lable at ScienceDirect

Learning and Instruction 29 (2014) 103e114

Contents lists avai

Learning and Instruction

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

Procrastination and self-efficacy: Tracing vicious and virtuous circlesin self-regulated learning

Kristin Wäschle a,*, Anne Allgaier a, Andreas Lachner a, Siegfried Fink b, Matthias Nückles a

aDepartment of Educational Science, Albert-Ludwigs-University of Freiburg, Rempartstr. 11, 79085 Freiburg, Germanyb Faculty for Forest and Environmental Science, Albert-Ludwigs-University of Freiburg, Bertoldstr. 17, 79085 Freiburg, Germany

a r t i c l e i n f o

Article history:Received 14 August 2012Received in revised form20 September 2013Accepted 24 September 2013

Keywords:ProcrastinationSelf-efficacyPerceived goal achievementFeedback loopsSelf-regulated learning

* Corresponding author. University of Freiburg, Depence, Rempartstr. 11, D-79098 Freiburg, Germany. Tel

E-mail addresses: kristin.waeschle@ezw.uni-fanne.allgaier@ezw.uni-freiburg.de (A. Allgaier),freiburg.de (A. Lachner), siegfried.fink@fobot.umatthias.nueckles@ezw.uni-freiburg.de (M. Nückles).

0959-4752/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.learninstruc.2013.09.005

a b s t r a c t

In the present study, we investigated how students react to self-assessed low goal achievement in self-regulated learning. Over a university term (19 weeks), 150 university students recorded self-efficacy,procrastination and perceived goal achievement in weekly web-based self-monitoring protocols. Usingmultilevel analyses for growth curve models, we investigated the reciprocal amplifying between pro-crastination and perceived goal achievement and self-efficacy and perceived goal achievement. Resultsindicated a vicious circle of procrastination and a virtuous circle of self-efficacy. Students who recordedhigh levels of procrastination assessed their goal achievement as being low. As a consequence of low goalachievement, they reinforced procrastination. Students who recorded high levels of self-efficacy assessedtheir goal achievement as being high. As a consequence of high goal achievement, self-efficacy increased.Self-efficacy mediated the effect of perceived goal achievement on procrastination. Thus, students withlow perceived self-efficacy are vulnerable for finding themselves in a vicious circle of procrastination.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Although even very young children are able to self-regulatetheir learning processes to some extent (e.g., Perels, Merget-Kullmann, Wende, Schmitz, & Buchbinder, 2009), in general, self-regulated learning is a demanding task for many learners. Self-regulated learning requires considerable knowledge about how toinstruct and regulate oneself effectively (Schraw, 1998). Further-more, it is indispensable that learners not only know about how tolearn, but really use self-regulation strategies adaptively to succeedin their learning process (Mayer, 2002; Weinstein & Mayer, 1986;Zimmerman, 2008a). These strategies encompass cognitive, meta-cognitive, motivational and behavioral strategies (Pintrich, 2004).Vermunt and Vermetten (2004) stated that a lack of regulationoccurs when students possess insufficient strategies for regulatingtheir learning processes and insufficient external support is madeavailable to them. At the university level, external support is typi-cally limited to setting some deadlines for students (such as exam

artment for Educational Sci-.: þ49 761 203 2460.reiburg.de (K. Wäschle),

andreas.lachner@ezw.uni-ni-freiburg.de (S. Fink),

All rights reserved.

dates or deadlines for submissions of assignments). Thus, theability of students to self-regulate should be even more important.A ubiquitous problem is students’ delay of academic tasks until thelast minute, that is, procrastination (Steel, 2007), which can beregarded as a failure-avoiding strategy (Helmke & van Aken, 1995).A classic counter-argument that students like to put forward todefend themselves against the criticism of procrastination is thatlearning is done best under pressure, and at any rate, that of coursethey are able to reduce their procrastination if it becomes reallyproblematic. Thus, many students seem to believe that they are ableto self-regulate effectively when it is required (Zimmerman, 2002).

In the present longitudinal study, we explored university stu-dents’ self-regulation abilities over a whole term. Specifically, weinvestigated how procrastination affects self-regulated learning, towhat extent procrastination tends to reinforce itself in a viciouscircle as a result of low goal achievement, and to what extent theperception of self-efficacy might result in a virtuous circle thathelps the students overcome the tendency to procrastinate.

In this article, we use the term procrastination to denote irra-tional postponing of important learning tasks. A distinction shouldbe made between trait and state procrastination, because actualpostponing (state procrastination) is influenced by a personaltendency to delay tasks (trait procrastination), situational aspectsand self-regulation strategies (Steel, 2007). The results of Steel’smeta-analysis showed that trait procrastination is stable across

K. Wäschle et al. / Learning and Instruction 29 (2014) 103e114104

situations and over a longer time span. Nevertheless, there isalso empirical evidence that postponing can be influenced bycontextual aspects, for example task attributes (e.g., aversive,complicated, boring). Learners who were asked to solve tasksassociated with low autonomy and low personal utility were morelikely to procrastinate (Lonergan &Maher, 2000). Strategies for self-regulation are generally expected to mediate the effect of personal(e.g., trait procrastination) and contextual characteristics (e.g., taskcharacteristics) on learning behavior (Pintrich, 2004; Vermetten,Vermunt, & Lodewijks, 1999). Thus, a personal tendency to pro-crastinate does not necessarily need to result in actual postponing.A medium correlation of r ¼ .51 (Stainton, Lay, & Flett, 2000) in-dicates that trait and state procrastination are indeed related butnot quite the same. Hence, strategies of self-regulation may in factcompensate for a personal tendency to procrastinate. On the otherhand, deficits in self-regulation could then result in procrastinationthat interferes with goal achievement (Steel, 2007). Thus, wetheoretically assume a reciprocal amplifying effect between pro-crastination and goal achievement. Bandura (1978) coined the termreciprocal determinism to denote that psychological functioninginvolves reciprocal interactions between behavioral, cognitive andenvironmental factors.

1.1. How procrastination sustains itself

Previous studies on procrastination identified numerous corre-lates of procrastination that are important components of self-regulated learning and strongly related to learning success, forexample goal setting (e.g., Steel, 2007) and cognitive strategy use(e.g., Howell & Watson, 2007). Goal setting is indeed a linchpin inself-regulated learning (Locke & Latham, 2002; Zimmerman,2008b). Goal setting differs from students’ goal orientation inthat goal orientation focuses on the reasons for engaging in aca-demic tasks (i.e., to learn or perform), whereas goal setting focuseson the self-regulatory act of setting a specific goal anchored incontext and time (Eccles & Wigfield, 2002; Zimmerman, 2008b).Goal setting enables learners to plan their learning processes(Boekaerts, 2011; Efklides, 2011; Zimmerman, 2008b). Self-setlearning goals help learners to decide which learning strategiesare beneficial and how much effort they need to invest. Further-more, self-set learning goals enable learners’ self-reflection of goalachievement. Hence, goal setting is an important strategy self-regulated learners use to guide their learning process (self-regu-lation as a top-down, goal directed regulation; Efklides, 2011).There is considerable evidence that learning goals guide learners’selection of cognitive strategies and their self-reflection on goalachievement (Locke & Latham, 2002; Schunk, 2001; Winne &Hadwin, 1998). When thinking about appropriate learning goals,it is not only important for learners to think about how, but alsowhy to achieve a learning goal. Therefore, learning goals that reflectreasons such as personal utility (i.e., why a content is relevant) ormastery (i.e., being able to understand or to do something) helplearners to start a task and to persist on the task, expending greateffort in meeting the requirements (Assor, Kaplan, & Roth, 2002;Belenky & Nokes-Malach, 2012; Wigfield, Eccles, Roeser, & Schie-fele, 2008). Taken together, appropriate learning goals can supportself-regulated learning and prevent deficits in motivation (Hofer,Schmid, Fries, Kilian, & Kuhnle, 2010; Seo, 2009).

A problem related to procrastination is deficient cognitivestrategy use (Howell & Watson, 2007). Learning strategy re-searchers typically distinguish between surface and deep learningstrategies (Leutner & Leopold, 2003; Marton & Saljö, 1997).Following Weinstein and Mayer (1986), cognitive strategies can becategorized into organization, elaboration, and rehearsal strategies.Rehearsal strategies mainly refer to the repetition of information in

order to support refreshment and retention in memory. Rehearsalstrategies rather support superficial information processing andcan therefore be regarded as lower in quality than organization andelaboration strategies (Leutner & Leopold, 2003). Organizationstrategies include, for example, identifying the main concepts ofthe newly learned contents as well as the structuring of the con-cepts (e.g., identifying the main concepts and relating them to eachother in a map). Elaboration strategies connect the learning con-tents with a learner’s prior knowledge (constructing examples andanalogies, for example). Due to their role in the process of knowl-edge construction (see Mayer, 2010), organization and elaborationstrategies are regarded as deep learning strategies that facilitate theenduring integration of learning content into existing cognitiverepresentations by changing or complementing them, therebyallowing for flexible use of knowledge (Mayer, 2002).

Several empirical studies showed that procrastination is indeednegatively related to the extent of cognitive strategy use. In a studyby Howell and Watson (2007), self-reported use of cognitivelearning strategies proved to be a strong predictor of procrastina-tion that remained stable even when controlling for motivationaltrait variables, such as achievement goal orientations (i.e., masteryapproach and mastery avoidance orientations, see Elliot &McGregor, 2001). In the study by Wolters (2003), the results weresomewhat less clear, as the negative relation between cognitivestrategy use and procrastination disappearedwhenwork avoidance(i.e., another type of achievement goal orientation) and perceivedself-efficacy (see Subsection 1.2) were introduced as additionalpredictors of procrastination. Given that Howell and Watson, aswell as Wolters, used cross-sectional designs by regressing pro-crastination on predictors such as reported cognitive strategy useand perceived self-efficacy, a more process-oriented analysis ofhow procrastination, learning goals, cognitive strategy use and self-efficacy affect each other was not possible. Therefore, analyzing theinterplay of these variables using a longitudinal study design couldbe illuminative.

