A PERSONAL CONSTRUCT ASSESSMENT OF STRUCTURAL FAMILY THERAPY TRAINING
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA 1985
I would like to take this opportunity to acknowledge and thank those individuals who have given their encouragement and guidance throughout the completion of my dissertation. Drs. Harry Grater and Greg Neimeyer, the chairpersons of my committee, have guided and directed this work from its inception to its completion. They have both been invaluable as sounding boards, as well as being my friends. Harry and Greg's commitments to excellence and professionalism have been important models during my years as a graduate student. Their enthusiasm and flexibility in support of my basic ideas, and cooperation over the many years and miles are greatly appreciated. I would also like to thank the members of my committee, Bill Froming, Carolyn Tucker, and Robert Jester, for their effort and insightful suggestions. I have enjoyed and profited from being a student in their respective courses and seminars. It has been my pleasure to know these outstanding individuals personally and professionally.
I would also like to thank all the people who were an integral part of this dissertation. Drs. Peggy Fong, Mary Fukuyama, Max Parker, Ellen Amatea, Robert Myrick and Harry Grater allowed me to utilize their classrooms to solicit subjects. Drs. Herb Steier and Andres Nazario gave their time and energy on short notice in the aid
of refining the stimulus materials. Florence Roess and Ces Bibby
were invaluable in coordinating subjects, experimenters, testing rooms and preparation of all dissertation stimulus materials. I would especially like to thank all the undergraduate research
assistants who gave long hours and were involved in all technical aspects of this study. Susan Lerner, April Metzler, Marc Levant, Mark Kaplan, Regina Davis, Gloria Pinzon, Joni Congdon and Jesus Llobet were invaluable aids to this project as well as good friends. Anita Moreles, Michele Majorek and Chere Ruquist also gave much of their time to this project.
I would like to give a very special acknowledgment to my husband Michael, for his constant love and support, and the many sacrifices he has made in his own career so that I may pursue mine. Throughout
his own graduate work, he has always ungrudingly found the time to give me many hours of technical and professional help on the various
aspects of this project. He has been the single greatest support in my work and in my life, and I am happy to be able to share this time with him. I would also like to thank my parents, Shirley and David Zaken, for their unconditional faith and support throughout my life
and the long years of my graduate career.
TABLE OF CONTENTS
ACKNOWLEDGMENTS ................................................. iii
ABSTRACT ........................................................ vii
ONE INTRODUCTION .......................................... 1
Commonalities and Differences in Family Therapy
Training ............................................ 3
Outcome Studies in Family Therapy Training ............ 9
The Present Study .................................... 10
TWO LITERATURE REVIEW .................................... 14
Family Therapy Training .............................. 14
Descriptive Studies ............................. 14
Empirical Studies ............................... 18
Personal Construct Theory ............................ 25
Personal Construct Psychology and Training Studies...29 The Present Study .................................... 34
Hypotheses ........................................... 35
THREE METHODOLOGY .......................................... 38
Subjects ............................................. 38
Instruments .......................................... 39
Family Experience Inventory ..................... 39
Family Repertory Grid ........................... 40
Therapist Responses to Simulated Family Tapes ... 45 Case Conceptualizations ............... 50
Procedure ............................................ 52
Design and Analyses ... o .............. o ............... 55
Structural Grid Analyses ........................ 56
Content Grid Measures .................. o ... o .... 56
Case Conceptualization Analyses ................. 57
Therapist Response Analyses ... o ........ oo- .... 57
FOUR RESULTS .............................................. 60
Structural Grid Analyses ............................. 60
Content Grid Analyses ................................ 63
Case Conceptualization Analyses...' 63
Family Conceptualization Analyses ............... 63
Family Intervention Analyses .................... 64
Therapist Response Analyses .......................... 64
Within Category Analyses ........................ 64
Between Category Analyses ....................... 66
FIVE DISCUSSION ........................................... 68
Perceptual Skills and Cognitive Complexity ........... 70
Perceptual Skills and Level of Construct
Abstraction ........................................ 75
Conceptual Skills and Case Conceptualizations ........ 76 Executive Skills and Therapist Responses ............. 78
Executive Skills Within Category ................ 80
Executive Skills Across Category ................ 83
Summary .............................................. 84
Future Considerations ................................ 87
A FAMILY THERAPY COURSE DESCRIPTION .................... 92
B FAMILY THERAPY EXPERIENCE INVENTORY .................. 96
C FAMILY REPERTORY GRID ................................ 98
D GRID CONTENT CODING INSTRUCTIONS ..................... 99
E GRID CONTENT EXAMPLES ............................... 101
F FAMILY A DYNAMICS SEEN BY ACTORS .................... 102
G FAMILY B DYNAMICS SEEN BY ACTORS .................... 107
H THERAPIST RESPONSE CODING SYSTEM .................... Ill
I THERAPIST RESPONSE ALTERNATIVES: FA14ILY A .......... 112
i THERAPIST RESPONSE ALTERNATIVES: FAMILY B .......... 119
K THERAPIST RESPONSES ANSWER SHEET .................... 126
L CASE CONCEPTUALIZATION QUESTIONS: FAMILY A ......... 135
M CASE CONCEPTUALIZATION QUESTIONS: FAMILY B ......... 139
N INFORMED CONSENT FOR FAMILY THERAPIST'S PROJECT ..... 143
0 FAMILY A DYNAMICS SEEN BY SUBJECTS .................. 144
P FAMILY B DYNAMICS SEEN BY SUBJECTS .................. 145
Q CASE CONCEPTUALIZATION ANSWER SHEET ................. 146
R FAMILY REP TEST INSTRUCTIONS ........................ 147
REFERENCES ...................................................... 149
BIOGRAPHICAL SKETCH ............................................. 157
Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy A PERSONAL CONSTRUCT ASSESSMENT OF STRUCTURAL FAMILY THERAPY TRAINING BY
Chairman: Dr. Harry Grater
Cochairman: Dr. Greg J. Neimeyer Major Department: Psychology
The investigation of the effects of training in family therapy is a new and relatively uncharted area of inquiry. There are only
two studies cited in the literature that begin to approach the experimental rigor and control necessary to explore the impact of family therapy training on therapy skills. The present investigation
attempted to control for the lack of experimental control and objective measures of family therapy skills that continue to be
reported in the literature.
Based upon the distinction among therapy skills cited in the
literature, the present study investigated perceptual, conceptual and executive skills associated with effective family therapy. Perceptual skills were measured using a modified version of the Role Repertory Grid, yielding measures of the structure and content of
individuals' personal cognitive systems. It was predicted that the structural complexity and level of abstraction in the family construct system would increase as a function of family therapy training.
Conceptual and executive skills were measured by analyzing
trainees' multiple choice responses to simulated family therapy tapes. It was predicted that the family therapy trainees would show increases in level of systemic conceptualization and therapist response sophistication over the course of training.
Results of the 2 x 2 Analyses of Variance on the pre-post scores for both family and nonfamily therapy trainees provided equivocal support for the predictions of increased complexity in family therapy trainee skills. The analyses revealed significantly decreased levels of differentiation in perceptual skills for family therapy trainees
and significantly increased levels of differentiation for the nonfamily therapy trainees, and no significant changes between groups for conceptual or executive skills.
These results suggest that family therapy skills may be
developmentally dependent upon each other such that higher level conceptual and executive skills cannot develop until lower level
perceptual skills have been adequately developed and consolidated. The results of this study were also discussed in relation to the unique contribution of family therapy training on therapy skills, and
future directions were suggested.
The past decade has witnessed the growing demand for clinical
and counseling psychology programs, professional psychology programs,
and specialized training programs for mental health and post-doctoral professionals. Foremost among these is the demand for training in family therapy which has resulted in the growth of a variety of family therapy training programs. Despite this increasing demand, there has been a dearth of research investigating important variables in the training of family therapists (Woody & Weber, 1984).
The task of investigating the impact of training on potential
family therapists is an ambitious one. The investigator must decide
what type of program to investigate, what theoretical orientation within the family therapy field to explore, and then, how to go about isolating the critical variables involved in order that general statements about training can be made. According to Tucker and Pinsof (1984), this task is made even more difficult due to the complexity of measuring conceptual, experiential, and behavioral learning in a stimulus situation that is never duplicated. In addition to the problem of not having a standard stimulus, the family therapy field has also lacked the necessary instruments to measure conceptual and behavioral skills (Gurman & Kniskern, 1978).
The present investigation focuses on training in structural
family therapy. As in all forms of therapy, the structural family therapist relies on his powers of observation when assessing a family's functioning. The distinguishing feature of structural family therapy, however, involves the active participant observation of the therapist. The focus of intervention according to the structural model is the patterns of structural relatedness occurring within the family system (White, 1979). The therapist's task is to discover the dysfunctional structure existing within the family and to design interventions to alter that structure. Structure is examined in terms of power hierarchy, coalitions, alliances,
individual and subsystem boundary regulation, and the family's flexibility for change (Minuchin, 1974). In order to observe these patterns, the structural family therapist must find a way to get the
family to demonstrate their patterns of behavior, and this is where the techniques of the therapy are first utilized. Once the therapist
has operationalized where in the structure the system fails to carry out its functions, the goals of therapy are to solve problems through changing the underlying systemic structure of the family.
The present study will adopt a personal construct approach to the study of training in structural family therapy. Personal construct psychology is a theoretical and empirical approach to the study of personality and interpersonal relationships. The personal construct methodology can provide the needed empirical rigor, which has been lacking in prior studies of family therapy training, and can
bridge the gap between the relatively uncharted area of family
therapy training and more established approaches to personality, counseling and interpersonal relationships.
Researchers in the field of family therapy have recently begun
to report some descriptive studies of training, typically ending with a discussion of the need for empirical research in this area. Before discussing the findings of the current literature on family therapy training, the reader needs to be acquainted with the programs currently available, the-goals of family therapy, and how those goals are translated into training and assessment. Further, the components of family therapy training that extend across the various schools of thought will be reviewed. In addition, a brief discussion of the outcome literature will be presented, followed by a rationale for the adoption of a Personal Construct Psychology approach to the study of training in family therapy.
Commonalities and Differences in Family Therapy Training
In a national survey of training programs in family therapy, Bloch and Weiss (1981) reported a growing and varied assortment of programs. In addition to the diversity of programs available, they
also discovered that the establishment of family therapy training programs has grown exponentially between 1942 and 1980. Levant (1984) has suggested that these training programs may be divided into two major groups. The first group involves training that is offered as part of the overall training program of one of the various mental health or human-service professions. The second group involves programs that are specifically designed to train family therapists.
According to Kniskern and Gurman (1979), these training models differ on three major dimensions. The first dimension involves whether or not the training leads to a degree or certificate in family therapy. Second, is family therapy the only form of therapy taught
at a given training center or does the program require some experience with individual and group therapy? Third, training programs vary as to the extent of previous experience in psychotherapy that is required for admission.
In order to evaluate family therapy training, investigators must specify what type of program is under study as well as how that program defines family therapy. Kniskern and Gurman (1979) argue
that the meaning which one attaches to the term family therapy will directly affect the way in which one trains family therapists and evaluates that training. These authors suggest that a training program that views family therapy as a technique will focus primarily on the acquisition of technical skill, and as such, the acquisition of specific technical skills will be viewed as the most appropriate criterion of the success of the training program (e.g., behavioral and communications approaches). However, if family therapy is viewed
as a conceptual approach to the understanding of behavior and behavioral pathology, Kniskern and Gurman (1979) argue that the training program will view the acquisition of specific intervention
techniques as a secondary goal (e.g., existential and strategic approaches) Thus, before general statements about the effectiveness of family therapy training can be made, differences between the goals
and methods of different training programs must be taken into consideration.
Although there exists a diversity of theoretical orientations
and training programs in family therapy, several investigators (e.g., Bloch & Weiss, 1981; Woody & Weber, 1984) have suggested that there
are several important basic concepts that are inherent to all training programs irrespective of the setting, experience of the student, and theoretical preferences of the trainers. Regardless of
the theoretical orientation, family therapy focuses on the interrelationships of the family members. Each theory provides guidelines for the therapist to assess and develop treatment
interventions that are intended to create systemic changes among the family members (Woody & Weber, 1984).
Many family therapists (e.g., Okun & Rappaport, 1980; Sedgwick,
1981; Stanton & Todd, 1979; Woody & Weber, 1984) agree that students need to learn to assess the family from a developmental model that considers the progression of the family through family life stages.
Since each stage of the family life cycle requires that specific tasks be successfully accomplished before the family can progress
into the next stage, failure on the part of the family to perform the tasks can result in becoming fixated, resulting in dysfunctional behavior. Family therapy trainees need to learn to consider the family's current level of functioning in relation to their life stages.
