Contemporary Educational Psychology 36 (2011) 257–267

Contents lists available at ScienceDirect

Contemporary Educational Psychology

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

Agency as a fourth aspect of students’ engagement during learning activities

Johnmarshall Reeve a,⇑, Ching-Mei Tseng b
a
Department of Education, Korea University, 633 Uncho-Useon Hall, Anam-Dong Seongbuk-Gu, Seoul 136-701, Republic of Korea
b
Department of Teaching and Learning, Science Education Program, N259 Lindquist Center, University of Iowa, Iowa City, Iowa 52242, United States

a r t i c l e i n f o a b s t r a c t

Article history: While a consensus has emerged to characterize student engagement during learning activities as a three-
Available online 17 May 2011 component construct featuring behavioral, emotional, and cognitive aspects, we propose adding agentic
engagement as an important new aspect, which we define as students’ constructive contribution into the
Keywords: flow of the instruction they receive. High school students (237 females, 128 males) from Taiwan com-
Engagement pleted surveys of their classroom motivation and the four hypothesized aspects of engagement while
Agency grades were obtained at the end of the semester. Structural equation modeling analyses showed that
Agentic engagement
agentic engagement was both a distinct and an important construct, one that was associated with stu-
Self-determination theory
Achievement
dents’ constructive motivation, related to each of the other three aspects of engagement, and predicted
independent variance in achievement. The discussion highlights the important, though currently
neglected, ways that students contribute constructively into the flow of the instruction they receive, as
by personalizing it and by enhancing both the lesson and the conditions under which they learn.
Ó 2011 Elsevier Inc. All rights reserved.

1. Introduction and support for autonomy (Birch & Ladd, 1997; Murray &
Greenberg, 2000; Reeve, Jang, Carrell, Jeon, & Barch, 2004; Skinner
Student engagement during learning activities is an important & Belmont, 1993; Skinner, Furrer, Marchand, & Kindermann, 2008;
and heavily researched educational construct (Christenson, Wentzel, 1997), just as it is responsive to students’ own motiva-
Reschly, & Wylie, 2011; Fredricks, Blumenfeld, & Paris, 2004; tional states, such as autonomy, competence, relatedness, and per-
Jimerson, Campos, & Grief, 2003; National Research Council, ceived control (Connell & Wellborn, 1991; Furrer & Skinner, 2003;
2004; Skinner, Kindermann, Connell, & Wellborn, 2009). It is an Gottfried, 1990; Miserandino, 1996; Skinner et al., 1998, 2008).
important educational outcome in its own right as a marker of stu- Recognizing that engagement is responsive to proximal condi-
dents’ positive functioning, but it is further important because it tions, researchers generally emphasize the directional flow that
predicts highly valued outcomes, such as students’ academic pro- teachers’ high-quality relationships and instructional supports
gress and achievement (Ladd & Dinella, 2009; Skinner, Zimmer- have on students’ subsequent behavioral, emotional, and cognitive
Gembeck, & Connell, 1998). Student engagement is also a well- engagement during learning activities. That said, this same body of
understood construct, as a general consensus has emerged to research also acknowledges that student engagement exerts a
characterize it as a 3-component construct featuring behavioral (bi-)directional effect on teachers’ subsequent motivating style
(on-task attention, effort, persistence, lack of conduct problems), and instructional behaviors (Pelletier, Seguin-Levesque, & Legault,
emotional (presence of interest and enthusiasm, absence of anger, 2002; Pelletier & Vallerand, 1996; Skinner & Belmont, 1993). For
anxiety, and boredom), and cognitive (use of strategic and sophis- instance, when students episodically display boredom, dispersed
ticated learning strategies, active self-regulation) aspects (e.g., see attention, and little effort, then teachers tend to change how they
Fredricks et al., 2004; Jimerson et al., 2003; National Research relate to those students (consciously or unconsciously) by lessen-
Council, 2004). Proximal influences on student engagement are ing their support and heightening their control (Pelletier et al.,
also well understood. For instance, student engagement rises and 2002).
falls in response to lessons that are challenging vs. too easy The reciprocal influence that student engagement has on teach-
(Davidson, 1999; Turner, Thorpe, & Mayer, 1998) and to varying ers’ interpersonal style is presumed to flow through teachers’
levels of a teacher’s expression of warmth, provision of structure, awareness of, observations of, and reactions to students’ behav-
ioral, emotional, and cognitive engagement. While this is almost
certainly true, it is also an incomplete understanding of these
⇑ Corresponding author. Address: 633 Uncho-Useon Building, Department of
Education, Korea University, Anam-Dong, Seongbuk-Gu, Seoul 136-701, Republic of
dynamic student–teacher interactions. In large, diverse, fluid, and
Korea. Fax: +82 2 3290 2926. multi-activity classrooms in which teachers are engrossed in
E-mail address: [email protected] (J. Reeve). instruction, teachers necessarily miss (are unable to monitor) a

0361-476X/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.cedpsych.2011.05.002
258 J. Reeve, C.-M. Tseng / Contemporary Educational Psychology 36 (2011) 257–267

good deal of students’ displays of engagement vs. disengagement. interesting, personable, or optimally challenging), and even creat-
What is missing from an understanding of how students intention- ing or requesting them in the first place, rather than merely react-
ally contribute into the instruction they receive is a direct (rather ing to them as a given. That is, students sometimes try to get ahead
than inferential) path. To better understand this process of how of the lesson-to-come so to offer input that might potentially guide
students contribute constructively into the flow of instruction they its flow toward that which will be more personalized or more en-
receive, as by personalizing it and by enhancing both the lesson riched (i.e., more challenging or more relevant to their needs, inter-
and the conditions under which they learn, we propose the concept ests, and priorities). Therefore, a fuller (and more accurate)
of agentic engagement. portrayal of what happens when the teacher presents students
with a math problem is that students not only react with varying
displays of behavioral, emotional, and cognitive engagement, but
2. Agentic engagement
they also more or less act agentically to try to enrich the learning
activity (look for an opportunity to make the task more enjoyable),
We define agentic engagement as students’ constructive contri-
modify it (make a suggestion, change the level of difficulty), per-
bution into the flow of the instruction they receive. What this new
sonalize what is to be learned (communicate likes and dislikes,
concept captures is the process in which students intentionally and
generate options), afford themselves greater autonomy (express a
somewhat proactively try to personalize and otherwise enrich both
preference, offer input), and gain greater access to the means
what is to be learned and the conditions and circumstances under
needed for better understanding (solicit resources, request
which it is to be learned. For instance, during the flow of instruc-
assistance).
tion, students might offer input, express a preference, offer a sug-
Current conceptualizations of student engagement that empha-
gestion or contribution, ask a question, communicate what they
size only students’ behavioral, emotional, and cognitive involve-
are thinking and needing, recommend a goal or objective to be pur-
ment fall short of capturing the extent to which students
sued, communicate their level of interest, solicit resources or learn-
contribute agentically into the on-going flow of the instruction
ing opportunities, seek ways to add personal relevance to the
they receive. It is one thing to try hard, enjoy, and enact sophisti-
lesson, ask for a say in how problems are to be solved, seek clarifi-
cated learning strategies when exposed to a learning activity, while
cation, generate options, communicate likes and dislikes, or re-
it is another to contribute constructively into modifying what is to
quest assistance such as modeling, tutoring, feedback,
be learned or how it is to be experienced and learned. To the extent
background knowledge, or a concrete example of an abstract
to which students act agentically, they initiate a process in which
concept.
they generate for themselves a wider array of options that expand
To quantify this aspect of student engagement, one pioneering
their freedom of action and increase their chances of experiencing
group of researchers developed the Hit-Steer Observation System
both strong motivation (e.g., autonomy, self-efficacy) and mean-
(Fiedler, 1975; Koenigs, Fiedler, & deCharms, 1977). This system
ingful learning (e.g., internalization, conceptual understanding)
assesses the frequency of students’ attempts to constructively
(Bandura, 2006).
influence the teacher (a ‘‘hit’’) as well as whether those influence
attempts are successful or not in changing the teacher’s subse-
quent behavior (a ‘‘steer’’). A ‘‘hit’’ (influence attempt) reflects
4. Goals and hypotheses of the present study
what students do, and it typifies agentic engagement. A ‘‘steer’’ re-
flects how teachers respond to students’ suggestions and inputs,
The present study had three goals and five hypotheses. The first
and it typifies a teacher’s motivating style (autonomy supportive
goal was to validate a new measure of agentic engagement. A valid
vs. controlling). When trained raters use the Hit-Steer Observation
measure would correlate positively with the other three aspects of
System to score students’ classroom engagement, they find that
engagement, with students’ classroom motivational status, and
students’ influence attempts (and the ratio of these attempts to
with important educational outcomes. First, we proposed that
all influence attempts that occur during instruction—students’
agentic engagement would reflect lesson engaging, rather than les-
and teachers’) (a) correlate positively with students’ perception
son evading or lesson rejecting (from Hansen, 1989). Thus, Hypoth-
of an origin learning climate, (b) occur more frequently in the
esis 1 predicted that agentic engagement would correlate positively
classrooms of autonomy-supportive rather than controlling teach-
and significantly with the other three previously-validated aspects
ers, and (c) correlate positively with students’ academic achieve-
of student engagement. While we did expect agentic engagement
ment (Fiedler, 1975; Koenigs et al., 1977; Reeve, Jang et al., 2004).
to correlate highly and positively with the other three aspects of
engagement, we did not expect the observed intercorrelations to
3. Why agency needs to be added as a fourth aspect of student be so high as to preclude conceptualizing agentic engagement as
engagement a distinct construct (as per Hypothesis 4).
Second, we proposed that agentic engagement would reflect
Students react to the learning activities teachers provide, and constructive aspects of students’ motivation to learn. Thus, Hypoth-
the existing concepts of behavioral engagement, emotional esis 2 predicted that agentic engagement would be closely associ-
engagement, and cognitive engagement nicely capture the extent ated with students’ underlying classroom motivation, as assessed
to which students react to teacher-provided learning activities. in the present study by the extent of students’ psychological need
That is, a teacher might present a math problem for students to satisfaction during instruction. We focused on students’ psycholog-
make sense of (e.g., find the volume of a cylinder) and students ical need satisfaction because we conceptualized it as an exemplar
might react by paying attention or not, enjoying the activity or of students’ constructive classroom motivation (following Ryan &
feeling anxious about it, and utilizing sophisticated or only super- Deci, 2000) and because our program of research emerged out of
ficial learning strategies. Such a linear model (teacher presents a the origin-pawn distinction (following deCharms, 1976, and his
learning activity ? students to some degree engage them- Hit-Steer Observation System).
selves ? students to some degree learn and profit from the experi- Third, we proposed that agentic engagement would contribute
ence) overlooks students’ agentic involvement in the learning positively to students’ learning and performance. Thus, Hypothesis
process (Bandura, 2006). In actuality, students not only react to 3 predicted that agentic engagement would predict academic
learning activities but they also act on them—modifying them, achievement, operationally defined by students’ grades. Impor-
enriching them (e.g., transforming them into something more tantly, Hypothesis 3 predicted that agentic engagement would
 J. Reeve, C.-M. Tseng / Contemporary Educational Psychology 36 (2011) 257–267 259

