Attachment and Supervision
1
© 2006 American Psychological Association
"This article may not exactly replicate the final version published in the APA journal. It is not the copy of record."
To access the full text of this publication,
please visit the Professional Psychology: Research and Practice journal home page at
http://www.apa.org/journals/pro/.
Attachment Processes in the Supervisory Relationship
Shelley Riggs and Karen Bretz
University of North Texas
Regular Mail:
Shelley A. Riggs, Ph.D.
Department of Psychology
P.O. Box 311280
University of North Texas
Denton, TX 76203-1280
Telephone:
Facsimile:
(940) 565-2672
(940) 565-4682
Electronic Mail: riggs@unt.edu
RUNNING HEAD: ATTACHMENT AND SUPERVISION
Attachment and Supervision
2
Abstract
How do relational variables influence the quality of the supervisory relationship from the
supervisee perspective? Recently, it has been suggested that attachment theory might provide
some insight into this question, so we explored adult attachment processes and the early parentchild attachment experiences of supervisees in relation to the supervisory working alliance.
Findings highlight the relevance of multiple attachment constructs in the supervision process.
Results suggest that a history of parental indifference, compulsive self-reliance in supervisees
and supervisor attachment style may be particularly important in shaping the supervisory
working alliance. Implications for training and supervision are discussed.
Attachment and Supervision
3
Attachment Processes in the Supervisory Relationship
Clinical supervisors are responsible for the education, guidance and encouragement of the
trainees they supervise (Ramos-Sanchez, Esnil, Goodwin, Riggs, Touster, Wright,
Ratanasiripong, & Rodolfa, 2002). Because supervisory relationships during training shape the
emerging professional identities of young psychologists, they may be the most formative
relationships of our professional lives. The foundational nature of these supervisory relationships
makes it critical that these experiences be positive for the trainee. Yet, regardless of similarities
between supervisors and their trainees in theoretical orientation, personality or relational style,
sometimes a positive supervisory working alliance between trainees and their supervisors can
remain elusive. Paradoxically, at other times, in spite of seemingly important individual
differences between supervisor and supervisee, the supervisory dyad develops into an
exceptionally positive relationship. Very little empirical research has explored the impact of
individual and interpersonal variables on the supervisory alliance, possibly due to the lack of a
suitable theoretical framework. Recently, experts have suggested that attachment theory may
provide a strong theoretical foundation from which to conceptualize the supervisory relationship
(Pistole & Watkins, 1995;Watkins, 1995). The current project was guided by the key question:
In what ways do attachment processes influence the quality of the supervisory working alliance?
Bordin (1983) suggested that there was a correlation between the quality of the
supervisory alliance and supervisee development. Subsequently, theorists focused on
formulating developmental models to explicate the underlying processes of both supervisee and
supervisor development (Hess, 1986; Stoltenberg & Delworth, 1987; Watkins, 1990). These
theorists hoped that developmentally matched pairings of supervisors and supervisees would
enhance the learning experience for both supervisors and supervisees. However, the concept of
Attachment and Supervision
4
supervisory developmental levels can be useful in facilitating positive matches only if those
levels are identifiable. In an analogue study of 48 supervisors, Chagnon and Russell (1995)
examined the ability of supervisors to accurately determine supervisee developmental level
based on taped counseling sessions. Regardless of their own experience level, supervisors
demonstrated the most difficulty correctly identifying the mid-levels of development (striving for
independence, needing a supervisory environment low in structure but high in support). Results
indicated that supervisors tended to make judgments about their supervisee’s developmental
level by making referent comparisons across cohort groups of supervisees rather than by
attending to individual supervisee characteristics.
Borders (1989) suggested that individual differences, such as supervisee thoughts and
emotions, are relevant to supervision because they may moderate the normative anxiety a
supervisee experiences during supervision and counseling. Borders suggested that supervisee
characteristics potentially influencing the supervisory alliance include interpersonal maturity,
clinical experience level, and relational style. He posited that in tailoring the supervision
experience to the needs of individual trainees, an understanding of individual differences across
supervisees may be far more informative than the general descriptions posited by developmental
stage models in the prediction of a positive supervisory working alliance. To elucidate individual
factors contributing to the supervisory experience, researchers have examined variables such as
gender (Lichtenberg & Goodyear, 2000; Twohey & Volker, 1993), identity issues (Ellis &
Douce, 1994), and trainee and supervisor emotional well-being (White & Queener, 2003). In a
recent national survey, psychology interns who reported a negative event in supervision
indicated that the experience deleteriously impacted their training and the supervisee-client
relationships (Ramos-Sanchez et al., 2002). Qualitative analyses indicated that the majority of
Attachment and Supervision
5
negative experiences reported by participants involved interpersonal stylistic differences between
supervisees and supervisors. The current study extends Ramos-Sanchez et al.’s research to a
quantitative analysis of individual differences in interpersonal style, conceptualized as
attachment style, in relation to the supervisory alliance.
Supervisory Working Alliance
Bordin (1983) argued that the quality of the working alliance between supervisor and
trainee was more vital to positive outcomes for the trainee than the particular supervision model
or approach to which the supervisor subscribed. He theorized that the relationship between
trainees and their supervisors could best be described as a type of attachment bond based on a
foundation of mutual trust, which facilitates agreement and understanding of the tasks and goals
of supervision. Consequently, the attachment processes activated in the supervisory relationship
may play a crucial role in determining the effectiveness of the supervisory working alliance
(Epps, 1999).
Bordin (1976, 1983) identified three main components to the working alliance: (1)
consensus and shared understanding regarding the goals of the change process; (2) the identified
tasks of each partner; and (3) the bonds founded upon the level of trust between partners, which
is necessary to sustain the working alliance. Bordin postulated that these three distinct
dimensions of the working alliance were not equally important during all phases of the helping
relationship. His work helped to generalize the concept of the working alliance to all changeproducing relationships, including the therapist-client relationship and the clinical supervisory
relationship (Horvath & Greenberg, 1986).
Horvath and Greenberg (1986) refined the operational definitions of Bordin’s concepts in
order to develop an instrument measuring the working alliance of change-producing
Attachment and Supervision
6
relationships, which they called the Working Alliance Inventory. They defined goals as the
outcomes mutually agreed upon by supervisor and supervisee as the targets of intervention.
Secondly, they described the supervisory tasks as the behaviors within session perceived by both
members as relevant and efficacious, and for which each member accepts responsibility. The
final contributor to the working alliance is the relationship bond, which they defined as involving
“issues of mutual trust, acceptance and confidence” (Horvath & Greenberg, 1986, p. 224).
Attachment Theory
Bowlby (1979) defined attachment behavior as any action that seeks to attain proximity
to a preferred individual, who is considered stronger and more capable of coping with the world.
Through early attachment experiences, the individual develops internal working models, or a set
of beliefs about the extent to which the individual is lovable and others are trustworthy
(Neswald-NcCalip, 2001). Although attachment-based working models of self and others are
subject to revision over time, they continue to be salient throughout the life course and are often
generalized to alternative attachment figures in new relational contexts (Bowlby, 1988). This is
particularly true in times of stress or emergency, which is precisely when individuals who have
had negative experiences with early caregivers may exhibit signs of insecure or pathological
attachment. Based on the idea that individuals seek security and comfort in attachment
relationships (Ainsworth, 1989), the principles of attachment theory have been expanded beyond
the infant-caregiver bond to adult romantic relationships (Hazan & Shaver, 1987), therapistclient relationships (Dolan, Arnkoff & Glass, 1993; Dozier, Cue & Barnett, 1994; Farber, Lippert
& Nevas, 1995; Ligiero & Gelso, 2002), and supervisory relationships (Neswald-McCalip, 2001;
Pistole & Watkins, 1995; Watkins, 1995; White & Queener, 2003).
