86,RMS = .07.

Figure 1 illustrates the four latent variables (EmotionalReactivity, I Position, Emotional Cutoff, and Fusion WithOthers), the individual indicator loadings on the latentvariables, and the residual error terms. Correlations amongthe four factors were negligible to moderate, ranging from

DIFFERENTIATION OF SELF 241

Table 3Correlation Matrix for Confirmatory Factor AnalysesVariable

1. ER,2. ER23. ER34. IP!5. IP26. IP37. ECi8. EC29. EC3

10. FO!11. FO212. FO3

1—.74.76.35.49.45.22.16.06.34.37.32

2

—.63.24.42.40.26.27.13.39.33.33

3

—.35.46.53.26.24.18.30.31.32

4

—.70.47.20.28.06.01.13

-.05

5

—.64.23.18.01.02.21.05

6

—.32.39.15.00.14.09

7

—.51.50.07.04

- .08

8

—.62.02

- .06-.05

9

—-.01-.12-.07

10

—.49.48

11

—.50

12

—Note. ER = Emotional Reactivity; IP = I Position; EC - Emotional Cutoff; FO = Fusion WithOthers; DSI = Differentiation of Self Inventory. ER, = DSI Items 1,14,26, 38; ER2 = DSI Items 6,18, 30,40; ER3 = DSI Items 10, 21,34; IPi = DSI Items 4,15,27,41; EP2 = DSI Items 7,19, 31,43;TP3 = DSI Items 11, 23, 35; EC, = DSI Items 2, 12, 24, 36; EC2 - DSI Items 3, 16, 28, 39; EC3 =DSI Items 8,20,32,42; FOj = DSI Items 5,17,29; FO2 = DSI Items 9,22,33; FO3 = DSI Items 13,25, 37. To estimate each indicator, individual items were summed and divided by number of items perindicator.

- .06 (Fusion With Others and Emotional Cutoff) to .59(Emotional Reactivity and I Position). Next, a related factormodel was also tested representing the four DSI subscales asfactors, with differentiation of self identified as a single,higher order latent factor. Results were also positive, x2(50,N = 137) = 94.58, p < .0001, GFI - .91, adjusted GFI =.86, and y^tdf = 1.89, for the DSI as representing a single,multidimensional construct.

Tests of Bowen theory. On the basis of the results ofpreliminary univariate regression and one-way analyses ofvariance conducted using a familywise alpha of .01, age andgender, respectively, were included as covariates in themajor analyses. Gender showed a significant relationshipwith Emotional Reactivity, F(l, 113) = 7.05, p = .01, withwomen reporting relatively more emotional reactivity(M = 3.18) than men (M = 3.69; see Table 1). (Recall thatbecause higher subscale scores reflect greater differentiation,a lower Emotional Reactivity score signifies greater emo-tional reactivity). Age showed a unique relationship withFusion Wilh Others, r(l, 104) - 3.08, p < .003, r = .45,with younger participants reporting greater difficulties withfusion. Level of educational attainment, marital status(single vs. married), parental status (yes vs. no), andtreatment history all failed to predict DSI scores at a = .01.(A table illustrating intercorrelations among predictor, covari-ate, and criterion variables is available from Elizabeth A.Skowron.)

A univariate multiple regression analysis was conducted,with four predictors (DSI subscales), two covariates, and thecriterion variable, GSI. A significant inverse relationshipemerged between scores on the four DSI subscales and theGSI, with age and gender controlled, F(4,104) = 18.73,p <.0001, tf^mipartia! = -42. In other words, higher differentiationscores predicted significantly less symptomatic distress.Examination of the t tests on each beta weight showed thatEmotional Reactivity and Emotional Cutoff made significantunique contributions, /3 = - .45 , t(\, 104) = -4.07 forEmotional Reactivity; /3 = - .32, t{\, 104) = -3.90 for

Emotional Cutoff; both p$ < .001. Valences of these betaweights indicated that greater emotional reactivity andcutoff predicted greater symptomatic distress.

Next, we performed a univariate multiple regressionanalysis on the set of four DSI predictors, two covariates,and the criterion, DAS. Only data from the married partici-pants (n = 91) were analyzed. A significant relationshipemerged between scores on the four DSI subscales and theDAS. With age and gender controlled, F(4, 84) = 6.79, p <.0001, semipartiai = -24» higher DSI scores predicted greatermarital satisfaction. Only Emotional Cutoff made a signifi-cant, unique contribution, /3 = .39, f(l, 84) = 3.86, p <.001, indicating that less emotional cutoff predicted greatersatisfaction.

