Supervision matters: Collegial, developmental and reflective approaches to supervision of teacher candidates

This self-study examined the communication approaches of 15 university supervisors who oversaw teacher candidates enrolled in year-long, co-taught P-12 clinical experiences. Supervisors attended 20 hours of professional learning on pre-service co-teaching, developmental supervision, and instructional coaching. Findings indicated that our supervisors primarily used collaborative and nondirective communication approaches to improve candidates’ teaching skills, while facilitating collegial relationships between collaborating teachers and teacher candidates engaging in co-teaching, co-reflection, and co-generative dialogs. Ultimately, our findings suggested that these collaborative and nondirective communicative approaches played a significant role to enable our candidates to demonstrate selfdirected and self-regulated learning. Subjects: Education Social Sciences; Post 16 Education; Primary Education – Teaching Practice; Secondary Education; Teacher Education & Training; Teachers & Teacher Education; Teaching Practice Education

Since 2010, an interdisciplinary team of teacher educators at a large, state university has researched the impact of a comprehensive initiative to transform traditional student teaching into a year-long, co-taught clinical experience. As part of this initiative, our researchers designed, developed, implemented, and evaluated the effectiveness of a set of interdependent practices, including year-long clinical placements, pre-service co-teaching, and developmental supervision. To document the effectiveness of our new practices, our researches used a variety of mixed methodologies in a number of small studies. While the overall findings are preliminary, they are being used effectively to identify promising practices and inform future development. Members of our research team have shared the results of these small studies at national and international conferences as well as independent schools in Brazil, Ecuador, and Costa Rica.

PUBLIC INTEREST STATEMENT
As teacher effectiveness is of international concern, it logically follows that the methods that are used to supervise and develop new teachers are also of concern to international audiences. This study examined the communication approaches used by university supervisors of undergraduate students who were seeking state certification to teach in either elementary, middle grades, or high school settings. The findings revealed that undergraduates demonstrated increased confidence and competence when their supervisors used communication approaches that were primarily collaborative and non directive. Collaborative and non directive approaches are those that encourage individuals to set their own professional goals and monitor their progress toward meeting those goals. These approaches are diametrically opposed to those where individuals are told directly what to do and how to do it. People who work outside of teacher education may find these results relevant to the supervision of their employees.

Introduction
Student teaching has been heralded as the most important aspect of teacher education (Evertson, 1990;Sullivan & Glanz, 2000), and potentially, the most problematic (Goodlad, 1990;Rodgers & Jenkins, 2010). With a goal to improve the student teaching experience, numerous authors have written about changes in the roles and responsibilities of cooperating teachers and university supervisors (Rodgers & Jenkins, 2010), effective pedagogy in supervision (Blanton, Berensen, & Norwood, 1999;Glickman, Gordan, & Ross-Gordan, 2014) and the integration of mentoring practices into supervision (Rogers & Keil, 2007). Recognizing the important role of student teaching in preparing teachers, National Council of Accreditation in Teacher Education (2010). The National Council of Accreditation in Teacher Education challenged colleges of education to "turn teacher education upside down" (p. 2) and focus on clinical practice rather than coursework. The emphasis on clinical practice created a critical need for the development of school-university partnerships to serve as the vehicle for shared responsibility in the preparation of the teacher workforce.
In response to the national call for change, many educator preparation programs across the United States formalized efforts to transform the field and clinical experiences required of candidates in initial certification programs (Darling-Hammond, 2000. These changes focused on improving candidate effectiveness by assuring coherence of clinical experiences with university coursework (Hubball & Gold, 2007), increasing the number of year-long, co-taught clinical placements (Bacharach & Washet-Heck, 2012;Williams, Gray, & Stockdale, 2012), and improving schooluniversity partnerships (Stephens & Boltd, 2004). While these improvements and reform efforts were carefully researched and documented in the literature (Bacharach et al., 2010;Badiali & Titus, 2012;Strieker et al., in press), little if anything, has been reported on efforts within teacher education programs to transform the pedagogical practices of university supervisors who evaluated the performance of candidates enrolled in these co-taught, clinical experiences.
Given this situation, our interdisciplinary team began a college-wide initiative to transform our practices in university supervision by aligning them with our co-taught model of clinical experience, thereby making them more developmental, reflective, and collegial in nature. After careful study, our team designed a set of practices and provided year-long professional learning for supervisors. The purpose of this study is to examine the communication practices of 15 university supervisors overseeing candidates in a year-long, co-taught, P-12 teaching experience. Specifically, the research questions that guided our study are: (1) What communication and behavioral approaches did our university supervisors exhibit during structured dialog on teaching and learning during supervisory sessions with candidates enrolled in co-taught clinical experiences?
(2) What is the content of the structured dialog on teaching and learning during the supervisory sessions?

