Destined to fail

Statistics might show a correlation between test results in KS1 and KS2, but to assume every child’s fate is sealed at the age of seven would be a gross misjudgement, says Richard Cowan...

Educational statistics and scientific research can make arresting headlines in newspapers and online digests. But news stories rarely provide enough information. Journalists use their imagination in writing stories and readers use their imagination to understand them. Unfortunately beliefs in determinism persist in our collective imagination.

Consider this opening to a story that appeared in the Observer in 2005.

Twins hold key to unravelling maths gene

Parents have long battled to persuade their children to master new spellings and learn their tables, but they may be wasting their time. A new study suggests that both maths and reading ability lies largely in the genes.

Even allowing for journalistic licence in singling out spelling and learning tables, which are very small parts of literacy and numeracy, it is just wrong on so many levels. The Twins Early Development Study (TEDS) researchers, whose work this describes, would not agree to this portrayal. First, there is no maths gene: there are many genes that contribute to individual differences in mathematical potential. Second, there is no single maths ability or reading ability. There are lots of different literacy and numeracy abilities that children develop as they progress through primary school. Third, even if genetic variation explained a lot of individual variation, this does not rule out the importance of the environment. Differences in height owe much more to genes than differences in maths, but children do not grow without food.

TEDS includes thousands of twins: some are identical and some are not. Identical twins are clones as they are genetically the same, but they defy popular myths and fears (remember Dolly the sheep). In reality, identical twins do not always do the same job, marry at the same age and call their dogs by the same name. In reality, if one identical twin scores very poorly on a maths test, then it is likely, but by no means certain, that her co-twin will do similarly. Your genes do not determine your destiny.

How about this article, which appeared in the Telegraph in October 2013?

Children’s future earnings predicted from age nine

Future earnings of children can already be predicted by the time they are nine simply by looking at their English and Maths scores, according to a new study.

Predictions that have a basis in social science are particularly alluring. However, if you think the study highlighted in the Telegraph shows that doing well at maths age nine has a causal relation with later earnings, you are using your imagination. All it really shows is a non-trivial correlation between primary educational achievement and later earnings. One would have to read the study to see what else they took into account. It is certainly not an experiment and so any talk about cause is just speculation.

As Steven J Gould pointed out in The Mismeasure of Man, “The invalid assumption that correlation implies cause is probably among the two or three most serious and common errors of human reasoning”.

Brain scans support findings that IQ can rise or fall significantly during adolescence

IQ, the standard measure of intelligence, can increase or fall significantly during our teenage years, according to new research, and these changes are associated with changes to the structure of our brains. The findings may have implications for testing and streaming of children during their school years.

This accurate description, reported in Science Daily in 2011, was issued by the Wellcome Trust, who funded the research. A group of adolescents were tested twice with an interval of three years between assessments. Despite a high correlation between the assessments, over a third showed substantial changes between the first and second test.

This is due to the difference between a correlation, which is a measure of association at a group level, and individual variation. Even when there are very high correlations between assessments, individuals can show very different paths with some making considerable improvements and others declining.

The correlation between the IQ assessments was .79. This is much higher than the correlations between KS1 and KS2 assessments that Kathryn Duckworth and Ingrid Schoon at the IOE found: .66 (Maths) and .67 (English). It is also higher than the correlations between PIPS baseline assessments and KS2. Assuming an individual child’s result in the KS1 assessment will be mirrored in his KS2 test would, therefore, be a mistake.

Children in the same school differ much more than schools differ on average. Accordingly, school statistics published by the Department for Education provide very little basis for predicting whether a child will do better in one school than another. Do your school’s parents know that?

About the author

Richard Cowan is Professor of Psychology of Education at the Institute of Education.