A third major point that family therapists agree upon is the emphasis on the individuation processes of the family members as a
primary task for the family to accomplish. If a family is unable to foster this individuation process, an individual's symptoms and problems may be reflective of family stress, conflict, or the internalization of other family breakdowns. Further, the less effective the family, the more limited the family will be in its capacity to facilitate the individuation process. The student
requires training that focuses on learning to accurately assess and develop goals for the family that will alter the family system and promote the individuation process.
Most importantly, family therapies are problem focused. The family therapist must be skilled in assessing the function of the problem for the family and the family system's response to it (e.g., enmeshed or disengaged relations, regression, rigidity, or internalization).
Although the differing family therapy orientations agree on
these four points, the conceptual orientation of the curriculum will influence how the problem resolution will be obtained. While some theories, such as problem-solving, learning, behavioral, and communication emphasize learning problem-solving skills as a major objective, other theories, such as structural and strategic approaches, use indirect, manipulative, and metacommunicative interventions to induce change for problem resolution (Woody & Weber, 1984). How these goals of therapy are translated into training remains unclear. Empirical research that investigates the process of change as well as the outcome of training may be able to overcome the
dilemma of differing programs and theoretical orientations, and
answer the question of what impact training has on family therapy trainees.
.Regardless of the type of training program involved, there are certain components of family therapy training that extend across the various schools of thought. All therapy training programs, including family programs, utilize four primary methods for training: didactic, supervisory, observational, and experiential. Since the field of family therapy is relatively new, most training programs
require their trainees to read a variety of sources that reflect the diversity of family therapy. However, there are also training
programs that concentrate on the work of a single well known therapist and school of thought. Kniskern and Gurman (1979) suggest that although programs with a single, integrated approach to family therapy lessens the confusion that results from sampling readings from the various schools of thought, these programs may run the risk of producing less creative and flexible therapists. In agreement with Kniskern and Gurman's argument, the present study will
investigate the training process that occurs within a program that places a major emphasis on structural family therapy (e.g., Minuchin), while also incorporating other schools of thought such as strategic (e.g., Haley), communications (e.g., Satir), and experiential (e.g., Whitaker).
The second mode of training involves supervision of the
trainees. Most family therapy trainers agree (Kniskern & Gurman, 1979; Liddle & Haiprin, 1978) that the primary teaching of family therapy occurs in supervision. However, the techniques of
supervision vary from a focus almost exclusively on the problems of
the family being treated to a focus almost entirely on the trainee independent of the family being treated, to the therapist's relationship with the family. Thus, the techniques utilized by the supervisor, as well as his or her style of supervision, are
influenced by both the theoretical and therapeutic orientation (Kniskern & Gurman, 1979). For example, a directive and problemoriented therapist would probably supervise in a direct, problemoriented way, while a process-oriented therapist would be more concerned with the personal growth of the trainee. Although no study
has investigated this impact of training, Kniskern and Gurman (1979) suggest that such stylistic differences in supervision will tend to result in differences in therapeutic style by the trainees. Thus, when interpreting shifts across training, it is necessary to keep the supervision style in mind. The supervision style utilized in the present study is based on a structural problem-solving approach to family therapy.
The observational component involves having the trainees view tapes of family therapists in action, viewing in vivo family therapy sessions through a one-way mirror, and viewing role-play interviews of other family therapy trainees. As with the supervision style,
what is observed is influenced by the style and theoretical orientation of the trainer.
The fourth mode of therapy training is classified as
experiential. Experiential methods may involve role-playing, participating in a simulated family, or actually working with a
clinical family. In the present study, trainees participate in roleplaying interview sessions with simulated families, and work with a clinical family throughout their second semester of training.
Although these training components extend across the various schools of thought, each program utilizes them in a different way.
Thus, it becomes necessary to concentrate on the processes of change involved in training rather than simply on which modes of training are effective, or which school of therapy produces more effective family therapists. Further, when attempting to generalize across the various theoretical orientations through the commonalities in
training methods, it is important to clarify and operationalize the structure and content of the program. Without this specificity, it will be impossible to account for discrepant findings, or even to integrate findings among the different schools of thought.
Outcome Studies in Family Therapy Training
Although there recently has been an emphasis placed on the need for process-outcome investigations (Gurman & Kniskern, 1981), there have been only a handful of empirical studies of family therapy training (e.g., Allred & Kersey, 1977; Breunlin, Schwartz, Krause & Selby, 1983; Byles, Bishop & Horn, 1983; Friedman, 1971; Lange, 1978; Lewis, 1977; Pinsof, 1977; Tomm, 1980; Tucker & Pinsof, 1984), or even descriptive analyses of training in family therapy (e.g., Ferber, 1972; Flint & Rioch, 1963; Flomenhaft & Carter, 1974; 1977; Nichols, 1979). In an extensive review of the literature on structural family therapy, Aponte and Van Deusen (1981) cite only
four studies designed to address issues related to training structural family therapists. Three of these (Betof, 1977; Flomenhaft & Carter, 1974; 1977) described results of uncontrolled post-hoc evaluations of training programs. The fourth (Kaplan, Rosman, Liebman & Honig, 1977) represented a multi-method assessment of a one year training program within the context of a small sample (20 trainees and 5 controls). These studies only begin to redress a
long-standing inattention to the empirical assessment of family therapy training programs. Another major difficulty encountered in this area of research is the lack of sensitive instruments for measuring change in therapist skills (Gurman, 1983). Therefore, the dearth of process-outcome studies and lack of a methodology sensitive enough to detect changes in the skills of family therapists support Gurman and Kniskern's recent conclusion that "we must acknowledge and underline the field's collective empirical ignorance about this domain" (1981, p. 772).
The Present Study
The theory and methodology underlying the present study is
adopted from Kelly's (1955) personal construct psychology. Personal construct theory was initially developed as a conceptual approach to the study of personality, psychopathology and psychotherapy. This theory has generated a wide body of empirical research and has recently reemerged as an important contribution to the study of human behaviors (e.g., Bannister, 1981; Bannister & Fransella, 1955; R.A. Neimeyer, 1985; Neimeyer, Klein, Gurman & Griest, 1983; Neimeyer &
Neimeyer, 1981). This theory has also been applied to various training contexts, including training in social work (e.g., Lifshitz, 1974), education (e.g., Ryle & Breen, 1974) and marriage and family therapy (e.g., Lewis, 1977; G.J. Neimeyer, 1985; Neimeyer & Hudson, 1984; Procter, 1981).
According to personal construct theory, individuals operate as personal scientists" whose aim is to render experience meaningful;
to better understand, predict and control the events with which they are confronted. After repeated experience with a particular type of event, an individual abstracts salient commonalities and characteristics of these situations. The abstracted features are meaningful only in relation to their opposites. For example, after several family therapy sessions, a family therapist may come to see certain families as enmeshed, a quality meaningful insofar as it contrasts with being disengaged. This personal dimension, or construct of "enmeshed-disengaged" serves as a continuum along which events (i.e., families) can be ordered, thereby lending structure and meaning to the family therapy experience. Taken together, many such
dimensions form the construct system, a unique conceptual template for ordering and anticipating experience.
Importantly, this system of dimensions is ordered (Organization Corollary, Kelly, 1955) and undergoes characteristic shifts over the course of personal and professional development (see Crockett, 1982, for a review). In particular, following Werner (1955), it can be
assumed that construct systems develop from a stage of globality to particularity; that is, moving from simple to complex levels of
cognitive differentiation over the course of continued experience. In general, this reasoning is consistent with Bieri's definition of complexity as "the tendency to construe social behavior in a multidimensional way such that a cognitively complex individual has available a more versatile system for perceiving the behavior of others than does a less complex person" (1955, p. 14). Following from this definition, subsequent work has noted predicted changes in cognitive complexity over the course of professional development.
For example, two studies have reported a curvilinear relationship between cognitive complexity and level of training for teachers (Runkel & Damrin, 1961) and social workers (Ryle & Breen, 1974). Training seemed to first restrict, and then to enlarge, the subjects' cognitive systems. Subsequent work has addressed the relationship between this level of cognitive complexity and therapeutic decision making among a group of social work students (Duehn & Procter, 1974). These investigators reported that the more complex students
specified a greater number of alternative therapeutic interventions, suggesting that greater complexity may be associated with greater flexibility in responding.
In addition to structural changes in cognitive complexity
associated with training, shifts in the content of the constructs are also predictable from personal construct theory. For example, Lifshitz (1974) compared the content of social work students' constructs with those of their supervisors. Results indicated that students used more concrete descriptions (e.g., age, sex, profession)
while their supervisors used more abstract dimensions (e.g., diligent, responsible).
Thus, the use of personal construct methodology lends itself well to the study of a variety of training methods. Extending this personal construct training literature to the training of family therapists suggests several lines of inquiry. These concern both
structural and qualitative changes in trainees' conceptual and clinical abilities as they gain exposure to, and practice with, a structural-systemic approach to family treatment.
First, based on the systems approach of structural family therapy, shifts are expected in the direction of greater use of systemic dimensions for the trainees (e.g., focusing on family interactions versus an individual focus), as well as shifts in the direction of greater overall complexity in the family construct systems of family therapy trainees (e.g., increases in the number of
constructs used and the interrelationships between those constructs) relative to a comparable control group. In addition, it is predicted that there will be similar shifts towards more complex and systemically oriented case conceptualization skills in the trainees, and more effective and systemically oriented therapist responses in trainees than in control subjects.
Family Therapy Training
Although there is little empirical evidence for the
effectiveness of family therapy training, there is some research concerning the specific therapist factors that influence the outcome of family therapy (e.g., Epstein, Sigal & Rakoff, 1968; Thomlinson, 1974; Tomm & Wright, 1979). Thus, the identification of those
factors that may be taught in training provide indirect support for the potential effectiveness of family therapy training programs. In an extensive review of the family therapy research literature, Gurman and Kniskern (1978) reported therapy structuring skills, experience
level, and relationship skills as the three most important factors that influence the outcome of family therapy. Obviously, experience level is not able to be taught in a training program; however, both therapy structuring skills and relationship skills may be focused on in training. Structuring skills have been investigated by several researchers (Alexander, Barton, Schiavo & Parsons, 1976; Epstein et al., 1968; Sigal, Guttman, Chagoya & Lasry, 1973). Based on thlis research, structuring skills of the therapist have been divided into
directiveness, clarity, self-confidence, information gathering, and
stimulating interaction. Further, Gurman and Kniskern (1978) have argued that the family therapist must generally be active and provide early structure without assaulting family defenses too soon.
In addition, the -ability of a family therapist to establish a
positive relationship with the family has received consistent support as the most important outcome-related therapist factor. Several investigators (Shapiro, 1974; Shapiro & Budman, 1973; Waxenburg,
1973) have reported that therapist empathy, warmth, and genuineness appear to be very important in keeping families in treatment beyond the first interview. In an impressive investigation of relationship
skills, Alexander et al. (1976) reported that while structuring skills discriminated between two levels of poor outcome, only relationship skills were able to discriminate between good and very good outcomes. They reported that these factors were related to positive outcome regardless of the theoretical orientation of the therapist. However, these are the same skills that have been found to be critical for the process of effective psychotherapy in general. Thus, research in family therapy training needs to identify and confirm the effectiveness of those variables specific to family therapy in order to conclude that this training is any more effective than traditional training in psychotherapy.
However, the goals of training and supervision and the skills of
the supervisor are dependent upon the theoretical orientation of the particular training program involved. The experientially oriented (Constantine, 1976; Ferber & Mendelsohn, 1969; Luthman & Kirschenbaum, 1974) and psychodynamically based programs (Ackerman,
1973; La Perriere, 1977) tend to emphasize the personal growth aspects of training and the affective experiences of the trainees.
For example, two studies which attempted to isolate changes in psychiatric residents undergoing psychodynamic family therapy training were based on trainee self-reports (Flint & Rioch, 1963; Schopler, Fox & Cochrane, 1967). Based on these self-reports, both
investigators concluded that the trainees gained increased awareness and appreciation for family dynamics. However, as Orlinsky and Howard (1978) point out, self-report is often unreliable due to the investment the individual has in perceiving change. In addition, since external judges were not employed and the variables being judged were often poorly defined, these results must be interpreted with caution. Self-report should not, however, be ignored, but should be reported in conjunction with more objective measures of change.
In those programs that operate more from structural (Minuchin, 1974), behavioral (Cleghorn & Levin, 1973), and strategic (Haley, 1976) therapeutic orientations, goals are more cognitively based and focus more on defining particular sets of therapist skills and ways of intervening into dysfunctional systems. The current trend in the family therapy training literature is in the direction of establishing operationally defined objectives and therapist competencies according to differing theoretical schools of thought (Garrigan & Bambrick, 1976).