predict student achievement in a way that was above and beyond 5.2.1. Engagement
the variance in academic achievement explained by the behavioral, We assessed four aspects of student engagement—agentic
emotional, and cognitive aspects—that is, agentic engagement engagement, behavioral engagement, emotional engagement, and
would predict independent or unique variance in achievement. cognitive engagement. All items for each of these four aspects of
The second goal of the study was to test whether agentic engagement appear in Table 1. For each measure, we used the
engagement was a distinct engagement component. Hypothesis 4 same 1–7 bipolar response scale that ranged from ‘‘strongly dis-
predicted that agentic engagement would function as a distinct as- agree’’ to ‘‘strongly agree’’ with ‘‘agree and disagree equally’’ serv-
pect of engagement—one that was intercorrelated with (as per ing as the midpoint (4).
Hypothesis 1), yet was conceptually and statistically distinct from, To assess agentic engagement, we could not rely on a previously
its behavioral, emotional, and cognitive aspects. This prediction validated measure because such a scale of this newly-proposed
follows from our observation that researchers are increasingly rec- concept did not exist. We based the conceptual framework of our
ognizing that each aspect of engagement is distinct in important new measure on the Hit-Steer Observation System. The Hit-Steer
ways. For instance, the behavioral dynamics of engagement are dif- Observation System is for classroom observational purposes, but
ferent from its emotional dynamics in several important ways (e.g., the concept can be extended to questionnaire purposes. To trans-
antecedents, year-to-year developmental change; Skinner et al., late this concept into a self-report scale, we inspected the class-
2008). room observation notes we had from two previous studies (Jang,
The third goal of the study was to test if agentic engagement Reeve, & Deci, 2010; Reeve, Jang et al., 2004) in which two teams
was an important educational construct. Student engagement is of five raters each used the Hit-Steer Observation System during
important principally because it functions to connect students’ a total of 198 different hour-long classroom sessions involving a
motivation to important and highly-valued outcomes (e.g., wide breadth of subject matters to observe, score, and take notes
achievement). Hence, agentic engagement should, just like the on the various ways that middle and high school students’
other three aspects of engagement, function as a mediator to ex- attempted to contribute constructively into the flow of the instruc-
plain the motivation-to-achievement relation. Hypothesis 5 pre- tion they received. From these notes, we identified the most fre-
dicted that agentic engagement, as a latent variable, would quent ways that students proactively and constructively engaged
mediate the effect that student motivation (i.e., psychological need themselves into the flow of the day’s instruction. The five items
satisfaction) might have on student achievement. Further, Hypoth- that emerged from this review reflected categories of behavior
esis 5 predicted that agentic engagement would mediate the rather than specific instances of behavior, and these items are
motivation-to-achievement relation even after including behav- shown at the top of Table 1. In the present study, this five-item
ioral engagement, emotional engagement, and cognitive engage- measure showed adequate internal reliability (alpha = .82).
ment as three additional and complementary mediators of this To assess behavioral engagement, we used Miserandino’s (1996)
same motivation-to-achievement relation. task involvement questionnaire that was based on Wellborn’s
(1991) items and conceptualization of behavioral engagement. This
measure represents the first (principal) factor from her psychomet-
5. Method
ric investigation of a larger ‘‘perceived behavioral engagement’’
questionnaire. In the present study, we removed the two
5.1. Participants and procedure
Table 1
Participants were 369 (65% females, 35% males) high school stu- Questionnaire items to assess the four aspects of engagement.
dents (38% 10th grade, 51% 11th grade, 11% 12th grade) from a
Items to assess agentic engagement
large, middle-class, urban high school in Taipei City, Taiwan. As 1. During class, I ask questions
part of a regularly scheduled study hall, students completed a con- 2. I tell the teacher what I like and what I don’t like
sent form and 3-page survey administered at the beginning of the 3. I let my teacher know what I’m interested in
class period. Participation was voluntary, and scores were confi- 4. During class, I express my preferences and opinions
5. I offer suggestions about how to make the class better
dential and anonymous. We collected the questionnaire data eight
weeks into the semester and the achievement data (semester Items to assess behavioral engagement
1. I listen carefully in class
grade) after the semester ended. Participants rated their learning 2. I try very hard in school
experiences in general across all the classes they were currently 3. The first time my teacher talks about a new topic, I listen very carefully
taking. We were unable to obtain achievement data for four partic- 4. I work hard when we start something new in class
ipants, leaving a final sample size of 365 students, 237 females and 5. I pay attention in class
128 males. Items to assess emotional engagement
1. I enjoy learning new things in class
2. When we work on something in class, I feel interested
5.2. Measures 3. When I am in class, I feel curious about what we are learning
4. Class is fun
Participants self-reported the extent of their classroom engage- Items to assess cognitive engagement
ment and psychological need satisfaction. For each measure, we 1. When doing schoolwork, I try to relate what I’m learning to what I already
know
began with a previously validated questionnaire (except for agen-
2. When I study, I try to connect what I am learning with my own experiences
tic engagement) and then translated that measure into Chinese 3. I try to make all the different ideas fit together and make sense when I
through a professional English–Chinese translator, following the study
guidelines recommended by Brislin (1980). Separate English 4. I make up my own examples to help me understand the important
back-translations were carried out by two graduate students who concepts I study
5. Before I begin to study, I think about what I want to get done
were fluent in both languages and were native Chinese. Any dis-
6. When I’m working on my schoolwork, I stop once in a while and go over
crepancies that emerged between the translators were discussed what I have been doing
until a consensus translation was reached. In addition to the self- 7. As I study, I keep track of how much I understand, not just if I am getting
report measures, we obtained students’ grades for the semester the right answers
8. If what I am working on is difficult to understand, I change the way I learn
in which they completed the survey from their school records so
the material
that we could attain an objective measure of achievement.
260 J. Reeve, C.-M. Tseng / Contemporary Educational Psychology 36 (2011) 257–267

