Journal of Counseling Psychology1982, Vol. 29, No. 5,468-477

Copyright 1982 by the American Psychological Association, Inc.0022-0167/82/2905-0468$00.75

Resolving Decisional Conflict by Gestalt Two-Chair Dialogue:Relating Process to Outcome

Leslie S. Greenberg and Michael C. WebsterUniversity of British Columbia, Vancouver, British Columbia, Canada

Thirty-one clients completed a six-week program using Gestalt two-chair dia-logue to work on intrapsychic conflict related to the making of a decision. Theclients were classified as resolvers or nonresolvers based on a pattern of in-ses-sion process indicators. The resolvers were identified as clients who had man-ifested three components of a proposed model of conflict resolution:. the ex-pression of criticism by one part of the personality, the expression of feelingand wants by another, and the softening in attitude of the "critic." These at-tributes were measured in terms of voice quality, depth of experience, andstructural analysis of social behavior. Resolvers were found to be significantlyless undecided and less anxious after treatment; they also reported greater im-provement on target complaints and behavior change. In addition, after thesession in which the "critic" softened, resolvers reported greater conflict reso-lution, less discomfort, greater mood change, and greater goal attainment thannonresolvers.

The centrality of decision making to thecounseling process has recently been de-scribed by a number of authors (Horan,1979; Ivey, 1980) and is seen by many as aprimary goal of counseling (Ivey 1980;Schlossberg, Troll, & Leibowitz 1978). Re-cently Janis and Mann (1977) highlightedthe importance of decisional conflict inhuman decision making. They describeddecisional conflict as "occurring when thereare simultaneous opposing tendencies withinthe individual to accept and reject a givencourse of action" (p. 42) and stressed thatpeople often avoided making decisions be-cause of the pain experienced in dealing withdecisional conflict.

Greenberg (1979), describing splits as themost appropriate context for Gestalt two-chair dialogue interventions, describedconflict splits as client statements made incounseling that indicate a struggle betweentwo aspects of the self that are in opposition.This description is highly similar to, al-though broader than, Janis and Mann's de-

This study was supported by Social Sciences andHumanities Research Council of Canada Grant 410-80-0210.

Requests for reprints should be sent to Leslie S.Greenberg, Department of Counselling Psychology,University of British Columbia, Vancouver, BritishColumbia, Canada V6T 1Z5.

scription of decisional conflict. Little hasbeen written in Gestalt therapy on decisionmaking per se, but from the gestalt per-spective, situations in which people en-counter blocks in carrying out tasks that theyhave set for themselves are conflict situa-tions in which the conflict is between twoparts of the personality.

This situation of conflict between parts ofthe personality comes vividly into play in theprocess of decision making and choice. In-spection of Tiedeman and O'Hara's (1963)two-stage model of decision making revealsthat conflict plays an important role in theanticipatory stages of decision making andcan impede completion of the crystallization,choice, and clarification substages. In thecrystallization substage, in which alterna-tives are evaluated, conflict can prevent theperson from choosing one of the alternatives,whereas in the clarification substage, con-flicts and doubts related to the newly madechoice must be resolved before decisionmaking can move to the implementationstage. Counselors who are able to promotethe integration of conflicting aspects ofclients' experiences in order to relieve theunderlying struggle in these stages can helpthem crystallize their choices, clarify theirdoubts, and pass from the anticipatory to theimplementation stage of decision making.

A method shown to be effective in the

468

RESOLVING DECISIONAL CONFLICT 469

reconciliation of opposites in the personalityis the Gestalt two-chair operation(Greenberg, 1979; Greenberg & Clarke, 1979;Greenberg & Dompierre, 1981; Greenberg &Rice, 1981). This procedure facilitates adialogue between the two sides in order tobring the opposing parts into contact.Resolution is achieved through a dialecticalprocess that integrates the opposing partsinto a new set of responses to the situation(Perls, 1970).

Task analyses of two-chair dialogue con-flict resolution performances have resultedin the construction of a model of successfulperformance (Rice & Greenberg, 1982) andthe delineation of some of the essentialcomponents of these performances(Greenberg 1982a, 1982b, Greenberg, inpress). Initial intensive analysis(Greenberg, 1980) revealed that conflictresolution occurred when the previouslyharsh critical side in the dialogue, referredto as the "other chair," softened its stancetoward the side that expressed the experi-encing self. This softening was shown by adeepening in the level of experiencing of thecritic, a change in its voice quality from a"lecturing at" to an "internally focused"voice, and an increase in its affiliative be-havior. This model of the critic becomingmore accepting of previously disavowedfeelings has been shown in previous studiesto discriminate between successful and un-successful conflict resolution performances(Greenberg, in press).

Further intensive analyses (Greenberg1982b) that used the methods of task anal-ysis produced a model of intrapsychic con-flict resolution in two-chair dialogue com-posed of six primary resolution components.The harsh critical side in successful di-alogues was found to go through a sequenceof blaming the self, then expressing itsstandards and values, and finally softeningits attitude towards the self. The experi-encing self in all successful performances wasfound to express its feelings and to followthese with an expression of wants. Thepresence of these components almost inevi-tably led to the final component of successfulperformance, a negotiation between the twosides that resulted in an integration of thetwo sides.

