The Supervisory Alliance: A Half Century of Theory, Practice, and Research in Critical Perspective

Introduction

The relationship between psychotherapy supervisor and supervisee has long been a matter of concern in the supervision literature (Bernard & Goodyear, 2014; Falender & Shafranske, 2012). Even in some of the earliest writings about supervision, the importance of the supervisory connection, if not explicitly emphasized, appears to have been implicitly conveyed (e.g., Eitingon, 1923, 1926; Watkins, 2013a). In contemporary practice, the supervisor-supervisee relationship continues to be assigned a place of nuclear significance; as now conceptualized, it encompasses a host of substantive variables, including supervisory style, real relationship, transference-countertransference configuration, supervisee anxiety, and issues of difference and diversity (Beinart, 2012; Crook Lyon & Potkar, 2010; Rodenhauser, 1997; Toldson & Utsey, 2008; Watkins, 2011b). But of the various elements that compose the supervision relationship, none seems to exert more power and influence on supervisor and supervisee than their jointly-forged supervisory alliance.

Over the course of the last half century, the supervisory alliance—which tends to be thought of as the supervision equivalent of the psychotherapeutic alliance—has emerged as supremely significant in the conceptualization and conduct of the supervision experience (Bordin, 1983; Falender & Shafranske, 2004; Fleming & Benedek, 1964; Teitelbaum, 1995, 2001). The alliance has been increasingly embraced as the very heart and soul of supervision, and its potential impact on the supervisee change process and supervision outcome has generally come to be regarded as affecting and far reaching (Inman & Ladany, 2008; Ladany & Inman, 2012). Across psychology competency frameworks, the formation and management of the supervisory alliance has been made a core competency internationally (Falender, Cornish, Goodyear, Hatcher, Kaslow, Leventhal, & Grus, 2004; Fleming, 2012; Gonsalvez & Milne, 2010; Psychology Board of Australia, 2011; Roth & Pilling, 2008; Turpin & Wheeler, 2011). Regardless of the supervision model employed (e.g., psychotherapy-focused, social role models), form of treatment being supervised (e.g., individual, family), or type of population served, the alliance tends now to be recognized as pivotal in making the work of supervision possible (Falender & Shafranske, 2008; Hess, Hess, & Hess, 2008; Stoltenberg & McNeil, 2010; Watkins, 2012). Some supervision approaches may weigh the role of alliance more heavily than others (cf. Beck, Sarnat, & Baranstein, 2008), but its place in supervision appears to be uniformly acknowledged and accepted across varied models of conceptual understanding.

In this paper, I would like to shine a critical light on the supervisory alliance and to consider its current status and future directions. With the supervisory alliance having reached the half century mark, and being regarded as such a pivotal, transtheoretical construct in guiding supervision practice, I thought it might be a good time to “take stock” and examine two basic questions:

I will address those questions by (a) defining and describing the two alliance perspectives that dominate supervision theory, practice, and research; (b) reviewing the supervision research that has emerged over the past generation; and (c) developing an integrative picture of the present status and future needs of the supervisory alliance.

Two Generative Visions of the Alliance in Supervision

To best understand the supervisory alliance in a contemporary perspective, it is important to first look back, examine how the construct began to take form in psychological thought, and reflect on how it came to inform our views about “good” supervision. As with so much in psychotherapy’s origin and evolution, the alliance is typically viewed as having begun with Freud; that beginning also appears to have laid the groundwork for the eventual emergence of alliance considerations in supervision.

Freud’S Contributions and Beyond

Although Freud did not specifically use the term “alliance” until his 1937 paper on analysis terminable or interminable, there is little question that his ongoing clinical work consistently revolved around matters of alliance. As Hatcher (2010) has nicely summarized: “Freud encountered alliance issues as soon as he began to use psychological methods with his patients.… we can identify as alliance issues his struggles to engage and keep his patients in treatment” (p. 8). In his writings, Freud (1913/1958, 1937/1964) made mention of the supreme importance of “a well-developed rapport”, pact, or compact between analyst and analysand for treatment advance. As his thinking evolved, the element of collaboration came to figure far more prominently into his vision of psychoanalysis (Hatcher, 2010). Building on those seminal concepts, other notable analysts such as Sterba (1934), Zetzel (1956), and Greenson (1965, 1967) introduced the terms of alliance, therapeutic alliance, or working alliance into the psychoanalytic lexicon, and the treatment relationship became increasingly conceptualized as a type of joint relational effort or therapeutic partnership. As the idea of a treatment alliance gained traction and acceptance in psychoanalytic thought, a related question soon emerged: How might the construct of alliance also apply to the psychoanalytic supervision relationship?

