family process volume 23 issue 1 1984 [doi 10.1111%2fj.1545-5300.1984.00075

9
Fam Proc 23:75-87, 1984 Patterns of Alliances in Nondistressed and Multiproblem Families RICHARD GILBERT, PH.D. a ANDREW CHRISTENSEN, PH.D. a GAYLA MARGOLIN, PH.D. b a Department of Psychology, University of California, Los Angeles, 90024. Reprint requests should be addressed to Dr. Christensen. b Department of Psychology, University of Southern California, Los Angeles. The goal of this study was empirically to investigate differences between distressed and nondistressed families in the strength, patterning, and cross-situational consistency of alliances. Audiotapes were made of 12 distressed and 12 nondistressed families in two observational situations. The resultant interaction was then coded for family alliances by trained observers. Distressed families were characterized by low overall levels of alliance behavior, weakness in the marital alliance relative to other family alliances, and discrepancies in parental alliances with the target child. There were generally no differences between groups in the cross-situational variability of alliance strength or patterning. The results are discussed in relation to the predictions of structural models of family process. The structural perspective on family systems (e.g. 19, 20, 24) has addressed marital and parent-child alliances in distressed and nondistressed families. Some statements have focused on differences in the overall level of alliance or interpersonal support across contrast groups. Concepts such as disengagement (19) or emotional divorce (24) emphasize a deficiency in emotional closeness and support in distressed family systems. Other theorists have focused on distinctions in the intrasystem patterning of alliances across groups. For example, Lidz and his colleagues (18) have maintained that many families with a schizophrenic child suffer from a pattern of "role distortion" termed family schism. Schismatic families are characterized by open hostility among spouses and attempts to force the child to take sides. Similarly, Minuchin and his colleagues (20) have implicated the presence of a weak marital coalition and extreme cross-generational or parent-child alliances as a dysfunctional structural pattern. Minuchin has discussed a number of implications of low levels of mutual support and skewed alliance patterns (19). An absence of support leads to negative affective states in individual family members. In addition, a system consisting of disengaged relationships will be less effective in adaptive processes such as negotiation, problem-solving, and conflict-resolution. This is especially true when the system members with the most knowledge and experience (i.e., the parents) are disengaged. With respect to skewed alliance patterns, extreme parent-child alliances are viewed as dysfunctional, in that (a) they provide the aligned parent with an alternate support system, thus reducing the motivation to address and resolve interspouse conflicts directly; (b) they place the child in a situation of divided loyalty, of having to choose between parents; and (c) they place the child in an undesirable position of power relative to the non-aligned spouse. Leibman, Honig, and Berger (15) have also discussed the dysfunctional aspects of extreme parent-child alliances. These authors note that the extreme parent-child alliance in distressed families is most often between the mother and the target or problem child. Hoffman (13) hypothesizes that the frequency of mother-child alliances may be due to the fact that the mother is generally the more dissatisfied partner under conditions of marital conflict and is thus more likely to rely upon the child for emotional compensation or an increase in interpersonal power. Finally, there has been some theoretical discussion concerning the differential flexibility of alliance patterns and levels across situations. Speer (29); Wertheim (31, 32); and Olson, Russell and Sprenkle (26) have proposed that disturbed families are characterized either by the inability to change (rigidity hypothesis) or extreme, cross-situational change (disorganization hypothesis). In contrast, functional families are characterized by the ability to make moderate alignment changes dictated by current situational demands. A variety of measures have been employed in previous empirical studies of family alliances. Alliances have been inferred on the basis of similarities in the verbal reports of family members. Family members with similar attitudes (4) or decision-making responses (25) were considered to be aligned. Questions can be raised as to whether similarities of verbal reports are acceptable operationalizations of interpersonal alliances. In addition, the investigator does not know what particular interactions contribute to the obtained differences. With respect to substantive findings, studies employing verbal-report measures of alliances have yielded inconsistent conclusions concerning differences in marital and parent-child alliances across contrast groups. The majority of empirical studies have evaluated family interaction as a means of inferring alliances. A large number of studies have inferred alliances on the basis of family process measures the relative frequency with which family interact (11, 23), the sequential patterning of speakers (22), and successful interruptions of a dyadic exchange by a third party (16). _____________________________________________________________________________________________________________ 1

Upload: ernesto-contreras

Post on 17-Aug-2015

213 views

Category:

Documents


1 download

DESCRIPTION

Family Process Volume 23

TRANSCRIPT

