Citation
Counseling process and premature termination

Material Information

Title:
Counseling process and premature termination
Creator:
Molnar, Suzanne L., 1945-
Copyright Date:
1982
Language:
English

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Suzanne L. Molnar. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
9320505 ( OCLC )
ABU6009 ( LTUF )

Downloads

This item has the following downloads:

counselingproces00moln.pdf

counselingproces00moln_0139.txt

counselingproces00moln_0104.txt

counselingproces00moln_0081.txt

counselingproces00moln_0132.txt

counselingproces00moln_0004.txt

counselingproces00moln_0051.txt

counselingproces00moln_0099.txt

counselingproces00moln_0026.txt

counselingproces00moln_0070.txt

counselingproces00moln_0030.txt

counselingproces00moln_0032.txt

counselingproces00moln_0039.txt

counselingproces00moln_0079.txt

counselingproces00moln_0144.txt

counselingproces00moln_0138.txt

counselingproces00moln_0047.txt

counselingproces00moln_0103.txt

counselingproces00moln_0074.txt

counselingproces00moln_0115.txt

counselingproces00moln_0090.txt

counselingproces00moln_0066.txt

counselingproces00moln_0072.txt

counselingproces00moln_0091.txt

counselingproces00moln_0093.txt

counselingproces00moln_0012.txt

counselingproces00moln_0077.txt

counselingproces00moln_0019.txt

counselingproces00moln_0089.txt

counselingproces00moln_0109.txt

counselingproces00moln_0013.txt

counselingproces00moln_0097.txt

counselingproces00moln_0000.txt

counselingproces00moln_0009.txt

counselingproces00moln_0018.txt

counselingproces00moln_0015.txt

counselingproces00moln_0084.txt

counselingproces00moln_0101.txt

counselingproces00moln_0035.txt

counselingproces00moln_0022.txt

counselingproces00moln_0028.txt

counselingproces00moln_0100.txt

counselingproces00moln_0135.txt

counselingproces00moln_0059.txt

counselingproces00moln_0136.txt

counselingproces00moln_0145.txt

counselingproces00moln_0124.txt

counselingproces00moln_0111.txt

counselingproces00moln_0096.txt

counselingproces00moln_0024.txt

counselingproces00moln_pdf.txt

counselingproces00moln_0083.txt

counselingproces00moln_0114.txt

counselingproces00moln_0082.txt

counselingproces00moln_0029.txt

counselingproces00moln_0121.txt

counselingproces00moln_0108.txt

counselingproces00moln_0046.txt

counselingproces00moln_0056.txt

counselingproces00moln_0125.txt

counselingproces00moln_0086.txt

counselingproces00moln_0113.txt

counselingproces00moln_0001.txt

counselingproces00moln_0058.txt

counselingproces00moln_0042.txt

counselingproces00moln_0007.txt

counselingproces00moln_0120.txt

counselingproces00moln_0129.txt

counselingproces00moln_0098.txt

counselingproces00moln_0064.txt

counselingproces00moln_0038.txt

counselingproces00moln_0078.txt

counselingproces00moln_0069.txt

counselingproces00moln_0071.txt

counselingproces00moln_0052.txt

EQ3X5CEJ6_MEELDD_xml.txt

counselingproces00moln_0044.txt

counselingproces00moln_0102.txt

counselingproces00moln_0041.txt

counselingproces00moln_0020.txt

counselingproces00moln_0128.txt

counselingproces00moln_0085.txt

counselingproces00moln_0141.txt

counselingproces00moln_0112.txt

counselingproces00moln_0094.txt

counselingproces00moln_0011.txt

counselingproces00moln_0045.txt

counselingproces00moln_0092.txt

counselingproces00moln_0034.txt

counselingproces00moln_0107.txt

counselingproces00moln_0025.txt

counselingproces00moln_0063.txt

counselingproces00moln_0068.txt

counselingproces00moln_0036.txt

counselingproces00moln_0133.txt

counselingproces00moln_0014.txt

counselingproces00moln_0050.txt

counselingproces00moln_0088.txt

counselingproces00moln_0116.txt

counselingproces00moln_0006.txt

counselingproces00moln_0008.txt

counselingproces00moln_0110.txt

counselingproces00moln_0060.txt

counselingproces00moln_0118.txt

counselingproces00moln_0023.txt

counselingproces00moln_0048.txt

counselingproces00moln_0062.txt

counselingproces00moln_0146.txt

counselingproces00moln_0054.txt

counselingproces00moln_0105.txt

counselingproces00moln_0053.txt

counselingproces00moln_0122.txt

counselingproces00moln_0126.txt

counselingproces00moln_0080.txt

counselingproces00moln_0073.txt

counselingproces00moln_0005.txt

counselingproces00moln_0106.txt

counselingproces00moln_0137.txt

counselingproces00moln_0134.txt

counselingproces00moln_0140.txt

counselingproces00moln_0061.txt

counselingproces00moln_0095.txt

counselingproces00moln_0087.txt

counselingproces00moln_0057.txt

counselingproces00moln_0075.txt

counselingproces00moln_0040.txt

counselingproces00moln_0119.txt

counselingproces00moln_0043.txt

EDCGAT0AP_LXG0OK_xml.txt

counselingproces00moln_0017.txt

counselingproces00moln_0130.txt

counselingproces00moln_0002.txt

counselingproces00moln_0076.txt

counselingproces00moln_0067.txt

counselingproces00moln_0131.txt

counselingproces00moln_0037.txt

counselingproces00moln_0031.txt

counselingproces00moln_0127.txt

counselingproces00moln_0055.txt

counselingproces00moln_0021.txt

counselingproces00moln_0065.txt

counselingproces00moln_0147.txt

counselingproces00moln_0016.txt

counselingproces00moln_0049.txt

counselingproces00moln_0143.txt

counselingproces00moln_0117.txt

counselingproces00moln_0003.txt

counselingproces00moln_0123.txt

counselingproces00moln_0010.txt

counselingproces00moln_0142.txt

counselingproces00moln_0033.txt

counselingproces00moln_0027.txt


Full Text












COUNSELING PROCESS AND PREMATURE TERMINATION


By

SUZANNE L. MOLNAR














A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY




UNIVERSITY OF FLORIDA


1982























To Terry, to Julia,

and

to my parents















ACKNOWLEDGMENTS


I count myself very fortunate to have received the help and friendship of many people throughout the course of my doctoral work. I cannot mention them all, but special thanks are due to some.

Most affectionate appreciation is extended to Dr. Harry Grater. Both as chairman of my dissertation committee and as clinical supervisor during my internship, he has been an exceptionally supportive mentor. He has given freely of his time, his expertise, and his sense of proportion. My graduate years would have been poorer without him.

The other members of my dissertation committee deserve warm thanks as well. Dr. Franz Epting has been a friend and a source of support since my first year at the University of Florida, and he provided invaluable assistance with many aspects of this dissertation. Dr. James Archer, Jr., has been a helpful committee member and an inspiring role model for whom to work. Dr. William Froming was especially helpful with design and statistical problems. Dr. Carolyn Tucker has been generous both with her long-term friendship and with her last-minute assistance.

I also want to express gratitude to Ernest Downs for the time and care he took with the scoring of my research tapes. He worked hard at a most difficult point in the research-and he even managed to help me enjoy the process.


iii











Finally, I want to express my deepest appreciation to my

husband, Terry Molnar. Without his faith in me, his unfailing support when I needed it most (and his endless hours of babysitting), this dissertation might not have been completed.


iv
















TABLE OF CONTENTS


Page

ACKNOWLEDGMENTS . . . . . . . . . . . . iii

ABSTRACT . . . . . . . . . . . . . . viii

CHAPTER

ONE INTRODUCTION . . . . . . . . . . . 1

The Problem . . . . . . . . . . 1
Need for This Study . . . . . . . . 1
Overview of the Design . . . . . . . 4
Dependent Variable . . . . . . . 4
Independent Variables . . . . . . 5
Hypotheses . ... . .. .. .. .. .. 7
Organization of the Study . . . . . . 8

TWO REVIEW OF THE LITERATURE . . . . . . . . 11

Premature Termination . . . . . . . 11
The Definitional Problem . . . . . 11
Premature Termination and Psychotherapy
Outcome . *. .. . . .. .. .. 18
Predicting/Explaining Premature Termination . . 21
General Factors . . . . . . . . 21
Client characteristics . . . . . 21
Therapist characteristics . . . . 25 Interaction of client and therapist characteristics . . . . . . . 26
Client-Therapist Dynamics . . . . . 32
Client reaction to therapist . . . . 33 Therapist reaction to client . . . . 35 Therapy process factors . . . . . 37 Studying Therapy Process . . . . . . . 38
General Problems . . . . . . . . 38
Process as a Predictor of Outcome . . . 39 Process variables derived from a theory of psychotherapy . . . . . 39 A-theoretical variables . . . . . 43
Hypotheses . . . . . . . . . . 51


v











CHAPTER

THREE METHODOLOGY . . . . . . . . . . .

Subjects . . . . . . . . . .
Data Collection . . . . . . . .
Instrument: The Leary Interpersonal Checklist .
Theoretical Basis . . . . . . .
Construction . . . . . . . .
Validating Research . . . . . .
Category integrity and arrangement . Dimensional integrity and orthogonality Structural similarity among the levels of personality . . . . . .
Concurrent validity . . . . . Reliability . . . . . . .
Use in This Study . . . . . . .
Tape Rating . . . . . . . . .
Procedures . . . . . . . . .
Selection and Training of Rater . . .
Rater Reliability . . . . . . .
Data Analysis . . . . . . . . .
Hypothesis Testing . . . . . . .
Investigating Potential Intervening
Variables . . . . . . . . .
Other Analyses . . . . . . . .
Significance Levels . . . . . .


FOUR RESULTS . . . . . . . . . .


Introduction . . . . . . . . .
Hypothesis Testing . . . . . . .
Client Behavior and Premature Termination:
Hypotheses 1 and 2 . . . . . .
Therapist Behavior and Premature Termination: Hypotheses 3 and 4 . . . .
Therapist-Client Complementarity and
Premature Termination: Hypotheses 5 and 6 Analysis of Selected Client and Therapist Factors . . . . . . . . . .
Client Factors . . . . . . . .
Therapist Factors . . . . . . .
Response Intensity . . . . . . .
Other Reanalyses . . . . . . . .


84 84

86

86

88

88 88 . 91
93 97


FIVE DISCUSSION . . . . . . . . . . .


99


Counseling Process and Premature Termination
Client Process and Premature Termination
Therapist Process and Premature Termination
Therapist-Client Complementarity and
Premature Termination . . . . . .


99 99
. 101

103


vi


Page


54


54 55 56 56 58 61 62 62

65 67 67 69 73 73 75 77 81 81

81 83 83


. . . 84











CHAPTER


Nonprocess Factors and Premature Termination .
Client Factors . . . . . . .
Therapist Factors . . . . . .
Limitations of This Study and Suggestions for Future Research . . . . . . .
General Conclusions . . . . . . .


APPENDIX

A THE LEARY INTERPERSONAL CHECKLIST . . . . .

B INTERPERSONAL PROCESS DIAGNOSIS . . . . .

C RATER TRAINING PROGRAM AND SUPPLEMENTARY CODING AID D DATA ON CASES COMPRISING THE STUDY SAMPLE . . REFERENCES . . . . . . . . . . . .

BIOGRAPHICAL SKETCH . . . . . . . . . .


vii


Page

105 105 106

107 110


. 113 117 120 124 127 135


I















Abstract of Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


COUNSELING PROCESS AND PREMATURE TERMINATION By

Suzanne L. Molnar

August 1982

Chairman: Harry A. Grater, Jr. Major Department: Psychology

Premature termination from psychotherapy is a problem for many mental health professionals, but little is known about how it occurs. This research focused on client-therapist process as a potential predictor of premature termination.

Thirty university counseling center client-therapist pairs,

half premature terminators and half remainers, were selected as subjects. All clients had been referred by intakers for ongoing psychotherapy. Number of sessions attended and therapist judgment were used to decide termination status.

Process was measured using Leary's interpersonal circumplex to rate behavior on two bipolar, orthogonal dimensions (dominancesubmission and affiliation-hostility). Fifteen-minute audiotaped segments from first postintake therapy sessions were scored. Each statement was located in one quadrant of the circumplex and then assigned an intensity rating of "1" (low) or "2" (high).


viii


I










On the basis of previous research, it was hypothesized that premature terminators' tapes would differ from remainers' tapes by containing:

1. Relatively high percentages of hostile-submissive

client statements

2. Relatively low percentages of affiliative-submissive

client statements

3. Relatively high percentages of hostile-submissive

therapist statements

4. Relatively low percentages of hostile-dominant therapist statements

5. Relatively low levels of therapist-client complementarity, both for dominance

6. And for affiliation

Hypotheses were tested using two-tailed t-tests for the significance of differences between means. No relationship was found between client process and premature termination. Therapist process differentiated the two outcome groups, but one difference reversed that predicted: both hostile-submissive and hostile-dominant behavior were more common among the therapists of premature terminators. The obtained complementarity of therapists' responses to their clients also reversed the direction hypothesized. Remainers' therapists behaved in the less

complementary fashion, and this difference was significant at the .06 level for dominance complementarity.

Additional analyses found no client or therapist characteristic predictive of termination status at the .05 significance level.


ix


I










However, n's were very small. Numerical differences favored clients from business-professional backgrounds and favored female therapists.

It was concluded that client-therapist process probably is one component of the multiply determined process of premature termination.


x















CHAPTER ONE
INTRODUCTION


The Problem

Premature termination from psychotherapy is a problem of concern to many. Mental health centers, student counseling centers, and private therapists all are troubled by the frequency with which clients make contact, begin therapy, then terminate before they have received what professionals consider to be adequate treatment. Practitioners ask why this happens and what, if anything, they can do about it. Some researchers have attempted to answer these questions, but much remains to be done.


Need for This Study

To begin with, most research attempting to "explain" premature

termination actually just explains it in the statistical sense by looking for correlates. Client characteristics and therapist characteristics are the usual independent variables. This line of work has been fairly successful. A variety of client characteristics, plus therapist experience, has been established as reliable correlates of premature termination (Baekeland and Lundwell, 1975).

However, this type of information is of limited utility. Knowing that a poor, black woman of limited intelligence is likely to leave therapy before it has had much impact helps little with case management.


I




2


Knowing that inexperienced therapists lose many clients prematurely gives the profession no assistance since the cure it suggests is inseparable from routine provision of service.

A group of studies with greater potential usefulness focuses on the interaction of client characteristics with therapist characteristics. Demographic characteristics have not proven very valuable in this research. On the other hand, various similarities in the personalities of therapist and client have been found to increase client retention rates (Mendelsohn, 1966; Landfield, 1971). Also, clienttherapist similarity with regard to therapy goals, the priority of those goals, and perception of the client's problem all have been found to make premature termination less likely (Butler, 1977; Raschella, 1975). All of these results suggest ways in which service facilities and/or training facilities might be able to improve their programs to increase client retention rates.

In a sense, though, doing that type of study at this point

is like furnishing a house before the floor is put in. A more logical approach is to first tackle the question, "How does premature termination occur?" This question can be rephrased as, "How does therapy process leading to premature termination differ from process leading to ongoing therapy?" Answers can focus on the clients, on the therapists, or on the interaction between them. Verbal behavior, non-verbal behavior, or both may be targeted. Behavior may receive much or little interpretation. In any case, process research results are needed as a base for manipulative research efforts.





3


In their 1970 review, Meltzoff and Kornreich declared inattention to process a fundamental weakness in research on premature termination. In spite of their urgings, however, only a little effort has gone into this area.

A variety of studies approach therapy process indirectly by asking clients and/or therapists to record their reactions to, and interpretations of, what took place in a session. True process studies, however, are almost nonexistent. Only four were located in which process was measured directly and was compared with termination status.

Winder et al. (1962) pioneered the terrain with their study of therapist response to dependency content in clients' verbalizations. A later researcher tested the hypothesis that interpreting transferencebased behavior would make early termination less likely (Goldstein, 1975). Another looked at self-disclosure as a predictor of early termination (Heilbrun, 1973). Still more recently, Schiller used the Leary Interpersonal Checklist to measure therapist-client process complementarity and used complementarity as a predictor variable in his study of early termination (Schiller, 1978).

All four are interesting studies. Winder et al. (1962) and

Goldstein (1975) emphasized therapist behavior-a focus with real potential usefulness to clinicians. Heilbrun looked at (female) client behavior and found intriguing interactions between self-disclosure and sex of therapist when they were used as independent variables to predict early termination versus remaining in therapy. Schiller (1978) examined therapist behavior, client behavior and the complementarity of therapist behavior to client behavior-an approach commendable for its thoroughness.




4


It is important to note, however, that only Winder et al. (1962) actually studied premature termination. The others defined termination status entirely by the number of sessions attended; they

are studies of early-not premature-termination. This distinction is discussed at length in the first section of Chapter Two.

At this point, then, the field is wide open. Connecting

therapy process with premature termination has been declared a promising and potentially valuable endeavor, but almost no work has been done on it. This research was designed to begin filling that void.


Overview of the Design

The intent of this study was to learn more about how premature termination occurs. Therapeutic process was audiotaped, and comparisons were drawn between two groups of subjects: client-therapist dyads whose clients terminated prematurely and dyads whose clients remained in treatment.

Subjects were clients and counselors at the University of Florida Psychological and Vocational Counseling Center. Taping was done as clients began ongoing, individual psychotherapy. Assignment to termination category was made after completion of counseling. Dependent Variable

One major difficulty with this field of research appeared immediately: Premature termination enjoys no standard definition.

Even among researchers of similar theoretical persuasions, the concept is operationalized in a wide variety of ways.




5


At one extreme are the many researchers who operationally

equate early termination with premature termination. This gives them

simple, objective definitions, usually based on the number of sessions a client attended. However, this simplicity is simplistic to a selfdefeating extent if one's aim is to uncover significant differences in therapy process.

At another extreme are researchers who leave judgment of termination status entirely to the clinician. Without some strict controls, this can lead to chaos. Psychoanalytic clients with several hundred hours of therapy could be classified as premature terminators while clients from a crisis-oriented clinic were discharged as successes after two sessions.

Given this lack of consensus about a central concept, it seemed important to review others' efforts with care before choosing or constructing a definition of one's own. This review will be presented in the first part of Chapter Two. Based on its conclusions, a definition was constructed for this study which combines objective and subjective factors and which discards questionable cases rather than forcing them into a Procrustean bed.


Independent Variables

Since this field of research is largely unexplored, the literature offered little help in choosing independent variables. Fortunately, some assistance-was available from the field of social psychology. Reviewers covering a variety of factor analytic studies have concluded that dominance-submission and affiliation-hostility emerge as the major





6


factors in dyadic interaction. The descriptive terms differ slightly from one study to the next, but the dimensions remain intact and appear to be orthogonal (Foa, 1961).

Choosing variables based on their prominance in general interaction, rather than in the therapy setting, eliminates biases which could originate from practice of a particular type of therapy. It also

eliminates therapist and client roles; behaviors were rated similarly for both. However, the relevance of the dimensions to therapeutic interaction needs to be demonstrated.

That need was met in this study by the way in which dominancesubmission and affiliation-hostility were operationalized. The Leary Interpersonal Checklist was used to rate taped therapeutic process on an interaction-by-interaction basis. The instrument was designed for this purpose (Leary, 1957), and its validity and usefulness are well established (cf. Crowder, 1970; Mueller, 1969; Mueller and Dilling, 1969; Schiller, 1978; Spierling, 1972).

In fact, the Leary Interpersonal Checklist has a combination of general strengths and specific attributes which made it an ideal instrument for this research. The general strengths, which will be discussed at length in Chapter Three, include a solid theoretical grounding, competently done validation studies, and tested relevance for psychotherapy research. Specific assets-and the use made of them in the design of this research-will be discussed next.

One attribute of particular importance in this research is Leary's provision for rating therapist and client behavior both in




7


isolation and in interaction with each other. On the basis of literature reviewed, this research hypothesized that therapist behavior, client behavior and therapist responses to specific client behaviors all would differ significantly when terminator dyads were compared with retainer dyads.

Furthermore, Leary interaction scores can be interpreted in terms of complementarity-an important concept in process research (cf. Carson, 1969). Again on the basis of the literature review, it was hypothesized that differences in therapist-client interaction would include differences in the complementarity of therapist response to client behavior between the two groups of therapists.

Finally, process ratings using the Leary can be "fine tuned." The researcher decides how broad or narrow categories are to be both in terms of dimensional purity and in terms of intensity shown. Previous research suggested the use of broad scoring for dimensional location plus differentiation of two or three levels of response intensity. In this study, broad (quadrant) dimensional scoring was done, and two levels of intensity were distinguished.


Hypotheses

In general, it was hypothesized as follows:

1. Therapist dominance-submission and hostilityaffiliation will differ significantly between

initial interviews with clients who prematurely terminate and interviews with clients who remain

in counseling.




8


2. Client dominance-submission and hostilityaffiliation in the same interviews will differ

significantly when premature terminators are

compared with remainers.

3. The complementarity of therapist behavior to

client behavior along the dimensions of

dominance-submission and hostility-affiliation

will differ when the therapists of premature

terminators are compared with therapists

interacting with remainers.

Specific expectations about the nature of these differences are based in the related literature. Operational hypotheses will be formulated at the conclusion of the literature review.


Organization of the Study

Chapter Two consists primarily of a review of literature serving as background material for the study. This review begins by considering the problem of defining premature termination. It then moves into a review of research attempting to predict/explain premature termination. Nonprocess research is covered, first as general background and second to support later decisions concerning research design. The lack of substantial process research is documented. Next, there is a consideration of the general problems of doing therapy process research and a brief review of research using process to predict therapy outcome in general. The chapter ends with formulation of the research hypotheses.




9


Chapter Three covers the design and methodology of the research. Subject selection is discussed, and data collection procedures are detailed. Next, the Leary Interpersonal Checklist is examined for its validity and reliability. Its construction and its theoretical basis are covered, and the specific use made of it in this study is explained. Next, tape rating procedures and procedures for selecting and training the tape rater are presented. Finally, data analysis is discussed. Specific analyses are planned for hypothesis testing, and others are presented for use in answering other questions.

Chapter Four presents the data summaries and analyses. This

begins with an overview of the data obtained in the project as a whole and that used in the final study sample. Next, the data on which hypothesis tests were based and the results of those tests are given. Other planned analyses are reported. Finally, several post-hoc analyses are reported, and the reasons for their inclusion are mentioned.

Chapter Five interprets the data, discusses their implications in terms of past research, and delineates limitations of this study plus recommendations for future research. The first section of this chapter is devoted to the relationship between process variables and termination status. This includes the testing of research hypotheses and related analyses. The second section examines the relationship of nonprocess client and therapist characteristics to termination status. Agreements and disagreements between this study and others in the area are noted. The third section discusses general problems, most of which this study shares with others in the area. In some





10


cases, further research is suggested which could ameliorate the problems. Finally, general conclusions are drawn about the role of process in predicting therapy termination.















CHAPTER TWO
REVIEW OF THE LITERATURE


Premature Termination

The Definitional Problem

Premature termination appears to be a meaningful concept. It is used frequently by professional psychologists, and most seem to find its definition as intuitively clear as that of any other concept. If asked whether or not a client terminated prematurely, the typical practitioner answers, "Yes," "No," or "Maybe" and can provide reasons supporting that judgment.

However, intuitive meaningfulness does not equal researchability. Ideally, a research concept first has consensual validity and then is operationalized in a way which does not compromise that validity.

In the case of premature termination, definitional problems abound. Clinicians disagree with each other about the standards for judging a termination premature. Most researchers want objective criteria, while clinicians prefer to make subjective judgments. Some reviewers want the client's judgment to be considered in addition to those of the clinician and the researcher.

This lack of consensus is usual for concepts originating with therapists rather than theorists. However, it does dictate a reversal of the ideal procedure. In the absence of consensus, a researcher must choose from among the offered definitions, operationalize the chosen


11





12


definition conscientiously, and then record everything as clearly as possible. Only if others agree with those procedures will the work be generally meaningful.

In examining the literature for a definition of premature termination, one finds a range of usage. In some articles it refers to termination occurring for any reason before a given number of sessions are completed. One recent example is the work of Stern et al. (1975) in which premature termination was defined simply as failure to attend six consecutive sessions. Similar but more stringent tests were applied by Heilbrun (1974) and by Fiester (1977). They defined premature termination as failing to return for the second interview and the third interview, respectively.

At the other extreme, premature termination may be defined

entirely subjectively, based on the case therapist's judgment or on a combination of therapist and client judgments. An early example using therapists' judgments is Landis's study (1937) of psychoanalytic psychotherapy. More recently, Raschella (1975) published a study of "premature client dropout from therapy" in which prematurity was strictly a matter of therapists' judgments. Similarly, McGrogan (1976) used termination without therapist approval as his sole criterion of prematurity.

Landfield (1971) devised a set of criteria which are far more complex than those discussed previously but are equally subjective. According to Landfield, premature termination has occurred when

(1) both therapist and client agree that therapy goals have not been




13


met, and (2) therapist and/or client feel(s) that therapy could be beneficial, but (3) one or the other is unwilling to continue treatment.

The first, "objective," type of definition implies that premature termination is equivalent to early termination. Probably few clinicians would endorse such an equivalency. Nevertheless, it is important to consider whether-and how-the two can be differentiated from each other.

"Premature," as opposed to "early," termination implies a judgment about the timing involved. Some researchers sidestep this issue by avoiding the term "premature terminator." The most common substitute is "dropout." Recent examples include the work of Saltzman et al. (1976), of Fiester and Rudestam (1975), of Caligor (1975), and of Vail (1974). As the discussion section of Baekeland and Lundwell's (1975) review article indicates, however, this semantic sidestepping solves nothing. A proper definition of "dropping out" should specify how termination occurred, not just length of treatment.

Since length of treatment alone cannot define premature termination satisfactorily, a question naturally arises: What about leaving the decision to the case therapist? The answer is much debated.

One objection stems from research such as Rosen's (1978) in which counselor perception of termination was found to be unrelated to client perceptions of either therapy outcome or termination. Countering this is the fact that therapists coined the term in the first place, labeling a phenomenon they identified and considered conceptually important to them. It can be argued that premature termination is not a meaningful concept to most clients.





14


However, there is a potential problem for researchers with allowing either therapists or clients to judge termination status: noncomparability of classification criteria. The general solution seems obvious. Termination status should be decided on a case-by-case basis, but objective guidelines are needed if cases are to be grouped for the purposes of research. A number of researchers have put some effort into constructing such multiple criteria.

One approach is to have therapists contract with clients for a minimum number of sessions and then compare clients completing the contract with those who fail to complete it. Saltzman et al. (1976) did this using a fixed number of sessions, while Norkus (1976) allowed therapists and clients to write their own contracts. Each variant has its advantages and its disadvantages. Saltzman's approach differs little from the simple session-counting type of criterion; the only difference is that since the standard for prematurity is shared knowledge among clients, therapists and researchers, a client who terminates prematurely has decided to accept less than the recommended course of treatment. Norkus, on the other hand, allowed for individualization on a case-by-case basis but failed to insure any uniformity in contracting practices among her participating therapists.

Another general approach uses session-counting but increases

its sophistication by adding other criteria. For example, Sasseen (1976) defined premature termination as attending fewer than five sessions and terminating without contacting one's therapist. Sandler (1975) used the number of sessions attended plus therapist's judgment about





15


termination status. Winder et al. (1962) did essentially the same thing.

These two procedures have strengths and weaknesses analogous

to those in the preceding paragraph. That is, Sasseen's (1976) work is a definite improvement over the use of a fixed session criterion alone, but it remains somewhat arbitrary. For example, a client who quit in a fit of temper would not be defined as a premature terminator since the therapist was contacted. On the other hand, Sandler's (1975) method provides the flexibility necessary to correctly classify the unusual case but fails to assure uniformity beyond the general control provided by session counting. However, both procedures are much better than the methodologies discussed previously. They narrow substantially the range of possible error without introducing any impracticable requirements.

More complicated methodologies can refine the operational

definition of premature termination still further, but some seem to suffer from a diminishing return of benefits in relation to effort required. An example is Landfield's (1971) complex definition, previously detailed. It requires extensive querying of both therapists and clients for their views on the appropriateness of therapy in general, its completeness at termination, and their desires regarding termination. Even assuming honest and insightful answers to questions on such an emotionally charged subject, the results do not seem worth the effort. For one thing, the definition is entirely subjective. For another thing, it is overly inclusive. It would classify any therapy




16


failure as premature termination unless both therapist and client recognized the failure for what it was. Moreover, it would do the same for some cases in which the therapist (or the client) no longer considered therapy appropriate.

Jachim (1974) utilized a combination of criteria which avoids most of the pitfalls discussed so far. He defined premature termination both by number of sessions and by a subjective criterion, lack of therapist agreement. In addition, Jachim telephoned terminators to be sure that they had no extrinsic reason for dropping out. Best ofall, Jachim avoided the usual practice of maximizing n's by classifying every case, whether or not the classification is a meaningful one. His "terminators" left within four sessions, while his "remainers" stayed for at least eleven. Those cutoff points were arbitrary, chosen by discarding the middle third of all the cases sampled. Although the cutoff points are debatable, this is a laudable attempt to eliminate "noise" from the comparison of terminators and remainers.

In this study, premature termination will be defined in a way similar to Jachim's in that both objective and subjective criteria will be used and intermediate cases will be discarded. However, the cutoff points that serve as objective criteria will be chosen based on theoretical and empirical considerations.

A starting point is Landfield's (1971) observation that "Premature termination indicates that members of a therapy dyad are unable to maintain interpersonal communication" (p. 66). One way of insuring that this condition is met is to limit the designation "premature" to




17


cases which terminated so quickly that establishment of an interpersonal relationship of therapeutic quality is improbable.

Other researchers using this approach have chosen anywhere from one (Heilbrun, 1974) to nine (Saltzman et al., 1976) or ten (Winder et al., 1962) as the maximum number of sessions a person could attend and still be considered a premature terminator. Neither extreme seems desirable. Saltzman et al. observed themselves that most clients who terminated unilaterally did so before the sixth session. On the other hand, using one session as the upper limit for premature terminators is questionable, as there is some evidence that failure to return after the first session is a rather different phenomenon from early termination in general (Schiller, 1978).

The most frequently used cutoff point in the literature reviewed was failure to return for the fifth session (cf. Jachim, 1974; Sandler, 1975; Sasseen, 1976). The sixth session was used similarly by two others researchers (Schiller, 1978; Stern et al., 1975). However, with the exception of Jachim all of these researchers set up criteria which classified every case as either a terminator or a retainer. Had they allowed themselves the luxury of discarding intermediate cases, the cutoff points probably would have been lower.

In this study, premature termination was defined in part as failure to return for the fourth (or earlier) post-intake session. Following the conclusions of Saltzman et al. (1976), remainers were required to attend six or more post-intake sessions. Intermediate cases were discarded from the final sample.





18


In addition to these objective guidelines, subjective factors were considered as well. First, only clients referred by an intaker as appropriate for ongoing psychotherapy were included in the sample. Second, no client was considered a remainer if he or she canceled or skipped one of the first six sessions and returned only after a counselor initiated contact. Finally, counselors were querried for their opinion of each client's termination status. Clients were considered premature terminators only if they terminated against the counselor's judgment or without consulting the counselor. They were considered remainers only if the therapist concurred with the termination. Again, indeterminate cases were dropped from the final sample.

The procedures were unusually rigorous. It was hoped that they would insure a high level of validity in the final assignment of clients to categories of the dependent variable.


Premature Termination and
Psychotherapy Outcome

Premature termination has an uncertain place in psychotherapy research for reasons other than controversies over its definition. A major problem is deciding how its study relates to classical psychotherapy outcome research.

The earliest major study of psychotherapy outcomes recognized premature termination, though not by name. In the 1930 report summarizing ten years of work at the Berlin Psychoanalytic Institute, premature terminators were described as having "left treatment," and this was recognized as a distinct category of therapy outcome (Landis, 1937).