The Fleming and Benedek Perspective on the Learning Alliance

In the mid-1960s, Fleming and Benedek (1964, 1966) first proposed the concept of a learning alliance for psychoanalytic supervision, though they readily acknowledged Freud’s contributions in stimulating their own supervision alliance formulations.

For Fleming and Benedek, the learning alliance—a supervision partnership, pact, or compact—was sine qua non for supervisor and supervisee to work profitably together.

Their vision of the learning alliance brought into focus the necessity of key features such as supervisor empathic perceptiveness and responsiveness, creation of a state of rapport in supervision, determination of a supervision goal or goals in common, and placement of learning tasks that stimulate goal accomplishment. The primary goals that provided guidance for supervision were: (a) insuring patient care; (b) enhancing the analyst’s knowledge; and (c) developing and enhancing the use of self as analytic instrument during the treatment process. Some of the learning tasks involved were: (a) the analyst/student’s working up case material for presentation in supervision; (b) the supervisor’s educational diagnosis of student’s learning needs for remediation; and (c) the supervisor’s sharing of instructive and corrective feedback. The Fleming/Benedek conceptualization accentuated, as never before, the ideas that supervision was itself a type of collaboration and the nature of that collaboration could be expected to substantially affect the entirety of the supervisory process and its eventual outcome (Watkins, 2011b). The concept of learning alliance has since heartily endured, proven quite durable down through the decades, and continues to be widely regarded as being of pivotal, nuclear significance for psychoanalytic supervision practice today (see Dewald, 1987, 1997; Frawley-O’Dea & Sarnat, 2001; Gill, 2001; Hyman, 2008; Kernberg, 2010; Oberman, 1990; Rock, 1997; Sarnat, 2012; Teitelbaum, 1990b, 2001; Watkins, 2013b).

The Bordin Perspective on the Supervision Working Alliance

As prelude to and model for his supervision working alliance conceptualization, Bordin (1979)—building on the work of Greenson (1967), Menninger (1958), Sterba (1934), and Zetzel (1956)—first proposed a highly heuristic, pan-theoretical vision of the treatment working alliance that hinged on three crucial components: The therapist-patient bond, the collaboratively established goals that guide the treatment process, and the collaboratively agreed-upon tasks that facilitate pursuit of goal attainment. In his presidential address to the Society for Psychotherapy Research, Bordin (1980) incorporated the importance of rupture and repair events into his alliance formulations (a substantive elaboration upon his earlier proposal). While the working alliance has long been a variable of preeminent concern in contemporary psychotherapy research, Bordin’s tripartite reconceptualization is credited with opening up untold possibilities for taking the alliance agenda into new and unexplored directions (Horvath, 2001); psychotherapy research across the last few decades has clearly borne strong testament to that reality (Lambert, 2013; Orlinsky, Ronnestad, & Willutzki, 2004).

Drawing on his treatment working alliance model, Bordin (1983) proposed a similar pan-theoretical vision of the supervision working alliance, which accordingly was composed of three core elements: The supervisor-supervisee bond, the collaboratively established goals that guide the supervision process, and the collaboratively agreed-upon tasks that facilitate pursuit of supervision goal attainment. As with the therapeutic working alliance, Bordin also envisioned rupture and repair events as playing a significant role in the development and maintenance of the supervision working alliance. The supervisor and supervisee bond was considered to involve their shared “feelings of liking, caring, and trusting” (Bordin, 1983, p. 36) and to “typically fall somewhere between… teacher to class members and therapist to patient” (p. 38). Bordin (1983) identified eight possible goals that could be used to guide supervision process:

He also identified three tasks by which those goals could be pursued in supervision: (a) report (oral or written) prepared by the therapist of the hour or hours to be reviewed; (b) treatment session observation through audio-recordings, video-recordings, or live viewing; and (c) presentation of problems or issues in supervision selected by the supervisee (p. 38).