Inasmuch as extensive procrastination is associated with non-optimal use of learning strategies, and deep learning strategies inparticular, procrastination has also been found to impair students’learning outcomes (Klassen, Krawchuk, & Rajani, 2008; Lay &Schouwenburg, 1993; Tice & Baumeister, 1997). As a consequence,in cases of substantial state procrastination, the extent of havingachieved one’s learning goals should be reduced in students’ self-perceptions. Moon and Illingworth (2005) conducted growthcurve analyses to model the effect of the course of procrastinationon academic performance as measured by five multiple-choicetests in introductory psychology courses. The results showed thatmeasures of state procrastination were significant predictors ofacademic performance. Similar results were, for example, found inlongitudinal analyses conducted by Tice and Baumeister (1997),who showed that procrastination had negative effects on students’grades. Given these results, we wondered whether procrastinationis also related to a student’s self-reflection of goal achievement.This self-reflection informs students about whether they haveachieved their learning goals or whether adaptations of learningactivities are necessary. Thus, perceived goal achievement is animportant reference for students that helps them to regulate theirlearning process. Furthermore, we were interested in whether itwould be possible to detect reciprocal amplifying of procrastinationand perceived goal achievement. Thus, we were interested inwhether students who perceive their goal achievement as ratherlow are able to regulate themselves by reducing procrastination, orwhether they continue or even reinforce procrastination. Studentswho continue or reinforce procrastination after low goal achieve-ment would be at particular risk of becoming involved in a viciouscircle of procrastination.

K. Wäschle et al. / Learning and Instruction 29 (2014) 103e114 105

1.2. How self-efficacy sustains itself

An important source of learning motivation is the perception ofself-efficacy, that is, the anticipation of success based on one’s owncompetencies (e.g., Zimmerman, 2008a). Perceived self-efficacy isnegatively related to procrastination (Steel, 2007): the more self-efficacious one feels, the less is one’s inclination to procrastinate.Accordingly, perceptions of high self-efficacy in academic settingspositively affect the quality of studying and performance (Capraraet al., 2008). Reciprocally, perceived self-efficacy of academicfunctioning depends on the experience of former success in aca-demic tasks (Helmke & van Aken, 1995; Marsh & Craven, 2006;Williams & Williams, 2010). In line with this assumption,Sitzmann and Yeo (2013) found in their recent meta-analysis that,in diverse domains including sports, statistics, logical thinking andwork-place learning, self-efficacy was rather a product than adriver of performance. For academic achievement, Caprara et al.(2008) showed with growth-curve analyses that highly self-efficacious students seem to become involved in a virtuous circleof perceived self-efficacy and academic success, because eachvariable sequentially alternates as the cause and the effect. Thus,the experience of success could help students to further improvetheir behavior (as bottom up, experience-directed regulation;Efklides, 2011). Caprara and colleagues investigated the develop-mental course of perceived self-efficacy from junior high school tohigh school and its reciprocal relation to academic achievements.Over a period of 10 years (from age 12 to 22 of the learners), theyassessed perceived self-efficacy at six points of measurement, andacademic achievement at two of them (measurement points T3and T6). The results indicated that perceived self-efficacy pre-dicted academic achievement and academic achievement (T3:finish of junior high school) was a predictor of subsequentlymeasured perceived self-efficacy. Given these results, wewondered whether it would be possible to detect analogousreciprocal amplifying of self-efficacy and achievement in a morefine-grained longitudinal analysis of self-regulated learning with,for example, weekly points of measurement and also with mea-sures reflecting a student’s self-evaluation of the extent to whichthe self-set weekly learning goals had been achieved. Such ananalysis would allow an even more accurate portrayal of the psy-chological dynamics of goal achievement and self-efficacy in astudent’s everyday study life.

1.3. Motivational feedback-loops in models of self-regulatedlearning

Self-regulated learning involves regular reflections of the indi-vidual student’s goal achievement over the course of a whole uni-versity term (Zimmerman, 2002). This reflection is an importantreference for students, helping them to decide whether they couldgo on with the learning process or whether improvements arenecessary. We expected that high self-efficacy should sustain itselfin a positive reciprocal amplifying with goal achievement e that is,in a positive feedback loop. At the same time, the personal tendencyto procrastinate might contain the risk of procrastination sustain-ing itself in a negative reciprocal amplifying with goal achievement,or a negative feedback loop.

Such feedback loops are at the heart of recent theoretical modelsof self-regulated learning (Schmitz & Wiese, 2006; Winne &Hadwin, 1998; Zimmerman, 2000). According to Zimmerman’s(2000) cyclical-interactive model, self-regulation of one’s learningencompasses three consecutive phases: a forethought phase, aperformance phase and a self-reflection phase. Each phase buildsupon the preceding phase and influences the subsequent phase.The three phases offer a theoretical framework, in which we can

locate the sub-processes that interact in the aforementionedfeedback loops of self-efficacy and procrastination.

Students estimate their self-efficacy in the forethought phasewhen they are confronted with a learning task and have to appraisewhether they will be able to succeed on this task. Perceptions ofhigh self-efficacy increase the motivation to tackle the task. Sub-sequently, in the performance phase, highly self-effective studentsimplement cognitive deep learning strategies that increase thelikelihood of succeeding in the task. As a result, these students willperceive high degrees of goal achievement in the self-reflectionphase. Consequently, perceptions of high goal achievement willcontribute to and raise the students’ self-efficacy (Helmke & vanAken, 1995; Marsh & Craven, 2006; Williams & Williams, 2010).Increased self-efficacy, in turn, will positively impact the students’motivation to tackle new tasks in the next learning cycle. Thus, self-efficacy acts as a protective factor in self-regulated learning thatsustains itself in positive feedback-loop and can therefore bedescribed as a “virtuous circle of self-efficacy”.

Procrastination similarly occurs in the forethought phase, eitherwhen students are confronted with tasks associated with low au-tonomy and low personal utility (Lonergan & Maher, 2000), or, incase of low self-efficacy (Steel, 2007; Van Eerde, 2003). If the stu-dents perceive the task as aversive or experience low self-efficacy inregards to the task, students will be less likely to tackle the task andto use cognitive deep learning strategies. Therefore, they will tendto procrastinate, that is, delay the performance phase, and/orrealize the performance phase in a qualitatively less than optimalfashion. As a consequence, these students presumably will perceivetheir goal achievement as rather low in the subsequent reflectionphase of the learning cycle. As the perception of low goalachievement might cause emotions of disappointment, the ten-dency that the students will anticipate failure in the next learningcycle might increase and the students may become tempted toavoid this failure e even for a short time e by avoiding the taskentirely (e.g., Alexander & Onwuegbuzie, 2007; Helmke & vanAken, 1995). Such a defensive reaction would lead into amplifiedprocrastination that, in turn, increases the likelihood of repeatedfailure (Zimmerman, 2002). Thus, students who procrastinateregularly could in that way become involved in a negativefeedback-loop in which procrastination and the perception of lowgoal achievement amplify each other. For these reasons, we coinedthis negative feedback loop in the present study the “vicious circleof procrastination”. Evidence for such a negative feedback loopwould indicate a serious vulnerability in students’ ability to self-regulate their learning process.

1.4. Investigating feedback loops with growth-curve modeling inHLM

Feedback loops in self-regulated learning such as the negativeloop between procrastination and goal achievement as well as thepositive loop between self-efficacy and goal achievement obviouslyrely on theoretical assumptions about causality. Generally, estab-lishing a causal model requires satisfaction of the following threecriteria: (a) the cause temporarily precedes the effect (temporalprecedence criterion), (b) the cause and the effect are related(covariance criterion), and (c) disqualification of alternative expla-nations by referring to a third, confoundingvariable that accounts forthe relationship (Shadish, Cook, & Campbell, 2002). Experimentalstudies, in which the hypothesized cause is manipulated and itssubsequent effect on the outcomes is measured, generally meetthese criteria (Duckworth, Tsukayama,&May, 2010). Confounds (e.g.,socioeconomic status) can effectively be controlled by randomlyassigning participants to experimental conditions. Unfortunately,procrastination cannot easily be investigated experimentally. As it is

K. Wäschle et al. / Learning and Instruction 29 (2014) 103e114106

theoretically defined as irrational postponing (Steel, 2007), it wouldmake no sense, for example, that a researcher requests postponingfrom the participants in one experimental condition, because in thatcase irrationality of the observed behavior could not be assumed.Consequently, researchers interested in the causes and effects ofprocrastination typically resort to correlative field designs (cf., Steel,2007) that measure procrastination in the natural learning envi-ronment of the students. Correlative studies, especially if they arecross-sectional, however, only fulfill the covariance criterion. Incomparison with cross-sectional designs, longitudinal designs aremore appropriate to investigate reciprocal relationships betweenvariables, because they additionally fulfill the precedence criterion.The predictor and the outcome variable can bemeasured at differentpoints in time. To analyze reciprocity in longitudinal data, typically,cross-lagged panel analyses are used (Duckworth et al., 2010). Cross-lagged panel analyses account for autoregressive effects that appearwhen relatively stable variables linearly depend on their own pre-vious value. For example, to investigate the effect of previous goalachievement on procrastination, we need to simultaneouslyconsider the stability of procrastination over time, that is, the so-called autoregressive effect of procrastination. Thus, in order toassess a potentially causal influence of goal achievement on pro-crastination in a cross-lagged panel analysis, we need to investigatethe relationship between preceding goal achievement and subse-quent changes in procrastination. Hence, cross-lagged-panel ana-lyses thus allow for satisfaction of the precedence and covariancecriteria. However, although potential third variables can bemeasured and statistically controlled, it is theoretically impossible toanticipate all potential confounds (Duckworth et al., 2010). There-fore, the possibility to disqualify alternative explanations is limited.Growth curve analysis usinghierarchical linearmodeling (HLM)goesone step further and provides a solution for controlling for third-variable confounds. Within hierarchical linear modeling, a growthcurve for changes in an outcome variable over time is modeled foreach individual. Each growth curve conveys estimations of an in-dividual’s baseline (i.e., the intercept), and the change fromonepointof measurement to the next (i.e., the slope). Because of a repeatedmeasurement of both the predictor and the outcome variable onecan show the relation of two time-varying variables more than oncefor each individual. Thus, in away, each individual serves as his orherown control. Therefore, time-invariant, between-individual con-founds (e.g., socioeconomic status) can that way be disqualified asalternative explanations (Duckworth et al., 2010). Hence, bymodeling of growth-curvemodels for longitudinal data, wemay stillnot have randomization of participants such as in experimental de-signs. Nonetheless, we come close to meeting the preconditions fortesting causal hypotheses regarding potential effects of prior factorson subsequent outcomes.

