Exploring the Relationship Between Attitudes of Preservice Primary Science Teachers Toward Integrated STEM Teaching and Their Adaptive Expertise in Science Teaching

The call for the implementation of integrated science, technology, engineering, and mathematics (iSTEM) teaching has been on the rise. This teaching approach helps students to develop innovative skills and meet twenty-first century challenges. However, teachers need to possess positive attitudes toward iSTEM teaching to implement it. Therefore, studying factors that influence these attitudes is important. So far, investigated factors have focused on teachers’ personal and professional experiences. Predisposing and enabling factors related to iSTEM teaching are yet to be investigated. One putative construct that includes predisposing and enabling factors is adaptive expertise (AE) which allows professionals to highly perform in new tasks. Here, we argue that AE in science teaching is linked to teachers’ iSTEM attitudes through its three dimensions: perceived relevance (PR), self-efficacy (SE), and anxiety (Anx). We utilized a cross-sectional survey method to verify the proposed relationship on fourth year preservice primary science teachers (n = 91). They completed two online surveys measuring AE in science teaching and iSTEM attitudes, respectively. Results from multivariate regression modeling indicate a significant positive effect of AE in science teaching was found significantly positive on all three dimensions of iSTEM attitudes. These findings draw teacher educators’ attention to the importance of developing adaptive expertise while preparing iSTEM teaching advocates.


Introduction
The call for integrated science, technology, engineering, and mathematics (iSTEM) education has lately been on the rise as a response to the growing twenty-first century challenges (Bybee, 2018). Climate change, cyber security, renewable energy, re-/emergence of infectious diseases, among others require STEM-related competencies to be addressed. Accordingly, these competencies were considered by many nations as a critical component of education (UNE-SCO, 2017). Some of these competencies are problem-solving skills, ability to design solutions, competence to work in multidisciplinary teams, and being capable of carrying out successful inquiries (Thibaut et al., 2018). However, teachers' attitudes toward iSTEM teaching are key determinants of whether they will prepare citizens for such competencies (Thibaut et al., 2019). For instance, teachers' perceived relevance of and capability for iSTEM teaching were shown to be positively correlated with classroom activities that boost STEM-related competencies (Thibaut et al., 2018).
Various factors may influence these attitudes. So far, research has focused on factors related to aspects of the school context and teachers' background characteristics such as educational level, years of experience, gender, and professional development (Han et al., 2015;Thibaut et al., 2019). Predisposing and enabling factors are yet to be explored. A well-established construct that includes both factors is adaptive expertise. Adaptive experts show willingness and ability to learn and innovate practices that meet the challenges of the time (Bohle Carbonell et al., 2016). In contrast, routine experts tend to stick to routines they developed over the years, and disregard new teaching approaches. Bowers et al. (2020) reported lower adaptiveness to science reforms by experienced teachers. Thibaut et al. (2019) found that the more the teaching experience, the less positive are the teachers' attitudes toward iSTEM teaching. Should they developed adaptive expertise, they would have probably demonstrated an innovative attitude toward iSTEM teaching. To investigate this hypothesis, we conducted a correlational study on fourth-year preservice science primary teachers. Although in-service teachers are crucial in promoting iSTEM teaching, we focus on pre-service science teachers for two reasons. First, attitude toward reformse.g. iSTEM teaching-tends to fluctuate with years of professional teaching experience (Thibaut et al., 2019). Unlike in-service teachers, PSTs are yet to be licensed before any professional teaching experience takes place. Second, inservice teachers' attitudes vary based on educational level (holding undergraduate/graduate degrees) (Clark et al., 2014). Hence, utilizing PSTs with the same educational level (i.e. working on their undergraduate degree) and with no years of professional teaching experience allows us to study the relationship in isolation from these confounding variables. PSTs are soon to become in service. So, teacher educators need to design research-based programs that feed iSTEM teaching advocates into the system. Overall, our research question is: does adaptive expertise in science teaching impact attitude toward iSTEM teaching of preservice primary science teachers?