Like Fleming and Benedek’s conception of the learning alliance, Bordin’s conception of the supervision working alliance has proven highly durable, is embraced as being of pivotal, nuclear significance in much psychotherapy supervision practice, and is increasingly heuristic in stimulating supervision research (Inman & Ladany, 2008; Ladany, 2004; Ladany & Inman, 2012).

The Two Alliance Conceptualizations in Perspective

Fleming and Benedek focused their alliance attention on psychoanalytic supervision exclusively, whereas Bordin (while psychodynamically influenced) proposed a pan-theoretical alliance perspective. But as is clear from a reading of these two supervision alliance descriptions, both visions are highly similar in content—invoking a shared bond or rapport as requisite, shared goals as critical, and learning tasks as mandatory. While Bordin may have identified more goals overall, the goals across both visions are alike, as are the tasks involved in their pursuit. The essence of the Fleming/Benedek and Bordin supervision alliance proposals is much the same if not identical. In my reading, I have found that Fleming/Benedek’s learning alliance understandably appears far more apt to be familiar to and used by mental health professionals—particularly psychoanalysts and psychiatrists—who readily identify themselves as psychoanalytic, whereas Bordin’s supervision working alliance appears far more apt to be familiar to and used by mental health professionals—particularly psychologists, social workers, and counselors—who reflect a host of varied theoretical leanings. Exceptions can certainly be found, but if examined across disciplines, those differences appear to hold up quite well.

What then are the operational specifics that seem to cut across these two alliance perspectives in supervision? And how might those specifics be captured in a succinct but meaningful way? Based on my reading of the Fleming/Benedek and Bordin visions, some core features (after Rogers, 1957) that appear requisite for early alliance formation in supervision might best be formulated as follows:

While not exhaustive, those 12 postulates—which are explicitly stated within or suggested by the Fleming/Benedek (1964, 1966) and Bordin (1983) visions—paint a portrait of some essentials needed for supervision alliance formation to occur. Where compromise occurs at any point (e.g., a lack of openness or receptivity), alliance formation will be negatively affected and, in turn, can be slowed in its development and rendered less likely to occur. Approaches to considering supervision goals and tasks can conceivably be quite varied in scope, ranging from the more formalized and contractual to the relatively informal (see Alonso, 2000; Bordin, 1983; Gordan, 1996; Jacobs, David, & Meyer, 1995; Teitlebaum, 1990a). What appears most important is having a mutually clarifying discussion and reaching a mutually clarifying agreement.

The Fleming/Benedek and Bordin visions provide us with two highly similar ways by which to understand the supervision alliance and its development. Although having psychoanalytic provenance, the supervisory alliance has since been embraced as a construct of transtheoretical import (cf. Carroll, 2009, 2010; Farber, 2012; Hawkins & Shohet, 2012; Ladany, Friedlander, & Nelson, 2005; Levenson & Ladany, 2012; Reiser & Milne, 2012; Sarnat, 2012; Scaturo, 2012; Stoltenberg & McNeil, 2010; Watkins, 2012; Watkins & Milne, 2014). As now conceptualized in psychotherapy supervision, the supervision alliance has come to be viewed as the very heart and soul of the supervision endeavor itself. Extrapolating Bordin’s (1979, p. 253) propostions about the therapeutic working alliance to supervision, the following statements appear to capture current thinking about the transtheoretical applicability of the supervisory alliance:

What Have We Learned from Supervision Alliance Research?

Context and Method

In complementing the theoretical/practical material considered thus far, I would like to next examine the research that has been done on the supervisory alliance. The primary questions that I will address are: Across the last 20 plus years, what have we learned empirically about the supervisory alliance? What do we not know? What do we need to know? My specific focus will be on the supervisory alliance as studied in psychological treatment supervision (i.e., the supervision of psychotherapy or counseling)—as opposed to other forms of clinical supervision, such as occupational therapy or speech pathology, where psychological treatment is not the primary subject of concern. Psychological treatment supervision will be defined as: A “distinct professional activity” (Falender & Shafranske, 2004, p. 3) or “intervention” (Bernard & Goodyear, 2014) in which a senior professional (supervisor) serves as mentor or guide to a junior professional (supervisee) who is in the process of learning and practicing psychological treatment; its primary objective is enhancement of the supervisee’s professional functioning, it involves evaluation of that professional functioning by the supervisor, and it is a hierarchical monitoring process (supervisor to supervisee) that serves a protective function for both patients and profession (Milne, 2007; Thomas, 2010).

