news article

lable at ScienceDirect

Teaching and Teacher Education 73 (2018) 43e55

Contents lists avai

Teaching and Teacher Education

journal homepage: www.elsevier.com/locate/tate

Professional learning communities among vocational school teachers: Profiles and relations with instructional quality

Julia Warwas a, *, Christoph Helm b

a Chair of Business Education, Georg-August-Universit€at G€ottingen, Germany, Platz der G€ottinger Sieben 5 37073 G€ottingen, Germany b Department of Educational Research, Johannes Kepler University of Linz, Austria, Altenberger Straße 69, 4040 Linz, Austria

h i g h l i g h t s

� Core PLC dimensions distinguish behavioural, ideational, and structural elements. � Departmental profiles convey Advanced, Structurally embedded, and Rudimentary PLCs. � Teachers from Advanced PLCs provide higher levels of application-orientation (AO). � AO includes instruct. practices (e.g. explanations) and methods (e.g. problem solving).

a r t i c l e i n f o

Article history: Received 1 July 2017 Received in revised form 16 March 2018 Accepted 21 March 2018 Available online 30 March 2018

Keywords: Professional learning communities Vocational schools Instructional quality Multilevel latent profile analysis

* Corresponding author. E-mail addresses: [email protected]

jku.at (C. Helm).

https://doi.org/10.1016/j.tate.2018.03.012 0742-051X/© 2018 Elsevier Ltd. All rights reserved.

a b s t r a c t

Although improving teachers' classroom strategies presents the primary goal of PLCs, empirical evidence is still scarce, mainly derived from case-study material and confined to (pre-) K12 settings. We use profiling techniques and a comparative design to examine (a) distinct configurations in which PLCs occur within vocational school departments, and (b) their relations to instructional quality. Multilevel latent profile analysis, based on teacher assessments of core PLC dimensions, reveals three configurations. Multilevel multiple group analysis of students’ instructional ratings shows that teachers from Advanced PLC departments create more authentic, application-oriented learning environments than teachers from other departments.

© 2018 Elsevier Ltd. All rights reserved.

1. Introduction

Recent literature reviews and meta-analytic findings indicate that high-performing schools feature complex forms of teacher cooperation such as professional learning communities (PLCs) (Fulton & Britton, 2011; Lomos, Hofman, & Bosker, 2011b; Scheerens, 2014). However, despite converging evidence regarding the impact of PLCs on important outcomes of pedagogical activities, as measured by student achievement, comparably little is known about how PLC membership affects pedagogical activities per se, as reflected in indicators of instructional quality (Stoll, Bolam, McMahon, Wallace, & Thomas, 2006; Vangrieken, Dochy, Raes, & Kyndt, 2015). Investigations of teachers’ instructional practices and methods that are associated with working in a PLC thus present

(J. Warwas), christoph.helm@

a largely missing but highly relevant piece of empirical knowledge, because “[a]t its core, the concept of a PLC rests on the premise of improving student learning by improving teaching practice” (Vescio, Ross, & Adams, 2008, p. 82; Supovitz, 2002).

This deficit is particularly salient in the vocational education and training (VET) system, which has been largely overlooked by re- searchers despite the major challenges it poses to the teaching profession. Professional demands in the VET system are to provide students with occupational knowledge and skills they need in their future workplaces and, thus, to ensure that they are well equipped to compete on the labour market (Hellwig, 2006). Continuous and accelerating changes in many workplaces put high pressures on teachers to constantly enhance their domain-specific expertise and instructional repertoire (Grollmann & Bauer, 2008). Moreover, concerted actions of stable teams are needed to coherently design and implement curricular modifications that represent construc- tive pedagogical answers (referred to as action- and application- oriented ‘learning fields’) to changes of the occupational world

J. Warwas, C. Helm / Teaching and Teacher Education 73 (2018) 43e5544

(i.e., corresponding ‘activity fields’ in the workplace) (Tenberg, 2017). In German vocational schools, departments tackle these permanent tasks, involving staff members who manage, adapt and improve the instructional program for a specific vocational domain (Pahl, 2007). As in the present study, these highly specialized organizational sub-units may concentrate on occupations in the technical-industrial sector, in the social sector, in the commercial sector, or in the technological sector.1 Owing to the diversity of Germany's VET system, however, departments operate in different types of vocational schools, i.e., in different institutional frame- works, and are responsible for students with divergent educational backgrounds and qualification goals (see Schneider, 2008; Solga, Protsch, Ebner, & Brzinsky-Fay, 2014 for detailed descriptions and statistics). The current investigation, conducted in the Federal State of Bavaria/Germany, includes part-time and full-time vocational schools (belonging to secondary education level II), commercial colleges (sec I), and upper vocational schools (sec II and tertiary level).

Both part-time and full-time vocational schools provide full vocational qualifications and therefore award nationally recognized certificates of ‘skilled occupations’, allowing adolescents and young adults to take up a suitable position in an employing organization (e.g., as a trained retail salesman or a trained paediatric nurse). Full- time vocational schools offer VET courses that take place only at school and often prepare for occupations in the social sector. The clear majority of students who enrol in these courses holds an in- termediate or even upper secondary school degree (of general education). Part-time vocational schools function as equal partners of training companies in Germany's ‘dual system’ of VET and cover many vocational domains. Entering the dual system requires stu- dents to have an apprenticeship contract but no formal school- leaving certificate, which is why roughly a third of them has a lower secondary (general) school degree or no school degree at all. Upper vocational schools primarily prepare students for higher academic tracks (particularly universities of applied science), but they do so by offering specialization in distinct vocational domains. Consequently, they require students to do internships or even to have successfully completed a certified (dual or school-based) training in the chosen area of specialization. Commercial colleges in Bavaria e representing one of several school types within Ger- many's VET system that exist in one or a few federal states only e ‘anticipate’ occupational specialization by preparing students to take up a fully qualifying VET course in the commercial sector but also to acquire a certificate of intermediate general education.

This brief outline suggests that available evidence on the prev- alence, sophistication and instructional impact of departmental PLCs does not necessarily apply to vocational schools. Relevant findings stem from secondary schools in the general education system (for overviews, see Lomos, Hofman, & Bosker, 2011a; Visscher & Witziers, 2004). But although in both education sys- tems, departments represent teachers’ main area of professional (inter-)action and influence (M€arz & Kelchtermans, 2013), voca- tional school departments face professional demands and institu- tional environments that differ from those of general, academically oriented schools. Transferability of extant findings is further limited by the fact that most of them were obtained from small-scale, qualitative investigations (see section 3.1).

Against this background, our paper pursues two objectives. The first is to explore if and to what extent departments in German vocational schools operate as PLCs, i.e., in accordance with generic

1 See Germany's Federal Institute of VET (BIBB) for a detailed classification of over 300 occupations into vocational domains: https://metadaten.bibb.de/klassifikation/ 9.

features of PLCs. We do so by adopting a profiling approach and examining profile composition and occurrence. The second objec- tive is to analyse if and to what extent teacher membership in different PLC profiles explains differences in instructional quality.

Since concepts and measures of PLCs vary considerably in the literature (see Sleegers, den Brok, Verbiest, Moolenaar, & Daly, 2013; Vangrieken et al., 2015), we start with a theoretical frame- work that explicates core dimensions of PLCs as well as analytical strategies to investigate their existence and sophistication. We further summarize extant evidence on how PLCs affect classroom instruction and highlight features that may serve as indicators of instructional quality in vocational classrooms. The empirical part of the paper comprises two subsequent survey studies. To analyse our first research question, we draw on teacher ratings of core PLC di- mensions for their respective organizational units (Study 1: 395 teachers from 47 departments). Multilevel latent profile analyses are run to identify departmental PLC profiles. Latent profile random effects models are employed to test the predictive value of school, department, and teacher characteristics for profile membership. To answer the second question, we use student assessments of teachers’ instructional practices and methods. Student question- naires were administered about five months after the teacher sur- vey (Study 2: 1243 students). Multilevel multiple group analysis serves to test if instructional features vary systematically between PLC profiles while accounting for the nested structure of student ratings in classes and departments. We discuss main findings synoptically at the end of our paper, highlighting their contribu- tions to current research on PLCs and their implications for future studies.

2. The PLC architecture and its measurement

2.1. Core dimensions of professional learning communities

Professional communities represent conceptually and opera- tionally heterogeneous fields of research (Sleegers et al., 2013; Stoll et al., 2006; Vangrieken, Meredith, Packer, & Kyndt, 2017). This concerns, inter alia, the scale and institutional anchoring of a community. Some studies focus on small, interdisciplinary teams that evolve outside formal organizational structures while carrying out innovative projects (e.g., Owen, 2014; Schaap & de Bruijn, 2017). Others consider the entire staff of schools (e.g., Bolam, McMahon, Stoll, Thomas, & Wallace, 2005; Hord, 1997) or distinct organizational sub-units (e.g., Hallam, Smith, Hite, Hite, & Wilcox, 2015; Visscher & Witziers, 2004) as PLCs, again in varying evolu- tionary stages. Research in German schools predominantly refers to PLCs that are installed at a departmental level, where shared areas of responsibility and expertise both necessitate and facilitate close cooperation (e.g., Bonsen & Rolff, 2006; Buhren, 2015). Against this background, the present study concentrates on generic features in the work of PLC members rather than on their number or organi- zational affiliation. Thus, we focus on dimensions that describe the way collective professional practice is orchestrated within a PLC, and which should all be very pronounced in strong, fully developed PLCs.

