Perspectives on Class Size Reduction

Full text

Perspectives on Class Size Reduction
Elizabeth Graue, Denise Oen, Kelly Hatch, Kalpana Rao, Erica Fadali
University of Wisconsin, Madison
A paper presented at the symposium Early Childhood Policy in Practice: The Case of Class Size
Reduction at the annual meeting of the American Educational Research Association, Tuesday, April 12,
2005, Montreal, Canada.
Primary contact information: Elizabeth Graue, Professor, Wisconsin Center for Education Research,
University of Wisconsin Madison, 1025 W. Johnson, Madison, WI 53706, 608 262-7435,
graue@education.wisc.edu
.
Support for this paper was provided by the Wisconsin Department of Public Instruction.

Perspectives on CSR 2
Gov. Jeb Bush on Monday abandoned his effort to repeal Florida's popular class size
amendment and unveiled a scaled-back plan that also increases teacher pay.
Conceding that repealing the amendment would be difficult, Bush asked lawmakers to
put a measure before voters that would essentially freeze the caps on class sizes to the
current standard. That would give school districts more flexibility, he said.(James &
Johnson, 2005)
The governor's proposal to increase state funding for public education wouldn't solve all
of the Madison School District's budget woes, but it would provide significant help in
dealing with an estimated shortfall of $6 million to $8 million next year, Superintendent
Art Rainwater said.
The district would gain an added $1.3 million for the Student Achievement Guarantee in
Education program, which limits class sizes in kindergarten through third grade to an
average of 15 students.
Doyle wants to raise the reimbursement for low-income SAGE students from $2,000 to
$2,500 per pupil. Madison has 2,656 low- income SAGE students and a total 5,620
students in the SAGE program. (Cullen, 2005)
Class size reduction (CSR) is one of the most frequently researched and policy implemented
reforms in recent educational history. It has been studied by a wide range of researchers
representing interests ranging from economics to administration and has focused on short term
and long term outcomes for students, costs relative to other reforms, teacher attitudes and
practices. In the last 10 years, 40 states have implemented CSR programs, the federal
government had a short term program that infused teachers into the employment ranks to reduce
class size, and a number of reviews have suggested that children learn more in smaller classes,
their teachers are more satisfied, and discipline problems are reduced. Anchored by the only
large scale randomized experiment in fieldbased educational research (Jeremy D. Finn &
Achilles, 1990), the knowledge base on class size reduction is voluminous, highly varied in
quality and foci.

Perspectives on CSR 3
In our work analyzing the literature on class size reduction we found that a number of
substantive reviews had already been completed on the subject. One might wonder why
researchers would undertake another review. In fact, we were told just that by Alex Molnar who
strongly suggested we were just reinventing the wheel – that we knew what we needed to know
ten years ago.
Despite this advice, we can think of several reasons to revisit this literature. The first is a
pragmatic one in relation to policy. Given the difficult economic context that has many states
and local governments cutting programs and reallocating the funding to other initiatives, it is
important to assess the knowledge base about the efficacy of class size reduction so that good
decisions can be made. The second reason might be seen as more rhetorical, or political. We
consider a review to be much more and much less than attempting to find out what the literature
“says.” The literature has no voice – it is researchers who take a position and speak about topics
in their area of interest. We follow the idea, first suggested by Grant & Graue (Grant & Graue,
1999) that reviews represent conversations in the educational research community – as such they
are sites of interaction that often present multiple perspectives on a topic. To reflect this
approach, we structure our review in a slightly different way than most reviews. We take an
interpretive approach to research syntheses, exploring how different researchers have taken up a
topic, operationalized it in research methods, and argued for a particular understanding of the
outcomes of the study. From this perspective, a research synthesis is an analysis of the discourse
of the research community on a topic. We add to this a layer of inquiry that is not typically part
of a review. We present researchers’ perspectives on our knowledge base through analysis of
interviews with a sample of researchers who have worked on the topic of class size reduction.

Perspectives on CSR 4
This exploration of understandings of understandings is a methodological instantiation of the
interpretive status we use here.
Our review is organized around two questions:
 How has class size reduction been conceptualized in the literature?
 What methods and measures have shaped our knowledge?
Methods
We should be clear about the catalyst for this review. As part of an evaluation of the Student
Achievement Guarantee in Education, we recognized that having a firm grasp of the literature
would be important. Given the number of reviews that had been done previously and the mixed
messages coming out of those reviews, we decided to take a mixed method approach to
exploring the literature. We began in a traditional way, doing a systematic search of the
literature. But one of the first that we found is that what is “there” depends on “where” you look.
With the changes in ERIC it is rarely the case that one can find a single database that covers the
range of journals of interest in education. Therefore, multiple searches across multiple databases
were required. We did searches of the major electronic databases (Education Full text, ERIC,
EBSCO, Ingenta, JSTOR and Proquest) using the descriptor class size reduction. Initially we
searched for any types of written text (journals, reports, popular press) and found close to 400
pieces. We narrowed our search to peer reviewed published works to make the task more
manageable. We then began working from bibliography to bibliography, searching out
published work that our initial searches had missed.
We worked to describe the literature by focusing on 4 general categories of the literature.
We summarized the research related to the three major policy initiatives for class size
reduction—STAR in Tennessee, SAGE in Wisconsin, and the California class size reduction

Perspectives on CSR 5
program. We described the parameters of the policy, the methods used to evaluate the programs
and the suggested results. In addition we analyzed the synthetic work on class size reduction,
reading and analyzing reviews on the topic. Finally, we identified key researchers who have
contributed to the scholarship on class size and interviewed individuals who represented varied
perspectives and expertise. Our interviews included scholars instrumental in evaluation of the
major class size initiatives: Charles Achilles (STAR), Alex Molnar (SAGE), Joan McRobbie
(California). In addition, we talked with people who had synthesized the CSR literature: Bruce
Biddle, David Grissmer, and Adam Gamoran. We interviewed an economist who analyzes the
cost of class size relative to other investments in education – Eric Hanushek and a researcher
who did secondary and follow on analysis of SAGE data – Norman Webb. We recognize that
our interview participants represent a sample of the research on CSR. It is important to say that
some researchers declined participation. Therefore, silences come for a variety of reasons and
we are comforted in the idea that we do not have to talk with everyone to have some sense of the
conversations related to CSR.
Major class size initiatives
STAR
Tennessee’s Student Teacher Achievement Ratio project began as a pilot in 1984 and was
mandated in 1985. A large scale, randomized within-school experiment (Nye, Hedges, &
Konstantopoulos, 2001), pupils entering kindergarten in participating schools were randomly
assigned to a small class (13-17), a full size class (22-25) or a full size class with a full time
teacher aide. Teachers were also assigned at random to the classrooms. Pupils were kept in the
same condition for up to 4 years, with a new teacher assigned at random each year (Finn, 2002).
The four year, $13 million project was completed in 1990.