Within the behavioral and structural schools, Cleghorn and Levin (1973) have been influential in proposing operational objectives for
the assessment of training in family therapy. According to their model, therapist skills may be classified into three groups: perceptual, conceptual, and executive. Perceptual skills refer to what the therapist observes in the therapy session, how the therapist
perceives interactions, and the meaning and effect of them on family members and the family system. Conceptual skills basically involve what the therapist thinks about in the therapy session and how those thoughts are organized. Cleghorn and Levin argue that in order to formulate a problem in terms of the family system, one must specify the implicit rules of behavior in the family that make sequences of family interactions predictable. In addition to the perceptual and conceptual skills required of the family therapist, he/she must also
develop skills that influence the family to demonstrate the way it functions. Cleghorn and Levin have labeled these skills as executive skills. Executive skills involve what the therapist says and does in the therapy session in order to influence the family's sequences of transactions and thus alter the way the family functions. The
present study will attempt to assess the process of change in these three skills delineated by Cleghorn and Levin (1973).
This descriptive model proposed by Cleghorn and Levin (1973) has yet to be empirically tested, although several articles on family therapy training have based their discussions on this model (e.g., Barton & Alexander, 1977; Woody & Weber, 1984). Although the current trend is to objectify the skills of family therapy trainees, the majority of the family therapy training literature continues to be based on an impressionistic perspective (Tomm, 1980). For example, a
review of the training literature revealed tnat studies either described training methods based on the clinical observations of the trainer (e.g., Aponte & Van Deusen, 1931; Beal, 1976; Ferber & Mendelsohn, 1969; Nichols, 1979), described the historical development of marital and family therapy training (e.g., Constantine, 1977; Nichols, 1979; Tucker, Hart, & Liddle, 1976), or provided a sociological comparison of supervision methods based on trainee self-reports (e.g., Tomm & Leahey, 1980). Thus, rigorous empirical investigations are required, including the addition of
appropriate control groups, adequate number of subjects, multiple objective measures, and operational ization of variables that are specific to family therapy.
A popular method of assessing trainees' knowledge of family therapy course content and theory involves paper and pencil tests such as multiple choice questions and essay questions (Friedman, 1971; Tomm, 1980). Friedman reported that mental health professionals significantly increased in factual and theoretical knowledge between pre- and posttraining tests. Similarly, Tomm
reported that first year medical students demonstrated significant increases in their knowledge of a Family Categories Scheme devised by Epstein and his associates following their training experiences (Epstein, Sigal & Rakoff, 1968). An additional method popular in the
family therapy training literature involves assessing changes in trainees' attitudes. Poelstra and Lange (1978) and Lange and
Brinckman (1976) reported that trainees' attitudes shifted significantly towards acceptance of behavioral family therapy as a result of training in this model. In addition, Flomenhaft and Carter
(1974; 1977) reported that mental health professionals trained in family therapy reported a significant increase in the amount of time spent in family therapy over individual therapy. Although these findings suggest that training leads to an increased knowledge of
course content and an acceptance of new theoretical positions, these investigations actually offer little more than an assurance that professionals are capable of learning new concepts and are more favorably disposed to those concepts with increased familiarity.
Several studies have attempted to measure changes in the
conceptual and perceptual skills of family therapy trainees by having them respond to videotaped segments of family therapy sessions as if they were the therapists (e.g., Friedman, 1971; Lange & Brinckman, 1976; Lange & Zeegers, 1978; Tomm, 1980). In each of these studies, the trainees' responses were coded by independent judges using preselected coding schemes. The majority of these studies reported
significant increases in trainees' perceptual and conceptual skills.
A third method for evaluating the therapy skills of family therapy trainees is the family genogram measure proposed by Tom (1980). Using this method, trainees are required to determine the
level of family development and associated family developmental tasks from genograms that are provided to them. Tomm reported significant increases in both perceptual and conceptual skills based on the
Family Categories Schema (Epstein et al., 1968).
However, there are two major limitations of the empirical
research discussed so far. First, these research designs did not
include comparable control groups and thus any changes in the trainees may be attributable to factors other than the training program (e.g., spontaneous improvement, maturation, attention-placebo effects, Cook & Compbell, 1979). Second, the variables which have
been selected for evaluation measure whether or not trainees have assimilated the instructional material and not whether their therapeutic skills have been influenced by training. In addition, those studies that do assess therapeutic skills are not specific to family therapy, but may be generalized to any form of psychotherapy.
In a recent attempt to eliminate these two major limitations, namely, the lack of a comparable control group and the use of paper and pencil tests of increased knowledge of training, Breunlin et al. (1983) reported the development of an instrument designed to measure the perceptual, conceptual and executive skills of family therapists. This instrument consists of a videotape of an enacted family's first session and a series of multiple choice questions regarding the subjects' perceptions, conceptualizations and therapeutic recommendations about the tape. The experimental
subjects consisted of 22 psychiatric residents who were given one month of family therapy training, and the control subjects consisted
of 11 pediatric residents who were not given family therapy training or any formal training in psychotherapy. A pre-post assessment revealed a significant increase in conceptualization skills for only the family therapy trainees. However, there were no changes in
either perceptual or executive skills for either group. Breunlin et al. (1983) suggested that their instrument may not have been sensitive enough to detect change in skill level.
The study of Breunlin et al. (1983) represents a considerable advance over previous investigations; however, there are several flaws in their methodology that need to be addressed. First, the
simulated therapy tapes showed a therapist interacting with the family. It might be expected that viewing another therapist might influence the responses of the subjects, thus confounding the results. Further, the authors do not adequately describe how subjects' responses were coded, or along what dimension. Third, the actual construction of the tape is not adequately described. Four family therapists reviewed the tape and arrived at a consensus that it was representative of the structural -strategic model. Although this procedure may provide a high level of content validity, the reliability of this measure is open to question. Fourth, their control group was not comparable for level of therapeutic sophistication. It is logical to expect that psychiatric residents would show greater levels of conceptual skill concerning family therapy dynamics than a group of pediatric residents with no therapy training. Finally, prior to developing multiple choice responses to
the simulated tape, open-ended test responses were subjectively analyzed and determined to have substantially improved in both complexity of answers and correct application of training knowledge from pre- to posttest. Again, there was no mention of any objective coding system and thus the results may have been biased toward the
logical expectation that trainees' responses would become more complex after training.
More promising have been recent advances in the development of objective instruments to assess behavioral changes of family therapists. There have recently been two analysis systems developed that target therapist executive skills behavior (e.g., Allred & Kersey, 1977; Pinsof, 1977). Both of these measures have been shown
to demonstrate appropriately high levels of reliability and validity. Pinsof (1977) provided empirical evidence that the therapeutic verbal behavior of novice and advanced family therapists can be coded reliably and shown to be significantly different. At the present time, the Family Therapist Coding System (FTCS, Pinsof, 1977) is the only family therapist verbal coding instrument that has provided researchers with the ability to distinguish among trainees according to their training level (Tucker & Pinsof, 1984). However, there is a major limitation involved in the use of this measure. Due
to the complexity of the instrument, a considerable amount of practice administration is required in order to ensure reliable measurement and therefore the use of the FTCS is often prohibitive.
The second measure, the Allred Interactional Analysis for
Counsellors (AIAC, Allred & Kersey, 1977), has also been shown to differentiate among trainees' levels of training. Several studies have reported this measure of verbal behavior to be both highly reliable (Kersey, 1976; Sanders, 1974; Watson, 1975) and valid (Kersey, 1976; Sanders, 1974; Tripp, 1975; Watson, 1975). Compared
to the FTCS, the AIAC is much less complex and produces reliable results after a short period of coding practice.
This measure has recently been used to compare the executive skills of family therapy trainees with those of nonfamily therapy trainees. Zaken-Greenberg and Neimeyer (1934) investigated the executive skills of family therapy trainees and comparable controls in a pre-post design using a modified version of the AIAC. Subjects
were required to observe videotapes of a simulated family therapy session and instructed to assume the role of the primary therapist. At different intervals, a family member looked into the camera and asked the therapist (i.e., the subjects) a question to which the
subject was to respond to in writing. Each response was rated by two independent judges on a modified version of the AIAC. The original categories were broken down into more refined and mutually exclusive categories ordered along a dimension from obstructive responses through levels of functional family therapy responses (e.g., relationship building, structural realignment). An overall average of response level was obtained.
The results of this investigation revealed that family therapy trainees responded with significantly more structural maneuvering techniques than did the control subjects at posttest. In addition, the family therapy trainees' use of these responses increased significantly from pre- to posttest with no change indicated for the control group. Zaken-Greenberg and Neimeyer (1984) suggested that the structural training emphasis on behavioral manipulation of the
family influenced the way in which trainees revised their executive skills when responding to a family.
A major problem with this approach is the unreliability
contributed by the judges. This approach is also weakened by the difficulty in scoring and the lack of standardization. As Breunlin et al. (1983) point out, a multiple choice format in which subjects
choose an alternative in response to a simulated videotape constitutes a reasonable compromise in that it can reliably measure therapist skills within a standardized and easily scramble methodology. The present study will use a Breunlin et al. (1983) paradigm, combined with a modified version of the AIAC, to assess and compare the executive skills of family therapy trainees and individual therapy trainees.
After reviewing this literature, it becomes clear that the
research in family therapy training is only beginning to contribute a firm empirical grounding for the exploration and assessment of the effectiveness of family therapy training. In order to provide a
comprehensive literature capable of integrating the disparate findings, there needs to be a unified methodology that allows for comparison across the different established training programs as well as across the varying theoretical orientations. In addition, this methodology must include comparative measures of family and nonfamily therapy trainees. If there is no difference between individual therapy experience and family therapy experience in producing effective family therapists, then the skills proposed for family therapy need to be reevaluated. The present study suggests the use
of Kelly's (1955) personal construct approach as a basis for the empirical study of family therapy training.
Personal Construct Theory
George Kelly's (1955) personal construct psychology has been
described as a comprehensive theoretical framework for the study of human personality (Mancuso & Adams-Webber, 1982). Personal construct
theory takes the position that an individual's interpretation of the world is explained in terms of that person's organized system of personal constructs. Each person has a unique way of construing the world, and people are similar to each other to the extent that they construe events in a similar way. Thus, the personal construct is the basic unit of analysis in this system, and was originally defined as "a way in which some things are construed as alike and yet different from others" (Kelly, 1955, p. 105). Thus, the construct is a bipolar dimension upon which a decision is made between a pair of alternative acts (e.g., a family therapist may construe a family as It rigid versus permissive").
Kelly based his psychology of personal constructs on a
fundamental postulate and 11 corollaries. This fundamental postulate states that "A person's processes are psychologically channelized by the ways in which he anticipates events" (Kelly, 1955, p. 46). Any act or behavior can be viewed as hypotheses which are continuously validated or invalidated by experience. However, a validated hypothesis does not imply that an event has only one interpretation. Kelly's theory argues that a totally different
interpretation may have led to an equally successful prediction. This concept, labeled "constructive alternativesm" is an important reason why personal construct psychology lends itself so well to the investigation of process-outcome variables. In family therapy training, for example, each theoretical orientation perceives the process of change differently, and yet the outcome (i.e., an effective family therapist) will hopefully be the same. Even within the field of family therapy itself, the therapist may construe a family differently and through the therapeutic process bring that family to the same outcome; a change from dysfunctional behavior to
Personal construct psychology has recently been extended to the area of family therapy. Although this theory was designed for individual psychotherapy and intrapsychic processes, Procter (1981)
argues that it can easily be extended to the area of multiperson relationships. There are two important corollaries that allow for this extension. The first is the sociality corollary, which states that, "To the extent that one person construes the construction process of another, he may play a role in a social process involving the other person" (Kelly, 1955, p. 95). This corollary states that a person makes choices or behaves in accordance with how he anticipates that another will respond, regardless of the accuracy of his construal. Thus, the sociality corollary allows for an understanding of the complementary aspects of dyadic relationships. In order to
look at families, Procter introduced the group corollary which states that, "To the extent that a person can construe the relationships
between members of a group, he may take part in a group process with them" (1981, P. 354). This new corollary is applicable to training as the trainee must learn to construe the relationships between the members of the family and himself if he is to take part in a group process with them.
The second corollary that extends to multiperson relationships involves the choice corollary, which states that, "A person chooses for himself that alternative in a dichotomized construct through which he anticipates the greater possibility for extension and definition of his (construct) system" (Kelly, 1955, P. 64). Kelly further emphasized that choices are not necessarily conscious or verbalized. An extension of this logic is Procter's family corollary, which states that "For a group of people to remain together over an extended period of time, each must make a choice, within the limitations of his system, to maintain a common construction of the relationships in the group" (p. 354). Procter (1981) argues that a negotiation of a common family reality occurs in families, which he labeled as the family construct system (FCS). This family construct system involves a hierarchically organized set of family constructs used by the family members to make choices and anticipations. The family therapy trainee must somehow learn to perceive and conceptualize the common family reality (i.e., FCS), and through his executive skills, to change that reality. According to Procter, this will lead to new understandings about what has occurred in the past,
thus allowing the family to negotiate a new construction.
