Multiple Goals, Multiple Pathways: The Role Of Goal Orientation In .

This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.MULTIPLE GOALSlevels of both of these goals would be the most adaptive. In thiscase, following the logic of multiplicative interaction effects, ifthere are two positive main effects for mastery and approachperformance goals, then it may be that a focus on mastery alongwith a focus on trying to do better than others at the same time (ahigh-mastery/high-performance pattern) would result in enhancedpositive outcomes. That is, as suggested by Harackiewicz et al.(1998), it may not matter what type of goals are pursued, but ratherthat the goals lead to affective and cognitive involvement in thetask. In this enhancement view, with mastery goals leading tointrinsic task involvement and with approach performance goalsleading to involvement based on competition and trying harder todo better than others, the overall net effect would be a boost ininvolvement in the task with a variety of positive outcomes.On the other hand, normative goal theory would suggest that anyconcern with performance, even an approach performance orientation, could have negative effects on involvement due to distractions fostered by attention to comparisons with others or to negative judgments regarding the self. Under this dampening orreduction perspective, the overall level of involvement fostered bya mastery goal would be less when students simultaneously endorse an approach performance goal. This lower level of overallinvolvement would then result in less positive outcomes. Accordingly, under this normative model, it would be hypothesized thatthe most adaptive pattern of multiple goals would be a highmastery, low approach performance combination.In classroom research regarding these issues, Wotters, Yu, andPintrich (1996), using a two-wave correlational design with juniorhigh students, found that higher levels of an approach performancegoal predicted higher levels of self-efficacy, task value, and the useof cognitive and metacognitive strategies. Moreover, this maineffect of performance goals was independent of the positive maineffect of mastery goals, paralleling the findings of Harackiewiczand Elliot (Elliot & Church, 1997; Harackiewicz et al., 1997).Wolters et al. (1996) did not find many significant interactionsbetween the two goals as indexed by multiplicative effects in aregression model (a variable-centered analysis). The few interactions that did come out significantly were in line with normativegoal theory, with a high-mastery/low-performance pattern beingthe most adaptive in terms of self-efficacy, task value, and cognitive and metacognitive strategy use.In contrast, Midgley and her colleagues (Kaplan & Midgley,1997; Middleton & Midgley, 1997) have not found many positiverelations between approach performance goals and adaptive motivation or cognition in correlational studies of middle school students. Middleton and Midgley (1997), using a variable-centeredanalysis, found that approach performance goals did not relate toself-efficacy or to self-regulation but were positively related tomore test anxiety. Kaplan and Midgley (1997) reported similarresults, with performance goals being unrelated to the use ofadaptive strategies and positively related to the use of maladaptivestrategies for learning.In other classroom studies with more person-centered analyses(e.g., using median splits or clustering procedures) to create groupsof students in contrast to variable-centered analyses, the findingsalso have been mixed. For example, Meece and Holt (1993)observed that a high-mastery/low-performance group of elementary students had the most adaptive pattern of cognitive strategyuse as well as actual achievement, in line with normative goal545theory predictions. Pintrich and Garcia (1991), also using clusteranalysis with a sample of college students, found that the highmastery/low-performance group had the most adaptive profile. Atthe same time, they noted that their low-mastery/high-performancegroup did display some positive signs of motivation and cognition,at least in contrast to the low-mastery/low-performance group.They suggested that in the absence of a mastery goal, at least aconcern with performance motivated their college students toengage in their courses to some degree. In contrast, Bouffard,Boisvert, Vezeau, and Larouche (1995), who used median splits toform groups in another study of college students, found that thehighest levels of motivation, cognitive strategy use, selfregulation, and achievement were displayed by the high-mastery/high-performance group. The next best pattern was the highmastery/low-performance group, followed by the low-mastery/high-performance group, and the least adaptive pattern was foundfor the low-mastery/low-performance group. These results aremore in line with a revised perspective on goal theory that proposes a more adaptive role for performance goals.In an attempt to synthesize these divergent findings, Pintrich(2000) recently suggested that there may be multiple "pathways,"or developmental trajectories, that are fostered by different goalorientations. He suggested that mastery and performance goalscould set up and foster different patterns of motivation, affect,strategy use, and performance over time. In this sense, studentswho adopt different goals might follow different pathways, ortrajectories, over time, with some of them ending up in the "same"place in terms of actual achievement or performance but having avery different experience on the way to this overall outcome. Forexample, to continue the metaphor of a trip and the experiences ona trip, mastery goal students might generally have a "smoother ornicer" experience in terms of their motivation, positive affect,effort, and strategy use along the way to good levels of achievement. That is, given their continued focus on mastery and