Training and Education in Professional Psychology

© 2018 American Psychological Association 2018, Vol. 12, No. 3, 188 –195
1931-3918/18/$12.00 http://dx.doi.org/10.1037/tep0000201

Supervisee Perception of Power in Clinical Supervision: The Power

Dynamics in Supervision Scale
Ryan M. Cook W. Bradley McKibben
The University of Alabama Nova Southeastern University

Stefanie A. Wind
The University of Alabama
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Power dynamics between supervisors and supervisees are inherent as a result of the hierarchical structure
This document is copyrighted by the American Psychological Association or one of its allied publishers.

of supervision. Failure to adequately attend to issues of power in supervision can result in ineffective or
even harmful supervision. Currently, supervisors do not have an objective measure of power dynamics
within the supervisory relationship or a tool that allows for ongoing measurement and discussion of
power in supervision. The authors developed the Power Dynamics in Supervision Scale (PDSS) and
investigated its psychometric properties with a sample of 267 supervisees. Results from the polytomous
Rasch model indicated that the PDSS items (n ⫽ 16) explained 51.77% of the variance as a single factor.
Item locations on the logit scale indicate supervisees perceive differing levels of power, meaning that
PDSS items adequately identified differences in power perceptions among supervisees. Supervisees
perceived themselves as possessing the most power on maintaining healthy boundaries with their
supervisors, a willingness to feel vulnerable in supervision, and feeling empowered in supervision. They
also perceived their supervisors as possessing the most power in identifying interventions to use with
clients, setting goals for supervision, and providing feedback about clinical skills in supervision.
Implications for supervisors and supervision research are provided in light of the findings.

