The Classroom Environment and Students' Reports of Avoidance Strategies in Mathematics: A Multimethod Study

Journal of Educational Psychology
Copyright 2002 by the American Psychological Association, Inc.
2002, Vol. 94, No. 1, 88 –106
0022-0663/02/$5.00
DOI: 10.1037//0022-0663.94.1.88
The Classroom Environment and Students’ Reports of Avoidance
Strategies in Mathematics: A Multimethod Study
Julianne C. Turner
Carol Midgley
University of Notre Dame
University of Michigan
Debra K. Meyer
Margaret Gheen
Elmhurst College
University of Michigan
Eric M. Anderman
Yongjin Kang
University of Kentucky
University of Michigan
Helen Patrick
Northern Illinois University
The relation between the learning environment (e.g., students’ perceptions of the classroom goal structure
and teachers’ instructional discourse) and students’ reported use of avoidance strategies (self-
handicapping, avoidance of help seeking) and preference to avoid novelty in mathematics was examined.
Quantitative analyses indicated that students’ reports of avoidance behaviors varied significantly among
classrooms. A perceived emphasis on mastery goals in the classroom was positively related to lower
reports of avoidance. Qualitative analyses revealed that teachers in high-mastery/low-avoidance and
low-mastery/high-avoidance classrooms used distinctively different patterns of instructional and moti-
vational discourse. High incidence of motivational support was uniquely characteristic of high-mastery/
low-avoidance classrooms, suggesting that mastery goals may include an affective component. Impli-
cations of the results for both theory and practice are discussed.
By early adolescence some students have begun to purposefully
mine performance and may contribute to the devaluation of learn-
withdraw effort, resist novel approaches to learning, and avoid
ing and dropping out of school. Why would students engage in
seeking academic help when they need it. These avoidance strat-
behaviors that undermine their performance and limit their ability
egies, often adopted to deflect attention from low ability, under-
to learn in different and original ways? Anecdotal information
gathered in conversations with teachers suggests that they may
ascribe avoidance strategies to factors such as laziness, devaluing
Julianne C. Turner, Institute for Educational Initiatives and Department
of school, and lack of parental support.
of Psychology, University of Notre Dame; Carol Midgley, Margaret
The aim of this study was to examine aspects of the learning
Gheen, and Yongjin Kang, Combined Program in Education and Psychol-
environment—the goals for achievement that are emphasized in
ogy, University of Michigan; Debra K. Meyer, Department of Education,
the classroom and instructional practices—that are related to stu-
Elmhurst College; Eric M. Anderman, Department of Educational and
dents’ reports of the use of avoidance strategies in mathematics.
Counseling Psychology, University of Kentucky; Helen Patrick, Department
Using a multimethod approach (e.g., Tashakkori & Teddlie, 1998)
of Educational Psychology and Foundations, Northern Illinois University.
classroom contexts related to three avoidance strategies were ex-
Helen Patrick is now at the Department of Educational Studies, Purdue
amined: students’ use of strategies aimed at withdrawing effort,
University.
resisting novelty, and avoiding seeking academic help. If aspects
The research reported in the article was made possible by Grant
199800210 from the Spencer Foundation to Julianne C. Turner and Carol
of the learning environment exacerbate or ameliorate the use of
Midgley. The data presented, the statements made, and the views expressed
avoidance strategies, teachers can adjust their practices and
are solely the responsibility of the authors.
thereby have a positive influence on student performance. These
Carol Midgley, our colleague, mentor, and friend, passed away on
research questions are addressed:
November 23, 2001. We will always value her clear thinking, her incisive
1. How do students’ perceptions of the classroom goal structure
analyses, and her dedication to young adolescents. We thank all of the
relate to their reports of the use of avoidance strategies?
teachers and students who participated in this research as well as Haya
2. How does teachers’ use of instructional discourse relate to
Shamir, Christine Willard, Michelle Kramer, Jessica Ziembroski, Eileen
students’ perceptions of the classroom goal structure and to their
McConnell, Pam Veldman, Denise Talotta, and Tina Durocher-Schudlich,
reports of the use of avoidance strategies?
who provided invaluable assistance with data collection.
The main contribution of the present study is providing a qual-
Correspondence concerning this article should be addressed to Julianne
C. Turner, 270 Institute for Educational Initiatives Building, University of
itative examination of teachers’ discourse practices to help explain
Notre Dame, Notre Dame, Indiana 46556. E-mail: turner.37@nd.edu
why students use avoidance strategies. Another contribution is the
88