A review of the family therapy and training literatures suggests that family therapists either restrict themselves to one theoretical school, or find themselves in a confusing eclectic mixture of theories and techniques. Haley (1976) has stated that a direct synthesis of the various theoretical orientations in family therapy is not possible due to the fundamental differences in their basic postulates. However, common to both family therapy and family therapy training is the process of change. Kelly explained change as a process of reconstruction. According to this view, reconstruction involves a change in the linkages between constructs, the addition of new constructs, the elaboration of subsystems or a change in the range of convenience (i.e., inclusiveness of a particular construct dimension) which the construct system covers. Thus, family therapy training consists of giving the trainees experiences which enable them to elaborate and revise their own construct system such that they are capable of objectifying the constructions of the family and expediting the change process. It may be argued that personal construct psychology subsumes the different approaches in a theoretical framework that lies outside and at a greater level of
abstraction to the various present alternatives (Procter, 1981). Further, personal construct theory allows a synthesis of various theoretical alternatives without any compromise of theoretical rigor and precision. This provides a strong rationale for the use of personal construct psychology as a method to investigate the training of family therapists.
Personal Construct Psychology and Training Studies
With the increased recognition of personal construct psychology
as a viable context within which to study human behavior and relationships (i.e., systems of personal constructs), there has been a burgeoning of research based on Kelly's Role Construct Repertory Grid (Rep Grid). This Rep Grid measure has been adapted to the study
of a variety of issues ranging from analyses of individuals with psychiatric disorders to the perceptions of therapists, friends, marital couples, and trainees in various fields. Many of these studies have focused on a concept labeled cognitive complexity which Bieri (1955) has defined in terms of the differentiation of (i.e.,
the number of constructs in) an individual's construct system. Basically, Bieri argued that as the construct system develops it becomes more differentiated. This notion of progressive differentiation has been investigated in many training studies based
on Kelly's (1955) organization corollary, suggesting that a construct system is hierarchically organized and undergoes shifts over the course of personal and professional development.
However, several studies investigating the training of
professionals have suggested that the relationships among the constructs undergo periods of decreased complexity during the early
phases of training, followed by a reorganization of the constructs at higher levels of complexity during advanced training stages (e.g., Baldwin, 1972; Pope, 1978; Runkel & Damrin, 1961; Ryle & Breen, 1974). These findings offer some suggestions concerning the
conceptual difficulties that family therapy trainees often report during the early phases of training (Gurman & Kniskern, 1978). According to personal construct psychology, a person develops new ideas through shifts in construction. This shift in construction,
known as the Creativity Cycle, starts with loosened construction and terminates with tightened and validated constructions. Constructs must be loosened in order that hypotheses may be tested, and then they must be tightened after the hypotheses have been either validated or invalidated so that they may be acted upon. Extending this logic to family therapy training, if the trainees' constructs pertaining to families become less organized during early phases of training such that the relationships among the constructs decrease,
then efforts to interpret family behavior would be based on constructs that are loosely related and lacking meaningful connections. Consequently, trainees may experience difficulty in formulating their conceptualizations of families, resulting in the temporary decline in trainee performance cited in the training literature (Duhl, 1978; Gurman & Kniskern, 1978).
Contrary to the above findings, Adams-Webber and Mirc (1976)
reported no decrease in construct relationships during the early phases of training. These findings were based on "integration" scores that provided estimates of the overall degree of intercorrelations among supplied constructs (e.g., activities of educational personnel). The results indicated that there was a
gradual increase in the level of integration (i.e., cognitive complexity) of "teacher role" subsystems as the trainees acquired
classroom experience. These findings may indicate that integration either increases in an irregular fashion during periods of training,
or actual experience accelerates the integration and tightening of new constructs. However, these discrepant findings may also be the result of using different measures than Pope (1978), Ryle and Breen (1974), and Runkel and Damrin (1961) to estimate cognitive complexity. The present study will use pre-post measures of
cognitive complexity based on both differentiation and integration scores as a method of investigating the cognitive complexity issue in training.
Crockett (1982) has suggested that a second major feature of a developing construct system involves the appearance of abstract constructs. One method used to assess the development of abstract constructs involves a content analysis of the construct dimensions elicited by the Rep Grid. Lifshitz (1974) investigated whether constructs became more abstract as a consequence of training. This study used factor analysis to identify the construct with the highest loading on the first factor in the grids of social work trainees and their supervisors. Each construct was then assigned to one of seven
categories and the distribution across the categories was examined. The results indicated that the trainees used more concrete
descriptive categories while their supervisors showed more abstract constructs of themselves, others, and their work. An increase of more abstract constructs in the construct subsystems of trainees during training was also revealed. However, this study provided
content information about only one construct in each trainee's
construct system. The present study will investigate a larger sample of the constructs in the training subsystem in order to evaluate the hypothesis of increased abstraction over the course of training.
Kelly's Rep Grid methodology has only recently been applied to the area of family therapy training. Lewis (1977) conducted a study of family therapy trainee's perceptual and conceptual abilities before and after a twelve-hour structural family therapy workshop. Basically, Lewis examined how trainees structured their thoughts about families and the type of constructs that they used when thinking about families. This study examined both the structural and content changes of family therapy trainees. Lewis used a structural measure devised by Landfield (1971) called ordination. Ordination is a measure of the hierarchical organization of the trainees' system of thoughts about the elements under investigation (e.g., families). Lewis also employed a measure called the functional independence of
constructs (Landfield, 1971) that measures the degree of uniqueness between family constructs, along with Bonarius' (1970) measure of the meaningfulness of the constructs to the individual. The results showed a significant change from pre- to posttraining, indicating
that the structure of the trainees' thoughts about families had become more cognitively complex. Further, a content analysis of the
trainee's family constructs revealed that a significant number of trainees increased their use of interactional family constructs. Interactional constructs were considered to be the most abstract in the family construct subsystem as they impart more information about
a family and thus allow for the generation of more hypotheses about the family system.
In contrast to the findings reported by ewis (1977), ZakenGreenberg and Neimeyer's (1984) investigation of the structure and content of family therapy trainees' family construct systems revealed that training in family therapy was generally associated with lower levels of differentiation and integration. These results suggest that such training initially channelizes and rigidifies perception of the family. This interpretation is consistent with the related body of work which has shown that the construct system undergoes periods of decreased complexity during the early periods of training,
followed by a reorganization of the constructs at higher levels during advanced training stages (Baldwin, 1972; Pope, 1978; Ryle & Breen, 1974).
Zaken-Greenberg and Neimeyer also reported a concomitant
decrease in the level of abstraction of the constructs themselves. These authors suggest that this content shift toward more concrete,
descriptive construing that occurred in both groups may reflect the tendency for trainees to favor more behaviorally evident constructs (e.g., single-parent family vs. two-parent family) over more abstract ones (e.g., undifferentiated family ego mass) as they are confronted with the need to effectively understand actual clinical experience.
The effect of this gravitation towards concrete dimensions might be to reduce the structure of the system since the dichotomous nature of many concrete descriptions (e.g., dual career vs. single career family) prevents their flexible, and hence more integrated, use in
the system. The present study will use both structural and content analyses of family rep grids in an attempt to examine any changes in cognitive complexity or level of abstraction as a result of structural family therapy training.
The Present Study
Kniskern and Gurman (1979) suggest that the assessment of the outcome of training procedures in family therapy should be made on several dimensions. First, it is important to assess if the trainee increases his/her conceptual knowledge of family dynamics and interaction patterns. The second major area for the assessment of change involves the trainees' in therapy behavior (i.e., executive skills). Third, it is important to evaluate if the trainees' perceptions of families have changed as a function of training. The present study will address all of these points.
This study is based on a structural training foundation. The subjects have been divided into two groups: (1) an experimental group consisting of 12 graduate students in counselor education and counseling psychology undergoing a 16-week (45 hours) training program in structural family therapy, and (2) a control group consisting of 21 graduate students in counselor education and counseling psychology undergoing individual psychotherapy training.
These control subjects have not taken any graduate courses in family therapy. Participants in the present study completed (1) a family therapy experience inventory, (2) a family repertory grid, (3) a multiple choice questionnaire made up of therapist response
alternatives in response to a series of simulated-family therapy videotapes (with the response alternatives designed to reflect the classification system of the Allred Family Interaction Scale), and
(4) a multiple choice questionnaire designed to measure the case conceptualization of the simulated family dynamics and treatment intervention, both before and after the 16-week interval.
There are four basic components that are being analyzed in this study. The first is based on the family repertory grid which consists of a clinical interview in which participants systematically
compare and contrast various families to determine dimensions representative of their evaluations of family dynamics. A structural analysis (FIC and chi square scores, see Landfield, 1971) will provide measures of cognitive complexity. Based on the training
literature reviewed, it is hypothesized that family trainees will significantly increase in cognitive complexity after training compared to the control group.
The second component of this study involves a content analysis of the repertory grid data. The content analysis is based on an adaptation of Duck's (1973) procedure. Basically, constructs are coded as either physical, role, psychological, or interactional. A
content analysis will provide measures of abstraction and it is hypothesized that there will be a significant increase in the use of
abstract (i.e., interactional) constructions in the experimental
group. It is expected that the control group will remain at
consistent levels at pre- and posttesting.
The third component of this study will investigate trainees' conceptual skills in three major areas. These areas include a conceptualization of the family, a conceptualization of the intervention, and a systemic dimension (e.g., focusing on an individual vs. the entire system). Subjects will be required to
answer eight multiple choice questions that reflect these three areas of interest. Based on the construction corollary, Kelly (1955)
stated that a person's anticipations are a matter of ascribing certain recurrent themess" as he successively construes events. This logic suggests that recurrent themes for family dynamics are learned in training. Further, personal construct psychology suggests that training should lead to more successful prediction of the experience of the family in treatment, as well as creating shared meanings among family therapists. Case conceptualization analyses will indicate if the family therapy trainees are capable of conceptualizing family
dynamics, predicting family behaviors, prescribing treatment, and describing these constructions of families using the shared meanings of structural family therapists. It is hypothesized that these conceptual skills will increase for the family therapy trainees, and remain stable across time for the nonfamily therapy trainees.
One of the most important skills of family therapists involves executive skills. The family therapist must be capable of translating perceptual and conceptual skills into active behavior
that will expedite change in the family rather than result in a
deterioration in family behavior and interactions. The present study will investigate whether structural training aids in increasing these executive skills through the use of a set of family therapy videotapes consisting of 20-minute simulations of family therapy sessions in which the viewer is oriented as the therapist. The sessions is halted periodically, and participants are asked to select a preferred response as primary therapist to the family. Measures of response effectiveness will be obtained by coding these responses
according to a modified version of the Allred Family Interaction Scale. It is hypothesized that there will be an increase in family therapy trainee executive skills as compared to the control group.
Each dependent measure will involve a 2(Group) x 2(Time) mixed factorial design. The results of these analyses will test the predictions that, compared to controls, trainees will become more complex (differentiated and integrated) and systemic in conceptualizing family dynamics, and more effective in therapeutic responding to simulated cases. Such findings would suggest that training in structural family therapy influences both the quality and
effectiveness of therapeutic interventions in a manner consistent with the objective of the training program.
In summary, this dissertation will investigate family therapy training by the personal construct methodology. There has been a dearth of empirical research in this area, with most studies being either descriptive in nature or lacking empirical rigor and sophistication. The present study provides a sounder empirical basis for investigating the training of family therapists.
Thirty-three subjects were recruited voluntarily from the
counselor education and counseling psychology graduate programs at the University of Florida in Gainesville, Florida. The experimental
group consisted of 12 graduate students currently enrolled in a graduate seminar in structural family therapy given through the counseling psychology program (see Appendix A for a description of the seminar). This group consisted of four males and eight females, with a mean age of 32 (range = 23 to 41). The control group
consisted of 21 graduate students from the identical programs who had not had previous training in family therapy and who had not yet taken the graduate seminar in structural family-therapy. There were seven
males and fourteen females in the control group, with a mean age of 29 (range = 21 to 49). The experimental subjects had an average of
four semesters of graduate school and the control subjects had an average of two semesters of graduate school. Thus, both groups are comparable in their graduate training, with the experimental group receiving additional training in structural family therapy.
Family Experience Inventory
Each subject was required to complete a modified version of the Family Therapy Experience Inventory (FTEI) in order to assess any contact with family therapy prior to the current family training seminar (Connors, 1984). This modified inventory provides information on age, sex, educational level in their graduate program, and amount of prior counseling experience (see Appendix B). Connors (1984) originally developed this inventory in order to assess both
the type of prior exposure with family therapy as well as the extent of that experience. There are three major categories of prior exposure. The first involves prior training level in family therapy and is determined by the number of courses taken where there was a major focus on the "family," as well as any training sessions or workshops that focused on work with families. The amount of prior
reading in family therapy comprised the second category of reading (e.g., the number of books or articles read on family therapy). The
third category, experience, consisted of direct experiences with family therapy, and the number of these cases that were supervised. Table 1 presents the mean family therapy experience scores for experimental and control subjects. In the present investigation, there are no significant differences between the two groups for prior exposure to some form of family therapy training.
