38
THE CONDITIONAL AND INTERACTION EFFECTS OF EPISTEMOLOGICAL BELIEFS ON THE SELF-REGULATED LEARNING OF COLLEGE STUDENTS: Motivational Strategies Michael B. Paulsen* , and Kenneth A. Feldman** ................................................................................................ ................................................................................................ This study examines the conditional and interaction effects of each of four dimensions of the epistemological beliefs of college students regarding the ability to learn, the speed of learning, the structure of knowledge, and the stability of knowledge on six measures of the motivational components of self-regulated learning strategies (intrinsic goal orientations, extrinisic goal orientation, task value, self-efficacy, control of learning and test anxiety). Students with more sophisticated beliefs about the nature of knowledge and learning were more likely than their peers to use educationally productive motivational strategies in their learning. Beliefs about one’s ability to learn and the structure of knowledge had the most significant and substantial effects on students’ use of self-regulated motivational strategies. Although a student’s belief about the stability of knowledge by itself had a statistically significant effect on only one motivational strategy, this belief did have four statistically significant interaction effects with beliefs about ability to learn and the structure of knowledge. Implications of these findings for theory, research, policy and practice are examined. ................................................................................................ ................................................................................................ KEY WORDS: epistemological beliefs; self-regulated learning; student motivation. Increased attention on the learner-centered classroom—as a founda- tion for the design of learning environments and reform of teaching practice in higher education—has stimulated the need for deeper under- standing of how college students learn (Cross and Steadman, 1996; *Department of Educational Policy & Leadership Studies, The University of Iowa, Iowa City, IA. **Department of Sociology, State University of New York at Stony Brook, Stony Brook, NY.  Address correspondence to: Michael B. Paulsen, Professor of Higher Education, Educational Policy & Leadership Studies, College of Education, The University of Iowa, N491 Lindquist Center North, 52242-1529, Iowa City, IA. E-mail: michael-paulsen@ uiowa.edu 731 0361-0365 05 1100-0731 0 Ó 2005 Springer Science+Business Media, Inc. Research in Higher Education, Vol. 46, No. 7, November 2005 (Ó 2005) DOI: 10.1007/s11162-004-6224-8

The Conditional and Interaction Effects of Epistemological Beliefs on the Self-Regulated Learning of College Students: Motivational Strategies

Embed Size (px)

Citation preview

  • THE CONDITIONAL AND INTERACTION EFFECTSOF EPISTEMOLOGICAL BELIEFS ON THESELF-REGULATED LEARNING OF COLLEGESTUDENTS: Motivational Strategies

    Michael B. Paulsen*, and Kenneth A. Feldman**

    ................................................................................................................................................................................................

    This study examines the conditional and interaction effects of each of fourdimensions of the epistemological beliefs of college students regarding the ability tolearn, the speed of learning, the structure of knowledge, and the stability ofknowledge on six measures of the motivational components of self-regulatedlearning strategies (intrinsic goal orientations, extrinisic goal orientation, task value,self-efficacy, control of learning and test anxiety). Students with more sophisticatedbeliefs about the nature of knowledge and learning were more likely than theirpeers to use educationally productive motivational strategies in their learning.Beliefs about ones ability to learn and the structure of knowledge had the mostsignificant and substantial effects on students use of self-regulated motivationalstrategies. Although a students belief about the stability of knowledge by itself hada statistically significant effect on only one motivational strategy, this belief did havefour statistically significant interaction effects with beliefs about ability to learn andthe structure of knowledge. Implications of these findings for theory, research,policy and practice are examined.

    ................................................................................................................................................................................................KEY WORDS: epistemological beliefs; self-regulated learning; student motivation.

    Increased attention on the learner-centered classroomas a founda-tion for the design of learning environments and reform of teachingpractice in higher educationhas stimulated the need for deeper under-standing of how college students learn (Cross and Steadman, 1996;

    *Department of Educational Policy & Leadership Studies, The University of Iowa, Iowa

    City, IA.

    **Department of Sociology, State University of New York at Stony Brook, Stony Brook,

    NY.Address correspondence to: Michael B. Paulsen, Professor of Higher Education,

    Educational Policy & Leadership Studies, College of Education, The University of Iowa,N491 Lindquist Center North, 52242-1529, Iowa City, IA. E-mail: [email protected]

    731

    0361-0365 05 1100-0731 0 2005 Springer Science+Business Media, Inc.