---

THE CLASSROOM ENVIRONMENT AND AVOIDANCE STRATEGIES
89
specific focus on both instructional and affective–motivational
practices that provide insights into these students’ perceptions of
dimensions of teacher discourse as they relate to student avoidance
the goal structure and their use of avoidance strategies are
strategies.
examined.
Roots of Avoidance Behaviors
Students’ Perceptions of the Classroom Goal Structure
and Avoidance Behaviors
Covington (1992) asserted that the search for self-acceptance is
the highest human priority and that “in schools self-acceptance
Many messages about the purposes for achievement behaviors
comes to depend on one’s ability to achieve competitively” (p. 74).
are communicated by teachers and perceived by students (e.g.,
For many students, “to be able is to be worthy, but to do poorly is
Ames, 1992b; Maehr & Midgley, 1996; Midgley, 1993; Nolen,
evidence of inability and is reason to despair” (p. 78). To protect
1988). One message that may be communicated to students is that
self-worth, students who are uncertain about their ability to
demonstrating ability and outperforming others are the reasons for
achieve competitively may develop strategies that deflect attention
engaging in academic behavior (i.e., a performance goal structure).
from their ability.
Another message may be that understanding, intellectual develop-
There are several strategies available to students who seek to
ment, and improvement are the reasons for engaging in academic
protect self-worth by deflecting attention from their ability. Cov-
behavior (i.e., a mastery goal structure). When the emphasis is on
ington called these strategies “ruses and artful dodges” that are
relative ability and competition, students may seek ways to dem-
used as ploys in “the struggle to escape being labeled as stupid”
onstrate that they are more able than others or at least to demon-
(1992, p. 85). They include avoiding seeking help, resisting novel
strate they are not less able than others. They may perceive that
approaches to academic work, and purposefully withdrawing effort
asking for help, trying hard, and approaching their work in novel
(self-handicapping). For example, many students perceive a threat
ways is a threat to their self-worth and thus purposefully avoid the
to self-worth from both teachers and classmates when contemplat-
use of strategies that might enhance their understanding and
ing seeking help (Butler & Neuman, 1995; Ryan & Pintrich, 1997).
achievement. When the emphasis is on learning, understanding,
Butler (1998) found that 10 –12-year-olds with concerns about
and intellectual development, students are less likely to feel threat-
their competency were the least likely to request help when it was
ened and may not perceive a need to use these avoidance
needed. Similarly, some students may prefer to avoid novel ways
strategies.
of solving problems and doing their work, fearing that they may
Both experimental and survey-based research has shown that
make mistakes and appear unable. Finally, working hard can put
students are less likely to seek help under performance goal
self-worth at risk because trying hard and failing to do as well as
conditions than under mastery goal conditions (Butler & Neuman,
others is compelling evidence of low ability (Covington & Om-
1995; Ryan, Gheen, & Midgley, 1998). Studies examining predic-
elich, 1979). By not trying, the student is able to stave off the
tors of students’ use of self-handicapping mirror the results re-
public judgment of low ability and the causes of failure become
ported previously for avoiding help seeking (e.g., Midgley &
uncertain. These avoidance strategies may protect students from
Urdan, 2001). Perceptions of a mastery goal structure in the
negative judgments by others, but they are also likely to undermine
classroom negatively predicted handicapping, and perceptions of a
performance.
performance goal structure positively predicted handicapping. It is
Given recent work by Dweck and her colleagues on the early
important to replicate these findings with a new sample of students
development of motivational problems in some children (e.g.,
and extend the findings by adding a qualitative component.
Burhans & Dweck, 1995; Cain & Dweck, 1995), avoidance be-
Although few studies have examined students’ preference to
haviors may be demonstrated by young children as well. However,
avoid novel approaches to academic work, Dweck and her col-
they may become more frequent in young adolescents because at
leagues (e.g., Farrell & Dweck, 1985) found that students with
this age children move from the conception of ability as modifiable
mastery goals worked much harder and scored significantly higher
with effort to an understanding of the notion of ability as fixed
in solving novel problems than did students with performance
(Nicholls, 1984; Nicholls & Miller, 1984). Having developed the
goals. These measures of effort and achievement are different from
schema that can lead to the use of avoidance strategies, young
expressing a preference to undertake or to avoid novel tasks.
adolescents may forego their efforts to succeed to protect their
However, it provides a rationale for extending the research exam-
public image of competence. However, as Nicholls (1984) pointed
ining the relation between the classroom goal structure and avoid-
out, the nature of the context may influence which notion of ability
ance strategies by considering whether the same pattern emerges
students will invoke. Thus aspects of the classroom context may be
for avoiding novel ways to approach academic tasks.
related to students’ use of avoidance strategies.
Students may perceive a high emphasis on both goals in the
classroom, an emphasis on one goal more than another, or a low
Relation Between the Classroom Context and the Use of
emphasis on both goals. The issue of “multiple goals” is one that
Avoidance Strategies
has been addressed in studies of students’ personal goal orienta-
tions but not in studies assessing students’ perceptions of the goal
A main purpose of this study was to examine the relation
structure in the classroom. In most studies examining students’
between the classroom learning environment and students’ reports
personal goals, the most facilitative pattern has been high mastery/
of avoidance strategies by focusing on different, but complemen-
low performance (e.g., Meece & Holt, 1993; Pintrich & Garcia,
tary, aspects of this environment. First, the relation between stu-
1991; Wolters, Yu, & Pintrich, 1996) and one expected to be
dents’ perceptions of the goal structure in the classroom and their
associated with low avoidance. There is also some evidence that
use of avoidance strategies is discussed. Second, instructional
espousing both high-mastery and high-performance goals is facil-