selfimprovement across time and tasks, there could be a cumulativeeffect of the mastery emphasis such that even in the face ofdifficulties, these mastery students maintain their focus on learningand involvement with the tasks and sustain an adaptive profile ofaffect, strategy use, and achievement (Dweck & Leggett, 1988).In contrast, but in line with a revised goal theory perspective,performance goal students might arrive at the same level ofachievement, or even higher achievement, given Harackiewicz etal.'s (1998) findings for approach performance students, accompanied by high levels of efficacy. However, as suggested bynormative goal theory, these students may experience less interest,less positive affect, and perhaps more anxiety or negative affectgiven their concerns about doing better than others. In addition,they may be less likely to demonstrate effort because of their goalof looking smarter than others. If they experience any difficultiesor failures along the way, there could be costs for them in terms oftheir affect (lower interest, more negative affect), or they mayinvoke different types of effort or strategies to attain their goal ofbeing better than others. Accordingly, although there could becumulative positive benefits for achievement and efficacy whenapproach performance goals are adopted, there also could becumulative costs in terms of increased anxiety or negative affect orloss of interest over time. Thus, both normative and revised goaltheory predictions may be accurate, but the accuracy would depend

This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.546PINTRICHon which outcome is under consideration over time (Harackiewiczet al., 1998; Pintrich, 2000).Longitudinal data are important to examine the proposeddifferent trajectories or pathways fostered by different goals.That is, it may not be sufficient to examine different outcomesat one time point only. It may be that some of the positive ornegative outcomes do not occur until time has allowed forpotential goal effects to accumulate, highlighting the need forlongitudinal data. Moreover, the idea of accumulating multiplegoal effects over time suggests the importance of a personcentered analysis because the effects would accumulate to theindividual person over time. Magnusson (1998; Magnusson &Staltin, 1998) has suggested that the logic of person-centeredanalysis is best suited to address these types of questionsconcerning individual development over time.In this study, data are presented for different outcomes overthree waves for junior high students in mathematics classrooms.The main research question concerns how the developmentaltrends or multiple pathways vary as a function of multiplegoals. Following the logic of a person-centered analysis, fourgroups of students were examined in a repeated measuresdesign as a function of creating a 2 X 2 matrix of mastery andperformance goals. Accordingly, comparisons were made between high-mastery/high-performance students, high-mastery/low-performance students, low-mastery/high-performance students, and low-mastery/low-performance students. The keyissue was whether group membership moderated the developmental trends in outcomes. The outcomes included four sets ofvariables: motivational beliefs, affect, strategy use, and classroom performance.For motivational beliefs, I included self-efficacy, task value, andtest anxiety as three motivational outcomes (cf. Pintrich & DeGroot, 1990). Given the findings of a general decrease in adaptivemotivational beliefs over time, especially in junior high and middleschools (Eccles, Wigfield, & Schiefele, 1998; Wigfield, Eccles, &Pintrich, 1996), I predicted that efficacy and value would decreaseand anxiety would increase over time (a time effect). However,these general developmental trends over time should be moderatedby group membership. For self-efficacy, I expected that the highmasteryAow-perforniance group and high-mastery/high-perfbrmance group would have similar trajectories over time, but that thetwo low-mastery groups would have a more significant decrease inefficacy over time (cf. Bouffard et al., 1995; Wolters et al., 1996).I expected that the high-mastery/low-performance group wouldhave higher levels of interest and task value over time in comparison to the high-mastery/high-performance group (Harackiewicz etal., 1998). Given their concern for performance, I hypothesizedthat the high-mastery/high-perfonnance group would increasemore in test anxiety over time than the high-mastery/low-performance group (Middleton & Midgley, 1997).The second general category of outcome included two affectivescales. Positive affect referred to feeling happy, proud, and goodabout oneself during school. Negative affect concerned the frequency of feeling ashamed, embarrassed, or angry during school.I expected that there would be a time effect with positive affectdecreasing and negative affect increasing over time (Eccles et al.,1998; Wigfield et al., 1996). For time by goal group interactioneffects, I predicted that the high-mastery/low-performance groupwould show the most adaptive pattern followed by the high-mastery/high-performance group, low-mastery/high-performancegroup, and then the low-mastery/low-performance group (Dweck& Leggett, 1988; Pintrich & Garcia, 1991).The third general category of outcome variables included fourstrategy variables. Motivational strategies included student attempts to control their own effort levels as well as their willingnessto take risks in the classroom setting. One motivational strategy,self-handicapping, included students' withdrawal of effort in theface of difficult tasks or procrastination in doing school work. Ipredicted that there would be an increase in self-handicapping overtime but that this developmental trend would be moderated by goalgroup. The predicted interaction was that the high-mastery/lowperformance group would report the