Attachment and Supervision
7
Hazan and Shaver (1987) argued that romantic love involves affective bonds rooted in
attachment processes. Based on Ainsworth, Blehar, Waters and Wall’s (1978) description of
three infant attachment classifications, Hazan and Shaver (1987) developed a forced-choice, selfreport measure of attachment consisting of short paragraphs identifying relationship features
characteristic of the three primary attachment styles. Bartholomew (1990) subsequently
expanded Hazan and Shaver’s work, adding a fourth category of adult attachment. Using
Bowlby’s (1973) notion that attachment processes are reflected in mental representations of the
self and other important people, Bartholomew conceptualized adult attachment style as an
interaction between positive or negative internal working models of self and other. Working
models of self combine with working models of other to define four adult attachment styles: (a)
Secure adults have positive self and other models, so the self is seen as worthy of love and others
are seen as trustworthy and available, (b) Dismissing-avoidant adults have a positive model of
self, but they have a negative model of others so that they view others as unavailable or rejecting,
(c) Preoccupied adults have a negative model of self and view themselves as unlovable and
unworthy, but they have a positive model of others, and (d) Fearful-avoidant adults have a
negative model of both self and other. Based on this new four-category typology, Bartholomew
and Horowitz (1991) developed a measure of adult attachment called the Relationship
Questionnaire, which requires participants to self-select one of four paragraphs that best
describes their relational style and also to rate each paragraph on a 7-point Likert scale according
to how well it describes them.
Bowlby (1977, 1980) also suggested that there were distinct forms of pathological
attachment behavior: compulsive self-reliance, compulsive care-giving, angry withdrawal and
compulsive care-seeking. Compulsive self-reliance is a pattern of behavior that involves doing
Attachment and Supervision
8
everything for oneself whatever the conditions. Bowlby theorized that compulsively self-reliant
individuals experience poor early attachment relationships and subsequently develop a tendency
to inhibit attachment feelings and behavior, demonstrating deep mistrust and fear of depending
on others. As an extreme form of avoidant attachment behavior, compulsive self-reliance
involves self-sufficiency and distancing behavior. In contrast, compulsive care-giving is reflected
in a pattern of exclusively taking on care-giving roles while simultaneously being unable or
unwilling to receive care. Although individuals who develop a compulsive care-giving pattern
have frequent feelings of martyrdom and self-sacrifice, they nevertheless prioritize others needs
before their own whether or not it is requested or even desired by the recipient (West & SheldonKeller, 1994). Compulsive care-seeking, which is analogous to extremely anxious attachment as
it was first discussed by Bowlby (1977), involves a pattern of overactive seeking and excessive
reliance on attachment figures for care and assistance. Lastly, angry withdrawal was identified
by Bowlby as a pattern of negative reactions (e.g., anger, withdrawal) that surface when the
attachment figure is perceived as unavailable or unresponsive.
West and Sheldon-Keller (1994) developed the Reciprocal Attachment Questionnaire
(RAQ) to measure pathological attachment behavior. Although limited, research using the RAQ
has addressed several psychological issues with different populations. Using a sample of
undergraduate college students, Lapsley, Varshney, and Aalsma (2000) demonstrated that RAQ
pathological attachment scales were significantly related in theoretically expected ways with
adult attachment style, assessed by Bartholomew and Horowitz’s (1991) Relationship
Questionnaire, although the two instruments appeared to tap different aspects of the attachment
system. Secure adult attachment style was negatively correlated with compulsive self-reliance,
and moderately positively correlated with compulsive care-giving. Fearful attachment was
Attachment and Supervision
9
positively correlated with angry withdrawal and compulsive self-reliance, while dismissing
attachment was negatively correlated with compulsive care-seeking and positively correlated
with compulsive care-giving. Preoccupied attachment was positively associated with compulsive
care-giving, angry withdrawal, and compulsive care-seeking. Furthermore, RAQ pathological
attachment behavior was significantly associated with adjustment problems, pathology of
separation-individuation, and psychopathological symptoms.
In another study examining pathological attachment in cases of spousal bereavement,
Field and Sundin (2001) found that compulsive care-seeking was correlated with an inability to
cope and the experience of emotional numbness. Compulsive self-reliance was correlated with
positive thoughts about the spouse, whereas angry withdrawal was correlated with negative
thoughts about the spouse; both were correlated with unwanted thoughts and feelings. In
addition, higher scores of psychopathology were associated with both compulsive care-seeking
and angry withdrawal. Similarly, in their study comparing the attachment processes of eating
disorder patients with non-eating disorder patients controls, Ward, Ramsey, Turnbull,
Benedettini, and Treasure (2000) found that eating disorder patients reported more compulsive
care-seeking and compulsive self-reliance behaviors than controls. Overall, the scarce literature
available indicates that individuals demonstrating high levels of pathological attachment are
likely to be insecure (Lapsley et al., 2000) and may be more vulnerable to adjustment difficulties
and psychopathological behaviors (Field & Sundin, 2001; Ward et al., 2000).
Attachment Processes and Clinical Supervision
Pistole and Watkins (1995) have argued that attachment theory is useful for
conceptualizing interpersonal relationships, as well as counseling and supervision processes,
because the theory is based on the principle that emotionally significant relationships play a
Attachment and Supervision 10
critical role in development. The primary goal of clinical supervision is to facilitate the
supervisee’s professional development, which often corresponds with the supervisee’s personal
development. Consequently, important elements of an individual’s attachment style may emerge
in the context of supervision. Similar to the role of the clinician in the therapeutic alliance, the
supervisor should function as a secure base from which supervisees can freely explore and
develop their professional identities as therapists, knowing that a dependable and reliable
supervisor/attachment figure will be available to support them (Pistole & Watkins, 1995).
Yet, even when the supervisor provides safe and reliable support, supervisees’ insecure
attachment processes may surface in various forms during the supervision process (Pistole &
Watkins, 1995). For example, the compulsively self-reliant supervisee may be distant and have
difficulty asking for or receiving supervisory assistance, perhaps even challenging the
supervisor’s authority but feeling wronged by perceived supervisor criticism. Supervisees
manifesting an angry withdrawal pattern may be hyper-vigilant to signs that the supervisor is
inattentive to their needs and paradoxically distance themselves from the supervisor. In contrast,
the compulsively care-giving supervisee may be overly solicitous of the supervisor or clients,
may have difficulty in receiving help from the supervisor, seeking instead to be of assistance to
the supervisor. Conversely, the compulsively care-seeking supervisee may seek to retain
proximity to the supervisor, and may expect the supervisor to take primary responsibility for the
supervisee’s professional development. Because attachment processes are so fundamental and
mostly unconscious, a close examination of the supervisory relationship may uncover
pathological attachment patterns that are affecting the working alliance. According to Watkins
(1995), when a clinical supervisor begins questioning his or her professional judgment,
Attachment and Supervision 11
agonizing over lack of supervisory progress, or experiencing strong emotions in the supervisory
relationship, insecure attachment may be present in the supervisory relationship.