Keywords: supervision, power dynamics, Rasch modeling, psychology

Supplemental materials: http://dx.doi.org/10.1037/tep0000201.supp

Power dynamics in supervision have been routinely discussed in
clinical supervision literature (e.g., Murphy & Wright, 2005; Por-
ter & Vasquez, 1997; Szymanski, 2003, 2005). The power dynam-
ics between supervisors and supervisees inherently exist as a result
This article was published Online First June 28, 2018.
RYAN M. COOK is an Assistant Professor in the Department of Educa-
tional Studies in Psychology, Research Methodology, and Counseling at
the University of Alabama. His research interests include intentional non-
disclosure, counselor decision-making, and clinical supervision.
W. BRADLEY MCKIBBEN is an Assistant Professor in the Department of
Counseling at Nova Southeastern University. His research interests include
professional counselor development and relational/cultural issues in clini-
cal supervision. He has supervised master’s-level counseling practicum and
internship students through a variety of supervision modalities, and he has
taught and supervised doctoral-level counseling supervisors-in-training.
STEFANIE A. WIND is an Assistant Professor of Educational Measure-
ment in the Department of Educational Research at The University of
Alabama. She conducts methodological and applied research in the area of
psychometrics, with an emphasis on issues related to raters, rating scales,
and parametric and nonparametric item response theory models for ratings.
THE AUTHORS WOULD LIKE to acknowledge the research assistance pro-
vided by Bryant Abbott and Morgan Drake, The University of Alabama.
CORRESPONDENCE CONCERNING THIS ARTICLE should be addressed to
Ryan M. Cook, Educational Studies in Psychology, Research Methodology
& Counseling, The University of Alabama, 310A Graves Hall, Tuscaloosa,
AL 35487. E-mail: rmcook@ua.edu
188
of the hierarchical structure of supervision (Porter & Vasquez,
1997). Supervisors often have more years of training and experi-
ence compared to their supervisees, which might lend more power
to the supervisor (Bernard & Goodyear, 2014). Some supervisees
view the supervisor’s level of expertise as advantageous to their
clinical development (De Stefano, Hutman, & Gazzola, 2017;
Murphy & Wright, 2005). However, clinical supervisors also serve
as evaluators of supervisees’ clinical work and gatekeepers to
protect clients and the public (American Psychological Associa-
tion [APA], 2014; Bernard & Goodyear, 2014), which places
supervisees in a vulnerable position (Hernández & McDowell,
2010; Murphy & Wright, 2005; Szymanski, 2003). Failure to
adequately attend to power dynamics in supervision can result in
ineffective or even harmful supervision that presents legal and
ethical risks to supervisors and supervisees (Ellis et al., 2014).
Thus, both supervisors and supervisees should be aware of the
influence of power in supervision and address these issues when
possible (Porter & Vasquez, 1997; Szymanski, 2003, 2005).
Supervisors are encouraged to attend to the inherent power
differentials from the outset of supervision. Informed consent
documents and disclosure agreements are intended to define the
supervisory roles and responsibilities as well as outline the expec-
tations of supervision for both the supervisor and supervisee (APA,
2014; Borders et al., 2014). Even with these resources, supervisors
and supervisees may not fully understand what is expected of them
in supervision (Ellis et al., 2014) or predict with reasonable accu-
racy issues of power that might emerge in the supervisory rela-
 POWER DYNAMICS IN SUPERVISION SCALE
tionship. Currently, supervisors do not have an objective measure
of power dynamics, supervisees’ understanding of power dynam-
ics, or supervisees’ perceptions of sources of power within the
supervisory relationship. Accordingly, we sought to create the
Power Dynamics in Supervision Scale (PDSS), an instrument
developed using Rasch modeling that is designed to measure
where supervisees perceive power lying between them and their
supervisors across multiple sessions. The PDSS is a brief, 16-item
tool that supervisees and supervisors can easily utilize to under-
stand and explore the supervision experience.
Conceptual Framework of Power Dynamics
in Supervision
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.
Power in supervision has been most commonly discussed through
the lens of feminist supervision with emphasis on the supervisor using
power diligently (e.g., Nelson et al., 2006; Porter, 1995; Porter &
Vasquez, 1997; Szymanski, 2003, 2005). Power analysis, whereby
supervisors and supervisees openly discuss inherent power differen-
tials in the supervisory relationship, is a core function of a feminist-
oriented supervisory approach (Szymanski, 2003, 2005). Through this
approach, power can be used in the supervisory relationship to benefit
the supervisee by facilitating their autonomy and empowerment (Por-
ter & Vasquez, 1997). Also, supervisors create an environment of
safety whereby supervisees feel their ideas and input are valued and
in which supervisees feel that their strengths are recognized (Porter,
1995; Porter & Vasquez, 1997; Szymanski, 2003).
To amply address power dynamics in supervision, Szymanski
(2003) suggested that supervisors explain the supervision process,
define the roles and confines of supervision, and encourage dis-
cussions of potential boundary issues (e.g., sexual attractions,
multiple relationships). She also noted the importance of supervi-
sors reflecting on their privilege, which might contribute to the
power dynamics in supervision. Hernández and McDowell (2010)
also highlighted cultural identities and social norms as particularly
salient because supervisors and supervisees influence the supervi-
sory relationship by virtue of privilege or lack thereof.
In a study especially important to the current investigation,
Murphy and Wright (2005) explored supervisees’ perceptions of
positive and negative uses of power in supervision. They found
that supervisees perceived positive uses of power as including (a)
facilitation of discussions of power, (b) sharing ideas, (c) provid-
ing feedback, (d) evaluations, (e) empowering them as supervisees,
(f) promoting safety in the supervisory relationship, (g) a collab-
orative partnership, and (h) defining the expectations of supervi-
sion. In contrast, supervisees perceived negative uses of power as
including (a) preferential treatment toward one supervisee, (b)
supervisors imposing their clinical style, (c) abusing power, (d)
violating the supervisees’ confidentiality, and (e) supervisors pri-
oritizing their needs over the supervisees. Regarding their own use
of power, supervisees perceived their power as deciding what
information to share with or withhold from their supervisors.
Supervisees in this study also viewed themselves as consumers of
supervisory services whereby they had an opportunity to learn
from and provide feedback to their supervisors. Seemingly, both
supervisors and supervisees exert power, which influences the
outcomes of supervision.
The conceptual and empirical literature, although limited, high-
lights the importance of attending to power dynamics, as well as
189
positive and negative uses of power in supervision. Although the
literature informs the importance of identifying and attending to
power dynamics in supervision, supervisors still lack a tool to help
them examine power dynamics within their own supervisory rela-
tionships. Supervisors and supervisees are recommended to reflect
on their supervision experience and issues of power that they
identify (Porter & Vasquez, 1997). This leaves supervisors without
the ability to operationalize important questions unanswered, such
as the following: When do supervisees perceive power as being
beneficial versus problematic? To what degree should a supervisor
or supervisee possess power in the supervisory relationship? Most