For'the purpose of this study, in order to

relate process to outcome, three essentialcomponents- of resolution were constructedfrom the above six-component model. Thefirst component regarded as essential toresolution is criticism in the other chair.This component integrates blaming andstandards and values of the above model andensures that the client has truly engaged inthe dialogue by expressing the opposition ofone side to the other. The second essentialcomponent is the expression of felt wants inthe experiencing chair. This componentintegrates the feelings and wants compo-nents of the above model into a-componentthat requires that the client express, fromthe experiencing chair, wants in the contextof feeling. This ensures that the client isdeeply experiencing this side of the polarity.The final essential component is a softeningin the other chair in which the previouslyharsh critic softens in its attitude toward theexperiencing part.

It was the purpose of this study to relatethe process of intrapersonal conflict resolu-tion in resolving decisional conflicts to de-cision making outcome. This was done bycomparing treatment outcome effects of in-session resolvers and nonresolvers. Re-solvers were defined as those clients whomanifested all three of the essential com-ponents in the sessions. Specifically, it washypothesized that the occurrence of thethree critical components of criticism, felt-wants, and softening in two-chair dialoguesessions would relate to treatment effects.These effects were measured in the form ofthe immediate effects of the session in whichthe final resolution component, softening,was expressed; the prolonged effects of thisresolution session; and the final treatmenteffects at termination and follow-up.

Method

Subjects

The clients were urban adults, 31 women and 5 men,ranging in age from 17 to 65 years (M = 34.72, SD =12.38) seeking counseling for their difficulties in makinga decision. The sample was fairly well educated; of 36participants, 34 had graduated from high school and 19had some university education. Acceptance was basedon Malan's (1976) criteria for brief analytic treatment.Only persons who were relatively well functioning andwere experiencing a decisional conflict were acceptedfor the treatment. Six counselors (three men and three

470 LESLIE S. GREENBERG AND MICHAEL C. WEBSTER

women) were used in the study. All had a minimum of100 hours of training in the Carkhuff/Egan model ofcounseling. In addition all had at least 96 hours oftraining in the gestalt two-chair method. The coun-selors' experience with the method ranged from 2 to 7years.

Measures

Process instruments. The in-session level, of expe-riencing in the dialogue was measured on the Experi-encing Scale (ES) (Klein, Mathieu, Kiesler, & Gendlin,1969). This scale has been shown to be a highly reliablemeasure of client involvement or "experiencing" incounseling. As one moves up the scale there is a pro-gression from limited, impersonal references to self toan inner descripton of feelings to a synthesis of newlyemerged feelings to solve problems.

The degree of affiliation and interdependence in thesides of the conflict was measured by Benjamin's (1979)circumplex model, called the Structural Analysis ofSocial Behavior (SASB). This system has shown highinterjudge reliabilities and has been extensively vali-dated. .In this system, each statement is characterizedas belonging to one of 36 categories that belong to oneof four quadrants on one of two grids.x Affiliation(measured along the horizontal axis) intersects withinterdependence (measured along the vertical axis),forming the four quadrants of the circumplex. The gridis determined by whether the focus of the statement ison the speaker or the receiver of the statement.

The Client Vocal Quality system (CVQ; Rice, Koke,Greenberg, & Wagstaff, 1979) was used to measure thevocal quality of the statements from each chair. Thissystem contains four mutually exclusive voice patterns:focused, emotional, externalized, and limited. Focusedvoice indicates inward deployment of attention. TheCVQ has been shown to possess satisfactory interjudgeagreement, and focused voice has been shown to be anindex of good process in client-centered counseling.

Raters. Two raters and one judge, all unaware of thehypothesis to be tested, rated the process data. Theraters and the judge had 4 hours of training on theconflict split resolution model, 16 hours of training onthe ES, 16 hours of training on the CVQ, and 13 hoursof training on the SASB. Interrater reliability at theend of training yielded a Pearson product-momentcorrelation of .86 for experiencing, and Cohen's Kappasof .89 on SASB and .50 on CVQ. Both Kappas wereshown to be statistically significant. In addition theraters were checked on their identification of the twochairs and showed a 100% agreement on chair identifi-cation.

Relationship instruments. The task subscale of theWorking Alliance Inventory (WAI; Horvath, 1981,Horvath & Greenberg, in press), based on Bordin's(1979) ideas on the components of the working alliance,was used to establish if clients perceived the two-chairprocess as relevant to their goals. In addition, onesubscale from Barrett-Lennard's (1962) RelationshipInventory (BLRI) was used to ensure that clients per-ceived their counselor as empathic.

Treatment outcome instruments. The Scale ofVocational Indecision (Osipow, Carney, & Barak, 1976)was adapted to suit the purposes of the present study.1

The scale consists of eighteen items to be rated on a4-point scale from 1 (not at all like me), to 4 (exactly likeme). A high score reflects indecision; a low score re-flects decidedness. Since most of the items on the scalewere particular to educational and/or vocational inde-cision some revision of wording was necessary to makeit appropriate for this study. The essential meaning ofeach item was maintained, however.