least amount of selfhandicapping over time, whereas the high-mastery/high-performance group and the low-mastery/high-performance groups mightincrease in self-handicapping over time (Midgley, Arunkumar, &Urdan, 1996).The other motivational strategy included risk taking (see Clifford, 1988) and focused on students' willingness to offer their ownideas in classroom discussions or to try tasks that were new ordifficult. I expected that risk taking would decrease over time(Eccles et al., 1998; Wigfield et al., 1996) but that this developmental decrease would be moderated by goal-group membership.As predicted by normative goal theory, the high-mastery/low-performance group was hypothesized to report the least decrease inrisk taking, with the high-mastery/high-performance group reporting more of a decrease over time given their concern with lookingsmarter than others. I predicted that the low-mastery/high-performance group would report the most decrease in risk taking (Dweck& Leggett, 1988).In terms of cognitive strategies, I investigated two measures,used by Pintrich and De Groot (1990). Cognitive strategy usereflected the students' active cognitive engagement in the tasksin terms of their use of rehearsal, elaboration, and organizational strategies. The second measure concerned the use ofmetacognitive strategies for planning, monitoring, and regulating cognition. Because these scales index general cognitiveinvolvement, I predicted that they would decrease over time(Eccles et al., 1998; Wigfield et al., 1996) but that the decreasewould be moderated by goal-group membership. I expected thatboth high-mastery groups would report similar decreases instrategy use over time, whereas the two low-mastery groups,especially the low-mastery/low-performance group, would report the most decrease in cognitive strategy use over time(Bouffard et al., 1995; Pintrich, 2000; Pintrich & Garcia, 1991;Wolters et al., 1996).The final outcome measure that was included in the studywas the students' actual classroom performance as indexed bytheir grades in their math classes. Paralleling the predictions forcognitive strategy use, I expected that grades would decreaseover time but that this decrease would change as a function ofgoal-group membership. I expected that both high-masterygroups would have the least decline and higher levels of performance than the two low-mastery groups with the low-mastery/low-performance group having the lowest level of performanceand the largest decrease in grades over time (Pintrich, 2000;Pintrich & Garcia, 1991).

MULTIPLE GOALSMethodParticipantsThe participants were 150 students (78 girls, 52%) in the eighth andninth grades in one junior high school (seventh to ninth grades) in southeastern Michigan. The district was primarily working class in terms ofsocioeconomic status, and the sample was over 95% Caucasian. At thebeginning of the study at Wave 1, the average age of the sample was 13years, 4 months.This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.ProcedureThe data were collected in three waves. Wave 1 at the beginning of theyear in eighth grade (October), Wave 2 at the end of the year in eighthgrade (May), and Wave 3 at the end of the year in ninth grade (May).Students were given a self-report questionnaire, an adapted version of theMotivated Strategies for Learning Questionnaire (MSLQ; Pintrich & DeGroot, 1990; Pintrich, Smith, Garcia, & McKeachie, 1993), to fill outduring regular class time. After students were read a set of instructionsregarding the confidentiality of their self-reports and the importance ofbeing accurate in their ratings, each item was read aloud to the whole classby trained research assistants. Each class had two research assistantsinvolved in data collection, one reading the items aloud and the other onewalking around monitoring progress, ensuring quality responses, answering questions, and helping students with the questionnaire.There was some attrition over the course of the study, due to studentsmissing one of the waves of survey data collection. There was very littlemovement out of the district in terms of the collection of the mathematicsgrade data. Students were included in the study if they filled out all threequestionnaires at the three waves, resulting in a overall sample of 150students. Comparisons of this final sample with the larger sample revealedno large differences on the measures. However, there was still somemissing survey data because some students skipped items or skipped a pageon the surveys at one of the waves. The data analytic strategy required allcases at all three waves, so these cases were then dropped from the analysesof variance (ANOVAs). For the two cognitive strategies scales, 1 case wasmissing; for the grade analyses, 2 cases were missing; and for the twoaffect scales, 26 cases were missing.MeasuresAll items on the questionnaire were rated on 7-point Likert scales (1 strongly disagree to 7 strongly agree). Items were worded to havestudents focus on their mathematics classroom; for example, phrases like"in this class" were part of the stem of the item. The only exception to thiswas for the two affect scales, which asked students about their affectiveexperience in school in general. See the Appendix for sample items for allself-report scales.Goal orientation. The measures of personal goal orientation wereadapted from Midgley et al. (1998) and included two scales, one focusedon mastery and one focused on performance. The items focused on thestudents' personal goals, not on their perception of the classroom-level goalorientation. The mastery scale had six items pertaining to a personal goalconcerned with mastery and learning of the class work (a .70). Performance goal items reflected an approach performance orientation to classroom work, such as trying to be smarter than or outperform others. Therewere no items included in this scale that referred to an