Perspectives on CSR 6
Students in small classes performed better on all achievement measures, and the positive
benefits of smaller classes were cumulative. In kindergarten, slightly more than half of the small
classes were in the top scoring 10% of STAR classes. By third grade, that proportion had grown
to 78% (Bain et al, 1992). Small class advantages were greater children in poverty and of color.
The gap between black and white students were reduced in the smaller classes, with African
American students in smaller classes 4%ile points below white students but those in large classes
were behind by 14%ile points. While there was no evidence that smaller classes improved
student motivation, small classes had fewer discipline problems, more questions and greater
participation and concentration. Teacher quality, including practices developed through
professional development, had little effect on student outcomes. The addition of teacher aides to
regular sized classes did not improve achievement (Achilles, 1997, Biddle et al., 2002, Finn,
1998,2002). Charles Achilles explains why:
Teacher aides are not a help in the classroom at early grades except as required by
legislation and/or to help teachers out with materials and things. They shouldn’t be
around kids unless they are well-trained and then there is some question because the
teacher sends troubled kids to the teacher aide. (Achilles interview)
While the STAR project appeared to be very positive, there were several criticisms of the
study. The primary concerns about the project were that the student sample wasn’t
representative of the U.S. population (few Hispanic, Native American and Non-English speaking
students) and that the study suffered from the Hawthorne effect when news about the greater
achievement of small classes leaked out during the project. Critics argued that it was likely that
this may have led participating teachers to work harder to get positive results (Gilman & Kiger,
2003). Another critique of the project was that students were randomly assigned not randomly
selected and that the assignments changed over time. Only volunteers were part of the random
assignment pool and schools had to be large enough to accommodate at least three classes in

Perspectives on CSR 7
each grade. Hanushek (1999) argued that the results of the study depended fundamentally on the
choice of teachers. There was little description of how teachers were randomly assigned to
treatment groups. The huge differences generally found among teachers could dramatically
influence the results. Because of the ongoing need to assign new teachers to the various
treatment groups across the 4 years, it seems entirely plausible (in Hanushek’s estimation) that
elements of teacher and principal preferences for different classes entered into the process. He
also criticized the original researchers for their lack of consideration of attrition of the treatment
groups. Of the initial experimental group starting in kindergarten, only 48% remained in the
experiment for the entire 4 years. In our interview, Eric Hanushek described the kind of study
that should be done to alleviate concerns:
A more careful description of the randomization process and how they randomized
teachers and students. More careful assessment of achievement differences before and
after reduced class size; allowing for some variation in the number of years that people
are in different treatment categories; trying to assess better any variation in teacher
quality that lie behind this; trying to follow the attrition from the program better; trying to
maintain the study design that keeps people in their particular treatment groups.
(Hanushek interview)
Bruce Biddle recognized the weaknesses in the study but suggests that in the context of a
practice based field like education, STAR might be a good enough study to inform policy:
The issue is that within a field setting, with human beings, this one was a pretty good
piece of research. It was done on a state-wide basis. It involved matched classrooms
within schools, which were subjected to three different kinds of treatments, standard-
sized classrooms, classrooms that were standard sized where a teacher was supplemented
in her or her efforts by a teacher’s aide and then reduced size classes. These were done in
the early grades. It was begun in kindergarten and in the best cases the kids were
continued in these various treatment conditions for four years, through first, second, and
third grade. Thereafter they were returned to standard classrooms in their schools.
(Biddle interview)
A follow on study, the Lasting Benefits study, conducted from 1989-1996 tracked
achievement effects as students progressed into grades higher than those observed by the original

Perspectives on CSR 8
study (Nye et al., 1989, 1994-1999). This study yielded two primary findings: First, students in
smaller class performed better in all academic areas studied. Second, students from the small
classes expended more effort in the classroom, took greater initiative in learning activities, and
displayed disruptive or inattentive behavior (U.S. Department of Education, 1998).
SAGE
We believe that state and local policies can and should ensure that teachers have the time
and resources to get to know their students and parents and to give each child sufficient
attention, ensuring academic achievement at the highest possible level. We believe that it
is important to fully prepare children for the diverse civic and employment environment
where they will live and work. And we believe it is time to free educators to educate,
holding both teachers and all other people who hold a stake in the schools-
administrators, board members, parents, and students-accountable for the results. (Molnar
& Zmrazek, 1994)
The Student Achievement Guarantee in Education, SAGE, much like other Class Size Reduction
reforms, was envisioned as a way to narrow the achievement gap for children living in poverty.
Class size reduction was only one aspect of a multifaceted program that also includes a
family/school component, curriculum reform, and professional development. How these
components are studied provides insight into how SAGE is being conceptualized and
implemented.
Student outcomes were a central indicator of program efficacy. Students in SAGE
classrooms outperformed the comparison group in reading, language arts and mathematics
through third grade, with the most pronounced effects found for African-American students (P.
Smith, Molnar, & Zahorik, 2003). Webb and colleagues reanalyzed the initial evaluation data
and found similar patterns of advantage in the original data which did not sustain in other data
sources in third and fourth grade (Webb, Meyer, Gamoran , & Fu, 2004). Hruz questioned
whether the size of the effects were in fact worth the cost of the program and suggested that

Perspectives on CSR 9
implementation be limited to the subgroups that had proven effects (K and 1
st
grade and African
American students) (Hruz & Wisconsin Policy Research Institute., 2000).
Bruce Biddle contrasted the design of SAGE with that of STAR:
It involved not a clean experimental design but rather, a massive intervention study in
which class size reduction was a part of what was initially to be a multi-faceted
intervention study but the other facets of the intervention essentially xxx out and much of
the efforts are clearly associated with class size. This was a very good piece of research.
. . .There were more controls built into it. It was focused specifically on schools where a
lot of minority kids were present. (Biddle interview)
One advantage of the SAGE evaluation was its attempt to represent the complexity of the
reform, with attention to teaching practices and professional development opportunities.
Through observations in math and reading instruction, Zahorik et al found that SAGE classrooms
had fewer discipline problems, more time spent on instruction and more individualized, explicit
instruction, the teachers had more knowledge of students and more satisfaction with teaching
(John Zahorik, Halbach, Ehrle, & Molnar, 2003). Despite these enumerated changes, researchers
suggested that teacher practice does not change in smaller classes (J. Zahorik, Molnar, & Smith,
2003). Examinations of the professional development components of SAGE found that they
were only partially met (Molnar, 2002). The authors also explored how SAGE professional
development guidelines aligned with both the Wisconsin Teacher Professional Standards (which
went into effect July 1, 2004) as well as the mandates of No Child Left Behind. It is their belief
that current professional development practices will not fully comply.
Kiger (2002) argued that smaller classes create a “coherence” of the all four SAGE
components and that all components have a synergistic effect on one another. He goes on to
suggest that research needs to explore why smaller class sizes seem to have a positive effect.
Missing from the research on SAGE is attention to English Language Learners, to
administrative practices at the school and district level, to the effects the programs had on home-