notably, where do supervisees perceive the power lying between
them and their supervisor in regards to specific aspects of power?
With no way to address these questions, supervisors may be, at a
minimum, unknowingly providing ineffective supervision, but
they might also be placing their supervisees and themselves at risk
for legal and ethical ramifications (Ellis et al., 2014).
The Current Study
Mangione, Mears, Vincent, and Hawes (2011) found that super-
visees were most often the ones who initiated conversations of
power, but rarely did this dialogue result in substantive change in
the relationship. An instrument that measures power perhaps could
create an opportunity for supervisors and supervisees to engage in
purposeful dialogue about the presence of power in supervision
that would result in meaningful change. Thus, we sought to create
the PDSS, which will provide supervisors with a measure of how
their supervisees perceive power dynamics within their supervi-
sory relationships. The PDSS is designed to be given to supervis-
ees at the end of a supervision session to provide quick, objective
feedback regarding the supervisees’ experience of their supervi-
sion session that can be utilized immediately.
We hypothesized a unidimensional structure of power dynam-
ics, where higher scores indicate that supervisees perceived their
supervisor possessing more power and lower scores indicating that
the supervisees perceived themselves possessing power. Specifi-
cally, we sought to investigate the following research questions:
(a) To what extent can the PDSS items be ordered to reflect
different levels of power between supervisees and supervisors?
and (b) What evidence is there that the PDSS items demonstrate
acceptable psychometric properties, such that the instrument can
be interpreted as a meaningful measure of power dynamics be-
tween supervisors and supervisees?
Method
Development of the PDSS
We used the following guidelines from DeVellis (2016) to
develop the PDSS: (a) determine what to measure, (b) generate an
item pool, (c) determine the measurement format, (d) expert re-
view, (e) administer the survey to a development sample, and (f)
optimize scale length. In determining what to measure, our goal
was to develop an instrument that can identify where the power is
perceived to lie between supervisor and supervisee. Coming from
this perspective, we made three assumptions: (a) Supervision
power dynamics are fluid and may transition between supervisee
and supervisor; (b) power may be held more by the supervisor in
 190 COOK, MCKIBBEN, AND WIND
some areas, by the supervisee in others, and equally shared in
others; and (c) power lying more or less with a supervisor or
supervisee, or being equally shared, is not inherently positive or
negative.
Based on the intended scope and purpose of the PDSS, we
generated an initial item pool of 21 items derived from the liter-
ature on power in clinical supervision. Each item contained two
dichotomous statements placed at opposing ends of a visual analog
scale (VAS; see Table 1). We used this measurement format to
score items rather than a unidirectional Likert scale (e.g., 0 – 4) in
order to more clearly represent our view of power dynamics as a
continuum between supervisors and supervisees. We also utilized
a VAS rather than a Likert scale to better capture the fluidity of
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
power dynamics. The VAS presents a sliding scale response for-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
mat for each item. Item statements on the left anchor of the VAS
reflected power held by the supervisee and were scored as 1. Item
statements on the right anchor of the VAS reflected power held by
the supervisor and were scored as 4. Participants also had the
option of selecting not applicable for each item statement. These
responses were scored as zeros and treated as missing in the
analysis. Accordingly, we interpreted responses to PDSS items on
the VAS such that higher ratings indicated that supervisees per-
ceived their supervisor as possessing more power and lower rat-
ings indicated that supervisees perceived themselves as possessing
more power, with the midpoint as the “tipping point” toward
supervisee or supervisor.
Next, we submitted the items and VAS to expert review. Spe-
cifically, we asked three professors, chosen based on their exten-
sive research agendas in clinical supervision, to review the items
and VAS for clarity and for content validity. Based on their
feedback, we added 6 items and edited 25 to improve clarity. All
experts commented that the VAS seemed to fit the items and that
it seemed to measure where power dynamics lie in supervision.
Following expert review, we administered the items to a develop-
ment sample.
Procedure
To sample trainees receiving clinical supervision, we obtained a
list of counseling psychology and school psychology programs
accredited by the APA and counselor education programs accred-
ited by the Council for Accreditation of Counseling and Related
Educational Programs (CACREP) by using the APA and CACREP
websites, respectively. We then identified the publicly available
Table 1
Sample Power Dynamics in Supervision Scale Items With the Visual Analog Scale
I identified the goals for
this supervision
session
I had the power in our
supervisory
relationship in this
supervision session
I was able to speak
freely in this
supervision session
e-mail addresses of all department heads and clinical faculty at the
APA programs and all faculty members from CACREP programs.
This resulted in us contacting faculty at 64 APA counseling or
school psychology programs and 274 counselor education pro-
grams after obtaining institutional review board approval. We
emailed all faculty members and requested that they forward a
participation request and Qualtrics survey link to any students who
were enrolled in a counseling practicum, internship, or postdoc-
toral placement (school and counseling psychology programs
only). The online survey included the PDSS, a demographic ques-
tionnaire, and two other instruments not analyzed for the current
study.
Participants
In total, 267 master’s-level and doctoral-level trainees from
counseling psychology, school psychology, and counselor educa-
tion programs who were currently enrolled in counseling practi-
cum, internship, or postdoctoral internship participated in this
study. An additional 108 participants opened the study but did not
complete any items. The participants’ ages ranged from 21–51
(M ⫽ 29.69, SD ⫽ 7.93). The majority of participants identified as
White or Caucasian (n ⫽ 174, 65.2%) and female (n ⫽ 193,
72.3%). Most participants were enrolled in a master’s-level coun-
selor education program (n ⫽ 158, 59.2%). The average length of
the supervisory relationship was 7 months (SD ⫽ 9.26, Mdn ⫽ 4.0,
n ⫽ 228). Full demographic information for both supervisees and
supervisors is included in Table 2.
Analytic Approach
Because our focus was on exploring the psychometric properties
of the PDSS with the goal of improving the instrument for future
administrations such that it can be used as a communicative device
in practical counseling settings, we used Rasch measurement the-
ory (Rasch, 1960) to guide our analytic approach. Empirical evi-
dence of fit to the Rasch model allows researchers to identify items
that do not function consistently across persons and persons who
are not measured consistently across items. Furthermore, because
the Rasch model is based on invariant measurement, item and
person estimates can be described on a common linear scale, thus
facilitating communication about a variety of features of latent
variables, including item ordering and person ordering.
My supervisor
identified the goals
for this supervision
session
My supervisor had the
power in our
supervisory
relationship in this
supervision session
I withheld information
in this supervision
session
 POWER DYNAMICS IN SUPERVISION SCALE 191

