Journal of Counseling Psychology Copyright 2006 by the American Psychological Association

2006, Vol. 53, No. 1, 26 –35 0022-0167/06/$12.00 DOI: 10.1037/0022-0167.53.1.26

Concept Mapping the Client’s Perspective on Counseling

Alliance Formation
Robinder P. Bedi
University of British Columbia

The purpose of the present study was to identify, categorize, and model clients’ understanding of early
counseling alliance formation factors. Forty participants who had received counseling services were
interviewed and asked about what observable behaviors and verbalizations they thought had helped
establish the alliance with their counselor. Alliance formation factors were recorded on index cards, and
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

31 participants returned and sorted these statements into self-defined, conceptually homogeneous
This document is copyrighted by the American Psychological Association or one of its allied publishers.

categories. Multivariate concept-mapping statistical techniques were used to compute the “average” sort
across the participants. Seventy-four client-identified common factors were elicited and reliably orga-
nized into 11 categories: setting, presentation & body language, nonverbal gestures, emotional support
& care, honesty, validation, guidance & challenging, education, referrals & recommended materials,
client’s personal responsibility, and session administration.

Keywords: working alliance, therapeutic alliance, counseling alliance, concept mapping, client’s
perspective

Despite the assortment of theories about the counseling alliance ers still know relatively little about how clients, themselves, un-
(e.g., Bordin, 1979; Gelso & Carter, 1985; Luborsky, 1976), there derstand and psychologically experience the counseling alliance in
is a conspicuous absence of theorizing and empirical research that comparison to how practitioners and researchers interpret the cli-
represents clients’ subjective understandings. Present theories of ent’s perspective on the alliance.
the alliance and the measures derived from them are biased by This deficiency in the literature is problematic for a number of
practitioner and investigator understandings and frequently limited reasons. For example, (a) clients and counselors appear to under-
to variables that they hypothesize are important (a point echoed by stand the quality and strength of the alliance differently (e.g.,
Bachelor, 1995; Mohr & Woodhouse, 2001). Although it is true Bachelor, 1995; Cecero, Fenton, Frankforter, Nich, & Carrol,
that the client’s perspective is indirectly included in some mea- 2001; Horvath & Marx, 1990; Tichenor & Hill, 1989); (b) clients’
sures of the alliance, as clients are sometimes consulted in devel- judgments of the strength of the alliance tend to be stronger
oping the initial item pool, this is often not the case (see the predictors of a positive counseling outcome than counselors’ judg-
development of two of the most popular alliance instruments: the ments (see meta-analyses of Horvath & Bedi, 2002; Horvath &
Working Alliance Inventory [Horvath & Greenberg, 1989] and the Symonds, 1991); and (c) although counselors tend to construe the
Helping Alliance Scales [Luborsky, 2000]). Nevertheless, using alliance differently according to their particular theoretical orien-
indirect measures of client experience constrains researchers’ un- tation, leading alliance researchers are proposing that clients’
derstanding to concepts consistent with counselors’ or researchers’
understandings are more homogeneous and relatively independent
conceptual frameworks (cf. comments made by Elliott & James,
of the counselor’s theoretical orientation (e.g., Horvath, 2001;
1989; Paulson, Truscott, & Stuart, 1999). Consequently, research-
Horvath & Bedi, 2002; Horvath & Luborsky, 1993). These differ-
ences between counselor and client accounts of the alliance can be
so marked that Horvath and Bedi (2002) were compelled to state
Robinder P. Bedi, Department of Educational and Counselling Psychol- that “each of these perspectives throws light on a qualitatively
ogy and Special Education, University of British Columbia, Vancouver,
different aspect of the phenomena. . .” (p. 41). Therefore, it ap-
British Columbia, Canada.
Portions of this article were presented at the annual convention of the
pears that clients understand the alliance somewhat differently
Society for Psychotherapy Research, Rome, Italy, June 2004. This article from counselors and that further elucidation of the client’s per-
is based on the doctoral dissertation of Robinder P. Bedi, conducted under spective holds the potential for helping promote better counseling
the primary supervision of Beth E. Haverkamp. I thank the members of the outcome.
alliance research team (Meris Williams, Michael Davis, Dave Alexander, There is a dearth of empirical research in which the client’s
Laura Paetkau, Shauna Marlow, Naomi Kolinsky, Tamara Hilscher, Beth subjective perspective on the alliance has been investigated. The
E. Haverkamp, and Bruno Zumbo) for their dedication and assistance with results of these few studies are particularly enlightening because
this project. Appreciation must also be given to the 40 participants who
they advance unique client-driven conceptualizations of the alli-
generously shared their time and expertise.
Correspondence concerning this article should be addressed to Robinder ance that are not readily commensurable with existing theories. For
P. Bedi, who is now at the Counseling Center for Human Development, example, the work of Bachelor (1995) and Mohr and Woodhouse
University of South Florida, SVC 2124, 4202 East Fowler Avenue, Tampa, (2001) converge in noteworthy ways: (a) Clients’ understandings
FL 33620. E-mail: [email protected] of the alliance can be broken down into types that do not corre-

26
 CONCEPT MAPPING CLIENT’S PERSPECTIVE 27

spond to the existing literature (i.e., nurturant, insight oriented, or 5 out of 10 and if they had participated in at least three counseling sessions.
collaborative; professional or personal), (b) clients identify vari- In regard to where participants were recruited from, 37.5% received
ables that are not well accounted for by existing alliance theories services at a university counseling center, 27.5% at a private practitioner’s
(e.g., counselor friendliness, humor, advice, client self- office, 22.5% at a community mental health clinic, and 12.5% from another
setting (e.g., church). The only measured statistically significant demo-
understanding), and (c) factors considered essential by the majority
graphic difference (␣ ⫽ .05) between participants receiving services from
of alliance researchers (e.g., collaboration, mutuality) appear to be
these four locations was that those receiving services in another setting
less important in the minds of clients. Although these two studies spent significantly less time receiving counseling than those receiving
shed light on clients’ subjective understandings of alliance types, services in a community clinic. There were no significant differences
the factors that clients subjectively consider most helpful in alli- across gender, age, ethnicity, number of sessions, and status of currently
ance formation have yet to be sufficiently investigated. receiving counseling or not.
Two recent studies built on this new avenue of research. Bedi, The participants were primarily women (77.5%) and ranged in age from
Davis, and Arvay (2005) and Bedi, Davis, and Williams (2005) 18 to 62 years (M ⫽ 35.8, SD ⫽ 13.3). The sample was primarily White
investigated clients’ perspectives on the factors that play a critical (70%) or of Asian descent (15%) and participated in a median of 15.5
role in the formation of a quality alliance early in counseling. Bedi, sessions (M ⫽ 58.1, SD ⫽ 127.6). The majority of participants (80%) were
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

