Dramatic improvements in primary English, maths and science test results during Labour's first term have not been replicated at GCSE, a TES analysis shows.
Pupils who registered staggering gains in key stage 2 assessments in 1999 have failed to reap the benefits five years later.
The findings raise fresh questions over one of the Government's biggest education success stories - improvements in primary schools. They come as a leading academic suggested that the primary tests had failed to measure pupils' achievements accurately during the 1990s.
Professor Peter Tymms, of Durham university, said KS2 gains were a result more of pupils becoming better prepared for the tests than of any underlying teaching improvements.
When this year's 16-year-olds took their KS2 tests five years ago, the results were stunning. The proportion of youngsters achieving the benchmark level 4 in English, maths and science rose by an average of nine percentage points in 1999.
Yet when the same cohort sat their GCSEs this summer, improvements were much more modest (see box). Across all subjects, the proportion of grades awarded a C or better rose only one percentage point. The improvement was even smaller in English (0.3 points). There was a one point increase in science. In maths the rise was 1.8 points, but results were still only 0.7 points up on those of two years ago.
KS3 results for the same cohort, in 2002, also showed only modest gains, the proportion reaching the benchmark level 5 rising two points in English, and one each in maths and science.
Experts suggest two explanations for the disparity: that the KS2 improvements were illusory, or that they were accurate but that pupils have not sustained improvements in secondary school.
Dr Robert Coe, who works with Professor Tymms, said the former explanation was more plausible.
Professor Tymms has looked at 11 sets of data on the achievements of primary and early secondary pupils in 1995-2003. Results that appear in the British Educational Research Journal show he found that although there was some evidence that pupils' KS2 maths and reading skills had improved, official test scores overestimated gains for writing. He writes that much of the KS2 gains in the years 1995-2000 could have come about as a result of inadequate checks on test standards, and better teaching to the test.
However, Sheila Dainton, education policy adviser at the Association of Teachers and Lecturers, said the lack of any leap forward in GCSE results was more likely to reflect the fact that pupils had slipped back in secondary school. The Government had since sought to address this, introducing the KS3 strategy in 2001, which had come too late for this year's GCSE cohort.
A spokeswoman for the Department for Education and Skills said: "This summer's results showed good progress in raising school standards and steady progress in education since 1997."