Table 2 Model-data fit statistics. Model-data fit analysis in Rasch

Supervisee and Supervisor Demographic Information measurement is conducted using residuals or discrepancies be-
tween observed responses and expected responses based on the
Supervisee, Supervisor, model. Residuals are summarized using two mean square error
Variable n (%) n (%)
(MSE) statistics: Infit MSE and Outfit MSE; standardized versions
Gendera of these statistics are also available. These statistics can be calcu-
Androgynous 1 (0.4) — lated for individual items and supervisees, such that researchers
Female 193 (72.3) 165 (61.8)
can identify and explore misfit related to specific attitude state-
Heterosexual 1 (0.4) —
Male 32 (12) 59 (22.1) ment and supervisees. Because Infit MSE is weighted by statistical
Postgender male 1 (0.4) — information (i.e., variance or measurement precision), it is sensi-
Queer 1 (0.4) — tive to less extreme unexpected observations. Outfit MSE is not
Transgender male — 1 (0.4) weighted, so it is sensitive to more extreme unexpected observa-
Did not respond 37 (13.9) 42 (15.7)
Race tions.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

African American/Black 11 (4.1) 20 (7.5) The generally agreed-upon expected value for Infit and Outfit
This document is copyrighted by the American Psychological Association or one of its allied publishers.