1.5. Overview of the present study, research questions andhypotheses

1.5.1. OverviewFor data collection in our longitudinal field study, we asked

students to record their class preparation in a web-based self-monitoring protocol once aweek during awhole university term. Ineach of the 19 protocols, the learners documented and evaluateddifferent components of their class preparation based onstandardized and open questions about the learning process. Theseself-monitoring protocols enabled us to assess the course of theo-retically relevant variables such as procrastination as well as theirrelationships to other variables, such as goal achievement andperceived self-efficacy. As the students completed the self-monitoring protocols in their natural study environment at homeor at the university, this should contribute to ecological validity of

the collected data (Schmitz & Wiese, 2006). Compared to cross-sectional questionnaire studies, regular self-monitoring protocolsallow for a more contextualized assessment of a student’s learningbehavior, because the students do not have to mentally averageabout many different learning situations when answering ques-tions about self-regulation strategies and processes (Cleary, 2006;Hadwin, Winne, Stockley, Nesbit, & Woszczyna, 2001).

1.5.2. Research questions and hypothesesThe main scope of our study was to investigate reciprocally

amplifying motivational feedback loops in self-regulated learning.

1.5.2.1. Hypothesis 1: vicious circle hypothesis. First, we assumed afeedback loop between procrastination and perceived goalachievement. Therefore, we expected that procrastination wouldimpair perceived goal achievement and low goal achievementwould reinforce future procrastination. The following relationsshould be satisfied. (1) In a first growth-curve model, the degree ofself-reported procrastination that precedes the self-evaluation ofgoal achievement should be a significant negative predictor ofperceived goal achievement, when auto-regression of goalachievement is controlled. (2) In a second growth-curve model, thedegree of perceived goal achievement should be a significantnegative predictor of future procrastination, when auto-regressionof procrastination is controlled. (3) When including potential third-variable confounds in the growth-curve models, the relations (1)and (2) should remain stable. To disqualify important time-varyingconfounds, we controlled for auto-regression and the time factor,representing the general time courses of procrastination and goalachievement related to the examinations.

1.5.2.2. Hypothesis 2: virtuous circle hypothesis. Second, weassumed a reciprocally amplifying feedback loop between self-efficacy and perceived goal achievement. Therefore, we expectedthat self-efficacy would improve perceived goal achievement andhigh goal achievement would reinforce future perceptions of self-efficacy. The following relations should be satisfied. (1) In a firstgrowth-curve model, the degree of self-efficacy that precedes theself-evaluation of goal achievement should be a significant positivepredictor of perceived goal achievement. (2) In a second growth-curve model, the degree of perceived goal achievement should bea significant positive predictor of future self-efficacy. (3) Whenincluding potential third-variable confounds in the growth-curvemodels, the relations (1) and (2) should remain stable. Todisqualify important time-varying confounds, we controlled forauto-regression and the time factor, representing the general timecourses of procrastination and goal achievement related to theexaminations.

1.5.2.3. Hypothesis 3: self-efficacy mediates effect of goal achieve-ment on procrastination. Third, we investigated whether the vi-cious and the virtuous circles are inter-related. To answer thisquestion, we analyzed whether the effect of goal achievement onprocrastination would be mediated by self-efficacy. The perceptionof high goal achievement should lead to increased self-efficacy(Caprara et al., 2008). Beliefs about self-efficacy are related tolower procrastination (Steel, 2007).We expected that an increase inself-efficacy should be related to a decrease in procrastination.Thus, we expected self-efficacy to be a mediator between goalachievement and procrastination.

1.5.2.4. Supplementary explorative analyses concerning the viciousand virtuous circle hypotheses. Fourth, in order to investigate theprocesses underlying the relationships postulated by the viciouscircle and virtuous circle hypotheses in more detail, we conducted

K. Wäschle et al. / Learning and Instruction 29 (2014) 103e114 107

additional HLM-analyses. Inasmuch as procrastination was ex-pected to reduce perceived goal achievement (see vicious circlehypothesis) whereas self-efficacy should improve goal achieve-ment (see virtuous circle hypothesis), we wondered whether theseeffects would at least partly depend on the use of cognitive learningstrategies. More specifically, we investigated whether higher de-grees of procrastination were associated with a decrease in the useof cognitive learning strategies, whereas, on the other hand, higherdegrees of self-efficacy were associated with an increase in the useof cognitive strategies. On the other hand, due to their central rolein the process of knowledge construction (see Mayer, 2010), theamount of cognitive learning strategies as reported in the self-monitoring protocols should also predict a student’s perceivedgoal achievement. As there is considerable evidence that thelearning goals guide a learners’ selection of cognitive strategies(Locke & Latham, 2002; Schunk, 2001; Winne & Hadwin, 1998), wealso explored whether the self-reported frequency of cognitivestrategies employed to reach one’s self-determined learning goalsdepended on specific attributes of these goals, such as, the articu-lation of personal utility (i.e., why a goal is personally relevant) ormastery (e.g., “being able to explain a concept”).

1.5.2.5. Validation of the self-report data. Fifth, in order to validateour results based on self-report data, we analyzed how self-reported procrastination and self-efficacy affected the students’objective learning outcomes as measured by their examinationgrades at the end of the university term. Based on previous research(e.g., Caprara et al., 2008; Moon & Illingworth, 2005; Tice &Baumeister, 1997), we expected that self-efficacy should be a pos-itive predictor and procrastination a negative predictor of learningoutcomes. The reported frequency of the use of deep-learningstrategies should also predict learning outcomes. Evidence forthese relationships would provide support for the validity of theself-report data documented in the self-monitoring protocols.

2. Method

2.1. Sample and design

One-hundred-fifty students in forest and environmental scienceof the Albert-Ludwigs-University of Freiburg with an average age of22.3 years (SD ¼ 2.73) at the first point of measurement voluntarilyparticipated in the study. The large majority were Bachelor stu-dents. Twenty-six pursued a Master’s degree. The proportion offemale participants was 56%. We asked our participants to fill in aself-monitoring-protocol once a week over a time span of 19 weeks(a whole term) in order to self-report their class preparation. Thefirst entry was written in the second week after the beginning oflectures. On average, participants submitted 15.43 (SD ¼ 4.41) of 19possible self-monitoring protocol entries. Dependent measurescomprised the students’ self-reported procrastination, theirperceived goal achievement, self-efficacy, the self-reported fre-quency of cognitive strategy use and the number and content of theself-set learning goals.

2.2. The self-monitoring protocol

We implemented the self-monitoring protocol within the onlinelearning management system which lecturers and students at ourdepartment regularly use as part of their lectures and courses(Moodle, see http://docs.moodle.org). The web-based imple-mentation had the advantage that students could decide when andwhere to make their entries. To keep the amount of missing entriesas low as possible, participants who failed to complete their weeklyself-monitoring protocol in time were sent a friendly reminder via

email. We limited the length of the self-monitoring protocol to asmall number of questions in order to encourage continuousparticipation. Therefore, we selected e based on the theoreticalconsiderations outlined above e variables and items from morecomprehensive questionnaires (see the following subsections fordetails). The self-monitoring protocol consisted of a reflection and aplanning section.

2.2.1. Reflection of class preparationTo answer the reflection items, students were asked to think

about their class preparation in the previous week and to evaluatetheir learning behavior retrospectively on a number of rating scales.

2.2.1.1. Perceived goal achievement. We asked the students toindicate the extent to which they had been able to achieve theirlearning goals on a five-point rating scale ranging from 1 (¼notachieved) to 5 (¼completely achieved). They rated their goalachievement separately for each goal that they had set in the pre-vious week. An overall individual achievement scorewas computedby averaging across the learning goals.

2.2.1.2. Cognitive learning strategies. To investigate how the stu-dents sought to achieve their goals, we asked them to estimatetheir cognitive strategy use on several items that were adapted andtranslated from the Motivated Strategies for Learning Question-naire (MSLQ; Pintrich, Smith, Garcia, & McKeachie, 1993). The sixitems used to self-assess cognitive strategy use included twoelaboration items (e.g., “I thought about examples that illustrateabstract learning content”; Cronbach’s a ¼ .78), two organizationitems (e.g., “I tried to structure the learning material”; Cronbach’sa ¼ .82), and two rehearsal strategy items (e.g., “I repeated thelearning materials more than once”; Cronbach’s a ¼ .73). The stu-dents rated the frequency of the use of cognitive strategies on afive-point rating scale ranging from 1 (¼seldom) to 5 (¼very often).

2.2.1.3. Perceived self-efficacy. We used three items fromZimmerman, Bandura, and Martinez-Pons (1992) (e.g., I was surethat I could cope with the academic demands in the last week”;Cronbach’s a ¼ .78). The students indicated their perceived self-efficacy on a five-point rating scale ranging from 1 (¼low agree-ment) to 5 (¼high agreement).

2.2.1.4. State procrastination. Subsequently, the students estimatedthe extent to which they had procrastinated during the previousweek and postponed their academic tasks. Based on four items ofthe dilatory behavior questionnaire by Lay and Silverman (1996;e.g., “I postponed my tasks until the very last minute”; Cronbach’sa ¼ .67), the students estimated their state procrastination on a 5-point rating scale ranging from 1 (¼low agreement) to 5 (¼highagreement). An overall individual procrastination score wascomputed by averaging across the items.