currently receiving counseling when they were interviewed, and 70% had
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Davis, and Arvay (2005) found that client-identified early alliance

last attended a session within 30 days. There were no statistically signifi-
formation factors could be organized into eight mutually exclusive
cant differences between those currently receiving counseling services and
categories (general counseling skills, expression of positive affect those who were no longer receiving services.
and sentiment, tracking the counseling process, counseling envi- To further characterize the sample, participants were provided defini-
ronment, punctuality and use of time, going beyond normative tions of alliance types on the basis of previous phenomenological research
expectations, personal attributes of the counselor, and positive first and asked to endorse their ideal types. Using Bachelor’s (1995) typology,
encounters). However, the credibility of this categorization scheme 50% of the sample preferred a collaborative counselor, 35% preferred an
is questionable because of a reliance on only 9 participants. A insight-oriented counselor, and 7.5% preferred a nurturant counselor (7.5%
larger sample size replication (Bedi, Davis, & Williams, 2005) provided no response). Using Mohr and Woodhouse’s (2001) typology,
yielded a comprehensive 25-category system for organizing close 55% preferred a more personal counselor, and 42.5% endorsed a more
professional counselor (2.5% provided no response).
to 400 client-identified alliance formation factors.1 Although both
Thirty-one participants (77.5% of the original sample) returned between
studies appeared to reliably elicit client-identified variables that
about 1 month and 4 months for the sorting and rating procedures (to be
are important for early alliance formation, they still relied on discussed in the Procedures section). There were also no statistically
investigator comprehension to categorize these factors (as did significant differences between those 31 individuals who returned for the
Bachelor, 1995) and to name these factors (as did both Bachelor, sorting and rating tasks and those 9 participants who did not. Of the 31
1995, and Mohr & Woodhouse, 2001). Allowing participants to participants who returned for sorting procedures, 87.1% provided feedback
use their own meaning systems in creating and naming categories on the final results of the study.
ensures that the determined categorical system is a better reflection
of their conceptual structures rather than the conceptual structure
of the investigator. Because clients’ meaning systems in organiz- Interviewers and Research Assistants
ing and naming the categories were not used in any of these
Five interviewers (Robinder P. Bedi was not one of the interviewers)
studies, the extent to which the results truly represent the client’s interviewed 8 participants each. The interviewers (four women, and one
subjective perspective is uncertain. man) held at least a bachelor’s degree, scored at least an “A” grade in an
The present study extends researchers’ description of the alli- undergraduate counseling interviewing skills course, and ranged in age
ance as understood by clients. In addition to the use of a client- from 26 to 34 years (M ⫽ 28.2, SD ⫽ 3.3). Only one of the interviewers
driven organization of client-identified early alliance formation belonged to an ethnic minority. The interviewers received 18 hr of training
factors, the present study also attempts to model clients’ concep- and supervised practice. Three of these interviewers and one other research
tual structures of these factors using their own words. At the assistant (a White woman in her early 20s) received approximately 10
outset, it is important to recognize that this study is intended to be additional hours of training and conducted the sorting and rating proce-
exploratory, descriptive, and inductive rather than confirmatory, dures (and each worked with 7– 8 participants).
causally explanatory, or deductive. In other words, it is intended to
aid in further understanding the construct of the alliance as under- Procedures
stood by clients and does not purport to offer any causal claims
with respect to this psychological phenomenon. Potential participants were asked prescreening questions addressing the
inclusion criteria outlined earlier. Any participants not meeting these
Method
1
Participants The 25 categories were technical activity, nonverbal communication,
active listening, choices, psychotherapy environment, client agency, per-
A sample of 40 participants who received counseling services within the sonal characteristics, service beyond normative expectation, self-
last year were recruited through leaflets posted at an urban West Coast disclosure, positive commentary, greetings and farewells, positive initial
university, at various local mental health clinics, and at local independent contact, normalization and validation, respecting agreements, positive sen-
practitioners’ offices inviting clients to participate. Through a prescreening timent, recommendations, confidentiality, candor, experience of crying in
telephone interview, participants were included if they believed that they front of the psychotherapist, role induction, humor, emphasizing client
had experienced a “positive working relationship” with their counselor, as expertness, previous knowledge of psychotherapist, openness to personal
evidenced by a rating of the relationship’s strength greater than or equal to criticism, and external contact.
 28 BEDI

criteria were thanked for their time and not invited to participate further. asked to give each pile a label that they believed most accurately repre-
After informed consent was obtained, selected participants were paid $20 sented the statements in it and to rate the importance of each of these
to complete a demographic questionnaire and to take part in an audiotaped statements for establishing a “good working relationship” with them on a
research interview (usually lasting about 1 hr). Specifically, they were 5-point Likert-type scale ranging from 1 (not important) to 5 (extremely
asked to retrospectively recall their understanding of observable occur- important).
rences that contributed to their experience of establishing the alliance. Multivariate concept-mapping (MVCM) statistical techniques were used
Because clients could be somewhat confused by the term counseling to compute the “average” sort across the participants, which represents the
alliance, phrases like “working relationship” and “counseling relationship” most typical conceptual structure used by participants in understanding the
were used instead (cf. Bachelor, 1995; Mohr & Woodhouse, 2001). The client-identified factors important for alliance formation. A review of
following prompting question was used: MVCM studies in the area of counseling and a more detailed elaboration
of MVCM are available in Bedi and Alexander (in press) and in Goodyear,
Please think back over the meetings you had with your mental health Tracey, Claiborn, Lichtenberg, and Wampold (2005). In synopsis, MVCM
professional, paying particular attention to the working relationship is a structured research strategy that includes analyzing sorting information
that was developing between you and the mental health professional. on the basis of a sequential multidimensional scaling and cluster analysis
approach. To help eliminate spurious statistical relationships, a statement
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

What were the things that helped form and strengthen the counseling
must have been sorted with another statement by at least 2 participants in
This document is copyrighted by the American Psychological Association or one of its allied publishers.