American European or European decent 3 (1.1) — MSE statistics is 1.00, where values below 1.00 suggest less
American Indian 1 (0.4) 3 (1.1) variation than expected and values above 1.00 suggest more vari-
Asian 8 (3) 5 (1.9)
ation than expected. In this study, we view the fit statistics as
Eastern European — 1 (0.4)
Hispanic/Latino/a 13 (4.9) 13 (4.9) continuous variables, while recognizing the generally accepted
Jamaican — 1 (0.4) values for rating scale items of about 0.60 –1.40 (Bond & Fox,
Middle Eastern 1 (0.4) — 2015; Engelhard, 2013) for MSE values. We also examined the
Multiracial 18 (6.7) 5 (1.9) standardized versions of these statistics, where acceptable fit is
Turkish — 1 (0.4)
White/Caucasian 174 (65.2) 181 (67.8) defined as ⫺2.00 ⱕ Infit Z or Outfit Z ⱕ 2.00.
Did not respond 38 (14.2) 39 (14.6) Separation statistics. Next, separation statistics describe the
Training background degree to which individual items and supervisee locations are
Clinical psychology — 34 (12.7) distinct. Specifically, the reliability of separation statistic (Rel) is
Counseling psychology 53 (19.8) 52 (19.5)
an index of the differences between individual items and individ-
Counselor education 205 (76.8) 101 (37.8)
Multiple — 1 (0.4) ual supervisees. When there is good model-data fit, Rel can be
School psychology 4 (1.9) 21 (7.9) interpreted similar to Cronbach’s alpha coefficient. A chi-square
Social work — 12 (4.5) statistic can also be calculated that describes the degree to which
Other — 7 (2.6) these differences are statistically significant.
Did not respond 5 (1.9)
Academic level
Master’s level 158 (59.2) —
Doctoral level 72 (26.9) — Results
Did not respond 37 (13.9) —
a
Ten (3.7%) supervisees specifically identified as cisgender in addition to Preliminary Analysis
their affirmed gender pronoun, and nine (3.4%) supervisors were identified
as cisgender in addition to their gender pronoun. First, we examined the presence of missing data, and across all
27 items, 2.4% of the data was missing. It was not necessary to
impute missing values because the Rasch model can provide
estimates as long as there is sufficient connectivity between items
Data Analysis
and persons (i.e., each person responds to at least one item in
We estimated supervisee (person) and attitude statement (item) common with another person, and each item has responses from at
locations using the Rating Scale (RS) model (Andrich, 1978): least two people; Schumacker, 1999). Next, we conducted a pre-
liminary analysis to identify any items that demonstrated signifi-
ln关Pni(x⫽k) ⁄ Pni(x⫽k⫺1)兴 ⫽ ␪n ⫺ ␦i ⫺ ␶k
cant model misfit and thus warranted removal from the final scale.
where ␪n is location of supervisee n on the latent variable (i.e., the With all items included, there was adequate overall fit, where 49%
strength of the supervisees’ perception of their supervisor possess- of the variance in responses was explained by the model. However,
ing power across the PDSS attitude statements), and ␦i is the fit statistics for individual items revealed 11 items with fit statistics
location of attitude statement i on the latent variable (i.e., the that exceeded the range of generally accepted values: Items 1–3, 5,
degree to which a particular attitude statement is associated with a 7–9, 11, and 22–24. Accordingly, we removed these items from
power balance in favor of the supervisor across supervisees). The the scale in the subsequent analyses.
last term in Equation 1 is the location of the threshold between Prior to our final analysis, we confirmed the unidimensionality
adjacent rating scale categories. When the RS model is applied, of the final set of PDSS items (n ⫽ 16) using a principal compo-
locations of each supervisee, attitude statement, and threshold are nents analysis (Kaiser-Meyer-Olkin measure of sampling ade-
estimated on a common linear scale. The RS model also provides quacy ⫽ .938; Bartlett’s test of sphericity p ⬍ .001), and the
a variety of indices that summarize the degree to which the data results showed 51.77% of variance explained by a single compo-
correspond to the expectations of the Rasch model. In this study, nent. The amount of variance explained by the first factor is well
we focus on two indices of psychometric quality: (1) model-data above the 20% critical value suggested by Reckase (1979) for
fit statistics and (2) separation statistics. Rasch analysis of potentially multidimensional scales. Further-
 192 COOK, MCKIBBEN, AND WIND

more, analysis with the RS model on the 16 items revealed that the
RS model explained 66.84% of the variance in PDSS responses.
We then checked for differences in participant responses based
on gender, race, training background, educational level, and length
of the supervisory relationship. We found no statistically signifi-
cant differences in main effects for gender, race, training back-
ground, and length of the supervisory relationship, but we did find
that master’s-level students responded statistically different from
doctoral students (␹2(2) ⫽ 18.7, p ⬍ .0001). The doctoral students
perceived themselves as possessing more power than the master’s-
level students. This finding may be explained by the differing
developmental levels of the supervisees. Whereas master’s-level
students are novices, doctoral students may be more developmen-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
researchers have established as expected when data fit the model.
For supervisees, the reliability of separation statistic was 0.91,
␹2(266) ⫽ 2,457.0, p ⬍ .001; this value suggests that the final set
of PDSS items was sensitive to supervisees with different levels of
power perceptions. For items, the reliability of separation statistic
was 0.98, ␹2(15) ⫽ 962.0, p ⬍ .001; this value suggests that the
PDSS items were distinct in terms of their difficulty to endorse.
Supervisee and Item Locations
The variable map (see online supplemental Figure S1) is a
graphical summary of the supervisee and item estimates. The first
column is the common scale on which supervisee and item loca-

tally advanced supervisees given prior experiences in supervision tions were estimated. The second column shows the locations of
This document is copyrighted by the American Psychological Association or one of its allied publishers.