Literature Review iSTEM Teaching
Literature embraces different approaches to iSTEM teaching in terms of teaching strategies. One group of researchers emphasizes integration strategies among STEM subjects, but at different levels. Vasquez (2014) described a continuum of integration levels that helps organize these strategies. At one extreme of the continuum, students learn the content and skills of STEM subjects in separate classes. The second focal point is thematic integration, where subjects are still taught separately but linked through one common theme or topic. The third focal point is about applying content and skills from different disciplines for the purpose of deepening understanding of integrated subjects. The fourth focal point is also about application, but for the purpose of solving real-world problems.
Another group of researchers underscore the utility of teaching strategies that develop twenty-first century skills such as critical thinking, creativity, communication, and collaboration. In their review, Breiner et al. (2012) emphasized inquiry-based and problem-based strategies. Along these, many researchers also push for design-based strategies (such as Guzey et al., 2016;Siverling et al., 2019). Others further call for collaborative-based strategies within design activities (e.g. McFadden & Roehrig, 2017;Wongta et al., 2021). Thibaut et al. (2018) conducted a review of papers describing multiple STEM teaching strategies. Interestingly, they noticed consistent emphasis on integration strategies along with strategies meant to develop twenty-first century skills. Accordingly, Thibaut et al. (2018) articulated a theoretical framework for iSTEM teaching that consists of five distinct but related teaching strategies: integration of STEM content (INT), problem-centered (PCL), inquiry-based (IBL), designbased (DBL), and cooperative (COL) learning. Integration of STEM content refers to the alignment of learning objectives and activities from different STEM disciplines (Bryan et al., 2015). Problem-centered learning entails that teachers involve their students in solving authentic problems to ensure deep conceptual understanding (Christensen et al., 2015). Inquiry-based learning indicates that teachers encourage students to pose and investigate questions, make sense of collected information, and generate new understandings (Diggs, 2009). In designbased learning, teachers assign students to build their own artifacts to help them devise solutions to a given problem (Ackerman, 1996). Finally, in cooperative learning, teachers design learning environments that encourage students to collaborate in the various learning activities to deepen their knowledge (Christensen et al., 2015).
This framework embraces a comprehensive set of iSTEM teaching strategies and offers a clear description of required teaching practices that teachers can conform to and implement (see appendix in Thibaut et al., 2018). However, for teachers to implement iSTEM teaching, a set of personal and environmental factors must be in place. Personal factors include knowledge, expectations, and attitudes toward a certain behavior (i.e. iSTEM teaching) (Henson, 2001). It also includes self-efficacy (thinking that they are capable) toward performing that behavior (Ajzen, 2005), and positive background experiences related to this behavior (Henson, 2001). Environmental factors include subjective norm (Ajzen, 2005), school context (Thibaut et al., 2019), and general teaching contexts (Zint, 2002). Cooper and Carr (2018) reported the saliency of some of these factors in STEM teaching. Pre-service teachers participating in their study reported positive social influence to teach STEM, low levels of self-efficacy toward certain STEM teaching strategies, and positive attitudes toward STEM teaching. The attitude factor is of special interest to us especially because, in another study (Thibaut et al., 2018), teachers' attitudes turned to be positively linked to actual iSTEM teaching practices rather than just their intentions to teach STEM (Cooper & Carr, 2018).

Teachers' Attitudes Toward iSTEM Teaching
Attitude can be defined as an affective evaluation of an attitude object-feelings toward iSTEM teaching (Fishbein & Ajzen, 1975). However, a cognitive component for attitude became well established afterwards such as a teacher's opinion about iSTEM teaching (Ajzen et al., 1980;Eagly & Chaiken, 1993). Many theorists add behavior as the third component to attitude (see Bergman, 1998). However, Ajzen and Fishbein (1980) argue that attitude is an antecedent of behavior and therefore, behavior does not fit in the theoretical construct of attitude. Even behavioral intention is rather a direct outcome of the affective and cognitive dimensions of attitude, and not a dimension of attitude itself (Ajzen, 2001). If one has positive feelings and opinions about a behavior, then one might intend to pursue this behavior. Hence, attitude is an antecedent of behavioral intention and behavioral intention is an antecedent of actual behavior. In his theory of planned behavior, Ajzen (1991) recognizes perceived control of behavior as a third dimension of attitude. The concept of perceived control entails an individual's perceptions of personal (e.g. self-efficacy) and environmental factors (e.g. school context) that may influence their behavioral intention (e.g. intention to teach iSTEM).
In short, the present analysis leads to a tripartite framework for attitude: affect, cognition, and perceived control of behavior. Van Aaalderen-Smeets et al. (2012) validated this framework for teachers' attitudes toward science teaching and found evidence for the three dimensions. However, most research on teachers' attitudes toward STEM teaching fall short of satisfying this framework. Cooper and Carr (2018) were concerned solely about how teachers felt about STEM teaching, entertaining the affective dimension. Mobley (2015) only measured self-efficacy, which belongs to the perceived control dimension. Wei and Maat (2020) considered a three-dimensional framework to attitude including cognition and affect. However, they considered behavior as the third dimension, which does not conform with the above analysis. Relying on Van Aalderen-Smeets et al. (2012) work, Thibaut et al. (2018) satisfied the tripartite framework with one subscale representing each dimension: feelings of anxiety a teacher may experience while teaching iSTEM represented the affective dimension, a teacher's perceived relevance and importance of iSTEM teaching reflected the cognitive dimension, and self-efficacy represented the S184 perceived control dimension. Based on this framework, Thibaut et al. (2018) proposed and validated a questionnaire measuring teachers' attitudes toward iSTEM teaching. Other than satisfying the tripartite framework, we adopt Thibaut et al.'s (2018) questionnaire in our study because it identifies multiple teaching strategies for iSTEM instead of reducing it to one. Their questionnaire also offers a clear description of required teaching practices for each strategy making it easier for a teacher to develop an informed attitude toward iSTEM teaching. Informed attitudes tend to be more stable and consistent with subsequent behavior; hence increasing the possibility that PSTs (reporting positive attitudes) will implement iSTEM teaching (Fabrigar et al., 2006).