Fam Proc 23:75-87, 1984Patterns of Alliances in Nondistressed and Multiproblem FamiliesRICHARDGILBERT, PH.D.aANDREWCHRISTENSEN, PH.D.aGAYLAMARGOLIN, PH.D.baDepartment of Psychology, University of California, Los Angeles, 90024. Reprint requests should be addressed to Dr. Christensen.bDepartment of Psychology, University of Southern California, Los Angeles.The goal of this study was empirically to investigate differences between distressed and nondistressed families in thestrength, patterning, and cross-situational consistency of alliances. Audiotapes were made of 12 distressed and 12nondistressed families in two observational situations. The resultant interaction was then coded for family alliances bytrained observers. Distressed families were characterized by low overall levels of alliance behavior, weakness in themarital alliance relative to other family alliances, and discrepancies in parental alliances with the target child. Therewere generally no differences between groups in the cross-situational variability of alliance strength or patterning. Theresults are discussed in relation to the predictions of structural models of family process.The structural perspective on family systems (e.g. 19, 20, 24) has addressed marital and parent-child alliances indistressed and nondistressed families. Some statements have focused on differences in the overall level of alliance orinterpersonal support across contrast groups. Concepts such as disengagement (19) or emotional divorce (24) emphasize adeficiency in emotional closeness and support in distressed family systems. Other theorists have focused on distinctions inthe intrasystem patterning of alliances across groups. For example, Lidz and his colleagues (18) have maintained that manyfamilies with a schizophrenic child suffer from a pattern of "role distortion" termed family schism. Schismatic families arecharacterized by open hostility among spouses and attempts to force the child to take sides. Similarly, Minuchin and hiscolleagues (20) have implicated the presence of a weak marital coalition and extreme cross-generational or parent-childalliances as a dysfunctional structural pattern.Minuchin has discussed a number of implications of low levels of mutual support and skewed alliance patterns (19). Anabsence of support leads to negative affective states in individual family members. In addition, a system consisting ofdisengaged relationships will be less effective in adaptive processes such as negotiation, problem-solving, andconflict-resolution. This is especially true when the system members with the most knowledge and experience (i.e., theparents) are disengaged. With respect to skewed alliance patterns, extreme parent-child alliances are viewed asdysfunctional, in that (a) they provide the aligned parent with an alternate support system, thus reducing the motivation toaddress and resolve interspouse conflicts directly; (b) they place the child in a situation of divided loyalty, of having tochoose between parents; and (c) they place the child in an undesirable position of power relative to the non-aligned spouse.Leibman, Honig, and Berger (15) have also discussed the dysfunctional aspects of extreme parent-child alliances. Theseauthors note that the extreme parent-child alliance in distressed families is most often between the mother and the target orproblem child. Hoffman (13) hypothesizes that the frequency of mother-child alliances may be due to the fact that themother is generally the more dissatisfied partner under conditions of marital conflict and is thus more likely to rely upon thechild for emotional compensation or an increase in interpersonal power.Finally, there has been some theoretical discussion concerning the differential flexibility of alliance patterns and levelsacross situations. Speer (29); Wertheim (31, 32); and Olson, Russell and Sprenkle (26) have proposed that disturbedfamilies are characterized either by the inability to change (rigidity hypothesis) or extreme, cross-situational change(disorganization hypothesis). In contrast, functional families are characterized by the ability to make moderate alignmentchanges dictated by current situational demands.A variety of measures have been employed in previous empirical studies of family alliances. Alliances have been inferredon the basis of similarities in the verbal reports of family members. Family members with similar attitudes (4) ordecision-making responses (25) were considered to be aligned. Questions can be raised as to whether similarities of verbalreports are acceptable operationalizations of interpersonal alliances. In addition, the investigator does not know whatparticular interactions contribute to the obtained differences. With respect to substantive findings, studies employingverbal-report measures of alliances have yielded inconsistent conclusions concerning differences in marital and parent-childalliances across contrast groups.The majority of empirical studies have evaluated family interaction as a means of inferring alliances. A large number ofstudies have inferred alliances on the basis of family process measuresthe relative frequency with which family interact(11, 23), the sequential patterning of speakers (22), and successful interruptions of a dyadic exchange by a third party (16)._____________________________________________________________________________________________________________1In addition to yielding highly inconsistent findings, there are major interpretive problems associated with content-freemeasures of family process. Specifically, high rates of exchange between two family members can signal the presence ofopen conflict as well as a close, involved relationship. Similarly, successful interruptions of an ongoing exchange can havevarious alliance implications. In either case, an assessment of verbal content would facilitate discrimination between thesealternate interpretations. Other studies have evaluated family alliances using global ratings of interactive content. Forexample, Lewis, Beavers, Gossett, and Phillips (17) rated videotapes of family interaction and concluded that healthyfamilies are characterized by a firm parental alliance, whereas distressed systems have conflictual marital relationships andstrong parent-child alliances (most often between the mother and the identified patient).To date, there has been only one study reported in the literature that employed coded interactions to assess familyalliances (28). In this