Theoretical framework and literature review
In large part, the theoretical framework is undergirded by Glickman's (1981) and Watkins' (1995) theories of how the interpersonal behavior and communication approaches used by a supervisor can be selected and/or altered to meet the developmental level of the teacher (candidate). According to these theories, in the early stages of development, the supervisor provides "a safe, secure, and structured environment … and provides direction, guidance, and advise" (Watkins, 1995, p. 650). As the individual progresses through the middle stages, the supervisors become less direct, providing the individual with more freedom and flexibility to try more approaches. When the individual reaches the advanced stages, the supervisor functions in a collaborative/facilitative role by encouraging the candidate to become self-directed in his or her own learning. Throughout the entire developmental process, the university supervisor guides the course of the conversation by asking questions that engage the individual's thinking, while concurrently guiding the metacognitive development and the ability to think independently and critically (Glickman et al., 2014).

Developmental supervision
Supervision is generally viewed as an oversight function, where one is responsible for evaluating the performance of a subordinate. Developmental Supervision is a specific approach to supervising practicing teachers in schools wherein the supervisor adjusts his or her communication and style of interaction based on the adult and professional needs of the teacher (Glickman et al., 2014). Inherent in Developmental Supervision is the assumption that because teachers operate at varying levels of conceptual understanding, ability, and effectiveness, they need to be supervised in ways consistent with their needs. As the teachers become self-regulated and self-directed learners, Developmental Supervisors naturally shift from an expert to a facilitator role, which is more collegial, cooperative, and nondirective. Research on the effectiveness of Developmental Supervision (Zellermayer & Margolin, 2005) indicated that supervisors who use this approach are flexible because they are able to shift their approaches based on the needs of teachers and groups.
Developmental supervisors use the nondirective approach (Glickman et al., 2014) to provide a safety net for teachers to engage in a series of interdependent activities: teachers establish professional goals, create self-improvement plans, conduct progress monitoring, and think critically about teaching and learning. With this recursive nature of the components of the non-directive approach, developmental supervisors are more reflective in their practice than supervisors in control groups (Siens & Ebmeier, 1996).

Changes in conceptual language and framework for collaboration
In our own attempt to make our approach to university supervision more collegial and reflective, we adopted Knight's (2007) Partnership Principles as the common language and conceptual framework for developing substantive relationships to support collaboration, co-teaching, and co-generative dialog. The Partnership Principles are: equality, choice, voice, dialog, reflection, praxis, and reciprocity.
In terms of co-taught clinical experiences, equality implies that even though the candidates do not hold state teacher certification, their knowledge, skills, experience, and contributions are of equal value to those of the collaborating teachers. Furthermore, each person has voice and choice in decision-making; therefore, the candidates' voices are heard throughout the cycle of co-teaching, including co-reflection (Center for Educational Partnerships and Placements [CEPP], 2015). After the candidate's monthly observations, the supervisor joins the dialog and adjusts his or her approach to meet the developmental needs of the candidate. The fluid nature of the approach supports the candidate in applying the knowledge and skills learned in university coursework to the realities of the classroom (praxis). As the collaborating teacher, candidate, and supervisor share experiences and engage in collaborative inquiry (Williams et al., 2012), they form a professional learning community (PLC). Note that heretofore all references to the candidate, collaborating teacher, and supervisor as a collective group will be written as PLC.