2.2.2. Planning of class preparationTo plan their learning activities in the following week, the stu-

dentswere asked to formulate up to three learning goals at the end ofeach self-monitoring protocol. For this purpose, the self-monitoringprotocol offered several text fields inwhich the students could entertheir learning goals. The students’descriptions of their learning goalswere subjected to a content-analysis. A preliminary inspection of thedescriptions suggested categorizing the learning goals according tomotivational aspects. Accordingly, we defined “personal utility” and“mastery” as analytic categories (inter-coder reliability as deter-mined by Cohen’s Kappa was k ¼ .73 and k ¼ .89, respectively).Goal descriptions with personal utility included information whystudents wanted to perform a task (“repeating facts to performwell

Table 1The unconditional means models for dependent variables.

Dependent variables Unconditional means model

Mean Variancebetween (s)

Variancewithin (s2)

ICC

Procrastinationa 2.89 0.28 0.42 .40Self-efficacya 2.71 0.34 0.49 .41Mastery goalsb 0.19 0.05 0.20 .20Personal utility goalsb 0.30 0.02 0.08 .20Elaborationa 3.00 0.54 0.45 .54Organizationa 2.88 0.49 0.95 .34Rehearsal strategiesa 2.29 0.28 0.42 .40Goal achievementa 3.22 0.16 0.73 .18

The mean represents an average person’s mean value across the 14 points ofmeasurement. The ICC represents the proportion of variance between persons.

a Rated on a five-point rating scale (1e5).b Number of goals with this attribute.

1 Growth-curve modeling is a special application of hierarchical linear modelingthat can also be estimated using structural equation modeling (Duckworth et al.,2010).

K. Wäschle et al. / Learning and Instruction 29 (2014) 103e114108

in an exam”). Thus, in these goals, students explained to themselvesthe relevance they attributed to the task (why they pursued a goal).Goal descriptions articulating a need for mastery focused oncompetence outcomes (e.g., “knowing”, “being able”), rather than onperformance only. Each goal description could include statementsregarding personal utility as well as mastery.

2.2.3. Recording of exam performanceIn the last, that is, the 19th, self-monitoring protocol, the par-

ticipants were asked to record their exam grades to the best of theirknowledge up to that point. The exams students passed at the endof the semester typically included recall and transfer tasks. Theexams were designed by the lecturers of the faculty of forest andenvironmental sciences. The exams were similar to those oneswritten in previous semesters and they were not specificallyadapted to the purposes of the present study. As the last self-monitoring protocol fell into the term holidays, a substantialnumber of the students participating in our study unfortunatelymissed that entry. Therefore, we obtained performance data onlyfrom 92 out of the 150 participants (for more details see theparagraph on treatment of missing data below). The average gradewas 2.00 (SD ¼ 0.56), with low grade values representing highperformance and high values representing low performance(grades: 1¼ very good, 2¼ good, 3¼ satisfactory, 4¼ pass, 5¼ fail).

2.3. Procedure

At the beginning of the study, we informed our participants thatour intention was to learn more about students’ self-regulatedlearning in order to better understand potential problems and toprovide support. During the 19-weeks-intervention period, thestudents recorded their class preparation in the web-based self-monitoring protocol once a week while they attended their regularcourses. To do so, the students logged on to the Moodle learningmanagement system. In the protocols, the students evaluated theirgoal achievement and estimated their cognitive strategy use,perceived self-efficacy and procrastination, and subsequentlydefined their personal learning goals for the upcoming week. Thestudents were able to complete one protocol per week betweenFriday and the following Tuesday. On Monday mornings, partici-pants who had not yet completed their protocol were reminded viaemail. Making an entry required about 10e15 min to complete.

3. Results

3.1. Missing data

Prior to HLM analyses, we excluded five points of measurementthat took place during vacations (Christmas and after the end ofterm) because many students did not study at that time (between25% and 50% of the students had no learning goals in this period)and many entries (>25%) were missing. Hence, a total of 14 entrieswere included in the analyses. The participating studentscompleted 12.07 (SD ¼ 2.75) of these 14 protocol entries onaverage. To check for any biases resulting from missing data on thelevel of the individual student, we analyzed whether the number ofprotocol entries a student missed to fill in was systematicallyrelated to the student’s procrastination, perceived self-efficacy andperceived goal achievement scores. Accordingly, we computedPearson correlations between the number of completed entriesduring the surveyed time span and the degree of self-reportedprocrastination, self-efficacy and perceived goal achievement atthe beginning of the study. None of these correlations was signifi-cant, procrastination: r ¼ �.05 (N ¼ 150), p ¼ .50; self-efficacy:r ¼ .06 (N ¼ 150), p ¼ .50; and perceived goal achievement:

r ¼ .02 (N ¼ 150), p ¼ .76. These results suggest that there were nosystematic relationships between missing entries and the mainvariables under investigation.

As we had substantial missing data especially with regard to theexam grades reported in the last self-monitoring protocol, weconducted several ANOVAS to check whether there were any sys-tematic differences between students who did or did not reporttheir grades. None of the F-tests approached statistical significance,procrastination: F(1, 148) ¼ 0.32, p ¼ .57, partial h2 ¼ .02; perceivedself-efficacy: F(1, 148) ¼ 0.05, p ¼ .81, partial h2 ¼ .00; perceivedgoal achievement: F(1, 148) ¼ 0.05, p ¼ .82, partial h2 ¼ .00.Together, these missing data analyses suggest that whether stu-dents reported their exam grades and how many protocol entriesthey completedwas evidently independent of our main variables ofinterest, that is, procrastination, perceived self-efficacy andperceived goal achievement.

3.2. The statistical modeling

We applied growth-curve modeling1 to estimate the baselineandweekly change in observed outcomes for each student to assessthe hypothesized relationships. The models in HLM consider thepoints of measurement to be nested within individuals. Therandom coefficient approach permits capturing within-personrandom effects and between-person effects in the model. Thepoints of measurement represent the first level of the model,whereas individuals represent the second level. As our participantstypically failed to complete about two of the 14 self-monitoringprotocols on which we based our analyses, we had missing dataon the first level of the models but no missing data on the secondlevel. The degrees of freedom vary across estimates, representingthe rate of the data for each estimate. Missing entries are notconsidered for estimation and have therefore no biasing impact inHLM (Raudenbush & Bryk, 2002).

In the first step of the HLM analyses, unconditional meansmodels that included only the dependent variables were specified.They outlined the average persons’ mean level across the obser-vation period and provided estimates of the variability of thedependent variables at both the points of measurement and theindividual levels. These variances were used to compute the intra-class coefficient (ICC). The ICC, representing the percentage ofvariance between persons, indicated that approximately 18%e54%of the variance in the dependent variables was between persons.

Table 2Zero order correlations between the dependent variables averaged across all pointsof measurement and persons.

Variables 1 2 3 4 5 6 7 8

1. Self-efficacy e �.41** .49** .44** .43** .40** .08 .032. Procrastination e �.30** �.27** �.19* �.49** �.06 �.033. Elaboration

strategiese .54** .45** .27** .23** .18*

4. Organizationstrategies

e .62** .25** .47** .42**

5. Rehearsalstrategies

e .18* .49** .40**

6. Goal achievement e .17* .23*

7. Mastery goals e �.068. Personal

utility goalse

Note. *p < .05, **p < .01.

2 Different degrees of freedom are due to a slightly different number of partici-pants at different points of measurement.

K. Wäschle et al. / Learning and Instruction 29 (2014) 103e114 109

Therefore, we controlled individual effects by modeling persons onthe second level in HLM. In turn, 46%e82% of variance was withinpersons indicating different degrees of the variables at differentpoints of measurement. The differences between the points ofmeasurement were stronger than the average differences betweendifferent students (except of elaboration). Table 1 illustrates thevariances within and between persons as well as the ICCs. Table 2shows the zero order correlations.

In the second step, we added potential first-level predictors tothe models according to our hypotheses. As recommended byRaudenbush and Bryk (2002), we centered all predictors on theirgroup mean that is e in case of longitudinal studies e the person’smean. To account for the hypothesized reciprocal relationshipsbetween variables, we specified contemporary effects, assessed bythe current self-monitoring protocol and lagged effects, assessed bythe self-monitoring protocols one week before. A lagged effect didnot exist for the first point of measurement. As such, the data of thefirst point of measurement served only to determine the laggedeffects for the second point of measurement. By specifyingcontemporary and lagged effects, we were able to investigate theeffects of preceding variables on the outcome measures similar tocross-lagged panel analyses.

3.3. Hypothesis 1: reciprocal amplification between procrastinationand perceived goal achievement (vicious circle hypothesis)

To test the hypothesized relationships that would make up a vi-cious circle of procrastination and perceived goal achievement (seehypotheses subsection), we first modeled perceived goal achieve-ment in a growth-curve model, with the self-reported degree ofpreceding procrastination as predictor. To account for auto-regression of goal achievement, we included the previously re-ported goal achievement as a second predictor. To account for theexpected changes over time, we further included the time factor as athird predictor. Controlling for auto-regression of perceived goalachievement, b ¼ .06, t(1142) ¼ 2.06, p < .05, and the time factor,b ¼ .03, t(1142) ¼ 4.65, p < .01, indicating an increase of goalachievement over time, the preceding degree of procrastinationwasa negative predictor of goal achievement, b ¼ �.39, t(1142) ¼ �9.96,p< .01. Following the second part of the vicious circle hypothesis, wecomputed another growth-curve model, in which we modeled theself-reported degree of procrastination with goal achievement re-ported in the preceding protocol as predictor. To account for auto-regression of procrastination, we included previous procrastinationas a second predictor. To account for the expected changes over time,we further included the time factor as a third predictor. Controllingfor auto-regression of procrastination, b¼ .13, t(1167)¼ 3.78, p< .01,and the time factor, b ¼ �.03, t(1167) ¼ �5.01, p < .01, indicating a

decrease in procrastination over time, perceptions of goal achieve-ment in the preceding protocol were a significant negative predictorof self-reported procrastination in the subsequent protocol, b¼�.05,t(1167) ¼ �2.23, p < .05.2 Thus, as predicted by the vicious circlehypothesis, we found evidence for a reciprocal and negative feed-back loop between goal achievement andprocrastination (see Fig.1).Preceding procrastination was a negative predictor of subsequentgoal achievement, and, reciprocally, the perceived degree of goalachievementwas anegative predictorof subsequent procrastination.Thus, students who tended to procrastinate, subsequently experi-enced less achievement of their learning goals, and this lower goalachievement reinforced their inclination to further procrastinate thestudy tasks.