Taking stock of the literature, we can identify three strands of modelling the work of PLCs. The first strand contains established “five-component-models” and can be subdivided into two similar but not congruent segments. In one segment, we find papers that adopt the five components put forth by Kruse, Louis, and Bryk (1995), namely reflective dialogue, deprivatization of practice, collaboration, shared norms and values, and collective focus on stu- dent learning (e.g., Lomos et al., 2011a; Vescio et al., 2008). In the other segment, researchers largely follow Hord's (1997) classifica- tion, which proposes shared personal practice, collective creativity,

J. Warwas, C. Helm / Teaching and Teacher Education 73 (2018) 43e55 45

shared values and vision, supportive and shared leadership, and supportive conditions (e.g., Hipp & Huffman, 2010; Lee, Zhang, & Yin, 2011; Stegall, 2011). The second strand in the literature em- braces eclectic models, which merge and extend both five- component models to create exhaustive lists of PLC elements (e.g., Bolam et al., 2005; Mitchell & Sackney, 2007; Sigurdardottir, 2010). The third strand distinguishes core dimensions that bundle and structure the various proposed elements (e.g., Sleegers et al., 2013). The present paper contributes to the third, integrative strand of research and draws on Lavi�e’s (2006) and Achinstein's (2002) work to discern three core dimensions of PLCs. We use the labels collaborative development, normative agreement, and sup- portive infrastructure to describe their constitutive elements.

Collaborative development denotes the behavioural dimension of a PLC, as its members engage in genuinely collaborative efforts to improve pedagogical processes (Little, 2012). This means that teachers not only occasionally exchange instructional materials, but also co-constructively design and consistently implement instruc- tional units, thereby establishing a growing set of shared instruc- tional practices and methods (e.g., Bolam et al., 2005). Moreover, PLC members exercise activities of continuous collective inquiry (Hord, 1997) and, thus, cultivate processes of teacher learning (Little, 2012). These activities often include reflective dialogue about instructional issues (Kruse et al., 1995; Penner-Williams, Diaz, & Gonzales Worthen, 2017) but also mutual visits and consultation, peer coaching, or joint evaluation of instructional processes and outcomes (e.g., Hipp & Huffman, 2010; Stegall, 2011; Visscher & Witziers, 2004).

Normative agreement describes the ideational dimension of a PLC, as collaborative efforts need to serve shared purposes and to be closely aligned with consented standards of professional conduct to be successful and enduring. Therefore, teachers who work together in a PLC hold a common set of pedagogical values and goals, and they strongly agree in their personal beliefs about effective teaching and learning (e.g., Andrews & Lewis, 2007; Kruse et al., 1995; Stoll et al., 2006).

Supportive infrastructure depicts the structural dimension of a PLC, as substantial collaborative activities necessitate operational structures that facilitate and promote them (e.g., Hord, 1997). A unit's “organizational capacity” (Mitchell & Sackney, 2007, p. 630; Sleegers et al., 2013, p. 7) may again include a wide range of aspects but two of them are particularly prominent in the literature. Firstly, operational routines as well as project workflows should run smoothly and be closely coordinated, which requires clear rules, procedures and responsibilities, adequate information supply, and fast and transparent communication channels. Secondly, sufficient timely and spatial resources should be available to meet and work in teams (Hipp & Huffman, 2010; Sleegers et al., 2013; Stegall, 2011).

Recent but few surveys among vocational school teachers in Germany indicate moderate levels of instruction-focused cooper- ation despite high satisfaction with structural prerequisites (sum- marized by Tenberg, 2017). They further suggest that individual characteristics of teachers might be conducive to establishing and maintaining cooperation. Conforming with arguments on advan- tageous “personal capacity of individuals to learn, teach, question, reflect, and grow” (Mitchell & Sackney, 2007, p. 630), interviewees in Koschmann's (2013) investigation of nine teams emphasize that work-related attitudes such as trust, openness and confidence provide fertile grounds for working as a PLC. In R€oder's (2017) questionnaire-based study with 342 vocational school teachers, the respondents' age and gender, but also their VET domain partly predicted their preference for cooperation. Female teachers,

younger teachers, and teachers specialized in the commercial sector were stronger advocates of cooperation than their male and older colleagues as well as teachers from the technical-industrial sector.

2.2. The profiling approach to distinguish PLC configurations

To prove the existence and sophistication of PLCs and their re- lations with pedagogical processes or outcomes, previous studies have employed different methods. Most scholars investigated separate or additive effects of singular PLC dimensions (e.g., Bolam et al., 2005; Lee et al., 2011; Visscher & Witziers, 2004) while others created a composite index by summing up or averaging the scores of each dimension (e.g., Louis & Marks, 1998; Sun-Keung Pang, Wang, & Lai-Mei Leung, 2016). Still others focused on distinct phenomenological configurations of PLC dimensions by employing profiling techniques. The latter include statistical clus- tering (Lomos et al., 2011a), qualitative, interview-based typologies (DeMatthews, 2014), and even contrastive case studies that docu- ment divergent trajectories in the development of PLCs over time (Schaap & de Bruijn, 2017).

The profiling approach assumes that between a fully-developed or ideal-type PLC, represented by maximum scores on all its defining features, and the total absence of these features, several discrete scoring patterns coexist that characterize common varia- tions of collective professional practice. Consequently, the aim is to detect prevalent ‘real-types’ of PLCs (see also Lee, 2013, for a broader discussion on organizational configurations of schools). We adopt this approach for three reasons. Firstly, profiling techniques provide appropriate diagnostic tools for multidimensional concepts of professional practice. For instance, a department whose mem- bers largely agree on pedagogical purposes but hardly engage in well-organized, tightly coordinated efforts to enhance their instructional repertoire may bear a strong ideational basis but does not embody a full-fledged PLC (e.g., Owen, 2014). Secondly, inves- tigating prevalent configurations is one feasible way of analysing the combined effects of multiple indicators (Fiss, 2007). The fact that in several previous studies, substantial bivariate correlations be- tween singular PLC dimensions and target variables are missing may partly be due to such multiplicative effects (Bolam et al., 2005; Lee et al., 2011; Visscher & Witziers, 2004). Thirdly, profile com- parisons not only help to identify configurations that are particu- larly conducive or detrimental to instructional quality (Lee, 2013), which could be expected for configurations with maximal or minimal scores on all classification variables and might as well be tested by using a composite index. The surplus value here is that profile comparisons can also detect phenomenologically distinct but equally (un)successful patterns of collective professional practice (Gresov & Drazin, 1997). They may reveal real-type PLCs that differ in their configuration of dimensional values but not in their impact on student learning.

3. The relevance of PLCs for classroom instruction

3.1. Extant findings

Although improvements in educational quality and effective- ness describe the primary objective of PLCs (Stoll et al., 2006), robust evidence of how participation in a PLC influences classroom instruction is sparse and derived from general education settings (Lomos et al., 2011b). Available studies are criticized for methodo- logical weaknesses (Fulton & Britton, 2011) and a lack of precision about which instructional features are affected (Vescio et al., 2008).

J. Warwas, C. Helm / Teaching and Teacher Education 73 (2018) 43e5546

Reviews of school-restructuring programs in the United States quite broadly suggest that “in those schools that demonstrated characteristics consistent with strong professional learning com- munities, teachers' classroom pedagogy was more likely to change in accordance with ongoing reform efforts” (Westheimer, 2008, p. 760). Among those evaluations that exemplify instructional effects in more detail is Louis and Marks’ (1998) investigation of features they refer to as authentic pedagogy and social support for achieve- ment in 24 elementary, middle, and high schools. Here, data from teacher and student questionnaires, together with observer ratings of lessons and assessment tasks, entered multilevel regression models. Based on a composite index to measure PLC activities within each school, results document that these activities explain substantial amounts of between-school variation regarding authentic pedagogy and social support. Where PLCs were strongly pronounced, instructional interactions and assessment tasks required students to process content deeply and meaningfully, and to deal with realistic problems of everyday life. Furthermore, classroom environments were more appreciative, supportive, and friendly than in schools with low levels of PLC activities.

Vescio et al.’s literature review (2008) documents a prepon- derance of qualitative designs for researching PLCs. These authors criticize that by reporting interview statements, field notes, and meeting transcriptions in a case-study format, most studies only touch upon temporal changes in the instructional strategies of PLC members, or upon instructional variations associated with the presence or absence of PLCs. Nevertheless, when taken together, the reviewed studies create the impression that instructional strategies of PLC members become “more student centered” (p. 88). For example, teachers gain flexibility in classroom arrangements and the pacing of lessons to adjust instructional processes to differing ability levels of their students.

Only recently, Penner-Williams et al. (2017) refined this picture. They used structured classroom observations to elucidate how instructional behaviours changed during a professional develop- ment program for teaching English as a second language, which included PLCs as an integral component. The authors conclude that the participants (41 pre-K and K-12 teachers) employed more culturally and linguistically responsive strategies in their class- rooms. For example, teachers implemented more scaffolding techniques, better adjusted task requirements to students’ back- ground knowledge, and stimulated lively, substantive conversa- tions more often than at the beginning of the program.

Fulton and Britton's (2011) synthesis of research, focusing on STEM (science, technology, engineering, and mathematics) teach- ers across grades K-12, further adds to this impression. Based on a very inclusive definition of PLCs, the authors infer that instructional quality benefits in at least three aspects. STEM teachers who un- dertake collaborative and coordinated efforts to improve peda- gogical activities (a) use more research-based instructional methods such as student inquiry to explore subject matter, (b) show more interest in students' reasoning and understanding, and (c) employ more diverse modes of engaging students in problem- solving techniques than other teachers. However, as most reviewed studies concentrate on communities whose imple- mentation was a planned and guided part of a training program, Fulton and Britton urge researchers to more carefully attend to “naturally occurring PLCs” (p. 13).

