By Gary Eason
Education editor, BBC News website
Changes have been made to the marking process
The proportion of higher achieving 11-year-olds in national curriculum tests in English fell five percentage points this year to 29%.
Yet the proportion achieving this Level 5 in the reading component stayed the same, at 48%, and the proportion doing so in writing went up, from 19% to 20%.
So how can it be that the overall result fell so sharply?
Two factors are in play: the way the marks are combined, and the ending of the practice known as "borderlining".
Firstly: how is the overall English result arrived at?
At any level in English pupils can do badly in one component - and it is usually the writing - but well enough in the other to bump up their overall score.
For example: in writing, the range of marks regarded as indicating Level 4 attainment this year is from 25 to 36 (out of a maximum of 50). The Level 5 range is from 37 to 50.
In reading, Level 4 goes from 18 to 31, and level 5 from 32 to 50.
Overall for English Level 4 is 43 to 68 and Level 5 is 69 to 100.
Let us say a child reads well and obtains 45 marks - a good Level 5.
Their writing test accrues only 24, however. That is a middling Level 4 performance.
Yet add the two together to make 69 and they are awarded Level 5 - exactly the same, incidentally, as someone who scores 50 out of 50 in both tests.
2008 KEY STAGE 2 RESULTS
English: 81% (80% last year)
Maths: 78% (77%)
Science: 88% (88%)
English: 29% (34%)
Maths: 31% (32%)
Science: 44% (47%)
This appears to have happened on a large scale in 2007, with a strong reading performance bumping up the overall proportion of Level 5s.
This year did not go so well and the proportion fell - perhaps by a little over two percentage points.
This is where the borderlining effect took over.
Previously, test scripts that were up to three marks below what was required for the next level were double checked to see if any more marks should be awarded.
It was a one-way street: efforts were made to see that pupils got extra marks where merited - but there was no double checking of those who were a few marks above the level to see if they had been awarded any they did not deserve.
The Qualifications and Curriculum Authority (QCA) had recommended ending the practice four years ago.
This year, the new test contractor, ETS, introduced more checks of markers' work - five instead of two - during the marking process.
The National Assessment Agency (part of the QCA) said: "The improvements in quality checking in 2008 mean that the process of checking test scripts that fall close to level thresholds, known as borderlining, is no longer necessary."
So borderlining, which had been in use since 1995, was dropped - and the assessment agency is making no effort to defend past practice.
Government statisticians have calculated the impact of removing borderlining and applied it to recent years' results too.
They believe its removal will have reduced the proportion of children getting Level 5 in English by fully 2.9 percentage points this year.
It would have been 2.6 points last year, 2.5 in 2005 and 2.4 in 2004 - which is when the Qualifications and Curriculum Authority recommended ending the practice.
(The effect in maths is reckoned to be no more than 0.4 percentage points this year and in science, 1.5 points.)
In other words if borderlining had remained this year, the proportion attaining Level 5 in English might have fallen only from 34% to 32% - the same as it had been in 2006 and a far less dramatic drop.