19


The concept of premature termination has not always been accorded this respect in more recent work. A prominent example, Eysenck's (1952) famous and damning summary, "The Effects of Psychotherapy: An Evaluation," consigned all early terminations to the category of therapy failures. Indeed, premature termination from therapy continues to be an embarrassment to researchers who want to locate all cases along a success-to-failure continuum. Some, following Eysenck, have deemed it a type of failure. Many simply have eliminated most cases by such means as dropping from their studies all clients who attend less than a given number of sessions. There still is no consensus as to the place of premature termination in outcome research.

A plausible argument can be made for treating voluntary premature termination as failure. Logically, one can argue, the therapist has failed to make the desired impact on the client if unilateral termination occurs. This logic is bolstered by the empirical observation that most factors which predict therapy failure (cf. Luborsky et al., 1971) also predict a high rate of premature termination (cf. Baekeland and Lundwell, 1975). These factors will be discussed individually in a later section of this paper.

In addition to general correlations, a coincidence of factors connected first with client dissatisfaction and second with premature termination occasionally has been established within the same subject sample. An example is Sasseen's (1976) conclusion that disconfirmation of clients' expectations and preferences predicts both premature termination and client dissatisfaction with the interview.




20


In spite of these logical and empirical arguments relating voluntary premature termination to therapy failure, the most crucial type of data remains sparse. Very few efforts have been made to evaluate directly the post-treatment status of premature terminators. The majority of such follow-ups conclude that most premature terminators are therapy failures (cf. Jachim, 1974). However, one elaborate investigation of dependent female clients concluded that many of them terminated not from frustration but due to symptomatic relief of anxiety following a cathartic therapy session (Heilbrun, 1974). A later investigator also found cathartic relief to be a factor in premature termination (Fiester, 1977).

At this point it seems that premature termination occurs for a wide variety of reasons. Even when voluntary and unilateral, it does not always imply failure. Particularly with low-readiness clients, it may mean that the client received the symptomatic relief he or she sought. Such a client might even return at a later time for more extensive treatment, but no research has been done on this to date.

This conclusion is worth serious attention because most theory and research assumes that premature termination is an expression..of client discontent and/or anxiety. Theoretically, this complexity means that it is pointless to attempt to assign premature termination a fixed spot on any success-to-failure continuum. Empirically, it means that conclusions valid for the majority (presumably the discontent/ anxious clients) are being weakened statistically by the heterogeneity of most "terminator" samples.




21


On the other hand, there is no support at all for the not uncommon practice of dropping premature terminators from outcome research. However inconveniently heterogeneous they appear, premature terminators definitely do not form a representative cross sample of therapy cases. Unless investigated in more detail, their termination itself is one type of therapy outcome.


Predicting/Explaining Premature Termination

Logic dictates treating premature termination as a type of

therapy outcome, and this logic has not been lost on researchers. Therapy researchers acknowledge the connection in a variety of ways. One of the more subtle is the overall similarity of premature termination research designs to those of general outcome research.

Historically, outcome researchers have focused most heavily on client factors as predictor variables. Therapist factors rank second. Investigations of interaction among client and therapist factors are relatively rare and so are designs using process factors as predictors of outcome. The same set of priorities is evident in research aimed at predicting/explaining premature termination.


General Factors

Client characteristics. To judge from the literature, professionals see premature termination from (or dropping out of) therapy to be largely the responsibility of the client. To begin with, the nouns "premature terminator" and "dropout" are used freely. Therefore, it is not surprising that the vast majority of predictor variables




22


investigated are relatively fixed characteristics of clients such as age, sex, socioeconomic status and diagnosis.

In 1975, Baekeland and Lundwell reviewed 62 studies of dropping out among adults in individual psychotherapy. Sixty of these studies found at least one significant predictor variable. The majority were client factors. Much replication remains to be done, as many successful predictor variables were investigated in only one study. However, a few were well substantiated by several studies utilizing different populations and different definitions of dropping out.

The most extensively investigated variable, socioeconomic status, was associated consistently with dropout rate. Whether defined by

education, occupation, or income, higher status predicted lower dropout rates in 16 of 18 studies. Race and IQ, generally associated with socioeconomic status, also are predictive in the expected direction (Baekeland and Lundwell, 1975).

Another consistently predictive variable is sex of client. All four studies reviewed found that females dropped out more often than males did. This is one of the few points at which remaining in therapy is (statistically speaking) a different phenomenon from making progress. Most outcome studies have found females to be the more successful clients (Luborsky et al., 1971).

Client age predicts dropping out somewhat less well. Three of four studies found age predictive, but their results are not completely consistent. In one study of brief psychotherapy, client retention rates varied directly with client age. However, a project studying longterm therapy found that people in their 30s were more likely to remain




23


than were either younger or older clients (Baekeland and Lundwell, 1975). Interestingly enough, most outcome studies find that, among adults, the younger the client the brighter the prognosis. Apparently younger clients, like female clients, are relatively likely to drop out, but those who stay make particularly promising candidates for therapy.

According to Baekeland and Lundwell (1975), diagnosis is somewhat predictive of dropping out. Five of nine studies found sociopathy, paranoia, alcoholism and extreme depression predictive of premature termination. Anxiety and moderate depression, on the other hand, improved the chances of remaining in therapy.

Related predictors studied only once each include referral

source, psychological mindedness, defensiveness, stereotypic masculinity or femininity, and personality disturbance in a significant other (Baekeland and Lundwell, 1975).

None of these findings has been challenged seriously by more recent research. However, some have been reinterpreted. In addition, more recent studies have added to the list of predictor variables.

Fiester and Rudestam (1975) found social class status predictive of premature termination rate at one facility studied but not at a second facility. The two facilities had similar dropout rates and served similar clienteles. They did differ in types of therapies employed and in therapy offered. Social class was predictive in a setting with older, traditionally trained therapists doing analytic therapy and was not predictive where more eclectic therapy was practiced by younger therapists with a variety of types of training.




24


Trait anxiety was supported as a predictor of continuation in

therapy by two recent studies. In one, Barr (1977) noted that it seemed to explain longer attendance in therapy by first-borns. In the second study, Stern et al. (1975) found that the predictive value of the Taylor Manifest Anxiety Scale remained even when the effect of social class was removed. At the same time they found that two other diagnostic variables-sociopathy (as measured by the Behavior Disturbance Scale) and authoritarianism (as measured by the California F Scale)-washed out when the effect of social class was removed.

In addition to his findings regarding trait anxiety, Barr (1977) added to the information relating client retention to diagnosis while making more questionable its connection with client age. In Barr's sample, the younger clients attended more sessions, Also, significant differences were found among diagnostic categories. Psychotics averaged longest in attendance, while people with marital problems left most quickly. It is important to note, however, that Barr simply averaged attendance figures to characterize his client groups. One cannot draw conclusions about comparative rates of premature termination from such data.

The same caution applies to a study in which Nacev (1977)

found dependency and ego strength (as measured by the MMPI) to be of no value in predicting length of stay in therapy. That caution is underscored by the work of Goldenholz (1975). Goldenholz separated "number of sessions" from "type of termination" and used both as dependent variables. She found no difference between males and females in the




25


tendency to terminate unilaterally. Males averaged more therapy sessions, however, a result the researcher attributed to their being more severely disturbed initially.

McGrogan (1976) screened demographic data, diagnosis, MMPI

scales, CPI scales, and therapist experience as predictors of number of visits and of termination with or without therapist's approval. In line with earlier research, he found demographic and diagnostic data to be his most powerful predictors. However, he observed that little of the total variance was explained in the multiple regression analyses. In fact, he recommended abandoning this type of research.

A final word about assessing client proclivities is provided by Koran and Costell (1973). In their investigation of the value of test scores, therapists' predictions, and so forth, the most valuable predictor of early termination was found to be refusal of the client to complete pretest questionnaires. The results were quite powerful: Refusal carried an 80 percent risk of premature termination, versus 17 percent overall, and it identified 53 percent of the early terminators.

Therapist characteristics. In 1975, Baekeland and Lundwell found only one therapist factor which predicted client retention in more than two studies. Interestingly, it is the sole reliable predictor of therapist success as well (Luborsky et al., 1971). The factor, of course, is therapist experience. In six of the seven relevant studies they reviewed, inexperienced therapists lost a significantly disproprotionate number of clients.




26


Only two other "pure" therapist factors predicted client retention in any study reviewed. In one study, therapists low in ethnocentricity held more clients. In two studies, female therapists had a significantly lower rate of premature termination than did male therapists (Baekeland and Lundwell, 1975). The other "therapist" factors reported actually are process-related factors, so they will be discussed later.

More recent research has little to add in this area. Its most

important result, probably, is to render even more questionable the tentative connection between therapist sex and premature termination reported previously. Andrews (1975) found no relationship between sex of therapist and number of sessions attended. Goldenholz (1975) got the same result. In addition, the latter also found sex of therapist to be unconnected with type of termination.

In the realm of recent null results, Butler (1977) found therapist training unrelated to clients' remaining in therapy. Fiester (1977) checked therapist demographic characteristics, training, and experience and found nothing predicting whether or not clients would stay past the second session. Of course, neither of these researchers actually studied premature termination.

Interaction of client and therapist characteristics. Some of the results in this area are intriguing. It is important to note that most have not been replicated. This is not damning evidence, however; for most, no replication has been attempted.

Two interaction variables form an interesting exception to that generalization. Each has been tested repeatedly for its value as a




27


predictor of premature termination, and the results consistently are negative. Neither predicts therapy success, either. Those null results so contradict the convictions of many therapists and researchers, however, that they continue to be questioned. The two variables, of course, are client-counselor matching by sex and client-counselor matching by race.

For example, Vail (1974) hypothesized that black, lower-class

clients would remain in treatment longer if matched for race and matched for sex. In fact, matching for race made no difference, and matching for sex actually was associated with early termination.

Using an unusually good experimental design, Safer (1975) tested sex-of-therapist by sex-of-client interaction for both weekly and biweekly therapy schedules. In earlier research, Safer had found that female-female pairs averaged the largest number of sessions. This time, such pairs were indistinguishable from male-male pairs in weekly therapy and had lower client retention when therapy was biweekly. In addition, Safer found that male therapists with female clients averaged more sessions than did female therapists with male clients.

Andrews (1975) found no differences at all for the sex-oftherapist by sex-of-client interaction. Mendelsohn (1966) also found none.

A related hypothesis attempts to explain high rates of premature termination among working-class clients as a matching failure. However, Butler (1977) found that social class origin of her therapists made no difference in the attendance patterns of working-class clients.





28


On the other hand, there is solid evidence that therapy or

therapist matching a client's preference makes early or premature termination less likely. In contrast to the null results obtained when similarity of sex is the independent variable, Norkus (1976) found that gratifying a client's preference for a male or female therapist made the client more likely to remain. All of her clients were female. Another researcher, who found no demographic variables predictive of termination type among his subjects, found that therapist behavior which matched clients' preferences or expectations made premature termination less likely (Sasseen, 1976). Sasseen was fairly careful with his terminology; to be defined as a premature terminator, one of his subjects had to attend counseling four times or less and terminate without contacting the therapist.

Ross's (1977) discussion of the possible meanings of clienttherapist similarity offers one explanation for the data presented so far in this section. Ross points out that (a) similarity probably matters only for characteristics of importance to the client, and

(b) the valence will depend on whether those characteristics are positively or negatively valued by the client. Incidentally, the latter is related to self-esteem, and many clients are symptomatically low in self-esteem, at least at the beginning of therapy. If a sample contains enough clients who are indifferent to the characteristic studied or who value it negatively, a matching hypothesis will be disconfirmed.

Demographic matching is poorly supported. On the other hand, a variety of researchers have found client retention superior when clients





29


and therapists are similar in some less obvious ways. Various personality variables have been used in this research. Similarity of goals for therapy and of role expectations also predict better client retention.

One early study of personality matching was done by Mendelsohn (1966). Mendelsohn tested clients and therapists using the Myers-Briggs Type Indicator. No main effects were found relating either client or counselor personality type to early termination. However, overall client-counselor dissimilarity strongly predicted early termination.

Landfield (1971) uses a complex set of criteria to categorize clients as remainers or terminators (see page 12). His predictor variables are complex, as well: content similarity of constructs and similarity in the organization of constructs on the Role Repertory Test. Both predict a client's remaining in therapy rather than terminating

prematurely.

One can argue plausibly that client-therapist similarity on therapeutic issues would be far more important than similarity on general characteristics for predicting the likelihood of a client's remaining in therapy. Whether or not that is the relevant dimension, a number of studies have found similar ideas of therapy goals and of expected roles to be reliably predictive of client retention.

Raschella (1975) found that client-therapist similarity in both the content of therapy goals and the priority of those goals made premature termination less likely. (Raschella used "dropping out" synonymously with "premature termination.") However, he also found




30


client-therapist similarity predictive of session length and of the spacing between sessions, suggesting that goal similarity is related

to client comfort-and possibly to therapist comfort, as well.

Butler (1977) studied retention of working-class clients.

Neither therapist training nor therapist background predicted client retention in her study. Retention was more likely when the client's problem was perceived similarly by client and therapist. It also was related to good matching between client expectations and therapist behavior.

Horenstein (1973) distinguished between "confirmation" and

"failure to disconfirm" expectations. For his college student population, specific confirmation of expectations regarding therapist insession behavior seemed unimportant. However, a clear violation of expectations was associated with increased tendencies to drop out or to come late or attend irregularly.

Sandler (1975) tested both therapists and clients pretherapy,

before the third session, and before the twelfth session regarding their role expectations. He found that terminators had significantly higher pretherapy discrepancy scores. Also, while discrepancies generally dropped over time, they dropped most for remainers and their therapists. Since Sandler used both number of sessions attended and therapist's judgment of termination type, these results can be properly considered to relate to premature, not just early, termination.

Based on his results, Sandler suggested either matching clients to therapists or using pretraining to align clients' expectations more




31


closely to those of their therapists. Client-therapist matching in this area has not been researched, but there is a small body of literature on the effects of pretraining. Its results divide fairly evenly between those showing a positive effect from pretraining and those showing no significant effect.

An example on the positive side is Martin's (1975) experiment using 40 community mental health center clients. Half were treated, using a videotape with both didactic and modeling content, and half were no-treatment controls. Viewing the videotape significantly reduced the rate of "unilateral termination."

On the negative side, Holliday (1978) used live, group preparation with half of his subjects and kept the rest as no-treatment controls. These were community mental health center clients also. The treatment had no effect on the dependent variable, number of sessions attended. Of course, this study differs from Martin's (1975) in at least two important ways. However, the inconsistency in results is typical for these studies as a group.

As was mentioned previously, sex-of-client by sex-of-therapist interaction has failed as a predictor variable. However, sex-of-client considered in combination with other client or therapist characteristics has led to some intriguing results.

One early investigation found that low-autonomy females assigned to dominant counselors and high-autonomy females with average counselors are less likely to terminate early. Neither combination made any difference for male clients (Heilbrun, 1961a). The same researcher





32


found that low-autonomy females are likely to terminate earlier than autonomous females, while the reverse is true for males (Heilbrun, 1961b).

Attempting to find an explanation for these results, Heilbrun replicated his earlier research and also asked clients to fill out a questionnaire eliciting their degree of satisfaction following initial interviews. He found that the dependent female clients were more satisfied with relatively nondirective counselors, whereas autonomous females were more positive about directive counselors. The reverse was true for male clients. The interview satisfaction seemed to be predicting dropping out! Heilbrun interpreted these results as supporting the hypothesis that premature termination, defined here as failure to return following the initial interview, is due at least in part to cathartic relief of anxiety produced by a satisfying interview (1974).

Baekeland and Lundwell (1975) also reviewed one study in which sex-of-therapist was crossed with Rorschach productivity scores of clients. There was a main effect for sex of therapist, with female therapists superior in client retention. However, it also appeared that female therapists were keeping primarily the less "productive" clients, while male therapists were better at retaining those with high productivity scores. Unfortunately, this work has been neither replicated nor tied to other research data, so interpretation is difficult. Client-Therapist Dynamics

So far, this section has concerned itself exclusively with

general characteristics and/or predispositions on the part of therapists and clients. What remains is to consider the dynamics of the specific





33


dyad involved in therapy. Within the body of research on premature termination these dynamics have been approached in two ways. Some researchers have looked at the client's reaction to the therapist or the therapist's reaction to the client. Others have used the thirdperson point of view and measured process variables.

Client reaction to therapist. The bulk of premature termination literature assumes that clients quit prematurely because of dissatisfaction with their therapists or with therapy. Nevertheless, it is rare for an investigator to test that assumption and rarer still for one to delve into the content of terminators' perceptions of their therapists. Those researchers who did so question clients got a few surprises, including evidence that there may be more than one major type of premature termination.

Most often, terminators did view their therapists more negatively than did remainers. Saltzman et al. (1976) said that dropout subjects saw their therapists as less respectful and understanding of them than did remaining subjects. Interestingly enough, the therapists' self-ratings differed in the same fashion.

Similarly, Caligor's (1975) dropouts rated their therapists

relatively unaccepting and argumentative. Like Saltzman et al. (1976), Caligor had data suggesting that this represented true differences in therapists' behavior, not just differing client perceptions. Therapists in Caligor's study who had many dropouts were rated relatively critically ("unaccepting," "distant") by both dropouts and remainers.

Rosen's (1978) more recent study got conflicting results. Global attraction to therapist was significantly related to





34


client-reported completeness of counseling. However, these client reports were unrelated to counselors' perceptions and also to number of sessions attended. Therefore, it is doubtful that Rosen measured anything more than general client positiveness toward the counseling experience.

More significant and challenging is Heilbrun's (1974) conclusion, based on a series of investigations reported previously, that termination after the first interview may signal satisfaction. The evidence was strongest for dependent female clients seeing nondirective therapists, but the data suggested that the same dynamic might be operating for other client-therapist matchings.

In the same vein, a pair of studies using multivariate analysis suggest that there really are at least two quite different types of dropout dynamics. Feister (1977) divided a group of 17 community mental health center therapists at the median by dropout rate and obtained demographic, training, and experience data for them. No differences appeared. He then gave clients Orlinsky and Howard's Therapy Session Report to complete after their first interviews. The clients of high-attrition therapists saw themselves as more inhibited and their therapists as more anxious and more directive than did the clients of low-attrition therapists. Two interactional factors appeared as well. The high-attrition group was relatively overstocked with (a) ineffectual therapists seeing confrontive clients and

(b) anxious therapists providing cathartic relief for their clients.

Fiester and Rudestam (1975) found more "positive" early

terminations. In their factor analytic study, several factors suggested





35


a limited sort of success. The clients were satisfied, saw their therapists as warm and effective, and terminated because they felt better. However, they also found a group of terminations in which clients, most of them seriously disturbed, were given inexperienced therapists. The clients were hostile in the interviews and dissatisfied afterwards.

Therapist reaction to client. Baekeland and Lundwell (1975) reported several therapist reaction factors to be predictive of clients' dropping out. Only one had appeared in more than one article as of their 1975 publication date, but all of them probably are related to each other.

According to Baekeland and Lundwell, three studies found therapist dislike of clients, globally defined, to be predictive of early and/or premature termination. Two other factors, investigated in one project apiece, also predicted dropping out. Those two are low expectation of client improvement (Goldstein, 1960) and lack of interest in the patient's problems (Baekeland and Lundwell, 1975). It seems reasonable to assume that most therapists prefer clients who they see as having a good prognosis and interesting problems, so these factors may be regarded as specific elements in global therapist attraction to client.

More recent studies have not been unanimous in connecting therapist attraction to client with client retention. Notably, Hochstadt (1975) found that, for his sample of community mental health center clients seeing either professional or paraprofessional counselors,




36


global attraction to client failed to predict the number of sessions clients would attend. On the other hand, Rosen's (1978) investigation using a sample of college students and therapists did find global attraction significantly related to client retention.

It is hard to reconcile those results. Possibly Hochstadt's

(1975) inclusion of paraprofessional counselors made a difference, but this is just a speculation.

Saltzman et al. (1976) provided one of the most detailed and methodologically sophisticated investigations of client perceptions in relationship to treatment duration. At the same time, they investigated therapists' perceptions of their clients. The researchers found that therapists saw themselves as less respectful, understanding, involved, and emotionally reactive to their predropout clients. In addition, they saw those clients as less open, less self-responsible, less likely to improve, and they saw their relationships with those clients as less continuous than their relationships with clients who turned out to be remainers.

It is important to note that all of the work in this section is correlational only. Prediction does not imply causality. One would expect experienced therapists to be less positive toward clients who are likely-for whatever reason-to be therapy failures or to reject therapy entirely. At the same time, it may be that clients respond to perceived therapist negativity by dropping out. This is a chicken-andegg problem. The answer may be that an unspecifiable mix of therapist and client perceptions of each other contribute toa therapy interaction in which client retention becomes progressively more unlikely.





37


Therapy process factors. In 1975, Baekeland and Lundwell found

only three studies which connected true therapy process with client retention rate. One simply established that gross disruption of the therapy process, through patient reassignment, made premature termination more likely. The other two dealt with more intricate variables.

Winder et al. (1962) studied therapists' responses to client statements. They found that therapist avoidance of dependency content in patients' verbalizations made premature termination more likely.

Heilbrun (1973), working with an entirely female sample of

clients, found that dropouts had disclosed more than remainers if their therapists were male. Among clients with female therapists, however, remainers were the more disclosing.

Neither of these studies has been replicated, and neither is very useful standing alone. Nevertheless, they are praiseworthy for their specificity.

Unfortunately, few later researchers have followed their lead regarding specificity. One did use a design much like that of Winder et al. (1962) to test the hypothesis that interpreting transferencebased behavior would increase clients' length of stay in therapy. His sample was composed of 39 clients seen by 17 social workers. The hypothesis was not supported at the .05 level of significance (Goldstein, 1975).

Most other research using process variables as predictors of premature termination has had those variables assessed by clients, by





38


therapists, or both. In other words, they are studies of clients' and therapists' reactions to each other.

One outstanding exception is a recent study by Schiller (1978). Schiller used tape recordings of sessions and had both client and counselor statements rated, using the Leary Interpersonal Checklist, on the dimensions of hostility-affiliation and dominance-submissiveness. Client statements, therapist statements and the complementarity of therapist statements to preceding client statements were checked for their power to predict early termination. Schiller found that early termination (attending five sessions or less) was significantly related to relatively low client-therapist complementarity. It was unrelated to either hostility or dominance expressed by either clients or therapists.

In 1970, Meltzoff and Kornreich commented forcefully on the

desirability of studying interpersonal process leading to premature termination from psychotherapy. To date, few researchers have heeded their call.


Studying Therapy Process

General Problems

Research critics agree that there is a woeful shortage of

process research in many areas where process seems of obvious relevance. There is a good reason for this, though. Doing competent process research is difficult. Data gathering must be planned in advance. It is relatively invasive, so client and therapist cooperation is critical. Obtaining process ratings usually means either direct observation or tape/transcript analysis. All are time-consuming and must be preceded




39


by rater training. Furthermore, useful ratings require good instruments. As of yet, few have been carefully developed and validated. Finally, any good piece of research is theoretically relevant and operationalizes theoretical constructs in a plausible manner. Existing operationalizations of many process constructs are controversial, to say the least.


Process as a Predictor of Outcome

In spite of the difficulties inherent in process research and those associated with studying psychotherapy outcomes, a body of research does exist investigating relationships between the two. Little of it touches specifically on premature termination. Nevertheless, it provides a most relevant background for discussion of therapy process

leading to premature termination.

Process variables derived from a theory of psychotherapy. At one time or another, diverse process variables have been tested for their relationship to psychotherapy outcome. Most have been the object of only transitory interest on the part of researchers, and at this point many results are contradictory. On the other hand, a few variables have been investigated with exceptional thoroughness. The more miscellaneous results will be presented first.

According to the research, interpretation may or may not facilitate psychotherapeutic progress. One highly regarded research project found a negative relationship between frequency of interpretations and therapy outcome (Sloane et al., 1975). Another respected researcher found outcome unrelated to the frequency of interpretations in general but positively related to the frequency of classical transference interpretations (Malan, 1976).




40


Confrontation also has received mixed reviews. Godbole and Verinis (1974) found it more effective than supportive interventions in treating the emotional disorders of physically ill geriatric patients. On the other hand, Padfield (1976) found no significant difference between the two in his study of depressed, lower-class women.

Catharsis generally is viewed as a positive process in therapy (Gomes-Schwartz et al., 1978), but recent research challenges that as a blanket assumption. Bierenbaum et al. (1976) found that the amount of catharsis experienced in therapy interviews was positively related to outcome for clients assigned to weekly, one-hour sessions. However, clients seen biweekly for half-hour sessions improved most when catharsis was minimal.

Another interesting study compared the effects on outcome of

three process variables, each originating in a different theory of psychotherapy. The clients, members of two therapy groups, rated each session for the amount of therapist acceptance, desensitization, and abreaction they experienced. Observers made similar ratings. Patientobserver agreement was high. Symptom amelioration was used as the sole dependent variable. In a newly formed therapy group, all three process factors were related to improvement, abreaction being the most important. For the ongoing group, only acceptance predicted patient improvement (Cabral et al., 1975).

Promising as some of those leads appear, none has been followed up, to date. In powerful contrast stands the substantial body of research concerning therapists' provision of Rogerian core conditions.





41


There are problems and flaws in the work itself, but at present it represents a unique achievement in the field of psychotherapy evaluation.

The Carkhuff Scales are a set of directions for observer evaluation of therapist performance along the critical Rogerian dimensions of accurate empathy, nonpossessive warmth, and genuineness. According to client-centered therapy, it is the therapist's provision of the "core conditions" which is the essence of any successful therapy.

This hypothesis has been tested in various settings and with

various types of clients and therapists. In general, results have been positive: High conditions are associated with improvement, while low conditions bring little change or deterioration.

A good example is provided by Schauble and Pierce's (1974) study of university student clients. Forty-one students and their 34 counselors were rated from early and late tapes of therapy. Counselors were rated on the Carkhuff Scales for empathy, positive regard, genuineness, and concreteness. Clients were rated for depth of self-exploration, locus of control, owning of feelings, commitment to change, and differentiation of stimuli. Clients were also pre- and post-tested with the Minnesota Multiphasic Personality Inventory. Comparisons (t-tests) between successful and unsuccessful groups showed that counselors differed significantly on all four scales for their late tapes and on all but empathy for their early tapes. Clients had no significant differences on the early tapes, but all measures differed significantly on the later tapes. Schauble and Pierce thus took the problem beyond





42


simple correlation and demonstrated that it was the counselors, not the clients, who differed at the onset of counseling.

A number of criticisms have been aimed at this body of work. Early criticism focused on rater inaccuracy and experimenter bias. Truax and Carkhuff (1967) reported that these objections were countered in later research. However, other problems remain.

One source of confusion is the fact that many Rogerians (and some critics, as well) apparently assume that core conditions are perceived similarly by clients and observers. Researcher results from projects using Carkhuff's scales for trained raters (or Truax's reworking of those scales) often are intermingled with those from projects using the Barrett-Lennard Relationship Inventory, a client-report measure. One of Truax's (1966) own studies invalidates such an equivalency, for Truax found that clients apparently do not need to perceive the therapist as warm, empathetic, and genuine in order to benefit from high levels of those conditions.

Also, factor analytic studies question the integrity and

separateness of the core conditions (Gladstein, 1977; Lambert, de Julio, and Stein, 1978). Given that the three conditions are highly intercorrelated, relatively stable across time, and independent of the patient, do they represent anything more profound than a general "good guy" factor (Gladstein, 1977)?

Another problem is evaluation methodology. Most studies

involve numerous outcome measures, many admittedly thrown in with no theoretical reason (Meltzoff and Kornreich, 1970). This approach to evaluation suggests that "core conditions" are seen as a panacea. While





43


more than a chance proportion show significant change in the predicted direction, the presence of numerous measures showing no difference does weaken overall results. The question, "What don't core conditions do?" is never answered, let alone interpreted (Gladstein, 1977).

This is not an exhaustive list of the criticisms which have been leveled at core conditions outcome research. Certainly many more specific criticisms can be directed at individual studies. On the other hand, it should be noted that this body of work has received an unusual amount of critical attention.

More important than specific results or criticisms are the

contributions made to process research in general by this major research effort. Rating scales were developed, checked for reliability, and validated empirically. The scales were used with a variety of populations: schizophrenics, delinquents, college underachievers, as well as the more usual clinic populations. Research procedures were refined to minimize or eliminate a variety of biases which go uncontrolled in the average study. For these reasons, the body of research as a whole forms a benchmark against which other studies of therapeutic process can be measured.

A-theoretical variables. Most therapy process research-including all of the aforementioned studies-uses process variables suggested by one or another theory of psychotherapy. In part, they serve to test the assumptions of theoreticians. Certainly this is a valuable endeavor. However, every such study faces one difficult dilemma. If it examines only therapy of the school from which its constructs are drawn,




44


generalizability is limited. If, on the other hand, a variety of therapeutic approaches are included, the researcher risks obtaining only null results. For example, the level of Rogerian core conditions provided may be quite unrelated to outcome in Gestalt therapy.

One solution to this dilemma is to use process variables which

are relevant to therapy in general rather than to a particular school of therapy. Another approach is to treat therapy simply as dyadic interaction and use variables found by social psychologists to describe critical aspects of such interaction.

The first approach was taken in a recent study by Gomes-Schwartz (1978). The methodology of that study is exceptionally good. Relevant, a-theoretical process variables were used: patient exploration, therapist exploration, patient participation, patient hostility, therapist warmth and friendliness, negative therapist attitude, and therapist directiveness. Client variability was limited to males with depression and social introversion. The therapists all were very experienced and belonged to distinct categories: dynamic psychiatrists, Rogerian psychologists, and "inherently helpful" professors. Client assignment was random, and therapy lasted a defensible length of time: 17 sessions, on the average. The outcome criteria were equally good: therapists' ratings, observer clinicians' ratings, patients' ratings, and MMPI scores.

Predictable process differences were found among the three categories of therapists. However, all types were equally effective. Moreover, the same relationships between process variables and outcome




45


variables obtained across therapist categories. Nothing predicted patient-rated change or MMPI change. Several variables predicted therapist and observer-rated changes. However, with the effect of other

processes partialed out, only patient involvement correlated with any change measure at the .05 level of significance. It correlated with four change measures.

The author notes that these results support Frank and his coworkers at the expense of both Rogers and Freud (Gomes-Schwartz, 1978). In addition, the doctrine of client responsibility is bolstered, though the inclusion of only very experienced therapists controlled therapist variability to a significant extent.

A second approach to studying psychotherapy process without

recourse to variables linked to one or another theory of psychotherapy involves using variables found by researchers in other areas of psychology to be central to dyadic interaction. This second approach was taken by Leary (1957) when he constructed the Leary Interpersonal Checklist.

The Leary Interpersonal Checklist locates behavior on a circumplex formed by two orthogonal, bipolar dimensions: dominance-submission and hostility-affiliation. These dimensions are based in social psychology research. Foa (1961) has documented the background research, reviewing a variety of studies in which factor analysis was applied to behavioral observations. For group members' ratings of each other, three bipolar factors tended to appear: individual prominance and achievement, sociability, and aiding group achievement (Carter, 1954). In studies of dyadic behavior, however, the third disappeared while the





46


first two remained under various names: "individual assertiveness" and "sociability" (Borgatta et al., 1958); "control-autonomy" and "love-hostility" (Schaefer, 1959).

In addition to locating behavior on the circumplex, the Leary Interpersonal Checklist provides a means for assessing the complementarity of one person's behavior to that of another, either in general or on an interaction-by-interaction basis. Complementarity scores are included in much of the therapy process research utilizing the Checklist.

Mueller and his coworkers performed some of the earliest

research using the Checklist to measure psychotherapeutic process. They were interested in methodological problems. Also, they wanted to know whether or not Leary's assumptions about behavioral reciprocity held true for psychotherapeutic interviews.