General DiscussionOur aim, to construct a reliable, valid self-report measure

of differentiation of self for adults using a construct ap-proach to test development, was realized in these investiga-tions. While a portion of the items were created or rewordedduring the Study 2 revision of the DSI, its multidimensionalstructure was retained. Confirmatory factor analyses demon-strated support for the DSI subscales, Emotional Reactivity,I Position, Emotional Cutoff, and Fusion with Others, asidentifiable, empirically distinct dimensions of a singleconstruct, differentiation of self. Subsequent analyses withthe DSI subscales supported the internal consistency reliabil-ity and initial construct validity of the measure. Tests ofBowen theory supported the hypothesized relations betweenself-reported differentiation, symptomatology, and maritalsatisfaction, providing important psychometric support forthe DSI.

Differentiation of self, estimated by the DSI subscales,correlated significantly with amount and intensity of symp-tomatic distress. The unique predictors of global maladjust-ment were Emotional Reactivity and Emotional Cutoff, the

242 SKOWRON AND FRIEDLANDER

er, er,

.17r 1

.36' 1

32r i f ^

.15f 1

.51r y

.58 34 .44r i

.52 .51 .50

ER1 ER2 ER3 IP1 IP2 IP3 EC1 EC2 EC3 FO1 FO2 FO3

.83.92 .82 .70

Figure 1. Study 3: four-factor model confirmatory factor loadings, error variances, and correlationsamong the latent variables (ER = Emotional Reactivity; IP = I Position; EC = Emotional Cutoff;FO = Fusion With Others).

two subscales whose items reflect difficulties in handlingaffect. (Differentiation scores obtained in Study 1 alsosignificantly predicted trait anxiety, a noteworthy indicatorof construct validity because, according to Bowen theory,lack of differentiation is closely equated with chronicanxiety; Kerr & Bowen, 1988.) These results, taken together,suggest support for Bowen's (1976, 1978; Kerr & Bowen,1988) contention that differentiation of self is an importantaspect of psychological well-being.

Taken together, the four DSI subscales also showed astrong relationship with marital satisfaction. According toBowen theory (Bowen, 1978; Kerr & Bowen, 1988), maritalproblems are created when spouses are less differentiated,and the resulting reactivity or cutoff heightens their anxiety.Emotional Cutoff scores also were found to uniquely predictmarital satisfaction. Such a result is consistent with Gottmanand KrokofTs (1989) longitudinal study of marital interac-

tion, in which "withdrawal from interaction," or cutoff,predicted concurrent marital distress and deterioration inmarital satisfaction over time (p. 49). Perhaps further studyof differentiation and its relationship to marital satisfactionmay reveal that separation and divorce represent extremebehavioral manifestations of emotional cutoff in relation-ships. Could increases in emotional cutoff or emotionallyavoidant defenses be a common cause or result of divorce? Ifso, would such changes be state dependent or longstandingin nature?

Other results revealed some interesting patterns in differ-entiation across age and gender. First, there was no relation-ship observed between age and scores on the I Positionsubscale, suggesting that younger adults are just as capableas their older counterparts of defining a self and behavingautonomously (Bowen, 1978). With respect to gender,Bowen (1978) asserted that no gender differences exist on

DIFFERENTIATION OF SELF 243

levels of differentiation. In contrast, sex role socialization,self-in-relation, and feminist family theorists (e.g., Carter &McGoldrick, 1988; Chodorow, 1978; Gilligan, 1982; Jossel-son, 1988; Luepnitz, 1988; Miller, 1976) would posit thatgender differences exist on differentiation, with womenmore likely to endorse Emotional Reactivity items and menmore likely to endorse Emotional Cutoff items. The resultsof our investigations revealed that women reported signifi-cantly greater emotional reactivity than did men. Con-versely, no significant gendeT differences were observed onEmotional Cutoff. Given that significant gender differenceswere present on level of education (with men reportinggreater levels than women) and that samples comprisedunequal numbers of men and women, caution is advised ininterpreting the observed gender effects. We encourageinvestigators to continue testing the DSI for gender differ-ences, using equal numbers of men and women from othersocioeconomic strata and matched on relevant third vari-ables to clarify the ways in which differentiation problemsare manifested across genders.