Perspectives on CSR 10
school connections, to the specific implementation of the pupil teacher ratio in its varied forms,
and systemic analysis of the politics and interconnections that facilitate or inhibit the successful
implementation of this type of complex reform. While the designs of the SAGE evaluations did
mirror the components that served to structure the policy, most attention was focused on the
student outcomes, with the structural and process analyses examined at a diffuse level of many
sites rather than at a level that provided in-depth knowledge. This attention to knowledge about
many is important for population generalization but may be less informative for informing
particular practice.
California
CSR in California is conceived in the literature as something that was born at a good
time, with good intentions, but poor planning. Its enactment was motivated by a ten-year decline
in achievement in California, tying for last place on NAEP in 1994. A snapshot of California
classrooms at the time would have shown overcrowded classes, a shortage of qualified teachers,
and minority students, ELLs and poor students falling more and more behind (Stecher,
Bohrnstedt, Kirst, McRobbie & Williams, 2001). With a budget surplus and the impressive
claims of the STAR study, the California government dove head first into CSR as a statewide
program. This approach had its downside, outlined by David Grissmer:
We have the California situation – California has kind of taken a bum rap on some of this
stuff especially when they had a huge budget surplus. They didn’t have an option of let’s
do a little this year, a little this year, a little next year, that sort of not the way politics
works. What this meant was that implementation was going to be swampy for a while.
They are not going to have the supply of teachers they need, all of the classroom size is
not going to be available, and so forth. They reduced from 28 to 20. they did it on all
socioeconomic levels, there was no sort of distinction and what they ran into was that it
created a lot of teaching positions and hiring in schools and inner city teachers moved to
those. So it got implemented fairly shabbily in the beginning but in my sense it was still
worthwhile. (Grissmer interview)

Perspectives on CSR 11
Researchers emphasize that CSR is only one part of a comprehensive plan to improve
learning (McRobbie, 1996). It is essential to also change teaching and learning behaviors
(McRobbie). This means providing teacher support and professional development—about
individualized instruction, better use of methods and materials, improved organization, a better
variety of activities, better assessment and a stronger curriculum (McRobbie). Researchers also
point out that effective, specific strategies for teaching smaller classes have yet to be identified
(McRobbie, year?). Moreover, teacher quality matters, because they feel high quality instruction
is essential for CSR to succeed.
In California, the gains earned from CSR do not necessarily outweigh the costs of the
policy, according to researchers. To be sure, there were gains: a slight increase in test scores
after the second and third years, more time teaching and less disciplining, and more parent-
teacher contact time (reported) (Stecher, et al., 2001). However, the costs of the CSR policy
include, most prominently, that educational inequities were exacerbated, particularly in districts
already considered disadvantaged. There was a 38% increase in the teaching force, accompanied
by a serious decline in the qualifications of these teachers particularly in the highest poverty
districts and schools (McRobbie, Finn & Harman, 1998; Stecher, et al.). It is also more
expensive to implement the policy in these districts. Space for new classrooms was taken from
special education, childcare, specials and library/computer facilities (McRobbie, Finn &
Harman; Stecher, et al.). And, many teachers of English Language Learners (ELLs) and special
education students switched to regular education to help fill the gaps (Stecher, et al.; Wexler, Izu,
Calrlos, Fuller, Hayward & Kirst, 1998).
From the California implementation researchers suggest that investments be made in
ongoing professional development with a focus on literacy (McRobbie, year?; McRobbie, 1996)

Perspectives on CSR 12
and on instructional strategies for a diverse student population, i.e. language acquisition and
communication strategies across languages and cultures. Further, a focus on prevention would
help teachers identify and respond to student learning issues. Finally a systemic induction model
would support new teachers, especially emergency credentialed teachers added to staff CSR
classrooms (McRobbie, 1996). Good CSR practice broadens notions of defining policy success.
While including measures of student achievement it also involves using resources creatively and
establishing new district and community collaboratives to reach goals for student achievement
(McRobbie, 1996; McRobbie, Finn & Harman, 1998).
California CSR researchers are skeptical supporters. They remind consumers of the
literature that it is unclear how CSR affects student achievement, and that we know that CSR
effects are coupled with teacher quality, instructional strategies, curriculum, student-teacher
interactions, family background and economic situation (of students). Their discussions tend to
focus on implementation critique. Given the differences between STAR and the California
policy, they suggest that the Tennessee study was not a good model for California to try to
duplicate. STAR was a limited controlled study and California implemented a statewide program
for 1.8 million children. Availability of qualified teachers was not considered, the baseline class
size in California was much larger than in Tennessee, and California’s curriculum standards were
not as fully developed. One of the most glaring differences, though, is the diversity in
California, especially concerning English Language Learners (McRobbie, Finn & Harman, 1998;
Stecher, et al., 2001). Joan McRobbie that not all of California’s children gained:
It’s fair to say that probably some kids have lost because the way the policy was setup,
certain kids who all along have been more likely to get the least qualified, least
experienced teachers, by virtue of class size reduction, are even more likely to get the
least qualified, least experienced teachers because of the shortage you’ve created and the
number of people brought in. . . At least initially, it probably made for a more negative
situation for some kids (McRobbie interview).

Perspectives on CSR 13
The California experience suggests some guidelines for large scale implementation of
CSR. Focus on equity. Develop policies that fully recognize the complexity of practice.
Improve teaching strategies. Plan for sufficient staffing. Implementations that ignore these
elements compromise the potential that CSR can have for student achievement. (McRobbie,
year?; McRobbie, Finn & Harman, 1998).
British Implementation of CSR
In the 1990’s the British government claimed that educational attainment was a function
of teacher quality rather than class size. Teachers and the public resisted this conclusion, and
through meetings and talks, it became clear that “a proper and thorough study of educational
consequences of class size differences” (p. 4) was needed. It was in this context that Peter
Blatchford aimed to bridge the gap between professional experience and research-based
evidence. Rather than trying to settle the debate, he focused on ‘why’ and “provide a sustained
inquiry” (Blatchford, 2003, p.1).
Blatchford believes that teachers interact with students in context, so that CSR is only
one influence on the classroom experience. Moreover, the nested contexts of schools means that
different sizes of classes can mean different dynamics and these might influence teachers and
students. His research, then, aimed to 1) determine if CSR affects students’ achievement; and 2)
study the relationships between class size and classroom processes, which might illustrate any
differences found in achievement. He constructed a three year longitudinal study that followed
students from school entry, i.e. 4 years. . Both quantitative and qualitative methods were used
including case studies, teachers reports, observations, teacher ratings of behavior, and teachers’
estimates of time use.

Perspectives on CSR 14
In Reception Year (kinder), all ability groups had positive outcomes related to smaller
classes, with larger effects in literacy for students with lower baseline. Long terms effects were
maintained when students moved to small or similar sized classrooms after the first year but they
were lost if students went from from a small class to a large class.
In larger classes, students were more likely to be off task, distracted, and to spend more
time with their peers. In smaller classes, students had more interaction with their teachers—child
initiations and answering the teacher. With the exception of Reception year students, peer
relations were often worse in smaller classes because of aggression and rejection.. As class size
increases, teachers perceive less time for individual reading support and rarely mention guided
reading groups as an instructional option. They felt that the quality of interactions was better in
smaller classes; with teachers knowing and supporting their students better. Teachers used more
procedural talk in larger classes, and more social or personal interactions to manage the group in
smaller classes.
In first and second year classes students read more frequently and for longer periods.
There was less differentiated instruction with larger classes and questioning was used for
controlling rather than to teach. The feedback in smaller classes was more immediate, but such
feedback takes time from lessons and disrupts the flow of activities and instruction.
It was harder to extend reading strategies with larger classes and there were fewer
individual reading opportunities, monitoring and checking for understanding. More time was
used for teaching in smaller classes including more one-to-one teaching, more teaching in
groups, more times when students were focus of teachers’ attention, more student-teacher
interactions, and more participatory interactions between students and teachers.