In spite of the increasing number of theorists and researchers who have recommended the
exploration of relational variables in supervision (e.g., Pistole & Watkins, 1995; Epps, 1999),
few studies have examined the role of attachment processes in the context of the clinical
supervisory relationship. Fewer still have used Bordin’s (1983) theory of the working alliance to
operationalize supervision variables and none have explored the potential influence of childhood
attachment experiences or pathological attachment behaviors on supervisory processes.
However, Epps (1999) examined adult attachment style and supervisory rapport among master’s
level supervisee/supervisor dyads using the Supervisory Working Alliance Inventory (Efstation,
1990; SWAI). Epps found that securely attached supervisees rated supervisory rapport higher
than insecurely attached supervisees. Of the supervisory dyads, 39% had secure supervisors and
secure supervisees, 24% had insecure supervisors and supervisees, 20% of the dyads had
insecure supervisors and secure supervisees, and 17% of the dyads had insecure supervisors and
secure supervisees. Findings indicated that dyads in which one or both members were insecurely
attached had greater variations in scores than those dyads in which both were securely attached.
Compared to insecure supervisees, securely attached supervisees perceived a better supervisory
bond with both secure and insecure supervisors, suggesting that securely attached supervisees
may experience more positive relational interactions regardless of their supervisors’ attachment
styles. Similarly, Kim (1997) found that trainee attachment characteristics accounted for 11% of
the variance in the trainees’ satisfaction with supervision, and 19% of the variance in the
trainees’ rating of the quality of the supervisory bond. Results also indicated that trainee
attachment security explained a significant amount of the variance in trainee perceptions of the
Attachment and Supervision 12
supervisor’s interpersonal style. Trainees who worried that others would not care for them as
much as they cared for others tended to be less satisfied with supervision, while trainees secure
in their ability to relate to others were more likely to be satisfied with supervision.
Some research suggests that the supervisor characteristics may be more important than
supervisee characteristics in determining the quality of the supervisory relationship than the
supervisee characteristics. In a sample of 67 supervisor-supervisee dyads, White and Queener
(2003) reported that the supervisors’ ability to make healthy adult attachments predicted both
their own and their trainees’ perceptions of the supervisory working alliance. Furthermore, the
supervisor’s ability to make attachments and accept social support was more predictive of the
quality of the supervisory alliance than the same characteristics in supervisees. Similarly,
Summerel and Borders (1996) found that the strength of the supervisory working alliance was
more highly correlated with the supervisor’s interactional style than individual supervisee
characteristics.
In summary, while empirical evidence is slowly emerging to support the notion that
attachment processes actively contribute to the quality of supervision and the supervisory
relationship, many questions remain regarding the contribution of various attachment processes
to different aspects of the supervisory working alliance. We were interested in several different
attachment constructs that theoretically are likely to influence the supervisory working alliance
(goal, task, and bond), including parent-child attachment experiences, pathological attachment
behaviors, and adult attachment style. Based on the assumption that early childhood attachment
experiences continue to affect relational behavior throughout the lifespan, particularly in
relationships that in many ways mirror the hierarchical organization of the parent-child
relationship, we explored the potential associations between childhood attachment and the
Attachment and Supervision 13
working alliance. Next, we examined Pistole and Watkins’ (1995) hypotheses regarding the
impact of supervisee pathological attachment behavior on the supervisory alliance. Then, we
tested the predictions related to supervisor and supervisee adult attachment styles: (a) we
expected that both supervisor and supervisee adult attachment style would be associated with
aspects of the working alliance, especially supervisory bond, and (b) we expected that working
alliance ratings for supervisory dyads with secure supervisors and supervisees would be higher
than ratings for supervisory dyads composed of at least one (or both) insecure member. After
these preliminary analyses, we explored a path model linking the various attachment processes to
the three components of the working alliance (See Figure 1).
National Supervision Study
Attachment and Supervision 14
Two hundred internship sites from all 50 states were randomly selected from the 2003
APPIC manual. The internship directors of these sites were sent email, which described the study
and requested that they forward to their current interns an invitation to participate and the online
address for the survey. The directors were also asked to reply to the principal investigators with
the number of interns who had received the invitation to participate in the study. Sixty-one
directors complied with this request (30% response rate for directors), indicating that a total of
173 interns received the invitation to participate. The interns were given 2 months to complete
the survey. Approximately one month after the original message, the directors were sent a second
email asking them to remind their interns of the deadline for participation. A total of 87 interns
completed the online survey before the two-month deadline elapsed. After one case was deleted
due to missing data, a final sample of 86 participants produced a 50% response rate for the
interns.
A five-part survey requiring approximately 30 minutes to complete included the
following: (a) a demographic form; (b) the Working Alliance Inventory (WAI; Horvath &
Greenberg, 1994) to assess the task, bond and goal components of the supervisory working
alliance; (c) the Relationship Questionnaire (RQ; Bartholomew & Horowitz, 1991) to assess
respondents’ adult attachment styles (i.e., Secure, Fearful, Dismissing, Preoccupied) and their
perceptions of their supervisors’ attachment styles; (d) the Memory of Parental Styles (MOPS;
Parker et al., 1997) to assess early attachment experiences with parents, including parental
indifference and over-control; and (e) the Reciprocal Attachment Questionnaire (RAQ; West &
Sheldon-Keller, 1994) to assess pathological attachment behaviors, including compulsive selfreliance, compulsive care-giving, compulsive care-seeking and angry withdrawal.
Participants
Attachment and Supervision 15
Survey data were collected from psychology interns at their internship placement in the
Spring of 2004 (N=86). The total sample consisted of 66 women (77%) and 20 men (23%), who
were predominantly of Caucasian (78.2%) descent, ranging in age from 25 to 54 (M = 32.6).
Forty-five percent of respondents were completing their Ph.D.’s in Clinical Psychology, 17.2%
in Counseling Psychology, 2.3% in School Psychology, and 33.3% were completing a Psy.D. in
Clinical Psychology. Approximately 30% of respondents were working at a Veteran’s
Administration Hospital, 13.8% at a Counseling Center, 10.3% at an Armed Forces Medical
Center, 10.3% at a Medical School, 9.2% at State or County Hospital, 8.0% at a Consortium, and
the remaining 17.2% were at federal prisons, state hospitals, or at institutions they identified as
“other.” In order to protect the anonymity of respondents, information regarding the geographic
location or name of facility was not requested. Fifty-four percent of the sample indicated that
they were married, 27.6% were dating someone exclusively, 3.4% were dating someone
casually, and 13.8% were not presently involved.
Findings
Preliminary Analyses
Prior to testing the path model, preliminary analyses were conducted to provide
descriptive statistics (See Table 1), explore interrelations among attachment constructs, and test
direct associations between attachment processes and the supervisory alliance. Consistent with
previous research using non-clinical samples (e.g., Brennan, Clark & Shaver, 1998; Brennan &
Shaver, 1998), over half (59.9%) of the respondents identified themselves as Secure, 18.4%
Fearful, 9.2% Avoidant, 5.7% Preoccupied, and six participants did not self-identify a
predominant adult attachment style. In terms of the participants’ perceptions of their
supervisor’s attachment style, 50.6% identified their supervisor as Secure, 12.6% Fearful, 19.5%
Attachment and Supervision 16
Avoidant, 12.6% Preoccupied, and four participants did not indicate supervisor attachment style.
Participant adult attachment style was not associated with participant gender, ethnicity, age, or
sexual orientation.