reverse-scored items from her scale—‘‘When I have a hard question We used participants’ scores on the AFS scales to serve as three sep-
or problem in class, I don’t even try’’ and ‘‘When I’m in class, I just act arate indicators of the latent variable, ‘‘Psychological need satisfac-
like I’m working’’—due to non-English speaking students’ difficulty tion.’’ Past research has provided reliability and validity evidence
with these two items in past data sets (i.e., Jang, Reeve, Ryan, & for each AFS scale in that each scale produces scores that are sensi-
Kim, 2009). In the present study, this five-item measure showed tive to classroom variables known to affect psychological need sat-
strong internal reliability (alpha = .94). We chose this particular isfaction (e.g., teachers’ motivating styles), correlates highly with
scale because it represents behavioral engagement as an expression other corresponding measures of psychological need satisfaction
of students’ on-task attention, lesson involvement, and effort (i.e., (e.g., Basic Needs Scale; Gagne’, 2003), and predicts student out-
task involvement rather than school engagement or prosocial con- comes such as classroom engagement and course grades (Hardre
duct, as is sometimes done for the assessment of behavioral engage- & Reeve, 2003; Jang et al., 2009; Reeve, Nix, & Hamm, 2003).
ment in other studies), because it is a near-equivalent to other
widely used and validated behavioral engagement scales (Skinner, 5.2.3. Achievement
Kindermann, & Furrer, 2009), and because past research has shown To assess academic achievement, we used the actual school re-
the scale to be both reliable and valid in terms of its capacity to pre- cord of each student’s overall semester grade, scored at the end of
dict student achievement (Jang et al., 2009; Miserandino, 1996). the semester on a 100-point scale.
To assess emotional engagement, we used several of the posi-
tively-valenced items from Wellborn’s (1991) conceptualization
5.3. Data analyses
of students’ emotional engagement. Items from our adapted mea-
sure did not reflect the entire range of positive and negative aca-
We tested Hypotheses 1–3 with standard statistical methods
demic emotions students experience during task engagement
(zero-order correlations, multiple regression), while we tested
but, rather, reflect those associated with energized emotional
Hypotheses 4 and 5 with structural equation modeling (using LIS-
states (i.e., enjoyment, interest, curiosity, and fun). In the present
REL 8.8; Joreskog & Sorbom, 1996). For Hypothesis 4, we used both
study, this four-item measure showed adequate internal reliability
exploratory and confirmatory factor analyses to test the factor
(alpha = .78). We chose this particular scale because Skinner and
structure and underlying measurement model of the 22-item
her colleagues (2009) showed that their near-equivalent scale
engagement questionnaire (five items for agentic engagement, five
was distinct from, yet supplemental to, the behavioral aspect of
items for behavioral engagement, four items for emotional engage-
engagement and because these researchers showed that their emo-
ment, and eight items for cognitive engagement). For Hypothesis 5,
tional engagement scale correlated both with students’ construc-
we tested the structural model to evaluate the hypothesized
tive motivation (e.g., perceived control beliefs) and with
engagement mediation model (i.e., Motivation ? Engagement ?
important educational outcomes (e.g., achievement).
Achievement).
To assess cognitive engagement, we used Wolters’ (2004) learn-
To evaluate model fit for Hypotheses 4 and 5, we relied on the
ing strategies questionnaire which is a briefer instrument derived
chi-square test statistic and multiple indices of fit (as recommended
from the widely-used Motivated Strategies for Learning Question-
by Kline (2011)), including the standardized root-mean-square
naire (Pintrich, Smith, Garcia, & McKeachie, 1993). This adapted
residual (SRMR; Hu & Bentler, 1999), the root-mean-square error
measure features two subscales, one with items to assess the use
of approximation (RMSEA; Steiger, 1990) and the comparative fit in-
of sophisticated (elaboration-based) learning strategies (items
dex (CFI; Bentler, 1990). In general, statistical values that indicate a
1–4 in Table 1) and a second with items to assess the use of meta-
good fit are .08 or less for the SRMR and RMSEA and .95 or more for
cognitive self-regulation strategies such as planning, monitoring,
the CFI, though it is further important that the set of indicators con-
and revising one’s work (items 5–8 in Table 1). In the present
firm one another to show a good overall collective fit (Hu & Bentler,
study, this combined eight-item measure showed high internal
1999; Kline, 2011). When multiple models are compared (as with
reliability (alpha = .88). We chose this particular scale because
Hypothesis 4), the Akaike information criterion is further used such
scores from this measure have been shown to correlate with
that the lower the AIC value, the better the fit is.
students’ constructive motivational states (mastery goals, self-effi-
cacy), with non-cognitive indices of engagement (effort, persis-
tence), and with course grades (Wolters, 2004). 6. Results

5.2.2. Psychological need satisfaction Prior to testing our hypotheses, we explored for possible gender
We conceptualized the quality of students’ motivation as the ex- and grade level effects on our assessed measures. Gender predicted
tent of psychological need satisfaction they reported experiencing behavioral engagement, t(363) = 2.64, p < .01 [Ms, 5.12 (females) vs.
during instruction. To assess students’ perceived autonomy, 4.79 (males)], but it did not predict any of the other seven measures.
perceived competence, and perceived relatedness (following self- Grade level predicted two measures: agentic engagement,
determination theory; Ryan & Deci, 2000), we used the Activity- F(2, 362) = 5.68, p < .01 [Ms, 3.77 (10th grade) vs. 3.63 (11th grade)
Feelings States (AFS; Reeve & Sickenius, 1994). The AFS offers the vs. 3.06 (12th grade)] and achievement, F(2, 362) = 3.22, p < .05 [Ms,
stem, ‘‘During class, I feel:’’ and lists 14 items. Four items assessed 73.6 (10th grade) vs. 69.9 (11th grade) vs. 69.7 (12th grade)]. Table 2
perceived autonomy: ‘‘free’’; ‘‘I’m doing what I want to be doing’’; shows the descriptive statistics and intercorrelation matrix for gen-
‘‘free to decide for myself what to do’’; and ‘‘I do this with my full der, grade level, and the eight measured variables—four engage-
personal endorsement’’ (alpha = .84). Three items assessed per- ment scales, three psychological need satisfaction scales, and
ceived competence: ‘‘capable’’; ‘‘competent’’; and ‘‘my skills are achievement. As can be seen in the table, correlations among all of
improving’’ (alpha = .83). Three items assessed perceived related- the assessed variables were positive, significant, and in the expected
ness: ‘‘I belong and the people here care about me’’; ‘‘involved with direction.
close friends’’; and ‘‘emotionally close to the people around me’’
(alpha = .85). [The remaining four items assessed perceived pres- 6.1. Agentic engagement’s relation to other aspects of engagement,
sure, which was not included in the present study’s focus on stu- student motivation, and achievement
dents’ constructive motivation.] Each item featured a 1–7 bipolar
response scale that ranged from ‘‘strongly disagree’’ to ‘‘strongly Agentic engagement correlated positively and significantly with
agree’’ with ‘‘agree and disagree equally’’ serving as the midpoint. the other three aspects of engagement (see Table 2), thereby
 J. Reeve, C.-M. Tseng / Contemporary Educational Psychology 36 (2011) 257–267 261

Table 2
Descriptive statistics for and intercorrelations among all the measures.

Variable 1 2 3 4 5 6 7 8 9 10
a * **
1. Gender – .12 .05 .14 .07 .04 .03 .05 .08 .01
2. Grade levelb – .15** .02 .09 .00 .08 .10 .09 .12*
3. Agentic engagement – .36** .46** .48** .35** .43** .35** .48**
4. Behavioral engagement – .42** .59** .25** .40** .32** .41**
5. Emotional engagement – .42** .43** .49** .57** .47**
6. Cognitive engagement – .32** .49** .32** .50**
7. Perceived autonomy – .60** .61** .38**
8. Perceived competence – .64** .42**
9. Perceived relatedness – .42**
10. Achievement –
M – – 3.61 5.01 4.10 4.30 4.88 5.00 5.08 71.3
SD – – 1.24 1.16 1.14 1.05 1.22 1.23 1.32 13.9

N = 365.
a
Gender scored as 0 for males, 1 for females.
b
Grade level scored as 1 for 10th grade, 2 for 11th grade, and 3 for 12th grade.
*
p < .05, two-tailed.
**
p < .01, two-tailed.