---

90
TURNER ET AL.
itative (Bouffard, Boisvert, Vezeau, & Larouche, 1995; Elliot &
performance focus were characterized by the frequent mention of
Church, 1997; Pintrich, 2000; Wentzel, 1991, 1993). A high em-
grades and evaluation.
phasis on performance goals may not be associated with the use of
avoidance strategies if mastery goals are also salient. In contrast,
Relation Between Teacher Discourse and Students’ Use
perceiving a high emphasis on performance goals and a low
of Avoidance Strategies
emphasis on mastery goals in the classroom may be associated
with the highest incidence of avoidance strategies. This question is
There has been no research, to our knowledge, on the relation
important not only in terms of specific recommendations to teach-
between the instructional practices teachers actually use in class-
ers but also should provide insights into the current controversy
rooms and students’ use of avoidance strategies. What do teachers
regarding the facilitative nature of performance goals (e.g., Harac-
say during instruction and how does this relate to students’ use of
kiewicz, Barron, & Elliot, 1998; Midgley, Kaplan, & Middleton,
avoidance strategies? Are all practices equal or do some practices,
2001).
or some combination of practices, appear to have a power-
ful relation to avoidance? Discourse analysis assumes that what
Relation Between Instructional Practices and the
teachers say sends powerful messages about what counts as learn-
Classroom Goal Structure
ing in their classrooms, thus creating different instructional
environments.
Although the relation between students’ perceptions of the
classroom goal structure and avoidance behaviors has been inves-
Classroom Discourse
tigated, there has been little research about how instructional
practices influence classroom goal structures. Ames (1992a,
The teacher’s role in classroom discourse may signal to students
1992b) has focused our attention on the relation among instruc-
whether teachers think that they are capable of learning and
tional practices in the classroom, the classroom goal structure, and
whether they are succeeding in meeting the teacher’s expectations.
students’ beliefs and behaviors. Through their instructional prac-
If students perceive teachers as supporting their learning through
tices, teachers send messages to students about the reasons for
what they say, the students may be less likely to adopt defensive
engaging in achievement behaviors. Ames argued that the goal
measures such as avoidance strategies. Conversely, if students
structure is conveyed by a constellation of instructional strategies
perceive teacher discourse as nonsupportive—as suggesting that
that are conceptually related, rather than by a particular instruc-
they cannot or will not meet such expectations—they may then
tional method. One of the most obvious needs is to describe actual
adopt avoidance strategies. The research literature has suggested
(not hypothetical) instructional practices in real classrooms with
certain classroom discourse practices as supportive (e.g., “scaf-
all their complexity.
folded discourse”; Hogan & Pressley, 1997; Turner et al., 1998) or
Very few researchers have used classroom observations to pro-
nonsupportive of learning (e.g., Mehan, 1985). In this study,
vide insights into the instructional practices that are associated
teacher discourse was examined as it related to (a) instruction, (b)
with students’ perceptions of the saliency of different achievement
motivation, and (c) organizing classroom time and activities.
goals. Meece (1991) aggregated fifth- and sixth-grade students’
survey reports of their personal achievement goals. She then used
Supportive and Nonsupportive Forms
observational data to describe the differences between high- and
of Instructional Discourse
low-mastery classrooms. Instructional practices that characterized
high-, but not low-, mastery-focused classrooms were as follows:
Scaffolded instructional discourse provides for negotiation of
emphasizing the meaningfulness of learning, adapting instruction
meaning and transfer of responsibility (Gallimore & Tharp, 1990;
to students’ developmental levels, providing teacher support for
Hogan & Pressley, 1997; Meyer, 1993) for learning to students.
instructional activities, deemphasizing ability-related information,
Negotiation is reflected both in teachers’ attempts to build under-
and emphasizing intrinsic reasons for learning. Patrick and her
standing with their students and to help them attain higher levels of
colleagues (Patrick, Anderman, Ryan, Edelin, & Midgley, 2001)
competence. As students demonstrate increasing competencies,
observed instruction in four fifth-grade classrooms that differed in
teachers withdraw assistance and they transfer responsibility for
students’ perceptions of the emphasis on mastery and performance
learning to the student. Transfer of responsibility increases student
goals based on survey responses (e.g., high mastery/high perfor-
ownership while it holds students accountable for their learning.
mance, high mastery/low performance, low mastery/high perfor-
Negotiation, or building understanding, is a necessary prerequisite
mance, and low mastery/low performance). Two features distin-
to transfer of responsibility. Supportive discourse patterns that
guished classrooms with a high-mastery focus. First, teachers in
reflect scaffolding—with their emphasis on learning, improve-
high-mastery-oriented classrooms spoke about learning as an ac-
ment, and understanding—are expected to contribute to the per-
tive process, whereas those in low-mastery-oriented classrooms
ception of a mastery-focused goal structure and to lower reports of
endorsed a transmission model of learning. Second, the teachers in
avoidance strategies. When teachers send messages that they will
high-mastery-oriented classrooms expressed strong positive affect
help students learn and students are able to assume responsibility
about learning and positive expectations for their students, whereas
for their increased competence, then students should not fear
those in low-mastery-oriented classrooms expressed little enthusi-
appearing unable.
asm about learning and did not convey high expectations for all.
In contrast, nonscaffolded forms of instructional discourse are
This second finding is interesting because it suggests that affect is
less oriented to assisting learning and are more focused on direct-
a potentially important component of mastery goal structures. Yet
ing and assessing. For example, in the Initiation–Response–
it has received little attention. Finally, classrooms with a strong
Evaluation cycle (I–R–E; Mehan, 1985), the teacher addresses