In the first project, ten-minute segments of taped interviews were scored, with a major (overt) theme and a minor (subtle) theme named for each interaction unit. Scores were grouped by quadrant. Analysis was restricted to comparisons of people's response totals by quadrant. The researchers found, for example, that the more frequently a therapist's major themes fell in the hostile-dominant quadrant the more hostile-dominant and the fewer affiliative-submissive major themes the client would exhibit. This procedure gave the research problems with interjudge reliability. In particular, coding of minor themes tended to be unreliable (Mueller and Dilling, 1968). Moreover, no conconclusions could be drawn about microprocess from this sort of analysis.

In a second project, Mueller (1969) changed several procedures. Raters were allowed to score as many themes per statement as seemed




47


appropriate to them. Instead of looking at totals per subject, complementarity was scored on an interaction-by-interaction basis. Interjudge reliability was figured likewise, and reliability improved.

Moreover, Mueller analyzed the data once with themes grouped by quadrant and once with grouping by octant. Mueller preferred the use of octants, pointing out that more significant relationships were found that way (18 versus 8). Of course, the percentage of possible relationships found significant was higher when scoring was done by quadrant (50 percent versus 28 percent). A stronger point in favor of octant scoring is its ability to establish more specific complementarities than those detected by quadrant scoring.

Two associates of Mueller's have taken the additional step of attempting to connect therapist and client behavior with success in psychotherapy. Crowder published his results in 1970, and Spierling

followed suit in 1972.

Crowder (1970) predicted that out-of-role behavior on the part of either therapist or client would make successful therapy less likely. He defined "reality-based," in-role therapist behavior as that falling in the affiliative-dominant quadrant and defined "reality-based" client behavior as that rated affiliative-submissive. Like Mueller (1969), Crowder did not score reflexes for their intensity.

Crowder's (1970) hypothesis was not supported -at the .05

level of significance. However, Crowder did find significant differences in both therapists' and clients' behavior when he compared successful and unsuccessful cases. In successful cases, therapists were more likely




48


to be hostile-dominant and less likely to be hostile-submissive during early sessions. During closing sessions they were more likely to be affiliative-dominant and less likely to be either hostile-dominant or hostile-submissive. Successful clients were more hostile-dominant or affiliative-submissive and less hostile-submissive or affiliativedominant during early sessions than were unsuccessful clients. During mid-therapy, they continued to be more affiliative-submissive and less hostile-submissive. No client differences appeared during closing sessions.

These results are not easy to interpret. The most clearcut pattern suggests that hostile-submissive (otherwise known as passiveaggressive) behavior, on the part of either therapist or client, is associated with unsuccessful therapy.

It is interesting to note that Crowder (1970) found both

client and therapist process predictive of therapeutic outcome. In a study reviewed earlier, Schauble and Pierce (1974) found poor process in the early stages typical of the therapists-but not the clients-of unsuccessful dyads. By contrast, Crowder's work suggests that therapists and clients share the responsibility for failure.

Spierling (1972) examined clients' reports of their relationships with others as well as the in-therapy behavior of both clients and therapists. Clients' reports were taken from the early stage of therapy, while in-therapy behavior was sampled from early, mid-therapy, and late sessions. Ten successful and ten unsuccessful client-therapist pairs were studied.





49


Spierling (1972) compared (1) similarity between self-reported client behavior toward others and observed client behavior toward the therapist, by outcome and by therapy stage, and (2) similarity between client-reported behavior received from others and observed therapist behavior toward the client, by outcome and by therapy stage. In addition, all analyses which included information on reported interactions with significant others were performed twice, once including parents and once excluding them.

In the first comparison, no main effect was found for either

therapy success or therapy stage. However, one significant interaction appeared. Successful clients behaviorally differentiated their therapists more from nonparental others as therapy progressed, while unsuccessful clients did not.

The second comparison found one main effect and no interactions. As therapy progressed, all therapists tended to treat clients more as clients had reported others treating them (Spierling 1972). Since the reporting was taken from early sessions only, this result documents a significant increase in countertransference as therapy progressed. The lack of either a main effect for outcome or an interaction of outcome with therapy stage again underscores the importance of client, rather than therapist, behavior in the production of psychotherapeutic success.

A better known outcome study using the Leary Interpersonal Checklist was done in 1975 by Dietzal and Abeles. Dietzal and Abeles focused on the complementarity of therapists' responses to clients' statements. They defined complementarity in Leary's terms so that, for example, a




50


hostile-submissive statement is complemented by a hostile-dominant reply.

Dietzal and Abeles (1975) found that complementarity levels of

early sessions and of closing sessions had no relationship to therapeutic outcome. However, during the middle stage of therapy those cases which would later be judged successful were characterized by significantly less complementarity than were unsuccessful cases. Thus, they found successful therapists more likely to avoid the high probability response during mid-therapy.

Seen another way, Dietzal and Abeles's results could be interpreted to mean that successful cases are characterized by more interpersonal control on the part of therapists. Giving a noncomplementary response is refusing to follow the other's lead.

However, a note of caution is in order. It is important to realize that both clients and therapists stay in-role [as Crowder (1970) describes it] most of the time. That is, the majority of client statements are affiliative-submissive, and most therapist statements are affiliative-dominant (Schiller, 1978). Moreover, since those two quadrants are complementary, the complementarity of therapist-client interactions would be expected statistically to be above the theoretical average.

This observation brings us to the recent study of Schiller's connecting client-therapist behavior and length of therapy. Schiller used the Interpersonal Checklist to rate three five-minute segments from each audiotaped intake interview. Ratings were made by quadrant.




51


Research hypotheses asked whether length of therapy could be connected with therapist process, with client process or with therapist-client complementarity.

Schiller (1978) got highly stereotyped behavior. Ninety-six percent of his therapists' responses were judged affiliative-dominant, and 72 percent of the clients' statements were judged affiliativesubmissive. Furthermore, 72 percent of the response pairs were fully complementary. These figures are higher than those obtained by other researchers (cf. Mueller, 1969).

Schiller (1978) distinguished three outcome groups: nonreturners (attended intake only), early terminators (attended one to five postintake sessions), and remainers (attended six or more sessions after intake). No therapist or client behavior predicted any outcome. However, early terminators enjoyed significantly less complementarity with their therapists than did either nonreturners or remainers. With its use of "early termination" as an outcome category, Schiller's research currently comes closest to using the Interpersonal Checklist to predict premature termination.


Hypotheses

Studies connecting psychotherapeutic process with premature

termination are virtually nonexistent. Therefore, it has been necessary to refer to related research in order to obtain guidelines for choosing independent variables. Recent studies connecting therapeutic process either with early termination or with general therapy outcome have been





52


reviewed. Based on that review, the following choices have been made.

First, it seems desirable to use independent variables which are not associated with any particular school of psychotherapy. For one thing, therapists at the target facility vary considerably in their theoretical orientations. For another, the field is so wide open that designing studies for maximal generalizability also maximizes their potential usefulness.

The literature reviewed strongly suggests as independent variables the two bipolar dimensions operationalized in the Leary Interpersonal Checklist. The two dimensions have emerged in several factor analytic studies (Foa, 1961), and Leary's (1957) operationalization of them is well tested for validity and reliability (see Chapter Three).

The hypotheses cover three categories of predictor variables:

client behavior, therapist behavior and the complementarity of therapist behavior to preceding client behavior. Specific hypothesized relationships with premature termination are based on research results obtained by Crowder (1970, 1972) and by Schiller (1978).

Hypothesis 1:

Weighted Percentages of Interaction in the hostile-submissive quadrant will average higher

for premature terminators than for remainers.

Hypothesis 2:

Weighted Percentages of Interaction in the affiliative-submissive quadrant will average higher

for remainers than for premature terminators.




53


Hypothesis 3:

Weighted Percentages of Interaction in the

hostile-submissive quadrant will average higher for

the therapists of premature terminators than among

the therapists of remainers. Hypothesis 4:

Weighted Percentages of Interaction in the

hostile-dominant quadrant will average higher for

the therapists of remainers than they will among the

therapists of premature terminators. Hypothesis 5:

Therapist Complementarity of Response for

Dominance scores will average higher among therapists

interacting with remainers than they will among the

therapists of premature terminators.

Hypothesis 6: Therapist Complementarity of Response for

Affiliation scores will average higher among therapists interacting with remainers than they will

among the therapists of premature terminators.















CHAPTER THREE
METHODOLOGY


Subjects

A sample of 30 student clients and their therapists was

selected. The primary selection criterion for clients was application and acceptance for ongoing individual psychotherapy at the Psychological and Vocational Counseling Center of the University of Florida. Since this research was intended to uncover predictive factors found in the first postintake session, clients who had received other ongoing therapy within the previous year were excluded. Clients enrolled in the university's counseling psychology program or counselor education department also were excluded because of the possibility of their being known to the experimenter and/or the tape rater.

From the pool of eligible clients, clients were further

selected based on their willingness to participate and on their assignment to a participating therapist. Inclusion in the final sample depended as well on therapists' completion of the research materials and on the need to balance the sample for type of termination.

The therapist sample included all willing staff (senior staff, interns, practicum students and other trainees) at the Psychological and Vocational Counseling Center except for the researcher herself. No therapist was paired with more than three clients, a stipulation designed to limit the influence of therapist idiosyncracies on the data.


54





55


Data Collection

At the time of intake, potential research subjects were given a consent form explaining the data collection in general terms. Intake therapists were instructed to discuss the consent form with their clients, if necessary, and to obtain clients' signatures if possible. In addition, clients completed standard intake forms which include demographic data.

The first postintake session of each subject was audiotaped by the therapist. The taping was presented as routine in order to minimize reactive bias (see Hollenbeck, 1978). Fifty-nine clients and 24 therapists participated in this initial phase of the study.

Therapists kept attendance records recording each appointment and whether or not the client attended. Clients were considered to be continuing in therapy so long as they attended their scheduled appointments or rescheduled any they missed. However, if a therapist initiated rescheduling (by contacting a client who canceled or failed to appear), client intentionality was considered debatable; such cases were discarded from the final sample unless the client already qualified as a remainer or unless termination occurred before the fourth postintake session in spite of the therapist's efforts.

After termination, the therapist completed a section at the bottom of the attendance record giving his or her view of how termination took place.

Data collection proceeded until 15 cases of premature termination and 15 cases of a client's remaining in therapy were completely documented. The final study sample was drawn to include all 15 cases





56


of premature termination in which taping had been adequate and 15 cases randomly selected from among the 25 with audible tapes where criteria for remaining in therapy had been met. This final sample included 16 therapists, each working with one, two, or three clients.

The 30 cases in the final study sample are listed in Appendix D by study number. For each case, client sex, race, years of education completed, and parental occupation are listed. Therapist sex, race, and experience level are provided, and the termination status of each case is specified.


Instrument:
The Leary Interpersonal Checklist

The Leary Interpersonal Checklist has an unusual combination of virtues which recommend it for analysis of therapy process. For one thing, it is solidly grounded in a theory of personality, Leary's Interpersonal Theory (Leary, 1957). Second, it was constructed rationally, rather than empirically. That makes validating research necessary but assures that any research done will do more than simply duplicate instrument construction procedures. Furthermore, the Leary has been well validated in a variety of studies. Finally, it already has been used in a variety of psychotheraphy research. Therefore, any further studies using the Leary can be placed within an ongoing field of research.


Theoretical Basis

Leary's Interpersonal Theory assumes that

What a person does in any social situation is a function
of at least two factors, (1) his multilevel personality





57


structure and (2) the activities and effect of the "other one," the person with whom he is in interaction. (Leary,
1957, p. 91)

It goes on to say that the basic unit of behavior is the "interpersonal effect" or "interpersonal mechanism." This is defined as the interpersonal function of a unit of social behavior. For example, "He is rejecting help" is an interpersonal effect. Leary categorized these effects into a list of "interpersonal reflexes" (Freedman et al., 1951).

Leary's (1957) theory is interactional in two senses. First, it specifies interpersonal interaction as the arena in which "personality" has most meaning. Second, it focuses on interaction as a source of personality. According to Leary, each interpersonal reflex has a reciprocal which it tends to evoke. Therefore, interpersonal exchanges both create and reinforce behavior/personality.

In Leary's theory, personality is described using just two

dimensions, hostility-affiliation and dominance-submissiveness. However, it is seen as existing at five levels. Level I is the level of public communication: what people do when others are present. The others are the levels of II, conscious self-description; III, private symbolization such as fantasies; IV, unexpressed unconscious; and V, values.

Although he always expected individuals' ratings to vary

across levels, Leary considered it very important to be able to assess personality functioning comparably across levels. For example, connections between Levels I and II can be related to Harry Stack Sullivan's focus on the relationship between self-perception and consensual validation (Leary and Coffey, 1955).




58


Leary never has claimed that one level is more "real" than

another. However, he has described discrepancies among the different levels as resulting in part from the operation of defenses (Leary, 1957).

Finally, Leary has described reflexes as so automatic that

they generally are not available for direct self-report. They tend to operate as a background to the verbal content, and the individual may well be unconscious of them.


Construction

Leary (1957) tried five methods for measuring personality at Level I. Only two proved satisfactorily consistent: psychologists' minute-by-minute ratings and more global ratings from those interacting with the subject. Finally, Leary developed the Interpersonal Checklist to systematize the measuring ofinterpersonal reflexes.

The Checklist was constructed rationally. Leary and his

colleagues assembled a list of 334 adjectives from psychology literature. These were narrowed down to 106 "generic interpersonal motives" by the pooled judgments of five psychologists. They aimed for a complete and balanced representation of interpersonal behavior. Leary converted as many as possible to transitive verb phrases, but many adjectives remain. The descriptors then were grouped into 16 themes,

each reflected in both "normal" and "pathological" behavior.

Within each theme, the adjectives were arranged according to "intensity." Intensity (or pathology) was operationalized as the inverse of the frequency with which an adjective was used by subjects to describe themselves or others. Four levels of intensity were





59


suggested theoretically. (In practice, it is the rare researcher who uses more than two.) Testing the list on hundreds of subjects also led to elimination of words which were too difficult for some subjects or which were interpreted ambiguously (Freedman et al., 1951; LaForge and Suczek, 1955; Leary, 1957).

At one point, Leary and Coffey (1955) stated that division of behavior among the 16 themes was best for both clinical and research purposes. However, they admitted that combining them into eight categories might be far more convenient. In fact, octant scoring quickly became standard in research, including Leary's own (Shannon and Guerney, 1973). Many projects studying therapy interaction use quadrants.

As was mentioned previously, Leary (1957) envisioned each

interpersonal reflex as the sum of two orthogonal vectors, hostilityaffiliation and dominance-submissiveness. These vectors are plotted on an x axis and a y axis, with point 0,0 representing neutrality on both dimensions. However, in his visual representations of the dimensions Leary has drawn a circle rather than the square which would be necessary to encompass all possible points were the vector addition actually being performed. This apparently was done to keep all extreme points equidistant from the center. The result is a pie sliced into eight pieces (or 16 pieces or four pieces, depending on the researcher). Behavior located toward the point of a slice is normal, of mild intensity; out toward the crust it becomes more intense, pathological (see Appendix A).




60


Leary (1957) has detailed two types of ratings which can be obtained from administration of the Checklist. One is characterization of a given unit of behavior by the name of the octant (or other division) within which it falls. The octants are managerialautocratic (high dominance, neutral affiliation); competitivenarcissistic (dominant and hostile); aggressive-sadistic (neutral dominance, high hostility); rebellious-distrustful (submissive and hostile); self-effacing-masochistic (high submission, neutral affiliation); docile-dependent (submissive and affiliative); cooperativeoverconventional (neutral dominance, high affiliation); responsiblehypernormal (dominant and affiliative).

Six of these octants have been connected to nosological classifications. Managerial-autocratic and competitive-narcissistic individuals have not been so classified. Leary and Coffey (1955) speculated that those individuals simply are too autonomous to seek therapy.

Alternatively, octant or other ratings may be used to measure a subject on just one of the two dimensions. When this is done, scores in "pure" octants are given more power than are scores in "mixed" octants. The ratio is 10:7 (Leary, 1957). This is consistent with Leary's use of a circle, rather than a square, to represent the field of interpersonal behavior.

The Interpersonal Checklist is used to obtain Level II (selfreport) as well as Level I (observer-rated) data. However, as a selfreport instrument it has different problems of validation and a




61


different body of literature reporting its use. From here on, this review will focus on the use of the Interpersonal Checklist as a source of Level I data.


Validating Research

As a rationally constructed measure, the Leary Interpersonal Checklist needed to prove its empirical validity, as well as its construct validity and its reliability. If the observer-rated Checklist is fully valid, five questions can be answered affirmatively. First, do the items belong in their assigned categories? Second, do those categories relate to each other empirically as their (rationally assigned) position indicates that they should? Third, do scores on the Checklist relate to measures of other levels of personality as Leary theorized that they would? Fourth, do scores on the Checklist dimensions correlate highly with other measures of hostility-affiliation and of dominance-submission? Finally, are scores reliable?

At this point it is important to distinguish between two methods for using the Checklist to gather Level I data. Both were sanctioned by Leary (1957). In one system, global ratings are made using the Checklist in the same fashion that it is used for self-ratings. Raters usually are relatives or acquaintances of those rated. In the other system, the Checklist is used as a process measure. Trained raters categorize behavior on an interaction-by-interaction basis.

These two very different uses of the same measure never have

been distinguished from each other very clearly. Leary (1957) apparently assumed that they were tapping the same phenomenon (Level I





62


personality) since he speculated about which was the more valid. In current literature both are given the same name, and one must read an article closely to see which method was used. Most research has been done with global ratings. This review will include studies using both types of ratings. The data source of each will be specified.

Category integrity and arrangement. Only one study was

located which checked the validity of item assignment on a category-bycategory basis. McCormick (1977) had 234 college students sort the Checklist items. He found a high level of agreement between his student raters and Leary's panel of psychologists. This study also supported the circular arrangement of categories since students not assigning an item to Leary's category most often placed it in one adjacent.

The circular arrangement of categories was validated by another method when the Leary Interpersonal Checklist was constructed. Item intercorrelations were compared by LaForge and Suczek (1955). They reported "an orderly descending relationship as one moves around the circle from any reference point" (p. 109).

In his review, Beutler (1965) accepted this evidence of circularity. However, he pointed out that (1) average endorsement of the categories is uneven and (2) social desirability is uneven also. In other words, the circle is somewhat warped. It needs standardization.

Dimensional integrity and orthogonality. The factor structure of the Interpersonal Checklist has been the subject of repeated testing and much discussion. This probably happened because an early study found three main factors rather than two. This finding challenged





63


Leary's theory as well as the validity of the Checklist. All research discussed in this section used global ratings.

In 1963 Briar and Bieri factor analyzed 250 self-ratings using the Checklist. They got three factors: "aggressive dominance," "love," and "masochistic inferiority." They then checked this level II data against Level I data by asking 40 social workers to rate a set of profiles constructed to illustrate the polar positions of Leary's two factors. Again, their data supported the existence of three factors.

To counter this conclusion, Lange (1970) developed two telling arguments. First, he presented evidence of a two-factor structure emerging from other researchers' analyses of behavior. Second, he noted a flaw in Briar and Bieri's methodology which could account for their discrepant results.

Foa reviewed a variety of studies in which factor analysis was applied to behavioral observations. For group members' ratings of each other, three bipolar factors tended to appear: individual prominance and achievement, sociability, and aiding group achievement (Carter, 1954). In studies of dyadic behavior, however, the third disappeared while the first two remained under various names: "individual assertiveness" and "sociability" (Borgatta et al., 1958); "control-autonomy" and "love-hostility" (Schaefer, 1959). Foa (1961) concluded that, ". . a circumplex structure around the two orthogonal axes of Dominance-Submission and Love-Hostility" appears to be the best description of empirical analyses of behavior (p. 348).

The reviewer Beutler agreed that Foa's (1961) conclusion holds if one corrects for subjects' "tendency to agree with anything"





64


(Beutler, 1965, p. 268). The importance of making this correction was underscored by Truckenmiller more recently. Truckenmiller used Briar and Bieri's (1963) methodology and replicated their results (Truckenmiller, 1971).

Lange (1970) has pointed out that correcting for average intensity of ratings, a further refinement, makes the argument for a twofactor structure stronger still. He added an "inferential investigation" to earlier factor analytic analyses, gathering data in the form of ratings made by undergraduates watching taped enactments by characters designed to represent one of the four polar extremes of Leary's two dimensions. His data support both the integrity and the orthogonality of those dimensions.

One final assumption of Leary's (1957) remains to be covered in this section. Leary theorized that the poles of his two dimensions can be placed in such a way that one bit of behavior is judged for its complementarity to another bit. Specifically, he theorized that hostileaffiliative behavior is complementary to itself but dominant-submissive behavior is complementary to its opposite.

Only one study was located which provided a fairly thorough

test of Leary's assumptions regarding response complementarity. In it, 14 groups of six women each were given three questions to discuss. The researchers concluded that responses, categorized by octant, definitely were nonrandom. Not all of the associations predicted by Leary were obtained, but many were. For example, docile-dependent behavior did tend to elicit leadership-advice, and the reverse was true as well.





65


However, self-enhancing-competitive behavior did not elicit its complement, and neither did aggressive-rejecting behavior (Shannon and Guerney, 1973).

Shannon and Guerney also tested the complementarity theory on the person-to-person level. In doing this, they were asking whether a person who frequently emits x behavior receives y responses with exceptional frequency or infrequency. Of the ten response pairs found to be significantly related in the first analysis, seven were significant in the person-to-person analysis as well. Leadership elicited cooperative behavior; competitiveness and aggressiveness both elicited more competition and aggression; docility both elicited leadership and made a docile response unlikely.

Shannon and Guerney's data discourage an unquestioning acceptance of Leary's theory of complementarity. However, their results should not be regarded as definitive. For one thing, coding was based on typescripts rather than the tapes Leary considered preferable (Leary, 1957).

One further possibility emerges from the fact that Shannon and Guerney (1973) allowed neither extra credit for exactness or partial credit for near misses. Perhaps octant scoring divides responses into

either too few or too many categories for significant relationships to emerge. The "too many categories" hypothesis is favored by the fact that Shannon and Guerney had to discard one stimulus octant for poor intercoder reliability and two stimulus octants plus two response octants for infrequency.

Structural similarity among the levels of personality. The existence of five levels of personality is central to Leary's




66


Interpersonal Theory. Therefore, it was very important to him that his Interpersonal Checklist facilitate the process of drawing comparisons between and among levels (Leary and Coffey, 1955). The validity of such comparisons is a subject for empirical investigation.

In general, the results have been positive but limited in scope. The comparability of Level I and Level II data has been established most securely. Comparisons between Level I and other levels are more risky. Again, these researchers restricted themselves to obtaining global ratings, so no conclusions can be drawn about the Interpersonal Checklist used as a process measure.

Zimmerman and Vestre (1975) compared patients' self-ratings on

the Checklist with ratings made by family members of each patient. They found the factor structure to be the same. There was little content agreement between patients and their relatives, supporting Leary's

contention that self and observer ratings represent quite different aspects of a person.

Truckenmiller (1971) did a similar study using normal subjects.

He compared self-ratings with ratings made by subjects' sorority sisters. In addition, he gathered Level III data by administering the Thematic Apperception Test. Truckenmiller got three trait factors (see previous section) which matched across Levels I and II. However, only one of those traits emerged from the T.A.T. data.

Too, Briar and Bieri's (1963) study supports the contention that Level I and Level II data are comparable. This research differs from Truckenmiller's in that self-ratings on one population (live subjects)




67


were compared factor analytically with ratings from quite a different source (social workers' ratings of constructed profiles). The comparability of factors across these sources suggests that Leary's factors may arise not from the structure of human personality but from the structure of our conceptions of people. A cross-cultural study would be an interesting addition to the literature at this point.

Concurrent validity. No research was found which correlated

scores obtained on the Interpersonal Checklist with any other measure(s) of dominance, submission, hostility or affiliation. The concurrent validity of those dimensions, then, remains untested.

Reliability. The uses made of the Interpersonal Checklist for gathering Level I data present different problems in reliability. When the Checklist is used globally, split-half reliability and test-retest reliability are most relevant. When ratings are made on an interaction-by-interaction basis, interjudge reliability is a major concern.

Armstrong (1958) tested the split-half reliability for global ratings. In fact, he checked the reliability of Level II data (selfratings) and Level IV data (ideal self-ratings) as well as that of three Level I ratings for each subject. All 100 of his subjects were male; half were alcoholics and half were "normal." Using the KuderRichardson estimate of reliability, Armstrong obtained very high reliability coefficients ranging from .953 to .876.

Interjudge reliability for interaction ratings is far more

complex. It is common for studies using the Interpersonal Checklist in




68


this fashion to report reliability figures. However, as Mueller and Dilling (1969) have pointed out, these figures can come from a wide variety of places. One could report (1) the percentage of scores which agree, across items; (2) coefficients of item agreement; or (3) rankorder correlations based on the percentage of reflexes occurring within a given number of units. Also, any of these methods can be based on

(a) reflex agreement, (b) section agreement (octant, quadrant, etc.), or (c) agreement within a given range, for example, scores coming within one reflex of that marked by the criterion judge. Furthermore, any combination of the above could be made far more complex by considering the intensity of each rating. Still another complication is introduced if raters are permitted to ascribe more than one reflex to a given unit of communication.

Mueller and Dilling (1969) claimed that their literature review made 70 percent item agreement at the reflex level look feasible. However, Mueller's (1969) own study of therapist-client interaction, published the same year, reported only 49.4 percent item agreement at that level. Mueller obtained 69.1 percent agreement when he allowed ratings to vary within one reflex of each other and 76.1 percent agreement when the permissible range was extended to two reflexes in each direction.

The reliability figures actually obtained by Mueller are similar to others found when therapist-client interaction has been studied. Dietzal and Abeles (1975) obtained 82 percent mean agreement for responses classified by quadrant. Crowder (1972) reported 75 percent agreement for quadrant scoring.




69


Use in This Study

In this study, the independent variables are measures of client behavior, of therapist behavior, and of therapist response to client behavior. All were obtained by using the Leary Interpersonal Checklist as a process measure.

Both the literature review in Chapter Two and the preceding

description of validation studies mention that the Leary Interpersonal Checklist has been used in a variety of ways by different researchers. It is a very flexible instrument, and that flexibility gives the researcher an interesting variety of choices. The choices made in designing this research will be discussed next.

To begin with, the data were collected on audiotapes. When

presenting the Checklist and its preliminary research, Leary referenced a study suggesting that Checklist ratings are more valid if based on audiotapes rather than typescripts. He added that "sound movies" would be optimal but had not been used to date (Leary, 1957). Twentyfive years later, no change has occurred. Videotapes are much admired as a data source (cf. Mueller and Dilling, 1969), but collecting them presents so many practical difficulties that their use is abjured. In addition, as Hollenbeck (1978) has noted, videotaping is so intrusive that it can be a source of reactive bias.

Using the Checklist as a process measure requires rating

behavior on an interaction-by-interaction basis. The actual unit of analysis can be defined in different ways. In this study, the basic scoring unit was defined as one person's statement(s) from the time





70


the other stopped talking until he or she started again. However, more than one unit was declared if the tone of the response changed to an important degree while the same person was talking. This methodology follows the guidelines outlined by Crowder (1970). Mueller's (1969) experimental procedure in which both an overt and a covert theme were identified proved unreliable. Therefore, the rater was required to choose one reflex to describe each scoring unit.

The Interpersonal Checklist offers researchers a variety of options when it comes to calibrating responses. The circumplex distinguishes 16 "interpersonal reflexes," and four levels of intensity are described for each (Appendix A). Theoretically, one could work with 64 categories. In practice, most researchers have grouped the 16 reflexes either into octants or into quadrants. Moreover, most have made no distinctions by level of intensity.

In psychotherapy research, the usual practice has been to

locate each communication unit at one of the 16 points of the circumplex but then to total scores by quadrant. A person or group of people is described by the percentage of responses located in each quadrant with no consideration given to levels of intensity. This system was used in the four studies which bear most directly on this one, three relating therapist and client behavior to therapy outcome (Crowder, 1970, 1972; Dietzal and Abeles, 1975; Spierling, 1972) and one relating behavior to early termination (Schiller, 1978).

In those four studies, however, one problem emerged consistently. Client and, particularly, therapist behavior was sufficiently stereotyped





71


that little variability was obtained. In all four studies, the majority of client behavior fell in the affiliative-submissive quadrant, while therapists were lodged even more firmly in the affiliativedominant quadrant. This low variability of the behaviors as measured may well be responsible for the fact that Schiller (1978) obtained no differences between early and late terminators on any variable except response complementarity.

These results suggest that the categories should be refined. They could be refined by using octant scoring, but that would not help distinguish routine, "in role" behavior from the more extreme variety. Response intensity differentiation-ignored by the referenced psychotherapy researchers-offers a promising solution. "Appreciative" behavior (low intensity rating) seems healthy and appropriate from a client, and it should be useful to distinguish it from higher intensity behavior ("likes to be taken care of" or "will believe anyone"). Likewise, one expects therapists to be "helpful," but when the intensity of affiliative dominance increases to "overprotective" or "spoils people with kindness," it probably is significantly atypical behavior.

Therefore, in this study the measures of client behavior and

of therapist behavior were constructed as follows. First, each scoring unit was assigned to the category (one of the 16 circumplex points) which best represents it. In addition, it was given an intensity rating of one (low) or two (high).

In order to do this, the behavioral categories developed by Freedman et al. (1951) and elaborated by Crowder (1970) were sorted





72


into intensity levels with the help of Leary's global descriptions. Appendix A presents Leary's circumplex and the adjectives he used, by reflex and intensity level. Appendix B presents the behaviorally oriented circumplex constructed by Crowder plus the verbs used in this research, by reflex and intensity level.

Quadrant scores, plus intensity ratings, were used to test all hypotheses.

For client process scores (Hypotheses 1 and 2), each client's intensity scores within a quadrant were summed, and the four sub-totals which resulted then were converted into percentages of the client's grand total of scores. These percentages are referred to as Weighted Percentages of Interaction (WPIs).

Therapist process scores (Hypotheses 3 and 4) were obtained in a similar fashion from therapist statements. They, too, are expressed as WPIs. Of course, separate sets of WPIs were constructed to demonstrate how a therapist interacted with each client he or she interviewed.

Therapist complementarity of response measures (Hypotheses

5 and 6) were constructed somewhat differently. To begin with, Leary (1957) postulated that his dimensions describe behavior which is not only interpersonally relevant but also self-cuing. Behavior as rated on the affiliation dimension is said to cue like behavior, while behavior rated on the dominance dimension cues its opposite. Complementarity of therapist response, then, means the extent to which a therapist acts on this cuing.




73


Complementarity scores can be constructed from simple quadrant ratings. Dietzal and Abeles did this (1975), and so did Schiller (1978). Neither reported complementarity on the two dimensions separately, but this separation is easy to make and has been made in this study.

Furthermore, provision has been made to include the additional information gained from intensity ratings. Since each statement was

rated either "1" or "2" for its intensity, two statements could differ from each other by as much as four points on each dimension. Therefore, a therapist's statement could be rated from 0 to 4 for its complementarity to the dominance-submissiveness expressed by the client and from 0 to 4 for its complementarity to the affiliation-hostility expressed.

When a client's statement was composed of two or more scorable units, the last unit scored was used to judge complementarity. When a therapist's statement contained two or more scorable units, the first was used. This means that the units compared were those adjacent in the conversation. This procedure duplicates that used and recommended by Mueller (1969). Two complementarity scores were computed for each therapist-client pair: the therapist's average complementarity of response for dominance and average complementarity of response for affiliation.


Tape Rating

Procedures. A 15-minute segment of each tape (minutes 16-30) was rated. This is similar to, but longer than, the segment used by





74


Mueller and Dilling (1968) and identical to Spierling's procedure (Spierling, 1972). A continuous segment was chosen, in preference to several short segments such as Schiller (1978) used, to improve the rater's comprehension of tape content. The particular segment was chosen because it avoids the "social" and/or administrative content often found in the opening and closing minutes of therapy sessions. Therapy sessions at the target facility usually last from 45 to 60 minutes.