Contrary to expectations, results of Pearson correlations for the three supervisory alliance
scales and both parent-child and pathological attachment scales were non-significant.
Subsequently, a series of MANOVAs were performed to assess the relationship of the
supervisory working alliance (goal, task, bond) to participant attachment style, supervisor
attachment style, and the dyadic configuration of the supervisory alliance (See Table 2).
Although results of the MANOVA examining participant attachment style and the supervisory
alliance were significant, F (81) = 2.30, p < .02, follow-up univariate tests were non-significant.
With respect to perceived supervisor attachment style, MANOVA results indicated that
supervisor attachment style was also significantly related to the supervisory alliance, F (83) =
3021, p = .001. In this case, however, follow-up ANOVAs showed significant associations
between supervisor attachment style and supervision task, F (83) = 4.29, p = .007, as well as the
supervisory bond, F (81) = 5.32, p = .002, but not supervision goal, F (83) = 1.02, p = .38. These
findings indicate that participants who perceived their supervisors to have a Secure attachment
style rated the supervisory task and bond significantly higher than participants who perceived
their supervisors to be Preoccupied or Dismissing. Finally, we conducted a MANOVA
comparing secure-secure supervisory dyads (n = 31), insecure-insecure dyads (n=17), and two
types of mixed dyads composed of either secure participants/supervisees and insecure
supervisors (n = 18) or insecure participants/supervisees and secure supervisors (n = 12). Results
showed that dyadic composition was significantly associated with the supervisory alliance, F
(78) = 3.82, p < .000. Subsequent univariate tests with post-hoc comparisons indicated that
Attachment and Supervision 17
secure-secure dyads had significantly higher scores on supervision task than dyads composed of
a secure participant/supervisee and insecure supervisor, F (78) = 3.29, p = .025. Similarly,
secure-secure dyads and dyads composed of an insecure participant/supervisee and secure
supervisor had significantly higher scores on supervisory bond than dyads with a secure
participant/supervisee and insecure supervisor, F (81) = 5.33, p = .002.
Path Analysis
The Latent Variable Path Analysis with Partial Least squares estimation procedure (PLS)
(Lohmoeller, 1984) was used to examine the association between supervisee attachment
processes and their supervisory relationships. We chose this approach because PLS allows the
examination of hypothesized relationships between theoretical constructs free from the
restrictions of measurement and distributional assumptions (Chin, 1998). Additionally, this
procedure is ideally suited for use on small samples with multiple measures (Cowan, Cohn,
Cowan & Pearson, 1996). In PLS, hypotheses about the directionality of relationships between
latent variables are identified based on knowledge of theory and previous research. Based on
path coefficients and correlations between latent constructs, as well as the variance accounted for
on latent variables, the model(s) explaining the most amount of variance is selected for
interpretation.
For the present analyses, several models were generated based on the theoretical and
empirical literature. Per Falk and Miller’s (1992) recommendations for interpretation of PLS
analyses, manifest variables in the outer model demonstrating a low factor loading (e.g., below
.55) or negative correlation with other manifest variables measuring the same latent construct
were excluded from the analyses. When path coefficients between latent variables in the inner
model did not explain greater than 5% of variance, pathways were deleted and the trimmed
Attachment and Supervision 18
model was re-run. To establish a more stable path coefficient than the initial analysis, the
bootstrapping procedure was run, which involved the model being re-run one thousand times and
calculating the mean path coefficient for each pathway in the model.
Based on the theoretical and empirical literature, an initial model including the constructs
of interest was hypothesized (See Figure 1). The two hypotheses underpinning the initial model
were as follows: 1) supervisee’s parent-child attachment experiences shape the supervisees’ adult
attachment style, which in turn may influence the development of pathological attachment
behavior, and 2) the supervisory bond will predict the goal and tasks of supervision, as well as
the supervisees’ perceptions of his or her supervisor’s attachment style. Based on the results of
analyses using the initial model, several more theoretically-sound models were subsequently
tested. The models exhibiting the maximal fit with the data will be presented. Although PLS
output does not include an overall goodness of fit statistic between the model and the data, we
utilized the formula for an overall F test of the R² statistic recommended by Falk and Miller
(1992) to calculate the goodness of fit statistic. Additionally, we chose to calculate Cohen’s f²
effect size statistic for multiple partial correlations to examine the magnitude of the effect present
in the model. All models presented can be interpreted as representing the pathways from the
supervisee’s parent-child attachment to current adult attachment processes and to the supervisory
relationship.
MODIFY TO INCLUDE PARTICIPANT ATTACHMENT STYLE. In Figure 1, the
latent variables of the supervisee’s experiences of parental attachment (i.e., parental indifference
and over-control) and current pathological attachment (i.e., compulsive self-reliance, care-giving,
care-seeking, and angry withdrawal) are shown to influence the supervisory bond, which in turn
affects the supervisees’ perceptions of the tasks and goals of supervision as well as his or her
Attachment and Supervision 19
perceptions of the supervisor’s attachment style. Results indicated that the supervisees’
experiences of parental indifference predicted the manifestation of compulsive self-reliance (path
weight [pw] = .492), and compulsive care-seeking (pw = .433), as well as angry withdrawal (pw
= .349). However, experiences of parental over-control predicted only compulsive care-seeking
(pw = .249). Next, results indicated that the largest path coefficients leading to the supervisory
bond were compulsive care-giving (pw = .165) and angry withdrawal (pw = .150). Finally, the
path coefficients leading from the supervisory bond to the supervisee’s perceptions of the
supervisor’s attachment style, as well as the tasks and goals of supervision were examined. The
supervisory bond predicted the supervision tasks (pw = .884), supervision goals (pw = .848), and
the supervisee’s perception of the supervisor’s attachment style (pw = .352). However, the F test
for the overall goodness of fit indicates non-significance, F (15, 71) = 10.549, ns.
Figure 2 represents another theoretically-sound model developed to explain the data. Two
pathways leading to the task component of supervision are shown. The first pathway shows the
supervisees’ recollection of indifferent parent-child attachment leading to compulsive selfreliance, whereas the second pathway shows self-reported attachment style leading to
perceptions of the supervisor’s attachment style. As shown in Table 3, the supervisees’ early
attachment experiences of parental indifference predicted the manifestation of supervisee
compulsive self-reliance (path weight [pw] = .43), which then predicted supervisees’ perceptions
of the tasks in supervision (pw = .191). Participants’ self-reported adult attachment styles
(secure, fearful, anxious, and avoidant) predicted their perceptions of their supervisor’s
attachment classification (pw = .359), which then predicted the supervisees’ perception of the
tasks of supervision (pw = .361). Contrary to hypotheses, findings indicated that attachment
processes most strongly predicted the in-session tasks and behaviors of supervision. In turn,
Attachment and Supervision 20
supervisory task ratings strongly predicted the supervisees’ perceptions of shared supervisory
goals (pw = .935), and supervisory bond (pw = .894). According to Cohen’s (1992) delineation
of the magnitude of effect sizes, the effect size of f² = .555 for the second model qualifies as
large. The large effect size in this model indicates that changes in the endogenous latent variables
are likely to cause significant change in the latent variables (i.e., task, goal, bond) to which the
endogenous variable pathways point. The F test for the overall goodness of fit indicates F (6, 79)
= 44.638, p < .0001.