Conceptual language and framework for university supervision
We adopted Developmental SuperVision (Glickman et al., 2014) as the conceptual framework, common language and benchmark in the revision of our own approach to university supervision. Inherent in Developmental SuperVision are four communication and behavior approaches and two roles for that supervisors use when working with teachers in P-12 schools. For example, supervisors select the approaches and roles based on the developmental needs of the teachers. When teachers are new to the profession, supervisors should assume an expert role and engage in Directive Control and Directive Informational approaches. As the teachers become more mature and improve with knowledge and skills, Glickman et al. (2014) recommended that supervisors assume a facilitative role and engage in collaborative approach, and nondirective approaches.
The Directive Control approach is necessary when the teacher (candidate) demonstrates or expresses as urgent need for support. When this occurs, the supervisor assumes an expert role and provides high levels of input, direct explanation, and frequent observations. Even in a co-taught classroom, the university supervisor tends to conduct the Directive Control approach without input from the collaborating teacher. The Directive Informational approach is typically used when the candidate enters a time of exploration, with a high learning curve. This situation also calls for the supervisor to assume an expert role to provide explanations of methods, resources, and recommendations for alternative solutions to problems of practices. In co-taught classrooms, the collaborating teachers are asked to make recommendations, demonstrate preferred practices, and model teacher thinking.
As the candidate gains confidence and capacity, the supervisor shifts from the expert role to the facilitator role. The collaborative approach provides the vehicle for the candidate to continue to explore new options, while thinking more critically about teaching and learning. With the collaborative approach, the supervisor reduces the amount of advice given to the candidate and begins to co-reflect and co-generate solutions to problems of practice. The candidate's collaborating teacher is also encouraged to actively participate and contribute to co-reflection and co-generative dialogs.
As the candidate becomes increasingly self-directed, the supervisor continues to facilitate the dialog, but listens more and talks less. With the nondirective approach, the supervisor has a low level of input (i.e. directions), asking clarifying questions that are reflective and encouraging. Nondirective approaches provide the vehicle for the candidates to establish goals and plans for self-improvement and self-monitor their progress toward meeting the goals. We adopted Whitemore's (2002) GROW model for the candidates to establish their goals and selfmonitor their progress. We believed that the goal-setting process was critical because it fostered the candidate's ability to become more confident and self-directed. During the goal-setting session, the supervisor and the collaborating teacher guide the conversation to assist the candidate in the following areas: (a) establishing goals for personal improvement; (b) sharing their knowledge, skills, and resources; (c) co-constructing assessment; (d) co-constructing instruction; and (e) co-generating solutions to complex problems of practice. As candidates learn to monitor their progress, their supervisors often agree to collect data during the classroom observations, which creates a shift in the role of the supervisor from facilitator of GROW to one with co-assessment.
Co-assessment, in turn, creates opportunities for the supervisor to actively participate in co-reflection and critical dialog. Once the supervisor actively engages in the dialog, the candidate has increased opportunities to observe models of thinking about practice and learn from two experienced, practicing professionals.

Classroom coaching
Classroom coaching is different from supervision in that it is a form of professional development that is nonevaluative (Tschannen-Moran & Tschannen-Moran, 2011). While coaching is most commonly employed in public schools to improve the performance of practicing teachers, and ultimately, their P-12 students, it also implies a desire for growth on the part of the teacher. It is not some sort of disciplinary action. Classroom coaching is a generic term for a number of practices that include instructional coaching (Knight, 2007), differentiated coaching (Kise, 2006), literacy coaching (Stover, Kissel, Haag, & Shoniker, 2011), and content coaching (Tschannen-Moran & Tschannen-Moran, 2011). Knight and van Nieuwerburgh (2012) reported instructional coaching has demonstrated a positive impact on student achievement.
According to Bearwald (2011), effective coaches guide the conversation by asking critical questions, rather than offering solutions or making recommendations. Coaches encourage teachers to use high-leverage, evidence-based practices (Knight, 2013) in one or more of the following areas: (a) planning lessons based on high standards; (b) using formative assessment to monitor student performance; (c) employing high-leverage instructional practices; and/or (d) relationship and community building (Knight & van Nieuwerburgh, 2012). In doing so, the coach facilitates a dialog where teacher candidates and collaborating teachers have the opportunity to theorize the lesson and seek to understand the theory-to-practice (or practice-to-theory) implications as they co-generate ideas for improving teaching and learning (Roth, Tobin, Camambo, & Dalland, 2004).
Based on this review of the literature and theoretical framework, our interdisciplinary team of faculty designed, developed, and offered a year-long series of opportunities for professional learning to 15 university supervisors. They were assigned to one partner school district as a field site to host year-long, co-taught clinical experiences to 45 teacher candidates (P-12). The purpose of this study was to examine and describe the pedagogical practices of university supervisors who participated in the professional learning and supervised teacher candidates in that district.

Self-study in teacher education research
Within the teacher education research community, there has been growing interest in employing reflective, self-study approaches (Grossman, 2005;Loughran, 2007) to better understand teaching and learning. Self-studies encompass a broad range of qualitative research traditions such as narrative inquiry, case study, or action research (Lassonde, Galman, & Kosnik, 2009). Zeichner (2007) called upon teacher education researchers to engage in self-study in an effort to "… contribute to the improvement of teacher education practice and to our broader knowledge about particular questions of significance to teacher educators and policy makers" (p. 43). While our study focused on teacher education, the conducted self-study was also a case study that used a dynamic action process. Our research and analysis allowed us to make changes to our program, contributions to our teaching, feedback, and development for our supervisors. We followed this research tradition since it has had a strong influence on self-study research and has been referred to as a "useful tool for self-study" because it provides a method to conduct systematic inquiry into one's practices (Feldman, Paugh, & Mills, 2004, p. 970). The current investigation responds to his call by situating the present study in the larger context of research on the pedagogical practices of supervisors of practicing teacher candidates, enrolled in year-long, co-taught clinical experiences.