    Research in Higher Education, Vol. 46, No. 7, November 2005 ( 2005)DOI: 10.1007/s11162-004-6224-8

  • Paulsen, 1999; Stage, Muller, Kinzie, and Simmons, 1998; Svinicki, 2004;Weimer, 2002). The social-cognitive mediation model of college studentlearning developed by McKeachie, Pintrich and their colleagues (McKeachie,Pintrich, Lin, Smith, and Sharma, 1990; Pintrich, 1989; Pintrich andZusho, 2002) has emerged as a particularly productive framework forexamining various factors that inuence meaningful learning in the col-lege classroom. This model assumes that the eects on learning outcomesof the prior background of students and classroom context are mediatedby the motivational, cognitive and behavioral strategies students use intheir learning, including their self-regulation of these strategies.Within this framework, self-regulated learning is viewed as a process

    in which individual students actively and constructively monitor andcontrol their own motivation, cognition and behavior toward the suc-cessful completion of academic tasks (Butler and Winne, 1995; DuBoisand Staley, 1997; Pintrich, 1995; Valle et al., 2003; Winne, 1995;Zimmerman, 1994). Research in this area has consistently demonstratedthat students use of eective motivational, cognitive and behavioralprocesses and strategies is directly related to students academic achieve-ment (Paulsen and Gentry, 1995; Pintrich and Garcia, 1991; Pintrichand Schrauben, 1992; Pintrich and Zusho, 2002). A relatively new, butvery promising, development in research on the student mediationmodel of learning is based on investigating how dierences in epistemo-logical beliefs of students might help explain variation in their use andself-regulation of their motivational, cognitive, and behavioral processesand strategies for learning (Hofer and Pintrich, 1997; Paulsen andFeldman, 1999a, 1999b; Pintrich, 2002; Schommer, 2002).Educational psychologists typically have viewed epistemological beliefs

    as systems of implicit assumptions and beliefs held by students about thenature of knowledge and its acquisition (Bruning, Schraw, and Ronning,1999; Hofer and Pintrich, 1997). Most traditional studies of the episte-mological beliefs of college students are generally embedded indevelopmental perspectives. Such studies have viewed students as devel-opmentally advancing through a sequence of positions or stages in theirepistemological beliefs. For the most part, these studies see students asprogressing along a primarily unidimensional continuum from dualisticbeliefs to more relativistic beliefs about the nature of knowledge andlearning (Baxter Magolda, 1992; Belenky, Clinchy, Goldberger, andTarule, 1986; Kitchener and King, 1981; Perry, 1970).In the 1990s, Schommer (1990, 1994a, 1994b; Schommer-Aikins,

    2004) moved beyond the traditional unidimensional approaches to theconceptualization and measurement of epistemological beliefs, proposingan expanded view that re-conceptualized the construct as a multidimen-

    732 PAULSEN AND FELDMAN

  • sional system of beliefs about the nature of knowledge and learning.This multidimensional system comprises beliefs about the structure ofknowledge, the stability of knowledge, the speed of knowledge acquisi-tion and the control of knowledge acquisition. Schommer developed thismultidimensional approach through her eorts to integrate the work ofa number of scholars who had worked independently of one anotherwhile they investigated dierent dimensions of epistemological beliefs(e.g., Dweck and Leggett, 1988; Perry, 1970; Schoenfeld, 1983). As ithappens, her multidimensional perspective on the conceptualization andmeasurement of epistemological beliefs has helped stimulate and redirectresearch in the eld (e.g., see Hofer and Pintrich, 2002).Research has demonstrated that the origins and development of episte-

    mological beliefs are formed and shaped by various pre-college factors(Unger, Draper, and Pendergrass, 1986), such as background character-istics like age and gender (Paulsen and Wells, 1998; Schommer, 1990),parental education and socioeconomic status (Schommer, 1990, 1993a),characteristics of upbringing in early home and family environments,including encouragement of independence in decisions and discussions orstrictness of adherence to rules and guidelines (Schommer, 1990, 1993a),and pre-college schooling experiences (Schommer, 1993b; Schommer,Calvert, Garigliette, and Bajaj, 1997). Yet, research also indicates thatcollege-related experiences can impact students epistemological beliefs aswell (Hofer, 2000; Jehng, Johnson, and Anderson, 1993; Kardash andHowell, 2000; Paulsen and Wells, 1998; Schommer, 1993a).The development and expansion of theories of epistemology have par-

