I can clearly remember, as an educational psychologist, giving an intelligence test to a surly 13-year-old girl who was faring badly at school. She got an IQ of 70 and the result confirmed the school's view of her as pretty dim. Several years later I heard that she had picked up three A levels and gone to university. Anecdotes like this are not as rare as one would wish. They have in recent years persuaded most psychologists and educators to place less faith in formal assessments of cognitive ability. But changing practice is not predicated solely on anecdote: over the past 20 or 30 years a substantial evidence base has grown which has warned against placing too much reliance on tests.

But so enticing is the notion that one ought to be able to assess, diagnose and prescribe that it bubbles away irrepressibly. An industry is devoted to devising better tests - tests that more accurately rank children by attainment, predict subsequent achievement, or diagnose weaknesses supposedly necessary for skilled performance in a related area. The "science" of psychometrics is about enhancing the accuracy of tests in these areas. In fact, one of the editors of this compilation prefaces one of his own chapters with a science-invoking quotation from an eminent psychologist: "Prediction is the most obvious hallmark of a successful science". Well, it depends what you mean by "prediction" - and psychometric prediction is not hugely impressive. Ptolemy constructed an intricate calculus for explaining and predicting the anomalous movement of the planets in what he assumed to be earth-centred universe. It was accurate enough, yet few would now call his calculus "science".

Indeed, the increasing refinement of Ptolemy's system - with all its epicycles and equants - comes to mind with the discussion of different kinds of dyslexia in the first chapter of this book, as developmental dyslexia is divided into developmental surface dyslexia and developmental phonological dyslexia. Elsewhere the refinement continues and we hear also about dyseidectic, dysphoneidectic and dysphonetic dyslexias - and even, believe it or not, mixed dysphonetic-dyseidectic dyslexia-alexia.

Only in a few chapters of this varied compilation does one get a feeling for the theoretical frailty of the assessment arena. Take, for instance, the question of the capacity of measured subskills to predict later performance in reading - or indeed of the significance of supposed weakness in these sub-skills for explaining existing weakness in reading. These relationships have been at the centre of much controversy - the problem being that reading (and the accompaniments of reading in verbal, literate families) tends to train up children in the supposed subskills (things like auditory memory), so any interesting-looking association between sub-skill and reading ability may have its roots in the reading rather than the supposedly important subskill. Direction of causality is the issue. Causality is not usually difficult for us to figure out in real life: not many people believe that eggs are the right size for egg cups because of good planning by hens. But direction of causality seems to engender interminable logical tangles for psychometricians. In this volume, the important phenomenon of reciprocal causation is discussed seriously only in John P. Rack's coherent and well-argued chapter. There are a few other interesting chapters.

̽»¨ÊÓƵ

Much of this compilation rests on the claim of psychometrics to be taken seriously as science. One might therefore have expected some humility about the fragility of the subject matter being debated. But there are too many clod-hopping forays into areas where it would be better to tiptoe. And having a four-page reference list behind your chapter makes neither your approach nor your conclusions scientific.

Gary Thomas is professor of education, University of the West of England, Bristol.

̽»¨ÊÓƵ