Factors Related to Attitude Toward iSTEM Teaching
Knowledge about factors influencing teachers' attitudes toward iSTEM teaching is useful when attempting to improve these attitudes. However, there is a paucity of studies which investigated these factors. Al Salami et al. (2017) explored if change in attitudes toward interdisciplinary teaching and teamwork through professional development would differ across teachers' gender, school level, discipline taught, and educational level. Variations in initial attitudes as well as changes in attitudes after the workshop were observed across the abovementioned factors reflecting their potential role in shaping teachers' attitudes. However, this work only considered two iSTEM teaching strategies (integration and teamwork). Further, Al Salami et al. (2017) did not provide a conceptual model for the attitude construct. Additionally, environmental factors, that might have influenced teachers' attitudes, were excluded. Thibaut et al. (2017) later addressed these three issues through employing the tripartite attitude construct on all the five iSTEM teaching strategies. They also considered the school context as a putative environmental factor. Results of multiple regression analysis revealed five factors that are positively correlated to teachers' attitudes: participation in professional development, supportive school social context, personal relevance of science, and personal relevance of technology. The same analysis also revealed two factors that are negatively linked to teachers' attitudes toward iSTEM teaching: teaching experience in mathematics and having over twenty years of teaching experience.
One insightful observation emerges from these two studies. In Al Salami et al.'s (2017), teachers who taught science initially reported the highest attitudes toward interdisciplinary teaching and teamwork compared to those who taught mathematics or technology/engineering. In Thibaut et al.'s (2017), personal relevance of science was positively linked to attitudes toward iSTEM teaching. Therefore, it seems that expertise in science teaching might be positively related to teachers' attitude toward iSTEM teaching. But there was a negative correlation between years of teaching experience and attitudes toward iSTEM teaching in Thibaut et al.'s (2017). The more the expertise, the less positive the attitudes. In other words, teachers' instructional choices demonstrated a routine type of expertise they developed over the years disregarding the new iSTEM teaching approach. Should they exhibit an adaptive type of expertise, they would have probably shown an innovative attitude toward iSTEM teaching. Prior research described teachers with adaptive expertise as reform-oriented (Stylianides & Stylianides, 2014). Thus, we hypothesize a positive correlation between adaptive expertise in science teaching and attitude toward iSTEM teaching. The next section explains our theoretical conceptualization of this relationship.

Adaptive Expertise
Adaptive expertise was first conceptualized by Hatano and Inagaki (1984). Adaptive expertise allows professionals to highly perform in new tasks, setting it apart from routine expertise. Routine experts exhibit efficiency in performing tasks that are previously experienced. Adaptive experts not only exhibit efficiency toward experienced tasks, but also exhibit innovation when confronted with non-experienced ones. While routine expertise allows for efficiency, adaptive expertise allows for efficiency and innovation. Bohle Carbonell et al. (2016) proposed a framework for adaptive expertise comprising the two dimensions: efficiency and innovation. The former emphasizes acquisition of domain knowledge as prerequisites for efficiency. It also focuses on epistemology as a predisposing factor. The epistemic stance an adaptive expert takes predisposes them for a positive attitude toward continuous acquisition of domain knowledge. This predisposition is one distinction between adaptive and routine experts. The innovation dimension underscores cognitive flexibility as an enabling factor for innovation. With cognitive flexibility, adaptive experts can decontextualize acquired knowledge through dismantling and reassembling it to serve new contexts. This ability is another factor that distinguishes adaptive experts from their counterparts. The inclusion of both predisposing and enabling factors, made Bohle Carbonell et al.'s (2016) framework a good fit for cross-sectional survey studies measuring personal factors associated with adaptive expertise at a given point in time; hence its adoption in this study.
Note that in our work, the term "expertise" is not used in the traditional sense of professional experience (e.g. teaching experience) as in novice versus expert. Rather, "expertise" is utilized here as a type of professionalism (Anthony et al., 2015). If an individual's expertise allows them to highly perform when confronted with novel tasks and methods at work, then this type of expertise is adaptive (Carbonell et al., 2014). Otherwise, it is considered as routine. In this sense, several researchers studied the development of adaptive expertise in teacher preparation programs using preservice teachers (Anthony et al., 2015;De Arment et al., 2013;Mason-Williams et al., 2015).