(Loganbill, Hardy, & Delworth, 1982; Stoltenberg & McNeill, each supervisee, where an asterisk represents two supervisees and
2010). Given that our population included all of these subgroups of a period represents one supervisee. Supervisee locations ranged
participants and our dimensionality analysis suggested that the from ⫺6.23 for the supervisee with the highest perception of
construct was similar for all of the participants, we examined all supervisee power to 5.62 for the supervisee with the highest
educational-level groups as a single group. perception of supervisor power. The third column shows the lo-
cations of the items. Item 6 (“I decided which interventions will be
Summary Statistics used with my client(s) in this supervision session”) has the highest
location (␦ ⫽ 3.29), suggesting that supervisees reported that their
After we removed the severely misfitting items, we examined supervisor possessed the most power related to this item. On the
summary statistics for the RS model (see Table 3). To provide a other hand, Item 27 (“I think my supervisor maintained healthy
frame of reference for interpreting locations, we centered the items boundaries with me in this supervision session”) has the lowest
at zero logits (M␦ ⫽ 0.00, SD ⫽ 0.80). The average supervisee location (␦ ⫽ ⫺1.58), suggesting that supervisees reported they
location was lower than the average item location (M␪ ⫽ ⫺2.48, possessed the most power most often related to this item. The final
SD ⫽ .76), which suggests that the supervisees did not report a columns in online supplemental Figure S1 illustrate the calibration
strong imbalance of power in favor of their supervisors. For of the rating scale categories for the PDSS items.
supervisees and items, the average values of the standardized and Item-fit statistics. The primary purpose of the current analy-
unstandardized fit statistics were within the range that previous sis is to explore the psychometric properties of a pilot administra-
tion of the PDSS; accordingly, we focus on item-fit statistics rather
than person-fit statistics. Table 4 includes item locations and
Table 3 model-data fit statistics for each of the 16 items included in the
Summary Statistics final set of PDSS items. To identify items with the most extreme
unexpected responses, we focused on the Outfit statistics. The
Supervisee Item
Summary statistics (n ⫽ 267) (n ⫽ 16) most frequent unexpected responses were observed for Item 6
(Outfit MSE ⫽ 1.41, Outfit Z ⫽ 2.16), Item 4 (Outfit MSE ⫽ 1.43,
Calibrations Outfit Z ⫽ 3.75), and Item 27 (Outfit MSE ⫽ 1.34, Outfit Z ⫽
Measure (logits) 1.24). These values suggest that, for these items, supervisees
M ⫺2.48 .00
SD 2.23 1.20 provided ratings that were unexpected given their measures. An
Standard error examination of the residuals revealed that most of the unexpected
M .57 .15 responses for Item 6 consisted of supervisees providing lower
SD .20 .01 ratings than expected based on the model. Conversely, most of the
Model-data fit statistics
Infit MSE
unexpected responses for Items 4 and 27 consisted of supervisees
M 1.01 1.01 providing higher ratings than expected based on the model.
SD .48 .27
Standard infit
M ⫺.01 ⫺.09 Discussion
SD 1.22 2.51
Outfit MSE The purpose of this study was to provide psychometric evidence
M .95 .95 for the PDSS, a measure of power between supervisees and su-
SD .56 .30 pervisors. Based on the results, we retained a single factor, 16-item
Standard outfit measure, which was consistent with our hypothesized unidimen-
M ⫺.08 ⫺.42
SD 1.17 2.15
sional structure. The results suggest differing levels of perceived
Separation statistics power by supervisees, meaning that PDSS items adequately iden-
Reliability of separation .93 .98 tified differences in power perceptions among supervisees. Super-
Chi-square 3,024.0ⴱ 1,131.0ⴱ visees perceived themselves as possessing the most power on
df 266 15 maintaining healthy boundaries with their supervisors, a willing-
ⴱ
p ⬍ .001. ness to feel vulnerable in supervision, and feeling empowered in
 POWER DYNAMICS IN SUPERVISION SCALE 193

Table 4
Calibration of the Item Facet

Item number Average rating Measure SE Infit MSE Infit Z Outfit MSE Outfit Z

6 2.76 3.29 .13 1.26 2.45 1.41 2.16

4 2.09 1.45 .13 1.40 3.63 1.43 3.75
19 2.07 1.09 .14 0.95 ⫺0.50 0.90 ⫺0.85
17 1.92 0.66 .13 1.22 2.13 1.15 1.28
15 1.80 0.30 .14 0.68 ⫺3.56 0.61 ⫺3.74
16 1.74 0.05 .14 0.85 ⫺1.51 0.78 ⫺1.84
13 1.70 ⫺0.03 .14 1.11 1.09 1.25 1.75
14 1.70 ⫺0.07 .14 0.68 ⫺3.38 0.77 ⫺1.78
18 1.63 ⫺0.42 .14 1.03 0.29 0.89 ⫺0.69
12 1.59 ⫺0.45 .15 1.19 1.64 1.18 1.10
10 1.59 ⫺0.56 .15 0.90 ⫺0.86 0.75 ⫺1.54
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