Program context
The supervisory programs and practices examined in this study were designed, developed, and analyzed at a state university, located in the metropolitan area of a large city in the southern region of the United States. At this institution, each year approximately 1,000 prospective teachers complete initial certification programs in early childhood, elementary, middle and secondary education, special education, and instructional technology. Of those who graduated in 2015, approximately 85% self reported as white (non-Hispanic), 9% as black (non-Hispanic), 2% as Hispanic, 1% as Asian, 1% as multi-racial, and 2% are undeclared. The teacher education unit is fully accredited by National Council for Accreditation of Teacher Education/Council for the Accreditation of Educator Preparation. Their respective national professional associations also accredit individual programs.

Informants in the study
All of the informants met the prevailing state guidelines for university supervision in that they had taught a minimum of three years and were certified in the area that they supervised, including Elementary, Social Studies, Math, English, or Physics Education. The cadre of university supervisors was comprised of 13 females and 2 males. Based on their self-reports, 4 were retired school principals, 11 were retired teachers, and 1 was a member of the university faculty. Thirteen were Caucasian, 1 was Hispanic, and 1 was African American. While most of the supervisors held a master's degree, one-third held doctoral degrees. Overall, the average number of years of experience supervising coteachers and/or assisting new teachers through induction was 25.
The supervisors who participated in this study were assigned to 41 pairs of teacher candidates (along with their collaborating teachers) in 5 elementary schools, 2 middle schools and 4 high schools. Each supervisor was assigned four to six candidates enrolled in co-taught, year-long clinical experiences throughout the school year. The supervisors were assigned to schools affiliated with the formal district-university agreement that clinical experiences offered during a university candidate's senior year would be year-long in duration and co-taught by the candidates and collaborating teachers. Furthermore, the district agreed to assign our teacher candidates to collaborating teachers who would not only co-teach the entire year, but who would participate in professional development series that consisted of a six-hour interactive seminar on pre-service co-teaching and four online modules on how to use co-teaching to support classroom management, differentiated instruction, and student engagement. The candidates were required to complete the professional development series, along with their collaborating teachers.
The university agreed to send all of the university supervisors and candidates who were assigned to this district to the same seminar. Throughout the entire year, each supervisor who participated in our study, also agreed to attend monthly sessions that addressed research-based practices in supervision, including (but not restricted to) Partnership Principles, Goal-Setting, Data Collection and Progress Monitoring, Effective Communication and Questioning, Video and Virtual Approaches to Supervision, Differentiated Coaching and Supervision, Developmental Supervision, and edTPA (American Association of Colleges for Teacher Education, 2013). Additionally, supervisors were responsible for submitting monthly reports, protocols, and reflections as well as brief surveys on their monthly activities. For the additional responsibilities, the supervisors were awarded a stipend through our federal project.

Data sources
The data consisted of artifacts gleaned from the supervisors' reports and protocols, personal reflections, and surveys collected at the monthly meetings. Specifically, each month all supervisors uploaded their completed Observation and Reflection Protocols (adapted from Costa & Garmston, 2002; see Appendix 1), Problem-Solving Protocols (adapted from Costa & Garmston, 2002; see AppendiX 2), and GROW (adapted from Whitemore, 2002; see Appendix 3) as well as the monthly surveys on a web-based data repository. Regular submission of this information provided a set of qualitative data-set for our analysis; which subsequently, described the communication approaches used by the participants as well as the content of the dialog.

Development of coding instrument
Given that the context for our study is relatively new, there were few instruments available that could accurately and systematically classify and quantify the information from our extensive dataset. Therefore, our research team created an instrument that would ensure consistency in the analysis of multiple researchers. To that end, the first author developed a set of codes and categories aligned with promising practices in clinical experiences, currently known as Developmental SuperVision (Glickman et al., 2014), Instructional Coaching (Knight, 2007(Knight, , 2013 and Pre-service Co-teaching (Strieker et al., in press). Center for Educational Placement and Partnerships (2015) Once the initial codes and categories were established, the second author, acting as a peer de-briefer, independently reviewed the coding scheme and made recommendations for revisions. Next, four co-authors independently used the second draft of the instrument to code data collected from the observation and goal-setting protocols. After the first round of data analysis, the co-authors met to review, compare, and revise the codes and categories through dialog and agreement, ultimately creating the Coding Instrument for Supervisory Data (CISD; see Appendix 4) The final version of the instrument reflects the revised codes and categorizations based on the co-authors' shared-understanding of the literature and group consensus.