In a broad-based review of psychoanalytic constructs in psychotherapy supervision, I (Watkins, 2010) examined the features of 17 supervisory alliance studies (along with parallel process and countertransference research) appearing up through the early part of 2010. The focus, with one study excepted, was on the individual therapy supervision relationship (as opposed to group and marital/family therapy supervision). As part of this half-century retrospective, and in order to provide a more complete picture of where alliance research stands now, I would like to build on and render the alliance portion of the earlier review current. Although the primary alliance conclusions were recently made (and will be referenced in my discussion), the alliance portion of the review could benefit from updating for two reasons: (a) several important alliance studies that merit scrutiny were not included; and (b) in the last several years supervision alliance studies have shown a considerable comparative increase. In the earlier review, studies were not included for three reasons: (a) investigations were published at or shortly before the review’s close date; (b) investigations did not focus on the supervisory alliance as the primary variable of interest; and (c) investigations were missed due to oversight. I have attempted here to cast my net as widely as possible and include all studies where the supervisory alliance was a variable of interest in any way. As with the earlier review, my focus will be on individual supervision; it remains the most prevalently-used form of supervision learning—what Bernard and Goodyear (2014) have referred to as “the cornerstone of professional development.” Thus, to get the most contemporary, informed, and comprehensive picture of supervision alliance research, I subsequently consider the entirety of such studies done so far; to make that possible, I have taken material from Watkins (2010; with permission), identified missed and new research studies, and accordingly combined the two sources of research information for review purposes.

To identify those new and missed articles, four steps were taken: (a) PsycInfo, MedLine, Education Research Complete, and Google Scholar database searches were conducted using “supervisory alliance” or “supervision alliance” as the key search words; (b) reference sections of identified studies were examined to further identify other appropriate articles for inclusion that might have been missed (“ancestry approach”; Cooper, 1989); (c) supervision journals or journals that publish some supervision material were examined for any recent articles that might have appeared; and (d) recent supervision texts (e.g., Ladany & Bradley, 2010) were also examined to further find any other possible missed work. Based on those steps, a total of 24 new or missed studies was identified for inclusion, which were then combined with 16 of the earlier-identified studies (Watkins, 2010; the one study from that review on marital/family supervision [Inman, 2006] was not included here). The review time period spanned from 1/1990 through the early part of 2013. Each article was reviewed to determine: Setting/sample characteristics, measures used, procedure, analyses, findings/conclusions, and limitations/strengths. Table 1 provides a summary of those seven features for each study. Table 2 provides a more focused snapshot of study similarities across several selected variables of empirical import.

Review Results and Discussion

In examining the study information, the results will be considered according to the seven features identified in Table 1.

Setting/Sample Characteristics

The vast majority of investigations (34 out of 40) took place in university settings or settings where graduate students were being trained; master’s or doctoral students from psychology, counseling, or social work programs were the subjects. Only six studies focused on supervision in the work place. While the supervisee’s perspective only was assessed in most studies (28), both supervisor and supervisee perspectives were assessed in ten of the investigations. The supervisor’s perspective of the alliance was the exclusive subject of interest in two studies. Across the 40 studies, 4,563 supervisees and 837 supervisors, respectively, participated. Where complete gender and race/ethnicity demographics were provided, the supervisees were 75% female and 80% White/Caucasian; the supervisors were 65% female and 85% White/Caucasian. The mean age ranged, respectively, from 28 to 51 years for supervisees (with most falling between 28 to 35 years) and 37 to 51 years for supervisors.

Measures Used

In measuring the alliance in supervision, the Supervisory Working Alliance Inventory (SWAI; Efstation, Patton, & Kardash, 1990) and supervision version of the Working Alliance Inventory (WAI)—both self-report inventories—were routinely used. The SWAI has parallel supervisee and supervisor versions. The SWAI supervisee version consists of two subscales, Rapport and Client Focus; the SWAI supervisor version consists of three subscales, Rapport, Client Focus, and Identification. The supervisee and supervisor versions of the WAI (Bahrick, 1990; Baker, 1991) each consist of three subscales: Bond, Goals, and Tasks. While not psychometrically flawless, the WAI and SWAI appear to both be generally accepted as providing adequate measure of the supervisory alliance. Of the 40 studies, 19 used some version of the WAI, 17 used the SWAI, 3 used both measures, and one used a self-designed questionnaire (Bucky, Marques, Daly, Alley, & Karp, 2010).