supporting Hypothesis 1 and the notion that agentic engagement Table 3 shows the results from the exploratory factor analysis,
overlaps meaningfully with the other three previously-validated using the 22 items listed in Table 1. As expected, four factors
aspects. Agentic engagement also correlated positively and signif- emerged, based on eigenvalue >1, and these four factors accounted
icantly with all three measures of psychological need satisfaction, for 66.6% of the total variance in the questionnaire. All factor load-
thereby supporting Hypothesis 2 and the notion that agentic ings .30 or greater are shown in the table. Three cross-loadings
engagement was associated with students’ constructive motiva- emerged, and all three involved items from the cognitive engage-
tional status. Agentic engagement further correlated positively ment scale cross-loading onto the behavioral engagement scale.
and significantly with achievement, thereby supporting Hypothe- This suggests that while the four items assessing elaboration-based
sis 3 and the notion that the new agentic engagement measure learning strategies (items 1–4) assessed cognitive engagement—or
possessed predictive validity for a key student outcome.1 at least an aspect of engagement that was statistically distinct from
To extend Hypothesis 3 to a test as to whether agentic engage- its behavioral, emotional, and agentic aspects, three of the four
ment could explain independent (i.e., unique) variance in student items assessing metacognitive self-regulation strategies (items 5–
achievement, the achievement measure was regressed simulta- 8) reflected behavioral engagement as much as they reflected cog-
neously on the four measures of engagement (plus gender and nitive engagement.2 Crucial to the purposes of the present paper,
grade level, which were added as control variables). Collectively, however, all five items from the agentic engagement scale loaded
the four components significantly and rather substantially pre- as hypothesized (see factor 2), did not cross-load onto any other
dicted achievement, F(6, 358) = 36.29, p < .01 (R2 = .38). More engagement factor, and no item from the other three aspects of
importantly (for Hypothesis 3), after controlling for the contribu- engagement cross-loaded onto the agentic engagement factor.
tions from the other three engagement components, agentic Table 4 shows the set of statistics used to evaluate the fit of the
engagement explained independent variance in achievement, 12 possible models, using confirmatory factor analysis. The 12 pos-
F(1, 358) = 16.17, p < .01 (beta = .21; change R2 = .031). Further, sible models were as follows: (a) a single-factor model in which all
the extent of its unique contribution compared relatively favorably 22 indicators listed in Table 1 loaded onto a single latent variable
to the extent of unique contribution made by each of the other three (i.e., engagement consists of one unitary factor), (b) all possible
components: behavioral engagement, F(1, 358) = 3.47, p < .07 two-factor models, (c) all possible three-factor models, and (d) a
(beta = .10; change R2 = .005); emotional engagement, F(1, 358) = four-factor model consisting of four separate latent factors (as
19.30, p < .01 (beta = .22; change R2 = .035); and cognitive engage- characterized in Table 4). While none of the models fit the data
ment, F(1, 358) = 20.57, p < .01 (beta = .25; change R2 = .035). according to the chi-square statistic (all ps < .01), the fit indices
(SRMR, RMSEA, CFI, and AIC) collectively suggested that the four-
factor model (model 12) adequately fit the data. Importantly, the
6.2. Four distinct aspects of engagement four-factor model fit the data significantly better than did each of
the other 11 alternative models, as it fit the data significantly bet-
Hypothesis 4 predicted that agentic engagement would be a ter than did the one-factor model, DX2 (D6 df) = 1292.52, p < .01,
distinct aspect of engagement. To assess this, we first conducted significantly better than all four two-factor models, range of
an exploratory factor analysis and then used a series of confirma- DX2’s (D5 df) = 314.77 to 1212.01, ps < .01, and significantly better
tory factor analyses to compare all possible models of how the var-
ious aspects of engagement might combine to define a best-fitting
2
structure. Given that the items assessing metacognitive self-regulatory strategies loaded on
both the cognitive and behavioral factors, we conducted a follow-up exploratory
factor analysis that excluded these four items (items 5–8 from Tables 2 and 3). The
1
Item 1 on the agentic engagement scale (‘‘During class, I ask questions.’’) might factor analysis of the remaining 18 items showed a four-factor solution, as expected,
seem to assess a more reactive response to instruction than the more proactive and accounted for 71.2% of the total variance. Importantly, no cross-loadings emerged
responses assessed by items 2–5. To explore this further, we conducted supplemental on either the behavioral or cognitive factors. The one cross-loading to emerge
analyses with versus without item 1, but the results observed with the four-item involved the ‘‘curiosity’’ item from the emotional engagement scale, as it cross-loaded
scale (that excluded item 1) were virtually identical to the results observed with the .31 on the agentic engagement factor, though the item’s primary loading continued to
full five-item scale. This equivalency applied to the scale alpha coefficients, the be on the emotional factor (.45). Overall, this follow-up analysis suggests that the
correlations reported in Table 2, the exploratory and confirmatory factor analyses reduced 18-item scale successfully and cleanly assessed four distinct aspects of
reported in Tables 3 and 4, and the mediation model illustrated in Fig. 1. engagement.
262 J. Reeve, C.-M. Tseng / Contemporary Educational Psychology 36 (2011) 257–267

Table 3
Factor loadings from an exploratory factor analysis of all 22 items to assess the various aspects of student engagement.

Questionnaire item Factor 1 Factor 2 Factor 3 Factor 4

(38.5%) (12.1%) (9.2%) (6.8%)
Behavioral engagement items
I listen carefully in class .89
I pay attention in class .89
The first time my teacher talks about a new topic, I listen very carefully .89
I try very hard in school .86
I work hard when we start something new in class .84
Agentic engagement items
During class, I express my preferences and opinions .91
During class, I ask questions .90
I tell the teacher what I like and what I don’t like .89
I let my teacher know what I am interested in .68
I offer suggestions about how to make the class better .45
Cognitive engagement items
When doing schoolwork, I try to relate what I’m learning to what I already know .85
When I study, I try to connect what I am learning with my own experiences .83
I try to make all the different ideas fit together and make sense when I study .83
I make up my own examples to help me understand the important concepts I study .71
When what I am working on is difficult to understand, I change the way I learn the material .68
When I’m working on my schoolwork, I stop once in a while and go over what I have been doing .47 .47
As I study, I keep track of how much I understand not just if I am getting the right answers .47 .45
Before I begin to study, I think about what I want to get done .48
Emotional engagement items
When I am in class, I feel curious about what we are learning .88
When we work on something in class, I feel interested .84
I enjoy learning new things in class .78
Class is fun .45
Factor intercorrelations
1. Factor 1 – .29 .44 .30
2. Factor 2 – .36 .27
3. Factor 3 – .25
4. Factor 4 –

than all six three-factor models, range of DX2’s (D4 df) = 61.91– autonomy, perceived competence, and perceived relatedness), gen-
1062.66, ps < .01.3 der, and grade level were entered as predictors of achievement
showed that the data fit the direct-effect motivation-to-achieve-
ment model well, X2 (6) = 9.39, ns, SRMR = .021, RMSEA = .039,
6.3. Agentic engagement as a mediator of the motivation-to- CFI = 0.99. Importantly, the magnitude of the direct effect of moti-
achievement relation vation on achievement was strong (beta = .45, p < .01) and ac-
counted for 21% of the variance in student achievement (while
Hypothesis 5 predicted that agentic engagement would serve as neither gender nor grade level were individually predictive of
a mediator to explain the effect that student motivation might achievement; betas = 0.01, ns, and .0.06, ns, respectively).
have on academic achievement. Students’ mid-semester motiva- To test the prediction that engagement—and agentic engage-
tion did predict their end-of-the-semester achievement, as shown ment in particular—would mediate the motivation-to-achievement
in Table 2. That is, achievement correlated positively and signifi- relation, we performed an analysis in which all four aspects of en-
cantly with perceived autonomy, perceived competence, and per- gagement were included as hypothesized mediators. We also al-
ceived relatedness (range of r’s = .38 to .42, all p’s < .01). A lowed gender to predict behavioral engagement and grade level
preliminary structural equation modeling analysis in which moti- to predict agentic engagement, following the earlier exploratory
vation (i.e., the latent variable of ‘‘psychological need satisfaction’’ analyses involving these two control variables (see Table 2). The
which was composed of participants’ three scores for perceived four-mediator model fit the data fairly well, X2 (328) = 978.25,
p < .01, SRMR = .083, RMSEA = .073, CFI = .96, and, importantly, this
3
To assess whether agentic engagement could be distinguished specifically from model increased the proportion of explained variance in achieve-
behavioral engagement, we compared a two-factor model that constrained the five ment to 30%.
AE items on one factor and the five BE items on a second factor against a one-factor Because we wanted to test if engagement mediated the motiva-
model that constrained all 10 items on a single factor. The two-factor model fit
significantly better than did the one-factor model, DX2 (D1 df) = 915.24, p < .01. To
tion-to-achievement relation, we conducted an additional analysis
assess whether agentic engagement could be distinguished specifically from in which we added psychological need satisfaction and grade level
emotional engagement, we compared a two-factor model that constrained the five as two direct-effect predictors of achievement. We added the path
AE items on one factor and the four EE items on a second factor against a one-factor from psychological need satisfaction to achievement to test if its
model that constrained all 9 items on a single factor. The two-factor model fit
direct effect dropped to nonsignificant after the inclusion of the
significantly better than did the one-factor model, DX2 (D1 df) = 200.25, p < .01. To
assess whether agentic engagement could be distinguished specifically from cognitive four engagement mediators, and we added the path from grade le-
engagement, we compared a two-factor model that constrained the five AE items on vel to achievement because it showed a significant zero-order cor-
one factor and the eight CE items on a second factor against a one-factor model that relation with achievement (see Table 2). Adding these two direct
constrained all 13 items on a single factor. The two-factor model fit significantly paths to the engagement mediation model did not produce a re-
better than did the one-factor model, DX2 (D1 df) = 677.69, p < .01. These analyses
show that the data fit best when the agentic engagement items are kept separate to
vised model that fit significantly better than the four-mediator
load on their own unique latent factor rather than when these same items are merged model reported above, DX2 (D2 df) = 2.98, ns., and the betas
into a latent factor that includes the items from any other engagement scale. for both added paths were non-significant (beta = .08, ns, for
 J. Reeve, C.-M. Tseng / Contemporary Educational Psychology 36 (2011) 257–267 263

