Peter Tymms, reader in education at the University of Durham, said that children achieving 50 per cent in the key stage 1 reading comprehension test could have a true score of anything from 17 to 79 per cent. "This is an extraordinary range of uncertainty to accompany an official reading test, " Dr Tymms said.
Seven-year-olds' tests are shorter than those given to older children. He said: "No one wants to have long assessments for seven-year-olds, and yet reliable assessments have many items in them. If an assessment has just a few questions some less able pupils will be lucky and get them right. The opposite can also happen when very competent pupils get seemingly easy questions wrong."
Dr Tymms believes that the test's unreliability raises questions about the quality of all statutory assessments of seven-year-olds, although he sees England as a world leader in national testing.
He said: "Many children are likely to have been misjudged and parents will be unaware of this. It is a problem for schools because they are unable to benefit from high-quality information. The unreliability of this test means that value-added scores will be suspect and there will inevitably be difficulties with target-setting."
However, the Qualifications and Curriculum Authority, which is responsible for the tests, rejects the criticisms. It insists that the reading test is extremely accurate, considering the age of the children.
A spokesman said that if a child had a standardised score of 110 in the level 2 reading comprehension test a teacher could be 90 per cent certain that the true score lay between 102 and 118.
He said:"Generally, longer assessments are more reliable, but not for young children. They tire quickly and longer tests can be less reliable for this reason."
Dr Tymms' critique of testing appears in the latest issue of the British Journal of Curriculum and Assessment, vol 8, no 3