Table 1. Mean Family Therapy Experience Scores for Experimental and
Group N Courses Workshops Books Articles Family SuperTaken Attended Read Read Therapy vised
Experimental 12 1.82 1.36 2.09 2.00 2.00 1.82
(.78)* (.67) (1.22) (.45) (.63) (.40) Control 21 1.90 1.57 1.62 2.24 1.29 1.24
(.83) (1.16) (1.20) (.89) (.40) (.44)
Numbers in parentheses are standard deviations.
Family Repertory Grid
The Family Repertory Grid (Connors, 1984) is a modified version of Kelly's Role Construct Repertory Grid (Rep Grid), originally devised as a measure of different aspects of the structure and content of individuals' personal cognitive systems. When using this
personal construct methodology, elements are selected that represent the cognitive constructs that are to be investigated. The elements chosen for the present study were family role titles adopted from Connor's (1984) Family Repertory Grid. These elements have been designed to represent various types of families that are familiar to most individuals (e.g., the family you grew up in; a happy family; a single parent family). The present study is based on the minimum
context card form in which the subjects are required to give names to ten preselected elements (i.e., family role titles), and then write
the names on index cards numbered from one to ten. The subjects are
then presented with three of these elements written on the index cards and asked to specify some important way in which two of the families are "alike in some way and yet different from the third family." For example, after they respond with how two of the elements are alike, they are required to specify "in what way the third family differs from the other two families." Following the
elicitation of the constructs, the subjects are required to rate each family along each construct dimension using a 7-point Likert type scale (see Appendix C).
Kelly's repertory grid procedure has been used extensively in
research on various issues related to personality theory and clinical applications. As a result of this procedure's increasing popularity,
attention has been directed toward obtaining psychometric data, with special reference to reliability and construct validation. Generally, reliability coefficients based on the consistency of
ratings across time and the consistency of constructs across time (given identical elements in the repertory grid) have been found to be relatively high. For example, Hunt (1951) elicited constructs to
fit 41 role titles by the triadic method and reported a test-retest reliability of .70 over an interval of a week. Similarly, Fjeld and
Landfield (1961) repeated Hunt's basic design and demonstrated that, given the same elements, there was a correlation of .80 between the two sets of elicited constructs over a two week interval. Validity, however, has proven much more difficult to assess. The repertory
grid is not a static assessment tool in that it may be adapted to a
variety of content areas (e.g., schizophrenia, depression, suicide, education and training). Consequently, validity studies have been generated in these various content areas. In general, however, Fransella and Bannister (1977) argue that the validity of this
technique has been shown to be acceptable, as the grid has been demonstrated to discriminate between different diagnostically defined groups as well as between before and after treatment groups.
The Family Repertory Grid provided data for the first two
components of this investigation (i.e., cognitive complexity and
construct content). A structural analysis of the subjects' ratings provides measures of cognitive complexity, and a content analysis of their constructs provides measures of construct abstraction.
Structural grid measures
Cognitive integration. Based on a cluster analysis designed by Landfield (1971), the family repertory grid ratings result in
ordination scores that are basically an integration measure of the hierarchical organization of the trainees' system of thoughts about the family elements. This Landfield analysis also provides chi
square scores that measure integration and level of flexibility in the use of constructs. These chi square scores measure the extent to which the personal constructs are arranged into interrelated systems of constructions. According to Landfield and Schmittdiel (1983), the
chi square score is a more reliable measure of hierarchical potential than the ordination score.
Cognitive differentiation. Structural analyses also provide FIC scores. The FIC scores reflect the functional independence of constructs which measure the degree of uniqueness between family constructs. Landfield (1971) originally developed the FIC score as a measure of cognitive differentiation such that an individual who is
highly differentiated has a large number of different constructs that are used in different ways. By using constructs in different ways, one is capable of viewing a situation from a number of different perspectives. Thus, FIC scores provide information on the level of differentiation (i.e., how many different ways are the constructs used). According to this system, differentiation and integration provide indices of cognitive complexity, such that higher levels of differentiation and integration imply a cognitively complex individual.
Content grid measures
The second component of this study involved a content analysis of the family repertory grid constructs elicited from the subjects.
The content analysis was based on an adaptation of Duck's (1973) procedure in which constructs are coded as either physical, role, psychological, or interactional. Based on the pilot data and feedback from both the subjects and the examiners, it was determined
that independent judges' coding of the subjects' construct content was often misleading and inconsistent. Subjects frequently provided
constructs that fit into one of the category codes (e.g., psychological), while also making comments to the examiner concerning
what the construct implied (e.g., discussing the interactional properties of the construct). Consequently, the manner of coding the constructs was not always consistent with the inferred meaning the constructs had for the subjects. In view of the marginal interrater reliabilities (range = .56 to .82), the present study was designed to
exclude judges, and to base the content coding on the subjects' perceptions of their own constructs (see Appendix D). In addition, the coding label "psychological" was changed to "personality" since it is believed to be less abstract in its meaning, and therefore more
familiar to the average subject.
A test-retest reliability analysis was performed using this new
procedure in order to establish a strong empirical rationale for its use. One-hundred-thirty-four undergraduate volunteers were given a repertory grid measuring their perception of various types of interpersonal relationships. These subjects were required to use Duck's (1973) coding system to code their own constructs. After an 18-day delay, subjects were again asked to code their constructs. With no training or prior exposure to this sytem, an analysis of the
data revealed a 73% agreement across time, suggesting that the selfcoding procedure is as reliable as using independent judges. A major advantage to using this procedure is that the data will be a more
accurate reflection of the subjects' perceptions of their personal constructs.
Subjects in the present investigation were provided with
instructions and examples of how to use the coding system. They were given five constructs to code using Duck's system, and were corrected
by the experimenter if they made an error. Any subject who made two or more coding errors was given five more constructs to practice on. No subjects made any errors on the second set of trial constructs (see Appendix E). Analyses of content provided measures of abstraction (i.e., shifts in the use of personality and interactional constructs as a function of training). Therapist Responses to Simulated Family Tapes
Subjects were required to observe videotapes of a simulated family therapy session, and instructed to assume the role of the primary therapist. At four different intervals a family member looked into the camera and asked the therapist (i.e., the subjects) a question to which the subject was to respond by selecting one of several provided multiple choice answers.
There are two 20-minute simulated-family videotapes that were used in this study. These tapes are based on the dynamics of
simulated families used by Connors (1984) to investigate family therapy training. Student volunteers from an undergraduate psychology course and research assistants in the family therapy training project made up the simulated families. Each "actor" received a written description of the family dynamics and their role in particular which served as a guide for their performance (see Appendices F and G). They were given the family dynamics the evening before taping, and were allowed one practice trial in order to feel comfortable with their role. The actors were signaled when they were
to speak into the camera and ask the "therapist" a question concerning the family.
The family scenarios consisted of disturbances in the family structure which required intervention by the family therapist. The first family is composed of two parents, two adolescent daughters and two young sons. The oldest daughter is presented as the identified patient by the parents, the mother as ineffective and the father as peripheral. The second family is composed of a single mother with an adolescent daughter and son. In this family, the son is presented as the identified patient, and the mother focuses on her bitter
relationship with her ex-husband and her inability to control her son's behavior. Mother and daughter are in a coalition against the
Connors (1984) reported that the two sets of family dynamics used in his study were equated for level of difficulty. Level of
difficulty was determined through consultation with professionals acquainted with family therapy. In addition, presentation of the
videotapes were counterbalanced at pre- and posttraining to distribute any possible order effects.
Based upon the results of the pilot investigation, several revisions were made for the improvement of this dissertation. A major difficulty with the pilot study involved the use of naive and unsophisticated judges concerning family therapy. Judges were
undergraduate students without any exposure to family therapy prior to participating in this project. In addition, the pilot data for
therapist responses revealed a paucity of information for the amount
of data collected. Allred's coding system did not prove to be reliably used by the judges (r = .52 at Time 1 and r = .60 at Time
2), requiring that the data be collapsed across categories for a composite score for each subject. This composite score simply
reflected what direction subjects were moving toward in terms of their level of executive skills. Thus, the present study was revised
to both eliminate the unreliability contributed by the judges, and to provide a greater array of information concerning the executive skills of the two different training groups.
The revisions for the dissertation study involved providing
subjects with alternative therapist responses to choose from at each of the four intervals. These responses represent the categories found to be meaningful in the pilot study. There were four main
categories used for coding therapist responses: Gathers Information, Supports and Reflects, Interprets and Confronts, and Structural Real ignment. In order to represent each of the main categories, four alternative therapist responses were generated within each category. These responses represent a dimension from obstructive to the most sophisticated response level within that category. For example, the category of interprets and confronts has four alternative responses ranging from obstructive +1, direct
interpretation = +2, strategic interpretation +3, and confrontive interpretation = +4 (see Appendix H for a description of the coded categories). Thus, subjects are required to answer four multiple choice questions at each of the four intervals. Except for the obstructive responses, all response alternatives were designed to be
appropriate responses to the dynamics of the family tape that the subjects have viewed (see Appendices I and J). Within each of the four response groups, subjects were also required to rank the confidence level they felt concerning their chosen response on a scale of 1 to 10 (very unconfident to very confident). These rankings provide a measurement of perceived confidence across time for both groups of subjects. Subjects were also required to rank their four choices from most to least preferred. The most preferred therapist response was also ranked for confidence level on a scale of 1 to 10. Thus, in addition to measuring whether subjects' confidence changes across time, it is also possible to measure if their confidence about their most preferred response changes across time (see Appendix K).
Two professional structural family therapists in the community
were solicited to evaluate these materials in order to establish empirical reliability. Both therapists were asked to rank order the
response alternative category from least to most preferred therapist response. These rankings were correlated between the two therapists, and between the therapist rankings and the rank ordering generated by the experimenter using a modified version of Allred's AIAC coding system. Therapist responses were based on the actual responses provided by the subjects in the pilot study. For a therapist response to be chosen, it had to be coded identically by the two independent judges. In addition, the actual responses were also
evaluated using the procedure just outlined (see Tables 2 and 3 for reliability coefficients).
Table 2. Family Expert Ratings for Therapist Response Stimuli-Family Tape I
Modified Expert 1 Expert 2
Modified Allred Scheme .61* .45*
Expert I .65*
Table 3. Family Expert Ratings for Therapist Response Stimuli-Family Tape II
Modified Expert 1 Expert 2
Modified Allred Scheme .58* .68*
Expert 1 .76*
Case Conceptual izations
The reliabilities for the case conceptualization pilot data were, at best, marginal (range = .12 to .67). Therefore, this
component of the dissertation was revised to exclude the use of judges. Based on the pilot data, it was determined that there were three major areas of information that were not being adequately tapped. These areas include a conceptualization of the family, a
conceptual ization of the intervention, and a systemic dimension (e.g., focusing on an individual versus the entire system). Rather than have subjects simply write their impressions of the family dynamics, subjects were required to answer eight multiple choice questions that reflect these three areas of interest. Each question was provided with four alternative choices ranging from individual, dyad, subgroup, and system. Four of the questions target the
conceptual ization of the family and four of the questions target the conceptualization of the intervention with the family. Each of the
response alternatives is appropriate to the dynamics of the family videotapes (see Appendices L and M for a description of the response alternatives). Again, two structural family therapists evaluated these questions and response alternatives in terms of the
appropriateness of the responses as well as whether they accurately reflect the systemic dimensions they have been constructed to reflect (see Tables 4 and 5 for reliability coefficients).
Table 4. Expert Ratings of Case Conceptualization Stimuli-Family Tape I
Cases Expert 1 Expert 2
Cases Coding Scheme .38 *53*
Expert 1 563*
Table 5. Expert Ratings of Case Conceptualization Stimuli-Family Tape II
Cases Expert 1 Expert 2
Cases Coding Scheme .33 .69*
Expert 1 569*
All subjects were scheduled by phone for a testing time at their convenience. Subjects were tested in small groups ranging from two to four subjects depending on their availability. Each subject was told that he/she would be participating in a family therapy training study that would last approximately two hours at both pre- and
Subjects were met in an outer office by the experimenter and led into a small testing room containing a portable video monitor, a table to write on, and several chairs. Each subject was requested to sit at a table facing the video monitor. All subjects received identical instructions regardless of whether they were in the control group or the experimental group.
The first task required of the subjects was reading and signing the informed consent form (see Appendix N), followed by the completion of the Family Therapy Experience Inventory. This was followed by instructions for the first phase of the testing session which involved viewing the videotape of a simulated family therapy session. All subjects were given the following instructions:
You are about to view a simulated family in an initial family therapy session. You are to view the family as
though you were the primary therapist and were actually
involved in the session. At different points in the
session, the tape will be stopped, a family member will ask the therapist a question, and you will be provided
with four groups of therapist response alternatives.