Perspectives on CSR 15
Syntheses
Focusing on class size alone is like trying to determine the optimal amount of butter in a
recipe without knowing the nature of the other ingredients. (Berger, 1982)
The knowledge base on class size reduction is framed in relation to a pair of papers that
presented a relatively new type of analysis of relationships among class size, student
achievement, attitudes, and instruction. Utilizing meta-analysis, Glass & Smith and Smith &
Glass used the power generated by aggregating results across studies to refine our understanding
of the knowledge base. Based on 59 studies yielding hundreds of effect measures, the authors
provided a large scale overview of how class size shapes educational outcomes. With studies
that ranged across academic content, levels of schooling, everything from tutoring to large
lectures, the meta-analyses indicated a generally positive effect for student achievement with
larger effects in studies of secondary programs and in more well designed inquiry. The effects
ranged from 30 percentile ranks when comparing a size of 40 and a size of 1 to 10 percentile
ranks between a class of 10 and 20 (Glass & Smith, 1979). The effects for attitudes and
instructional practices were even larger, with an average effect size of .49. To make this more
concrete, Smith and Glass note that if we think of a class size of 30 as the 50
th
percentile, then
reducing the class size to 15 would shift the percentile to 63
rd
and increasing it to 40 would make
it drop to 45. Larger effects were found for teachers and for younger students (M. L. Smith &
Glass, 1980). In contrast to the achievement analysis, less well designed studies had larger
outcomes than more well designed work. The positive outlook presented on reducing class size
in these studies was a major building block in the later large scale implementation. Several key
issues could be noted however. The later work on class size reduction has focused almost
entirely on the early primary grades, while the Glass and Smith analyses explore work through
the secondary school.

Perspectives on CSR 16
Following the Glass & Smith meta-analyses, synthetic reviews of class size reduction or
class size effects became more pervasive but also less inclusive. Rather than sampling the
universe of studies on the topic, researchers made choices that narrowed the field, which in turn
narrowed the dimensions of the problem considered. Hedges and Stock reanalyzed a subset of
the Smith and Glass data using an alternate method of meta-analysis and found smaller, but still
sustainable effects for class size reduction (Hedges & Stock, 1983). Concerned about the
adequacy of study designs, Slavin completed a best evidence synthesis comprised of studies that
met particular methodological requirements. With this constraint the number of studies
considered in the original meta-analysis narrowed from 59 to 11 (Slavin, 1989). On the other
hand, Robinson completed a related cluster analysis that grouped studies by like findings and
similar characteristics. For example, Robinson did analysis of the studies by grade level,
typically considering divisions such as K-3, 4-8, and 9-12. He did not however, consider the
nature of the evidence provided (Robinson, 1990). Syntheses in the 1990’s and later focused on
each other and on the major class size reduction initiatives approaching the literature from a
variety of perspectives. In the next section we present our readings of the major themes in this
literature:
Class size reduction vs pupil teacher ratio
One of the major issues in this literature has been definitional – what exactly is class size
reduction? Is it when you reduce the number of students in a particular classroom or is it when
you reduce the number of students per teacher? While these two are related, they are not the
same (Addonizio & Phelps, 2000; Biddle & Berliner, 2002; Ehrenberg, Brewer, Gamoran, &
Willms, 2001; Jeremy D. Finn, 2002; J.D. Finn, Pannozzo, & Achilles, 2003; Odden, 1990). For
example, many analyses of class size take the number of students related to the number of staff

Perspectives on CSR 17
to argue that class sizes have shrunk over the years (E. Hanushek & Rivkin, 1997). But that
estimate often includes additional support staff who do not systematically provide instruction to
most students. Further, because these analyses rely on data collected for other purposes, the
estimates include effects not related to class size such as student ability, with lower ability
students overrepresented in smaller classes (Addonizio & Phelps, 2000). Adam Gamoran made
this point clearly in his interview:
Many of the correlational studies have been of pupil/teacher ratios rather than class sizes.
Those are two very different things. For example, they basically mix in funds for special
ed with class size so that a school with a lot of special ed teachers would appear to have a
smaller expenditure. But while those special ed teachers are probably going to be good
for the special ed kids, they probably are not going to raise average test scores in the
school by any substantial amount. So that’s a methodological problem in much of the
correlational work. (Gamoran interview)
In contrast some implementations of CSR increase both group size and the number of
teachers, reducing the pupil-teacher ratio. This is the case in many of the classrooms
participating in the SAGE program (Grissmer, 1999). Charles Achilles points out that this is an
administrative solution to an instructional problem:
Pupil/ teacher ratio is to assure equitable distribution of funds. To make sure that every
building and every youngster, whatever the state formula is, gets his or her share of the
money. It’s not an organization for instruction. It’s an administrative device to trace
money and to deal with desegregation issues and things like that. This is the thing that
amazes me, then to have people assume that those two are the same and then using them
as the same. (Achilles interview)
One reason that the patterns in the literature appear to be unclear is that the treatment in
question is muddled. What assumptions shape advocacy of PTR or group size? The focus on
pupil teacher ratio has embedded in it a number of assumptions. The first is that quality in
education should be tied to investment and when the number of professional staff goes up so
should the quality. Hanushek is a primary advocate of that argument and he uses historical
analysis of the staffing patterns to argue that despite significant influx of new staff in US

Perspectives on CSR 18
schools, student achievement has remained flat or decreased over time. Ehrenberg et al (2001)
point out the problems with this approach, noting that indicators like drop out rates should be
considered in school performance; that increasing numbers of students are coming to school with
risk factors that compromise their potential achievement; teaching is attracting less able
candidates; and changes in special education rules has increased the number of staff supporting
students with disabilities. Further, Jeremy Finn disagrees with this rationale, arguing that:
Because pupil teacher ratios are usually computed for large, heterogeneous units (i.e.,
school districts, states, or countries), it is little surprise that they have a weak relationship
with academic achievement. These levels of analysis may be appropriate for an
economists work, but they are not useful for educators concerned with teaching and
learning in individual classrooms. (Jeremy D. Finn, 2002)
The second assumption for using PTR is related to the mechanisms that change student
achievement – from this perspective, researchers assume that reducing the ratio of students to
teachers provides more access to teacher expertise, it provides more hands and eyes to support
student behavior. Odden argues for an array of strategies that allow targeted reductions in
specific content such as tutoring or the addition of a teacher for reading and mathematics
instruction (Odden, 1990). Targeted addition of staff is based on the idea that smaller groups are
effective only for specific content areas. In contrast, a number of implementations have chosen
to add students and teachers within a classroom. We could find no researchers who advocated
this strategy. In contrast, those that argue for group size suggest that the mechanism that is most
potent in changing achievement is teacher knowledge of individual children and through building
relationships among teachers, children and families. With large classes paired with more than
one teacher, this might be less likely.
What we see in these contrasting assumptions is that researchers have used multiple
theories of action related to class size reduction. These theories of action lead to valuing
different criteria and the use of different measures. Even the same measures are interpreted