Protocol analysis
Our research team used the CISD to analyze the GROW (goal-setting protocols), the Problem-solving Protocols, and the Observation and Reflection Protocols created by our supervisors to indicate the presence and degree of evidence of emerging practices during feedback sessions with their candidates while facilitating goal-setting sessions, co-generative dialog, and co-reflection. Table 1 illustrates examples of codes with anchored evidence found in the original data. Finally, the monthly reflections were qualitatively analyzed to identify comments of the supervisors related to the emerging practices described above.
The analysis of monthly protocols and surveys (4 × 15 = 60 documents) provided clear evidence of each supervisor's emerging practice as well as the content of their discussions. By quantifying the evidence in percentages, we see emerging patterns of practices typically associated with Developmental SuperVision and pre-service co-teaching. Specifically, the researchers examined the frequency of the following behaviors, including: (a) the use of standard protocols; (b) approach to facilitation of dialog in goal-setting and problem-solving sessions; (c) support for candidates as they integrate and apply the knowledge and skills learned from coursework to their daily practice; (d) description of the co-teaching practices of the collaborating teachers and teacher candidate; and (e) the language used to bring closure to feedback sessions (see Table 2).

Use of adapted protocols
Through our CISD analysis, we found that all of our supervisors used our adaptation of the GROW framework (Whitemore, 2002) to support their candidates in establishing personal improvement goals, developing plans to meet their goals, and monitor their progress (see Appendix 1). Furthermore, 50% of the Grow protocols were fully completed, which indicated that our supervisors had attempted to complete all of the steps in the GROW process. Of the protocols that were partially completed, 93% contained an assessment of the current reality, 93% identified data collection procedures to monitor candidate progress or impact on P-12 students, and 87% stated the goals for teacher candidate. Candidate goals included co-planning for classroom management and instruction.
Similarly, all of our supervisors used our adaptations of Costa and Garmston's (2002) protocols to document their observations and provide feedback to their candidates (see Appendix 2 and 3). The content of the structured dialog used with these approaches is as following: (a) co-instruction (100%); (b) co-planning (80%); (c) development of candidate-collaborating teacher partnerships (73%); (d) co-reflection on shared experiences of co-instruction (60%); and (e) co-assessment (33%) (see Table 3).
In addition, our analysis indicated that, in general, our supervisors closed the feedback sessions in the following pattern: (1) reflection on the progress of the candidates and the P-12 students in meeting their goals; (2) candidate and collaborating teacher's sharing their perceptions of the progress; and (3) negotiation of a set of recommendations for action steps, along with roles and responsibilities of each member of the PLC. At that point, approximately 40% of our candidates established new goals; and nearly all of the supervisors closed the session with a compliment. http://dx.doi.org/10.1080/2331186X.2016.1251075

Use of Glickman et al.'s (2014) communication and behavioral approaches
Figure 1 depicts our findings of evidence that our supervisors used communication and behavior approaches that were closely aligned those described by Glickman et al. (2014). Note Figure 1 depicts the variant distribution of supervisor behavior and only provides a single snapshot of one particular point in time. This snapshot does not suggest the trajectory of approaches from Directive/Control to Nondirective that was reported by Glickman (1981) and Watkins (1995). Rather, our analysis identified instances in which supervisors did demonstrate different approaches. One elementary supervisor reflected upon how she adjusted her approach based on the perceived needs of the candidates,

Figure 1. A variant distribution of four approaches to developmental supervision.
Notes: This figure depicts the variant distribution of the four communications and behavioral approaches described by Glickman et al. (2014). Note that this depiction provides the reader with a snapshot of supervisor behavior, rather than a trajectory as suggested by Glickman (1981) and Watkins (1995).
"By using more specific questions and wait time, I kept learning (how to) work with the candidate in ways that would draw her into more confident articulation." Based on our findings, 53% of the data provide evidence of the nondirective approach, 47% of the Directive Informational approach, 40% of the collaborative approach, and 13% of the Directive Control approach. By conducting an analysis of the monthly reflections, we found one supervisor, who provided evidence of using the Directive Control approach stating, "She (candidate) has no idea how to conduct a guided reading lesson. No prediction … No preview of vocabulary … Sounding out is the only word attack skill used." Later, the same supervisor reported, "She relies on the CT to determine who should be in the guided reading groups. I re-iterated the need for her (the candidate) to use assessment data to determine student needs."

Evidence of nondirective approach
As noted above, the composite of distribution indicated that over half of the evidence documented the use of nondirective approach (see Figure 2). Furthermore, our evidence suggested that our candidates established one or more goals related to improving student engagement by means of (2) Teacher decision-making [27%] (

3) Theorization of practice [20%]
Co-instruction Supervisor facilitates a discussion of: • Willingness of the collaborating teacher to engage in co-teaching [20%] improving their own abilities to manage their classrooms and/or differentiate instruction. In terms of classroom management, 67% of the evidence on candidate goals specifically addressed classroom management; and another 67% addressed student engagement.
In terms of instructional practices, 47% of the evidence on candidate goals specifically addressed data-driven instruction; 33% addressed effective questioning; 20% addressed student learning; and 13% addressed time management.

