COLLEGIALITY AS UNCERTAINTY MANAGEMENT: MULTILEVEL
CONTEXTS OF COLLABORATIVE TEACHER INTERACTIONS
By
Seung-Hwan Ham

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Educational Policy
2011

ABSTRACT
COLLEGIALITY AS UNCERTAINTY MANAGEMENT: MULTILEVEL
CONTEXTS OF COLLABORATIVE TEACHER INTERACTIONS
By
Seung-Hwan Ham
This study aims to explore how collaborative teacher interaction is contingent
upon teachers’ immediate classroom and school-organizational contexts as well as
broader societal and policy environments. Notwithstanding the widely acknowledged
connection between teacher collegiality in schools and its beneficial effects, little
systematic effort has been made to understand what types of teachers, under what
contextual conditions, build and/or sustain collegial relationships with other teachers.
In order to understand better a range of multilevel contextual factors associated with
collaborative teacher interaction, this study intends to both elaborate and test an
uncertainty management perspective as a theoretical framework. In addition, four
conventional perspectives – a school climate perspective, a principal instructional
leadership perspective, a curriculum policy perspective, and a national culture
perspective – are also considered to see if a set of primary hypotheses derived from the
uncertainty management perspective is empirically supported even after a range of
other plausible hypotheses are simultaneously taken into account.
Overall, the main findings from this study support an uncertainty management
perspective which understands collegiality as a byproduct of the collective sensemaking
of uncertainties, which reduces organizational members’ feelings of discomfort with

given uncertainties and thus helps them manage those uncertainties without actually
removing them completely from given situations. Extensive international and U.S. data
analyzed in this study substantiate the central proposition of the uncertainty
management perspective that teacher collegiality has to do with teachers’ collective
effort to deal with various instructional and classroom-contextual uncertainties that
they confront in their teaching. This result sheds light on the possibility that lateral
collegial relationships among teachers may serve as a source of information processing,
sensemaking, and problem-solving, whereby teachers can better manage to go through
given situations of uncertainty.
This study provides useful implications for educational policy and
administration. In particular, given the fact that it is not always either possible or
desirable to eliminate the sources of uncertainty in teaching, finding ways to help
teachers manage the uncertainty through collaborative interaction with their colleagues
is an important task for educational administrators and policy makers. Further,
considering the wide variety of possible situations in which teachers confront
instructional and classroom-contextual uncertainties, it is important for principals to
explore various ways to actively share leadership roles and responsibilities with many
teachers, so that individual teachers can become empowered agents to foster and
sustain a collegial work environment.

Copyright by
SEUNG-HWAN HAM
2011

ACKNOWLEDGMENTS

This dissertation benefited from a great deal of support, help, and
encouragement from many people in different parts of the world. First of all, I cannot
fully express my gratitude to my dissertation committee, Professors Dorothea
Anagnostopoulos, Amita Chudgar, Brian DeLany, Robert Floden, and Lynn Paine, for
their supportive guidance, invaluable comments, and constructive suggestions. They
broadened my intellectual horizons at the intersection of educational policy and
administration and teacher education. It was truly a great honor for me to have
Professor Lynn Paine serve as my academic advisor, who enthusiastically encouraged
me to explore outside my comfort zones to help me further strengthen every aspect of
my academic and professional experience. Working with her in research projects
constructively challenged, expanded, and enriched my perspective on various aspects
of educational policy and practice.
I am deeply grateful to Professors John Schwille, Michael Sedlak, Sharif Shakrani,
and Peter Youngs at Michigan State University, Professor Sigrid Blömeke at the
Humboldt University of Berlin, and Vice President Kil Gyeun Yoon of Bucheon
University for their encouragement and constant support. My special gratitude also
goes to Professor Yun-Kyung Cha at Hanyang University, Seoul, who first ignited my
interest in school organization and educational policy and profoundly influenced my
intellectual growth and development. He has given me every possible help as an
unfailing source of encouragement and guidance ever since I first met him as an

v

undergraduate freshman.
My heartfelt appreciation is extended to my dear friends for their cheerful
optimism and encouragement, including Amy Damrow, Todd Drummond, Sedat
Gumus, Eun Hye Ham, Chong Min Kim, Rae Young Kim, Wang Jun Kim, Jay Lee,
Sheila Marquardt, Betty Okwako, Veronica Son, Min Sun, Paul Tanner, and Yisu Zhou
at/from Michigan State University, as well as many other friends and colleagues whom
I was fortunate enough to meet in seminars, workshops, conference sessions, and
research projects.
My doctoral work in Educational Policy was supported by multiple funds from
Michigan State University. In addition to graduate research assistantships from the
College of Education, a four-year Erickson Research Fellowship from the Education
Policy Center at Michigan State University generously supported my doctoral studies.
Also, an Anderson-Schwille Fellowship, a Summer Research Fellowship, and many
travel funds from the College of Education encouraged me to be more ambitious in
developing my research and enriching my professional experience. It was a great honor
for me to receive such generous support throughout the entire course of my doctoral
studies.
Moreover, for the last two semesters of my doctoral studies, I had an invaluable
opportunity to teach an undergraduate-level social foundations course in teacher
education, TE 250, at Michigan State University. I am grateful to Professors Dorinda
Carter, Lynn Paine, and Suzanne Wilson for all kinds of support, without which I
would not have had the privilege to enjoy such a great opportunity to teach, and learn
vi

from, my students. Interacting with these future teachers, I truly enjoyed learning a lot,
not only from their questions and ideas, but also from the ways in which they inquired
into education, schooling, and society.
Also, I am heartily thankful to my brother, Cheol-Hwan, for his warm wishes.
Finally, no words can fully express my deep gratitude to my parents for their endless
support and unconditional trust. I dedicate this dissertation to my parents, and to the
memory of my grandparents.

vii

TABLE OF CONTENTS

LIST OF TABLES .......................................................................................................................... x
LIST OF FIGURES ...................................................................................................................... xii
CHAPTER 1
INTRODUCTION .........................................................................................................................1
1.1. Purpose of the Study: Examining Collegiality in Context..........................................1
1.2. Significance of the Study .................................................................................................5
CHAPTER 2
CONCEPTUAL FRAMEWORK .................................................................................................9
2.1. Theoretical Background...................................................................................................9
2.1.1. School Climate Perspective............................................................................11
2.1.2. Principal Instructional Leadership Perspective ..........................................14
2.1.3. Curriculum Policy Perspective .....................................................................17
2.1.4. National Culture Perspective ........................................................................20
2.1.5. Uncertainty Management Perspective: An Alternative Approach..........22
2.2. Hypotheses ......................................................................................................................28
2.2.1. Primary Hypotheses .......................................................................................28
2.2.2. Additional Hypotheses ..................................................................................32
2.3. Summary..........................................................................................................................35
CHAPTER 3
DATA AND METHODOLOGY................................................................................................37
3.1. Description of the Data ..................................................................................................37
3.1.1. Sample Design and Weights..........................................................................37
3.1.2. Units of Analysis .............................................................................................40
3.2. Dependent Variables: Measures of Teacher Collegiality ..........................................40
3.3. Independent Variables ...................................................................................................43
3.3.1. Primary Independent Variables ....................................................................43
3.3.2. Additional Independent Variables ...............................................................51
3.3.3. Control Variables ............................................................................................54
3.3.4. U.S.-Specific Contextual Variables ...............................................................59
3.4. Descriptive Statistics ......................................................................................................60
3.5. Analytic Approaches .....................................................................................................60
CHAPTER 4
DESCRIPTIVE AND PRELIMINARY ANALYSES ...............................................................65
4.1. Analyses of Aggregated Country-Level Data ............................................................65
4.2. Analyses of Multilevel Data..........................................................................................74
viii

4.3. Discussion ........................................................................................................................81
CHAPTER 5
MULTILEVEL MODELING ANALYSES ................................................................................84
5.1. Model Specification ........................................................................................................84
5.2. Results ..............................................................................................................................89
5.3. Discussion ........................................................................................................................98
CHAPTER 6
ANALYSES OF THE CASE OF THE UNITED STATES .....................................................102
6.1. Within-U.S. Analyses ...................................................................................................102
6.2. Comparison with Other Countries ............................................................................115
6.3. Discussion ......................................................................................................................121
CHAPTER 7
CONCLUSION AND DISCUSSION ......................................................................................124
7.1. Summary of the Findings ............................................................................................124
7.2. Implications for Educational Policy and Administration.......................................132
7.3. Suggestions for Future Research ................................................................................136
APPENDIX .................................................................................................................................139
A.1. Additional Analyses ...................................................................................................140
BIBLIOGRAPHY .......................................................................................................................155

ix

LIST OF TABLES
Table 3-1. Exploratory factor analysis of the level-1 primary independent variables .....47
Table 3-2. Summary of variable definitions and data sources ............................................56
Table 3-3. Descriptive statistics for variables .........................................................................61
Table 3-4. Descriptive statistics for variables used in within-U.S. analyses ......................62
Table 4-1. Level-2 variables: Values by country ....................................................................70
Table 4-2. Pearson’s bivariate correlations between country-means of TCHRCOL and
level-2 variables........................................................................................................72
Table 4-3. Ordinary least squares regressions for country-means of TCHRCOL, with all
level-2 variables entered into the equation ..........................................................73
Table 4-4. Unconditional hierarchical (generalized) linear models for TCHRCOL .........76
Table 4-5. Bivariate associations between TCHRCOL_LESSON and level-2 variables:
Coefficients from multilevel linear regressions ...................................................79
Table 4-6. Bivariate associations between TCHRCOL_CLASS and level-2 variables:
Coefficients from multilevel linear regressions ...................................................79
Table 4-7. Bivariate associations between TCHRCOL_VARIETY and level-2 variables:
Coefficients from multilevel ordinal logistic regressions ..................................80
Table 5-1. Multilevel linear regressions for TCHRCOL_LESSON ......................................90
Table 5-2. Multilevel linear regressions for TCHRCOL_CLASS .........................................91
Table 5-3. Multilevel ordinal logistic regressions for TCHRCOL_VARIETY ...................96
Table 6-1. Ordinary least squares regressions for TCHRCOL_LESSON: The case of the
United States ...........................................................................................................103
Table 6-2. Ordinary least squares regressions for TCHRCOL_CLASS: The case of the
United States ...........................................................................................................106

x

Table 6-3. Ordinal logistic regressions for TCHRCOL_VARIETY: The case of the United
States ........................................................................................................................113
Table A1. Exploratory factor analysis of the level-1 primary independent variables:
Orthogonal rotation ...............................................................................................140
Table A2. Pearson’s bivariate correlations between level-2 variables..............................140
Table A3. Multilevel linear regressions for TCHRCOL_LESSON, with all slopes fixed at
level 2 .......................................................................................................................141
Table A4. Multilevel linear regressions for TCHRCOL_CLASS, with all slopes fixed at
level 2 .......................................................................................................................142
Table A5. Multilevel ordinal logistic regressions for TCHRCOL_VARIETY, with all
slopes fixed at level 2.............................................................................................143
Table A6. Multilevel linear regressions for TCHRCOL, with all variables centered at
their country means ...............................................................................................144
Table A7. Ordinary least squares regressions for TCHRCOL_CLASS: The case of the
United States: Additional models .......................................................................145
Table A8. Within-nation regressions for TCHRCOL_LESSON: Ordinary least squares
regression coefficients for UNC_TCHUNDS and UNC_CONTEXT, after
controlling for all other level-1 variables ...........................................................151
Table A9. Within-nation regressions for TCHRCOL_CLASS: Ordinary least squares
regression coefficients for UNC_TCHUNDS and UNC_CONTEXT, after
controlling for all other level-1 variables ...........................................................152
Table A10. Within-nation regressions for TCHRCOL_VARIETY: Ordinal logistic
regression coefficients for UNC_TCHUNDS and UNC_CONTEXT, after
controlling for all other level-1 variables ...........................................................153
Table A11. Summary of results, by hypothesis ....................................................................154

xi

LIST OF FIGURES
Figure 4-1. Eighth-grade mathematics teachers’ engagement in collaborative lesson
planning, by country .............................................................................................66
Figure 4-2. Eighth-grade mathematics teachers’ engagement in collaborative classroom
observation, by country ........................................................................................67
Figure 4-3. Variety in collaborative interaction among eighth-grade mathematics
teachers, by country ...............................................................................................68
Figure 6-1. Within-nation ordinary least squares regressions for TCHRCOL_LESSON:
Standardized coefficients for UNC_TCHUNDS and UNC_CONTEXT ......116
Figure 6-2. Within-nation ordinary least squares regressions for TCHRCOL_CLASS:
Standardized coefficients for UNC_TCHUNDS and UNC_CONTEXT ......117
Figure 6-3. Within-nation ordinal logistic regressions for TCHRCOL_VARIETY: Odds
ratios for UNC_TCHUNDS and UNC_CONTEXT .........................................119

xii

CHAPTER 1
INTRODUCTION

1.1. Purpose of the Study: Examining Collegiality in Context

It has been widely acknowledged that the virtues of teacher collegiality are
widespread. Different stakeholders, including teachers, educational administrators, and
policy makers, may have a range of reasons to believe that teacher collegiality is
beneficial to various aspects of school success and improvement. Indeed, a sizable body
of empirical research links collegiality among teachers to their enhanced feelings of
efficacy, increased teaching quality, and greater accountability for student achievement
(Bryk, Camburn, & Louis, 1999; Dunne, Nave, & Lewis, 2000; Lee & Smith, 1996; Louis
& Marks, 1998; Newmann, Rutter, & Smith, 1989; Saunders, Goldenberg, & Gallimore,
2009; Stevens & Slavin, 1995; Yasumoto, Uekawa, & Bidwell, 2001). Taking the form of
collaborative lesson planning, classroom observation, peer coaching, or mentoring,
teacher collegiality has often been conceptualized as an important factor contributing to
teacher development whereby teachers can be provided with opportunities to learn
from each other; rather than dependent on external experts, teachers are expected to
share and develop their expertise together, so that they can be provided with
opportunities to revisit, and take advantage of, their own reflections on teaching
(Cochran-Smith & Lytle, 1999; Fernandez & Yoshida, 2004; Grossman, Wineburg, &
1

Woolworth, 2001; McLaughlin & Talbert, 2001; Paine & Ma, 1993; Paine, Pimm, Britton,
Raizen, & Wilson, 2003).
Drawing on definitions in common use in the literature, I conceptualize teacher
collegiality as collaborative interaction among teachers which involves joint work
concerning their core task in school, i.e., the practice of classroom teaching. Such joint
work implies collective action, strong interdependence, shared responsibility, and a
great degree of readiness to participate in reflective inquiry into practices (Little, 1990).
While emphasizing joint work as a central feature of collegiality, I do not view
collegiality as a condition for consensus and solidarity. Indeed, it is important to note
that teachers frequently run into considerable conflicts over professional beliefs about
teaching and learning when involved in joint work (Achinstein, 2002; de Lima, 2001);
however, the contention resulting from different viewpoints can ultimately become the
potential source of organizational learning if teachers can manage to embrace, rather
than suppress, differences and disagreements (Achinstein, 2002; Fullan, 1993). Further,
such organizational learning is more viable in a collegial atmosphere where
collaborative working relationships among teachers evolve from teachers‟ own
perceived needs, rather than from administrative imposition (Bidwell, 2001; Hargreaves,
1991). This is because collaborative working relationships as “[c]ommunities of practice
rely on the voluntary participation of their members to make them useful” (Galbraith,
Downey, & Kates, 2002, p. 73).
Considering that teachers‟ professional growth and development can be fostered
by collegiality, it is not surprising that collegiality can be an important source of
2

organizational learning (Mohr & Dichter, 2001; Rait, 1995). Since teachers are the very
persons who finally interpret and implement reform (Anagnostopoulos & Rutledge,
2007; Coburn, 2001; Spillane, Reiser, & Reimer, 2002), the successful transfer of reform
ideas into practice is likely to depend greatly on how well collegial working
relationships have been created and sustained among teachers. It is, thus, reasonable to
expect both school-based and externally-introduced reform initiatives to be
implemented more effectively in schools where productive collegial relationships are
the norm than in other schools which lack enough organizational capacity to elicit
shared understandings among teachers. In this respect, many argue that collegiality in
schools is an important prerequisite for the effective implementation of various reform
initiatives (Cosner, 2009; Frank, Zhao, & Borman, 2004; Hargreaves, 1989; Little, 1993;
Louis, Marks, & Kruse, 1996).
Indeed, newly emerging social psychological perspectives posit that cognition is
not simply a property of an individual person, but it is distributed across people
(Hutchins, 1995; Salomon, 1993). Such distributed cognition perspectives commonly
suggest the possibility that teacher collegiality may help “diverse groups of teachers
with different types of knowledge and expertise . . . draw upon and incorporate each
other‟s expertise to create rich conversations and new insights into teaching and
learning” (Putnam & Borko, 2000, p. 8). Similarly, images of leadership have also
changed in a way that teachers are encouraged to actively engage in leadership roles.
Recent literature on school organization challenges conventional conceptualizations of
leadership as monopolized by administrators, emphasizing new styles of leadership,
3

whereby leadership is shared with, and distributed among, teachers (Scribner, Sawyer,
Watson, & Myers, 2007; Somech, 2005; Spillane, Halverson, & Diamond, 2004). These
alternative forms of leadership are characterized by shared decision making,
collaborative problem solving, and supportive relationships, which all relate to
collegiality.
Notwithstanding the widely acknowledged connection between teacher
collegiality in schools and its beneficial effects, little systematic effort has been made to
understand what types of teachers, under what contextual conditions, build and/or
sustain collegial relationships with other teachers. In terms of research design, teacher
collegiality, in most studies, has typically been conceptualized and treated as an
independent variable explaining a certain desirable outcome (e.g., Lee & Smith, 1996;
Saunders, et al., 2009; Stevens & Slavin, 1995; Yasumoto, et al., 2001). Although those
studies help us recognize various positive effects of teacher collegiality, we also need
systematic investigations of what contextual factors are related to teacher collegiality; in
order to understand how to promote the beneficial effects of teacher collegiality, it is an
important prerequisite to develop a comprehensive knowledge base that provides
insight into how the level and type of collaborative teacher interaction varies depending
on various factors.
As a systematic investigation in this direction, this study explores an alternative
account of teacher collegiality as a dependent variable contingent upon various factors
at multiple levels of abstraction ranging from the teacher level to the national sociocultural and policy environment level. In other words, this study is an attempt to
4

address the following research question: Why do some teachers participate to a greater
degree in collaborative interaction with their colleagues than other teachers do? Put
differently, under what contextual conditions are teachers more likely to engage in
collaborative interaction with other teachers? Identifying a range of multilevel factors
significantly associated with teacher collegiality as a dependent variable, this study
aims to provide new insights into the nature of teacher collegiality on the basis of
systematic empirical evidence from extensive national and international data from
multiple sources.

1.2. Significance of the Study

Collegiality among teachers is widely seen as beneficial because it helps teachers
engage in reflective inquiry about teaching and learning (Bryk & Schneider, 2002;
Kazemi & Franke, 2004; Lassonde & Israel, 2010; Little, Gearhart, Curry, & Judith, 2003;
Slavit, Nelson, & Kennedy, 2009). Considering that it is important for each teacher to be
empowered as an active and reflective educational theorist and practitioner rather than
seen as just a passive agent of a curriculum package (Shulman, 2004; Silcock, 1994;
Westbury, Hopmann, & Riquarts, 2000), collegiality can provide teachers with various
opportunities to collaboratively produce useful local knowledge about the practice of
classroom teaching. One of the most valuable benefits of collaborative teacher
interaction lies in the increased opportunity provided to teachers to interact with one
another regarding instructional issues; it has been well documented that teachers who
5

regularly participate in collaborative interaction are provided with opportunities to
listen to a wide range of ideas about the practice of teaching as they exchange helpful
feedback with their colleagues regarding how to deal with various instructional
dilemmas and constraints they are confronted with in their teaching (Dunne, et al., 2000;
Fernandez & Yoshida, 2004; Grossman, et al., 2001; McLaughlin & Talbert, 2001; Paine
& Ma, 1993).
As a result, teachers can develop useful local knowledge and skills that they can
effectively use in their classrooms. Teachers can promote their professional growth
“when they generate local knowledge of practice by working within the contexts of
inquiry communities” (Cochran-Smith & Lytle, 1999, p. 250). Teacher collegiality in
schools, as a form of “relational” capacity (Bryk & Schneider, 2002), may facilitate
teachers to build their professional knowledge. Further, the development of teacher
competence is largely acquired through collective efforts embedded in the “ethic of care”
(Gilligan, 1993) in schools. In schools where collaboration is constantly fostered, the
ongoing communication is likely to increase the awareness that virtually every
professional in school has a range of difficult tasks to do; when this awareness is
combined with the sharing of knowledge and skills, teachers find themselves
“connect[ed] with each other as a result of felt interdependencies, mutual obligations,
and other ties” (Sergiovanni, 1994, p. 5).
The rarity of U.S. schools engaged in meaningful collaborative work around
teaching has been well documented (Akiba & LeTendre, 2009; Leonard & Leonard, 2003;
Lortie, 2002; Rosenholtz, 1989). Many studies decry limited opportunities for teachers to
6

have supportive collegial relationships through which they can share their “personal
knowledge” (Clandinin, 1985) with other teachers. As an empirical attempt to explain
differences among schools regarding the extent to which teacher collegiality is present,
this study employs a cross-national multilevel research design. Since teaching is not
only a complex non-routine practice but also a culturally embedded activity (Alexander,
2000; Anderson-Levitt, 2003; Hiebert et al., 2003; Paine & Fang, 2006; Rowan, 2002;
Tobin, Hsueh, & Karasawa, 2009), the simultaneous consideration of multilevel contexts
around teaching will lead to a more comprehensive understanding of factors
contributing to teacher collegiality.
In order to examine a series of hypotheses derived from different theoretical
perspectives, this study involves a range of empirical analyses using a large-scale
quantitative dataset with an international scope. In addition, this study focuses on the
case of the United States by conducting another set of within-U.S. analyses. Analyzing a
series of large-scale data with an international scope can provide deeper
understandings of the interplay of education and culture and produce further
intriguing hypotheses, precisely because of its broader perspective (Alexander, 2001;
Blömeke & Paine, 2008; Bray, Adamson, & Mason, 2007; Porter & Gamoran, 2002).
Providing new empirical evidence concerning the nature of collaborative teacher
interaction, this study attempts to make a unique contribution to the literature on
teachers‟ organizational behavior in schools as well as to provide substantive
implications for educational policy and administration in relation to teacher
development and school improvement. I hope this study will stimulate further studies
7

based on different sets of quantitative and qualitative data to both corroborate and
reexamine the findings from this study.

8

CHAPTER 2
CONCEPTUAL FRAMEWORK

2.1. Theoretical Background

Research on teacher collegiality has given little analytic attention to why some
teachers frequently participate in collaborative interaction with their colleagues
whereas other teachers do not. Teacher collegiality has usually been treated as an
independent variable that is presumed to yield some positive outcomes, irrespective of
how it is conceptualized. Conceptualizing teacher collegiality as a dependent variable
contingent upon a range of multilevel contextual factors has attracted little scholarly
attention despite its significance for understanding not only the nature of teacher
collegiality but also the ways in which teacher collegiality arises and is sustained in
schools.
In order to understand better a range of multilevel contextual factors associated
with collaborative teacher interaction, it is important to consider various independent
variables ranging from teacher-level variables to national-contextual variables. Thus, I
explain five perspectives in this study, each of which provides useful insight into
teacher collegiality as a dependent variable. They represent a school climate perspective,
a principal instructional leadership perspective, a curriculum policy perspective, a
national culture perspective, and an uncertainty management perspective. Of course,
9

these are not an exhaustive list of frameworks for understanding teacher collegiality in
relation to contextual factors; various other perspectives may also exist that proffer
different avenues to explore the contextual contingencies of teacher collegiality.
Nevertheless, I believe that the five perspectives considered in this study can provoke
further theoretical and empirical inquiry into the complexities involved in
understanding multiple layers of social context in which collaborative relationships are
formed and sustained among teachers.
In particular, this study intends to both elaborate and test an uncertainty
management perspective as a theoretical framework for understanding teacher
collegiality. Although there have been some conceptual and empirical studies that
allude to or suggest a possible relationship between uncertainty in teaching and teacher
collegiality (e.g., Bidwell, 2001; Collinson & Cook, 2007; Gamoran, Secada, & Marrett,
2000; Rowan, 1990), there has never been a systematic attempt to elucidate teachers‟
exposure to uncertainties in teaching as a contextual factor related to teacher collegiality.
On the basis of an uncertainty management perspective, this study posits that
collegiality is largely a byproduct of the collective sensemaking of uncertainties, which
reduces organizational members‟ feelings of discomfort with given uncertainties and
thus helps them manage those uncertainties without actually removing them
completely from given situations. The other four perspectives, which are some of the
most insightful conventional accounts of teacher collegiality as a dependent variable
contingent upon certain contextual factors, are also considered to see if a set of primary
hypotheses of this study derived from the uncertainty management perspective is
10

empirically supported even after some other plausible hypotheses are simultaneously
taken into account.
In what follows in this chapter, I first discuss the strengths and weaknesses of
each of the four conventional perspectives. Next, I introduce the uncertainty
management perspective as an alternative conceptual framework for understanding
teacher collegiality. Finally, a series of primary hypotheses of this study is formulated
on the basis of the uncertainty management perspective, and a range of additional
plausible hypotheses are derived from each conventional perspective. It should be
noted that exceptions and complexities abound within each perspective. Despite this
limitation, it will be helpful to identify the main defining qualities of these varying
perspectives and their underlying assumptions, so that different hypotheses can be
advanced and then empirically tested and further explored.

2.1.1. School Climate Perspective
Extensive research and literature in organizational analysis has suggested that
the extent to which teacher collegiality is fostered and sustained in a school depends
largely on the organizational climate of the school. Although how to conceptualize and
measure school climate is still in debate (Anderson, 1982; Freiberg, 1998; Hoy, Tarter, &
Bliss, 1990; Marshall, 2004), a central underlying assumption that is generally shared in
the school climate literature is that a school‟s climate is determined by a set of relatively
enduring organizational characteristics of the school that is collectively perceived and
experienced by teachers working in that school. Acknowledging that the climate of a
11

school is conceptually inseparable from “teachers‟ [shared] perceptions of the general
work environment of the school” (Hoy & Miskel, 2005, p. 185), the school climate
literature commonly posits that teachers‟ behavior in a school is constantly influenced
by the climate of the school; that is, teachers‟ behavioral patterns in a school are
assumed to reflect the inter-subjective realities co-constructed by the teachers working
in that school.
In this line of inquiry, collegial teacher behavior has been understood as an
important indicator of a positive school climate. Relying on Miles‟ (1965)
conceptualization of “organizational health” in educational organizations, Sergiovanni
and Starratt (2002) emphasize that a healthy school operates as “a system of dynamic
interaction characterized by a high degree of interdependence” (p. 314) and “a sense of
togetherness that bonds people together” (p. 312). That is, the concept of school climate
draws our attention to teachers‟ interpersonal work life in schools. Empirical research
has also suggested that a positive school climate can facilitate collaborative interaction
among teachers, particularly because teachers working in such a school climate tend to
show greater willingness to invest their time and energy in contributing to the
organizational goals of their school (Bryk & Schneider, 2002; Hoy, 2002; TschannenMoran, 2001).
Despite the popularity of school climate as a useful explanation for differences in
teacher collegiality between schools, a closer look at the school climate literature reveals
that it provides only partial insight into teacher collegiality. In fact, a strong association
between a positive school climate and teacher collegiality would not be surprising,
12

precisely because school climate is an organization-level variable to capture the
interpersonal dimensions of teachers‟ work life in schools. It is widely accepted that a
positive school climate comes from teachers‟ collective perception of their school as a
workplace where teacher morale is high and organizational goals are clearly
communicated and well implemented (Freiberg, 1999; Hoy & Miskel, 2005; Sergiovanni
& Starratt, 2002). However, high teacher morale, clear goal focus, and many other
factors that are purported to be of significance to a positive school climate are, indeed,
good examples of school organizational characteristics that either presuppose or
necessitate collaborative teacher behavior in one way or another. In other words, it is
often hard to imagine either high teacher morale or clear goal focus without assuming
the presence of teacher collegiality that helps teachers communicate effectively within a
given school.
In this respect, the school climate framework for understanding teacher
collegiality is not firmly grounded in a substantive explanation of who participates in
collaborative teacher interaction and why. Although it appears highly plausible to
speculate that there exists a strong relationship between a positive school climate and
teacher collegiality, such an association per se does not provide a clear explanation of
why some teachers are motivated to participate in collaborative interaction with other
teachers whereas others are not. In other words, while it is clear from a school climate
perspective that schools with a positive organizational climate are likely to exhibit a
high level of teacher collegiality, this does not necessarily mean that teacher collegiality
arises in a positive school climate; rather, it could just be that teacher collegiality is an
13

integral constituent of a positive school climate.

2.1.2. Principal Instructional Leadership Perspective
Principal instructional leadership has usually been conceptualized as a blend of
the principal‟s participation in curriculum development and her/his engagement in
instructional supervision and evaluation (Blase & Blase, 1999; DiPaola & Hoy, 2007;
Hallinger & Murphy, 1987). The role of a principal as an instructional leader is
understood as different from that of traditional school principals. While a traditional
principal is portrayed as spending the majority of her/his time dealing with
administrative duties, a principal who is an instructional leader is expected to redefine
her/his role to become the primary learner in a school-based inquiry community
seeking excellence in teaching. In order to become an instructional leader, the principal
should recognize the importance of what Blase (1987) calls “leadership competencies”
for working with teachers in addition to “administrative competencies” for dealing
with the technical aspects of work. That is, it becomes the principal‟s responsibility to
work closely with teachers in an atmosphere of shared learning (Cogan, Anderson, &
Krajewski, 1993; Printy, 2008; Somech, 2005).
For instance, Reitzug (1994) examined closely the instructional leadership
behaviors of one principal who appreciated teachers‟ opinions based on their personal
practical knowledge, asked inquiry-oriented questions, encouraged risk taking, and
provided staff development opportunities and ideas. These instructional leadership
behaviors helped teachers reflect on their own practice, consider alternative frameworks
14

for understanding teaching and learning processes, and build teamwork among
colleagues. Similarly, in their analysis of a nationally sampled dataset on school
restructuring, Marks and Printy (2003) reaffirm the importance of the principal
instructional leadership model in which “the principal seeks out the ideas, insights, and
expertise of teachers . . . [to work] for school improvement . . . [by] shar[ing] [with
teachers] responsibility for staff development, curricular development, and supervision
of instructional tasks” (p. 371). In such a school setting, teachers can be empowered to
“share their experiences and knowledge [with their colleagues and principal] in freeflowing, creative ways that foster new approaches to problems” (Wenger & Snyder,
2000, p. 140).
Although it appears evident that principal instructional leadership is an
important factor that helps change the school into a “collaborative inquiry” community
(Reitzug, 1997), one may plausibly suspect that this effect may be rather limited,
especially because teachers‟ instructional activities are fraught with numerous
complexities and uncertainties for which principals do not necessarily have something
helpful to offer to teachers in terms of either clear ideas or practical solutions (Floden &
Buchmann, 1993; Jackson, 1986; Labaree, 2000; Wood, Cobb, & Yackel, 1990); this
accounts for why many principals frequently feel uncertain about the soundness of their
judgments when they supervise teachers‟ instructional practices (Lortie, 2009). In this
respect, Bidwell (2001) has developed an insightful hypothesis – he postulates that
“teachers‟ problem-solving networks are likely to differentiate according to kinds of
problems and kinds of pedagogical and subject matter knowledge and experience.
15

When faculty networks differentiate in this way, the unifying effectiveness of formal
leadership should be reduced” (p. 107).
In relation to this hypothesis, Horng, Klasik, and Loeb‟s (2010) study is very
suggestive. In their analysis of observational time-use data for all high school principals
in one district in Miami, they found that while principals‟ time spent on school
management activities was significantly associated with positive school outcomes, their
time spent on day‐to‐day instruction-related activities was marginally related to
improvements in student performance and often even negatively associated with
teacher and parent assessments of the school. These findings are counterintuitive from a
principal instructional leadership perspective because what these findings suggest is
that principals‟ well-intended instructional leadership behaviors may have a fairly
limited, if not negative, influence on teaching. These findings, however, seem consonant
with Bidwell‟s (2001) hypothesis that principals‟ instructional leadership behaviors are
likely to have only a limited effect on unifying teachers into a community of inquiry
into teaching.
In sum, the central logic in the principal instructional leadership literature is, of
course, useful for understanding the organizational context in which collegiality can be
fostered in schools. However, the literature on schools as organizations and the
literature on teachers‟ work in schools commonly point to the possibility that the effect
of principals‟ instructional leadership behaviors on collegiality among teachers in
schools may not be as strong as suggested in the principal instructional leadership
literature; this is especially because many classroom-situational and pedagogical
16

specifics that beget a variety of complexities and uncertainties confronted by individual
teachers are likely to lead to differentiation of teacher networks “according to kinds of
problems and kinds of pedagogical and subject matter knowledge and experience”
(Bidwell, 2001, p. 107) rather than leading to consolidation of community solidarity in a
given school. In this situation, the principal‟s instructional leadership behaviors may
not have a strong and stable effect on unifying teachers into a cohesive inquiry
community.

