University-based teacher educators are a professional group who are responsible for the initial and ongoing education of pre-service teachers (European Commission, 2012; Kelchtermans et al., 2018). Identifying the role of teacher educators can be challenging since they work as mediators between the academic world, the world of teacher education, and the world of practicing teachers (Reynolds et al., 2013). In almost all European countries, becoming a teacher educator requires no formal preparation and often only minimum support from more experienced colleagues (Wilson, 1990).

Teacher educators are hired by universities based on their teaching qualifications and teaching experience (e.g., Bourke et al., 2018; Guberman and Mcdossi, 2019). Because their main task is teaching, the ones who have teaching experience in schools feel more confident since this is their chosen career, and that is the main reason they are recruited as teacher educators. However, teaching pre-service teachers about teaching is different from teaching pupils in school. Teacher educators are expected to familiarize themselves with research in the teacher education field and to engage in research in order to improve their own quality of practice (Cochran-Smith, 2005; Loughran, 2014; Vanassche and Kelchtermans, 2015).

Tack and Vanderlinde (2014) investigated school-based teacher educators' attitudes toward research and found out that active researchers have positive attitudes toward research, and because they are involved in research themselves, they contribute to the knowledge base of teacher education. This is because researchers who are also working as teacher educators are able to supervise their students' research projects to conduct their own research and to publish their findings, and, thus, they actively advance the evidence base in teacher education.

In contrast to the positive attitudes toward teaching, teacher educators' attitudes regarding professional development in research are mixed (Griffiths et al., 2014). Teacher educators with less exposure in research, especially the ones who come from schools rather than academia or the ones who are not engaged in research, feel underestimated by their colleagues and the university structures. Their lack of motivation and low self-efficacy beliefs leads them to neglect participation in research activities for the improvement of their research skills (Meeus et al., 2018).

Several challenges such as the lack of resources or skills and sometimes institutional support can lead to discrepancies between attitudes and actual behavior (Cochran-Smith, 2005; Brown and Zhang, 2016). As aforementioned, given the research exposure and academic background of teacher educators, it is expected that they will face different challenges regarding EBT implementation (Czerniawski et al., 2017). To summarize, teacher educators with higher research exposure and positive experiences in research might feel more confident about research while their counterparts might have difficulty coping as well with research (Zeichner, 2005; Czerniawski et al., 2017). To support teacher educators' EBT professional development our study sheds light on their views about the biggest challenges and incentives to increase the frequency of EBT implementation in university classrooms.

In the context of current EBT reforms, we discuss teacher educators' professional development from the perspective of developing teacher educators' practical knowledge and self-efficacy beliefs toward research. Since EBT has its roots in medicine, it is reasonable to build on the knowledge base of the field. Thus, the current investigation was based on two theoretical models, one stemming from medical education (Miller, 1990) and Bandura's social cognitive theory (1977) stemming from education.

Miller's pyramid has its roots in constructivism and situated learning theories and its hierarchical structure is similar to Bloom's taxonomy (Constantinou et al., 2018). This model helps to match learning outcomes with one's expectations about the abilities of the learners at any stage of their professional development. The model is usually described as having four levels: knowing, knowing how, showing, and doing. The first two levels are categorized as cognitive aspects and the second two as behavioral aspects. In order for professionals to apply new reforms such as the EBT reform in practice, they need to fulfill all four levels stated above (McKinley, 2006). In the present study, the first two levels or the cognitive aspect are investigated by measuring teacher educators' practical knowledge and self-efficacy beliefs. Also, one behavioral aspect, the “doing” level, is investigated by measuring how often teacher educators implement EBT practices in their classrooms.

The above-mentioned model can be greatly supplemented by Banduras' social cognitive theory (1977) which emphasized the major role of certain cognitive processes in encoding and performing behaviors. Self-efficacy beliefs refer to the teacher's perceived ability and confidence to perform a given behavior (Bandura, 1977). Literature suggests that a number of parameters may affect self-efficacy beliefs (e.g., stress, well-being) among those, experiences have been identified as strongly related with teachers change (Holzberger et al., 2013). Thus, self-efficacy beliefs may change in response to teachers' previous experiences.