Designating minutes 16-30 as the time sample was an approximation. Since client-therapist interaction was being studied, it would have been pointless to rate incomplete interactions. For the purposes of this study, an interaction unit was defined as a client statement and the therapist's response. Therefore, the rater was instructed to start with the first client statement begun after the 15-minute mark and to continue after the 30-minute mark to the end of a therapist response.

All of the tapes were scored by a single rater, the criterion judge. One-third were scored also by the researcher, acting as reliability judge. The reliability tapes were selected randomly but with the limitation that two were drawn from each group of six tapes rated. The criterion judge was not told which tapes would be rerated. Examination of the raters' level of agreement on these tapes provided an ongoing check for rater drift over time. This procedure duplicates that of Crowder (1970) and is in line with Hollenbeck's (1978) observation that continuous monitoring increases rater reliability.





75


Selection and training of rater. A beginning graduate student in counseling psychology, blind to the research hypotheses, was trained and used to rate the tapes. A counseling graduate student was used first because graduate status reduced the likelihood of his knowing any of the (largely ungraduate) subjects and second because his status as a professional-in-training obligates him to conform to American Psychological Association ethical standards regarding confidentiality of client material. As a further safeguard of confidentiality, each tape was identified only by a number assigned sequentially when the researcher obtained the tape from the case therapist.

The rater was taught according to the principles outlined by Mueller (Mueller, 1969; Mueller and Dilling, 1969) and explained more fully in codebook format by Crowder (1970). The rater first was oriented to the circumplex. This included defining a scoring unit and clarifying the 16 points on the circumplex as the products of two bipolar dimensions.

Next, the rater was taught to rate items. He was told to

"empathize with the person exhibiting the behavior, from the position of the target of the behavior" (Crowder, 1970, pp. 43-44, after Freedman et al., 1951). From this point of view, the rater was instructed first to decide whether a behavior was dominant or submissive, next to decide whether it was affiliative or hostile, and finally to locate its specific point on the circumplex. For this investigation a fourth step had to be added: rating the intensity as "1" (low) or "2" (high).

The next part of training involved giving the rater examples

of difficult items and teaching him to score them. A primary principle




76


of Mueller's is to score the "emotional intention" of a statement, rather than its overt content, if the two conflict (Mueller and Dilling, 1969). Crowder added the following clarifications: "Context takes precedence over affect . and interpretation does not go beyond the immediate context" (Crowder, 1970, p. 109).

Finally, the rater rated practice tapes until a preset criterion of reliability was reached. In this study, Crowder's (1970) achieved level of 75 percent was used as the criterion for acceptable average interrater agreement on quadrant assignment. No test (or later, ongoing reliability check) was considered adequate unless quadrant agreement was above 70 percent. In addition, agreement concerning intensity level was checked for all items where reflex assignment matched. Here, the minimum acceptable level of agreement for any test or reliability check was set at 75 percent, and it was expected that average agreement would exceed 80 percent.

Rater training required approximately 30 hours of the experimenter's time and somewhat more time for the trainee. About four hours were required for orientation to the circumplex and scoring aids and for two scoring demonstrations. The rest of the time was spent by the investigator and the trainee in doing practice tape ratings and discussing discrepancies. All training was done using tapes not included in the final study sample. At the completion of training interrater agreement for quadrant placement averaged 73 percent (across six, 5-minute tape segments), and agreement on intensity level averaged 88 percent (across three of the same 5-minute segments). The rater training program is detailed in Appendix C.


I





77


No rater retraining was required once formal rating began

since all five checks produced agreements above the minimum acceptable levels. However, quadrant agreement figures from the first two checks were borderline. Therefore, the investigator examined those tests, located systematic errors being made by the rater, and clarified the relevant guidelines in an attempt to improve the validity, as well as the reliability, of the ratings. Those guidelines, a supplement to Crowder's (1970) codebook, are included in Appendix C.

Rater reliability. The five ongoing reliability checks conducted during the course of tape rating produced average agreements which improved from 71 percent quadrant and 90 percent intensity agreements on the first check to 79 percent quadrant, 88 percent intensity and 74 percent quadrant, 92 percent intensity agreements on the last two checks. Each of the five tests checked reliability on two entire tape segments. Overall average reliability on the ten segments checked was 76.3 percent quadrant agreement and 88.5 percent intensity agreement (see Table 1). Each of these figures exceeds its preset standard for an acceptable level of agreement.

Evaluation of interrater reliability was taken one step

further by computing agreement percentages corrected for chance agreement between raters. Since this was not done in comparable studies, no criteria were available for determining an acceptable level of agreement.

Correcting for chance agreement between raters can be based on one of two assumptions. One may assume that the proportion of items


I




78


Table 1. Percentages of item agreement, for both quadrant and intensity ratings, both unadjusted and after correction for
chance agreement, by test period


Quadrant Agreement Intensity Level Agreement Test Period Unadjusted Corrected Unadjusted Corrected First 70.5 60.7 90.4 80.8 Second 69.9 59.9 83.6 67.2 Third 88.3 84.4 88.5 77.0 Fourth 78.7 71.5 87.7 75.4 Fifth 74.0 65.3 92.1 84.2 Means 76.3 68.4 88.5 76.9





79





a rater assigns to a particular category is a fixed property of that rater or of the coding system. In that case, chance agreement for a category is found by a means such as multiplying the proportion of one rater's codes falling in that category by the proportion of the second rater's codes falling there. This is the approach used by Cohen in creating his Kappa statistic (Cohen, 1960).

On the other hand, one may assume that raters approach each item with no preconceptions, no propensity to rate it in a particular fashion apart from what the item itself suggests. In that case, chance agreement is synonymous with random scoring. The reciprocal of the number of categories among which the raters must choose gives the likelihood of two raters agreeing if one or both codes randomly. Using this approach, it is very simple to adjust an obtained agreement figure. If P0 is the obtained proportion of item agreement and Pr is the proportion of agreement expected from random coding, the corrected agreement figure is (P )/(l r

This formula gives a result identical to that of Cohen's (1960) Kappa when ratings are distributed evenly among coding categories. The two formulas produce quite different results, however, if item distribution is markedly uneven.

In this research, the rating system had two aspects which needed to be checked for reliability: quadrant assignment and designation of an intensity level. In neither case did the criterion judge have a reason to expect any particular proportion of items to appear. In fact, rater instructions to consider the Leary Interpersonal Checklist dimensions separately and then to code a reflex (rather than a quadrant)





80


meant that the quadrants were not even presented conceptually to the criterion judge. Therefore, the formula defining chance agreement as that which would be produced by random coding has been used to correct percentages of item agreement.

The corrected percentages of item agreement average 68.4 percent for quadrant agreement and 76.9 percent for intensity agreement. The range of the first is 61 to 84 percent, and the range of the second is 67 to 84 percent. Table 1 presents both simple percentages of item agreement and the corrected percentages.

In actuality, the adjusted figure for quadrant assignment would have differed little had Cohen's Kappa been used. Quadrant codes were not distributed completely evenly, but the distribution was even enough that Kappa would have subtracted only 2 to 4 more percentage points from most of the interrater agreement scores for quadrants.

On the other hand, intensity codes tended to be low. "One" was coded more frequently than "2" on all protocols, and on the reliability protocols as a group "1" was coded six times as often as was "2." For this reason, applying Kappa to intensity coding would have subtracted from 11 to 37 additional percentage points from the corrected interrater agreement figures. The uneven distribution of intensity codes raises questions about the use of this coding system, and these questions will be examined in Chapter Five. However, at this point there seems to be no compelling reason to lower interrater reliability estimates because of this uneven distribution.





81


Data Analysis

Hypothesis testing. Hypotheses 1 through 6 were tested using two-tailed t-tests for the significance of differences between means.

Investigating potential intervening variables. The literature review in Chapter Two discussed a variety of client and therapist factors which have been investigated for their relationship to premature termination. These will be considered here in light of their potential for acting as "nuisance" variables in this study.

The two (nonindependent) client factors cited most frequently as influencing rates of premature termination are socioeconomic status and educational level. In this study, it was expected that those factors would be controlled for to a large extent by using a sample composed entirely of university students. In a similar fashion, the variable of race should be largely eliminated. Due to differential utilization rates of the counseling facilities, plus the presence of a predominantly Caucasion student body at the target university, it was expected that the study sample would be self-limited almost entirely to Caucasians. Those assumptions were checked by gathering socioeconomic, educational, and racial data and performing chi-square analyses for these factors by termination status.

One remaining client factor is an indentified, potentially significant intervening variable: sex of client. Results have not been consistent, but some studies have found males less inclined than females to terminate prematurely. As it seemed impractical to balance the study for sex of client, a chi-square analysis was performed to





82


see whether or not women would be found disproportiately among the premature terminators in this study.

The one therapist factor which has been found fairly consistently predictive of premature termination rates is amount of experience. Frequently, experience has been confounded with amount of training, so that needs to be considered a possible factor as well.

In this study, however, there is no real justification for treating either experience or training as an intervening variable. Generally speaking, therapists at the target facility announce neither their degrees nor their years of experience to potential clients. Some holding doctorates include this information on their door plaques, but not all do. The absence of a doctorate is never indicated. Moreover, the most obvious clue to experience, age, varies little within the group of participating therapists.

It was anticipated that both the amount of experience and the amount of training enjoyed by therapists would affect the likelihood of their clients terminating prematurely. Primarily, though, this influence should have operated not as an intervening variable but as a source of variability in client-therapist process. Therefore, controlling for experience/training might have artificially eliminated much variability of the independent variables utilized in Hypothesis 3 through Hypothesis 6. Instead, the expected association between therapist experience and termination status was sought by chi-square analysis. Had it been found, its roots would have been explored via two-way analyses of variance comparing mean levels of the independent





83


variables by experience level of the therapist as well as by termination status.

One other therapist factor needs to be mentioned here: sex

of therapist. Some weak support exists for the hypothesis that female therapists have lower rates of premature termination than do male therapists. No explanation has been offered, so one can only speculate as to whether this effect operates through client-therapist process or through a client's internal reactions to the idea of his or her therapist's gender.

Due to the size of this study and an imbalance between male and female therapists at the facility studied, it would have been difficult to balance the sample for sex of therapist. Therefore, its relationship to termination status was examined in a post hoc chisquare analysis similar to that performed for sex of client by termination status.

Other analyses. Exploratory work was done utilizing the concept of response intensity. Two-tailed t-tests were performed comparing the mean intensity of responses in cases terminated prematurely with those of cases where clients remained in treatment. For both clients and therapists, separate analyses were performed by quadrant. This totaled eight analyses concerning response intensity.

Significance levels. The .05 level of significance was used in all tests of association as the criterion for rejection of the null hypothesis. However, in the interest of thoroughness, significance levels as low as p < .25 were reported. Below p = .25, results were described simply as nonsignificant.















CHAPTER FOUR
RESULTS


Introduction

Fifty-nine clients and 24 therapists participated in the datagathering phase of this study. This phase continued until there were 15 cases with complete data which clearly could be classified as premature terminations and more than 15 which could be classified as remaining in treatment. Table 2 breaks down the case classification by termination status and by tape usability.

The final study sample included all 15 premature terminations in which taping had been successful and 15 cases randomly selected from among those remaining in treatment where taping was adequate. Sixteen therapists had cases in this final sample.

Six hypotheses were tested. In each case, termination status was the dependent variable. The independent variables included two measures of client behavior, two measures of therapist behavior, and two measures of the complementarity of therapist behavior to the preceding client behavior.

Other analyses were performed to test the predictive strength of client and therapist characteristics which previous research has associated with case termination status. The characteristics tested are sex of client, sex of therapist, experience level of therapist, two elements of client socioeconomic status (years of education


84





85


Table 2. Number of cases, by termination status and tape usability Cases with
Cases with Tapes Lost Termination Status Usable Tapes or Inaudible Total Premature termination 15 2 17 Remaining in treatment 25 3 29 Unclassifiable 13 1 13




Table 3. Mean client WPIs, by quadrant and by case termination
status


Premature
Quadrant Terminators Remainers df t-test Significance Affiliativedominant 8.800 10.867 Hostiledominant 14.867 14.000 Affiliativesubmissive 55.533 54.400 28 -.053 n.s. Hostilesubmissive 20.867 20.800 28 .008 n.s.





86


completed and parental occupation), and client race. Data were gathered on therapist race as well.

Finally, exploratory analysis was done in the area of response intensity. Eight analyses were performed, four using measures of client intensity and four using measures of therapist intensity.


Hypothesis Testing

Client Behavior and Premature
Termination: Hypotheses 1 and 2

Table 3 presents the mean client Weighted Percentages of Interaction (WPIs) in all four Leary Interpersonal Checklist quadrants for premature terminators and for remainers. In addition, t-statistics and significance test results are reported for the two quadrants (affiliative-submissive and hostile-submissive) where significant differences were hypothesized. In both quadrants, mean differences were slight and did not approach significance; hypotheses 1 and 2 were not supported.


Therapist Behavior and Premature
Termination: Hypotheses 3 and 4

Table 4 presents the mean therapist Weighted Percentages of

Interaction for the therapists of premature terminators and for those interacting with remainers. Again, t-statistics and significance tests results are reported for mean comparisons where significant differences were hypothesized. In both cases, sizable mean differences were found, but variances were large. Moreover, in the hostiledominant quadrant the obtained difference is in the opposite direction of that hypothesized. Hypothesis 3 was supported only at the .25 level of confidence, and hypothesis 4 was not supported at all.





87


Table 4. Mean therapist WPIs, by quadrant
status


and by case termination


Premature
Quadrant Terminators Remainers df t-test Significance Affiliativedominant 82.467 91.933 Hostiledominant 13.333 4.867 28 -1.472 <.15 Affiliativesubmissive 1.467 1.933 Hostilesubmissive 2.733 1.133 28 1.178 <.25




Table 5. Mean complementarity of therapists' responses, by dimension
and by case termination status


Premature
Dimension Terminators Remainers df t-test Significance Dominance 3.430 3.261 28 -1.998 <.06 Affiliation 3.229 3.156 28 -.899 n.s.




Table 6. Chi-square analysis of case termination status by sex of
client


Premature 2 Sex of Client Terminators Remainers df X Significance Male 5 4 Female 10 11 3 .159 n.s.





88


Therapist-Client Complementarity and Premature Termination: Hypotheses 5 and 6

Table 5 presents mean complementarity scores for therapists'

responses to clients. Complementarity means are figured separately on the two Leary Interpersonal Checklist dimensions, affiliation (affiliation-hostility) and dominance (dominance-submission). Neither hypothesis 5 nor hypothesis 6 was supported. One two-tailed t-test found the mean difference in affiliation complementarity to be nonsignificant. The second test found the mean difference in dominance significant at the .06 level of confidence, but that difference reversed the direction hypothesized.


Analysis of Selected Client
and Therapist Factors

Client Factors

Sex of client was considered a possible intervening variable in this study. Table 6 presents a chi-square analysis of termination status by sex of client. It can be seem that, for the sample utilized in this study, there is no significant relationship between the two.

Tables 7, 8, 9, and 10 present descriptive statistics and chisquare analyses for other client characteristics, all related to socioeconomic status, which have been found predictive of premature termination. It was expected that the use of a university counseling center clientele would eliminate most of the variability here. Specifically, it was anticipated that the sample would be largely Caucasian, educated beyond high school level, and middle to upper class in background as measured by parental occupation. As it turned out, the first two assumptions were accurate, but the third was not.





89


Table 7. Chi-square analysis of case termination
client


status by race of


Premature
Racea Terminators Remainers df X2 Significance Black 2 0 Caucasian 13 15 3 2.33 n.s. aSelf-identified



Table 8. Chi-square analysis of client education level by case termination status


Premature 2 Years of Education Terminators Remainers df x Significance At least 12 but
less than 16 13 12

16 or more 2 3 3 .240 n.s.




90


Table 9. Parental occupation by case termination status


Premature
Occupation Terminators Remainers Total

Farm 0 0 0 Semi-skilled
non-farm 2 0 2 Skilled
non-farm 5 2 7 Clerical and
sales 1 2 3 Business and
professional 4 8 12 Data missing or
not codableb 3 3 6


aFather's or stepfather's occupation was used, if known; otherwise, mother's or stepmother's occupation was used. bIncludes four instances of missing data and two descriptions of parental occupation as "retired."



Table 10. Chi-square analysis of parental occupation by case termination status


Premature 2 Occupation Terminators Remaining df X Significance

Semi-skilled,
skilled, clerical
or sales 8 4 Business or
professional 4 8 3 2.67 n.s.




Full Text

PAGE 1

COUNSELING PROCESS AND PREMATURE TERMINATION By SUZANNE L. MOLNAR A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1982

PAGE 2

To Terry, to Julia, and to my parents

PAGE 3

ACKNOWLEDGMENTS I count myself wery fortunate to have received the help and friendship of many people throughout the course of my doctoral work. I cannot mention them all, but special thanks are due to some. Most affectionate appreciation is extended to Dr. Harry Grater. Both as chairman of my dissertation committee and as clinical supervisor during my internship, he has been an exceptionally supportive mentor. He has given freely of his time, his expertise, and his sense of proportion. My graduate years would have been poorer without him. The other members of my dissertation committee deserve warm thanks as well. Dr. Franz Epting has been a friend and a source of support since my first year at the University of Florida, and he provided invaluable assistance with many aspects of this dissertation. Dr. James Archer, Jr., has been a helpful committee member and an inspiring role model for whom to work. Dr. William Froming was especially helpful with design and statistical problems. Dr. Carolyn Tucker has been generous both with her long-term friendship and with her last-minute assistance. I also want to express gratitude to Ernest Downs for the time and care he took with the scoring of my research tapes. He worked hard at a most difficult point in the research— and he even managed to help me enjoy the process. m

PAGE 4

Finally, I want to express my deepest appreciation to my husband, Terry Molnar. Without his faith in me, his unfailing support when I needed it most (and his endless hours of babysitting), this dissertation might not have been completed. IV

PAGE 5

TABLE OF CONTENTS Page ACKNOWLEDGMENTS iii ABSTRACT viii CHAPTER ONE INTRODUCTION 1 The Problem 1 Need for This Study 1 Overview of the Design 4 Dependent Variable 4 Independent Variables 5 Hypotheses 7 Organization of the Study 8 TWO REVIEW OF THE LITERATURE 11 Premature Termination 11 The Definitional Problem 11 Premature Termination and Psychotherapy Outcome 18 Predicting/Explaining Premature Termination .... 21 General Factors 21 Client characteristics 21 Therapist characteristics 25 Interaction of client and therapist characteristics 26 Client-Therapist Dynamics 32 Client reaction to therapist 33 Therapist reaction to client 35 Therapy process factors 37 Studying Therapy Process 38 General Problems 38 Process as a Predictor of Outcome 39 Process variables derived from a theory of psychotherapy 39 A-theoretical variables 43 Hypotheses 51

PAGE 6

CHAPTER Page THREE METHODOLOGY 54 Subjects 54 Data Collection 55 Instrument: The Leary Interpersonal Checklist ... 56 Theoretical Basis 56 Construction 58 Validating Research 61 Category integrity and arrangement 62 Dimensional integrity and orthogonality . 62 Structural similarity among the levels of personality 65 Concurrent validity 67 Reliability 67 Use in This Study 69 Tape Rating 73 Procedures 73 Selection and Training of Rater 75 Rater Reliability 77 Data Analysis 81 Hypothesis Testing 81 Investigating Potential Intervening Variables 81 Other Analyses 83 Significance Levels 83 FOUR RESULTS 84 Introduction 84 Hypothesis Testing 84 Client Behavior and Premature Termination: Hypotheses 1 and 2 86 Therapist Behavior and Premature Termination: Hypotheses 3 and 4 86 Therapist-Client Complementarity and Premature Termination: Hypotheses 5 and 6 . 88 Analysis of Selected Client and Therapist Factors 88 Client Factors 88 Therapist Factors 91 Response Intensity 93 Other Reanalyses 97 FIVE DISCUSSION 99 Counseling Process and Premature Termination .... 99 Client Process and Premature Termination .... 99 Therapist Process and Premature Termination . 101 Therapist-Client Complementarity and Premature Termination 103 VI

PAGE 7

CHAPTER Page Nonprocess Factors and Premature Termination .... 105 Client Factors 105 Therapist Factors 106 Limitations of This Study and Suggestions for Future Research 107 General Conclusions 110 APPENDIX A THE LEARY INTERPERSONAL CHECKLIST 113 B INTERPERSONAL PROCESS DIAGNOSIS 117 C RATER TRAINING PROGRAM AND SUPPLEMENTARY CODING AID . 120 D DATA ON CASES COMPRISING THE STUDY SAMPLE 124 REFERENCES 127 BIOGRAPHICAL SKETCH 135 vn

PAGE 8

Abstract of Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy COUNSELING PROCESS AND PREMATURE TERMINATION By Suzanne L. Molnar August 1982 Chairman: Harry A. Grater, Jr. Major Department: Psychology Premature termination from psychotherapy is a problem for many mental health professionals, but little is known about how it occurs. This research focused on client-therapist process as a potential predictor of premature termination. Thirty university counseling center client-therapist pairs, half premature terminators and half remainers, were selected as subjects, All clients had been referred by intakers for ongoing psychotherapy. Number of sessions attended and therapist judgment were used to decide termination status. Process was measured using Leary's interpersonal circumplex to rate behavior on two bipolar, orthogonal dimensions (dominancesubmission and affiliation-hostility). Fifteen-minute audiotaped segments from first postintake therapy sessions were scored. Each statement was located in one quadrant of the circumplex and then assigned an intensity rating of "1" (low) or "2" (high). viii

PAGE 9

On the basis of previous research, it was hypothesized that premature terminators' tapes would differ from remainers' tapes by containing: 1. Relatively high percentages of hostile-submissive client statements 2. Relatively low percentages of affiliative-submissive client statements 3. Relatively high percentages of hostile-submissive therapist statements 4. Relatively low percentages of hostile-dominant therapist statements 5. Relatively low levels of therapist-client complementarity, both for dominance 6. And for affiliation Hypotheses were tested using twotailed t^tests for the significance of differences between means. No relationship was found between client process and premature termination. Therapist process differentiated the two outcome groups, but one difference reversed that predicted: both hostile-submissive and hostile-dominant behavior were more common among the therapists of premature terminators. The obtained complementarity of therapists' responses to their clients also reversed the direction hypothesized. Remainers' therapists behaved in the less complementary fashion, and this difference was significant at the .06 level for dominance complementarity. Additional analyses found no client or therapist characteristic predictive of termination status at the .05 significance level. ix

PAGE 10

However, n^'s were very small. Numerical differences favored clients from business-professional backgrounds and favored female therapists. It was concluded that client-therapist process probably is one component of the multiply determined process of premature termination.

PAGE 11

CHAPTER ONE INTRODUCTION The Problem Premature termination from psychotherapy is a problem of concern to many. Mental health centers, student counseling centers, and private therapists all are troubled by the frequency with which clients make contact, begin therapy, then terminate before they have received what professionals consider to be adequate treatment. Practit'loners ask why this happens and what, if anything, they can do about it. Some researchers have attempted to answer these questions, but much remains to be done. Need for This Study To begin with, most research attempting to "explain" premature termination actually just explains it in the statistical sense by looking for correlates. Client characteristics and therapist characteristics are the usual independent variables. This line of work has been fairly successful. A variety of client characteristics, plus therapist experience, has been established as reliable correlates of premature termination (Baekeland and Lundwell, 1975). However, this type of information is of limited utility. Knowing that a poor, black woman of limited intelligence is likely to leave therapy before it has had much impact helps little with case management. 1

PAGE 12

Knowing that inexperienced therapists lose many clients prematurely gives the profession no assistance since the cure it suggests is inseparable from routine provision of service. A group of studies with greater potential usefulness focuses on the interaction of client characteristics with therapist characteristics. Demographic characteristics have not proven very valuable in this research. On the other hand, various similarities in the personalities of therapist and client have been found to increase client retention rates (Mendelsohn, 1966; Landfield, 1971). Also, clienttherapist similarity with regard to therapy goals, the priority of those goals, and perception of the client's problem all have been found to make premature termination less likely (Butler, 1977; Raschella, 1975). All of these results suggest ways in which service facilities and/or training facilities might be able to improve their programs to increase client retention rates. In a sense, though, doing that type of study at this point is like furnishing a house before the floor is put in. A more logical approach is to first tackle the question, How does premature termination occur?" This question can be rephrased as, "How does therapy process leading to premature termination differ from process leading to ongoing therapy?" Answers can focus on the clients, on the therapists, or on the interaction between them. Verbal behavior, non-verbal behavior, or both may be targeted. Behavior may receive much or little interpretation. In any case, process research results are needed as a base for manipulative research efforts.

PAGE 13

In their 1970 review, Meltzoff and Kornreich declared inattention to process a fundamental weakness in research on premature termination. In spite of their urgings, however, only a little effort has gone into this area. A variety of studies approach therapy process indirectly by asking clients and/or therapists to record their reactions to, and interpretations of, what took place in a session. True process studies, however, are almost nonexistent. Only four were located in which process was measured directly and was compared with termination status. Winder et al (1962) pioneered the terrain with their study of therapist response to dependency content in clients' verbalizations. A later researcher tested the hypothesis that interpreting transferencebased behavior would make early termination less likely (Goldstein, 1975). Another looked at self-disclosure as a predictor of early termination (Heilbrun, 1973). Still more recently, Schiller used the Leary Interpersonal Checklist to measure therapist-client process complementarity and used complementarity as a predictor variable in his study of early termination (Schiller, 1978), All four are interesting studies. Winder et al. (1962) and Goldstein (1975) emphasized therapist behavior— a focus with real potential usefulness to clinicians. Heilbrun looked at (female) client behavior and found intriguing interactions between self-disclosure and sex of therapist when they were used as independent variables to predict early termination versus remaining in therapy, Schiller (1978) examined therapist behavior, client behavior and the complementarity of therapist behavior to client behavior— an approach commendable for its thoroughness.

PAGE 14

It is important to note, however, that only Winder et al (1962) actually studied premature termination. The others defined termination status entirely by the number of sessions attended; they are studies of early— not premature— termination. This distinction is discussed at length in the first section of Chapter Two. At this point, then, the field is wide open. Connecting therapy process with premature termination has been declared a promising and potentially valuable endeavor, but almost no work has been done on it. This research was designed to begin filling that void. Overview of the Design The intent of this study was to learn more about how premature termination occurs. Therapeutic process was audiotaped, and comparisons were drawn between two groups of subjects: clienttherapist dyads whose clients terminated prematurely and dyads whose clients remained in treatment. Subjects were clients and counselors at the University of Florida Psychological and Vocational Counseling Center. Taping was done as clients began ongoing, individual psychotherapy. Assignment to termination category was made after completion of counseling. Dependent Variable One major difficulty with this field of research appeared iimiediately: Premature termination enjoys no standard definition. Even among researchers of similar theoretical persuasions, the concept is operationalized in a wide variety of ways.

PAGE 15

At one extreme are the many researchers who operationally equate early termination with premature termination. This gives them simple, objective definitions, usually based on the number of sessions a client attended. However, this simplicity is simplistic to a selfdefeating extent if one's aim is to uncover significant differences in therapy process. At another extreme are researchers who leave judgment of termination status entirely to the clinician. Without some strict controls, this can lead to chaos. Psychoanalytic clients with several hundred hours of therapy could be classified as premature terminators while clients from a crisis-oriented clinic were discharged as successes after two sessions. Given this lack of consensus about a central concept, it seemed important to review others' efforts with care before choosing or constructing a definition of one's own. This review will be presented in the first part of Chapter Two. Based on its conclusions, a definition was constructed for this study which combines objective and subjective factors and which discards questionable cases rather than forcing them into a Procrustean bed. Independent Variables Since this field of research is largely unexplored, the literature off ered 1 i ttl e help in choosing independent variables. Fortunately, some assistance was available from the field of social psychology. Reviewers covering a variety of factor analytic studies have concluded that dominance-submission and affiliation-hostility emerge as the major

PAGE 16

factors in dyadic interaction. The descriptive terms differ slightly from one study to the next, but the dimensions remain intact and appear to be orthogonal (Foa, 1961). Choosing variables based on their prominance in general interaction, rather than in the therapy setting, eliminates biases which could originate from practice of a particular type of therapy. It also eliminates therapist and client roles; behaviors were rated similarly for both. However, the relevance of the dimensions to therapeutic interaction needs to be demonstrated. That need was met in this study by the way in which dominancesubmission and affiliation-hostility were operational i zed. The Leary Interpersonal Checklist was used to rate taped therapeutic process on an interaction-by-interaction basis. The instrument was designed for this purpose (Leary, 1957), and its validity and usefulness are well established (cf. Crowder, 1970; Mueller, 1969; Mueller and Dilling, 1969; Schiller, 1978; Spierling, 1972). In fact, the Leary Interpersonal Checklist has a combination of general strengths and specific attributes which made it an ideal instrument for this research. The general strengths, which will be discussed at length in Chapter Three, include a solid theoretical grounding, competently done validation studies, and tested relevance for psychotherapy research. Specific assets— and the use made of them in the design of this research— will be discussed next. One attribute of particular importance in this research is Leary's provision for rating therapist and client behavior both in

PAGE 17

isolation and in interaction with each other. On the basis of literature reviewed, this research hypothesized that therapist behavior, client behavior and therapist responses to specific client behaviors all would differ significantly when terminator dyads were compared with remainer dyads. Furthermore, Leary interaction scores can be interpreted in terms of complementarity— an important concept in process research (cf. Carson, 1969). Again on the basis of the literature review, it was hypothesized that differences in therapist-client interaction would include differences in the complementarity of therapist response to client behavior between the two groups of therapists. Finally, process ratings using the Leary can be "fine tuned." The researcher decides how broad or narrow categories are to be both in terms of dimensional purity and in terms of intensity shown. Previous research suggested the use of broad scoring for dimensional location plus differentiation of two or three levels of response intensity. In this study, broad (quadrant) dimensional scoring was done, and two levels of intensity were distinguished. Hypotheses In general, it was hypothesized as follows: 1. Therapist dominance-submission and hostilityaffiliation will differ significantly between initial interviews with clients who prematurely terminate and interviews with clients who remain in counseling.

PAGE 18

2. Client dominance-submission and hostilityaffiliation in the same interviews will differ significantly when premature terminators are compared with remainers. 3. The complementarity of therapist behavior to client behavior along the dimensions of dominance-submission and hostility-affiliation will differ when the therapists of premature terminators are compared with therapists interacting with remainers. Specific expectations about the nature of these differences are based in the related literature. Operational hypotheses will be formulated at the conclusion of the literature review. Organization of the Study Chapter Two consists primarily of a review of literature serving as background material for the study. This review begins by considering the problem of defining premature termination. It then moves into a review of research attempting to predict/explain premature termination. Nonprocess research is covered, first as general background and second to support later decisions concerning research design. The lack of substantial process research is documented. Next, there is a consideration of the general problems of doing therapy process research and a brief review of research using process to predict therapy outcome in general. The chapter ends with formulation of the research hypotheses.

PAGE 19

Chapter Three covers the design and methodology of the research. Subject selection is discussed, and data collection procedures are detailed. Next, the Leary Interpersonal Checklist is examined for its validity and reliability. Its construction and its theoretical basis are covered, and the specific use made of it in this study is explained. Next, tape rating procedures and procedures for selecting and training the tape rater are presented. Finally, data analysis is discussed. Specific analyses are planned for hypothesis testing, and others are presented for use in answering other questions. Chapter Four presents the data summaries and analyses. This begins with an overview of the data obtained in the project as a whole and that used in the final study sample. Next, the data on which hypothesis tests were based and the results of those tests are given. Other planned analyses are reported. Finally, several post-hoc analyses are reported, and the reasons for their inclusion are mentioned. Chapter Five interprets the data, discusses their implications in terms of past research, and delineates limitations of this study plus recommendations for future research. The first section of this chapter is devoted to the relationship between process variables and termination status. This includes the testing of research hypotheses and related analyses. The second section examines the relationship of nonprocess client and therapist characteristics to termination status. Agreements and disagreements between this study and others in the area are noted. The third section discusses general problems, most of which this study shares with others in the area. In some

PAGE 20

10 cases, further research is suggested which could ameliorate the problems. Finally, general conclusions are drawn about the role of process in predicting therapy termination.