    alleled the growing research in this area. An increasingly prominentview in the literature is that the epistemological beliefs of college stu-dents play a potentially critical and central role in facilitating or con-straining the eectiveness of students self-regulated learning. Morespecically, reviews of recent research and emerging theory have con-cluded that students beliefs about the nature of knowledge and learningcan function as standards, criteria, implicit theories, or cognitive lensesthrough which students (1) assess academic tasks and task-related per-formance or learning, (2) monitor, select, generate, and interpret inter-nal feedback regarding what constitutes appropriate and eectiveapproaches to such tasks, and (3) develop motivational, cognitive, andbehavioral strategies for eective self-regulated learning (Butler andWinne, 1995; Hofer and Pintrich, 1997; Pintrich, 2002; Ryan, 1984;Schommer-Aikins, 2002; Winne, 1995).Showing steady support for these propositions, research has consis-

    tently demonstrated signicant relationships between students epistemo-logical beliefs and a variety of learning outcomes including academic

    EPISTEMOLOGICAL BELIEFS AND MOTIVATIONAL STRATEGIES 733

  • achievement (Bruning et al., 1999; Hofer, 1999, 2000; Hofer andPintrich, 1997; Ryan, 1984; Schommer, 1994a, 1994b; Schommer-Aikins,2002, 2004; Schutz, Pintrich, and Young, 1993; Wood and Kardash,2002). In addition, several initial studies in the area have already pro-vided preliminary evidence that students epistemological beliefs may berelated to their engagement in either the motivational, cognitive orbehavioral aspects of self-regulated learning, which constitute importantaspects of learning. For example, Schutz et al. (1993) found that a one-dimensional measure of epistemological beliefs was related to studentsuse of two types of cognitive learning strategies and one type of motiva-tional strategy. Valle et al. (2003) found that a one-dimensional measureof epistemological beliefs about learning was related to students use oftwo types of motivational strategies. Schommer, Crouse, and Rhoads(1992) showed that the eect of one measure of epistemological beliefsabout learning was mediated by students use of one type of cognitivelearning strategy, while Kardash and Howell (2000) found that severaldimensions of the epistemological beliefs of students were related totheir use of several types of cognitive strategies. Finally, Hofer (1999)showed that students beliefs about the nature of mathematics knowl-edge were related to one type of cognitive strategy and two types ofmotivational strategies.None of these studies, however, has performed a comprehensive

    examination of the relationships between a wide range of students useof multiple strategies from all three of the primary categories of self-reg-ulated learning strategiesmotivational, cognitive, and behavioralandseparate measures of each of the distinct components of a well-devel-oped, multidimensional system of students beliefs about the nature ofknowledge and learning. Earlier studies have tended to focus on sam-ples of students taking classes in only one or two specic elds, such aspsychology or mathematics, rather than examining students takingcourses across a wide range of elds spanning the humanities, social sci-ences and natural sciences.1 Additionally, initial studies have providedpreliminary evidence about only the main eects of epistemologicalbeliefs on students self-regulation, thereby ignoring the potential condi-tional and interaction eects of epistemological beliefs on self-regula-tion. Our own research program has been designed to address thesegaps in the emerging research literature. Our earlier and preliminarystudies, which were based on the experiences of students taking under-graduate courses in education, examined four distinct components of amultidimensional system of epistemological beliefs and their relation-ships with six measures of motivational learning strategies (Paulsen andFeldman, 1999b), four measures of cognitive learning strategies and

    734 PAULSEN AND FELDMAN

  • four measures of behavioral strategies (Paulsen and Feldman, 1999a).In these two studies, two dimensions of epistemological beliefs weresignicantly related to more than half of the motivational, cognitive andbehavioral learning strategies examined, a third dimension was relatedto about one-third of all strategies, and a fourth measure was unrelatedto such strategies.The present studyone in a planned set of analysesfurther ad-

    vances progress in this line of inquiry by using a sample more thantwice as large as in our previous studies and by examining students tak-ing classes in a variety of disciplines representing the humanities, natu-ral sciences, social sciences and education. It examines both theconditional and interaction eects of four dimensions of epistemologicalbeliefs on motivational learning strategies onlybecause of the com-plexity of the analysis and consequent space considerations. The eectsof epistemological beliefs on cognitive and behavioral learning strategiesare analyzed in a separate article as a companion piece to the presentreport.