However, based on the aforementioned fundamental paper (Holzberger et al., 2013), social-cognitive theory goes well-beyond self-efficacy, this is why it is important to see self-efficacy in a larger framework including further cognitive determinants of behavior initiations (Lippke, 2017). Building on the theoretical model from Miller (1990), we base our study on the four different levels described above to frame our variables under behavioral and cognitive processes. Accordingly, based on the social cognitive theory (Bandura, 1977), we include experience as an important aspect interacting with self-efficacy beliefs and practical knowledge to build our mediation model. Thus, based on the aforementioned theoretical frameworks (for an overview see Figure 1), we expect that teacher educators' research experience affects their self-efficacy beliefs and practical knowledge which in turn impact their teaching practices (frequency of EBT use).

Understanding the relationship between teacher educators' research experience, practical knowledge, self-efficacy beliefs, and frequency of EBT implementation plays a pivotal role in fostering teacher educators' professional development and improving their teaching quality. This study attempts to understand whether increased research exposure can reinforce and strengthen the frequency of EBT implementation in university classrooms. We also aim to understand whether teacher educators' practical knowledge and self-efficacy can be relevant mediators in the interplay between research exposure and EBT use. In addition, we seek to understand teacher educators' views regarding the biggest challenges and facilitators to increasing EBT practices in universities. Thus, we propose the following research questions:

A total sample of N = 243 teacher educators (60% female) completed the study, with ages ranging from 23 to 68 (M = 44, SD = 11.55). The sample included teacher educators from Germany (n = 152), the German-speaking part of Switzerland (n = 40), Austria (n =22), and the United Kingdom (n = 28). Teacher educators were recruited into the study via e-mail or through an institutional research subject pool. A total of 691 teacher educators entered the survey and 243 completed it (35% response rate). One participant did not state his or her country of origin. Research exposure was measured based on teacher educators' university position, and we identified the five following groups: school mentors (group 1: n = 33), teaching associates (group 2: n =17), PhD candidates with teaching obligations (group 3: n = 80), postdoctoral candidates with teaching obligations (group 4: n = 23), and professors (group 5: n = 90). Teaching experience in the university ranged from 5 months to 43 years (M = 11, SD = 8.80).

We used a correlational design and survey methods to quantitively investigate the role of research experience, practical knowledge, and self-efficacy beliefs of teacher educators toward the use of evidence-based teaching practice. In addition, we explored teacher educators' views about the challenges and incentives to the use of EBT. As teacher education contexts, we chose German-speaking cultures (Germany, Austria, and the German-speaking part of Switzerland) and the United Kingdom, all of which have long histories of teacher education. Teacher educators gave informed consent before participation.

The scales and subscales used in this study were taken out of a newly developed instrument named Evidence-Based Teaching Scale (EBTS) measuring teacher educators' practical knowledge, self-efficacy beliefs, and attitudes toward Evidence-based teaching (Georgiou, 2020). Below we will provide a brief overview of the scale development, the factorial structure and convergent validity of the original evidence-based teaching scale.

A review of the literature was conducted to identify potential scales that could measure teachers' or teacher educators' knowledge, beliefs, and attitudes toward the use of scientific evidence in their teaching practice. A search was conducted in PsycINFO and Web of Science databases using the keywords “evidence-based practice” or “evidence-based teaching” combined with “knowledge,” “believes,” “attitudes,” “measures,” or “instruments.” Because the first search yielded no scales for teachers or teacher educators, a new search was conducted using the keywords “evidence-based practice” and “measures” or “instruments.” This search produced six potential Evidence-based practice scales and a semi-structured interview guide that have been used in medicine (for an overview see Table 1).

A three-step approach was followed for the selection of the items. This process included two expert validation faces and a pilot study with 30 teacher educators. After the above-mentioned process, the questionnaire consisted of 61 items in total. In detail, 47 items aimed to assess teacher educators' practical knowledge, self-efficacy beliefs, and attitudes toward EBT. Sixteen additional items developed to investigate teacher educators' frequency of and confidence in EBT practices.