Although efforts were made to sample three separate,heterogeneous groups of adults, future tests of the DSI areneeded, with new samples representing a wider range ofdemographic variables. Results of the exploratory andconfirmatory factor analyses may have differed with differ-ent samples. Cross-validation is needed with larger samples,and demographic characteristics of the respondents need tobe taken into account. The current respondents were, onaverage, middle-aged, White, educated, employed individu-als who were married and had children. In addition, notest-retest reliability estimates were obtained for the DSI orits subscales.

Ethnicity is another characteristic that should be takeninto account, because only approximately 10% of thepresent samples identified themselves as minority members.Further independent studies are needed with adults fromdifferent ethnic groups, to test Bowen's (1978; Kerr &Bowen, 1988) assertion that differentiation of self is univer-sally applicable. We suspect that in Latino-Latina, NativeAmerican, or Asian cultures, for example, Fusion WithOthers might not correlate significantly with chronic anxi-ety, psychological symptoms, or marital dissatisfaction.Perhaps in Asian cultures, where autonomy and self-assertion are less valued that in Western societies, I Positionscores may correlate positively (rather than negatively, asreported here) with symptomatology and chronic anxiety.These questions await future investigation.

One of the greatest challenges in creating a self-reportmeasure of differentiation is the difficulty inherent inmeasuring a systemic construct in a reliable and validmanner. Given the possibility for error due to monomethodbias in the current studies, further support for the validity ofthe DSI could be obtained by comparing adults' self-ratingswith Bowen experts' ratings of their levels of differentiationobtained through structured clinical interviews (cf. Kerr &Bowen, 1988). If the therapists' ratings closely reflect theirclients* self-reported scores, such a result would furthersupport the instrument's construct validity. Likewise, givenBowen's (1978) proposition that chronic, debilitating disor-

ders occur with more frequency among those with lowerdifferentiation, further evidence for the DSI's constructvalidity may be found by determining whether differentia-tion scores covary with severity of psychiatric diagnoses,that is, major depressive disorder vs. dysthymic disorder;alcohol dependence vs. abuse; or Axis II vs. Axis I disorders(according to the Diagnostic and Statistical Manual ofMental Disorders; American Psychiatric Association, 1994).

It is expected that the DSI may contribute to the practiceof family therapy in several ways. First, the instrument mayprovide a means for identifying individual differences invarious aspects of functioning that are purportedly stableand central to a client's intrapsychic and interpersonalwell-being. Second, Bowen (1976, 1978) recommendedworking only with the most differentiated person in a family,theorizing that change achieved by this individual willindirectly benefit the entire family system. Researcherscould test Bowen's assertion by comparing the effectivenessof individual therapy with the most differentiated familymember versus traditional family systems therapy with allfamily members present (cf. Szapocznik, Kurtines, Foote,Perez-Vidal, & Hervis, 1983). Perhaps the DSI could beused as a screening device to identify the family membermost likely to enter into and benefit from treatment.

Third, because the construct differentiation is multidimen-sional, a comparative analysis of a client's scores on the DSIsubscales may help pinpoint which aspect of differentiationis most problematic (e.g., emotional reactivity or problemstaking an I position) and whether the client copes with his orher interpersonal difficulties through, for example, fusion oremotional cutoff. Indeed, an understanding of differentiationprovided by use of the DSI in future investigations couldresult in important implications for treatment. For example,emotional cutoff displayed by spouses in marital therapymight suggest the need for interpersonal, experiential inter-ventions (e.g., Greenberg & Johnson, 1988), whereas emo-tionally reactive young adults who are fused with theirparents might benefit more from an individual, insight-oriented approach. Clients in the midst of marital separationor postdivorce adjustment may benefit from family-of-originwork (e.g., Framo, 1992) to gain insight into the nature oftheir partner choices and decrease emotional cutoff andreactivity in their intimate relationships.