study, the Bales Interpersonal Process Coding System was used to evaluate the decision-makingcommunication of groups of distressed and nondistressed family triads (2). The results indicated that distressed triads hadweaker marital alliances and more frequent cross-generational alliances. The cross-generational alliances, however, weremost frequently between the father and the child.In the currect investigation, samples of nondistressed and distressed family systems were observed while involved in twoprocesses presumed central to effective adaptationnegotiation and problem-solving. Subsequently, interaction from eachobservational context was coded, employing an observational system designed specifically to study family alliances. Theresults of the coding process were then used to test hypotheses regarding the strength and patterning of family alliances.Moreover, obtaining interaction in two contexts permitted the first cross-situational study of alliances and thus generateddata relevant to the question of structural flexibility or rigidity in family systems. Finally, the use of a detailed coding systemenabled the analysis of specific behaviors contributing to any obtained differences in alliance strength and patterning.METHODSubjectsTwelve distressed families with concurrent problems in both the marital and parent-child subsytems and 12nondistressed families were recruited through local clinics and media announcements. All families were paid for theirparticipation in the study. For the distressed families, the research was part of a pretreatment assessment.All families had both parents and at least one school-aged child (5, 6, 7, 8, 9, 10, 11, 12, 13) living at home.Classification as a nondistressed or distressed family was based primarily on a number of parent-report measures of maritaland child behavior, each of which has normative data. Specifically, to be considered a nondistressed family, questionnairescores had to be within one standard deviation of the mean for nondistressed families on (a) a general measure of maritalsatisfaction: an average parent score of 97 or less on the Dyadic Adjustment Scale (29); (b) a measure of desired-maritalbehavior change: a total parent score of 14 or less on the Areas of Change Questionnaire (30); (c) a subjective rating scaleof child problems: an average parent score of 20 or less for each child on the Becker Bipolar Adjective Checklist (3); and(d) a behavior checklist of child problems: an average parent score of 60 or less for each child on the externalizing scale ofthe Achenbach Child Behavior Checklist (1). In addition, following a screening interview, two interviewers had to agreethat the family manifested no signs of meaningful child or marital problems. To be considered a distressed family, parentscores had to be more than one standard deviation from the mean for non-distressed families on at least two questionnaires.Further, the two questionnaires indicating distress had to apply to different subsystems of the family (i.e., at least one childand one marital measure). In addition, two clinicians had to agree, following an intake interview, that the family wasexperiencing moderate to severe marital and child problems.Table I summarizes the means and standard deviations for nondistressed and distressed families on the four parent-reportcriterion measures. T-comparisons on each criterion measure across groups were significant at the .001 level or below. Inaddition, Table I summarizes 10 demographic characteristics of the samples under investigation. Families in the studytended to be middle-class with two school-aged children and parents in their mid to late 30s who had some collegeeducation. Two demographic characteristics refer to the "target child." The target child was the child whose combined scoreon the Becker and the Achenbach child questionnaires was most negative. A target child was determined for nondistressedas well as problem families. In all cases, the empirically defined target child in distressed families was the same as the childclinically reported by parents as most problematic. Target children in the current sample tended, to be about 9 years of ageand were most often male. T-comparisons were conducted on each demographic variable. The results indicated that onlymother's age, t = -2.78, df = 22, p < .01) was significantly different across groups. However, the mean difference was just 4years, and age of mother did not correlate significantly with any of the major alliance measures (all r's were below .23).Table 1Criterion Data and Demographic CharacteristicsNondistressed Families Distressed Families_____________________________________________________________________________________________________________2Variable M SD M SDCRITERION MEASURESDyadic Adjustment Scalea123.0 15.6 83.0 13.6Areas of Changeb4.0 3.8 28.0 14.5Becker Adjective Checklista-9.0 18.1 26.0 11.1Achenbach Child Behavior Checklista49.0 5.0 67.0 6.9DEMOGRAPHIC MEASURESMother's Age 37.0 3.7 33.0 4.2Father's Age 39.0 6.8 35.0 3.7Years of Education-Mother 15.0 2.4 14.0 2.3Years of Education-Father 16.0 1.9 15.0 2.8Family Incomec43.0 24.6 31.0 20.9Socioeconomic Statusd53.0 9.8 50.0 13.4Family Size 4.2 1.2 4.3 1.0Age of Children 9.0 2.8 8.0 2.2Age of Target Childe9.3 3.1 9.3 2.5Sex of Target Childf1.7 0.5 1.5 0.4a Average score across mother and father data.b Total score across mother and father data.c In thousands of dollars.d Based on the Hollingshead Two-Factor Index (14)e Target Child = the child whose combined score on the Becker Adjective Checklist and the Achenbach Child Behavior Checklist wasthe most negative.f Where 1 = female and 2 = male.ProcedureAll families participated in an extensive assessment that included interviews, questionnaires, two weeks of randomaudio-recordings in the home (5), two weeks of daily telephone calls monitoring important child and marital behaviors, anda number of structured interaction tasks. A detailed account of these assessment procedures is provided by Christensen andMargolin (6).In the currect study, data from two, 10-minute, structured