2.1.3. Curriculum Policy Perspective
The literature on the relationship between policy and classroom instruction
suggests that the absence of a standardized national curriculum in a country may
hinder teachers from engaging in collaborative interaction with other colleagues. This is
because the paucity of a shared technical language and “a common technical culture”
(Lortie, 2002, p. 76) in the profession of teaching is hard to be alleviated without a
standardized curriculum policy as a common coherent framework for teaching and
learning (Schmidt, Wang, & McKnight, 2005; Stevenson & Baker, 1991). Put differently,
one may plausibly expect that the presence of a national curriculum may facilitate
collaborative interaction among teachers in a given country insofar as the curriculum
serves as a framework that provides teachers with a shared technical language, by
which they can secure at least a certain degree of efficiency in communicating with
other teachers about teaching and learning.
Further, the literature on “sensemaking” in organizations elaborated by Weick
17

and his associates (Weick, 1995; Weick, Sutcliffe, & Obstfeld, 2005) also suggests the
possibility that the presence of a national curriculum policy may be in a positive
relationship with a high level of collaborative interaction among teachers in a given
country. That is, teachers may often need to collaboratively make sense of the content
and intention of a given curricular policy because the policy is unlikely to be free of
ambiguity in terms of its language (Hill, 2001; Spillane, et al., 2002). A curricular policy
is usually tied to many “reform ideals” (Kennedy, 2005, p. 17) that are often perceived
by teachers as remote from everyday teaching and “classroom exigencies” (Lortie, 2002,
p. 69), and such reform ideals, containing various definitions, slogans, and metaphors
(Berendt, 2008; Scheffler, 1960), may conflict with one another, adding to already
existing ambiguity in policy language (Green, 1994; Shulman, 2004). In this respect, it is
reasonable to expect that teachers often need to help one another make sense of many
aspects of a given national curriculum in order to share in the effort to respond
efficiently to the curriculum policy.
In addition, previous studies of the association between governance structure
and classroom instruction suggest the possibility that the amount of uncertainty in
teaching may be different across countries, depending on curriculum policy
arrangements (Astiz, Wiseman, & Baker, 2002; Stevenson & Baker, 1991). In particular,
the presence of a standardized national curriculum in a country may function to
alleviate some uncertainty perceived by teachers because the national curriculum may
serve as a safety net that teachers can depend on when making various decisions. In
countries with a standardized national curriculum, teaching usually occurs within a
18

nationally prescribed range of curricular content, which is likely to reduce a certain
amount of uncertainty in teaching that may exist in countries without a standardized
national curriculum.
Although this curriculum policy perspective provides a useful account of the
possible relationship between curricular policy arrangements and collaborative teacher
interaction patterns, one may plausibly believe that this relationship may not be very
strong; this is because a sizeable body of research has shown that schools are “loosely
coupled organizations” (Weick, 1976) where formal structures and internal activities are
often not closely connected (Deal & Celotti, 1980; Elmore, 2004; Meyer & Rowan, 1991).
The curriculum policy perspective presumes a tight linkage between curricular policy
arrangements and teachers‟ cognitive and behavioral patterns; however, this may not be
the case, given the possibility that schools as organizations strive to maintain
“ceremonial conformity” (Meyer & Rowan, 1991, p. 41) to the rules and values
embedded in the external policy environment (DiMaggio & Powell, 1991; Scott, 2005),
whereby their internal technical activities are buffered from pressures and influences
from the external environment. Indeed, such a loose coupling between formal
structures and internal activities is not very surprising in educational organizations
around the world (Meyer & Ramirez, 2003), considering that teaching is inherently a
complex and non-routine task for which teachers‟ professional autonomy is inevitably
necessary even in the presence of central policy control over instruction (Rowan,
Raudenbush, & Cheong, 1993; Shulman, 2004). Thus, one may reasonably expect the
relationship between curricular policy arrangements and collaborative teacher
19

interaction to be modest, if any.

2.1.4. National Culture Perspective
The literature on the cross-national differences in terms of individualist versus
collectivist orientations also provides an important insight. According to Hofstede and
his associates (Hofstede, 2001; Hofstede & Bond, 1984; Hofstede & Hofstede, 2005), a
central difference between typical individualist and collectivist societies lies in whether
people see themselves as an integral part of a social group with primary alliance to the
group or as separate individual persons with primary responsibility for themselves and
their families. In typical collectivist societies, there are a range of informal, yet socially
institutionalized, ways of creating family-like ties with other persons who are not
biological relatives but who are socially integrated into the same group. This accounts
for why “the workplace itself may become an in-group in the emotional sense of the
word . . . [which] resembles a family relationship, with mutual obligations of protection
in exchange for loyalty. . . . In the individualist society, [however,] the relationship [in
the workplace] . . . is primarily conceived as a business transaction” (Hofstede, 2001, p.
237).
The rarity of U.S. schools engaged in collaborative work around teaching is often
understood to have resulted at least in part from the individualist culture of the nation.
Leonard and Leonard (2003), for instance, attend to the possibility that the macro-level
culture of individualism in the United States penetrates the culture of U.S. schools. They
postulate that this is one of the factors contributing to a low level of collaborative
20

interaction among teachers in U.S. schools. In contrast, many studies of teaching
cultures in some East Asian countries, for example, have shown that teachers in those
countries have benefited from their traditions whereby collaborative interaction has
been institutionalized as an integral part of their work life in schools. Collaborative
“lesson study” activities in Japan (Fernandez & Yoshida, 2004; Lewis, 2000) and schoolbased “teaching research groups” and their “public lesson” activities in China (Han &
Paine, 2010; Paine & Ma, 1993) are illustrative examples here. Although it is unclear
whether or not, or to what extent, these forms of collaborative teacher interaction in
these countries have been sustained by their collectivist cultural orientations, it would
not be unreasonable to suspect that there may be some relationship between their high
levels of teacher collaboration and their collectivist orientations grounded in their
cultural norms and values.
Despite the useful insight gained from this national culture perspective on
teacher collegiality, there is some empirical evidence to believe that the level of teacher
collegiality in schools may not be heavily dependent on national cultural contexts. In
their study of teachers‟ working conditions and beliefs in the United States, Germany,
and Japan, LeTendre and his colleagues (LeTendre, Baker, Akiba, Goesling, & Wiseman,
2001) report considerable cross-national similarities in “core teaching practices and
teacher beliefs” despite “some differences in how teachers‟ work is organized” (p. 3).1

1. Further, some sociologists argue that “education is an institution . . . that at a deeper
level is strongly affixed to global norms and rules about what education is and how schools
should operate” (Baker & LeTendre, 2005, p. 8). Extensive empirical studies provide substantial
evidence supporting the notion that many aspects of schooling are, by and large, the
21

Although it is still in debate to what extent teaching is a cultural practice contextualized
within individual countries (Alexander, 2000; Anderson-Levitt, 2003; Baker & LeTendre,
2005; Givvin, Hiebert, Jacobs, Hollingsworth, & Gallimore, 2005; Kim, Ham, & Paine,
2011; McEneaney & Meyer, 2000), the cross-national similarities LeTendre and his
colleagues have found in teachers‟ core teaching practices and beliefs point to the
possibility that the influence of national cultural orientations on teachers‟ collaborative
interaction with their colleagues may be rather limited, as long as the patterns of
interaction among teachers are unlikely to be detached from their core teaching
practices and beliefs that may be fairly similar across countries despite cultural
differences.2

2.1.5. Uncertainty Management Perspective: An Alternative Approach
Teaching has often been seen as an uncertain technology; instructional decisions
with respect to how to promote student learning in a particular classroom environment
can never be made with absolute certainty (Jackson, 1986; Lortie, 2002; McLaughlin &
Talbert, 2001; Rosenholtz, 1989). Although teachers are expected to possess a set of

embodiments of institutional rules and values that are embedded in the cultural dimension of
world society (Benavot, Cha, Kamens, Meyer, & Wong, 1991; Boli & Ramirez, 1986; Cha &
Ham, 2011; Ham & Cha, 2009; Meyer & Ramirez, 2003). On the basis of this line of
sociological thought, LeTendre and his colleagues’ findings about considerable cross-national
similarities in “core teaching practices and teacher beliefs” are not very surprising. For critiques
and further discussions, see Anderson-Levitt (2003), Ham, Paine, and Cha (2011), Paine and
Fang (2006), and Schriewer (2003), for example.
2. See also Anderson-Levitt’s (2002) discussion and critique of LeTendre and his
colleagues’ findings.
22

specialized knowledge and skills, such a social expectation contrasts with the relatively
weak core technology teachers actually have. Indeed, classroom situations are full of
uncertainties, which are often not adequately explained by general theories taught in
teacher preparation programs (Floden & Buchmann, 1993; Floden & Clark, 1988;
Labaree, 2000).3 The practice of teaching is, by nature, context-specific and full of
situated-complexities, so that it cannot always be theorized in a generic way of
explanation (Lampert, 2001; Munthe, 2007; Putnam & Borko, 2000; Wood, et al., 1990).
Numerous possibilities to lose lesson momentum are the hallmark of classroom
teaching (Kennedy, 2005). Teachers are always confronted with the vivid realities of
classroom teaching that are often inexplicable in terms of theoretical conceptualizations
of educational practice. Indeed, the practice of teaching is often seen as “the very
prototype of the idiographic, individual, clinical enterprise” (Shulman, 2004, p. 139).
This accounts for why teachers usually have a wider range of concerns than educational
administrators and policy makers, whose attention is usually focused on a relatively
small number of educational issues in the form of abstract reform ideals (Kennedy,
2005).
Although teaching is inherently an uncertain task as described above, I postulate
that there is another kind of uncertainty whose amount may vary depending on
instructional strategies teachers employ and classroom contexts in which teaching
3. This may be one of the reasons why many of the formal qualification profiles of
teachers such as training and certification often do not significantly matter to student
performance as shown in both U.S. and international research findings (Aslam & Kingdon, 2011;
Hanushek & Rivkin, 2006). However, for a different perspective, see, e.g., Kennedy, Ahn, and
Choi (2008).
23

occurs. In this study, I conceptualize what I call “uncertainty in teaching” as a state of
doubt or a feeling of incertitude about particular instructional or classroom situations as
perceived by teachers. When operationalized in this way, uncertainty in teaching is
likely to increase when details of given situations appear complex or unpredictable due
to insufficient information, knowledge, or resources readily available to teachers. More
specifically, I posit that uncertainty in teaching may come from at least two
conceptually separable, yet interrelated, sources. First, an important source of
uncertainty in teaching is teachers‟ recognition of the instructional complexities that
result from their effort to use varying teaching strategies to foster students‟ engagement
in meaningful learning experiences. For example, when teachers try to provide their
students with creative learning tasks to elicit higher-order thinking and deeper
understanding, teaching becomes less reducible to predictable routines, which
inevitably increases what I call here instructional uncertainty. In other words, when
teachers use teaching strategies that open up possibilities for students to engage in
intellectually challenging questions, develop and explain their own ideas, and relate
what they have learned to examples from their daily lives, teachers are likely to
confront a greater amount of instructional uncertainty.4 Another important source of
uncertainty in teaching is the difficulty in effectively dealing with various complex
issues arising from a range of classroom-contextual conditions and constraints. Of
course, classroom teaching as a task dealing with a group of students makes teaching
4. This kind of instructional strategies is often collectively referred to as “teaching for
understanding” (McLaughlin & Talbert, 1993; Wiske, 1998). See section 2.2.1 in this chapter for
a further discussion.
24

inherently uncertain; however, I emphasize the possibility that teachers may face
additional uncertainty whose amount varies depending on how certain classroomcontextual factors facilitate or constrain their teaching. For example, if teaching occurs
in a classroom environment with a high level of student variability or in an
environment with insufficient classroom resources to effectively promote student
leaning, teachers can be faced with a variety of difficult challenges and complex class
dynamics they need to deal with in their everyday teaching practice. I call such
challenges and complexities classroom-contextual uncertainty in this study. That is, the
amount of classroom-contextual uncertainty teachers confront may vary from one
classroom to another depending on the situations of the classroom context in which
teachers are positioned to implement their teaching practice.5
There is some sociological literature that sheds light on the relationship between
the nature of teaching as a complex and uncertain task and the collegial behaviors of
teachers. Rowan and his colleagues (Rowan, 1990; Rowan, et al., 1993), for instance,
posit that when teaching is understood to be a complex activity, “organic” structures
are likely to emerge in schools; he and his colleagues assert that teachers who recognize
5. It should be noted that there may be various ways to conceptualize the uncertainties
teachers may confront in their work life (Floden & Buchmann, 1993; Labaree, 2000; Lortie,
2002; Munthe, 2007; Rosenholtz, 1989). The notion of uncertainty in teaching that this study
relies on is, therefore, only one of many possible conceptualizations. Alternative avenues to
frame the notion of uncertainty deserve further systematic research in order to examine how
varying conceptualizations of uncertainty are similarly or differently related to collaborative
teacher interaction patterns. In addition, further studies are needed to see if (or to what extent)
what I have described above as sources of uncertainty in teaching really cause teachers to
experience feelings of uncertainty. Qualitative case studies, accompanied by some psychological
testing, would be very useful in this respect.
25

the complexities of teaching are more prone to form collegial networks for sharing
useful information and mutual support. In their analysis of longitudinal data from a
San Francisco Bay Area, Cohen, Deal, Meyer, and Scott (1979) also found that their
school-level index capturing the level of “differentiation of teaching materials and
students” as their “indicator of technological complexity within classrooms” (p. 23) was
significantly associated with increased communication and teaming within schools.
Similarly, Bidwell (2001) postulates that teachers are likely to turn to other colleague
teachers for guidance and support when they are situated “in an environment of
substantive uncertainty, [where] pedagogical doctrines rarely provide procedural
templates of sufficient specificity to guide [their] day-to-day practice effectively” (p.
106); in other words, when they feel “unsure of their technical footing, . . . teachers who
are similarly situated in a school . . . [are likely to] pool the resources of their individual
training and experience in collective problem solving” (pp. 105-106).
It is worth noting that this line of sociological literature is consistent with
organizational contingency theories positing that the degree of complexity of an
organization‟s structure is contingent upon the degree of complexity of the task
performed and the level of uncertainty involved in making decisions about what to do
in order to successfully complete the task (Galbraith, 1973; Perrow, 1986). In March and
Simon‟s (1993) words, collegial networks, thus, emerge due to “the limits of human
intellective capacities in comparison with the complexities” (p. 190) that organizational
members are faced with. According to this line of theoretical reasoning, “[i]f the
complexity of the task generates uncertainty, then lateral relations between workers can
26

serve as a source of problem-solving and processing of information as well as
coordination” (E. G. Cohen, et al., 1979, p. 21). This accounts for why “[n]etworks of
colleague-to-colleague consultation and advice . . . [are often] more capable of
coordinating the work of colleagues than the formal administrative hierarchy” (Bidwell,
2001, p. 105).
In addition, another related body of literature in organizational analysis suggests
that, when faced with complex and unpredictable situations, organizational members
frequently rely on collective sensemaking, which is a collaborative process through
which the link between actions and beliefs is clarified (Weick, 1995). In other words,
collective sensemaking helps organizational members to link an action to a set of beliefs
that are already clear. Since uncertainty can be understood as a state of doubt about
particular situations as perceived by organizational members, uncertainty is likely to
arise when details of situations are complex or unpredictable due to insufficient or
inconsistent information available (Brashers, 2001; Perrow, 1986). If what action to take
is unclear to organizational members under a given uncertain situation, collective
sensemaking as part of collaborative effort to overcome individual organizational
members‟ “bounded rationality” (March & Simon, 1993; Simon, 1991) can serve as an
effective mechanism that “prevents an already uncertain world from becoming a
meaningless one” (Collinson & Cook, 2007, p. 38).
In this respect, collegiality can be seen largely as a byproduct of the collective
sensemaking of uncertainties, which reduces organizational members‟ feelings of
discomfort with given uncertain situations and thus helps them manage uncertainties
27

without actually removing them completely from given situations (Berger, 1995;
Brashers, 2001). Supportive others facilitate uncertainty management as “sources of
information, collaborators in information gathering, evaluators of information, or
buffers against information” (Brashers, 2001, p. 485). Further, it is expected that the
collective sensemaking of uncertainties allows organizational members to reduce costs
because they share in the effort to respond to given uncertainties (Smith-Doerr & Powell,
2005; Williamson, 1994). Collective sensemaking, thus, can be understood as an
interpretive cognitive process used by organizational members in order to share
understandings, whereby learning becomes “distinctly organizational [as] it relies on
the combined experiences, perspectives, and capabilities of a variety of organizational
members” (Rait, 1995, p. 72).

2.2. Hypotheses

2.2.1. Primary Hypotheses
On the basis of the uncertainty management perspective, this study postulates
that teacher collegiality is largely a collective response to uncertainty which teachers
face in their teaching. In this regard, three primary hypotheses are examined in this
study.
First of all, an important example showing the complexities teachers go through
is that many teachers often do not encourage their students towards deeper intellectual
engagement (Cusick, 1983; McNeil, 1986; Powell, Farrar, & Cohen, 1985). As teachers try
28

harder to encourage their students towards deeper intellectual engagement, they are
more likely to be exposed to “the uncertainties that are created when students offer
unexpected ideas” (Kennedy, 2005, p. 189); this is especially the case when teachers do
not have strong competence in content or pedagogical knowledge (Ball, Thames, &
Phelps, 2008; Shulman, 2006; Wilson, Floden, & Ferrini-Mundy, 2001). Further, in order
to help students have enough “cognitive flexibility” (Spiro, Feltovich, Jacobson, &
Coulson, 1992) to avoid oversimplification of complex concepts or topics, teachers
cannot rely solely on routinized teaching methods. With respect to mathematics
teaching, for example, it has been well documented that teachers‟ knowledge and skills
exert important influences on the extent to which they utilize instructional strategies
focusing on “teaching for understanding” (D. K. Cohen, McLaughlin, & Talbert, 1993;
Gamoran, et al., 2000; Wiske, 1998). Unlike conventional teaching accompanied by a
range of readily available routines that teachers can draw upon, teaching for
understanding requires teacher knowledge that goes beyond knowing the prescribed
curricular content supposed to be delivered to students; in particular, research
emphasizes that teachers should have deep knowledge about how students reason
mathematically (Fennema & Romberg, 1999; Hiebert & Carpenter, 1992; Prawat,
Remillard, Putnam, & Heaton, 1992; Silver, Mesa, Morris, Star, & Benken, 2009).
Rather than encouraging their students towards deeper intellectual engagement,
many teachers try to find “an optimal level of student engagement” (Kennedy, 2005, p.
183); from teachers‟ point of view, there may be considerable advantages to
conventional teaching because it helps protect them from the uncertainties that could
29

emerge from students‟ unexpected responses. Indeed, as teachers put more effort to
incorporate instructional strategies for teaching for understanding into their classroom
teaching, they must confront greater instructional uncertainty; that is, the practice of
teaching becomes inevitably more unpredictable, less reducible to predictable routines
or “defensive teaching” (McNeil, 1986) practices, thereby exposing teachers to a greater
extent to the notion of what Heaton (2000) calls “teaching as an improvisational activity”
(p. 60) that requires “moment-to-moment responsiveness” (p. 63) in interacting with
students. Consequently, teachers are likely to reach out to other colleague teachers,
whereby they can not only exchange practical suggestions but also “reduce
inappropriate pressures for certainty” (Floden & Buchmann, 1993, p. 380). This is why
collegial networks may emerge among teachers when they try to develop and use
instructional strategies for teaching for understanding. Therefore, the following
hypothesis is put forth:

Hypothesis 1: Teachers who exert greater effort to implement teaching for
understanding are more likely to engage in collaborative interaction with other
teachers.

Also, a range of classroom-contextual factors, including a low level of student
performance, a high level of student heterogeneity, and a shortage or inadequacy of
classroom resources, for example, are likely to lead to increased uncertainty in teaching.
In particular, how to accommodate diverse students from a wide range of
30

socioeconomic, cultural, and linguistic backgrounds inevitably involves a variety of
difficult questions with regard to both curricular content and instructional methods
(Gay, 2000; Ladson-Billings, 1999; Miller, Kostogriz, & Gearon, 2009), primarily because
“there are few well-established techniques – codified technical knowledge – to help
teachers meet students‟ widely varying needs” (Rosenholtz, 1989, p. 4). In a similar vein,
a teacher is likely to be exposed to a high level of uncertainty in teaching if she/he
teaches a classroom where students show widely varying academic performance levels
(Grimmett & MacKinnon, 1992; Hallinan, 1994). In this case, the teacher is expected to
make considerable extra effort in order to effectively accommodate the diverse
academic needs of individual students; however, this sharply contrasts with the reality
that teachers‟ instructional “goals must be met and relationships [with students]
managed in a group context . . . [in which teachers‟ educational] attempts to shape
children are continually constrained by the fact of „classness‟” (Lortie, 2002, p. 137).6
Further, the shortage and inadequacy of classroom resources experienced in many
schools is likely to add additional complexities to classroom teaching insofar as teachers
perceive how they teach their students is negatively affected by such resource
constraints (Anyon, 2001; Biddle & Berliner, 2002; D. K. Cohen, Raudenbush, & Ball,
2003). In sum, considering that teachers‟ uncertainty in teaching may be derived from a
wide range of classroom-contextual factors, it is possible to advance the following

6. Considering that the internal schooling process “occurs within an environment of
constraints, uncertainty, and change” (DeLany, 1991, p. 202), it is not surprising that dealing
with students from a wide range of backgrounds is at the intersection of various variables
exerting direct or indirect influences on teaching.
31

hypothesis:

Hypothesis 2: Teachers who face greater classroom-contextual uncertainty are more
likely to engage in collaborative interaction with other teachers.

Finally, extending the central logic of the uncertainty management perspective, it
is plausible to expect that teachers‟ uncertainty in teaching may arise in part from sociocultural heterogeneity at the societal level. Considering that education is an institution
on which different sectors of society impose multiple, often conflicting, purposes,
“teaching performances . . . can be free of controversy only in societies which are
marked by an extremely high degree of value consensus” (Lortie, 2002, p. 136). In a
socio-culturally heterogeneous society, the range of the goals of schooling envisioned
by different groups of people is likely to be very wide, and many of those goals may be
conflicting with one another, adding further uncertainty to the complexities of teaching.
This leads to the following hypothesis:

Hypothesis 3: Teachers in countries with a higher level of socio-cultural heterogeneity
are more likely to engage in collaborative interaction with other teachers.

2.2.2. Additional Hypotheses
When testing the primary hypotheses derived from the uncertainty management
perspective, it is important to control for other possibilities by simultaneously
32

considering other plausible hypotheses as well that are based on different theoretical
perspectives. This process makes it possible to see if the primary hypotheses turn out
empirically supported even after some other plausible hypotheses are simultaneously
taken into account. In this regard, five additional hypotheses are considered here. Each
of these additional hypotheses is based on one of the aforementioned theoretical
perspectives other than the uncertainty management perspective. They are the school
climate perspective, the principal instructional leadership perspective, the curriculum
policy perspective, and the national culture perspective.
First of all, on the basis of the school climate perspective on teacher collegiality,
the extent to which collegial work relationships are fostered and sustained among
teachers in a school is dependent on the organizational climate of the school. Therefore,
the following hypothesis is set forth:

Hypothesis A1: Teachers’ collaborative interaction with other teachers is more likely in
schools with a more positive organizational climate.

In addition, the principal instructional leadership perspective suggests the
possibility that teacher collegiality is facilitated to a greater degree in schools where
principals engage more in instructional leadership roles. In this respect, the following
hypothesis is derived:

Hypothesis A2: Teachers’ collaborative interaction with other teachers is more likely in
33

schools where the principal’s engagement in instruction-related activities is
greater.

Further, on the basis of the curriculum policy perspective on teacher collegiality,
it is postulated that the presence of a standardized national curriculum may function to
facilitate collaborative interaction among teachers insofar as it provides them with some
shared technical language for thinking about teaching and learning.7 Thus, the
following hypothesis is formulated:

Hypothesis A3: Teachers in countries with a more standardized national curriculum
are more likely to engage in collaborative interaction with other teachers.

At the same time, the curriculum policy perspective also suggests the possibility
that the presence of a standardized national curriculum may alleviate a certain amount
of uncertainty in teaching perceived by teachers. This is because curriculum
7. In addition, one may plausibly hypothesize that curricular policy stability, as
conceptualized as the degree to which the curriculum policy of a given country remain
unchanged over time, may be negatively related to collaborative teacher interaction in that
country. That is, in a highly stable curricular policy environment, teachers may not feel they
really need to communicate much with other teachers, precisely because each teacher can
anticipate fairly accurately how other teachers are thinking and feeling and what they are doing
in a given situation, thereby decreasing the chances for teachers to engage in collaborative
interaction. In other words, a high level of curricular policy instability as an indicator of frequent
policy changes in a given country is likely to be positively related to a high level of collaborative
teacher interaction. Although this study does not test this hypothesis due to the lack of available
data to measure curricular policy (in)stability, a useful contribution with regard to future research
would be to examine this hypothesis.
34

standardization may lessen a certain amount of uncertainty in teaching as the
curriculum standards may serve as a set of readily available and authoritative
guidelines that teachers can rely on when making various decisions about their teaching.
In this respect, this possibility is in line with the central logic of the uncertainty
management perspective as well. Thus, the following hypothesis is advanced:

Hypothesis A4: The positive association between teachers’ exposure to uncertainty in
teaching and their engagement in collaborative interaction with other teachers (=
hypotheses 1 and 2) is likely to be weaker in countries with a more standardized
national curriculum.

Finally, drawing on the national culture perspective, I consider the literature on
the cross-national differences in terms of individualist versus collectivist orientations,
which points to the following hypothesis regarding teacher collegiality:

Hypothesis A5: Teachers in countries with a more individualist culture are less likely to
engage in collaborative interaction with other teachers.

2.3. Summary

On the basis of five different conceptual perspectives, a series of hypotheses has
been formulated as an attempt to gain a comprehensive understanding of teacher
35

collegiality as a dependent variable contingent upon a range of multilevel contextual
factors. Three primary hypotheses of this study have been derived from an uncertainty
management perspective; from this perspective, teacher collegiality is understood
largely as a byproduct of the collective sensemaking of uncertainties in teaching, which
reduces individual teachers‟ feelings of discomfort with given uncertainties and thus
helps them deal with those uncertainties rather than simply trying to either ignore or
eliminate them from their teaching. Four additional perspectives – a school climate
perspective, a principal instructional leadership perspective, a curriculum policy
perspective, and a national culture perspective – have yielded five more hypotheses,
which represent some of the most popular conventional accounts of teacher collegiality
as a variable dependent on certain contextual conditions. In the next chapter, I provide
detailed descriptions of the data, specific measures, and analytic methods to test the
hypotheses advanced in this chapter.

36

CHAPTER 3
DATA AND METHODOLOGY

3.1. Description of the Data

3.1.1. Sample Design and Weights
The primary data for this study are from the TIMSS 2007 eighth-grade
mathematics dataset. This dataset, chiefly known for its student achievement data, also
contains extensive contextual data on schools and classes, including useful information
about collaborative interaction among teachers.8 The sample design of the TIMSS 2007
eighth-grade mathematics survey was intended to provide accurate estimates of the
nationally representative eighth-grade student population within each participating
country.9 In addition to data on sampled students, a large amount of information about
their classes and schools was also gathered. The hierarchical nature of these units
necessitated a two-stage sample design, whereby a systematic probability-proportionalto-size sample of schools was selected at the first sampling stage, and one or more intact
classes of students were sampled per school with an equal probability of selection at the
8. Using an earlier TIMSS dataset, Akiba and LeTendre (2009) provide some useful
descriptive information about collaborative teacher interaction in Australia, Japan, and the
United States.
9. Each country was expected to include all eighth-grade students in its definition of the
student population. Even in countries where some regions or school types could not be included
in the operational population for some practical reasons, such exclusions were predetermined not
to exceed five percent of the population (Joncas, 2008a).
37

second stage.10
Since this study is concerned primarily with collaborative teacher interaction, the
teacher is the primary unit of analysis. In terms of multilevel regression modeling,
teacher-level data constitute level-1 variables. The teachers selected in TIMSS 2007 were
those who taught the nationally representative sample of classes within each country. In
TIMSS 2007, teacher-level, classroom-level, and school-level data are virtually at the
same level in terms of data structure, while student-level data are distinguished as a
separate level nested within teacher/classroom/school-level data. The reason for this is
that only one or two intact classes were selected per school in most cases in TIMSS 2007,
and each class was taught mathematics by only one or two teachers in most cases.
Therefore, although teachers, classrooms, and schools are all separate levels
conceptually, it is unreasonable to treat them as separate levels in statistical analyses.
This study uses the classroom level as the data level to which teacher-level and
school-level data are matched together. That is, in places where the number of cases
differed between these levels, the number of cases was adjusted to be consistent with
the number of classrooms in the dataset. For example, if two classes were included for a
given school in the dataset, the same school-level data were matched to the two classes,
resulting in the number of cases remaining two; similarly, if one of those two
classrooms was taught mathematics by two teachers rather than one, the two teachers

10. While all participating countries adopted this basic two-stage sampling approach,
there were some acceptable variations. The Russian Federation introduced a preliminary stage to
select regions, and Singapore added a third sampling stage to subsample students within
classrooms rather than selecting intact classes (Joncas, 2008a, 2008b).
38

were weighted so that they would represent only one case statistically, making the
number of cases remain consistent with the number of classrooms.
This strategy is useful because the TIMSS 2007 dataset contains a weight variable
designed to be used for analyses where the unit of analysis is the classroom. This study
uses this weight variable after transforming it into a purely proportional weight
variable by removing the scaling effect within each country.11 This ensures the
weighted sample corresponds to the actual sample size and, thus, yields correct
standard errors in subsequent analyses. This weight variable, WEIGHT_WITHIN, is
used in this study when within-nation analyses are conducted, adjusting for differential
selection probabilities of individual teachers within a given country.
Although WEIGHT_WITHIN is a desirable weight variable for within-nation
analyses, it has a drawback for cross-national analyses. Using WEIGHT_WITHIN
would lead to biased results in cross-national analyses because results would be more
sensitive to countries with larger samples of teachers. In order to avoid this problem, it
is important to make each country treated equally by using an alternative weight
variable. Thus, WEIGHT_WITHIN was modified into an alternative weight variable,
WEIGHT_ACROSS, in order to make each country have the same sample size while
ensuring that the WEIGHT_ACROSS-weighted number of cases in the cross-national
dataset corresponds to the sum of the WEIGHT_WITHIN-weighted numbers of cases
from individual countries. In this respect, this alternative weight variable, like
11. The TIMSS 2007 dataset does not include a weight variable for teacher-level
analyses. Although there are several teacher-related weight variables in the dataset, they are all
designed for using teacher-background data in student-level analyses (Foy & Olson, 2009).
39

WEIGHT_WITHIN, is also a proportional weight variable that does not inflate the
sample size on the entire dataset level. In this study, when data are analyzed after
national datasets have been combined into a single cross-national dataset,
WEIGHT_ACROSS is applied to the combined dataset before analyses are conducted.