In the present paper, we turn to PLC profiles as they appear in vocational school departments without external interventions. These kinds of “formative” communities differ from planned and guided “formal” communities (Vangrieken et al., 2017, 52e53; see also Supovitz, 2002, for an illustrative example). They represent

task-centered, autonomous teams or organizational units, not governmental initiatives to establish educational standards or implement educational reforms; they do not receive systematic support from external experts such as researchers or trainers who clarify goals, pre-define desirable instructional practices and methods, and supervise compliance; and they are usually longer lasting. To assess their potential impact on teachers’ pedagogical activities, we build on renowned evaluation criteria for instruc- tional quality and adapt them to the specificities of vocational education.

3.2. Indicators of instructional quality in the vocational classroom

Current discussions on instructional quality draw on construc- tivist theories of learning (e.g., Loyens & Gijbels, 2008) and teacher effectiveness studies (e.g., Seidel & Shavelson, 2007). Constructivist theories substantiate that instructional quality reflects the degree to which teachers create favourable opportunities for students’ self- directed and insightful knowledge construction. Effectiveness studies help to establish the nature and effects of these opportu- nities empirically.

For general compulsory education, discussions converge on the assumption that three broad and basic dimensions reveal the ‘deep structures’ of high-quality instructional interaction (Kunter & Voss, 2013; Praetorius, Pauli, Reusser, Rakoczy, & Klieme, 2014). Effective classroom management describes the teacher's success in orches- trating multiple and simultaneous occurrences within a classroom, and in preventing and handling disruptions, distractions and disciplinary conflicts, thereby maximizing content-related learning time. Individual learning support describes the extent to which a teacher is sensitive to individual students' comprehension diffi- culties, encourages students to tackle even demanding tasks, and patiently but unobtrusively assists in detecting and rectifying misconceptions. Cognitive activation mirrors the degree to which task assignments and communication patterns stimulate higher- order thinking and thus, prompt students to process subject mat- ter thoroughly.

Helm (2016) demonstrated that these basic quality dimensions are predictive of learning outcomes in accounting instruction, too. Nevertheless, the overarching goal of vocational education is to create opportunities for students to develop competencies they need to succeed in their future workplaces (e.g., Achtenhagen & Winther, 2014). These competencies not only rest on advanced knowledge but also encompass technical mastery of procedures and instruments, as well as sound proficiency of communicative standards and collaborative strategies in the respective vocational domain (Sembill, Rausch, & K€ogler, 2013). Consequently, building and refining conceptual knowledge is an essential but not the sole aim of vocational education, and the extent to which vocational teachers provide action- and application-oriented learning envi- ronments becomes another key indicator of instructional quality (Gessler & Howe, 2015; Hellwig, 2006). Ideally, vocational class- rooms ensure high levels of authenticity by anchoring learning processes in work-related problems and allowing students to autonomously plan, completely perform, and carefully control coherent sequences of work activities over longer periods of time (e.g., De Bruijn & Leeman, 2011). Overt activity structures in the classroom are therefore quite informative about the implementa- tion status of action- and application-orientated learning envi- ronments. For example, the frequency in which students collaborate in small groups to solve work-related problems (De Bruijn & Leeman, 2011) or engage in simulation methods like experimental games or role plays (e.g., DeBourgh & Prion, 2011;

J. Warwas, C. Helm / Teaching and Teacher Education 73 (2018) 43e55 47

Khaled, Gulikers, Biemans, van der Wel, & Mulder, 2014) can serve as reasonable proxies.

4. Research aims and strategy

This paper investigates how distinct configurations of PLCs in vocational school departments relate to instructional quality in vocational classrooms. Our research interest therefore embraces two sub-goals. The first is to explore departmental PLC profiles, using dimensional scores on collaborative development, normative agreement, and supportive infrastructure as classification criteria. This allows us to investigate to which extent and in which combi- nation of dimensional scores departments operate in accordance with generic features of a PLC. We employ multilevel latent profile analysis and evaluate the fit indices of alternative profile solutions to identify characteristic patterns in teacher ratings of core PLC dimensions at the department level. We further examine the composition and occurrence of the profiles obtained in the final solution (Study 1). The second goal is to examine if and to what extent variations in instructional quality are attributable to teach- ers’ membership in distinct PLC profiles. Following extant findings reported in the previous sections, we expect teachers who belong to (comparably) strong PLCs, indicated by (comparably) high scores in all dimensions, to provide students with better opportunities to acquire occupational knowledge and skills. Multilevel multiple group analyses serve to test this assumption. We draw on student assessments of the degree to which teachers demonstrate effective classroom management, individual learning support, and application- oriented teaching, and of the frequency in which teachers employ action- and application-oriented methods such as problem-solving (Study 2).

5. Study 1 e teacher survey

5.1. Sample

The present sample can be labelled a convenience sample because questionnaires were sent to schools in Bavaria/Germany that had already participated in a survey on the professional de- mands of principals and had signalled their interest in subsequent studies among staff and students. That way, we collected responses from 395 teachers (45.5% female; 17 years of teaching experience on average, SD ¼ 10.7 years). They belong to 47 departments that are specialized in different vocational domains, namely, in the technical-industrial (11), social (10), commercial (20), and techno- logical (6) sectors of VET. On average, a department counts 8 teachers (min 4, max 25). The departments are located in 34 schools that represent different vocational school types: commer- cial colleges (N ¼ 31 teachers, 5 departments), upper vocational schools (77, 12), part-time vocational schools (192, 22) and full- time vocational schools (95, 8). This allows us to examine poten- tial dependencies of departmental profiles on school type or vocational domain.

To test the representativeness of the sample, we compared sample and population distributions regarding the number of teachers per school type, teacher sex, and teacher age (Statistisches Bundesamt, 2015). Chi-square tests show that the sample is only representative for Bavaria with respect to gender. However, de- viations in the number of teachers per school type and teacher age are limited to a few categories: Teachers in full-time vocational schools and teachers between 40 and 45 years are overrepresented in our sample; teachers between 60 and 65 years are

underrepresented.

5.2. Teacher questionnaires

To assess the degree to which departments display core PLC dimensions, teachers rated 21 items pertaining to collaborative development (CD), normative agreement (NA), and supportive infra- structure (SI). Items were adapted from school quality research in Germany (Ditton, n.d.; Steinert, Gerecht, Klieme, & D€obrich, 2003) because scales for measuring PLCs that are both widely accepted and psychometrically validated are missing (Vangrieken et al., 2015). Confirmatory factor analysis for a 3-factor model yields satisfying model fit (RMSEA 0.044, CFI 0.941, SRMR 0.052). Furthermore, significant but medium latent intercorrelations rCD- NA ¼ 0.633, rCD-SI ¼ 0.405, rNA-SI ¼ 0.677, and the fact that the 3- factor model outperforms a 1-factor model (more precisely, a 3- factor model in which all latent correlations are constrained to 1; Satorra-Bentler Scaled Chi-Square: Dc2 ¼ 189.753; Ddf ¼ 3; p < .000) support its structural validity. Table 1 states two exem- plary items for each factor, descriptive statistics, and measures of reliability and interrater agreement. Appendix A lists all items used.

Cronbach's alpha and ICC(2)-values indicate that the scales' in- ternal consistencies and the reliabilities of aggregate constructs are satisfactory (>0.70) or at least acceptable (>0.60) (Lüdtke, Trautwein, Kunter, & Baumert, 2006). ADM values well below the critical mark of 1.00 for a six-point answer format show good interrater agreement within each department (Burke, Finkelstein, & Dusig, 1999). ICC(1)-values indicate that substantial amounts of variability in the teachers' ratings (16e31%) represent between- department differences (Lüdtke et al., 2006).

In addition to PLC dimensions, we assessed teachers’ work- related efficacy beliefs, which may be predictive of their PLC profile membership, using four items from Abele, Stief, and Andr€a (2000). Confirmatory factor analysis yields a good model fit (c2 5.337, p ¼ .069, RMSEA 0.065, CFI 0.999, SRMR 0.003). Scale mean and standard deviation are 4.09 (0.66). Response options vary from 1 (does not apply at all) to 5 (fully applies).

5.3. Statistical procedure

Latent profile analysis (LPA) is a probabilistic approach “to identify subtypes of related cases using a set of categorical or continuous observed variables” (Henry & Muth�en, 2010, p. 193). In the present study, variables are latent factor scores (obtained from principal component analysis) for the three PLC dimensions. LPA assumes that observations are independent of each other. In the present sample, however, teachers are nested within departments and therefore exposed to similar contextual influences that foster converging perceptions within subgroups. To account for this hi- erarchical data structure, we performed multilevel LPA using Mplus (Muth�en & Muth�en, 1998e2014). Thus, we assume that a teacher's probability of belonging to a specific latent level-2- (i.e., depart- mental) profile varies across departments.

We base our model selection on a range of fit indices. According to Tein, Coxe, and Cham (2013), the Bayesian information criterion (BIC), Akaike information criterion (AIC), and entropy are well established to decide on the number of latent classes. We, too, consider these indices but rely on the sample size adjusted BIC (saBIC). The latter penalizes complex models to a lower degree than BIC (Tofighi & Enders, 2007). Furthermore, we consider the log- likelihood, the AIC with a higher penalty weight (AIC3), Bozdo- gan's consistent AIC (CAIC), the parametric bootstrapped likelihood

Table 1 Descriptive statistics, reliability, interrater agreement, and two sample items for each PCL dimension.

# Sample Items M SD a ICC(1) ICC(2) ADM

CD 7 We often prepare instructional units together. Self-assessments and mutual evaluations are a matter of course in our work.