For the development of the EBTS scale, an exploratory factor analysis, using principal axis factoring and an oblique rotation, was conducted on the 47-item evidence-based teaching scale for N = 210 respondents. The Bartlett test was significant (p < 0.000) and the Kaiser-Meyer-Olkin Measure of sampling adequacy (KMO = 0.84) indicated that the variables were suitable for factor analysis. To reveal optimal results, low loadings (λ < 0.4) or freestanding items were dropped. To decide on the adequate number of factors, the original study used the eigenvalue-criterion, the scree test and conducted a parallel analysis (Costello and Osborne, 2005). A three-factor solution showed the best fit for the analyzed data. Table 2 highlights the factor loadings for the three factors of the evidence-based teaching scale.

Cronbach's alpha was calculated for all three scales that loaded on the three factors. The results indicate that the three scales have a good internal consistency (George and Allery, 2003). In detail, for the practical knowledge scale α = 0.84, for the self-efficacy scale α = 0.85, and the attitudes scale α = 0.83. Pearson's r was used to determine the degree to which the three aforenoted factors of the evidence-based teaching scale are interrelated. Results showed moderate and significant positive correlations among the scales, supporting the notion that the variables are theoretically highly interrelated but at the same time distinct from each other. In particular the practical knowledge scale and the self-efficacy scale demonstrated the strongest relationship (r = 0.57, p = 0.00) while self-efficacy and attitudes the weakest (r = 0.23, p = 0.00).

To examine teacher educators' practical knowledge and self-efficacy beliefs toward the use of evidence-based teaching practices, we used the subscales practical knowledge and self-efficacy beliefs from the EBTS instrument (Georgiou, 2020). For both subscales a 6-point Likert scale was used, with 6 being the highest level of agreement. A similar 6-point Likert scale was used to measure challenges (r = 0.20; p < 0.001) and incentives of EBT use (r = 0.17; p < 0.001). Additionally, a 10-point frequency scale (α = 0.84) ranging from 0 (no use) to 10 (EBT use ten or more times a semester) measured teacher educators' strategies of EBT implementation in a time span of a semester (example item: In the past semester how often have you a) read published research reports, b) shared and discussed literature findings with your colleagues, c) interpreted results from experimental studies).

An explanation of the term “evidence-based practice” was provided in the introduction to the survey. This way we ensured that all participants had a similar information basis regarding the EBT concept. Finally, research experience was measured based on teacher educators' university positions (a) school mentors, (b) teaching associates, (c) PhD candidates with teaching obligations, (d) postdoctoral candidates with teaching obligations, (e) professors.

Mediation refers to the effect of the independent variable (predictor) on the dependent variable (outcome) through a third variable (mediator), which mediates the relationship between the predictor and outcome. In this study, we hypothesized that teacher educators' research experience would affect their practical knowledge and self-efficacy beliefs, which, in turn, would affect the frequency of EBT implementation in university classrooms (research question 1). To test our hypothesis, we ran a step-wise mediation analyses (Hayes, 2013). As literature suggests and because our study is correlational the proposed model is not causal in nature (e.g., Sobel, 2008; Hayes, 2013; Agler and De Boeck, 2017). Figure 2 presents the two hypothesized relationship models. The first model refers to the relationships among teacher educators' research experience and frequency of EBT implementation with practical knowledge as a mediator. While, the second model refers to the relationship among teacher educators' research experience and frequency of EBT implementation with self-efficacy beliefs as a mediator. As a third step, an overall model with both mediators added was calculated.

In the overall model, the path c is the original direct effect of research experience on frequency of EBT implementation without the mediators (practical knowledge and self-efficacy beliefs), while c' is the direct effect of research experience on frequency of EBT implementation when the mediators are included in the model. Paths a₁, a₂and b₁, b₂ represent the effect of research experience on practical knowledge and self-efficacy beliefs and that of both mediators on frequency of EBT implementation, respectively. The product of ab is the indirect effect of research experience on frequency of EBT implementation via the mediator's practical knowledge and self-efficacy beliefs, which is commonly known as the mediation effect. Statistically, the total effect is the sum of the direct effect and mediation effect (c = c'+ab). For testing the mediation effect ab, we used the bootstrapping procedure in PROCESS model 4 macro Version 3.3 (Hayes, 2018) by obtaining the bias-corrected bootstrapped confidence interval limits. Statistical significance was confirmed if the confidence interval was greater than zero.