PAGE 21

CHAPTER TWO REVIEW OF THE LITERATURE Premature Termination The Definitional Problem Premature termination appears to be a meaningful concept. It is used frequently by professional psychologists, and most seem to find its definition as intuitively clear as that of any other concept. If asked whether or not a client terminated prematurely, the typical practitioner answers, "Yes," "No," or "Maybe" and can provide reasons supporting that judgment. However, intuitive meaningfulness does not equal researchability. Ideally, a research concept first has consensual validity and then is operational i zed in a way which does not compromise that validity. In the case of premature termination, definitional problems abound. Clinicians disagree with each other about the standards for judging a termination premature. Most researchers want objective criteria, while clinicians prefer to make subjective judgments. Some reviewers want the client's judgment to be considered in addition to those of the clinician and the researcher. This lack of consensus is usual for concepts originating with therapists rather than theorists. However, it does dictate a reversal of the ideal procedure. In the absence of consensus, a researcher must choose from among the offered definitions, operational ize the chosen n

PAGE 22

12 definition conscientiously, and then record everything as clearly as possible. Only if others agree with those procedures will the work be generally meaningful. In examining the literature for a definition of premature termination, one finds a range of usage. In some articles it refers to termination occurring for any reason before a given number of sessions are completed. One recent example is the work of Stern et al. (1975) in which premature termination was defined simply as failure to attend six consecutive sessions. Similar but more stringent tests were applied by Heilbrun (1974) and by Fi ester (1977). They defined premature termination as failing to return for the second interview and the third interview, respectively. At the other extreme, premature termination may be defined entirely subjectively, based on the case therapist's judgment or on a combination of therapist and client judgments. An early example using therapists' judgments is Landis's study (1937) of psychoanalytic psychotherapy. More recently, Raschella (1975) published a study of "premature client dropout from therapy" in which prematurity was strictly a matter of therapists' judgments. Similarly, McGrogan (1976) used termination without therapist approval as his sole criterion of prematurity. Landfield (1971) devised a set of criteria which are far more complex than those discussed previously but are equally subjective. According to Landfield, premature termination has occurred when (1) both therapist and client agree that therapy goals have not been

PAGE 23

13 met, and (2) therapist and/or client feel(s) that therapy could be beneficial, but (3) one or the other is unwilling to continue treatment. The first, "objective," type of definition implies that premature termination is equivalent to early termination. Probably few clinicians would endorse such an equivalency. Nevertheless, it is important to consider whether— and how— the two can be differentiated from each other. "Premature," as opposed to "early," termination implies a judgment about the timing involved. Some researchers sidestep this issue by avoiding the term "premature terminator." The most common substitute is "dropout." Recent examples include the work of Saltzman et al. (1976), of Fiester and Rudestam (1975), of Caligor (1975), and of Vail (1974). As the discussion section of Baekeland and Lundwell's (1975) review article indicates, however, this semantic sidestepping solves nothing. A proper definition of "dropping out" should specify how termination occurred, not just length of treatment. Since length of treatment alone cannot define premature termination satisfactorily, a question naturally arises: What about leaving the decision to the case therapist? The answer is much debated. One objection stems from research such as Rosen's (1978) in which counselor perception of termination was found to be unrelated to client perceptions of either therapy outcome or termination. Countering this is the fact that therapists coined the term in the first place, labeling a phenomenon they identified and considered conceptually important to them. It can be argued that premature termination is not a meaningful concept to most clients.

PAGE 24

14 However, there is a potential problem for researchers with allowing either therapists or clients to judge termination status: noncomparability of classification criteria. The general solution seems obvious. Termination status should be decided on a case-by-case basis, but objective guidelines are needed if cases are to be grouped for the purposes of research. A number of researchers have put some effort into constructing such multiple criteria. One approach is to have therapists contract with clients for a minimum number of sessions and then compare clients completing the contract with those who fail to complete it. Saltzman et al. (1976) did this using a fixed number of sessions, while Norkus (1976) allowed therapists and clients to write their own contracts. Each variant has its advantages and its disadvantages. Saltzman 's approach differs little from the simple session-counting type of criterion; the only difference is that since the standard for prematurity is shared knowledge among clients, therapists and researchers, a client who terminates prematurely has decided to accept less than the recommended course of treatment. Norkus, on the other hand, allowed for individualization on a case-by-case basis but failed to insure any uniformity in contracting practices among her participating therapists. Another general approach uses session-counting but increases its sophistication by adding other criteria. For example, Sasseen (1976) defined premature termination as attending fewer than five sessions and terminating without contacting one's therapist. Sandler (1975) used the number of sessions attended plus therapist's judgment about

PAGE 25

15 termination status. Winder et al. (1962) did essentially the same thing. These two procedures have strengths and weaknesses analogous to those in the preceding paragraph. That is, Sasseen's (1976) work is a definite improvement over the use of a fixed session criterion alone, but it remains somewhat arbitrary. For example, a client who quit in a fit of temper would not be defined as a premature terminator since the therapist was contacted. On the other hand, Sandler's (1975) method provides the flexibility necessary to correctly classify the unusual case but fails to assure uniformity beyond the general control provided by session counting. However, both procedures are much better than the methodologies discussed previously. They narrow substantially the range of possible error without introducing any impracticable requirements. More complicated methodologies can refine the operational definition of premature termination still further, but some seem to suffer from a diminishing return of benefits in relation to effort required. An example is Landfield's (1971) complex definition, previously detailed. It requires extensive querying of both therapists and clients for their views on the appropriateness of therapy in general, its completeness at termination, and their desires regarding termination. Even assuming honest and insightful answers to questions on such an emotionally charged subject, the results do not seem worth the effort. For one thing, the definition is entirely subjective. For another thing, it is overly inclusive. It would classify any therapy

PAGE 26

16 failure as premature termination unless both therapist and client recognized the failure for what it was. Moreover, it would do the same for some cases in which the therapist (or the client) no longer considered therapy appropriate. Jachim (1974) utilized a combination of criteria which avoids most of the pitfalls discussed so far. He defined premature termination both by number of sessions and by a subjective criterion, lack of therapist agreement. In addition, Jachim telephoned terminators to be sure that they had no extrinsic reason for dropping out. Best of all, Jachim avoided the usual practice of maximizing n_'s by classifying e\jery case, whether or not the classification is a meaningful one. His "terminators" left within four sessions, while his "remainers" stayed for at least eleven. Those cutoff points were arbitrary, chosen by discarding the middle third of all the cases sampled. Although the cutoff points are debatable, this is a laudable attempt to eliminate "noise" from the comparison of terminators and remainers. In this study, premature termination will be defined in a way similar to Jachim's in that both objective and subjective criteria will be used and intermediate cases will be discarded. However, the cutoff points that serve as objective criteria will be chosen based on theoretical and empirical considerations. A starting point is Landfield's (1971) observation that "Premature termination indicates that members of a therapy dyad are unable to maintain interpersonal communication" (p. 66). One way of insuring that this condition is met is to limit the designation "premature" to

PAGE 27

17 cases which terminated so quickly that establishment of an interpersonal relationship of therapeutic quality is improbable. Other researchers using this approach have chosen anywhere from one (Heilbrun, 1974) to nine (Saltzman et al., 1976) or ten (Winder et al 1962) as the maximum number of sessions a person could attend and still be considered a premature terminator. Neither extreme seems desirable. Saltzman et al observed themselves that most clients who terminated unilaterally did so before the sixth session. On the other hand, using one session as the upper limit for premature terminators is questionable, as there is some evidence that failure to return after the first session is a rather different phenomenon from early termination in general (Schiller, 1978). The most frequently used cutoff point in the literature reviewed was failure to return for the fifth session (cf. Jachim, 1974; Sandler, 1975; Sasseen, 1976). The sixth session was used similarly by two others researchers (Schiller, 1978; Stern et al., 1975). However, with the exception of Jachim all of these researchers set up criteria which classified every case as either a terminator or a remainer. Had they allowed themselves the luxury of discarding intermediate cases, the cutoff points probably would have been lower. In this study, premature termination was defined in part as failure to return for the fourth (or earlier) post-intake session. Following the conclusions of Saltzman et al (1976), remainers were required to attend six or more post-intake sessions. Intermediate cases were discarded from the final sample.

PAGE 28

18 In addition to these objective guidelines, subjective factors were considered as well. First, only clients referred by an intaker as appropriate for ongoing psychotherapy were included in the sample. Second, no client was considered a remainer if he or she canceled or skipped one of the first six sessions and returned only after a counselor initiated contact. Finally, counselors were querried for their opinion of each client's termination status. Clients were considered premature terminators only if they terminated against the counselor's judgment or without consulting the counselor. They were considered remainers only if the therapist concurred with the termination. Again, indeterminate cases were dropped from the final sample. The procedures were unusually rigorous. It was hoped that they would insure a high level of validity in the final assignment of clients to categories of the dependent variable. Premature Termination and Psychotherapy Outcome Premature termination has an uncertain place in psychotherapy research for reasons other than controversies over its definition. A major problem is deciding how its study relates to classical psychotherapy outcome research. The earliest major study of psychotherapy outcomes recognized premature termination, though not by name. In the 1930 report summarizing ten years of work at the Berlin Psychoanalytic Institute, premature terminators were described as having "left treatment," and this was recognized as a distinct category of therapy outcome (Landis, 1937).

PAGE 29

19 The concept of premature termination has not always been accorded this respect in more recent work. A prominent example, Eysenck's (1952) famous and damning summary, "The Effects of Psychotherapy: An Evaluation," consigned all early terminations to the category of therapy failures. Indeed, premature termination from therapy continues to be an embarrassment to researchers who want to locate all cases along a success-to-failure continuum. Some, following Eysenck, have deemed it a type of failure. Many simply have eliminated most cases by such means as dropping from their studies all clients who attend less than a given number of sessions. There still is no consensus as to the place of premature termination in outcome research. A plausible argument can be made for treating voluntary premature termination as failure. Logically, one can argue, the therapist has failed to make the desired impact on the client if unilateral termination occurs. This logic is bolstered by the empirical observation that most factors which predict therapy failure (cf. Luborsky et al,, 1971) also predict a high rate of premature termination (cf. Baekeland and Lundwell, 1975). These factors will be discussed individually in a later section of this paper. In addition to general correlations, a coincidence of factors connected first with client dissatisfaction and second with premature termination occasionally has been established within the same subject sample. An example is Sasseen's (1976) conclusion that disconfirmation of clients' expectations and preferences predicts both premature termination and client dissatisfaction with the interview.

PAGE 30

20 In spite of these logical and empirical arguments relating voluntary premature termination to therapy failure, the most crucial type of data remains sparse. Very few efforts have been made to evaluate directly the post-treatment status of premature terminators. The majority of such follow-ups conclude that most premature terminators are therapy failures (cf. Jachim, 1974). However, one elaborate investigation of dependent female clients concluded that many of them terminated not from frustration but due to symptomatic relief of anxiety following a cathartic therapy session (Heilbrun, 1974). A later investigator also found cathartic relief to be a factor in premature termination (Fi ester, 1977). At this point it seems that premature termination occurs for a wide variety of reasons. Even when voluntary and unilateral, it does not always imply failure. Particularly with low-readiness clients, it may mean that the client received the symptomatic relief he or she sought. Such a client might even return at a later time for more extensive treatment, but no research has been done on this to date. This conclusion is worth serious attention because most theory and research assumes that premature termination is an expression ..of client discontent and/or anxiety. Theoretically, this complexity means that it is pointless to attempt to assign premature termination a fixed spot on any success-to-failure continuum. Empirically, it means that conclusions valid for the majority (presumably the discontent/ anxious clients) are being weakened statistically by the heterogeneity of most "terminator" samples.

PAGE 31

21 On the other hand, there is no support at all for the not uncommon practice of dropping premature terminators from outcome research. However inconveniently heterogeneous they appear, premature terminators definitely do not form a representative cross sample of therapy cases. Unless investigated in more detail, their termination itself is one type of therapy outcome. Predicting/Explaining Premature Termination Logic dictates treating premature termination as a type of therapy outcome, and this logic has not been lost on researchers. Therapy researchers acknowledge the connection in a variety of ways. One of the more subtle is the overall similarity of premature termination research designs to those of general outcome research. Historically, outcome researchers have focused most heavily on client factors as predictor variables. Therapist factors rank second. Investigations of interaction among client and therapist factors are relatively rare and so are designs using process factors as predictors of outcome. The same set of priorities is evident in research aimed at predicting/explaining premature termination. General Factors Client characteristics To judge from the literature, professionals see premature termination from (or dropping out of) therapy to be largely the responsibility of the client. To begin with, the nouns "premature terminator" and "dropout" are used freely. Therefore, it is not surprising that the vast majority of predictor variables

PAGE 32

22 investigated are relatively fixed characteristics of clients such as age, sex, socioeconomic status and diagnosis. In 1975, Baekeland and Lundwell reviewed 62 studies of dropping out among adults in individual psychotherapy. Sixty of these studies found at least one significant predictor variable. The majority were client factors. Much replication remains to be done, as many successful predictor variables were investigated in only one study. However, a few were well substantiated by several studies utilizing different populations and different definitions of dropping out. The most extensively investigated variable, socioeconomic status, was associated consistently with dropout rate. Whether defined by education, occupation, or income, higher status predicted lower dropout rates in 16 of 18 studies. Race and IQ, generally associated with socioeconomic status, also are predictive in the expected direction (Baekeland and Lundwell, 1975). Another consistently predictive variable is sex of client. All four studies reviewed found that females dropped out more often than males did. This is one of the few points at which remaining in therapy is (statistically speaking) a different phenomenon from making progress. Most outcome studies have found females to be the more successful clients (Luborsky et al 1971). Client age predicts dropping out somewhat less well. Three of four studies found age predictive, but their results are not completely consistent. In one study of brief psychotherapy, client retention rates varied directly with client age. However, a project studying longterm therapy found that people in their 30s were more likely to remain

PAGE 33

23 than were either younger or older clients (Baekeland and Lundwell, 1975) Interestingly enough, most outcome studies find that, among adults, the younger the client the brighter the prognosis. Apparently younger clients, like female clients, are relatively likely to drop out, but those who stay make particularly promising candidates for therapy. According to Baekeland and Lundwell (1975), diagnosis is somewhat predictive of dropping out. Five of nine studies found sociopathy, paranoia, alcoholism and extreme depression predictive of premature termination. Anxiety and moderate depression, on the other hand, improved the chances of remaining in therapy. Related predictors studied only once each include referral source, psychological mindedness, defensiveness, stereotypic masculinity or femininity, and personality disturbance in a significant other (Baekeland and Lundwell, 1975). None of these findings has been challenged seriously by more recent research. However, some have been reinterpreted. In addition, more recent studies have added to the list of predictor variables. Piaster and Rudestam (1975) found social class status predictive of premature termination rate at one facility studied but not at a second facility. The two facilities had similar dropout rates and served similar clienteles. They did differ in types of therapies employed and in therapy offered. Social class was predictive in a setting with older, traditionally trained therapists doing analytic therapy and v/as not predictive where more eclectic therapy was practiced by younger therapists with a variety of types of training.

PAGE 34

24 Trait anxiety was supported as a predictor of continuation in therapy by two recent studies. In one, Barr (1977) noted that it seemed to explain longer attendance in therapy by first-borns. In the second study. Stern et al (1975) found that the predictive value of the Taylor Manifest Anxiety Scale remained even when the effect of social class was removed. At the same time they found that two other diagnostic variables— sociopathy (as measured by the Behavior Disturbance Scale) and authoritarianism (as measured by the California F Scale)— washed out when the effect of social class was removed. In addition to his findings regarding trait anxiety, Barr (1977) added to the information relating client retention to diagnosis while making more questionable its connection with client age. In Barr's sample, the younger clients attended more sessions. Also, significant differences were found among diagnostic categories. Psychotics averaged longest in attendance, while people with marital problems left most quickly. It is important to note, however, that Barr simply averaged attendance figures to characterize his client groups. One cannot draw conclusions about comparative rates of premature termination from such data. The same caution applies to a study in which Nacev (1977) found dependency and ego strength (as measured by the MMPI) to be of no value in predicting length of stay in therapy. That caution is underscored by the work of Goldenholz (1975). Goldenholz separated "number of sessions" from "type of termination" and used both as dependent variables. She found no difference between males and females in the

PAGE 35

25 tendency to terminate unilaterally. Males averaged more therapy sessions, however, a result the researcher attributed to their being more severely disturbed initially. McGrogan (1976) screened demographic data, diagnosis, MMPI scales, CPI scales, and therapist experience as predictors of number of visits and of termination with or without therapist's approval. In line with earlier research, he found demographic and diagnostic data to be his most powerful predictors. However, he observed that little of the total variance was explained in the multiple regression analyses. In fact, he recommended abandoning this type of research. A final word about assessing client proclivities is provided by Koran and Costell (1973). In their investigation of the value of test scores, therapists' predictions, and so forth, the most valuable predictor of early termination was found to be refusal of the client to complete pretest questionnaires. The results were quite powerful: Refusal carried an 80 percent risk of premature termination, versus 17 percent overall, and it identified 53 percent of the early terminators. Therapist characteristics In 1975, Baekeland and Lundwell found only one therapist factor which predicted client retention in more than two studies. Interestingly, it is the sole reliable predictor of therapist success as well (Luborsky et al 1971). The factor, of course, is therapist experience. In six of the seven relevant studies they reviewed, inexperienced therapists lost a significantly disproprotionate number of clients.

PAGE 36

26 Only two other "pure" therapist factors predicted client retention in any study reviewed. In one study, therapists low in ethnocentricity held more clients. In two studies, female therapists had a significantly lower rate of premature termination than did male therapists (Baekeland and Lundwell, 1975). The other "therapist" factors reported actually are process-related factors, so they will be discussed later. More recent research has little to add in this area. Its most important result, probably, is to render even more questionable the tentative connection between therapist sex and premature termination reported previously. Andrews (1975) found no relationship between sex of therapist and number of sessions attended. Goldenholz (1975) got the same result. In addition, the latter also found sex of therapist to be unconnected with type of termination. In the realm of recent null results, Butler (1977) found therapist training unrelated to clients' remaining in therapy. Fiester (1977) checked therapist demographic characteristics, training, and experience and found nothing predicting whether or not clients would stay past the second session. Of course, neither of these researchers actually studied premature termination. Interaction of client and therapist characteristics Some of the results in this area are intriguing. It is important to note that most have not been replicated. This is not damning evidence, however; for most, no replication has been attempted. Two interaction variables form an interesting exception to that generalization. Each has been tested repeatedly for its value as a

PAGE 37

27 predictor of premature termination, and the results consistently are negative. Neither predicts therapy success, either. Those null results so contradict the convictions of many therapists and researchers, however, that they continue to be questioned. The two variables, of course, are client-counselor matching by sex and client-counselor matching by race. For example. Vail (1974) hypothesized that black, lower-class clients would remain in treatment longer if matched for race and matched for sex. In fact, matching for race made no difference, and matching for sex actually was associated with early termination. Using an unusually good experimental design. Safer (1975) tested sex-of-therapist by sex-of-client interaction for both weekly and biweekly therapy schedules. In earlier research. Safer had found that female-female pairs averaged the largest number of sessions. This time, such pairs were indistinguishable from male-male pairs in weekly therapy and had lower client retention when therapy was biweekly. In addition, Safer found that male therapists with female clients averaged more sessions than did female therapists with male clients. Andrews (1975) found no differences at all for the sex-oftherapist by sex-of-client interaction. Mendelsohn (1966) also found none. A related hypothesis attempts to explain high rates of premature termination among working-class clients as a matching failure. However, Butler (1977) found that social class origin of her therapists made no difference in the attendance patterns of working-class clients.

PAGE 38

28 On the other hand, there is solid evidence that therapy or therapist matching a client's preference makes early or premature termination less likely. In contrast to the null results obtained when similarity of sex is the independent variable, Norkus (1976) found that gratifying a client's preference for a male or female therapist made the client more likely to remain. All of her clients v;ere female. Another researcher, who found no demographic variables predictive of termination type among his subjects, found that therapist behavior which matched clients' preferences or expectations made premature termination less likely (Sasseen, 1975). Sasseen was fairly careful with his terminology; to be defined as a premature terminator, one of his subjects had to attend counseling four times or less and terminate without contacting the therapist. Ross's (1977) discussion of the possible meanings of clienttherapist similarity offers one explanation for the data presented so far in this section. Ross points out that (a) similarity probably matters only for characteristics of importance to the client, and (b) the valence will depend on whether those characteristics are positively or negatively valued by the client. Incidentally, the latter is related to self-esteem, and many clients are symptomatically low in self-esteem, at least at the beginning of therapy. If a sample contains enough clients who are indifferent to the characteristic studied or who value it negatively, a matching hypothesis will be disconfirmed. Demographic matching is poorly supported. On the other hand, a variety of researchers have found client retention superior when clients

PAGE 39

29 and therapists are similar in some less obvious ways. Various personality variables have been used in this research. Similarity of goals for therapy and of role expectations also predict better client retention. One early study of personality matching was done by Mendelsohn (1966). Mendelsohn tested clients and therapists using the Myers-Briggs Type Indicator. No main effects were found relating either client or counselor personality type to early termination. However, overall client-counselor dissimilarity strongly predicted early termination. Landfield (1971) uses a complex set of criteria to categorize clients as remainers or terminators (see page 12). His predictor variables are complex, as well: content similarity of constructs and similarity in the organization of constructs on the Role Repertory Test, Both predict a client's remaining in therapy rather than terminating prematurely. One can argue plausibly that client-therapist similarity on therapeutic issues would be far more important than similarity on general characteristics for predicting the likelihood of a client's remaining in therapy. Vihether or not that is the relevant dimension, a number of studies have found similar ideas of therapy goals and of expected roles to be reliably predictive of client retention. Raschella (1975) found that client-therapist similarity in both the content of therapy goals and the priority of those goals made premature termination less likely. (Raschella used "dropping out" synonymously with "premature termination.") However, he also found

PAGE 40

30 client-therapist similarity predictive of session length and of the spacing between sessions, suggesting that goal similarity is related to client comfort — and possibly to therapist comfort, as well. Butler (1977) studied retention of working-class clients. Neither therapist training nor therapist background predicted client retention in her study. Retention was more likely when the client's problem was perceived similarly by client and therapist. It also was related to good matching between client expectations and therapist behavior. Horenstein (1973) distinguished between "confirmation" and "failure to disconfirm" expectations. For his college student population, specific confirmation of expectations regarding therapist insession behavior seemed unimportant. However, a clear violation of expectations was associated with increased tendencies to drop out or to come late or attend irregularly. Sandler (1975) tested both therapists and clients pretherapy, before the third session, and before the twelfth session regarding their role expectations. He found that terminators had significantly higher pretherapy discrepancy scores. Also, while discrepancies generally dropped over time, they dropped most for remainers and their therapists. Since Sandler used both number of sessions attended and therapist's judgment of termination type, these results can be properly considered to relate to premature, not just early, termination. Based on his results, Sandler suggested either matching clients to therapists or using pretraining to align clients' expectations more

PAGE 41

31 closely to those of their therapists. Client-therapist matching in this area has not been researched, but there is a small body of literature on the effects of pretraining. Its results divide fairly evenly between those showing a positive effect from pretraining and those showing no significant effect. An example on the positive side is Martin's (1975) experiment using 40 community mental health center clients. Half were treated, using a videotape with both didactic and modeling content, and half were no-treatment controls. Viewing the videotape significantly reduced the rate of "unilateral termination." On the negative side, Holliday (1978) used live, group preparation with half of his subjects and kept the rest as no-treatment controls. These were community mental health center clients also. The treatment had no effect on the dependent variable, number of sessions attended. Of course, this study differs from Martin's (1975) in at least two important ways. However, the inconsistency in results is typical for these studies as a group. As was mentioned previously, sex-of-client by sex-of-therapist interaction has failed as a predictor variable. However, sex-of-client considered in combination with other client or therapist characteristics has led to some intriguing results. One early investigation found that low-autonomy females assigned to dominant counselors and high-autonomy females with average counselors are less likely to terminate early. Neither combination made any difference for male clients (Heilbrun, 1961a). The same researcher

PAGE 42

32 found that low-autonomy females are likely to terminate earlier than autonomous females, while the reverse is true for males (Heilbrun, 1961b). Attempting to find an explanation for these results, Heilbrun replicated his earlier research and also asked clients to fill out a questionnaire eliciting their degree of satisfaction following initial interviews. He found that the dependent female clients were more satisfied with relatively nondirective counselors, whereas autonomous females were more positive about directive counselors. The reverse was true for male clients. The interview satisfaction seemed to be predicting dropping out! Heilbrun interpreted these results as supporting the hypothesis that premature termination, defined here as failure to return following the initial interview, is due at least in part to cathartic relief of anxiety produced by a satisfying interview (1974). Baekeland and Lundwell (1975) also reviewed one study in which sex-of-therapist was crossed with Rorschach productivity scores of clients. There was a main effect for sex of therapist, with female therapists superior in client retention. However, it also appeared that female therapists were keeping primarily the less "productive" clients, while male therapists were better at retaining those with high productivity scores. Unfortunately, this work has been neither replicated nor tied to other research data, so interpretation is difficult. Client-Therapist Dynamics So far, this section has concerned itself exclusively with general characteristics and/or predispositions on the part of therapists and clients. What remains is to consider the dynamics of the specific

PAGE 43

33 dyad involved in therapy. Within the body of research on premature termination these dynamics have been approached in two ways. Some researchers have looked at the client's reaction to the therapist or the therapist's reaction to the client. Others have used the thirdperson point of view and measured process variables. Client reaction to therapist The bulk of premature termination literature assumes that clients quit prematurely because of dissatisfaction with their therapists or with therapy. Nevertheless, it is rare for an investigator to test that assumption and rarer still for one to delve into the content of terminators' perceptions of their therapists. Those researchers who did so question clients got a few surprises, including evidence that there may be more than one major type of premature termination. Most often, terminators did view their therapists more negatively than did remainers. Saltzman et al. (1976) said that dropout subjects saw their therapists as less respectful and understanding of them than did remaining subjects. Interestingly enough, the therapists' self-ratings differed in the same fashion. Similarly, Caligor's (1975) dropouts rated their therapists relatively unaccepting and argumentative. Like Saltzman et al (1976), Caligor had data suggesting that this represented true differences in therapists' behavior, not just differing client perceptions. Therapists in Caligor's study who had many dropouts were rated relatively critically ("unaccepting," "distant") by both dropouts and remainers. Rosen's (1978) more recent study got conflicting results. Global attraction to therapist was significantly related to

PAGE 44

34 client-reported completeness of counseling. However, these client reports were unrelated to counselors' perceptions and also to number of sessions attended. Therefore, it is doubtful that Rosen measured anything more than general client positiveness toward the counseling experience. More significant and challenging is Heilbrun's (1974) conclusion, based on a series of investigations reported previously, that termination after the first interview may signal satisfaction. The evidence was strongest for dependent female clients seeing nondirective therapists, but the data suggested that the same dynamic might be operating for other client-therapist matchings. In the same vein, a pair of studies using multivariate analysis suggest that there really are at least two quite different types of dropout dynamics. Feister (1977) divided a group of 17 community mental health center therapists at the median by dropout rate and obtained demographic, training, and experience data for them. No differences appeared. He then gave clients Orlinsky and Howard's Therapy Session Report to complete after their first interviews. The clients of high-attrition therapists saw themselves as more inhibited and their therapists as more anxious and more directive than did the clients of low-attrition therapists. Two interactional factors appeared as well. The high-attrition group was relatively overstocked with (a) ineffectual therapists seeing confrontive clients and (b) anxious therapists providing cathartic relief for their clients. Fiester and Rudestam (1975) found more "positive" early terminations. In their factor analytic study, several factors suggested

PAGE 45

35 a limited sort of success. The clients were satisfied, saw their therapists as warm and effective, and terminated because they felt better. However, they also found a group of terminations in which clients, most of them seriously disturbed, were given inexperienced therapists. The clients were hostile in the interviews and dissatisfied afterwards. Therapist reaction to client Baekeland and Lundwell (1975) reported several therapist reaction factors to be predictive of clients' dropping out. Only one had appeared in more than one article as of their 1975 publication date, but all of them probably are related to each other. According to Baekeland and Lundwell, three studies found therapist dislike of clients, globally defined, to be predictive of early and/or premature termination. Two other factors, investigated in one project apiece, also predicted dropping out. Those two are low expectation of client improvement (Goldstein, 1960) and lack of interest in the patient's problems (Baekeland and Lundwell, 1975). It seems reasonable to assume that most therapists prefer clients who they see as having a good prognosis and interesting problems, so these factors may be regarded as specific elements in global therapist attraction to client. More recent studies have not been unanimous in connecting therapist attraction to client with client retention. Notably, Hochstadt (1975) found that, for his sample of community mental health center clients seeing either professional or paraprofessional counselors,

PAGE 46

36 global attraction to client failed to predict the number of sessions clients would attend. On the other hand, Rosen's (1978) investigation using a sample of college students and therapists did find global attraction significantly related to client retention. It is hard to reconcile those results. Possibly Hochstadt's (1975) inclusion of paraprofessional counselors made a difference, but this is just a speculation. Saltzman et al (1976) provided one of the most detailed and methodologically sophisticated investigations of client perceptions in relationship to treatment duration. At the same time, they investigated therapists' perceptions of their clients. The researchers found that therapists saw themselves as less respectful, understanding, involved, and emotionally reactive to their predropout clients. In addition, they saw those clients as less open, less self-responsible, less likely to improve, and they saw their relationships with those clients as less continuous than their relationships with clients who turned out to be remainers. It is important to note that all of the work in this section is correlational only. Prediction does not imply causality. One would expect experienced therapists to be less positive toward clients who are likely— for whatever reason— to be therapy failures or to reject therapy entirely. At the same time, it may be that clients respond to perceived therapist negativity by dropping out. This is a chicken-andegg problem. The answer may be that an unspecifiable mix of therapist and client perceptions of each other contribute to a therapy interaction in which client retention becomes progressively more unlikely.