3.4. Hypothesis 2: reciprocal amplification between self-efficacyand perceived goal achievement (virtuous circle hypothesis)

To test the virtuous circle hypothesis, we first modeledperceived goal achievement in a growth-curvemodel, with the self-reported degree of preceding self-efficacy as predictor. To accountfor auto-regression of goal achievement, we included the previ-ously reported goal achievement as a second predictor. To accountfor the expected changes over time, we further included the timefactor as a third predictor. Controlling for auto-regression ofperceived goal achievement, b¼ .07, t(1142)¼ 2.36, p< .05, and thetime factor, b ¼ .04, t(1142)¼ 5.26, p < .01, indicating an increase ofgoal achievement over time, the degree of previously experiencedself-efficacy was a positive predictor of perceived goal achieve-ment, b ¼ .29, t(1142) ¼ 8.01, p < .01. Following the second part ofthe virtuous circle hypothesis, we computed another growth-curvemodel, in which we modeled self-efficacy with the goal achieve-ment reported in the preceding self-monitoring protocol as pre-dictor. To account for auto-regression of self-efficacy, we includedprevious self-efficacy as a second predictor. To account for the ex-pected changes over time, we further included the time factor as athird predictor. Controlling for the time factor, b ¼ .03,t(1168) ¼ 5.47, p < .01, indicating an increase in self-efficacy overtime, previous perceptions of goal achievement were a significantpositive predictor of subsequently reported self-efficacy, b ¼ .06,t(1168) ¼ 2.24, p < .05. The auto-regressive effect of previous self-efficacy on actual self-efficacy failed statistical significance, b ¼ .10,t(1167)¼ 1.92, ns. Taken together, as predicted by the virtuous circlehypothesis, we found evidence for a reciprocal and positive feed-back loop between goal achievement and self-efficacy (see Fig. 1).Preceding self-efficacy was a positive predictor of subsequent goalachievement, and, reciprocally, preceding perceptions of goalachievement were a positive predictor of subsequent self-efficacy.Thus, students who experienced higher degrees of self-efficacy,reported higher achievement of their learning goals subsequently,and this perception of higher goal achievement contributed to theirself-efficacy on future study tasks. Consequently, perceived goalachievement and self-efficacy indeed intensified each other withina reciprocally amplifying positive feedback loop (see Fig. 1).

3.5. Hypothesis 3: self-efficacy mediates effect of goal achievementon procrastination (mediation hypothesis)

To test this mediation hypothesis, we followed in our HLManalysis the three steps for mediator analyses recommended byBaron and Kenny (1986). In order to establish mediation, first, thedependent variable (procrastination) has to be regressed on the

Fig. 1. Overview over the main HLM regression coefficients illustrating the virtuous circle of self-efficacy and the vicious circle of procrastination. aHLM coefficient when controllingfor cognitive strategies; bHLM coefficient when controlling for self-efficacy; cWe analyzed organization, elaboration, and rehearsal strategies separately.

K. Wäschle et al. / Learning and Instruction 29 (2014) 103e114110

predictor (previous goal achievement). Second, the mediator(perceived self-efficacy) has to be regressed on the predictor(previous goal achievement). Third, the dependent variable (pro-crastination) has to be regressed simultaneously on both the pre-dictor (previous goal achievement) and the mediator (perceivedself-efficacy). The above-reported HLM analysis of the vicious cir-cle hypothesis already showed that previous goal achievement(predictor) was negatively related to subsequent procrastination,that is, the dependent variable (¼step 1). The test of the virtuouscircle hypothesis showed that previous goal achievement (predic-tor) was negatively related to subsequent self-efficacy, that is, themediator (¼step 2). In the third step of the mediator analysis, wecomputed a hierarchical linear model, in which we regressed pro-crastination simultaneously on both the predictor (previous goalachievement) and the mediator (perceived self-efficacy). To ensurethat the effects were not only effects of time and auto-regression ofprocrastination, the time factor and previous procrastinationwere included as in the HLM-analyses reported above. In this HLM-analysis, the time factor remained stable, b¼�.03, t(1166)¼�4.94,p < .01, as well as previous procrastination (auto-regression),b¼ .10, t(1166)¼ 4.83, p< .01. The HLM coefficient representing therelation between perceived self-efficacy (mediator) and procrasti-nation (dependent variable) was also significant, b ¼ �.08,t(1166) ¼ �2.82, p < .01, indicating that higher self-efficacy wasrelated to lower procrastination (see Fig. 1). At the same time, theHLM coefficient, representing the relation between perceived goalachievement (predictor) and procrastination (dependent variable)failed statistical significance, b ¼ �.04, t(1166) ¼ �1.88, ns.Following Baron and Kenny (1986), these results indicated media-tion because the HLM coefficient representing the relation betweenthe predictor and the dependent variable failed statistical signifi-cance when the mediator was controlled, while, simultaneously,the effect of the mediator on the dependent variable remainedsignificant when the predictor was controlled.

Hence, the effect of perceived goal achievement on procrasti-nation was mediated by self-efficacy. On the one hand, procrasti-nation reduced students’ perceptions of having achieved their self-determined learning goals. On the other hand, the perception oflow or insufficient goal achievement reduced the students’ feelingof self-efficacy and thereby contributed to the perpetuation ofprocrastination. Thus, altogether, our results provided evidence fora vicious circle of procrastination in which perceived self-efficacyevidently had a mediating role (see Fig. 1).

3.6. Supplementary explorative analyses concerning the vicious andvirtuous circle hypotheses

To investigate the processes underlying the relationshipspostulated by the vicious circle and virtuous circle hypotheses inmore detail, we conducted additional HLM-analyses. Specifically,

we investigated whether higher degrees of procrastination wereassociated with less frequent use of cognitive learning strategies,whereas, on the other hand, higher degrees of self-efficacy wereassociated with more frequent use of cognitive strategies. On theother hand, we tested whether the amount of cognitive learningstrategies as reported in the self-monitoring protocols predicted astudent’s perceived goal achievement. Apart from testing theserelations, we further explored whether the self-reported frequencyof cognitive strategies depended on specific attributes of thesegoals, such as, the articulation of personal utility and mastery.

For this purpose, we conducted separate HLM analyses withelaboration, organization and rehearsal strategies as dependentvariables. To account for auto-regression and general changesduring the surveyed time span, we included previous strategy useand the time factor as potential predictors. Additional predictorswere the students’ individual ratings for procrastination and self-efficacy as well as the frequencies by which they mentioned per-sonal utility and mastery as features of their learning goals.

3.6.1. Modeling elaboration strategiesControlling for the time factor, b ¼ .02, t(1200) ¼ 4.41, p < .01,

and auto-regression, b ¼ .31, t(1200) ¼ 12.84, p < .01, the reportedfrequency of elaboration strategies was positively related to thefrequency of mastery as feature of the learning goals, b ¼ .15,t(1200) ¼ 3.66, p < .01, the frequency of personal utility as featureof the learning goals, b ¼ .13, t(1200) ¼ 2.05, p < .05, and perceivedself-efficacy, b ¼ .12, t(1200) ¼ 4.37, p < .01. On the other hand,procrastination was a negative predictor of the use of elaborationstrategies, b ¼ �.08, t(1200) ¼ �2.72, p < .01.

3.6.2. Modeling organization strategiesControlling for the time factor, b ¼ .04, t(1200) ¼ 5.67, p < .01,

and auto-regression, b ¼ .25, t(1200) ¼ 8.79, p < .01, the self-reported frequency of organization strategies was positivelyrelated to the frequency of mastery as feature of the learning goals,b ¼ .31, t(1200) ¼ 5.53, p < .01, the frequency of personal utility asfeature of the learning goals, b ¼ .33, t(1200) ¼ 3.58, p < .01, andperceived self-efficacy, b ¼ .13, t(1200) ¼ 3.41, p < .01, whereasprocrastination turned out to be a negative predictor of theself-reported use of organization strategies, b ¼ �.21,t(1200) ¼ �4.93, p < .01.

3.6.3. Modeling rehearsal strategiesControlling for the time factor, b ¼ .04, t(1201) ¼ 4.89, p < .01,

and auto-regression, b ¼ .30, t(1201) ¼ 6.59, p < .01, self-reportedrehearsal strategies were positively related to the frequency ofmastery articulated in the learning goals, b ¼ .35, t(1201) ¼ 5.97,p < .01, and the frequency of personal utility as feature of thelearning goals, b¼ .35, t(1201)¼ 3.62, p< .01, but negatively relatedto procrastination, b ¼ �.23, t(1201) ¼ �6.07, p < .01. Self-efficacy

K. Wäschle et al. / Learning and Instruction 29 (2014) 103e114 111

was no significant predictor of rehearsal strategies, b ¼ .06,t(1200) ¼ 1.60, ns.

In summary, we found that higher degrees of procrastinationwere associated with less frequent use of cognitive learning stra-tegies, whereas higher degrees of self-efficacy were generallyassociated with more frequent use of cognitive learning strategies(except for rehearsal strategies). We also obtained evidence thatfeatures of the learning goals influenced the (reported) use oflearning strategies, that is, the more a learning goal was consideredas personally valuable and represented a mastery goal, the morethe students reported using cognitive learning strategies to reachthis goal.

3.6.4. Cognitive learning strategies as predictors of goalachievement

In the last supplementary analysis, we modeled perceived goalachievement in a growth-curve model with the self-reported use ofcognitive learning strategies as predictors. Controlling for the timefactor, b ¼ .01, t(1405) ¼ 2.21, p < .05, and auto-regression ofperceived goal achievement, b ¼ .06, t(1138) ¼ 1.92, ns, elaborationstrategies, b ¼ .08, t(1405) ¼ 2.11, p < .05, organization strategies,b ¼ .06, t(1405) ¼ 2.23, p < .05, rehearsal strategies, b ¼ .06,t(1405) ¼ 2.05, p < .05, and self-efficacy, b ¼ .15, t(1405) ¼ 4.19,p < .01, were included as positive predictors. At the same time,procrastinationwas a negative predictor, b¼�.28, t(1405)¼�7.25,p < .01.