or therapy relationship? We are most interested in specific behaviors

and other observable things. These can be things that either you or the order to be considered more than random error (see discussion of similarity
mental health professional did, things you did together, or something cut-off in Concept Systems Inc., 2003).
else that happened within or outside the sessions. Please describe each Each participant’s sorting results were placed into a symmetrical binary
behavior or event completely and in as much detail as possible. data matrix (called a binary similarity matrix; BSM) in which rows and
columns each equal the number of statements and in which endorsement in
In order to limit the reported events to the initial establishment of the this matrix indicates that the statements corresponding to that particular
row and column were sorted together in a pile by that particular person
alliance, only statements referring to the first six sessions or defined by the
(Shern, Trochim, & LaComb, 1995). Next, individual sort matrices (i.e.,
client as occurring “early” in counseling were included. This decision is
BSMs) across the 40 participants were summed to obtain a combined
open to critique but serves as one means of ensuring that the emphasis is
group similarity matrix (GSM). The value in this latter matrix for any pair
on early alliance. The interviewers were trained to use paraphrases, verbal
of statements indicates how many participants placed that pair of state-
probes, and open-ended questions to ensure adequate understanding of
ments together in a pile regardless of what other statements were included
participant statements and to translate the statements to meet the inclusion
or excluded from that pile. Unweighted nonmetric multidimensional scal-
criteria. In order to improve the potential for the results to have practical
ing (nMDS) of the GSM was then conducted.
value, it was decided that acceptable statements be defined as any occur-
It has been recommended that a two-dimensional solution be routinely
rence reported by the participants that could be translated into specific,
fit if the purpose of the nMDS is to visually display clustering results
observable, and behavioral terms. Statements that did not meet these
(Fitzgerald & Hubert, 1987; Jackson & Trochim, 2002; Kruskal & Wish,
criteria were only included if they yielded concrete information that was in
1978). Although this is not often done in the MVCM literature, it seems
a form easily communicable to others on how to develop an alliance. To be
prudent to assess the adequacy of the two-dimensional solution. This was
clear, the emphasis was on description (i.e., what happened) rather than on
done by computing and comparing the stress values of one-dimensional to
explanation (i.e., how or why this resulted in a stronger alliance).
six-dimensional solutions using SPSS 11 for Windows and by examining
After the interview, the interviewers transcribed the interviews, and two
a plot of stress values versus dimensional solution. The stress value
research assistants independently extracted relevant statements from the represents the extent to which the created nMDS pattern of distances
participant accounts. The two lists of extracted incidents (one from each matches the observed similarities and ranges from 0 (no discrepancy
research assistant) were compared and usually found to be at least 90% in between the proximity matrix and the distances on the nMDS map) to 1
line initially. Differences were resolved through discussion and consensus. (the nMDS distances are completely random).
In a few rare cases, the participant was telephoned for clarification or Agglomerative hierarchical cluster analysis (hCA), using Ward’s (1963)
another research team member was consulted and the issue voted upon minimum variance algorithm, was then applied to the nMDS coordinate
among these three individuals. To facilitate the comprehension of the data using The Concept System (Concept Systems Inc., 2003). Cluster
extracted statements, they were translated from nonformal spoken language solutions ranging from 20 to 3 (the default range for the program) were
(including “umms,” stuttering, word repetition) to standard written lan- examined, in that decreasing order, for interpretability. As an alternate
guage using the participant’s idiosyncratic words. decision-making tool, the mean of all standardized cluster bridging values
The participants identified 376 helpful factors in early alliance forma- for each cluster solution was examined. The standardized cluster bridging
tion.2 Duplicate factors mentioned by more than 1 participant were com- value ranges from 0 to 1. It mathematically indexes the average of how
bined, and, in an effort to exclude idiosyncratic factors with little gener- often statements within a particular cluster were sorted with other state-
alizability across participants, any factor not mentioned by at least 2 ments that are close to them on an nMDS plot compared with those
participants (i.e., by more than 2.5% of the sample) was eliminated from statements further away on this plot (Concept Systems Inc., 2003). The
further analyses. After these reductive strategies, 74 statements represent- formula is available in the Appendix (W. Trochim, personal communica-
ing client-identified common factors in early alliance formation remained. tion, February 1, 2004). Lower cluster bridging values indicate that the
These 74 statements were then placed on index cards. statements within this cluster were not commonly sorted with statements
Thirty-one participants returned and were paid $15–$20 (the latter from other clusters. As a third decision-making tool, cluster solutions close
amount was given to those who requested parking reimbursement) to sort in raw number to the average number of piles selected by participants were
these statements into self-defined, mutually exclusive, and conceptually given greater consideration in selecting a final cluster solution.
homogeneous categories in which no single pile was permitted to include The Concept System (Concept Systems Inc., 2003) provides labels for
more than 33% of all cards (for a detailed rationale of the 33% strategy, see
Campbell & Salem, 1999; Jackson & Trochim, 2002; Trochim, Cook, &
2
Setze, 1994). Each sorter decided how many categories to create and what A list of these statements is available from Robinder P. Bedi upon
each category contained. Upon completion of the sorting, participants were request.
 CONCEPT MAPPING CLIENT’S PERSPECTIVE 29

each cluster by supplying a list of titles that individuals provided to their from an 8-cluster solution to a 13-cluster solution. On the basis of
piles. These provided labels are rank ordered in terms of how close the the three used heuristics, an 11-cluster solution was selected. Final
centroid of each individual’s pile falls to the centroid of the aggregated pile labels for the clusters were then selected, as outlined in the Method
corresponding to a particular cluster when plotted in two-dimensional
section.
space. Each cluster was labeled by Robinder P. Bedi through examination
of the closest individual pile sort labels provided by participants for Eleven categories of alliance formation factors were identified:
themes, and choosing one that seemed to best describe the statements in the setting, presentation & body language, nonverbal gestures, emo-
cluster under consideration.3 Upon completion of the study, participants tional support & care, honesty, validation, guidance & challenging,
were mailed the final results of the study in layperson’s language and education, referrals & recommended materials, client’s personal
invited to offer feedback. responsibility, and session administration. The statement compo-
Testimonial validity was built into the research design. Testimonial
sition of each categorical cluster is presented in Table 1.
validity refers to agreement between one’s interpretations of the data and
the intended meanings of those whose experience it purports to represent
The concept map for the final solution is presented in Figure 1.
(Elliott, Fischer, & Rennie, 1999; Stiles, 1993). In other words, it refers to This map combines the nMDS plotted coordinates, the hCA cluster
the extent to which the phenomenon is understood from the perspective of solution, and the selected category labels in pictorial format.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

the participants, using their own language, rather than on the basis of the Because distance is an indicator of conceptual similarity, catego-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