Evidence of directive informational approach
As noted above, the composite of distribution indicated that slightly less than half of the evidence documented use of the Directive Informational approach (see Figure 2). Furthermore, our analysis indicated that the PLCs established goals related to data-driven instruction; 33% addressed effective questioning; 20% addressed student learning; and 13% addressed time management. In terms of instructional practices, 49% of the evidence described discussions on general notions of differentiated instruction, while another 40% discussed the related topics of flexible groups and 40% discussed tiered assignments. Evidence was also provided that approximately 20% of the conversations were about research-based practices, including inquiry based instruction and cooperative learning. Finally, we noted that 6% of the data described conversations in which supervisors facilitated theorization of practice and/or justification their instructional choices.

Evidence of the collaborative approach
As noted above, the composite of distribution indicated that slightly less than half of the evidence documented use of the collaborative approach (see Figure 2). Qualitative analysis of the reflections of the supervisors revealed that our supervisors often used the collaborative approach, particularly during feedback sessions. A supervisor of secondary math reflected, "I noticed during the lesson that the candidate stayed at the front of the room, so I suggested that he move more [around the classroom] to help refocus off task students." Later, the same supervisor described the contribution of the collaborating teacher; She explained [to the candidate] that it is important to write down all of the KEY terms on the board as they arise because some students need visual confirmation. Then [the CT] offered alternative approaches to the warm-up that identified the student misconceptions typically found [when teaching] this material. Figure 2 depicts a variable distribution of supervisory foci of content of the structured dialogs. Virtually all (100%), the content focused on some aspect of co-teaching (CEPP, 2015), particularly co-planning (80%) and co-reflection (60%), followed by co-assessment by the candidate and classroom teachers (33%). Approximately 78% of the distribution described dialog that addressed the development of effective partnerships among members of the PLC. For example, one elementary supervisor commented in her reflection, "Our partnership is strong and productive … Growth and professional development for the candidate and the collaborating teacher are evident qualitatively and quantitatively. My own growth and development in coaching classroom management has been significant." Given the nature of our clinical experience, it was not surprising that 100% of the data provided evidence of conversations about some aspect of the co-teaching cycle (CEPP, 2015). Using the CISD, the preponderance of evidence revealed conversations about one or more of the models of co-instruction described by Friend and Cook (2007). At least one middle school and one high school supervisor described situations where the collaborating teachers used models of co-instruction to provide a scaffold of support for their candidates. For example, one high school math supervisor described a variation on One Teach/One Observe and One Teach/One Assist (Friend and Cook, 2007). "During first period, they do One Teach/Candidate Observes. Second period, they do One Teach/ Candidate Assists. After that, they both solo teach the content to different sections." While the co-teaching was clearly a continuous topic of conversation, all of the shared perspectives were not positive. Two supervisors described situations where collaborating teachers were hestitant to co-teach with their candidates and to collaborate with the candidate's supervisor. In fact, one high school English teacher actively resisted by refusing to co-instruct and/or attend sessions with the candidate and the supervisor. In her reflection, the supervisor of this candidate provided the following rationale, "The CT said that he preferred the traditional model of student teaching to [coinstruction]; however, he did assist the candidate by co-planning regularly." Another supervisor of a candidate in a middle school social studies classroom reflected upon a similar situation.

Supervisory foci of dialog on co-teaching
"Although somewhat apprehensive about co-teaching, the CT and candidate have a natural proclivity to do so. The first lesson was station teaching … The second lesson was team teaching. The pair [CT and candidate] is effectively co-teaching, they just don't notice how naturally is comes to them." What is clear, is that all of these situations occurred in high school classrooms. What is unclear is whether these situation persisted throughout the yearlong placements.

Discussion
At our institution, the move to year-long, co-taught, clinical experiences created a need for more developmental, reflective, and collegial approaches to supervision. This need, coupled with the university's requirement for supervisors to support our candidate's during national assessments forced the faculty to re-think the existing supervisory practices and find legitimate ways to add effective pedagogical practices. These new practices had to meet not only the developmental needs of the candidates, but also the needs of the collaborating teachers who were asked to co-teach and serve as mentor-models for the candidates during the year-long placements. Therefore, this study was critical to our understanding of the emerging practices of our university supervisors who were assigned to candidates enrolled in year-long clinical experiences.
One of our most salient findings was that our supervisors, candidates, and collaborating teachers engaged in ongoing collaborative inquiry and reflective dialog, which gave our candidates the benefit of focused learning from two experienced professionals. Through co-reflection, the candidates were able to meet their goals to improve their practice, increase their use of research-based instructional approaches, and/or increase their ability to adjust their approaches to meet the needs of their P-12 students. Our findings that 43% of our supervisors worked closely with either the teacher candidate, the collaborating teacher, or both to establish a strong relationship was particularly insightful in light of previous studies which indicated that the relationships among members of the PLCs are "inevitably hierarchical … . Promoting shifting alliances, one with the university supervisor on top and one with the collaborating teacher on top" (Bullough & Draper, 2004, p. 407).