---

THE CLASSROOM ENVIRONMENT AND AVOIDANCE STRATEGIES
91
questions to the class, receives an answer from a volunteer, and
ganizational discourse to the use of avoidance strategies were not
then evaluates the answer. Alternatively, “telling” students what to
made, but it is acknowledged that organization is an important
think or do limits opportunity for student learning and autonomy
function of teacher discourse. In summary, it is assumed that it is
while establishing the teacher as authority (Deci & Ryan, 1985).
not only individual responses to students (e.g., a student who is
Such forms of instructional discourse usually do not offer “suffi-
having trouble learning) but also whether students perceive general
cient assistance, responsiveness, joint productive activity, or the
patterns of discourse as supportive or nonsupportive that may
building of common meanings and values” to support student
encourage or discourage avoidance strategies.
learning (Gallimore & Tharp, 1990, p. 188). An over use of
nonscaffolded, or controlling, instructional discourse patterns
Hypotheses
would be expected to contribute to the students’ perceptions of the
First, the relation between students’ perceptions of the goal
classroom goal structure as performance focused because they are
structure in the classroom and their use of and preference for
characterized by language that is more evaluative. Similarly, be-
avoidance strategies in a new sample of sixth-grade elementary
cause such nonscaffolded instructional patterns may not provide
school students is examined. Self-handicapping, the avoidance of
adequate support for learning, they might cause students to worry
help seeking, and a preference to avoid novelty are predicted to
about appearing unable and thus to adopt avoidance strategies.
vary between classrooms. It is also predicted that perceptions of an
emphasis on mastery goals in the classroom are negatively related
Supportive and Nonsupportive Forms
to the use of avoidance strategies, whereas perceptions of an
of Motivational Discourse
emphasis on performance goals in the classroom are positively
related to the use of avoidance strategies. Then the relationship
Supportive instructional discourse, directed to the cognitive
between teacher discourse and students’ reports of avoidance strat-
functions of teaching, may not be sufficient for students to feel
egies and the classroom goal structure is examined. It is expected
confident about learning. Students may also need motivational and
that supportive instructional and motivational discourse will be
affective support through interaction with their teachers and peers.
associated with lower reports of avoidance strategies and with
Recently, Goldstein (1999) reemphasized the motivational aspects
perceptions of a mastery goal structure. It is also expected that
of scaffolding (see also Wood, Bruner, & Ross, 1976). Goldstein
nonsupportive instructional and motivational discourse will be
asserted that when a teacher is pleasant and responsive, it builds
associated with higher reports of avoidance strategies and with
trust and maximizes students’ engagement and willingness to take
perceptions of a performance goal structure.
on challenging tasks. Similarly, Brophy (1999) suggested that
teachers can scaffold motivation to create an optimal match in the
“motivational zone of proximal development.” This motivational
Method
component would include such discourse practices as recruiting
Participants
the students’ interest, maintaining students’ persistence, minimiz-
ing frustration and risk, or enhancing students’ confidence (Lep-
This study is a part of a larger longitudinal study focusing on the relation
per, Drake, & O’Donnell-Johnson, 1997). Facets such as these
between the learning environment in mathematics classrooms and students’
should engage students and encourage effort and persistence, thus
beliefs and behaviors during the transition from elementary to middle
making it less likely that students would worry about appearing
school. Participants in this study included 1,197 sixth-grade elementary
school students and 65 sixth-grade classrooms in four ethnically and
unable.
economically diverse school districts in three Midwestern states. Students
Teacher discourse also may include nonsupportive motivational
were required to have parental permission to participate and 89% received
statements such as emphasizing grades over learning or correct-
permission. Only Euro-American and African American students were
ness over understanding, making negative remarks or using sar-
retained for the analysis of the survey data, and students who were
casm, or portraying errors as signs of incompetence (e.g., Ames,
Hispanic (4%), Asian (1%), and Other or Unknown (3.5%) were dropped.
1992a). Practices such as these would be more likely to highlight
Five additional students were dropped because of missing data. Thus the
students’ ability in a negative way and might encourage students to
present study includes 1,092 students (52% female, 70% Euro-American,
adopt strategies to protect their self-worth.
and 30% African American). In most cases teachers taught math and the
other core subjects to these students in self-contained classrooms.
In one district, researchers met with teachers in all 20 participating
Supportive and Nonsupportive Forms
classrooms to ask if they would be willing to have observers in their classes
of Organizational Discourse
in addition to participating in the survey component of the study. One
teacher declined observations. The observed classrooms were then chosen
Intertwined with goals for learning and motivational supports,
randomly from the 2 or 3 participating classrooms in each school, resulting
effective classroom teachers use supportive organizational dis-
in 10 sixth-grade classrooms from 9 schools. One teacher withdrew from
course so that the classroom can function smoothly and learning
the study before all the data were collected. The 9 observed teachers ranged
can take place (e.g., Evertson, Emmer, Clements, Sanford, &
in experience from 1 to 30 years with a mean of 16. Two of the teachers
Worsham, 1989). Such organizational decisions, which are re-
were male and 3 were African American. The rest were Euro-American.
flected in teacher discourse patterns, give predictability to the
instructional lesson and may help students feel confident and
Procedures
successful in their class. Nonsupportive organizational discourse
Surveys
may communicate to students that the teacher is not in control of
the learning environment or that she or he is unpredictable, thus
Students completed surveys in their classrooms in the late winter and
causing some apprehension. Predictions about the relation of or-
spring of 1998 –1999. Trained research assistants read the questions aloud.