In order to verify that these changes had not ad-versely affected the validity and reliability of the scale,the adapted form was administered to 136 graduatestudents in a pilot study. The students responded tothe questionnaire in terms of a current decision. AHoyt's analysis of variance (ANOVA) internal consis-tency reliability of ,85 indicated that the revised itemswere consistent with the other items in the test.

The State form of the State-Trait Anxiety Inventory(STAI) was used to measure state anxiety. This formhas been shown to be a valid and reliable index of cur-rent anxiety. It contains 20 brief items which assesshow "you feel right now, that is at this moment" and hasbeen shown to measure change in anxiety (Spielberger,Gorsuch, & Lushene, 1970). It has been shown thatdecisional conflict and anxiety are related and it wasassumed that resolving decisional conflict would leadto anxiety reduction (Janis & Mann, 1977).

The Target Complaints (TC) measure (Battle, et al.,1966), recommended in Waskow & Parloff (1975) as acore battery instrument for use in outcome research,consists of five-point scales on which the client is askedto rate the amount of change on up to three differentcomplaints selected by the client. The scales rangefrom ".worse" through "a little better" to "a lot better."The validity of this measure has been shown by its highcorrelations with four other outcome measures whilepre- and postsession severity and content ratings havebeen shown to be highly reliable (Battle et al., 1966).

An intensive personalized measure of symptom se-verity based on the work of Shapiro (1961) and extendedby Phillips (1970) was used to follow, session by session,each client's conflict-related symptom. This instru-ment, the Phillips Personalized Questionnaire (PPQ),elicits a symptom statement from each client, and a setof five statements representing different levels of thestatement are drawn up. On each administration of theindividualized instrument, the client compares thepresent intensity of his or her symptom with each levelof the scale. This provides for a consistency measureon each of the individualized measures.

A Behavioral Report (BR) questionnaire was devisedfor this study to assess the degree to which a decisionhad been implemented. The questionnaire was basedon Tiedeman and O'Hara's (1963) decision makingmodel, and consisted of six sets of questions adminis-tered in a follow-up interview. Each set of questionswas tied to one of the last six substages of the Tiedemanand O'Hara model: exploration, crystallization, choice,clarification, induction, reformation, or integration.The interviewer scored each stage (either 1 for yes, or0 for no), interpreting the clients' answers to questions

1 This scale is available from the first author at theDepartment of Counselling Psychology, University ofBritish Columbia, Vancouver, British Columbia, Can-ada V6T 1Z5.

RESOLVING DECISIONAL CONFLICT 471

to assess if they had attained that stage. This measurepossesses only face validity.

Session outcome instruments. Session outcomeinstruments fall into two categories, immediate effectsinstruments and prolonged effects instruments.

• Immediate effects instruments. The ConflictResolution Scale (CRS) consisted of a seven-point boxscale on which-clients indicated their feelings of reso-lution regarding the issue they identified as their con-flict in the session. The first box (starting from thebottom) is labeled "not at all resolved"; the seventh box,"totally resolved." The instrument has been shown tosuccessfuly discriminate between more and less resolvedsessions in a study, comparing the effects of two chairdialogue and empathic reflections on conflict resolution(Greenberg & Dompierre, 1981). The CRS was usedin this study to track clients' feelings of resolution afterevery session.

Clients completed the Target Complaints DiscomfortBox Scale (TCDBS) before and after each session. Theinitial discomfort rating was made on a column dividedinto thirteen boxes. The words "not at all" wereprinted beside the bottom box; "a little" by the fourthbox from the bottom; "pretty much" by the seventh box;and "couldn't be worse" by the top (or thirteenth) box.The scale has shown satisfactory pre-post session reli-ability (Battle et al., 1966),

Epstein's Prevailing Mood Scale (EPMS; Epstein,1979) was used to measure mood. Clients indicatewhich side of the nine-point scale best describes theirfeeling "right now, this very moment." Each side of thescale is identified by an opposing cluster of two or threeadjectives that were determined by factor analyses ofadjective checklists (Epstein, 1976). Satisfactorytemporal reliabilities over a 7-day period have beenreported (Epstein, 1979). In this study, attention wasfocused on three of the scales of the EPMS that wereperceived to be theoretically linked to intrapsychicconflict. These scales were (a) Pleased With Self-Ashamed (PWS), (b) Integrated-Disorganized (INT),and (c) Powerful-Weak (POW). It was assumed thatafter resolution, clients would feel more pleased withand accepting of themselves and would feel more inte-grated and more powerful.

• Prolonged effects instruments. The Goal At-tainment Scale (GAS) is a measure that facilitates goalsetting and goal measurement and has been shown tobe a reliable and valid instrument (Kiresuk & Sherman,1968). It is a method for obtaining specific observablegoals and for providing a common measure for indi-vidualized goals. The goal outcomes range from muchworse than expected (—2) through expected (0) to muchbetter than expected (+2), with each outcome leveldefined by an observable behavior.

The Post Resolution Session Questionnaire2 (PCSQ)is an instrument constructed by the authors for thisstudy to measure attitude change related to conflict.The instrument contained five questions asking theclient to identify the "core issue" and related attitudesthat he or she had worked on in the session. The in- .strument also obtained later reports on whether thespecified attitudes around this issue had changed as aresult of the session. Two five-point Likert scales onthe instrument were used to derive the PCSQ attitudechange score. These scales ranged from "no change"through "somewhat" to "very greatly changed." The

two scales showed a high correlation and were groupedtogether to form a single score. This instrument pos-sesses only face validity.

