The state courted controversy by suddenly slashing infant classes from more than 30 to 20 in 1996. Critics claim the vote-winning initiative has worsened staffing problems in the inner-cities where many teaching posts have been filled by unqualified recruits.

But a new study involving more than 42,000 Los Angeles children suggests that the policy has led to big improvements in maths, writing and spelling, and a moderate increase in reading scores. The researchers also found that disadvantaged children benefit most from smaller classes. On average, they gain twice as much as infants in more advantaged areas.

But the study shows teaching quality is important. “On average, the benefits of class-size reduction are about 50 per cent higher for teachers with six or more years of experience,” said Dr Harold Urman, who compared test scores of eight-year-olds in 1998-99 with those from 1996-97. He said experiened teachers were better at taking advantage of the benefits of smaller class He accepts that other factors could have affected the scores. “It is possible that the 1998-99 cohort of students experienced more ... practice examinations and teaching to the test than the 1996-97 comparison group,” he said.

But he suspects his study may have “greatly underestimated” the positive effects of smaller classes. In 1998-99 most third-graders had spent only two years in smaller classes and none had been in small kindergarten classes, he said.

This is significant because the Tennessee STAR experiment - the best-known class-size reduction programme - showed that it is the youngest children who gain most from smaller classes.

Though other research has been less emphatic in backing the policy Dr Urman, who works for Vital Research, a consultancy firm, has urged policy-makers to “stay the course”.

Contact: Dr Harold Urman urman@vitalresearch.com. Tennessee STAR programme www.nashville.netherosstar.htm