Andrew Sutherland makes interesting claims about the unreliability of 5-14 assessments (February 27).
One means of assessing test reliability is to measure (on a 0-1 scale) how well the results on one test correlate with the results on another test.
In 2007, I and a group of my then colleagues in the University of Strathclyde collected 5-14 national assessment data on 517 young people. We found (on the 0-1 scale) that the reading scores had a correlation of 0.9 with the writing scores; and that the maths scores had correlations of 0.8 both with the reading scores and with the writing scores.
On that evidence, 5-14 scores do have a moderate to high degree of reliability. Pairs of tests, where one or other or both is or are unreliable, simply do not exhibit high inter-test correlations.
The question of validity (that is, exactly what it is that these data are measuring) is another matter.
Ian Smith, vice-principal (learning and teaching), UHI Millennium Institute, Ness Walk, Inverness.