Likewise, researchers might investigate whether couplescould benefit from premarital counseling that incorporatesuse of the DSI. Could results of their scores on the foursubscales be used to facilitate discussion about whichaspects of differentiation are more difficult for each of themand how their characteristic differentiation problems areprimarily expressed? For example, with respect to thedistance-pursue pattern frequently observed in couples andnoted in the family therapy literature (e.g., Friedlander,Heatherington, Johnson, & Skowron, 1994; Guerin, Fogarty,Fay, & Kautto, 1996; Minuchin & Nichols, 1993), mightemotional cutoff in one spouse be complemented (andheightened) by greater fusion in the other, and vice versa?Would couples in premarital counseling, who learn abouttheir contrasting differentiation styles in the context ofcommunication and problem-solving skills training, supple-

244 SKOWRON AND FRIEDLANDER

ment their skill development with a deeper understanding ofone another's characteristic styles of reacting to stress?

Finally, the DSI may lend itself to the examination ofclient outcome in therapy. In future investigations, the DSIcould be used to test Bowen's (1978; Kerr & Bowen, 1988)proposition that psychotherapy can produce moderate in-creases in a person's level of differentiation. Given thatdifferentiation scores are strongly associated with overallpsychological adjustment (as Bowen asserted), if the DSIsubscales are also demonstrated to be sensitive to changes inthe client's differentiation over the course of therapy,increases in differentiation may be observed regardless ofthe approach to treatment.

References

American Psychiatric Association. (1994). Diagnostic and statisti-cal manual of mental disorders (4th ed.). Washington, DC:Author.

Anderson, J. C , & Gerbing, D. W. (1988). Structural equationmodeling in practice: A review and recommended two-stepapproach. Psychological Bulletin, 103, 411—423.

Anonymous. (1972). Toward the differentiation of a self in one'sown family. In J. Framo (Ed.), Family interaction (pp. 111-173).New York: Springer.

Bios, P. (1975). The second individuation process of adolescence.New York: International Universities Press.

Boszormenyi-Nagy, I., & Ulrich, D. N. (1981). Contextual familytherapy. In A. S. Gurman & D. P. Kniskera (Eds.), Handbook offamily therapy (pp. 159-186). New York: Brunner/Mazel.

Bowen, M. (1976). Theory in the practice of psychotherapy. In P. J.Guerin, Jr. (Ed.), Family therapy: Theory and practice (pp.42-90). New York: Garner Press.

Bowen, M. (1978). Family therapy in clinical practice. New York:Jason Aronson.

Bray, J. H., Harvey, D. M., & Williamson, D. S. (1987). Intergenera-tional family relationships: An evaluation of theory and measure-ment. Psychotherapy, 24, 516-528.

Bray, J. H., Williamson, D. S., & Malone, P. E. (1984). Personalauthority in the family system: Development of a questionnaireto measure personal authority in intergenerational family pro-cesses. Journal of Marital and Family Therapy, 10, 167-178.

Campbell, D. X, & Fiske, D. W. (1959). Convergent and discrimi-nant validation by the multitrait-multimethod matrix. Psychologi-cal Bulletin, 56, 81-105.

Carter, B., & McGoldrick, M. (1988). The changing family lifecycle: A framework for family therapy (2nd ed.). New York:Garner Press.

Chodorow, N. (1978). The reproduction of mothering: Psychoanaly-sis and the sociology of gender. Berkeley: University of Califor-nia Press.

Cole, D. A. (1987). Utility of confirmatory factor analysis in testvalidation research. Journal of Consulting and Clinical Psychol-ogy, 55, 584-594.

Cronbach, L. J., & Meehl, P. E. (1955). Construct validity inpsychological tests. Psychological Bulletin, 52, 281-302.

Crowne, D., & Marlowe, D. (1960). A new scale of socialdesirability independent of psychopathology. Journal of Consult-ing Psychology, 24, 349-354.

Crowne, D., & Marlowe, D. (1964). The approval motive. In D. P.Crowne (Ed.), The experimental study of personality (pp.153-183). New York: Wiley.

Derogatis, L. R., Lipman, R. S., Rickels, K., Uhlenhuth, E. H., &Covi, L. (1974). The Hopkins Symptom Checklist (HSCL): Aself-report symptom inventory. Behavioral Science, 19, 1-15.

Derogatis, L. R., Yevzeroff, H., & Wittelsberger, B. (1975). Socialclass, psychological disorder, and the nature of the psychopatho-logic indicator. Journal of Consulting and Clinical Psychology,43, 183-191.