---

92
TURNER ET AL.
Students were told that this was not a test and that only research staff would
motivational discourse (see Table 1). Two coding subcategories, support-
see their answers. The research staff solicited questions and put a sample
ive and nonsupportive, were developed within each of these categories to
item on the blackboard to illustrate the use of Likert-type scales. Research
capture the majority of teacher discourse during an instructional period.
assistants arranged make-up times for absentees.
The two instructional discourse categories included scaffolding for under-
standing (supportive) and nonscaffolding (nonsupportive). Scaffolding for
Survey Measures
understanding was coded when teachers helped students build understand-
ing (i.e., negotiation) and when teachers promoted student autonomy and
The present study included five scales from the student survey. Ryan
held them accountable for their learning (i.e., transfer of responsibility).
(Ryan et al., 1998; Ryan & Pintrich, 1997) developed the scale assessing
Nonscaffolding was coded when teachers asked questions with known
the avoidance of help seeking. All the other scales were taken from the
answers (i.e., I–R–E sequences) or told students what to do and how to do
Patterns of Adaptive Learning Survey (Midgley et al., 2000) and have
it, supporting low-level understanding or compliance. The two organiza-
proven to be reliable and valid in a number of studies with young adoles-
tional discourse categories included support for on-task behavior (e.g.,
cents. Students responded to the items on a scale of 1 (not at all true) to 5
teachers gave directions that helped maintain pacing and momentum,
(very true).
organized groups, or made smooth transitions between activities) and
Two scales from the student survey assessed students’ reports of the use
nonsupport (e.g., teachers interrupted learning because of off-task behav-
of avoidance strategies in the classroom, including self-handicapping and
iors or made an abrupt transition to a different task). Similarly, the two
the avoidance of help seeking, and one scale assessed students’ preference
motivational discourse categories included support (e.g., focus on learning,
to avoid novel approaches to doing academic work. The items in these
positive emotions, and peer support for collaboration) and nonsupport (e.g.,
three scales formed three distinct factors. Two scales from the student
focus on errorless learning, impersonal or negative affect, and individual
survey assessed students’ perceptions of the mastery and performance goal
success and failure). The forms of motivational support could appear
structure in their sixth-grade classrooms. Factor analysis confirmed that
independent of, or could cooccur with, instructional and organizational
these were distinct scales. Scales, items, and alpha coefficients are included
discourse. Therefore, any instructional or organizational response could be
in the Appendix. In addition, information about student gender, ethnicity,
coded simultaneously as motivationally supportive or nonsupportive.
and math standardized achievement scores and final grade in math in the
sixth grade was collected from school records. Math grades were coded on
Coding Procedures
a scale of 1 to 13 (13
A , 12
A, 11
A , and so on).
Using this coding system, Julianne C. Turner and Debra K. Meyer, who
also had served as two of the nine observers, independently analyzed
Discourse Collection
transcripts. Coders placed the transcripts in a database, read them, and
Mathematics instruction was observed and audiotaped during the same
marked the instructional, organizational, and motivational codes in a col-
two units of instruction in each of the nine classrooms. Classroom visits
umn to the right of the teacher discourse being classified. If any code
lasted for 5 days during a unit on factoring (e.g., least common multiple,
cooccurred with an instructional, organizational, or motivational code, then
greatest common factor, factor trees, etc.) in the fall of 1998 and for 5 days
it was identified in a second column. If a teacher response could not be
in the spring of 1999 for a unit on geometry (e.g., identifying and mea-
coded (e.g., an ambiguous statement, a remark to another adult in the room,
suring angles, turns, etc.). Because of tape recorder malfunctions, three
etc.), then no code was used. No code was used for 3% to 8% of teacher
classrooms did not have a complete set of transcripts (10 lessons) available
discourse, with a mean of 4.5% for the nine classrooms. A coded teacher
for analysis, but all classrooms had at least seven transcripts. All partici-
response could range from a single word to the entire speaking turn. A
pating classrooms were using the Connected Mathematics curriculum
coded response indicated that the categorization continued until the teach-
(Lappan, Fey, Fitzgerald, Friel, & Phillips, 1998).
er’s response ended or a different code was used.
Nine classroom observers were trained to take observation notes to
Transcript coding was completed in three parts to establish validity of
supplement audiotaped discourse. The notes provided context for the
codes and reliability between coders. First, one fall transcript from each
recordings so that the intent or consequences of a teacher statement or
classroom was coded independently and then codes were compared to
action could be discerned (e.g., when one teacher said “I’ll wait” that was
reach consensus on the use of discourse categories. This process allowed
a cue for student attention; observers noted “children laughed” or “teacher
for the coding categories to be further refined and for coders to discuss
put a silly expression on her face”). Observers received 10 hr of training,
ambiguity or disagreement. Then the coders established formal interrater
including comparing notes, to ensure a similar level of detail and focus. At
reliability on the major subcategories of instructional (supportive and
least two observers were assigned to a classroom in order to avoid bias;
nonsupportive) and organizational codes. Interrater reliability included
however, only one observer was in a classroom at one time. Classroom
instances in which one coder changed codes, but the other coder did not.
observers sat in the back of the classroom during instruction. The observers
One fall and one spring transcript for each classroom were chosen ran-
integrated field notes with transcriptions of the audiotaped lessons into a
domly to compute interrater agreement to determine whether the instruc-
document describing the entire lesson. Transcripts were typed using mod-
tional discourse categories were distinct from each other and organizational
ified standard transcription symbols (adapted from Psathas, 1995) to ensure
discourse. A Cohen’s kappa of .60 or higher was considered to be a good
comparability. Within transcripts, teacher discourse was labeled as whole
measure of interrater reliability and these kappas corresponded to the preset
class, small group, or individual instruction. In the present study only
goal of establishing approximately 80% or better agreement on each
discourse from whole-class instruction was analyzed. Because whole-class
transcript. The average kappa across all discourse categories for the nine
discourse simultaneously provides public messages about learning, perfor-
classrooms ranged from .67 to .75, representing a range of agreement from
mance, and expectations to all students, it was thought to be related most
78% to 84%. Once interrater agreement was established for a classroom,
directly to students’ reports of goal structures and avoidance strategies.
the remaining transcripts were coded independently by the two coders.
The third part of coding involved the newly developed categories for
motivational support and nonsupport, which were central to the research
Discourse Coding of the Observations
questions. For this coding category, all the transcripts were reread jointly
Discourse Codes
by the coders, and disagreements or coding errors were identified and
resolved through consensus. Therefore, all transcripts were analyzed by
A priori coding categories for teacher discourse were used to code the
both coders at least once and by one coder twice. Discourse analyses were
transcripts into three broad categories: instructional, organizational, and
completed prior to the analysis of the survey data to avoid bias.

---

THE CLASSROOM ENVIRONMENT AND AVOIDANCE STRATEGIES
93
Table 1
Discourse Codes
Code
Description
Example
Supportive instructional discourse (scaffolding)
Negotiating meaning
Adjusting instruction, simplifying, clarifying, or
“These—these numbers: 2
5
3. Or 3
5
2, 2
3
5,
elaborating; highlighting concepts or key
2
5
3 [pointing to the numbers on the board as he says
features or contrasts; modeling what students
them aloud]—they’re all the same numbers! You’re using the
should do—“thinking aloud” with students.
same combination; you’re just doing it in different orders. But
you’re still going to get 30, right?”
Transferring responsibility
Supporting strategic thinking and autonomous
“I am only talking about greatest common factors. She said
learning; holding students accountable for
that 2
2
3 are the greatest common factors of 24 and 60.
understanding.
Alex, do you agree with that? Why do you agree with that?
Why do you agree that these are the common factors of these
two numbers?”
Nonsupportive instructional discourse (nonscaffolding)
Telling
Prescribing how students should think and act
“That’s what [the book] says right there, right? It says, ‘You use
conceptually or emphasizing completion and
a shortcut to write 2
2
5
5.’”
accuracy over learning.
Initiating and evaluating
Asking a known-answer question or evaluating a
[Teacher question and answer regarding answers to math
(I–R–E sequence)
student response without demonstrating
problems.] “You had to find the greatest common factor and
understanding.
least common multiple for each pair of numbers. What is the
greatest common factor for the two numbers? The greatest
common factor?”
Organizational discourse
Supportive
Making transitions into and out of activities and
[Teacher instructions to try the next problem.] “Okay, now
giving directions or answering questions about
multiply these numbers together and see what the least
procedures.
common multiple is between these two numbers. Multiply it
together.”
[Teacher question regarding a student’s writing on the
overhead.] “Can you guys see that writing? Can you see?”
Nonsupportive
Commenting on student off-task or inappropriate
[Teacher instructions during student demonstration.] “Everybody
behavior that detracts from learning or
should be paying attention. I don’t want to see any pencils
interrupting learning (abrupt transitions).
going! I don’t want to see any chalk or anything! All eyes.
Marco [snaps her fingers], turn around and face the front.”
Motivational discourse (supportive)
Focus on learning
Focusing on the process of learning, challenging
April: (I don’t understand this.)
students, viewing errors as constructive, or
Teacher: You know what? That’s why we’re going to keep
supporting persistence.
working on it today and tomorrow. You’ll get it. Okay?
We’re just now starting it, April, so I don’t expect you to
fully understand it right away.
Positive emotions
Using enthusiasm or humor, or reducing anxiety;
A student gives an incorrect answer and teacher responds,
addressing emotional needs.
“Okay, he’s probably just checking to see if I was awake.”
Peer support and
Building collaboration, emphasizing joint goals—
“Marco, he’s your partner, so come up with him.”
collaboration
shared responsibilities.
Motivational discourse (nonsupportive)
Focus on errorless
Emphasizing goals for completion, perfection, or
“How you can tell, from the prime factorization of the number,
performance and
high scores; or labeling an activity as too
whether the least common multiple of two numbers is the
completion
difficult for the students; or viewing errors as
product of two numbers—oh, forget this one.” [The teacher is
detrimental to learning.
overcome with the complexity and trying to explain it to the
class. She looks at the observer and says, “Can you believe
this?”]
Impersonal, insignificant,
Using superficial, positive statements that
“Remember yesterday when we took 100? And we showed all
or negative affect
deemphasize authentic accomplishments; or
the different factor pairs that would make 100? How many
using threats, sarcasm.
recall that? Half of the (class) recalls that, the other half of
the class is brain dead.”
Individual success and
Emphasizing competition among students that
“Oh, how many got a hundred on the (test)? How many got a
failure
“excludes” or socially compares students.
hundred on the (test)? Shannon and Ciara got a hundred on
the (test).”

