Teachers have always been worried about class sizes, but recently governing bodies and parents’ groups have been expressing anxiety too. There are several reasons for this. There is a fear that changes to the management and funding of schools have created a tendency toward larger classes. The introduction of the national curriculum and associated assessment arrangements has meant that teachers in larger classes are spending more time in preparation. With local management of schools, governors face difficult decisions about class size when faced with competing demands on funds. Parents understandably want their children to be educated in small classes, comparable to the independent sector. Teacher associations and educational pressure groups are worried about large classes and are lobbying with increasing force for smaller classes.

Recent evidence to the House of Commons Select Committee on education shows clearly the central place of class sizes in debates over the unequal funding of primary and secondary sectors. And recent reports have set strongly worded targets on class sizes. The National Commission on Education, for example, stated that within five years no primary school should have classes bigger than 30, and that this should be 20 in deprived urban areas and in schools which have high proportions of children whose mother tongue is not English.

We have contributed to this debate by providing a “briefing” for the NCE in which we examined the research evidence on class sizes and pupil-teacher ratios. We showed that average class sizes at primary level have increased over the period 1982 to 1992, from an average of 25.4 to 26.8 (in England). The number of classes in 1992 with more than 30 pupils was 22 per cent (in England). There were four times the number of large classes (ie more than 30 pupils) at primary as opposed to secondary schools.

The latest provisional figures issued by the Department for Education are that in January 1994 the average class size for primary schools was 26.9 and that 28 per cent of primary pupils were being taught in classes of more than 30. At primary level, according to international comparisons the UK has one of the highest PTRs of all industrialised countries. Most concern, therefore, has been expressed about primary schools where, in any case, pupils are more likely to be taught as a whole group.

But what does research have to tell us about the crucial question concerning the effects of different sized classes on pupils’ attainments and behaviour? Is small better? There can be few areas where something as seemingly self evident as the benefit to pupils and teachers of small classes has not been confirmed by research. Given that the financial implications of reducing class sizes are obviously considerable, it is small wonder that policy makers at national, local education authority and (increasingly) at school level are cautious.

Yet, we argue in a recently published paper, “The issue of class size for young children in school”, (Oxford Review of Education, Vol. 20, No. 4) that there simply is no research evidence available in the UK on which to make informed judgments. This state of affairs is surely not acceptable.

We have reviewed the research evidence to try and see what we can conclude with any certainty. The first and most common research design used in British and US research has involved the analysis of associations between naturally occurring classes of different sizes on the one hand, and pupils’ attainments on the other. The problem with this kind of research is that we often do not know whether the results can be explained by another factor. Statistical analyses can go some way toward controlling for different factors, but these studies are still inadequate as the basis for drawing definite conclusions about the effects of class size.

The second type of research did lead researchers to think, for a time, that they had finally found answers to questions about the effects of class size. This research involves taking the results from large numbers of projects and putting them together into one analysis. For a while it seemed to have settled the issue in favour of smaller classes, but these studies are also difficult to interpret, principally because conclusions will inevitably depend on the quality of the studies included and some of these are suspect. But most reviewers are agreed that children in the early years of school are most likely to benefit from smaller classes, at least in the basic subjects of mathematics and reading, and this applies especially to low achievers and disadvantaged pupils.

This conclusion is supported by the third and only type of research that can give us conclusive answers to the question of whether children in smaller classes do better. Research must find some way of getting over the problem that it could be something about the kinds of pupils in small or large classes which might explain any differences found. This also applies to teachers because it could be that teachers in classes of different sizes are unlike in some way which might affect pupils. So what is needed is experimental research which compares the progress of pupils who have been randomly allocated to classes of different sizes. Furthermore, teachers would have to be allocated randomly to these classes in order to make sure that it was size of class and not something to do with the teachers which made the difference.

Clearly this is not an easy kind of study to do, and it is no surprise that it has never been contemplated in the UK. But it has been done in the United States. Project STAR is a state-wide survey in the state of Tennessee which included more than 7,000 pupils in 79 schools. The results are impressive and consistent. In both reading and mathematics, pupils in small classes (13-17) performed significantly better than pupils in larger classes (22-25). This was true from kindergarten to Grade 3. Small classes with one qualified teacher had pupils who did better than pupils in larger classes with an assistant. There also appeared to be a particular advantage for pupils from ethnic minorities.

There are still reasons to be cautious. One obvious one is the extent to which one can generalise from research conducted in another country with a different education system. But there are at least three other problems. First, how much difference does it make? Some critics have agreed that substantial reductions in class size do have positive effects, but that the size of the difference is moderate. It is argued that class reductions are less effective than other and less costly reforms like one-to-one tutoring. However, one does need to compare like with like: in the STAR research, only the number of pupils to adults was changed - no advice was given on instructional practice. It is also important to bear in mind the age of child when comparing the effectiveness of different initiatives. In view of the STAR and other research, it seems clear that class size reductions are likely to be more effective in the first years of school.

A second concern is that class size reductions are too costly. But a distinction needs to be drawn between across-the-board reductions in class size, which may well be too expensive, and targeting class size reductions at the earliest years in school, which will be both cheaper and happens to be supported by research evidence.

A third question is how much do class sizes have to be reduced in order to be effective? There is unlikely to be an exact answer, though there is some consensus from the American research that reducing class sizes by a few pupils across the board is unlikely to be effective and that effects are unlikely to be marked until classes are reduced to below 20.

If the STAR research does show that pupils in small classes do better, it is less clear how the effect works: in which ways do reduced class sizes affect classroom processes, the curriculum, behaviour and relationships? Here we have very little evidence. However, it seems that teachers do not always alter their teaching when faced with fewer pupils. They have to take deliberate advantage of the opportunities created to show some gain. Class size reductions and teaching practices therefore need to be considered together.

It seems to us that the case for research into the effects of class size differences is obvious. Although a focused study of differences between classes of naturally varying sizes is possible, there will always be the problem of ascertaining cause and effect. For this reason we believe that there is a strong case for an experimental approach which would replicate and extend the STAR project.

The cautious should not be fearful of research on class size. There is no guarantee that results would favour small (or larger) classes. But the central place of class size in affecting the experiences of all pupils and teachers surely requires us, deliberately, to collect as clear evidence as we can on this vital issue.

Professor Peter Mortimore is director of the Institute of Education, University of London and Peter Blatchford is a senior lecturer.