3.1.2. Units of Analysis
This study explores how collaborative teacher interaction is associated with
individual teachers‟ immediate classroom and school contexts as well as with the
broader societal and policy contexts by which they are surrounded. Thus, the research
design for this study requires the simultaneous consideration of statistical relationships
at both teacher and country levels. Consequently, there are two different unit-ofanalysis levels, which are the teacher level and the country level. Since teachers are
nested within each country in terms of the data structure, teachers constitute the level-1
data, which are clustered within individual countries at level 2.12

3.2. Dependent Variables: Measures of Teacher Collegiality13

A question in TIMSS 2007 asked eighth-grade mathematics teachers about the
12. The TIMSS 2007 dataset does not include either district-level or state-level data for
the United States. With regard to future research, a useful within-U.S. analysis would be to use
the School and Staffing Survey dataset compiled by the National Center for Education Statistics.
In this dataset, district-level data are also available in addition to some useful data on
collaborative teacher interaction.
13. For a conceptual discussion of teacher collegiality, see chapter 1 (section 1.1).
40

frequency of different types of collaborative interaction with other teachers. The
question asked: “How often do you have the following types of interactions with other
teachers?” The types of collaborative interaction given were: “discussions about how to
teach a particular concept,” “working on preparing instructional materials,” “visits to
another teacher‟s classroom to observe his/her teaching,” and “informal observations of
my classroom by another teacher.” For each interaction type, teachers were asked to
choose one of the following response options: “never or almost never,” “2 or 3 times per
month,” “1-3 times per week,” or “daily or almost daily.” These response options were
assigned with numeric values of zero, one, two, and three, respectively.
Based on these four types of collaborative interaction between teachers, two
teacher collegiality measures, collaborative lesson planning (TCHRCOL_LESSON) and
collaborative classroom observation (TCHRCOL_CLASS), were constructed.
TCHRCOL_LESSON captures the degree to which an eighth-grade mathematics teacher
regularly engages in collaborative lesson planning; that is, the mean of “discussions
about how to teach a particular concept” and “working on preparing instructional
materials” is used as one measure of teacher collegiality. Similarly, TCHRCOL_CLASS
captures the degree to which an eighth-grade mathematics teacher regularly
participates in collaborative classroom observation; this variable is the mean of “visits
to another teacher‟s classroom to observe his/her teaching” and “informal observations
of my classroom by another teacher.”14

14. Although these measures are based on quantitative data about how often teachers
participate in certain types of collaborative interaction, these measures do not capture various
41

Little (1990) argues that collaborative lesson planning and classroom observation
constitute strong forms of collegiality. Such joint work implies collective action, creating
strong interdependence, shared responsibility, and a great degree of readiness to
participate in reflective inquiry into practices. Failure and uncertainty are not hidden,
but they are shared and discussed. According to some previous studies, however, weak
forms of collegiality are dominant in schools (Little, 1990; Lortie, 2002); they show that
collegial work practices are usually limited to giving help only when asked or pooling
some existing ideas. This phenomenon has led to the emergence of what Fullan and
Hargreaves (1996) call “comfortable collegiality” (as cited in Fallon & Barnett, 2009),
whereby reassurance and sympathy are emphasized while close scrutiny and criticism
are discouraged.
In addition, an alternative collegiality measure, variety in collaborative teacher
interaction (TCHRCOL_VARIETY), was constructed based on the above two
collaboration measures. TCHRCOL_VARIETY is an ordinal-scale variable with values
of zero, one, and two, for none, low, and high variety, respectively, in collaborative
interaction among eighth-grade mathematics teachers. An eighth-grade mathematics
teacher belongs to the category of “high variety” if she/he participates in both
TCHRCOL_LESSON and TCHRCOL_CLASS; “low variety” if she/he participates in
only one of TCHRCOL_LESSON and TCHRCOL_CLASS; and “none” if she/he does
not participate in either TCHRCOL_LESSON or TCHRCOL_CLASS. This alternative
qualitative aspects of teacher interaction. Therefore, a fruitful area of future research would be to
explore conceptual and methodological issues concerning which qualitative aspects of teacher
interaction to observe and how to adequately measure those aspects.
42

measure captures how various types of collaborative interaction are used by a teacher,
regardless of how frequently she/he engages in collaborative interaction with other
teachers.

3.3. Independent Variables

3.3.1. Primary Independent Variables
In order to test hypotheses 1 and 2 from the uncertainty management perspective,
two level-1 composite variables, uncertainty from teaching for understanding and
uncertainty from classroom-contextual factors, were derived from the following seven
variables: teaching for understanding, student achievement level, variation in student
achievement, students’ educational capital, variation in educational capital, proportion of
language minority students, and shortage of classroom resources. The description of each of
these variables is as follows:15
Teaching for understanding (TCHUNDS): an interval variable that measures the
degree to which an eighth-grade mathematics teacher tries to use instructional

15. I do not assume that uncertainty in teaching arises to the same degree from the same
situations regardless of societal contexts. Thus, for the multilevel regression analyses in chapter
5, I permit all slopes and the intercept of the model to randomly vary across countries in light of
the possibility of significant random differences. The comparative education literature suggests
that there may be various country-specific factors that are unique to individual societies’
education systems and cultures (Anderson-Levitt, 2003; Crossley & Watson, 2003; Schriewer,
2003). Despite this statistical strategy used in this study, it would be a useful future study to
closely examine how the specific situations from which uncertainty arises may be different or
similar across countries.
43

strategies to foster students‟ engagement in inquiry-based learning and non-routine
problem solving. A question in TIMSS 2007 asked teachers, “How often do you usually
ask [students] to do the following?” Four items focused on teaching for understanding:
asking students to “work on problems for which there is no immediately obvious
method of solution,” “decide on their own procedures for solving complex problems,”
“explain their answers,” and “relate what they are learning in mathematics to their
daily lives.” The response options were: “never,” “some lessons,” “about half the
lessons,” and “every or almost every lesson.” These response options were given
numeric values of zero, one, two, and three, respectively. The mean of these four items
is used as the variable.16
Student achievement level (ACHIEVE): a continuous variable that measures the
overall mathematics achievement level of the classroom that an eighth-grade
mathematics teacher teaches. In order to construct this variable, the individual studentlevel mathematics achievement data from TIMSS 2007 were aggregated into the
classroom level, which was then linked to the eighth-grade mathematics teacher data.17
The mean score of mathematics achievement for a given classroom is used as the

16. An exploratory factor analysis of these four items was conducted with the combined
sample of all countries for this study after WEIGHT_ACROSS was applied to the dataset. The
result showed that all items were loaded on a single factor (eigenvalue = 2.05) and all items had
factor loadings exceeding 0.65 (factor loadings for the four items, in the order shown in the
preceding paragraph, were 0.67, 0.79, 0.67, and 0.72), indicating a high degree of consistency
among the items. Cronbach’s α was 0.68.
17. At the student level, the first plausible achievement score from TIMSS 2007 was
used for each student. Using different plausible achievement scores did not alter the results of the
analyses presented in this study.
44

variable (divided by 100 to make 100 score points as a one-unit).18
Variation in student achievement (ACHIVAR): a continuous variable that measures
the variation in mathematics achievement in the classroom that an eighth-grade
mathematics teacher teaches. To construct this variable, the student-level mathematics
achievement data from TIMSS 2007 were aggregated into the classroom level, which
was then linked to the eighth-grade mathematics teacher data. The standard deviation
of the mean score of mathematics achievement for a given classroom was changed into
the coefficient of variation, which is the ratio of the standard deviation to the mean in
terms of percentage.19 This coefficient of variation is used as the variable (divided by 10
to make 10% as a one-unit).
Students’ educational capital (EDCAPIT): a continuous measure that captures the
overall socioeconomic background of the classroom that an eighth-grade mathematics
teacher teaches. For the construction of this measure, a student-level index of
educational capital was first constructed from TIMSS 2007 as done by Chudgar and
Luschei (2009) using TIMSS 2003. That is, based on individual students‟ answers
regarding a range of family possessions related to learning, the student-level sum of
those family possessions was calculated from the following: “study desk,” “dictionary,”

18. Student achievement results from TIMSS are reported on a standardized scale
ranging from zero to 1,000. In order to compare student achievement over time across countries,
each TIMSS administration uses the same standardized scale, which has an international mean of
500 and a standard deviation of 100.
19. While the standard deviation must always be understood in the context of the mean of
a given group, the coefficient of variation is a dimensionless number that makes a comparison
possible regardless of varying means across groups.
45

“calculator,” “computer,” “internet connection” (for each of these items, “yes” = one;
“no” = zero), and “books” in the home (one = “none or very few” to five = “more than
200 books”).20 Next, this student-level information was aggregated into the classroom
level. Finally, the classroom-level mean was linked to the eighth-grade mathematics
teacher data. As a result, EDCAPIT is defined as an index of the overall socioeconomic
status background of a given classroom that an eighth-grade mathematics teacher
teaches.21
Variation in educational capital (EDCAVAR): a continuous variable that captures
the variation in students‟ socioeconomic status backgrounds in the classroom that an
eighth-grade mathematics teacher teaches. For this measure, a student-level index of
educational capital was first constructed from TIMSS 2007 based on individual students‟
answers regarding a range of family possessions related to learning. That is, the
student-level sum of those family possessions was calculated as done for the EDCAPIT
variable. Next, this student-level index was aggregated into the classroom level, which
was then linked to the eighth-grade mathematics teacher data. Finally, the classroomlevel standard deviation of this index was changed into the coefficient of variation, i.e.,

20. In their cross-national study of the school effect vis-à-vis the family background
effect on student achievement, Chudgar and Luschei (2009) constructed basically the same index
of educational capital from the TIMSS 2003 dataset, and they found the index to be “a fair proxy
for the family’s economic status and parental education” (p. 637).
21. Although EDCAPIT is a rough socioeconomic status (SES) index, the national mean
of this SES index is very highly correlated with the national economic development level,
supporting the reliability of the index. The correlation between the national mean of EDCAPIT
and the gross domestic product per capita (at purchasing power parity, 2007) is 0.76 (p ≤ .001, n
= 29 countries).
46

Table 3-1. Exploratory factor analysis of the level-1 primary independent variables
Factor 1:
Uncertainty from
classroom-contextual factors
(UNC_CONTEXT)
Students' educational capital (EDCAPIT)
Variation in educational capital (EDCAVAR)
Student achievement level (ACHIEVE)
Variation in student achievement (ACHIVAR)
Shortage of classroom resources (SHORTRES)
Proportion of language minority students (LANGMIN)
Teaching for understanding (TCHUNDS)

-.903
.856
-.855
.729
.546
.505
.131

Factor 2:
Uncertainty from
teaching for understanding
(UNC_TCHUNDS)
.082
.004
-.063
.178
-.063
-.471
.874

Note . Principal component extraction with promax rotation (kappa = 4) was used. Factor loadings are sorted by size. Factor
loadings greater than .50 in terms of their absolute values are indicated by dotted-line boxes. Bartlett’s test of sphericity is
significant at the p < .001 level. The correlation between factor 1 and factor 2 is -0.015. n = 6,624.

47

the ratio of the standard deviation to the mean in terms of percentage. This coefficient of
variation is used as the variable (divided by 10 to make 10% as a one-unit).
Proportion of language minority students (LANGMIN): a continuous variable that
captures the degree to which the classroom that an eighth-grade mathematics teacher
teaches is populated by language minority students. Uncertainty may arise if teachers
feel unsure of how to accommodate the special needs of language minority students. A
question in TIMSS 2007 asked eighth-grade students, “How often do you speak [the
language of the test] at home?” Students were asked to choose one of the following
response options: “always,” “almost always,” “sometimes,” or “never.” The percentage
of students who “never” or only “sometimes” speak the language of the test was
calculated at the classroom level, which was then linked to the teacher data. Thus, the
resulting variable is the percentage of language minority students in the classroom that
an eighth-grade mathematics teacher teaches (divided by 10 to make 10% as a one-unit).
Shortage of classroom resources (SHORTRES): an interval variable that measures
the degree to which an eighth-grade mathematics teacher perceives that her/his
teaching is negatively affected by a shortage or inadequacy of classroom resources.
Such resource constraints can add uncertainty to classroom teaching because teachers
may feel unsure of how to teach effectively without adequate resources. A question in
TIMSS 2007 asked teachers, “In your view, to what extent do the following limit how
you teach the TIMSS class?”22 The following four items were included regarding the

22. The TIMSS 2007 Teacher Questionnaire indicates on its “General Directions” page
(p. 2) that the term “TIMSS class” refers to “the class that is identified on the cover of this
48

shortage or inadequacy of basic classroom resources: “shortage of textbooks for
students‟ use,” “shortage of other instructional equipment for students‟ use,” “shortage
of equipment for your use in demonstrations and other exercises,” and “inadequate
physical facilities.” For each of these items, the response options were “not at all” or
“not applicable,” “a little,” “some,” and “a lot.” These response options were given
numeric values of zero, one, two, and three, respectively. The mean of the four items is
used as the variable.23
Since all of these seven variables have been constructed to measure teachers‟
uncertainty in teaching commonly on the basis of the uncertainty management
perspective, high interrelationships among the variables are expectable. An effective
statistical method to examine underlying structures of variable interrelationships is
exploratory factor analysis, which shows variable interrelationships with as few as
possible factors (Raykov & Marcoulides, 2008). An exploratory factor analysis of the
seven measures of teachers‟ uncertainty in teaching was conducted after
WEIGHT_ACROSS was applied to the dataset. As reported in Table 3-1, two factors
with an eigenvalue over 1.0 were derived from the seven variables.24 All variables

questionnaire, and that will be tested as part of TIMSS 2007 in your school.”
23. An exploratory factor analysis of these four items was conducted with the combined
sample of all countries for this study after WEIGHT_ACROSS was applied to the dataset. The
result showed that all items were loaded on a single factor (eigenvalue = 2.63) and all items had
factor loadings exceeding 0.70 (factor loadings for the four items, in the order shown in the
preceding paragraph, were 0.71, 0.87, 0.86, and 0.79), indicating a high degree of consistency
among the items. Cronbach’s α was 0.83.
24. The “promax” method, an oblique rotation, was used in Table 3-1, because “in many
social and behavioral research studies, . . . underlying latent variables typically tend to be related”
49

expect TCHUNDS were loaded heavily on factor 1 (eigenvalue = 3.38), while
TCHUNDS alone was heavily loaded on factor 2 (eigenvalue = 1.03).
Recall that only TCHUNDS is a measure of the uncertainty that can stem from
instruction (i.e., the core technology of teaching), whereas the other measures,
ACHIEVE, ACHIVAR, EDCAPIT, EDCAVAR, LANGMIN, and SHORTRES, commonly
capture the uncertainty that may arise from the classroom context. The result from the
factor analysis reported in Table 3-1 also suggests the possibility that teachers‟
uncertainty in teaching may come from at least two conceptually separable sources: first,
the instructional complexity involved in the implementation of teaching for
understanding; and second, the difficulty in effectively dealing with various issues and
challenges arising from a range of classroom-contextual factors. In other words, the
result in Table 3-1 indicates there may be two different kinds of uncertainty in teaching,
which can be called uncertainty from teaching for understanding (UNC_TCHUNDS) and
uncertainty from classroom-contextual factors (UNC_CONTEXT).25 Consequently, I use

(Raykov & Marcoulides, 2008, p. 267) rather than they are uncorrelated. Indeed, since all of the
seven variables in Table 3-1 have been constructed in this study to measure teachers’ exposure to
uncertainty in teaching, one may plausibly expect some correlation between the factors derived
from these variables. However, the result in Table 3-1 shows that the correlation between the two
factors is fairly small (r = 0.015) even when the promax method is used to allow the factors to be
correlated. In addition, using an orthogonal rotation based on the “varimax” method produces
almost the same result, as reported in Table A1 in the Appendix.
25. Since TCHUNDS alone is heavily loaded on factor 2 in the factor analysis reported in
Table 3-1, this factor (UNC_TCHUNDS) can be seen as basically the same variable with
TCHUNDS, with the correlation between the two being over 0.87. Thus, the factor label
uncertainty from teaching for understanding does not mean that UNC_TCHUNDS captures
instructional uncertainty, while TCHUNDS, or teaching for understanding in its full name, does
not. Rather, both UNC_TCHUNDS and TCHUNDS commonly reflect the degree to which a
50

these two component factors as the measures of uncertainty in teaching. These factors
are standardized composite variables with a mean of zero and a standard deviation of
one.
In addition to these level-1 primary variables, the following variable is used as a
level-2 primary variable in order to test hypothesis 3 derived from the uncertainty
management perspective:
Socio-cultural heterogeneity (SOCHET): the country-level ethnic fractionalization
index calculated by Alesina, Devleeschauwer, Easterly, Kurlat, and Wacziarg (2003).
This index ranges theoretically from zero for no fractionalization to one for perfect
fractionalization (multiplied by 10 to rescale the index to range between zero and 10). I
use this index as a proxy for the degree to which a country is heterogeneous in terms of
social and cultural values held by people in the country.

3.3.2. Additional Independent Variables
Pertaining to hypotheses A1 and A2, the following two additional independent
variables are used at level 1:
School climate (CLIMATE): a continuous variable that measures the

teacher exerts effort to foster students’ engagement in inquiry-based learning and non-routine
problem solving, which inevitably makes the teacher exposed to greater instructional uncertainty
as her/his teaching becomes less reducible to predictable routines. In other words, teaching for
understanding is assumed to accompany greater instructional uncertainty than conventional
teaching (see chapter 2, especially sections 2.1.5 and 2.2.1). The only difference worth
mentioning between UNC_TCHUNDS and TCHUNDS is that UNC_TCHUNDS is clearly
distinct from UNC_CONTEXT (r = -0.015, statistically insignificant), while TCHUNDS is
slightly correlated with UNC_CONTEXT (r = 0.131, p ≤ .001).
51

organizational climate of the school where an eighth-grade mathematics teacher works.
A question in TIMSS 2007 asked teachers, “How would you characterize each of the
following within your school?” The following four items were included regarding the
school climate: “teachers‟ job satisfaction,” “teachers‟ understanding of the school‟s
curricular goals,” “teachers‟ degree of success in implementing the school‟s curriculum,”
and “teachers‟ expectations for student achievement.” For each of these items, the
response options were “very low,” “low,” “medium,” “high,” and “very high.” These
response options were given numeric values of one, two, three, four, and five,
respectively. The mean of the four items is used as the variable.26
Principal’s instructional leadership (PRINLEAD): a continuous variable that
captures the degree to which an eighth-grade mathematics teacher‟s school principal
spends time on instructionally relevant activities. A question in TIMSS 2007 asked the
principal of each sampled school, “By the end of this school year, approximately what
percentage of time in your role as principal will you have spent on these activities?” The
following three pertained to instructionally relevant activities: “developing curriculum
and pedagogy,” “supervising and evaluating teachers and other staff,” and “teaching.”
The sum of the percentages allocated to these activities is used as the variable (divided
by 10 to make 10% as a one-unit).
26. An exploratory factor analysis of these four items was conducted with the combined
sample of all countries for this study after WEIGHT_ACROSS was applied to the dataset. The
result showed that all items were loaded on a single factor (eigenvalue = 2.27) and all items had
factor loadings exceeding 0.70 (factor loadings for the four items, in the order shown in the
preceding paragraph, were 0.73, 0.78, 0.82, and 0.67), indicating a high degree of consistency
among the items. Cronbach’s α was 0.74.
52

In addition, with respect to hypotheses A3, A4, and A5, the following two
additional independent variables are used at level 2:
Curriculum standardization (CURRSTD): a country-level index of curriculum
standardization based on data from the TIMSS 2007 Mathematics Curriculum
Questionnaire.27 This variable ranges from one for low standardization to six for high
standardization.28

27. For the value for each country, see Table 4-1 in chapter 4. One = The country does
not have “a national curriculum that covers mathematics instruction at the eighth grade of formal
schooling” and does not have either prescribed curricular “materials” or “mandated or
recommended textbooks.” Two = The country does not have “a national curriculum that covers
mathematics instruction at the eighth grade of formal schooling” but does have either prescribed
curricular “materials” or “mandated or recommended textbooks.” Three = The country does not
have “a national curriculum that covers mathematics instruction at the eighth grade of formal
schooling” but does have both prescribed curricular “materials” and “mandated or recommended
textbooks.” Four = The country has “a national curriculum that covers mathematics instruction at
the eighth grade of formal schooling” but does not have either prescribed curricular “materials”
or “mandated or recommended textbooks.” Five = The country has “a national curriculum that
covers mathematics instruction at the eighth grade of formal schooling” and also has either
prescribed curricular “materials” or “mandated or recommended textbooks.” Six = The country
has “a national curriculum that covers mathematics instruction at the eighth grade of formal
schooling” and also has both prescribed curricular “materials” and “mandated or recommended
textbooks.”
28. For example, the value for the United States is two, indicating a fairly low degree of
curricular standardization. According to the TIMSS 2007 Mathematics Curriculum
Questionnaire, the United States does not have “a national curriculum that covers mathematics
instruction at the eighth grade of formal schooling,” but the country does have “mandated or
recommended textbooks” in most states, in the sense that state education agencies and local
school districts “publish a list of preferred/recommended texts” in alignment with their own
curriculum standards. One might argue that the United States has a fairly standardized
curriculum within each state; however, Schmidt and his colleagues note that “even with the
presence of well-defined state standards and, more recently, the increasing presence of
corresponding state assessments, local districts still maintain de facto control of their curriculum”
(Schmidt, Cogan, Houang, & McKnight, 2011, p. 400). Unfortunately, the lack of state-level and
district-level information in the TIMSS 2007 dataset did not allow the construction of a more
53

Individualist culture (INDVCUL): the individualism index by Hofstede and
Hofstede (2005). This country-level index captures the extent to which people in a given
society tend to give emphasis on goals and interests of individuals rather than those of
groups (divided by 10 to rescale the index to range between zero and 10).29

3.3.3. Control Variables
In addition to the primary and additional independent variables described above,
the following four variables are also entered into the regression equation at level 1, as
these variables might influence teacher behavior in schools:
Teacher workload (WORKLOAD): a continuous measure of school-level teacher
shortage. This variable is used in this study as a proxy for individual teachers‟
workload in a given school.30 A question in TIMSS 2007 asked principals how difficult
it was “to fill eighth-grade teaching vacancies for this school year” for the subject area
of “mathematics.” The response options were “no vacancies,” “easy to fill vacancies,”
“somewhat difficult,” and “very difficult.” These response options were given numeric
values of zero, one, two, and three, respectively.
accurate measure of curricular standardization for this study.
29. This index is frequently used in the international management literature, but it has
been rarely used in the education literature. One recent exception is Chiu, Chow, and McbrideChang’s (2007) study, which found that compared to students in individualist societies,
achievement scores of students in collectivist cultures were linked more strongly to schoolmates’
use of meta-cognitive strategies.
30. This is based on the logic that individual teachers’ workload in a given school
becomes greater as the school’s teacher shortage level becomes higher, assuming that the amount
of total work in that school remains constant. Unfortunately, there is no direct measure of teacher
workload in the TIMSS 2007 dataset.
54

Teacher experience (TCHREXP): a continuous variable of how many years an
eighth-grade mathematics teacher has been in the profession of teaching according to
TIMSS 2007 (divided by 5 to make 5 years as a one-unit).
Mathematics major (MATHMJR): a binary variable indicating whether or not an
eighth-grade mathematics teacher majored in mathematics or mathematics education
during her/his post-secondary education. If an eighth-grade mathematics teacher
majored in either mathematics or mathematics education according to TIMSS 2007, the
teacher was coded one; otherwise, the teacher was coded zero.
Female teacher (FEMALE): a binary variable indicating whether an eighth-grade
mathematics teacher is female or male according to TIMSS 2007. Female teachers were
coded one, and male teachers were coded zero.
In addition to these level-1 control variables, the following two control variables
are also considered at level 2, as they have been widely used as basic indicators of
country-level characteristics that might affect educational policy and practice:
Public expenditure on education (EDEXPN): the amount of public expenditure on
education as a percentage of gross domestic product (2007). Public expenditure on
education consists of local, regional, and national governments‟ expenditure on
education plus subsidies to private education. This variable is available from the World
Bank‟s World Development Indicators dataset.
Economic development (GDPCAP): the gross domestic product per capita (2007)
adjusted by purchasing power parity (divided by 5,000 to make $5,000 as a one-unit).
This variable comes from the World Development Indicators dataset compiled by the
55

Table 3-2. Summary of variable definitions and data sources
Definition
Dependent variables
TCHRCOL_LESSON:
Collaborative lesson planning
TCHRCOL_CLASS:
Collaborative classroom
observation
TCHRCOL_VARIETY: Variety
in collaborative teacher
interaction

Level-1 variables
UNC_TCHUNDS: Uncertainty
from teaching for understanding

UNC_CONTEXT: Uncertainty
from classroom-contextual
factors

TCHUNDS: Teaching for
understanding

ACHIEVE: Student
achievement level
ACHIVAR: Variation in student
achievement

Source

The degree to which an eighth-grade
mathematics teacher regularly participates
in collaborative lesson planning.
The degree to which an eighth-grade
mathematics teacher regularly participates
in collaborative classroom observation.
The extent to which an eighth-grade
mathematics teacher regularly engages in
multiple types of collaborative interaction
with other teachers.

TTE

The factor score for the degree to which
an eighth-grade mathematics teacher
confronts uncertainties due to her/his effort
to use instructional strategies to foster
students’ engagement in inquiry-based
learning and non-routine problem solving.
The factor score for the degree to which
an eighth-grade mathematics teacher
confronts uncertainties due to various
complex issues arising from a range of
classroom-contextual conditions and
constraints.
The degree to which an eighth-grade
mathematics teacher uses instructional
strategies to foster students’ engagement in
inquiry-based learning and non-routine
problem solving.
The overall mathematics achievement
level of the classroom that an eighth-grade
mathematics teacher teaches.
The variation in mathematics achievement
in the classroom that an eighth-grade
mathematics teacher teaches.

T31

56

TTE

TTE

T31

TTE

TST

TST

Table 3-2. (cont'd)
Definition
Level-1 variables, cont'd
EDCAPIT: Students’
educational capital
EDCAVAR: Variation in
educational capital
LANGMIN: Proportion of
language minority students

SHORTRES: Shortage of
classroom resources

CLIMATE: School climate

PRINLEAD: Principal’s
instructional leadership

WORKLOAD: Teacher
workload
TCHREXP: Teacher
experience
MATHMJR: Mathematics
major

FEMALE: Female teacher

The overall socioeconomic background of
the classroom that an eighth-grade
mathematics teacher teaches.
The variation in students’ socioeconomic
backgrounds in the classroom that an
eighth-grade mathematics teacher teaches.
The degree to which the classroom that an
eighth-grade mathematics teacher teaches
is populated by language minority
students.
The degree to which an eighth-grade
mathematics teacher perceives that her/his
teaching is negatively affected by a
shortage or inadequacy of classroom
resources.
Teachers’ perception of the degree to
which teacher morale is high and
organizational goals are clearly
communicated and well implemented.
The degree to which an eighth-grade
mathematics teacher’s school principal
spends time on instructionally relevant
activities.
The degree of school-level teacher
shortage as a proxy for individual
teachers’ workload in a given school.
How long an eighth-grade mathematics
teacher has been in the profession of
teaching.
Whether or not an eighth-grade
mathematics teacher majored in
mathematics or mathematics education
during her/his post-secondary education.
Whether an eighth-grade mathematics
teacher is female or male.

57

Source
TST

TST

TST

TTE

TTE

TSC

TSC

TTE

TTE

TTE

Table 3-2. (cont'd)
Definition
Level-2 variables
SOCHET: Socio-cultural
heterogeneity

CURRSTD: Curriculum
standardization
INDVCUL: Individualist
culture

EDEXPN: Public expenditure
on education
GDPCAP: Economic
development

The country-level ethnic fractionalization
index as a proxy for the degree to which a
country is heterogeneous in terms of social
and cultural values held by people in the
country.
The degree to which a country is
characterized as having a standardized
curriculum for eighth-grade mathematics.
The degree to which people in a given
country tend to give emphasis on goals and
interests of individuals rather than those of
groups.
The amount of public expenditure on
education as compared to gross domestic
product.
The gross domestic product per capita
adjusted by purchasing power parity.

U.S.-specific variables
Course taught

Source
ALE

TCU

HOF

WDI

WDI

The mathematics course that an eighthUST
grade mathematics teacher teaches to a
given class of students.
School type
Whether a school is a public school, a
UST
religiously affiliated private school, or an
independent or unspecified private school.
Free or reduced-price lunch
The proportion of students at a given
UST
school who are eligible to receive free or
reduced-price lunches through the
National School Lunch Program.
Note . ALE = Alesina et al., 2003. HOF = Hofstede & Hofstede, 2005. T31 = Table 3-1. TCU =
TIMSS Curriculum Questionnaire 2007. TSC = TIMSS School Questionnaire 2007. TST =
TIMSS Student Questionnaire 2007. TTE = TIMSS Teacher Questionnaire 2007. UST = U.S.
National TIMSS 2007. WDI = World Development Indicators 2007.

58

World Bank.

3.3.4. U.S.-Specific Contextual Variables
Three U.S.-specific variables are from the supplementary U.S. national TIMSS
dataset prepared by the National Center for Educational Statistics. This dataset is
designed to be linkable to the TIMSS 2007 international dataset compiled by the
International Association for the Evaluation of Educational Achievement. The variables
are as follows:
Course taught: a categorical variable indicating which mathematics course an
eighth-grade mathematics teacher teaches to a given class of students. The courses were
categorized into basic eighth-grade mathematics, algebra, geometry, and other courses.
This variable is used in light of the possibility that there may be significant differences
in collaborative interaction among teachers depending on courses they teach.
School type: a categorical variable indicating whether a school is a public school, a
religiously affiliated private school, or an independent or unspecified private school.
This variable is used because many people frequently attribute differences in various
aspects of school culture to school types.
Proportion of free or reduced-price lunch recipients: the percentage of students at a
given school who are eligible to receive free or reduced-price lunches through the
National School Lunch Program.31 This is a widely used measure of school-level

31. Since the National School Lunch Program is a federally assisted meal program
operating in public and nonprofit private schools, data for this variable are available only for
59

poverty (divided by 10 to make 10% as a one-unit).

3.4. Descriptive Statistics

The definitions and data sources of all dependent and independent variables
used in this study are summarized in Table 3-2. This study uses multilevel data from 29
countries for which all level-1 and level-2 variables are available.32 These countries are
from a variety of geographic regions of the world and represent a range of cultural and
economic contexts.33 Table 3-3 represents the descriptive statistics for all variables in
the combined sample of 29 countries to which WEIGHT_ACROSS has been applied. In
addition, Table 3-4 shows the descriptive statistics for the WEIGHT_WITHIN-applied
U.S. data. Recall that both weight variables are proportional weight variables without
any scaling effect (i.e., they do not artificially inflate the size of the sample).

3.5. Analytic Approaches

This study first examines the overall cross-national patterns in collaborative

these types of schools.
32. The 29 countries are as follows: Australia, Bulgaria, Chinese Taipei, Colombia, the
Czech Republic, Egypt, England, Ghana, Hong Kong, Hungary, Indonesia, Iran, Israel, Italy,
Japan, Kuwait, Lebanon, Malaysia, Norway, Romania, the Russian Federation, Saudi Arabia,
Scotland, Singapore, South Korea, Sweden, Thailand, Turkey, and the United States.
33. See Table 4-1 in chapter 4 for national characteristics of these countries (i.e., values
of level-2 variables for each country).
60

Table 3-3. Descriptive statistics for variables
Mean

(SD )

Min.

