content to stop here, but the fundamental limitation of this analysis is that it tells us only
about what did happen in the STAR study, rather than what might be possible with reduced
class sizes. One of the most intriguing findings of the study was that the teachers in the
study do not appear to have changed their teaching approaches very much. Clearly some
teaching approaches that are feasible with 15 students are more difficult with classes of 22
and may be impossible for the average teacher with a class of 30 students. It is possible,
therefore, that the potential effects of CSRPs may be significantly greater than has been
found to date, because few studies have systematically investigated class-size reduction
combined with inservice training for teachers on how they can best make use of smaller
classes (although there are significant exceptions, such as Blatchford, Basset, Goldstein, &
Martin, 2003).
The lesson I wish to draw from this brief example is that in education, what we can learn
from the scientifically detached approach—in the terminology used by Labaree (2003), an
approach that emphasizes the analytic, the theoretical, the universal, and the intellectual—
provides only a part of the story and the scientifically detached account is partial in both
senses of the word. It provides only a partial picture, and, more importantly, it is a picture
that is actually misleading. As the work of Blatchford et al. (2003) shows, class-size
reduction programs do offer significant possibilities for improving educational outcomes at
scale, but only when they are combined with well-designed programs of inservice support
for teachers.
So, what seemed at the outset like a very clear simple question does not have a clear simple
answer. Given the equilibrium effects, and the age-specificity, found in most CSRPs, it
seems that the costs might well not justify the benefits, and other uses of the money could
well produce better returns. If, however, the question is changed from “by how much have
class-size reduction programs improved student outcomes?” to “by how much can class-
size reduction programs improved student outcomes in the past?” then we may get very
different answers, but only if we are prepared to tangle with the messy reality of the
contingent and the variable.
The relationship between physics and engineering may be an instructive parallel here. To
design a bridge that will be safe in operation, it is necessary to know the physical properties
of the materials to be used, such as knowing that stone is a reliable material to use in
compression, but behaves unpredictably in tension, while steel and, to a lesser extent,
wood, are relatively predictable in their behavior in both tension and compression.
However, precise knowledge of these kinds of properties does not provide any guidance
about what the bridge should look like. The detailed knowledge of the physical properties
of the materials will indicate whether a particular design is likely to be effective, but they
do not, by themselves, provide guidance about how to generate the design. Designing a
bridge, requires knowledge of the properties of the materials to be sure, but at its heart is a
fundamental creative process, substantially under-determined by the physics.
In the same way, the challenge of “engineering” effective learning environments requires
knowledge of the findings of educational research, but this research, no matter how well it
is done, underdetermines what is possible. Rather than limiting the enterprise of educational
research to the pursuit of “knowledge” (in the sense of collections of items of generally
agreed upon information) and the development of theories, it seems therefore that a more