Perspectives on CSR 19
differently when they are informed by contrasting values. The effects of these theories of action
will be seen in subsequent sections in which specific outcomes and mechanisms are described
and analyzed.
Timing, intensity, and duration of treatment
Prevention scientists typically analyze treatments in terms of 3 key elements: timing,
intensity and duration (Durlak, 1997). One way to assess the literature on class size reduction is
to use like criteria. Important questions, from this perspective would be when does class size
reduction have the most effect?(timing) How small does the group need to be to optimize the
advantage?(intensity) And How many years of small classes are necessary? Or conversely How
few years of small classes will elicit a sustainable effect?(duration).
The literature on class size reduction mirrors many of the policies implemented in the
past twenty years. In terms of the question of timing, researchers seem to agree that investments
in class size reduction are most effective in the primary grades (Addonizio & Phelps, 2000;
Biddle & Berliner, 2002; Jeremy D. Finn, 2002; J.D. Finn et al., 2003; Robinson, 1990). CSR is
consistently suggested in the range of kindergarten to grade three. Bruce Biddle pointed to the
clarity of that point in the research:
The evidence favoring the small classes in the early grades is just overwhelming,
absolutely overwhelming. There is just no way to wiggle around it. Critics have tried to
criticize individual studies, which you can always do, but that’s not the issue. The issue is
alright, what about the next study, and so forth. You can’t brush it all aside. (Biddle
interview)
He goes on to state however, that the findings related to timing might be an artifact of who has
received the treatment. Despite strong support by teachers organizations for implementation
across the board, most of the recent attention has focused on primary grades:
There is very little good research that has to do with class size reduction in the upper
grades. Even though the teachers unions are very much in favor reducing class size and

Perspectives on CSR 20
teachers are very much in favor of reducing class size at all grade levels and I would
prefer not to subject sophomores in high school to large lecture classes. It sounds
ridiculous but the fact of the matter is that there is just very little really good evidence at
that level. In other words, most of the stuff that has been done has been done at the early
grades. (Biddle interview)
Others were not so convinced of the early is better argument. Alex Molnar felt that targeting an
age group was a political decision to allocate resources rather than an empirical matter:
The argument over kindergarten and first grade is not an empirical argument. It’s an
argument about policy preferences and the relative cost of programming. That’s what it’s
about. It can’t be answered through social science, it’s a political argument. So it’s not in
that sense an evaluation problem. People don’t disagree about what the results say, they
disagree about what the implications of the results are. (Molnar interview)
Theorizing the early is better effect has been broadly debated. For example Ehrenberg et
al (2001) suggest that class size reduction might have a latent intervention effect—that the early
experience in a smaller group provides the opportunity for children to develop skills and
dispositions that allow them to take better advantage of the opportunities provided in later
grades. Others broaden the potential explanations by noting that interventions have two different
types of effects: they can change cognitive, psychological, or social development or they change
the future context in which the individual grows and develops (Biddle & Berliner, 2002;
Grissmer, 1999). Charles Achilles suggests that it is essentially socialization to the institution of
schooling – start it early and it will be more successful:
Small classes have to start when the kids start school… the same reason by analogy that
we do inductions for teachers or mentoring or we do apprenticeships or things like that,
kids have to learn what school is, they don’t know natively at age four and five what
school is. (Achilles interview)
The issue of how small has been connected to the grade level at which class size
reduction occurs (Addonizio & Phelps, 2000). Most researchers suggest that reductions of at
least to 20 are necessary with larger gains seen for reductions below 20 (Addonizio & Phelps,
2000; Biddle & Berliner, 2002; Jeremy D. Finn, 2002). Researchers increasingly contrast

Perspectives on CSR 21
literature on class size reduction (or pupil teacher ratio) and analyses of more intense reductions
of small group or one on one tutoring. Early analyses included these two types of research,
which probably over-estimate the positive effects of class size reduction (Glass & Smith, 1979).
Strong outcomes are asserted for tutoring or groups smaller than three (Odden, 1990; Slavin,
1989). While it certainly makes sense to compare class size reduction with tutoring when
making policy decisions, it is in some ways an apples and oranges comparison. Both are
strategies for allocating resources and for instruction but are quite different in intent and practice.
Joan McRobbie argues that there isn’t a specific size suggested by the literature – that this is a
wonderful example of pragmatic policy implementation:
There is no magic number but the small classes in the STAR study were between 13 and
17 students. So then states look at that and say how much can we afford and come up
with a number like California did, 20. There is nothing you can point to in research that
says well 20 is the magic number. (McRobbie interview)
Grissmer’s (1999) analysis of duration issues draws support from the early intervention
literature that links length of intervention with longevity of benefit, suggesting that more years in
small classes will produce larger and sustainable outcomes for students. Other researchers have
agreed (Biddle & Berliner, 2002; Jeremy D. Finn, 2002)
Teaching practices
It’s not just having small classes but the teachers have to change what they actually do. . .
.Teachers have to do things differently with a small class, they have to give students more
attention, that’s theory but if they are going to get the effects from that, then they have to
carry out the theory (Webb interview)
One of the most pervasively held beliefs among researchers is that class size reduction, in
and of itself, is only one piece of a complex puzzle of schooling. Although the policy implies
that teaching practices would shift in response to the smaller group, researchers have consistently
found that teachers use the same strategies in smaller and larger groups (Ehrenberg et al., 2001;

Perspectives on CSR 22
J.D. Finn et al., 2003; Robinson, 1990; Slavin, 1989). While teachers suggest that they provide
more individualized instruction, observations of teaching practice do not bear this out. Finn et al
suggest that teacher practice changes in quantity (more or less of what they already do) rather
than quality (doing things differently).
If we take the no change finding as true, it is important to theorize it. Most researchers
suggest structural explanations for this lack of change, asserting that schools, as complex
systems, are resistant to long standing practices that have served the system well.
Overall, the weight of the evidence tilts strongly toward a conclusion that reducing class
size, by itself, does not typically affect the instructional activities that occur in
classrooms. . .The finding that teaching practices do not vary with class size is consistent
with recent work on school restructuring. Observers report that teaching methods are
highly resistant to changes in school structure. . . This does not mean teachers cannot
change practice along with class size reductions, but it may take time, and may require
opportunities for teachers to learn about other approaches to teaching. (Ehrenberg et al.,
2001)
Hanushek suggests another explanation – asserting that teachers do not seem to change their
practice, he highlights the fact that teacher quality is more important than group size:
The fact that learning in a classroom is so dominated by the quality of the teacher that it
overwhelms any of the small differences in class size we are normally talking about. And
secondly, the fact that for the most part, the evidence seems to suggest that with changes
in class size, teacher don’t change much what they do. So there is no reaction.
(Hanushek interview)
For Hanushek, we should be focusing on teacher quality rather than class size. The variation
among teachers accounts for the variation in student performance and when class size reduction
is implemented at the state level, it increases the number of teachers needed in what he would
characterize as a shallow pool of talent. Employing poor quality teachers in more classrooms
compounds the problem of low teacher quality (E. A. Hanushek, 1999).
This finding is contrasted with the assertion that small groups, regardless of the perceived
teaching practices, differ in quality from larger groups (Biddle & Berliner, 2002). This was