Max.

n

1.388
.486

(.765)
(.641)

.000
.000

3.000
3.000

6785
6747

.059
.476
.464

…
…
…

.000
.000
.000

1.000
1.000
1.000

6743
6743
6743

Level-1 independent variables
UNC_TCHUNDS
UNC_CONTEXT
TCHUNDS
ACHIEVE
ACHIVAR
EDCAPIT
EDCAVAR
LANGMIN
SHORTRES
CLIMATE
PRINLEAD
WORKLOAD
TCHREXP
MATHMJR
FEMALE

.000
.000
1.686
4.672
1.442
6.619
2.403
1.972
.921
3.653
5.027
.827
3.077
.822
.559

(1.000)
(1.000)
(.582)
(.936)
(.565)
(1.662)
(1.039)
(2.889)
(.806)
(.596)
(1.584)
(.952)
(2.178)
…
…

-3.306
-2.548
.000
1.897
.000
1.081
.000
.000
.000
1.000
.000
.000
.000
.000
.000

2.886
3.851
3.000
7.451
4.562
10.000
9.321
10.000
3.000
5.000
10.000
3.000
10.000
1.000
1.000

6624
6624
6725
7080
7080
7079
7079
7079
6671
6791
6655
6793
6538
6689
6795

Level-2 independent variables
SOCHET
CURRSTD
INDVCUL
EDEXPN
GDPCAP

3.083
4.621
4.534
4.569
4.984

(2.316)
(1.474)
(2.577)
(1.155)
(3.109)

.020
1.000
1.300
2.594
.273

7.351
6.000
9.100
6.745
10.981

29
29
29
29
29

Dependent variables
TCHRCOL_LESSON
TCHRCOL_CLASS
TCHRCOL_VARIETY
None
Low variety
High variety

61

Table 3-4. Descriptive statistics for variables used in within-U.S. analyses
Mean
Dependent variables
TCHRCOL_LESSON
TCHRCOL_CLASS
TCHRCOL_VARIETY
None
Low variety
High variety
Independent variables
UNC_TCHUNDS
UNC_CONTEXT
TCHUNDS
ACHIEVE
ACHIVAR
EDCAPIT
EDCAVAR
LANGMIN
SHORTRES
CLIMATE
PRINLEAD
WORKLOAD
TCHREXP
MATHMJR
FEMALE
U.S.-specific independent variables
Course taught
Basic 8th-grade math
Algebra
Geometry
Others
School type
Public school
Private, religious
Private, others
Free or reduced-price lunch

(SD )

Min.

Max.

n

1.236
.269

(.901)
(.508)

.000
.000

3.000
3.000

477
477

.163
.576
.261

…
…
…

.000
.000
.000

1.000
1.000
1.000

477
477
477

.057
-.725
1.681
5.086
.918
7.567
1.993
.845
.390
3.882
4.700
.984
2.812
.698
.703

(.849)
(.588)
(.574)
(.627)
(.305)
(1.062)
(.874)
(1.335)
(.510)
(.647)
(1.774)
(1.067)
(2.097)
…
…

-2.341
-2.548
.250
2.593
.000
4.000
.000
.000
.000
1.500
.500
.000
.000
.000
.000

2.205
.809
3.000
7.178
1.945
10.000
5.533
7.500
2.500
5.000
8.500
3.000
9.000
1.000
1.000

482
482
484
532
532
531
531
530
485
484
437
464
480
481
487

.200
.739
.016
.045

…
…
…
…

.000
.000
.000
.000

1.000
1.000
1.000
1.000

532
532
532
532

.884
.076
.040
4.289

…
…
…
(2.813)

.000
.000
.000
.000

1.000
1.000
1.000
10.000

520
520
520
486

62

teacher interaction by comparing the aggregated country-level data across the 29
countries in the sample. These patterns are, then, analyzed vis-à-vis variations in level-2
variables. This preliminary analysis will provide a broad picture of the relationships
between a range of country-level contextual variables and collaborative teacher
interaction. Next, a series of multilevel modeling analyses is conduced (Raudenbush &
Bryk, 2002), whereby both level-1 and level-2 variables are simultaneously considered
to explain collaborative interaction among teachers. Using this analytic approach is
aimed at providing a more comprehensive and accurate understanding of how teacher
collegiality is associated with various factors at multiple levels of social context.
Considering the hierarchical structure of the data where teachers (i.e., the level-1 unit)
are clustered within countries (i.e., the level-2 unit), using multilevel modeling is an
effective approach, whereby differential variance between countries is taken into
account to yield improved estimates of regression coefficients and standard errors.
Finally, I also conduct a set of similar analyses using only the U.S. portion of the TIMSS
2007 dataset. In these analyses, I examine collaborative teacher interaction as a function
of level-1 variables to examine the case of the United States, where teachers‟
engagement in meaningful collaborative work around teaching has been reported to be
the exception rather than the norm (Bryk & Schneider, 2002; Fullan & Hargreaves, 1996;
Lortie, 2002; Rosenholtz, 1989).34

34. Conducting within-U.S. analyses is also beneficial for another important reason. The
cross-national multilevel analyses of teacher collegiality in this study are based on the
assumption that the meaning of collaborative teacher interaction described in the TIMSS 2007
Teacher Questionnaire is reasonably constant across countries in the sample. However, there are
63

varying perspectives on this assumption. For example, while some sociological perspectives
highlight the embeddedness of education in world-cultural models (Baker & LeTendre, 2005;
LeTendre, et al., 2001; Meyer & Ramirez, 2003), many studies suggest that educational practices
are often fundamentally local cultural activities (Alexander, 2000; Anderson-Levitt, 2003;
Britton, Paine, Pimm, & Raizen, 2003). Although the collaborative teacher interaction types
included in the TIMSS 2007 Teacher Questionnaire seem to be relatively low-inference generic
descriptions, conducting within-U.S. analyses will provide opportunities to analyze the data
without the need for assuming the validity of this assumption.
64

CHAPTER 4
DESCRIPTIVE AND PRELIMINARY ANALYSES

4.1. Analyses of Aggregated Country-Level Data

Figure 4-1 compares across countries eighth-grade mathematics teachers‟
engagement in collaborative lesson planning, i.e., the country-means of
TCHRCOL_LESSON. Recall that TCHRCOL_LESSON is an index that captures the
degree to which an eighth-grade mathematics teacher engages in collaborative lesson
planning, ranging from zero for no engagement to three for the highest level of
engagement.35 As shown in Figure 4-1, the country-mean of TCHRCOL_LESSON
varies considerably, ranging from 0.96 in Czech Republic to 2.03 in Kuwait. The United
States is 1.24, ranked 22nd among the 29 countries in the sample. Similarly, Figure 4-2
shows a cross-national comparison of eighth-grade mathematics teachers‟ engagement
in collaborative classroom observation. That is, this figure reports the country-means of
TCHRCOL_CLASS, an index that captures the degree to which an eighth-grade
mathematics teacher regularly participates in collaborative classroom observation,
which ranges from zero for no participation to three for the highest level of

35. See chapter 3 (section 3.2) for more details about how TCHRCOL variables (i.e.,
TCHRCOL_LESSON, TCHRCOL_CLASS, and TCHRCOL_VARIETY) were constructed for
measures of teacher collegiality. See Table 3-3 in chapter 3 for the descriptive statistics for these
variables.
65

2.25

2.00

1.75

1.50

1.25

1.00

.75

.50

Figure 4-1. Eighth-grade mathematics teachers’ engagement in collaborative lesson planning, by country (i.e., country-mean of
TCHRCOL_LESSON with the 95% confidence interval). Note . The index theoretically ranges from 0.0 for no engagement to 3.0 for
the highest level of engagement.

66

1.75

1.50

1.25

1.00

.75

.50

.25

.00

Figure 4-2. Eighth-grade mathematics teachers’ engagement in collaborative classroom observation, by country (i.e., country-mean of
TCHRCOL_CLASS with the 95% confidence interval). Note . The index theoretically ranges from 0.0 for no engagement to 3.0 for the
highest level of engagement.

67

100%
90%

80%
70%
60%
50%
40%

30%
20%
10%
0%

High variety

Low variety

None

Figure 4-3. Variety in collaborative interaction among eighth-grade mathematics teachers, by country (i.e., country-level breakdown of
TCHRCOL_VARIETY).

68

participation. As shown in this figure, there is considerable cross-national variation in
country-means of TCHRCOL_CLASS, ranging from 0.13 in Czech Republic to 1.42 in
Ghana. The United States is 0.27, ranked 21st among the 29 countries in the sample.
Figure 4-3 provides an alternative description of cross-national variation in
collaborative interaction among eighth-grade mathematics teachers on the basis of
TCHRCOL_VARIETY, an ordinal variable capturing how various types of collaborative
interaction are used by the teacher, regardless of how frequently she/he engages in
collaborative interaction with other teachers. Teachers are categorized into three
different groups, depending on whether they engage in both collaborative lesson
planning and collaborative classroom observation (i.e., high variety), only collaborative
lesson planning or only classroom observation (i.e., low variety), or neither of these (i.e.,
none). As shown in Figure 4-3, there is substantial variation in the country-level
breakdown of TCHRCOL_VARIETY. For example, 88.3% of eighth-grade mathematics
teachers belong to the “high variety” category in the Russian Federation, while only
13.9% belong to the same category in Czech Republic. The United States is ranked 23rd
among the 29 countries in the sample in terms of the percentage of eighth-grade
mathematics teachers who belong to the “high variety” category, with the percentage
being 26.0%. In terms of the percentage who fall into the “none” category, the United
States is 16.4%, which is second only to 18.5% in Italy, meaning that 16.4% of U.S.
eighth-grade mathematics teachers never engage in either collaborative lesson planning
or collaborative classroom observation. In sum, compared to other countries, the United
States is characterized as having a very high proportion of eighth-grade mathematics
69

Table 4-1. Level-2 variables: Values by country
SOCHET CURRSTD
INDVCUL
EDEXPN
Australia
.929
1
9.0
4.651
Bulgaria
4.021
4
3.0
4.135
Chinese Taipei
2.744
6
1.7
4.350
Colombia
6.014
1
1.3
4.052
Czech Rep.
3.222
4
5.8
4.607
Egypt
1.836
6
3.8
3.685
England
1.211
4
8.9
5.559
Ghana
6.733
6
2.0
5.437
Hong Kong
.620
6
2.5
3.520
Hungary
1.522
5
8.0
5.417
Indonesia
7.351
3
1.4
3.534
Iran
6.684
3
4.1
5.492
Israel
3.436
5
5.4
6.359
Italy
1.145
4
7.6
4.317
Japan
.119
6
4.6
3.449
Kuwait
6.604
5
3.8
3.760
Lebanon
1.314
6
3.8
2.594
Malaysia
5.880
5
2.6
4.533
Norway
.586
4
6.9
6.745
Romania
3.069
6
3.0
4.365
Russian Fed.
2.452
5
3.9
3.872
Saudi Arabia
1.800
6
3.8
6.416
Scotland
1.211
5
8.9
5.559
Singapore
3.857
6
2.0
2.795
South Korea
.020
5
1.8
4.211
Sweden
.600
4
7.1
6.673
Thailand
6.338
5
2.0
4.004
Turkey
3.200
6
3.7
2.863
United States
4.901
2
9.1
5.535
Note . For correlations between the variables, see Table A2 in the Appendix.

70

GDPCAP
7.271
2.308
6.194
1.708
4.907
1.011
7.156
.273
8.487
3.852
.748
2.197
5.306
6.350
6.707
9.673
2.145
2.709
10.981
2.533
3.362
4.525
7.156
10.565
5.237
7.686
1.558
2.633
9.292

teachers who do not engage in any type of collaborative teacher interaction while
having a very low proportion of eighth-grade mathematics teachers who regularly
engage in multiple types of collaborative teacher interaction.
In order to examine the relationship between national contexts and crossnational variations illustrated in Figures 4-1, 4-2, and 4-3, a range of country-level
variables were used. They are socio-cultural heterogeneity (SOCHET), curriculum
standardization (CURRSTD), individualist culture (INDVCUL), public expenditure on
education (EDEXPN), and economic development (GDPCAP). The values for these
variables are presented in Table 4-1.36 Table 4-2 shows bivariate correlations between
each of these variables and the country-means of TCHRCOL_LESSON,
TCHRCOL_CLASS, and TCHRCOL_VARIETY.37 Notable is that both
TCHRCOL_LESSON and TCHRCOL_CLASS, when aggregated to country-level means,
are significantly positively correlated with SOCHET, which is consistent with
hypothesis 3 from the uncertainty management perspective that teachers in countries
with a higher level of socio-cultural heterogeneity are more likely to engage in
collaborative interaction with other teachers.
In addition, both TCHRCOL_CLASS and TCHRCOL_VARIETY, when
aggregated to country-level means, are significantly positively correlated with

36. See chapter 3 (section 3.3) for detailed descriptions of these variables. For
correlations between these variables, see Table A2 in the Appendix.
37. TCHRCOL_VARIETY, an ordinal-scale variable coded two, one, or zero for “high
variety,” “low variety,” and “none,” respectively, was treated as an interval-scale variable when
aggregated to the country level in Table 4-2.
71

Table 4-2. Pearson's bivariate correlations
between country-means of TCHRCOL and level-2
variables (n = 29)

SOCHET
CURRSTD
INDVCUL
EDEXPN
GDPCAP

Country mean of
TCHRCOL_LESSON
r
.403 *
.049
-.273
-.320 +
-.317 +

SOCHET
CURRSTD
INDVCUL
EDEXPN
GDPCAP

Country mean of
TCHRCOL_CLASS
r
.315
.352
-.319
-.135
-.390

SOCHET
CURRSTD
INDVCUL
EDEXPN
GDPCAP
+p ≤ .10; *p ≤ .05.

+
+
+
*

Country mean of
TCHRCOL_VARIETY
r
.284
.415 *
-.373 *
-.249
-.373 *

72

Table 4-3. Ordinary least squares regressions for country-means of TCHRCOL, with all level-2 variables entered
into the equation (n = 29)
Y=
Y=
Y=
Country mean
Country mean
Country mean
of TCHRCOL_
of TCHRCOL_
of TCHRCOL_
LESSON
CLASS
VARIETY
B
β
B
β
B
β
SOCHET
.043 +
.423
.050 +
.381
.038
.349
(.024)
(.029)
(.024)
CURRSTD
.021
.133
.098 *
.477
.087 *
.502
(.035)
(.042)
(.035)
INDVCUL
.019
.203
.017
.147
.012
.125
(.026)
(.031)
(.026)
EDEXPN
-.063
-.304
.004
.016
-.021
-.094
(.044)
(.053)
(.043)
GDPCAP
-.012
-.155
-.026
-.269
-.019
-.228
(.016)
(.019)
(.016)
Intercept
1.419 ***
-.089
1.017 **
(.321)
(.387)
(.319)
.27
.34
.37
R²
Note . B = Unstandardized coefficients with standard errors in parentheses. β = standardized coefficients.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

73

CURRSTD and significantly negatively correlated with INDVCUL, which are in line
with hypotheses A3 and A5 from the curriculum policy perspective and the national
culture perspective, respectively. That is, these results appear to substantiate the
possibility that teachers in countries with the presence of a more standardized national
curriculum, and in countries with a less individualist culture, are more likely to engage
in collaborative interaction with other teachers. Further, as shown in Table 4-3, the
significant positive association between socio-cultural heterogeneity and both
TCHRCOL_LESSON and TCHRCOL_CLASS aggregated to national-level means
remains to be statistically significant even when all other country-level contextual
variables are held constant. Table 4-3 also shows that the significant positive connection
between curriculum standardization and both national-level aggregated
TCHRCOL_CLASS and TCHRCOL_VARIETY persists even after all other country-level
contextual variables are controlled for.

4.2. Analyses of Multilevel Data

Although providing useful preliminary information, analyses of the aggregated
national-level data reported in Tables 4-2 and 4-3 ignore within-nation variation in each
country, which is problematic because within-nation variation may vary across
countries. For more accurate analyses, the multilevel nature of the data structure should
be taken into account. One possible statistical method in this regard is hierarchical
(generalized) linear modeling (Raudenbush & Bryk, 2002). Based on this method, the
74

unconditional model (i.e., the simplest possible hierarchical linear model) is as follows,
which is equivalent to a one-way analysis of variance with random effects.
At level 1, for teacher i in country j,

TCHRCOLij = β0j + rij,

where β0j is the mean outcome for country j, and rij is a random error which is assumed
to be normally distributed with a mean of zero and a variance, σ squared.
At level 2,

β0j = γ00 + u0j,

where each country-mean of TCHRCOL, β0j, is represented as a function of the grand
mean, γ00, plus a random error, u0j, which is assumed to be normally distributed with a
mean of zero and a variance, τ00.
Here, the intra-class correlation, ρ, which is the proportion of the variance in
TCHRCOL that exists at level 2 (i.e., between countries) rather than at level 1 is
calculated by:

2

ρ = τ00 / (τ00 + σ ),
where ρ can theoretically range from zero to one. A higher ρ coefficient indicates a
75

Table 4-4. Unconditional hierarchical (generalized) linear models for TCHRCOL

Fixed effect
Intercept, γ 00

Yij =
TCHRCOL_LESSON
Coeff.
(SE )
1.405 ***
(.051)

Yij =
TCHRCOL_CLASS
Coeff.
(SE )
.512 ***
(.059)

Threshold difference, δ (2)

Random effect
Intercept, u 0j
Level-1 random error, r ij

Var.
.061
.534

Intraclass correlation

ρ
.103

df
28

χ²
670.4 ***

Var.
.094
.323
ρ
.225

Note . OR = odds ratios.
***p ≤ .001.

76

df
χ²
28 1866.5 ***

Yij = ŋ (m)ij for
TCHRCOL_VARIETY
Coeff.
OR
(SE )
-.022
.979
(.193)
3.223 ***
25.115
(.130)
Var.
.929

ρ
.220

df
χ²
28 1274.4 ***

greater degree of importance of level-2 contextual information when explaining the
variance in the dependent variable.38
As reported in Table 4-4, 10.3% of the variance in TCHRCOL_LESSON, 22.5% of
the variance in TCHRCOL_CLASS, and 22.0% of the variance in TCHRCOL_VARIETY
lie between countries, indicating that collaborative classroom observation is more than
twice more contingent upon national contexts than is collaborative lesson planning. In
addition, Table 4-4 also shows that the average country-mean of TCHRCOL_LESSON is
1.405, which is almost three times greater than the average country mean of
TCHRCOL_CLASS, which is 0.512. That is, on average across the countries in the
sample, teachers engage in collaborative lesson planning to an almost three times
greater extent than in collaborative classroom observation.
To examine how significantly each country-level contextual variable is associated
with the three dependent variables, some additional preliminary analyses were
conducted using multilevel linear regressions where only one level-2 variable was
entered into the model for each time. That is, at level 1, for teacher i in country j,

TCHRCOLij = β0j + rij,

38. Since TCHRCOL_VARIETY is an ordinal-scale variable, the intra-class correlation
cannot be calculated based on the same unconditional model for TCHRCOL_LESSON and
TCHROL_CLASS. Thus, the intra-class correlation for TCHRCOL_VARIETY was calculated
based on the unconditional multilevel ordinal logistic regression model as explained by
O’Connell and her colleagues (O’Connell, Goldstein, Rogers, & Peng, 2008). For details about
the multilevel ordinal logistic regression used in this study, see chapter 5 (section 5.1).
77

and at level 2,

β0j = γ00 + γ01(level-2 variable)j + u0j,

where the level-2 variable is one of SOCHET, CURRSTD, INDVCUL, EDEXPN, or
GDPCAP for models I through V, respectively.
As reported in Tables 4-5 and 4-6, SOCHET at level 2 is significantly positively
related to both TCHRCOL_LESSON and TCHRCOL_CLASS at level 1, meaning that
collaborative teacher interaction, as a form of either collaborative lesson planning or
collaborative classroom observation, is significantly more likely in countries of higher
socio-cultural heterogeneity, which is consistent with hypothesis 3 derived from the
uncertainty management perspective. In addition, CURRSTD is significantly positively
associated with TCHRCOL_CLASS in line with hypothesis A3, suggesting that teachers
in countries with a more standardized national curriculum are more likely to engage in
collaborative classroom observation. INDVCUL has a significant negative relationship
with TCHRCOL_CLASS, indicating that teachers in countries with a more individualist
culture are less likely to engage in collaborative classroom observation, as suggested by
hypothesis A5.
With regard to TCHRCOL_VARIETY, hierarchical generalized linear modeling
was used based on the cumulative probability model for ordinal data (Raudenbush &
Bryk, 2002, pp. 317-325). The coefficients in Table 4-7 indicate the amount of increase in
the predicted ordered log odds of moving to the next higher level in the dependent
78

Table 4-5. Bivariate associations between TCHRCOL_LESSON and level-2 variables: Coefficients (γ 01, df =27) from
multilevel linear regressions
I
Coeff. (SE )
.049 (.019) **

SOCHET
CURRSTD
INDVCUL
EDEXPN
GDPCAP
Intercept, γ 00
1.391 (.041) ***
*p ≤ .05; **p ≤ .01; ***p ≤ .001.

II
Coeff. (SE )

III
Coeff. (SE )

IV
Coeff. (SE )

V
Coeff. (SE )

-.011 (.025)
-.026 (.016)
-.076 (.030) *
1.406 (.050) ***

1.391 (.047) ***

1.392 (.045) ***

-.023 (.021)
1.395 (.054) ***

Table 4-6. Bivariate associations between TCHRCOL_CLASS and level-2 variables: Coefficients (γ 01, df =27) from
multilevel linear regressions
I
Coeff. (SE )
.046 (.024) +

SOCHET
CURRSTD
INDVCUL
EDEXPN
GDPCAP
Intercept, γ 00
.499 (.051) ***
+p ≤ .10; **p ≤ .01; ***p ≤ .001.

II
Coeff. (SE )

III
Coeff. (SE )

IV
Coeff. (SE )

V
Coeff. (SE )

.062 (.032) +
-.037 (.019) +
-.008 (.048)
.508 (.056) ***

.493 (.051) ***

79

.511 (.063) **

-.035 (.022)
.498 (.054) ***

Table 4-7. Bivariate associations between TCHRCOL_VARIETY and level-2 variables: Coefficients (γ 01, df =27) from
multilevel ordinal logistic regressions

SOCHET
CURRSTD

I
Coeff. (SE )
[OR]
.133 (.067) +
[1.142]

II
Coeff. (SE )
[OR]

III
Coeff. (SE )
[OR]

IV
Coeff. (SE )
[OR]

V
Coeff. (SE )
[OR]

.233 (.077) **
[1.263]

INDVCUL

-.124 (.054) *
[.883]

EDEXPN

-.124 (.141)
[.883]

GDPCAP

-.081 (.070)
[.922]
-.042 (.191)
-.057 (.198)
3.228 (.129) *** 3.230 (.127) ***

Intercept, γ 00
-.059 (.177)
-.039 (.182)
TD, δ (2)
3.225 (.129) *** 3.228 (.127) ***
Note . OR = odds ratios. TD = threshold difference.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

-.087 (.181)
3.232 (.126) ***

80

variable TCHRCOL_VARIETY by a one-unit increase in an independent variable. Again,
consistent with hypothesis 3, SOCHET is significantly positively associated with
TCHRCOL_VARIETY. In addition, CURRSTD appears to have a significant positive
relationship with TCHRCOL_VARIETY, while INDVCUL has a significantly negative
relationship; these results are again in accordance with hypotheses A3 and A5,
respectively.

4.3. Discussion

A series of preliminary analyses presented in this chapter suggests that there is
considerable cross-national variation in eighth-grade mathematics teachers‟
engagement in collaborative teacher interaction. According to the unconditional
multilevel modeling of the data, about 10 to 23% of the variance in collaborative teacher
interaction appears to be attributable to national contextual factors, depending on how
to measure collaborative teacher interaction. The United States, when compared to
other countries, is characterized as having a very high proportion of eighth-grade
mathematics teachers who do not engage in any type of collaborative teacher
interaction while having a very low proportion of eighth-grade mathematics teachers
who regularly engages in multiple types of collaborative teacher interaction.
Very interesting is that, regardless of how to measure collaborative teacher
interaction, the results from the multilevel bivariate regressions indicate that teachers in
countries of higher socio-cultural heterogeneity are significantly more likely to engage
81

in collaborative teacher interaction. This finding gives credence to the uncertainty
management perspective in which collaborative teacher interaction is understood as a
collective response to uncertainty; it seems that teachers in a socio-culturally more
heterogeneous society are more likely to confront various complex issues in teaching
due to a lower level of value consensus at the societal level, which appears to
necessitate a higher level of engagement in collaborative interaction with their colleague
teachers.
Also notable in the results from multilevel bivariate regressions is that teachers
in countries with a more standardized national curriculum appear more likely to
engage in collaborative classroom observation; further, they are likely to participate in
more varied types of collaborative interaction. This result, supporting the curriculum
policy perspective, substantiates the hypothesis that the presence of a standardized
national curriculum may facilitate collaborative teacher interaction insofar as such a
curriculum provides teachers with some shared technical language for thinking about
teaching and learning. In addition, in line with the national culture perspective, teachers
in countries with a more individualist culture appear less likely to engage in
collaborative classroom observation; also, they are less likely to participate in multiple
types of collaborative interaction.
In the next chapter, another set of multilevel modeling analyses is conducted in
order to see if the significant effects found in the preliminary bivariate analyses in this
chapter persist even when a range of other variables are held constant. These further
analyses, although more complex in terms of statistical modeling, will provide more
82

accurate and comprehensive information about all three primary and five additional
hypotheses.

83

CHAPTER 5
MULTILEVEL MODELING ANALYSES

5.1. Model Specification

Building upon the extensive preliminary analyses conducted in the previous
chapter, a series of more elaborate and comprehensive analyses is carried out in this
chapter based on the multilevel modeling approach elaborated by Raudenbush and
Bryk (2002). These analyses aim at testing all three main and five additional hypotheses
simultaneously, with a range of other variables held constant.39
At level 1, for teacher i in country j, TCHRCOLij is a function of
UNC_TCHUNDSij and UNC_CONTEXTij as the primary independent variables and
CLIMATEij and PRINLEADij as additional independent variables; in addition,
WORKLOADij, TCHREXPij, MATHMJRij, and FEMALEij are also entered into the
equation as control variables. Thus, the level-1 model is specified as follows:

TCHRCOLij = β0j + β1j(UNC_TCHUNDS)ij + β2j(UNC_CONTEXT)ij
+ β3j(CLIMATE)ij + β4j(PRINLEAD)ij
+ β5j(WORKLOAD)ij + β6j(TCHREXP)ij
39. See chapter 2 (section 2.2) for all these hypotheses derived from different theoretical
perspectives.
84

+ β7j(MATHMJR)ij + β8j(FEMALE)ij + rij,

where β0j is the mean outcome for country j, and βaj for 1 ≤ a ≤ 8 is the slope for each
level-1 variable; rij is a random error.
At level 2, the intercept, β0j, is modeled as a function of SOCHETj, CURRSTDj,
INDVCULj, EDEXPNj; in addition, CURRSTDj × EDEXPNj is also entered into the
equation as an interaction term, given the possibility that a national curriculum is likely
to be implemented with greater impact on practice when the amount of public
educational funding is greater. Thus, the intercept is specified as follows:

β0j = γ00 + γ01(SOCHET)j + γ02(CURRSTD)j + γ03(INDVCUL)j
+ γ04(EDEXPN)j + γ05(CURRSTD × EDEXPN)j + u0j,

where γ00 is a constant; γ0b for 1 ≤ b ≤ 4 is the slope for each level-2 variable, and γ05 is
the slope for the interaction term, CURRSTDj × EDEXPNj. A random error, u0j, was
added to the model in light of the possibility that the mean of TCHRCOL may vary
randomly between countries due to some country-specific factors that are unique to
individual societies.
In addition, the slopes for UNC_TCHUNDSij and UNC_CONTEXTij, which are
β1j and β2j, are treated randomly varying as a function of CURRSTDj to test hypothesis
A4; EDEXPNj and CURRSTDj × EDEXPNj are also entered into the equation, given the
85

possibility that a national curriculum is likely to be implemented with greater impact on
practice when the amount of public educational funding is larger. The remaining β‟s,
with no cross-level interaction hypothesized, are treated simply as random. One might
want these slopes to be treated as fixed, but I permit them to randomly vary across
countries in light of the possibility of significant random differences in relationships; as
suggested by a large body of comparative education literature, there may be various
country-specific factors that are unique to individual societies‟ education systems and
cultures (Anderson-Levitt, 2003; Crossley & Watson, 2003; Schriewer, 2003).40 Thus, the
slopes are specified as follows:41

βpj = γp0 + γp1(CURRSTD)j + γp2(EDEXPN)j
+ γp3(CURRSTD × EDEXPN)j + upj
βqj = γq0 + uqj

for 1 ≤ p ≤ 2,

for 3 ≤ q ≤ 8,

All level-1 and level-2 independent variables are grand-mean-centered, which
makes the intercept, β0j, more interpretable (Raudenbush & Bryk, 2002). The intercept is
now defined as the expected TCHRCOL value when all independent variables are set
40. Since all slopes are permitted to randomly vary in the model, the resultant reduction
in the degrees of freedom at level 1 corresponding to the level-2 sample size results in reduced
statistical power (for a related discussion, see Bickel (2007), especially chapter 5). Thus, I report
results significant at p ≤ 0.10 in addition to p ≤ 0.05, p ≤ 0.01, and p ≤ 0.001.
41. Treating all slopes as fixed, however, does not change much the results reported in
this chapter. Compare Tables 5-1, 5-2, and 5-3 presented in this chapter with Tables A3, A4, and
A5 in the Appendix.
86

equal to their grand means, as expressed as (Xij – ̅ ..) for each level-1 variable and as
(Wj – ̅ .) for each level-2 variable. Thus, one may interpret the intercept as the expected
degree of collaborative teacher interaction as measured by TCHRCOL for a typical
teacher in a typical country in the sample.
This hierarchical linear model, however, is not appropriate when the dependent
variable is TCHRCOL_VARIETY because this is an ordinal-scale variable with only
three levels (i.e., “high variety,” “low variety,” and “none”), while TCHRCOL_LESSON
and TCHRCOL_CLASS are treated as interval-scale variables in this study. Thus,
multilevel ordinal logistic regression, as a form of hierarchical generalized linear
modeling, is used as an alternative statistical method as suggested by O‟Connell and
her colleagues (O‟Connell, et al., 2008) and Raudenbush and Bryk (2002). Specifically,
let

P(Rij = 1) = Prob(TCHRCOL_VARIETYij = “high variety”),
P(Rij = 2) = Prob(TCHRCOL_VARIETYij = “low variety”),
P(Rij = 3) = Prob(TCHRCOL_VARIETYij = “none”),

and let

P(Rij ≤ m) = P(Rij = 1) + . . . + P(Rij = m).

87

Then, the level-1 model is specified as

η(m)ij = ln[P(Rij ≤ m) / P(Rij > m)]
= β0j + β1j(UNC_TCHUNDS)ij + β2j(UNC_CONTEXT)ij
+ β3j(CLIMATE)ij + β4j(PRINLEAD)ij + β5j(WORKLOAD)ij
+ β6j(TCHREXP)ij + β7j(MATHMJR)ij + β8j(FEMALE)ij
+ D(2)ijδ(2),

where D(2)ij is a dummy variable indicating whether m = 2 (then, D(2)ij = 1) or not (i.e., if
m = 1, D(2)ij = 0), and δ(2) is the threshold difference which equals η(2)ij – η(2)ij.42
At level 2, the intercept, β0j, and the slopes, βkj for 0 ≤ k ≤ 8, remain the same as
specified earlier. That is,

β0j = γ00 + γ01(SOCHET)j + γ02(CURRSTD)j + γ03(INDVCUL)j

42. When the dependent variable has M categories, only M – 1 cumulative probabilities
are of interest because P(Rij ≤ M) = 1 in all studies. Since M = 3 in this study, P(Rij ≤ 3) = 1 is
redundant. Note that when the number of categories of the dependent variable is M and the
number of level-1 independent variables is N, the level-1 model can be generalized into the
following form:
η(m)ij = β0j + ∑

β(n)jX(n)ij + ∑

D(m)ijδ(m),

where D(m)ij is an indicator for the mth category, and δ(m) is the threshold difference which
equals η(m−1)ij – η(m)ij. See O’Connell, Goldstein, Rogers, and Peng (2008), and Raudenbush
and Bryk (2002) for more details and discussions.
88

+ γ04(EDEXPN)j + γ05(CURRSTD × EDEXPN)j + u0j,
βpj = γp0 + γp1(CURRSTD)j + γp2(EDEXPN)j
+ γp3(CURRSTD × EDEXPN)j + upj
βqj = γq0 + uqj

for 1 ≤ p ≤ 2,

for 3 ≤ q ≤ 8,

where, as in the earlier multilevel model for TCHRCOL_LESSON and
TCHRCOL_CLASS, all level-1 and level-2 independent variables are grand-meancentered.