The second model was trimmed to include only those manifest variables or indicators
loading most heavily on the endogenous latent variable constructs. Because the latent variable of
self-reported adult attachment style did not meaningfully predict any aspect of the supervisory
working alliance, it was deleted from the model. As a result, the trimmed model included twelve
indicators on the parental indifference scale of parental attachment, and six indicators on the
compulsive self-reliance scale of pathological attachment. The trimmed model was re-run and
the result is presented in Figure 3. As in the previous model, all pathways met statistical
significance at the .05 level (see Table 4). Results indicated that experiences of parental
indifference predicted compulsive self-reliance in supervisees (pw = .564), which then predicted
their perceptions of the supervisor’s attachment style, (pw = .326). In turn, supervisor
attachment style predicted the evaluation of the in-session tasks in supervision (pw = -.355),
which then predicted the reported quality of the supervisory bond (pw = .874), and goals of
supervision (pw = .945). The third model evidences a larger effect size than the previous model
at f² = .749. The F test for the overall goodness of fit, F (5, 80) = 59.972, p < .0001, indicates that
the model is a very good fit with the data.
Implications
Attachment and Supervision 21
The main premise of attachment theory is that attachment relationships form the
foundation of ongoing development, and that attachment processes are salient throughout the
lifespan (Bowlby, 1988). Because the purpose of supervision is to facilitate the growth and
development of the supervisee, the supervisory relationship can incorporate important elements
of other significant relationships and therefore elicit attachment responses (Pistole & Watkins,
1995), especially when involving issues around autonomy, authority and individuation (Watkins,
1995). The findings of the present study support the theoretical and empirical literature (Epps,
1999; Kim, 1997; Neswald-McCalip, 2001; Pistole & Watkins, 1995; Watkins, 1995; White &
Queener, 2003) and extend our understanding of attachment processes at work in the supervisory
working alliance by providing empirical evidence that attachment constructs representing diverse
levels of experience in childhood and adulthood are relevant to clinical supervision even at
advanced stages of training and clinical experience. Specifically, the final path model showed
that the a childhood history characterized by parental indifference leads to compulsively selfreliant attachment behavior, which influences perceptions of supervisor attachment styles, which
in turn first affects the supervisory task, then the supervisory bond and goal. This model
provides a logical representation of the developmental chain that contributes to the supervisee’s
experience in the supervisory alliance. The following discussion will consider the implications of
current findings for clinical training and supervision.
Contrary to expectations, current results indicated that neither parent-child attachment
experiences nor pathological attachment behaviors were significantly correlated with ratings for
supervisory task, goal, and bond. In addition, while an overall MANOVA was significant, selfrated adult attachment style was not significantly related to any of the supervisory alliance scales
in follow-up univariate analyses. Because MANOVA takes into account correlations among
Attachment and Supervision 22
variables and considers differences on all variables jointly whereas univariate tests do not
(Stevens, 1996), this finding suggests that the significant comparison may involve some linear
combination of the three supervisory variables that does not emerge when each scale is examined
separately. That is, it appears that the supervisee’s adult attachment style is not directly
associated with a specific aspect of the supervisory alliance, but rather is associated with an
overall view of the supervisory process that integrates task, goal, and bond, representing the “big
picture,” so to speak. An informal examination of scale means in search of patterns among the
attachment groups showed that the Dismissing group of supervisees had the lowest mean ratings
for all three supervisory alliance scales; they reported less effective in-session task-related
behaviors, less consensus regarding supervision goal, and an inferior bond with their supervisor.
Consistent with theoretical expectations for Dismissing adults to minimize the importance of
relationships, viewing themselves as “perfect” and others as unworthy (Bartholomew, 1990;
Bowlby, 1973), this finding suggests that Dismissing supervisees may devalue the fundamental
nature of the supervisory alliance, possibly considering themselves “above” supervision and
acting independently without regard to supervisory guidance. Supervisors working with these
individuals may initially be pleased with the independence they display, but should be cautious
in allowing too much independence and may wish to check in regularly to ascertain if the
supervisee has incorporated the supervisor’s suggestions in work with clients. Special efforts to
strengthen the supervisory relationship and sensitively challenge noncompliance may be
necessary to avoid defensiveness and promote a good working alliance with Dismissing
supervisees.
Although analyses showed no direct linear correlation between the supervisory alliance
and parent-child attachment experiences or pathological attachment behavior in adulthood,
Attachment and Supervision 23
results of the path analyses indicated that these two attachment constructs contribute indirectly to
the quality of the supervisory alliance via their association with the supervisees’ perceptions of
their supervisors’ attachment styles. Current findings are consistent with theoretical assumptions
regarding the importance of the parent-child attachment relationship in the development of adult
attachment behavior. In particular, the model supports theoretical predictions (e.g., Bowlby’s,
1977; West & Sheldon-Keller, 1994) that early parental indifference and/or rejection play a
critical role in promoting compulsively self-reliant behavior. Attachment processes may be
particularly relevant to the supervisory process when they are pathological in nature (Watkins,
1995). Of the four pathological attachment styles identified by Bowlby (1977), the present
results suggest that compulsive self-reliance may have the most impact on the supervisory
relationship. This finding is consistent with the pattern of alliance ratings described above for
Dismissing individuals, who are theoretically more likely to demonstrate compulsive selfreliance (West & Sheldon-Keller, 1994). According to Watkins (1995), the compulsively selfreliant supervisee refuses help, is defiant towards authority (e.g., disparaging the supervisor,
disregards the supervisor’s comments) and actively distances him- or herself from the
supervisory relationship. Compulsive self-reliance may have the most impact on the supervisory
relationship because it most directly challenges the trainees’ ability to be open to the supervisor’s
feedback. Compulsively self-reliant supervisees may experience extreme discomfort with
evaluation and alleviate this discomfort by devaluing the quality of supervision they are
receiving. In contrast, compulsive care-seekers or care-givers may not be as inclined to discount
or subvert the supervision process. Because self-reliance is an adaptive and valued quality in
many graduate programs, pathological forms may be less readily perceptible to supervisors and
training directors until it overtly intrudes in the supervision process. Once recognized, given
Attachment and Supervision 24
childhood experiences with parental indifference and/or rejection, it would be extremely
important for the supervisor to avoid repeating these interaction patterns in the supervisory
relationship. Direct criticism can be eschewed in favor of constructive feedback, and a balance
between the consistent demonstration of empathic guidance and encouragement of the
supervisee’s autonomy can be sought.
Consistent with findings of White and Queener (2003), results of the current study
highlighted the importance of the supervisor’s role in setting the tone of the supervisory working
alliance. In addition to path analyses, both multivariate and two univariate follow-up analyses for
supervisor attachment style were significant. Supervisees who saw their supervisors as securely
attached tended to evaluate the supervisory task and bond more positively than supervisees who
saw their supervisors as preoccupied or dismissing. Similarly, results showed that secure-secure
dyads were associated with higher ratings on the supervisory task than mixed dyads composed of
secure supervisees and insecure supervisors. Insecure participants with secure supervisors rated
the supervisory task only slightly lower than their secure counterparts. It is interesting that
ratings by secure participants represented the polar extremes of the quality of the supervisory
alliance. Perhaps secure individuals are more attuned to interpersonal behaviors, and thus more
likely to notice and respond to the attachment styles of others. Importantly, regardless of their
own attachment style, supervisees with secure supervisors rated the supervisory bond more
positively than supervisees with insecure supervisors.