preexisting biases of the researchers. “When the interviewee’s own under- ries falling closer together in Figure 1 are conceptually more
standing of a statement is asked for, the validity of the researcher’s similar. For example, according to the participants in this study,
interpretations is, in principle, decided by the subject” (Kvale, 1996, p.
the category guidance & challenging is understood as being more
217). In the present study, testimonial validity was established in three
ways: (a) At the end of the interview, the interviewer shared his or her similar to the categories education and to validation than it is to the
understanding of the statements identified by the participant and invited the category session administration or to presentation & body
participant to modify the list of documented statements, (b) the original language.
participant’s language was preserved as much as possible in all extracted The average bridging value for each category represents how
statements, and (c) participants judged how well the final results repre- similarly participants understood each category, with lower values
sented their understandings. Means of establishing reliability for this study representing categories of greater conceptual homogeneity across
are as follows:
participants. That is, a category with a high bridging value has
1. The reliability of individual sorting was determined by correlat- more varied meaning for different participants, whereas a category
ing all BSMs and taking the average correlation (analogous to the with low bridging value is understood more similarly across par-
average of interitem correlations). ticipants. This information is presented in Table 2. Categories such
as guidance & challenging, validation, and presentation & body
2. The association of each individual sort with the aggregated sort
language are the most homogeneous constructs.
solution (a form of average item-total correlation) was deter-
mined by correlating each BSM with the GSM and taking the The average rating of importance of each category in facilitating
average. alliance formation, as understood by participants, was calculated
by taking the average rating of importance that participants pro-
3. The split-half reliability was examined by recalculating two vided to the statements within each category, which is presented in
GSMs on the basis of both (a) an odd-even split of participants Table 3. Validation was rated as the most important category for
and (b) a male-female split and correlating the two respective
alliance formation.
GSM matrices.
The results appear to have adequate to high reliability. The
4. The “internal consistency” of the two-dimensional solution was average phi correlation between individual sort matrices (BSMs)
examined by evaluating stress values computed on both (a) an was .17 (SD ⫽ .09), indicating significant (at ␣ ⫽ .05) but low
odd-even split of the participants and (b) a male-female split of correlation between individual sorts. This denotes much variability
the participants. in how participants understand the relationships among the factors
underlying early alliance formation. Considering that the average
Results BSM correlation in MVCM studies appears to be only .12 without
the correction factor, used by Trochim (1993), the obtained result
Participants sorted the statements into an average of about 12
piles (M ⫽ 11.7, SD ⫽ 4.8). On the basis of the plot of stress
3
values versus dimensionality, it was reasonable to conjecture a A list of the eight “best” labels for each cluster is available from
two-dimensional solution because a notable bend in the plot could Robinder P. Bedi upon request.
4
be judged to occur at this junction. The stress value for the nMDS According to Trochim (1989), a value of less than .30 represents a
two-dimensional solution was .24.4 Because the average number stable MDS solution for MVCM studies, whereas a stress value over .40
of piles used by participants was 11.7, 11-, 12-, and 13-category indicates an unstable solution. This heuristic was supported by a subse-
solutions, which fall within about 0.25 SDs from the mean, were quent meta-analysis (Trochim, 1993) that found, across 33 MVCM studies,
that the average stress value was 0.29 (SD ⫽ 0.04). This standard is
considered most in line with participant conceptualizations. In
considerably higher than standard conventions for MDS, which define
examining interpretability, starting with a 20-cluster solution in adequate fit as ⱕ 0.15 (Stalans, 1995). However, it must be remembered
decreasing order, it was noted that participant labels for the clus- that this recommendation is premised on research with more stable phe-
ters became less comprehensible after the 11-cluster solution. In nomenon (e.g., physiological perceptions of color similarities), fewer en-
addition, the mean cluster bridging value for an 11-cluster solution tities, and more precise measurement methods (e.g., paired comparisons)
was slightly smaller than all other average cluster bridging values (Trochim, 1993).
 30 BEDI

Table 1
Statement Composition of Each Category

Category Statement

Nonverbal gestures The counselor nodded.

The counselor leaned forward.
The counselor made eye contact with me.
Emotional support & care The counselor let me cry.
The counselor and I laughed together.
The counselor gave me a tissue when I cried.
The counselor provided verbal prompts (e.g.“uh-huh”, “hm-mmm”)
The counselor made jokes.
The counselor offered to hug me, but only if asked.
The counselor took notes.
Presentation & body The counselor shook my hand at the first session.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

language The counselor did not sit behind a desk.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

The counselor walked out of the office/counseling room with me.

The counselor sat facing me (i.e., sat directly across from me).
The counselor sat still (i.e., did not fidget).
The counselor greeted me at the beginning of each session (e.g., “Hello,” “How are you today?”).
The counselor greeted me with a smile.
The counselor sat up straight.
The counselor recognized me (e.g., remembered my name, made eye contact with me while I was in the
waiting room).
Setting The counselor’s office had flowers/plants.
The counselor’s office was decorated with little objects (e.g., seashells).
The counselor wore casual but appropriate office clothing (e.g., a button-up shirt not a t-shirt, pants not jeans).
The counselor wore either a suit or a shirt and tie.
The counselor was hygienic and well-groomed (e.g., clean nails, clean clothes).
The counselor’s support staff greeted me.
The counselor served me food and drinks.
The counselor’s office had books on shelves (counseling books and/or noncounseling books).
The counselor let me choose which chair I could sit in.
The counselor took my preferences for the office into account (e.g., chair position, blinds, air conditioning).
The counselor was a woman.
Session admin. The counselor kept the administration (e.g., fees, scheduling of appointments, paperwork) outside of the
sessions.
The counselor assured me about confidentiality (and its exceptions).
The counselor explained the process of counseling (e.g., how it works, how many sessions we can have, how to
cancel a session).
The counselor told me that it is my decision when counseling would end.
The counselor invited me to call or e-mail in between sessions, if I needed to.
The counselor did not turn me away, even though it was reasonable for him/her to do so (e.g., he/she saw me
before the grant money came in, he/she was my high school counselor, but I already graduated).
Client’s personal The counselor came highly recommended by my friend(s).
responsibility The counselor worked in the same office as another counselor who I thought was a good counselor.
I had previous contact with the counselor (i.e., I knew him/her or about him/her before starting counseling).
I found or chose my own counselor rather than being appointed one.
I attended sessions regularly as scheduled (e.g., once per week).
I came on time for sessions.
I informed the counselor in advance if I was going to be late or miss a session.
I had noncounseling contact with the counselor after sessions or in between sessions.
Referrals & recommended The counselor made referrals to available and useful services (e.g., AA, mental health advocate, vocational
materials services society).
The counselor suggested additional materials (e.g., books, Web sites, videos) for outside of the sessions.
The counselor helped me with my medication issues (e.g., calling the psychiatrist).
The counselor followed through on what he/she said he/she would do (e.g., brought a questionnaire to the next
session, called with the name of a book).
Guidance & challenging The counselor asked me, “How did that make you feel?”
The counselor summarized what I said.
The counselor kept me on topic.
The counselor confronted or challenged me.
The counselor referred to details talked about in previous sessions.
The counselor asked me about my goals.
The counselor told me about similar experiences that he/she had.
The counselor provided suggestions.
The counselor de-emphasized his/her expertness (e.g., stated he/she was more like a coach, stated that I know
myself the best, admitted when he/she did not know something)
The counselor let me decide what to talk about (or not to talk about).
 CONCEPT MAPPING CLIENT’S PERSPECTIVE 31