Thus, we can conclude that, on the one hand, the reported fre-quency of cognitive learning strategies was influenced by motiva-tional variables such as procrastination and self-efficacy. On theother hand, the frequency, with by a student had employedcognitive learning strategies to reach her learning goals, also pre-dicted her experienced goal achievement. Hence, the use ofcognitive learning strategies seems to play a significant role withinthe feedback loop between procrastination and goal achievementas well as within the feedback loop between self-efficacy and goalachievement. However, the analysis above also indicates thatcognitive learning strategies did not mediate the effects of pro-crastination and self-efficacy on goal achievement in the sense ofBaron and Kenny (1986), inasmuch as procrastination and self-efficacy evidently had independent effects on perceived goalachievement beyond cognitive strategy use.

3.7. Validating the self-report data: prediction of learning outcomes

To validate the aforementioned results that are based on stu-dents’ self-reports of their learning processes, we tested whetherthe self-report data predicted more objective learning outcomes asmeasured by exams grades (high values of exam grades representlow performance in the German university system). To this end, wecomputed average scores for our twomotivational variables, that is,perceived self-efficacy and procrastination, as well as for thefrequency of cognitive learning strategies, across all 14 self-monitoring protocol entries. Then, we conducted an exploratorymultiple regression analysis (backward method) with the examgrades as the criterion and procrastination, self-efficacy as well aselaboration, organization and rehearsal strategies as predictors. Inthis analysis,15% (11% adjusted) of the total variance in exam gradeswas explained by the predictors, F(4, 87) ¼ 3.69, p < .01. As ex-pected, procrastination, b ¼ .27, t(87) ¼ 2.38, p < .05, and self-efficacy, b ¼ �.23, t(87) ¼ �2.04, p < .05, predicted exam grades,indicating that higher grades (¼lower performance) were associ-ated with higher procrastination and lower grades (¼higherperformance) were associated with higher self-efficacy. Of thethree predictors representing the frequency of cognitive learningstrategies, only elaboration strategies turned out to be a statistically

significant predictor of exam grades; elaboration strategies,b ¼ �.19, t(87) ¼ �2.05, p < .05, organization strategies, b ¼ .04,t(87) ¼ 0.80, ns, rehearsal strategies, b ¼ .18, t(87) ¼ 1.68, ns.Together, these results suggest that the motivational and cognitiveaspects of self-regulated learning as assessed by self-reports withinself-monitoring protocols also predicted examination grades, thatis, objective learning outcomes, to a considerable extent. We takethis as evidence for the validity of our results based on self-reportsof self-regulated learning processes.

4. Discussion

The main goal of the present study was to test theoretical as-sumptions regarding two reciprocally amplifying feedback loopsthat play a major role in current cyclical-interactive models of self-regulated learning. Based on the research literature on procrasti-nation and self-efficacy, we postulated a negative feedback loopbetween procrastination and perceived goal achievement whichwe called “vicious circle of procrastination” and a second antago-nistic and positive feedback loop between self-efficacy andperceived goal achievement, we referred to as “virtuous circle ofself-efficacy”. Whereas for the virtuous circle of self-efficacy someevidence already existed (e.g., Caprara et al., 2008), evidence for thevicious circle of procrastination, and the mediating role of self-efficacy in this circle, was still missing.

To test our hypotheses regarding the vicious and virtuous cir-cles, we conducted a longitudinal study inwhich we had universitystudents complete a series of self-monitoring protocols over awhole term. In the self-monitoring protocols, the students recordedtheir learning goals, learning strategies, their goal achievement aswell as motivational variables such as their procrastination andself-efficacy. These repeated self-reports allowed for the applica-tion of growth curve modeling in HLM. The results of our study canbe summarized as follows.

First, our predictions with regard to a reciprocally amplifyingfeedback-loop between procrastination and perceived goalachievement (Hypothesis 1) were confirmed. A growth-curvemodel in HLM showed that the more students postponed theirtasks, the lower they assessed their personal goal achievement.Reciprocally, another growth-curve model showed that the lowerthey assessed their previous goal achievement, the more theyprocrastinated in the following week. The effect of procrastinationon goal achievement can be explained by a reduction of time forstudying. Irrational postponing might have reduced time on task(Tabak, Nguyen, Basuray, & Darrow, 2009) and this, in turn, lead tothe observed decrease in the use of cognitive learning strategies.Although we found evidence that procrastination was related to areduced frequency of the use of cognitive learning strategies, pro-crastination still had an impact on perceived goal achievement thatwas apparently not, or at least not completely mediated by cogni-tive strategy use. Thus, the perception of one’s goal achievementmight also depend on other than cognitive variables. For example,Grunschel, Patrzek, and Fries (2012) found in their study thatespecially those students who were worried about their procras-tination on study-related tasks were actually impaired in their ac-ademic performance. Hence, affective variables such as worry andpsychological strain might likewise contribute to both a student’sperceived and actual goal achievement. Further research is neededto fully explain the underlying processes that mediate the rela-tionship between previous procrastination and subsequent goalachievement.

Second, our predictions with regard to a reciprocally amplifyingfeedback-loop between self-efficacy and perceived goal achieve-ment (Hypothesis 2) were also confirmed. A growth-curve modeltesting the relation between self-efficacy and perceived goal

K. Wäschle et al. / Learning and Instruction 29 (2014) 103e114112

achievement showed that the more self-efficacious the studentsfelt, the more positively they evaluated their personal goalachievement. Reciprocally, another growth-curve model showedthat the more positively the students evaluated their previous goalachievement, the more self-efficacious they felt in the followingweek. The relation between perceived self-efficacy and perceivedgoal achievement can partially be explained by the positive effectsof perceived self-efficacy on cognitive strategy use. Students whofelt more self-efficacious reported more frequent use of elaborationand organization strategies. As a consequence, theymade alsomoreoptimistic evaluations with regard to the achievement of theirlearning goals. In sum, perceived self-efficacy encouraged appro-priate learning behavior and learning outcomes and in this waysustained itself within a positive feedback loop of perceived goalachievement and self-efficacy. Caprara et al. (2008) used in their10-years longitudinal study rather large time intervals to detect areciprocal relationship between self-efficacy and the final grades atthe end of junior high (1st measurement point) and high school(2nd measurement point). In the present study, we replicated andextended this finding using a more fine-grained analysis of weeklyself-monitoring protocols. Our results showed that the dynamicsbetween self-efficacy and perceived goal achievement similarlybuilds up within rather small time intervals comprising just a fewdays. Thus, the experience of low goal achievement or failure madein one learning episode may have significant negative conse-quences for the trust in one’s ability to master the tasks in the nextlearning episode. In this respect, our results are in line with themeta-analysis of Sitzmann and Yeo (2013) who showed that pastperformance and academic success are the basis for the formationof the belief in one’s self-efficacy. At the same time, our resultsunderscore the dynamics affected by the perception of self-efficacywithin the cycle of self-regulated learning.

Third, we tested the hypothesis that self-efficacy would mediatethe effect of perceived goal achievement on procrastination (Hy-pothesis 3). Suchmediationwould underline the crucial role of self-efficacy within the negative feedback-loop between goal achieve-ment and procrastination. Furthermore, it would indicate that, in away, the vicious circle of procrastination and the virtuous circle ofself-efficacy are interrelated. In fact, the results of the mediationanalysis (Baron & Kenny, 1986) conducted in HLM showed that lowgoal achievement reduced perceived self-efficacy and that wayincreased procrastination. One reason for this pattern of resultsmight be the expectation of repeated failure and negative emotions.The perception of personally unsatisfactory goal achievementdecreased the students’ perceived self-efficacy. But instead ofincreasing their learning effort and raising cognitive strategy use,the students tended to irrationally postpone their studying. Thus,the learners tended to show a rather dysfunctional, defensive re-action instead of employingmore beneficial and adaptive strategiesto overcome the problems (Zimmerman, 2002). Hence, our resultssupporting the negative feedback-loop between procrastinationand goal achievement point to a critical vulnerability in student’sself-regulated learning. The mediation analysis, in particular, em-phasizes the importance of perceived self-efficacy in self-regulatedlearning, inasmuch as perceptions of lower self-efficacy strength-ened the tendency to procrastinate and thereby inhibited moreeffective self-regulation. On the other hand, perceptions of highself-efficacy could help to reduce procrastination and therefore actas a protective factor against dysfunctional learning behavior suchas the tendency to procrastinate.

Fourth, a multiple regression analysis with examination gradesas criterion showed that self-reported procrastination and self-efficacy as well as estimated frequencies of elaboration strategiespredicted the students’ grades in the examinations they took at theend of the term. Although we were able to obtain information

about the grades only from a subsample of our participants, theresults of the regression analysis nevertheless underscore theexternal validity of the self-report data which students provided intheir weekly self-monitoring protocols. In line with our theoreticalassumptions, we found, in particular, that higher degrees of self-reported procrastination were associated with worse examinationgrades, whereas higher degrees of reported self-efficacy wereassociated with better examination grades. Also, more frequent useof elaboration strategies as reported in the protocols was associatedwith better examination grades. The latter result is in line withmodern theories of knowledge acquisition that emphasize the roleof elaboration strategies in the process of knowledge constructionby enabling integration of new information into a learner’s priorknowledge (see Mayer, 2002, 2010).

4.1. Study limitations and future research

Despite these promising results, several limitations have to beacknowledged. The first limitation refers to the self-monitoringprotocols and a potential reactivity effect by increasing meta-cognitive awareness (Schmitz & Wiese, 2006). Generally, increasedmetacognitive awareness is the result of every method that sup-ports making implicit processes explicit for the researcher and/orthe learners (for example, in think-aloud protocols and self-reports;Gerjets, Kammerer, & Werner, 2011). Accordingly, it is likely thatthe self-monitoring protocols operated as a kind of intervention bydirecting students’ awareness to their learning processes andlearning behavior. Thus, using self-monitoring protocols to inves-tigate the vicious circle of procrastination might possibly haveunderestimated its “real” effects on self-regulated learning. It cantherefore be speculated that without the protocols, students’vulnerability to procrastination might be even greater.