Attitude and Adaptive Expertise
Several researchers spoke of a relationship between attitude and adaptive expertise (Martin et al., 2015;Schwartz et al., 2005;Walker et al., 2006). However, there were theoretical concerns. For instance, Martin et al. (2015) did not consider attitude as a distinct construct from adaptive expertise. To gauge adaptive expertise, they simply sorted statements of a design survey into two categories, those indicating efficiency attitudes (e.g. "Good designers get it right the first time") or innovative attitudes (e.g. "Creativity is integral to design"). Schwartz et al. (2005) named confidence as an attitudinal construct related to adaptive expertise. However, Bandura et al. (1999) argue that confidence does not specify an individual's certainty about performing a new task. It only reflects the strength of their belief. Confidence, then, is rather a colloquial term than a construct. Walker et al.'s (2006) work provides support for Bandura et al.'s (1999) argument. They found that only confidence coupled with high capability was positively linked to innovation. This said, we saw merit in embedding the relationship between attitude and adaptive expertise in a theoretical system aiming to further understanding of this relationship, and eventually informing daily practice in a systematic manner.
In this paper, we introduce three propositions to frame the relationship. We support each proposition with empirical evidence from the literature. Then, we employ two psychological theories to explain this evidence. Our first proposition is that selfefficacy, rather than confidence, is an attitudinal variable related to adaptive expertise. Self-efficacy is the degree to which an individual believes they can successfully perform a specific task. Not only does self-efficacy reflect the strength of a belief about performing a task (i.e. confidence), it also is an affirmation of an expert's capability level (Bandura et al., 1999). Several empirical studies showed positive correlations between self-efficacy and adaptive expertise (Bell & Kozlowski, 2008;Stokes et al., 2010).
Our second proposition is that perceiving relevance (e.g. positive impact) in performing a specific task is another attitudinal variable related to adaptive expertise. Literature showed that teachers with adaptive expertise commit to continuous improvement of their teaching approaches and pay constant attention to the impact of their teaching practices (Bowers et al., 2020). The social cognitive theory provides theoretical ground for the first two propositions (Bandura et al., 1999). The theory states that there are two key determinants to perform a task: outcome expectancy and self-efficacy. Outcome expectancy refers to the perceived relevance and consequences of performing the task, such as perceived impact of iSTEM teaching on student learning. Self-efficacy is the degree to which an individual feels capable of performing the task, such as feeling capable of teaching iSTEM. Hence, if teachers expect positive outcome(s) of iSTEM teaching and feel self-efficacious about it, then they are more likely to demonstrate adaptiveness toward iSTEM teaching.
Finally, we propose anxiety as a third attitudinal variable related to adaptive expertise. In a recent study (Smith et al., 2021), a group of ten teachers expressed high anxiety toward teaching the design process prior to a 15-week course in which they explored the engineering design process in different contexts. However, after recurring experiences with the design process, teachers reported a significant 46.9% reduction in their anxiety toward design coupled with high expectations to succeed in teaching the design process. The flow model explains this phenomenon (Csikszentmihalyi, 2000). Flow is a psychological concept that signifies enjoyment and complete involvement in a task as long as the level of challenge posed by the task matches the current set of skills. If mismatched, then either anxiety (high challenge; low skills) or relaxation (low challenge; high skills) will result. Rebello and Zollman (2013) saw an overlap between the flow model and expertise. For instance, anxiety corresponds to the confusion of a novice. Similarly, relaxation coincides with boredom of an expert. Hence, the presented case demonstrates an initial imbalance between challenge (high) and skills (low) resulting in anxiety. In terms of expertise, it demonstrates an initial imbalance between innovation (high) and efficiency (low) that was better matched after 15 weeks of training.
Based on this analysis, we conceptualize adaptive expertise to be related to all three attitudinal variables in the tripartite attitude framework: perceived relevance, self-efficacy, and anxiety. Tied to the abovementioned observations about the positive links between science teaching and interdisciplinary/iSTEM teaching from Al Salami et al.'s (2017) and Thibaut et al.'s (2017) work, we propose that adaptive expertise in science teaching is a potential factor that is positively related to teachers' attitudes toward iSTEM teaching. Figure 1 illustrates our theoretical framework of the relationship.

Participants
Ninety-one PSTs in their fourth year of the primary science teaching program participated in this study. Fourth-year PSTs were purposely sampled to ensure minimal domain expertise. By the fourth year, they have already had three teaching practice placements in public schools as an essential part of their teacher training program. In these placements, PSTs spend 60 days planning and teaching lessons, and observing in-service teachers. They reflect on their own teaching and lesson plans, and receive continuous feedback from cooperative in-service teachers. Additionally, they hold seminars with their peers to discuss and reflect on video clips of their own teaching. Aside from teaching practice, PSTs worked on various science teaching projects in the five science methods courses required by the program. They plan activities that integrate math and science content and skills, evaluate educational tools that integrate technology into science teaching, plan for and practice inquiry-based science teaching, design collaborative learning activities, develop project-based Fig. 1 Proposed theoretical framework for the relationship between adaptive expertise in science teaching and iSTEM teaching attitudes S188 science learning units, and explore resources for problem-solving activities, among other projects. Hence, working on these projects, fourth-year PSTs must have developed minimal expertise in science teaching that is respective of iSTEM teaching strategies such as integration, inquiry-based, collaborative-based, and problem-centered learning.