Eddy, J. M., Heyman, R. E., & Weiss, R. L. (1991). An empiricalevaluation of the Dyadic Adjustment Scale: Exploring thedifferences between marital "satisfaction" and "adjustment."Behavioral Assessment, 13, 199-220.

Framo, J. L. (1992). Family of origin therapy. New York: Brurmer/Mazel.

Friedlander, M. L., Heatherington, L.. Johnson, B., & Skowron, E.A. (1994). "Sustaining engagement": A change event in familytherapy. Journal of Counseling Psychology, 41, 438-448.

Gilligan, C. (1982). In a different voice: Psychological theory andwomen's development. Cambridge, MA: Harvard UniversityPress.

Gottman, J. M., & Krokoff, L. J. (1989). Marital interaction andsatisfaction: A longitudinal view. Journal of Consulting andClinical Psychology, 57, 47-52.

Greenberg, L. S., & Johnson, S. M. (1988). Emotionally focusedtherapy for couples. New York: Guilford Press.

Greene, G. J., Hamilton, N., & Rolling, M. (1986). Differentiationof self and psychiatric diagnosis: An empirical study. FamilyTherapy, 8, 187-194.

Guerin, P. J., Fogarty, T. F, Fay, L. F., & Kautto, J. G. (1996).Working with relationship triangles: The one-two-three of psycho-therapy. New York: Guilford Press.

Gurman, A. S. (1978). Contemporary marital therapies. A critiqueand comparative analysis of psychoanalytic, behavioral, andsystems theories. In T. J. Paolino & B. S. McCrady (Eds.),Marriage and marital therapy (pp. 506-518). New York:Brunner/Mazel.

Gurman, A. S. (1991, March/April). Family therapy's neglectedprophet. The Family Therapy Networker, 77, 25-37.

Harvey, D. M., Curry, C. J., & Bray, J. H. (1991). Individuation andintimacy in intergenerational relationships and health: Patternsacross two generations. Journal of Family Psychology, 5,204-236.

Hoffman, J. A. (1984). Psychological separation of late adolescentsfrom their parents. Journal of Counseling Psychology, 31,170-178.

Hovestadt, A. J., Anderson, W. T., Piercy, F. P., Cochran, S. W., &Fine, M. (1985). A family-of-origin scale. Journal of Maritaland Family Therapy, 11, 287-297.

Jackson, D. N. (1970). A sequential system for personality scaledevelopment. In C. D. Spielberger (Ed.), Current topics inclinical and community psychology (Vol. 2, pp. 61-96). NewYork: Academic Press.

Jackson, D. N., & Messick, S. (1958). Content and style inpersonality assessment. Psychological Bulletin, 55, 243-252.

Jacobson, N. S., Follette, W. C, & McDonald, D. W. (1982).Reactivity to positive and negative behavior in distressed andnondistressed married couples. Journal of Consulting and Clini-cal Psychology, 50, 706-714.

Jacobson, N. S., Follette, W. C , Revenstorf, D., Baucom, D. H.,Hahlweg, K., & Margolin, G. (1984). Variability in outcome andclinical significance of behavioral marital therapy: A reanalysisof outcome data. Journal of Consulting and Clinical Psychology,52, 497-504.

DIFFERENTIATION OF SELF 245

Jacobson, N. S., Waldron, H., & Moore, D. (1980). Toward abehavioral profile of marital distress. Journal of Consulting andClinical Psychology, 48, 696-703.

Jdreskog, K. G., & Sorbom, D. (1986). LISREL IV: Analysis oflinear structural relationships by the method of maximumlikelihood. Chicago: National Educational Resources.

Josselson, R. (1988). The embedded self: I are thou revisited. InD. L. Lapsly & F. C. Power (Eds.), Self, ego, and identity:Integrative approaches (pp. 91-108). New York: Springer.

Kear, J. S. (1978). Marital attraction and satisfaction as a functionof differentiation of self Unpublished doctoral dissertation,California School of Professional Psychology, Fresno.

Kerr, M. E. (1985). Obstacles to differentiation of self. In A. S.Gurman (Ed.), Casebook of marital therapy (pp. 111-153). NewYork: Guilford Press.

Kerr, M. E. (1988, September). Chronic anxiety and defining a self.Atlantic Monthly, 9, 35-58.