psychological need satisfaction and beta = .05, ns, for grade level). Table 4
The R2 for achievement did not increase, and none of the three fit Fit indices associated with 12 models from the confirmatory factor analyses, using
maximum likelihood estimation.
indicators improved (SRMR = .083, RMSEA = .074, CFI = .96). Thus,
collectively, the four aspects of engagement fully mediated the X2 df SRMR RSMEA CFI AIC
motivation-to-achievement relation. One-factor model
The path from agentic engagement to academic achievement Model 1: AE/BE/EE/CEa 1868.89 197 .16 .18 .70 2572.0
was individually significant in both models (beta = .13, p < .01). Two-factor models
The path diagram showing the standardized parameter estimates Model 2: AE + BE/EE/CE 891.14 196 .12 .11 .87 1113.5
in the four-mediator model that also includes the two direct, but Model 3: BE + AE/EE/CE 1056.15 196 .17 .12 .84 1329.6
Model 4: EE + AE/BE/CE 1788.38 196 .13 .17 .71 2487.2
not hypothesized, paths involving psychological need satisfaction Model 5: CE + AE/BE/EE 1684.19 196 .14 .17 .73 2349.3
and grade level appears in Fig. 1.4 As can be seen in the figure, emo-
Three-factor models
tional engagement and cognitive engagement, like agentic engage- Model 6: AE + BE + EE/CE 637.38 194 .087 .079 .92 757.6
ment, explained independent variance in achievement, though, Model 7: AE + EE + BE/CE 806.77 194 .11 .097 .89 978.9
behavioral engagement, in these data, did not.5 Model 8: AE + CE + BE/EE 674.45 194 .11 .082 .91 791.4
Model 9: BE + EE + AE/CE 987.21 194 .16 .11 .86 1199.7
Model 10: BE + CE + AE/EE 659.60 194 .11 .081 .92 774.1
7. Discussion Model 11: EE + CE + AE/BE 1639.03 194 .14 .16 .74 2207.0
Four-factor model
The present study pursued three goals—namely, to validate a Model 12: 576.37 191 .082 .073 .93 681.5
new measure of agentic engagement, to test whether agency was AE + BE + EE + CE
a distinct engagement component, and to determine if agentic
N = 365.
engagement was educationally important by assessing the extent Note. X2 = chi-square statistic; df = degrees of freedom; SRMR = standardized root-
to which it mediated the motivation-to-achievement relationship. mean-square residual; RSMEA = root-square-mean error of approximation;
Results supported all three goals, as agentic engagement (1) covar- CFI = comparative fit index; AIC = Akaike information criterion; AE = agentic
ied with students’ motivation, with other indices of engagement, engagement; BE = behavioral engagement; EE = emotional engagement; and
CE = cognitive engagement.
and with achievement, (2) was conceptually and statistically dis- a
To read the models, when an engagement scale appears by itself (AE), the items
tinct from the three other aspects of engagement, and (3) predicted from that scale were constrained to load only on that factor. When more than one
student achievement even after taking out the variance in achieve- engagement scale appears as a grouped set (AE/BE), the items from those scales
ment that could otherwise be attributed to students’ behavioral, were constrained to load onto one common factor. The + sign means separate fac-
tors; the / sign means a combined factor.
emotional, and cognitive engagement.
Adding agency as a new aspect of student engagement is an
important and worthwhile advance for two key reasons. First,
agentic engagement explained unique and meaningful variance conceptualization that excludes it. The reason why agentic engage-
in students’ achievement (as shown in Fig. 1). This means that, ment contributes uniquely to achievement is presumably because
even after accounting for the contribution of the other three as- it is through intentional, proactive, and constructive acts that stu-
pects of engagement (as well as psychological need satisfaction dents find ways to improve their opportunity to learn by enriching
and grade level), there remained unexplained variance in students’ the learning experience and by enhancing the conditions under
achievement that agentic engagement was able to explain. Hence, which they learn.
a conceptualization of student engagement that includes agentic Second, adding agency as a new aspect of engagement is impor-
engagement is better able to explain achievement than is a tant because it allows for a fuller portrayal of how students engage
themselves in learning activities. Recognizing that students con-
4
We also tested the alternative model that psychological need satisfaction might structively contribute into the instruction they receive clarifies
mediate the direct effect that the four aspects of engagement had on academic the picture of how students learn and profit from potential learn-
achievement. This alternative ‘‘reverse causation’’ model fit the data notably worse ing opportunities. Of course, students voice themselves in all sorts
than did the hypothesized model, as it produced a higher X2 value even though it
of ways—constructively but also defensively and counter-produc-
featured fewer degrees of freedom (i.e., X2 (326) = 1023.71, p < .01), and it explained
only 26% of the variance in achievement. The reason this alternative model fit the data tively, as during lesson evading (e.g., turning to off-task interests
worse than did the hypothesized model (according to an examination of the and entertainments) and lesson rejecting (e.g., resisting a disliked
modification indices from the gamma matrix) was because it failed to include the four teacher; Hansen, 1989). Students might also voice themselves in
otherwise direct and significant paths (unmediated by psychological need satisfac- ways that seem to challenge the teacher’s authority and imply a
tion) from each aspect of engagement to achievement.
5 degree of teaching incompetence, as in ‘‘You are shouting at us
The nonsignificant path from behavioral engagement to achievement (beta = .01,
ns) was surprising. We wondered if the reason why this path failed to reach statistical and I don’t think that kids should be shouted at. We don’t deserve
significance might be attributed to a potential multicollinearity problem with the this and you shouldn’t do it. No one likes it and stop it right now!’’
cognitive engagement scale (i.e., notice the high beta = .62 covariance between the (Winograd, 2002, p. 358). So, it is important to center any discus-
two mediators in Fig. 1). To pursue this possibility, we tested an alternative model sion of agentic engagement on the construct’s conceptual defini-
that included only the four cognitive engagement items assessing learning strategies
(items 1–4 in Tables 1 and 3) and therefore excluded the four items assessing
tion—students’ constructive contribution into the flow of the
metacognitive self-regulatory strategies (items 5–8 in Tables 1 and 3), because these instruction they receive.
latter items loaded as much on the behavioral factor as they did on the cognitive
factor in the factor analysis reported in Table 2. This alternative model did fit the data
fairly well, X2 (233) = 600.39, p < .01, SRMR = .071, RMSEA = .063, CFI = .97, and it did 7.1. Agentic engagement within the student–teacher dialectical
decrease the covariance between the cognitive engagement and behavioral engage-
framework
ment latent variables (beta = .62 from Fig. 1 decreased to beta = .49). This reduced
model did not, however, explain any additional variance in achievement (R2 actually
decreased to 29%). As suspected, the magnitude of the standardized betas for the four One theoretical framework to conceptualize the mutual effects
engagement mediators did somewhat increase in the reduced model (betas for the that teachers and students have on each other is the student–
behavioral, emotional, cognitive, and agentic engagement latent variables predicting teacher dialectical framework within self-determination theory
the achievement outcome increased from .13, .01, .19, and .26 in the original model,
as shown in Fig. 1, to .15, .07, .27, and .31, respectively, in the reduced model), but the
(SDT; Reeve, Deci, & Ryan, 2004). In this framework, (a) a teacher’s
individual path from behavioral engagement to achievement remained non- motivating style (and classroom contextual factors more generally)
significant, beta = .07, ns (t = 1.54). affects students’ motivation, (b) changes in students’ underlying
264 J. Reeve, C.-M. Tseng / Contemporary Educational Psychology 36 (2011) 257–267

-.15 Agentic
Grade Level Engagement
(R 2 = .24)

-.12 .38
.48 .13
-.05
.15 Behavioral
-.11 Gender Engagement
(R 2 = .20) Academic
.01 Achievement
(R 2 = .30)
.04 .08
.46 .44
.42 .19
Psychological
Need .77 Emotional .26
Satisfaction Engagement
(R 2 = .61)
.51 .62
.50
.40

Cognitive
Engagement
(R 2 = .26)

