National test results apparently showing a transformation in 11-year-olds' English achievements from 1996 to 2000 came about largely because the standard of the reading test fell, official research has revealed.
A three-year study, finally released this week by the test regulator after The TES disclosed it had remained unpublished for nearly two years, reveals that gains in reading were to some extent "illusory".
As revealed in The TES in October, the research, commissioned by the Qualifications and Curriculum Authority, found that the pass mark for key stage 2 English was set five marks too low in both 1999 and 2000.
If it had been set as the study indicates it should have been, the proportion of pupils performing to expected levels would have fallen by 10 percentage points.
And Labour would not have been able to claim such startling improvements in primary English at the last election. Official figures showed the proportion of pupils achieving expected levels rising from 57 per cent in 1996 to 75 per cent in 2000. The correct 2000 figure should have been 65 per cent.
The study also found that published results for KS1 English and maths, KS2 maths and KS3 English and science were reliable, but that KS3 maths pass marks may have been too low.
The conclusions come in research led by Alf Massey of the University of Cambridge Local Examinations Syndicate, released this week by the QCA. It received the findings in January 2002.
Asked why the report had taken so long to appear, the QCA said that "the findings were being acted on internally before the report was circulated more widely".
A QCA spokesman said: "In future, we will aim to share research publications more widely and more quickly."
The investigation involved pupils in Northern Ireland sitting KS1, 2 and 3 tests in 1996 and then compared the results to those of similar ability groupings of pupils taking 1999, 2000 and 2001 tests. In KS2 English, a first experiment saw 420 pupils taking the 1996 tests scoring four marks higher, on average, than 424 taking the 1999 test.
A second experiment saw a further 495 pupils taking the 1996 test averaging three marks more than those sitting the 2000 version.
This suggests that the 1996 tests were easier. But the QCA overcompensated in setting much lower-level thresholds or pass marks for the tests, the research says. It lowered the threshold for level 4 by nine marks between 1996 and 1999 and by eight marks between 1996 and 2000.
As a result, in the first experiment, 75 per cent of the Northern Ireland pupils taking the 1999 tests achieved the level 4 threshold, compared to only 65 per cent among those taking the 1996 test.
In the second experiment, 67 per cent of pupils taking the 2000 test gained level 4 or above, compared to 65 per cent taking the 1996 version.
Separate analyses of test results in six English local authorities, and interviews with a group of 10 teachers, supported the findings.
Ministers have highlighted improving results in reading, apparently revealed by national test figures, as a reason for focusing on improving pupils' writing skills.
But the study said that most of the "disparity of standards" between the 1996 test and later version came because the reading pass marks had been set too low.
The 245-page research report raises questions about whether making tests more user-friendly or "accessible" for pupils has affected levels of difficulty.
But it found no evidence that ministers had deliberately made the tests easier.