The second limitation refers to the selection of variables coveredby our self-monitoring protocols. Self-regulated learning is a verycomplex process involving numerous cognitive, motivational, af-fective, behavioral and contextual variables (for an overview, seePintrich, 2004). Asking students to regularly answer a compre-hensive self-monitoring protocol encompassing all these variableswould have taken a very long time and probably would havedecreased the students’ motivation to regularly participate in thestudy. Therefore, based on our research intention, we decided tofocus on a few variables that could be assigned to the differentphases of Zimmerman’s cyclical model of self-regulated learning:goal setting, perceived self-efficacy, procrastination (¼forethoughtphase), cognitive strategy use (¼performance phase), and self-assessment of goal achievement (¼self-reflection phase). Thesevariables, which are related to self-regulated learning, had beenshown to be strongly associated with procrastination in previousresearch (Steel, 2007). Besides motivational beliefs such as self-efficacy, which we considered in this study as a state variable, itwould also be interesting for further research to include motiva-tional trait variables such as achievement goal orientations (Elliot &McGregor, 2001). For example, Wolters (2003) found that a generalwork avoidance orientation, that is, the tendency to minimize theeffort one must provide for academic tasks, predicted procrasti-nation in addition to self-efficacy. Howell and Watson (2007) pro-vided evidence that mastery approach as well as mastery avoidanceorientations were associated with procrastination. Students whowere interested in truly understanding and mastering an academictask (mastery approach orientation) reported less procrastinationthan students who were mainly preoccupied with avoidingmisunderstanding academic tasks (mastery-avoidance orienta-tion). It is likely that such general goal orientations influence theextent to which an academic task is perceived by the student asrather aversive or as attractive (see Howell & Watson, 2007). It

K. Wäschle et al. / Learning and Instruction 29 (2014) 103e114 113

would therefore be interesting to include such time-invariant,motivational trait variables as potential moderators of self-regulated learning processes in future research.

A third caveat refers to the question of to what extent our an-alyses allow for a causal interpretation. In the present study, wefollowed the approach by Duckworth et al. (2010) to use growth-curve modeling in HLM in order to disqualify all time-invariantconfounds. What our analyses could not rule out is the possibilityof an unmeasured time-varying third variable that changes syn-chronously with self-efficacy, procrastination and goal achieve-ment. A controlled randomized experimental study would go onefinal step further in order to establish causality (Reinhart, Haring,Levin, Patall, & Robinson, 2013). We acknowledge the importanceof experimental studies. Nevertheless, in order to investigate thedynamics involved in procrastination and its impact on self-regulated learning, we decided in the present study to assess pro-crastination as ecologically valid as possible in the natural learningenvironment of the students. Thus, it is up to future research to findways in how to experimentally investigate the dynamics of pro-crastination, or more generally, self-regulation of motivation.

4.2. Practical implications

As students who had higher beliefs in perceived self-efficacyshowed lower degrees of procrastination, focusing on the role ofperceived self-efficacy could be a promising starting point toreduce procrastination. To strengthen the perception of self-efficacy, students need to experience success in self-regulatedlearning. According to the results of our study (see Subsection3.6), setting learning goals which are viewed as personally rele-vant and which focus on mastery lead to an increase in the use ofcognitive learning strategies. Self-reported use of cognitive learningstrategies, in turn, raised the students’ perception of having ach-ieved their learning goals. Thus, supporting students in settingadequate learning goals (i.e., personally valued mastery goals)might have positive effects on the learning process as well as on theperception of self-efficacy (i.e., the expectancy that one will be ableto master future learning tasks) as an outcome of successfullearning. Positive training effects on goal setting were obtained intraining evaluation studies such as by Cleary and Zimmerman(2004), as well as Brown and Latham (2000). In summary,improved academic achievement and satisfaction with academicachievement contribute to improvements in self-efficacy (Capraraet al., 2008; Schmitz & Wiese, 2006) and thereby put students onthe road to a virtuous circle.

4.3. Conclusion

Are students right when they claim that they are able to reducetheir procrastination if it becomes really problematic? We do notwant to deny that there are students who prefer to work underpressure and still gain academic success (see Grunschel et al., 2012).Nevertheless, our study showed that in a sample of prototypicaluniversity students both dynamics were traceable e the viciouscircle of procrastination and the virtuous circle of self-efficacy. Thetwo dynamics are inter-related, and they may be expected e to acertain extent and in different constellations between individualseas being present in every student. If a student’s perceived self-efficacy and goal achievement are rather high, the vicious circleof procrastinationmay spin at a rather low pace. On the other hand,if a student’s procrastination is high, her or his perceived self-efficacy will be rather low as a consequence of low goal achieve-ment, and the vicious circle may unfold its harmful dynamics. Thus,for a substantial number of students, the claim that they are able toreduce their procrastination when it negatively affects learning

outcomes, may be regarded as a face saving statement rather thanas accurate self-perception. Therefore, it is up to educators to pro-vide instructional support, for example by supporting the studentsin formulating appropriate learning goals and in using cognitivelearning strategies. Providing instructional support is necessary topromote the virtuous circle of self-efficacy and to help studentsavoid the vicious circle of procrastination within the dynamics ofself-regulated learning.

Acknowledgments

The project is part of the “Exzellente Lehre” [Excellent Instruc-tion] initiative at the University of Freiburg. We would like to thankall students who participated for their cooperation and engage-ment as well as our research assistants Ruth Deutschländer andJasmin Leber for coding the free statements in the self-monitoringprotocols and all the other tasks that they mastered. We thankWesley Dopkins for proof reading the manuscript.

References

Alexander, E. S., & Onwuegbuzie, A. J. (2007). Academic procrastination and the roleof hope as a coping strategy. Personality and Individual Differences, 42(7), 1301e1310. http://dx.doi.org/10.1016/j.paid.2006.10.008.

Assor, A., Kaplan, H., & Roth, G. (2002). Choice is good, but relevance is excellent:autonomy-enhancing and suppressing teacher behaviours predicting students’engagement in schoolwork. British Journal of Educational Psychology, 72(2),261e278. http://dx.doi.org/10.1348/000709902158883.

Bandura, A. (1978). The self system in reciprocal determinism. American Psycholo-gist, 33(4), 344e358. http://dx.doi.org/10.1037/0003-066X.33.4.344.

Baron, R. M., & Kenny, D. A. (1986). The moderatoremediator variable distinction insocial psychological research: conceptual, strategic, and statistical consider-ations. Journal of Personality and Social Psychology, 51(6), 1173e1182. http://dx.doi.org/10.1037/0022-3514.51.6.1173.

Belenky, D. M., & Nokes-Malach, T. J. (2012). Motivation and transfer: the role ofmastery-approach goals in preparation for future learning. Journal of the LearningSciences, 21(3), 399e432. http://dx.doi.org/10.1080/10508406.2011.651232.

Boekaerts, M. (2011). Emotions, emotion regulation and self-regulation of learning.In B. J. Zimmerman, & D. H. Schunk (Eds.), Handbook of self-regulation of learningand performance (pp. 408e425). New York: Taylor & Francis.

Brown, T. C., & Latham, G. P. (2000). The effects of goal setting and self-instructiontraining on the performance of unionized employees. Industrial Relations, 55,80e94.

Caprara, G. V., Fida, R., Vecchione, M., Del Bove, G., Vecchio, G. M., Barbaranelli, C.,et al. (2008). Longitudinal analysis of the role of perceived self-efficacy for self-regulated learning in academic continuance and achievement. Journal ofEducational Psychology, 100(3), 525e534. http://dx.doi.org/10.1037/0022-0663.100.3.525.

Cleary, T. J. (2006). The development and validation of the self-regulation strategyinventory-self-report. Journal of School Psychology, 44(4), 307e322. http://dx.doi.org/10.1016/j.jsp.2006.05.002.

Cleary, T. J., & Zimmerman, B. J. (2004). Self-regulation empowerment program: aschool-based program to enhance self-regulated and self-motivated cycles ofstudent learning. Psychology in the Schools, 41(5), 537e550. http://dx.doi.org/10.1002/pits.10177.

Duckworth, A. L., Tsukayama, E., & May, H. (2010). Establishing causality usinglongitudinal hierarchical linear modeling: an illustration predicting achieve-ment from self-control. Social Psychological and Personality Science, 1(4), 311e317. http://dx.doi.org/10.1177/1948550609359707.

Eccles, J. S., & Wigfield, A. (2002). Motivational beliefs, values, and goals.Annual Review of Psychology, 53(1), 109e132. http://dx.doi.org/10.1146/annurev.psych.53.100901.135153.

Efklides, A. (2011). Interactions of metacognition with motivation and affect in self-regulated learning: the MASRL model. Educational Psychologist, 46(1), 6e25.http://dx.doi.org/10.1080/00461520.2011.538645.

Elliot, A. J., & McGregor, H. A. (2001). A 2 $ 2 achievement goal framework. Journal ofPersonality and Social Psychology, 80, 501e519.

Gerjets, P., Kammerer, Y., & Werner, B. (2011). Measuring spontaneous andinstructed evaluation processes during web search: integrating concurrentthinking-aloud protocols and eye-tracking data. Learning and Instruction, 21(2),220e231. http://dx.doi.org/10.1016/j.learninstruc.2010.02.005.

Grunschel, C., Patrzek, J., & Fries, S. (2012). Exploring different types of academicdelayers: a latent profile analysis. Learning and Individual Differences, 23, 225e233. http://dx.doi.org/10.1016/j.lindif.2012.09.014.

Hadwin, A., Winne, P., Stockley, D., Nesbit, J., & Woszczyna, C. (2001). Contextmoderates students’ self-reports about how they study. Journal of EducationalPsychology, 93, 477e487. http://dx.doi.org/10.1037/0022-0663.93.3.477.