From each group of response alternatives, you are to
select the therapist response that you feel you might use
with that family at that point in time. All of the
responses are viable alternatives and there are no right
or wrong answers. We are simply interested in which
responses you would choose from each group of response
After you have viewed the tape and selected your
therapist response alternatives, you will be requested to
answer some multiple choice questions concerning your
case conceptualization of the family as though you were
the primary therapist for the family.
Following these instructions, the subjects were provided with the written family descriptions (see Appendices 0 and P). When the
subjects indicated that they had finished reading the family notes, they were given a therapist response form to use for recording their responses to the family (see Appendix K). The examiner stopped the
videotape at four intervals and requested the subjects to write their responses on the designated pages. The booklet of four therapist response forms was then removed by the examiner and replaced with the
following written case conceptual ization instructions:
We are interested in your case conceptualization of the
family as though you were the primary therapist for this family. There are 8 multiple choice questions regarding
the dynamics that you see as important in this family.
We would like you to choose the alternatives that best reflect how you perceive this family. Again, there are
no right or wrong answers to these questions, each choice simply reflects a preferred style of viewing the dynamics
of the case.
You will have 10 minutes to answer these questions.
When the subjects indicated that they finished reading the instructions, they were given the case conceptualization questions and a form to use for recording their responses (see Appendix Q).
Upon completion of this task, subjects were administered the
family repertory grid. This task consists of a structured interview in which subjects systematically compare and contrast various families to determine dimensions representative of their evaluations of family dynamics (see Appendix R for a complete description of administration). The administration of this family repertory grid
involves the experimenter asking the subject to give names to role titles (e.g., the name of the family you grew up in, the name of a happy family, etc.). The subjects are then presented with three of
these elements and asked to specify "some important way in which two of the families are alike in some way and yet different from the third family." After they respond with how two of the elements are alike, they are required to "give a word or short phrase that describes how the third family is different from the other two."
Following the elicitation of the constructs, subjects are required to rate each pole of the construct as either positive or negative. The experimenter then takes these descriptions and uses them as end points on a 7-point Likert type scale (e.g., Happy +3, +2, +1, 0, -1,
-2, -3 Sad). Subjects then use these construct dimensions to rate all ten families across each of the 15 construct dimensions.
After the administration of the family repertory grid, subjects
were told that they would be contacted in approximately 10 weeks for another two-hour session. The second session involved the same procedure as the first session, with the exception that all subjects viewed a different simulated family therapy videotape. The presentation of the therapy tapes were counterbalanced to reduce
testing effects. Subjects were then debriefed and told tnat the study was designed to investigate the effects of family therapy training on how counselor-trainees respond to and conceptualize various family problems. They were also given a brief rationale for
undertaking this investigation, followed by a question and answer session if desired by the subject. Subjects were also given the
opportunity to sign up for feedback sessions on their participation in the investigation.
Design and Analyses
This investigation has been divided into four major
components. The first component involves the management of trainees' perceptual skills. Perceptual skills may be measured by using structural grid analyses (see Landfield, 1971) on the following dependent measures: chi square scores (measuring level of cognitive integration) and FIC scores (measuring level of cognitive differentiation). The second component involves the measurement of the level of abstraction of trainees' perceptual skills. This component involves two dependent measures: Abstractness (reflected by a Weighted Average score) and Modal Response. The third component of this investigation involves the measurement of trainees' conceptualization skills for both family dynamics and treatment for the following two dependent variables: Weighted Average and Modal Response. The fourth component of this investigation involves the measurement of trainees' executive skills. Executive skills may be
measured by analyzing trainees' therapist responses to the simulated
family tapes. There are two sets of dependent measures for therapist responses. The first set involves analyses within each category of
interest (e.g., gathers information, reflects and supports) for the following dependent variables: Weighted Average, Modal Response and Average Confidence Level. The second set involves analyses across the four categories of interest on the following two dependent variables: Mode for Preferred Response and Average Confidence for Preferred Response. For all four major components of this design, all dependent measures were analyzed with a 2(Group) x 2(Time) Analysis of Variance with repeated measures on the last factor. Structural Grid Analyses
Subjects' ratings from the 7-point Likert type scales were
cluster analyzed by Landfield's (1971) computerized method. This analysis provided FIC and chi square scores for each subject, thus providing the measures of cognitive complexity previously discussed (see pages 42-43). These scores comprise the dependent measures for
the structural grid data.
Content Grid Analyses
Each of the 15 constructs elicited from the subjects were selfcoded along a dimension of abstractness ranging from physical, role,
personality and interactional characteristics (i.e., physical = +1, role = +2, personality = +3, and interactional = +4). A single score
representing a subject's overall level of abstraction was obtained by averaging the weighted ratings across the 15 constructs. Each
subject was also assigned a score based on their modal response category.
Case Conceptualization Analyses
Therapist conceptual skills were evaluated along a systemic dimension. Subjects were required to respond to eight multiple choice questions which were provided with four alternative choices ranging from individual, dyad, subgroup and system. Four of the
questions targeted the conceptual ization of the family and four of the questions targeted the conceptual ization of the intervention with the family. This design provides for two categories (family
conceptual ization and family intervention) that are analyzed separately for the following dependent measures: Weighted Average and Modal Response. The Weighted Average is an overall average of systemic response level obtained by totaling a subjects' coded ratings and then dividing the total by the number of questions within a category. The Modal Response represents that response level used most frequently within each category. Therapist Response Analyses
Each subject responded to the videotape with multiple choice
questions at four separate intervals. For each interval, the four multiple choice questions reflected a modified classification system designed by Allred and Kersey (1977). The four categories selected were found to be the most meaningful of Allred's system based on the pilot investigation. The four categories reflected by the questions
were Gathers Information, Reflects and Supports, Interprets and Confronts, and Structural Realignment. Within each category there
are four possible response alternatives that represent a dimension from obstructive to the most sophisticated response level within that category (see Appendix H).
Within category analyses
All four intervals of therapist responses were combined for the following three dependent measures: Weighted Average, Modal Response and Average Confidence Level. These three measures were analyzed separately for each of the four therapist response categories. The Weighted Average is an overall average of response level obtained by
totaling a subject's coded ratings and then dividing the total by 4 (i.e., the number of intervals). The Modal Response represents that response level used most frequently within each category across the four intervals. Within each of the four therapist response categories, subjects rated how confident they felt about the response they had chosen on a scale of I to 10 (very unconfident to very confident). The Average Confidence Level was obtained by taking an average across the four intervals.
Between category analyses
In addition to investigating possible changes within each of the
four therapist response categories, two dependent measures were designed to reflect any changes in therapeutic skills or confidence across these four categories. Subjects were asked to rank the four
therapist responses they chose for working with the family in order of preference from the most to the least preferred. That category most frequently endorsed by the subject as their preferred response was labeled as the Mode for Preferred Response. This score provided a measure of response preference across the four different categories. The final dependent measure represents the Average Confidence for Preferred Response. Subjects were required to rank how confident they felt about their most preferred response on a scale of 1 to 10 (very unconfident to very confident). An average
score was obtained across the four intervals.
Structural Grid Analyses
A 2(Group) x 2(Time) analysis of variance for Functionally
Independent Construct (FIC) scores on family constructs revealed a significant interaction, F (1,31) = 12.63, p<~.001 (see Table 6). Followup analyses revealed a significant effect of Time for the control subjects, F (1,20) = 8.42, p<.008, such that the nonfamily therapy trainees became significantly more differentiated in their family constructs across time. Followup analyses also revealed a significant effect of Time for the experimental subjects, F (1,11)
5.25, p<.04. These results suggest that the family therapy trainees became significantly less differentiated across time (see Figure 1).
In addition to looking at levels of differentiation, analyses also revealed some significant findings based on integration scores. Landfield's chi square score is viewed as a measure of integration and flexibility. A 2(Group) x 2(Time) analysis of variance on chi square scores revealed a significant main effect of Group, F (1,31) = 6.27, p<.02, such that the control subjects were significantly less integrated in their family constructs than were the family therapy trainees. An analysis of variance also revealed a trend toward a main effect of Time, F (1,31) = 2.71, p=.10,
Table 6. Analysis of variance for FIC scores
Source Sum of Squares df Mean Square F p
Mean 6127.286 1 6127.286 145.32 .000
Group 31.953 1 31.953 0.76 .391
Error 1307.077 31 42.164
Time 0.091 1 0.091 0.01 .909
Time x Group 71.304 1 71.304 10.17 *QQ3*
Error 217.363 31 7.012
Table 7. Analysis of variance for chi square scores
Source Sum of Squares df Mean Square F p
Mean 141671.093 1 141671.093 483.82 .000
Group 1835.680 1 1835.680 6.27 .012*
Error 9077.315 31 292.816
Time 543.736 1 543.736 2.71 .109
Time x Group 23.829 1 23.829 0.12 .732
Error 6220.378 31 200.657
Q experimental Control
10.0 9.5 I
FIC 8.0 Scores 7.
6.0 5.5 5.0
Figure 1. Interaction for FIC scores
suggesting that both groups tended to become less integrated across time (see Table 7).
Content Grid Analyses
The 2(Group) x 2(Time) analysis of variance revealed no significant findings for Modal Response. However, these analyses did reveal a trend toward a main effect of Group, F (1,31) = 2.94, p=.09 for the Weighted Average score of abstraction such that the family therapy trainees appeared to have higher levels of abstraction in their family constructs than did the control subjects. Analyses also revealed a trend toward a main effect of Time, F (1,31) = 2.94, p=.09, such that both groups became somewhat more abstract across time.
Case Conceptualization Analyses
Family Conceptualization Analyses
Analyses of the Weighted Average score revealed no significant differences in level of systemic conceptualization of family dynamics for the two groups. However, an analysis of variance on the Modal Response scores did reveal a trend toward a main effect of Time, F(1,31) = 3.40, p=.07. The direction of the means suggests that both groups were viewing families along subgroup-systemic dimensions at Time 1 (M=3.51), while at Time 2 they were viewing families along more dyadic-subgroup dimensions (M=3.09). These results suggest that both groups had a tendency to narrow the focus of their family conceptualizations across time.
Family Intervention Analyses
The analysis of variance did not reveal any significant findings for either Weighted Average or Modal Response scores. The mean response levels, however, suggest that both groups focused on
subgroup relationships when conceptualizing an intervention for a family (means range from 2.87 to 3.64).
Therapist Response Analyses
Within Category Analyses
The analysis of variance revealed no significant findings for the Modal Response and Average Confidence scores. However, the results did reveal a trend toward a main effect of Time for the Weighted Average score, F (1,31) = 3.64, p=.06. These results suggest that both groups had a tendency to ask for information about a subsystem relationship less frequently at Time 2 (M=3.15) than at Time l(M=3.39).
Reflects and supports
The analysis of variance revealed a significant main effect of Time for the Weighted Average score, F (1,31) = 9.22, p<.004, such
that both groups had a higher frequency of supportive responses at Time 1 (M=3.15) than at Time 2 (M=2.73) which revealed a higher frequency of reflecting responses (see Table 8).
Table 8. Therapist Response ANOVA for Reflects and Supports Weighted
Source Sum of Squares df Mean Square F p
Mean 522.883 1 522.883 1191.76 .000
Group 0.156 1 0.156 0.36 .550
Error 13.601 31 0.439
Time 3.208 1 3.208 9.22 .004*
Time x Group 0.239 1 0.239 0.69 .410
Error 10.792 31 0.348
An analysis of variance also revealed a significant main effect of Time for Modal Response (see Table 9), F (1,31) = 17.90, p<.001 such that the modal response for both groups at Time 1 was a sophisticated combination response of reflecting and supporting (M=4.00), while the modal response for both groups at Time 2 was a much less sophisticated response involving simple reflection responses (M=2.94). These results suggest that both groups became less sophisticated across time. There was no significant difference in their Average Confidence Level suggesting that they were equally confident across time.
Table 9. Therapist Response ANOVA for Reflects and Supports Modal
Source Sum of Squares df Mean Square F p
Mean 719.377 1 719.377 487.51 .000
Group 1.195 1 1.195 0.81 .375
Error 45.744 31 1.476
Time 21.001 1 21.001 17.90 .001*
Time x Group 2.576 1 2.576 2.20 .148
Error 36.363 31 1.173
Interprets and confronts
The 2(Group) x 2(Time) analysis of variance revealed no
significant findings for Weighted Average, Modal Response, or Average Confidence Level.
The analyses of variance did not reveal any significant findings for Weighted Average, Modal Response, or Average Confidence Level. Between Category Analyses
Mode of preferred category response analyses
The analysis of variance revealed no significant findings for
the Modal Preference score. It is interesting to note, however, that
the preferred category for both groups was the Reflects and Supports
category of therapist responses (Control M=2.86, Experimental M=3.00).