Perspectives on CSR 23
specifically analyzed in Finn et al’s review (2003) where they suggested that teachers in smaller
classes get to know their students better because there are more opportunities for interaction and
connection between life at home and school. Further, teachers tolerate a wider range of student
behavior in smaller classes. These changes provide a context in which students become more
engaged and therefore have more opportunity to learn.
Gamoran suggested a differential effect that was developmentally and institutionally
fixed.
The kinds of activities that go on in first grade and kindergarten classes are ones that
benefit from smaller classes. Whereas the kinds of activities that predominate in classes
with older kids, doesn’t make so much of a difference. By this I mean, by third, fourth,
fifth, certainly middle school, teachers are basically lecturing and engaging in kind of
question/answer recitation, and assigning seatwork. Well, those are activities that it
doesn’t matter if you have 27 or 17 kids, if you are lecturing and doing seatwork. . . .
Whereas in kindergarten, first grade, . . . there is a lot more small group work, there
might be more one-on-one-contact with teachers. Certainly there is more of a concern
with establishing a nurturing environment. So those are the kinds of activities that might
especially benefit from having small classes. (Gamoran interview)
One suggestion made by researchers in the literature and interviews related to teacher
professional development was that any investment in a structural change needs to be
accompanied by support for teacher change. Joan McRobbie made the case that:
You often hear people say what should accompany smaller classes is professional
development of teachers so that they learn strategies for teaching in small classes but then
no none can seem to answer the question what are those strategies, what do you do
differently if you have a class of 35 versus a class of 20 or whatever. (McRobbie
interview)
Echoing the findings on teacher practice, Odden (1990) pointed to work from the STAR study
that noted that professional development related to teaching strategies for smaller groups had
little effect on grade three teaching practice. Jeremy Finn alluded to the issue of teacher quality,
noting that an influx of new teachers in a class size reduction context paired with the need for
retraining current teachers makes a system of professional development especially important:

Perspectives on CSR 24
When inexperienced teachers are placed in classrooms without adequate preparation, the
result can be a level of disorganization that requires weeks or even months to remedy.
These dynamics can easily offset the academic benefits of small classes. Brian Stecher
and his colleagues recommend that CSR initiatives be undertaken slowly and with careful
planning. Well designed programs of professional support and development can also
help. (Finn, 2002)
McRobbie did have some ideas at the level of professional community. She thought of
professional development systemically, as something that was not a “make and take” approach
provided in an isolated inservice but instead was based on and generated a stronger community
of colleagues:
The key issue is figuring out how to make sure the teachers get the kinds of professional
development that they need in order to be able to do the job, as opposed to the kind where
you just go off to a hotel to a conference all day . . ., the kind that is sort of embedded in
your daily work that has to do with mentoring and modeling and meeting with other
teachers and sitting down and saying, “Here’s what we know the kids need to learn, how
do we know if they are learning it and what do we do if they are not?” A staff working
together, the staff who work all with the same students group sitting down. (McRobbie
interview)
This is a very different image of professional development – instead of delivered to teachers by
“experts” this collaborative approach suggests that cultural change and leadership are vitally
important to the development enterprise.
Who benefits from class size reduction?
The question of who benefits can be addressed from a variety of perspectives. At the
level of policy analysis the typical perspective is focused on who has the largest outcomes.
When we compare the varied subgroups that have participated in CSR studies, researchers are
typically united behind the idea that smaller classes are most beneficial for children living in
poverty and for children of color. For this reason, most analysts suggest targeting the resources
allocated for CSR to these groups.
Class size reduction in the early grades helps kids who are at risk primarily. In other
words, it’s not an across the board thing. That doesn’t mean that middle class kids aren’t

Perspectives on CSR 25
helped, they are helped by class size reduction, but the evidence suggests that the amount
of help is greater for kids from minority backgrounds and from impoverished homes
(Biddle interview)
In our interview, David Grissmer distinguished middle class students from middle class parents,
noting that it might be “sort of an insurance policy for high income parents,” because they will
have to spend less time on homework and out of school support of student learning. A number
of middle class parents found they wanted such an insurance policy in Wisconsin when the initial
SAGE studies found positive outcomes for children living in poverty. The program was
subsequently offered to all districts with reimbursement provided only for the low income
students served.
Almost all researchers pointed to the incredible political popularity of class size reduction
– that voters, particularly parents, found it an easy to understand, compelling program. This was
surely the case in Florida, where voters mandated class size caps through a popular vote. Eric
Hanushek described the political payoff in California
Well, my colleague Pete Wilson started this whole thing as governor of California, his
popularity doubled within a two or three week period when he announced class size
reduction policy for the state. And so, for that reason, 20 plus governors had announced
their own class size reduction policies within a month and the President of the United
States announced his class size reduction policy within a month and a half. (Hanushek
interview)
Teachers were seen as the other beneficiaries of class size reduction. Increases in job
satisfaction, reduction in stress and a perceived reduction in work load were the benefits
suggested for teachers. Norman Webb noted SAGE’s popularity with teachers and suggested
that its support from the state teachers union was instrumental in its strong position in the state
educational program. Changes in California’s program have met with much resistance by
teacher and parent groups according to McRobbie:

Perspectives on CSR 26
Every time it comes up to make some kind of change, groups like the CTA and PTA (the
Teachers’ Association and the Parent Teaching Association) slam the brakes on because
they are afraid of losing it. They are afraid that it will get watered down and that that
would be a first step toward losing it. (McRobbie interview)
The political power of teacher organizations to mobilize and leverage policy was often framed as
pork barrel politics—on the one hand it diminished the notion that policy is implemented in an
impartial manner relying on “science.” For researchers like Eric Hanushek, it was guild
mentality – it made the job of teaching easier and increased the demand for new teachers. For
others it biased teacher reports about their practice and therefore the results of many studies–
they were likely to react positively to CSR because it was a good thing for them. This was the
underlying image in Gamoran’s caution about relying on teacher report data:
So it’s really important to get beyond teachers’ opinions. Their opinions are more
favorable because the experience is better, because it’s an easier job. It’s really important
to just not take what they say at face value and try to make an objective assessment of
what they are experiencing (Gamoran interview).
On the other hand, the recognition that this reform had the backing of powerful political groups;
that it was part of deals forged for both educational and political reasons during a period of
relative budget affluence was surprisingly refreshing.
What research do we need?
In the most glaring example of how the literature does not talk, we found a cacophonous
set of when we asked what kind of research was needed to settle the question of class size
reduction. Responses ranged from Hanushek’s design that addressed all the weaknesses in the
STAR study, noted earlier in the paper. Adam Gamoran also favored another experiment,
building on the use of high stakes tests for a measure of efficacy and paired with tightly designed
observational studies nested within the experiment to detail instructional practices.
But I would like to see more quantitative work that would use the high stakes tests the
states are always using, that are already using and would try to quantitatively or