5.2. Results

Tables 5-1 and 5-2 report the results from multilevel linear regressions for
TCHRCOL_LESSON and TCHRCOL_CLASS, respectively. In each table, model I
includes only those variables that pertain to the three primary hypotheses in this study,
while model II includes all variables as specified above. A very clear pattern is that
UNC_TCHUNDS is significantly positively associated with both TCHRCOL_LESSON
and TCHRCOL_CLASS, indicating that teachers who face greater instructional
uncertainty due to their effort to implement teaching for understanding are more likely
to engage in collaborative interaction with other teachers. This significant pattern
persists even when all additional hypotheses are simultaneously taken into account.
Since UNC_TCHUNDS is a standardized factor score with a mean of zero and a
89

Table 5-1. Multilevel linear regressions for TCHRCOL_LESSON
Fixed effect
Intercept, γ 00
SOCHET, γ 01
CURRSTD, γ 02
INDVCUL, γ 03
EDEXPN, γ 04
CURRSTD × EDEXPN, γ 05
UNC_TCHUNDS, γ 10
× CURRSTD, γ 11
× EDEXPN, γ 12
× CURRSTD × EDEXPN, γ 13
UNC_CONTEXT, γ 20
× CURRSTD, γ 21
× EDEXPN, γ 22
× CURRSTD × EDEXPN, γ 23
CLIMATE, γ 30
PRINLEAD, γ 40
WORKLOAD, γ 50
TCHREXP, γ 60
MATHMJR, γ 70
FEMALE, γ 80

I
Coeff. (SE )
1.417 (.045) ***
.066 (.019) **

.126 (.013) ***

-.009 (.026)

Coeff.
1.406
.054
.042
.034
-.116
.104
.103
.005
.016
-.018
.031
.007
.017
-.023
.165
.009
-.006
.004
.054
.063

II
(SE )
(.039)
(.020)
(.023)
(.022)
(.042)
(.029)
(.015)
(.011)
(.011)
(.009)
(.025)
(.016)
(.022)
(.016)
(.030)
(.008)
(.012)
(.008)
(.036)
(.039)

***
**
+
**
**
***

*

***

Random effect
Var.
df
χ²
Var.
df
χ²
Intercept, u 0j
.054
27
220.1 *** .038
23
98.5 ***
UNC_TCHUNDS, u 1j
.002
28
42.1 *
.004
25
44.2 **
UNC_CONTEXT, u 2j
.011
28
73.9 *** .013
25
60.4 ***
CLIMATE, u 3j
.013
28
65.0 ***
PRINLEAD, u 4j
.001
28
43.2 *
WORKLOAD, u 5j
.001
28
30.1
TCHREXP, u 6j
.001
28
82.7 ***
MATHMJR, u 7j
.009
28
32.7
FEMALE, u 8j
.028
28
83.5 ***
Level-1 random error, r ij
.519
.488
Note . All independent variables are grand-mean-centered. Coeff. = Unstandardized coefficients
with robust standard errors in parentheses.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

90

Table 5-2. Multilevel linear regressions for TCHRCOL_CLASS
Fixed effect
Intercept, γ 00
SOCHET, γ 01
CURRSTD, γ 02
INDVCUL, γ 03
EDEXPN, γ 04
CURRSTD × EDEXPN, γ 05
UNC_TCHUNDS, γ 10
× CURRSTD, γ 11
× EDEXPN, γ 12
× CURRSTD × EDEXPN, γ 13
UNC_CONTEXT, γ 20
× CURRSTD, γ 21
× EDEXPN, γ 22
× CURRSTD × EDEXPN, γ 23
CLIMATE, γ 30
PRINLEAD, γ 40
WORKLOAD, γ 50
TCHREXP, γ 60
MATHMJR, γ 70
FEMALE, γ 80

I
Coeff. (SE )
.487 (.055) ***
.026 (.026)

.068 (.010) ***

.077 (.019) ***

Coeff.
.478
-.007
.080
-.002
.043
.009
.058
-.013
-.014
-.007
.097
-.004
-.019
-.030
.086
.016
.006
.013
.027
.006

II
(SE )
(.048)
(.011)
(.021)
(.015)
(.027)
(.026)
(.011)
(.005)
(.008)
(.006)
(.016)
(.007)
(.017)
(.012)
(.016)
(.007)
(.013)
(.006)
(.038)
(.022)

***
***

***
*

***

*
***
*
*

Random effect
Var.
df
χ²
Var.
df
χ²
Intercept, u 0j
.090
27 664.4 *** .059
23 230.6 ***
UNC_TCHUNDS, u 1j
.001
28
45.8 *
.002
25
40.5 *
UNC_CONTEXT, u 2j
.005
28
76.0 *** .004
25
53.9 ***
CLIMATE, u 3j
.002
28
36.9
PRINLEAD, u 4j
.001
28
43.4 *
WORKLOAD, u 5j
.002
28
50.2 **
TCHREXP, u 6j
.001
28
60.8 ***
MATHMJR, u 7j
.022
28
63.2 ***
FEMALE, u 8j
.008
28
42.5 *
Level-1 random error, r ij
.312
.298
Note . All independent variables are grand-mean-centered. Coeff. = Unstandardized coefficients
with robust standard errors in parentheses.
*p ≤ .05; **p ≤ .01; ***p ≤ .001.

91

standard deviation of one, the multilevel regression coefficients can be interpreted as
the amount of expected increase in TCHRCOL when there is a one-standard-deviation
increase in UNC_TCHUNDS. After all other hypotheses are taken into account, a onestandard-deviation increase in UNC_TCHUNDS is significantly linked to an expected
increase in TCHRCOL_LESSON by 0.103, which is equivalent to 13.5% of one standard
deviation of TCHRCOL_LESSON because the standard deviation of
TCHRCOL_LESSON is 0.765 as reported in Table 3-3 in chapter 3. Similarly, a onestandard-deviation increase in UNC_TCHUNDS is significantly related to an expected
increase in TCHRCOL_CLASS by 0.058 with all other things in the model held constant,
which translates into 9.0% of one standard deviation of TCHRCOL_CLASS because the
standard deviation of TCHRCOL_CLASS is 0.641 as shown in Table 3-3 in chapter 3.
These results strongly support hypothesis 1 that teachers who exert greater effort to
implement teaching for understanding are significantly more likely to engage in
collaborative interaction with their colleague teachers.
With respect to the effect of UNC_CONTEXT, Table 5-2 reports that, with all
other things in the model held constant, a one-standard-deviation increase in this
variable is significantly associated with an expected increase in TCHRCOL_CLASS by
0.097, which represents 15.1% of one standard deviation of TCHRCOL_CLASS.
Although UNC_CONTEXT does not appear to have a significant effect on
TCHRCOL_LESSON as shown in Table 5-1, the significant positive association found
between UNC_CONTEXT and TCHRCOL_CLASS provides some evidence to
substantiate the relationship between teachers‟ exposure to complexities and difficulties
92

emerging from classroom contexts and their engagement in collaborative interaction
with their colleague teachers. This result partly supports hypothesis 2 that teachers who
face greater classroom-contextual uncertainty are more likely to engage in collaborative
interaction with other teachers.
Regarding the effect of national-level socio-cultural heterogeneity, it has a
significant positive effect on TCHRCOL_LESSON after all other things in the model are
controlled for as shown in Table 5-1. A one-unit increase in SOCHET at level 2 is
significantly related to an expected increase in TCHRCOL_LESSON at level 1 by 0.054.
Since the standard deviation of SOCHET is 2.316 as reported in Table 3-3 in chapter 3, it
is possible to say that a one-standard-deviation increase in SOCHET is related to an
increase in TCHRCOL_LESSON by 0.054 × 2.316 = 0.125, which translates into 16.3% of
one standard deviation of TCHRCOL_LESSON. Such a significant effect of SOCHET,
however, does not exist on TCHRCOL_CLASS as shown in Table 5-2. It appears that
teachers in countries characterized by socio-cultural heterogeneity tend to participate to
a greater degree in collaborative lesson planning although not necessarily in
collaborative classroom observation. In this respect, hypothesis 3 is partly supported, in
which teachers in countries with a higher level of socio-cultural heterogeneity are
expected to be more likely to engage in collaborative interaction with other teachers.
Concerning the five additional hypotheses, the significant positive association of
CLIMATE with both TCHRCOL_LESSON and TCHRCOL_CLASS is very strong as
reported in Tables 5-1 and 5-2. Teachers‟ collaborative interaction with other teachers
appears to be more likely in schools with a more positive organizational climate, as
93

suggested by hypothesis A1. Further, the effect of PRINLEAD is significant on
TCHRCOL_CLASS although not on TCHRCOL_LESSON. This result partly supports
hypothesis A2, in which it was expected that teachers‟ collaborative interaction with
other teachers is more likely in schools where the principal‟s engagement in instructionrelated activities is greater.
In addition, as indicated in Tables 5-1 and 5-2, CURRSTD at level 2 is also
significantly positively connected to both TCHRCOL_LESSON and TCHRCOL_CLASS,
supporting hypothesis A3 that teachers in countries with a more standardized national
curriculum are more likely to engage in collaborative interaction with other teachers. It
is also important to note that there is a cross-level interaction effect between CURRSTD
at level 2 and UNC_TCHUNDS at level 1 on TCHRCOL_CLASS as shown in Table 5-2;
further, similar cross-level interaction effects exist not only between the level-2
interaction term CURRSTD × EDEXPN and UNC_TCHUNDS at level 1 on
TCHRCOL_LESSON, but also between the same level-2 interaction term and
UNC_CONTEXT at level 1 on TCHRCOL_CLASS as reported in Tables 5-1 and 5-2,
respectively. These significant cross-level interaction effects substantiate hypothesis A4
that the positive association between teachers‟ exposure to uncertainty in teaching and
their engagement in collaborative interaction with other teachers is likely to be weaker
in countries with a more standardized national curriculum. These results are also in
alignment with the uncertainty management perspective in light of the possibility that
the presence of a standardized national curriculum may alleviate a certain amount of
uncertainty in teaching as the curriculum may serve as a safety net that teachers can
94

depend on when making various decisions about their teaching.
Finally, Tables 5-1 and 5-2 show that the effect of INDVCUL at level 2 is not
statistically significant on either TCHRCOL_LESSON or TCHRCOL_CLASS. Although
a significant negative bivariate association exists between INDVCUL and
TCHRCOL_CLASS as reported in Table 4-6 in chapter 4, such a significant effect
disappears as other variables are introduced in the model. Thus, hypothesis A5 is not
supported. It appears that a teacher‟s engagement in collaborative interaction with
other colleague teachers is fairly independent of her/his country‟s individualistcollectivist cultural norms. However, this result does not preclude the possibility that
teacher interaction may differ qualitatively, although not quantitatively, depending on
varying individualist-collectivist cultural norms across countries.43
Table 5-3 reports the results from multilevel ordinal logistic regressions for
TCHRCOL_VARIETY. The coefficients in this table indicate the amount of increase in
the predicted ordered log odds of moving to the next higher level in
TCHRCOL_VARIETY by a one-unit increase in an independent variable, with all other
things in the model held constant. Similar to the findings reported in Tables 5-1 and 5-2,
both UNC_TCHUNDS and UNC_CONTEXT are significantly positively related to

43. See, e.g., Wang, Strong, and Odell’s (2004) study, in which they found some
qualitative differences between Chinese and U.S. mentor-novice conversations. For example,
their data showed that U.S. mentor-novice pairs paid substantial attention to individual students,
but their discussion about student issues was not very well connected to their reflections on
teaching. Chinese mentor-novice pairs, on the other hand, talked much about subject matter in
relation to teaching and student learning, but they tended to discuss students in groups or
categories.
95

Table 5-3. Multilevel ordinal logistic regressions for TCHRCOL_VARIETY
Fixed effect
Intercept, γ 00
SOCHET, γ 01
CURRSTD, γ 02
INDVCUL, γ 03
EDEXPN, γ 04
CURRSTD × EDEXPN, γ 05
UNC_TCHUNDS, γ 10
× CURRSTD, γ 11
× EDEXPN, γ 12
× CURRSTD × EDEXPN, γ 13
UNC_CONTEXT, γ 20
× CURRSTD, γ 21
× EDEXPN, γ 22
× CURRSTD × EDEXPN, γ 23
CLIMATE, γ 30
PRINLEAD, γ 40
WORKLOAD, γ 50
TCHREXP, γ 60
MATHMJR, γ 70
FEMALE, γ 80
Threshold difference, δ (2)

I
Coeff. (SE )
-.019 (.214)
.090 (.078)

OR
.981
1.094

Coeff.
-.038
.043
.323
.076
-.139
.088
.257 (.041) 1.293 *** .232
-.040
-.074
-.034
.213 (.079) 1.237 **
.358
.011
-.027
-.116
.457
.117
.062
.013
.168
-.005
3.258 (.127) 25.994 *** 3.329

II
(SE )
OR
(.181)
.963
(.063) 1.044
(.084) 1.382
(.070) 1.079
(.150)
.870
(.100) 1.092
(.045) 1.262
(.024)
.961
(.028)
.929
(.022)
.966
(.071) 1.430
(.031) 1.011
(.057)
.974
(.033)
.890
(.073) 1.579
(.028) 1.124
(.042) 1.064
(.028) 1.013
(.083) 1.182
(.084)
.995
(.130) 27.907

***

***
*
***

**
***
***

*
***

Random effect
Var.
df
χ²
Var.
df
χ²
Intercept, u 0j
1.206
27 620.6 *** .839
23 223.1 ***
UNC_TCHUNDS, u 1j
.020
28
53.5 **
.039
25
46.8 **
UNC_CONTEXT, u 2j
.085
29
77.8 *** .059
25
43.6 **
CLIMATE, u 3j
.062
28
44.8 *
PRINLEAD, u 4j
.010
28
46.3 *
WORKLOAD, u 5j
.018
29
36.6
TCHREXP, u 6j
.011
29
79.0 ***
MATHMJR, u 7j
.088
29
28.1
FEMALE, u 8j
.099
29
39.7 +
Note . All independent variables are grand-mean-centered. Coeff. = Unstandardized ordered logit
coefficients with robust standard errors in parentheses. OR = odds ratios.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

96

TCHCOL_VARIETY. The ordered log-odds estimate for the effect of UNC_TCHUNDS
is 0.232, which translates into exp(0.232) = 1.262 in terms of the odds ratio. Since one
unit equals one standard deviation in UNC_TCHUNDS, it is possible to say that, for a
one-standard-deviation increase in UNC_TCHUNDS, we can expect to see a 26.2% (=
(1.262 – 1) × 100) increase in the odds of moving to the next higher level in
TCHRCOL_VARIETY, after all other things in the model are taken into account.
Similarly, the ordered log-odds estimate for the effect of UNC_CONTEXT is 0.358,
which is equivalent to exp(0.358) = 1.430 in terms of the odds ratio. This means that a
one-standard-deviation increase in UNC_CONTEXT corresponds to a 43.0% (= (1.430 –
1) × 100) increase in the odds of moving to the next higher level in
TCHRCOL_VARIETY, after all other things in the model are controlled for. Thus,
hypotheses 1 and 2 are corroborated.
However, the effect of national-level socio-cultural heterogeneity is insignificant
in Table 5-3. That is, different from hypothesis 3, teachers‟ engagement in collaborative
interaction appears to be fairly independent of the socio-cultural homogeneityheterogeneity dimension of national context. Although there is a significant positive
bivariate relationship between SOCHET and TCHRCOL_VARIETY as reported in Table
4-7 in chapter 4, the effect of SOCHET does not remain significant when a range of other
variables are introduced in the model.
In addition, both CLIMATE and PRINLEAD are significantly positively related
to TCHRCOL_VARIETY. In accordance with hypotheses A1 and A2, it appears that
teachers are likely to engage in more varied types of collaborative interaction with other
97

teachers if they work in a school where its organizational climate is more positive
and/or principal instructional leadership is stronger. Also, the effect of CURRSTD is
significantly positive, meaning that teachers in countries with a more standardized
national curriculum are likely to collaborate with other teachers through more various
types of collaborative interaction, which supports hypothesis A3. Further, there is a
significant cross-level interaction effect between UNC_CONTEXT at level 1 and
CURRSTD × EDEXPN at level 2, in alignment with hypothesis A4. That is, the positive
association between teachers‟ exposure to classroom-contextual uncertainty and the
range of collaborative interaction types in which they engage is likely to be weaker in
countries with a more standardized national curriculum. Finally, INDVCUL is not in a
significant relationship with TCHRCOL_VARIETY, indicating that how various the
types of collaborative interaction in which a teacher participates is fairly independent of
her/his country‟s individualist-collectivist cultural norms. This result does not support
hypothesis A5 that teachers in countries with a more individualist culture would be less
likely to engage in collaborative interaction with other teachers.

5.3. Discussion

The results from a series of multilevel regression analyses presented in this
chapter support the possibility that, as expected from the uncertainty management
perspective discussed in chapter 2, teacher collegiality may be related with teachers‟
collective effort to deal with uncertainties with which they are confronted in their
98

teaching. The results suggest that teachers who place greater emphasis on instructional
strategies to implement teaching for understanding, and thus confront greater
instructional uncertainty, are more likely to engage in both collaborative lesson
planning and collaborative classroom observation; in addition, they also tend to
participate in more various types of collaborative interaction. Similarly, teachers who
face greater classroom-contextual uncertainty tend to engage to a greater degree in
collaborative classroom observation, and they are more likely to collaborate with other
teachers through multiple types of collaborative interaction.
Further, teachers in countries with a higher level of socio-cultural heterogeneity
are more likely to engage in collaborative lesson planning with other teachers.
Considering that a higher level of socio-cultural heterogeneity of a country is likely to
indicate a lower degree of societal-level value consensus with respect to various aspects
of schooling as a social institution, it is not surprising from the uncertainty management
perspective that teachers in socio-culturally more heterogeneous countries tend to
collaborate to a greater degree with their colleague teachers in order to collectively
manage the consequent uncertainty.
In addition, also notable is the strong and significant effect of school climate on
collaborative interaction among teachers. Regardless of how to measure collaborative
teacher interaction, the results from a series of multilevel regression analyses presented
in this chapter substantiate a consistently strong positive association between school
climate and teachers‟ engagement in collaborative interaction with their colleagues.
Another clear pattern is the significant effect of national curriculum standardization on
99

teachers‟ engagement in collaborative interaction, no matter how their collaborative
interaction is measured. It appears that a standardized national curriculum may
facilitate collaborative interaction among teachers insofar as it provides them with some
shared technical language about teaching and learning.
At the same time, some results in this chapter indicate that the effect of teachers‟
exposure to uncertainty in teaching on their engagement in collaborative interaction
with other teachers is likely to lessen in the presence of a highly standardized national
curriculum. This sheds light on the possibility that the presence of a standardized
national curriculum may alleviate a certain amount of uncertainty in teaching perceived
by teachers to the degree to which the standardized curriculum may serve as a safety
net that teachers can depend on when making various decisions about how to deal with
a variety of uncertainties they confront in their classroom teaching.
Finally, it is very interesting that the results reported in this chapter do not
support the popular hypothesis that collaborative teacher interaction is highly
dependent on the degree to which the country has a collectivist culture as opposed to
an individualist culture.44 Of course, this insignificant effect does not preclude the
possibility that teacher interaction may differ qualitatively depending on varying
individualist-collectivist cultural norms across countries; however, it appears that,
when a range of other factors are taken into account simultaneously, the effect of

44. For example, Leonard and Leonard (2003) discuss the possibility that the culture of
U.S. schools, embedded in the macro-culture of individualism in the United States, contributes to
a low level of teacher collaboration. See also, e.g., Hofstede (2001) and Schein (1985) for further
discussions.
100

societal individualism does not remain significant. Despite the popular assumption that
the macro-culture of individualism in the United States is one of the important factors
leading to a low level of collaboration among teachers in U.S. schools, this assumption
is not corroborated by the results in this chapter.
In the next chapter, a set of regression analyses is performed based on the U.S.only data in order to examine the case of the United States. Although the results
reported in the present chapter provide valuable information about the nature of
collaborative teacher interaction based on a hierarchically-structured international
dataset, another series of analyses, conducted with all level-1 variables used in this
chapter and some additional U.S.-specific contextual variables, will proffer further
insights into collaborative teacher interaction within the context of the United States.

101

CHAPTER 6
ANALYSES OF THE CASE OF THE UNITED STATES

6.1. Within-U.S. Analyses
In this chapter, I conduct a series of analyses using the U.S. portion of the data to
examine the case of the United States, where teachers‟ engagement in meaningful
collaborative work around teaching has been reported to be the exception rather than
the norm (Bryk & Schneider, 2002; Fullan & Hargreaves, 1996; Lortie, 2002; Rosenholtz,
1989). While the results reported in chapters 4 and 5 provide valuable information
about the nature of collaborative teacher interaction based on a multinational dataset,
this chapter aims to proffer further useful insights into collaborative teacher interaction
within the context of the United States. As discussed in chapter 3, the availability of
some additional U.S.-specific TIMSS data prepared by the National Center for
Educational Statistics makes it possible to undertake a more fine-grained examination
of collaborative interaction among U.S. teachers.
Tables 6-1 and 6-2 report the results from multiple regressions for
TCHRCOL_LESSON and TCHRCOL_CLASS, respectively. Four different models are
reported in each table. Model I includes all level-1 variables used in multilevel
modeling in chapter 5 and two additional U.S.-specific variables, which are the course
taught and the school type. Model II is identical to model I except that UNC_TCHUNDS
and UNC_CONTEXT are replaced with the seven variables from which these two
102

Table 6-1. Ordinary least squares regressions for TCHRCOL_LESSON: The case of the United
States
B
.171
.071

I
(SE )
(.052)
(.081)

β
.163 ***
.047

B

II
(SE )

β

UNC_TCHUNDS
UNC_CONTEXT
TCHUNDS
.271 (.077)
.176
ACHIEVE
.064 (.117)
.043
ACHIVAR
-.071 (.178) -.023
EDCAPIT
.029 (.080)
.034
EDCAVAR
.126 (.075)
.121
LANGMIN
.034 (.036)
.049
SHORTRES
-.051 (.085) -.030
CLIMATE
.206 (.070)
.145 **
.190 (.072)
.134
PRINLEAD
.005 (.025)
.010
.010 (.026)
.020
WORKLOAD
.076 (.041)
.089 +
.072 (.041)
.085
TCHREXP
-.054 (.021) -.125 **
-.051 (.021) -.119
MATHMJR
.021 (.095)
.011
.034 (.096)
.017
FEMALE
.194 (.097)
.096 *
.205 (.097)
.101
Course taught
Basic math (ref.)
Algebra
-.086 (.106) -.042
-.101 (.107) -.049
Geometry
.416 (.354)
.059
.347 (.365)
.049
Others
.045 (.265)
.009
.050 (.269)
.009
School type
Public (ref.)
Private, religious
-.759 (.153) -.254 ***
-.729 (.158) -.245
Private, others
-.498 (.224) -.110 *
-.486 (.235) -.108
Free or reduced-price lunch
Intercept
.520 (.293) +
-.690 (.763)
R²
.18
.20
n
392
392
Note . B = Unstandardized coefficients with standard errors in parentheses. β = standardized
coefficients. Ref. = reference category.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

103

***

+

**
+
**
*

***
*

Table 6-1. (cont'd)

UNC_TCHUNDS
UNC_CONTEXT
TCHUNDS
ACHIEVE
ACHIVAR
EDCAPIT
EDCAVAR
LANGMIN
SHORTRES
CLIMATE
PRINLEAD
WORKLOAD
TCHREXP
MATHMJR
FEMALE
Course taught
Basic math (ref.)
Algebra
Geometry
Others
School type
Public (ref.)
Private, religious
Private, others
Free or reduced-price lunch
Intercept
R²
n

B
.162
.081

III
(SE )
(.053)
(.100)

β
.152 **
.053

β

.265
.050
-.120
.022
.138
.035
-.037
.206
.002
.087
-.054
.074
.215

(.081)
(.123)
(.196)
(.090)
(.080)
(.038)
(.089)
(.074)
(.026)
(.043)
(.022)
(.099)
(.100)

.169
.034
-.038
.025
.129
.050
-.022
.144
.004
.102
-.126
.038
.106

***

+

.223
-.003
.090
-.059
.059
.209

(.073)
(.026)
(.043)
(.021)
(.098)
(.100)

.156
-.006
.106
-.138
.030
.103

-.081
.424
.072

(.109)
(.362)
(.278)

-.040
.061
.013

-.093
.347
.065

(.111)
(.372)
(.283)

-.045
.050
.012

-.702
.008
-.002
.468

(.168)
(.353)
(.021)
(.330)
.17
377

-.225 ***
.001
-.006

-.678
.043
-.007
-.590

(.173)
(.356)
(.023)
(.861)
.18
377

-.217 ***
.006
-.023

104

**

B

IV
(SE )

*
**
*

**
*
**
*

composite factors have been constructed as presented in Table 3-1 in chapter 3. Finally,
models III and IV are identical to models I and II, respectively, except that one
additional U.S.-specific contextual variable, the proportion of free or reduced-price lunch
recipients, is added to these models as a measure of school-level poverty.45 It should be
noted that models III and IV have a slightly smaller sample size than models I and II
because data for this newly added variable are available only for public and nonprofit
private schools.46
A very clear pattern is the significant effect of UNC_TCHUNDS across different
models. With all other things held constant, UNC_TCHUNDS is in a significant positive
relationship with both TCHRCOL_LESSON and TCHRCOL_CLASS, indicating that
teachers who face greater instructional uncertainty due to their effort to implement
teaching for understanding tend to engage to a greater degree in both collaborative
lesson planning and collaborative classroom observation. The standardized regression
coefficient for UNC_TCHUNDS in model I in Table 6-1 indicates that, for a onestandard-deviation increase in UNC_TCHUNDS, we can expect an increase in
TCHRCOL_LESSON by 16.3% of one standard deviation. This is higher than the
estimate of 13.5% based on the international data analyzed in chapter 5. Similarly,
model I in Table 6-2 indicates that, for a one-standard-deviation increase in
UNC_TCHUNDS, the amount of expected increase in TCHRCOL_CLASS is 19.7% of
45. For detailed information about these U.S.-specific variables, see chapter 3 (section
3.3.4).
46. That is, the National School Lunch Program operates in public and nonprofit private
schools.
105

Table 6-2. Ordinary least squares regressions for TCHRCOL_CLASS: The case of the United
States
B
.116
.197

I
(SE )
(.029)
(.046)

β
.197 ***
.232 ***

B

II
(SE )

β

UNC_TCHUNDS
UNC_CONTEXT
TCHUNDS
.190 (.044)
.220
ACHIEVE
-.099 (.066) -.119
ACHIVAR
.027 (.101)
.016
EDCAPIT
-.055 (.045) -.116
EDCAVAR
-.017 (.043) -.028
LANGMIN
-.004 (.020) -.009
SHORTRES
.115 (.049)
.121
CLIMATE
.090 (.040)
.114 *
.107 (.041)
.134
PRINLEAD
.003 (.014)
.009
.000 (.015)
.002
WORKLOAD
-.025 (.024) -.053
-.025 (.024) -.053
TCHREXP
.004 (.012)
.018
.005 (.012)
.020
MATHMJR
-.012 (.054) -.011
-.015 (.055) -.014
FEMALE
.029 (.055)
.026
.028 (.056)
.025
Course taught
Basic math (ref.)
Algebra
.012 (.061)
.011
.020 (.061)
.017
Geometry
.797 (.202)
.200 ***
.736 (.208)
.185
Others
.393 (.151)
.133 **
.372 (.153)
.126
School type
Public (ref.)
Private, religious
-.017 (.087) -.010
-.021 (.090) -.013
Private, others
.332 (.128)
.131 **
.336 (.134)
.133
Free or reduced-price lunch
Intercept
-.007 (.167)
.374 (.435)
R²
.16
.17
n
392
392
Note . B = Unstandardized coefficients with standard errors in parentheses. β = standardized
coefficients. Ref. = reference category.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

106

***

*
**

***
*

**

Table 6-2. (cont'd)

UNC_TCHUNDS
UNC_CONTEXT
TCHUNDS
ACHIEVE
ACHIVAR
EDCAPIT
EDCAVAR
LANGMIN
SHORTRES
CLIMATE
PRINLEAD
WORKLOAD
TCHREXP
MATHMJR
FEMALE
Course taught
Basic math (ref.)
Algebra
Geometry
Others
School type
Public (ref.)
Private, religious
Private, others
Free or reduced-price lunch
Intercept
R²
n

B
.093
.155

III
(SE )
(.029)
(.055)

.105
.000
-.028
.001
-.022
.084

(.040)
(.014)
(.023)
(.012)
(.053)
(.055)

.004
.776
.336

(.060)
(.198)
(.152)

-.012
.239
.017
-.164

(.092)
(.193)
(.011)
(.181)
.14
377

107

β
.161 **
.189 **

.136 **
.001
-.060
.004
-.021
.077

.003
.206 ***
.115 *

-.007
.062
.100

B

IV
(SE )

β

.138
-.024
.103
-.048
.005
-.024
.128
.122
-.002
-.027
.001
-.033
.082

(.044)
(.068)
(.108)
(.049)
(.044)
(.021)
(.049)
(.041)
(.015)
(.024)
(.012)
(.054)
(.055)

.163 **
-.030
.060
-.102
.009
-.065
.140 **
.159 **
-.006
-.060
.002
-.031
.075

.000
.704
.279

(.061)
(.204)
(.155)

.000
.187 ***
.096 +

.000
.252
.018
-.202

(.095)
(.195)
(.012)
(.472)
.16
377

.000
.066
.104

one standard deviation. This is much higher than the estimate of a 9.0% increase based
on the international data analyzed in chapter 5. These results suggest the possibility that
the effect of teachers‟ effort to implement teaching for understanding on their
engagement in collaborative interaction is stronger in the United States than in a
majority of other countries.47 Thus, hypothesis 1 is strongly supported.
Another related finding is the significant association between UNC_CONTEXT
and TCHRCOL_CLASS, indicating that teachers who face greater classroom-contextual
uncertainty tend to engage to a greater degree in collaborative classroom observation.
As reported in model I in Table 6-2, a one-standard-deviation increase in
UNC_CONTEXT is significantly associated with an expected increase in
TCHRCOL_CLASS by 23.2% of one standard deviation. Again, this is higher than the
estimate of a 15.1% increase based on the international data analyzed in chapter 5.
Although UNC_CONTEXT does not appear to have a significant effect on
TCHRCOL_LESSON as shown in model I in Table 6-1, model II in the same table
indicates that EDCAVAR has a significantly positive relationship with
TCHRCOL_LESSON. This means that teachers confronted with greater student
heterogeneity in terms of socioeconomic background tend to engage more in
collaborative lesson planning. With all other things in the model held constant, a onestandard-deviation increase in EDCAVAR is significantly associated with an expected
increase in TCHRCOL_LESSON by 12.1% of one standard deviation. These results