3.39 0.86 0.81 .26 0.79 0.58

NA 5 It is hard to find common ground in the pedagogical beliefs of our teaching staff. (inverse) Every colleague has his/her own ideas of instructional design. (inverse)

3.63 0.77 0.75 .16 0.62 0.56

SI 9 Responsibilities are clearly defined. Workflows are well structured.

4.14 0.76 0.86 .31 0.75 0.52

Notes. # ¼ number of items, M ¼ mean, SD ¼ standard deviation, a ¼ Cronbach's alpha, ICC ¼ Intraclass correlation, ADM ¼ mean absolute deviation index, CD ¼ collaborative development, NA ¼ normative agreement, SI ¼ supportive infrastructure. Six-point rating format from 1 ¼ does not apply at all to 6 ¼ fully applies.

Table 2 Model fit indices of latent profile analysis in step 1.

M # LP Par Log-LL AIC AIC3 CAIC saBIC Entropy LRT LMR

1 1 9 �1679 3377.881 3386.881 3422.691 3385.133 1.000 e e 2 2 13 �1652 3331.745 3344.746 3396.472 3342.222 0.810 0.000 0.005 3 3 17 �1645 3325.944 3342.944 3410.585 3339.644 0.756 0.077 0.339 4 4 21 �1643 3328.883 3349.832 3433.389 3345.756 0.773 1.000 0.531 5 5 25 �1640 3331.200 3356.200 3455.672 3351.347 0.765 1.000 0.517

Note. M ¼ model. # LP ¼ number of latent profiles. The lowest values of comparative fit indices are in italics.

Table 3 Model fit indices of latent profile analysis in step 2.

M # LP 2 # LP 1 Par Log-LL AIC AIC3 CAIC saBIC Entropy

1 1 3 17 �1645 3325.947 3342.946 3410.587 3339.647 0.757 2 2 3 20 �1629 3299.173 3319.174 3398.752 3315.291 0.814 3 3 3 23 �1622 3289.548 3312.548 3404.062 3308.083 0.844 4 4 3 26 �1621 3294.731 3320.730 3424.181 3315.683 0.838 5 5 3 29 �1621 3300.723 3329.724 3445.112 3324.094 0.850

Note. M ¼ model. # LP 1, LP 2 ¼ number of latent profiles at level 1 and level 2. LMR and LRT are not applicable in multilevel latent profile analysis. The lowest values of comparative fit indices are in italics.

J. Warwas, C. Helm / Teaching and Teacher Education 73 (2018) 43e5548

ratio test (LRT), and the Lo-Mendell-Rubin adjusted LRT (LMR) when identifying the best fitting models.

Following the stepwise procedure recommended by Luko�cien_e, Varriale, and Vermunt (2010), we first determine the number of teacher-level profiles, ignoring the multilevel structure (Table 2). In a second step, we fix the number of teacher-level profiles to the value of step 1 and determine the number of department-level profiles (Table 3).

As set out in chapters 1 and 2, the PLC profiles investigated in this paper represent distinct modes of collective professional practice at a departmental level but may partly depend on indi- vidual or institutional factors. Given the heterogeneity of our sample, we therefore examine the composition and prevalence of departmental PLC profiles with latent profile random effects models (Section 5.5). Teachers’ work-related efficacy beliefs, gender and seniority, the school type, and the vocational domain of a depart- ment serve as potential predictors.

5.4. Results: PLC profiles of vocational school departments

Based on single-level LPA, Table 2 lists alternative models that assume one to five patterns in vocational teachers’ perceptions of collective professional practice. Fit indices are ambiguous regarding the best solution. While CAIC, entropy, and LMR indicate a better fit for the 2-profile-solution, AIC, AIC3, and saBIC are in favour of the 3-profile-solution. Since LRT (when testing the 2- vs. 3-profile-so- lution) only slightly fails significance and the 2-profile solution divides teachers into highly unbalanced und poorly differentiated subgroups (340 vs. 55 teachers), we fixed the value of level-1

profiles to three to enhance within-homogeneity and between- heterogeneity among the subgroups.

In the second step (Table 3), model 3 with three level-2- and three level-1-profiles shows the most favourable AIC, AIC3, and saBIC values. Although model 2 has a lower CAIC value, it is inferior with respect to entropy. Therefore, we fixed the number of level-2 profiles to three.

Since our research question relates to level-2 profiles (i.e., departmental profiles) only, we concentrate on these profiles when interpreting characteristic scoring patterns. Based on mean scores in each PLC dimension (Table 4), we label the investigated de- partments as follows:

Profile 1 ¼ Structurally embedded PLCs. Most departments in our sample (n ¼ 33; 259 teachers) are well equipped with supportive organizational resources. Contrary to this rather strong structural component, ideational and behavioural dimensions are moderately pronounced.

Profile 2 ¼ Advanced PLCs. On average, departments clustered in this profile (n ¼ 9, 96 teachers) score highest on all PLC dimensions when compared to the other profiles. Considering the six-point rating format, all scores are above the scale mean but not sugges- tive of full-fledged PLCs. Therefore, this configuration is the most sophisticated one in our sample but still not a prime example of how “strong PLCs” ideally look like in the conceptual literature.

Profile 3 ¼ Rudimentary PLCs. The smallest subgroup contains five departments (40 teachers), scoring comparably low in any PLC dimension. Since all scores fall below the theoretical scale mean but still do not depict serious deficiencies, this configuration presents the inverse picture of the Advanced PLC. In these organizational

Table 4 Profiles of professional learning communities in vocational school departments (mean levels and standard deviations on six-point scales).

Collaborative Development Normative Agreement Supportive Infrastructure

LP2: Advanced PLCs (n ¼ 9) 3.98 (.32) 3.72 (.30) 4.25 (.23) LP3: Rudimentary PLCs (n ¼ 5) 3.19 (.72) 3.27 (.38) 3.34 (.22) LP1: Structurally embedded PLCs (n ¼ 33) 3.20 (.31) 3.65 (.35) 4.22 (.30)

Note. LP ¼ latent profile.

Table 5 Predictors of latent profile membership.

predictor Rudimentary PLC

Structurally embedded PLC

b p b p

M1 level1 teacher gender 0.024 0.117 1.445 0.485 level2 share of female teachers 0.185 0.171 0.179 0.140

M2 level1 teacher seniority �0.015 0.141 �0.011 0.128 level2 average seniority �0.014 0.887 0.006 0.956

M3 level2 school type c2 (6, N ¼ 47) ¼ 8.49, p ¼ .204) M4 level2 technological sector c2 (1, N ¼ 47) ¼ .84, p ¼ .366) M5 level2 social sector �1.819 0.278 0.120 0.942 M6 level2 technical-industrial sector 0.767 0.634 0.563 0.743 M7 level2 commercial sector �1.115 0.676 0.489 0.830 M8 level1 teacher work-relat. efficacy �1.219 0.063 ¡1.556 0.018

level2 average work-relat. efficacy ¡9.631 0.014 ¡9.249 0.027 Note. b ¼ unstandardized coefficients; reference category is the Advanced PLC. Significant coefficients (at p � .10 because of the low number of level-2 clusters) are bold. M3/M4: A model with vocational school type and a model with technological sector as L2-predictor could not be identified due to a few empty cells. Thus, an ANOVA was performed instead.

J. Warwas, C. Helm / Teaching and Teacher Education 73 (2018) 43e55 49

units, all distinguishing features of work within a PLC are under- developed but not completely absent.

2 This interval was a pragmatic solution that arose when negotiating possible time slots for data collection with the participants of our study.

5.5. Results: profile composition and prevalence

To predict profile membership, we performed latent profile random effects models (Henry & Muth�en, 2010), introducing pre- dictors at level 1 (gender, seniority, work-related efficacy beliefs of teachers) and level 2 (share of women within a department, average seniority and average work-related efficacy beliefs of department staff, school type, and vocational domain of the department). Due to the small sample size at level 2 (47 de- partments), we performed separate models for the predictors. Though the model estimations terminated normally, level-2 pa- rameters might be biased because of the low number of profile members. Therefore, we also compared the profile means regarding the mentioned variables. Since both analytic strategies yield similar findings, we mainly report unstandardized co- efficients of latent profile random effects models in Table 5. As a rule for interpretation, a one-unit change in the predictor leads to a change of b logits. These logits represent the probability of belonging to the Rudimentary or Structurally embedded profile (with the Advanced profile being the reference category). Level-1 results refer to within-department differences between individual teach- ers, conditional on a specific department; level-2 results depict differences between the departments.

Results show that neither gender nor a department's share of female teachers significantly predict profile membership. The same is true for (individual and average) seniority of teachers, school type and vocational domain. However, department-specific de- grees of work-related efficacy beliefs predict profile membership. Staff in Advanced PLCs hold stronger efficacy beliefs than in Struc- turally embedded and Rudimentary PLCs.

6. Study 2 e student survey

6.1. Sample

Students assessed their teachers’ instructional practices and methods approximately five months after the teacher survey.2

Student ratings are available for teachers from 34 of the 47 de- partments. On average, two classes per department completed the questionnaires, totalling 61 classes that comprise of 1243 voca- tional students (male: 41.8%, female: 58.2%; age groups: 15e16 years ¼ 17%; 17e18 years ¼ 43.9%; 19 years or older ¼ 36.8%). The number of students per class averages 21 (min ¼ 11; max ¼ 34). Each class rated one teacher only.

