Research, commissioned by a key government adviser, shows that pupils rated among the brightest prospects at primary school go on to under-achieve at GCSE, The Times has learnt. Some do only nearly half as well as their peers in good schools.

The most politically explosive finding was of a direct relationship between the number of bright children in a school and individual achievements.

The study highlights the scale of the challenge facing Ms Kelly in tackling poor secondary schooling, particularly in deprived urban areas.

It emerged as the latest edition of The Times Good University Guide shows that universities plan to devote huge sums of money trying to satisfy government demands that they widen access to students from poor backgrounds.

The research, by David Jesson, of York University, used government data to track the progress of 28,000 children who scored the highest marks in national curriculum tests of English and mathematics at the age of 11.

They represented the top 5 per cent from more than half a million pupils in England who take Key Stage 2 tests in primary schools each year.

Professor Jesson found that nearly 6,000 pupils who took the tests in 1999 were admitted to 167 selective grammar schools and 5,800 went on to 223 high-achieving comprehensives. The remaining 16,500 went into 2,407 comprehensives, many in urban areas, with lower achievement.

When the same students took their GCSEs last summer, many had effectively been lost because schools failed to push them to reach their potential.

Professor Jesson found that success rates declined in line with the numbers of bright children in a school, and dipped sharply when there were fewer than five.

Where 20 pupils from the most able 5 per cent were clustered together in a year group, each achieved an average of nearly seven GCSE passes at A* and A grade last year.

But where there was just one child from this group in a school, he or she passed fewer than four GCSEs at these grades. This is likely to have a severe impact on prospects for university admission. The children’s performance at A level will be followed to establish how many of those who could be expected at 11 to be candidates for Oxford, Cambridge and other top universities actually achieve the necessary grades.

An analysis of results in 2002 showed that comprehensive students had only a 5 per cent chance of getting three A grades, the standard expected at Oxbridge and many other elite universities. Nearly 20,500 18-year-olds achieved three A grades. But of the 110,000 who took A levels in comprehensives, only 5,821 reached this standard. This compared with 3,394 out of 18,265 (19 per cent) among sixth-formers who took A levels in grammar schools. At independent schools, 7,565 gained three A grades out of 32,873 candidates (23 per cent).

Professor Jesson found that individual pupils in high-achieving comprehensives scored slightly better at GCSE than those in grammar schools. This suggests that provision for the most able children within schools, rather than selection, at 11, is the critical issue.

Sir Cyril Taylor, chairman of the Specialist Schools Trust, which commissioned the study, said that it showed the need for a systematic search to ensure that the brightest pupils fulfilled their potential.

“Clearly, the implication is that if you have only one or two children of high ability, then they get lost in the system and don’t get the support they need. A lot of these kids are not delivering their potential,” he said. He added that the Government was spending £300 million on widening access in universities to students from state schools and poor backgrounds, much of it on programmes to prevent them dropping out. Some of the money would be better spent on supporting the most able children at school.

“The way to solve the access issue is to identify these kids, find out which schools they have gone to and insist that they get the nurturing their potential requires. If we have the data, then why not use it?

“The Key Stage 2 scores in maths and English could be a very quick and inexpensive way of identifying the top 5 per cent of the ability range. Then you could track those kids through the national tests at 14 and on to GCSE and A level.”

Explore Education

• Schools

• GCSE & A Level Results

• Good University Guide

• Sunday Times University Guide

• Student

• Young Times

Times recommends

• Google generation has no need for rote learning
• Left-handers do 'less well' in exams
• Nativity comic books for multicultural Britain