Fig. 1. Standardized parameter estimates for the engagement mediation model of the motivation-to-achievement relation. Bold lines represent hypothesized paths, while curved
lines represent intercorrelations among the predictor variables. Solid straight lines represent significant paths, p < .01, while dashed lines represent non-significant paths. The
numbers adjacent to the lines represent standardized parameter estimates. Ovals represent latent variables, while rectangles represent observed variables. For clarity of
presentation, the 25 individual indicators for the five latent variables are not shown. The correlations of the disturbances among the four mediators are shown so to illustrate
the extent of statistical overlap among the four aspects of engagement.

motivational states (e.g., interest, psychological need satisfaction, Agentic engagement is important not only to a SDT conceptual-
goals) are expressed through changes in students’ engagement, ization of student motivation but to perhaps all major theories of
and (c) changes in engagement in turn feedback to affect on-going student motivation. All motivation theories of interest to educa-
changes in the teacher’s motivating style toward the student. For tional psychologists highlight the unobservable psychological pro-
instance, when teachers are autonomy supportive (rather than cesses that energize and direct students’ observable effort, interest,
controlling) early in the semester, students’ psychological need and strategic involvements, and the concepts of behavioral engage-
satisfaction and engagement increase by mid-semester, and teach- ment, emotional engagement, and cognitive engagement, respec-
ers then adjust their motivating styles by the end of the semester tively, correspond nicely to these three categories of observable
in response to students’ rising or falling engagement as they be- motivated action. What adding the concept of agentic engagement
come significantly more autonomy supportive with engaged stu- can do for any view of student motivation is to draw greater atten-
dents but significantly more controlling with disengaged tion to students’ intentional, proactive, and origin-like motivated
students (Jang, Kim, & Reeve, in preparation). involvement in these same learning activities. While the present
It makes sense to put student agency at the center of the study focused on students’ psychological need satisfaction, several
student–teacher dialectic because students’ agentic engagement agency-based motivational constructs seem especially ripe to ben-
can be conceptualized as the ideal complement to a teacher’s efit from attention to students’ agentic engagement, including self-
autonomy-supportive motivating style. That is, agentic engage- efficacy, personal goals, possible selves, individual interests, and a
ment involves students expressing opinions, communicating inter- mastery goal orientation. For instance, Bandura (1997) argued that
ests, and asking questions, while autonomy support involves self-efficacy is the very foundation of human (i.e., student) agency.
creating the classroom conditions in which students feel free to ex-
press opinions, pursue interests, and ask questions. For instance, 7.2. Educational constructs similar to agentic engagement
consider the following three items that appear on the Learning Cli-
mate Questionnaire, a measure widely used to assess students’ We propose agentic engagement as a new educational con-
perceptions of how autonomy-supportive their teachers are (e.g., struct—a newly-proposed fourth aspect of students’ engagement
Black & Deci, 2000; Jang et al., 2009): (1) My teacher listens to during learning activities. Still, other programs of research have fo-
how I would like to do things; (2) My teacher tries to understand cused on similar student behaviors and classroom processes. Some
how I would like to do things before suggesting a new way to do of these behaviors are near-equivalents to the current concept of
them; and (3) My teacher encourages me to ask questions. These agentic engagement. One near-equivalent is student input using
items pair up strikingly well to items 4, 3, and 1 from the agentic instructional technologies such as a personal response system (or
engagement measure shown in Table 1. ‘‘clickers’’) or classroom response systems (CRSs) in which students
 J. Reeve, C.-M. Tseng / Contemporary Educational Psychology 36 (2011) 257–267 265

press a button on a hand-held remote control device to communi- research pursuit would be to explore ways that might more fully
cate their answer to, or to express their opinion on, a teacher- characterize the process in which students intentionally, proac-
prepared question projected on a screen (Mayer et al., 2009). While tively, and constructively contribute into the flow of instruction
such student input does contribute constructively into the on- they receive.
going flow of instruction, it is nevertheless reactive (rather than To help advance that goal, we identify here what we believe to
proactive). That said, a study using the Turning Point active re- be five essential characteristics of agentic engagement: (1) It is
sponse system (whole-class questions built into PowerPoint lec- proactive (occurs before or during, rather than after, the learning
ture slides) showed that students afforded this type of input did activity); (2) it is intentional (deliberate and purposeful); (3) it
initiate a greater number of verbal responses (one likely manifes- tries to enrich the learning opportunity (by making it more per-
tation of high agentic engagement) than did a control group of stu- sonal, interesting, challenging, or valued); (4) it contributes con-
dents (Harper, 2009). structive input into the planning or on-going flow of instruction
Formative assessment is another near-equivalent concept. Dur- so that the student has a say in the conditions under which he or
ing formative assessment, teachers utter verbal prompts or hand she learns; and (5) it does not connote teacher incompetence or
out index cards asking, ‘‘Any suggestions?’’ Indeed, each of the five ineffectiveness. Some possible items that may meet these criteria
items listed in Table 1 to represent agentic engagement might include the following: ‘‘I let the teacher know what I am thinking
serve well as an open-ended formative assessment (e.g., ‘‘Any and needing’’; ‘‘I make whatever we are learning as relevant to
questions?’’, ‘‘What about this class did you like or dislike?’’, and my life as possible’’; ‘‘I speak up whenever I think I can add some-
‘‘Any suggestions about how we might make tomorrow’s class bet- thing important to the flow of the class’’; ‘‘When a lesson is excit-
ter?’’). Teachers elicit and obtain student feedback in a number of ing and interesting, I let my teacher know that I like it’’; and ‘‘When
ways, and it is likely that each of these affords students an oppor- I need something, I’ll ask the teacher for it instead of just suffering
tunity for agentic engagement (for a review of these student feed- quietly.’’
back techniques, see Richardson, 2005). Future improvements are needed not only for the assessment of
Other existing educational constructs overlap, yet are distinct agentic engagement, but for the assessment of cognitive engage-
from, our concept of agentic engagement. Instrumental help seek- ment as well. In the present study, items assessing the metacogni-
ing (or ‘‘adaptive help seeking’’) involves students actively seeking tive and self-regulatory aspects of cognitive engagement
out teacher-provided assistance (e.g., hints when stuck) so that unexpectedly cross-loaded onto the behavioral factor (see Table
they can complete an assignment (Karabenick, 1998; Karabenick 3). Further, when we used only the items assessing the use of
& Newman, 2006; Pajares, Cheong, & Oberman, 2004). Unlike agen- sophisticated learning strategies, the cross-loadings disappeared
tic engagement, instructional help seeking does not generally cor- (see footnote 2). This implies that it might be helpful to narrow
relate with academic achievement, a finding that is likely due to its the conceptualization of cognitive engagement down to the use
reactive nature. A somewhat similar construct is the use of behav- of sophisticated learning strategies (e.g., elaboration, paraphrasing,
iorally-oriented boredom-related coping strategies (Nett, Goetz, & summarizing) that enable deep and personally meaningful, rather
Hall, 2011). Strategies such as ‘‘asking the teacher if we can do than superficial, learning. Students’ metacognition and self-
something else’’ seem to represent constructive contributions into regulation during learning activities are certainly important as-
the flow of instruction, at least from the student’s point of view. pects of engagement, though they seem to reflect engagement’s
The concept of ‘‘strategies for regulating motivation’’ is also similar behavioral aspect as much as its cognitive aspect.
(Wolters, 2003), as it, like the concept of agentic engagement, fo- The theoretical effort to clarify the conceptual nature of cogni-
cuses on the process in which students take a purposive role in tive engagement has been an ongoing debate (Pintrich, 2000,
their own learning. Taking a purposive role in one’s learning is a 2004; Pintrich & De Groot, 1990; Zimmerman, 2002), and the
concept that is closely related to our concept of making an inten- findings in the present study suggest that learning strategies tap
tional, proactive, and constructive contribution into the flow of uniquely into cognitive engagement while metacognitive self-
the instruction one receives. regulatory strategies tap into a more general construct that
The literature on student–teacher negotiations of classroom confounds cognitive engagement with behavioral engagement. If
power is also related (e.g., Schrodt et al., 2008; Sproston, 2008), future studies are to better understand how cognitive engagement
as the basic principle of negotiated power is that students need uniquely contributes to students’ positive outcomes, then they
to be allowed by teachers to negotiate various aspects of the class- would be well advised to narrow their conceptual and operational
room curriculum and decision-making, such as rules and evalua- definitions of cognitive engagement to include only the use of
tions. Like similar literatures on responsive or authoritative sophisticated learning strategies. That said, perhaps additional as-
teaching (Wentzel, 2002) and on constructivist approaches to pects of cognitive engagement that have not yet been the focus of
teaching (Glaserfeld, 1989; Prawat, 1992), the focus is on what research might be considered, including perhaps mental simula-
teachers need to do to empower (and engage) students in class- tions as emphasized within the talent development literature
room activities. Thus, these literatures relate to the present con- (Ericsson, Krampe, & Tesch-Romer, 1993), critical thinking (Elliot,
cept of agentic engagement in that they may well predict McGregor, & Gable, 1999), and the higher aspects of Bloom’s taxon-
teacher-related behaviors that facilitate relatively high levels of omy (analysis, evaluation, synthesis).
students’ agentic engagement.
7.4. Limitations
7.3. Assessment issues for future research
We acknowledge four limitations and potential criticisms with-
Perhaps the most pressing issue for future research is to more in our investigation. First, while we conceptualized agentic engage-
adequately assess the agentic engagement construct. In the present ment as a class-specific phenomenon (or even as a learning
study, we created a brief measure based on our observations of activity-specific phenomenon), we actually assessed it in a way
students’ actual classroom behavior that represented a student that collapsed students’ engagement ratings across all their current
‘‘hit’’ (an influence attempt) within the Hit-Steer Observation classes. Assessing engagement this way was a necessary procedure
System (discussed in the Introduction). While adequate for the in our study, however, because we knew in advance of the data col-
present purposes, it is conceivable that the agentic engagement lection effort that we would have access only to students’ semester
construct is a richer one than we portrayed. A promising future grades (not to their individual class grades). Because our top
266 J. Reeve, C.-M. Tseng / Contemporary Educational Psychology 36 (2011) 257–267