Table 1 (continued )

Category Statement

Education The counselor asked me what steps that I thought I needed to take.
The counselor taught me skills.
The counselor not only assigned homework but also would review it at the next session.
I expanded on the issues that I brought up in counseling and/or on my feelings about the issues.
Honesty The counselor told me personal information about himself/herself.
The counselor was honest (i.e., he/she shared negative information truthfully).
The counselor listened to my truthful negative personal reactions to him/her.
Validation The counselor validated my experience (e.g., he/she said that my reaction was understandable and reasonable,
that it was okay to feel this way).
The counselor normalized my experiences.
The counselor identified and reflected back my feelings.
The counselor paraphrased (i.e., repeated back in different words) what I said.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

The counselor made encouraging comments.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

The counselor would ask me about other parts of my life (e.g., family, friends) and not just only about my
problem(s).
The counselor asked questions.
The counselor agreed with what I said.
The counselor made positive comments about me.

Note. admin. ⫽ administration; AA ⫽ alcoholics anonymous.

is still favorable compared with the population of MVCM studies.5 The stress values for the odd-even split of participants were .27
The average point-biserial correlation of each BSM with the GSM (n ⫽ 16) and .26 (n ⫽ 15), respectively— both considered to be
was rpb ⫽ .45 (SD ⫽ .11), indicating an almost moderate corre- stable values for MVCM studies and both below the average
lation between the individual sorts and the aggregated group sort. split-half stress value in existing MVCM studies (Trochim, 1993).
This can be taken to suggest that the GSM is a fairly trustworthy The stress values for female participants was .24, and the stress
indicator of individual sorting. According to Trochim’s (1993) value for male participants was .29, and both these values are also
meta-analysis, the average correlation between BSMs and the considered stable.
GSM was .93 (SD ⫽ .02), but without the unconventional use of The participants overwhelmingly endorsed the proposed cate-
the Spearman-Brown correction, it was .29 —much lower than gorical scheme of client-identified factors as representing their
what was found in the present study. The correlation between the
understanding, as evidenced by statements such as “I think the
odd-even split GSMs (i.e., the coefficient of equivalency) was high
factors chosen are certainly the primary relevant factors, and in my
at .76 ( p ⬍ .001), and when appropriately corrected for attenuation
relationship with my counselor, the picture herein is quite repre-
with the Spearman-Brown formula (for the formula, see Trochim,
sentative,” “. . . I am gratified to see the results, which are so much
1993) the correlation was .86, which is considered good reliability.
The correlation between the GSM of the male participants with the in tune with my own experience,” “I completely agree with all of
GSM of the female participants was high at .74 ( p ⬍ .001). them [categories]. I think they’re all important . . . they seem
accurate,” “Very well comprised list of categories. I like the names
of the categories because they are put in words that regular people
understand,” and “I am also sending a copy of this to the counselor
I had seen to spread this important and necessary info!”

5
According to a meta-analysis by Trochim (1993), the average correla-
tion among BSMs was .82 (SD ⫽ .07). However, it appears that Trochim
unconventionally applied the Spearman-Brown correction, thereby inflat-
ing this value. Standard conventions in the area of reliability estimation
indicate that the Spearman-Brown correction should be applied in cases in
which all the test items are not used in calculating a reliability correlation,
such as in a split-half reliability analysis. On the basis of Trochim’s
description of his analytical method (and confirmed through subsequent
e-mail contact), in the calculation of the corrected value, it appears that
Trochim applied the Spearman-Brown correction at the level of sample
size (using the number of studies in the meta-analysis as the total) and
considered the averaged correlation across the studies as only one value out
of the total number in the study. This does not seem appropriate for the
traditional Spearman-Brown correction. Therefore, eliminating this correc-
Figure 1. Concept map of categories of helpful factors in alliance for- tion factor, the average correlation between individual sorts seems to be
mation. approximately .12.
 32 BEDI

Table 2 ing the client’s perspective, which can inform future theorizing.
Bridging Values of Each Category Particular emphasis is placed on how the perspective of the par-
ticipants in this study differs from the perspective of counselors
Category Bridging value and researchers, as implied by popular theories and measures of
Guidance & challenging .13 the alliance. The following section is provisional and for the
Validation .15 purposes of putting forth preliminary propositions consistent with
Presentation & body language .27 a newly emerging model of the client’s perspective (see also
Setting .43 Bachelor, 1995; Bedi, Davis, & Arvay, 2005; Bedi, Davis, &
Nonverbal gestures .45
Williams, 2005; Mohr & Woodhouse, 2001).
Education .47
Honesty .53 First, although collaboration and mutuality are central to most
Emotional support & care .63 researcher-derived and practitioner-derived theories (Horvath &
Client’s personal responsibility .69 Bedi, 2002), clients do not characteristically affirm the importance
Session admin. .85 of collaborative efforts in alliance formation. Instead, they more
Referrals & recommended materials .95
often choose to assign the majority of the responsibility to the
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Note. admin. ⫽ administration. counselor (this point is supported by all five studies on the client’s
perspective, including the present one, that were identified in the
previous paragraph). This appears to be the case because clients
may be so vulnerable or distressed that they sometimes see them-
Several participants provided elaboration on the category of selves, in the words of one of the participants in the study, as
setting, as evidenced by the following two sets of comments: “unable to handle much responsibility.” Second, counselor tech-
After I’d finished the sorting the other day I came in, and after I went nical activities may be, for the most part, inextricable from the
home, I was surprised at the importance I’d put on setting and counseling alliance because techniques were seen by the partici-
presentation and body language. For my own experience, in looking pants in this study to fundamentally contribute to building the
back, I see how it had a profound influence on my feeling safe to counseling alliance (this finding was also replicated across all five
continue with my counselor. I wouldn’t have been able to verbalize it studies). Third, existing theories of the alliance place little, if any,
in the beginning. explicit emphasis on the importance of session administration,
I think the setting should be more important [It was the lowest rated
setting, education (particularly responsive education, as implied by
category]. You want to see it as a professional place and not dirty and the composition of this category), referrals, recommended materi-
grungy. You want to see that the person cares about the office. The als, guidance, and challenge in alliance formation (all of which
office helps make you feel comfortable. were categories found in the present study). Fourth, many key
factors that clients understand to be important for alliance forma-
Two participants also provided an elaboration of personal re- tion may be deceptively simple and reflect counselor microskills
sponsibility: such as nonverbal gestures and empathic listening (see also Bedi,
Davis, & Arvay, 2005; Bedi, Davis, & Williams, 2005). Fifth,
Regarding the clients’ responsibility—it seems it’s not that important
from the clients’ point of view. . . . From the counselor’s point of
alliance formation, as understood by clients, actually begins before
view, it’s probably frustrating to have a client who shows up late, the counselor fully engages the client, as clients may develop
doesn’t show up, or doesn’t inform in advance. But I guess the study predispositions or impressions on the basis of the counselor’s
is looking at how each factor ‘helps develop good working relation- attire, the counselor’s nonverbal gestures, the counselor’s greeting,
ships’ from the point of view of the client. They may not be aware that
it is a problem that can affect their ‘good working relationship.’
Table 3
I understand why client’s personal responsibility is not as ‘highly Rated Importance of Each Category in Facilitating Early
rated’ as other categories, because as a client, I sometimes saw myself Alliance Formation
as unable to handle much responsibility.
Average rating
Discussion Category of importance