In addition to testing the mediation effect ab, statistical evidence for the mediating role of teacher educators' practical knowledge and self-efficacy beliefs was enhanced by comparing the total effect c (i.e., direct effect of research experience on frequency of EBT implementation without the mediator in the model) with path c' (i.e., direct effect of research experience on frequency of EBT implementation with the mediator in the model). Statistically, if a mediator is significant, it is expected that c' would be smaller than c. Traditionally, if c' becomes statistically non-significant, this would be considered as full mediation otherwise, it would be considered as partial mediation (Baron and Kenny, 1986). Recently, the concept of full mediation is considered problematic because it excludes the possibility of other mediation mechanisms via other variables between the independent and dependent variables of interest (Pituch et al., 2005). Therefore, literature supports the idea that all mediations can be considered as partial mediations (Preacher and Hayes, 2004).

To capture any differential effects among teacher educator groups based on their research exposure, we first performed a multivariate analysis of variance (MANOVA) on practical knowledge and self-efficacy beliefs as dependent variables, with teacher educators' university rank being defined as the independent variable. Accordingly, univariate comparisons were calculated in order to explore the differences across the teacher educator groups. Post-hoc tests and effect sizes were also calculated for all groups using Hedge's g (Lakens, 2013). A Hedges' g < 0.5 was defined as a “small,” 0.5 to 0.8 as a “moderate,” and > 0.8 as a “large” effect size. The effect sizes for all significant differences ranged from g = 0.3 to 1.8.

In order to answer the second research question of the study, we first descriptively analyzed teacher educators' views on challenges and facilitators of EBT use. Accordingly, we performed analysis of variance (ANOVA) on all challenges and facilitators as dependent variables, with teacher educators' research exposure defined as the independent variable. Next, we performed univariate comparisons to explore the differences across the teacher educator groups.

The results of the mediation analyses for all three models are reported in Figure 3. The total effect c of teacher educators' research experience on the frequency of EBT use was statistically significant for all models. Next, a significant coefficient a relating teacher educators' research experience to the hypothesized mediators—practical knowledge and self-efficacy beliefs—was confirmed. The coefficient b for both practical knowledge (b = 0.45) and self-efficacy beliefs (b = 0.32) was statistically significant (Preacher and Hayes, 2004). The mediation effect was shown to be statistically significant (a₁b₁ = 0.149, p < 0.05; a₂b₂ = 0.073, p < 0.05). When the mediators were included in the model, the direct effect c' was still statistically significant for each of the two models (c' = 0.31). However, when both mediators were included in the same model, only self-efficacy beliefs remained a significant mediator.

These findings support the prediction that teacher educators' research experience is related to the frequency of EBT use in university classrooms. Furthermore, teacher educators' practical knowledge and self-efficacy beliefs function as mediators between their research experience and their practice. However, self-efficacy beliefs seem to be a stronger predictor of how frequently one uses EBT practices as compared to practical knowledge. The mediation effect of practical knowledge on the frequency of EBT use accounts for 25% of the total effect, and the mediation effect of self-efficacy beliefs on the frequency of EBT use accounts for 33% of the total effect. Thus, our hypothesis that self-efficacy beliefs act as a mediator between research exposure and frequency of EBT use is confirmed. Concerning practical knowledge, our hypothesis is partially confirmed, since practical knowledge does not remain a significant mediator when both mediators are added in the same model.

Table 3 shows the descriptive statistics of teacher educators' practical knowledge, self-efficacy beliefs, and frequency of EBT use for the whole sample and for every group of teacher educators based on their research experience measured by their university position. Overall, teacher educators reported to be relatively knowledgeable about EBT practices. The sample mean for the practical knowledge scale was 4.74 (SD = 0.59) out of a possible 6 points, with a range of 1–6. Accordingly, participants also reported rather high self-efficacy beliefs (M = 5.10, SD = 0.71) about the implementation of EBT practices. Professors reported the highest practical knowledge of the five groups and teaching associates the lowest. Professors also reported the highest values for self-efficacy beliefs, and this time school mentors reported the lowest. Teacher educators reported moderate use of EBT practices. The sample mean was 7.00 (SD = 2.34) out of a possible 10 points, with a range of 0–10. Professors reported to use EBT practices more often than all other groups of teacher educators, with teaching associates being the group with the lowest reported frequency of EBT use. In general, professors who were more exposed to research showed descriptively higher values for all three scales, while teaching associates and school mentors with less exposure to research reported the lowest values.