interaction tasks were used. In the Problem-SolvingInteraction, family members select and discuss a salient problem from a standard list of family problems (e.g., problemsinvolving chores, bedtime, teasing, etc.). In the Negotiation Skills Interaction, family members negotiate a mutuallyagreeable dinner menu. According to members of Minuchin's systems-oriented family research group, the menu taskgenerates the highest rate of interaction applicable to alliance patterns of the five systems-based interaction tasks they haveused.1Verbatim transcripts from each of the 48, 10-minute interactions (24 families in two situations) were made. Thetranscripts indicated who was speaking and the content of the speech. Each transcript was coded using the Family AlliancesCoding System (9). This system consists of 6 content codes of positive alliance (e.g., Affection, Defend/Protect), 8 contentcodes signaling "negative alliance" (e.g., Attack, Disaffiliate), as well as codes of positive and negative affect.Within a particular transcript, each event or speech act was coded. A new event occurred each time there was a changein speaker. For each event, the coder indicated To Whom the speaker was communicating, the Content of thecommunication, and its Affective quality. The coding of Affect required that the coder listen to the raw interaction as well asrefer to the transcripts. In addition, family members often talked to one member about another family member. In suchcases, the person About Whom the speaker was referring was coded as well as the Content of this reference.Numerical values or weightings for all content and affect codes were determined by providing 20 advanced graduatestudents in clinical psychology with brief descriptions of each code and asking them to rate (from +10 to -10) the degree towhich the code indicated a positive or negative alliance. The average rating of the 20 graduate students determined theweighted alliance value of a particular code. For example, Attack was given a mean rating of -9, whereas Defend/Protectearned an average rating of +5._____________________________________________________________________________________________________________3For each event, alliance scores for all dyads implicated in the event were computed using the weighted values of thecodes. Thus, mother attacks daughter and defends father would result in alliance scores of mother-daughter, -9, andmother-father, +5, for this event. These event scores were summed across the entire interactionan average of 200 events.The result of this summation was a set of "total alliance scores" for each dyad in the family reflecting the "interpersonalstructure" or alliance pattern of the family. Figure 1 depicts the alliance pattern of one of the distressed families thatparticipated in the study. The family depicted in Figure 1 has an alliance pattern that includes (a) moderate to high levels ofnonsupport in the overall system; (b) weak within-generation alliances (marital and sibling), with the marital alliance theweakest in the family; (c) equivalent support of the nontarget child (the daughter) by both parents; and (d) mother andfather in highly dissimilar relationships with the target child (the son), with the mother-target relationship far more negative(i.e., a skewed parent-target-child alliance pattern).Figure 1.An Alliance Diagram of a Distressed Family of FourFor purposes of reliability, each transcript was independently coded by three observers, each of whom had partipated in asix-week training program in the use of the Family Alliances Coding System. Observer agreement was based on thepercentage of events in which there was Consensus Agreement or unanimity in coding (i.e., at least two out of three codersagreed on a codesee Cohen and Christensen [7]). Consensus Agreement was computed for all possible coding decisions(i.e., To Whom, Content, Affect, About Whom, About Whom Content). During training, coders had to achieve a criterionperformance of 90 per cent consensus agreement. In the actual study, overall consensus agreement (across all families,situations, and coding decisions) was 97 per cent. In the small number of cases in which there was no consensus for aparticular decision, the response of one of the three individual coders was selected at random to be the code used foranalyses.Four groups of coders (three coders per group) were trained. The assignment of coders to transcripts was random, withthe constraints that (a) all coders code an equal number of the four types of transcripts (Distressed Problem-Solving,Nondistressed Problem-Solving, Distressed Negotiation, Nondistressed Negotiation), and (b) no coder code the samefamily in two situations. All coders were kept uninformed concerning the treatment status of the families they coded, the_____________________________________________________________________________________________________________4numerical values of the codes, and the specific purposes of the investigation.RESULTSStrength of AlliancesIt was hypothesized that nondistressed families would be characterized by stronger alliances than distressed families. Inorder to conduct these analyses, the total alliance scores (i.e., the totals for each dyad across all events in an interaction)were adjusted for differences in family size and amount of speech to create an alliance rate for each dyad.Two-way analyses of variance examining treatment status (distressed/nondistressed) and situation(problem-solving/negotiation) were performed separately on the marital, mother-child, father-child, sibling, and overallfamily alliance rates. In addition, because of the structural view that the target child in distressed families is involved in thedysfunctional parent-child alliance, the mother-child and father-child alliances were broken down into mother- andfather-target alliances and mother- and father-nontarget alliances. The results indicate main effects for Treatment Status forall alliance rates, with the exception of the sibling and father-target alliances. In all significant cases, the alliance rates werehigher in