PAGE 47

37 Therapy process factors In 1975, Baekeland and Lundwell found only three studies which connected true therapy process with client retention rate. One simply established that gross disruption of the therapy process, through patient reassignment, made premature termination more likely. The other two dealt with more intricate variables. Winder et al. (1962) studied therapists' responses to client statements. They found that therapist avoidance of dependency content in patients' verbalizations made premature termination more likely. Heilbrun (1973), working with an entirely female sample of clients, found that dropouts had disclosed more than remainers if their therapists were male. Among clients with female therapists, however, remainers were the more disclosing. Neither of these studies has been replicated, and neither is very useful standing alone. Nevertheless, they are praiseworthy for their specificity. Unfortunately, few later researchers have followed their lead regarding specificity. One did use a design much like that of VJinder et al. (1962) to test the hypothesis that interpreting transferencebased behavior would increase clients' length of stay in therapy. His sample was composed of 39 clients seen by 17 social workers. The hypothesis was not supported at the .05 level of significance (Goldstein, 1975). Most other research using process variables as predictors of premature termination has had those variables assessed by clients, by

PAGE 48

38 therapists, or both. In other words, they are studies of clients' and therapists' reactions to each other. One outstanding exception is a recent study by Schiller (1978). Schiller used tape recordings of sessions and had both client and counselor statements rated, using the Leary Interpersonal Checklist, on the dimensions of hostility-affiliation and dominance-submissiveness. Client statements, therapist statements and the complementarity of therapist statements to preceding client statements were checked for their power to predict early termination. Schiller found that early termination (attending five sessions or less) was significantly related to relatively low client-therapist complementarity. It was unrelated to either hostility or dominance expressed by either clients or therapists. In 1970, Meltzoff and Kornreich commented forcefully on the desirability of studying interpersonal process leading to premature termination from psychotherapy. To date, few researchers have heeded their call Studying Therapy Process General Problems Research critics agree that there is a woeful shortage of process research in many areas where process seems of obvious relevance. There is a good reason for this, though. Doing competent process research is difficult. Data gathering must be planned in advance. It is relatively invasive, so client and therapist cooperation is critical. Obtaining process ratings usually means either direct observation or tape/transcript analysis. All are time-consuming and must be preceded

PAGE 49

39 by rater training. Furthermore, useful ratings require good instruments. As of yet, few have been carefully developed and validated. Finally, any good piece of research is theoretically relevant and operational izes theoretical constructs in a plausible manner. Existing operationalizations of many process constructs are controversial, to say the least. Process as a Predictor of Outcome In spite of the difficulties inherent in process research and those associated with studying psychotherapy outcomes, a body of research does exist investigating relationships between the two. Little of it touches specifically on premature termination. Nevertheless, it provides a most relevant background for discussion of therapy process leading to premature termination. Process variables derived from a theory of psychotherapy At one time or another, diverse process variables have been tested for their relationship to psychotherapy outcome. Most have been the object of only transitory interest on the part of researchers, and at this point many results are contradictory. On the other hand, a few variables have been investigated with exceptional thoroughness. The more miscellaneous results will be presented first. According to the research, interpretation may or may not facilitate psychotherapeutic progress. One highly regarded research project found a negative relationship between frequency of interpretations and therapy outcome (Sloane et al., 1975). Another respected researcher found outcome unrelated to the frequency of interpretations in general but positively related to the frequency of classical transference interpretations (Malan, 1976).

PAGE 50

40 Confrontation also has received mixed reviews. Godbole and Verinis (1974) found it more effective than supportive interventions in treating the emotional disorders of physically ill geriatric patients. On the other hand, Padfield (1976) found no significant difference between the two in his study of depressed, lower-class women. Catharsis generally is viewed as a positive process in therapy (Gomes-Schwartz et al 1978), but recent research challenges that as a blanket assumption. Bierenbaum et al. (1976) found that the amount of catharsis experienced in therapy interviews was positively related to outcome for clients assigned to weekly, one-hour sessions. However, clients seen biweekly for half-hour sessions improved most when catharsis was minimal Another interesting study compared the effects on outcome of three process variables, each originating in a different theory of psychotherapy. The clients, members of two therapy groups, rated each session for the amount of therapist acceptance, desensitization, and abreaction they experienced. Observers made similar ratings. Patientobserver agreement was high. Symptom amelioration was used as the sole dependent variable. In a newly formed therapy group, all three process factors were related to improvement, abreaction being the most important. For the ongoing group, only acceptance predicted patient improvement (Cabral et al., 1975). Promising as some of those leads appear, none has been followed up, to date. In powerful contrast stands the substantial body of research concerning therapists' provision of Rogerian core conditions.

PAGE 51

41 There are problems and flaws in the work itself, but at present it represents a unique achievement in the field of psychotherapy evaluation. The Carkhuff Scales are a set of directions for observer evaluation of therapist performance along the critical Rogerian dimensions of accurate empathy, nonpossessive warmth, and genuineness. According to client-centered therapy, it is the therapist's provision of the "core conditions" which is the essence of any successful therapy. This hypothesis has been tested in various settings and with various types of clients and therapists. In general, results have been positive: High conditions are associated with improvement, while low conditions bring little change or deterioration. A good example is provided by Schauble and Pierce's (1974) study of university student clients. Forty-one students and their 34 counselors were rated from early and late tapes of therapy. Counselors were rated on the Carkhuff Scales for empathy, positive regard, genuineness, and concreteness. Clients were rated for depth of self-exploration, locus of control, owning of feelings, commitment to change, and differentiation of stimuli. Clients were also preand post-tested with the Minnesota Multiphasic Personality Inventory. Comparisons (t-tests) between successful and unsuccessful groups showed that counselors differed significantly on all four scales for their late tapes and on all but empathy for their early tapes. Clients had no significant differences on the early tapes, but all measures differed significantly on the later tapes. Schauble and Pierce thus took the problem beyond

PAGE 52

42 simple correlation and demonstrated that it was the counselors, not the clients, who differed at the onset of counseling. A number of criticisms have been aimed at this body of work. Early criticism focused on rater inaccuracy and experimenter bias. Truax and Carkhuff (1967) reported that these objections were countered in later research. However, other problems remain. One source of confusion is the fact that many Rogerians (and some critics, as well) apparently assume that core conditions are perceived similarly by clients and observers. Researcher results from projects using Carkhuff 's scales for trained raters (or Truax 's reworking of those scales) often are intermingled with those from projects using the Barrett-Lennard Relationship Inventory, a client-report measure. One of Truax 's (1966) own studies invalidates such an equivalency, for Truax found that clients apparently do not need to perceive the therapist as warm, empathetic, and genuine in order to benefit from high levels of those conditions. Also, factor analytic studies question the integrity and separateness of the core conditions (Gladstein, 1977; Lambert, de Julio, and Stein, 1978). Given that the three conditions are highly intercorrelated, relatively stable across time, and independent of the patient, do they represent anything more profound than a general "good guy" factor (Gladstein, 1977)? Another problem is evaluation methodology. Most studies involve numerous outcome measures, many admittedly thrown in with no theoretical reason (Meltzoff and Kornreich, 1970). This approach to evaluation suggests that "core conditions" are seen as a panacea. While

PAGE 53

43 more than a chance proportion show significant change in the predicted direction, the presence of numerous measures showing no difference does weaken overall results. The question, "What don't core conditions do?" is never answered, let alone interpreted (Gladstein, 1977). This is not an exhaustive list of the criticisms which have been leveled at core conditions outcome research. Certainly many more specific criticisms can be directed at individual studies. On the other hand, it should be noted that this body of work has received an unusual amount of critical attention. More important than specific results or criticisms are the contributions made to process research in general by this major research effort. Rating scales were developed, checked for reliability, and validated empirically. The scales were used with a variety of populations: schizophrenics, delinquents, college underachievers, as well as the more usual clinic populations. Research procedures were refined to minimize or eliminate a variety of biases which go uncontrolled in the average study. For these reasons, the body of research as a whole forms a benchmark against which other studies of therapeutic process can be measured. A-theoretical variables Most therapy process research— including all of the aforementioned studies— uses process variables suggested by one or another theory of psychotherapy. In part, they serve to test the assumptions of theoreticians. Certainly this is a valuable endeavor. However, e\/ery such study faces one difficult dilemma. If it examines only therapy of the school from which its constructs are drawn.

PAGE 54

44 generalizability is limited. If, on the other hand, a variety of therapeutic approaches are included, the researcher risks obtaining only null results. For example, the level of Rogerian core conditions provided may be quite unrelated to outcome in Gestalt therapy. One solution to this dilemma is to use process variables which are relevant to therapy in general rather than to a particular school of therapy. Another approach is to treat therapy simply as dyadic interaction and use variables found by social psychologists to describe critical aspects of such interaction. The first approach was taken in a recent study by Gomes-Schwartz (1978). The methodology of that study is exceptionally good. Relevant, a-theoretical process variables were used: patient exploration, therapist exploration, patient participation, patient hostility, therapist warmth and friendliness, negative therapist attitude, and therapist directiveness. Client variability was limited to males with depression and social introversion. The therapists all were very experienced and belonged to distinct categories: dynamic psychiatrists, Rogerian psychologists, and "inherently helpful" professors. Client assignment was random, and therapy lasted a defensible length of time: 17 sessions, on the average. The outcome criteria were equally good: therapists' ratings, observer clinicians' ratings, patients' ratings, and MMPI scores. Predictable process differences were found among the three categories of therapists. However, all types were equally effective. Moreover, the same relationships between process variables and outcome

PAGE 55

45 variables obtained across therapist categories. Nothing predicted patient-rated change or MMPI change. Several variables predicted therapist and observer-rated changes. However, with the effect of other processes partial ed out, only patient involvement correlated with any change measure at the .05 level of significance. It correlated with four change measures. The author notes that these results support Frank and his coworkers at the expense of both Rogers and Freud (Gomes-Schwartz, 1978). In addition, the doctrine of client responsibility is bolstered, though the inclusion of only very experienced therapists controlled therapist variability to a significant extent. A second approach to studying psychotherapy process without recourse to variables linked to one or another theory of psychotherapy involves using variables found by researchers in other areas of psychology to be central to dyadic interaction. This second approach was taken by Leary (1957) when he constructed the Leary Interpersonal Checklist. The Leary Interpersonal Checklist locates behavior on a circumplex formed by two orthogonal, bipolar dimensions: dominance-submission and hostility-affiliation. These dimensions are based in social psychology research. Foa (1961) has documented the background research, reviewing a variety of studies in which factor analysis was applied to behavioral observations. For group members' ratings of each other, three bipolar factors tended to appear: individual prominance and achievement, sociability, and aiding group achievement (Carter, 1954). In studies of dyadic behavior, however, the third disappeared while the

PAGE 56

46 first two remained under various names; "individual assertiveness" and "sociability" (Borgatta et al., 1958); "control-autonomy" and "love-hostility" (Schaefer, 1959). In addition to locating behavior on the circumplex, the Leary Interpersonal Checklist provides a means for assessing the complementarity of one person's behavior to that of another, either in general or on an interaction-by-interaction basis. Complementarity scores are included in much of the therapy process research utilizing the Checklist. Mueller and his coworkers performed some of the earliest research using the Checklist to measure psychotherapeutic process. They were interested in methodological problems. Also, they wanted to know whether or not Leary 's assumptions about behavioral reciprocity held true for psychotherapeutic interviews. In the first project, ten-minute segments of taped interviews were scored, with a major (overt) theme and a minor (subtle) theme named for each interaction unit. Scores were grouped by quadrant. Analysis was restricted to comparisons of people's response totals by quadrant. The researchers found, for example, that the more frequently a therapist's major themes fell in the hostile-dominant quadrant the more hostile-dominant and the fewer affiliative-submissive major themes the client would exhibit. This procedure gave the research problems with inter judge reliability. In particular, coding of minor themes tended to be unreliable (Mueller and Dilling, 1968). Moreover, no conconclusions could be drawn about microprocess from this sort of analysis. In a second project, Mueller (1969) changed several procedures. Raters were allowed to score as many themes per statement as seemed

PAGE 57

47 appropriate to them. Instead of looking at totals per subject, complementarity was scored on an interaction-by-interaction basis. Interjudge reliability was figured likewise, and reliability improved. Moreover, Mueller analyzed the data once with themes grouped by quadrant and once with grouping by octant. Mueller preferred the use of octants, pointing out that more significant relationships were found that way (18 versus 8). Of course, the percentage of possible relationships found significant was higher when scoring was done by quadrant (50 percent versus 28 percent). A stronger point in favor of octant scoring is its ability to establish more specific complementarities than those detected by quadrant scoring. Two associates of Mueller's have taken the additional step of attempting to connect therapist and client behavior with success in psychotherapy. Crowder published his results in 1970, and Spierling followed suit in 1972. Crowder (1970) predicted that out-of-role behavior on the part of either therapist or client would make successful therapy less likely. He defined "reality-based," in-role therapist behavior as that falling in the affiliative-dominant quadrant and defined "reality-based" client behavior as that rated affiliative-submissive. Like Mueller (1969), Crowder did not score reflexes for their intensity. Crowder's (1970) hypothesis was not supported jat the .05 level of significance. However, Crowder did find significant differences in both therapists' and clients' behavior when he compared successful and unsuccessful cases. In successful cases, therapists were more likely

PAGE 58

48 to be hostile-dominant and less likely to be hostile-submissive during early sessions. During closing sessions they were more likely to be affiliative-dominant and less likely to be either hostile-dominant or hostile-submissive. Successful clients were more hostile-dominant or affiliative-submissive and less hostile-submissive or affiliativedominant during early sessions than were unsuccessful clients. During midtherapy, they continued to be more affiliative-submissive and less hostile-submissive. No client differences appeared during closing sessions. These results are not easy to interpret. The most clearcut pattern suggests that hostile-submissive (otherwise known as passiveaggressive) behavior, on the part of either therapist or client, is associated with unsuccessful therapy. It is interesting to note that Crowder (1970) found both client and therapist process predictive of therapeutic outcome. In a study reviewed earlier, Schauble and Pierce (1974) found poor process in the early stages typical of the therapists— but not the clients— of unsuccessful dyads. By contrast, Crowder's work suggests that therapists and clients share the responsibility for failure. Spierling (1972) examined clients' reports of their relationships with others as well as the in-therapy behavior of both clients and therapists. Clients' reports were taken from the early stage of therapy, while in-therapy behavior was sampled from early, mid-therapy, and late sessions. Ten successful and ten unsuccessful client-therapist pairs were studied.

PAGE 59

49 Spierling (1972) compared (1) similarity between self-reported client behavior toward others and observed client behavior toward the therapist, by outcome and by therapy stage, and (2) similarity between client-reported behavior received from others and observed therapist behavior toward the client, by outcome and by therapy stage. In addition, all analyses which included information on reported interactions with significant others were performed twice, once including parents and once excluding them. In the first comparison, no main effect was found for either therapy success or therapy stage. However, one significant interaction appeared. Successful clients behavioral ly differentiated their therapists more from nonparental others as therapy progressed, while unsuccessful clients did not. The second comparison found one main effect and no interactions. As therapy progressed, all therapists tended to treat clients more as clients had reported others treating them (Spierling 1972). Since the reporting was taken from early sessions only, this result documents a significant increase in countertransference as therapy progressed. The lack of either a main effect for outcome or an interaction of outcome with therapy stage again underscores the importance of client, rather than therapist, behavior in the production of psychotherapeutic success. A better known outcome study using the Leary Interpersonal Checklist was done in 1975 by Dietzal and Abeles. Dietzal and Abeles focused on the complementarity of therapists' responses to clients' statements. They defined complementarity in Leary 's terms so that, for example, a

PAGE 60

50 hostile-submissive statement is complemented by a hostile-dominant reply. Dietzal and Abeles (1975) found that complementarity levels of early sessions and of closing sessions had no relationship to therapeutic outcome. However, during the middle stage of therapy those cases which would later be judged successful were characterized by significantly less complementarity than were unsuccessful cases. Thus, they found successful therapists more likely to avoid the high probability response during mid-therapy. Seen another way, Dietzal and Abeles 's results could be interpreted to mean that successful cases are characterized by more interpersonal control on the part of therapists. Giving a noncomplementary response is refusing to follow the other's lead. However, a note of caution is in order. It is important to realize that both clients and therapists stay in-role [as Crowder (1970) describes it] most of the time. That is, the majority of client statements are affiliative-submissive, and most therapist statements are affiliative-dominant (Schiller, 1978). Moreover, since those two quadrants are complementary, the complementarity of therapist-client interactions would be expected statistically to be above the theoretical average. This observation brings us to the recent study of Schiller's connecting client-therapist behavior and length of therapy. Schiller used the Interpersonal Checklist to rate three five-minute segments from each audiotaped intake interview. Ratings were made by quadrant.

PAGE 61

51 Research hypotheses asked whether length of therapy could be connected with therapist process, with client process or with therapist-client complementarity. Schiller (1978) got highly stereotyped behavior. Ninety-six percent of his therapists' responses were judged affiliative-dominant, and 72 percent of the clients' statements were judged affiliativesubmissive. Furthermore, 72 percent of the response pairs were fully complementary. These figures are higher than those obtained by other researchers (cf. Mueller, 1969). Schiller (1978) distinguished three outcome groups: nonreturners (attended intake only), early terminators (attended one to five postintake sessions), and remainers (attended six or more sessions after intake). No therapist or client behavior predicted any outcome. However, early terminators enjoyed significantly less complementarity with their therapists than did either nonreturners or remainers. With its use of "early termination" as an outcome category, Schiller's research currently comes closest to using the Interpersonal Checklist to predict premature termination. Hypotheses Studies connecting psychotherapeutic process with premature termination are virtually nonexistent. Therefore, it has been necessary to refer to related research in order to obtain guidelines for choosing independent variables. Recent studies connecting therapeutic process either with early termination or with general therapy outcome have been

PAGE 62

52 reviewed. Based on that review, the following choices have been made. First, it seems desirable to use independent variables which are not associated with any particular school of psychotherapy. For one thing, therapists at the target facility vary considerably in their theoretical orientations. For another, the field is so wide open that designing studies for maximal generalizability also maximizes their potential usefulness. The literature reviewed strongly suggests as independent variables the two bipolar dimensions operationalized in the Leary Interpersonal Checklist. The two dimensions have emerged in several factor analytic studies (Foa, 1961), and Leary 's (1957) operationalization of them is well tested for validity and reliability (see Chapter Three) The hypotheses cover three categories of predictor variables: client behavior, therapist behavior and the complementarity of therapist behavior to preceding client behavior. Specific hypothesized relationships with premature termination are based on research results obtained by Crowder (1970, 1972) and by Schiller (1978). Hypothesis 1 : Weighted Percentages of Interaction in the hostile-submissive quadrant will average higher for premature terminators than for remainers. Hypothesis 2: Weighted Percentages of Interaction in the affiliative-submissive quadrant will average higher for remainers than for premature terminators.

PAGE 63

53 Hypothesis 3: Weighted Percentages of Interaction in the hostile-submissive quadrant will average higher for the therapists of premature terminators than among the therapists of remainers. Hypothesis 4: Weighted Percentages of Interaction in the hostile-dominant quadrant will average higher for the therapists of remainers than they will among the therapists of premature terminators. Hypothesis 5: Therapist Complementarity of Response for Dominance scores will average higher among therapists interacting with remainers than they will among the therapists of premature terminators. Hypothesis 6: Therapist Complementarity of Response for Affiliation scores will average higher among therapists interacting with remainers than they will among the therapists of premature terminators.

PAGE 64

CHAPTER THREE METHODOLOGY Subjects A sample of 30 student clients and their therapists was selected. The primary selection criterion for clients was application and acceptance for ongoing individual psychotherapy at the Psychological and Vocational Counseling Center of the University of Florida. Since this research was intended to uncover predictive factors found in the first postintake session, clients who had received other ongoing therapy within the previous year were excluded. Clients enrolled in the university's counseling psychology program or counselor education department also were excluded because of the possibility of their being known to the experimenter and/or the tape rater. From the pool of eligible clients, clients were further selected based on their willingness to participate and on their assignment to a participating therapist. Inclusion in the final sample depended as well on therapists' completion of the research materials and on the need to balance the sample for type of termination. The therapist sample included all willing staff (senior staff, interns, practicum students and other trainees) at the Psychological and Vocational Counseling Center except for the researcher herself. No therapist was paired with more than three clients, a stipulation designed to limit the influence of therapist idiosyncracies on the data. 54

PAGE 65

55 Data Collection At the time of intake, potential research subjects were given a consent form explaining the data collection in general terms. Intake therapists were instructed to discuss the consent form with their clients, if necessary, and to obtain clients' signatures if possible. In addition, clients completed standard intake forms which include demographic data. The first postintake session of each subject was audiotaped by the therapist. The taping was presented as routine in order to minimize reactive bias (see Hollenbeck, 1978). Fifty-nine clients and 24 therapists participated in this initial phase of the study. Therapists kept attendance records recording each appointment and whether or not the client attended. Clients were considered to be continuing in therapy so long as they attended their scheduled appointments or rescheduled any they missed. However, if a therapist initiated rescheduling (by contacting a client who canceled or failed to appear), client intentionality was considered debatable; such cases were discarded from the final sample unless the client already qualified as a remainer or unless termination occurred before the fourth postintake session in spite of the therapist's efforts. After termination, the therapist completed a section at the bottom of the attendance record giving his or her view of how termination took place. Data collection proceeded until 15 cases of premature termination and 15 cases of a client's remaining in therapy were completely documented. The final study sample was drawn to include all 15 cases

PAGE 66

56 of premature termination in which taping had been adequate and 15 cases randomly selected from among the 25 with audible tapes where criteria for remaining in therapy had been met. This final sample included 16 therapists, each working with one, two, or three clients. The 30 cases in the final study sample are listed in Appendix D by study number. For each case, client sex, race, years of education completed, and parental occupation are listed. Therapist sex, race, and experience level are provided, and the termination status of each case is specified. Instrument: The Leary Interpersonal Checklist The Leary Interpersonal Checklist has an unusual combination of virtues which recommend it for analysis of therapy process. For one thing, it is solidly grounded in a theory of personality, Leary' s Interpersonal Theory (Leary, 1957). Second, it was constructed rationally, rather than empirically. That makes validating research necessary but assures that any research done will do more than simply duplicate instrument construction procedures. Furthermore, the Leary has been well validated in a variety of studies. Finally, it already has been used in a variety of psychotheraphy research. Therefore, any further studies using the Leary can be placed within an ongoing field of research. Theoretical Basis Leary 's Interpersonal Theory assumes that What a person does in any social situation is a function of at least two factors, (1) his multilevel personality

PAGE 67

57 structure and (2) the activities and effect of the "other one," the person with whom he is in interaction. (Leary, 1957, p. 91) It goes on to say that the basic unit of behavior is the "interpersonal effect" or "interpersonal mechanism." This is defined as the interpersonal function of a unit of social behavior. For example, "He is rejecting help" is an interpersonal effect. Leary categorized these effects into a list of "interpersonal reflexes" (Freedman et al., 1951). Leary's (1957) theory is interactional in two senses. First, it specifies interpersonal interaction as the arena in which "personality" has most meaning. Second, it focuses on interaction as a source of personality. According to Leary, each interpersonal reflex has a reciprocal which it tends to evoke. Therefore, interpersonal exchanges both create and reinforce behavior/personality. In Leary's theory, personality is described using just two dimensions, hostility-affiliation and dominance-submissiveness. However, it is seen as existing at five levels. Level I is the level of public communication: what people do when others are present. The others are the levels of II, conscious self-description; III, private symbolization such as fantasies; IV, unexpressed unconscious; and V, values. Although he always expected individuals' ratings to vary across levels, Leary considered it very important to be able to assess personality functioning comparably across levels. For example, connections between Levels I and II can be related to Harry Stack Sullivan's focus on the relationship between self-perception and consensual validation (Leary and Coffey, 1955).

PAGE 68

58 Leary never has claimed that one level is more "real" than another. However, he has described discrepancies among the different levels as resulting in part from the operation of defenses (Leary, 1957). Finally, Leary has described reflexes as so automatic that they generally are not available for direct self-report. They tend to operate as a background to the verbal content, and the individual may well be unconscious of them. Construction Leary (1957) tried five methods for measuring personality at Level I. Only two proved satisfactorily consistent: psychologists' minute-by-minute ratings and more global ratings from those interacting with the subject. Finally, Leary developed the Interpersonal Checklist to systematize the measuring of interpersonal reflexes. The Checklist was constructed rationally. Leary and his colleagues assembled a list of 334 adjectives from psychology literature. These were narrowed down to 106 "generic interpersonal motives" by the pooled judgments of five psychologists. They aimed for a complete and balanced representation of interpersonal behavior. Leary converted as many as possible to transitive verb phrases, but many adjectives remain. The descriptors then were grouped into 16 themes, each reflected in both "normal" and "pathological" behavior. Within each theme, the adjectives were arranged according to "intensity." Intensity (or pathology) was operationalized as the inverse of the frequency with which an adjective was used by subjects to describe themselves or others. Four levels of intensity were

PAGE 69

59 suggested theoretically. (In practice, it is the rare researcher who uses more than two.) Testing the list on hundreds of subjects also led to elimination of words which were too difficult for some subjects or which were interpreted ambiguously (Freedman et al 1951; LaForge and Suczek, 1955; Leary, 1957). At one point, Leary and Coffey (1955) stated that division of behavior among the 16 themes was best for both clinical and research purposes. However, they admitted that combining them into eight categories might be far more convenient. In fact, octant scoring quickly became standard in research, including Leary's own (Shannon and Guerney, 1973). Many projects studying therapy interaction use quadrants. As was mentioned previously, Leary (1957) envisioned each interpersonal reflex as the sum of two orthogonal vectors, hostilityaffiliation and dominance-submissiveness. These vectors are plotted on an x axis and a y axis, with point 0,0 representing neutrality on both dimensions. However, in his visual representations of the dimensions Leary has drawn a circle rather than the square which would be necessary to encompass all possible points were the vector addition actually being performed. This apparently was done to keep all extreme points equidistant from the center. The result is a pie sliced into eight pieces (or 16 pieces or four pieces, depending on the researcher). Behavior located toward the point of a slice is normal, of mild intensity; out toward the crust it becomes more intense, pathological (see Appendix A).

PAGE 70

60 Leary (1957) has detailed two types of ratings which can be obtained from administration of the Checklist. One is characterization of a given unit of behavior by the name of the octant (or other division) within which it falls. The octants are managerialautocratic (high dominance, neutral affiliation); competitivenarcissistic (dominant and hostile); aggressive-sadistic (neutral dominance, high hostility); rebellious-distrustful (submissive and hostile); self-effacing-masochistic (high submission, neutral affiliation); docile-dependent (submissive and affiliative); cooperativeoverconventional (neutral dominance, high affiliation); responsiblehypernormal (dominant and affiliative). Six of these octants have been connected to nosological classifications. Managerial-autocratic and competitive-narcissistic individuals have not been so classified. Leary and Coffey (1955) speculated that those individuals simply are too autonomous to seek therapy. Alternatively, octant or other ratings may be used to measure a subject on just one of the two dimensions. When this is done, scores in "pure" octants are given more power than are scores in "mixed" octants. The ratio is 10:7 (Leary, 1957). This is consistent with Leary's use of a circle, rather than a square, to represent the field of interpersonal behavior. The Interpersonal Checklist is used to obtain Level II (selfreport) as well as Level I (observer-rated) data. However, as a selfreport instrument it has different problems of validation and a

PAGE 71

61 different body of literature reporting its use. From here on, this review will focus on the use of the Interpersonal Checklist as a source of Level I data. Validating Research As a rationally constructed measure, the Leary Interpersonal Checklist needed to prove its empirical validity, as well as its construct validity and its reliability. If the observer-rated Checklist is fully valid, five questions can be answered affirmatively. First, do the items belong in their assigned categories? Second, do those categories relate to each other empirically as their (rationally assigned) position indicates that they should? Third, do scores on the Checklist relate to measures of other levels of personality as Leary theorized that they would? Fourth, do scores on the Checklist dimensions correlate highly with other measures of hostility-affiliation and of dominance-submission? Finally, are scores reliable? At this point it is important to distinguish between two methods for using the Checklist to gather Level I data. Both were sanctioned by Leary (1957). In one system, global ratings are made using the Checklist in the same fashion that it is used for self-ratings. Raters usually are relatives or acquaintances of those rated. In the other system, the Checklist is used as a process measure. Trained raters categorize behavior on an interaction-by-interaction basis. These two very different uses of the same measure never have been distinguished from each other very clearly. Leary (1957) apparently assumed that they were tapping the same phenomenon (Level I

PAGE 72

62 personality) since he speculated about which was the more valid. In current literature both are given the same name, and one must read an article closely to see which method was used. Most research has been done with global ratings. This review will include studies using both types of ratings. The data source of each will be specified. Category integrity and arrangement Only one study was located which checked the validity of item assignment on a category-bycategory basis. HcCormick (1977) had 234 college students sort the Checklist items. He found a high level of agreement between his student raters and Leary's panel of psychologists. This study also supported the circular arrangement of categories since students not assigning an item to Leary's category most often placed it in one adjacent. The circular arrangement of categories was validated by another method when the Leary Interpersonal Checklist was constructed. Item intercorrelations were compared by LaForge and Suczek (1955). They reported "an orderly descending relationship as one moves around the circle from any reference point" (p. 109). In his review, Beutler (1965) accepted this evidence of circularity. However, he pointed out that (1) average endorsement of the categories is uneven and (2) social desirability is uneven also. In other words, the circle is somewhat warped. It needs standardization. Dimensional integrity and orthogonality The factor structure of the Interpersonal Checklist has been the subject of repeated testing and much discussion. This probably happened because an early study found three main factors rather than two. This finding challenged

PAGE 73

63 Leary's theory as well as the validity of the Checklist. All research discussed in this section used global ratings. In 1963 Briar and Bieri factor analyzed 250 self-ratings using the Checklist. They got three factors: "aggressive dominance," "love," and "masochistic inferiority." They then checked this level II data against Level I data by asking 40 social workers to rate a set of profiles constructed to illustrate the polar positions of Leary's two factors. Again, their data supported the existence of three factors. To counter this conclusion, Lange (1970) developed two telling arguments. First, he presented evidence of a two-factor structure emerging from other researchers' analyses of behavior. Second, he noted a flaw in Briar and Bieri 's methodology which could account for their discrepant results. Foa reviewed a variety of studies in which factor analysis was applied to behavioral observations. For group members' ratings of each other, three bipolar factors tended to appear: individual prominance and achievement, sociability, and aiding group achievement (Carter, 1954). In studies of dyadic behavior, however, the third disappeared while the first two remained under various names: "individual assertiveness" and "sociability" (Borgatta et al 1958); "control -autonomy" and "love-hostility" (Schaefer, 1959). Foa (1961) concluded that, ". .a circumplex structure around the two orthogonal axes of Dominance-Submission and Love-Hostility" appears to be the best description of empirical analyses of behavior (p. 348). The reviewer Beutler agreed that Foa's (1961) conclusion holds if one corrects for subjects' "tendency to agree with anything"

PAGE 74

64 (Beutler, 1965, p. 268). The importance of making this correction was underscored by Truckenmiller more recently. Truckenmiller used Briar and Bieri's (1963) methodology and replicated their results (Truckenmiller, 1971). Lange (1970) has pointed out that correcting for average intensity of ratings, a further refinement, makes the argument for a twofactor structure stronger still. He added an "inferential investigation" to earlier factor analytic analyses, gathering data in the form of ratings made by undergraduates watching taped enactments by characters designed to represent one of the four polar extremes of Leary's two dimensions. His data support both the integrity and the orthogonality of those dimensions. One final assumption of Leary's (1957) remains to be covered in this section. Leary theorized that the poles of his two dimensions can be placed in such a way that one bit of behavior is judged for its complementarity to another bit. Specifically, he theorized that hostileaffiliative behavior is complementary to itself but dominant-submissive behavior is complementary to its opposite. Only one study was located which provided a fairly thorough test of Leary's assumptions regarding response complementarity. In it, 14 groups of six women each were given three questions to discuss. The researchers concluded that responses, categorized by octant, definitely were nonrandom. Not all of the associations predicted by Leary were obtained, but many were. For example, docile-dependent behavior did tend to elicit leadership-advice, and the reverse was true as well.