Procedure

Announcements in the local media described a brieffree program for people experiencing difficulty inmaking a decision. In an induction session the clientswere briefly introduced to the Gestalt theory of in-trapsychic conflict (as opposition between parts of thepersonality) and the two-chair technique; they alsoengaged in an exercise designed to create an awarenessof each client's own "top dog" and "underdog." Thiswas done to facilitate ease of entry into treatment.

Instrumentation, At the end of the induction ses-sion the clients constructed the Target Complaintsmeasure and the Phillips Personalized Questionnaireand completed the first administration of the latter.They were given the Scale of Indecision and theState-Trait Anxiety Inventory to take home with in-structions to complete both instruments the night be-fore their first counseling session. The repeated mea-sure, the Phillips Personalized Questionnaire, was ad-ministered before each session and at termination andfollow-up. The Conflict Resolution Scale, the TargetComplaints Discomfort Box Scale, and Epstein's Pre-vailing Mood Scale were administered in that orderbefore and after each session and at termination andfollow-up.

The resolution session test administrations differedfrom a regular interview by the addition after the ses-sion of the Post Critical Session Questionnaire and theGoal Attainment Scale. Clients were also given sixcopies of Epstein's Prevailing Mood Scale with the in-struction to complete one at the same time each dayuntil the next session (termination) one week away.

At termination and follow-up, clients completed thesix session instruments already mentioned plus thethree treatment outcome measures, Target Complaints,the Scale of Indecision, and the State-Trait AnxietyInventory. During the one-month follow-up, clientscompleted the Behavior Report Questionnaire inter-view.

At the conclusion of the second session the task andempathy scales from the WAI and RI were given to theclients who filled them out at home and returned them,sealed in the provided envelope, at the next session.The data generated by these instruments would permitthe, division of the clients into engagers and nonen-gagers, so that any differences between resolvers andnonresolvers could not be attributed to the nonen-gagement of the nonresolvers. All of the 31 clients whocompleted the project were, however, identified as en-gagers due to their attainment of a criterion score of 36on perceived task relevance and 16 on perceived em-pathy. The 5 female clients who dropped out were lowon perceived task relevance scores.

Determining the resolution session. The clients

2 This scale is available from the first author at theDepartment of Counselling Psychology, University ofBritish Columbia, Vancouver, British Columbia, Can-ada V6T1Z5.

472 LESLIE S. GREENBERG AND MICHAEL C. WEBSTER

were seen once a week for six weeks or until they re-solved their conflict (whichever came first). All coun-seling sessions were audiotaped. When in the coun-selor's judgement the client had resolved (that is, thethree essential elements had occurred: an expressionof criticism, felt wants, and a softening), the session inwhich softening occurred was defined as the counselorjudged resolution session. If a client failed to manifestthe essential elements, the fifth session was declared theresolution control session. At these resolution sessionsboth those who had manifested the resolution elementsand those who had not were reminded that the nextsession was the termination session. Fifteen clientswere identified by the counselors as "softening." Themodal number of sessions for resolution was 4, with arange of 1. All clients reported one month after thetermination session for a follow-up interview.

Identification of resolvers. In order to be classifiedas a resolver, clients had to manifest the three compo-nents: criticism, felt wants, and softening. Resolverswere determined by the following procedure devised forthis study:

• Counselor identified components. The tapes ofpotential resolution sessions were identified by havingthe counselor indicate on a questionnaire, following eachsession, the occurrence of any resolution components.At the end of treatment, the counselors were asked toreturn to the session audiotapes marked as containingthe most advanced component tfnd to indicate the bestexample of each of the components that had occurredduring the hour. The experimenter used the indicatedcomponent as a midpoint for an 8-minute segment.That is, the segment was composed of the 4 minutesbefore the indicated component plus the following 4minutes including the component. This 8-minutesegment was regarded as the critical episode and waspresented to the two trained raters. It was possible forone, two, or three of the essential components to bepresent in a single episode. If only softening waspresent and the other two elements necessary to qualifythe segment as a resolution episode were not present ina single segment, the necessary number of precedingcritical episodes (tape segments) were presented to theraters to establish if the other two components werepresent. The critical episodes were presented randomlyto the raters.

To identify the possible presence of a softening in thenonresolvers, the experimenter selected for each clientthe session that the counselor had indicated containedthe most advanced component. In addition to thecounselor-indicated session, the experimenter selectedthe session in which the client reported being most re-solved on the Conflict Resolution Scale. These tapeswere broken into 8-miiiute segments using the begin-ning of the two-chair dialogue as an entry point. Thesesegments were presented in random order to the ratersto determine if a possible softening had occurred in anyof the nonresolvers.

The remaining unused tapes from the nonresolversand several sessions from resolvers already rated ascontaining a softening component were sent to a judge(who was trained in the model and unaware of the hy-potheses of the study) to be scanned for the occurrenceof a softening component in the nonresolvers that mighthave been overlooked by the counselors. If any addi-tional softening components were identified—which did

not occur—the tape in question would have been seg-mented in the manner described above and sent to theraters.