avoidance performance goal orientation such as avoiding looking incompetent. There werefive items in this scale with an alpha of .76 at Wave 1. These two scaleswere positively related (r — .23, p .01) to each other.To examine the interactions between these two goal orientations and tolink the interactions to three waves of data over time in a person-centeredanalysis, the Wave 1 mastery and performance goal scales were dichotomized using median splits. This procedure allowed for the use of repeated547measures ANOVAs with multiple dependent variables, in contrast to theuse of regression (and cross-product multiplicative terms to index theinteraction), which can handle only a single dependent variable. Accordingly, simple linear regressions would not allow for the examination ofchange over time within groups of students, but the ANOVAs would, albeitthey require categorical predictors, hence the median splits. Studentsscoring below 4.8 on mastery were classified as low mastery and those 4.8and above were categorized as high mastery. For performance goals, thelow-performance group had scores at or below 4.6, and the high-performance group had scores above 4.6. This resulted in 50% (n 75 each) ofthe sample being classified as either low or high mastery and 47% (n 70)classified as low performance and 53% (n 80) categorized as highperformance. Neither of these two categorical predictors was related togender: mastery goals-gender, 1 , N 150) .10, p .74, andperformance goals- gender, x*(l,N 150) 1.89,p .17.1ntermsof theassociation between the two categorical variables of mastery and performance goals, the chi-square was not significant at a conventional level,suggesting some orthogonality between the two goals, x*(,l, N 150) 2.67, p .10, although the phi coefficient of .13 was positive,paralleling the positive zero-order correlation between the continuous goalmeasures. The four cells had the following distribution of participants:low-mastery/low-performance n 40; low-mastery/high-performance n 35; high-mastery/low-performance n 30; and high-mastery/high-performance n 45.Motivation dependent variables. There were three motivational beliefscales: Self-Efficacy, Task Value, and Test Anxiety. Self-Efficacy(items 4, as ranged from .81 to .90 across the three waves) concernedstudents' confidence that they could do the work in the class. Task Valueitems (items 6, as .75-.S5) referred to students' personal interest inthe course content and perceived utility of math. Test Anxiety (items 4,as .78-.82) focused on die cognitive interference or worry componentof anxiety and concerned worry about doing poorly on tests. These threescales were correlated as in previous studies (e.g., Pintrich & De Groot,1990), with Self-Efficacy and Task Value positively related to each other(rs ranged from .42 to .46) and low to moderate correlations with TestAnxiety (rs ranged from - .29 to — .52 for Self-Efficacy-Test Anxiety, andfrom - . 0 3 to - . 1 1 for Task Value-Test Anxiety).Affect dependent variables. There were two affect scales included, onefor negative affect and one for positive affect experienced in school. TheNegative Affect Scale (as .66-.72) included four items about how oftenparticipants felt angry, ashamed, embarrassed, or frustrated in school. ThePositive Affect Scale (items 4, as .80-.86) concerned feeling happy,having fun, feeling proud about oneself, and being in a good mood duringschool. These two scales were moderately correlated with each other (r — .39, — .46, and — .48 at the three waves, respectively) but were distinctscales in factor analyses and represented different aspects of affect considered separately in the analyses.Strategy dependent variables. There were four strategy variables withtwo motivational strategies of self-handicapping and risk taking. TheSelf-Handicapping Scale included five items with alphas ranging from .50to .57. The items concerned procrastinating or holding back effort as afunction of concerns about poor grades. Risk taking (items 4, as .55to .62) was adapted from Clifford (1988) and referred to students' attemptsto try to do the work even if they were not sure of the answer and mightbe wrong.The Cognitive Strategies Scales were adaptations of the Pintrich and DeGroot (1990) MSLQ scales and included a scale focused on generalcognitive strategies for learning and a scale on metacognitive strategies forlearning. The items (items 9, as .86 to .88) for cognitive strategy useincluded strategies for rehearsing or memorizing course material as well asattempts to engage the material in a deeper manner through the use ofelaborative or organizational strategies. Metacognitive strategy items(items 7, as .64 to .74) included planning, setting goals, monitoringcomprehension, and regulating cognition.

This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.548PINTRICHThese motivational and cognitive strategy scales were moderately correlated with each other, but were distinct enough constructs and scales toconsider separately. For example, rs ranged from - . 2 0 to -.60 for selfhandicapping, with the more adaptive risk taking, cognitive strategy use,and metacognitive strategy use scales. Risk taking, cognitive strategy use,and metacognitive strategy use were positively related to one another (rsranged from .42 to .66).Math grade. Students' semester grades for mathematics were collectedfrom school records. The Wave 1 math grade was the first semestereighth-grade math letter mark, which was assigned in the January following the Wave 1 questionnaire administration. The Wave 2

The Role of Goal Orientation in Learning and Achievement Paul R. Pintrich University of Michigan Mastery goals have been linked to adaptive outcomes in normative goal theory and research; performance goals, to less adaptive outcomes. In contrast, approach performance goals may be adaptive for some outcomes under a revised goal theory perspective.