25 1.59 ⫺0.58 .15 0.61 ⫺4.12 0.52 ⫺3.48

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

26 1.54 ⫺0.81 .15 0.68 ⫺3.27 0.54 ⫺2.97

20 1.51 ⫺0.92 .15 0.99 ⫺0.08 0.87 ⫺0.64
21 1.41 ⫺1.42 .17 1.17 1.36 0.86 ⫺0.51
27 1.38 ⫺1.58 .17 1.46 3.20 1.34 1.24
Mean 1.75 0.00 .15 1.01 ⫺0.09 0.95 ⫺0.42
SD .34 1.20 .01 0.27 2.51 0.30 2.15
Note. Items are ordered by measure, from high (supervisor perceived as possessing more power) to low
(supervisee possessing the most power).

supervision. These three items provide important details regarding
how supervisees perceive their use of power in supervision. It is
important to note that the participants in the current study had
worked with their supervisors for, on average, 7 months. Most
field experiences are less than a year in length, so the findings from
the current study may reflect how supervisees perceive power in
more well-established supervision relationships.
Although maintaining healthy boundaries was where supervis-
ees perceived possessing the most power, this item also elicited the
most extreme unexpected responses, suggesting that supervisees
provided higher scores (more supervisor power) than anticipated.
Maintaining healthy boundaries is an important ethical responsi-
bility of supervisors (APA, 2014; Murphy & Wright, 2005; Szy-
manski, 2003, 2005) that protects supervisee well-being. Supervi-
sors are encouraged to define the roles and responsibilities for their
supervisees clearly from the outset of the supervisory relationship
(APA, 2014; Borders et al., 2014). The supervisees’ perceptions of
healthy boundaries might be influenced by the large percentage of
master’s-level students (59.2%) included in the current study. As
novice supervisees, master’s-level students are less able to identify
and articulate their developmental needs to their supervisors (Lo-
ganbill et al., 1982; Stoltenberg & McNeill, 2010). They may be
confused about supervisors’ expectations of them and uncertain of
what constitutes “healthy boundaries,” and they may therefore rely
on their supervisors for guidance (Stoltenberg & McNeill, 2010).
It is plausible that healthy boundaries existed in the supervisory
relationships for the supervisees in the current study; however, the
supervisors did not engage their supervisees in conversations about
their roles and responsibilities in supervision adequately (Ellis et
al., 2014).
Participants also viewed their power as being able to be vulner-
able with their supervisors. A willingness to be vulnerable is
imperative to supervisee personal and professional growth in su-
pervision, but it also means opening up to criticism and feedback
from supervisors (Porter & Vasquez, 1997). Some supervisees
may be concerned about negative consequences associated with
their openness (Ladany, Hill, Corbett, & Nutt, 1996). Supervisees
in this study also perceived themselves as being empowered in the
supervisory relationship, which aligns with feminist supervision
theory’s position that supervisors should create an empowering
environment that facilitates supervisee autonomy (Porter &
Vasquez, 1997). Supervisees who feel vulnerable or disempow-
ered may feel that the supervisor holds more power in the rela-
tionship and/or is misusing power (Murphy & Wright, 2005; Porter
& Vasquez, 1997). There are other potential explanations for our
findings, such as Mehrabian’s three-dimensional model of plea-
sure, arousal, and dominance (Mehrabian, 1980). Perhaps the
supervisees in this study experienced low dominance but high
arousal due to the normative process of continually being evalu-
ated (Bernard & Goodyear, 2014; Mehrabian, 1980).
In this study, participants perceived their supervisors as possess-
ing the most power when identifying interventions to use with
clients, setting goals for supervision, and providing feedback about
clinical skills in supervision. These findings speak to the supervi-
sors’ roles as experts and evaluators in the supervisory relationship
(Bernard & Goodyear, 2014). Such sources of supervisor power
may explain why supervisees perceived their supervisors in posi-
tions of power in general. Interestingly, although the supervisees
perceived their supervisors as possessing more power for these
aspects of supervision, they still viewed these aspects of clinical
supervision as shared between them and their supervisors. This
finding may provide empirical evidence regarding the importance
of collaboration and cooperation within the supervisory relation-
ship (Rousmaniere & Ellis, 2013; Szymanski, 2003). Overall, we
found evidence that supervisees are not only aware of power
dynamics in supervision (Mangione et al., 2011; Murphy &
Wright, 2005) but also able to detect differences in power between
themselves and their supervisors across multiple areas. The PDSS
allows supervisors to understand how supervisees perceive power
in supervision and to attend to issues when needed (Porter &
Vasquez, 1997).
 194 COOK, MCKIBBEN, AND WIND
Implications
Failure to address power is at the very least against best prac-
tices (APA, 2014) but may even result in ineffective or harmful
supervision (Ellis et al., 2014). Ellis et al. (2014) suggested that
attending to power dynamics in supervisory relationships is a
requirement of supervisors to provide minimally adequate clinical
supervision. Feminist tenets provide guidance for how supervisees
can best attend to power dynamics in supervision. Supervisors
should reflect on the supervisory process and invite feedback from
their supervisees regarding their experiences in supervision (Por-
ter, 1995; Porter & Vasquez, 1997). However, issues of power may
go unnoticed or unaddressed by supervisors. Mangione et al.
(2011) found that supervisors might be less aware of issues of
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
power compared to their supervisees and infrequently engage their
This document is copyrighted by the American Psychological Association or one of its allied publishers.
supervisees in a purposeful conversation about issues of power in
the supervisory relationship. The PDSS creates an opportunity for
formalized feedback to be quickly solicited, which allows for
conversations about power in numerous aspects of clinical super-
vision as recommended by feminist scholars (e.g., Porter, 1995;
Porter & Vasquez, 1997), thereby helping supervisors to provide
adequate clinical supervision (Ellis et al., 2014).
The PDSS can also help to normalize power dynamics that may
otherwise be difficult to discuss overtly (Murphy & Wright, 2005).
Supervisees are often aware of the power dynamics in supervision,
but some supervisees may be hesitant to share these concerns with
supervisors due to fears of repercussions or negative evaluations
(Ladany et al., 1996). The supervisor needs to take the lead in
initiating these discussions and naming specific aspects of power
directly (Mangione et al., 2011), which can be facilitated with the
PDSS.
Supervisees perceived themselves as holding the most power in
maintaining healthy boundaries, being willing to be vulnerable,
and feeling empowered. Given that feminist supervision empha-
sizes supervisee autonomy (Szymanski, 2003), supervisors may
focus on the aspects of clinical supervision in which supervisees
perceive themselves as holding power as a means to facilitate
autonomy (Porter & Vasquez, 1997). Supervisors can ask ques-
tions of their supervisees to better understand the factors that
facilitate or inhibit these aspects of clinical supervision. Similarly,
supervisees may perceive some aspects of supervision as respon-
sibilities that supervisors should hold. Since supervisors have more
experience and training (Bernard & Goodyear, 2014), supervisees
may look to their supervisors for guidance in some areas. Even
though supervisees may need additional assistance on certain
tasks, it may be important for supervisors to approach these tasks
as a shared experience (Porter & Vasquez, 1997). Supervisors may
find it helpful to encourage supervisees to conceptualize a case
before interjecting and to assist them in identifying interventions
rather providing solutions (Mangione et al., 2011). In sum, a
supervisee who perceives himself or herself as not holding power
in some facet of supervision does not indicate a problem; super-
visors need to have open conversations to determine how all
parties can utilize power responsibly for the betterment of the
supervisee and the protection of clients.
Limitations and Future Research
Limitations of the current study create an opportunity for future
research. First, we recruited participants via faculty members and
are thus unable to calculate a response rate. While we know 108
individuals opened the survey link and did not take the survey, we
have no way of knowing who chose not to participate in the study
and why. Also, we only collected data regarding the supervisees’
perceptions of power. Although the PDSS provides information
about supervisees’ perceptions of power balance, it does not reveal
the implications of power balance for individual supervisees or
supervisors. In future studies, researchers should examine PDSS
results alongside qualitative data to gain a better understanding of
how supervisors and supervisees interpret information about
power balance in supervision. Future researchers should explore
how supervisors perceive the presence of power in supervision in
relation to their supervisees’ perceptions of power. Future re-
searchers should also investigate differences in perceived power
dynamics based on professional and personal identities. Demo-
graphically, we sampled psychology and counseling trainees who
were predominantly White (65.2%) females (72.3%). Our results
may not generalize evenly across diverse cultural identities. Issues
of privilege and oppression may be salient to how supervisees
experience power in the supervisory relationship in more diverse
supervision dyads. Relatedly, we found that master’s-level stu-
dents perceived their supervisors as possessing more power in the
supervisory relationship than did the doctoral-level supervisees.
Thus, future researchers should more closely examine validity
evidence for the PDSS in more diverse samples. Specifically,
studies are needed that include investigations of person fit and
differential item functioning across professional disciplines, cul-
tural identities, and supervisee developmental levels. Finally, we
believe that the PDSS has promise as an instrument that can be
administered after each supervision session to improve outcomes
in supervision, an investigation that was beyond the scope of the
current study. Future studies should examine if the PDSS can be
administered across multiple sessions to mitigate potential issues
and improve the quality of supervision. Future studies should
examine the relationship between the PDSS and other variables of
interest (e.g., working alliance) in order to understand how power
dynamics affect the supervisory process and outcomes.
References
American Psychological Association. (2014). Guidelines for clinical su-
pervision in health service psychology. Washington, DC: Author. Re-
trieved from http://www.apa.org/about/policy/guidelines-supervision
.pdf
Andrich, D. A. (1978). A rating formulation for ordered response catego-
ries. Psychometrika, 43, 561–573. http://dx.doi.org/10.1007/BF0
2293814
Bernard, J. M., & Goodyear, R. K. (2014). Fundamentals of clinical
supervision. Boston, MA: Pearson.
Bond, T. G., & Fox, C. M. (2015). Applying the Rasch model: Fundamen-
tal measurement in the human sciences (3rd ed.). New York, NY:
Routledge.
Borders, L. D., Glosoff, H. L., Welfare, L. E., Hays, D. G., DeKruyf, L.,
Fernando, D. M., & Page, B. (2014). Best practices in clinical supervi-
sion: Evolution of a counseling specialty. The Clinical Supervisor, 33,
26 – 44. http://dx.doi.org/10.1080/07325223.2014.905225
De Stefano, J., Hutman, H., & Gazzola, G. (2017). Putting on the face: A
qualitative study of power dynamics in clinical supervision. The Clinical
Supervisor, 36, 223–240. http://dx.doi.org/10.1080/07325223.2017
.1295893
 POWER DYNAMICS IN SUPERVISION SCALE 195