Very little is known about the counseling events that clients Validation 4.06
Education 3.88
deem most helpful for forming a quality counseling alliance. This Nonverbal gestures 3.80
study represents the first formal research attempt not only to Presentation & body language 3.80
systematically list the factors that clients believe are important in Guidance & challenging 3.76
early alliance formation but also to model clients’ conceptual Referrals & recommended materials 3.73
structure regarding alliance formation factors. The results of this Honesty 3.69
Emotional support & care 3.65
study are very much in line with the eight alliance formation Session administration 3.63
categories found by Bedi, Davis, and Arvay (2005) and greatly Client’s personal responsibility 3.23
overlap with the 25 categories found by Bedi, Davis, and Williams Setting 3.03
(2005).
Note. Statements within each category were rated for importance on a
The literature lacks a coherent theoretical framework of the 5-point Likert-type scale in which 1 ⫽ not important, 2 ⫽ slightly impor-
counseling alliance, derived predominantly through client concep- tant, 3 ⫽ moderately important, 4 ⫽ very important, and 5 ⫽ extremely
tions. The results of this study offer a preliminary model describ- important.
 CONCEPT MAPPING CLIENT’S PERSPECTIVE 33

the office environment, and the reception staff (see also Bedi, mation but that they have no quantitatively measurable effect on
Davis, & Arvay, 2005; Bedi, Davis, & Williams, 2005). Sixth, alliance formation. Whether or not these variables actually have a
constructs such as validation, guidance and challenging, and pre- measurable impact on alliance formation is beyond the scope of
sentation and body language were found to be the most concep- this study—which is to broaden researchers’ understanding of the
tually homogeneous categories and could serve as generalizable client’s perspective on the alliance. Fifth, even if it is assumed that
constructs in a model of the client’s perspective of the alliance. these factors will eventually be supported by controlled research,
Finally, validation, education, nonverbal gestures, and presentation the information should still not be translated into a mere prescrip-
& body language were rated as being the most important catego- tive checklist for counselors to adhere to with each client because
ries in this study and could also play a central role in any devel- considerable skill, adaptability, and appropriate timing are in-
oping model of the client’s perspective on the alliance. volved in providing such facilitative behaviors and conditions.
Because most of the alliance measures presently in use are based The results of the present study underscore the need to fully
on theories and constructs derived from practitioner and investi- investigate the process of alliance formation from the client’s
gator understandings or interpretations of client understandings perspective. The findings remain tentative, however, considering
and because of the likelihood that clients and counselors use the preliminary, descriptive, and exploratory nature of this study.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

different indicators when considering the quality and strength of Several methodological limitations of the study should be noted.
the alliance, present measures of the alliance may not tap a number First, only very cautious generalizations should be made beyond
of the areas of the alliance that are important to clients. The the particular sample of clients used in this study. Generalizability
omission or underrepresentation in these scales of constructs sub- is limited to the demographics of the clients because the categories
jectively deemed important by the participants may result in test of alliance development factors could differ across client popula-
scores that inadequately reflect the strength of the alliance as tions, age, ethnicity, gender, setting, presenting issue, and alliance
experienced by the client (cf. comments of Bachelor, 1995). By strength. Second, only those statements that were thought to have
this reasoning, clients’ scores on most existing alliance measures a positive influence on the formation of the alliance were obtained,
would be less relevant for measuring their experience of the whereas a richer understanding of the client’s perspective on the
alliance than for measuring counselors’ perceptions of the alliance. alliance requires understanding negative factors as well. Third, the
Consequently, if investigators wish to claim to measure a client’s present study only focused on the initial formation and strength-
subjective experience of the alliance, then constructs that reflect ening of the alliance, whereas the factors that help to develop the
the client’s perspective need to be included into future measures. alliance may be substantially different from those that are required
On the basis of the client accounts provided in this study, it to maintain it. Fourth, by limiting the identified factors to those
seems that several variables have been largely ignored in previous that were concrete, specific, and behavioral, this study may have
alliance research. These underresearched variables include the overlooked more indeterminate, interactional, or abstract factors.
counselor’s personal characteristics, the counselor’s subtle nonver- Fifth, in limiting the statements to 74 by a participation rate greater
bal behaviors and messages, and the counseling environment. than 5% and by combining repetitive statements, the general
According to the review of Horvath and Bedi (2002), there are relevance of the items for sorting and the procedural validity of the
very few investigations of possible counselor contributions to the study were increased. In doing so, however, the study may have
alliance, for which there is a most urgent need for researchers to omitted some valuable influences on alliance formation that are
attend to. This deficiency is well addressed by the present study very rare. It is important not to equate the frequency of a variable
because it provides at least 65 potential counselor variables that with its importance or impact. That is, a low-frequency variable
should be subject to further investigations, especially in controlled (e.g., one occurring only for 1% of participants) may still have a
and prospective research designs. monumental effect in the formation of the alliances of this minor-
Several interpretation caveats are important to recognize. First, ity of individuals. Regardless, such a factor is of little relevance for
we should not automatically assume that these client-identified the vast majority of clients and therefore also for describing a
factors are exhaustive because they are bound by clients’ aware- typical conceptual structure. Sixth, the helpful factors identified by
ness (i.e., other factors may have operated outside of clients’ participants were not verified as being accurate reports of actual
awareness). Second, counselors should not unconditionally pro- events that occurred in their counseling. Seventh, because random-
vide clients with whatever type of alliance they desire. There are sampling procedures were not used and because demographic
clinical (e.g., transference) and ethical considerations (Norcross & information was not available for those in the solicited settings
Beutler, 1997). Third, only sampling those with favorable alli- who did not participate in the study, it is problematic to confidently
ances, although necessary to limit the scope of this study, presents conclude that the present sample represents the population of
a skewed portrayal that may not necessarily generalize to those clients with positive alliances.
with alliances of less strength. We cannot assume that “good” and Additional limitations are typical of self-report studies. Retro-
“bad” alliances fall on a continuum. It is an assumption that the spective recall can sometimes result in important aspects or nu-
factors involved in fostering highly positive alliances are just more ances of experiences being forgotten or misremembered (Nisbett
of or less of what is present in weaker alliances—there may be & Wilson, 1977) because people recreate the past on the basis of
some distinct factors present in each (i.e., differences in kind). other memories, their general knowledge, and the social demands
Fourth, given the nonexperimental research method used, there is of the retrieval situation (Hyman & Loftus, 1998). “Although
no defensible guarantee that these factors will result in stronger memory errors are the result of normal memory construction, this
alliances when manipulated in controlled studies. Instead, it is does not mean that everything a person remembers is erroneous.
possible that clients believe these factors are key “characters” in Most memories will be generally accurate” (Hyman & Loftus,
their narrative understandings of what happened in alliance for- 1998, p. 945). Clients’ reports may also be biased by response sets,
 34 BEDI