the nondistressed dyads. In addition, there was a significant situation effect for the father-child, mother-child,mother-nontarget and overall family alliance. In all significant cases, alliance rates in the negotiation skills task were morepositive than in the problem-solving interaction. Finally, there was one significant interaction of treatment status bysituation for the marital alliance rate, F(1,22) = 8.63, p < .01. Sheffe post hoc comparisons indicated that there was nosignificant difference in the marital alliance rates in the negotiation skills task. However, distressed couples hadsignificantly more negative alliance rates in the problem-solving interaction (p < .05).Patterns of AllianceIt was hypothesized that, in distressed families, the relative strength of the marital alliance would be lower and therewould be a greater discrepancy between the mother-child and father-child alliance rates (i.e., a skewed pattern ofparent-child support). In addition, because of the structural view that alliance problems often involve the target child, it wasspecifically hypothesized that the parent-child alliance rate discrepancies would be most pronounced in the case of thetarget child.In order to test the hypothesis concerning the relative position of the marital alliance, the "alliance rate" scores for eachdyad were converted into ranks. Subsequently, in order to control for differences in family size, the rank of each dyad wasdivided by the total number of communication pathways in the family. For example, a marital rank score of 2 in a family ofthree would be divided by 6 (the number of communication pathways in a family of three), with a result of .33. Theobtained score, multiplied by one hundred, reflects the percentage of communication pathways in the family that are morepositive than the dyad under study. The lower the alliance rank percentage score, the greater the relative position of thealliance within the family system. By employing a measure based upon rank data, one can evaluate the intrasystem positionof the marital alliance independent of the overall level of positivity or negativity of the family.Two-Way (Treatment Status X Situation) analyses of variance were performed on the alliance rank percentages of alldyads. The results indicated no main effects for either Treatment Status or Situation. However, there was a significantTwo-Way Interaction for the marital alliance rank, F(1,22) = 8.27, p < .01. Sheffe post hoc comparisons indicated thatthere was a significant difference in the problem-solving interaction (p < .05) but none in the negotiation tasks.Two-Way (Treatment Status X Situation) analyses of variance were also employed to assess differential discrepancyscores between mother-child and father-child alliance rates. Discrepancy scores are also independent of the overall level ofpositivity or negativity of the system. Thus, high discrepancy between two positive alliance rates would signal a skewedparent-child alliance pattern, whereas low discrepancy between two negative rates would not. Separate analyses of variancewere performed for target child and non-target child data. The results indicate no main effects for Treatment Status orSituation for either target or nontarget data. However, a significant Two-Way Interaction was found using the target childscores, F(1,22) = 11.07, p < .003. Sheffe post hoc comparisons indicated that distressed families had significantly greaterdiscrepancies between the parent-target child alliance rates only in the problem-solving interaction (p < .05). Thedifferences in the rates of support to the target provided by the mother versus the father were more than twice as great fordistressed families. An inspection of the raw data indicated that the distressed parent-target discrepancy was due to greatertension and negativity in the mother-target relationship, rather than maternal overinvolvement and excessive suport. Intwo-thirds of the cases, the mother-target rate was more negative than the father-target rate.The two-pattern variables (the relative strength of the marital alliance and parent-target child alliance rate discrepancy)can be combined into a single measure termed the Maladaptive Family Structure Index. This index is the marital alliancerank percentage score plus the absolute value of the difference between the parent-target child alliance rates. Expressed asan equation this index is_____________________________________________________________________________________________________________5Maladaptive Family Structure = Martial Alliance Rank Percentage Score (Mother-Target Alliance Rate) (Father-TargetAlliance Rate)Conceptually, if many alliances are stronger than the marital alliance and the parent-target alliances are discrepant, theMaladaptive Family Structure Index will be high. Of course, because both components of the index were found to besignificantly different in the problem-solving task, the combined index would be as well.Cross-Situational Consistency of AlliancesDifferential variability of alliances across situations were evaluated in the following manner: For all Strength of Alliancemeasures (alliance rates) and Pattern measures (alliance ranks and the Maladaptive Family Structure Index), differencescores were computed across situations for each family. T-comparisons across groups were then performed on the absolutevalues of the difference scores. The results did not lend strong support for either greater rigidity (no situational change) ordisorganization (extreme situational change) in distressed families. With respect to alliance rates, only the situationalvariability of the marital alliance was significantly different across groups, t2.37, df22, p < .03. In addition, there was amarginally significant effect for the rank of the mother-target alliance, t1.85, df22, p < .07. In both cases, there was morevariability in the distressed dyad, and the problem-solving score was more negative. There were no significant differencesin situational variability across groups on either the Maladaptive Family Structure Index target or nontarget