Perspectives on CSR 27
statistically link instructional activities with student outcomes. That would be key.
(Gamoran interview)
Though he didn’t call for a randomized experiment, Norman Webb focused on large scale
comparative studies and multimethod studies that linked teacher practice and student outcomes.
Variations on this theme came from those who thought we needed good cost-benefit analyses
(Achilles, Webb, McRobbie) or longitudinal analyses that followed students into later schooling
(Achilles, Biddle; Molnar). All of these researchers wanted to fine tune our knowledge of the
effects of class size reduction through more detailed studies.
Experiments were much less important to others. For example, both Biddle and Achilles
had a “been there done that” mentality – they felt that STAR had done the job of telling the story
that could be told with an experiment:
I don’t think we need another experiment. I think STAR answered that question. If you
want to do one, if you don’t believe that when they put hydrogen and oxygen together
you get water, then you can go ahead and do it. We actually ran that experiment
manipulative no variables but class size, despite what people say with a caveat that we
must let schools run as they run. (Achilles interview)
Instead of seeing experiments as the gold standard for research (a pervasive concept in education
research recently), David Grissmer came at the issue from a theory of experimentation—he knew
the limits of what we can know from experimental studies
While more experimentation seems essential to making progress in educational research,
experiments can never be depended on to solve all the complex and contextual effects.
Educational research will probably never follow health research where trials are needed
for every new intervention before implementation. (Grissmer, 1999, p. 239)
In its place, Grissmer suggests that we should be focusing on the why question at this point – that
by asking why we get the outcomes we do in research on class size reduction, we can begin to
develop what we called earlier a theory of action:
One of the key problems in the research community is that we are not oriented to
answering the question “Why do things work?” We tend to measure class size effects
and go home rather than say, “Well, why do class size effects work, why do smaller

Perspectives on CSR 28
classes work?”. . . [We need to] begin to develop a theory of what happens that improves
achievement. That is, experiments only give us sort of a single measurement. What
science tries to do is develop a theory to explain several measurements. Parents do this
and teachers do this . . . But it’s a theory, the scientific consensus forms around a theory.
(Grissmer interview).
A similar approach was suggested by Joan McRobbie, who thought we needed to be much more
synthetic and derivative in understanding of class size reduction. Rather than focusing
exclusively on new studies, McRobbie thought we needed a better understanding of the existing
literature, with attention to distilling it for policymakers
If you don’t [make recommendations] then you have policy makers just sort of drifting
and it seems like you really have to try to give them information about what’s been
learned by other policy experiments. So I put emphasis on what you can take from all of
the sort of state wide policies, or even local policies, that have been designed so far that
are about class size reduction. What can you make of all of that in terms of the wisdom
that has now been built about designing a policy based on the research that’s been done.
(McRobbie interview)
Reading the literature and suggesting next steps for researchers is very much a situated activity.
The range of responses to this question illustrates that what we know about a particular topic is
shaped by how we know it – by the perspectives we bring to the literature, by the criteria we use
for judging its adequacy, and the communities of practice to whom we speak. For some, the
aggregation of data and findings and increasingly tight experimental designs are key to our
knowledge. Through replication and better specification of findings we can assert with surety
the outcomes of treatments. For others, longitudinal designs examining long-term effects are
favored. And others look for policy relevant synthetic analyses that focus on the how rather than
the what. From this perspective, treatments happen in contexts that make their effects contingent
and often involve tradeoffs of other services and investments. While it would be tempting to
suggest that perhaps all these researchers are right, that in a perfect world, we would have all
these types of studies going on at the same time, we are not exactly in that perfect place. In

Perspectives on CSR 29
contrast to the context that prompted much of the work on the large state implementation,
characterized by budget surplus and interest in multiple types of inquiry we are now in a time of
budget retrenchment and increasingly narrow views of what counts as data or as science. In the
final section we add our voices to this conversation, suggesting what we have taken away from
our reading and interviews related to class size reduction.
Discussion
The large and varied literature related to class size reduction rests on early meta-analyses and re-
analyses that variously find positive effects on student outcomes. More recently, three major
state initiatives have given us a practiced based view of CSR. STAR provides a state-sampled
experiment of various CSR treatments, SAGE provides a phased-in implementation of a targeted
intervention, and the California implementation provided resources to the entire state for a
midrange class size reduction of 20:1 in kindergarten to grade three. At this point, we know that
none of the research is perfect – each had strengths and weaknesses that reflect the politics and
compromise of the moment. But across these research conversations, we would have to say that
we think we are getting good enough research – that taken as a group, we can begin to form a
fuzzy picture of socially held knowledge of class size reduction.
The biggest take home message is that class size reduction is not pupil teacher ratio. As
long as these two frameworks for allocating resources are conflated we will have continued
disagreement about how CSR affects student outcomes. While PTR is the simplest indicator, it
often does not reflect the day-to-day life in the classroom and is a good example of how a proxy
is not always the best representation of a construct.
Considering CSR as a prevention strategy leads to consideration of how researchers have
examined the intervention in terms of timing, intensity, and duration. While most researchers

Perspectives on CSR 30
agree that earlier is better, it has also been pointed out that the research base would be stronger if
more studies were done at the intermediate and middle school level. The question of how small
groups have to be generally is answered as at least 20 (California’s strategy) but better at 15.
One confound in these conversations was the inclusion of extremely small groupings – tutoring
or pull-out programs. There is certainly a choice in terms of investment that is implied (do you
reduce class size overall, do you reduce for particular content, or do you provide pull-out small
groups) but including it within the literature seems to mix apples and oranges.
Researchers generally state that what teachers do is as important as how many students
they have—that instructional practices must change if CSR is to be effective. Most researchers
we read and talked with could not point to work that showed changed – in fact, they held that no
change occurred. One question we find ourselves asking is what constitutes a change in teacher
practice? What has to change for researchers to recognize it as something different? If the
quantity of certain practices change, aren’t things different? One way that we might better
understand the outcomes of class size reduction is to have a better sense of the nature of
instruction in the varied CSR instructional contexts.
If CSR is an intervention, should it be targeted to those most in need or should it be
universally administered as a benefit to all? Whose benefit should be considered when such a
program is implemented? Most researchers agreed that CSR had the greatest potential for
students typically seen as at risk—children of poverty and of color. Others extend the notion of
benefit to middle class families, suggesting that it is a kind of insurance policy for students who
are likely to do well in school anyway. The immense popularity of CSR was also recognized –
parents wanted the benefit for their children and teachers wanted it to make teaching easier. This