    METHODS

    The Sample

    The subjects for this study were 502 students at a large urban publicuniversity. They were enrolled in 16 undergraduate classes oered dur-ing the spring semester of 1996 from a diverse set of elds including thehumanities (philosophy and English), the natural sciences (biology andgeology), the social sciences (psychology and history), and education.About 61% of the students were women and 39% were men; 70.3%were between 17 and 24 years old and 28.7% were 25 years or older. Inthe sample, 51.4% were lower-division undergraduate students, 41.8%were upper-division undergraduates, and 6.8% were graduate students,but all were surveyed while taking undergraduate courses. Finally, 9.6%of the students in the sample were African American, 3.8% were AsianAmerican, 75.5% were Caucasian, 6.2% were Hispanic, and about 5%were from Native American or other backgrounds.

    The Variables and their Measurement

    This study examines the relations between four measures of epistemo-logical beliefs (independent variables) and six measures of motivationallearning strategies (dependent variables) college students use in their

    EPISTEMOLOGICAL BELIEFS AND MOTIVATIONAL STRATEGIES 735

  • courses. The nature and measurement of these variables are explained inthe sections below.

    Epistemological Beliefs

    We assessed students epistemological beliefs using the EpistemologicalQuestionnaire (EQ) developed by Schommer (1990). This 64-item ques-tionnaire, which is the written instrumentation for measuring personalepistemology that currently has the most widespread use (Hofer, 2002,p. 10), has often been used in the study of college students epistemologi-cal beliefs (Bendixen, Dunkle, and Schraw, 1994; Kardash and Howell,2000; Kardash and Scholes, 1996; Paulsen and Feldman, 1999a, 1999b;Paulsen and Wells, 1998; Schommer, 1990, 1993a; Schommer-Aikins,2004; Schommer-Aikins, Duell, and Barker, 2003; Schommer et al.,1992).The EQ was administered to all 502 students in the sample during the

    second half of the spring semester of 1996. Students responded to eachitem by rating it on a 5-point Likert scale. The items on the question-naire were designed to measure students epistemological beliefs on fourdistinct and largely independent dimensions: simple knowledge; certainknowledge; xed ability; and quick learning. The rst two dimensionsmeasure students beliefs about the nature of knowledgein particular,its structure and its stability. Scores on the scale for simple knowledgevary along a continuum ranging from beliefs that knowledge is orga-nized as isolated bits and pieces to knowledge is organized as highlyinterwoven concepts (Schommer, 1994a, p. 28). Scores on the scale forcertain knowledge vary along a continuum ranging from beliefs thatknowledge is absolute to knowledge is tentative (p. 28). The last twodimensions measure students beliefs about the nature of learninginparticular, the ability to learn and the speed of learning. Scores on thescale for xed ability vary along a continuum ranging from beliefs thatthe ability to learn is xed at birth to the ability to learn can bechanged (p. 28). Scores on the scale for quick learning vary along acontinuum ranging from beliefs that knowledge is acquired quickly ornot-at-all to knowledge is acquired gradually (p. 28). Each of thesefour continua of epistemological beliefs is anchored by a more navebelief at one end and a more sophisticated belief at the other end.Following Schommers (1990) practice, higher scores on each of thesescales correspond to more nave beliefs, while lower scores representmore sophisticated beliefs. In similar fashion, the labels for each dimen-sion are expressed in terms of the nave form of each beliefthat is,nave beliefs that knowledge is simple in structure and certain or