child scores.Interactive Contributions to Family AlliancesTwo-Way (Treatment Status X Situation) analyses of variance were performed for each of the Family Alliances CodingSysdtem content and affect codes. The results indicate that, although the trends were all in the expected direction, the onlymain effect for Treatment Status was for positive affect, F(1,22)5.72, p < .05. In addition, there was a marginallysignificant effect for negative affect, F(1,22)3.92, p < .06. At the same time, there were a large number of main effects forsituation. Examinations of the cell means revealed that more negative codes were emitted in the problem-solvinginteraction and more positive codes in the negotiation-skills task. There were no significant interaction effects.DISCUSSIONThe results of the present investigation provide empirical support for a number of central ideas in structural familysystems theory (e.g., 19). The data also point toward a need to revise certain theoretical conceptions.With respect to Strength of Alliances, the present findings support the structural view that distressed families arecharacterized by low levels of mutual support (26). Specifically, the alliance rates for distressed families were morenegative than those for nondistressed families. This lack of interpersonal support is thought to reduce the family's ability tochange effectively in response to internal and external requirements (i.e., adaptation), resulting in progressive escalation ofdifficulty. It is interesting to note that the differences in alliance rates across groups were general or systemwide rather thanlocalized in any particular subsystems. The only exception to this overall difference in supportiveness were the sibling andfather-target alliances. No significant differences were found in the alliance rates of these dyads across groups. In addition,there was no evidence that the alliance rates of theoretically prominent dyads (e.g., the marital or parent-target dyads) morestrongly discriminated across groups.More subtle and interesting group differences were found when the focus shifted to the intrasystem patterning ororganization of alliances without regard to the overall level of positivity or negativity of the systems. According to structuralsystems theory, the marital alliance should occupy a significantly lower position relative to other family dyads in distressedfamilies (18, 20, 19, 21). In addition, there should be greater discrepancy between the mother-child and father-childalliance rates in distressed families (i.e., a skewed pattern of parent-child alliance) (e.g., 18, 19). Moreover, discrepanciesin parent-child alliances across groups should involve the target-child (15), often because of overinvolvement between themother and the target (13). The mother is thought to use her relationship with the child as an alternate support system(when (a) there are marital conflicts along the dominant-submissive or closeness-distance dimensions, and (b) the mother isthe more dissatisfied partner in the marital relationship.The results of the current study support the structural view concerning the marital relationship. The rank of the maritaldyad was significantly lower in distressed families during the problem-solving interactions. To the extent that the maritalalliance is weak, the family no longer has its most experienced, knowedgeable, and resourceful subsystem fully contributingto regulatory matters (e.g., child-rearing, household tasks, financial responsibilities) and adaptive needs (e.g.,problem-solving, conflict resolution, etc.). The reduced regulatory and adaptive ability of the system places additional stresson the family and contributes to the maintenance and exacerbation of existing difficulties.The current data also support the presence of a more skewed pattern of parent-child support in distressed families undercertain conditions. Specifically, there was a significantly greater discrepancy between the mother-target and father-targetalliance rates in distressed families in the problem-solving situation. In this conflictual situation, differences in the reates of_____________________________________________________________________________________________________________6support provided by parents to the target child were more than twice as great for distressed families. An inspection of thedata indicates that in two-thirds of the cases, the mother-target rate was more negative than the father-target rate. It is highlyunlikely that this situation is indicative of father-target overinvolvement in distressed families, given that there were nodifferences in the father-target alliance rates across groups. The finding of excessive tension in the mother-target dyad isopposite to the prediction of structural theory, which has stressed the role of the mother-target alliance in achievingsubstitute gratification for the mother under conditions of marital dissatisfaction. The alternative view suggested by thecurrent data is that the mother-target relationship often functions as the locus of displaced maternal tension. A number ofparallel findings support this alternative view. Hazzard, Christensen, and Margolin (12), in a study of children's perceptionsof parental behavior, noted that (a) children from distressed families viewed their parents as treating them less similarlythan did children from nondistressed families, and (b) children in distressed families perceived significantly greatermaternal negative behavior than children in nondistressed families. In addition, correlational data reported by Christensenand Margolin are inconsistent with the substitute gratification perspective (6). These authors report that there is a positivecorrelation between maternal ratings of relationship satisfaction with spouse and relationship satisfaction with the targetchild in distressed families. Acceptance of the substitute gratification view would lead to the expectation that ratings ofmother-target satisfaction would rise when marital satisfaction fell. This inverse variation was not found.Thus, the current data support the