likely complex relationships between class size and pupils’ educational progress, while
controlling for likely confounding factors. One advance over previous research was the
modelling and graphical representation of relationships throughout the distribution of
class sizes. A simple linear relationship cannot be assumed.
There are a number of difficult issues involved in any comparison of non-experimental
vs. experimental research designs, which cannot be entered into here. One of the
potential problems with non-experimental designs is that there may be other possibly
influential variables which have not been included in the analysis which might explain
relationships between class size and outcome. This is a possibility the reader would need
to bear in mind, though every effort was made to include likely factors. The use of
previous attainments test scores makes it very unlikely that our results could be
explained by the allocation of children of different abilities to different class sizes - a
problem that bedevilled some earlier UK research (Blatchford, Goldstein and
Mortimore, 1998). A connected issue concerns whether there was anything systematic
about the allocation of teachers to classes of different sizes - for example that teachers
allocated larger classes were more experienced or more effective. We did not measure
directly quality of teaching in different classes and so we cannot be entirely sure about
any differences between teachers in different classes. However, in separate analyses we
have found no relationships between class size and a number of teacher characteristics,
including age, length of teaching experience, participation in in-service training and a
judgement by headteachers of teachers’ effectiveness.
We feel that the use of a multi method approach to data collection was vindicated. Again
the longitudinal nature of the research helped because it was possible to adapt methods
in the light of emerging results so that methods became progressively compatible,
internally consistent, and deliberately complementary. So, for example, results from the
systematic observations during the reception year were clear about the importance of
individualised instruction and the greater amount of peer interactions in larger classes,
and these informed the questions asked in the end of year teacher questionnaires.
Methods were adjusted and fine-tuned in the light of early field visits and data analyses.
Another connected feature was the presentation of early results to teachers involved in
the first year of the research. We were able to discuss provisional findings with them and
together work through some possible explanations and possibilities of changes to data
collection techniques and additional information needed.
We avoided the choice of instruments/methods on the basis of apriori decisions about
the appropriate way of combining qualitative and quantitative approaches. Rather, our
choice was informed by debates in previous research and what methods of data
collection seemed best able to address specific research questions and dimensions
related to classroom processes considered important. In particular, in examining
relations between class size and teaching, we feel even more strongly that there is
considerable value in combining in the same study data that stems from classroom
observation, teacher report, and informed fieldworker notes.
But there are difficult issues involved in the use of mixed method approaches, that
require further attention. One issue concerns the way that different techniques are used
together. Our approach was to develop methods that were deliberately complementary.
An alternative approach, though, would be to use a particular method of data collection
to test the claims from other forms of data. There are also problematic issues when it
23