• Rater identification of resolvers. The raters firstdecided which "chair" was speaking in the dialogue andthen rated each segment for softening and/or criticismsif it was the other chair and felt wants if it was the ex-periencing chair. The basic unit for these ratings wasthe client statement. The ratings of the essentialcomponents were based on ratings of the experiencing,voice quality, and structural analysis of social behaviorof each chair. In the case of any disagreement betweenraters the segment was given to the judge.

Using the above procedure, the session was deemeda resolution session if it contained a softening that hadbeen preceded in that session or previous ones by theother two components. Two of the counselor-deter-mined resolvers were found not to meet the criteria.Thirteen resolvers and eighteen nonresolvers wereidentified by the raters. The following criteria (basedon our knowledge of the nature of the phenomenon fromthe intensive analysis; Greenberg, 1982b) were used foridentifying the components:

Softening1. At least two consecutive client statements from the

other chair had to be rated as affiliative on SASB,that is, in Quadrants 1 or 4. (A client statementwas defined as everything occurring between twocounselor statements.)

2. At least one client statement in focused voice hadto be coincident with the above SASB rating (i.e.,one of the two consecutive SASB statements).

3. At least one statement in the other chair must havereached Level 4 or above on the Experiencing Scale:"Description of feelings and! personal experiences"(Klein et al, 1969, p. 64). Level 4 was used ratherthan Level 5, as in previous studies, because of thesmaller unit (i.e., one statement).

Only if all three criteria were satisfied was the episodedefined as an instance of softening. A comparison be-tween the two raters on their overall ratings was per-formed. There was agreement on 12 of 13 of the cases.A third person, the judge, was called in to break thedisagreement in favor of a 13th instance of softening.

Felt Wants1. At least one client statement in the experiencing

chair on SASB had to be rated in Categories 243(Ask, trust, count on); 217 (Assert on own); or 216(Put cards on table).

2. At least one client statement had to have reachedLevel 4 on the Experiencing Scale. A secondaryjudgment was made on whether there was only oneor more than one statement at Level 4. Onestatement at Level 4 satisfied the criterion, andmore than one provided a stronger statement ofexperiencing.

3. At least one client statement coincident with theSASB statement had to be in focused voice.

4. The client must have met the semantic criterion ofexpressing directly to the other chair a desire, for

RESOLVING DECISIONAL CONFLICT 473

example, "I want.,.", "I need...", or "I'd like..."or some other equivalent statement of want.

Criticisms1. At least two client statements in the other chair had

to be rated as hostile, that is, in either Quadrant 2or 3 of the Other grid on the SASB.

2, At least two client statements had-to be rated asbeing in external voice on Client Vocal Quality.

The identification process yielded 158 8-minutesegments. The raters independently rated two-thirdsof the total number of segments providing a one-thirdoverlap for a reliability check. These overlappingsegments were designed to contain all the counselor-indicated softenings. To determine the extent ofagreement between the raters, Cohen's Kappa (Cohen,1960), a procedure that yields a coefficient of agreementfor nominal scales, was used. The coefficient ofagreement for these data was .925, indicating very highagreement between raters on the identification of theresolution components. This suggested that the criteriawere clear, independent, and mutually exclusive withrespect to the defining of resolvers and nonresolvers,

Results

Due to the unequal sample size of 13 re-solvers and 18 nonresolvers, tests for homo-geneity of variance were conducted using theBartlett-Box Homogeneity of DispersionTest (Kirk, 1968) on the Scale of Indecision,A-State, Target Complaints and Goal At-tainment. The failure to attain statisticalsignificance between the two groups on theScale of Indecision and A State led to theacceptance of the assumption of homoge-neity of variance and made only these twomeasures eligible for analysis by parametricstatistical procedures. The data from thesetwo measures were tested for equality ofvariance-covariance matrices in tests sug-gested by Box (1950). The results revealedthat the data on both measures exhibitedmatrices equal for both groups and fulfilledthe symmetry conditions thus qualifying forunivariate statistical procedures. The re-maining measures, because of the ordinaland quasi-interval nature of their scales,were deemed appropriate for nonparametricanalysis. The hypothesis of homogeneity ofvariance on these measures was checked onHartley's Fmax and rejected for all.

Treatment Outcomes

The Scale of Indecision and A-State scoreswere submitted to a two-way analysis of

variance (ANOVA) with repeated measureson one factor. Group membership wasviewed as a two-level fixed factor, and mea-surement occasions were the repeated mea-sure. Analysis revealed significant GroupX Occasion interaction effects for the Scaleof Indecision data, F(2, 58) = 10.489 p <.001, and for the A-State data, F(2, 58) =9.399, p < .001. The simple main effectswere therefore examined by constructingBonferroni confidence intervals around themean differences of the two groups at eachoccasion. The two groups showed no dif-ferences on Scale of Indecision or A-StateScores on the pretest but were significantlydifferent at the .05 level on the Scale of In-decision at termination and follow-up and onA-State at follow-up. The means andstandard deviations of the groups on Scaleof Indecision and A-State are shown in Table1. Inspection of the means indicated a re-duction of Indecision and State Anxiety forthe Resolvers.