To test the representativeness of the sample, we compared sample and population distributions (Statistisches Bundesamt, 2015) regarding student gender and the number of students per school type. Chi-square tests show that the sample is not fully representative for Bavaria. Whereas the number of students in part- time vocational schools is underrepresented, the number of stu- dents in full-time vocational schools is overrepresented. With respect to gender, our sample contains a lower share of male stu- dents than one would expect based on population values.

6.2. Student questionnaires

We asked students to answer 15 items that serve as indicators of instructional quality in vocational education as outlined in section 3.2. These items stem from previous research on school quality and instructional quality in Germany (Ditton, n.d.; Gruehn, 2000; Seeber & Squarra, 2003). Twelve items tap the extent to which the teacher's instructional practices convey individual learning support (5 items), effective classroom management (4 items), and application orientation regarding the explanations, examples, and materials he/ she provides (3 items). Three items measure the frequency in which the teacher implements action- and application-oriented instruc- tional methods, namely problem solving in small groups and simu- lation-based learning (experimental games, role-plays). Table 6 presents sample items, descriptive statistics, and measures of reli- ability and interrater agreement.

As evident in ADM values below 1.00 and ICC(2) values above 0.70, student ratings within each class are consistent and reliable for most dimensions (see Lüdtke et al., 2006). Only the problem- solving dimension slightly and expectedly falls below the 0.70 threshold because its variation at class-level is comparably low. Still, all ICC(1) values indicate that ratings of instructional practices and methods deviate substantially between classes, as class-level differences account for 19e37% of total variance in student ratings.

6.3. Statistical procedure

To examine differences in instructional quality dependent on teachers' PLC profile membership, we performed four steps: (1) We

Table 6 Descriptive statistics, reliability, interrater agreement, and sample items for dimensions of instructional quality.

Measures M SD a ICC(1) ICC(2) ADM Sample Items

classroom management (a) 4.25 1.13 0.77 0.27 0.87 0.71 Our teacher ensures a calm and orderly learning atmosphere. individual learning support (a) 4.21 1.11 0.89 0.26 0.88 0.69 Our teacher supports students who have difficulties understanding subject matter. application orientation (a) 4.05 1.27 0.87 0.37 0.92 0.64 Our teacher explains the practical usefulness of subject matter. problem solving (b) 3.82 0.62 e 0.19 0.66 0.89 We solve work-related problems collaboratively in small groups. simulation (b) 2.16 0.65 .80 0.26 0.75 0.79 We conduct experimental games during lessons.

Note. M ¼ mean, SD ¼ standard deviation, a ¼ Cronbach's alpha, ICC ¼ Intraclass correlation, ADM ¼ mean absolute deviation index. (a) Six-point rating format from 1 ¼ does not apply at all to 6 ¼ fully applies. (b) Six-point rating format from 1 ¼ almost never to 6 ¼ almost always.

J. Warwas, C. Helm / Teaching and Teacher Education 73 (2018) 43e5550

started with simple comparisons of scale means between the PLC profiles obtained in study 1 and calculated effect size estimates for each instructional feature. (2) We specified features that consist of multiple indicators as doubly latent constructs (Marsh et al., 2009) to account for measurement and sampling errors. In contrast, we specified features that we assessed with one or two indicators as manifest-latent constructs (Marsh et al., 2009); i.e., we partitioned the indicator variance of students' estimates at class-level to con- trol for sampling error. (3) We tested measurement invariance of all doubly latent constructs across the three PLC profiles. For one/two- item constructs, this step was omitted. (4) By using multilevel multiple group analysis, we investigated whether class-level means of instructional features deviate significantly between teachers from different PLC profiles. This was achieved by testing modera- tion effects that we defined in the model constraint commands in Mplus. To account for the nested data structure, we used the twolevel complex and the stratification command. By defining classrooms as clusters, and departments as strata, we corrected standard errors for the violation of the “independent and identi- cally distributed variable”-assumption. Considering the modest number of clusters (61 classes), we use p values < .10 and the latent mean delta's share of standard deviation at level 2 to determine whether teachers' membership in different PLC profiles systemat- ically and substantially relates to differing instructional practices and methods.

We assess model fit using Chi2/df-ratio, RMSEA, CFI, TLI, SRMR L1, and SRMR L2 as recommended in the literature (Little, 2013). To evaluate measurement invariance, we concentrate on changes in CFI (Cheung & Rensvold, 2002).

6.4. Results: profile-dependent differences in instructional quality

Step 1: Differences in the scale means of the assessed quality dimensions between departmental PLC profiles are reported in Table 7. According to Cohen's classification (1988), effect sizes (eta square) are small for individual learning support and classroom management but medium to large for application-orientated teaching, and for the frequencies in which students engage in problem solving and simulation-based learning. On average, these

Table 7 Mean comparison of instructional quality measures between latent PLC profiles.

Mean LP1 Mean LP2 Mean LP3 eta2

individual learning support 4.216 4.292 4.124 0.006 classroom management 4.256 4.155 4.147 0.005 application orientation 3.967 4.405 3.854 0.079 problem solving 3.720 4.300 3.882 0.139 simulation 2.072 2.463 2.200 0.059

Note. LP ¼ latent profile, indicating departmental configurations of PLCs (1 ¼ Structurally embedded; 2 ¼ Advanced; 3 ¼ Rudimentary).

three aspects are considerably more pronounced in Advanced PLCs (LP2) than in Structurally embedded PLCs (LP1) and Rudimentary PLCs (LP3).

Step 2: Fit indices obtained from confirmatory factor analysis for the doubly latent constructs of instructional practices suggest that all specified models accurately reflect the structure of the data (Table A in Appendix B). Thus, reliable constructs of instructional quality are available for further analyses.

Step 3: Consistent measures of latent constructs across com- parison groups are indispensable for interpreting mean level dif- ferences. Measurement invariance (MI) can be assumed if CFI values do not decrease more than 0.01 when testing the more restricted model (strong MI) against the less restricted one (weak MI) (Cheung & Rensvold, 2002). As documented in Table B in Appendix B, all latent constructs fulfil this requirement. Thus, constraining item loadings and even intercepts across groups of teachers from different PLC profiles does not lead to a substantial loss in model fit.

Step 4: Table 8 presents the results of multilevel multiple group analyses. For the doubly latent measures of teachers' instructional practices, they document excellent fit with respect to Chi2/df values below 3, CFI and TLI values above 0.95, RMSEA, SRMR L1 and L2 below 0.08 (with only one exception). Statistically significant and practically substantive differences in the instructional practices of teachers from different PLC profiles appear for the extent of appli- cation-oriented teaching (p < .05; effect size ¼ latent-mean delta's share of standard deviation at level 2: 0.707 and 0.888). More precisely, vocational students experience markedly higher degrees of application-orientated teaching if teachers stem from Advanced PLCs (LP2) then they do if teachers stem from Structurally embedded PLCs (LP1) or Rudimentary PLCs (LP3). Regarding the two manifest indicators of action- and application-oriented methods of instruc- tion, students taught by teachers from Advanced PLCs report engaging considerably more frequently (at p < .01, effect size ¼ 1.015) in problem solving in small groups, and (at p < .10, effect size ¼ 0.664) in simulation-based learning than students from Structurally embedded PLCs.

7. Discussion

7.1. Main results

By investigating if instructional quality in vocational classrooms varies by teachers’ membership in different departmental PLC profiles, the present study aimed to address several conceptual and operational weaknesses of extant research that were identified in literature reviews.

Firstly, we attempted to integrate the manifold concepts of professional learning communities by extracting three ‘core’ di- mensions of working in a PLC. The behavioural dimension, collab- orative development, comprises activities of instruction-focused

Table 8 Model fit indices and multiple latent mean comparisons of instructional quality dimensions for the 3-latent profile solution.

Modell IQ Par Chi2 df Chi2/df p RMSEA CFI TLI SRMR L1 SRMR L2 LP mean LP mean Abs. Deltaa p Effect Sizeb

M1 IS 51 115.004 54 2.13 0.000 0.052 0.984 0.982 0.030 0.044 LP1: 0.000 vs. LP2: 0.055 0.055 0.724 0.106 LP1: 0.000 vs. LP3: �0.125 0.125 0.682 0.241 LP2: 0.055 vs. LP3: �0.125 0.180 0.575 0.346

M2 CM 40 50.435 32 1.58 0.020 0.037 0.987 0.986 0.043 0.127 LP1: 0.000 vs. LP2: 0.042 0.042 0.847 0.052 LP1: 0.000 vs. LP3: �0.060 0.060 0.828 0.074 LP2: 0.042 vs. LP3: �0.060 0.102 0.718 0.127

M3 AO 30 29.335 15 1.96 0.015 0.048 0.991 0.989 0.018 0.004 LP1: 0.000 vs. LP2: 0.437 0.437 0.006 0.707 LP1: 0.000 vs. LP3: �0.112 0.112 0.662 0.181 LP2: 0.437 vs. LP3: �0.112 0.549 0.040 0.888

M4 PS 9 e e e e e e e e e LP1: 3.714 vs. LP2: 4.290 0.576 0.001 1.015 LP1: 3.714 vs. LP3: 3.869 0.155 0.590 0.273 LP2: 4.290 vs. LP3: 3.869 0.421 0.168 0.742

M5 SI 9 e e e e e e e e e LP1: 2.069 vs. LP2: 2.473 0.404 0.056 0.664 LP1: 2.069 vs. LP3: 2.168 0.099 0.713 0.163 LP2: 2.473 vs. LP3: 2.168 0.305 0.328 0.501

Note. IQ ¼ instructional quality dimensions; IS ¼ individual learning support, CM ¼ classroom management, AO ¼ application-orientated teaching, PS ¼ problem solving in small groups SI ¼ simulation-based learning. LP ¼ latent profile, indicating departmental configurations of PLCs (1 ¼ Structurally embedded; 2 ¼ Advanced; 3 ¼ Rudimentary).

a Differences in the latent means of instructional practices (M1-M3) between teachers from different profiles were calculated using multilevel group comparison in Mplus. Thereby, the latent mean of the first group (latent profile 1) was fixed to 0. Accordingly, the differences displayed in the column “Absolute Delta” represent the absolute distance between the latent means of the compared groups. M4 and M5 contain manifest indicators. Hence, model fit indices are not reported and mean levels of the comparison groups appear in the original metric.

b Effect sizes were calculated as the delta's share of the level-2 standard deviation for each instructional feature.