Procedures

In 31 of the 40 studies, a one-shot set of self-report survey questionnaires was distributed to participants either online or by mail. Four of the investigations (Deal, Bennett, Mohr, & Hwang, 2011; Livni, Crowe, & Gonsalvez, 2012; Newgent, Davis, & Farley, 2004; Renfro-Michel, & Sheperis, 2009) involved the completion of the same set of self-report survey questionnaires at three points in time, usually over the course of one or two semesters. Only five of the investigations were supervision process studies (Bilodeau & Lecomte, 2010, 2012; Burke, Goodyear, & Guzzardo, 1998; Chen & Bernstein, 2000; Patton & Kivlighan, 1997), where questionnaire data were collected on a session by session basis; in 3 of those 5 studies, some form of session interaction/behavior rating or coding was also employed. The number of tracked supervision sessions across investigations was 3 (Chen & Bernstein, 2000), 4 (Patton & Kivlighan, 1997), 5 (Bilodeau & Lecomte, 2010, 2012), and 10 (Burke et al., 1998).

Analyses

The analyses used ran the gamut of possibilities, ranging from the simple (e.g., determination of frequencies) to complex (e.g., path analysis, hierarchical linear modeling).

Findings/Conclusions

The findings/conclusions were largely as hypothesized and highly consistent with supervisory alliance theory. A strong or favorably rated supervisory alliance was found to be linked to such variables as: higher supervisee self-efficacy and well-being, greater willingness to self-disclose during supervision, more satisfaction with supervision, more job satisfaction, greater perceived effectiveness of supervision, more availability of coping resources, secure attachment style, more supportively-perceived gender events during supervision, an attractive, interpersonally sensitive supervisor style, higher interactional complementarity between supervisee and supervisor, higher supervisee and supervisor racial identity statuses, more discussions of culture in supervision, more favorable perceptions of supervisor ethical behaviors, greater supervisor relational ability, and more frequent yet appropriate supervisor self-disclosures. A weak or unfavorably rated supervisory alliance was found to be related to such variables as: Supervisee avoidant attachment style, higher degree of perceived stress, more exhaustion and burnout, greater amount of role conflict and role ambiguity, and more frequently perceived occurrences of negative supervision events. While some support for all measured dimensions of the supervisory alliance was found (i.e., the Bond, Goal, and Task of Working

Alliance Inventory; Rapport and Client Focus of Supervisory Working Alliance Inventory), the Bond and Rapport elements emerged most strongly and consistently across studies.

Study Limitations

As Table 2 brings into focus, the vast majority of supervisory alliance studies have been ex post facto and cross-sectional in nature. The supervisee’s perspective has largely been the subject of interest, with the supervisor’s perspective being tapped in less than a third of the investigations. Only one study (Deal et al., 2011) was an intervention-based randomized controlled trial. Any attention to supervision process has been virtually absent. The one study, in which alliance measurements were taken across ten supervision sessions, was published more than 15 years ago (Burke, Goodyear, & Guzzardo, 1998), with the actual data having been collected around 1990 or 1991 (Burke, 1991).

Discussion

The supervisory alliance has come to be increasingly regarded as the crucial and pivotal component in the successful prosecution of the supervision relationship, considered to substantively affect and contribute to its ongoing process and eventual outcome. If we define clinical validity as the convergence of informed professional opinion over time about the significance of a particular variable for practice efficacy, then it seems the supervisory alliance now enjoys a unanimous endorsement of its clinical importance for supervisory conceptualization and conduct, and the endorsement transcends discipline lines, theoretical boundaries, and type of work setting (Brown, 2012; Dewald, 1997; DiGiuseppe, 2011; Farber, 2012; Gilbert & Evans, 2000; Goldman, 2012; Karon, 2008; Levenson, 2012; Morgan, Soetaert, & Heinrichs, 2008; Newman, 2012; Wood, 2005). The clinical usefulness and validity of the supervisory alliance for supervision do not seem to be disputed.