The Target Complaint and Reported Be-havior Change scores were analyzed using aWilcoxon rank sum nondirectional test.The groups were found to be significantlydifferent on target complaints improvementat termination (Z = 2.336 p < .05), and fol-low-up (Z = 2.88, p < .01) and on reportedbehavior change (Z = 4.67, p < .01). In-spection of the means indicated higher scoresfor the resolvers. Means and standard de-viations are shown in Table 1.'

Symptom Reduction

The repeated measures on the PhillipsPersonalized Questionnaire of each clientwere analyzed using a two-way ANOVA toestablish an estimate of reliability (Phillips,1970). Reliable symptom measures wereobtained for only 12 resolvers and 14 non-resolvers. Trend analyses were then per-formed on the symptom data for each of theclients. Of the 12 resolvers, 11 had a sig-nificant positive linear trend (which indi-cated overall symptom improvement). Theresolver who failed to attain a significantpositive linear trend manifested significanceon a quadratic trend (which indicated apattern of improvement followed by wors-ening in the symptom). Of the 14 nonre-solvers 9 had a significant positive linear

474 LESLIE S. GREENBERG AND MICHAEL C. WEBSTER

Table 1Means and Standard Deviations for Resolvers and Nonresolvers on Scale of Indecision, StateAnxiety, Target Complaints, Behavioral Report, Goal Attainment, and Post Critical SessionAttitude

Indecision

Occasion M SD

State Anxiety

M SD

Target BehavioralComplaints Report

M SD M SD

Postcritical

Goal sessionAttainment attitude

M SD M SD

Resolvers

PretestTerminationFollow-up

36.6924.4624.38

4.976.025.09

46.6131.5429.08

10.258.255.94

4.694.85

.63

.38 5.61 .5155.38 8.77 9.31 .6360.77 6.4 9.38 .77

Nonresolvers

PretestTerminationFollow-up

39.3333.6133.83

7.187.407.25

42.5538.2239.78

7.3210.3111.22

3.963.94

.94

.94 2.78 1.2247.22 4.61 7.28 1.6749.44 7.25 7.17 1.38

trend. Three clients manifested significantquadratic trends, and 2 clients producedsignificant negative linear trends (an overallworsening of the symptom). Fisher's test ofexact proportions showed that the twogroups did not differ significantly (p = .10)on the proportion of resolvers and nonre-solvers who showed a positive trend. It was

concluded therefore that the groups were notsignificantly different on symptom reliefover the length of the study as assessed onthe PPQ.

Resolution Session Measures

Wilcoxon rank sum tests (two-tailed) were

Table 2Means and Standard Deviations for Resolvers and Nonresolvers Over the Resolution Session onConflict Resolution, Target Complaints, Discomfort, and Mood Indicators

ConflictResolution

Occasion M SD

TargetComplaintsDiscomfortBox Scale

M SD

Pleasedwith self

M SD

Integrated

M SD

Powerful

M SD

Resolvers

Presession 4.91 1.5Postsession 6.46 .66Mean score

overfollowingweek

9.85 2.2711.92 1.11

6.467.61

6.95

1.51.77

1.15

6 1.637.85 .69

6.74 1.25

6.697.46

6.88

1.11.97

.964

Nonresolvers

Presession 5.0 1.61Postsession 4.94 1.73Mean score

overfollowingweek

9.22 2.8610.05 3.08

6.556.55

5.85

1.21.38

1.36

5.8 1.56.44 1.38

5.65 1.33

6.56.53

5.93

1.151.25

1.15

RESOLVING DECISIONAL CONFLICT 475

performed on the pretest and posttest reso-lution session scores. The groups were notsignificantly different on the presessionscores on the five measures. The groupswere found to be significantly different on allfive postsession measures (Z = 2.47, p < .05for CRS; Z - 1.97, p < .05 for TCDBS; 2 »2.19, p < .05 for PWS; £ - 3.04, p < .01 forINT; and 2 = 2.03, p < .05 for POW). Themeans and standard deviations are shown inTable 2. Inspection of the means indicatedthat differences were in the direction of im-proved scores of resolvers.

A Wilcoxon rank sum nondirectional testindicated that the groups were significantlydifferent at the .05 level on the mood scales,PWS, INT, and POW over the week fol-lowing the resolution session. An inspectionof the group's means on the three scales forthe week indicated that this difference wasin the direction of improved scores for theresolvers, reflecting increased self accep-tance, integration, and feelings of power.Means and standard deviations of the groupsare shown in Table 2. It was found (Wil-coxon test) that the groups were significantlydifferent at the .05 level on GAS andPCSQat the termination and follow-up sessions.An inspection of the groups' means at ter-mination and follow-up indicated that dif-ferences were in the direction of higherscores for resolvers, reflecting greater successat goal attainment and attitude change.Means and .standard deviations are shown inTable 1.

Discussion

The analyses revealed a consistent patternof statistically significant differences be-

• tween resolvers and nonresolvers. The re-solvers were significantly less undecided attermination and follow-up and significantlyless anxious at follow-up than the nonresol-vers. The resolvers, in addition, were sig-nificantly more improved on their targetcomplaints at termination and follow-up andon a report of behavior change at follow-up.Over the resolution session the resolvers re-vealed a significantly greater sense of conflictresolution, less target complaint discomfort,a greater sense of self acceptance, greaterintegration and greater feelings of power.The mood changes endured for the resolvers

during the week following the resolutionsession, and resolvers showed superior goalattainment and attitude change at bothtermination and follow-up.