J. Warwas, C. Helm / Teaching and Teacher Education 73 (2018) 43e55 51

cooperation, monitoring and improvement. The ideational dimen- sion, normative agreement, depicts the degree to which PLC mem- bers hold common pedagogical beliefs, values and goals that guide collaborative efforts. The structural dimension, supportive infra- structure, indicates the presence of organizational resources that facilitate and promote collaboration.

Secondly, we identified PLC profiles that exist in vocational school departments without external interventions. Multilevel latent profile analysis, based on teacher ratings of core PLC di- mensions (N ¼ 395) in 47 departments, reveal three distinct con- figurations. Structurally embedded PLCs have a highly supportive infrastructure but exhibit only modest levels of collaborative devel- opment and normative agreement. Finding this profile to be the prevailing pattern of collective professional practice at the depart- ment level is in line with previous surveys among vocational school teachers in Germany. These suggest that many teachers judge structural conditions for cooperative activities as good, but are, nevertheless, reluctant to engage in such activities (Tenberg, 2017). Rudimentary PLCs score comparably low on all PLC dimensions, whereas Advanced PLCs convey the highest scores. We scrutinized profile composition and prevalence but could not detect systematic relations with gender or seniority of staff, school type, or vocational domain. Only teachers' work-related efficacy beliefs proved to be predictive of profile membership. Thus, teachers who are very confident of their abilities to master even difficult demands of their profession are more likely to be members of Advanced PLCs than teachers who are not. This result is well compatible with concep- tions of advantageous “personal capacity” of PLC members (e.g., Mitchell & Sackney, 2007, p. 630), even though it gives no unam- biguous answer on whether strong efficacy beliefs are a prerequi- site or a consequence of establishing elaborate forms of PLCs. Although we consider mutual reinforcement to be a realistic pos- sibility, Schaap and de Bruijn's (2017) longitudinal case-analyses indeed suggest that work-related attitudes of PLC members, particularly beliefs about ownership and alignment, affect the developmental trajectories of PLCs and, thus, varying levels of so- phistication at a given point in time. Furthermore, our finding that work-related efficacy beliefs are the only significant predictor of PLC membership corroborates the idea that the obtained profiles mainly represent emergent modes of collective professional prac- tice at the department level, which are not predetermined by

formal or hard-to-change factors such as the vocational domain or the gender ratio of a department.

Thirdly, we examined how PLC membership relates to instruc- tional quality, since creating optimal learning opportunities for students represents the primary objective of PLCs (e.g., Westheimer, 2008) but only a marginal subject of empirical research that goes beyond descriptive reports (e.g., Vescio et al., 2008). We expected teachers from comparably ‘strong’ PLC pro- files to provide better opportunities for vocational students to ac- quire occupational competencies than teachers from ‘weaker’ profiles. Survey data from 1243 students in 61 classes served to investigate this issue. Students assessed instructional practices of their respective teacher (classroom management, individual learning support, application-oriented teaching) as well as the frequency in which he/she employs action- and application- oriented methods of instruction (work-related problem solving in small groups; simulation-based learning). Preliminary profile comparisons, based on aggregated manifest criterion variables for each class, reveal substantial deviations concerning application- oriented practices and methods with medium to large effect sizes. Subsequent multilevel analyses corroborate these findings while accounting for the nested data structure and controlling for measurement and sampling errors. Hence, teachers from Advanced PLCs demonstrate significantly and substantially higher levels of application-oriented teaching (by offering explanations, examples, and materials that illustrate the practical usefulness of subject matter) than teachers from other profiles. Furthermore, these teachers employ substantially more frequently instructional methods that allow students to actively and cooperatively deal with occupation-specific tasks, instruments and procedures. Teachers from Advanced PLCs score highest on both problem- and simulation-based methods, which leads to very consistent de- viations from the other profiles, even though statistical significance can be established only when compared to Structurally embedded PLCs. On the one hand, this finding underscores the specific pro- fessional demands of teachers in vocational schools, where pro- moting the employability of graduates has high priority. On the other hand, we need to consider that available studies from the general education system examine a range of different target var- iables themselves, which further limits the comparability of results. Although many of these studies suggest that PLCs foster

J. Warwas, C. Helm / Teaching and Teacher Education 73 (2018) 43e5552

individually supportive strategies of instruction, there is also evi- dence for increased levels of student problem solving, particularly in STEM classrooms (see section 3.1).

Another interesting finding is that teachers from Structurally embedded and Rudimentary PLCs do not differ systematically regarding their instructional practices and methods. This finding suggests that these two configurations of collective professional practice are phenomenologically different but equally (un-)suc- cessful in vocational schools. Moreover, it could not have been detected by investigating linear relationships between instructional features and singular PLC dimensions or a composite PLC index under the assumption that increases in any one of multiple constitutive elements of a PLC should be associated with enhanced instructional quality. In this respect, our findings correspond with those of Lomos et al.’s (2011) study in 130 general secondary schools in the Netherlands. The authors identified four PLC profiles based on Kruse et al.’s (1995) five PLC components and compared them regarding students' math achievement. Only the strongest profile (labelled Professional community schools) contributed significantly to explaining between-school variations in math achievement. Although achievement scores were lower in all remaining profiles, they did not form a clear rank order, with the conceptually most deficient profile (Non-professional community schools) occupying the last position. Instead, profiles that displayed one-sided strengths of collective professional practice (Collaborate- activity schools, Deprivatization-of-practice schools) performed even worse. In line with our results, it seems plausible that educationally successful PLCs need to be balanced and strong in all their consti- tutive elements.

7.2. Limitations

Investigating ‘naturally occurring’ PLCs certainly has the po- tential to uncover the ways in which department staff orchestrate collective professional practice without external interventions (Fulton & Britton, 2011). However, the ‘formative’ character of these communities (Vangrieken et al., 2017) also confronts researchers with a major difficulty. Without pre-set agendas and design prin- ciples, the substantive foci of teacher collaboration inevitably vary across naturally occurring PLCs and may even change over time (DeMatthews, 2014). In the present study, we therefore opted for broad and basic dimensions of instructional quality to portray teachers' classroom performance. The underlying rationale was that creating authentic, work-related learning environments (application-orientation), maximizing students' active learning time (classroom management) and providing targeted assistance (individual learning support) undoubtedly are of vital interest to any department in vocational schools. However, relations between profile membership and instructional quality might even be stronger if criterion variables correspond more closely with the specific agendas of different PLCs, which reveal selected sub-goals under the connecting top theme of improving student learning. For example, if a PLC explicitly concentrates on fostering business process orientation in classes for commercial trainees, then the extent to which its members integrate Enterprise Resource Planning systems into classwork presents a more specific criterion for assessing instructional consequences than the comparably broader criterion of application-orientation.

Another weakness of the present study is the lack of repeated measurements, which inhibits causal interpretations of between- profile differences in instructional quality, even though the teacher survey on collective professional practice preceded the student survey on instructional features. Without rigorous control of potential confounders and a pre-post-evaluation, we cannot rule out the possibility that teachers who provide students with

superior learning opportunities also cultivate more collaboration within their department than their pedagogically less versatile colleagues do. Furthermore, generalizability is limited because (a) checks for the representativeness of our samples are unsatisfactory, and (b) the relatively low number of teachers in the profile of Rudimentary PLCs tends to increase standard errors and, thus, makes significant results more difficult to obtain.

Finally, teacher self-selection might have biased available data on instructional quality in Study 2, since merely 61 out of 395 teachers who took part in the teacher survey consented to the subsequent student survey. Therefore, study 2 may include a disproportionately high number of effective teachers because teachers who knew about their pedagogical weaknesses may have refrained from further data collection among their students.

8. Conclusion and outlook

Studies of PLCs make up a vital field of research on teachers' professional development and effectiveness in the general educa- tion sector. By investigating teachers in vocational school de- partments, the present study gained insights into collective professional practice in an educational setting that has received little attention so far. Our results suggest that teachers from Advanced PLCs e compared to teachers from other departmental profiles e provide better opportunities for vocational students to acquire occupational knowledge and skills by creating more authentic, application-oriented learning environments. However, they do not seem to demonstrate superior classroom management or individual learning support. This finding underlines the need for further investigations on context-dependencies in the work of PLCs. Demonstrable consequences for instructional quality might be conditional on a community's educational setting. To date, only few empirical investigations explicitly focus on revealing differ- ential effects of PLCs on teachers' strategies for preparing and conducting classroom instruction. Among them is Supovitz's (2002) study, which documents systematic differences in lesson planning, such as the use of flexible student grouping strategies, between PLCs in elementary, middle grade, and high schools of the same district. In order to examine potential moderators of PLCs' instructional impact, comparative study designs are needed that span different educational sectors and school types, while con- trolling for individual teacher competencies that might affect the quality of both collaborative action and classroom instruction. Nevertheless, our approach to distinguish departmental profiles, based on characteristic scoring patterns on behavioural, ideational, and structural PLC dimensions, can offer useful ideas for educa- tional research and practice. It can help researchers to systemati- cally describe and critically evaluate collective professional practice within departments in the light of desirable features of work in a PLC. Most departments in the present sample do not convey the profile of a fully developed, ideal-type PLC. At the same time, profile analysis can help teachers to identify potential deficiencies in their respective departments and to plan targeted measures for improvement.