The vision of the supervisory alliance as tripartite in nature—involving a solid or “good enough” bond or rapport, shared goal or goals, and shared tasks between supervisor and supervisee—also appears to have gained wide acceptance. During the past 25-year period, those three elements have been increasingly presented as sine qua non dimensions of the supervisory relationship, and supervisors have been accordingly urged to attend to them when beginning supervision and to monitor them over the course of the relationship (Alonso, 2000; Bernard & Goodyear, 1992, 2014; Ellis, 2012; Ellis, Siembor, Swords, Morere, & Blanco, 2008; Falender & Shafranske, 2004; Gilbert & Evans, 2000; Gordan, 1996; Hughes, 2012; Jacobs, David, & Meyer, 1995; Mead, 1990; Teitlebaum, 1990a; Watkins, 2013b; Williams, 1995). In addition to providing guidance, supervision goals and tasks largely derive their significance from the mutuality of the supervisor-supervisee discussion and agreement process that leads to their very establishment. In that sense, the mutuality of goal/task establishment also contributes to the building of supervisory bond or rapport.

Research across the last two decades increasingly regards the supervisory alliance as a highly robust (if not the most robust) empirical variable of substantial import within supervision scholarship (Inman & Ladany, 2008; Ladany & Inman, 2008, 2012). Yet considering the description and summary of studies provided in Table 1 and 2, it seems reasonable to ask: How truly robust is it? More than 15 years ago, in what remains one of the most rigorous, comprehensive reviews of the supervision literature ever done, Ellis and Ladany (1997) critiqued the emerging empirical studies on the supervisory alliance; they concluded that the studies suffered from a lack of randomization, failed to address numerous threats to internal validity, and failed to control for many potentially competing, confounding extraneous variables. Some 12 years later, Bernard and Goodyear (2009) offered their own critique on what had then grown to 19 supervisory alliance investigations; they accentuated that, while the results were generally supportive, the studies were correlational in nature and that any causal inferences could not be drawn. The conclusions presented by Ellis and Ladany (1997) and Bernard and Goodyear (2009) find an echo in what is said here. While the data were generally supportive of the supervisory alliance, these studies were virtually all correlational, lacked randomization, and were subject to possible influences from a number of uncontrolled variables.

Almost all the investigations were “one-shot efforts” in which participants completed a set of questionnaires either by mail or online. In the vast majority of the research, only the supervisee’s perspective was assessed; client/patient perspective was assessed in but one study; supervision in the workplace was studied minimally, and the supervisory alliance in process (or over time) also was studied minimally. These findings are consistent with those identified in the earlier review (Watkins, 2010), further accentuate the limited range that defines the investigations, and bring into focus the lack of attention given to alliance areas of eminent concern (e.g., alliance rupture and repair). The supervisory alliance may be reasonably regarded as a robust supervision variable; however, any such view must be tempered by the features of available data and what those features indicate about the current state of knowledge. In many respects, supervisory alliance study is still in its formative stage. Our knowledge base largely remains descriptive and correlational.

The current state of supervision research has been likened to psychotherapy research in the 1950s and 1960s, where focus was on developing valid measures and demonstrating effectiveness (Milne, Leck, James, Wilson, Procter, Ramm, … & Weetman, 2012). As Milne et al. (2012) have indicated, “we are currently about ‘half-way there’, working on the ‘search for scientific rigour’…” (p. 144). That 1950s/1960s comparison seems to capture where supervisory alliance research is in its development. Two other factors that may contribute to this developmental delay are: (a) the general lack of programmatic research in supervision; and (b) the slow output of such supervision research. Unfortunately, with few exceptions (Milne, 2009, 2014; Milne & Reiser, 2012, 2014), programmatic research in supervision is more rarity than reality; that has long been the case and very much remains so (Ellis, 2010; Ellis & Ladany, 1997). Ladany has been the only programmatic force in the supervision alliance research thus far (see Table 1); but most of his work over the past decade, Ladany, Mori, & Mehr, 2013, excepted has been in more mentoring capacity than otherwise, and having now moved into a university deanship position, it is unclear how that change might affect his continued pursuit of alliance research. As for the matter of supervision study output, it is quite limited overall, with only around ten investigations per year (Ladany & Inman, 2008). With that being the case, it is easier to understand why alliance research is at such an early phase in its development.

