THERE is a highly significant education statistic that every parent, teacher and policy-maker should know about. Pupils in the most effective classes learn at four times the speed of those in the least effective classes.

What accounts for this remarkable difference? Dozens of explanations could be offered, but one frequently-cited factor is class size. Teachers of smaller classes can, it is said, respond to pupils more quickly and more effectively, and can develop a better understanding of individual children's abilities
and needs.

Studies on both sides of the Atlantic have, however, repeatedly found that having a good teacher is much more important to children's
progress than class size. Provided pupils are not unruly, classes of 30 can be as effective as those with 20.

It is true that five and six-year-olds tend to thrive in classes of no more than 15. It is unclear why this should be so – the
extra help with reading that teachers are able to give pupils
in small classes seems to be a key factor.

But we do know that such classes speed up the learning rate by 50 per cent, on average, and by up to 100 per cent for disadvantaged and minority-ethnic pupils. The extra expenditure that very small infant classes incur can, therefore, be justified.

However, my colleague, Professor Peter Blatchford, who led the biggest UK study into the effects of class size, found no evidence that smaller classes produced learning gains for older pupils.

One American study that reported similar findings concluded that teaching 120 eight- and nine-year-olds in classes of 20 rather than 30 would result in five more pupils passing a test.

The cost of this improvement? Some £60,000 a year.

Did anyone think that smaller classes can be created cheaply?
A recent report for the Scottish Government estimated that reducing the country's primary classes to 20 could add £309m
to the annual salaries bill for teachers.

Recruiting all those additional teachers would be far from easy, of course. And unless recruitment has been mishandled in the past, the extra teachers you bring in will not be as good as the ones you already have, so teacher quality will go down.

California found that when it reduced all infant classes to 20.
It also discovered, too late, that the creation of thousands of
new teaching posts encouraged many of the best teachers in inner-city schools to head for the wealthier suburbs.

Should we therefore employ more classroom assistants
rather than slash class sizes? Perhaps, surprisingly, there is no evidence that having an additional adult in a classroom improves learning, even though an assistant can reduce a teacher's stress levels.

We should, therefore, concentrate on what has been described as the "love the one you're with" strategy. By that, I mean improving the quality of the teachers we have.

But is it really possible to raise the quality of serving teachers? Ten or 15 years ago, the answer might have been "no". Most of the countless millions spent on teacher workshops and conferences had led to no significant change in practice.

Within the past few years, however, a clearer picture of effective professional development for teachers has begun to emerge. We now know that teacher development is more effective when it is closely related to a school's local circumstances

We also know that it has to be sustained over perhaps a couple of years rather than taking the form of one-day workshops.

Finally, we are also now sure that teachers learn best in collaboration with their colleagues. Simply telling teachers how to change their practice does not work.

For example, most teachers have heard about research from the 1980s which shows that if they wait three to five seconds after asking a question, their pupils' performance improves because they have been given some time to think. Even so, many teachers are still allowing less than a second for pupils to respond.

Knowing what to do is the easy part of teaching, it seems. Actually doing it is what's hard – particularly if you have taught in a different way for 30 years. Professional development of the kind I have outlined will help to counter this problem. And the most effective training of all will encourage teachers to adjust their instruction through minute-by-minute and day-by-day analysis of their pupils' learning.

Research shows that this approach, known as formative assessment, could prove 20 times as cost-effective as reducing
class size in terms of raising achievement.

But, as I have indicated, if I had to spend money cutting class sizes, I would direct it to infants from disadvantaged and minority-ethnic backgrounds. That might be politically contentious but it would produce substantial achievement gains.

It would also have one less-obvious bonus that would delight headteachers: it would encourage the more capable teachers to stay in the most "difficult" schools.

Professor Dylan Wiliam is deputy director of the Institute of Education, University of London.