Kerr, M. E., & Bowen, M. (1988). Family evaluation. New York:Norton.

Levine, J. B., Green, C. J., & Millon, T. (1986). Separation-Individuation Test of Adolescence. Journal of Personality Assess-ment, 50, 123-137.

Loevinger, J. (1957). Objective tests as instruments of psychologi-cal theory. Psychological Reports, 3, 635-694.

Luepnitz, D. A. (1988). The family interpreted. New York: BasicBooks.

MacCallum, R. (1986). Specification searches in covariance struc-ture modeling. Psychological Bulletin, 100, 107-120.

Mahler, M., Pine, F., & Bergman, A. (1975). The psychologicalbirth of the human infant: Symbiosis and individuation. NewYork: Basic Books.

Marsh, H. W., & Hocevar, D. (1985). Application of confirmatoryfactor analysis to the study of self-concept: First and higher orderfactor models and their invariance across groups. PsychologicalBulletin, 97, 562-582.

McCollum, E. E. (1986). Bowen's concept of emotional connected-ness to spouse and family of origin as a moderator ofthe relationship between stress and individual well-being.Unpublished doctoral dissertation, Kansas State University,Manhattan.

McCollum, E. E. (1991). A scale to measure Bowen's concept ofemotional cutoff. Contemporary Family Therapy, 13, 247-254.

Miller, J. B. (1976). Toward a new psychology of women. Boston:Beacon Press.

Minuchin, S., & Nichols, M. P. (1993). Family healing: Tales ofhope and renewal from family therapy. New York: Free Press.

Nichols, M. P. (1984). Family therapy: Concepts and methods (1sted.). Boston: Allyn & Bacon.

Nichols, M. P., & Schwartz, R. C. (1998). Family therapy:Concepts and methods (4th ed.). Boston: Allyn & Bacon.

Niedermeier, C , Handal, P. J., Brown, N. Y, Searight, H. R., &Manley, C. M. (1992, August). Psychological distress andfamilyfunctioning: Relationship in adolescent psychiatric inpatients.Paper presented at the 100th Annual Convention of the AmericanPsychological Association, Washington, DC.

Nunnally, J. C. (1978). Psychometric theory. New York: McGraw-Hill.

Olver, R. R., Aries, E., & Batgos, J. (1990). Self-other differentia-tion and the mother-child relationship: The effects of sex andbirth order. Journal of Genetic Psychology, 150, 311-321.

Papero, D. V. (1990). Bowen family systems theory. Boston: Allyn& Bacon.

Rickels, K., Lipman, R. S., Park, L. C , Covi, L., Uhlenhuth, E. H.,& Mock, J. E. (1971). Drug, doctor warmth, and clinic setting inthe symptomatic responses to minor tranquilisers. Psychophar-macologica, 20, 128-152.

Robinson, J. P., & Shaver, P. R. (1973). Measures of socialpsychological attitudes. Ann Arbor, MI: University of Michigan,Institute for Social Research.

Spanier, G. B. (1976). Measuring dyadic adjustment: New scalesfor assessing the quality of marriage and similar dyads. Journalof Marriage and the Family, 47, 15-28.

Spanier, G. B. (1988). Assessing the strengths of the DyadicAdjustment Scale. Journal of Family Psychology, 2, 92-94.

Spielberger, C. D., Gorsuch, R. L., & Lushene, R. E. (1970).Manual for the State-Trait Anxiety Inventory. Palo Alto, CA:Consulting Psychologists Press.

Szapocznik, J., Kurtines, W. M., Foote, F. H., Perez-Vidal, A., &Hervis, O. (1983). Conjoint versus one-person family therapy:Some evidence for the effectiveness of conducting familytherapy through one person. Journal of Consulting and ClinicalPsychology, 51, 889-899.

{Appendix follows)

246 SKOWRON AND FRIEDLANDER

Appendix

Differentiation of Self Inventory

These are questions concerning your thoughts and feelings about yourself and relationships with others. Please read each statement carefully and decidehow much the statement is generally true of you on a 1 (nor at all) to 6 (very) scale. If you believe that an item does not pertain to you (e.g., you are notcurrently married or in a committed relationship, or one or both of your parents are deceased), please answer the item according to your best guess aboutwhat your thoughts and feelings would be in that situation. Be sure to answer every item and try to be as honest and accurate as possible in your responses.