Results of the path analysis also indicated that the supervisee’s assessment of his or her
supervisor’s attachment style predicted the supervisee’s assessment of the task of supervision,
and indirectly, the supervisory bond and goal. Consistent with preliminary analyses and
previous research (Summerel & Borders, 1996; White & Queener, 2003), the final model
Attachment and Supervision 25
indicated that rather than the supervisee’s own attachment style, it was his/her perception of the
supervisor’s attachment style that had the most direct impact on the supervisory alliance. From a
theoretical and applied standpoint, this is an extremely important point that should not be
overlooked. Previous theorists have focused on the role of supervisee attachment behavior (e.g.,
Watkins, 1995); however, current results suggest that it is the supervisor’s attachment style that
is critical to the success of the supervision process. Specifically, like the parent-child attachment
relationship, the supervisory relationship is inherently a hierarchical one, in which the supervisor
has more power in the relationship than the supervisee. Attachment researchers have
documented the process of intergenerational transmission whereby a parent’s attachment style
influences their caregiving behavior (e.g., responsiveness vs. inconsistency/role-reversal or
rejection/neglect), which in turn contributes to the development of secure or insecure attachment
in their infants (see van IJzendoorn, 1995 for meta-analytic review). Like a parent, the
supervisor’s attachment style sets the tone of the relationship via its impact on the supervisor’s
in-session task-related behavior, which in turn influences the attachment bond with the
supervisee. By virtue of his/her greater power and knowledge, the bulk of responsibility for the
quality of the supervisory alliance lies with the supervisor, not the supervisee.
Contrary to theoretical hypotheses that attachment processes would most strongly predict
the supervisory bond, in-session behaviors and tasks were found to strongly predict the ability of
the supervisee to build a bond with his or her supervisor, and to subjectively assess the extent to
which the supervisory relationship is goal-oriented. It is possible that attachment processes may
be more directly related to the supervisory bond among less advanced trainees, who presumably
require more nurturance and interpersonal support than the clinical interns in this sample. An
alternative explanation for this finding is that the supervisor’s in-session task-related behavior is
Attachment and Supervision 26
analogous to parental caregiving behavior, which is assumed to mediate the parents’ internal
working models of attachment and attachment behavior in their infants. Current findings indicate
that in-session behaviors are reflections of the supervisors’ attachment style, and as such may be
characterized by “caregiving” behaviors that indicate whether the supervisor can be trusted to
respond sensitively to supervisee cues, or conversely demonstrate inconsistency/role-reversal or
rejection/neglect. For example, if a supervisee perceives his or her supervisor as being
disorganized or inattentive to the agenda during session, the trainee might perceive the
supervisor as untrustworthy or unable to supervise effectively, which would negatively affect the
supervisory bond and goal. One implication of this finding is that the supervisor should closely
monitor their own in-session behaviors because the identification of supervisory tasks that are
tailored to supervisee needs and the manner in which those tasks are performed will provide the
foundation on which the supervisory bond will be built and the supervisory goals agreed upon.
Current findings underscore the need for supervisors to attend to specific aspects of how time in
supervision is spent (e.g., the number of times supervisors watch tapes, the quantity of clients
discussed, the clarity with which expectations are conveyed and responsibilities discussed, the
agreement upon and adherence to an agenda for each supervision meeting), which can provide
clarity of purpose within and across supervision sessions that may improve the supervisees’
perception of the supervision sessions as both structured and useful, contributing to a strong
supervisory bond and consensus regarding goals.
Limitations and Conclusions
Current findings should be interpreted in light of several limitations. Most important are
limitations related to sample selection. In order to obtain a national sample, supervisors were not
included in the study and therefore an important element of the supervisory dyad is absent from
Attachment and Supervision 27
these analyses. Because supervisors were excluded from our study, we were only able to include
an indirect measure of the supervisor’s attachment style (i.e., asking the supervisee to speculate
about his or her supervisors’ attachment orientation). In addition, at the recommendation of
several previous researchers interested in attachment and supervision, we focused on supervisees
at advanced levels of training (i.e., internship). As a result, these findings may not be
generalizable to trainees with less clinical and supervision experience. Future research should
include both members of the supervisory dyad and clinicians at all levels of training with their
respective supervisors to gain a more comprehensive picture of the role of attachment processes
in the context of clinical supervision. Another limitation of the present study involves
measurement issues. First, self-report measures were the only type of measurement utilized,
which may have introduced subjective bias. Researchers have recommended caution when
relying on self-reports with Dismissing/avoidant individuals, who may have overly positive selfimages and defensive strategies designed to minimize emotion and self-evaluation, making selfdisclosure less likely (Dozier, 1990; Dozier, Stevenson, Lee, & Villegan, 1991; Kobak & Sceery,
1988; Pianta et al., 1996; Riggs & Jacobvitz, 2002; Rothbard & Shaver, 1994). There are also
some important limitations inherent in the correlational and Partial Least Squares data analytic
strategies we chose for the present analyses. Because PLS makes no measurement, distributional
or sample size assumptions, like the correlational findings, these results cannot be interpreted to
suggest causality. The goal of PLS is to make modest predictions rather than causal claims.
Despite these limitations, the current study shed some light on the role of attachment
processes in clinical supervision and is the first study to examine the associations of early
attachment experiences and pathological attachment behavior to the quality of the supervisory
alliance. Path analyses generated a model that showed a logical predictive progression of
Attachment and Supervision 28
different attachment processes leading to the supervisory alliance. Results suggested that early
attachment experiences with parents contribute to pathological attachment behavior that
influences the supervisee’s perception of their supervisor’s attachment style, which in turn is a
factor in the perceived quality of in-session task-related behaviors that affect the supervisory
bond and goal. Of particular significance, findings emphasize the importance of the supervisor’s
attachment style in the functioning of the supervisory alliance. Based on these findings, future
theory and research concentrating on supervisor qualities that contribute to the ability to create
an effective working alliance may provide us with a better understanding of the supervision
process and improve training. Moreover, current results highlight the critical importance of
formal graduate training in the provision of clinical supervision, a course that is still not required
in many graduate clinical and counseling programs. Formal supervision coursework, as well as
self-exploration or personal psychotherapy, which educates trainees regarding the potential
impact of personal qualities like attachment style on their supervision skills will improve clinical
training and may allow some clinicians to override an insecure attachment style in the service of
providing effective supervision.
Attachment and Supervision 29
References
Ainsworth, M. (1989). Attachments beyond infancy. American Psychologist, 44, 709–716.
Andersson, P., & Eisemann, M. (2004). Parental rearing and substance related disorders—A
multi-factorial controlled study in a Swedish sample. Clinical Psychology &
Psychotherapy, 11, 392–400.
Andersson, P., & Perris, C. (2000). Perceptions of parental rearing and dysfunctional attitudes:
The link between early experiences and individual vulnerability? Nordic Journal of
Psychiatry, 54, 405–409.
Bartholomew, K. (1990). Avoidance of intimacy: An attachment perspective. Journal of Social
and Personal Relationships, 7, 147–178.
Bartholomew, K., & Horowitz, L. (1991). Attachment styles among young adults: A test of the
four-category model. Journal of Personality andSocial Psychology, 61, 226–245.
Borders, E. (1989). A pragmatic agenda for developmental supervision research. Counselor
Education & Supervision, 29, 16–24.
Bordin, E. (1983). Supervision in counseling: II Contemporary models of supervision: A
working alliance based model of supervision. Counseling Psychologist, 11, 35–42.
Bowlby, J. (1973). Attachment and loss: Vol. 2. Separation. New York: Basic Books.