concern going into the study was to test if agentic engagement 8. Conclusion
could predict students’ achievement, we decided to assess engage-
ment at the same level as the achievement data. That said, we Students vary in how they react to the learning activities their
acknowledge that assessing student engagement (and achieve- teachers provide, as some students work harder, with greater joy,
ment) at the classroom level is the better and more appropriate and more strategically. These behavioral, emotional, and cognitive
procedure, as it is entirely possible that a student might show differences are important in predicting students’ learning and
strong engagement in one class and with one teacher yet show achievement. But students further vary in how much or how little
weak engagement in another class. Future research, therefore, they purposively work to have a say in their learning opportunities,
would be best served by obtaining students’ engagement and indi- as by offering suggestions as to how they might be enriched, per-
ces of achievement (grades, performance, learning, skill develop- sonalized, or generally improved upon. The findings in the present
ment, academic progress) at the class (or learning activity) level. study showed that such agentic engagement was conceptually dis-
Second, our sample of participants included only high-school tinct from the three other three aspects of engagement, that it cor-
students from Taiwan. It is not yet clear how constructive agentic related significantly with a constructive aspect of students’
engagement might evidence itself to be when the setting changes motivation, and that it predicted independent variance in students’
to students of different grade levels and to students with different achievement. Such a pattern of results opens the door to future
ethnic and cultural backgrounds. Indeed, an interesting question work that seeks to more fully understand how students learn
for future research would be to ask if agentic engagement is more and also how educational psychologists can better appreciate stu-
predictive of student outcomes at one grade level rather than an- dents’ constructive contributions into their own learning.
other. After all, those who study elementary students tend to focus
disproportionally on the behavioral aspects of engagement
(Alexander, Entwisle, & Dauber, 1993; Birch & Ladd, 1997), while Acknowledgments
those who study high school students focus disproportionally on
the cognitive aspects of engagement (Greene, Miller, Crowson, We thank Mimi Bong, Editor Krista R. Muis, and four anony-
Duke, & Akey, 2004). While we recognize this sample limitation mous reviewers for their insights and comments on earlier ver-
as both real and important, our decision to sample students from sions of the article.
an Eastern nation was actually an intentional one, as the prototyp- This research was supported by the World Class University
ical classroom script in Chinese schools is highly teacher-centered (WCU) Program funded by the Korean Ministry of Education,
and somewhat antagonistic to students’ classroom agency (though Science and Technology, consigned to the Korea Science and
the mean agentic engagement score in the present study was a Engineering Foundation (Grant No. R32-2008-000-20023-0).
respectable 3.61 on a 1–7 scale; see Table 2). While the generaliz-
ability of our findings to more diverse samples is in question, we
References
nevertheless believe that future research will show that our data
from secondary students in an Eastern nation actually underesti- Alexander, K. L., Entwisle, D. R., & Dauber, S. L. (1993). First-grade classroom
mate the role that agentic engagement plays in students’ learning behavior: Its short- and long-term consequences for school performance. Child
and achievement. Development, 64, 801–814.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: W.H. Freeman.
Third, our study assessed only the positive face of engagement Bandura, A. (2006). Toward a psychology of human agency. Perspectives on
during instruction—engagement rather than disaffection. It is not Psychological Science, 1, 164–180.
clear how much of a limitation this omission was to the present Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological
Bulletin, 107, 238–246.
study, because our positively-valenced measures did rather Birch, S. H., & Ladd, G. W. (1997). The student–teacher relationship and children’s
adequately explain the motivation-to-achievement relation. Still, early school adjustment. Journal of School Psychology, 35, 61–79.
other research has shown that the disaffected face of engagement Black, A. E., & Deci, E. L. (2000). The effects of instructors’ autonomy support and
students’ autonomous motivation on learning organic chemistry: A self-
is important to understanding how students behave, learn, and determination theory perspective. Science Education, 84, 740–756.
achieve during learning activities (Skinner, Kindermann, & Furrer, Brislin, R. (1980). Translation and content analysis of oral and written material. In H.
2009). This observation raises the issue of how important it may C. Triandis & J. W. Berry (Eds.), Handbook of cross-cultural psychology (Vol. 2,
pp. 389–444). Boston: Allyn & Bacon.
be to conceptualize and assess the disaffected side of agentic
Christenson, S. L., Reschly, A. L., & Wylie, C. (2011). The handbook of research on
engagement—that is, those occasions when students sit passively student engagement. New York: Springer Science.
and simply take whatever instruction teachers provide them. It is Connell, J. P., & Wellborn, J. G. (1991). Competence, autonomy, and relatedness: A
not yet clear that students suffer from simply and unquestionably motivational model of self-system processes. Minnesota Symposia on Child
Psychology, 23, 43–77.
taking whatever instruction teachers provide, and it is not yet Davidson, A. (1999). Negotiating social differences: Youth’s assessments of
clear just what the opposite (disaffected face) of agentic engage- educator’s strategies. Urban Education, 34, 338–369.
ment is. deCharms, R. (1976). Enhancing motivation: Change in the classroom. New York:
Irvington.
Fourth, one argument against adding agency as an engagement Elliot, A. J., McGregor, H. A., & Gable, S. (1999). Achievement goals, study strategies,
component might be that some students are more vocal or more and exam performance. A meditational analysis. Journal of Educational
assertive than are others. Hence, agentic engagement might be Psychology, 91, 549–563.
Ericsson, K. A., Krampe, R. T., & Tesch-Romer, C. (1993). The role of deliberate
too confounded with student characteristics such as extraversion. practice in the acquisition of expert performance. Psychological Review, 100,
While this may be true, the same criticism applies in equal mea- 363–406.
sure to the other three aspects of engagement. The observations Fiedler, M. L. (1975). Bidirectionality of influence in classroom interaction. Journal of
Educational Psychology, 67, 735–744.
that some students are more behaviorally active than are others, Fredricks, J. A., Blumenfeld, P. C., & Paris, A. H. (2004). School engagement: Potential
are emotionally happier and less anxious than are others, and are of the concept, state of the evidence. Review of Educational Research, 74, 59–109.
more intelligent than are others are all parallel confounds inherent Furrer, C., & Skinner, E. A. (2003). Sense of relatedness as a factor in children’s
academic engagement and performance. Journal of Educational Psychology, 95,
within the concepts of behavioral engagement, emotional engage-
148–162.
ment, and cognitive engagement as well. Overall, however, it Gagne’, M. (2003). The role of autonomy support and autonomy orientation in
seems that educators’ understanding of how students learn and prosocial behavior engagement. Motivation and Emotion, 27, 199–223.
profit from potential learning experiences can only be enhanced Glaserfeld, E. (1989). Cognition, construction of knowledge, and teaching. Synthese,
80, 121–140.
by adding agency as a fourth aspect of students’ engagement dur- Gottfried, A. E. (1990). Academic intrinsic motivation in young elementary school
ing learning activities. children. Journal of Educational Psychology, 82, 525–538.
 J. Reeve, C.-M. Tseng / Contemporary Educational Psychology 36 (2011) 257–267 267