47. This possibility is substantiated by the results from a set of further analyses
conducted in the next section of this chapter (section 6.2).
108

provide evidence to substantiate the relationship between teachers‟ exposure to
classroom-contextual uncertainties and their engagement in collaborative interaction
with other colleague teachers. Thus, hypothesis 2 is largely supported, in which
teachers who face greater classroom-contextual uncertainty are expected to exhibit
greater engagement in collaborative interaction with other teachers.
In addition to these findings supporting the primary hypotheses of this study, a
noticeable pattern shown in Tables 6-1 and 6-2 is the significant effect of CLIMATE on
both TCHRCOL_LESSON and TCHRCOL_CLASS. Such a significant relationship is
evident across all models in Tables 6-1 and 6-2, with the standardized regression
coefficients ranging from 0.134 to 0.156 in Table 6-1 and from 0.114 to 0.159 in Table 6-2.
These results mean that, for every one-standard-deviation increase in CLIMATE, one
can expect an increase in TCHRCOL_LESSON by at least about 13% of one standard
deviation and an increase in TCHRCOL_CLASS by at least about 11% of one standard
deviation, after various other things are controlled for. This supports hypothesis A1 that
teachers‟ collaborative interaction with other teachers is more likely in schools with a
more positive organizational climate.
It is also notable that there are some teacher characteristics that are significantly
related to teachers‟ engagement in collaborative lesson planning. As reported in Table
6-1, with all other things in the model held constant, female teachers tend to engage
more in collaborative lesson planning than male teachers, introducing the possibility
that male teachers are isolated from their colleagues to a greater degree than are female
teachers. Further, it appears that teachers who have been in the profession of teaching
109

for a longer period of years tend to engage less in collaborative lesson planning. One
plausible explanation on the basis of the uncertainty management perspective is that the
amount of uncertainty felt by a teacher in a given complex situation may be gradually
reduced as she/he stays longer in the profession of teaching; this is because teachers
with more experience tend to have a greater amount of “situated knowledge and
expertise in teaching” (Leinhardt, 1988), with which they can better manage their
uncertainty without seeking help from other teachers. Finally, teachers with more
workload appear to engage to a greater extent in collaborative lesson planning. This
result seems counterintuitive because one might plausibly expect heavy workload to be
a factor that deters teachers from engaging in collaborative interaction with other
teachers. However, from the uncertainty management perspective, this result is not
surprising, for it is very likely that teachers confront a greater amount of task
uncertainty as their workload becomes heavier; in other words, they may want to deal
with the uncertainty caused by the complexity and diversity of tasks by collaborating
with their colleagues.
Also, some difference is observed among eighth-grade mathematics teachers in
their engagement in collaborative classroom observation depending on which
mathematics subject one teaches. With all other things in the model being equal,
geometry teachers appear to participate to a greater degree in collaborative classroom
observation than teachers who teach basic eighth-grade mathematics, while no
significant difference is observed between algebra teachers and basic eighth-grade
mathematics teachers. Although it is unclear why such a significant difference is found,
110

this result points to the possibility that there may be varying degrees of collegiality
among teachers depending on certain characteristics of the curricular content they
teach.48
With respect to the effect of the school type, teachers working in religiouslyaffiliated private schools appear to engage to a significantly lesser degree in
collaborative lesson planning when compared to teachers working in public schools, as
reported in Table 6-1; in addition, they do not participate to a greater degree in
collaborative classroom observation as shown in Table 6-2. These results contrast with
the popular belief that private schools are more likely to be places where teachers
collaboratively build and sustain professional communities than are public schools.
Although further investigation is necessary, one possible explanation may be that
teachers working in religiously-affiliated private schools already share a relatively clear
teaching philosophy, which lessens the perceived need to seek help or advice from
other teachers.49
Finally, all these patterns reported in models I and II in Tables 6-1 and 6-2 do not

48. Although further systematic studies are needed to substantiate this possibility, it is
notable that the standardized regression coefficient for “geometry” is even greater than that for
CLIMATE in all models in Tables 6-1 and 6-2. Given the strong and stable relationship between
a positive school climate and collaborative teacher interaction, an even stronger association
observed between geometry teaching and the level of engagement in collaborative teacher
interaction is very interesting. It would be a useful future study to examine this relationship more
closely.
49. As another possibility, teachers working in religiously-affiliated private schools may
have significantly more extra duties at school other than classroom teaching, resulting in
different forms or patterns of collaborative teacher interaction that are not adequately captured
by the notion of collaborative lesson planning or collaborative classroom observation.
111

change even if the proportion of free or reduced-price lunch recipients, a school-level
poverty measure, is newly introduced to each model. Both the statistical significance
level and the magnitude of effect of each independent variable little differ between
model I and model III and between model II and model IV in either table.50 It appears
that after controlling for a range of classroom-level socioeconomic contextual variables,
school-level poverty is not a significant factor related to collaborative teacher interaction
patterns.
Table 6-3 reports the results from a set of ordinal logistic regressions for
TCHRCOL_VARIETY. The coefficients in this table indicate the amount of increase in
the predicted ordered log odds of moving to the next higher level in
TCHRCOL_VARIETY by a one-unit increase in an independent variable, with the other
things in the model held constant. Similar to the findings reported in Tables 6-1 and 6-2,
both UNC_TCHUNDS and UNC_CONTEXT are significantly positively related to
TCHCOL_VARIETY. That is, in accordance with hypotheses 1 and 2, it appears that
teachers who face greater instructional uncertainty and/or classroom-contextual
uncertainty are likely to participate in more varied types of collaborative teacher

50. In models II and IV of Tables 6-1, 6-2, and 6-3, ACHIEVE, ACHIVAR, EDCAPIT,
EDCAVAR, LANGMIN, and SHORTRES, which are considerably inter-correlated to one
another as evident in the factor analysis reported in Table 3-1 in chapter 3 where they all are
heavily loaded on UNC_CONTEXT, appear to have statistically less significant effects than
expected from the uncertainty management perspective. This seems to be largely due to the
multicollinearity among these variables. If one of these variables is entered into a regression
equation at a time, with all other things remaining unchanged in the model, the effects of these
variables become more discernible, especially in the regressions for TCHRCOL_CLASS. See
Table A7 in the Appendix.
112

Table 6-3. Ordinal logistic regressions for TCHRCOL_VARIETY: The case of the United States

B
.418
.532

I
(SE )
(.129)
(.202)

OR
1.518 ***
1.702 **

B

II
(SE )

OR

UNC_TCHUNDS
UNC_CONTEXT
TCHUNDS
.704 (.194) 2.023 ***
ACHIEVE
-.442 (.292)
.643
ACHIVAR
-.282 (.443)
.755
EDCAPIT
-.180 (.202)
.835
EDCAVAR
-.093 (.189)
.911
LANGMIN
.026 (.088) 1.027
SHORTRES
.112 (.212) 1.119
CLIMATE
.509 (.175) 1.663 **
.527 (.179) 1.695 **
PRINLEAD
.048 (.062) 1.049
.036 (.063) 1.037
WORKLOAD
.102 (.102) 1.107
.103 (.102) 1.108
TCHREXP
-.076 (.051)
.927
-.077 (.052)
.926
MATHMJR
-.333 (.233)
.717
-.286 (.237)
.751
FEMALE
.417 (.240) 1.517 +
.428 (.243) 1.535 +
Course taught
Basic math (ref.)
Algebra
-.274 (.261)
.760
-.212 (.264)
.809
Geometry
2.398 (.938) 11.005 **
2.332 (.969) 10.295 *
Others
.887 (.659) 2.427
1.014 (.673) 2.758
School type
Public (ref.)
Private, religious
-2.264 (.398)
.104 ***
-2.272 (.413)
.103 ***
Private, others
.801 (.544) 2.227
.792 (.578) 2.207
Free or reduced-price lunch
Threshold 1
-.469 (.721)
-2.833 (1.963)
Threshold 2
2.740 (.736) ***
.406 (1.959)
Nagelkerke R ²
.27
.28
n
392
392
Note . B = Unstandardized ordered logit coefficients with standard errors in parentheses. OR =
odds ratios. Ref. = reference category.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

113

Table 6-3. (cont'd)
*************************************************************************************
III
IV
B
(SE )
OR
B
(SE )
OR
UNC_TCHUNDS
.351 (.131) 1.421 **
UNC_CONTEXT
.275 (.242) 1.316
TCHUNDS
.577 (.198) 1.780 **
ACHIEVE
-.353 (.301)
.703
ACHIVAR
-.227 (.478)
.797
EDCAPIT
-.056 (.222)
.946
EDCAVAR
-.031 (.196)
.969
LANGMIN
-.026 (.091)
.974
SHORTRES
.146 (.217) 1.158
CLIMATE
.571 (.179) 1.770 ***
.585 (.183) 1.796 ***
PRINLEAD
.013 (.063) 1.013
.007 (.064) 1.007
WORKLOAD
.117 (.104) 1.124
.115 (.104) 1.122
TCHREXP
-.094 (.052)
.910 +
-.093 (.053)
.911 +
MATHMJR
-.283 (.236)
.753
-.254 (.241)
.775
FEMALE
.575 (.245) 1.776 *
.549 (.247) 1.732 *
Course taught
Basic math (ref.)
Algebra
-.322 (.264)
.725
-.266 (.267)
.766
Geometry
2.224 (.947) 9.242 *
2.197 (.973) 8.995 *
Others
.899 (.668) 2.458
.914 (.683) 2.495
School type
Public (ref.)
Private, religious
-1.976 (.419)
.139 ***
-1.995 (.432)
.136 ***
Private, others
1.392 (.841) 4.024 +
1.384 (.852) 3.990 +
Free or reduced-price lunch
.097 (.051) 1.102 +
.083 (.055) 1.086
Threshold 1
.274 (.797)
-1.032 (2.126)
Threshold 2
3.488 (.824) ***
2.191 (2.132)
Nagelkerke R ²
.24
.25
n
377
377

114

interaction. The ordered log-odds estimate for the effect of UNC_TCHUNDS is 0.418,
which translates into exp(0.418) = 1.518 in terms of the odds ratio. Thus, it is possible to
say that, for every one-unit increase in UNC_TCHUNDS, we can expect a 51.8%
(=(1.518 – 1) × 100) increase in the odds of moving to the next higher level in
TCHRCOL_VARIETY, with all other things in the model held constant. Also, the
ordered log-odds estimate for the effect of UNC_CONTEXT is 0.532, which is
equivalent to exp(0.532) = 1.702 in terms of the odds ratio. This means that a one-unit
increase in UNC_CONTEXT is associated with a 70.2% (= (1.702 – 1) × 100) increase in
the odds of moving to the next higher level in TCHRCOL_VARIETY, after all other
things in the model are controlled for. In addition, also notable is the significant positive
association of CLIMATE with TCHRCOL_VARIETY. As expected by hypothesis A1, it
appears that teachers are likely to engage in more various types of collaborative
interaction with their colleague teachers when they work in schools where teacher
morale is high and organizational goals are clearly communicated and well
implemented.

6.2. Comparison with Other Countries

Figures 6-1 and 6-2 show standardized beta coefficients for UNC_TCHUNDS
and UNC_CONTEXT estimated by ordinary least squares regressions conduced within
each country in the sample. Each within-nation analysis includes all other level-1
variables (i.e., school climate, principal‟s instructional leadership, teacher workload,
115

.40
Standardized beta coefficient for UNC_TCHUNDS
Standardized beta coefficient for UNC_CONTEXT

.30

.20

.10

.00

-.10

-.20
Figure 6-1. Within-nation ordinary least squares regressions for TCHRCOL_LESSON: Standardized coefficients for
UNC_TCHUNDS and UNC_CONTEXT. Note . Independent variables included in the model are CLIMATE, PRINLEAD,
WORKLOAD, TCHREXP, MATHMJR, and FEMALE, in addition to UNC_TCHUNDS and UNC_CONTEXT. Statistically
insignificant results are not reported. Full results are available in Table A8 in the Appendix.

116

.40

Standardized beta coefficient for UNC_TCHUNDS
Standardized beta coefficient for UNC_CONTEXT
.30

.20

.10

.00

Figure 6-2. Within-nation ordinary least squares regressions for TCHRCOL_CLASS: Standardized coefficients for UNC_TCHUNDS
and UNC_CONTEXT. Note . Independent variables included in the model are CLIMATE, PRINLEAD, WORKLOAD, TCHREXP,
MATHMJR, and FEMALE, in addition to UNC_TCHUNDS and UNC_CONTEXT. Statistically insignificant results are not
reported. Full results are available in Table A9 in the Appendix.

117

teacher experience, mathematics major, and female teacher) as control variables in order
to allow a cross-national comparison of the extent to which teachers‟ exposure to
uncertainty in teaching is significantly related to collaborative lesson planning and
collaborative classroom observation, even after a range of factors are taken into
account.51 The results reported in Figures 6-1 and 6-2 indicate that one or both of
UNC_TCHUNDS and UNC_CONTEXT are significantly positively associated with
collaborative lesson planning and/or collaborative classroom observation in a majority
of countries in the sample, which is in accordance with the results obtained from the
multilevel modeling analyses conduced in chapter 5. A very interesting finding is that
the United States is one of only two countries where both of these uncertainty measures
are significantly positively connected to both collaborative lesson planning and
collaborative classroom observation, after all the aforementioned control variables are
included in the regression model. Such a robust relationship between these uncertainty
measures and collaborative teacher interaction within the context of the United States
suggests that the uncertainty management perspective is very important in explaining
the case of the United States.
Similarly, Figure 6-3 reports ordered odds ratios for UNC_TCHUNDS and
UNC_CONTEXT estimated by ordinal logistic regressions conduced within each
country, after all other level-1 variables are also included in the model. The patterns
shown in Figure 6-3 are not very different from those in Figures 6-1 and 6-2. In most of
51. In order to make the regression model identical across all countries in the sample, the
U.S.-specific variables used in the previous section (section 6.1) were not used here when
analyzing the U.S. data.
118

5.00
Odds ratio for UNC_TCHUNDS
Odds ratio for UNC_CONTEXT

4.00

3.00

2.00

1.00

.00

Figure 6-3. Within-nation ordinal logistic regressions for TCHRCOL_VARIETY: Odds ratios for UNC_TCHUNDS and
UNC_CONTEXT. Note . Independent variables included in the model are CLIMATE, PRINLEAD, WORKLOAD, TCHREXP,
MATHMJR, and FEMALE, in addition to UNC_TCHUNDS and UNC_CONTEXT. Statistically insignificant results are not
reported. Full results are available in Table A10 in the Appendix.

119

the countries in the sample, teachers‟ exposure to uncertainty from teaching for
understanding and/or uncertainty from classroom-contextual factors is in significant
positive relationships with the extent to which they regularly engage in multiple types
of collaborative interaction with other teachers. All significant odds ratios are greater
than one, indicating that no country in the sample shows a significantly negative
association of either UNC_TCHUNDS or UNC_CONTEXT with the degree to which
teachers collaborate through varied types of teacher interaction. In addition, the United
States is one of only six countries where both of these uncertainty measures turn out to
be significantly positive, after all other level-1 variables are taken into account. This
result once again highlights the importance of the uncertainty management perspective
to understand the nature of collaborative interaction among teachers within the context
of the United States.
Overall, teachers in many countries in the sample appear to exhibit a greater
level of engagement in collaborative interaction with other teachers as the level of
uncertainty they face in their teaching becomes higher. The significant relationship
between teachers‟ exposure to uncertainty in teaching and their participation in
collaborative interaction with other teachers is particularly strong in the United States
compared to many other countries in the sample. Such a strong relationship observed in
the United States sheds light on the uncertainty management perspective in which
collaborative interaction among teachers is understood as a collective effort to make
sense of, and lessen, their feelings of discomfort with the perceived elements of
uncertainty in their teaching, whereby they can better manage to go through given
120

situations of uncertainty rather than simply choosing to either avoid or ignore those
situations.

6.3. Discussion

The results from a series of regression analyses presented in this chapter are not
very different the findings from the international data reported in chapter 5. The U.S.only data analyzed in the present chapter also give credence to the uncertainty
management perspective in which teacher collegiality is assumed to be related with
teachers‟ collective effort to deal with uncertainties in their teaching. The results in this
chapter shed light on the possibility that teachers who exert greater effort to implement
teaching for understanding and thus confront greater instructional uncertainty may
engage to a greater degree in both collaborative lesson planning and collaborative
classroom observation; in addition, these teachers seem to collaborate with other
teachers through more various types of interaction. Similarly, teachers who face greater
classroom-contextual uncertainty appear to be more likely to engage in collaborative
classroom observation, and they tend to participate in more various types of
collaborative interaction.
In addition, comparison of these results with the findings from the multilevel
cross-national analyses conducted in chapter 5 suggests that the magnitude of the
significant effect of teachers‟ exposure to uncertainty in teaching, in terms of either
instructional uncertainty or classroom-contextual uncertainty, is larger in the United
121

States than in most other countries in the sample; this is further substantiated by direct
comparison of the case of the United States with other countries in the sample.
Although further research is needed to explain why this is the case, this result reaffirms
the significance of the uncertainty management perspective to unravel the nature of
collaborative teacher interaction within the context of the United States.
Also noticeable is the significant and persistent association of school climate with
collaborative teacher interaction. Regardless of how to measure collaborative teacher
interaction, the results from a series of regression analyses presented in this chapter
indicate a strong relationship between school climate and teachers‟ engagement in
collaborative interaction with their colleagues. However, the results reported in this
chapter do not show a significant effect of principal instructional leadership. This
insignificant effect contrasts with the findings from the multilevel cross-national
analyses performed in chapter 5, where a significant positive connection was found
between principal instructional leadership and both collaborative classroom
observation and variety in collaborative interaction. The insignificant effect of principal
instructional leadership in the United States may be in part because many principals do
not have strong confidence in their knowledge about the practice of teaching; as shown
in Lortie‟s (2009) analysis of principals‟ perception of difficulties in their tasks,
principals in the United States frequently feel uncertain about the soundness of their
judgments when they supervise teachers‟ instructional practices. Considering the
limited amount of useful instructional suggestions principals can actually provide to
teachers, principals‟ instructional leadership behaviors are unlikely to have a strong and
122

sustained effect on unifying teachers into a cohesive inquiry community.52

52. To clarify why the effect of principal instructional leadership is insignificant in the
U.S.-only data while significant in the international data, it would be a useful future study to
closely examine how the effect of principal instructional leadership varies depending on various
factors, including principal characteristics and school-level and national-level contextual
variables.
123

CHAPTER 7
CONCLUSION AND DISCUSSION

7.1. Summary of the Findings

The primary aim of this study was to explore how collaborative teacher
interaction is contingent upon teachers‟ immediate classroom and school contexts as
well as broader societal and policy environments. Although much research has
identified “the collaboration associated with a professional community of teachers as a
key element of successful schools” (Darling-Hammond, 2010, p. 261), little systematic
effort has been made to understand what types of teachers, under what contextual
conditions, are likely to show a greater tendency to build and/or sustain collegial
relationships with other teachers. In order to develop concrete ideas about formulating
administrative and policy strategies to promote the beneficial effects of teacher
collegiality, it is important to have a comprehensive knowledge base about how the
level and type of collaborative teacher interaction varies depending on various factors
at multiple levels ranging from teacher-level variables capturing individual
characteristics to national-level contextual variables describing socio-cultural and policy
environments.
In order to understand better a range of multilevel contextual factors associated
with collaborative teacher interaction, I considered various independent variables based
124

on five different conceptual perspectives – a school climate perspective, a principal
instructional leadership perspective, a curriculum policy perspective, a national culture
perspective, and an uncertainty management perspective. In particular, this study
intended to both elaborate and test an uncertainty management perspective as a
theoretical framework for understanding teacher collegiality. The other four
perspectives, which are some of the most promising conventional accounts of teacher
collegiality as a dependent variable, were also considered to see if a set of primary
hypotheses derived from the uncertainty management perspective is empirically
supported even after some other plausible hypotheses are simultaneously taken into
account.
The data for this study came primarily from the TIMSS 2007 eighth-grade
mathematics dataset. This quantitative dataset, chiefly known for its student
achievement data across various countries, also contains extensive contextual
information about schools and classes, including useful data on collaborative teacher
interaction. This dataset was linked to a set of national-level contextual data from
multiple sources.53 In order to examine a series of hypotheses derived from the five
different perspectives, this study first involved a range of descriptive and cross-national
multilevel analyses. Then, another set of within-U.S. analyses was conducted to
examine closely the case of the United States, where teachers‟ engagement in
meaningful collaborative work around teaching has been known to be the exception

53. For a summary of the variables used in this study and their data sources, see Table 32 in chapter 3. See also Tables 3-3 and 3-4 for the descriptive statistics of the variables.
125

rather than the norm (Bryk & Schneider, 2002; Fullan & Hargreaves, 1996; Lortie, 2002;
Rosenholtz, 1989).
Overall, the findings from this study support an uncertainty management
perspective which understands collegiality largely as a byproduct of the collective
sensemaking of uncertainties.54 With this view of uncertainty, collegiality is assumed to
reduce organizational members‟ feelings of discomfort with given uncertain situations
and thus help them manage those situations. Extensive international and U.S.-only data
analyzed in the present study give credence to the view that teacher collegiality has to
do with teachers‟ collective effort to deal with various uncertainties they face in their
teaching. This sheds light on the possibility that lateral collegial relationships among
teachers may serve as a source of information processing, sensemaking, and problemsolving, whereby teachers can better manage to go through given situations of
uncertainty rather than simply trying to avoid those situations.
Indeed, much literature has shown that the practice of teaching is, by nature,
context-specific and full of situated-complexities, so that it cannot always be theorized
in a generic way of explanation (Lampert, 2001; Munthe, 2007; Putnam & Borko, 2000;
Wood, et al., 1990); further, teachers are always confronted with the vivid realities of
classroom teaching that are often inexplicable in terms of theoretical conceptualizations
of educational practice (Floden & Buchmann, 1993; Floden & Clark, 1988; Kennedy,
2005; Labaree, 2000). Although it is important to understand teaching as an inherently

54. For a summary of all statistical results of the present study, see Table A11 in the
Appendix.
126

uncertain task, this study postulates that the amount of uncertainty teachers confront in
their classroom teaching may vary depending on the instructional strategies they try to
incorporate into their teaching practice and the classroom contexts in which they are
positioned to implement their teaching, which this study conceptualizes as instructional
and classroom-contextual uncertainty, respectively.
More specifically, it appears that teachers who exert greater effort to implement
instructional strategies to promote students‟ engagement in inquiry-based learning and
understanding tend to engage to a greater degree in collaborative interaction with their
colleague teachers. Unlike conventional teaching accompanied by a range of readily
available routines that teachers can draw upon, an alternative way of teaching that may
be called “teaching for understanding” (D. K. Cohen, et al., 1993; Gamoran, et al., 2000;
Wiske, 1998) requires teacher knowledge that goes beyond knowing the prescribed
curricular content supposed to be delivered to students (Hiebert & Carpenter, 1992;
McLaughlin & Talbert, 1993). From teachers‟ point of view, there are, in fact,
considerable advantages to conventional teaching in the sense that it may help protect
them from instructional uncertainties that could emerge from unexpected ideas and
responses from students; this accounts for why many teachers often try to find “an
optimal level of student engagement” (Kennedy, 2005, p. 183) rather than encouraging
students towards deeper intellectual engagement (Silver, et al., 2009). As teachers put
greater emphasis on teaching for understanding, they must confront greater
instructional uncertainty because teaching becomes inevitably less reducible to
predictable routines. As a result, it is likely that these teachers engage to a greater
127

degree in collaborative interaction with their colleagues in order to share in the effort to
manage the consequent uncertainty in teaching. Both across countries and in the United
States, the findings from this study indicate that the greater the effort teachers put on
incorporating instructional strategies for teaching for understanding into their
classroom teaching, the more they tend to engage in collaborative interaction with other
teachers.55
In a similar vein, the findings from this study also indicate that classroomcontextual uncertainty, as measured by a composite factor score derived from a wide
range of variables that capture teachers‟ exposure to complex instructional
environments, is in a significant positive relationship with collaborative interaction
among teachers. A range of classroom-contextual factors, such as a low level of student
performance, a high level of student heterogeneity, and a shortage or inadequacy of
55. In the United States, teaching for understanding is a set of reform-oriented
instructional strategies (National Council of Teachers of Mathematics, 2000, 2007; National
Research Council, 2002). Given the possibility that the extent to which teaching for
understanding has been part of a reform movement may differ depending on educational systems
and traditions, the amount of instructional uncertainty perceived by teachers may vary across
countries for a given degree of effort to use teaching for understanding. Indeed, different
magnitudes of the effect of UNC_TCHUNDS (uncertainty from teaching for understanding) in
different countries, as shown in Figures 6-1, 6-2, and 6-3 in chapter 6, appear to be in line with
this possibility; in this respect, it would be a valuable future study to examine how the meaning
of teaching for understanding as perceived by teachers is qualitatively different and/or similar
across countries. Despite this possibility, however, the multilevel analysis reported in Table A6
(in the Appendix) shows that the association between teaching for understanding and
collaborative teacher interaction remains strong even when all variables are centered at their
country means to estimate teacher-level relationships net of country-level effects; this result is
not surprising, considering that if teachers use instructional strategies to open up possibilities for
students to engage in intellectually challenging tasks through inquiry-based learning, teachers
inevitably confront a greater amount of instructional uncertainty than they would while using
conventional teaching. For a related discussion, see chapter 2, especially sections 2.1.5 and 2.2.1.
128

classroom resources, for example, are likely to lead to increased uncertainty in teaching,
resulting in a variety of difficult questions with regard to both curricular content and
instructional methods (Hallinan, 1994; Ladson-Billings, 1999; Miller, et al., 2009). A
series of both cross-national multilevel analyses and within-U.S. analyses conducted for
this study shows that teachers who face greater classroom-contextual uncertainty are
likely to participate to a greater degree in collaborative classroom observation; they also
tend to collaborate with their colleagues through more varied types of collaborative
teacher interaction.
The uncertainty management perspective is further supported by the result that
teachers in countries with a higher level of socio-cultural heterogeneity are likely to
engage more in collaborative lesson planning with other teachers. Considering that
education is an institution on which different sectors of society impose multiple, often
conflicting, purposes, “teaching performances . . . can be free of controversy only in
societies which are marked by an extremely high degree of value consensus” (Lortie,
2002, p. 136). In a socio-culturally heterogeneous society, the range of the goals of
schooling envisioned by different groups of people is likely to be very wide, and many
of those goals are likely to be conflicting with one another, adding further uncertainty
to the complexities of teaching. That is, since a higher level of socio-cultural
heterogeneity of a country is likely to indicate a lower degree of societal-level value
consensus on various aspects of teaching and learning, it is not surprising from the
uncertainty management perspective that teachers in such national contexts are more
likely to collaborate with their colleagues in order to collectively deal with various
129

complexities and difficulties emanating from divergent social norms and values about
what schooling is and what teaching should look like.
Many findings from this study also provide evidence to support some
conventional perspectives on collaborative teacher interaction. Extensive research and
literature in organizational analysis has suggested that the extent to which teacher
collegiality is fostered and sustained in a school depends largely on the organizational
climate of the school (Bryk & Schneider, 2002; Hoy, 2002; Sergiovanni & Starratt, 2002;
Tschannen-Moran, 2001). The results from a series of both cross-national multilevel
analyses and within-U.S. analyses performed in this study substantiate a strong positive
relationship between school climate and all measures of teachers‟ engagement in
collaborative interaction with their colleagues.56
Another noticeable pattern is the significant positive connection between
national curriculum standardization and all measures of collaborative teacher
interaction. This result supports the conventional perspective that a standardized
national curriculum may facilitate collaborative interaction among teachers insofar as it
provides them with some shared technical language about teaching and learning (Lortie,
2002; Schmidt, et al., 2005; Stevenson & Baker, 1991). Also, some cross-level interaction
effects found in the multilevel analyses of the data suggest that the effect of teachers‟
exposure to uncertainty in teaching on their engagement in collaborative interaction
with other teachers is likely to lessen in the presence of a highly standardized national
56. This finding, however, does not necessarily indicate that teacher collegiality arises as
a consequence of a positive school climate; rather, it could just be that teacher collegiality is an
integral constituent of a positive school climate, as discussed in detail in chapter 2 (section 2.1.1).
130

curriculum. This gives credence to the possibility that curriculum standardization may
alleviate a certain amount of uncertainty in teaching as the curriculum standards may
serve as a set of readily available and authoritative guidelines that teachers can rely on
when making various decisions about their teaching.
This study‟s findings about the effect of principal instructional leadership are
mixed. One the one hand, the findings from the cross-national multilevel data provide
some evidence to support the conventional perspective in which principal instructional
leadership is seen as an important factor that helps make the school a collaborative
inquiry community (Printy, 2008; Reitzug, 1997; Somech, 2005). When operationalized
as the degree to which a school principal spends time on instructionally relevant
activities, principal instructional leadership is significantly positively related to the
degree to which teachers engage in collaborative classroom observation as well as to the
extent that teachers collaborate through various types of interaction. On the other hand,
such a significant effect of principal instructional leadership is not observed in withinU.S. analyses. This insignificant effect of principal instructional leadership may be in
part attributable to the fact that many principals do not have strong confidence in their
knowledge about the practice of teaching; principals in the United States frequently feel
uncertain about the soundness of their judgments when they supervise teachers‟
instructional practices (Lortie, 2009). Given the limited amount of practical instructional
suggestions principals can actually provide to teachers, principals‟ instructional
leadership behaviors are unlikely to have a strong and sustained effect on unifying
teachers into a cohesive inquiry community.
131

In addition, it is very interesting that the results reported in this study do not
provide convincing evidence to support the popular hypothesis that collaborative
teacher interaction is highly dependent on the degree to which the country has a
collectivist culture as opposed to an individualist culture (Hofstede, 2001; Leonard &
Leonard, 2003).57 When a range of other factors are simultaneously taken into account,
the effect of societal individualism does not remain statistically significant. This result
suggests that the popular view of the macro-culture of individualism in the United
States as one of the decisive factors leading to a low level of collaborative teacher
interaction in U.S. schools needs further investigation.

7.2. Implications for Educational Policy and Administration

A growing body of research suggests that teachers‟ learning can continue to be
generative over time if it is done in collaborative ways, as illustrated by “lesson study”
groups in Japan (Fernandez & Yoshida, 2004; Lewis, 2000) and school-based “teaching
research groups” in China (Han & Paine, 2010; Paine & Ma, 1993), for example.
57. It is important to note that this result does not preclude the possibility that teacher
interaction may differ qualitatively depending on varying individualist-collectivist cultural norms
across countries. For example, in their study of mentor-novice conversations, Wang, Strong, and
Odell (2004) observed important qualitative differences between Chinese and U.S. teachers in
terms of conversational foci and forms. However, their findings, based on only four Chinese
teachers and four U.S. teachers, do not allow us to determine whether their findings really reflect
some cultural differences or simply individual differences. Further systematic research is needed
to examine how various aspects of collaborative teacher interaction, such as the kinds of
information exchanged during collaborative interaction and the discursive strategies used to
exchange information, differ depending on national cultural contexts.
132

However, the paucity of U.S. schools engaged in meaningful collaborative work around
teaching has been well documented; many studies decry limited opportunities for
teachers to have supportive collegial relationships through which they can share their
knowledge about the practice of teaching with other teachers (Akiba & LeTendre, 2009;
Bryk & Schneider, 2002; Darling-Hammond, 2010; Paine, et al., 2003). The cross-national
data analyzed in this study also show that the United States, when compared to other
countries, is characterized as having a very high proportion of teachers who do not
engage in any type of collaborative teacher interaction while having a very low
proportion of teachers who regularly participate in multiple types of collaborative
teacher interaction.
The main findings from this study point to the possibility that teachers‟ exposure
to uncertainty in their classroom teaching is significantly related to their engagement in
collaborative interaction with other colleague teachers. It is important to note that
teachers‟ exposure to uncertainty in teaching is not necessarily negative; rather,
“uncertainty is an essential driving force in teaching” (Floden & Buchmann, 1993, p. 380)
because recognition of uncertainty in teaching makes teachers “stop and think and want
to find out more. . . . Being aware of the uncertainties [involved in] teaching . . . can be
an attitude towards the profession of teaching” (Munthe, 2007, p. 17). In this respect,
providing teachers with the necessary administrative and professional support for them
to become “professionally creative and autonomous” (Shulman, 2004, p. 151) enough to
develop and use instructional strategies to implement teaching for understanding will
help them better recognize and inquire into various kinds of instructional uncertainty.
133

This, in turn, is likely to contribute to an increase in teachers‟ motivation to engage in
collegial interaction with other teachers in order to collaboratively manage the resultant
uncertainty rather than simply trying to eliminate possible sources of uncertainty from
their classroom teaching.58
With regard to the case of the United States, the degree of emphasis teachers
place on teaching for understanding appears to have a strong positive relationship with
their engagement in collaborative interaction; further, its effect size is similar to or
greater than the effect size of school climate when school climate is operationalized as
teachers‟ perception of the degree to which teacher morale is high and organizational
goals are clearly communicated and well implemented. Depending on regression
models, the expected increase in collaborative teacher interaction ranges from about 11%
to 16% of one standard deviation for every one-standard-deviation increase in school
climate, while the increase ranges from about 15% to 22% for every one-standard58. In this respect, educational policy makers need to ruminate on what kinds of federal,
state, and district policies can create institutional environments where teachers are provided with
high-quality learning opportunities pertinent to instructional strategies to promote students’
engagement in inquiry-based learning and understanding. Indeed, policy discourses on school
curricula around the world are shifting their emphasis “from authoritative structures of fact and
skill and discipline to . . . broadened individual participation in . . . a rationalized knowledge
system that functions as an umbrella of understanding and comprehension” (McEneaney &
Meyer, 2000, p. 207). Sustained policy efforts to translate these policy discourses into
professional development opportunities would stimulate collaborative teacher interaction as a
collective effort to manage instructional uncertainty. Similarly, teacher preparation programs
could do more to prepare future teachers to become “scholar teachers” (Leung, 2001) as well as
competent practitioners, who have rigorous content knowledge in addition to extensive general
and pedagogy-related knowledge (Kim, et al., 2011); this would help teachers to become more
perceptive of instructional uncertainty, thereby leading to a greater degree of collaborative
interaction with other colleague teachers as a collective effort to make sense of, and lessen, their
feelings of discomfort with their perceived elements of uncertainty in teaching.
134

deviation increase in teachers‟ effort to implement teaching for understanding.
Compared to the large amount of research and literature about the relationship between
school climate and teacher collegiality (Hoy, 2002; Sergiovanni & Starratt, 2002;
Tschannen-Moran, 2001), the paucity of a similar kind of research focusing on
instructional uncertainty is very striking, given that its effect size appears to be no less
than that of school climate.
Also, the main findings from this study shed new light on the beneficial effect of
teacher collegiality. In addition to the widely acknowledged linkage of teacher
collegiality to their enhanced feelings of efficacy, increased teaching quality, and greater
accountability for student achievement (Bryk, et al., 1999; Lee & Smith, 1996; Newmann,
et al., 1989; Stevens & Slavin, 1995; Yasumoto, et al., 2001), the findings from this study
provide extensive evidence to support the possibility that teacher collegiality may arise
as a collective effort to manage uncertainty in teaching. Given the fact that it is not
always either possible or desirable to eliminate the sources of uncertainty in teaching
(Floden & Buchmann, 1993; Jackson, 1986; Lampert, 2001; Munthe, 2007; Shulman, 1987),
finding ways to help teachers manage the uncertainty through collaborative interaction
with their colleagues is an important task for educational administrators and policy
makers.
Finally, educational administrators need to ponder how to help teachers sustain
and expand already-existing informal network ties among teachers. As suggested by
the main findings from this study, teachers are likely to establish collegial networks
with other colleague teachers as a collective response to uncertainty in teaching. Given
135

the wide variety of possible situations in which teachers confront instructional and/or
classroom-contextual uncertainty, it is important for educational administrators to find
ways to actively share leadership roles and responsibilities with many teachers, so that
individual teachers can become empowered agents to foster and sustain a collegial
work environment in which they may “informally discuss problems they face, share
ideas, help one another in preparing lessons, exchange tips, and provide other support
to one another” (Sergiovanni & Starratt, 2002, p. 247).