Bowlby, J. (1977). The making and breaking of affectional bonds: I Aetiology and
psychopathology in the light of attachment theory. British Journal of Psychiatry, 130,
201–210.
Bowlby, J. (1979). On knowing what you are not supposed to know and feeling what you are not
supposed to feel. Canadian Journal of Psychiatry, 25, 403–408.
Attachment and Supervision 30
Bowlby, J. (1988). Developmental psychiatry comes of age. American Journal of Psychiatry,
145, 1–10.
Brennan, K., Clark, C., & Shaver, P. (1998). Self-report measurement of adult attachment: An
integrative overview. In J. Simpson & W. Rholes (Eds.), Attachment theory and close
relationships (pp. 46–76). New York: Guilford Press.
Brennan, K., & Shaver, P. (1998). Attachment styles and personality disorders: Their
connections to each other and to parental divorce, parental death, and perceptions of
parental care-giving. Journal of Personality, 66, 835–878.
Carnelley, K., Peitromonaco, P., & Jaffe, K. (1994). Depression, working models of others, and
relationship functioning. Journal of Personality and Social Psychology, 66, 127–140.
Chin, W. (1998). The partial least squares approach to structural equation modeling. In G. A.
Marcoulides (Ed.), Modern methods for business research (pp. 295–336). Mahwah, NJ:
Erlbaum.
Chin, W., & Newsted, P. (1998). Structural equation modeling analysis with small samples using
partial least squares. In R. H. Hoyle (Ed.), Statistical strategies for small sample research
(pp. 307–341). Thousand Oaks, CA: Sage.
Cobb, R., Davila, J., & Bradbury, T. (2001). Attachment security and marital satisfaction: The
role of positive perceptions and social support. Personality and Social Psychology
Bulletin, 27, 1131–1143.
Cowan, P., Cohn, D., Cowan, C., & Pearson, J. (1996). Parents’ attachment histories and
children’s externalizing and internalizing behaviors: Exploring family systems models of
linkage. Journal of Consulting and Clinical Psychology, 64, 53–63.
Dolan, R., Arnkoff, D., & Glass, C. (1993). Client attachment style and the psychotherapist’s
Attachment and Supervision 31
interpersonal stance. Psychotherapy: Theory, Research, Practice, Training, 30, 408–412.
Dozier, M., Cue, K., & Barnett, L. (1994). Clinicians as caregivers: Role of attachment
organization in treatment. Journal of Consulting and Clinical Psychology, 62, 793–800.
Ellis, M., & Douce, L. (1994). Group supervision of novice clinical supervisors: Eight recurring
issues. Journal of Counseling & Development, 72, 520–525.
Epps, M. (1999). The effect of attachment styles on the working alliance in counselor
supervision. (Doctoral dissertation, Texas A&M. University, 1999). Dissertation
Abstracts International, 60, 5833.
Farber, B., Lippert, R., & Nevas, D. (1995). The therapist as attachment figure. Psychotherapy:
Theory, Research, Practice, Training, 32, 204–212.
Feeney, J. A. (1999). Adult romantic attachment and couple relationships. In J. Cassidy & P.
Shaver (Ed.), Handbook of attachment: Theory, research, and clinical applications (pp.
355–377). New York: Guilford Press.
Field, N., & Sundin, E. (2001). Attachment style in adjustment to conjugal bereavement. Journal
of Social and Personal Relationships, 18, 347–361.
Griffin, D., & Bartholomew, K. (1994). Models of the self and other: Fundamental dimensions
underlying measures of adult attachment. Journal of Personality and Social Psychology,
67, 430–445.
Hanson, W., Curry, K., & Bandalos, D. (2002). Reliability generalization of working alliance
inventory scale scores. Educational and PsychologicalMeasurement, 62, 659–673.
Hazan, C., & Shaver, P. (1987). Romantic love conceptualized as an attachment process. Journal
of Personality and Social Psychology, 52,511–524.
Hess, A. (1987). Psychotherapy supervision: Stages, Buber, and a theory of relationship.
Attachment and Supervision 32
Professional Psychology: Research and Practice, 18, 251–259.
Horvath, A., & Greenberg, L. (1986). The development of the working alliance inventory. In L.
Greenberg & W. Pinsof (Eds.), Psychotherapeutic process: A research handbook (pp.
529–556). New York. Guilford Press.
Horvath, A., & Greenberg, L. (1989). Development and validation of the working alliance
inventory. Journal of Counseling Psychology, 36,223–233.
Horvath, A. O., & Greenberg, L. S. (1994). The working alliance: Theory, research, and
practice. Oxford, England: Wiley.
Kim, S. (1998). Influence of trainee attachment on satisfaction with supervision, the supervisory
relationship, and perceptions of supervisory style. Unpublished doctoral dissertation,
University of Maryland.
Kirkpatrick, L., & Hazan, C. (1994). Attachment styles and close relationships: A four-year
prospective study. Personal Relationships, 1, 123–142.
Lapsley, D., Varshney, N., & Aalsma, M. (2000). Pathological attachment and attachment style
in late adolescence. Journal of Adolescence, 23, 137–155.
Lichtenberg, J., & Goodyear, R. (2000). The structure of supervisor–supervisee interactions.
Clinical Supervisor, 19, 1–24.
Ligiero, D., & Gelso, C. (2002). Countertransference, attachment, and the working alliance: The
therapist’s contribution. Psychotherapy: Theory, Research, Practice, Training, 39, 3–11.
Lohmo¨ller, J.-B. (1989). Latent variable path modeling with partial least squares. Heidelberg,
Germany: Physica-Verlag.
Mikulincer, M., Florian, V., & Weller, A. (1993). Attachment styles, coping strategies, and
posttraumatic psychological distress: The impact of the Gulf War in Israel. Journal of
Attachment and Supervision 33
Personality and Social Psychology, 64, 817–826.
Murray, S., Holmes, J., & Griffin, D. (1996). The self-fulfilling nature of positive illusions in
romantic relationships: Love is not blind, but prescient. Journal of Personality and Social
Psychology, 71, 1155–1180.
Neswald-McCalip, R. (2001). Development of the secure counselor: Case examples supporting
Pistole & Watkins’ (1995) discussion of attachment theory in counseling supervision.
Counselor Education & Supervision, 41, 18–27.
Parker, G., Roussos, J., Hadzi-Pavlovic, D., Mitchell, P., Wilhelm, K., & Austin, M. (1997). The
development of a refined measure of dysfunctional parenting and assessment of its
relevance in patients with affective disorders. Psychological Medicine, 27, 1193–1203.
Parker, G., Tupling, H., & Brown, L. (1979). A parental bonding instrument. British Journal of
Medical Psychology, 52, 1–10.
Pistole, C., & Watkins, C. E. (1995). Attachment theory, counseling process, and supervision.
The Counseling Psychologist, 23, 457–478.
Ramos-Sanchez, L., Esnil, E., Goodwin, A., Riggs, S., Touster, L., Wright, L., et al. (2002).
Negative supervisory events: Effects on supervision satisfaction and supervisory alliance.
Professional Psychology: Research and Practice, 33, 197–202.
Scharfe, E., & Bartholomew, K. (1998). Do you remember? Recollections of adult attachment
patterns. Personal Relationships, 5, 219–234.
Simoneau, T., & Le, P. L. (Eds.). (2003). The internship and postdoctoral programs in
professional psychology. Washington, DC: Association of Psychology Postdoctoral and
Internship Centers.