Perspectives on CSR 31
popularity made changes in the policy difficult to leverage as the electorate and lobbying groups
worked hard to maintain an easy to understand policy alternative.
One of the first things we noticed when we got into the literature on class size reduction
is that it is anything but dry and dispassionate. There is more sniping and namecalling than we
ever thought there would be. One interview was practically unusable for this paper as the
researcher spewed so much venom about state education officials he disagreed with and about
other scholars who have researched the topic. Why do people get hot under the collar when
talking about how many kids to put in a classroom? Our hunch, after much reflection, is that it
turns on two key issues. The first is that the conversations about class size reduction are
flashpoints for the issue of what counts as evidence in education research. When researchers
don’t see their value system operationalized in inquiry they raise flags about the work’s
adequacy. The rules of the game are not unitary, but are shaped to the perspectives of varied
communities of researchers. No wonder policy makers have a hard time making sense of what
we do – researchers are not a unified “we” but instead are more likely characterized as tribes of
likeminded scholars. This played out specifically in terms of what researchers saw as the
findings of the research and their suggestions for next steps.
The second point is that the policy of class size reduction turns on basic issues of equity.
It forces us to think about the basic inequality of schooling, the mechanisms that might mediate
those differences, and the investments that we are willing to make as a society to put success in
reach of all children. The very idea that people would think that achievement would go up just
by changing the number of students in a classroom is typical of the pie in the sky thinking that
afflicts much of education. Class size reduction is not a silver bullet that can ameliorate the
damage that poverty, violence, inadequate child care, etc set in the path of children coming to

Perspectives on CSR 32
school for the first time. It is enacted in the institutional context of high stakes testing,
crumbling buildings, increasing numbers of children with high needs, and competition from
other programs for teacher attention and effort. Adding resources for class size reduction is a
nod toward the inequities that form the foundation of schooling but it is not enough. CSR is part
of a system of reforms and problems that will be most fruitfully considered in a coordinated
manner—in terms of both the practice and research of schooling. It necessitates implementation
that connects the utilization of the resources for class size reduction with all curricular,
administrative, and institutional efforts that shape teaching and learning. It will require many
different kinds of inquiry – large scale studies that allow disaggregation of results as well as
small scale studies of process and mapping of social relations and practice. Berger was right –
focusing on class size reduction alone is like trying to decide the right amount of butter in a
recipe. We have a better sense of the butter and now it is time to take a look at the other
ingredients.

Perspectives on CSR 33
References
Addonizio, M. F., & Phelps, J. L. (2000). Class Size and Student
Performance: A Framework for Policy Analysis. Journal of
Education Finance, 26(2), 135-156.
Berger, M. A. (1982, April 17-20). Class size is not the issue.
Paper presented at the National School Board Association,
Atlanta, GA.
Biddle, B. J., & Berliner, D. C. (2002). Small Class Size and Its
Effects. Educational Leadership, 59(5), 12-23.
Cullen, S. (2005, February 10, 2005). District says proposal
would help with shortfall. Wisconsin State Journal, p. A5.
Durlak, J. A. (1997). Successful prevention programs for children
and adolescents. New York: Plenum.
Ehrenberg, R. G., Brewer, D. J., Gamoran, A., & Willms, J. D.
(2001). Class size and student achievement. Psychological Science
and the Public Interest, 2(1), 1-30.
Finn, J. D. (2002). Small Classes in American Schools: Research,
Practice, and Politics. Phi Delta Kappan, 83(7), 551-560.
Finn, J. D., & Achilles, C. M. (1990). Answers and Questions
about Class Size: A Statewide Experiment. American Educational
Research Journal, 27(3), 557-577.
Finn, J. D., Pannozzo, G. M., & Achilles, C. (2003). The "whys"
of class size: Student behavior in small classes. Review of
Educational Research, 73(3), 321-368.
Gilman, D. A., & Kiger, S. (2003). Should We Try To Keep Class
Sizes Small? Educational Leadership, 60(7), 80-85.
Glass, G., & Smith, M. L. (1979). Meta-analysis of research on
class size and achievement. Educational Evaluation and Policy
Analysis, 1(1), 2-16.
Grant, C., & Graue, M. E. (1999). (Re)viewing the review: A case
study of the Review of Educational Research. Review of
Educational Research, 69(4), 384-396.
Grissmer, D. (1999). Class Size Effects: Assessing the Evidence,
Its Policy Implications, and Future Research Agenda. Conclusion.
Educational Evaluation and Policy Analysis, 21(2), 231-248.
Hanushek, E., & Rivkin, A. (1997). Understanding the twentieth-
century growth in U.S. school spending. Journal of Human
Resources, 31(1), 35-68.
Hanushek, E. A. (1999). Some Findings from an Independent
Investigation of the Tennessee STAR Experiment and from Other
Investigations of Class Size Effects. Educational Evaluation and
Policy Analysis, 21(2), 143-163.
Hedges, L. V., & Stock, W. (1983). The effects of class size: An
examination of rival hypotheses. American Educational Research
Journal, 20(1), 63-85.
Hruz, T., & Wisconsin Policy Research Institute. (2000). The
costs and benefits of smaller classes in Wisconsin : a further
evaluation of the SAGE program. Thiensville, WI: Wisconsin Policy
Research Institute Inc.
James, J., & Johnson, C. (2005, February 15, 2005). Bush yields
on class sizes. St. Petersburg Times.

Perspectives on CSR 34
Molnar, A. (2002). School Reform Proposals: The Research
Evidence. Research in Educational Productivity Series (No. Vol-
2). Connecticut.
Molnar, A., & Zmrazek, J. (1994). Improving the Achievement of
Wisconsin's Students. Urban Initiative Task Force Recommendations
and Action Plan. Bulletin No. 95079. Wisconsin: Wisconsin State
Dept. of Public Instruction, Madison. Bureau for Policy and
Budget.
Nye, B., Hedges, L. V., & Konstantopoulos, S. (2001). The Long-
Term Effects of Small Classes in Early Grades: Lasting Benefits
in Mathematics Achievement at Grade 9. Journal of Experimental
Education, 69(3), 245-257.
Odden, A. (1990). Class Size and Student Achievement: Research-
Based Policy Alternatives. Educational Evaluation and Policy
Analysis, 12(2), 213-227.
Robinson, G. E. (1990). Synthesis of Research on the Effects of
Class Size. Educational Leadership, 47(7), 80-90.
Slavin, R. (1989). Achievement effects of substantial reductions
to class size. In R. Slavin (Ed.), School and classroom
organization (pp. 247-257). Hillsdale, NJ: Earlbaum.
Smith, M. L., & Glass, G. V. (1980). Meta-analysis of research on
class size and its relationship to attitudes and instruction.
American Educational Research Journal, 17(4), 419-433.
Smith, P., Molnar, A., & Zahorik, J. (2003). Class-Size Reduction
A Fresh Look at the Data., Educational Leadership (Vol. 61, pp.
72-74): Association for Supervision & Curriculum Development.
Webb, N., Meyer, R., Gamoran , A., & Fu, J. (2004). Participation
in the Student Achievement Guarantee in Education (SAGE) Program
and Performance on State Assessments at Grade 3 and Grade 4 for
Three Cohorts of Students —Grade 1 Students in 1996–97, 1997–98,
and 1998–99. Madison, WI: Wisconsin Center for Education
Research.
Zahorik, J., Halbach, A., Ehrle, K., & Molnar, A. (2003).
Teaching Practices for Smaller Classes. Educational Leadership,
61(1), 75-77.
Zahorik, J., Molnar, A., & Smith, p. (2003). SAGE advice:
Research on teaching in reduced-size classes. Tempe: Edudcation
Policy Studies Laboratory.