Data Collection
PSTs completed two online surveys (adaptive expertise and iSTEM attitude surveys) prepared through Google forms, which took 20 min of class time. We chose the cross-sectional correlational survey method to infer correlations that consequently encourage in-depth investigations (Brewer, 2009). To increase internal validity, we considered three variables that might confound the relationship between adaptive expertise and iSTEM attitude: gender, high school (HS) major, and professional development (PD) participation. Gender was considered because female teachers reported significantly higher attitudes toward the collaborative learning strategy in iSTEM teaching (Thibaut et al., 2019). However, their general attitudes toward STEM education did not differ from male teachers (Wei & Maat, 2020). HS major (science vs. other) was regarded as teachers who took more science courses reported more comfort and less pedagogical discontentment with teaching STEM (Nadelson et al., 2012). PD participation was examined as it was positively linked to teachers' attitudes toward iSTEM teaching (Thibaut et al., 2017).

(i) Adaptive Expertise Survey
To identify their type of expertise in science teaching, PSTs completed the 10-item survey developed and validated by Bohle Carbonell et al. (2016). Bohle Carbonell et al. (2016) stipulated that an instrument measuring adaptive expertise (AE) should be specific to the work domain. Accordingly, items were slightly adapted to focus on the science teaching domain. All ten items were formulated as, "During past science teaching projects, I …" The efficiency and innovation dimensions were each represented by 5 items (Appendix). Participants reported their level of agreement with each item on a 5-point Likert scale (1 = totally disagree; 5 = totally agree). To examine the validity of our adapted survey, we used principal component analysis with oblique rotation. The suitability of the data for exploratory factor analysis (EFA) was satisfied with the Kaiser-Myer-Olkin (KMO) measure of sampling adequacy (KMO ≥ 0.6) and the Bartlett test of sphericity (p < 0.05) (Tabachnick & Fidell, 2018). Appendix reports the structure for the adaptive expertise survey comprising 10 items with 2 factors (efficiency and innovation). Except for item_5, the pattern matrix in Appendix reveals a two-factor solution. This indicates that adapted items are appropriately loading in their corresponding factors, efficiency or innovation. Item_5 was meant to be an efficiency item, "During past science teaching projects, I was able to develop and integrate new knowledge with what I learned in earlier science teaching classes." However, it loaded strongly (0.73) and inappropriately onto the innovation factor, and barely loaded (0.07) on the efficiency factor. It seems that PSTs interpreted the phrase "develop and integrate new knowledge" as a sign for being innovative. Item_5 was therefore excluded from any further analysis. The reliability scores of PSTs' responses were acceptable for the 9-item survey (α = 0.733) and its dimensions; see Table 1 (Kline, 2000).

(ii) iSTEM Attitudes Survey
To express their attitudes toward iSTEM teaching, PSTs took the 75-item survey developed and validated by Thibaut et al. (2018) based on the adopted frameworks for teachers' attitudes and iSTEM teaching. Respondents reported their level of agreement with each item on a 5-point Likert scale. Attitude items in the Perceived Relevance (PR) subscale started with "How important do you think it is to …," anxiety (Anx) subscale with "How anxious do you feel to do this?" and self-efficacy (SE) with "How capable do you feel to do this?" Likert scale labels for the PR subscale were (1 = not important; 5 = very important), Anx (1 = very anxious; 5 = not anxious), and SE (1 = not capable at all; 5 = very capable). Sample items for each of the five aspects of iSTEM teaching are (i) integration, "linking mathematical, scientific, and/or technological concepts;" (ii) problem-centered, "working on authentic/real-world problems with students;" (iii) inquiry-based, "students find an answer to research questions;" (iv) design-based, "students justify their design choices;" and (v) collaborative learning, "students are guided throughout teamwork." Reliability scores of PSTs' responses were good to excellent for the 75-item survey (α = 0.937) and its dimensions (see Table 1). Adjoining was a one item surveying gender (1 = male; 2 = female), another item for HS major (1 = science; 2 = other), and another surveying the number of attended teaching PD sessions (1 = none-5 = four sessions, 6 = More than four sessions).