Stevens, J. (1996). Applied multivariate statistics for the social sciences (3rd ed.). Mahwah, NJ:
Attachment and Supervision 34
Erlbaum.
Stoltenberg, C., & Delworth, U. (1987). Supervising counselors and therapists: A developmental
approach. San Francisco: Jossey-Bass.
Summerel, M., & Borders, L. (1996). Addressing personal issues in supervision: Impact of
counselors’ experience level on various aspects of the supervisory relationship.
Counselor Education and Supervision, 35, 268–286.
Twohey, D., & Volker, J. (1993). Listening for the voices of care and justice in counselor
supervision. Counselor & Education, 32, 3–17.
van IJzendoorn, M. (1995). Adult attachment representations, parental responsiveness, and infant
attachment: A meta-analysis on the predictive validity of the Adult Attachment Interview.
Psychological Bulletin, 117, 387–403.
Ward, A., Ramsay, R., Turnbull, S., Benedettini, M., & Treasure, J. (2000). Attachment patterns
in eating disorders: Past in the present. International Journal of Eating Disorders, 28,
370–376.
Watkins, C. E. (1990). The separation/individuation process in psychotherapy supervision.
Psychotherapy: Theory, Research, Practice, Training,27, 202–209.
Watkins, C. E. (1995). Pathological attachment styles in psychotherapy supervision.
Psychotherapy, 32, 333–340.
West, M., & Sheldon-Keller, A. (1994). Patterns of relating: An adult attachment perspective.
New York: Guilford Press.
White, V., & Queener, J. (2003). Supervisor and supervisee attachments and social provisions
related to the supervisory working alliance. Counselor Education & Supervision, 42,
203–218.
Attachment and Supervision 35
Table 1
Means, Standard Deviations, and Correlations for Subscales
Subscale Correlations
(n = 86)
Mean
1
2
3
4
5
6
7
8
9
__
.46***
-.31**
.12
-.49***
-.30**
-.02
-.10
-.11
__
__
-.18
.24*
-.25*
.02
-.04
-.06
-.07
__
__
__
-.02
.45***
.36***
-.03
.05
.10
__
__
__
__
-.22*
.10
.07
.02
-.07
__
__
__
__
__
.28**
-.16
-.04
.06
__
__
__
__
__
__
-.08
.02
-.05
__
__
__
__
__
__
__
__
__
__
__
__
__
__
__
.81***
__
__
__
__
__
__
__
__
__
(SD)
1. Parent
6.53
Indifference
(6.95)
2. Parent
11.99
Overcontrol
(5.48)
3. Angry
29.06
Withdrawal
(4.52)
4. Compulsive
20.77
Care-Giving
(2.18)
5. Compulsive
29.14
Self-Reliance
(3.82)
6. Compulsive
29.67
Care-Seeking
(3.46)
7. Goal
44.13
.84*** .76***
(6.02)
8. Task
48.29
(6.29)
9. Bond
60.90
(10.74)
* p < .05, ** p < .01, *** p < .001
Attachment and Supervision 36
Table 2
MANOVAs and Follow-up Tests for Attachment Style and Supervisory Alliance
Participant Attachment Style
N = 81
Supervisory Alliance
Secure
Fearful
Preoccupied
M (SD)
M (SD)
M (SD)
Dismissing
F
M (SD)
MANOVA
2.30*
Goal
43.12 (6.26)
46.62 (3.40)
46.35 (3.89)
42.75 (9.25)
1.73
Task
49.38 (6.19)
48.87 (4.13)
47.83 (5.64)
45.62 (11.65)
.49
Bond
60.72(11.29)
63.56 (5.27)
61.33 (8.14)
54.12 (17.25)
1.32
Supervisor Attachment Style
N = 83
Secure
Fearful
Preoccupied
M (SD)
M (SD)
M (SD)
Dismissing
MANOVA
3.21***
Goal
45.06 (4.80)
45.27 (4.47)
43.18 (6.79)
42.35 (9.01)
1.02
Task
50.16 (4.27)a
49.36 (4.65)
45.00 (6.68)b
44.94 (9.59)b
4.29**
Bond
64.57 (6.82)a
60.74 (11.80)
57.64 (9.49)b
53.47 (15.55)b
5.32**
Dyadic Configuration
N = 78
Insecure
SecP/InsS
InsP/SecS
Secure
M (SD)
M (SD)
M (SD)
F
M (SD)
MANOVA
3.82***
Goal
44.48 (5.12)
44.94 (6.79)
41.05 (7.91)
46.32 (3.88)
2.17†
Task
50.31 (4.18)a
46.59 (8.22)
44.94 (8.22)b
49.51 (4.74)
3.29*
Bond
64.67 (7.76)a
57.95 (13.15)
53.44 (13.69)b
64.29 (4.19)a
5.32*
Note. Means with different subscripts differ significantly at p < .05 or greater. SecP = Secure
Participant/Supervisee, InsS = Insecure Supervisor, InsP = Insecure Participant/Supervisee, SecS = Secure
Supervisor.
†
F
M (SD)
p < .10, * p < .05, ** p < .01, *** p < .001
Attachment and Supervision 37
Table 3
Path Coefficients, Standard Deviations and Confidence Intervals for Predictors of Supervisory
Task, Bond, and Goal in Model 2 (N = 86)
______________________________________________________________________________
Pathway
Path Coefficient
SD
95% Confidence Interval
Self Reliance to Task
.209*
.104
.1, .417
Supervisor Attachment to Task
-.350*
.108
-.566, -.134
Task to Bond
.894*
.031
.832, .956
Task to Goal
.928*
.017
.894, .962
Parent Indifference to Self Reliance
-.431*
.139
-.153, .-709
Attachment to Supervisor Att
.377*
.131
.246, .508
______________________________________________________________________________
Note. * = pathway met statistical significance at the .05 level.
Attachment and Supervision 38
Table 4
Summary of Path Coefficients, Standard Deviations and Confidence Intervals for Variables
Predicting the Supervisory Task, Bond, and Goal in Model 3 (N = 86)
______________________________________________________________________________
Pathway
Path Coefficient
SD
95% Confidence Interval
Self Reliance to Task
.326*
.105
.116, .536
Task to Bond
.876*
.034
.808, .944
Task to Goal
.945*
.011
.923, .967
Task to Supervisor Attachment
-.372*
.093
-.186, -.558
Parent Attachment to Self Reliance
-.564*
.070
-.424, -.704
______________________________________________________________________________
Note. * = pathway met statistical significance at the .05 level.
Attachment and Supervision 39
Figure 1. Latent Variable Path Analysis (path model 1)
Figure 1. Latent variable path analysis: Hypothesized full model of the influence of parent-child
attachment and pathological attachment on the supervisee’s perception of the supervisory bond
which in turn influences how the supervisee perceives his or her supervisor’s attachment style, as
well as the tasks and goals of supervision.
Attachment and Supervision 40
Figure 2. Latent Variable Path Analysis (path model 2)
Figure 2. Latent variable path analysis: Influence of parental attachment and pathological
attachment, as well as self-reported attachment style and the supervisee’s perception of the
supervisor’s attachment style on the supervisory task, goal, and bond.
Attachment and Supervision 41
Figure 3. Latent Variable Path Analysis (path model 3)
Figure 3. Latent variable path analysis: Influence of parental indifference and compulsive selfreliance on the supervisory task, goal, bond and perceptions of the supervisor’s attachment style.