    736 PAULSEN AND FELDMAN

  • stable over time, that learning takes place quickly or not at all, andthat individuals have a xed ability to learn.The psychometric validity and reliability of the EQ have been estab-

    lished previously in a number of ways. Its underlying four-factor struc-ture has been consistently replicated with dierent samples of collegestudents (Bendixen et al., 1994; Schommer, 1990; Schommer et al.,1992), adults with high school, college, or graduate school levels of edu-cation (Schommer, 1998), and high school students (Schommer, 1993b;Schommer and Dunnell, 1994). The underlying factor structure has beensupported by both exploratory and conrmatory factor analyses(Schommer et al., 1992). The inter-item reliabilities for the sets of itemsconstituting each underlying factor range in magnitude from .63 to .85and tests have revealed an eight-week testretest reliability of .70(Schommer et al., 1997). Finally, a series of studies has demonstratedthe instruments ability to predict a variety of meaningful learning out-comes, such as students motivation to learn, use of cognitive andbehavioral learning strategies, comprehension, metacomprehension, testperformance, and level of complexity of reective judgment (Bendixenet al., 1994; Paulsen and Feldman, 1999a, 1999b; Schommer, 1990;Schommer et al., 1992; Schommer and Walker, 1995). Although therehave been some recent eorts to develop alternatives to SchommersEQ, these remain works in progress with no one highly-regarded alter-native yet available (e.g., see Schraw, Bendixen, and Dunkle, 2002;Wood and Kardash, 2002). Therefore, to make the ndings of the pres-ent study comparable with prior work in our own ongoing researchprogram on epistemological beliefs and self-regulated learning, we con-tinued to use Schommers EQ to measure epistemological beliefs in thepresent study. As in our previous work, students scores on each dimen-sion of belief were computed by applying factor scoring coecientsdeveloped by Schommer for use with college students to the actualmeans and standard deviation measures derived from the sample exam-ined in the present study.2

    Self-regulated Learning Strategies

    We assessed students use of self-regulated learning strategies in theirclasses using the Motivated Strategies for Learning Questionnaire(MSLQ) developed by Pintrich and colleagues (Pintrich, Smith, Garcia,and McKeachie, 1991). The items on the questionnaire were designed tomeasure students use of three sets of self-regulated learning strate-giesmotivational, cognitive, and behavioral. The 81-item survey wasadministered to all 502 students during the second half of the spring

    EPISTEMOLOGICAL BELIEFS AND MOTIVATIONAL STRATEGIES 737

  • semester of 1996. Each respondent rated themselves on their use of spe-cic strategies along a 7-point Likert scale, ranging from not at all trueof me to very true of me. Students reported on their attitudes,thoughts and behaviors regarding the specic academic tasks theyencountered in the context of the particular classes in which they com-pleted the MSLQ. The psychometric validity and reliability of theMSLQ have been well established previously through multiple adminis-trations in small and large classes, in a wide variety of subjectareas, and at dierent types of institutions (Paulsen and Feldman,1999a, 1999b; Paulsen and Gentry, 1995; Pintrich, 1989; Pintrich andGarcia, 1991; Pintrich and Zusho, 2002), and through conrmatory fac-tor analysis (Pintrich, Smith, Garcia, and McKeachie, 1993). In addi-tion, we calculated internal reliability coecients (Cronbachalphas)which ranged from .64 to .90for each of the six scales usedin the present study, as follows.Thirty one of the 81 survey items assessed the extent to which stu-

    dents used each of six distinct types of motivational learning strate-gies. The four items of the intrinsic goal orientation scale (a=.71)assessed the extent to which students perceived themselves to beengaged in academic tasks to pursue internal rewards like mastery orseeking to be challenged (e.g., In a class like this, I prefer course mate-rial that really challenges me so I can learn new things). The fouritems of the extrinsic goal orientation scale (a=.64) assessed the extentto which students perceived themselves to be engaged in academic tasksto pursue external rewards like approval from others or getting goodgrades (e.g., Getting a good grade in this class is the most satisfyingthing for me right now). The six items of the task value scale (a=.90)measured students assessments regarding how interesting, important, oruseful they perceived the course to be (e.g., I think I will be able to usewhat I learn in this course in other courses). The four items of thecontrol of learning scale (a=.68) measured the extent to which studentsbelieved that their academic performance was dependent on factors theycontrolled, such as the amount of their study or eort (e.g., If I tryhard enough, then I will understand the course material). The eightitems of the self-ecacy scale (a=.90) measured the extent to which stu-dents believed that they were competent in terms of task-related abilitiesand skills and had a high likelihood of a successful academic perfor-mance (e.g., Considering the diculty of this course, the teacher, andmy skills, I think I will do well in this class). The ve items of the testanxiety scale (a=.80) assessed the extent to which students experienceddiscomfort or had negative thoughts that could interfere with their test

    738 PAULSEN AND FELDMAN

  • performance (e.g., When I take a test, I think about items on otherparts of the test I cant answer).