preexisting beliefs, and presentation style. Nevertheless, one study of inquiry. Upon the accumulation of enough research on the
(Martin & Stelmaczonek, 1988) found that clients were able to client’s subjective perspective on the alliance, it seems important
recall events that they had previously identified as important with to eventually develop a new measure designed to assess a client’s
surprising accuracy up to 6 months after they had completed subjective experience of the alliance.
counseling.
Future research should address the limitations of this research References
study already outlined. For example, these results should be rep- Bachelor, A. (1995). Clients’ perception of the therapeutic alliance: A
licated in different populations (e.g., inpatients) and with samples qualitative analysis. Journal of Counseling Psychology, 42, 323–337.
comprising very different demographics (e.g., a larger proportion Bedi, R. P., & Alexander, D. A. (in press). Using multivariate concept-
of men, ethnic minorities). Narrative analyses of the categories can mapping for assessing client understandings of counselling and psycho-
assist researchers in understanding the nuances of how clients therapy. In B. D. Zumbo (Ed.), Developments in theories and applica-
make sense of these diverse factors and may enlighten researchers tions of measurement, evaluation, and research methodology across the
about the reasons why these elicited factors are seen to result in a disciplines (Vol. 2). Edgeworth Laboratory: Vancouver, British Colum-
bia.
quality alliance. Moreover, clients should be allowed to review
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Bedi, R. P., Davis, M. D., & Arvay, M. A. (2005). The client’s perspective
This document is copyrighted by the American Psychological Association or one of its allied publishers.

videotaped sessions in order to identify alliance formation factors on forming a counselling alliance and implications for research on
that are known to have actually occurred. In addition, because counsellor training. Canadian Journal of Counselling, 39, 71– 85.
these factors can only be taken to represent client understandings, Bedi, R. P., Davis, M. D., & Williams, M. (2005). Critical incidents in the
experimental and controlled investigations of each of these factors formation of the therapeutic alliance from the client’s perspective. Psy-
and of each of the categories needs to be completed. It would also chotherapy: Theory, Research, Practice, Training, 42, 311–323.
appear fruitful to correlate the factors and categories of factors Bordin, E. S. (1979). The generalizability of the psychoanalytic concept of
found in this study with existing alliance measures to better de- the working alliance. Psychotherapy, 16, 252–260.
termine overlapping and unique content. Discovering factors that Campbell, R., & Salem, D. A. (1999). Concept mapping as a feminist
do not correlate well with existing alliance measures would further research method: Examination of the community response to rape.
Psychology of Women Quarterly, 23, 65– 89.
bolster the claim that the client’s perspective is unique.
Cecero, J. J., Fenton, L. R., Frankforter, T. L., Nich, C., & Carrol, K. M.
Perhaps the biggest drawback to the methodological choices (2001). Focus on therapeutic alliance: The psychometric properties of
made in this study is the inability to speak about a broader client’s six measures across three treatments. Psychotherapy, 38, 1–11.
perspective on the alliance. It would be interesting to investigate Concept Systems Inc. (2003). The concept system facilitator training
negative factors for those with strong alliances. It would also be seminar manual. Ithaca, NY: Author.
interesting to discover the client’s perspective on what is contrib- Elliott, R. E., Fischer, C. T., & Rennie, D. L. (1999). Evolving guidelines
uting to or missing in weaker alliances. Clients who unilaterally for publication of qualitative research studies in psychology and related
terminated counseling could provide significant factors involved in fields. British Journal of Clinical Psychology, 38, 215–229.
alliance dissolution, and this information could be compared with Elliott, R. E., & James, E. (1989). Varieties of client experience in
those factors provided by clients in weaker alliances. This way, psychotherapy: An analysis of the literature. Clinical Psychology Re-
view, 9, 443– 467.
some indication of the differences between the client’s perspective
Fitzgerald, L. F., & Hubert, L. J. (1987). Multidimensional scaling: Some
on poor alliances and dissolved alliances may emerge. Neverthe- possibilities for counseling psychology. Journal of Counseling Psychol-
less, only by investigating the client’s perspective in both good and ogy, 34, 469 – 480.
poor alliances and for both positive and negative factors will the Gelso, C. J., & Carter, J. A. (1985). The relationship in counseling and
best understanding of the client’s perspective emerge. By doing so, psychotherapy: Components, consequences, and theoretical antecedents.
researchers will be better able to determine whether factors under- The Counseling Psychologist, 13, 155–243.
stood to be present in strong alliances are on a continuum with Goodyear, R. K., Tracey, T. J. G., Claiborn, C. D., Lichtenberg, J. W., &
those present in weaker alliances or whether alliance development Wampold, B. E. (2005). Idiographic concept mapping in counseling
is understood to be a discontinuous process (i.e., a strong alliance psychology research: Conceptual overview, methodology, and an illus-
may have distinct factors in it and therefore is not merely more of tration. Journal of Counseling Psychology, 52, 236 –242.
Horvath, A. O. (2001). The therapeutic alliance: Concepts, research and
the things present in weaker alliances). Researchers would also be
training. Australian Psychologist, 36, 170 –176.
better able to determine whether factors understood to contribute Horvath, A. O., & Bedi, R. P. (2002). The alliance. In J. C. Norcross (Ed.),
to alliance development are at one end of the spectrum compared Psychotherapy relationships that work: Therapist contributions and
with those related to alliance deterioration or whether qualitative responsiveness to patient needs (pp. 37– 69). London: Oxford University
differences exist between helpful and harmful factors (e.g., pro- Press.
viding psychoeducation could be a helpful factor, but not provid- Horvath, A. O., & Greenberg, L. S. (1989). Development and validation of
ing psychoeducation may not necessarily be a harmful factor). the Working Alliance Inventory. Journal of Counseling Psychology, 36,
Finally, researchers could also determine whether there is an 223–233.
interaction between alliance quality and the experienced nature Horvath, A. O., & Luborsky, L. (1993). The role of the therapeutic alliance
(helpful or harmful) of factors in the client’s perspective. in psychotherapy. Journal of Consulting and Clinical Psychology, 61,
561–579.
The results of the present study, in combination with Bachelor
Horvath, A. O., & Marx, R. W. (1990). The development and decay of the
(1995); Bedi, Davis, and Arvay (2005); Bedi, Davis, and Williams working alliance during time-limited counseling. Canadian Journal of
(2005); and Mohr and Woodhouse (2001), substantiate the distinc- Counseling, 24, 240 –259.
tiveness of the client’s perspective and the value of modeling the Horvath, A. O., & Symonds, B. D. (1991). Relation between working
client’s subjective perspective on the alliance. It is hoped that this alliance and outcome in psychotherapy: A meta-analysis. Journal of
study can serve as an impetus for further opening up this new area Counseling Psychology, 38, 139 –149.
 CONCEPT MAPPING CLIENT’S PERSPECTIVE 35