(iii) Validity of Instruments
Construct validity of both instruments was examined by confirmatory factor analysis (CFA) using AMOS version 25. Table 1 summarizes CFA results. The fit of models was evaluated by Chi-square statistics and fit indices. Chi-square statistics ( 2 df ) values < 2 for both instruments indicating a sufficient model fit (Marsh et al., 1996). Fit indices included the Comparative Fit Index (CFI: Bentler, 1990) and Tucker Lewis Index (TLI: Bentler & Bonett, 1980). Both CFI and TLI range from 0 (indicating poor fit) to 1 (indicating a perfect fit). CFI and TLI values for both instruments ranged between 0.91 and 0.94, which indicate a psychometrically accepted fit to the data (Hu & Bentler, 1999). Root mean square error of approximation (RMSEA) is one of absolute fit indices and a measure of discrepancy between the observed and model implied covariance matrices adjusted for degree of freedom. RMSEA value was 0.060 for the AE ST instrument indicating a reasonable fit, and 0.073 for the attitude instrument indicating an acceptable fit (Browne & Cudeck, 1993). Another fit index is the standardized root mean square residual (SRMR). SRMR values of both instruments were < 0.08 indicating an acceptable-fitting model (Byrne, 2016). Furthermore, we assessed convergent validity of subscale items through examining the composite reliability (CR) of each subscale and the average variance extracted (AVE) by its items. CR values for all subscales were above 0.70 reflecting adequate reliability (Hair et al., 2010). AVE values were all above 0.50, suggesting that more than half of the variance observed in the items was accounted for by their corresponding subscales (Hair et al., 2010). Overall, the figures of both CFA models shown in Table 1 indicate the validity of scores given by both instruments, adaptive expertise in science teaching and teachers' attitudes toward iSTEM teaching.

Data Analysis
Using SPSS 26.0, we ran a paired-sample t-test to compare the means of the two adaptive expertise subscales (efficiency and innovation). We also ran a one-way repeated measures ANOVA to compare the means of the three attitude subscales, followed by a post hoc test (LSD). These two tests helped better shape our understanding of PSTs' adaptive expertise and attitudes based on their constituting subscales. Since the sphericity assumption for the repeated measures ANOVA was violated, the Greenhouse-Geisser correction was used to report the F-test as the lower-bound epsilon value was below 0.75. A set of bivariate correlations were run for preliminary examination of the relationships among the variables: gender, HS major, PD, AE ST , attitude, PR, SE, and Anx (see Table 1 for a definition of the acronyms). Spearman's rank correlation was used because some variables are not continuous. A simple linear model regressing AE ST on iSTEM attitude was run to examine its share in explaining the variance in teachers' attitudes. This was followed by a multivariate model simultaneously regressing AE ST on the three attitude subscales PR, SE, and Anx. This step was done to test our hypothesis that AE ST is a potential factor positively related to iSTEM attitudes through the tripartite framework. Utilizing the multivariate model reduced the risk of committing Type I error, which  (Field, 2009). The independence of observations assumption is validated through examining variance inflation factor (VIF) values. The normality assumption for each score (attitude, PR, SE, and Anx) is examined through the Shapiro-Wilk test. The homoscedasticity assumption for the regression analyses is validated based on the Breusch-Pagan test. Table 2 shows statistics of participants and their survey responses. The 91 PSTs were mostly females (88%) and high school science majors (88%). Only 47% of the sample attended one or more PD sessions with an overall average of two sessions attended. Efficiency scores (M ± SE = 4.30 ± 0.05) indicate PSTs' agreement that they had a positive stance toward continuous acquisition of domain knowledge in science teaching. Innovation scores (M ± SE = 3.70 ± 0.06) reflect PSTs' agreement that they were able to successfully apply their science teaching knowledge in novel situations. However, the difference in means between both scores, meant that PSTs felt more efficient than innovative, t(90) = 9.248, p < 0.001. This can mean that PSTs felt more comfortable acquiring new domain knowledge than taking the risk of applying their knowledge in new situations within the context of graded projects. Alternatively, it can also mean that PSTs needed more opportunities to apply their knowledge in novel situations. Either way, overall survey scores (M ± SE; 4.00 ± 0.04) indicate PSTs' agreement that they demonstrated adaptive expertise in science teaching as they worked on their science teaching projects. iSTEM attitude survey scores reveal positive attitudes toward iSTEM teaching

S194
(M ± SE; 4.09 ± 0.04). As per Likert subscale labels, computed means indicated that PSTs perceived iSTEM teaching as "important to an extent" (4.39 ± 0.03), felt "capable to an extent" of teaching iSTEM (4.02 ± 0.05), and were "anxious to an extent" toward iSTEM teaching (3.86 ± 0.06). The differences among the means of the three variables were significant, F(1.63,146.21) = 57.898, p < 0.001. Post hoc analysis showed that PR scores were the highest followed by SE and Anx respectively, p < 0.001. This difference indicates that PSTs' reported "extent" of importance, capability, and anxiety is not the same across the three subscales. Perceived importance of iSTEM teaching is relatively higher than felt capabilities. In turn, PSTs' levels of anxiety were higher than both, perceived importance and capability (lower Anx score means higher anxiety). Table 3 shows significance of correlations (if any) among study variables. Gender, high school major, and number of attended PD sessions were not correlated with any studied variable. The lack of correlation between these variables and iSTEM attitudes resulted in excluding them as potential confounding variables in subsequent regression analyses.

Correlations Among Study Variables
There was a strong positive correlation between AE ST and iSTEM attitudes, which lends support to our hypothesis (r s = 0.521, p < 0.01). Furthermore, AE ST was positively linked to all three attitude subscales, PR (r s = 0.502, p < 0.05), SE (r s = 0.407, p < 0.01), and Anx (r s = 0.399, p < 0.05). These findings collectively and individually agree with our theoretical analysis that adaptive expertise in science teaching is positively related to iSTEM teaching attitude through its three dimensions, perceived relevance, self-efficacy, and anxiety.

Main Findings
In our theoretical framework, we also proposed that adaptive expertise in science teaching is a potential factor that is positively related to teachers' attitudes toward iSTEM teaching. The following regression analyses supported this proposal. All assumptions for multivariate regression analysis were met. First, the correlations shown in Table 2 between SE and PR (r s = 0.325, p < 0.05) as well as SE and Anx (r s = 0.745, p < 0.01) do not raise concerns F(1,90) = 0.81, p = 0.37). The latter assumption being met increased the precision of the confidence intervals (CI) for the regression weight estimates displayed in Table 4 (Hayes, 2022). Table 4 shows parameters for two regression models. The first column summarizes the output of a simple linear model regressing AE ST on iSTEM attitude. The remaining columns summarize the output of a multivariate model simultaneously regressing AE ST on the three attitude subscales PR, SE, and Anx. As displayed in Table 4, the effect of AE ST is significantly positive on iSTEM attitude and each of its three dimensions: iSTEM attitude (b = 0.537, p < 0.001), PR (b = 0.338, p < 0.001), SE (b = 0.486, p < 0.001), and Anx (b = 0.537, p < 0.001). In further support, none of the corresponding 95% CI straddle "zero" (no effect) as a possible value of the true effects of AE ST on all four variables. Along with bivariate correlation analyses, this body of evidence supports the hypothesis that PSTs' adaptive expertise in science teaching is positively correlated with their attitudes toward iSTEM teaching. It also supports our argument that this correlation come to pass through the three attitudinal dimensions, perceived relevance, self-efficacy, and anxiety.
To put these findings into perspective, AE ST standardized regression coefficient (b = 0.537, p < 0.001) tells us that one unit increase in adaptive expertise in science teaching translates into good 0.537 units increase in positive attitude toward iSTEM teaching. In turn, adjusted R 2 reveals that 28% of the variance in PSTs' attitudes toward iSTEM teaching can be explained by their adaptive expertise in science teaching. These figures suggest that adaptive expertise can be an important factor impacting PSTs' attitudes.

Discussion and Conclusion iSTEM Attitudes
Primary pre-service teachers in our study reported positive attitudes toward iSTEM teaching. This is a promising indicator for whether they are willing to prepare citizens for STEM-related competencies (Appleton, 2003;Stains & Vickrey, 2017;Thibaut et al., 2018). Scores of the S196 three attitude variables followed a linear trend with perceived relevance being the highest and anxiety being the lowest (the lower anxiety score the higher the anxiety). The same trend recured with in-service teachers in Thibaut et al.'s (2018). The authors did not provide an explanation in this regard. However, the relatively high anxiety did not seem to deter those teachers from pursuing iSTEM teaching, because the sample was exclusive for those who reported being involved in teaching iSTEM. Accordingly, we perceive this trend as less of a threat, especially that our PSTs' attitudes seem relatively higher (mean ± sd = 4.09 ± 0.04 vs. 3.43 ± 0.44) with a large effect size, d = 2.11 (Cohen, 1988). Subsequent qualitative studies might better explain this recurring trend and its nuance consequences.

iSTEM Attitudes and Background Characteristics
Gender, high school major, and number of attended PD sessions were not correlated with PSTs' attitudes. For gender, this aligns with Wei and Maat's report (2020). However, other research showed differences by gender. For instance, male teachers reported higher self-efficacy toward iSTEM teaching (Lee et al., 2019). Female teachers' attitudes were higher toward cooperative strategies in iSTEM (Thibaut et al., 2019). This discrepancy might be attributed to the fact that samples are not comparable. Our PSTs and Wei and Maat's (2020) teachers have yet to teach iSTEM, while those in Thibaut et al. (2019) were already teaching it. Wei and Maat's (2020) were all primary mathematics teachers while those in Lee et al.'s (2019) were secondary science, technology, or mathematics teachers. Male teachers were underrepresented (11%) in Wei and Maat's (2020) and our sample (12.1%), while they constituted more than half the sample in Lee et al.'s (2019) and Thibaut et al.'s (2019) (58% and 53%, respectively). Hence, further research is needed in this regard. Lack of correlation with high school major came contrary to expectations (Nadelson et al., 2012). We attribute this finding to the science courses taken in our B.Ed. program, which might have bridged the gap between science and non-science high school majors. As stated earlier, college science courses made teachers feel more comfortable in teaching STEM (Nadelson et al., 2012). The absence of correlation with PD participation also came in contrary to other studies (e.g. Aldemir & Kermani, 2017;Thibaut et al., 2017). Differences in the number of attended PD sessions may explain this inconsistency. For instance, in Aldemir's and Kermani's (2017) study, all teachers attended three sessions, whereas most of our PSTs (71.4%) either never attended (52.7%) or participated in a single PD session (18.7%).