    The Models and their Estimation

    This study uses multiple regression analysis to estimate the condi-tional and interaction eects of a set of four measures of epistemologi-cal beliefs (independent variables) on six measures of students use ofmotivational learning strategies (dependent variables) in their collegecoursework. Because these independent variables and dependent vari-ables are continuous measures, the use of conventional analysis of vari-ance approaches would be problematic. The arbitrary dichotomization(or some other form of categorization) of independent variablesrequired by ANOVA-based procedures (see e.g., Cohen and Cohen,1983) reduces cross-student variation and statistical power in key vari-ables. Therefore, multiple regression was used to capitalize on all thecross-student variation in variables available with the continuous mea-sures employed. Additionally, all dependent and independent variableswere standardizedbefore any interaction terms were createdso thatthe estimates of regression coecients would measure eect sizes instandard deviation units.

    Deriving the Full Two-Way Interaction, Product-Term Model

    Equation (1) below represents the general form of a conventionalmain-eects model estimated using ordinary least squares regression. Itprovides a useful starting point for the derivation of the full two-way inter-action, product-term model used in this study to estimate the conditionaland interaction eects of epistemological beliefs on students use of moti-vational learning strategies. Equation (1) includes the dependent vari-ablea particular motivational learning strategyand four independentvariables, each corresponding to one of the measures of four distinctdimensions of epistemological beliefs examined in this study.

    Sij a b1E1i b2E2i b3E3i b4E4i eij 1where i=1,..., n students; j=1,..., J motivational learning strategy vari-ables; and

    Sij =value of the jth motivational learning strategy variable for theith student;

    E1i=value of the rst epistemological belief variable in the model forthe ith student;

    EPISTEMOLOGICAL BELIEFS AND MOTIVATIONAL STRATEGIES 739

  • E2i=value of the second epistemological belief variable in the modelfor the ith student;

    E3i=value of the third epistemological belief variable in the model forthe ith student;

    E4i=value of the fourth epistemological belief variable in the modelfor the ith student;

    a=the intercept term for the model;b1=estimated eect of E1 (the 1st epistemological belief variable in

    model) on S (jth motivational learning strategy variable in themodel);

    b2=estimated eect of E2 (the 2nd epistemological belief variables inmodel) on S (jth motivational learning strategy variable in themodel);

    b3=estimated eect of E3 (the 3rd epistemological belief variables inmodel) on S (jth motivational learning strategy variable in themodel);

    b4=estimated eect of E4 (the 4th epistemological belief variables inmodel) on S (jth motivational learning strategy variable in the model);

    eij=the residual or error term for the equation, which is assumed tomeet the standard assumptions for OLS regression: its values areindependently and normally distributed, have a mean of zero, and avariance that is unrelated to one or more of the independent variables.

    Next, the conventional main-eects model of Eq. (1) is used as a basisfor deriving the full two-way interaction, product-term model used inthis study to estimate the conditional and interaction eects of episte-mological beliefs on students use of motivational learning strategies (seee.g., Aiken and West, 1991; Cohen and Cohen, 1983; Jaccard andTurrisi, 2003; Jaccard, Turrisi, and Wan, 1990). We follow one popu-lar school of thought [that] conceptualizes interaction eects in terms ofmoderated relationships (Jaccard and Turrisi, 2003, p. 3). For example,an interaction eect would exist if the eect of a particular epistemolog-ical belief on a motivational learning strategy changed or varied accord-ing to the values of one or more other epistemological belief variables.In such cases, the eect of one epistemological belief on the motiva-tional strategy is moderated by one or more of the other epistemolog-ical belief variables.Thus, the specication of our model expresses the hypotheses that

    the estimated eects (b1, b2, b3, b4) of each of the epistemologicalbelief variables in each model (E1, E2, E3, E4,) on the motivationallearning strategy variable (Sij) for a given model is a function of, or ismoderated by, the other epistemological belief variables in each

    740 PAULSEN AND FELDMAN

  • modeli.e., b1=f(E2, E3, E4), b2=f(E1, E3, E4), b3=f(E1, E2, E4),b4=f(E1, E2, E3). Eq. (2) below presents the complete two-way interac-tion, product-term model, which includes the four epistemological beliefvariables themselves and all six of the possible two-way interactionsbetween them.