K. Wäschle et al. / Learning and Instruction 29 (2014) 103e114114

Helmke, A., & van Aken, M. A. G. (1995). The causal ordering of academicachievement and self-concept of ability during elementary school: a longitu-dinal study. Journal of Educational Psychology, 87(4), 624e637. http://dx.doi.org/10.1037/0022-0663.87.4.624.

Hofer, M., Schmid, S., Fries, S., Kilian, B., & Kuhnle, C. (2010). Reciprocal relationshipsbetween value orientation and motivational interference during studying andleisure. British Journal of Educational Psychology, 80(4), 623e645. http://dx.doi.org/10.1348/000709910X492180.

Howell, A. J., & Watson, D. C. (2007). Procrastination: associations with achievementgoal orientation and learning strategies. Personality and Individual Differences,43(1), 167e178. http://dx.doi.org/10.1016/j.paid.2006.11.017.

Klassen, R. M., Krawchuk, L. L., & Rajani, S. (2008). Academic procrastination ofundergraduates: low self-efficacy to self-regulate predicts higher levels ofprocrastination. Contemporary Educational Psychology, 33(4), 915e931. http://dx.doi.org/10.1016/j.cedpsych.2007.07.001.

Lay, C., & Schouwenburg, H. (1993). Trait procrastination, time management, andacademic behavior. Journal of Social Behavior and Personality, 8, 647e662.

Lay, C., & Silverman, S. (1996). Trait procrastination, anxiety, and dilatory behavior.Personality and Individual Differences, 21(1), 61e67. http://dx.doi.org/10.1016/0191-8869(96)00038-4.

Leutner, D., & Leopold, C. (2003). Selbstreguliertes Lernen: Lehr-/lerntheoretischeGrundlagen [Self-regulated learning: teaching/learning theory foundations]. InU. Witthaus, W. Wittwer, & C. Espe (Eds.), Selbst gesteuertes Lernen e Theoreti-sche und praktische Zugänge [Self-regulated learning e Theoretical and practicalapproaches] (pp. 43e67). Bielefeld: Bertelsmann.

Locke, E. A., & Latham, G. P. (2002). Building a practically useful theory of goalsetting and task motivation: a 35-year odyssey. American Psychologist, 57(9),705e717. http://dx.doi.org/10.1037/0003-066X.57.9.705.

Lonergan, J. M., & Maher, K. J. (2000). The relationship between job characteristicsand workplace procrastination as moderated by locus of control. Journal ofSocial Behavior and Personality, 15, 213e224.

Marsh, H. W., & Craven, R. G. (2006). Reciprocal effects of self-concept and per-formance from a multidimensional perspective: beyond seductive pleasure andunidimensional perspectives. Perspectives on Psychological Science, 1(2), 133e163. http://dx.doi.org/10.1111/j.1745-6916.2006.00010.x.

Marton, F., & Saljö, R. (1997). Approaches to learning. In F. Marton, D. Hounsell, &N. J. Entwistle (Eds.), The experience of learning (pp. 39e58). Edinburgh: ScottishAcademic Press.

Mayer, R. E. (2002). Rote versus meaningful learning. Theory Into Practice, 41, 226e232. http://dx.doi.org/10.1207/s15430421tip4104_4.

Mayer, R. E. (2010). Merlin C. Wittrock’s enduring contributions to the science oflearning. Educational Psychologist, 45(1), 46e50. http://dx.doi.org/10.1080/00461520903433547.

Moon, S. M., & Illingworth, A. J. (2005). Exploring the dynamic nature of procras-tination: a latent growth curve analysis of academic procrastination.Personality and Individual Differences, 38(2), 297e309. http://dx.doi.org/10.1016/j.paid.2004.04.009.

Perels, F., Merget-Kullmann, M., Wende, M., Schmitz, B., & Buchbinder, C. (2009).Improving self-regulated learning of preschool children: evaluation of trainingfor kindergarten teachers. British Journal of Educational Psychology, 79(2), 311e327. http://dx.doi.org/10.1348/000709908X322875.

Pintrich, P. R. (2004). A conceptual framework for assessing motivation and self-regulated learning in college students. Educational Psychology Review, 16,385e407. http://dx.doi.org/10.1007/s10648-004-0006-x.

Pintrich, P. R., Smith, D. A. F., Garcia, T., & McKeachie, W. J. (1993). Reliability andpredictive validity of the Motivated Strategies for Learning Questionnaire(MSLQ). Educational and Psychological Measurement, 53(3), 801e813. http://dx.doi.org/10.1177/0013164493053003024.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications anddata analysis methods. SAGE.

Reinhart, A. L., Haring, S. H., Levin, J. R., Patall, E. A., & Robinson, D. H. (2013). Modelsof not-so-good behavior: yet another way to squeeze causality and recom-mendations for practice out of correlational data. Journal of Educational Psy-chology, 105(1), 241e247. http://dx.doi.org/10.1037/a0030368.

Schmitz, B., & Wiese, B. S. (2006). New perspectives for the evaluation of trainingsessions in self-regulated learning: time-series analyses of diary data.Contemporary Educational Psychology, 31(1), 64e96. http://dx.doi.org/10.1016/j.cedpsych.2005.02.002.

Schraw, G. (1998). Promoting general metacognitive awareness. Instructional Sci-ence, 26, 113e125. http://dx.doi.org/10.1023/A:1003044231033.

Schunk, D. H. (2001). Social cognitive theory and self-regulated learning. InB. J. Zimmerman, & D. H. Schunk (Eds.), Self-regulated learning and academicachievement: Theoretical perspectives (2nd ed.). (pp. 125e152) Mahwah, NJ:Erlbaum.

Seo, E. H. (2009). The relationship of procrastination with a mastery goal versus anavoidance goal. Social Behavior & Personality: An International Journal, 37(7),911e919.

Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs for generalized causal inference. Boston: Houghton-Mifflin.

Sitzmann, T., & Yeo, G. (2013). A meta-analytic investigation of the within-personself-efficacy domain: is self-efficacy a product of past performance or a driverof future performance? Personnel Psychology. http://dx.doi.org/10.1111/peps.12035.

Stainton, M., Lay, C. H., & Flett, G. L. (2000). Trait procrastinators and behavior/trait-specific cognitions. Journal of Social Behavior and Personality, 15(5), 297e312.

Steel, P. (2007). The nature of procrastination: a meta-analytic and theoretical re-view of quintessential self-regulatory failure. Psychological Bulletin, 133(1), 65e94. http://dx.doi.org/10.1037/0033-2909.133.1.65.

Tabak, F., Nguyen, N., Basuray, T., & Darrow, W. (2009). Exploring the impact ofpersonality on performance: how time-on-task moderates the mediation byself-efficacy. Personality and Individual Differences, 47(8), 823e828. http://dx.doi.org/10.1016/j.paid.2009.06.027.

Tice, D. M., & Baumeister, R. F. (1997). Longitudinal study of procrastination, per-formance, stress, and health: the costs and benefits of dawdling. PsychologicalScience, 8, 454e458. http://dx.doi.org/10.1111/j.1467-9280.1997.tb00460.x.

Van Eerde, W. (2003). A meta-analytically derived nomological network of pro-crastination. Personality and Individual Differences, 35(6), 1401e1418. http://dx.doi.org/10.1016/S0191-8869(02)00358-6.

Vermetten, Y. J., Vermunt, J. D., & Lodewijks, H. G. (1999). A longitudinal perspectiveon learning strategies in higher education e different view-points towardsdevelopment. British Journal of Educational Psychology, 69(2), 221e242. http://dx.doi.org/10.1348/000709999157699.

Vermunt, J. D., & Vermetten, Y. J. (2004). Patterns in student learning: relationshipsbetween learning strategies, conceptions of learning, and learning orientations.Educational Psychology Review, 16(4), 359e384. http://dx.doi.org/10.1007/s10648-004-0005-y.

Weinstein, C. E., & Mayer, R. E. (1986). The teaching of learning strategies. InM. C. Wittrock (Ed.), Handbook of research on teaching (3rd ed.). (pp. 315e327)New York: Macmillan.

Wigfield, A., Eccles, J. S., Roeser, R., & Schiefele, U. (2008). Development ofachievement motivation. In W. Damon, & R. M. Lerner (Eds.), Child andadolescent development: An advanced course (pp. 933e1002). New Jersey, NJ:Wiley & Sons.

Williams, T., & Williams, K. (2010). Self-efficacy and performance in mathematics:reciprocal determinism in 33 nations. Journal of Educational Psychology, 102(2),453e466.

Winne, P. H., & Hadwin, A. (1998). Studying as self-regulated learning. InD. J. Hacker, J. Dunlosky, & A. Graesser (Eds.),Metacognition in educational theoryand practice (pp. 277e304). Hillsdale, NJ: Erlbaum.

Wolters, C. A. (2003). Understanding procrastination from a self-regulated learningperspective. Journal of Educational Psychology, 95(1), 179e187. http://dx.doi.org/10.1037/0022-0663.95.1.179.

Zimmerman, B. J. (2000). Attaining self-regulation: a social cognitive perspective. InHandbook of self-regulation (pp. 13e39). San Diego, CA, US: Academic Press.

Zimmerman, B. J. (2002). Becoming a self-regulated learner: an overview. TheoryInto Practice, 41(2), 64. http://dx.doi.org/10.1207/s15430421tip4102_2.

Zimmerman, B. J. (2008a). Investigating self-regulation and motivation: historicalbackground, methodological developments, and future prospects. AmericanEducational Research Journal, 45(1), 166e183. http://dx.doi.org/10.3102/0002831207312909.

Zimmerman, B. J. (2008b). Goal setting: a key proactive source of academic self-regulation. In D. H. Schunk, & B. J. Zimmerman (Eds.), Motivation and self-regulated learning: Theory, research, and applications (pp. 267e295). NewYork: Lawrence Erlbaum Associates.

Zimmerman, B. J., Bandura, A., & Martinez-Pons, M. (1992). Self-motivation foracademic attainment: the role of self-efficacy beliefs and personal goal setting.American Educational Research Journal, 29(3), 663e676. http://dx.doi.org/10.3102/00028312029003663.