The Scottish Government's target of a maximum of 18 pupils in P1-3 is "nowhere near enough", according to one of the UK's leading experts who labelled part of its strategy as "stupid".
Dylan Wiliam, deputy director of the Institute of Education in London, argues that classes in the first two years of schooling should have no more than 15 pupils.
"If they (the Scottish Government) don't have enough money to do that, they should just cut class sizes in P1 and P2, rather than P1-3," he said.
Basing his comments on the Tennessee Star project, which he describes as the "gold standard" of research into class sizes, he argues that the American research showed there was no additional benefit in cutting pupil numbers in the equivalent of P3.
The rationale behind having very small classes in P1 and P2 is that it allows pupils to develop good learning habits, such as socialisation and reading skills, and these gains are maintained in future years.
However, at other stages of children's education, Professor Wiliam argues that it is far more cost-effective to focus resources on embedding formative assessment.
He suggests that cutting a class of 30 to 20 gives children the equivalent of four extra months of learning per year but the additional costs of teachers' salaries and school buildings make this a very expensive option of around pound;20,000 per class, per year.
On the other hand, the implementation of a formative assessment approach could provide eight extra months of educational development for an outlay of only pound;2,000, he says. This includes the costs of training, supply cover and development. That calculation depends on setting up "teaching learning communities" of groups of 8-10 teachers who meet once a month to share good practice.
Professor Wiliam, who was speaking at the annual Chartered London Teachers Conference this week, said the only caveat in his argument, that reducing class sizes to raise pupil achievement was a waste of money, occurred when pupils were badly behaved.
"Smaller classes do confer a benefit if pupils are unruly, because having fewer pupils in a class means less disruption," he commented. "But as long as pupils are well-behaved, then what you can do with a class of 20 is generally possible with a class of 30."
Professor Wiliam also argues that research from America shows that the best teachers are four times more effective than the least effective teachers.
He told The TESS: "If we are willing to pay a bad teacher pound;25,000, then, rationally, we should be willing to pay an excellent teacher pound;100,000. Teacher quality trumps almost everything else. What you do when you do reduce class size is dilute teacher quality."
Training large numbers of teachers to teach smaller classes is not cost-effective, he argues. The SNP-led Government has pledged to train 20,000 teachers by 2011, but Professor Wiliam describes that strategy as "a stupid way to spend the money". A further way to raise both teacher quality and pupil attainment would be to reduce teachers' non-contact time from the current 22.5 hours per week to 15 hours per week, he suggests. This would give teachers more time to prepare and deliver better lessons.
Ronnie Smith, general secretary of the Educational Institute of Scotland, which wants all classes reduced to 20 pupils or fewer, disputed Professor Wiliam's cost-benefit analysis. He pointed out that the average primary class size in Scotland was now just above 23, and that it would not cost as much as pound;20,000 per class to cut pupil numbers to 20.
As Scotland operated a standards-based system for teachers, the only way that teacher quality would drop was if the standard had been lowered in recent years - and that was not the case, said Mr Smith.
He argued that a good teacher teaching a class of 20 would still have more impact than if he or she was teaching a class of 40. This would particularly be the case if they were using formative assessment, because of its reliance on personalisation and individualised teaching.
Meanwhile, pupil census statistics published by the Scottish Government this week showed that the average primary class size fell from 23.6 to 23.3 last year and that 95 per cent of P1 pupils were in classes of 25 or fewer - up from 66 per cent. Only 12 per cent of P1-3 pupils were in classes of 18 or below.
Maureen Watt, the Schools Minister, said: "While it's good news that these statistics show class sizes in P1 are coming down, we are determined to do much, much more."
Leader, page 22.