Average confidence for preferred category response analyses
The 2(Group) x 2(Time) analyses of variance revealed a
significant main effect of Time, F (1,31) = 5.60, pcz.02, such that both groups became significantly more confident concerning their preferred responses across time (M at Time 1 = 7.45, M at Time 2
7.94). See Table 10 for the results of the ANOVA table.
Table 10. Therapist Response ANOVA for Preferred Category Response
Source Sum of Squares df Mean Square F p
Mean 3602.506 1 3602.506 1296.49 .000
Group 0.307 1 0.307 0.11 .742
Error 86.138 31 2.779
Time 5.325 1 5.325 5.60 .020*
Time x Group 2.161 1 2.161 2.27 .140
Error 29.476 31 0.951
The primary purpose of the present investigation was to
determine whether or not training in structural family therapy has an impact on the perceptual, conceptual and executive skills of family therapy trainees. This division of therapy skills is based on Cleghorn and Levin's (1973) model for the assessment of training in family therapy. Perceptual skills refer to how the therapist perceives family interactions and the effects of them on family members and the family system. Conceptual skills involve what the therapist thinks about in the therapy session and how those thoughts are organized. Executive skills involve how the therapist actually intervenes in therapy in order to influence the family's
dysfunctional patterns of interactions.
A review of the literature revealed that a majority of family therapy training studies reported significant increases in trainees' perceptual and conceptual skills (e.g., Friedman, 1971; Lange & Brinckman, 1976; Lange & Zeegers, 1978; Tomm, 1980; Tucker & Pinsof, 1984). Several studies have also reported significant increases in trainees' executive skills (e.g., Allred & Kersey, 1977; Pinsof, 1977; Watson, 1975). However, one limitation of this research
concerns the absence of comparable control groups against which the
effects of family therapy training can be measured. This leaves open the question as to whether factors other than the training program were responsible for the noted changes (see Cook & Campbell, 1979). The unique contribution of the present investigation is the assessment of change in the family therapy skills of family therapy trainees compared with a comparable control group.
In general, the results of this study did not provide compelling support for the major hypotheses. For example, it was predicted that the family trainees would become more cognitively complex than the
control subjects. The results of this investigation revealed that the family therapy trainees became less cognitively complex, while the control subjects became significantly more cognitively complex across time. It was also predicted that the trainees' family construct system would become more abstract than the family construct
system of the control subjects. Although the results revealed some trends in support of this hypothesis, the data do not argue cogently in support of the expected increase in abstraction among the family therapy trainees. In addition to predicting increases in the level
of perceptual skills of the family trainees, it was hypothesized that family trainees would show an increase in conceptual skills as a function of training compared to the control group. The results revealed that both groups changed across time, becoming somewhat more narrowly focused than predicted. Lastly, it was predicted that the family trainees wo uld show more sophisticated levels of executive skills compared to control subjects as a function of structural family therapy training. Although the results revealed some
interesting changes in executive skills for both groups across time, there were no significant differences between the two groups. These results are contrary to the prediction of increased skill level of family trainees compared to controls as a function of structural family therapy training. The remainder of this discussion will be organized around each of the major components and hypotheses of the study.
Perceptual Skills and Cognitive Complexity
Cognitive differentiation refers to the number of different
constructs that an individual has available for viewing and interpreting a situation. According to Werner (1955), construct systems develop from simple to complex levels of cognitive differentiation over the course of continued.experience. This reasoning is consistent with Bieri's definition of complexity as "the tendency to construe social behavior in a multidimensional way such that a cognitively complex individual has available a more versatile
system for perceiving the behavior of others than does a less complex person" (1955, p. 14). The present study investigated whether family therapy trainees' construct systems would become more cognitively complex as a function of structural family therapy training.
Analyses of differentiation revealed that the family therapy trainees became significantly less complex as a function of training. Interestingly, the control subjects showed significant increases in their level of complexity across time. These results
suggest that when students in the mental health fields are exposed to
family therapy training, they begin to construe families in less differentiated ways. The fact that the family therapy trainees were
less differentiated than those students without that training suggests that they are more channel ized in their constructions of
families, and have not yet integrated their new experiences into a larger framework within which to anticipate and predict families in therapy.
These results are consistent with the literature which has reported that trainees' constructs undergo periods of decreased complexity during the early phases of training, often followed by a
reorganization of the constructs at higher levels of complexity during advanced training stages (e.g., Baldwin, 1972; Lipshitz, 1974; Pope, 1978; Ryle & Breen, 1974). These findings are also consistent with the conceptual difficulties that family therapy trainees often report during the early phases of training (Gurman & Kniskern, 1978). Family therapy trainees may experience more difficulty than nonfamily therapy trainees in formulating their conceptualizations of families, thereby resulting in the temporary decline in trainee performance cited in the training literature (Duhi, 1978; Gurman & Kniskern, 1978).
Further, these findings may offer support for the operation of
the Creativity Cycle in structural family therapy, such that there is a shift that starts with loosened construction and terminates with tightened and validated constructions. Constructs must be loosened in order that hypotheses may be tested, and then they must be
tightened after the hypotheses have been either validated or
invalidated so that they may be acted upon. If the family therapy trainees' family construct system becomes less organized during early
phases of -training such that the relationships among the constructs decrease, then efforts to interpret family behavior would be based on
constructs that are loosely related and lacking meaningful connections. To explore this hypothesis further, future research needs to measure the impact of training on therapy skills at different intervals throughout training and after training.
In addition to looking at levels of cognitive differentiation,
analyses also revealed some significant findings concerning cognitive integration. The results revealed that the nonfamily therapy
trainees were significantly less integrated in their family constructs than the family therapy trainees. Further, both groups of
trainees tended to become somewhat less integrated across time regardless of the form of therapy training engaged in.
The results of the integration measures suggest that the family therapy trainees are more flexible in their use of their family construct system than are the nonfamily therapy trainees. However, this does not appear to be a unique function of training in structural family therapy. Although the family trainees appear to have fewer constructs available for perceiving families than the nonfamily therapy trainees, it may be that they are more invested in
tightening their family construct system in an attempt to make the constructs more meaningful. Further, it is interesting that both groups of trainees had a tendency to become somewhat less integrated
across time, suggesting that any training in therapy initially channelizes and rigidifies trainees' clinical perceptions.
Taken as a whole, the results of the measures of cognitive differentiation and cognitive integration are mixed. On the one hand, the results suggest that family therapy trainees become less
differentiated as a function of structural family therapy training while the nonfamily therapy trainees become more differentiated across time. However, the results also revealed that the family therapy trainees are more cognitively integrated than the nonfamily therapy trainees. Both sets of results are consistent with the training literature. Those studies reporting decreases in cognitive complexity based their conclusions on the results of differentiation scores (e.g., Baldwin, 1972; Pope, 1978; Ryle & Breen,1974) while those studies reporting increases in cognitive complexity based their
conclusions on the results of integration scores (e.g., Adams-Webber & Mirc, 1976). Thus, these discrepant findings reported in the
literature, as well as those revealed in the current investigation, may be the result of using different measures of cognitive complexity. These two measures may reflect different aspects of cognitive complexity that develop at differential rates as trainees are exposed to new training and therapy experiences.
However, there is an alternative hypothesis that may explain the discrepancy between cognitive differentiation and cognitive integration. As family therapy trainees learn a coherent and systemic conceptualization of family dynamics, they may begin to see families as less different from each other. Further, the family
therapy trainees irkay begin to see the commonalities and similar patterns across different families. This hypothesis is consistent with the goal of structural family therapy which is to look for the structural patterns that are supporting what appear to be the differing dynamics of different families (e.g., looking for the homeostatic principles that maintain the presenting symptom). Thus, the decreased levels of differentiation for the family therapy trainees suggests that.they see families as more alike as they begin to learn to interpret family dynamics with a structural template. However, the increased levels of differentiation for the nonfamily
therapy trainees suggests that their interpretations of family dynamics may be based on several different conceptualizations that they are utilizing in their individual therapy training. Thus, the lower levels of differentiation in the family therapy trainees may
suggest that structural family therapy training does not necessarily lead to lower levels of cognitive complexity, but influences family
therapy trainees' family construct system in such a way that they are able to tighten those family construct dimensions compared to therapy trainees without that training. This interpretation is also consistent with the higher level of cognitive integration found for the family therapy trainees. The family therapy trainees appear to
be more flexible in their use of their family construct system than the nonfamily therapy trainees. Perhaps the family therapy trainees are more willing to explore the possibilities of interrelationships among their family constructs and in that sense, they show higher
levels of cognitive complexity in their perceptual skills than the nonfamily therapy trainees.
Perceptual Skills and Level of Construct Abstraction
According to Crockett (1982), a major feature of a developing
construct system involves the appearance of abstract constructs. One method frequently used to assess the development of abstract
constructs involves a content analysis of the construct dimensions elicited by the Rep Grid (Duck, 1973). A review of the literature
revealed that those studies using a content analysis to investigate the level of construct abstraction as a function of training reported significant increases in level of abstraction as a function of training (e.g., Lewis, 1977; Lifshitz, 1974). The present study investigated whether family therapy trainees' family constructs would become more abstract as a function of structural family therapy training.
Level of abstraction is measured with a weighted average score based on Duck's (1973) system of content analysis. Contrary to expectations, there were no significant findings on measures of abstraction. However, there was a trend toward a main effect of Group, suggesting that the family therapy trainees tended to have
relatively higher levels of abstraction in their family constructs than did the nonfamily therapy trainees. Although these results are
in the predicted direction, the data do not argue cogently for differences between the two groups of therapy trainees. Perhaps a larger sample size would provide enough power to detect any changes
that may be taking place. It must be kept in mind, however, that family therapy training itself may not lead to higher levels of family construct abstraction.
Conceptual Skills and Case Conceptualizations
According to Cleghorn and Levin's (1973) model of therapy
skills, conceptual skills involve what the therapist thinks about in the therapy session and how those thoughts are organized. They argue that in order to formulate a problem in terms of the family system, the family therapist must specify the implicit rules of behavior in
the family that make sequences of family interactions predictable. Although the majority of the studies reported in the literature found increases in levels of conceptual skills for family therapy trainees,
these studies have been limited by the lack of a comparable control group (e.g., Breunlin et al., 1983; Friedman, 1971; Lange & Brinckman, 1976; Lange & Zeegers, 1978; Tom, 1980; Tucker & Pinsof, 1984) and the lack of a reliable, valid and standardized assessment instrument (e.g., Breunlin et al., 1983; Friedman, 1971; Lange & Brinckman, 1976).
In the present study, conceptual skill level was measured along a systemic dimension ranging from an individual focus to a systems focus. Trainees' conceptual skill level was measured for their ability to conceptualize intervention dynamics. it was predicted that the family therapy trainees would show an increase in their level of family and intervention conceptualization skills while the nonfamily therapy trainees would remain stable across time. The
results of the family intervention measures revealed no significant differences between the two training groups and no significant difference across time. The mean level of responses suggests that both groups focused on subgroup relationships when conceptualizing an
The results of the family conceptualization measures revealed a trend toward a main effect of Time, suggesting that both groups of subjects had a tendency to use slightly lower levels of conceptual skills at posttesting. There are several possible explanations for these findings. First, it is possible that the nature of the task biased the subjects to focus on family dynamics. Subjects were instructed to select case conceptualization response alternatives
that best reflected how they perceived the dynamics that were important in the family. By instructing subjects to focus on family dynamics, it is not surprising that subjects did just that, regardless of their training orientation.
A second explanation may involve the instrument itself. This instrument may not be sensitive enough to detect subtle changes, in part, because asking subjects to select their conceptualization of a segment of behavioral data is only a partial evaluation of conceptual skills. Such an evaluation may not predict how a trainee would conceptualize family dynamics if actually in a similar clinical situation. Since this measure is a compromise in that it can be
reliable, valid and standardized, it is important that it be made as reliable and valid as possible. Future research needs to establish such reliability and validity in their family therapy skills
instruments before changes in skill level of family therapy trainees can be adequately investigated.
Third, it may also be possible that the conceptual skills of the family therapy trainees were only beginning to undergo change at the time of the second measurement. It is possible that Cleghorn and Levin's (1973) model of therapy skills may be arranged along a dimension of skill level such that perceptual skills are a prerequisite for conceptual skill development and, in turn, conceptual skills are a prerequisite for the development of executive skills. Since the family therapy trainees are only beginning to change their perceptions of families, they may not have had enough time to develop family conceptual skills. Future research needs to
take a longitudinal approach to the investigation of the development of family therapy skills as a function of training. Lastly, it must be kept in mind that the development of family conceptual skills may not be unique to training in family therapy.