In light of these findings it appears thatthe ability of the resolvers to resolve an in-trapsychic conflict in the session accordingto the specified process criteria was sugges-tive of differential outcome. The in-sessionprocess of criticizing the self from one sideof the conflict, expressing a desire embeddedin a feeling context from the opposite side,and expressing a softening in the attitude ofthe critic toward the self appears to be re-lated in this sample, to positive outcome inGestalt two-chair work.

It remains possible that some variableother than the in-session performance of theresolution components could explain therelationship between group membership andoutcome. However, even if a third variablesuch as an individual difference variable ora relationship factor accounted for a greaterdegree of the variance than the process in-dicators, it would be unlikely that the per-formance pattern in the two chairs was achance corelation that did not possess anyexplanatory power regarding the mecha-nisms of change (Rice & Greenberg, 1982) inthe process of Gestalt dialogue.

It appears that the nonspecific or generalbenefits of treatment experienced by thenonresolvers were enough to result in reliefof symptoms such as tension, anger, loss ofappetite, insomnia, and discouragement thatwere associated with indecision. Having theopportunity to experience the persecutionof their critic, and to realize they were doingthis to themselves, or to express their feelingsand wants to their critics (as did nine of thenonresolvers), and to do this in an acceptingand understanding atmosphere may havebeen enough to have some impact on thenonresolvers' symptoms, even though theyfailed to attain resolution. This repeatedmeasure instrument, however, is most suitedto the study of individual change over time;with the loss of data from five clients (due toinstability of the measures), the group pro-portion scores were not a good measure ofgroup differences.

Findings on the session measures sug-gested that for resolvers, softening was thekey component. When the critic softened,

476 LESLIE S. GREENBERG AND MICHAEL C. WEBSTER

the underlying sense of struggle was relieved,the person felt less discomfort, and felt moreself-accepting and in control. Intrapsychicconflict resolution in the session, in addition,appeared to promote the attainment of aspecific goal related to the resolution of theconflict. Nonresolvers, who also set a goalat the end of their fifth session as to whatbehavior they would like to change, did notattain their goal as well as the resolvers.

Inspection of the session content in con-junction with the process analysis suggeststhat the obstacle to making a decision andimplementing it was an underlying conflictbetween the person's standards and valuesand his or her needs and wants. When thesecame "face to face" and a process of reeval-uation and softening by the critic took place,the conflict was resolved, and the person wasfree to decide. It appears that individualswho possessed crystallized alternatives butwere unable to make a choice were able to doso after resolving an internal conflict. Ex-amples of this process are given below:

A middle-aged divorced woman with college-boundchildren was unable to decide whether to return to hernative country or not. Conflicts about her motheringrole appeared in the first session. On the one hand shewanted to "do something for herself" and on the pthershe felt "a good mother should always put her childrenabove herself." The resolution of this conflict camewhen her critic no longer made her feel guilty and wasmore understanding and accepting of her desire to "dosomething for me for a change, before I'm too old."Having resolved her conflict she decided to return to hernative country and at follow-up had spoken to herchildren about their options, put her house up for sale,and purchased a plane ticket.

Another client, a younger woman, was undecidedabout whether to return to university to complete herstudies or travel. The initial conflict revealed in thesessions was between a harsh critic, "You've wasted fouryears of your life," and her underdog response, "I can'tseem to get motivated." The client, having movedthrough feelings of hopelessness in response to thecriticisms of being a failure, began to fight for herself.She asserted that she had done a menial job because shehad not been ready for anything else after dropping outof university, but that now she was ready, and going toschool was what she wanted. The resolution of theconflict came when her critic changed from a perse-cuting stance to one in which support was expressed forher newly emerged sense of direction. She improvedon the measuring instruments from being initially un-decided, anxious, uncomfortable, and minimally selfaccepting to feeling good, and deciding to return toschool. At follow-up she was making preparations toreturn to school.

The results of this study, relating processto outcome, provided support for the validityof the intrapersonal conflict resolution per-formance model proposed by Greenberg(1982b, in press)—that in the context of agood working alliance, conflict resolutionperformances follow a particular path. Thispath is characterized by an initial expressionby one part of the personality of criticismtoward another part, followed by a deepen-ing of experiencing and an assertion of de-sires in the previously unaccepted part. Thefinal necessary component of resolutionperformance is a softening in attitude of thecritic. This results in a resolution of conflictthrough some form of integration of the twoaspects of the personality.

References

Barrett-Lennard, G. T. Dimensions of therapist re-sponse as causal factors in therapeutic change.Psychological Monographs, 1962, 76(43, Whole No.562).

Battle, C., et al. Target complaints as criteria of im-provement. American Journal of Psychotherapy,1966,20,184-192.

Benjamin, L. S. Use of structural analysis of socialbehavior (SASB) and Markov Chains to study dyadicinteractions. Journal of Abnormal Psychology,1979,88, 303-319.