---

94
TURNER ET AL.
Results
in our outcomes are similar to those reported in other published
HLM studies on similar topics (e.g., Anderman et al., 2000;
The results have been organized to move from general findings
Anderman & Young, 1994).
to specific ones. First, the results of the analysis of the student
The same procedures were followed to determine the degree to
survey data for the whole sample are presented. Second, the results
which individual students’ perceptions of their classroom goal
of the analysis of the student survey data and the teacher discourse
structures were consistent within each classroom and were differ-
data for the nine classrooms that were observed are presented.
ent between classrooms. Estimates of adjusted ICCs indicated that
Third are presented detailed analyses of the discourse in four
approximately 25% of the variance in students’ perceptions of the
classrooms that differed from each other in terms of students’
classroom mastery goal structure and 35% of the variance in
perceptions of the goal structure and students’ use of avoidance
students’ perceptions of the classroom performance goal structure
strategies.
lay between classrooms. Individual student perceptions were ag-
gregated to the group level to measure goal structures from the
Results From Analyses of the Student Data
perspective of the students in the classrooms.
for the Sample as a Whole
In the second step, the HLM analyses were extended. The goal
Table 2 includes descriptive statistics for all variables and
was to determine whether classroom characteristics (aggregated
correlations among variables. Variables assessing student percep-
student reports of the goal structures) could explain variation in the
tions of the classroom goal structure were aggregated to the
average level of avoidance strategies between classrooms after
classroom level. Hierarchical linear modeling (HLM) was used to
controlling for student background characteristics (gender and
examine students nested in classrooms. HLM divides the total
ethnicity). The relationship of avoidance strategies with students’
variation in each student behavior into a within-classroom com-
math grades, relative to other students in their own classrooms,
ponent and a between-classroom component. Student-level and
was also taken into account. Teachers in each classroom may use
classroom-level parameters are estimated simultaneously. In the
different grading standards for assigning grades, and students are
first step of the analysis, it was determined whether students’
more likely to compare their own achievement with students
reports of handicapping, avoiding help seeking, and avoiding
within their own classes rather than across other classrooms. Thus,
novelty varied between classrooms. That is, did students in some
student math grades were centered at the group mean (i.e., within
classrooms report using avoidance strategies more than students in
classrooms) to provide a within-classroom estimate of the relation-
other classrooms? A fully unconditional HLM analysis, analogous
ship of grades to avoidance outcomes. Aggregated perceptions of
to an analysis of variance (ANOVA) with the classrooms acting as
mastery and performance goal structures were centered at the
the grouping variable, was run for each avoidance strategy. The
grand mean of classrooms. Gender and ethnicity were dummy
between-class chi-square variance estimates from these models
variables and were not centered. Parameter estimates represent
indicated that the variance in avoidance strategies differed signif-
slopes in the original metric.
icantly between classrooms: self-handicapping, 2(64, N
65)
Results of the final models of aggregated student reports of the
155.13, p
.001; avoiding help seeking,
2(64, N
65)
mastery and performance goal structures are presented in Table 3,
101.03, p
.01; avoiding novelty, 2(64, N
65)
112.91, p
Table 4, and Table 5. The intercept coefficient represents the
.001. Next, the intraclass correlation (ICC) for each variable was
estimated level of the avoidance outcome for male, Euro-American
calculated and adjusted for the reliability of the variance estimate
students (gender and ethnicity
0), with an average math grade
( ) to determine the proportion of variance in each outcome that
among their own classmates (group-centered grades
0), in
lay between classrooms. Results indicated the ICCs to be 13% for
classrooms with an average level of perceived mastery and per-
self-handicapping, 9% for avoiding help seeking, and 10% for
formance goal emphases (grand-mean centered goal structures
avoiding novelty. The percentages of between-classroom variance
0). Coefficients of class-level predictors of the intercept represent
Table 2
Descriptives and Bivariate Correlations
Variable
M
SD
1
2
3
4
5
6
7
8
Student-level variables (n
1,092)
1. Self-handicapping
1.90
0.85
—
2. Avoiding help seeking
2.13
0.91
0.49***
—
3. Avoiding novelty
2.99
1.10
0.35***
0.50***
—
4. Gender (1
female)
0.52
—
0.04
0.01
0.08**
—
5. Ethnicity (1
African American)
0.30
—
0.12***
0.08*
0.02
0.06*
—
6. Math grades
7.52
2.90
0.27***
0.33***
0.19***
0.11***
0.18***
—
Classroom-level variables: Aggregated student perceptions (n
65)
7. Mastery goal structure
3.79
0.42
—
8. Performance goal structure
2.92
0.71
0.40**
—
* p
.05.
** p
.01.
*** p
.001.