The selected teacher educator groups systematically differed regarding EBT practical knowledge, self-efficacy beliefs, and frequency of EBT use. The MANOVA showed an overall significant effect regarding group differences on EBT measures, F_{(12, 624)} = 4.67, p = 0.000; Wilk's Λ = 0.797, η² = 0.07. Univariate comparisons (ANOVA) and follow up Tukey's HSD post-hoc pair comparisons revealed further significant differences across different teacher educator groups in practical EBT knowledge, F_{(4, 238)} = 5.54, p = 0.000, η² = 0.08 and self-efficacy beliefs, F_{(4, 238)} = 10.85, p = 0.000, η² = 0.15.

The largest effect sizes were observed for differences in EBT knowledge between professors (M = 4.95, SD = 0.49) and school mentors (M = 4.55, SD = 0.47) (Hedge's g_s = 0.80, p = 0.000), and between professors (M = 4.95, SD = 0.49) and teaching associates (M = 4.48, SD = 0.46) (Hedge's g_s = 0.92, p = 0.000). Professors practical knowledge (M = 4.95, SD = 0.49) also significantly differed from postdoctoral candidates (M = 4.73, SD = 0.44) (Hedge's g_s = 0.44, p = 0.005), and PhD candidates (M = 4.67, SD = 0.73) (Hedge's g_s = 0.45, p = 0.000).

For self-efficacy beliefs, the largest effect size was observed for differences between professors (M = 5.41, SD = 0.54) and school mentors (M = 4.55, SD = 0.69) (Hedge's g_s = 1.44, p = 0.000), between professors M = 5.41, SD = 0.54 and teaching associates (M = 4.75, SD = 0.71) (Hedge's g_s = 1.10, p = 0.000), and between professors (M = 5.41, SD = 0.54) and PhD candidates (M = 5.00, SD = 0.76) (Hedge's g_s = 0.62, p = 0.000). Significant differences were also found between postdoctoral candidates (M = 5.17, SD = 0.55) and teaching associates (M = 4.75, SD = 0.71) (Hedge's g_s = 0.64, p = 0.004), and postdoctoral candidates (M = 5.17, SD = 0.55) with school mentors (M = 4.55, SD = 0.69) (Hedge's g_s = 0.95, p = 0.000). PhD candidates self-efficacy beliefs (M = 5.00, SD = 0.76) also differed significantly from school mentors self-efficacy beliefs (M = 4.55, SD = 0.69) (Hedge's g_s = 0.59, p = 0.000).

Tables 4, 5 summarize the mean overall scores for the items referring to challenges and facilitators to EBT implementation. Differences between the groups are depicted also in the tables. The findings indicate that staying up to date with the newest literature, EBT training and evidence evaluation skills were perceived as the biggest challenges to the implementation of EBT practices among teacher educators, followed by lack of critical appraisal skills. Significant differences were identified between professors and teaching associates concerning the need for EBT training, where professors scored significantly higher than their counterparts. Significant differences were also identified between professors, PhD candidates, and school mentors in regard to the evaluation skills one needs to find the best available evidence.

Regarding challenges related to time constraints, significant differences were identified between professors, teaching associates, and school mentors. Professors reported feeling more time pressured than teaching associates and school mentors. The lack of evidence in the literature and the gap between research and practice were also perceived as challenges to EBT implementation.

As for the facilitators, teacher educators reported that access to literature and critical appraisal skills are equally important to the implementation of EBT. Teacher educators also reported that research experience, teamwork, and development of evidence-based databases can facilitate the use of EBT. Concerning teacher educators' research experience, significant differences were identified between professors and teaching associates, where professors scored significantly higher than their counterparts. Teacher educators did not agree with the idea that the more involved they were in the development of evidence-based databases (e.g., What Works Clearinghouse database) the more they implemented EBT practices.