    Sij a b1E1i b2E2i b3E3i b4E4i b5E1iE2i b6E2iE3i b7E1iE3i b8E2iE4i b9E1iE4i b10E3iE4i eij 2

    In this study, Eq. (2) is estimated using multiple regression analysis, andthe coecient estimates constitute the conditional and interaction eectsof four distinct dimensions of students epistemological beliefs on eachof the six measures of motivational learning strategies.

    Interpreting the Coecients on the Interaction Terms

    The coecients on the interaction terms (b5, b6, b7, b8, b9, b10) in Eq.(2) can be readily interpreted. A complete examination of the eect ofE1 on Sij provides an illustration that generalizes in a straightforwardmanner to parallel expressions of the eects of E2, E3, or E4 on Sij.Taking the rst derivative of Sijwith respect to E1i in Eq. (2) (Sij/E1)indicates that in the full, two-way interaction, product-term model, theeect of the rst epistemological belief variable, E1, on a particularmotivational learning strategy Sij equals b1 + b5 E2 + b7 E3 + b9 E4and is fully expressed as follows:

    Effect of E1 on Sij oSij=oE1 b1 b5E2 b7E3 b9E4 2aTherefore, in this two-way interaction model, the eect of E1 on Sij canchange according to, or be moderated by, the values of E2, E3, or E4.In this illustration, then, E1 is viewed as the focal independent vari-able and E2, E3, and E4 are the moderator variables (Jaccard andTurrisi, 2003, p. 3). This means the eect of E1 on Sij could vary acrossthe values of E2 with the eect of E1 on Sij changing by the amount b5for each one-unit increase in E2. Similarly, the eect of E1 on Sij couldvary across the values of E3 with the eect of E1 on Sij changing by theamount b7 for each one-unit increase in E3, and the eect of E1 on Sijcould vary across the values of E4 with the eect of E1 on Sij changingby the amount b9 for each one-unit increase in E3. In other words, b5,b7, and b9 are estimates of interaction eects and measure theamount by which the eect of E1 on Sij (i.e., the slope of Sij on E1)changes per one-unit change in the values of the moderator variables,E2, E3, or E4, respectively.

    EPISTEMOLOGICAL BELIEFS AND MOTIVATIONAL STRATEGIES 741

  • Furthermore, because all the variables were standardized and thereforemean-centeredprior to the creation of any interaction termsthemean of each variable is equal to zero. This mean-centering provides astraightforward process for isolating and considering one of the interac-tion eect estimates, such as b5, as distinct from the others (b7, and b9),as desired for purposes of interpretation. As an illustration, one couldview E1 as the focal independent variable and E2 as the moderator vari-able and set the other potential moderator variables equal to zeroi.e.,equal to their mean values (E3 E4 0). Under these conditions, theeect of the rst epistemological belief, E1, on a particular motivationallearning strategy Sij in Eq. (2) equals b1 + b5 E2 and is fully expressed asfollows:

    Effect of E1 on Sij given E3 E4 0 oSij=oE1 b1 b5E2 2bMoreover, under these conditions (E3 E4 0), Eq. (2) can be rewrit-ten in a reduced form useful for interpreting and clarifying the meaningof selected interaction eects. Continuing with the illustration above,with E1 as the focal independent variable and E2 as the moderator vari-able, Eq. (2) (given E3 E4 0) can be written as follows:

    Sij a b1E1 b2E2 b5E1E2 3To further facilitate and clarify interpretation of the interaction eect,Eq. (3) can be rewritten in interceptslope form as

    Sij a b2E2 b1 b5E2E1 3awhere the intercept equals (a + b2E2) and the slope equals (b1 + b5E2).If the interaction eect b5 is statistically signicant, then Eqs. (3)and (3a) can be used to explore the meaning of the interaction eect.Cohen and Cohen (1983) and Aiken and West (1991) recommendsubstituting into Eq. (3a) values of the moderator variable E2 equal toone standard deviation below the mean, equal to the mean, and equalto one standard deviation above the mean, which in the case of thestandardized moderator variable E2 would be )1, 0, and 1, respectively.This approach provides insights into the nature of the signicant inter-action eect by simulating selected changes in the intercept and slopeterms of Eq. (3a), and these simulated changes in the equation for vari-ous values ()1, 0, 1) of a moderator variable can be plotted on a graphto display the interaction. When presenting the results of our study, weuse this approach to supplement traditional tabular and narrativepresentations of ndings about interaction eects.