DeVellis, R. F. (2016). Scale development: Theory and applications.
Thousand Oaks, CA: Sage.
Ellis, M. V., Berger, L., Hanus, A. E., Ayala, E. E., Swords, B. A., & Siembor,
M. (2014). Inadequate and harmful clinical supervision: Testing a revised
framework and assessing occurrence. The Counseling Psychologist, 42,
434 – 472. http://dx.doi.org/10.1177/0011000013508656
Engelhard, G. (2013). Invariant measurement: Using Rasch Models in the
social, behavioral, and health Sciences. New York, NY: Routledge.
Hernández, P., & McDowell, T. (2010). Intersectionality, power, and
relational safety in context: Key concepts in clinical supervision. Train-
ing and Education in Professional Psychology, 4, 29 –35. http://dx.doi
.org/10.1037/a0017064
Ladany, N., Hill, C. E., Corbett, M. M., & Nutt, E. A. (1996). Nature,
extent, and importance of what psychotherapy trainees do not disclose to
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Porter, N. (1995). Supervision of psychotherapists: Integrating anti-racist,
feminist, and multicultural perspectives. In H. Landrine (Ed.), Bringing
cultural diversity to feminist psychology: Theory, research, and practice
(pp. 163–175). Washington, DC: American Psychological Association.
http://dx.doi.org/10.1037/10501-008
Porter, N., & Vasquez, M. (1997). Covision: Feminist supervision, process,
and collaboration. In J. Worell & N. G. Johnson (Eds.), Shaping the
future of feminist psychology: Education, research, and practice (155–
171). Washington, DC: APA. http://dx.doi.org/10.1037/10245-007
Rasch, G. (1960). Probabilistic models for some intelligence and achieve-
ment tests. Copenhagen, Denmark: Danish Institute for Educational
Research.
Reckase, M. D. (1979). Unifactor latent trait models applied to multifactor
tests: Results and implications. Journal of Educational Statistics, 4,

their supervisors. Journal of Counseling Psychology, 43, 10 –24. http:// 207–230. http://dx.doi.org/10.3102/10769986004003207
Rousmaniere, T. G., & Ellis, M. V. (2013). Developing the construct and
This document is copyrighted by the American Psychological Association or one of its allied publishers.

dx.doi.org/10.1037/0022-0167.43.1.10
Loganbill, C., Hardy, E., & Delworth, U. (1982). Supervision: A concep-
tual model. The Counseling Psychologist, 10, 3– 42. http://dx.doi.org/10
.1177/0011000082101002
Mangione, L., Mears, G., Vincent, W., & Hawes, S. (2011). The supervi-
sory relationship when women supervise women: An exploratory study
of power, reflexivity, collaboration, and authenticity. The Clinical Su-
pervisor, 30, 141–171. http://dx.doi.org/10.1080/07325223.2011
.604272
Mehrabian, A. (1980). Basic dimensions for a general psychological the-
ory: Implications for personality, social, environmental, and develop-
mental studies. Cambridge, MA: Oelgeschlager, Gunn & Hain.
Murphy, M. J., & Wright, D. W. (2005). Supervisees’ perspectives of
power use in supervision. Journal of Marital and Family Therapy, 31,
283–295. http://dx.doi.org/10.1111/j.1752-0606.2005.tb01569.x
Nelson, M. L., Gizara, S., Hope, A. C., Phelps, R., Steward, R., &
Weitzman, L. (2006). A feminist multicultural perspective on supervi-
sion. Journal of Multicultural Counseling and Development, 34, 105–
115. http://dx.doi.org/10.1002/j.2161-1912.2006.tb00031.x
measure of collaborative clinical supervision: The supervisee’s perspec-
tive. Training and Education in Professional Psychology, 7, 300 –308.
http://dx.doi.org/10.1037/a0033796
Schumacker, R. E. (1999). Many-facet Rasch analysis with crossed, nested,
and mixed designs. Journal of Outcome Measurement, 3, 323–338.
Stoltenberg, C. D., & McNeill, B. W. (2010). IDM Supervision: An
integrated developmental model for supervising counselors & therapists
(3rd ed.). New York, NY: Routledge.
Szymanski, D. M. (2003). The feminist supervision scale: A rational/
theoretical approach. Psychology of Women Quarterly, 27, 221–232.
http://dx.doi.org/10.1111/1471-6402.00102
Szymanski, D. M. (2005). Feminist identity and theories as correlates of
feminist supervision practices. The Counseling Psychologist, 33, 729 –
747. http://dx.doi.org/10.1177/0011000005278408
Received August 28, 2017
Revision received March 1, 2018
Accepted March 23, 2018 䡲

E-Mail Notification of Your Latest Issue Online!

Would you like to know when the next issue of your favorite APA journal will be available
online? This service is now available to you. Sign up at https://my.apa.org/portal/alerts/ and you will
be notified by e-mail when issues of interest to you become available!