Hyman, I. E., Jr., & Loftus, E. F. (1998). Errors in autobiographical Paulson, B. L., Truscott, D., & Stuart, J. (1999). Clients’ perceptions of
memories. Clinical Psychology Review, 18, 933–947. helpful experiences in counseling. Journal of Counseling Psychology,
Jackson, K. M., & Trochim, W. M. (2002). Concept mapping as an 46, 317–324.
alternative approach for the analysis of open-ended survey responses. Shern, D. L., Trochim, W. M., & LaComb, C. A. (1995). The use of
Organizational Research Methods, 5, 307–336. concept mapping for assessing fidelity of model transfer: An example
Kruskal, J., & Wish, M. (1978). Multidimensional scaling. Beverly Hills, from psychiatric rehabilitation. Evaluation and Program Planning, 18,
CA: Sage. 143–153.
Kvale, S. (1996). InterViews: An introduction to qualitative research Stalans, L. J. (1995). Multidimensional scaling. In L. G. Grimm & P. R.
interviewing. London: Sage. Yarnold (Eds.), Reading and understanding multivariate statistics (pp.
Luborsky, L. (1976). Helping alliances in psychotherapy. In J. Claghorn 137–168). Washington, DC: American Psychological Association.
(Ed.), Successful psychotherapy (pp. 92–111). New York: Brunner/
Stiles, W. B. (1993). Quality control in qualitative research. Clinical
Mazel.
Psychology Review, 13, 593– 618.
Luborsky, L. (2000). A pattern-setting therapeutic alliance study revisited.
Tichenor, V., & Hill, C. E. (1989). A comparison of six measures of
Psychotherapy Research, 10, 17–29.
working alliance. Psychotherapy, 26, 195–199.
Martin, J., & Stelmaczonek, K. (1988). Participants’ identification and
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Trochim, W. M. (1989). An introduction to concept mapping for planning

recall of important events in counseling. Journal of Counseling Psychol-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

ogy, 35, 385–390. and evaluation. Evaluation and Program Planning, 12, 1–16.
Mohr, J. J., & Woodhouse, S. S. (2001). Looking inside the therapeutic Trochim, W. M. (1993, November). The reliability of concept mapping.
alliance: Assessing clients’ visions of helpful and harmful psychother- Paper presented at the annual conference of the American Evaluation
apy. Psychotherapy Bulletin, 36, 15–16. Association, Dallas, TX.
Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we know: Verbal Trochim, W. M., Cook, J. A., & Setze, R. J. (1994). Using concept
reports on mental processes. Psychological Review, 84, 231–259. mapping to develop a conceptual framework of staff’s views of a
Norcross, J. C., & Beutler, L. E. (1997). Determining the therapeutic supported employment program for individuals with severe mental ill-
relationship of choice in brief therapy. In J. N. Butcher (Ed.), Personality ness. Journal of Consulting and Clinical Psychology, 62, 766 –775.
assessment in managed health care: Using the MMPI-2 in treatment Ward, J. H. (1963). Hierarchical grouping to optimize an objective func-
planning. New York: Oxford University Press. tion. Journal of the American Statistical Association, 58, 236 –244.

Appendix

Standardized Bridging Value

The standardized bridging value (Bi) ranges from 0 to 1 and mathemat- d ij ⫽ ⻫共 x i ⫺ x j 兲 2 ⫹ 共 y i ⫺ y j 兲 2 (A3)
ically indexes how often a item was sorted with others that are close to it
on an MVCM map compared to those statements further away on the map in which
(Concept Systems Inc., 2003). It is expressed in Equation A1 (W. Trochim, xi ⫽ MDS x-coordinate for point i
personal communication, February 1, 2004). A cluster bridging value is the xj ⫽ MDS x-coordinate for point j
average bridging value of all the items in a particular category and reflects yi ⫽ MDS y-coordinate for point i
the homogeneity in meaning of that cluster across participants in terms of yj ⫽ MDS y-coordinate for point j
how often items in that cluster were sorted with items from other catego- and where Pij is calculated by Equation A4:
ries.
Pij⫽sij/m (A4)
b i ⫺ min共b兲
Bi ⫽ (A1)
max(b) ⫺ min(b) in which
sij ⫽ number of sorters who placed point i and j together in the same pile
in which
m ⫽ total number of sorters
Bi ⫽ the (standardized) bridging value for point i
min(b) ⫽ minimum of the bi values No normative cutoffs are yet available, and the standardized bridging
max(b) ⫽ maximum of the bi values index should be considered noninferential and on at least an ordinal scale.
bi ⫽ the raw bridging value for point i, and is calculated by the formula In a descriptive sense, lower bridging values indicate a closer relationship
in Equation A2: with other statements in the cluster (i.e., this statement is not commonly
¥¥(pij ⫻ dij) sorted with other statements). A good cluster solution will have many low
bi ⫽ (A2) bridging values.
¥¥pij
in which
pij ⫽ proportion of sorters who placed item i and item j in the same pile Received December 10, 2004
dij ⫽ standardized Euclidean distance between points i and j, and is Revision received April 27, 2005
calculated by the formula in Equation A3: Accepted April 27, 2005 䡲