Executive Skills and Therapist Responses
The structural family therapist must develop therapy skills that influence the family to demonstrate the way it functions. Cleghorn and Levin (1973) have labeled these skills as executive skills. They argue that these skills involve what the therapist says and does in the therapy session in order to influence the family's sequences of transactions and consequently alter the way the family functions. Investigating the executive skills of family therapy trainees has proven to be a very difficult task. A review of the literature
reveals many methodological flaws in this area of investigation.
These flaws have consisted of the lack of comparable control groups and the absence of reliable, valid, and standardized measures sensitive to changes in skill level that are unique to family therapy. Those studies reporting increases in family therapy trainees' executive skills were unable to attribute those changes to training in family therapy as they did not include a comparable control group (e.g., Allred & Kersey, 1977; Tucker & Pinsof, 1984).
In those studies that did include some type of control group, the results have typically not shown significant differences in executive skill level of the trainees (e.g., Breunlin et al., 1983). However, these studies did not use a reliable coding instrument and thus their lack of results may be due to the instruments used rather than there being no differences. The present study investigated whether the executive skills of family therapy trainees, as compared with nonfamily therapy trainees, would increase when measured by a reliable arid valid coding system.
Executive skills were measured along four categories of
therapist responses that were found to be the most meaningful of Allred and Kersey's (1977) system based on a pilot investigation. The four categories reflected by the therapist response alternatives were Gathers Information, Reflects and Supports, Interprets and Confronts, and Structural Realignment. All measures of executive skill level were analyzed separately for each of the four the rapist response categories.
Executive Skills Within Category
Contrary to expectations, there were no significant differences between the two groups for level of executive skill within this category of therapist responses. Both groups of subjects were able to ask for information about subsystem relationships regardless of
their theoretical orientation (i.e., systems versus intrapsychic). This lack of significant findings suggests that these executive skills may not be a unique function of training in family therapy. If basic therapy skills, such as information gathering, can be adapted to use with families, then it will be important for future
research to explore which therapy skills cannot make that transition without additional training in family therapy. The question that continues to concern the field revolves around what skills are unique
to family therapy.
Reflects and supports
The results of these analyses of the Weighted Average scores and the Modal Response scores were also contrary to expectations. Both groups of therapy trainees showed a significantly higher frequency of
supportive responses at Time 1, which shifted to a significantly higher frequency of less sophisticated reflective responses at Time 2. A possible interpretation of these results may be that as the family therapy trainees attempt to make sense of their new family construct system, and the nonfamily therapy trainees attempt to
effectively understand actual clinical experience that they are not prepared for, they revert back to their first learned style of
interacting in therapy. The fact that there were no significant differences between the two groups of therapy trainees suggests that reverting to previously learned levels of executive skills may not be a function of structural family therapy training, but a function of attempting to interact effectively in a new situation. This lack of significant differences between the two groups of trainees again raises the question concerning the unique impact of family therapy training. Although the subjects reverted to a lower level of their executive skills, it is interesting to note that there were no changes in their average confidence level concerning their choice of therapist response to the family.
Interprets and confronts
The results of these analyses revealed that there were no
significant differences in level of executive skWs between the two groups of therapy trainees. Both groups of subjects tended to use strategic and confrontive interpretations, and there were no changes across time. This lack of change in executive skill level for
interpretive responses suggests that these skills may not be unique to family therapy training. If these skills are unique to family
therapy, it is possible that the task in the present investigation was biased toward not detecting changes. Since all the subjects were required to respond to the family by selecting an interpretive response, it is not possible to determine whether the nonfamily
therapy trainees would have responded differently than the family therapy trainees if allowed to select an interpretive response with an individual rather than a family focus. Future research needs to adjust for this factor of response bias in a forced choice format. One possibility may be to use a crossed design similar to the present investigation with the addition of an individual therapy tape. This would allow for an assessment of whether family and individual therapy trainees differ in their level of interpretive executive skills when interacting with an individual versus interacting with a family.
According to Allred and Kersey's (1977) classification scheme,
structural realignment consists of the most sophisticated levels of therapy responses that can be made to a family. It was predicted that the family therapy trainees would show an increase in their level of structural realignment responses while the nonfamily therapy trainees would remain stable across time. Contrary to these predictions, the results revealed that there were no significant differences in response level for the two groups of trainees, and there were no significant differences across time. A possible explanation for this lack of change in executive skill level may be
that these responses were too sophisticated for even beginning family therapy trainees to attempt. It would be interesting to reassess the executive skill level of these trainees after the family therapy trainees had gained more training and exposure in family therapy.
Executive Skills Across Category
In addition to investigating possible changes within each of the
four therapist response categories, analyses were also performed to assess any possible changes in executive skills or confidence across these four categories. An analysis of modal responses for trainees' preferred category of therapy responses revealed no significant differences between the two groups or across time. Both groups of therapy trainees preferred to use Reflecting and Supporting responses. However, the results for the Average Confidence Level for the Preferred Response revealed that both groups of therapy trainees became significantly more confident concerning their preferred responses across time.
Taken together, these results suggest that there are no changes in executive skill level for family therapy trainees as a function of structural family therapy training. Both groups of therapy trainees maintained stable levels of executive skill for interpretive and structural realignment therapist responses. Perhaps these two categories of responses were too sophisticated to reflect changes in executive skill level of beginning therapists. Consistent with this hypothesis, changes in executive skill level were revealed for the therapist response categories of Gathers Information and Reflects and Supports. The results revealed that the level of executive skill decreased across time for both groups of therapy trainees. Further, the results revealed that the preferred category of therapist responses were Reflecting and Supporting responses. One possible
explanation for these results is that regardless of the therapy training involved, when confronted with a new situation, therapy trainees will revert back to more familiar and comfortable lower level therapist responses. Future research in this area will need to take a longitudinal view of the executive skill level of family therapy trainees (as compared to nonfamily therapy trainees) as they gain more exposure to family therapy.
In a recent review of the family therapy training field, Kniskern and Gurman (1979) emphasized the lack of empirical, evaluative studies. The majority of the studies that have been reported have been based on self-report and uncontrolled investigations. Although these studies have reported increases in family therapy skills, it has not been possible to attribute these perceived changes to training in family therapy. Recently there has been a trend toward more empirically sound measurement of the effects of family therapy training (e.g., Breunlin et al., 1983; Tucker & Pinsof, 1984). However, even these studies have been hampered by the lack of reliable and valid measures of therapy skills and the lack of
adequate control groups. The present investigation attempted to provide both reliable measures of family therapy skills as well as a comparable control group. Although the results of this investigation did not provide compelling support for the major hypotheses, they were suggestive of several alternative hypotheses that require additional research in the future.
Based on Cleghorn and Levin's (1973) model of family therapy
skills, the present study investigated three major skills associated with effective family therapy (i.e., perceptual, conceptual and executive skills). Using Kelly's (1955) Role Repertory Grid measure, the results of the present investigation revealed changes in
perceptual skills that at first glance appear to be opposite of those predicted. Measures of cognitive differentiation increased for the nonfamily therapy trainees and decreased for the family therapy trainees. This would appear to suggest that training in structural family therapy has a negative impact on trainees, resulting in decreased levels of cognitive complexity. However, measures of
cognitive integration revealed that the family therapy trainees were much more integrated and flexible than the nonfamily therapy
trainees. Taken as a whole, these results suggest that as trainees begin to learn about family dynamics they form a structural template that is used as the basis for their family interpretations. It appears that the family therapy trainees are beginning to see the commonalities and patterned similarity across the different families,
while the nonfamily therapy trainees continue to view families based on several different theoretical orientations. Thus, the decrease in differentiation for the family therapy trainees may be a reflection of the tightening of the new family construct system, while the increase in integration reflects their flexibility in the use of
their new constructs. In contrast, the increase in differentiation for the nonfamily therapy trainees may simply reflect their continued use of a diversity of therapy constructs, while their lower level of
integration reflects their less flexible use of therapy constructs when applied to families.
Interestingly, although the family therapy trainees are
beginning to experience changes in their perceptions of families, they have not yet developed these skills to the point where they can begin to conceptualize the family differently. Regardless of the therapy training involved, all trainees were able to focus on subsystem dynamics when required to conceptualize the family. However, the family therapy trainees were not yet able to conceptualize the family in completely systemic terms. It is possible that they had simply not advanced to this point. This hypothesis is consistent with the results of the content analysis of trainees' perceptual skills. Although trainees were beginning to show perceptual changes, there were no significant increases in level
of abstraction, suggesting that the trainees' perceptual skills were still developing.
Based on the results of the measures of perceptual and
conceptual skills, it is not surprising that there were no changes in their executive skill level. This further suggests that executive skills would be the last skills to develop. Before the family therapist can act differently in therapy, he or she must perceive the family differently and conceptualize the family dynamics differently. The fact that both groups of therapy trainees reverted back to relatively basic reflection of feeling responses suggests that dealing with a family was a new and possibly confusing situation that
prompted subjects to respond in ways that they felt more confident wi th.
In conclusion, although the results of this study did not
support the hypotheses of increased family therapy skill levels, the results did not negate these hypotheses either. Rather than
indicating an increase in these skills, the results suggest that family therapy skills may be developmentally dependent upon each other such that higher level skills can not develop until lower level skills have been adequately developed and consolidated. Thus, as a function of training in structural family therapy, trainees may first show changes in their perceptual skills, followed by changes in conceptual and then executive skills. Future research needs to take a longitudinal approach to the investigation of the impact of training on the family therapy skills of trainees in this field.
The investigation of the effects of training in family therapy is a new and relatively uncharted area of inquiry. There are only two studies cited in the literature that approach the empirical sophistication and rigor necessary to explore the impact of family therapy training on therapy skills (e.g., Breunlin et al., 1983; Tucker & Pinsof, 1984). Both of these investigations have several flaws in their methodology that need to be addressed. In the Breunlin et al. (1983) study, there was no mention of any objective coding system and, although they are beginning to develop an instrument for measuring family therapy skills, without an objective
measure their results may have been biased toward the logical expectation that trainees' responses would become more complex after training. Further, this study did not employ a comparable control group. In the Tucker and Pinsof (1984) investigation, the authors reported the use of a measure found to reliably discriminate between the family therapy skills of beginning and advanced family therapists. However, they did not employ any control group and, consequently, it is difficult to attribute their results to the specific effects of training in family therapy. Future research must combine these two aspects of empirical investigation, namely, the use
of a reliable measurement and experimental control.
While the present investigation attempted to control for these
two major limitations, the instruments used may not have been sensitive enough to detect subtle changes. The reliabilities for the therapist response executive skill measure ranged from .45 to .76, while the reliabilities for the conceptual skill measure only ranged from .33 to .69. These two instruments need to be revised and reliabilities increased before they can be expected to reliably measure changes in therapy skill level. A second limitation in the present investigation consisted of the small sample size (N=33), reducing the power of the analyses to detect changes that may be present. For example, analyses of power for structural measures of
cognitive integration and cognitive differentiation revealed that the power of the F tests were .11 and .72, respectively. In the present
study, adequate power (e.g., .80 to .90) would require an approximate sample size of 50 to 70 subjects. Future studies must include an
adequate sample size of both family therapy trainees and control subjects in order to determine if changes in therapy skills are a unique function of training in family therapy.
In addition to the above two changes in methodology, future
research may also benefit from taking a longitudinal approach to the study of family therapy training. It may be that changes in conceptual and executive skills do not begin to change until the trainees gain actual clinical experience with families. Future research will need to investigate this issue of past clinical experience more closely. Does the amount of past experience with individual therapy affect how family therapy trainees integrate their new knowledge? This could be investigated by comparing four groups of subjects: family therapy trainees with no prior therapy experience, family therapy trainees with two to five years of prior therapy experience, nonfamily therapy trainees with no prior therapy experience, and nonfamily therapy trainees with two to five years of
prior therapy experience. It would also be interesting to investigate the relationship between trainees' own family background (e.g., only child versus several siblings) and their family conceptual skills as a function of training in family therapy.
Further, future research may also benefit from taking a slightly different approach to the study of therapy skills in family therapy trainees. The present investigation may have biased trainees toward focusing on family dynamics. According to structural family
therapists (e.g., Minuchin, 1974), one can take a structural systems approach even when working with an individual. In order to control
for biasing subject responses simply by being in a family therapy project, future research can use a crossed design such that subjects are required to view an individual as well as a family on rape in order to measure conceptual and executive skills. It would be
expected that control subjects would not take a systems view of the individual, while the family therapy trainees may begin to conceptualize and respond to an individual along a more systemically oriented dimension.
If the field of family therapy training is going to claim that it has a unique impact on the family therapy skills of its trainees, then future research must not ignore the proven, traditional methodological approaches (Gurman, 1983). According to London and Klerman (1982), the basic principles of empirical investigation are measurability and replicibility. As the area of family therapy moves toward traditional methodological investigation, the area of training in family therapy must also make this move toward controlled comparative investigation.