Although important findings were outlined in this study, several limitations must be addressed in future research to better understand which factors should be the focus of EBT professional development efforts for the heterogeneous group of teacher educators. This correlational study only measured the mediating role of practical knowledge and self-efficacy beliefs in the interplay between teacher educators' research exposure and frequency of use of empirical evidence. Teacher knowledge can include pedagogical knowledge, content knowledge, and knowledge of the learner (Shulman, 1987). Beliefs can also include self-concept and self-esteem and can also be subject-specific (Pajares, 1992). Because the role of personal factors is scarcely investigated in the teacher educators' literature, researchers are encouraged to include other mediators in the model. Moreover, experimental designs are needed to further verify our first results. Future research focusing on experimental designs will be beneficial to test the hypotheses put forward in this study and to reduce the likelihood of confounding variables. Also, interview studies may prove beneficial to investigate in-depth possible factors that affect teacher educators research experience like commitment, self-regulation or motivation.

Additionally, we used self-reported measures, which are prone to social desirability biases, a risk that must be addressed with any form of subjective data collection (Desimone, 2009). However, in the instrument development, piloting, and data collection process, measures were taken to reduce social desirability bias (e.g., assurance of “no right or wrong answers” and absolute confidentiality; exclusion of items including polarizing phrasing). With that said, the presence of response biases cannot be excluded. However, and because no ceiling effects were reported, this consideration is rather limited given the high variation expressed in the standard deviation. Further validation of the new EBTS scale is recommended for future research. Future research is also required with respect to the complexities of personal domain variables. Examining teacher educator's personal domain variables toward EBT over time, for example in two time points during an academic year, may prove beneficial for teacher educators. Such an approach may help teacher educators to better understand how personal domain variables affect their practices. This can lead to improvements in teacher education curricula in the long run and to changes regarding their understanding of evidence-informed teaching (Levin et al., 2013).

A final consideration refers to the definition and measurement of practical knowledge. We recognize that practical knowledge has been defined and measured in various ways (Clandinin, 1985; Connelly et al., 1997; Elbaz, 2018). In this study, we defined practical knowledge as the amalgam of experiential knowledge, formal knowledge, and personal beliefs (Cochran-Smith and Lytle, 1999; Van Tartwijk et al., 2009). Based on this definition and similar studies in medical education, we developed the items of our scale. Because the role of personal factors is significant in the EBT professionalization of teacher educators, and because practical knowledge is a messy construct, future research may focus on the development of EBT instruments targeting personal factors such as practical knowledge.

Teacher educators belong to a heterogeneous group of teachers with various work duties, which involve either only teaching or teaching and research; thus, there is a variation on how research-exposed they are. Because of these differences, teacher educators may also experience differences in their practical knowledge and self-efficacy beliefs, differences that may affect the implementation of evidence in their teaching practice. In the interplay between self-efficacy beliefs and practical knowledge, self-efficacy plays a more important role concerning EBT implementation in practice than knowledge; the fact that teacher educators know how to apply EBT practices seems to be less significant to EBT implementation. Like teachers, teacher educators need to feel able to implement certain practices (in this case, EBT practices) in order to apply them in practice.

Moreover, because teacher educators come from different backgrounds, they have different needs for professional development. To design trainings tailored upon their individual needs, it is important to understand the challenges teacher educators face and the potential facilitators that could foster EBT use. University structures may consider investing in professional trainings for less research-exposed teacher educators to boost their self-efficacy. This measure, along with possible collaborations between high and low-experienced teacher educators, could further support teacher educators' EBT professional development and EBT implementation. Finally, as in medicine with the Cochrane library (Jadad and Haynes, 1998), successful application of research evidence to teacher education can be fostered with the support of evidence-based databases such as the clearinghouse databases in the USA and Europe.

Our research provides first insights on teacher educators' practical knowledge and self-efficacy beliefs regarding their evidence-informed professional development. Although we are not yet able to provide readymade solutions for the improvement of their EBT skills, we contribute to the scarce literature in this field by establishing a knowledge base for teacher educators' further professional development. We hope that this initial work will spark interest and stimulate further studies on teacher educators' competencies such as practical knowledge and self-efficacy beliefs which can contribute largely in teacher's instructional quality (e.g., Kunter et al., 2007).