7.3. Suggestions for Future Research

This study relies primarily on data from the TIMSS 2007 survey to introduce and
substantiate the uncertainty management perspective on teacher collegiality. One may
replicate similar analyses based on data from earlier and future TIMSS surveys to see to
what extent similar results are reproduced. Another useful source of data that can be
used to conduct related studies is the OECD Teaching and Learning International
Survey. A similar analysis based on this dataset would provide a valuable opportunity
to examine if the result patterns observed in the present study are corroborated in a
different international dataset as well. In addition, since this study focuses only on
eighth-grade mathematics teachers, a potentially fruitful area of research would also be
to explore how the relationships between various contextual factors and collaborative
teacher interaction patterns are dependent upon subject areas and grade and school
levels.
136

One methodological limitation of this study arises from the cross-sectional nature
of the TIMSS dataset, which makes causal inference difficult because it is usually not
possible to identify and measure all confounding variables and include each as a
covariate in a quantitative model (Frank, 2000; Schneider, Carnoy, Kilpatrick, Schmidt,
& Shavelson, 2007). Thus, the availability of reliable large-scale quantitative datasets
that can possibly be used to carry out longitudinal studies of teacher collegiality needs
to be scrutinized. In this respect, one possibility is to use the School and Staffing Survey
compiled by the National Center for Education Statistics. This U.S. dataset, as a
longitudinal one, contains some useful data on collaborative teacher interaction as well
as extensive information about individual teachers and their schools and districts.
Alternatively, one may want to collect a small-scale longitudinal dataset that is
specifically designed to examine the effect of various contextual factors on collaborative
teacher interaction.
Finally, future studies need to examine how collaborative teacher interaction
patterns differ qualitatively depending on various factors. Although this study uses
measures of collaborative teacher interaction constructed based on quantitative data
about how often teachers participate in certain types of interaction with other teachers,
these measures do not capture various qualitative aspects of teacher interaction.
Qualitative data on what kinds of information are actually exchanged during
collaborative teacher interaction and what specific discursive strategies are used to
exchange information, for example, would allow deeper analyses. Conceptual and
methodological issues concerning which qualitative aspects of teacher interaction to
137

observe and how to adequately measure those aspects need further systematic
exploration.

138

APPENDIX

139

APPENDIX
Additional Analyses

Table A1. Exploratory factor analysis of the level-1 primary independent variables: Orthogonal
rotation
Factor 1:
Uncertainty from
classroom-contextual factors
(UNC_CONTEXT)
EDCAPIT
ACHIEVE
EDCAVAR
ACHIVAR
SHORTRES
LANGMIN
TCHUNDS

Factor 2:
Uncertainty from
teaching for understanding
(UNC_TCHUNDS)

-.901
-.857
.856
.734
.544
.489
.159

.098
-.048
-.012
.165
-.073
-.480
.871

Note . Principal component extraction with varimax rotation was used. Factor loadings are sorted
by size. Each dotted-line box indicates the factor on which a given variable is loaded to a greater
degree. Bartlett’s test of sphericity is significant at the p < .001 level. n = 6,624.

Table A2. Pearson's bivariate correlations between level-2 variables (n = 29)
SOCHET
CURRSTD
INDVCUL
EDEXPN

CURRSTD
-.231

INDVCUL
-.458 **
-.358 +

+p ≤ .10; *p ≤ .05; **p ≤ .01.

140

EDEXPN
-.126
-.247
.526 **

GDPCAP
-.424 *
-.072
.481 **
.234

Table A3. Multilevel linear regressions for TCHRCOL_LESSON, with all slopes fixed at level 2
Fixed effect
Intercept, γ 00
SOCHET, γ 01
CURRSTD, γ 02
INDVCUL, γ 03
EDEXPN, γ 04
CURRSTD × EDEXPN, γ 05
UNC_TCHUNDS, γ 10
× CURRSTD, γ 11
× EDEXPN, γ 12
× CURRSTD × EDEXPN, γ 13
UNC_CONTEXT, γ 20
× CURRSTD, γ 21
× EDEXPN, γ 22
× CURRSTD × EDEXPN, γ 23
CLIMATE, γ 30
PRINLEAD, γ 40
WORKLOAD, γ 50
TCHREXP, γ 60
MATHMJR, γ 70
FEMALE, γ 80

i
Coeff. (SE )
1.400 (.044) ***
.056 (.021) **

.126 (.013) ***

-.014 (.024)

Coeff.
1.381
.043
.011
.030
-.128
.050
.113
.000
.009
-.009
.034
.003
.019
-.018
.164
.011
-.010
.005
.045
.044

ii
(SE )
(.038)
(.020)
(.023)
(.024)
(.054)
(.038)
(.014)
(.012)
(.013)
(.011)
(.024)
(.019)
(.027)
(.020)
(.030)
(.007)
(.013)
(.008)
(.034)
(.038)

***
*

*
***

***

Random effect
Var.
df
χ²
Var.
df
χ²
Intercept, u 0j
.052
27 645.4 *** .032
23 406.6 ***
Level-1 random error, r ij
.524
.510
Note . All independent variables are grand-mean-centered. Coeff. = Unstandardized coefficients
with robust standard errors in parentheses.
*p ≤ .05; **p ≤ .01; ***p ≤ .001.

141

Table A4. Multilevel linear regressions for TCHRCOL_CLASS, with all slopes fixed at level 2
Fixed effect
Intercept, γ 00
SOCHET, γ 01
CURRSTD, γ 02
INDVCUL, γ 03
EDEXPN, γ 04
CURRSTD × EDEXPN, γ 05
UNC_TCHUNDS, γ 10
× CURRSTD, γ 11
× EDEXPN, γ 12
× CURRSTD × EDEXPN, γ 13
UNC_CONTEXT, γ 20
× CURRSTD, γ 21
× EDEXPN, γ 22
× CURRSTD × EDEXPN, γ 23
CLIMATE, γ 30
PRINLEAD, γ 40
WORKLOAD, γ 50
TCHREXP, γ 60
MATHMJR, γ 70
FEMALE, γ 80

i
Coeff. (SE )
.479 (.049) ***
.029 (.025)

.065 (.010) ***

.076 (.017) ***

Coeff.
.466
.031
.101
.018
-.014
.070
.058
-.011
-.009
-.012
.106
-.001
-.009
-.024
.085
.014
.006
.012
.011
-.008

ii
(SE )
(.042)
(.022)
(.034)
(.022)
(.059)
(.038)
(.009)
(.006)
(.009)
(.007)
(.016)
(.008)
(.018)
(.013)
(.016)
(.007)
(.011)
(.006)
(.026)
(.020)

***
**

+
***
+
+
***

+
***
*
*

Random effect
Var.
df
χ²
Var.
df
χ²
Intercept, u 0j
.071
27 1483.7 *** .047
23 950.2 ***
Level-1 random error, r ij
.314
.307
Note . All independent variables are grand-mean-centered. Coeff. = Unstandardized coefficients
with robust standard errors in parentheses.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

142

Table A5. Multilevel ordinal logistic regressions for TCHRCOL_VARIETY, with all slopes
fixed at level 2
Fixed effect
Intercept, γ 00
SOCHET, γ 01
CURRSTD, γ 02
INDVCUL, γ 03
EDEXPN, γ 04
CURRSTD × EDEXPN, γ 05
UNC_TCHUNDS, γ 10
× CURRSTD, γ 11
× EDEXPN, γ 12
× CURRSTD × EDEXPN, γ 13
UNC_CONTEXT, γ 20
× CURRSTD, γ 21
× EDEXPN, γ 22
× CURRSTD × EDEXPN, γ 23
CLIMATE, γ 30
PRINLEAD, γ 40
WORKLOAD, γ 50
TCHREXP, γ 60
MATHMJR, γ 70
FEMALE, γ 80
Threshold difference, δ (2)

i
Coeff. (SE )
-.113 (.184)
.089 (.073)

OR
.893
1.093

Coeff.
-.154
.112
.345
.097
-.215
.258
.250 (.042) 1.283 *** .228
-.037
-.062
-.044
.231 (.067) 1.260 *** .371
-.003
.014
-.150
.430
.092
.047
.017
.128
-.062
3.240 (.126) 25.540 *** 3.277

ii
(SE )
OR
(.168)
.858
(.084) 1.118
(.116) 1.412
(.093) 1.101
(.214)
.806
(.138) 1.294
(.042) 1.256
(.032)
.963
(.033)
.940
(.030)
.957
(.067) 1.449
(.032)
.997
(.063) 1.014
(.035)
.860
(.070) 1.537
(.026) 1.096
(.041) 1.048
(.023) 1.017
(.076) 1.137
(.084)
.939
(.124) 26.500

**

+
***
+
***

***
***
***

+
***

Random effect
Var.
df
χ²
Var.
df
χ²
Intercept, u 0j
.832
27 1141.3 *** .597
23 747.4 ***
Note . All independent variables are grand-mean-centered. Coeff. = Unstandardized ordered logit
coefficients with robust standard errors in parentheses. OR = odds ratios.
+p ≤ .10; **p ≤ .01; ***p ≤ .001.

143

Table A6. Multilevel linear regressions for TCHRCOL, with all variables centered at their
country means

Fixed effect
Intercept, γ 00
UNC_TCHUNDS, γ 10
UNC_CONTEXT, γ 20
CLIMATE, γ 30
PRINLEAD, γ 40
WORKLOAD, γ 50
TCHREXP, γ 60
MATHMJR, γ 70
FEMALE, γ 80

Yij =
TCHRCOL_
LESSON
Coeff. (SE )
-.003 (.005)
.111 (.015) ***
.033 (.024)
.160 (.030) ***
.008 (.008)
-.007 (.012)
.005 (.009)
.054 (.035)
.065 (.040)

Yij =
TCHRCOL_
CLASS
Coeff. (SE )
-.011 (.006) +
.062 (.011) ***
.102 (.020) ***
.086 (.016) ***
.015 (.007) *
.006 (.013)
.013 (.006) *
.035 (.039)
.008 (.022)

Random effect
Var.
df
χ²
Var.
Intercept, u 0j
.000
28 12.0
.000
UNC_TCHUNDS, u 1j
.003
28 44.6 *
.001
UNC_CONTEXT, u 2j
.010
28 60.8 ***
.006
CLIMATE, u 3j
.013
28 65.2 ***
.002
PRINLEAD, u 4j
.001
28 43.2 *
.001
WORKLOAD, u 5j
.001
28 30.3
.002
TCHREXP, u 6j
.001
28 82.9 ***
.001
MATHMJR, u 7j
.009
28 32.8
.022
FEMALE, u 8j
.028
28 83.8 ***
.007
Level-1 random error, r ij
.486
.298
Note . All dependent and independent variables are country-mean-centered.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

144

df
28
28
28
28
28
28
28
28
28

χ²
17.4
44.1
60.7
36.9
43.3
50.3
60.9
63.6
42.7

*
***
*
**
***
***
*

Table A7. Ordinary least squares regressions for TCHRCOL_CLASS: The case of the United
States: Additional models
i-a
B (SE )
.194 (.044)
-.166 (.044)

β
.224 ***
-.199 ***

i-b
B (SE )
.138 (.044)
-.092 (.049)

β
.163 **
-.115 +

TCHUNDS
ACHIEVE
ACHIVAR
EDCAPIT
EDCAVAR
LANGMIN
SHORTRES
CLIMATE
.084 (.040)
.105 *
.102 (.040)
.131
PRINLEAD
.003 (.014)
.011
.001 (.014)
.003
WORKLOAD
-.024 (.023) -.050
-.025 (.024) -.054
TCHREXP
.004 (.012)
.015
-.001 (.012) -.003
MATHMJR
-.011 (.054) -.010
-.024 (.054) -.023
FEMALE
.029 (.055)
.026
.084 (.055)
.077
Course taught
Basic math (ref.)
Algebra
.027 (.061)
.023
.002 (.061)
.002
Geometry
.809 (.197)
.211 ***
.739 (.195)
.204
Others
.429 (.151)
.145 **
.365 (.153)
.125
School type
Public (ref.)
Private, religious
-.063 (.085) -.038
-.040 (.092) -.024
Private, others
.253 (.126)
.100 *
.212 (.194)
.055
Free or reduced-price lunch
.024 (.011)
.140
Intercept
.395 (.268)
-.047 (.310)
R²
.16
.14
n
393
378
Note . B = Unstandardized coefficients with standard errors in parentheses. β = standardized
coefficients. Ref. = reference category.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

145

**

***
*

*

Table A7. (cont'd)

TCHUNDS
ACHIEVE
ACHIVAR
EDCAPIT
EDCAVAR
LANGMIN
SHORTRES
CLIMATE
PRINLEAD
WORKLOAD
TCHREXP
MATHMJR
FEMALE
Course taught
Basic math (ref.)
Algebra
Geometry
Others
School type
Public (ref.)
Private, religious
Private, others
Free or reduced-price lunch
Intercept
R²
n

ii-a
B (SE )
.178 (.044)
.173 (.086)

.075
.011
-.015
.005
-.036
.035

(.040)
(.014)
(.024)
(.012)
(.054)
(.056)

β
.205 ***

ii-b
B (SE )
.128 (.044)

β
.151 **

.101 *

.159 (.091)

.093 +

.094 +
.040
-.032
.023
-.033
.030

.103
.004
-.023
.000
-.038
.092

-.006 (.061)
.681 (.197)
.387 (.153)

-.005
.178 ***
.131 **

-.015 (.060)
.712 (.192)
.345 (.153)

-.099 (.085)
.253 (.128)

-.059
.099 *

-.045
.205
.029
-.667

-.554 (.207) **
.13
393

146

(.040)
(.014)
(.023)
(.012)
(.054)
(.055)

.132 **
.013
-.051
.000
-.036
.084 +

-.014
.197 ***
.118 *

(.092) -.027
(.194)
.053
(.010)
.168 **
(.207) ***
.13
378

Table A7. (cont'd)

TCHUNDS
ACHIEVE
ACHIVAR
EDCAPIT
EDCAVAR
LANGMIN
SHORTRES
CLIMATE
PRINLEAD
WORKLOAD
TCHREXP
MATHMJR
FEMALE
Course taught
Basic math (ref.)
Algebra
Geometry
Others
School type
Public (ref.)
Private, religious
Private, others
Free or reduced-price lunch
Intercept
R²
n

iii-a
B (SE )
.182 (.043)

β
.210 ***

iii-b
B (SE )
.135 (.044)

-.096 (.027)

-.201 ***

-.070 (.034)

-.148 *

.084
.002
-.019
.002
-.003
.036

.106 *
.006
-.040
.009
-.003
.032

.099
.000
-.022
-.001
-.018
.089

.128 **
.001
-.048
-.003
-.017
.081

(.040)
(.015)
(.023)
(.012)
(.054)
(.056)

(.040)
(.014)
(.023)
(.012)
(.054)
(.055)

β
.160 **

.001 (.061)
.681 (.198)
.421 (.151)

.001
.171 ***
.143 **

-.007 (.060)
.679 (.195)
.366 (.153)

-.006
.180 ***
.125 *

.000 (.089)
.372 (.131)

.000
.147 **

-.006
.264
.017
.054

-.003
.069
.102

.300 (.258)
.15
392

147

(.093)
(.195)
(.012)
(.328)
.13
377

Table A7. (cont'd)

TCHUNDS
ACHIEVE
ACHIVAR
EDCAPIT
EDCAVAR
LANGMIN
SHORTRES
CLIMATE
PRINLEAD
WORKLOAD
TCHREXP
MATHMJR
FEMALE
Course taught
Basic math (ref.)
Algebra
Geometry
Others
School type
Public (ref.)
Private, religious
Private, others
Free or reduced-price lunch
Intercept
R²
n

iv-a
B (SE )
.177 (.044)

.071 (.031)

.074
.010
-.017
.004
-.022
.031

(.040)
(.014)
(.024)
(.012)
(.055)
(.056)

β
.205 ***

iv-b
B (SE )
.131 (.044)

β
.154 **

.122 *

.062 (.034)

.107 +

.093 +
.034
-.037
.018
-.020
.027

.098
.003
-.024
.000
-.026
.089

-.010 (.061)
.652 (.201)
.383 (.152)

-.008
.164 ***
.130 **

-.017 (.060)
.675 (.195)
.339 (.153)

-.052 (.089)
.300 (.129)

-.031
.119 *

-.013
.248
.025
-.623

-.532 (.198) **
.14
392

148

(.040)
(.014)
(.024)
(.012)
(.054)
(.055)

.126 *
.012
-.052
.001
-.025
.081

-.016
.179 ***
.116 *

(.093) -.008
(.195)
.065
(.010)
.148 **
(.197) **
.13
377

Table A7. (cont'd)

TCHUNDS
ACHIEVE
ACHIVAR
EDCAPIT
EDCAVAR
LANGMIN
SHORTRES
CLIMATE
PRINLEAD
WORKLOAD
TCHREXP
MATHMJR
FEMALE
Course taught
Basic math (ref.)
Algebra
Geometry
Others
School type
Public (ref.)
Private, religious
Private, others
Free or reduced-price lunch
Intercept
R²
n

v-a
B (SE )
.166 (.044)

β
.193 ***

.027 (.019)

.069

v-b
B (SE )
.124 (.044)

β
.146 **

-.013 (.021)

-.034

.089 +
.048
-.026
.021
-.029
.034

.098
.004
-.018
-.002
-.037
.090

.127 *
.014
-.039
-.009
-.035
.082

-.011 (.061)
.591 (.199)
.405 (.154)

-.010
.148 **
.137 **

-.029 (.060)
.596 (.193)
.345 (.154)

-.103 (.085)
.256 (.128)

-.061
.101 *

-.031
.223
.036
-.503

.070
.014
-.012
.005
-.033
.038

(.040)
(.014)
(.024)
(.012)
(.055)
(.056)

-.394 (.185) *
.13
392

149

(.040)
(.014)
(.024)
(.012)
(.054)
(.055)

-.026
.158 **
.118 *

(.093) -.019
(.195)
.058
(.010)
.208 ***
(.187) **
.12
377

Table A7. (cont'd)

TCHUNDS
ACHIEVE
ACHIVAR
EDCAPIT
EDCAVAR
LANGMIN
SHORTRES
CLIMATE
PRINLEAD
WORKLOAD
TCHREXP
MATHMJR
FEMALE
Course taught
Basic math (ref.)
Algebra
Geometry
Others
School type
Public (ref.)
Private, religious
Private, others
Free or reduced-price lunch
Intercept
R²
n

B
.168

vi-a
(SE )
.044

.130
.095
.012
-.015
.006
-.047
.029

.048
.041
.014
.023
.012
.054
.056

.136 **
.119 *
.042
-.031
.025
-.042
.026

.131
.121
.003
-.024
.000
-.045
.088

.048
.041
.014
.023
.012
.053
.055

.142 **
.156 **
.011
-.052
.002
-.042
.081

-.012
.551
.321

.061
.189
.154

-.011
.144 **
.108 *

-.018
.591
.274

.059
.183
.155

-.016
.163 ***
.093 +

-.103
.274

.085
.128

-.062
.108 *

-.492

.188 **
.14

-.031
.227
.029
-.615

393

150

β
.194 ***

B
.120

vi-b
(SE )
.043

β
.141 **

.092 -.019
.193
.059
.010
.170 **
.189 ***
.14
378

Table A8. Within-nation regressions for TCHRCOL_LESSON: Ordinary least squares regression
coefficients for UNC_TCHUNDS and UNC_CONTEXT, after controlling for all other level-1
variables
UNC_TCHUNDS
UNC_CONTEXT
B
(SE )
β
B
(SE )
β
Australia
-.010 (.056) -.013
.130 (.115) .088
Bulgaria
.029 (.068) .030
.187 (.072) .196 **
Chinese Taipei
.140 (.074) .166 +
-.119 (.096) -.108
Colombia
.142 (.110) .121
-.119 (.118) -.099
Czech Rep.
.155 (.069) .165 *
.147 (.140) .079
Egypt
.164 (.057) .204 **
-.184 (.082) -.162 *
England
.234 (.081) .224 **
-.164 (.098) -.134 +
Ghana
-.011 (.081) -.012
-.145 (.117) -.119
Hong Kong
.021 (.100) .024
.004 (.114) .004
Hungary
.048 (.069) .045
-.031 (.086) -.024
Indonesia
.139 (.060) .203 *
.024 (.109) .019
Iran
.181 (.046) .278 ***
.049 (.066) .058
Israel
.155 (.050) .190 **
.371 (.082) .309 ***
Italy
-.031 (.058) -.034
.201 (.112) .114 +
Japan
-.036 (.071) -.036
.006 (.150) .003
Kuwait
.018 (.084) .026
-.091 (.233) -.053
Lebanon
.032 (.115) .029
.159 (.164) .102
Malaysia
.098 (.057) .147 +
.097 (.098) .082
Norway
.144 (.069) .155 *
.131 (.172) .056
Romania
.110 (.061) .119 +
.032 (.070) .032
Russian Fed.
.269 (.067) .246 ***
-.009 (.078) -.007
Saudi Arabia
.144 (.068) .194 *
-.040 (.127) -.028
Scotland
.145 (.064) .155 *
.382 (.095) .259 ***
Singapore
.140 (.049) .176 **
.019 (.068) .018
South Korea
.122 (.063) .134 *
.216 (.154) .098
Sweden
.056 (.049) .060
-.061 (.124) -.026
Thailand
.108 (.053) .172 *
.102 (.068) .128
Turkey
.198 (.069) .235 **
.124 (.098) .119
United States
.205 (.052) .195 ***
.180 (.077) .119 *
Note . B = Unstandardized coefficients with standard errors in parentheses. β = standardized
coefficients. Independent variables included in the model are CLIMATE, PRINLEAD,
WORKLOAD, TCHREXP, MATHMJR, and FEMALE, in addition to UNC_TCHUNDS and
UNC_CONTEXT. Each row is a separate analysis.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

151

n
210
210
139
117
185
187
184
139
102
244
129
195
243
277
207
82
105
159
188
227
264
124
211
314
222
390
145
128
392

Table A9. Within-nation regressions for TCHRCOL_CLASS: Ordinary least squares regression
coefficients for UNC_TCHUNDS and UNC_CONTEXT, after controlling for all other level-1
variables
UNC_TCHUNDS
UNC_CONTEXT
B
(SE )
β
B
(SE )
β
Australia
.085 (.044) .137 *
.135 (.089) .117
Bulgaria
.031 (.032) .065
.060 (.034) .131 +
Chinese Taipei
.008 (.054) .014
.047 (.070) .061
Colombia
.159 (.091) .171 +
.093 (.098) .099
Czech Rep.
.013 (.035) .027
-.003 (.071) -.003
Egypt
.011 (.057) .015
.109 (.082) .104
England
.165 (.066) .194 **
.122 (.079) .124
Ghana
-.020 (.077) -.023
.110 (.111) .095
Hong Kong
.241 (.083) .315 **
.047 (.094) .053
Hungary
.034 (.041) .055
-.003 (.052) -.004
Indonesia
-.013 (.062) -.019
.186 (.113) .153 +
Iran
.054 (.030) .131 +
.151 (.044) .284 ***
Israel
-.007 (.034) -.013
.190 (.055) .244 ***
Italy
.028 (.036) .049
-.052 (.069) -.047
Japan
.082 (.070) .083
.503 (.146) .242 ***
Kuwait
.026 (.057) .058
-.017 (.158) -.015
Lebanon
.091 (.092) .103
.427 (.127) .346 ***
Malaysia
.125 (.046) .232 **
.155 (.078) .162 *
Norway
.080 (.081) .073
.014 (.202) .005
Romania
.052 (.054) .062
.151 (.062) .167 *
Russian Fed.
.156 (.049) .199 **
.062 (.058) .073
Saudi Arabia
.007 (.062) .012
-.148 (.115) -.125
Scotland
.196 (.058) .250 ***
.205 (.086) .166 *
Singapore
.032 (.036) .057
-.018 (.050) -.024
South Korea
.133 (.034) .259 ***
-.098 (.082) -.079
Sweden
-.004 (.028) -.007
.035 (.070) .027
Thailand
.064 (.051) .111
.161 (.066) .221 *
Turkey
.101 (.049) .175 *
.193 (.070) .272 **
United States
.140 (.029) .237 ***
.150 (.044) .177 ***
Note . B = Unstandardized coefficients with standard errors in parentheses. β = standardized
coefficients. Independent variables included in the model are CLIMATE, PRINLEAD,
WORKLOAD, TCHREXP, MATHMJR, and FEMALE, in addition to UNC_TCHUNDS and
UNC_CONTEXT. Each row is a separate analysis.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

152

n
208
208
137
117
185
187
182
139
98
239
129
194
240
277
207
81
104
159
188
227
262
124
210
314
222
389
145
128
392

Table A10. Within-nation regressions for TCHRCOL_VARIETY: Ordinal logistic regression
coefficients for UNC_TCHUNDS and UNC_CONTEXT, after controlling for all other level-1
variables
UNC_TCHUNDS
UNC_CONTEXT
B
(SE )
OR
B
(SE )
OR
n
Australia
.116 (.168) 1.123
.289 (.343) 1.335
208
Bulgaria
.146 (.173) 1.157
.346 (.183) 1.414 +
208
Chinese Taipei
.157 (.235) 1.171
.001 (.308) 1.001
137
Colombia
.714 (.313) 2.042 *
.105 (.323) 1.111
117
Czech Rep.
.319 (.237) 1.375
.462 (.480) 1.588
185
Egypt
-.176 (.239) .839
.294 (.330) 1.342
187
England
.473 (.232) 1.606 *
.417 (.278) 1.517
182
Ghana
-.203 (.250) .817
-.319 (.360) .727
139
Hong Kong
.731 (.401) 2.076 +
.590 (.441) 1.804
98
Hungary
.071 (.200) 1.073
.125 (.253) 1.133
239
Indonesia
.142 (.180) 1.152
.361 (.324) 1.435
129
Iran
.060 (.156) 1.061
.581 (.233) 1.788 **
194
Israel
.059 (.198) 1.061
1.543 (.336) 4.677 ***
240
Italy
-.139 (.149) .870
-.050 (.288) .951
277
Japan
.414 (.205) 1.512 *
1.448 (.448) 4.256 ***
207
Kuwait
.290 (.390) 1.336
-1.295 (1.112) .274
81
Lebanon
.403 (.308) 1.496
.744 (.424) 2.105 +
104
Malaysia
.455 (.200) 1.576 *
1.033 (.358) 2.810 **
159
Norway
.406 (.212) 1.502 +
.224 (.518) 1.251
188
Romania
-.154 (.236) .857
.900 (.287) 2.458 **
227
Russian Fed.
.935 (.310) 2.547 **
.435 (.388) 1.546
261
Saudi Arabia
-.104 (.203) .901
-.467 (.378) .627
124
Scotland
.346 (.183) 1.414 +
.707 (.277) 2.027 **
210
Singapore
.354 (.149) 1.425 *
.099 (.207) 1.104
313
South Korea
.915 (.228) 2.498 ***
.147 (.525) 1.158
222
Sweden
.124 (.143) 1.132
-.328 (.366) .721
389
Thailand
.474 (.215) 1.606 *
.579 (.294) 1.784 *
145
Turkey
.405 (.240) 1.500 +
1.197 (.360) 3.309 ***
128
United States
.565 (.123) 1.759 ***
.654 (.182) 1.924 ***
392
Note . B = Unstandardized ordered logit coefficients with standard errors in parentheses. OR =
odds ratios. Independent variables included in the model are CLIMATE, PRINLEAD,
WORKLOAD, TCHREXP, MATHMJR, and FEMALE, in addition to UNC_TCHUNDS and
UNC_CONTEXT. Each row is a separate analysis.
+p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001.

153

Table A11. Summary of results, by hypothesis

H-1
H-2
H-3
H-A1
H-A2
H-A3

Hypothesized
relationship w/
collaborative
interaction
Positive
Positive
Positive
Positive
Positive
Positive
Positive
Negative
Negative
Negative

Uncertainty (from teaching for understanding)
Uncertainty (from classroom-contextual factors)
Socio-cultural heterogeneity
School climate
Principal’s instructional leadership
Curriculum standardization
× Public expenditure on education
H-A4 Uncertainty × Curriculum standardization
× Public expenditure on education
H-A5 Individualist culture
Note . Results from U.S.-only data are in parentheses.
Source . Tables 4-2, 4-3, 4-5, 4-6, 4-7, 5-1, 5-2, 5-3, 6-1, 6-2, and 6-3.