---

THE CLASSROOM ENVIRONMENT AND AVOIDANCE STRATEGIES
95
Table 3
Goal Structures as Predictors of Self-Handicapping
Estimation of intercept and slopes
Predictor
Coefficient
SE
t
Intercept
1.89
0.05
41.14***
Class-level predictors of intercept
Mastery goal structure: Aggregated student perceptions
0.28
0.09
2.95**
Performance goal structure: Aggregated student perceptions
0.01
0.06
0.13
Student-level controls
Gender (fixed between-class variance)
0.03
0.05
0.63
Ethnicity (fixed between-class variance)
0.12
0.06
2.02*
Math grades (free between-class variance)
0.09
0.01
6.95***
Class-level predictors of variation in math grades slope
Mastery goal structure: Aggregated student perceptions
0.01
0.03
0.39
Performance goal structure: Aggregated student perceptions
0.02
0.02
1.26
Estimation of between-class variance
df
2
Intercept
62
145.94***
Math grades slope
62
122.67***
* p
.05.
** p
.01.
*** p
.001.
the unit change in students’ avoidance associated with a one-unit
ethnicity varied significantly between classrooms. In other words,
change in the classroom goal structure. Coefficients for the effects
in some classrooms the effect of grades or ethnicity on self-
of student-level controls represent the unit change in the avoidance
handicapping was significantly steeper (or shallower) than in other
outcome associated with a one-unit change in the student charac-
classrooms. Preliminary analyses also revealed that the relation-
teristic. In preliminary analyses, the relationship between self-
ship between avoiding help seeking and math grades varied be-
handicapping and math grades and between self-handicapping and
tween classrooms. In these cases, the between-class slope variance
Table 4
Goal Structures as Predictors of Avoiding Help Seeking
Estimation of intercept and slopes
Predictor
Coefficient
SE
t
Intercept
2.09
0.04
47.79***
Class-level predictors of intercept
Mastery goal structure: Aggregated student perceptions
0.29
0.10
3.05**
Performance goal structure: Aggregated student perceptions
0.03
0.06
0.57
Student-level controls
Gender (fixed between-class variance)
0.06
0.05
1.14
Ethnicity (free between-class variance)
0.02
0.07
0.34
Class-level predictors of variation in ethnicity slope
Mastery goal structure: Aggregated student perceptions
0.28
0.17
1.61
Performance goal structure: Aggregated student perceptions
0.02
0.10
0.15
Math grades (free between-class variance)
0.12
0.01
10.05***
Class-level predictors of variation in math grades slope
Mastery goal structure: Aggregated student perceptions
0.05
0.03
1.46
Performance goal structure: Aggregated student perceptions
0.02
0.02
1.02
Estimation of between-class variance
df
2
Intercept
52
69.92*
Ethnicity slope
52
70.56*
Math grades slope
52
91.94***
Note.
The chi-square statistics are based on 55 of 65 classrooms that had sufficient data for computation.
* p
.05.
** p
.01.
*** p
.001.

---

96
TURNER ET AL.
Table 5
variance in all models indicated that a significant proportion of
Goal Structures as Predictors of Avoiding Novelty
variance between classrooms in avoidance remained unexplained.
To test the fit of the models, procedures suggested by Snijders and
Estimation of intercept and slopes
Bosker (1999) were followed. Specifically, the deviance statistics
from the fully unconditional models and the full models were
Predictor
Coefficient
SE
t
compared using a chi-square distribution, with the differences in
Intercept
2.87
0.06
49.95***
the number of parameters used for degrees of freedom. Results of
Class-level predictors of
the deviance tested indicated that the full models provided an
intercept
improved fit of the data for all of the outcomes. The model fit
Mastery goal structure:
0.19
0.12
1.70†
Aggregated student
significantly improved for self-handicapping, 2(9)
121.33, p
perceptions
.01, avoidance of help seeking,
2(14)
144.64, p
.01, and
Performance goal structure:
0.09
0.07
1.32
avoidance of novelty, 2(5)
62.46, p
.01.
Aggregated student
The goal of this study was to examine classroom characteristics
perceptions
and student avoidance strategies, and thus the issue of variance
Student-level controls
Gender (fixed between-class
0.25
0.06
3.84***
between classrooms is important. Although estimates suggest that
variance)
the total explained variance between classrooms in avoidance
Ethnicity (fixed between-
0.04
0.08
0.48
behaviors is small, it is important to note that HLM provides rather
class variance)
conservative estimates of between-level relationships. This study
Math grades (fixed
0.09
0.01
6.80***
between-class variance)
is about the classroom environment and avoidance strategies.
Consequently, although large portions of variance were not ac-
Estimation of between-class variance
counted for, the HLM analyses are still highly appropriate, given
that the goal of this article was to examine classroom level effects
df
2
on individual students’ reported use of avoidance strategies. The
Intercept
62
110.98***
fact that our models explained 23%, 20%, and 9% of the between-
classroom variances is important and supports the goals and re-
† p
.10 (marginally significant).
*** p
.001.
search questions posed in this study. Most striking is the finding
that students’ aggregated perceptions of a classroom emphasis on
mastery goals emerged as a significant negative predictor of all
of the student level predictor was set “free,” and whether the
three avoidance strategies, above and beyond students’ gender,
classroom goal structures helped explain this variation in slopes
ethnicity, and grades. Thus, in classrooms in which students per-
was examined. Therefore, coefficients of class-level predictors of
ceived that there was a greater emphasis on learning, improving,
student-level slopes represent the effect that a one-unit change in
and understanding, there were also lower levels of reported self-
classroom goals has on the relationship between a freely varying
handicapping, avoiding help seeking, and preference for avoiding
student-level predictor and the avoidance outcome. Chi-square
novelty. Students’ aggregated perceptions of the performance goal
estimates of between-class variance indicate whether the level of
structure in the classroom did not emerge as a significant predictor
the avoidance outcome (and any free varying student-level slopes)
of these outcomes. This nonsignificant relationship held even
still differ significantly between classrooms.
when performance goals were considered as the sole classroom-
To investigate the possibility of interaction effects of perceived
level predictor, without mastery goal structure in the model.
classroom performance and mastery goals on student avoidance,
an interaction term for perceived performance and mastery em-
phases as a class-level predictor of the intercept was included. To
Results From Analyses of the Nine Observed Classrooms
create this interaction term, perceptions of classroom mastery and
performance goals were first aggregated and then centered at the
Student Survey Data
grand mean of classrooms, and their product was calculated. The
classroom mastery and performance goal interaction term was not
Separate analyses were performed to establish the generalizabil-
a significant predictor for any of the avoidance outcomes. To
ity of the HLM results found in the full sample to the subset of nine
confirm these results, the analysis was replicated at the student
observed classrooms. A one-way ANOVA using students’ percep-
level with ordinary least squares regression analysis, using un-
tions of the mastery goal structure as the dependent variable
grouped data and unaggregated perceptions. Again, interaction
demonstrated that there were significant differences among the
effects of perceived mastery and performance goals were not
nine classrooms, F(8, 156)
10.99, p
.001. This construct was
significant for any of the avoidance outcomes. Thus, the final
selected because it emerged as a classroom-level predictor of
HLM models do not contain classroom mastery and performance
avoidance strategies in the HLM analysis. A multivariate analysis
goal interaction terms.
of variance using self-handicapping, avoiding help seeking, and
Comparing between-class variance estimates from the fully un-
avoiding novelty as dependent variables revealed significant dif-
conditional models and the final HLM models revealed that the
ferences among the 9 observed teachers, F(24, 447)
2.16; p
final models explained 23% of the between-class variance in
.00. Means and standard deviations for the student survey variables
self-handicapping, 20% of the between-class variance in avoiding
for all nine classrooms are provided in Table 6 to illustrate how the
seeking help, and 9% of the between-class variance in avoiding
four classrooms selected for further analysis compared with the
novelty. The significant chi-square estimates of between-class
other classrooms in the study.