PAGE 75

65 However, self-enhancing-competitive behavior did not elicit its complement, and neither did aggressive-rejecting behavior (Shannon and Guerney, 1973). Shannon and Guerney also tested the complementarity theory on the person-to-person level. In doing this, they were asking whether a person who frequently emits x behavior receives ^ responses with exceptional frequency or infrequency. Of the ten response pairs found to be significantly related in the first analysis, seven were significant in the person-to-person analysis as well. Leadership elicited cooperative behavior; competitiveness and aggressiveness both elicited more competition and aggression; docility both elicited leadership and made a docile response unlikely. Shannon and Guerney 's data discourage an unquestioning acceptance of Leary's theory of complementarity. However, their results should not be regarded as definitive. For one thing, coding was based on typescripts rather than the tapes Leary considered preferable (Leary, 1957). One further possibility emerges from the fact that Shannon and Guerney (1973) allowed neither extra credit for exactness or partial credit for near misses. Perhaps octant scoring divides responses into either too few or too many categories for significant relationships to emerge. The "too many categories" hypothesis is favored by the fact that Shannon and Guerney had to discard one stimulus octant for poor intercoder reliability and two stimulus octants plus two response octants for infrequency. Structural similarity among the levels of personality The existence of five levels of personality is central to Leary's

PAGE 76

66 Interpersonal Theory. Therefore, it was very important to him that his Interpersonal Checklist facilitate the process of drawing comparisons between and among levels (Leary and Coffey, 1955). The validity of such comparisons is a subject for empirical investigation. In general, the results have been positive but limited in scope. The comparability of Level I and Level II data has been established most securely. Comparisons between Level I and other levels are more risky. Again, these researchers restricted themselves to obtaining global ratings, so no conclusions can be drawn about the Interpersonal Checklist used as a process measure. Zimmerman and Vestre (1975) compared patients' self-ratings on the Checklist with ratings made by family members of each patient. They found the factor structure to be the same. There was little content agreement between patients and their relatives, supporting Leary's contention that self and observer ratings represent quite different aspects of a person. Truckenmiller (1971) did a similar study using normal subjects. He compared self-ratings with ratings made by subjects' sorority sisters. In addition, he gathered Level III data by administering the Thematic Apperception Test. Truckenmiller got three trait factors (see previous section) which matched across Levels I and II. However, only one of those traits emerged from the T.A.T. data. Too, Briar and Bieri's (1963) study supports the contention that Level I and Level II data are comparable. This research differs from Truckenmi Tier's in that self-ratings on one population (live subjects)

PAGE 77

67 were compared factor analytically with ratings from quite a different source (social workers' ratings of constructed profiles). The comparability of factors across these sources suggests that Leary's factors may arise not from the structure of human personality but from the structure of our conceptions of people. A cross-cultural study would be an interesting addition to the literature at this point. Concurrent validity No research was found which correlated scores obtained on the Interpersonal Checklist with any other measure(s) of dominance, submission, hostility or affiliation. The concurrent validity of those dimensions, then, remains untested. Reliability The uses made of the Interpersonal Checklist for gathering Level I data present different problems in reliability. When the Checklist is used globally, split-half reliability and test-retest reliability are most relevant. When ratings are made on an interaction-by-interaction basis, interjudge reliability is a major concern. Armstrong (1958) tested the split-half reliability for global ratings. In fact, he checked the reliability of Level II data (selfratings) and Level IV data (ideal self-ratings) as well as that of three Level I ratings for each subject. All 100 of his subjects were male; half were alcoholics and half were "normal." Using the KuderRichardson estimate of reliability, Armstrong obtained very high reliability coefficients ranging from .953 to .876. Interjudge reliability for interaction ratings is far more complex. It is common for studies using the Interpersonal Checklist in

PAGE 78

68 this fashion to report reliability figures. However, as Mueller and Dilling (1969) have pointed out, these figures can come from a wide variety of places. One could report (1) the percentage of scores which agree, across items; (2) coefficients of item agreement; or (3) rankorder correlations based on the percentage of reflexes occurring within a given number of units. Also, any of these methods can be based on (a) reflex agreement, (b) section agreement (octant, quadrant, etc.), or (c) agreement within a given range, for example, scores coming within one reflex of that marked by the criterion judge. Furthermore, any combination of the above could be made far more complex by considering the intensity of each rating. Still another complication is introduced if raters are permitted to ascribe more than one reflex to a given unit of communication. Mueller and Dilling (1969) claimed that their literature review made 70 percent item agreement at the reflex level look feasible. However, Mueller's (1969) own study of therapist-client interaction, published the same year, reported only 49.4 percent item agreement at that level. Mueller obtained 69.1 percent agreement when he allowed ratings to vary within one reflex of each other and 76.1 percent agreement when the permissible range was extended to two reflexes in each direction. The reliability figures actually obtained by Mueller are similar to others found when therapist-client interaction has been studied. Dietzal and Abeles (1975) obtained 82 percent mean agreement for responses classified by quadrant. Crowder (1972) reported 75 percent agreement for quadrant scoring.

PAGE 79

69 Use in This Study In this study, the independent variables are measures of client behavior, of therapist behavior, and of therapist response to client behavior. All were obtained by using the Leary Interpersonal Checklist as a process measure. Both the literature review in Chapter Two and the preceding description of validation studies mention that the Leary Interpersonal Checklist has been used in a variety of ways by different researchers. It is a very flexible instrument, and that flexibility gives the researcher an interesting variety of choices. The choices made in designing this research will be discussed next. To begin with, the data were collected on audiotapes. When presenting the Checklist and its preliminary research, Leary referenced a study suggesting that Checklist ratings are more valid if based on audiotapes rather than typescripts. He added that "sound movies" would be optimal but had not been used to date (Leary, 1957). Twentyfive years later, no change has occurred. Videotapes are much admired as a data source (cf. Mueller and Dilling, 1969), but collecting them presents so many practical difficulties that their use is abjured. In addition, as Hollenbeck (1978) has noted, videotaping is so intrusive that it can be a source of reactive bias. Using the Checklist as a process measure requires rating behavior on an interaction-by-interaction basis. The actual unit of analysis can be defined in different ways. In this study, the basic scoring unit was defined as one person's statement(s) from the time

PAGE 80

70 the other stopped talking until he or she started again. However, more than one unit was declared if the tone of the response changed to an Important degree while the same person was talking. This methodology follows the guidelines outlined by Crowder (1970). Mueller's (1969) experimental procedure in which both an overt and a covert theme were identified proved unreliable. Therefore, the rater was required to choose one reflex to describe each scoring unit. The Interpersonal Checklist offers researchers a variety of options when it comes to calibrating responses. The circumplex distinguishes 16 "interpersonal reflexes," and four levels of intensity are described for each (Appendix A). Theoretically, one could work with 64 categories. In practice, most researchers have grouped the 16 reflexes either into octants or into quadrants. Moreover, most have made no distinctions by level of intensity. In psychotherapy research, the usual practice has been to locate each communication unit at one of the 16 points of the circumplex but then to total scores by quadrant. A person or group of people is described by the percentage of responses located in each quadrant with no consideration given to levels of intensity. This system was used in the four studies which bear most directly on this one, three relating therapist and client behavior to therapy outcome (Crowder, 1970, 1972; Dietzal and Abeles, 1975; Spierling, 1972) and one relating behavior to early termination (Schiller, 1978), In those four studies, however, one problem emerged consistently. Client and, particularly, therapist behavior was sufficiently stereotyped

PAGE 81

71 that little variability was obtained. In all four studies, the majority of client behavior fell in the affiliative-submissive quadrant, while therapists were lodged even more firmly in the affiliativedominant quadrant. This low variability of the behaviors as measured may well be responsible for the fact that Schiller (1978) obtained no differences between early and late terminators on any variable except response complementarity. These results suggest that the categories should be refined. They could be refined by using octant scoring, but that would not help distinguish routine, "in role" behavior from the more extreme variety. Response intensity differentiation— ignored by the referenced psychotherapy researchers— offers a promising solution. "Appreciative" behavior (low intensity rating) seems healthy and appropriate from a client, and it should be useful to distinguish it from higher intensity behavior ("likes to be taken care of" or "will believe anyone"). Likewise, one expects therapists to be "helpful," but when the intensity of affiliative dominance increases to "overprotective" or "spoils people with kindness," it probably is significantly atypical behavior. Therefore, in this study the measures of client behavior and of therapist behavior were constructed as follows. First, each scoring unit was assigned to the category (one of the 16 circumplex points) which best represents it. In addition, it was given an intensity rating of one (low) or two (high). In order to do this, the behavioral categories developed by Freedman et al (1951) and elaborated by Crowder (1970) were sorted

PAGE 82

72 into intensity levels with the help of Leary's global descriptions. Appendix A presents Leary's circumplex and the adjectives he used, by reflex and intensity level. Appendix B presents the behaviorally oriented circumplex constructed by Crowder plus the verbs used in this research, by reflex and intensity level. Quadrant scores, plus intensity ratings, were used to test all hypotheses. For client process scores (Hypotheses 1 and 2), each client's intensity scores within a quadrant were summed, and the four sub-totals which resulted then were converted into percentages of the client's grand total of scores. These percentages are referred to as Weighted Percentages of Interaction (WPIs). Therapist process scores (Hypotheses 3 and 4) were obtained in a similar fashion from therapist statements. They, too, are expressed as WPIs. Of course, separate sets of WPIs were constructed to demonstrate how a therapist interacted with each client he or she interviewed. Therapist complementarity of response measures (Hypotheses 5 and 6) were constructed somewhat differently. To begin with, Leary (1957) postulated that his dimensions describe behavior which is not only interpersonal ly relevant but also self-cuing. Behavior as rated on the affiliation dimension is said to cue like behavior, while behavior rated on the dominance dimension cues its opposite. Complementarity of therapist response, then, means the extent to which a therapist acts on this cuing.

PAGE 83

73 Complementarity scores can be constructed from simple quadrant ratings. Dietzal and Abeles did this (1975), and so did Schiller (1978). Neither reported complementarity on the two dimensions separately, but this separation is easy to make and has been made in this study. Furthermore, provision has been made to include the additional information gained from intensity ratings. Since each statement was rated either "1" or "2" for its intensity, two statements could differ from each other by as much as four points on each dimension. Therefore, a therapist's statement could be rated from to 4 for its complementarity to the dominance-submissiveness expressed by the client and from to 4 for its complementarity to the affiliation-hostility expressed. When a client's statement was composed of two or more scorable units, the last unit scored was used to judge complementarity. When a therapist's statement contained two or more scorable units, the first was used. This means that the units compared were those adjacent in the conversation. This procedure duplicates that used and recommended by Mueller (1969). Two complementarity scores were computed for each therapist-client pair: the therapist's average complementarity of response for dominance and average complementarity of response for affiliation. Tape Rating Procedures. A 15-minute segm.ent of each tape (minutes 16-30) wa s rated. This is similar to, but longer than, the segment used by

PAGE 84

74 Mueller and Dilling (1968) and identical to Spierling's procedure (Spierling, 1972). A continuous segment was chosen, in preference to several short segments such as Schiller (1978) used, to improve the rater's comprehension of tape content. The particular segment was chosen because it avoids the "social" and/or administrative content often found in the opening and closing minutes of therapy sessions. Therapy sessions at the target facility usually last from 45 to 60 minutes. Designating minutes 16-30 as the time sample was an approximation. Since client-therapist interaction was being studied, it would have been pointless to rate incomplete interactions. For the purposes of this study, an interaction unit was defined as a client statement and the therapist's response. Therefore, the rater was instructed to start with the first client statement begun after the 15-minute mark and to continue after the 30-minute mark to the end of a therapist response. All of the tapes were scored by a single rater, the criterion judge. One-third were scored also by the researcher, acting as reliability judge. The reliability tapes were selected randomly but with the limitation that two were drawn from each group of six tapes rated. The criterion judge was not told which tapes would be rerated. Examination of the raters' level of agreement on these tapes provided an ongoing check for rater drift over time. This procedure duplicates that of Crowder (1970) and is in line with Hollenbeck's (1978) observation that continuous monitoring increases rater reliability.

PAGE 85

75 Selection and training of rater A beginning graduate student in counseling psychology, blind to the research hypotheses, was trained and used to rate the tapes. A counseling graduate student was used first because graduate status reduced the likelihood of his knowing any of the (largely ungraduate) subjects and second because his status as a professional-in-training obligates him to conform to American Psychological Association ethical standards regarding confidentiality of client material. As a further safeguard of confidentiality, each tape was identified only by a number assigned sequentially when the researcher obtained the tape from the case therapist. The rater was taught according to the principles outlined by Mueller (Mueller, 1969; Mueller and Dilling, 1969) and explained more fully in codebook format by Crowder (1970). The rater first was oriented to the circumplex. This included defining a scoring unit and clarifying the 16 points on the circumplex as the products of two bipolar dimensions. Next, the rater was taught to rate items. He was told to "empathize with the person exhibiting the behavior, from the position of the target of the behavior" (Crowder, 1970, pp. 43-44, after Freedman et al., 1951). From this point of view, the rater was instructed first to decide whether a behavior was dominant or submissive, next to decide whether it was affiliative or hostile, and finally to locate its specific point on the circumplex. For this investigation a fourth step had to be added: rating the intensity as "1" (low) or "2" (high). The next part of training involved giving the rater examples of difficult items and teaching him to score them. A primary principle

PAGE 86

76 of Mueller's is to score the "emotional intention" of a statement, rather than its overt content, if the two conflict (Mueller and Dilling, 1969). Crowder added the following clarifications: "Context takes precedence over affect . and interpretation does not go beyond the immediate context" (Crowder, 1970, p. 109). Finally, the rater rated practice tapes until a preset criterion of reliability was reached. In this study, Crowder's (1970) achieved level of 75 percent was used as the criterion for acceptable average interrater agreement on quadrant assignment. No test (or later, ongoing reliability check) was considered adequate unless quadrant agreement was above 70 percent. In addition, agreement concerning intensity level was checked for all items where reflex assignment matched. Here, the minimum acceptable level of agreement for any test or reliability check was set at 75 percent, and it was expected that average agreement would exceed 80 percent. Rater training required approximately 30 hours of the experimenter's time and somewhat more time for the trainee. About four hours were required for orientation to the circumplex and scoring aids and for two scoring demonstrations. The rest of the time was spent by the investigator and the trainee in doing practice tape ratings and discussing discrepancies. All training was done using tapes not included in the final study sample. At the completion of training interrater agreement for quadrant placement averaged 73 percent (across six, 5-minute tape segments), and agreement on intensity level averaged 88 percent (across three of the same 5-minute segments). The rater training program is detailed in Appendix C.

PAGE 87

77 No rater retraining was required once formal rating began since all five checks produced agreements above the minimum acceptable levels. However, quadrant agreement figures from the first two checks were borderline. Therefore, the investigator examined those tests, located systematic errors being made by the rater, and clarified the relevant guidelines in an attempt to improve the validity, as well as the reliability, of the ratings. Those guidelines, a supplement to Crowder's (1970) codebook, are included in Appendix C. Rater reliability The five ongoing reliability checks conducted during the course of tape rating produced average agreements which improved from 71 percent quadrant and 90 percent intensity agreements on the first check to 79 percent quadrant, 88 percent intensity and 74 percent quadrant, 92 percent intensity agreements on the last two checks. Each of the five tests checked reliability on two entire tape segments. Overall average reliability on the ten segments checked was 76.3 percent quadrant agreement and 88.5 percent intensity agreement (see Table 1). Each of these figures exceeds its preset standard for an acceptable level of agreement. Evaluation of interrater reliability was taken one step further by computing agreement percentages corrected for chance agreement between raters. Since this was not done in comparable studies, no criteria were available for determining an acceptable level of agreement. Correcting for chance agreement between raters can be based on one of two assumptions. One may assume that the proportion of items

PAGE 88

78 Table 1. Percentages of item agreement, for both quadrant and intensity ratings, both unadjusted and after correction for chance agreement, by test period

PAGE 89

79 a rater assigns to a particular category is a fixed property of that rater or of the coding system. In that case, chance agreement for a category is found by a means such as multiplying the proportion of one rater's codes falling in that category by the proportion of the second rater's codes falling there. This is the approach used by Cohen in creating his Kappa statistic (Cohen, 1960). On the other hand, one may assume that raters approach each item with no preconceptions, no propensity to rate it in a particular fashion apart from what the item itself suggests. In that case, chance agreement is synonymous with random scoring. The reciprocal of the number of categories among which the raters must choose gives the likelihood of two raters agreeing if one or both codes randomly. Using this approach, it is very simple to adjust an obtained agreement figure. If P is the obtained proportion of item agreement and P^ is the proportion of agreement expected from random coding, the corrected agreement figure is (P^ P^)/(l P^) This formula gives a result identical to that of Cohen's (1960) Kappa when ratings are distributed evenly among coding categories. The two formulas produce quite different results, however, if item distribution is markedly uneven. In this research, the rating system had two aspects which needed to be checked for reliability: quadrant assignment and designation of an intensity level. In neither case did the criterion judge have a reason to expect any particular proportion of items to appear. In fact, rater instructions to consider the Leary Interpersonal Checklist dimensions separately and then to code a reflex (rather than a quadrant)

PAGE 90

80 meant that the quadrants were not even presented conceptually to the criterion judge. Therefore, the formula defining chance agreement as that which would be produced by random coding has been used to correct percentages of item agreement. The corrected percentages of item agreement average 68.4 percent for quadrant agreement and 76.9 percent for intensity agreement. The range of the first is 61 to 84 percent, and the range of the second is 67 to 84 percent. Table 1 presents both simple percentages of item agreement and the corrected percentages. In actuality, the adjusted figure for quadrant assignment would have differed little had Cohen's Kappa been used. Quadrant codes were not distributed completely evenly, but the distribution was even enough that Kappa would have subtracted only 2 to 4 more percentage points from most of the interrater agreement scores for quadrants. On the other hand, intensity codes tended to be low. "One" was coded more frequently than "2" on all protocols, and on the reliability protocols as a group "1" was coded six times as often as was "2." For this reason, applying Kappa to intensity coding would have subtracted from 11 to 37 additional percentage points from the corrected interrater agreement figures. The uneven distribution of intensity codes raises questions about the use of this coding system, and these questions will be examined in Chapter Five. However, at this point there seems to be no compelling reason to lower interrater reliability estimates because of this uneven distribution.

PAGE 91

81 Data Analysis Hypothesis testing Hypotheses 1 through 6 were tested using two-tailed t^tests for the significance of differences between means. Investigating potential intervening variables The literature review in Chapter Two discussed a variety of client and therapist factors which have been investigated for their relationship to premature termination. These will be considered here in light of their potential for acting as "nuisance" variables in this study. The two (nonindependent) client factors cited most frequently as influencing rates of premature termination are socioeconomic status and educational level. In this study, it was expected that those factors would be controlled for to a large extent by using a sample composed entirely of university students. In a similar fashion, the variable of race should be largely eliminated. Due to differential utilization rates of the counseling facilities, plus the presence of a predominantly Caucasion student body at the target university, it was expected that the study sample would be self-limited almost entirely to Caucasians. Those assumptions were checked by gathering socioeconomic, educational, and racial data and performing chi-square analyses for these factors by termination status. One remaining client factor is an indentified, potentially significant intervening variable: sex of client. Results have not been consistent, but some studies have found males less inclined than females to terminate prematurely. As it seemed impractical to balance the study for sex of client, a chi-square analysis was performed to

PAGE 92

82 see whether or not women would be found disproportiately among the premature terminators in this study. The one therapist factor which has been found fairly consistently predictive of premature termination rates is amount of experience. Frequently, experience has been confounded with amount of training, so that needs to be considered a possible factor as well. In this study, however, there is no real justification for treating either experience or training as an intervening variable. Generally speaking, therapists at the target facility announce neither their degrees nor their years of experience to potential clients. Some holding doctorates include this information on their door plaques, but not all do. The absence of a doctorate is never indicated. Moreover, the most obvious clue to experience, age, varies little within the group of participating therapists. It was anticipated that both the amount of experience and the amount of training enjoyed by therapists would affect the likelihood of their clients terminating prematurely. Primarily, though, this influence should have operated not as an intervening variable but as a source of variability in client-therapist process. Therefore, controlling for experience/training might have artificially eliminated much variability of the independent variables utilized in Hypothesis 3 through Hypothesis 6. Instead, the expected association between therapist experience and termination status was sought by chi-square analysis. Had it been found, its roots would have been explored via two-way analyses of variance comparing mean levels of the independent

PAGE 93

83 variables by experience level of the therapist as well as by termination status. One other therapist factor needs to be mentioned here: sex of therapist. Some weak support exists for the hypothesis that female therapists have lower rates of premature termination than do male therapists. No explanation has been offered, so one can only speculate as to whether this effect operates through client-therapist process or through a client's internal reactions to the idea of his or her therapist's gender. Due to the size of this study and an imbalance between male and female therapists at the facility studied, it would have been difficult to balance the sample for sex of therapist. Therefore, its relationship to termination status was examined in a post hoc chisquare analysis similar to that performed for sex of client by termination status. Other analyses Exploratory work was done utilizing the concept of response intensity. Two-tailed t^-tests were performed comparing the mean intensity of responses in cases terminated prematurely with those of cases where clients remained in treatment. For both clients and therapists, separate analyses were performed by quadrant. This totaled eight analyses concerning response intensity. Significance levels The .05 level of significance was used in all tests of association as the criterion for rejection of the null hypothesis. However, in the interest of thoroughness, significance levels as low as £ < .25 were reported. Below £ = .25, results were described simply as nonsignificant.

PAGE 94

CHAPTER FOUR RESULTS Introduction Fifty-nine clients and 24 therapists participated in the datagathering phase of this study. This phase continued until there were 15 cases with complete data which clearly could be classified as premature terminations and more than 15 which could be classified as remaining in treatment. Table 2 breaks down the case classification by termination status and by tape usability. The final study sample included all 15 premature terminations in which taping had been successful and 15 cases randomly selected from among those remaining in treatment where taping was adequate. Sixteen therapists had cases in this final sample. Six hypotheses were tested. In each case, termination status was the dependent variable. The independent variables included two measures of client behavior, two measures of therapist behavior, and two measures of the complementarity of therapist behavior to the preceding client behavior. Other analyses were performed to test the predictive strength of client and therapist characteristics which previous research has associated with case termination status. The characteristics tested are sex of client, sex of therapist, experience level of therapist, two elements of client socioeconomic status (years of education 84

PAGE 95

85 Table 2. Number of cases, by termination status and tape usability

PAGE 96

86 completed and parental occupation), and client race. Data were gathered on therapist race as well. Finally, exploratory analysis was done in the area of response intensity. Eight analyses were performed, four using measures of client intensity and four using measures of therapist intensity. Hypothesis Testing Client Behavior and Premature Termination: Hypotheses 1 and 2 Table 3 presents the mean client Weighted Percentages of Interaction (WPIs) in all four Leary Intsrpersonal Checklist quadrants for premature terminators and for remainers. In addition, _t-statistics and significance test results are reported for the two quadrants (affiliative-submissive and hostile-submissive) where significant differences were hypothesized. In both quadrants, mean differences were slight and did not approach significance; hypotheses 1 and 2 were not supported. Therapist Behavior and Premature Termination: Hypotheses 3 and 4 Table 4 presents the mean therapist Weighted Percentages of Interaction for the therapists of premature terminators and for those interacting with remainers. Again, t^-statistics and significance tests results are reported for mean comparisons where significant differences were hypothesized. In both cases, sizable mean differences were found, but variances were large. Moreover, in the hostiledominant quadrant the obtained difference is in the opposite direction of that hypothesized. Hypothesis 3 was supported only at the .25 level of confidence, and hypothesis 4 was not supported at all.

PAGE 97

87 Table 4. Mean therapist WPIs, by quadrant and by case termination status Premature Quadrant Terminators Remainers df jt-test Significance Affiliative-

PAGE 98

88 Therapist-Client Complementarity and Premature Termination: Hypotheses 5 and 6 Table 5 presents mean complementarity scores for therapists' responses to clients. Complementarity means are figured separately on the two Leary Interpersonal Checklist dimensions, affiliation (affiliation-hostility) and dominance (dominance-submission). Neither hypothesis 5 nor hypothesis 6 was supported. One twotailed ^-test found the mean difference in affiliation complementarity to be nonsignificant. The second test found the mean difference in dominance significant at the .06 level of confidence, but that difference reversed the direction hypothesized. Analysis of Selected Client and Therapist Factors Client Factors Sex of client was considered a possible intervening variable in this study. Table 6 presents a chi-square analysis of termination status by sex of client. It can be seem that, for the sample utilized in this study, there is no significant relationship between the two. Tables 7, 8, 9, and 10 present descriptive statistics and chisquare analyses for other client characteristics, all related to socioeconomic status, which have been found predictive of premature termination. It was expected that the use of a university counseling center clientele would eliminate most of the variability here. Specifically, it was anticipated that the sample would be largely Caucasian, educated beyond high school level, and middle to upper class in background as measured by parental occupation. As it turned out, the first two assumptions were accurate, but the third was not.

PAGE 99

89 Table 7. Chi-square analysis of case termination status by race of client Premature P Race^ Terminators Remainers df x Significance Black 2 Caucasian 13 15 3 2.33 n.s. ^Selfidentified Table 8. Chi-square analysis of client education level by case termination status Premature 2 Years of Education Terminators Remainers df x Significance At least 12 but less than 16 13 12 16 or more 2 3 3 .240 n.s.

PAGE 100

90 Table 9. Parental occupation by case termination status Occupation*^ Premature Terminators Remainers Total Farm Semi-skilled non-farm Skilled non-farm Clerical and sales Business and professional Data missing or not codablel^ 12 ^Father's or stepfather's occupation was used, if known; otherwise, mother's or stepmother's occupation was used. Includes four instances of missing data and two descriptions of parental occupation as "retired. Table 10. Chi-square analysis of parental occupation by case termination status Premature 2 Occupation Terminators Remaining df x Significance Semi-skilled, skilled, clerical or sales Business or professional 2.67 n.s,

PAGE 101

91 Table 7 presents a chi-square analysis of termination status by self-identified race of client. Table 8 does the same thing with achieved education level. Table 9 presents clients' parents' occupations, grouped by socioeconomic status and by client termination status. For Table 10, occupations were regrouped into working class/ lower middle class and upper middle class/upper class groups in order to make cell n^'s meet the requirements for chi-square analysis with a test of significance (Peatman, 1963). These two occupational groupings then were compared for client termination status. None of the chi-square tests returned significant results. However, n^'s were very small. Therapist Factors Sex of therapist was considered another possible intervening variable in this study. Table 11 presents a chi-square analysis of termination status by sex of therapist. The difference favored female therapists, but the test of association rated this difference nonsignificant. Experience level was not considered to be an intervening variable in this study. However, it was expected that therapist experience would be related to termination status through affecting process. Table 12 presents a chi-square analysis of termination status by experience level of therapist. Since its results do not approach significance, no further analysis was performed utilizing experience level as a variable. Race of therapist also was unrelated to outcome in this study. Four clients were seen by black therapists, and 26 were seen by

PAGE 102

92 Table 11. Chi-square analysis of therapist experience by case termination status Premature 2 Experience Level Terminators Remainers df x Significance Practicum student 7 9 Pre-doctoral intern 6 3 Staff^ 2 3 5 1.45 n.s. ^Includes Ph.D. psychologists and "ABD" counseling associates. Table 12. Chi-square analysis of case termination status by sex of therapist Premature 2 Sex of Therapist Terminators Remainers df x Significance Kale 5 2 Female 10 13 3 1.68 n.s.

PAGE 103

93 Caucasian therapists. Both groups divided equally between terminating prematurely and remaining in treatment. Response Intensity Response intensity was not scored, let alone analyzed, in any study closely related to this one. Therefore, analysis in this area is completely exploratory. Prior to gathering the data, two sets of comparisons were planned. It was planned that mean response intensity for clients would be compared by quadrant and by termination status, and similar comparisons were planned utilizing therapist intensity scores. Tables 13 and 14 present those tvjo sets of comparisons. No significant differences were found when mean differences were subjected to t^-tests. Examination of the raw data and of Table 13 suggests that high intragroup variability might be responsible for the complete lack of statistical significance. In particular, client responses in the affiliative-submissive quadrant were both numerous and distributed with marked unevenness between the two outcome groups. Therefore, it was decided that the data would be reanalyzed using chi-square analysis to test the significance of the difference between premature terminators and remainers in the total number of intensely affiliative-submissive messages they sent to their therapists. Using this methodology, the difference was found significant at the .25 level of confidence (see Table 15). Reanalysis was limited to the one quadrant because other client quadrants and all therapist quadrants either showed little difference between premature terminators and

PAGE 105

95 Table 15. Chi-square analysis of the frequency of level 2 (high intensity) affiliative-submissive client responses, by case termination status Premature 2 Terminators Remainers df x Significance Response Frequency 26 36 1 1.6 <.25 Table 16. Hostile therapist responses, by case number and termination status Case Termination Hostile-Dominant Hostile-Submissive Number Status'^ Therapist Responses Therapist Responses 3 1 6 2 9 7 2 3 14 3 4 5 14 1 4 5 3 1 n 1 02-02

PAGE 107

97 remainers for intensity level or else contained wery few instances of high-intensity behavior. Other Reanalyses In the same fashion, it seems likely that high intragroup variability was responsible for the low significance levels obtained when comparing the quite considerable numerical differences between therapists of premature terminators and therapists of remainers in the frequency with which they emitted hostile behaviors, both hostiledominant and hostile-submissive (see Table 4). Therefore, those data also were reanalyzed using chi-square to test the significance of differences with which the behaviors occurred in the two outcome groups as wholes. Table 16 presents the frequencies of hostile therapist behavior on a case-by-case basis. Tables 17 and 18 present the chi-square analyses. Both differences were found to be significant at better than the .05 level of confidence. Again, reanalysis was limited to quadrants displaying substantial mean differences between premature terminators and remainers.

PAGE 108

98 Table 17. Chi-square analysis of the frequency of hostile-dominant therapist responses, by case termination status Premature 2 Terminators Remainers df x Significance Response Frequency 75 28 1 21.45 <.001 Table 18. Chi-square analysis of the frequency of hostile-submissive therapist responses, by case termination status Premature 2 Terminators Remainers df x Significance Response Frequency 22 8 1 6.53 <.02

PAGE 109

CHAPTER FIVE DISCUSSION Counseling Process and Premature Termination Client Process and Premature Termination On the basis of the data and methodology of this study, early client process cannot be regarded as a predictor of psychotherapy termination status. The two termination groups did not differ in the relative frequency with which they produced either affiliativesubmissive or hostile-submissive behavior, the two categories where significant differences were hypothesized. Hypotheses 1 and 2, therefore, were not supported. Significant differences between premature terminators and remainers also failed to appear for other behaviors measured but not the subject of hypotheses. The relative frequencies of affiliativedominant and hostile-dominant behavior differed only slightly between the two termination groups. Upon examining the data more closely, it appears that the homogeneity of quadrant assignment percentages across termination groups is an indisputable fact. Neither the fairly small sample size nor high variability within groups accounts (in a statistical sense) for the null results. With the borderline exception of dominantaffiliative behavior— which remainers displayed approximately 25 percent 99

PAGE 110

100 more often than did premature terminators— differences between the two groups are miniscule. The situation is somewhat less clearcut with the comparison of intensity levels. Although t^-tests found no significant differences between group means, remainers were almost one and a half times as likely as premature terminators to have their affiliative-submissive responses scored as high intensity. Affiliative-submissive behavior is the only category where high intensity client responses were frequent enough for a comparison to be meaningful. The usefulness of the intensity data probably was limited by the coding system used. In order to rate behavior for its intensity, the phrases used by other process investigators (Freedman et al 1951; Crowder, 1970) to locate behavior by reflex were sorted into two categories by intensity level. Behaviors related to the two lov/er levels of Leary's descriptive phrases (Appendix A) were coded "1," while those related to the two high intensity levels were coded "2." (Appendix B includes the complete list of coded behaviors.) As it turned out, the "2" code was used relatively infrequently. Future research focusing on intensity level probably ought to use a different coding system. It is more difficult to explain the extreme similarity of the two outcome groups when quadrant scores were compared. Crowder (1970) found significant differences initially in all four quadrants when he compared successful and unsuccessful clients' statements. Differences in the affiliative-submissive and in the hostile-submissive quadrants remained into midtherapy. On the other hand, Schiller (1978) found no

PAGE 111

101 significant differences when he compared early terminators with remainers. Schiller's results agree with those obtained in this study, but they do not explain them. Quite possibly, client interaction along the dimensions of dominance and affiliation is a therapeutic element— like sex of client— where premature terminators prove themselves to be a different group from therapy failures. Perhaps behaviors such as hostilesubmissiveness do not predispose people to leaving therapy prematurely but rather just impede therapeutic progress. Fiester's (1977) work in which both highly satisfied and highly dissatisfied terminator groups emerged suggests another explanation. If this study's sample included both types of premature terminators, their process scores could be canceling each other out in the analysis of means. Replication of this study with the addition of a client satisfaction questionnaire would check this explanation. Therapist Process and Premature Termination Therapist behavior was found to relate to premature termination in this study. However, one research hypothesis (Hypothesis 3) was supported at the .05 level of significance only when an additional data analysis was performed, and the other (Hypothesis 4) was not supported at all. On the basis of Crowder's (1970) research, it was hypothesized that premature terminators would display less hostile-dominant and more hostile-submissive behavior than vyould remainers. In fact, it was the

PAGE 112

102 remainers who displayed less hostile dominance — as well as less hostile submissiveness. Neither difference was significant at the .05 level when the data were analyzed by averaging client response percentages (WPIs). However, when total frequencies were compared directly, both differences were significant at better than the .05 level. The total number of statements made by the therapists of remainers and the therapists of premature terminators were quite similar, so the results mentioned above are not artifacts of a difference in therapist loquacity. Rather it seems that high therapist variability within groups makes it unsafe to predict that differences will appear when particular therapists are compared. However, it does seem safe to predict that overall differences will appear when groups of therapists are compared as wholes. It is interesting that the results on hostile-dominant behavior among therapists contradict Crowder's findings. Again, it is important to consider that Crowder was comparing successful ness of therapy among remainers rather than comparing termination status. In any case, it is Crowder's result that is the more surprising. "Common sense," general therapeutic theory, and Leary's Interpersonal Theory (1957) all suggest that therapist hostility, particularly in early interviews, is likely to prevent client engagement. Specifically, Leary's theory suggests that hostile-dominance on the therapist's part would provoke hostile-submissiveness from the client (Leary, 1957). Unilateral termination is a prototypical ly hostile-submissive client behavior.