154

Num. of
effects
12
12
15
15
15
12
3
6
6
12

Results
Statistically significant
Positive
Negative
12 (6)
0 (0)
7 (3)
0 (0)
9
0
15 (12)
0 (0)
2 (0)
0 (0)
9
0
1
0
0
1
0
3
0
4

Statistically
insignificant
0 (0)
5 (3)
6
0 (0)
13 (12)
3
2
5
3
8

BIBLIOGRAPHY

155

BIBLIOGRAPHY

Achinstein, B. (2002). Conflict amid community: The micropolitics of teacher
collaboration. Teachers College Record, 104(3), 421-455.
Akiba, M., & LeTendre, G. K. (2009). Improving teacher quality: The U.S. teaching force in
global context. New York, NY: Teachers College Press.
Alesina, A., Devleeschauwer, A., Easterly, W., Kurlat, S., & Wacziarg, R. (2003).
Fractionalization. Journal of Economic Growth, 8(2), 155-194.
Alexander, R. (2000). Culture and pedagogy: International comparisons in primary education.
Oxford, UK: Blackwell.
Alexander, R. (2001). Border crossings: Towards a comparative pedagogy. Comparative
Education, 37(4), 507-523.
Anagnostopoulos, D., & Rutledge, S. A. (2007). Making sense of school sanctioning
policies in urban high schools: Charting the depth and drift of school and
classroom change. Teachers College Record, 109(5), 1261-1302.
Anderson-Levitt, K. M. (2002). Teaching culture as national and transnational: A
response to Teachers' Work. Educational Researcher, 31(3), 19-21.
Anderson-Levitt, K. M. (Ed.). (2003). Local meanings, global schooling: Anthropology and
world culture theory. New York, NY: Palgrave Macmillan.
Anderson, C. S. (1982). The search for school climate: A review of the research. Review of
Educational Research, 52(3), 368-420.
Anyon, J. (2001). Inner cities, affluent suburbs, and unequal educational opportunities.
In J. A. Banks & C. A. M. Banks (Eds.), Multicultural education: Issues and
perspectives (4th ed., pp. 85-98). New York, NY: John Wiley and Sons.
Aslam, M., & Kingdon, G. (2011). What can teachers do to raise pupil achievement?
Economics of Education Review, 30(3), 559-574.
Astiz, M. F., Wiseman, A. W., & Baker, D. P. (2002). Slouching towards decentralization:
Consequences of globalization for curricular control in national education
systems. Comparative Education Review, 46(1), 66-88.
Baker, D. P., & LeTendre, G. K. (2005). National differences, global similarities: World
culture and the future of schooling. Stanford, CA: Stanford University Press.

156

Ball, D. L., Thames, M. H., & Phelps, G. (2008). Content knowledge for teaching: What
makes it special? Journal of Teacher Education, 59(5), 389-407.
Benavot, A., Cha, Y.-K., Kamens, D. H., Meyer, J. W., & Wong, S.-Y. (1991). Knowledge
for the masses: world models and national curricula, 1920-1986. American
Sociological Review, 56(1), 85-100.
Berendt, E. A. (Ed.). (2008). Metaphors for learning: Cross-cultural perspectives. Amsterdam,
Netherlands: John Benjamins.
Berger, C. R. (1995). Inscrutable goals, uncertain plans, and the production of
communicative action. In C. R. Berger & M. Burgoon (Eds.), Communication and
social influence processes (pp. 1-28). East Lansing, MI: Michigan State University
Press.
Bickel, R. (2007). Multilevel analysis for applied research. New York, NY: Guilford Press.
Biddle, J., & Berliner, D. (2002). Unequal school funding in the United States. Educational
Leadership, 59(8), 48-59.
Bidwell, C. E. (2001). Analyzing schools as organizations: Long-term permanence and
short-term change. Sociology of Education, 74(Extra Issue), 100-114.
Blase, J. (1987). Dimensions of effective school leadership: The teacher's perspective.
American Educational Research Journal, 24(4), 589-610.
Blase, J., & Blase, J. (1999). Principals‟ instructional leadership and teacher development:
Teachers‟ perspectives. Educational Administration Quarterly, 35(3), 349-378.
Blömeke, S., & Paine, L. W. (2008). Getting the fish out of the water: Considering
benefits and problems of doing research on teacher education at an international
level. Teaching and Teacher Education, 24(8), 2027-2037.
Boli, J., & Ramirez, F. O. (1986). World culture and the institutional development of
mass education. In J. G. Richardson (Ed.), Handbook of theory and research for the
sociology of education (pp. 65-92). New York, NY: Greenwood Press.
Brashers, D. E. (2001). Communication and uncertainty management. Journal of
Communication, 51(3), 477-497.
Bray, M., Adamson, B., & Mason, M. (Eds.). (2007). Comparative education research:
Approaches and methods. Hong Kong, China: CERC-Springer.

157

Britton, E., Paine, L. W., Pimm, D., & Raizen, S. (Eds.). (2003). Comprehensive teacher
induction: Systems for early career learning. Dordrecht, Netherlands: Kluwer
Academic.
Bryk, A. S., Camburn, E. M., & Louis, K. S. (1999). Professional community in Chicago
elementary schools: Facilitating factors and organizational consequences.
Educational Administration Quarterly, 35(5), 751-781.
Bryk, A. S., & Schneider, B. (2002). Trust in schools: A core resource for improvement. New
York, NY: Russell Sage.
Cha, Y.-K., & Ham, S.-H. (2011). Educating supranational citizens: The incorporation of
English language education into curriculum policies. American Journal of
Education, 117(2), 183-209.
Chiu, M. M., Chow, B. W.-Y., & Mcbride-Chang, C. (2007). Universals and specifics in
learning strategies: Explaining adolescent mathematics, science, and reading
achievement across 34 countries. Learning and Individual Differences, 17(4), 344-365.
Chudgar, A., & Luschei, T. F. (2009). National income, income inequality, and the
importance of schools: A hierarchical cross-national comparison. American
Educational Research Journal, 46(3), 626-658.
Clandinin, D. J. (1985). Personal practical knowledge: A study of teachers' classroom
images. Curriculum Inquiry, 15(4), 361-385.
Coburn, C. E. (2001). Collective sensemaking about reading: How teachers mediate
reading policy in their professional communities. Educational Evaluation and
Policy Analysis, 23(2), 145-170.
Cochran-Smith, M., & Lytle, S. L. (1999). Relationships of knowledge and practice:
Teacher learning in communities. Review of Research in Education, 24, 249-305.
Cogan, M., Anderson, R. H., & Krajewski, R. (1993). Clinical supervision: Special methods
for the supervision of teachers (3rd ed.). Fort Worth, TX: Harcourt Brace.
Cohen, D. K., McLaughlin, M. W., & Talbert, J. E. (Eds.). (1993). Teaching for
understanding: Challenges for policy and practice. San Francisco, CA: Jossey-Bass.
Cohen, D. K., Raudenbush, S. W., & Ball, D. L. (2003). Resources, instruction, and
research. Educational Evaluation and Policy Analysis, 25(2), 119-142.
Cohen, E. G., Deal, T. E., Meyer, J. W., & Scott, W. R. (1979). Technology and teaming in
the elementary school. Sociology of Education, 52(1), 20-33.

158

Collinson, V., & Cook, T. F. (2007). Organizational learning: Improving learning, teaching,
and leading in school systems. Thousand Oaks, CA: Sage.
Cosner, S. (2009). Building organizational capacity through trust. Educational
Administration Quarterly, 45(2), 248-291.
Crossley, M., & Watson, K. (2003). Comparative and international research in education:
Globalisation, context and difference. London, UK: Routledge Falmer.
Cusick, P. A. (1983). The egalitarian ideal and the American high school. New York, NY:
Longman.
Darling-Hammond, L. (2010). The flat world and education: How America's commitment to
equity will determine our future. New York, NY: Teachers College Press.
de Lima, J. A. (2001). Forgetting about friendship: Using conflict in teacher communities
as a catalyst for school change. Journal of Educational Change, 2(2), 97-122.
Deal, T. E., & Celotti, L. D. (1980). How much influence do (and can) education
administrators have on classrooms? Phi Delta Kappan, 61(7), 472-473.
DeLany, B. (1991). Allocation, choice, and stratification within high schools: How the
sorting machine copes. American Journal of Education, 99(2), 181-207.
DiMaggio, P. J., & Powell, W. W. (1991). The iron cage revisited: Institutional
isomorphism and collective rationality in organizational fields. In W. W. Powell
& P. J. DiMaggio (Eds.), The new institutionalism in organizational analysis (pp. 6382). Chicago, IL: University of Chicago Press.
DiPaola, M. E., & Hoy, W. K. (2007). Principals improving instruction: Supervision,
evaluation, and professional development. Boston, MA: Pearson.
Dunne, F., Nave, B., & Lewis, A. (2000). Critical friends groups: Teachers helping
teachers to improve student learning. Phi Delta Kappa International Research
Bulletin, 28, 9-12.
Elmore, R. F. (2004). School reform from the inside out: Policy, practice, and performance.
Cambridge, MA: Harvard Education Press.
Fallon, G., & Barnett, J. (2009). Impacts of school organizational restructuring into a
collaborative setting on the nature of emerging forms of collegiality. International
Journal of Education Policy and Leadership, 4(9). Retrieved from
http://www.ijepl.org

159

Fennema, E., & Romberg, T. A. (Eds.). (1999). Mathematics classrooms that promote
understanding. Mahwah, NJ: Lawrence Erlbaum.
Fernandez, C., & Yoshida, M. (2004). Lesson study: A Japanese approach to improving
mathematics teaching and learning. Mahwah, NJ: Lawrence Erlbaum.
Floden, R. E., & Buchmann, M. (1993). Between routines and anarchy: Preparing
teachers for uncertainty. Oxford Review of Education, 19(3), 373-382.
Floden, R. E., & Clark, C. M. (1988). Preparing teachers for uncertainty. Teachers College
Record, 89(4), 505-524.
Foy, P., & Olson, J. F. (Eds.). (2009). TIMSS 2007 user guide for the international database.
Chestnut Hill, MA: TIMSS & PIRLS International Study Center, Boston College.
Frank, K. A. (2000). Impact of a confounding variable on a regression coefficient.
Sociological Methods and Research, 29(2), 147-194.
Frank, K. A., Zhao, Y., & Borman, K. (2004). Social capital and the diffusion of
innovations within organizations: The case of computer technology in schools.
Sociology of Education, 77(2), 148-171.
Freiberg, H. J. (1998). Measuring school climate: Let me count the ways. Educational
Leadership, 56(1), 22-26.
Freiberg, H. J. (Ed.). (1999). School climate: Measuring, improving and sustaining healthy
learning environments. London, UK: RoutledgeFalmer.
Fullan, M. (1993). Change forces: Probing the depths of educational reform. London, UK:
Falmer Press.
Fullan, M., & Hargreaves, A. (1996). What's worth fighting for in your school? New York,
NY: Teachers College Press.
Galbraith, J. R. (1973). Designing complex organizations. Reading, MA: Addison-Wesley.
Galbraith, J. R., Downey, D., & Kates, A. (2002). How networks undergird the lateral
capability of an organization: Where the work gets done. Journal of Organizational
Excellence, 21(2), 67-78.
Gamoran, A., Secada, W., & Marrett, C. (2000). The organizational context of teaching
and learning. In M. T. Hallinan (Ed.), Handbook of the sociology of education (pp. 3763). New York, NY: Kluwer Academic.
Gay, G. (2000). Culturally responsive teaching: Theory, research, and practice. New York, NY:
Teachers College Press.
160

Gilligan, C. (1993). In a different voice: Psychological theory and women's development.
Cambridge, MA: Harvard University Press.
Givvin, K. B., Hiebert, J., Jacobs, J. K., Hollingsworth, H., & Gallimore, R. (2005). Are
there national patterns of teaching? Evidence from the TIMSS 1999 video study.
Comparative Education Review, 49(3), 311-343.
Green, T. F. (1994). Policy questions: A conceptual study. Education Policy Analysis
Archives, 2(7). Retrieved from http://epaa.asu.edu/ojs/article/view/670
Grimmett, P. P., & MacKinnon, A. M. (1992). Craft knowledge and the education of
teachers. Review of Research in Education, 18, 385-456.
Grossman, P., Wineburg, S., & Woolworth, S. (2001). Toward a theory of teacher
community. Teachers College Record, 103(6), 942-1012.
Hallinan, M. T. (1994). Tracking: From theory to practice. Sociology of Education, 67(2),
79-84.
Hallinger, P., & Murphy, J. (1987). Assessing and developing principal instructional
leadership. Educational Leadership, 45(1), 54-62.
Ham, S.-H., & Cha, Y.-K. (2009). Positioning education in the information society: The
transnational diffusion of the information and communication technology
curriculum. Comparative Education Review, 53(4), 535-557.
Ham, S.-H., Paine, L. W., & Cha, Y.-K. (2011). Duality of educational policy as global
and local: The case of the gender equity agenda in national principles and state
actions. Asia Pacific Education Review, 12(1), 105-115.
Han, X., & Paine, L. W. (2010). Teaching mathematics as deliberate practice through
public lessons. Elementary School Journal, 110(4), 519-541.
Hanushek, E., & Rivkin, S. G. (2006). Teacher quality. In E. Hanushek & F. Welch (Eds.),
Handbook of the economics of education (Vol. 2, pp. 1051-1078). Amsterdam,
Netherlands: North-Holland.
Hargreaves, A. (1989). Curriculum and assessment reform. Toronto, ON: Ontario Institute
for Studies in Education Press.
Hargreaves, A. (1991). Contrived collegiality: The micropolitics of teacher collaboration.
In J. Blase (Ed.), The politics of life in schools (pp. 47-72). Newbury Park, CA: Sage.
Heaton, R. M. (2000). Teaching mathematics to the new standards: Relearning the dance. New
York, NY: Teachers College Press.
161

Hiebert, J., & Carpenter, T. P. (1992). Learning and teaching with understanding. In D.
A. Grouws (Ed.), Handbook of research on mathematics teaching and learning (pp. 6597). New York, NY: Macmillan.
Hiebert, J., Gallimore, R., Garnier, H., Givvin, K. B., Hollingsworth, H., Jacobs, J. K., . . .
Stigler, J. W. (2003). Teaching mathematics in seven countries: Results from the TIMSS
1999 video study. Washington, DC: National Center for Education Statistics.
Hill, H. C. (2001). Policy is not enough: Language and the interpretation of state
standards. American Educational Research Journal, 38(2), 289-318.
Hofstede, G. (2001). Culture's consequences: Comparing values, behaviors, institutions, and
organizations across nations (2nd ed.). Thousand Oaks, CA: Sage.
Hofstede, G., & Bond, M. H. (1984). Hofstede's culture dimensions: An independent
validation using Rokeach's value survey. Journal of Cross-Cultural Psychology,
15(4), 417-433.
Hofstede, G., & Hofstede, G. J. (2005). Cultures and organizations: Software of the mind
(2nd ed.). New York, NY: McGraw-Hill.
Horng, E. L., Klasik, D., & Loeb, S. (2010). Principal's time use and school effectiveness.
American Journal of Education, 116(4), 491–523.
Hoy, W. K. (2002). Faculty trust. Journal of School Public Relations, 23(2), 88-103.
Hoy, W. K., & Miskel, C. G. (2005). Educational administration: Theory, research, and
practice (7th ed.). New York, NY: McGraw-Hill.
Hoy, W. K., Tarter, C. J., & Bliss, J. R. (1990). Organizational climate, school health, and
effectiveness: A comparative analysis. Educational Administration Quarterly, 26(3),
260-279.
Hutchins, E. (1995). Cognition in the wild. Cambridge, MA: MIT Press.
Jackson, P. W. (1986). The practice of teaching. New York, NY: Teachers College Press.
Joncas, M. (2008a). TIMSS 2007 sample design. In J. F. Olson, M. O. Martin & I. V. S.
Mullis (Eds.), TIMSS 2007 technical report (pp. 77-92). Chestnut Hill, MA: TIMSS
& PIRLS International Study Center, Boston College.
Joncas, M. (2008b). TIMSS 2007 sampling weights and participation rates. In J. F. Olson,
M. O. Martin & I. V. S. Mullis (Eds.), TIMSS 2007 technical report (pp. 153-192).
Chestnut Hill, MA: TIMSS & PIRLS International Study Center, Boston College.

162

Kazemi, E., & Franke, M. L. (2004). Teacher learning in mathematics: Using student
work to promote collective inquiry. Journal of Mathematics Teacher Education, 7(3),
203-235.
Kennedy, M. M. (2005). Inside teaching: How classroom life undermines reform. Cambridge,
MA: Harvard University Press.
Kennedy, M. M., Ahn, S., & Choi, J. (2008). The value added by teacher education. In M.
Cochran-Smith, S. Feiman-Nemser & J. McIntyre (Eds.), Handbook of research on
teacher education: Enduring issues in changing contexts (3rd ed., pp. 1249-1273).
Mahwah, NJ: Lawrence Erlbaum.
Kim, R. Y., Ham, S.-H., & Paine, L. W. (2011). Knowledge expectations in mathematics
teacher preparation programs in South Korea and the United States: Towards
international dialogue. Journal of Teacher Education, 62(1), 48-61.
Labaree, D. F. (2000). On the nature of teaching and teacher education: Difficult
practices that look easy. Journal of Teacher Education, 51(3), 228-233.
Ladson-Billings, G. (1999). Preparing teachers for diversity: Historical perspectives,
current trends, and future directions. In L. Darling-Hammond & G. Sykes (Eds.),
Teaching as the learning profession: Handbook of policy and practice (pp. 86-123). San
Francisco, CA: Jossey-Bass.
Lampert, M. (2001). Teaching problems and the problems of teaching. New Haven, CT: Yale
University Press.
Lassonde, C. A., & Israel, S. E. (2010). Teacher collaboration for professional learning:
Facilitating study, research, and inquiry communities. San Francisco, CA: Jossey-Bass.
Lee, V. E., & Smith, J. B. (1996). Collective responsibility for learning and its effects on
gains in achievement for early secondary school students. American Journal of
Education, 104(2), 103-147.
Leinhardt, G. (1988). Situated knowledge and expertise in teaching. In J. Calderhead
(Ed.), Teachers' professional learning (pp. 146-168). London, UK: Falmer Press.
Leonard, L., & Leonard, P. (2003). The continuing trouble with collaboration: Teachers
talk. Current Issues in Education, 6(15). Retrieved from
http://cie.asu.edu/volume6/number15
LeTendre, G. K., Baker, D. P., Akiba, M., Goesling, B., & Wiseman, A. (2001). Teachers'
work: Institutional isomorphism and cultural variation in the U.S., Germany, and
Japan. Educational Researcher, 30(6), 3-15.

163

Leung, F. K. S. (2001). In search of an East Asian identity in mathematics education.
Educational Studies in Mathematics, 47(1), 35-51.
Lewis, C. (2000). Lesson study: The core of Japanese professional development. Paper
presented at the annual meeting of the American Educational Research
Association, New Orleans, LA.
Little, J. W. (1990). The persistence of privacy: Autonomy and initiative in teachers'
professional relations. Teachers College Record, 91(4), 509-534.
Little, J. W. (1993). Teachers' professional development in a climate of educational
reform. Educational Evaluation and Policy Analysis, 15(2), 129-151.
Little, J. W., Gearhart, M., Curry, M., & Judith, K. (2003). Looking at student work for
teacher learning, teacher community, and school reform. Phi Delta Kappan, 85(3),
184-192.
Lortie, D. C. (2002). Schoolteacher: A sociological study (2nd ed.). Chicago, IL: University of
Chicago Press.
Lortie, D. C. (2009). School principal: Managing in public. Chicago, IL: University of
Chicago Press.
Louis, K. S., & Marks, H. M. (1998). Does professional community affect the classroom?
Teachers‟ work and student experiences in restructuring schools. American
Journal of Education, 106(4), 532–575.
Louis, K. S., Marks, H. M., & Kruse, S. (1996). Teachers‟ professional community in
restructuring school. American Educational Research Journal, 33(4), 757-798.
March, J. G., & Simon, H. (1993). Organizations. Cambridge, MA: Blackwell.
Marks, H. M., & Printy, S. M. (2003). Principal leadership and school performance: An
integration of transformational and instructional leadership. Educational
Administration Quarterly, 39(3), 370-397.
Marshall, M. L. (2004). Examining school climate: Defining factors and educational influences.
Atlanta, GA: Georgia State University Center for School Safety, School Climate
and Classroom Management.
McEneaney, E., & Meyer, J. W. (2000). The content of the curriculum: An institutionalist
perspective. In M. T. Hallinan (Ed.), Handbook of the sociology of education (pp. 189211). New York, NY: Kluwer Academic.

164

McLaughlin, M. W., & Talbert, J. E. (1993). Introduction: New visions of teaching. In D.
K. Cohen, M. W. McLaughlin & J. E. Talbert (Eds.), Teaching for understanding:
Challenges for policy and practice (pp. 1-10). San Francisco, CA: Jossey-Bass.
McLaughlin, M. W., & Talbert, J. E. (2001). Professional communities and the work of high
school teaching. Chicago, IL: University of Chicago Press.
McNeil, L. M. (1986). Contradictions of control: School structure and school knowledge. New
York, NY: Routledge.
Meyer, J. W., & Ramirez, F. O. (2003). The world institutionalization of education. In J.
Schriewer (Ed.), Discourse formation in comparative education (2nd ed., pp. 111-131).
Frankfurt am Main, Germany: Peter Lang.
Meyer, J. W., & Rowan, B. (1991). Institutionalized organizations: Formal structure as
myth and ceremony. In W. W. Powell & P. J. DiMaggio (Eds.), The new
institutionalism in organizational analysis (pp. 41-62). Chicago, IL: University of
Chicago Press.
Miles, M. B. (1965). Planned change and organizational health: Figure and ground. In R.
O. Carlson, A. Gallaher, Jr., M. B. Miles, R. J. Pellegrin & E. M. Rogers (Eds.),
Change processes in the public schools (pp. 11-34). Eugene, OR: Center for the
Advanced Study of Educational Administration, University of Oregon.
Miller, J., Kostogriz, A., & Gearon, M. (Eds.). (2009). Culturally and linguistically diverse
classrooms: New dilemmas for teachers. Bristol, UK: Multilingual Matters.
Mohr, N., & Dichter, A. (2001). Building a learning organization. Phi Delta Kappan,
82(10), 744-747.
Munthe, E. (2007). Recognizing uncertainty and risk in the development of teachers'
learning communities. In M. Zellermayer & E. Munthe (Eds.), Teachers learning in
communities (pp. 15-26). Rotterdam, Netherlands: Sense.
National Council of Teachers of Mathematics. (2000). Principles and standards for school
mathematics. Reston, VA: NCTM.
National Council of Teachers of Mathematics. (2007). Mathematics teaching today (2nd
ed.). Reston, VA: NCTM.
National Research Council. (2002). Adding it up: Helping children learn mathematics.
Washington, DC: National Academy Press.

165

Newmann, F. M., Rutter, R. A., & Smith, M. S. (1989). Organizational factors that affect
school sense of efficacy, community, and expectations. Sociology of Education,
62(4), 221-238.
O‟Connell, A. A., Goldstein, J., Rogers, H. J., & Peng, C. Y. J. (2008). Multilevel logistic
models for dichotomous and ordinal data. In A. A. O‟Connell & D. B. McCoach
(Eds.), Multilevel modeling of educational data (pp. 199-242). Charlotte, NC:
Information Age.
Paine, L. W., & Fang, Y. (2006). Reform as hybrid model of teaching and teacher
development in China. International Journal of Educational Research, 45(4), 279-289.
Paine, L. W., & Ma, L. (1993). Teachers working together: A dialogue on organizational
and cultural perspectives of Chinese teachers. International Journal of Educational
Research, 19(8), 675-697.
Paine, L. W., Pimm, D., Britton, E., Raizen, S., & Wilson, S. (2003). Rethinking induction:
Examples from around the world. In M. Scherer (Ed.), Keeping good teachers (pp.
67-80). Washington, DC: ASCD.
Perrow, C. (1986). Complex organizations: A critical essay (3rd ed.). New York, NY:
McGraw-Hill.
Porter, A. C., & Gamoran, A. (Eds.). (2002). Methodological advances in cross-national
surveys of educational achievement. Washington, DC: National Academy Press.
Powell, A. G., Farrar, E., & Cohen, D. K. (1985). The shopping mall high school: Winners and
losers in the educational marketplace. Boston, MA: Houghton Mifflin.
Prawat, R. S., Remillard, J., Putnam, R. T., & Heaton, R. M. (1992). Teaching
mathematics for understanding: Case studies of four fifth-grade teachers.
Elementary School Journal, 93(2), 145-152.
Printy, S. M. (2008). Leadership for teacher learning: A community of practice
perspective. Educational Administration Quarterly, 44(2), 187-226.
Putnam, R. T., & Borko, H. (2000). What do new views of knowledge and thinking have
to say about research on teacher learning? Educational Researcher, 29(1), 4-15.
Rait, E. (1995). Against the current: Organizational learning in schools. In S. B.
Bacharach & B. Mundell (Eds.), Images of schools: Structures and roles in
organizational behavior (pp. 71-107). Thousand Oaks, CA: Corwin Press.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data
analysis methods (2nd ed.). Thousand Oaks, CA: Sage.
166

Raykov, T., & Marcoulides, G. A. (2008). An introduction to applied multivariate analysis.
New York, NY: Routledge.
Reitzug, U. C. (1994). A case study of empowering principal behavior. American
Educational Research Journal, 31(2), 283-307.
Reitzug, U. C. (1997). Images of principal instructional leadership: From super-vision to
collaborative inquiry. Journal of Curriculum and Supervision, 12(4), 324-343.
Rosenholtz, S. J. (1989). Teachers' workplace: The social organization of schools. New York,
NY: Longman.
Rowan, B. (1990). Commitment and control: Alternative strategies for the organizational
design of schools. Review of Research in Education, 16, 359-389.
Rowan, B. (2002). Large-scale, cross-national surveys of educational achievement:
Promises, pitfalls, and possibilities. In A. C. Porter & A. Gamoran (Eds.),
Methodological advances in cross-national surveys of achievement (pp. 319-350).
Washington, DC: National Academy Press.
Rowan, B., Raudenbush, S. W., & Cheong, Y. F. (1993). Teaching as a nonroutine task:
Implications for the management of schools. Educational Administration Quarterly,
29(4), 479-500.
Salomon, G. (1993). Distributed cognitions: Psychological and educational considerations.
Cambridge, UK: Cambridge University Press.
Saunders, W. M., Goldenberg, C. N., & Gallimore, R. (2009). Increasing achievement by
focusing grade-level teams on improving classroom learning: A prospective,
quasi-experimental study of Title I schools. American Educational Research Journal,
46(4), 1006-1033.
Scheffler, I. (1960). The language of education. Springfield, IL: Charles C. Thomas.
Schein, E. (1985). Organizational culture and leadership. San Francisco, CA: Jossey-Bass.
Schmidt, W. H., Cogan, L. S., Houang, R. T., & McKnight, C. C. (2011). Content
coverage differences across districts/states: A persisting challenge for U.S.
education policy. American Journal of Education, 117(3), 399-427.
Schmidt, W. H., Wang, H. C., & McKnight, C. C. (2005). Curriculum coherence: An
examination of U.S. mathematics and science content standards from an
international perspective. Journal of Curriculum Studies, 37(5), 525–559.

167

Schneider, B., Carnoy, M., Kilpatrick, J., Schmidt, W. H., & Shavelson, R. J. (2007).
Estimating causal effects using experimental and observational designs. Washington,
DC: American Educational Research Association.
Schriewer, J. (2003). Globalization in education: Process and discourse. Policy Futures in
Education, 1(2), 271-283.
Scott, W. R. (2005). Ecosystems and the structuring of organizations. In L. V. Hedges &
B. Schneider (Eds.), The social organization of schooling (pp. 37-47). New York, NY:
Russell Sage.
Scribner, J. P., Sawyer, R. K., Watson, S. T., & Myers, V. L. (2007). Teacher teams and
distributed leadership: A study of group discourse and collaboration. Educational
Administration Quarterly, 43(1), 67-100.
Sergiovanni, T. J. (1994). Building community in schools. San Francisco, CA: Jossey-Bass.
Sergiovanni, T. J., & Starratt, R. J. (2002). Supervision: A redefinition (7th ed.). New York,
NY: McGraw-Hill.
Shulman, L. S. (1987). The wisdom of practice: Managing complexity in medicine and
teaching. In D. C. Berliner & B. V. Rosenshire (Eds.), Talks to teachers (pp. 369-386).
New York, NY: Random House.
Shulman, L. S. (2004). Autonomy and obligation: The remote control of teaching. In S. M.
Wilson (Ed.), The wisdom of practice: Essays on teaching, learning, and learning to
teach (pp. 133-162). San Francisco, CA: Jossey-Bass.
Shulman, L. S. (2006). Knowledge and teaching: Foundations of the new reform. In D.
Hartley & M. Whitehead (Eds.), Teacher education: Major themes in education (pp.
119-145). New York, NY: Routledge.
Silcock, P. (1994). The process of reflective teaching. British Journal of Educational Studies,
42(3), 273-285.
Silver, E. A., Mesa, V. M., Morris, K. A., Star, J. R., & Benken, B. M. (2009). Teaching
mathematics for understanding: An analysis of lessons submitted by teachers
seeking NBPTS certification. American Educational Research Journal, 46(2), 501-531.
Simon, H. (1991). Bounded rationality and organizational learning. Organization Science,
2(1), 125-134.
Slavit, D., Nelson, T. H., & Kennedy, A. (Eds.). (2009). Perspectives on supported
collaborative teacher inquiry. New York, NY: Routledge.

168

Smith-Doerr, L., & Powell, W. W. (2005). Networks and economic life. In N. J. Smelser &
R. Swedberg (Eds.), The handbook of economic sociology (2nd ed., pp. 379-402).
Princeton, NJ: Princeton University Press.
Somech, A. (2005). Directive versus participative leadership: Two complementary
approaches to managing school effectiveness. Educational Administration
Quarterly, 41(5), 777-800.
Spillane, J. P., Halverson, R., & Diamond, J. B. (2004). Towards a theory of leadership
practice: A distributed perspective. Journal of Curriculum Studies, 36(1), 3-34.
Spillane, J. P., Reiser, B. J., & Reimer, T. (2002). Policy implementation and cognition:
Reframing and refocusing implementation research. Review of Educational
Research, 72(3), 387-431.
Spiro, R. J., Feltovich, P. J., Jacobson, M. J., & Coulson, R. L. (1992). Cognitive flexibility,
constructivism, and hypertext: Random access instruction for advanced
knowledge acquisition in ill-structured domains. In T. M. Duffy & D. H. Jonassen
(Eds.), Constructivism and the technology of instruction: A conversation (pp. 57-74).
Hillsdale, NJ: Lawrence Erlbaum Associates.
Stevens, R. J., & Slavin, R. E. (1995). The cooperative elementary school: Effects on
students' achievement, attitudes, and social relations. American Educational
Research Journal, 32(2), 321-351.
Stevenson, D., & Baker, D. P. (1991). State control of the curriculum and classroom
instruction. Sociology of Education, 64(1), 1-10.
Tobin, J., Hsueh, Y., & Karasawa, M. (2009). Preschool in three cultures revisited: China,
Japan, and the United States. Chicago, IL: University of Chicago Press.
Tschannen-Moran, M. (2001). Collaboration and the need for trust. Journal of Educational
Administration, 39(4), 308-331.
Wang, J., Strong, M., & Odell, S. J. (2004). Mentor-novice conversations about teaching:
A comparison of two U.S. and two Chinese cases. Teachers College Record, 106(4),
775-813.
Weick, K. E. (1976). Educational organizations as loosely coupled systems.
Administrative Science Quarterly, 21(1), 1-19.
Weick, K. E. (1995). Sensemaking in organizations. Thousand Oaks, CA: Sage.
Weick, K. E., Sutcliffe, K. M., & Obstfeld, D. (2005). Organizing and the process of
sensemaking. Organization Science, 16(4), 409-421.
169

Wenger, E. C., & Snyder, W. M. (2000). Communities of practice: The organizational
frontier. Harvard Business Review, 78(1), 139-145.
Westbury, I., Hopmann, S., & Riquarts, K. (Eds.). (2000). Teaching as a reflective practice:
The German Didaktik tradition. Mahwah, NJ: Lawrence Erlbaum Associates.
Williamson, O. E. (1994). The economics of organization: The transaction cost approach.
In W. R. Scott (Ed.), Organizational sociology (pp. 548-308). Aldershot, UK:
Dartmouth.
Wilson, S. M., Floden, R. E., & Ferrini-Mundy, J. (2001). Teacher preparation research:
Current knowledge, gaps, and recommendations. Seattle, WA: Center for the Study of
Teaching and Policy, University of Washington.
Wiske, M. S. (Ed.). (1998). Teaching for understanding: Linking research with practice. San
Francisco, CA: Jossey-Bass.
Wood, T., Cobb, P., & Yackel, E. (1990). The contextual nature of teaching: Mathematics
and reading instruction in one second-grade classroom. Elementary School Journal,
90(5), 496-513.
Yasumoto, J. Y., Uekawa, K., & Bidwell, C. E. (2001). The collegial focus and high school
students‟ achievement. Sociology of Education, 74(3), 181-209.

170