But if supervisory alliance research is to begin moving beyond the descriptive, correlational, single-shot study, what must be done? Some of the directions are indicated by what is lacking from these 40 studies: (a) investigation of the supervisory alliance in process, including attention to the alliance rupture and repair experience; (b) tapping multiple perspectives when measuring alliance; and (c) taking a methodologically diverse and diversified approach to alliance research.

Investigation of the Supervisory Alliance in Process

The supervisory alliance does not instantaneously occur; rather, it is a supervisor-supervisee construction that develops over time and is built through sustained interaction. Furthermore, ruptures to the supervisory alliance can be expected to occur over the course of supervision and repairs will be needed to realign the supervisor-supervisee relationship. If we are to understand the supervisory alliance as a process construction punctuated by rupture/repair events, then effort must be made to study the very process nature of the alliance itself. Such efforts have been minimal to this point, with any study of supervison for more than but the most meager number of sessions being non-existent. It is equally surprising that, with the supervision rupture/repair phenomenon being so integral to alliance theory, virtually no study of that has yet been done. The impressive body of work conducted on alliance rupture and repair in psychotherapy research (Safran, Muran, & Proskurov, 2009) could be used to stimulate thinking about research possibilities on supervision alliance rupture/repair (cf. Safran, Muran, Stevens, & Rothman, 2008). Milne et al. (2012) have also provided direction on how utilization of the fidelity framework, adapted from health behavior change research, could make supervision research more sound and move it beyond the 1950s/1960s empirical mentality. Other useful advice on conducting sound supervision research can be found in both older and more recent publications (e.g., Ellis & Ladany, 1997; Ladany & Malouf, 2010). While information on how to improve supervision research has been with us for a long time, it does not appear to have been closely heeded across many alliance studies.

Tapping Multiple Perspectives When Measuring Alliance

With the supervisory alliance being coconstructed, it would seem best to tap into the perspectives of the two primary parties. That has not been the norm. Based on the body of research, it appears that supervisor and supervisee alliance perspectives can indeed diverge (e.g., Deal et al., 2011). But what might such divergences mean? What differences might they make in the whole of the relationship? Those are unanswered questions that merit empirical scrutiny. If the alliance is to be understood, attention must be given to the two alliance perspectives that are represented, the ways in which they converge and diverge, and why.

In addition, to gain perspective on the potential impact of the supervisory alliance, the gathering of client/patient data could also be of tremendous benefit. How might the alliance affect clients in supervised treatment? Again, this has been ignored and it merits attention (cf. Hill & Knox, 2013). Although assessing impact of supervision on client outcome is a particularly thorny issue (Bennett & Deal, 2012; Lambert & Ogles, 1997), it remains a vital if not preeminent concern. As Lichtenberg (2007) has asked: How can we continue to justify psychotherapy supervision training if we cannot empirically show supervision’s effectiveness on client outcome? While we have but the most minimal data to suggest any such favorable outcome (Bambling, King, Raue, Schweitzer, & Lambert, 2006; Callahan, Almstrom, Swift, Borja, & Heath, 2009), it is clear that this type of challenging research can be done and be done well (Watkins, 2011a). But if we are to know what clients think and feel about treatment and consider the possible treatment impact of the supervisory alliance, then client data will need to be routinely gathered as a part of regular research procedure. It may be that any such supervision-outcome connections will have to be made by means of mediated or path models (Goodyear & Guzzardo, 2000; Wampold & Holloway, 1997), but for that to happen, client data will have to first become a standard component of the supervision research process.

Taking a Methodologically Diverse and Diversified Approach to Alliance Research

In reviewing these 40 studies, minimal methodological diversity was in evidence. While there are many empirical paths by which the supervisory alliance could be examined, the typical approach has been to administer a set of questionnaires to participants once. Only two studies here incorporated a case study design (Burke et al, 1998; Chen & Bernstein, 2000), one used a mixed-methods approach (Ladany et al., 2013), and one was a randomized controlled trial (Deal et al., 2011). We lack a diverse attack—quantitative, qualitative, and mixed methods—on the supervisory alliance, and if research in this area is to advance productively, the pursuit of a plurality of investigative avenues serves us best (cf. Ladany & Malouf, 2010; Milne et al., 2012). In conjunction with methodological diversity, alliance research has also lacked multi-method, multi-trait, multi-rater assessment. Research would benefit from observational data gathering and analysis. Behavioral data were coded in only three studies. Now our corpus of supervisory alliance data is based almost exclusively on self-report survey responses. While that form of information can be quite valuable, more diverse forms of data are needed if we are to build a more complete empirical picture of the supervisory alliance.