Not at all Very truetrue of me of me

1. People have remarked that I'm overly emotional. 1 2 3 4 5 62. I have difficulty expressing my feelings to people I care for. 1 2 3 4 5 63. I often feel inhibited around my family. I 2 3 4 5 64. I tend to remain pretty calm even under stress. 1 2 3 4 5 65. I'm likely to smooth over or settle conflicts between two people whom I care about. 1 2 3 4 5 66. When someone close to me disappoints me, I withdraw from him or her for a time. 1 2 3 4 5 67. No matter what happens in my life, I know that I'll never lose my sense of who I am. 1 2 3 4 5 68. I tend to distance myself when people get too close to me. 1 2 3 4 5 69. It has been said (or could be said) of me that I am still very attached to my parent(s). 1 2 3 4 5 6

10. I wish that I weren't so emotional. 1 2 3 4 5 611. I usually do not change my behavior simply to please another person. 1 2 3 4 5 612. My spouse or partner could not tolerate it if I were to express to him or her my true feelings

about some things. 1 2 3 4 5 613. Whenever there is a problem in my relationship, I'm anxious to get it settled right away. 1 2 3 4 5 614. At times my feelings get the best of me and I have trouble thinking clearly. 1 2 3 4 5 615. When I am having an argument with someone, I can separate my thoughts about the issue

from my feelings about the person. 1 2 3 4 5 616. I'm often uncomfortable when people get too close to me. 1 2 3 4 5 617. It's important for me to keep in touch with my parents regularly. 1 2 3 4 5 618. At times, I feel as if I'm riding an emotional roller coaster. 1 2 3 4 5 619. There's no point in getting upset about things I cannot change. 1 2 3 4 5 620. I'm concerned about losing my independence in intimate relationships. 1 2 3 4 5 621. I'm overly sensitive to criticism. 1 2 3 4 5 622. When my spouse or partner is away for too long, I feel like I am missing a part of me. 1 2 3 4 5 623. I'm fairly self-accepting. I 2 3 4 5 624. I often feel that my spouse or partner wants too much from me. 1 2 3 4 5 625. I try to live up to my parents' expectations. 1 2 3 4 5 626. If I have had an argument with my spouse or partner, I tend to think about it all day. 1 2 3 4 5 627. I am able to say no to odiers even when I feel pressured by them. 1 2 3 4 5 628. When one of my relationships becomes very intense, I feel the urge to run away from it, 1 2 3 4 5 629. Arguments with my parent(s) or sibling(s) can still make me feel awful. 1 2 3 4 5 630. If someone is upset with me, I can't seem to let it go easily. 1 2 3 4 5 631. I'm less concerned that others approve of me than I am about doing what I think is right. 1 2 3 4 5 632. I would never consider turning to any of my family members for emotional support. 1 2 3 4 5 633. Ifindmyselfthinkingalotaboutmy relationship with my spouse or partner. 1 2 3 4 5 634. I'm very sensitive to being hurt by others. 1 2 3 4 5 635. My self-esteem really depends on how others think of me. 1 2 3 4 5 636. When I'm with my spouse or partner, I often feel smothered. 1 2 3 4 5 637. I worry about people close to me getting sick, hurt, or upset. 1 2 3 4 5 638. I often wonder about the kind of impression I create. 1 2 3 4 5 639. When things go wrong, talking about them usually makes it worse. 1 2 3 4 5 640. I feel things more intensely than others do. 1 2 3 4 5 641. I usually do what I believe is right regardless of what others say. 1 2 3 4 5 642. Our relationship might be better if my spouse or partner would give me the space I need. 1 2 3 4 5 643. I tend to feel pretty stable under stress. 1 2 3 4 5 6

Differentiation of Self Inventory Subscale Composition (underlined means reverse scored):Emotional Reactivity: J_, 6, _10,14, _18, 21, 26, 30, 34, 38, 40I Position: 4, 7,11, 15, 19,23,27,31,35,41,43Emotional Cutoff: 2, 3, 8, .12,16, 20,24, 28, 32,36,39,42Fusion With Others: 5,9, 13, 17,22,25,29, 33, 37

Received September 22, 1997Revision received February 16, 1998

Accepted February 16,1998 •