PAGE 113

103 Prediction does not imply causality, of course. Moreover, since several premature terminators received little or no therapist hostility in the interview segments rated and one remainer received a relatively large amount, therapist hostility alone is neither necessary nor sufficient to produce premature termination. At this point, it is instructive also to look at the lack of correspondence between the results of Crowder's (1970) study and this one. A broad conclusion, which must be considered tentative since only two studies are available for comparison, is that the general similarity between predictors of remaining in therapy and predictors of therapeutic success may not extend to process variables. Finally, the intensity of therapists' responses in all quadrants was unrelated to case termination status. However, in only one quadrant (affiliative-dominant) were intense responses frequent enough to make a comparison meaningful. As was true for client intensity scores, the coding system's limitations may have prevented the emergence of meaningful differences. Therapi st-Cl i ent Compl ementari ty and Premature Termination On the basis of Leary's Interpersonal Theory (1957) and of Schiller's research (1978), it was hypothesized that remainers would receive more complementary responses from their therapists than would premature terminators. It was hypothesized that significant complementarity differences would be found both when behavior was rated for affiliation and when it was rated for dominance.

PAGE 114

104 Neither hypothesis (Hypotheses 5 and 6) was confirmed. In fact, it was premature terminators who received the greater amount of complementarity on both dimensions. In the case of the dominance dimension, that difference almost reached the .05 level of significance on a two-tailed t^test. The question is, "Why?" The answer seems to be related to the fact that client behavior differed "^ery little between the two termination groups while therapist behavior differed considerably. Crowder (1970) theorized that appropriate, in-role client behavior is that located in the affiliativesubmissive quadrant while affiliative-dominant behavior is appropriate for therapists. By those definitions, most out-of-role behavior in this study was exhibited by clients rather than by therapists. The situation, then, was that some clients in each termination group were prone to behaving "inappropriately," at least in their early interviews. Such inappropriate behavior may be considered an invitation to the therapist to step out of role. Apparently, some therapists were better than others at resisting these invitations, and such therapists also held clients better. Dietzal and Abeles (1975) also found that successful therapists displayed less complementarity to client behavior than did unsuccessful therapists. In their research, the difference was found not in early therapy but in mid-therapy. As was mentioned in Chapter Two, this result can be interpreted as indicating greater independence on the part of the successful therapists. The data from this research support that interpretation.

PAGE 115

105 Nonprocess Factors and Premature Termination Because it was impractical to control for many client and therapist characteristics, data were gathered on the most important to see whether or not they could be acting as intervening variables. Importance was judged by the presence or absence of research suggesting a relationship with premature termination. Client Factors Unlike many of the studies reviewed (Chapter Two), this study found no relationship between sex of client and termination status. In fact, clients of each sex divided themselves as equally as possible between the two termination groups. Of course, with the small ji's involved, this result need not be taken as a serious contradiction of the earlier studies. In previous research, client socioeconomic status has been the variable found most consistently predictive of termination status (Baekeland and Lundwell, 1975). Initially, it was assumed that restriction of the sample to university counseling center clients would almost eliminate variability in socioeconomic status. As it turned out, this assumption held true when the current status of the clients themselves was measured (by education level). However, it was not true when social class background was assessed by analyzing data on parents' occupations. Clients in this study came from families whose occupational status ranged from semi-skilled labor (coal mining; gardening) to high-ranking professional work (management consulting; dentistry). In fact, of the 24 clients for whom data were available, only half came

PAGE 116

106 from business or professional backgrounds. Clearly, researchers should not assume that homogeneity of socioeconomic status is assured by use of a university student population. Moreover, it appeared that socioeconomic status may well have been a factor in the termination status of some clients. Although the chi-square test did not reach the .05 level of significance, students vn'th business or professional backgrounds formed two-thirds of the remainer group and only one-third of the premature terminators; the reverse was true for those of nonprofessional backgrounds. Informal examination of the data disclosed no relationship between social class background and either client or therapist process variables. Therefore, knowledge of clients' social class background apparently adds more information to that supplied by process research concerning the likelihood of premature termination. Client race, a characteristic which often covaries significantly with socioeconomic status (Baekeland and Lundwell, 1975), was not a significant variable in this study. All but two clients were Caucasian. Both black clients terminated prematurely, but the n is far too small for that to be noteworthy. Therapist Factors Only one therapist factor has emerged in the research literature as fairly consistently predictive of termination status: therapist experience level (Chapter Two). In this study, the two were not related. However, there is a good reason for this apparent anomaly.

PAGE 117

107 The counseling center wehre data were collected does not assign clients at random. Senior staff and interns do all intakes, and clients are recommended for a particular experience level of counselor based, in part, on the anticipated difficulty of the case. Therefore, the moste experienced therapists tend to receive the most difficult cases. Such cases probably are prone to terminate prematurely. In some studies, premature termination rates have favored female therapists (Chapter Two). In this study, the same effect appeared as a large percentage difference in client retention rate by sex of therapist. However, that difference was not statistically significant due to the small n. Quite possibly a sample including more male therapists would have produced a statistically significant difference. Therapist factors are of particular interest in this research since most of the obtained process differences were attributable to therapists rather than to clients. However, it appears that factors other than the objectively defined characteristics reviewed are responsible for most of these differences in therapist process. Limitations of This Study and Suggestions for Future Research This study has important limitations in the areas of client and therapist sampling and of instrumentation. Future research would benefit by taking these into account. The client sample suffers from two problems. It is fairly small, so significance tests were failed where sizable effects were present. Moreover, it probably is not representative of its theoretical population, university counseling center clients.

PAGE 118

108 Many intakers at the target facility were reluctant to ask clients to participate in research. Some simply avoided the task altogether, and others asked only their most tractable clients. This made data collection slow, necessitated the use of a small n^, weighted the data pool heavily with remainers, and probably made the final study sample unrepresentative. It also made unattainable luxuries of desirable refinements such as balancing the sample for sex of client. The therapist sample is of satisfactory size, but it too has a composition problem. For a variety of reasons, the participation of senior staff therapists was relatively light. This is a particularly unfortunate limitation because it renders questionable the generalization of this study's results to cases seen by thoroughly experienced therapists. One improvement in future research would be to develop a client recruitment procedure which did not depend on intake counselors. Furthermore, an attempt should be made to insure the participation of a representative cross section of both clients and therapists. Two types of instrumentation problems appeared as this study progressed. The more circumscribed one has been mentioned previously: response intensity was rated "low" most of the time. This meant that the instrument was unlikely to differentiate among clients or therapists at a statistically significant level. It also meant that the Weighted Percentages of Interaction differed little from simple quadrant percentages. This problem invites a research project aimed at developing criteria for intensity in therapeutic interactions along the dimensions

PAGE 119

109 of dominance and affiliation. The guidelines offered by phrases descriptive of intensity at the global level of personality are not adequate. A more serious problem is the lack of standardized criteria for the training of raters. This problem, too, is particular to the use of the Leary Interpersonal Checklist as a process measure rather than as a globally descriptive checKlist. In this study, as in others, the criterion judge and the reliability judge achieved a satisfactory level of interrater reliability. However, at present there is no way of knowing to what extent the criterion judge in this study and the one in Crowder's (1970) study (for example) would agree with each other were they to rate the same tape. This is not so much a problem of reliability as it is of rating validity. Ideally, the next research project utilizing the Leary Interpersonal Checklist as a process measure would be aimed at developing training tapes. Their coding should represent the pooled judgments of current experts in the field. Additional tapes should be developed to provide ongoing checks for the validity, as well as the reliability, of judges' ratings in future research projects. With such a training package available, researchers anywhere could do process research, confident that their results would compare meaningfully with others'. At that point, the Leary Interpersonal Checklist could begin to make major contributions to the understanding of psychotherapeutic process.

PAGE 120

no General Conclusions The broad problem to which this research addressed itself was the prediction and explanation of premature termination from psychotherapy. The results of this research underscore the complexity of the phenomenon being studied. At this point it seems clear that premature termination has not one source but several — possibly many. Research cited in a previous section of this manuscript has established that a variety of client factors are liabilities. None determines termination status and none causes it directly, however. An example from this study was the apparent effect of social class origin. Presumably, something connected with social class influences some clients' expectations, perceptions, and/or behavior in a way that affects termination decisions. Similarly, some therapist characteristics appear to affect termination some of the time. Several studies have found therapist experience level to have this effect, and several others found it for sex of therapist. Finally, process variables appear to affect terminations with some of the people some of the time. In this study, clients were more likely to terminate prematurely if their therapists engaged in hostile, out-of-role behavior. However, although client hostility occurred far more frequently than did therapist hostility, its appearance was unrelated to termination status. Moreover, some clients remained in spite of therapist hostility and others left without receiving any.

PAGE 121

Ill Apparently, manipulation of a variety of factors could influence a facility's overall rate of premature termination. Understanding the decision to terminate, however, probably has to be done on a case-bycase basis.

PAGE 122

APPENDICES

PAGE 123

APPENDIX A THE LEARY INTERPERSONAL CHECKLIST The Circumplex Hostiledominant quadrant Affiliativedominant quadrant Hostilesubmissive quadrant Affiliativesubmissive quadrant Source: Adapted from Leary (1957, p. 135) 113

PAGE 124

114 Leary's Phrases for Global Personality Description at

PAGE 125

115 Quadrant and Reflex

PAGE 126

116 Quadrant and

PAGE 127

APPENDIX B INTERPERSONAL PROCESS DIAGNOSIS The Circumplex Hostile Affiliativedominant drant Source: Adapted from Freedman et al (1951) and Crowder (1970) 117

PAGE 128

118 Verbs for Process Ratings at Two Levels of Intensity by Quadrant and Reflex

PAGE 129

119 Quadrant and Reflex

PAGE 130

APPENDIX C RATER TRAINING PROGRAM AND SUPPLEMENTARY CODING AID The rater was a thirty-five year old Caucasian male who had just completed his first year as a doctoral student in the counseling psychology program at the University of Florida. Prior to training he had two years of experience in mental health work and was familiar with Sullivan's Interpersonal Theory of psychotherapy but was unacquainted with the Leary Interpersonal Checklist, The initial session with the rater involved orienting him to the circumplex and familiarizing him with the general task of coding. He was informed that all tapes were from first post-intake counseling sessions, but no other identifying information was given to him. At the close of this session, the trainee was given a copy of the circumplex, Crowder's scoring codebook (Crowder, 1970), and a coding summary for study at home. The second training session began with the investigator demonstrating some tape coding. At this point, training focused on quadrant and reflex assignment, ignoring intensity levels. Then the investigator and the trainee spent two hours jointly rating a practice tape, discussing each rating. The following four sessions alternated independent rating of tapes and replaying them together for discussion. Tape rating sessions 120

PAGE 131

121 ran between two and three hours each, while replaying sessions ran three or four hours. Finally, the trainee was given a test composed of five-minute sections from three different tapes. The investigator rated the same sections. Average quadrant agreement was 73 percent. The next stage of training involved teaching the trainee to distinguish between two levels of intensity per reflex. A new coding aid (similar to the list of verbs in Appendix B) was substituted for the one first given to him. The investigator and the trainee used one session to rate tapes together. Then the trainee was given a second test, constructed similarly to the first but using segments from three different tapes. Interrater agreement on quadrant assignment again reached 73 percent, and agreement on intensity level averaged 88 percent. At this point, the rater began to rate study tapes. Tapes were assigned a coding order entirely at random. After this assignment was made, two of each group of six were chosen, also at random, to be reliability checks. When the rater had completed his ratings on six tapes and the investigator had completed her ratings on the two used to check reliability, interrater agreement percentages were computed. This procedure was followed for each group of tapes. Table 1 in Chapter Three presents the results of those ongoing reliability checks. No reliability test was failed. However, quadrant agreement on the first two tests vjas marginal. The investigator examined the test protocols, isolated some problem areas, and developed clarifying guidelines to supplement those given by Crowder (1970). As Table 1 shows.

PAGE 132

122 interrater reliability improved. It was the impression of both investigator and rater that the validity of the ratings increased as well. The supplementary guidelines follow. Codes Confused Guidelines "I" vs. "L" "I" should be scored if there is clear absence of agreement in the voice and there is reason to suspect disagreement underneath. If the person sounds as though he or she agrees, "L" always is used. "B" vs. "L" Score "B" only if the client is not responsive to the counselor or clearly is intellectualizing (preferably with a hint of superiority). Otherwise, score "L." "B" vs. "P" Score "B" only if the therapist clearly is critical or if he or she wanders off into obfuscating complexities. "C" vs. "G" This is a question of skepticism versus flat disagreement. Be literal unless the tone clearly contradicts the content. "H" vs. "L" When the client is being responsive to the counselor, score "H" only if he or she is apologetic. Self-criticism that is part of the process of selfexploration and for which no apology is made is scored "L." Interruptions 1. Ignore the fact that an interruption occurred unless that seemed to be the intent. 2. If someone interrupts out of defensiveness (implying, "I can't stand this any longer; please stop") with comments such as, "Yes, I know" and a submissive tone, score "G." 3. If someone interrupts to disagree or express skepticism, score "C." Also score "C" if someone interrupts to return to a preferred topic. 4. If the therapist interrupts to get the client more on target, this would be "A" or "N" unless the therapist shows hostility ("D" or "E").

PAGE 133

123 5. If either client or therapist interrupts to correct a misunderstanding, this is scored "P"— again unless there is clear hostility ("D" or "E"). If the correction is elaborated on, a second rating probably will be added.

PAGE 134

I— I 1—1 Q. Q X -zi o to Q. Q. <: o o o ^ •rfO 1— ro QE

PAGE 135

125 rt3

PAGE 136

126 OJ I— rQJ

PAGE 137

REFERENCES Andrews, S. B. The effect of sex of therapist and sex of client on termination from psychotherapy (Doctoral dissertation. University of Illinois, 1975). Dissertation Abstracts International, 1976, 36, 4143B-4144B^ (University Microfilms No. 76-2276) Armstrong, R. G. The Leary Interpersonal Checklist: A reliability study. Journal of Clinical Psychology 1958, 14, 393-394, Baekeland, F., & Lundwell L. Dropping out of treatment: A critical review. Psychological Bulletin 1975, 82, 738-783. Barr, P. M. The effects of certain variables upon affiliation in psychotherapy (Doctoral dissertation, Adelphi University, 1977). Dissertation Abstracts International 1977, 38, 1868-B. (University Microfilms No. 77-22,563) Beutler, P. M. The Interpersonal Checklist. In Buros, 0. K. (Ed.), The sixth mental measurements yearbook Highland Park, N.J.: The Gryphon Press, 1965. Bierenbaum, H., Nichols, M. P., & Schwartz, A. J. Effects of varying session length and frequency in brief emotive psychotherapy. Journal of Consulting and Clinical Psychology 1976, 44, 790798. Borgatta, E. F., Cottrell, L. S., Jr., & Mann, J. M. The spectrum of individual interaction characteristics: An interdimensional analysis. Psychological Reports 1958, 4, 279, 319. Briar, S., & Bieri, J. A factor analytic and trait inference study of the Leary Interpersonal Checklist. Journal of Clinical Psychology 1963, 19, 193-198. Butler, P. Continuance in psychotherapy: Working class clients (Doctoral dissertation. University of Missouri, 1977). Dissertation Abstracts International, 1978, 39, 972B. (University Microfilms N. 78-14,106) Cabral, R. J., Best, J., & Paton, A. Patients' and observers' assissments of process and outcome in group therapy: A follow-up study. American Journal of Psychiatry 1975, 132 1052-1054. 127

PAGE 138

128 Caligor, J. A. Perceptions of the group therapist and the dropout from the group (Doctoral dissertation, Adelphi University, 1975). Dissertation Abstracts International, 1976, 2i> 3591B. (University Microfilms No. 76-1416) Carson, R. C. Interaction concepts of personality Chicago: Aldine Publishing Co., 1969. Carter, L. F. Evaluating the performance of individuals as members of small groups. Personnel Psychology 1954, 1_, 477-484. Cohen, J. A co-efficient of agreement for nominal scales. Educational and Psychological Measurement 1950, 20, 37-46, Crowder, J, E. Transference, transference dissipation, and identification in successful versus unsuccessful psychotherapy (Doctoral dissertation, Michigan State University, 1970). Dissertation Abstracts International 1971, 31, 6894B. (University Microfilms No. 71-11,812) Crowder, J. E. Relationship between therapist and client interpersonal behaviors and psychotherapy outcome. Journal of Counseling Psychology 1972, 19, 68-75. Dietzal, C. S., & Abeles, N. Client-therapist complementarity and therapeutic outcome. Journal of Counseling Psychology 1975, 22, 264-272. Eysenck, H. J. The effects of psychotherapy: An evaluation. Journal of Consulting Psychology 1952, 16, 319-324. Fiester, A. R. Clients' perceptions of therapists with high attrition rates. Journal of Consulting and Clinical Psychology 1977, 45, 954-955. Fiester, A. R., & Rudestam, K. E. A multivariate analysis of the early dropout process. Journal of Consulting and Clinical Psychology 1975, 43, 528-535. Foa, U. G. Convergences in the analysis of the structure of interpersonal behavior. Psychological Revie w, 1961, 68, 341-353. Freedman, M. B. Leary, T. F., Ossorio, A. G., & Coffey, H. S. The interpersonal dimension of personality. Journal of Personality 1951. 20, 143-161. Gladstein, G. A. Empathy and counseling outcome: An empirical and conceptual review. Counseling Psychologist 1977, 6^:4, 70-79.

PAGE 139

129 Godbole, A., & Verinis, J. S. Brief psychotherapy in the treatment of emotional disorders in physically ill geriatric patients. Gerontologist 1974, J4, 143-148. Goldenholz, N. The effect of the sex of therapist-client dyad upon outcome of psychotherapy (Doctoral dissertation, Kent State University, 1975). Dissertation Abstracts International, 1976, 4687B-4688B. (University Microfilms No. 76-4928) Goldstein, A. A. The client-social worker relationship: Transference behavior and interpretation (Doctoral dissertation, Michigan State University, 1975). Dissertations Abstracts International, 1975, 36, 1816A-1817A. (University Microfilms No. 75-20841) Goldstein, A. P. Therapist and client expectation of personality change in psychotherapy. Journal of Counseling Psychology 1960, 7, 180-184. Gomes-Schwartz, B. Effective ingredients in psychotherapy: Prediction of outcome from process variables. Journal of Consulting and Clinical Psychology 1978, 46, 1023-1035. Gomes-Schwartz, B., Hadley, S., & Strupp, H. H. Individual psychotherapy and behavior therapy. Annual Review of Psychology 1978, 29, 435-471. Heilbrun, A. B. Client personality patterns, counselor dominance, and duration of counseling. Psychological Reports 1961a, 9^, 15-25. Heilbrun, A. B. Male and female personality correlates of early termination in counseling. Journal of Counseling Psychology 1961b, 8, 31-36. Heilbrun, A. B. History and self-disclosure in females and early defection from psychotherapy. Journal of Counseling Psychology 1973, 20, 250-257. Heilbrun. A. B. Interviewer style, client satisfaction, and premature termination following the initial counseling contact. Journal of Counseling Psychology 1974, 2]_, 346-350. Hochstadt, N. J. Patient perception in a community mental health setting: A comparison between professional and paraprofessional workers (Doctoral dissertation, De Paul University, 1975). Dissertation Abstracts International, 1975, 36, 1436B. (University Microfilms No. 75-19,892) Hollenbeck, A. R. Problems of reliability in observational research. In Sackett, G. P. (Ed.), Observing behavior. Volume II: Data collection and analysis methods Baltimore: University Park Press, 1978.

PAGE 140

130 HoUiday, P. B. Effects of preparation for therapy on client expectations and participation (Doctoral dissertation. University of Georgia, 1978). Dissertation Abstracts International, 1979, 39, 3517B. (University Microfilms No. 79-1646) Horenstein, D. The effects of confirmation or disconfirmation of client expectations upon subsequent psychotherapy (Doctoral dissertation, University of Kansas, 1973). Dissertation Abstracts International 1974, 34, 6 21 IB. (University Microfilms No. 74-12,575) Jachim, D. P. The social history questionnaire as related to length of stay in psychotherapy. Journal of Clinical Psychology 1974, 30, 446-449. Koran, L. M., & Costell, R. M. Early termination from group psychotherapy. International Journal of Group Psychotherapy 1973, 23, 346-359. LaForge, R., & Suczek, R. F. The interpersonal dimension of personality: III. An interpersonal check list. Journal of Personal ity 1955, 24, 94-112. Lambert, M. J., de Julio, S. S., & Stein, D. M. Therapist interpersonal skills: Process, outcome, methodological considerations, and recommendations for future research. Psychological Bulletin 1978, 85, 467-489. Landfield, A. Personal construct systems in psychotherapy Chicago: Rand McNally & Company, 1971. Landis, C. A statistical evaluation of psychotherapeutic methods. In L. E. Hinsie (Ed.), Concepts and problems of psychotherapy New York: Columbia University Press, 1937. Lange, D. L. Validation of the orthogonal dimensions underlying the ICL and the octant constellations assumed to be their measure. Journal of Projective Techniques and Personality Assessment 1970, 34, 519-527. Leary, T. Interpersonal diagnosis of personality New York: The Ronald Press Co., 1957. Leary, T., & Coffey, H. S. Interpersonal diagnosis: Some problems of methodology and validation. Journal of Abnormal and Social Psychology 1955, 50, 110-124. Luborsky, L., Chandler, M., Auerbach, A. H., Cohn, J., & Bachrach, H. M. Factors influencing the outcome of psychotherapy: A review of quantitative research. Psychological Bulletin 1971, 75, 145-185.

PAGE 141

131 McCormick, C. C. A study of a procedure for the scaling and calibration of personality test items applicable to the circumplex model (Doctoral dissertation, Loyola University of Chicago, 1977). Dissertation Abstracts Inte rnational, 1977, 38, 2068A. (University Microfilms No. 77-22, 34T) McGrogan, D. T. Premature termination from psychotherapy— A continuing problem (Doctoral dissertation. University of Illinois, 1976). Dissertation Abstracts International 1976, 37, 2748A. (University Microfilms No. 76-24,213) Ma Ian, D. H. Toward the validation of dynamic psychotherapy: A replication New York: Plenum Press, 1976. Martin, D. E. Some effects of a pre-therapy procedure on the outcome of outpatient, individual psychotherapy (Doctoral dissertation. The University of Tulsa, 1975). Dissertation Abstr acts International 1975, 36, 1444B-1445B. (University Microfilms No. 75-19,920) Meltzoff, J., & Kornreich, M. Research in psychotherapy New York: Atherton Press, 1970. Mendelsohn, G. A. Effects of client personality and client-counselor similarity on the duration of counseling: A replication and extension. Journal of Counseling Psychology 1966, 1^, 228234. Mueller, W. J. Patterns of behavior and their reciprocal impact in the family in psychotherapy. Journal of Counseling Psychology 1969, J_6, monography supplement to no. 2. Mueller, W. J,, & Dilling, C. A. Therapist-client interview behavior and personality characteristics of therapists. Journal of Projective Techniques and Personality Assessment 1968, 32, 281-288. Mueller, W. J., & Dilling, C. A. Studying interpersonal themes in psychotherapy research. Journal of Counseling Psychology 1969, 16, 50-58. Nacev, V. Dependency and ego strength as indicators of patients' attendance in psychotherapy (Doctoral dissertation. United States International University, 1977). Dissertation Abstracts International 1979, 39, 5572B. (University Microfilms No. 799593) Norkus, A. G. Sex of therapist as a variable in short-term therapy with female college students (Doctoral dissertation, Boston University, 1976). Dissertation Abstracts International, 1976, 36, 6361B-6362B. (University Microfilms No. 76-13,556)

PAGE 142

132 Padfield, M. The comparative effects of two counseling approaches on the intensity of depression among rural women of low socioeconomic status. Journal of Counseling Psychology 1976, 23, 209-214. Peatman, J. G. Introduction to applied statistics New York: Harper & Row, Publishers, 1963. Raschella, 6. F. An evaluation of the effect of goal congruence between client and therapist on premature client dropout from therapy (Doctoral dissertation. University of Pittsburgh, 1975). Dissertation Abstracts International 1976, 36, 8301A. (Univeristy Microfilms No. 76-14,161) Rosen, D. A. Global attraction as a predictor of terminating and remaining in counseling (Doctoral dissertation, Iowa State University, 1978). Dissertation Abstracts International 1979, 39, 4050B. (University Microfilms No. 79-4016} Ross, M. B. Discussion of similarity of client and therapist. Psycho logical Reports 1977, 40, 699-704. Safer, J. Effects of sex of patient and therapist on length of therapy. International Mental Health Research Newsletter, 1975, 17_(2), 12-13. ~~~~ Saltzman, C, Luetgert, M. J., Roth, C. H., Creaser, J., & Howard, L. Formation of a therapeutic relationship: Experiences during the initial phase of psychotherapy as predictors of treatment duration and outcome. Journal of Consulting and Clinical Psychology 1976, 44, 546-555. Sandler, W. J. Patient-therapist dissimilarity of role expectations related to premature termination of psychotherapy with studenttherapists (Doctoral dissertation, The City University of New York, 1975). Dissertation Abstracts International 1975, 6111B-6112B. (University Microfilms No. 75-12,691) Sasseen, G. A. The relationship between the disconfirmation of client preferences and expectations to early premature termination (Doctoral dissertation, Rutgers University, 1976). Dissertation Abstracts International 1976, 37_, 815A-816A. (University Microfilms No. 76-17,318) Schaefer, E. S. A circumplex model for maternal behavior. Journal of Abnormal and Social Psychology 1959, 59, 226-235. Schauble, P. G., & Pierce, R. M. Client in-therapy behavior: A therapist guide to proaress. Psychotherapy: Theory, Research, and Practice, 1974, 11, 229-234.

PAGE 143

133 Schiller, L. S. The intake interview in psychotherapy: Clienttherapist complementarity and role behavior (Doctoral dissertation, Michigan State University, 1978). Dissertation Abstracts International 1978, 39, 1500B. (University Microfilms No. 7815,164) Shannon, J., & Guerney, B. Interpersonal effects of interpersonal behavior. Journal of Personality and Social Psychology 1973, 26, 142-150. Sloane, R. B., Staples, F. R., Cristol, A. H., Yorkston, N. J., & Whipple, K. Psychotherapy vs. behavior therapy Cambridge: Harvard University Press, 1975. Spierling, T. W. Success or failure in psychotherapy: The effects of comparable client-therapist and client-significant other interaction patterns upon the process and outcome of psychotherapy (Doctoral dissertation, Michigan State University, 1972). Dissertation Abstracts International 1973, 33, 4856A. (University Microfilms No. 73-5493) Stern, S. L., Moore, S. F. & Gross, S. J. Confounding of personality and social class characteristics in research on premature termination. Journal of Consulting and Clinical Psychology 1975, 43, 341-344. Truax, C. S. Therapist empathy, warmth, and genuineness and patient personality change in group psychotherapy: A comparison between interaction unit measures, time sample measures, and patient perception measures. Journal of Clinical Psychology 1966, 22, 225-229. Truax, C. B., & Carkuff, R. R. Toward effective counseling and psychotherapy Chicago: Aldine Press, 1967. Truckenmiller, J. L. Equivalence of personality structure in Leary's interpersonal system of diagnosis (Doctoral dissertation. West Virginia University, 1971). Dissertation Abstracts International 1972, 32, 6663B-6664B;^ (University Microfilms No. 72-14,085) Vail, A. F. Dropout from psychotherapy as related to patient-therapist discrepancies, therapist characteristics, and interaction in race and sex (Doctoral dissertation, Fordham University, 1974). Dissertation Abstracts International, 1974, 35, 2452B. (University Microfilms No. 74-25,087). Winder, C. L., Ahmad, F. Z. Bandura, A., & Rau, L. C. Dependency of patients' psychotherapists' responses and aspects of psychotherapy. Journal of Consulting Psychology 1962, 26, 129-134.

PAGE 144

134 Zimmerman, R. L., & Vestre, N. Comparisons of psychiatric patient selfevaluations with evaluations made by members of their immediate family. Catalog of Selected Documents in Psychology 1975, 5, 314.

PAGE 145

BIOGRAPHICAL SKETCH Suzanne Molnar was born March 29, 1945, in Washington, D.C. Her parents were New Englanders transplanted by World War II. Suzanne graduated from Antioch College in 1966 and studied sociology at the University of Southern California as an Oakley Fellow from 1966 to 1968. Then she taught public school for five years in Cypress, California. In 1974, Suzanne entered the University of California at Santa Barbara, studying client-centered, behavioral and Gestalt techniques of psychotherapy. She received her M.A, in counseling psychology and entered the University of Florida as a doctoral student in 1976. At the University of Florida, areas of study included Sullivanian theory, marital counseling, sex therapy, and the clinical use of hypnosis. Also, she first assisted with the teaching of undergraduates and then taught Introduction to Psychology for a year. Suzanne interned at the Psychological and Vocational Counseling Center of the University of Florida, developing specialties in the areas of individual counseling, couples' counseling, assertiveness training, and vocational development. During her internship year, she married Terry Molnar. Since completing her internship in 1979, Suzanne has remained at the Psychological and Vocational Counseling Center as a counseling 135

PAGE 146

136 associate. She has coordinated the Women's Walk-In Service, supervised trainees, conducted a variety of groups, and served as Groups Coordinator and as a member of the Center's research committee. During 1981, the year of her daughter's birth, she was absent from the Counseling Center but conducted a part-time private practice and taught assertiveness training classes at Santa Fe Community College in Gainesville, Florida, Following her graduation, Suzanne expects to continue to work in the areas of counseling practice, training, and research.

PAGE 147

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. James Archer, Jr. ~^ ''Associate Professor of Counselor Education I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Franz R^pting Professor of Psy

PAGE 148

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. William J. Fro Assistant Pro I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Carojyn M. ^-i^iick^r' Assistant Professor of Psychology This dissertation was submitted to the Graduate Faculty of the Department of Psychology in the College of Liberal Arts and Sciences and to the Graduate Council, and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. August 1982 Dean for Graduate Studies and Research

PAGE 149

UNIVERSITY OF FLORIDA 3 1262 08553 5812


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EQ3X5CEJ6_MEELDD INGEST_TIME 2017-11-15T17:39:57Z PACKAGE UF00102811_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EDCGAT0AP_LXG0OK INGEST_TIME 2017-07-11T21:40:07Z PACKAGE UF00102811_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES