Vous êtes sur la page 1sur 13

Contents lists available at ScienceDirect Learning and Individual Di ff erences journal homepage:

Contents lists available at ScienceDirect

Learning and Individual Di erences

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

erences journal homepage: www.elsevier.com/locate/lindif Constructivism and personal epistemology development in

Constructivism and personal epistemology development in undergraduate chemistry students

Michael M. Barger a , , Tony Perez b , Dorian A. Canelas c , Lisa Linnenbrink-Garcia d

b , Dorian A. Canelas c , Lisa Linnenbrink-Garcia d a University of Illinois at Urbana-Champaign,

a University of Illinois at Urbana-Champaign, Department of Psychology, 603 E. Daniel St., Champaign, IL 61820, United States

b Old Dominion, Department of Educational Foundations & Leadership, United States

c Duke University, Department of Chemistry, United States

d Michigan State University, Department of Counseling, Educational Psychology and Special Education, United States

ARTICLE INFO

Keywords:

Personal epistemology

Development

Constructivism

Epistemic climate

Interactive learning

ABSTRACT

Students' beliefs about knowledge, or their personal epistemologies, are critical components of the learning process. Researchers and educators need to understand how portrayals of knowledge in the classroom shape personal epistemology development. Using a quasi-experimental design, an organic chemistry instructor taught a traditional, lecture-based course and a constructivist-based interactive-learning course. Students ( N = 270) completed three surveys assessing personal epistemology and perceptions of constructivism in the classroom. Although the interactive-learning classroom did not seem to aect personal epistemology, evidence suggested that perceptions of a complex learning environment predicted changes in personal epistemology. Students' initial epistemic beliefs predicted how they perceived the classroom environment. Results also supported an epistemic alignment hypothesis: students performed better on the nal exam when their beliefs matched the course structure. Findings support an interactive model between students' personal epistemologies and epistemic cli- mate and highlight the challenges of changing beliefs through single-semester classroom interventions.

Knowledge construction is a central component of education. Through the process of building knowledge, each learner develops be- liefs about what knowledge is and how it is justi ed, a system of cog- nitions known as personal epistemology ( Hofer & Pintrich, 2002 ). Stu- dents' specic beliefs about knowledge, also referred to as epistemic beliefs, predict their self-regulation strategies (Bråten, Anmarkrud, Brandmo, & Strømsø, 2014 ; Muis, 2007 ; Muis & Franco, 2009 ) and are also closely interrelated with student motivation ( Chen & Barger, 2016 ; Bråten & Strømsø, 2004). Students' personal epistemologies also predict achievement in various contexts ( Bråten & Ferguson, 2014 ; Dai & Cromley, 2014 ; Muis, 2004 ; Trautwein & Lüdtke, 2007 ). College chemistry classrooms are one such context. For example, under- graduates who have chemistry-speci c epistemic beliefs (e.g., In chemistry truth is unchanging. ) that closely match their preferred general epistemic beliefs (e.g., I prefer to study subjects where truth is unchanging. ) tend to receive higher grades in chemistry ( Dai & Cromley, 2014 ). Given the importance of students' epistemic beliefs to the learning process, it is essential to understand the situational factors that might in uence changes in personal epistemology (e.g., Feucht, 2010; Muis & Du y, 2013 ). Researchers and educators have debated the virtues of

two types of college pedagogical practices: traditional, lecture-based classrooms and constructivist, active-learning-based classrooms ( Freeman et al., 2014 ; Friesen, 2011 ). Many researchers and educators have taken a strong stance on which teaching approach is more ap- propriate in STEM elds based on students' learning and achievement, but few of these researchers have considered the role of students' epistemic beliefs in this debate. Assessing epistemic beliefs can com- plement existing research on the e ectiveness of pedagogical techni- ques as an explanatory mechanism or describe how the classroom for- mats might a ect underlying beliefs that students can carry beyond the course. The current study examines the complex relations between students' perceptions of constructivist pedagogical practices and per- sonal epistemology development.

1. Theoretical background

1.1. Personal epistemology

Although there are many ways to conceptualize personal episte- mology (Alexander, 2016 ; Chinn, Buckland, & Samarapungavan, 2011; Hofer & Pintrich, 1997 ), one prominent conceptualization examines

An earlier version of this manuscript was presented at the Annual Conference of the American Educational Researchers Association in Chicago, April 2015. Corresponding author. E-mail address: mmbarger@illinois.edu (M.M. Barger).

Received 27 April 2017; Received in revised form 15 March 2018; Accepted 29 March 2018

M.M. Barger et al.

independent beliefs about the nature of knowledge (its structure) and how knowledge is justi ed (its sources). Philosophers often de ne knowledge as justi ed, true beliefs ( Greene, Azevedo, & Torney-Purta, 2008 ), but the focus in personal epistemology is on how individuals think about knowledge themselves. Students generally hold these be- liefs implicitly; they may not think about these beliefs until they are explicitly asked to articulate them. Students' beliefs are both domain- general (e.g., all knowledge is unchanging, ) and domain-speci c (e.g., chemistry knowledge is unchanging; Muis, Bendixen, & Haerle, 2006 ). To some extent, these di erent levels of beliefs relate in a hierarchical manner, such that someone who has a domain-general belief is more likely to hold a similar domain-speci c belief; however, individuals can, and often do, hold di erent beliefs across domains, such as chemistry and history (for a review, see Barger & Linnenbrink- Garcia, 2017 ; Muis et al., 2006 ). Three beliefs particularly relevant to the debate between lecture- based and active-learning-based classrooms have also received con- siderable attention by researchers studying epistemic beliefs more broadly (see Greene et al., 2008 ; Hofer, 2000 ; Hofer & Pintrich, 1997 ):

1) simple/certain knowledge, 2) justi cation by authority, and 3) per- sonal justi cation. Simple / certain knowledge involves the structure of knowledge, and indicates the belief that knowledge is composed of unchanging, unrelated facts, instead of complex, evolving information. The belief that chemistry knowledge involves a collection of un- changing facts that need to be memorized (e.g., equations and symbols that are learned through memorization) exempli es simple/certain knowledge. On the other hand, the belief that scientists' understanding of molecular structures and energetic processes constantly evolves would not. Both justi cation by authority and personal justi cation involve sources of knowledge. A belief in justi cation by authority im- plies that knowledge is handed down by authority gures. A belief in personal justi cation means that individuals construct knowledge for themselves and therefore can hold di erent understandings of what is true. Students with strong justi cation by authority beliefs would view chemistry teachers and textbooks as the primary sources of knowledge, whereas students with strong personal justi cation beliefs think that people might come to di erent understandings by approaching pro- blems from di erent perspectives. Theoretically, di erent types of beliefs are considered adaptive or maladaptive for learning and achievement (e.g., Muis, 2004 ). The belief that knowledge is simple/certain, for example, has been described as less availing, as it has been linked to lower achievement and learning ( Qian & Alvermann, 1995 ; Trautwein & Lüdtke, 2007 ). This connection has been attributed to students using shallower cognitive processing when they believe knowledge is simple/certain (Muis, 2004 ; Strømsø & Kammerer, 2016 ). The evidence regarding the relation between beliefs about sources and process of knowledge justi cation and achievement is less consistent across contexts (e.g., Bråten & Ferguson, 2014 ; Bråten, Ferguson, Strømsø, & Anmarkrud, 2013 ; Conley, Pintrich, Vekiri, & Harrison, 2004 ). In introductory science courses, it is especially im- portant to understand that experts have agreed upon trustworthy, foundational principles and that not every opinion in science is valid (particularly if that opinion is not based on reliable scienti c pro- cesses). However, over-reliance on teachers might prevent students from learning how to solve problems on their own or discover novel solutions, especially in more advanced scienti c study. Furthermore, it is also important to understand that scientists' knowledge and under- standing of scienti c principles is constantly being developed, and therefore scientists might have conicting ideas that deserve further inspection. Accordingly, the most sophisticated approach to the source of knowledge may be a balance between justi cation by authority and personal justi cation ( Greene et al., 2008 ). To summarize, beliefs about justi cation are not simply adaptive or maladaptive in science, but depend on how they guide the student to useful or unuseful behaviors in dierent contexts. The extent to which students match their epistemic beliefs to the

demands of the course may explain the con icting evidence ( Dai & Cromley, 2014 ). Epistemic cognition has been described as exible and dependent on contexts (e.g., Kienhues, Ferguson, & Stahl, 2016 ). However, researchers have not yet examined whether this exibility can be adaptive if students come to hold beliefs that match the epis- temic climate of the context. Following this line of thinking, we propose that students will learn more when their beliefs match the way that knowledge is portrayed in a particular class. Students beliefs about the source of knowledge may direct them to seek or construct knowledge using di erent strategies (e.g., memorizing what the instructor says or practicing problems on one's own), which would be di erentially ef- fective depending on how the course is designed. It is therefore possible that students with less constructivist beliefs (i.e., justi cation by au- thority beliefs) will perform better in more traditional, lecture-based college courses (i.e., courses that are more objectivist in nature because the teacher's lecture serves as the singular truth in the class), whereas students with more constructivist beliefs (i.e., personal justi cation beliefs) will perform better in a constructivist-based, interactive- learning context. Instead of describing certain beliefs as availing, we propose the epistemic alignment hypothesis, which suggests that beliefs are more adaptive in contexts that call for them. Successful students are not simply those that hold universally adaptive beliefs, but rather are the students whose beliefs align with the demands of the learning context.

1.2. Mechanisms of personal epistemology development

Researchers have long observed that students' personal epistemol- ogies can change over time (e.g., Kitchener & King, 1981 ; Perry, 1970 ; Schommer, Calvert, Gariglietti, & Bajaj, 1997 ). Recent work suggests two possible mechanisms for change, one explicit and one implicit ( Brownlee, Schraw, Walker, & Ryan, 2016 ; Lunn Brownlee, Ferguson, & Ryan, 2017 ). Explicit mechanisms derive from the conceptual change literature (e.g., Posner, Strike, Hewson, & Gertzog, 1982 ). When stu- dents' existing beliefs are directly confronted, this can trigger a dis- satisfaction with current beliefs called epistemic doubt (Bendixen & Rule, 2004 ). Examples of triggers include learners reading con icting texts ( Ferguson, Bråten, & Strømsø, 2012 ) or instructors directly con- fronting students' epistemic assumptions during class meetings ( Kienhues, Bromme, & Stahl, 2008 ). For example, a student may believe that experts have all the answers; but, if the student then reads con- icting papers about how scientists disagree on what the healthiest foods are, he or she learns to believe that experts are not omniscient. Alternatively, epistemic change might occur implicitly through stu- dents' experiences with the underlying epistemic assumptions of the classroom. In these cases, the conceptions about nature of knowledge are not directly presented; students' assumptions may subtly change in ways that they are not aware of until they are explicitly asked with a targeted interview or survey question. For example, a child who goes to a school that allows students to direct their own learning may not be explicitly told that knowledge can be personally constructed, or even hear about knowledge or truth at all. Nevertheless, such an en- vironment implicitly suggests that knowledge can be personally con- structed and is not just handed down by authority gures. Such a child could also develop the belief that experts are not omniscient. A classroom's epistemic climate is the amalgam of students' per- sonal epistemologies, the instructor's personal epistemology, the epis- temic messages in classroom instruction, and the way course materials present knowledge ( Feucht, 2010 ). Within this interactionist model, a classroom context that portrays knowledge as complex and originating from multiple sources should lead students to develop similar epistemic beliefs. Instructors' personal epistemology leads them to create tasks within the classroom that correspond to those portrayals of knowledge. Qualitative research has also found that even within the same subject, the portrayals of knowledge can di er signi cantly between classrooms ( Hofer, 2004). The structure of a course in turn implicitly provides

M.M. Barger et al.

students with evidence about how knowledge is created (Barger & Linnenbrink-Garcia, 2017 ), eliciting changes in students' epistemic be- liefs ( Muis, Trevors, & Chevrie, 2016). This is evident in one study that demonstrated that a web-supported, problem-based learning course structure decreased the belief that knowledge is simple/certain ( Tolhurst, 2007 ). Students in the same classroom can also interpret the instructor's goals around knowledge di erently ( Hofer, 2004). How- ever, the role of incoming beliefs in the epistemic climate remains understudied. Hypothetically, students' beliefs reciprocally in uence their environment, such that when students' beliefs change, it changes how they engage with peers, instructors, and course materials, further eliciting epistemic change (Bendixen & Rule, 2004 ). However, little existing empirical evidence supports this supposition.

1.2.1. Constructivism in classrooms and constructivist beliefs One possible type of epistemic climate that instructors establish incorporates the learning theory of constructivism. Constructivism generally refers to a philosophical viewpoint that assumes no single objective reality exists ( Jonassen, 1991 ), and that truth must be con- structed ( Hodson & Hodson, 1998 ). Knowledge can be constructed personally by an individual or constructed and shared socially ( Driver, Asoko, Leach, Scott, & Mortimer, 1994 ). In the present study, we refer speci cally to social constructivism, because the view that knowledge is socially constructed is the theoretical basis for interactive-learning classrooms in which students work in small-groups to solve problems and create an understanding of the material. We speci cally focus on students' perceptions of a constructivist class in this study because teaching from a constructivist paradigm lends itself to a particular epistemic climate. Some epistemic beliefs align with constructivism, such as the personal justi cation of knowledge belief, while other epistemic beliefs are less aligned with constructivism, such as simple/ certain knowledge and justi cation by authority beliefs. In practice, constructivism has meant many things in the classroom context (Hmelo-Silver, Duncan, & Chinn, 2007 ; Kirschner, Sweller, & Clark, 2006 ), including collaborative and problem-based learning. Translating constructivist learning theory into pedagogical practices in the classroom often includes several common components, such as authentic tasks that are comparable to the activities of scientists (like working through unsolved problems and conducting experiments), supporting learner ownership of problems within a complex learning environment, and building knowledge with peers ( Savery & Duy, 2001 ). For example, in the college setting, instead of an instructor lecturing the majority of the time, class time instead could be devoted to students engaged in solving challenging problems in small groups. Such an approach creates an authentic task or problem that better re- ects how scientists do their work ( Osborne, 2010 ) and increases col- laborative construction of knowledge between peers ( Chi, 2009 ). This experience may lead students to perceive knowledge in the domain as more complex and coming from multiple sources. Teaching using pedagogical techniques inspired by constructivism in college STEM courses has become increasingly popular, in part be- cause student populations tend to demonstrate more learning, on average, in these classrooms compared to traditional, lecture-based courses ( Freeman et al., 2014 ). As an illustration, Process Oriented Guided Inquiry Learning (POGIL) activities ( Farrell, Moog, & Spencer, 1999 ; Moog & Spencer, 2008 ), which were developed with social constructivism foundations, have gained momentum in science class- rooms; hundreds of published POGIL activities are now available. Ad- ditionally, ipped classrooms, in which students engage with tradi- tional lecture content outside of the classroom and engage in problem solving during class time, have received recent attention in college chemistry ( Goldwasser, Mosley, & Canelas, 2016 ; Canelas, Hill, & Novicki, 2017 ; Chase, Pakhira, & Stains, 2013 ; Fautch, 2015 ; Hein, 2012 ; Teo, Tan, Yan, Teo, & Yeo, 2014). Researchers have proposed many reasons for the advances in the average student's achievement observed after implementation of these types of teaching practices. The

reasons include inducing deeper learner understanding of the material ( Ruiz-Primo, Briggs, Iverson, Talbot, & Shepard, 2011 ), replacing competition with collaboration ( Springer, Stanne, & Donovan, 1999 ), and providing students with opportunities to resolve cognitive dis- sonance ( Jensen & Lawson, 2011 ). Phrases such as student-centered learning, active-learning, interactive-learning, and cooperative learning have dierent, nuanced meanings for di erent people, but these terms are often used interchangeably in the literature. We will employ the phrase interactive-learning throughout the remainder of the text for consistency it constitutes the most accurate description of the learning environment in this study, which involves actively engaging students in inquiry-guided problem solving activities in small groups to allow them to construct a shared understanding of organic chemistry. One experimental study provided evidence that constructivist ped- agogical practices can change epistemic beliefs among undergraduate statistics students ( Muis & Du y, 2013 ). This study demonstrated that, when compared to a typical classroom, a constructivist classroom led students to believe that statistics knowledge is complex and can be personally constructed. This supports the assertion constructivist learning environments shape epistemic beliefs and may serve as one way of fostering epistemic change. This prior study and others (e.g., Brownlee, Purdie, & Boulton-Lewis, 2001) involved interventions that explicitly targeted students' personal epistemologies; however, it is not clear whether epistemic development would occur in a more natur- alistic constructivist classroom that did not explicitly foster epistemic change ( Muis & Du y, 2013 , pp. 223 224). For example, a course structure that required students to learn on their own or gure out answers in teams would implicitly suggest that knowledge is complex and socially constructed, without explicitly addressing these issues with students. Furthermore, it is unclear from these studies what speci c aspects of the constructivist classroom led to epistemic changes. Did the opportunity to interact with peers, participating in authentic tasks, the complexity of the problems presented in class, or some other factors lead to changes in beliefs? The studies also did not explicitly test the interactive nature of students' preexisting personal epistemologies with the portrayal of knowledge in the classroom. A student's initial personal epistemology may color his or her interpretation of the epistemic climate. For ex- ample, the e ectiveness of one personal epistemology intervention depended on whether students already held constructivist views of knowledge ( Kienhues et al., 2008 ). Furthermore, students in traditional science classrooms who believed that science is socially negotiated were more likely to perceive their classroom as lacking opportunities to construct knowledge ( Tsai, 2000 ). However, it seems just as likely that under the context of a constructivist classroom, more constructivist beliefs would allow students to perceive (and thus take advantage of) opportunities to construct complex knowledge for oneself. If students are in a interactive-learning context, possessing constructivist epistemic beliefs should direct students to construct their own knowledge.

1.3. Current study

The current study sought to improve our understanding of the role of a constructivist-based, interactive-learning environment in epistemic change. We investigated college students' personal epistemology de- velopment in traditional lecture-based and interactive-learning organic chemistry classrooms. Examining students' epistemic beliefs and per- ceptions of a constructivist climate enabled us to: (1) investigate whe- ther or not perceptions of the epistemic climate implicitly cause epis- temic change and (2) clarify what role learners' initial epistemic beliefs play in the process. Finally, we wanted to examine an epistemic alignment hypothesis, postulating that epistemic beliefs di erentially predict academic achievement in lecture-based and interactive-learning contexts. If the hypothesis is correct, then an individual learner's aca- demic achievement to some extent depends upon the degree of matching between the learner's epistemic beliefs and the classroom

M.M. Barger et al.

M.M. Barger et al. Fig. 1. The conceptual model of the development of personal epistemology via

Fig. 1. The conceptual model of the development of personal epistemology via perceptions of a constructivist learning environment. These perceptions are divided into three dimensions in the tested model (see Fig. 2).

environment established by the instructor. We tested three research questions based on our conceptual model of epistemic change. (RQ1) Do students' epistemic beliefs, nal exam grades, and perceptions of dierent components of constructivism vary between interactive-learning and lecture-based classrooms? We hy- pothesized that the interactive-learning classroom would increase per- ceptions of constructivism such as having more opportunities to interact with peers, seeing problems as complex, and authentic scienti c tasks. We also hypothesized that students in the interactive-learning class- room would subsequently report more constructivist epistemic beliefs (i.e., higher personal justi cation, lower justi cation by authority and simple/certain knowledge) relative to those observed in the traditional, lecture-based classroom at the end of the course (as in Muis & Du y, 2013 ). (RQ2) Is there a reciprocal relation between perceptions of a constructivist classroom and changes in personal epistemology over the course of the semester (see Fig. 1)? We hypothesized that the percep- tions that the classroom had more constructivist teaching practices would predict changes in epistemic beliefs, and that this relation might be stronger in the interactive-learning classroom. We also expected that students' initial beliefs would predict their perceptions of the classroom environment in both classrooms, providing evidence for one possible mechanism for how learners' personal epistemology impacts the epis- temic climate ( Feucht, 2010 ). (RQ3) How do epistemic beliefs relate to achievement in organic chemistry in these di erent contexts? We ex- pected support for an epistemic alignment hypothesis: less con- structivist epistemic beliefs predict higher achievement in the lecture- based classroom, but more constructivist epistemic beliefs predict higher achievement in the interactive-learning classroom (Dai & Cromley, 2014 ).

Table 1 Demographic information: compositions of the lecture-based and interactive- learning classes; N-values and percentages by gender and self-reported race.

 

Lecture-based

%

Interactive-learning

%

class

class

Gender Male

40

36

87

61

 

Female

72

64

56

39

Race

Black

11

10

14

10

Asian

41

37

38

27

White

46

41

66

46

Hispanic

7

6

7

5

Other/

7

6

18

13

multiracial

2. Method

2.1. Participants

Participants ( N = 270) were undergraduates in two sections of an organic chemistry course at a selective university in the southeastern United States. The rst was a lecture-based format (n = 116, total en- rollment of 132), and the other section used a problem-based, inter- active-learning course format ( n = 154, total enrollment of 158). At this university, students typically take an advanced general chemistry course in the fall of their rst year, followed by two semesters of or- ganic chemistry in the spring of their rst year (the context of the current study) and the fall of their second year as previously described ( Hall, Curtin-Soydan, & Canelas, 2014; Canelas, 2015 ; Canelas et al., 2017 ). The overall sample included 85.1% rst-year students and re- ected the gender (47.2% male) and ethnic (41.6% Caucasian, 29.4% Asian/Paci c islander, 9.3% African American, 5.2% Hispanic, 9.3% other/multiracial, 5.2% did not report) make-up of the university. 1 The demographic characteristics of the lecture-based and interactive- learning classes are presented in Table 1 . The study was approved by the institutional review board in accordance with guidelines to protect human subjects.

2.2. Procedure

The third author was the instructor for both of the classes, covering the structures and reactions of carbon compounds and the impact of selected organic compounds on society. The use of di erent instruc- tional strategies was not advertised to students. Thus, students were not aware of the approaches to teaching when they enrolled for the course. In the lecture-based class, students attended three 50-minute lec- tures per week and the instructor primarily delivered course content via lecture. The instructor used visual aids (e.g., PowerPoint slides, docu- ment camera, and writing on board) while explaining the course con- tent. Students were also sometimes given time to think about a problem and then randomly selected to answer a question, but the instructor delivered a traditional lecture for the majority of class time while the students took notes. Students were advised to keep up with the course by reading and doing practice problems for 1 2 h each day outside of class. Students in the interactive-learning class also attended three 50- minute classes per week. In this case, student-centered, cooperative- learning activities replaced the lecture; the goal was to spend the ma- jority of class time with all students actively engaged in inquiry-guided problem solving in small groups rather than listening passively to a lecture. The primary mode of instruction for the interactive-learning class involved working in small groups on POGIL (POGIL; Farrell et al., 1999 ; Moog & Spencer, 2008) or Problem Manipulation activities ( Niaz & Robinson, 1992 ; Siburt, Bissell, & MacPhail, 2011 ). POGIL uses sca olded classroom activities and rotating team member group roles that are carefully designed and faculty peer-reviewed to contain a cycle of constructive learning: exploration, concept invention, and applica-

tion ( https://pogil.org/about ). These interactive-learning activities align with many components of constructivism ( Savery & Du y, 2001 ). First, it provides students the opportunity to work with their peers and socially construct understanding of the material. Second, working on problem solving tasks as opposed to listening to lectures more closely aligns with what scientists do compared, making classroom activities more authentic. Third, the exibility of the problem-solving tasks should allow students to take multiple pathways to solving problems,

1 The distribution of gender across the two sections of the course was unequal. Gender was controlled in initial analyses but was unrelated to epistemic beliefs (as other re- searchers have found; Hofer & Pintrich, 1997 ) and was therefore not included in nal analyses.

M.M. Barger et al.

making the scienti c process seem more complex than a collection of memorizable facts. Most activities were drawn from reviewed and published literature ( Ruder, 2015 ; Straumanis, 2009 ), although the instructor designed some activities (see Supplementary Materials for an example of instructor-designed activities). The instructor and several teaching assistants circulated through the classroom engaging in con- versations with groups of four to six students about the activities in order to prompt or sca old their thinking about the activities. Prior to coming to class, students had access to video lectures and were en- couraged to work on introductory activities related to the problems they would complete in small groups. Following a previously reported procedure ( Canelas et al., 2017), an independent coder observed ve randomly selected classes in both the lecture-based and interactive-learning sections. During these observa- tions, the coder assessed the amount of time that all students were engaged in active-learning activities other than listening or taking notes. Students in the interactive-learning class spent approximately 63% of class time working actively in small groups compared to about 17% of class time devoted to active-learning in the lecture-based class. The remainder of the times included students listening and taking notes, having tasks explained, and going over solutions. The laboratory sections (once per week), recitation sessions (once per week), homework assignments, textbook ( Organic Chemistry , 5th ed.; Loudon, 2009 ), course website, and other course materials were identical in both classes. This included identical online resources such as unit plans with learning objectives, online homework problems, worksheets for guiding recitations led by teaching assistants, a folder of student-centered activities, optional lecture videos, PowerPoint slides, and information about places to get extra help (peer tutoring, chemistry resource room, instructor o ce hours). Students in the lecture-based class had access to all of the activities that the students in the inter- active-learning class were completing in real time, and they were en- couraged to do them on their own for practice as an optional activity. Conversely, students in the interactive-learning class were encouraged to take advantage of the optional lecture videos when needing clar- i cation about concepts. Students completed surveys in their chemistry class at the beginning (T1), middle (T2), and end (T3) of the semester assessing their epis- temic beliefs about chemistry (T1 and T3) and their perceptions of the classroom environment (T2). Students completed a nal, cumulative exam in the week following the T3 survey. These time points were se- lected to assess students' initial and nal beliefs while getting a cross- section of their perceptions of how the class was proceeding in the middle of the semester. The instructor introduced the researchers at the beginning of the study but was not present while students were re- cruited to participate in the study or while they completed the surveys. Students were not compensated for participating.

2.3. Measures

All items were measured on a ve-point Likert scale (1 = Strongly Disagree ;5= Strongly Agree ). Items within each subscale were averaged to create a single score at each time point. We calculated Cronbach's alpha to determine the internal consistency reliability of each of the measures. Generally, an alpha 0.70 indicates su cient reliability ( Cortina, 1993 ); however, reliability scores lower than 0.70 may also be acceptable ( Schmitt, 1996 ). We used con rmatory factor analysis (CFA) to validate the factor structure of the measures. We report the speci c results of the CFA analyses in the results section.

2.3.1. Personal epistemology The epistemic beliefs about chemistry survey was adapted from a scale that was previously used in the domain of engineering (en- gineers and engineering were replaced by chemists and chem- istry ), developed by Barger, Wormington, Huettel, and Linnenbrink- Garcia (2016) via combining items from scales that were designed to

function across multiple domains, including psychology, science, math, and history ( Greene, Torney-Purta, & Azevedo, 2010 ; Hofer, 2000 ). The original engineering items had been inspected previously by an expert in the eld of epistemic cognition and were reviewed by the third au- thor, a chemistry professor, who found the adapted scales relevant to the organic chemistry context. We selected these scales because they

were domain-speci c and showed reasonable reliability with a similar demographic of students. There were 17 domain-speci c items, which assessed three subscales including simple/certain knowledge (e.g., Truth is unchanging in chemistry. ), justi cation by authority (e.g., If

a chemist says something is fact, I believe it.), and personal justi ca-

tion (e.g., In chemistry, what's a fact depends on a person's point of view.) at both T1 and T3. Based on the CFA results, we dropped one item from the justi cation by authority scale ( If you read something in

a chemistry textbook, you can be sure it's true. ) and two items from the

simple/certain knowledge scale ( In chemistry, most work has only one right answer and All chemistry professors would probably come up with the same answers to questions in chemistry. ). Thus, the nal scale included 14 items total (personal justi cation = 3 items, α T1 = 0.66, α T3 = 0.71; justi cation by authority = 5 items, α T1 = 0.73,

α T3 = 0.81;

α T3 = 0.81). Thus, the personal epistemology scales met the generally accepted standard for reliability of α 0.70 at all-time points except for the simple certain knowledge and personal justi cation scales at

time 1, which were only slightly under the 0.70 rule of thumb. See results for further details about the validity of the scales.

and simple certain knowledge = 6 items, α T1 = 0.67,

2.3.2. Perceptions of constructivist learning environment

We chose to measure perceptions of constructivism in order to as- sess whether students were able to perceive di erences in classroom practices and because we expected students' subjective perceptions to be more predictive of individual belief change than a more objective measure. A combination of items used in prior research ( Kahle, Meece, & Scantlebury, 2000 ) and new items representing additional con- structivist practices ( Savery & Duy, 2001 ) were used to measure stu- dents' perceptions of a constructivist environment in their organic chemistry classroom (i.e., perceptions of constructivism). The nal measure included 10 items. Students were asked to think about their current organic chemistry classroom when reading these items. Two items represented the role of peers in the learning process (Con- structivism - Peers, e.g. I learn from my classmates. α = 0.88). Four items measured the extent to which students felt the instructor believed in the complexity of knowledge ( Constructivism - Complex , e.g. My or- ganic chemistry professor thinks problems can be solved in more than one way.α = 0.72). The nal four items measured whether students found the learning environment to authentically replicate the work of scientists ( Constructivism - Authentic , e.g. My organic chemistry pro- fessor encourages me to think like a scientist when solving problems. α = 0.77). Overall, the perceptions of constructivism sub scales met the generally accepted standard for reliability of α 0.70. A CFA sup- ported the three-factor structure of the perceptions of constructivist learning environment scale (see results section for further details).

2.3.3. Final exam scores

Students in both courses completed an identical cumulative nal exam that was worth a maximum of 300 points ( M = 199.8, SD = 44.7). The exam included twenty multiple choice questions (33% of grade) as well as free-response questions such as essays for ex- plaining concepts, writing mechanisms, and interpreting spectral data (67%). Given the wide variety of question types and forms of assess- ment, this test conformed to norms in the discipline of chemistry and was considered an appropriate assessment of students' organic chem- istry content and learning gains in both classrooms. The use of the same exam, which measured key concepts of general chemistry including basic knowledge and higher level conceptual understanding, enabled us to directly compare learning in the two classroom contexts. Students

M.M. Barger et al.

completed the timed paper-and-pencil exam in class over 3 h, with no notes or books. The nal exam score was not curved and was roughly normally distributed (skewness = 0.44, kurtosis = 0.36).

2.4. Data analysis

2.4.1. Con rmatory factor analyses

It is important to establish construct validity and reliability of

measures used in any study. We used CFA to test the construct validity

of the hypothesized latent factor structure of the perceptions of con-

structivism and epistemic beliefs scales. The goal was to assure that observed covariance between items t a theoretical pattern in which covariance between items on a single scale are more related to each another than to items on a di erent scale, establishing construct va-

lidity ( Hoyle, 2000). We examined t indices to assess the t of the CFA models (see below for model t standards). We also examined mod- i cation indices for evidence of cross loading of items on other factors ( Byrne, 2012 ). A modi cation index of > 3.84 (suggesting specifying the parameter would result in a signi cant reduction in chi-square) and

a large expected parameter change (EPC) suggests possible cross-

loading with a factor that an item was not intended to measure. We evaluated the adequacy of items based on these factors and, im- portantly, based on the theoretical structure of the measure ( Byrne, 2012 ). We also examined correlations between di erent measures to

assess discriminant validity between epistemic beliefs and perceptions

of constructivism.

2.4.2. Primary analyses

Each class of students was considered a separate group in the ana- lysis (one group experienced the lecture-based class, while the other group experienced the interactive-learning class.) To analyze our rst research question (RQ1: Do students' epistemic beliefs, perceptions of di erent components of constructivism, and exam performances vary between interactive-learning and lecture-based classrooms), we used mixed repeated-measures MANOVA to examine any di erences be- tween the groups in their changes in epistemic beliefs over the time- frame of the semester. In the repeated-measures analysis, a signi cant time × group interaction would indicate dierences between the two groups in changes in epistemic beliefs from T1 to T3. MANOVA was used to examine mean di erences between the groups in their per- ceptions of constructivism variables. An independent samples t -test was used to examine di erences between the groups on nal exam scores. To account for multiple tests, we used p .016 (0.05 divided by three tests) as an indicator of a signi cant omnibus test using the Bonferroni correction ( Howell, 2012 ). For research question two (RQ2: Is there a reciprocal relation be- tween perceptions of a constructivist classroom and changes in personal epistemology over the course of the semester?) we used path analysis (see Fig. 2 ). A path modeling approach was most appropriate since this allowed us to examine the reciprocal e ects of epistemic beliefs on perceptions of constructivism and vice versa over the course of the semester in a single analysis ( Keith, 2015 ). Such an analysis would have otherwise required piecemeal multiple regression analyses, which is less ideal than using a single model. Finally, to examine research question 3 (RQ3: How do epistemic beliefs relate to achievement in organic chemistry in these di erent classroom contexts), we specied a multi-group path model ( Byrne, 2012 ). We analyzed the same model specied for research question 2 (see Fig. 2 ), but analyzed this model for the lecture-based class and the interactive-learning class separately in a single path analysis. This ap- proach allowed us to test whether the e ects of epistemic beliefs on achievement were the same in both classes ( Byrne, 2012 ). We rst al- lowed all coe cients between the two groups (lecture-based and in- teractive-learning classes) to vary (the con gural model) and examined the model for adequate t. Given an acceptable model t, we then xed the regression coe cients across the two groups to be equal and

compared the new model's t to the congural model using the change in chi-square test (Byrne, 2012 ). Using this approach, a statistically signi cant change in chi-square test indicates a signi cantly worse t- ting model compared to the congural model. Under such circum- stances, we examined the modi cation indices to determine which speci c coecients should be freely estimated across the two groups. We freed regression coe cients based on the modi cation indices until the change in model t was non-signi cant compared to the con gural model. In other words, the aim of this analysis was to nd the most parsimonious model (the model with the most degrees of freedom) that is not signi cantly worse tting compared to the con gural model. Given rules of thumb related to sample size in SEM analysis ( Mueller & Hancock, 2008 ), our sample size was relatively small for the path analyses described above (see Limitations for a discussion of this point). However, the path analysis approach is best when examining reciprocal e ects ( Keith, 2015), and reduced the number of statistical tests re- quired to answer our research questions. To mitigate our potential to be underpowered, we did not include a measurement model in the ana- lyses so there would be less complexity to the model. To determine the quality of model t, we followed guidelines uti- lized in prior research ( Pekrun, Hall, Goetz, & Perry, 2014 ; Trautwein et al., 2012 ). Accordingly, we considered models to demonstrate ac- ceptable model t when at least two indicators met the conventions of good model t ( 0.95 for CFI, < 0.08 for RMSEA, and < 0.10 for SRMR) outlined by Hu and Bentler (1999) . In the path analyses, missing data was handled using Full Information Maximum Likelihood estimation to t the models, which uses all available data instead of excluding participants with some missing data entirely ( Raykov, 2005). We used listwise deletion in the mean comparison analyses (MANOVAs and t -test).

3. Results

3.1. Con rmatory factor analyses

3.1.1. Epistemic beliefs

We examined the adequacy of a three-factor model at both T1 and

T3. However, modi cation indices suggested cross-loading of 3 items across the three-factors at T1 and T3 epistemic beliefs. Any model ad- justments based on modi cation indices should be made with the theory and prior research in mind. Since these three items were cross- loading on factors that were not appropriate given the construct they were intended to measure based on prior research, we removed these items from the analyses. The nal CFA models for T1 (χ 2

RMSEA = 0.068, 90% CI:

[0.051, 0.085], SRMR = 0.066) and for T3 (χ 2 [74] = 124.95, p < .001, CFI = 0.92, RMSEA = 0.052, 90% CI: [0.035, 0.067], SRMR = 0.054) demonstrated acceptable model t. We present the factor loadings for the epistemic beliefs scales in Table 2 . Both the model t and factor loadings (all factor loadings were statistically sig- nicant and > 0.35), combined with the acceptable Cronbach's alpha, provide evidence that our observed covariance matrix aligns with our theoretical model of these three epistemic beliefs constructs, which helps to establish construct validity ( Hoyle, 2000 ).

[74] = 142.68, p < .001, CFI = 0.92,

3.1.2. Perceptions of constructivism

The CFA on the perceptions of constructivism scale supported a

three-factor structure ( χ 2 [32] = 71.45.68, p < .001,

RMSEA = 0.074, 90% CI: [0.051, 0.097], SRMR = 0.056). While there was some evidence of potential cross-loading on two items, we decided to retain these items since there was not a strong theoretical rationale for removing them, and since removing these items would signicantly reduce model t. We present the factor loadings on the perceptions of constructivism scale in Table 3. Again, these CFAs provide evidence that our proposed theoretical model reasonably t the observed cov- ariance matrix, providing evidence for construct validity.

CFI = 0.92,

M.M. Barger et al.

M.M. Barger et al. Fig. 2. The initial model with all pathways shown. The initial two-group

Fig. 2. The initial model with all pathways shown. The initial two-group model constrained means and path coe cients to be equal across classes. Correlations between variables are not shown to reduce visual complexity.

3.2. Descriptive statistics and correlations

Descriptive statistics, including means and standard deviations for each of the classes, are presented in Table 4 . Correlations among the variables are presented in Table 5 . The correlations between epistemic belief and perceptions of constructivism were small ( r s 0.3), which suggests discriminant validity between the two sets of measures.

3.3. RQ1

We examined di erences in changes in personal epistemology from baseline to the end of the semester using repeated-measures MANOVA. The multivariate test indicated a non-signi cant time × class interac- tion indicating there were no signi cant di erences between the two groups in their change in personal epistemology over time (Pillai's Trace = 0.01, F [3, 190] = 0.93, p = .430, η 2 = 0.014). We used MANOVA to examine mean di erences between the groups in their perceptions of constructivism. The multivariate test indicated signi cant di erences in perceptions of constructivism by the two groups (Pillai's Trace = 0.33, F [3, 219] = 35.362, p < .001, η 2 = 0.33). The univariate tests indicated there were signi cant

Table 2 Standardized factor loadings for epistemic beliefs CFA models at time 1 and time 3.

Table 3 Standardized factor loadings for perceptions of constructivism.

 

123

I talk with my classmates about how to solve problems.

0.72

I learn from my classmates.

0.92

My organic chemistry professor asks questions that have

I

understand how new concepts in organic chemistry relate to

0.50

more than one answer. My organic chemistry professor thinks problems can be

0.59

solved in more than one way. My organic chemistry professor encourages us to develop our

0.57

own problems to solve. My organic chemistry professor encourages us to nd and

0.68

understand patterns on our own. The assignments serve a larger purpose than just busy work.

0.51

My organic chemistry professor encourages me to think like a

0.56

scientist when solving problems.

0.52

my previous knowledge. The process behind solving organic chemistry problems is very important.

0.56

Note . 1 = Constructivism-peers. 2 = Constructivism-complex. 3 = Constructivism- authentic. All factor loadings are statistically signicant at p < .001.

Time 1

Time 3

123123

All experts in chemistry understand the eld in the same way.

0.39

0.58

Truth is unchanging in chemistry.

0.62

0.72

Principles in chemistry are unchanging.

0.50

0.69

Most of what is true in chemistry is already known.

0.41

0.59

In chemistry, what is a fact today will be a fact tomorrow.

0.55

0.67

Chemists' knowledge of the facts about chemistry does not change.

0.53

0.65

If my personal experience conicts with ideas in a chemistry textbook, the book is probably right.

0.36

0.42

If a chemist says something is a fact, I believe it.

0.67

0.78

Things written in chemistry textbooks are true.

0.59

0.62

I believe everything I learn in chemistry classes.

0.78

0.76

If a chemistry professor says something is a fact, I believe it.

0.60

0.82

In chemistry, the truth means di erent things to dierent people.

0.71

0.83

In chemistry, everyone's knowledge can be dierent because there is no one absolutely right answer.

0.53

0.52

In chemistry, what's a fact depends upon a person's point of view.

0.66

0.68

M.M. Barger et al.

Table 4 Descriptive Statistics for Groups in the lecture-based and the interactive- learning classes.

 

Lecture-based class

Interactive-learning class

n

M

SD

n

M

SD

Time 1 Simple certain knowledge Justi cation by authority Personal justi cation Time 2 Constructivism peers Constructivism complex Constructivism authentic Time 3 Simple certain knowledge Justi cation by authority Personal justi cation End of semester Final exam grade

114

2.29

0.47

145

2.44

0.51

114

3.46

0.50

145

3.56

0.55

114

2.86

0.66

143

2.91

0.73

93

3.73

0.82

133

4.44

0.55

93

3.84

0.55

131

3.48

0.62

93

3.99

0.42

131

4.00

0.53

81

2.33

0.54

120

2.50

0.65

81

3.50

0.57

120

3.59

0.56

81

2.92

0.70

120

2.92

0.72

110

200.49

40.20

143

199.56

48.07

Note . Means that signi cantly di ered between classes are bolded.

when compared to the students in the lecture-based class. There were no mean di erences between the groups in either changes in epistemic beliefs over time or nal exam score. Overall, this pattern of ndings was contrary to our rst hypothesis. We elaborate on this further in the discussion section below.

3.4. RQ2

We hypothesized that constructivist perceptions would predict changes in personal epistemology (see path analysis results in Fig. 3). Indeed, perceiving a complex environment predicted decreases in simple/certain knowledge (b = 0.18, p = .018, β = 0.18) after controlling for initial epistemic beliefs. In other words, perceptions that the learning environment was complex predicted a decrease in the belief that knowledge is a collection of unchanging facts. However, learning from peers and perceiving class problems as authentic did not relate to changes in epistemic beliefs. We also hypothesized that students' initial personal epistemology would predict their perceptions of the classroom environment. As ex- pected, beliefs in simple/certain knowledge predicted lower percep- tions of an authentic learning environment ( b = 0.21, p = .002, β = 0.23). Therefore, if students started the semester believing that knowledge was a collection of unchanging facts, they perceived the classroom as less relevant to real world problems. Also in line with our hypotheses, personal justi cation predicted higher perceptions of complexity (b = 0.12, p = .038, β = 0.14). In other words, believing that people construct knowledge for themselves was associated with

Table 5 Correlations between all measures in study for students in the lecture-based and interactive-learning classrooms.

 

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

1. T1 SCK 2. T1 JA 3. T1 PJ 4. T2 peers 5. T2 authentic 6. T2 complex 7. T3 SCK 8. T3 JA 9. T3 PJ 10. Final exam

 

0.04

0.13

 

0.07

0.15

0.12

0.34

0.03

0.10

0.10

0.33

 

0.12

0.11

0.29

0.07

0.02

0.52

0.04

0.18

0.17

0.28

 

0.12

0.08

0.01

0.03

0.16

0.34

0.10

0.16

0.11

0.15

 

0.21

0.29

0.18

0.12

0.18

0.10

0.20

0.07

0.15

0.49

0.49

0.26

0.37

0.16

0.15

0.22

0.04

0.30

 

0.36

0.51

0.35

0.00

0.25

0.12

0.63

0.38

0.26

0.12

0.19

0.26

 

0.23

0.21

0.08

0.40

0.72

0.35

0.16

0.08

0.08

0.50

0.12

0.31

0.07

 

0.00

0.54

0.11

0.09

0.18

0.05

0.19 0.10

0.04

0.00

0.02

0.28

0.06

0.14

0.04

0.09

0.21

Note . Correlations between variables for students in the lecture-based class are located above the diagonal and those for the interactive-learning class are located below the diagonal. SCK = simple/certain knowledge; JA = justi cation by authority; PJ = personal justi cation. T2 variables are components of perceptions of constructivism. p < .10. p < .05. p < .01.

di erences between the groups on constructivism-peers ( F [1, 221] = 58.71, p < .001, η 2 = 0.21) and constructivism-complex ( F [1, 221] = 19.65, p < .001, η 2 = 0.08). The group in the interactive- learning classroom was, on average, signi cantly higher on con- structivism-peers when compared to the lecture-based class group (see Table 4 for mean values). However, the lecture-based class group had signi cantly higher mean constructivism-complex scores compared to the interactive-learning group. Finally, we used an independent samples t -test to examine any di erences in the average nal exam scores. Results indicated that the two groups did not perform di erently on the nal exam ( t [251] = 0.16, p = .870, d = 0.02; see Table 4 for means). In sum, mean group dierences were observed in perceptions of constructivism-peers and constructivism-complex by the students in the lecture-based and interactive-learning classes. On average, the group of students in the interactive-learning class reported more learning with peers and perceived the teacher as creating a less complex environment

perceiving the classroom as more complex. Contrary to our expecta- tions, justi cation by authority beliefs positively predicted perceptions of an authentic learning environment ( b = 0.13, p = .040, β = 0.16). The belief that knowledge comes from authorities was related to per- ceiving the classroom as more relevant to real world problems. Finally, when combining both classes, T1 personal justi cation predicted nal grade (b = 15.33, p = .001, β = 0.24). However, when analyzed across the two classes, this relation did not hold for both groups. We review this nding in the next section.

3.5. RQ3

After following the multi-group modeling procedure described previously, a multi-group model with freely estimated regression paths from T1 personal justi cation to nal exam score and from T3 justi - cation by authority to nal exam score while leaving the remaining regression paths constrained to be equal across groups was most

M.M. Barger et al.

M.M. Barger et al. Fig. 3. The fi nal model of personal epistemology devel- opment across

Fig. 3. The nal model of personal epistemology devel- opment across both courses. Only the paths between T1 personal justi cation and nal exam grade and T3 justi- cation by authority and grade di ered between students in the lecture-based and interactive-learning classrooms. All other paths were thus constrained to be equal across groups. Unstandardized path coe cients are shown. Final exam grades were graded on a scale of 300. Non- signi cant paths and correlations between variables at each time are not shown to reduce the complexity of the gure.

Table 6 Model t statistics for nested multi-group path models.

 

Free

χ 2

df

CFI

RMSEA SRMR

parameters

M1: congural model M2: x all regression paths

112

22.11

18

0.99

0.04

0.04

85

66.99

45

0.94

0.06

0.09

M3:

free T1PJ grade

86

61.03

44

0.96

0.05

0.08

M4: free T3JA grade

87

57.85

43

0.96 0.05

0.08

Note . Final model is in bold. Δχ 2 (25) = 35.74, p > .05.

parsimonious ( Δχ 2 [25] = 35.74, p > .05). 2 The nal model, shown in Fig. 3, met the criteria for acceptable t (χ 2 [43] = 57.85, p = .06, CFI = 0.96, RMSEA = 0.051, 90% CI: [0.000, 0.082]; SRMR = 0.079; Hu & Bentler, 1999 ). Table 6 presents model t information for each model in each step of our analysis. In the lecture-based classroom, T3 beliefs in justi cation by authority (a less constructivist belief) posi- tively predicted nal exam score ( b = 22.82, p = .005, β = 0.33). A one-unit increase on the T3 justi cation by authority scale corre- sponded with approximately a 23-point increase (7.6%) in nal exam score. The relation between T3 justi cation by authority and nal exam score was not statistically signi cant in the interactive-learning class- room ( b = 3.81, p = .684). However, in the interactive-learning class- room, higher scores on beliefs in personal justi cation (a more con- structivist belief) at the beginning of the semester positively predicted nal exam score ( b = 22.96, p < .001, β = 0.35). Speci cally, a one- unit increase in T1 personal justi cation corresponded with approxi- mately a 23-point increase (7.6%) in nal exam grade in the inter- active-learning class. There was no such relation in the lecture-based class between these two variables (b = 6.78, p = .26). These ndings support the epistemic alignment hypothesis, which we elaborate on further in the discussion section.

4. Discussion

This study adds to the broader research literature on the interplay between students' beliefs, the epistemic climate of the classroom, and achievement under di erent contexts. First, we examined the e ects of

2 The value of the modication indices for freeing the path from T1 personal justi - cation to nal grade and the path from T3 justi cation by authority to nal grade were very close in value, suggesting each would equally improve model t equally if freed. Therefore, we freed both paths to reach our nal model.

an interactive-learning course on perceptions of constructivism, per- sonal epistemology development, and nal exam grades. We identi ed three components of constructivism as perceived by students in the classroom: opportunities to learn from peers, an instructor who portrays complexity, and an authentic learning environment. The interactive- learning classroom increased perceptions of peer interaction. Surprisingly, students in the interactive-learning classroom reported that the instructor provided fewer opportunities to solve complex pro- blems relative to the lecture-based classroom. Perceiving classroom tasks as authentic was not aected by the class type. In this case, the interactive-learning classroom did not directly change personal episte- mology or grades. Despite these unexpected ndings, the second and third research questions provided an opportunity to examine the crucial relations between constructivist practices in the classroom, personal epistemology, and achievement. Perceptions of constructivist practices in the form of complexity in the classroom predicted changes in the belief in simple/certain knowledge, and all of students' initial epistemic beliefs predicted their perceptions of a constructivist classroom environment. The perception that the classroom a orded opportunities to take ownership of complex problems seemed to be a key ingredient for epistemic change. Those who perceived class materials as complex reported lower non-con- structivist belief of simple/certain knowledge by the end of the seme- ster. The results also provided novel evidence that students' beliefs play a role in the epistemic climate. In line with the interactive epistemic climate model (Feucht, 2010 ), we found evidence that initial personal epistemologies predicted students' perceptions of the classroom en- vironment. Whereas previous research found that students in a tradi- tional classroom with more constructivist beliefs found their classroom environments to a ord fewer opportunities to construct knowledge ( Tsai, 2000 ), the current study mostly found that students with initial constructivist epistemic beliefs reported more practices aligned with constructivist learning. First, students who initially believed that knowledge is complex and evolving rather than simple and certain were more likely to nd their classroom tasks as authentic. Second, students who started their class believing that knowledge is self-constructed (high personal justi ca- tion) were more likely to perceive class activities as complex. Perhaps believing that knowledge is self-constructed makes knowledge por- trayals in the classroom seem more complex since this belief orients the student to thinking about how multiple individuals play a role in creating that knowledge. In both cases, having more constructivist

M.M. Barger et al.

beliefs predicted students seeing more constructivist practices in the classroom, as hypothesized. However, justi cation by authority (a less constructivist belief) predicted higher perceptions of authentic pro- blems, contrary to expectations. Given the surprising nature of this result, the small e ect size, and the non-signi cant zero-order corre- lation in the interactive-learning class between these two variables,

future research will need to replicate and explicate this nding. To answer our third research question, we tested whether the re- lation between beliefs and achievement depends on the structure of the classroom. We found some support for the epistemic alignment hy- pothesis. Students who began the semester with more constructivist epistemic beliefs (i.e., higher personal justi cation beliefs) fared better on the nal exam in the interactive-learning class, but not in the lecture- based class. Conversely, students who ended the semester with less constructivist beliefs (i.e., higher justi cation by authority beliefs) in the lecture-based classroom achieved higher scores on the same nal exam. These results suggest that adaptive epistemic beliefs only bene t performance in cases where the epistemic climate of the class- room aligns with an individual student's own epistemic beliefs. These ndings are in line with recent ndings that personal epistemology alone does not predict achievement, but rather that a match between students' epistemic beliefs and the epistemic context of the classroom is key (e.g., Dai & Cromley, 2014 ; Fruge & Ropers-Huilman, 2008 ). An open question is whether levels of students' achievement reinforce be- liefs that match the epistemic climate. In other words, students who develop a belief that is adaptive in one classroom (e.g., justi cation by authority in a lecture-based classroom) may then carry that belief into future classes. This is a particularly important question, given the ndings that incoming students' beliefs frame their perceptions of the epistemic climate, which in turn might lead to epistemic change and play a role in learning. There were some unexpected ndings regarding the eects of the intervention on students' perceptions and beliefs. The nding that stu- dents perceived the interactive-learning classroom to be less complex was inconsistent with expectations. One interpretation is that requiring knowledge construction of less knowledgeable students necessitates a simpli cation of the material ( Kirschner et al., 2006 ), suggesting a gap between constructivist theory and constructivist practice. Alternatively, it is possible that students' perceptions di ered from reality. The in- teractive-learning classroom may have been objectively more complex, but it was perceived by learners to be less complex. POGIL activities are extensively reviewed by multiple instructors to e ectively sca old students from their existing knowledge to more complex ideas ( Farrell et al., 1999; Moog & Spencer, 2008). It seems plausible that class ac- tivities seemed less complex because the interactive-learning format and well-designed activities made organic chemistry seem less di cult, as evidenced by students' self-reported gains in understanding concepts ( Canelas et al., 2017 ). Even if this was the case, perceptions of com- plexity predicted changes in personal epistemology. Therefore, re- searchers need to keep an open-mind to the possibility that interven- tions may have unintended consequences (see also, Kienhues et al., 2008 ). Furthermore, prior work has demonstrated that epistemic beliefs can be changed through intervention, but the current intervention did not change students' beliefs. It may be that speci cally targeting epis- temic beliefs is critical to initiating change ( Brownlee et al., 2001 ; Muis

& Du y, 2013 ). In the present intervention, messages about the nature

of knowledge were only implicit to the course structure. Future studies should compare classrooms with explicit epistemic messages and

classrooms that do not explicitly send messages about knowledge. For example, instructors might explicitly address beliefs by directly con-

fronting students' beliefs, such as using a refutational text ( Gill, Ashton,

& Algina, 2004 ) or by using speci c linguistic cues (e.g., This is un-

decided in the eld; ” “ I'm not the only source of answers. ) that expose the nature of knowledge in the discipline (Barger & Linnenbrink-Garcia, 2017 ). The importance of students' initial epistemic beliefs found here

might also explain why our intervention was not e ective in changing beliefs. A one-size- ts-all approach to classroom interventions might be particularly ine ective since students bring di erent beliefs to the rst day of class.

4.1. Limitations

The lack of strong ndings for the intervention may also be due to weaknesses in the methodological design. It may be that the single instructor was actually sending constructivist messages to both classes; this could wash out the potential e ects from the structure of the in- teractive-learning class itself. Though using the same instructor was useful to control for likely instructor e ects on the classroom environ- ment ( Roth & Weinstock, 2013 ), it may have reduced the strength of the potential eects of being in an interactive-learning classroom. Re- searchers might also consider blinding instructors to the objectives of future interventions, given that the opposite problem could occur if instructors attempt to directly change students' beliefs. Likewise, the use of identical assessments in the course may have dampened the e ects of the course structures. Students in both classes may have focused on the knowledge that would be required on the exams, and so they may have not been as aected by the day-to-day epistemic climate. However, it is di cult to compare achievement in two separate classes if the same assessment is not given. It may be important to assess other learning and achievement outcomes (e.g., achievement in future courses to assess long-term learning) if the exams cannot be equated across two contexts. The current study is only able to shed light on development of personal epistemology in the context of introductory undergraduate organic chemistry. Personal epistemology has domain-speci c compo- nents ( Muis et al., 2006 ) and changes as students learn more about a topic ( Greene et al., 2008 ). Prior domain knowledge was also not as- sessed in this study and having more domain knowledge may play a role in more sophisticated epistemic cognition ( Ferguson & Bråten, 2013 ). Students with greater domain knowledge may also be more able to perceive the complexity in problems. Furthermore, students might be more likely to change their beliefs in a more elementary course (e.g., general chemistry) or in a subject in which they have no prior experi- ence. Initial beliefs may play a stronger role in students' perceptions of the classrooms in advanced coursework because students become more committed to their beliefs over time. Future research will need to in- vestigate these possibilities in a broader range of topics with varying levels of expertise. Future research should carefully consider issues of measurement. Epistemic belief measures often do not meet the criteria of being reli- able and valid ( DeBacker, Crowson, Beesley, Thoma, & Hestevold, 2008 ). Even though the measures of epistemic beliefs used in this study demonstrated similar reliability and t to other established measures (e.g., Greene et al., 2010), concerns about measurement cannot be ig- nored in future research. Using a Likert-type scale assumes that one end of a continuum is more adaptive, when in fact it may be that moderate levels of endorsement are ideal (e.g., a balance of beliefs that knowl- edge is justi ed by authority gures whom may also di er in their opinions; Barger et al., 2016 ; Greene et al., 2010). Supplementing surveys with in-depth interviews or more fully revising measures to better re ect beliefs relevant to di erent domains of study might help to address these issues ( Greene & Seung, 2014 ). Researchers might also consider a broader range of beliefs about knowledge (e.g., Chinn et al., 2011 ). For example, this study did not assess beliefs about whether collaboration can create knowledge, which may be particularly im- portant in an interactive-learning course. Future research might also improve on the assessment of perceptions of constructivism, as there are few established measures for students' perceptions of their class- rooms' epistemic climate. Finally, a larger sample size would have been more ideal for our path analyses. Some of our lack of signicant ndings may have been due to

M.M. Barger et al.

decreased power due to our relatively small sample. Future research should include a larger sample to test similar models, which would also enable the testing of a full SEM model rather than the path model and thus would enable us to also account for measurement error. Even so, we de- tected signicant eects despite a relatively small sample.

4.2. Implications

These results provide several preliminary insights for research and practice. First, learners' personal epistemologies do play a role in the classroom environment. As the eld moves towards a contextualized view of epistemic cognition ( Chinn et al., 2011; Sandoval, 2014), this serves as a reminder that individual di erences in students' incoming views of knowledge predict both how they view events in the classroom and how likely they are to succeed in the course. Therefore, individual learner di erences play a role in epistemic thinking in context . For educators who hope to change students' beliefs with innovative teaching techniques, some attempts may not be e ective for some students. It is therefore important to understand students' preexisting beliefs when designing interventions. Teachers might consider in- formally evaluating what students think about discipline-speci c knowledge and explicitly addressing beliefs that are not aligned with the epistemic nature of the course ( Muis et al., 2016 ). Teachers could also go one step further by making epistemic change an explicit com- ponent of their course objectives ( Maggioni & Parkinson, 2008). Using this study's ndings, it is also possible that establishing a complex classroom environment (e.g., asking questions that have more than one answer, encouraging students to understand material in their own way) might be helpful if instructors draw students' attention to why they are setting up the course in this way. These ndings should also be of importance to college educators as they consider how their teaching practices impact students. The debate over lecturing versus student-centered active learning has escalated recently ( Freeman et al., 2014 ; Friesen, 2011 ) and spilled into popular media (e.g., Abrams, 2012 ; Furedi, 2013 ; Worthen, 2015 ). Many re- searchers and college educators emphatically support one approach. Our ndings suggest that singularly championing one instructional approach ignores the role that students own beliefs play in how they interact with the course material. Individual di erences in students' epistemic beliefs are a moderating factor for the impact of the class format on student achievement. We advise against completely rejecting a teaching approach without further consideration of moderating fac- tors, such as students' epistemic beliefs. Finally, this study has implications for the socialization of beliefs from teachers to students. Teachers' beliefs do not necessarily align with their practices ( Maggioni & Parkinson, 2008 ), and the current study suggests that teachers' practices may not even align with students ex- periences. There are many bridges between teachers' beliefs and their students' beliefs about knowledge. Researchers will need to consider ways for educators to bridge these gaps, for example by having teachers consistently re ect on their own epistemic cognition and incorporate this into their classroom practices (e.g., Lunn Brownlee et al., 2017 ). Epistemic climate and personal epistemology development are in- teractive phenomena. Both the classroom structure and students' beliefs about knowledge work together, in tandem, to form an epistemic cli- mate which can bring about epistemic change. This makes interventions targeting epistemic change more challenging, but ultimately opens exciting new doors for how educators might change the way students think about knowledge.

Acknowledgements

The research was supported by a grant from the Duke University, Arts & Sciences Faculty Assessment Committee to Canelas, D., MacPhail, R., & Linnenbrink-Garcia, L. We would like to thank Prof. Keith Whit eld for his support and the nancial contributions from the

Provost's O ce that funded researchers' salaries while the analyses were being conducted. Andrea Novicki and Randy Riddle are thanked for their e orts in leading the 2013 14 Flipped Faculty Scholars group. We also thank Andrea Novicki for careful monitoring of the use of class time as either student-centered or lecturer-centered.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https:// doi.org/10.1016/j.lindif.2018.03.006 .

References

Abrams, M. (2012, September). The death of the lecture. The American Society of Mechanical Engineers. Retrieved from https://www.asme.org/career-education/ articles/undergraduate-students/the-death-of-the-lecture . Alexander, P. A. (2016). The arguments for and the reasoning about epistemic cognition. In J. A. Greene, B. Sandoval, & I. Bråten (Eds.). Handbook of epistemic cognition (pp. 100 110). New York, NY: Routledge . Barger, M. M., & Linnenbrink-Garcia, L. (2017). Developmental systems of students' personal theories about education. Educational Psychologist, 52 (2), 6383. http://dx. doi.org/10.1080/00461520.2016.1252264 . Barger, M. M., Wormington, S. V., Huettel, L. G., & Linnebrink-Garcia, L. (2016). Developmental changes in college engineering students' personal epistemology pro- les. Learning and Individual Dierences . http://dx.doi.org/10.1016/j.lindif.2016.04. 002 (Advance online publication). Bendixen, L. D., & Rule, D. C. (2004). An integrative approach to personal epistemology:

A guiding model. Educational Psychologist, 39, 69 80. http://dx.doi.org/10.1207/ s15326985ep3901_7 . Bråten, I., Anmarkrud, Ø., Brandmo, C., & Strømsø, H. I. (2014). Developing and testing a model of direct and indirect relationships between individual di erences, processing, and multiple-text comprehension. Learning and Instruction, 30, 9 24. http://dx.doi. org/10.1016/j.learninstruc.2013.11.002 . Bråten, I., & Ferguson, L. E. (2014). Investigating cognitive capacity, personality, and epistemic beliefs in relation to science achievement. Learning and Individual Dierences, 36, 124 130. http://dx.doi.org/10.1016/j.lindif.2014.10.003 . Bråten, I., Ferguson, L. E., Strømsø, H. I., & Anmarkrud, Ø. (2013). Justi cation beliefs and multiple-documents comprehension. European Journal of Psychology of Education, 28, 879 902. http://dx.doi.org/10.1007/s 10212-012-0145-2 . Bråten, I., & Strømsø, H. I. (2004). Epistemological beliefs and implicit theories of in- telligence as predictors of achievement goals. Contemporary Educational Psychology, 29, 371 388. http://dx.doi.org/10.1016/j.cedpsych.2003.10.001 . Brownlee, J., Purdie, N., & Boulton-Lewis, G. (2001). Changing epistemological beliefs in preservice teacher education students. Teaching in Higher Education, 6 , 247 268. http://dx.doi.org/10.1080/13562510120045221 . Brownlee, J., Schraw, G., Walker, S., & Ryan, M. (2016). Changes in preservice teachers' personal epistemologies. In J. A. Greene, B. Sandoval, & I. Bråten (Eds.). Handbook of epistemic cognition (pp. 300 317). New York, NY: Routledge . Byrne, B. M. (2012). Structural equation modeling with MPlus: Basic concepts, applications, and programming. Routledge . Canelas, D. A. (2015). Teaching College Chemistry to the Edges Rather than to the Average: Implications for Less Experienced Science Students. In K. Daus, & R. Rigsbee (Eds.). The Promise of Chemical Education: Addressing our Students' Needs (pp. 11 28). Oxford, UK: American Chemical Society Symposium Series 1193, Oxford University Press . Canelas, D. A., Hill, J. L., & Novicki, A. (2017). Cooperative learning in organic chemistry increases student assessment of learning gains in key transferable skills. Chemistry Education Research and Practice, 18, 441 456. http://dx.doi.org/10.1039/ c7rp00014f . Chase, A., Pakhira, D., & Stains, M. (2013). Implementing process-oriented, guided-in- quiry learning for the rst time: Adaptations and short-term impact on students' at- titude and performance. Journal of Chemical Education, 90 , 409416. http://dx.doi. org/10.1021/ed300181t . Chen, J. A., & Barger, M. M. (2016). Epistemic cognition and motivation. In J. A. Greene, W. Sandoval, & I. Bråten (Eds.). Handbook of epistemic cognition (pp. 425 438). New York, NY: Routledge . Chi, M. (2009). Active-constructive-interactive: A conceptual framework for dier- entiating learning activities. Trends in Cognitive Science, 1, 73 105. http://dx.doi.org/ 10.1111/j.1756-8765.2008.01005.x . Chinn, C. A., Buckland, L. A., & Samarapungavan, A. (2011). Expanding the dimensions of epistemic cognition: Arguments from philosophy and psychology. Educational Psychologist, 46, 141 167. http://dx.doi.org/10.1080/00461520.2011.587722 . Conley, A. M., Pintrich, P. R., Vekiri, I., & Harrison, D. (2004). Changes in epistemological beliefs in elementary science students. Contemporary Educational Psychology, 29, 186 204. http://dx.doi.org/10.1016/j.cedpsych.2004.01.004 . Cortina, J. M. (1993). What is coe cient alpha? An examination of theory and applica- tions. Journal of Applied Psychology, 78, 98104 . Dai, T., & Cromley, J. G. (2014). The match matters: Examining student epistemic pre- ferences in relation to epistemic beliefs about chemistry. Contemporary Educational Psychology, 39, 262 274. http://dx.doi.org/10.1016/j.cedpsych.2014.06.002 . DeBacker, T. K., Crowson, H. M., Beesley, A. D., Thoma, S. J., & Hestevold, N. L. (2008). The challenge of measuring epistemic beliefs: An analysis of three self-report

M.M. Barger et al.

instruments. The Journal of Experimental Education, 76, 281 312. http://dx.doi.org/ 10.3200/JEXE.76.3.281-314 . Driver, R., Asoko, H., Leach, J., Scott, P., & Mortimer, E. (1994). Constructing scientic knowledge in the classroom. Educational Researcher, 23, 512 . Farrell, J. J., Moog, R. S., & Spencer, J. N. (1999). A guided inquiry general chemistry course. Journal of Chemical Education, 76, 570 574. http://dx.doi.org/10.1021/ ed076p570 . Fautch, J. M. (2015). The ipped classroom for teaching organic chemistry in small classes: Is it eective? Chemistry Education Research and Practice, 16 , 179 186. http:// dx.doi.org/10.1039/C4RP00230J . Ferguson, L. E., & Bråten, I. (2013). Student proles of knowledge and epistemic beliefs:

Changes and relations to multiple-text comprehension. Learning and Instruction, 25, 49 61. http://dx.doi.org/10.1016/j.learninstruc.2012.11.003 . Ferguson, L. E., Bråten, I., & Strømsø, H. I. (2012). Epistemic cognition when students read multiple documents containing conicting scientic evidence: A think-aloud study. Learning and Instruction, 22, 103 120. http://dx.doi.org/10.1016/j. learninstruc.2011.08.002 . Feucht, F. C. (2010). Epistemic climate in elementary classrooms. In L. D. Bendixen, & F. C. Feucht (Eds.). Personal epistemology in the classroom: Theory, research, and educa- tional implications (pp. 5593). Cambridge, MA: Cambridge University Press . Freeman, S., Eddy, S. L., McDonough, M., Smith, M. K., Okoroafor, N., Jordt, H., & Wenderoth, M. P. (2014). Active learning increases student performance in science, engineering, and mathematics. Proceedings of the National Academy of Sciences, 111 , 8410 8415. http://dx.doi.org/10.1073/pnas.1319030111 . Friesen, N. (2011). The lecture as a transmedial pedagogical form: A historical analysis. Educational Researcher, 40, 95 102. http://dx.doi.org/10.3102/0013189X11404603 . Fruge, C. W., & Ropers-Huilman, R. (2008). Epistemological congruency in community college classrooms: E ects of epistemological beliefs on students' experiences. College

Teaching, 56, 121 127. http://dx.doi.org/10.3200/CTCH.56.2.121-128 . Furedi, F. (2013, December 10). In praise of the university lecture and its place in academic scholarship. The Guardian. Retrieved from http://www.theguardian.com/higher- education-network/blog/2013/dec/10/in-praise-of-academic-lecture . Gill, M. G., Ashton, P. T., & Algina, J. (2004). Changing preservice teachers' epistemo- logical beliefs about teaching and learning in mathematics: An intervention study. Contemporary Educational Psychology, 29, 164 185. http://dx.doi.org/10.1016/j. cedpsych.2004.01.003 . Greene, J. A., Azevedo, R., & Torney-Purta, J. (2008). Modeling epistemic and ontological cognition: Philosophical perspectives and methodological directions. Educational Psychologist, 43, 142 160. http://dx.doi.org/10.1080/00461520802178458 . Greene, J. A., & Seung, B. Y. (2014). Modeling and measuring epistemic cognition: A qualitative re-investigation. Contemporary Educational Psychology, 39, 12 28. http:// dx.doi.org/10.1016/j.cedpsych.2013.10.002 . Greene, J. A., Torney-Purta, J., & Azevedo, R. (2010). Empirical evidence regarding re- lations among a model of epistemic and ontological cognition, academic perfor- mance, and educational level. Journal of Educational Psychology, 102, 234 255. http://dx.doi.org/10.1037/a0017998 . Goldwasser, M., Mosley, P. L., & Canelas, D. A. (2016). Implementation of online lecture videos in introductory chemistry. Online Course Development and the E ect on the On- Campus Classroom (pp. 63 73). American Chemical Society . Hall, D. M., Curtin-Soydan, A. J., & Canelas, D. A. (2014). The Science Advancement through Group Engagement Program: Leveling the Playing Field and Increasing Retention in Science. Journal of Chemical Education, 91(1), 3747. http://dx.doi.org/ 10.1021/ed400075n . Hein, S. M. (2012). Positive impacts using POGIL in organic chemistry. Journal of Chemical Education, 89, 860 864. http://dx.doi.org/10.1021/ed100217v . Hmelo-Silver, C. E., Duncan, R. G., & Chinn, C. A. (2007). Sca olding and achievement in problem-based and inquiry learning: A response to Kirschner, Sweller, and Clark (2006). Educational Psychologist, 42, 99 107. http://dx.doi.org/10.1080/

Hofer, B. K. (2000). Dimensionality and disciplinary dierences in personal epistemology. Contemporary Educational Psychology, 25, 378 405. http://dx.doi.org/10.1006/ceps.

1999.1026 . Hofer, B. K. (2004). Exploring the dimensions of personal epistemology in diering classroom contexts: Student interpretations during the rst year of college. Contemporary Educational Psychology, 29(2), 129163. http://dx.doi.org/10.1016/j. cedpsych.2004.01.002 . Hofer, B. K., & Pintrich, P. R. (1997). The development of epistemological theories: Beliefs about knowledge and knowing and their relation to learning. Review of Educational Research, 67, 88 140. http://dx.doi.org/10.3102/00346543067001088 .

A Multidisciplinary Journal, 6 , 155. http://dx.doi.org/10.1080/

10705519909540118 . Jensen, J. L., & Lawson, A. (2011). Eects of collaborative group composition and inquiry instruction on reasoning gains and achievement in undergraduate biology. CBE Life Sciences Education, 10, 6473. http://dx.doi.org/10.1187/cbe.10-07-0089 . Jonassen, D. H. (1991). Objectivism versus constructivism: Do we need a new

philosophical paradigm? Educational Technology Research and Development, 39, 514 . Kahle, J. B., Meece, J., & Scantlebury, K. (2000). Urban African-American middle school science students: Does standards-based teaching make a di erence? Journal of Research in Science Teaching, 37, 1019 1041. http://dx.doi.org/10.1002/1098- 2736(200011)37:9<1019::AID-TEA9>3.0.CO;2-J . Keith, T. Z. (2015). Multiple regression and beyond: An introduction to multiple regression and structural equation modeling (2nd ed.). Routledge . Kienhues, D., Bromme, R., & Stahl, E. (2008). Changing epistemological beliefs: The unexpected impact of a short-term intervention. British Journal of Educational Psychology, 78, 545 565. http://dx.doi.org/10.1348/000709907X268589 . Kienhues, D., Ferguson, L., & Stahl, E. (2016). Diverging information and epistemic change. In J. A. Greene, B. Sandoval, & I. Bråten (Eds.). Handbook of epistemic cog- nition (pp. 318 330). New York, NY: Routledge . Kirschner, P. A., Sweller, J., & Clark, R. E. (2006). Why minimal guidance during in- struction does not work: An analysis of the failure of constructivist, discovery, pro- blem-based, experiential, and inquiry-based teaching. Educational Psychologist, 41, 75 86. http://dx.doi.org/10.1207/s15326985ep4102_1 . Kitchener, K. S., & King, P. M. (1981). Reective judgment: Concepts of justi cation and their relationship to age and education. Journal of Applied Developmental Psychology, 2 , 89 116. http://dx.doi.org/10.1016/0193-3973(81)90032-0 . Loudon, G. M. (2009). Organic chemistry (5th ed.). Englewood, CO: Roberts and Company Publishers . Lunn Brownlee, J., Ferguson, L. E., & Ryan, M. (2017). Changing teachers' epistemic cognition: A new conceptual framework for epistemic reexivity. Educational Psychologist . http://dx.doi.org/10.1080/00461520.2017.1333430 (Advance online publication). Maggioni, L., & Parkinson, M. M. (2008). The role of teacher epistemic cognition, epis- temic beliefs, and calibration in instruction. Educational Psychology Review, 20, 445 461. http://dx.doi.org/10.1007/s10648-008-9081-8 . Moog, R. S., & Spencer, J. N. (2008). POGIL: Process oriented guided inquiry learning. New York: Oxford University Press . Mueller, R. O., & Hancock, G. R. (2008). Best practices in structural equation modeling. In J. W. Osborne (Ed.). Best practices in quantitative methods (pp. 488 508). New York:

Sage Publications . Muis, K. R. (2004). Personal epistemology and mathematics: A critical review and synthesis of research. Review of Educational Research, 74, 317 377. http://dx.doi.org/ 10.3102/00346543074003317 . Muis, K. R. (2007). The role of epistemic beliefs in self-regulated learning. Educational Psychologist, 42, 173 190. http://dx.doi.org/10.1080/00461520701416306 . Muis, K. R., Bendixen, L. D., & Haerle, F. C. (2006). Domain-generality and domain- speci city in personal epistemology research: Philosophical and empirical reections in the development of a theoretical framework. Educational Psychology Review, 18, 354. http://dx.doi.org/10.1007/s10648-006-9003-6 . Muis, K. R., & Duy, M. C. (2013). Epistemic climate and epistemic change: Instruction designed to change students' beliefs and learning strategies and improve achieve- ment. Journal of Educational Psychology, 105 , 213 225. http://dx.doi.org/10.1037/ a0029690 . Muis, K. R., & Franco, G. M. (2009). Epistemic beliefs: Setting the standards for self- regulated learning. Contemporary Educational Psychology, 34, 306 318. http://dx.doi. org/10.1016/j.cedpsych.2009.06.005 . Muis, K. R., Trevors, G., & Chevrie, M. (2016). Epistemic climate for epistemic change. In J. A. Greene, B. Sandoval, & I. Bråten (Eds.). Handbook of epistemic cognition (pp. 331 359). New York, NY: Routledge . Niaz, M., & Robinson, W. R. (1992). Manipulation of logical structure of chemistry pro- blems and its eect on student performance. Journal of Research in Science Teaching, 29, 211 226. http://dx.doi.org/10.1002/tea.3660290303 . Osborne, J. (2010). Arguing to learn in science: The role of collaborative, critical dis- course. Science, 328 , 463 466. http://dx.doi.org/10.1126/science.1183944 . Pekrun, R., Hall, N. C., Goetz, T., & Perry, R. P. (2014). Boredom and academic achievement: Testing a model of reciprocal causation. Journal of Educational Psychology, 106 , 696710. http://dx.doi.org/10.1037/a0036006 . Perry, W. G. (1970). Forms of intellectual and ethical development in the college years; A scheme. New York: Holt, Rinehart and Winston . Posner, G. J., Strike, K. A., Hewson, P. W., & Gertzog, W. A. (1982). Accommodation of a scienti c conception: Toward a theory of conceptual change. Science Education, 66,

Qian, G., & Alvermann, D. (1995). Role of epistemological beliefs and learned help- lessness in secondary school students' learning science concepts from text. Journal of Educational Psychology, 87, 282 292. http://dx.doi.org/10.1037/0022-0663.87.2.

282 . Raykov, T. (2005). Analysis of longitudinal studies with missing data using covariance structure modeling with full-information maximum likelihood. Structural Equation Modeling, 12 , 493505. http://dx.doi.org/10.1207/s15328007sem1203_8 . Roth, G., & Weinstock, M. (2013). Teachers' epistemological beliefs as an antecedent of autonomy-supportive teaching. Motivation and Emotion, 37 , 402 412. http://dx.doi. org/10.1007/s11031-012-9338-x . Ruder, S. (2015). Organic chemistry: A guided inquiry. New York: Wiley . Ruiz-Primo, M. A., Briggs, D., Iverson, H., Talbot, R., & Shepard, L. A. (2011). Impact of undergraduate science course innovations on learning. Science, 331 , 1269 1270. http://dx.doi.org/10.1126/science.1198976 . Sandoval, W. (2014). Science education's need for a theory of epistemological develop- ment. Science Education, 98 , 383387. http://dx.doi.org/10.1002/sce.21107 . Savery, J. R., & Duy, T. M. (2001). Problem based learning: An instructional model and its constructivist framework. Indiana University: Center for Research on Learning and Technology1 17 . Schmitt, N. (1996). Uses and abuses of coecient alpha. Psychological Assessment, 8 ,

M.M. Barger et al.

Schommer, M., Calvert, C., Gariglietti, G., & Bajaj, A. (1997). The development of epis- temological beliefs among secondary students: A longitudinal study. Journal of Educational Psychology, 89, 37 40. http://dx.doi.org/10.1037/0022-0663.89.1.37 . Siburt, C. J. P., Bissell, A. N., & MacPhail, R. A. (2011). Developing metacognitive and problem-solving skills through problem manipulation. Journal of Chemical Education, 88, 1489 1495. http://dx.doi.org/10.1021/ed100891s . Springer, L., Stanne, M. E., & Donovan, S. S. (1999). Eects of small-group learning on undergraduates in science, mathematics, engineering, and technology: A meta-ana- lysis. Review of Educational Research, 69, 21 51. Straumanis, A. (2009). Organic chemistry: A guided inquiry. New York: Houghton Mi in Harcourt . Strømsø, H. I., & Kammerer, Y. (2016). Epistemic cognition and reading for under- standing in the internet age. In J. A. Greene, B. Sandoval, & I. Bråten (Eds.). Handbook of epistemic cognition (pp. 230 246). New York, NY: Routledge . Teo, T. W., Tan, K. C. D., Yan, Y. K., Teo, Y. C., & Yeo, L. W. (2014). How ip teaching supports undergraduate chemistry laboratory learning. Chemistry Education Research and Practice, 15, 550 567. http://dx.doi.org/10.1039/C4RP00003J .

Tolhurst, D. (2007). The in uence of learning environments on students' epistemological beliefs and learning outcomes. Teaching in Higher Education, 12, 219 233. http://dx. doi.org/10.1080/13562510701191992 . Trautwein, U., & Lüdtke, O. (2007). Epistemological beliefs, school achievement, and college major: A large-scale longitudinal study on the impact of certainty beliefs. Contemporary Educational Psychology, 32, 348 366. http://dx.doi.org/10.1016/j. cedpsych.2005.11.003 . Trautwein, U., Marsh, H. W., Nagengast, B., Lüdtke, O., Nagy, G., & Jonkmann, K. (2012). Probing for the multiplicative term in modern expectancy value theory: A latent interaction modeling study. Journal of Educational Psychology, 104, 763 777. http:// dx.doi.org/10.1037/a0027470 . Tsai, C.-C. (2000). Relationships between student scientic epistemological beliefs and perceptions of constructivist learning environments. Educational Research, 42, 193 205. http://dx.doi.org/10.1080/001318800363836 . Worthen, M. (2015, October 17). Lecture me. Really. The New York Times. Retrieved from

r=1 .