Acknowledgements

Christoph Weber (Teacher Training Institute, Linz, Austria) for valuable statistical advice.

Appendix A

PLC items

supportive infrastructure (SI) Workflows are well structured. Responsibilities are clearly defined. You often get the feeling that nobody feels in charge of any of our activities.a

Everyone is well informed about important decisions and developments. Teacher conferences are well planned and executed. One can safely assume that decisions will be implemented consistently. Our work consists of many incoherent and independent actions.a

It is always just a few teachers who take on effortful school-related tasks within our shared area of responsibility.a

collaborative development (CD) We often prepare instructional units together. Teachers often engage in joint projects. Cooperative planning of teaching topics is an exception in our unit.a

Teachers visit each other's classes regularly. Self-assessments and mutual evaluations are a matter of course in our work. In our unit, teachers rarely develop instructional content together in a multi-perspective way.a

Normally, teachers don't know exactly the topics colleagues deal with in their classes.a

normative agreement (NA) It is hard to find common ground in the pedagogical beliefs of our teaching staff.a

Students must adjust to different teaching strategies in lessons that are taught by different colleagues.a

Every colleague has his/her own ideas of instructional design.a

The teachers of our unit only comply with their individual working standards.a

Our students can be sure that all teachers place similar demands on instructional interaction.

a inversely coded.

J. Warwas, C. Helm / Teaching and Teacher Education 73 (2018) 43e55 53

Appendix B

Table A Fit indices of doubly latent models measuring teachers' instructional practices with multiple indicators.

Doubly Latent Par Chi2 df Chi2/df p RMSEA CFI TLI SRMR L1 SRMR L2

Individual learning support

27 15.069 8 1.884 .058 .027 .997 .991 .007 .035

Classroom management

20 5.175 4 1.294 .270 .015 .999 .997 .001 .018

application- oriented teaching

14 0.198 1 0.198 .656 .000 1.000 1.005 .000 .003

Note. L1 ¼ level 1; L2 ¼ level 2.

Table B Measurement invariance of doubly latent constructs of teachers' instructional practices across teachers from different PLC profiles.

N par df Chi2 CFI DCFI

IS configural invariance 80 25 43.922 0.995 metric invariance 64 41 83.990 0.989 0.006 scalar invariance 54 51 99.515 0.987 0.002

CM configural invariance 60 12 10.498 1.000 metric invariance 50 22 24.400 0.998 0.002 scalar invariance 42 30 42.818 0.991 0.007

AO configural invariance 42 3 0.853 1.000 metric invariance 37 8 22.028 0.991 0.009 scalar invariance 29 16 29.708 0.991 0.000

Note. IS ¼ individual learning support, CM ¼ classroom management, AO ¼ appli- cation orientation. N par ¼ number of parameters, df ¼ degrees of freedom.

References

Abele, A. E., Stief, M., & Andr€a, M. S. (2000). Zur €okonomischen Erfassung beruflicher SelbstwirksamkeitserwartungeneNeukonstruktion einer BSW-Skala. (On the economic measurement of occupational self-efficacy expectations - a new BSW

Scale). Zeitschrift für Arbeits- und Organisationspsychologie, 44(3), 145e151. Achinstein, B. (2002). Conflict amid community: The micropolitics of teacher

collaboration. Teachers College Record, 104(3), 421e455. Achtenhagen, F., & Winther, E. (2014). Workplace-based competence measurement:

Developing innovative assessment systems for tomorrow's VET programmes. Journal of Vocational Education and Training, 66(3), 281e295.

Andrews, D., & Lewis, M. (2007). Transforming practice from within: The power of the professional learning community. In L. Stoll, & K. S. Louis (Eds.), Professional learning communities: Divergence, depth and dilemmas (pp. 134e148). Maiden- head: Open University Press.

Bolam, R., McMahon, A., Stoll, L., Thomas, S., & Wallace, M. (2005). Creating and sustaining effective professional learning communities. Bristol: University of Bristol.

Bonsen, M., & Rolff, H.-G. (2006). Professionelle Lerngemeinschaften von Lehrer- innen und Lehrern. Zeitschrift für Padagogik, 52(2), 167e184.

Buhren, C. G. (2015). Personalentwicklung durch kollegiale Hospitation und Pro- fessionelle Lerngemeinschaften. In N. Berkemeyer, J. Berkemeyer, & F. Meetz (Eds.), Professionalisierung und Schulleitungshandeln. Wege und Strategien der Personalentwicklung an Schulen (pp. 160e184). Weinheim: Beltz Juventa.

Burke, M. J., Finkelstein, L. M., & Dusig, M. S. (1999). On average deviation indices for estimating interrater agreement. Organizational Research Methods, 2(1), 49e68.

Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling: A Multidisci- plinary Journal, 9(2), 233e255.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences. New York: Psychology Press.

De Bruijn, E., & Leeman, Y. (2011). Authentic and self-directed learning in vocational education: Challenges to vocational educators. Teaching and Teacher Education, 27(4), 694e702.

DeBourgh, G. A., & Prion, S. K. (2011). Using simulation to teach prelicensure nursing students to minimize patient risk and harm. Clinical Simulation in Nursing, 7(2), e47ee56.

J. Warwas, C. Helm / Teaching and Teacher Education 73 (2018) 43e5554

DeMatthews, D. (2014). Principal and teacher collaboration: An exploration of distributed leadership in professional learning communities. International Journal of Educational Leadership and Management, 2(2), 176e206.

Ditton, H. (n.d.). QuaSSU e Qualit€atsSicherung in Schule und Unterricht. (QuaSSU e Quality assurance in schools and instruction.) Retrieved from: http://www. quassu.net/. [19.10.2017].

Fiss, P. C. (2007). A set-theoretic approach to organizational configurations. Acad- emy of Management Review, 32(4), 1180e1198.

Fulton, K., & Britton, T. (2011). STEM teachers in professional learning communities: From good teachers to great teaching. Washington: National Commission on Teaching and America’s Future (NCTAF).

Gessler, M., & Howe, F. (2015). From the reality of work to grounded work-based learning in German vocational education and training: Background, concept and tools. International Journal for Research in Vocational Education and Training, 2(3), 214e238.

Gresov, C., & Drazin, R. (1997). Equifinality: Functional equivalence in organization design. Academy of Management Review, 22(2), 403e428.

Grollmann, P., & Bauer, W. (2008). Technical and vocational education and training research for the professionalisation of vocational teachers. In F. Rauner, & R. Maclean (Eds.), Handbook of technical and vocational education and training research (pp. 385e392). Dordrecht: Springer.

Gruehn, S. (2000). Unterricht und schulisches Lernen: Schüler als Quellen der Unter- richtsbeschreibung. (Instruction and academic learning: Students as sources for describing instruction.). Münster: Waxmann.

Hallam, P. R., Smith, H. R., Hite, J. M., Hite, S. J., & Wilcox, B. R. (2015). Trust and collaboration in PLC teams: Teacher relationships, principal support, and collaborative benefits. NASSP Bulletin, 99(3), 193e216.

Hellwig, S. (2006). The competency debate in German VET: An analysis of current reform approaches. International Journal of Training Research, 4(1), 1e16.

Helm, C. (2016). Zentrale Qualit€atsdimensionen von Unterricht und ihre Effekte auf Schüleroutcomes im Fach Rechnungswesen. Zeitschrift für Bildungsforschung, 6(2), 101e119.

Henry, K. L., & Muth�en, B. O. (2010). Multilevel latent class analysis: An application of adolescent smoking typologies with individual and contextual predictors. Structural Equation Modeling: A Multidisciplinary Journal, 17(2), 193e215.

Hipp, K. K., & Huffman, J. B. (2010). Demystifying the concept of professional learning communities. In K. K. Hipp, & J. B. Huffman (Eds.), Demystifying pro- fessional learning communities. School leadership at its best (pp. 11e21). New York: Rowman, Littlefield Education.

Hord, S. M. (1997). Professional learning communities. Communities of continuous inquiry and improvement. Austin: Southwest Educational Development Laboratory.

Khaled, A., Gulikers, J., Biemans, H., van der Wel, M., & Mulder, M. (2014). Charac- teristics of hands-on simulations with added value for innovative secondary and higher vocational education. Journal of Vocational Education and Training, 66(4), 462e490.

Koschmann, A. (2013). Kooperation von Lehrkr€aften zur Umsetzung der lernfeldor- ientierten Lehrpl€ane: Eine explorative Untersuchung an beruflichen Schulen in Niedersachsen. (Teacher cooperation for implementing learning-field curricula. An explorative study in vocational scholls in Lower Saxony). G€ottingen: Sierke.

Kruse, S. D., Louis, K. S., & Bryk, A. S. (1995). An emerging framework for analyzing school professional community. In K. S. Louis, & S. D. Kruse (Eds.), Profession- alism and community: Perspectives on reforming urban schools (pp. 23e44). Thousand Oaks: Corwin Press.

Kunter, M., & Voss, T. (2013). The model of instructional quality in COACTIV: A multicriteria analysis. In M. Kunter, J. Baumert, W. Blum, U. Klusmann, S. Krauss, & M. Neubrand (Eds.), Cognitive activation in the mathematics classroom and professional competence of teachers. Results from the COACTIV project (pp. 97e124). New York: Springer.