    742 PAULSEN AND FELDMAN

  • Estimating Main Eects versus Conditional Eects

    The main-eects model is illustrated by Eq. (1) given earlier. Thereare four independent variables, four corresponding regressioncoecients, and no higher-order interaction terms. An estimate of theregression coecient b1 in Eq. (1) measures the trends of change inSij per one-unit change in E1 at each value of E2, E3, or E4, or the gen-eral relationship or constant eect of E1 averaging across all levelsof the ranges of values for E2, E3, and E4. The meaning of the estimatesof b2, b3, and b4 in Eq. (1) would be analogous to b1 (Aiken and West,1991; Jaccard and Turrisi, 2003; Jaccard et al., 1990).The present study focuses on the estimation of the conditional and

    interaction eects of epistemological beliefs on motivational learningstrategies in a product-term model. This means that the coecients ofthe product-term interaction eects model presented in Eq. (2), insteadof Eq. (1), are estimated. In the presence of a signicant interactioneecti.e., b5, b6, b7, b8, b9, b10 0 in Eq. 2b1, b2, b3, and b4 areno longer estimates of main eects. Instead, b1, b2, b3, and b4 now rep-resent conditional relationships between the independent variables(epistemological beliefs) and the dependent variable (motivational learn-ing strategies) in the model. In other words, the coecient for E1 (b1)measures the relationship between E1 and Sij, or the eect of E1 on Sij,,for specic values of E2, E3, or E4that is, the magnitude of the rela-tionship is conditional on specic values for the moderator variables E2,E3, or E4. For example, the regression coecient b1 in Eq. (2) measuresthe eect of E1 on Sij when E2=E3=E4=0 and the meaning of the esti-mates of b2, b3, and b4 in Eq. (2) would be analogous to b1 (Aiken andWest, 1991; Jaccard and Turrisi, 2003). For some scales, such as certainLikert scales, the value of zero is not a meaningful value. That mighthave been the case in the present study. However, it is not, because thisstudy uses standardized values for all variables; they are all mean-cen-tered, thereby making zero the mean of the variables, as well as a mean-ingful value on the transformed scales for all variables in this study.

    RESULTS

    The following sections present the results of all the correlation andregression analyses. The results for the bivariate correlations provide anintroductory perspective on the relationships under study, whereas theresults for the regression analyses oer sets of controlled estimates ofthe eects of epistemological beliefsand their interactionson stu-dents use of various motivational learning strategies.

    EPISTEMOLOGICAL BELIEFS AND MOTIVATIONAL STRATEGIES 743

  • Results for the Bivariate Correlation Analysis

    Table 1 presents the bivariate correlations between the four epistemo-logical beliefs (independent variables in the study) and all sixmotivational learning strategies (dependent variables in the study), indi-cating the magnitude and statistical signicance for each of the bivariatecorrelations estimated. These results provide an introductory perspectiveon the relationships under study. The statistically signicant correlationsin Table 1 can be classied as being in the modest range (i.e., r

  • TABLE1.EpistemologicalBeliefs

    andMotivationalSelf-RegulatedLearningStrategies:Bivariate

    CorrelationalAnalysis

    MotivationalSelf-RegulatedLearningStrategies

    IntrinsicGoal

    Orientation

    ExtrinsicGoal

    Orientation

    Task

    Value

    Self-Ecacy

    Controlof

    Learning

    Test

    Anxiety

    EpistemologicalBeliefs

    (1)

    (2)

    (3)

    (4)

    (5)

    (6)

    SimpleKnowledge(SK)

    )0.27*

    0.26*

    )0.17*

    )0.20*

    )0.09

    0.27*

    Fixed

    Ability(FA)

    )0.27*

    )0.17*

    )0.37*

    )0.25*

    )0.39*

    )0.01

    Certain

    Knowledge(CK)

    0.05

    0.03

    0.03

    0.03

    0.10*

    )0.03

    QuickLearning(QL)

    )0.05

    )0.01

    )0.14*

    )0.09

    )0.13*

    0.01

    N(samplesize)=

    502,*=p