presence of a skewed pattern of parent-target support in distressed families underconditions of conflict. In addition, the data suggest the view that the mother-target relationship is crucial to understandingthis discrepancy. At the same time, the data are more consistent with a process of displaced maternal tension versussubstitute gratification. Perhaps additional research can identify family characteristics associated with one type ofintergenerational reaction to stress versus another.Another question addressed in the current study pertained to differences in cross-situational variability in alliancesacross groups. Cross-setting changes in alliance rates and ranks were noted for each dyad, and group differences in thesechange scores were evaluated. The results indicate few significant effects. Thus, strong overall support for either greaterrigidity (no change) or greater disorganization (extreme change) was not provided. The two strongest effects were in thedirection of more variability in distressed dyads. The most interesting cross-situational finding was that the greatestvariability across situations occurred in the distressed marital dyad (rate) and the distressed mother-target dyad (rank). Bothdyads had lower, more negative scores in the problem-solving interaction. These results suggest the possibility that thesetwo, theoretically central alliances are particularly affected by conditions of conflict or stress. When there is change fromconflictual to more neutral interaction, these dyads appear to vary most in their level of support.Finally, the use of coded observational data made possible an investigation of the particular responses contributing todifferences in family alliances. The results indicate that none of the 17 content codes significantly differed across groups. Atthe same time, variations in the frequency of alliance content codes were all in the expected directions. More positive andless negative alliance codes were emitted by nondistressed family members. The affect codes (positive and negative) werethe strongest discriminators across groups. Other investigators have also found that affect codes are more powerfuldiscriminators of clinic and nonclinic samples than content codes. For example, Gottman has reported parallel results in hisinvestigations of distressed and nondistressed marital interaction (10).A number of important methodological points should be mentioned. First, the age range of the children in the sample was5-13; the average age was 8. This range is somewhat wide and, most likely, the meaning and mechanics of parent-childalliances vary with the developmental status of the child. At the same time, the children in the present sample had a numberof important commonalities. For one, all the children were sufficiently verbal to participate actively in the interaction tasksemployed in the study. Second, all children in the sample were attending school. Thus, all the children spent part of the dayat home and part of the day with peers and other adults. Third, none of the children were in high school. It can be arguedthat the major transition from family-oriented to peer-oriented relationships generally occurs during high school. Thus,despite the age range, the children in the sample were all verbal, there was some commonality in their daily routine, andthey were probably still more family- than peer-oriented. It should also be noted that Minuchin and other structural theoristshave not stated that their views on parent-child alliances are only applicable when the children are within a certain agerange. The age range of the children in the study does not violate any existing assumptions of the theoretical model thecurrent study has sought to assess. Nevertheless, the possible impact of a child's developmental status on the dynamics ofparent-child alliances is an interesting idea that merits future theoretical and empirical consideration.Other methodological considerations pertain to the generalizability of the current findings. On one hand thegeneralizability of the study is limited by the relatively small number of families in the sample (N = 24). At the same time,obtaining significant effects on such a small number of cases may mean that the variables under investigation are importantones. Another cautionary note relates to the large number of statistical tests that were conducted. Although many of theeffects were significant at extreme confidence levels (i.e., .01, .001), a number of major findings only reached the .05 levelof significance (e.g., the child component of the Maladaptive Family Structure Index). Given the large number of testsconducted, these results must be interpreted cautiously. A final word of caution pertains to the extreme difference betweenthe two samples. As Table I indicates, the groups were different at the .001 level on all criterion measures of distress._____________________________________________________________________________________________________________7Incorporating a third, less distressed group of families into the design would provide information concerning the extent towhich the current findings can be generalized to a wider population of distressed families.CONCLUSIONIn sum, the present research could be extended by future studies with additional design features (e.g., a large sample size,adding a less distressed treatment group). Nevertheless, the current study makes a meaningful contribution by adding to theempirical development of structural systems theory. First, the current data provide support for a number of central ideas instructural systems theory (e.g., the presence of a relatively weak marital alliance in distressed families, the presence ofdiscrepant parent-target alliance rates in distressed families, and the locus of this discrepancy in the extremity of themother-target alliance). Second, the data should contribute to the refinement of the model by pointing toward areas in needof additional consideration (e.g., the nature of the distressed mother-target relationship, the situational flexibility or stabilityof alliance