Lavi�e, J. M. (2006). Academic discourses on school-based teacher collaboration: Revisiting the arguments. Educational Administration Quarterly, 42(5), 773e805.

Lee, L. C. (2013). The diversity of school organizational configurations. International Journal of Research and Method in Education, 36(4), 309e340.

Lee, J. C., Zhang, Z., & Yin, H. (2011). A multilevel analysis of the impact of a pro- fessional learning community, faculty trust in colleagues and collective efficacy on teacher commitment to students. Teaching and Teacher Education, 27(5), 820e830.

Little, J. W. (2012). Professional community and professional development in the learning-centered school. In M. Kooy, & K. van Veen (Eds.), Teacher learning that matters. International perspectives (pp. 22e43). New York, & London: Routledge.

Little, T. (2013). Longitudinal structural equation modeling. New York: The Guilford Press.

Lomos, C., Hofman, R. H., & Bosker, R. J. (2011a). The relationship between de- partments as professional communities and student achievement in secondary schools. Teaching and Teacher Education, 27(4), 722e731.

Lomos, C., Hofman, R. H., & Bosker, R. J. (2011b). Professional communities and student achievement e a meta-analysis. School Effectiveness and School Improvement, 22(2), 121e148.

Louis, K. S., & Marks, H. M. (1998). Does professional community affect the class- room? Teachers' work and student experiences in restructuring schools. American Journal of Education, 106(4), 532e575.

Loyens, S. M., & Gijbels, D. (2008). Understanding the effects of constructivist learning environments: Introducing a multi-directional approach. Instructional Science, 36(5e6), 351e357.

Lüdtke, O., Trautwein, U., Kunter, M., & Baumert, J. (2006). Reliability and agreement

of student ratings of the classroom environment e a reanalysis of TIMSS data. Learning Environments Research, 9(3), 215e230.

Luko�cien_e, O., Varriale, R., & Vermunt, J. K. (2010). The Simultaneous decision(s) about the number of lower- and higher-level classes in Multilevel Latent Class Analysis. Sociological Methodology, 40(1), 247e283.

Marsh, H. W., Lüdtke, O., Robitzsch, A., Trautwein, U., Asparouhov, T., Muth�en, B. O., et al. (2009). Doubly-latent models of school contextual effects: Integrating multilevel and structural equation approaches to control measurement and sampling errors. Multivariate Behavioral Research, 44(6), 764e802.

M€arz, V., & Kelchtermans, G. (2013). Sense-making and structure in teachers' reception of educational reform. A case study on statistics in the mathematics curriculum. Teaching and Teacher Education, 29(1), 13e24.

Mitchell, C., & Sackney, L. (2007). Building schools, building people: The school principal's role in leading a learning community. Journal of School Leadership, 16(5), 627e640.

Muth�en, L. K., & Muth�en, B. O. (1998e2014). Mplus user's guide (7th ed.). Los Angeles, CA: Muth�en & Muth�en.

Owen, S. (2014). Teacher professional learning communities: Going beyond contrived collegiality toward challenging debate and collegial learning and professional growth. Australian Journal of Adult Learning, 54(2), 54e77.

Pahl, J. P. (2007). Berufsbildende Schule: Bestandsaufnahme und Perspektiven. (Voca- tional schools: Status quo and future perspectives.). Bielefeld: W. Bertelsmann Verlag.

Penner-Williams, J., Diaz, E. I., & Gonzales Worthen, D. (2017). PLCs: Key PD component in learning transfer for teachers of English learners. Teaching and Teacher Education, 65, 215e229.

Praetorius, A. K., Pauli, C., Reusser, K., Rakoczy, K., & Klieme, E. (2014). One lesson is all you need? Stability of instructional quality across lessons. Learning and In- struction, 31, 2e12.

R€oder, L. (2017). Kollegiale Teamarbeit an berufsbildenden Schulen in Hessen. Empirische Befunde zu Implementierung und Qualit€at (Collegial Teamwork in vocational schools in Hesse. Empirical results on implementation and quality). Frankfurt a.M.: Peter Lang.

Schaap, H., & de Bruijn, E. (2017). Elements affecting the development of profes- sional learning communities in schools. Learning Environments Research, 1e26.

Scheerens, J. (2014). Evidence based educational policy and practice: The case of applying the educational effectiveness knowledge base. Journal of Educational, Cultural and Psychological Studies, 1(9), 83e99.

Schneider, S. L. (2008). Applying the ISCED-97 to the German educational qualifi- cations. In S. L. Schneider (Ed.), The International Standard Classification of Ed- ucation. An evaluation of content and criterion validity for 15 European countries (pp. 77e102). Mannheim: MZES.

Seeber, S., & Squarra, D. (2003). Lehren und Lernen in beruflichen Schulen. Schüler- urteile zur Unterrichtsqualit€at. (Instruction and learning in vocational schools. Student judgments on instructional quality.). Frankfurt am Main: Peter Lang.

Seidel, T., & Shavelson, R. J. (2007). Teaching effectiveness research in the past decade: The role of theory and research design in disentangling meta-analysis results. Review of Educational Research, 77(4), 454e499.

Sembill, D., Rausch, A., & K€ogler, K. (2013). Non-cognitive facets of competence: Theoretical foundations and implications for measurement. In K. Beck, & O. Zlatkin-Troitschanskaia (Eds.), From diagnostics to learning success. Pro- ceedings in vocational education and training (pp. 199e211). Rotterdam, Boston, & Taipei: Sense Publishers.

Sigurdardottir, A. K. (2010). Professional learning community in relation to school effectiveness. Scandinavian Journal of Educational Research, 54(5), 395e412.

Sleegers, P., den Brok, P., Verbiest, E., Moolenaar, N. M., & Daly, A. J. (2013). Toward conceptual clarity. A multidimensional, multilevel model of professional learning communities in Dutch elementary schools. The Elementary School Journal, 114(1), 118e137.

Solga, H., Protsch, P., Ebner, C., & Brzinsky-Fay, C. (2014). The German vocational education and training system: Its institutional configuration, strengths, and challenges. Discussion Paper No. SP I 2014-502. Berlin: WZB. Retrieved from: http://hdl.handle.net/10419/104536 29.09.2017.

Statistisches Bundesamt. (2015). Bildung und Kultur. Berufliche Schulen. (Education and culture. Vocational schools.). Wiesbaden: Statistisches Bundesamt. Retrieved from: https://www.destatis.de/DE/Publikationen/Thematisch/ BildungForschungKultur/Schulen/BeruflicheSchulen2110200167004.pdf? __blob¼publicationFile 19.10.2017.

Stegall, D. A. (2011). Professional learning communities and teacher efficacy: A cor- relation study. Doctoral thesis. United States of America: Appalachian State University. Retrieved from: https://libres.uncg.edu/ir/asu/f/Stegall,%20David_ 2011_Dissertation.pdf 23.06.2017.

Steinert, B., Gerecht, M., Klieme, E., & D€obrich, P. (2003). Skalen zur Schulqualit€at: Dokumentation der Erhebungsinstrumente. Materialien zur Bildungsforschung e Band 10. (Scales of school quality: Documentation of instruments. Materials on educational research e Volume 10.). Frankfurt/Main. Retrieved from http://www. pedocs.de/volltexte/2010/3121/pdf/MatBild_Bd17_D_A.pdf, 19.10.2017.

Stoll, L., Bolam, R., McMahon, A., Wallace, M., & Thomas, S. (2006). Professional learning communities. A review of the literature. Journal of Educational Change, 7(4), 221e258.

Sun-Keung Pang, N., Wang, T., & Lai-Mei Leung, Z. (2016). Educational reforms and the practices of professional learning community in Hong Kong primary schools. Asia Pacific Journal of Education, 36(2), 231e247.

Supovitz, J. A. (2002). Developing communities of instructional practice. Teachers College Record, 104(8), 1591e1626.

J. Warwas, C. Helm / Teaching and Teacher Education 73 (2018) 43e55 55

Tein, J.-Y., Coxe, S., & Cham, H. (2013). Statistical power to detect the correct number of classes in Latent Profile Analysis. Structural Equation Modeling: A Multidisci- plinary Journal, 20(4), 640e657.

Tenberg, R. (2017). Zum Stand der Lehrpersonenkooperation an berufsbildenden Schulen. (State of research on teacher cooperation in vocational schools.). Zeitschrift für Berufs- und Wirtschaftsp€adagogik, 113(2), 179e201.

Tofighi, D., & Enders, C. K. (2007). Identifying the correct number of classes in mixture models. In G. R. Hancock, & K. M. Samulelsen (Eds.), Advances in latent variable mixture models (pp. 317e341). Greenwich, CT: Information Age.

Vangrieken, K., Dochy, F., Raes, E., & Kyndt, E. (2015). Teacher collaboration: A systematic review. Educational Research Review, 15, 17e40.

Vangrieken, K., Meredith, C., Packer, T., & Kyndt, E. (2017). Teacher communities as a

context for professional development: A systematic review. Teaching and Teacher Education, 61, 47e59.

Vescio, V., Ross, D., & Adams, A. (2008). A review of research on the impact of professional learning communities on teaching practice and student learning. Teaching and Teacher Education, 24(1), 80e91.

Visscher, A., & Witziers, B. (2004). Subject departments as professional commu- nities. British Educational Research Journal, 30(6), 785e800.

Westheimer, J. (2008). Learning among colleagues: Teacher community and the shared enterprise of education. In M. Cochran-Smith, S. Feiman-Nemser, J. McIntyre, & K. Demers (Eds.), Handbook of research on education (pp. 756e783). New York: Routledge.