Figure 2. A variable distribution of supervisory foci of conversation.
Notes: The variant distribution depicted in this graphic provides the amount of evidence that the supervisors provided of facilitation of discussion with the candidates and collaborating teachers in a dialog about the cycle of pre-service co-teaching (e.g. co-teaching, co-planning, coassessment, and co-reflection) as well as discussing their co-teaching relationship as defined by Knight (2007).

Communication and behavioral approaches used by supervisors
As compared to Developmental SuperVision (Glickman et al., 2014), our supervisors provided evidence of equal engagement in approaches that approximated the Directive Informational approach, the collaborative approach and the nondirective approach. While our approaches reflected those in cited in Developmental SuperVision, the practices of our supervisor were inherently more collegial. Increased collegiality was due to the nature of pre-service co-teaching and the demands on the collaborating teacher to act as a mentor model for the teacher candidate (CEPP, 2015). Our findings revealed that our supervisors' facilitated dialog to troubleshoot issues related to co-planning, coassessment, co-instruction, and development of partnerships. As one elementary supervisor reflected, Our partnership is positive, but at the same time, I had the feeling today that the candidate is taking the role of being less than an equal partner, even though the collaborating teacher is encouraging and supportive. The growth of the candidate, in terms of … managing and guiding the students has been obvious and I have collected the data to back up this observation.
Because candidates in all of our teacher preparation programs are required to establish professional goals during their year-long clinical experience, virtually all of our university supervisors engaged in nondirective approaches (Glickman et al., 2014). When engaged in the nondirective approach, our supervisors typically assumed responsibility for measuring student engagement during the candidate's observations by collecting data on (1) on-task behavior; (2) disruptive behavior; (3) teacherpupil interactions; (4) positive to negative comments (Sprick, Knight, Reinke, Skyles, & Barnes, 2010). Thus, when employing the nondirective approach, the supervisors shifted their role from expert or facilitator to active participant in co-assessment. One elementary supervisor described the specific procedures that she used during a monthly reflection, The goal that our team [PLC] decided to work on first was increasing student engagement. During the observations, they asked me to examine the different levels of engagement in their students in whole group versus small group settings. I will also be watching [the candidate's] classroom management in general to help her build confidence … She knows that she has to develop confidence to positively impact her students.
Finally, even though our findings suggest that our supervisors used all four of Glickman et al. (2014) communication and behavioral approaches, there was little evidence of a the developmental trajectory reported by Glickman et al. (2014) and Watkins (1995). Based on the previous work, one might expect the trajectory of candidate learning and development to evolve from highly dependent in the early stages to self-directed in the latter. Our supervisors rarely used the Directive Control approach. As a matter of fact, most of them used nondirective approaches to facilitate goal setting, and later, they used more directive approaches to provide information to the candidates.

Co-teaching
Our findings indicated that virtually of the dialog addressed some aspect of co-teaching, with over 60% of the evidence describing planning and co-instruction. Approximately half of evidence addressed differentiated instruction and/or flexible grouping and tiered assignments. While the dialog also addressed research-based practices (e.g. inquiry-based instruction and cooperative learning) the amount of data describing theorization of practice and/or justification of instructional choices was negligible.

Challenges in candidates' efforts to integrate and apply course knowledge
As noted above, there was very little, if any, evidence of the supervisor providing intentional support to the candidate in linking previous coursework to the clinical experiences during goal-setting, action planning, observations, and reflections. During co-reflective feedback, the group tended to discuss topics or practices such as differentiated instruction without referring to the research that underlies the practice. There was only one person in a goal-setting session who attempted to theorize her daily practice. One reason for the lack of theorization could be a lack of supervisor' understanding of theory (Borko & Mayfield, 1995). According to Slick (1998), university supervisors are often unable to assist candidates in bridging theoretical content and pedagogical knowledge with their daily practice in the field experience.