patterns). Third, the data indicate that structural variables are able to discriminate between clinically relevantcomparison groups. Finally, and perhaps most important, the present study developed preliminary operationalizations ofabstract structural concepts (e.g., the relative strength of the marital alliance, maladaptive family structure) that can be usedas a point of departure for subsequent investigations. Although structural theory has been theoretically and clinicallyprominent for over a decade, the difficulty in developing valid operationalizations of central constructs has been a majorimpediment to the empirical development of the model. Perhaps because of these empirical difficulties, the majority offamily systems research has focused on relatively specific aspects of process (e.g., the rate and contingent nature ofreinforcement, the clarity of transactions, etc.). While the detailed study of specific process variables has meaningfullyadvanced the understanding and treatment of families, the present research suggests that attention to the interpersonalorganization or structure of the family system is also an important consideration.REFERENCES1. Achenbach,T.M., "The Child Behavior Profile: I. Boys Aged 6-11", J. Consult. Clin. Psychol., 46, 478-488,1978.2. Bales,R.F. (Ed.), Interaction Process Analysis, Cambridge, Mass., Addison-Wesley Press, 1950.3. Becker,W.C., "The Relationship of Factors in Parental Ratings of Self and Each Other to the Behavior ofKindergarten Children as Rated by Mothers, Fathers, and Teachers," J. Consult. Psychol., 24, 507-527, 1960.4. Cheek,F.E., "The %'Schizophrenic Mother' in Word and Deed," Fam. Proc., 3, 155-177, 1964.5. Christensen,A., "Naturalistic Observation of Families: A System for Random Audio Recordings in The Home,"Behav. Ther., 10, 418-422, 1979.6. Christensen,A. and Margolin,G., "Correlational and Sequential Analyses of Marital and Child Problems," Am.Psychol. Assn. 1981.7. Cohen,R.S. and Christensen,A., "Further Examination of Demand Characteristics in Marital Interaction," J.Consult. Clin. Psychol., 48, 121-123, 1980.8. Ferreira,H.J., Winter,W.D. and Poindexter,E.J., "Some Interactional Variables in Normal and AbnormalFamilies," Fam. Proc., 5, 60-75, 1960.9. Gilbert,R., Saltar,K., Deskin,J., Karagozian,A., Severance,G. and Christensen,A., "The Family AlliancesCoding System (FACS) Manual," unpublished manuscript, University of California, Los Angeles, 1981. 10. Gottman,J.M. (Ed.), Marital interaction: Experimental investigations, New York, Academic Press, 1979. 11. Haley,J., "Research on Family Patterns: An Instrument Measurement," Fam. Proc., 3, 41-65, 1964. 12. Hazzard,A., Christensen,A. and Margolin,G., "Children's Perceptions of Parental Behaviors," J. Abnorm. ChildPsychol., 11, 49-59, 1983. 13. Hoffman,L., "Breaking the Homeostatic Cycle," in P. Guerin, Jr., (ed.), Family Therapy: Theory and Practice,New York, Gardner Press, 1976. 14. Hollingshead,A.B., "Two-Factor Index of Social Status," working paper, Yale University, 1975. 15. Leibman,R., Honig,P. and Berger,H., "The Role of The Family in The treatment of Chronic Asthma," in P. J.Guerin, Jr. (ed.), Family Therapy: Theory and Practice, New York, Gardner Press, 1976. 16. Lennard,H.L., Beaulieu,M.R. and Embry,N.G., "Interaction in Families with a Schizophrenic Child," Arch.Gen. Psychiat., 12, 166-183, 1965. 17. Lewis,J.M., Beavers,W.R., Gossett,J.T. and Phillips,V.A. (Eds.), No Single Thread: Psychological Health inFamily Systems, New York, Brunner/Mazel, 1976. 18. Lidz,T., Cornelison,A., Fleck,S. and Terry,D., "The Intrafamilial Environment of Schizophrenia Patients: II.Marital Schism and Marital Skew," Am. J. Psychiatr., 114, 241-248, 1957. 19. Minuchin,S. (Ed.), Families and Family Therapy, Cambridge, Mass., Harvard University Press, 1974._____________________________________________________________________________________________________________8 20. Minuchin,S., Montalvo,B., Guerney,B., Rosman,B. and Schumer,F. (Eds.), Families of the Slums, New York,Basic Books, 1967. 21. Minuchin,S., Rosman,B.L. and Baker,L. (Eds.), Psychomatic Families: Anorexia Nervosa in Context,Cambridge, Mass., Harvard University Press, 1978. 22. Mishler,E.G. and Waxler,N.E., "The Sequential Patterning of Interaction in Normal and SchizophrenicChildren," Fam. Proc., 14, 17-50, 1975. 23. Murrell,S.A. and Stachowiak,J.G., "Consistency, Rigidity, and Power in the Interaction Patterns of Clinic andNonclinic Families," J. Abnorm. Psychol., 72, 265-272, 1967. 24. Nameche,G.F., Waring,M. and Ricks,D.F., "Early Indicators of Outcome in Schizophrenia," J. Nerv. Ment.Dis., 139, 232-240, 1964. 25. O'Connor,W. and Stachowiak,J., "Patterns of Interaction in Families with High Adjusted, Low Adjusted andMentally Retarded Members," Fam. Proc., 10, 229-241, 1971. 26. Olson,D.H., Russell,C.S. and Sprenkle,D.H., "Circumplex Model of Marital and Family Systems. II: EmpiricalStudies and Clinical Intervention," in J. Vincent (ed.), Advances in Family Intervention, Assessment and Theory,vol. 1, Greenwich, Conn., JAI Press, 1980. 27. Schuham,A., "Power Relations in Emotionally Disturbed and Normal Family Triads," J. Abnorm. Psychol., 75,30-37, 1970. 28. Spanier,G.B., "Measuring Dyadic Adjustment: New Scales for Assessing the Quality of Marriage and SimilarDyads," J. Marr. Fam., 38, 15-28, 1976. 29. Speer,D., "Family Systems: Morphostasis and Morphogenesis, or Is Homeostatis Enough?", Fam. Proc., 9,259-278, 1970. 30. Weiss,R.L. and Margolin,G., "Marital Conflict and Accord," in A. R. Ciminero, K. S. Calhoun, and H. E. Adams(eds.), Handbook for Behavioral Assessment, New York, John Wiley, 1977. 31. Wertheim,E., "Family Unit Therapy and the Science and Typology of Family Systems," Fam. Proc., 12, 361-376,1973. 32. Wertheim,E., "The Science and Typology of Family Systems: Further Theoretical and Practical Considerations,"Fam. Proc., 14, 285-308, 1975.Manuscript received May 17, 1983; Accepted July 1, 1983.1Personal Communication_____________________________________________________________________________________________________________9