iSTEM Attitudes and Adaptive Expertise in Science Teaching
Our PSTs reported demonstrating adaptive expertise in science teaching as they worked on their science teaching projects. Possessing adaptive expertise is considered a "gold standard" for becoming a teaching professional (Hammerness et al., 2005, p. 360). Correlation and regression analyses harmoniously answered our research question on the relationship between PSTs' adaptive expertise in science teaching and attitude toward iSTEM teaching. Adaptive expertise in science teaching was positively correlated with iSTEM attitudes, and it explained 28% of the variance in PSTs' attitudes. This suggests that adaptive expertise in science teaching stands out among factors affecting teachers' attitudes toward iSTEM teaching. Explained variance by other factors, such as personal background characteristics and school context, was only moderate to low (Thibaut et al., 2019).
To support our argument, the correlation between adaptive expertise in science teaching and teachers' iSTEM attitudes came to pass through the three attitudinal dimensions, perceived relevance, self-efficacy, and anxiety. To our knowledge, this is the first effort to conceptually bridge research between these two constructs. The added value of our theoretical framework lies in explaining this correlation: Working on the various projects, fourth-year PSTs developed expertise in science teaching that is respective of iSTEM teaching strategies (e.g. integration, inquiry-based, collaborative-based, and problem-based learning). In other words, these projects equipped PSTs with efficiency and innovative skills in teaching strategies related to iSTEM. The type of expertise that PSTs developed was adaptive in a sense that it made them willing (epistemic predisposition) and able to learn (cognitive flexibility) the iSTEM teaching practices described in the attitude questionnaire. In line, PSTs perceived relevance of, felt capable to an extent, and were anxious to an extent about iSTEM teaching.

Limitations and Perspectives for Further Research
Our findings contribute to the literature by introducing a significant factor influencing teachers' attitudes toward iSTEM teaching, adaptive expertise in science teaching. For generalization however, these findings need to be replicated with other pre-service and in-service STEM teachers in different contexts, education levels (primary and secondary), and career stages.
We perceived three limitations to this work. Both measures of adaptive expertise and iSTEM attitudes are self-reported. This may have encouraged socially desirable responses resulting in an artifactual covariation between the two measures. Hence, further research can benefit from other measures such as in-depth interviews about teachers' iSTEM attitudes and examination of S198 submitted projects to fetch indicators of alleged adaptive expertise. Another limitation is the consideration of iSTEM attitudes in general. Further research may employ a more detailed method to examine attitudes toward each of the five iSTEM strategies and their relations to adaptive expertise. Such research can provide insights into individual barriers to the implementation of iSTEM. However, a larger dataset is needed to conduct such analysis. A third limitation is the exclusion of environmental factors that might have influenced adaptive expertise and/or iSTEM attitudes. We did this on purpose to study the relationship between iSTEM attitudes and personal factors associated with adaptive expertise at a given point in time (epistemic stance and cognitive flexibility). Subsequent studies may include environmental factors to further our understanding of the phenomenon. For instance, providing PSTs with a safe environment for collaborative discourse as they work on their projects can be an enabling factor for developing adaptive expertise (Bowers et al., 2020). Considering the school contexts PSTs went to during their teaching practice may also be an environmental factor influencing their attitudes toward iSTEM (Thibaut et al., 2019).

Implications
These findings offer practical implications for teacher preparation programs. To improve iSTEM attitudes, projects can be designed to increase adaptive expertise in science teaching. One promising design is the STAR. Legacy Cycle in which PSTs can be challenged to solve a problem in small groups and present their solutions before the class (Schwartz et al., 1999). For instance, PSTs can be required to collaboratively construct a teaching aid addressing a common misconception, present it to their peers for feedback, and self-reflect on ways to improve their designs. As such they have opportunities to innovate and be efficient, i.e. develop adaptive expertise in science teaching. Self-reflections promote self-assessment and further innovation (Janssen et al., 2008). Peer feedback triggers reconsidering initial ideas allowing for further efficiency in the offered solutions (Soslau, 2012). STAR.Legacy course designs proved to increase adaptive expertise in fields, such as bioengineering and biomedical ethics (Martin et al., 2005;Pandy et al., 2004). A variant of the STAR.Legacy cycle (design-based approach) also showed to increase innovation with engineering teachers (Martin et al., 2015). Hence, we recommend testing the STAR.Legacy course design with science teachers. With a remarkably changing world, development of adaptive expertise should be a program outcome that enables science teachers to respond to the uncertainties associated with curricular reforms such as iSTEM teaching (Hammerness et al., 2005).
AppendixTable 5 Table 5 Factor loadings, percentage of variance and eigenvalues for the modified adaptive expertise survey