Greene, B. A., Miller, R. B., Crowson, M., Duke, B. L., & Akey, K. L. (2004). Predicting Pintrich, P. R., & De Groot, E. V. (1990). Motivational and self-regulated learning
high school students’ cognitive engagement and achievement: Contributions of components of classroom academic performance. Journal of Educational
classroom perceptions and motivation. Contemporary Educational Psychology, Psychology, 82, 33–40.
29, 462–482. Pintrich, P. R., Smith, D., Garcia, T., & McKeachie, W. (1993). Predictive validity and
Hansen, D. A. (1989). Lesson evading and lesson dissembling: Ego strategies in the reliability of the Motivated Strategies for Learning Questionnaire. Educational
classroom. American Journal of Education, 97, 184–208. and Psychological Measurement, 53, 801–813.
Hardre, P. L., & Reeve, J. (2003). A motivational model of rural students’ intentions to Prawat, R. S. (1992). Teachers beliefs about teaching and learning: A constructivist
persist in, versus drop out of, high school. Journal of Educational Psychology, 95, perspective. American Journal of Education, 3, 215–226.
347–356. Reeve, J., Deci, E. L., & Ryan, R. M. (2004). Self-determination theory: A dialectical
Harper, B. E. (2009). ‘I’ve never seen or heard it this way! Increasing student framework for understanding the sociocultural influences on student
management through the use of technology-enhanced feedback. Teaching motivation. In D. McInerney & S. Van Etten (Eds.), Research on sociocultural
Educational Psychology, 3, 1–8. influences on motivation and learning: Big theories revisited (Vol. 4, pp. 31–59).
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure Greenwich, CT: Information Age.
analysis: Conventional criteria versus new alternatives. Structural Equation Reeve, J., Jang, H., Carrell, D., Jeon, S., & Barch, J. (2004). Enhancing students’
Modeling, 6, 1–55. engagement by increasing teachers’ autonomy support. Motivation and Emotion,
Jang, H., Kim, E. J., & Reeve, J. (in preparation). Longitudinal test of self- 28, 147–169.
determination theory’s motivation mediation model in a naturally-occurring Reeve, J., Nix, G., & Hamm, D. (2003). Testing models of the experience of self-
classroom context. determination in intrinsic motivation and the conundrum of choice. Journal of
Jang, H., Reeve, J., & Deci, E. L. (2010). Engaging students in learning activities: It is Educational Psychology, 95, 375–392.
not autonomy support or structure but autonomy support and structure. Journal Reeve, J., & Sickenius, B. (1994). Development and validation of a brief measure of
of Educational Psychology, 102, 588–600. the three psychological needs underlying intrinsic motivation: The AFS scales.
Jang, H., Reeve, J., Ryan, R. M., & Kim, A. (2009). Can self-determination theory Educational and Psychological Measurement, 54, 506–515.
explain what underlies the productive, satisfying learning experiences of Richardson, J. T. E. (2005). Instruments for obtaining student feedback: A review of
collectivistically-oriented South Korean adolescents? Journal of Educational the literature. Assessment & Evaluation in Higher Education, 30, 387–415.
Psychology, 101, 644–661. Ryan, R. M., & Deci, E. L. (2000). Self-determination theory and the facilitation of
Jimerson, S. J., Campos, E., & Grief, J. L. (2003). Toward an understanding of intrinsic motivation, social development, and well-being. American Psychologist,
definitions and measures of school engagement and related terms. The 55, 68–78.
California School Psychologist, 8, 7–27. Schrodt, P., Witt, P., Myers, S., Turman, P., Barton, M., & Jernberg, K. (2008). Learner
Joreskog, K. G., & Sorbom, D. (1996). LISREL 8: Structural equation modeling with the empowerment and teacher evaluations as functions of teacher power use in the
SIMPLIS command language. Hillsdale, NJ: Scientific Software International. college classroom. Communication Education, 57, 180–200.
Karabenick, S. A. (Ed.). (1998). Strategic help seeking: Implications for learning and Skinner, E. A., & Belmont, M. J. (1993). Motivation in the classroom: Reciprocal
teaching. Mahwah, NJ: Lawrence Erlbaum. effects of teacher behavior and student engagement across the school year.
Karabenick, S. A., & Newman, R. S. (Eds.). (2006). Help seeking in academic settings: Journal of Educational Psychology, 85, 571–581.
Goals, groups, and contexts. Hahwah, NJ: Lawrence Erlbaum. Skinner, E. A., Zimmer-Gembeck, M. J., & Connell, J. P. (1998). Individual differences
Kline, R. B. (2011). Principles and practice of structural equation modeling (3rd ed.). and the development of perceived control. Monographs of the Society for
New York: Guilford Press. Research in Child Development, 63, (2–3, Whole No. 204).
Koenigs, S. S., Fiedler, M. L., & deCharms, R. (1977). Teacher beliefs, classroom Skinner, E. A., Furrer, C., Marchand, G., & Kindermann, T. (2008). Engagement and
interaction and personal causation. Journal of Applied Social Psychology, 7, disaffection in the classroom: Part of a larger motivational dynamic? Journal of
95–114. Educational Psychology, 100, 765–781.
Ladd, G. W., & Dinella, L. M. (2009). Continuity and change in early school Skinner, E. A., Kindermann, T. A., Connell, J. P., & Wellborn, J. G. (2009a). Engagement
engagement: Predictive of children’s achievement trajectories from first to and disaffection as organizational constructs in the dynamics of motivational
eighth grade? Journal of Educational Psychology, 101, 190–206. development. In K. Wentzel & A. Wigfield (Eds.), Handbook of motivation in
Mayer, R. E., Stull, A., DeLeeuw, K., Almeroth, K., Bimber, B., Chun, D., et al. school. Malwah, NJ: Erlbaum.
(2009). Clickers in college classrooms: Fostering learning with questioning Skinner, E. A., Kindermann, T. A., & Furrer, C. J. (2009b). A motivational perspective
methods in large lecture classrooms. Contemporary Educational Psychology, 34, on engagement and disaffection: Conceptualization and assessment of
51–57. children’s behavioral and emotional participation in academic activities in the
Miserandino, M. (1996). Children who do well in school: Individual differences in classroom. Educational and Psychological Measurement, 69, 493–525.
perceived competence and autonomy in above-average children. Journal of Sproston, C. (2008). When students negotiate: An action research case study of a
Educational Psychology, 88, 203–214. Year 8 English class in a secondary college in Victoria, Australia. Educational
Murray, C., & Greenberg, M. T. (2000). Children’s relationship with teachers and Action Research, 16, 187–208.
bonds with school. Journal of School Psychology, 38, 423–445. Steiger, J. H. (1990). Structural model evaluation and modification: An interval
National Research Council (2004). Engaging schools: Fostering high school students’ estimation approach. Multivariate Behavioral Research, 25, 173–180.
motivation to learn. Washington, DC: The National Academies Press. Turner, J., Thorpe, P., & Mayer, D. (1998). Students’ reports of motivation and
Nett, U. E., Goetz, T., & Hall, N. C. (2011). Coping with boredom in school: An negative affect: A theoretical and empirical analysis. Journal of Educational
experience sampling perspective. Contemporary Educational Psychology, 36, Psychology, 90, 758–771.
49–59. Wellborn, J. G. (1991). Engaged and disaffected action: The conceptualization and
Pajares, F., Cheong, Y. F., & Oberman, P. (2004). Psychometric analysis of computer measurement of motivation in the academic domain. Unpublished doctoral
science help-seeking scales. Educational and Psychological Measurement, 64, dissertation, University of Rochester.
496–513. Wentzel, K. R. (1997). Student motivation in middle school: The role of perceived
Pelletier, L. G., Seguin-Levesque, C., & Legault, L. (2002). Pressure from above and pedagogical caring. Journal of Educational Psychology, 89, 41–419.
pressure from below as determinants of teachers’ motivation and teaching Wentzel, K. R. (2002). Are effective teachers like good parents? Teaching styles and
behaviors. Journal of Educational Psychology, 94, 186–196. student adjustment in early adolescence. Child Development, 73, 287–301.
Pelletier, L. G., & Vallerand, R. J. (1996). Supervisors’ beliefs and subordinates’ Winograd, K. (2002). The negotiative dimension of teaching: Teachers sharing
intrinsic motivation: A behavioral confirmation analysis. Journal of Personality power will the less powerful. Teaching and Teacher Education, 18, 343–362.
and Social Psychology, 71, 331–340. Wolters, C. A. (2003). Regulation of motivation: Evaluating an underemphasized
Pintrich, P. R. (2000). The role of goal orientation in self-regulated learning. In M. aspect of self-regulated learning. Educational Psychologist, 38, 189–205.
Boekaerts, P. R. Pintrich, & M. Zeidner (Eds.), Handbook of self-regulation Wolters, C. A. (2004). Advancing achievement goal theory: Using goal structures
(pp. 451–502). San Diego, CA: Academic. and goal orientations to predict students’ motivation, cognition, and
Pintrich, P. R. (2004). A conceptual framework for assessing motivation and self- achievement. Journal of Educational Psychology, 96, 236–250.
regulated learning in college students. Educational Psychology Review, 16, Zimmerman, B. J. (2002). Becoming a self-regulated learner: An overview. Theory
385–407. into Practice, 41, 64–70.