Co-assessment: Competing priorities and focus
According to our analysis, one-third of the data provided evidence of co-assessment. While one reason for the depressed frequency may be because the supervisors prioritized assisting their candidates in developing their relationships with their collaborating teachers as well as the co-planning and co-instruction of the content. Another reason may be because the districts mandate multiple assessments used to measure student performance, so there is little room for intensive co-construction of key assessments.
Investigating the co-assessment practices of the teacher candidates and collaborating teachers posed numerous challenges to the researchers. While the supervisors focused on the assessment and evaluation of the candidate's content knowledge, pedagogical performance, and impact, the teacher candidates and the collaborating teachers focused on the assessment of the learning and engagement of P-12 students. These competing priorities first emerged during the initial goal-setting sessions when the supervisor facilitated the candidates' self-assessment of their capacity to teach their P-12 students. Typically, the candidate selected goals to improve the following: (a) teaching (e.g. effective questioning, differentiated instruction, smooth transitions, data-based decisionmaking); (b) classroom or time management; and/or (c) student learning or engagement. It would be logical to assume that the dialog naturally flowed from the assessment of the teaching and learning process. What did the candidate do? What was the evidence to show that the candidate taught effectively? What was the evidence that the students met the objectives? What was the evidence that the students were engaged? Did the candidate adjust the practice? However, there was no concrete evidence of this occurring. Thus, we can only surmise that if the cycle occurred, the evidence was not provided on the monthly protocol rendering this type of information difficult to collect, and ultimately, analyze.
The evidence provided, however, clearly indicates that nearly all of the supervisors were active participants in the co-assessment of student engagement in learning. The data provided multiple examples, K-12 of the supervisors collecting data on student engagement during large-and smallgroup instructions. Thus, it may be that this function was monopolized by the supervisors, with little room for collaborating teachers or the candidates, to collect data on this all important aspect of the teaching and learning process. Furthermore, there was ample evidence that co-generative dialog (among the collaborating teacher, the candidate and the supervisor) occurred regularly to resolve complex issues of practice, resulting from a lack of student engagement. Thus, it appears that although evidence of co-assessment by the collaborating teacher and the candidates only occurred in 30% of the time, co-assessment by the supervisors occurred nearly 100% of the time.

Vision for future efforts to prepare collegial supervisors
Based on our experiences in designing and researching new approaches to supervision of teacher candidates, the faculty developed a new model of collegial supervision that is more developmental, reflective, and collegial (see Figure 3). This model represents our initial attempt to design a university model of supervision that meets the criteria of Boundary Object, meaning that it is flexible "… enough to suit the needs and contextual constraints of users applying it, while robust enough to maintain a common identity across different communities, practices and trends in the field" (Star & Griesemer, 1989, p. 393).

Limitations of study
First, university faculty had limited input as to where teacher candidates should be placed, and with whom, while in the field. As a result, the voices of our faculty were marginalized leaving them with limited influence on the pedagogical practices of the collaborating teachers who hosted the teacher candidates, specifically in terms of the co-teaching and desire to collaborate with the university supervisor. Thus, some teacher candidates were exposed to highly effective collaboration and coteaching, while others were not. Some teacher candidates engaged in the entire cycle of co-planning, co-instruction, and co-planning, while others were not. At least two high school teachers refused to co-teach and collaborate. This particular limitation, when looking at this work through the lens of self-study, is a factor that is not likely to change. Our program will necessarily face the dichotomy between the faculty, researchers, supervisors, candidates, and the collaborative teachers from various educational settings. One clear outcome of such dichotomy is the inconsistent student teaching experience of our teacher candidates and their future teaching practice in the classroom.
The second limitation in the study was the lack of diversity in the participant sample. The sample was one of convenience where individuals were invited to participate based on the assignments of the supervisors to the single district with an articulated school-university agreement to host teacher candidates in co-taught classrooms. This agreement required all of teacher candidates, collaborating teachers, and university supervisors to complete professional development on pre-service coteaching instructional coaching and Developmental Supervision. Given the changing demographics in the national student body, the study should be replicated with a diverse demographic of candidates, teachers and supervisors. It is also not representative of the demographics of our programs.

Implications and future studies
Based on these findings, we recommend that university supervisors receive ongoing professional learning. Obviously, to increase the coherence of the university coursework to subsequent clinical experiences, professional learning should focus on applying the content of the coursework of the candidates to the classroom context with students. In addition to content, university supervisors should receive ongoing development on implementing effective communication, facilitation, and pedagogical practices to ensure that effective supervision be more collegial, developmental, and reflective. This is particularly true in universities that have adopted year-long co-taught clinical experiences.
Furthermore, it appears that the nondirective approach to supervision creates an effective vehicle for supervisors to support candidates in establishing their own self-improvement goals, becoming more self-directed in their own professional learning. The nondirective approach also provides the supervisor with various forms of data collection on the teaching and learning process, which provides the candidate with specific measures to monitor their progress toward meeting their goals as well as the progress of their P-12 students (Glickman et al., 2014).
Given the fact that there was little, to no, evidence of that our supervisors were able to assist their candidates in theorizing their practice and justifying their instructional decisions with research, we recommend that future research address the following question: To what degree do supervisors use best practices to support teacher candidates in implementing the methods taught in university coursework in their year-long clinical experiences? Potential sources of data include the following: (a) candidates' voices (interview or written comments); (b) video clips of co-generative dialog among the collaborating teacher, the candidate and the supervisor; (c) edTPA commentary; and (d) focus groups.