---

THE CLASSROOM ENVIRONMENT AND AVOIDANCE STRATEGIES
97
Table 6
Means (and Standard Deviations) for Survey Variables in Observed Classrooms
Mastery goal
Performance
Avoiding
Avoiding
Classrooma
structure
goal structure
Self-handicapping
help seeking
novelty
Anderson (n
14)
M
2.96
3.20
2.83
2.56
3.59
SD
1.19
1.28
0.85
0.97
1.09
Christian (n
16)
M
3.99
3.60
2.15
2.19
3.01
SD
0.72
1.05
0.75
0.92
1.23
Davis (n
23)
M
4.38
1.91
1.75
2.19
2.83
SD
0.45
0.90
0.75
0.91
1.02
Guthrie (n
21)
M
3.95
2.93
1.63
2.08
2.77
SD
0.83
0.96
0.79
1.13
1.00
Hayes (n
10)
M
3.98
1.66
2.25
2.08
2.92
SD
0.64
0.53
1.30
0.85
0.90
Marks (n
26)
M
3.48
3.82
1.69
2.04
2.80
SD
0.79
0.96
0.76
0.72
0.91
Parsons (n
16)
M
3.02
3.02
2.63
2.86
3.84
SD
0.95
0.95
0.91
0.97
1.20
Robinson (n
18)
M
4.79
3.94
1.39
1.71
2.97
SD
0.32
0.57
0.54
0.69
1.10
Weber (n
21)
M
3.78
4.28
2.15
2.14
2.63
SD
0.63
0.62
1.04
0.59
0.93
a All teacher names are pseudonyms.
Discourse Data
organizational moves than in conjunction with them. The fre-
quency of supportive motivational statements distinguished class-
Table 7 shows discourse percentages across all coding catego-
rooms more than the nonsupportive statements.
ries for each of the 9 observed teachers. These percentages repre-
Given the complex qualitative patterns of teacher discourse,
sent the within-teacher proportions for each category during
they were triangulated with the student self-reports of goal struc-
whole-class instruction. In the motivational discourse column, the
tures and avoidance strategies to answer the question: How is
percentages represent all discourse classified as supportive or
teacher discourse related to students’ perceptions of the classroom
nonsupportive, whether occurring independently of, or in conjunc-
goal structure and use of avoidance strategies?
tion with, an instructional or organizational code. The percentages
in parentheses represent the independent motivational codes.
As Table 7 illustrates, the majority of whole-class teacher dis-
Differing Patterns in Four Classrooms:
course, 52% to 68%, fell within the instructional discourse cate-
High- and Low-Avoidance Teachers
gories, although patterns of discourse varied. Teachers were sim-
Student Survey Data
ilar in their proportions of negotiating meaning, but they varied
considerably in transferring responsibility to the student and in
A series of planned comparisons were undertaken to compare
their use of nonscaffolded discourse (e.g., “What is 24 divided by
students’ perceptions of the mastery goal structure in the nine
3?”). Even though mathematics instructional discourse has been
observed classrooms as well as reports of avoidance strategies
criticized as mostly “right answer” teacher talk (Lampert, 1990),
across those classrooms. Family-wise alpha levels were controlled
only 2 of the 9 teachers had high percentages of nonscaffolded
at .05 with the use of a Bonferroni adjustment. The analysis
discourse. Not surprisingly, 20% to 30% of teacher discourse was
indicated that students in classrooms taught by Mr. Parsons and
organizational in nature and did not differ much among teachers.
Ms. Anderson1 perceived a significantly lower emphasis on mas-
This teacher discourse was generally positive and supported stu-
tery goals than those in most other classrooms. Mr. Parsons dif-
dent learning. Teacher organizational discourse that was classified
fered from all teachers except Ms. Anderson and one other; Ms.
as nonsupportive varied more, ranging from 1% to 14%. Table 7
Anderson differed from all teachers except Mr. Parsons. Students
also demonstrates that the proportions of teacher discourse in the
in classrooms taught by Ms. Davis and Ms. Robinson perceived a
motivational category varied considerably across classrooms.
Within the category of supportive motivational discourse, teachers
made more statements that were independent of instructional and
1 All teacher names are pseudonyms.

---