Other Factors

How might the supervisory alliance be affected by workplace considerations (e.g., particular work setting demands, type of clientele served)? For example, the “on–the-go” supervision that often occurs in medical settings can be quite different from supervision in a university psychology clinic or counseling center (cf. Morgan et al., 2008). The evaluation of supervision and alliance impact in work sites outside the university setting is quite limited, with only six workplace studies being included in this dataset. Workplace investigations are sorely needed for our understanding of alliance to advance (see Livni, Crowe, & Gonsalvez, 2012; Williams, Helm, & Clemens, 2012). As Livni et al. (2012) have recommended:

Those possible differences await study and seem important to address.

Admittedly, supervision is one of “the most complex of all activities associated with the practice of psychology” (Holloway & Wolleat, 1994, p. 30). For that reason, it can be a difficult proposition to investigate: Compared to psychotherapy research, problematic effects are multiplied in supervision research by the addition to the process of a third participant (the supervisor), a second level of intervention (supervisory techniques), and the resulting, complex interactions (Russell, Crimmings, & Lent, 1984, p. 668).

As Russell et al. (1984) have indicated, creative solutions—that will no doubt typically be labor-intensive affairs—will be required for proper supervision study to be conducted (e.g., gathering data at multiple sites, extending study duration over time so that sample size can be increased). But for that to happen, coordinated, professional (or interprofessional) collaboration is a necessity. Supervision is an important, required activity across clinical, clinical health, counseling, and school psychology training, as well as social work, psychiatry, counselor education, and nursing (e.g., Buss & Gonge, 2009; MacDonald & Ellis, 2012); but as Allen Hess (2008)—clinical psychologist and seminal contributor to the supervision literature—indicated in one of the last publications before his death, “the bulk of research in supervision is the fruit of counseling psychologists and appears in counseling journals” (p. 20). Such was the case with these 40 studies and, ideally, that would not be so: Supervision is the property of us all and would benefit immensely from being researched as such. If supervision alliance research is to overcome some of the challenges identified here, if “problematic effects” are to be avoided and “creative solutions” found, then working together to investigate empirically this “most complex”, yet supremely significant training staple would seem optimal if not requisite.

Conclusion

Competencies have now become the “zeitgeist of supervision discourse” (Holloway, 2012), and perhaps the most crucial competency of all lies in being able to establish and maintain an effective supervisory working alliance with supervisees. On the one hand, the supervisory alliance has come to be widely and broadly embraced as a practice variable of untold import, its clinical validity seemingly indisputable; on the other hand, however, the empirical validity of the alliance construct—while not without some foundation—tends to be somewhat limited overall and might be considered more tentative than otherwise at this time. In this review, I have described the two dominant supervisory alliance visions that continue to hold sway and critiqued the alliance research across the last 20 plus years. Compared to psychotherapy alliance research, which a decade ago was then identified as involving well over 1000 empirical findings (Orlinsky, Ronnestad, & Willutzki, 2004), the number of supervision alliance research findings pales pitifully by comparison. Let us hope that that will change.

Sue Wheeler (2007)—in an introductory editorial of a special issue on supervision—stated that the “future for research on supervision is wide open” (p. 1). I believe that her statement readily applies to where we are now with regard to the supervisory alliance. While supervision is a difficult subject to tackle, the potential impact of the supervisory alliance on the teaching and learning of psychotherapy is inestimable. It would be valuable for psychotherapy educators to know the various ways in which that potential impact occurs, the mechanisms by which the supervision alliance works, and how we as supervisors can more meaningfully and effectively contribute to making supervision a safe place and space within which supervisees can blossom and grow.

Psychotherapy supervision may well be the most substantial vehicle by which we teach, transmit, and perpetuate the traditions, practice, and culture of psychotherapy. The supervisory alliance is the medium by which that teaching, transmission, and perpetuation process occurs. Increasing and improving our empirical efforts is requisite if we are to: (a) have a more complete, anchored understanding of the supervisory alliance; and (b) accordingly enhance its practical application and implementation throughout the supervision process.