Bordin, E. The generalizability of the psychoanalyticconcept of the working alliance. Psychotherapy:Theory, Research and Practice, 1979,16, 222-229.

Box, G. Problems in the analysis of growth and wearcurves. Biometrics, 1950,6, 362-389.

Cohen, J. A system of agreement for nominal scales.Journal of Educational and Psychological Mea-surement, 1960,20, 37-46.

Epstein, S. Anxiety, arousal and the self. In I. G.Sarason & C. D. Spielberger (Eds.), Stress and anx-iety (Vol. 3). Washington D.C.: Hemisphere,1976.

Epstein, S. The stability of behavior: 1. On pre-dicting most of the people much of the time. Journalof Personality and Social Psychology, 1979, 37,1097-1126.

Greenberg, L. S. A task analytic approach to the eventsof psychotherapy (Doctoral dissertation, York Uni-versity [Canada], 1976). Dissertation AbstractsInternational, 1977,66,4647B.

Greenberg, L. S. Resolving splits: The two-chairtechnique. Psychotherapy: Theory, Research andPractice, 1979,16, 310-318.

Greenberg, L. S. An intensive analysis of recurringevents from the practice of Gestalt therapy. Psy-chotherapy: Theory, Research and Practice, 1980,17,143-152.

Greenberg, L. Task analysis: The general approach.In L. Rice & L. Greenberg (Eds.), Patterns of change:

RESOLVING DECISIONAL CONFLICT 477

Intensive anaysis of psychotherapy process. NewYork, Guilford, 1982. (a)

Greenberg, L. Task analysis of conflict resolution. InL. Rice & L. Greenberg (Eds.), Patterns of change:Intensive anaysis of psychotherapy process. NewYork, Guilford, 1982. (b)

Greenberg, L. S. Toward a task analysis of conflictresolution. Psychotherapy: Theory, Research andPractice, in press.

Greenberg, L. S., & Clarke, K. The differential effectsof the two-chair experiment and empathic reflectionsat a conflict marker. Journal of Counseling Psy-chology, 1979,26,1-8.

Greenberg, L. S., & Dompierre, L. Differential effectsof Gestalt two-chair dialogue and empathic reflectionat a split in counseling. Journal of CounselingPsychology, 1981,24, 288-294.

Greenberg, L. S., & Rice, L. The specific effects of aGestalt intervention. Psychotherapy: Theory,Research and Practice, 1981,78,210-216.

Horan, J. Counseling for effective decision making.Monterey, Calif.: Brooks/Cole, 1979.

Horvath, A. An exploratory study of the concept oftherapeutic alliance and its measurement. Un-published doctoral dissertation. University ofBritish Columbia, 1981.

Horvath, A., & Greenberg, L. The development of theworking alliance inventory. In L. Greenberg & W.Pinsof (Eds.), Psychotherapeutic process: A re-,search handbook. New York: Guilford, in press.

Ivey, A. Counseling and psychotherapy. New York,John Wiley, 1980. .

Janis, I. L., & Mann, L, Decision making: A psycho-logical analysis of conflict, choice and commitment.New York: Free Press, 1977.

Kiresuk, J. T., & Sherman, R. E. Goal attainmentscaling: General method for evaluating compre-hensive community mental health programs. Com-munity Mental Health Journal, 1968,4,443-453.

Kirk, R. E. Experimental design: Procedures for thebehavioral sciences. Monterey, Calif.: Brooks/Cole,1968.

Klein, M., Mathieu, P., Keisler, D., & Gendlin, E. Theexperiencing scale. Madison, Wis.: WisconsinPsychiatric Institute, 1969.

Malan, D. H. Toward the validation of dynamic psy-chotherapy: A replication. New York: PlenumPress, 1976.

Osipow, S. H., Carney, C. G., & Barak, A. A scale ofeducational vocational undecidedness: A typologicalapproach. Journal of Vocational Behavior, 1976,9,233-243.

Perls, F, Four lectures. In J. Fagan & I. Shepherd(Eds.), Gestalt therapy now. Palo Alto Calif.:Science & Behavior Books, 1970.

Phillips, J. P. N. A new type of personal questionnairetechnique. British Journal of Social and ClinicalPsychology, 1970,9,241-256.

Rice, L., & Greenberg, L. Patterns of change: In-tensive analyses of psychotherapy process. NewYork, Guilford, 1982.

Rice, K., Koke, C., Greenberg, L., & Wagstaff, A.Manual for client voice quality. Toronto, Canada:York University Counselling and DevelopmentCentre, 1979.

Schlossberg, N., Troll, L., & Leibowitz, Z. Perspectiveson counseling adults: Issues and skills. Monterey,Calif,: Brooks/Cole, 1978.

Shapiro, M. B. A method of measuring changes specificto the individual psychiatric patient. British Jour-nal of Medical Psychology, 1961,34,151-155.

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

Tiedeman, D. V., & O'Hara, R. P. Career development:Choice and adjustment. New York: College En-trance Examination Board, 1963.

Waskow, I. E., & Parloff, M. B. (Eds.), Psychotherapychange measures. Rockville, Md.: National Insti-tute of Mental Health, 1975.

Received October 23,1981Revision received March 4,1982 •