Monthly Archives: August 2007

IQ and intelligence

Today on Radio 4’s Today programme was a statement which brilliantly shows the problem of ‘correlation or causation’ in scientific statements, and social science especially. They were talking about IQ.

 

Stephen Murdoch (http://www.stephenmurdoch.com/) has a book out saying that IQ tests are problematic. Of course… IQ tests measure something, and we can call this intelligence, but we don’t know how much of this is cultural bias or training for particular skills and notations so we can’t say it’s the same as the ‘power’ of a person’s mind. There’s no way to design a test independent of our culture that can be used to test the IQ test.

In opposition, Colin Cooper (http://www.psych.qub.ac.uk/Staff/Profiles/cooper/index.aspx) of Queens, Belfast (and the BBC’s Test the Nation!) said:

‘There’s a correlation… which is huge… between a child’s intelligence aged about 11 [i.e. performance in IQ tests aged 11] and their subsequent performance on GCSE exams’

This is, of course, true and not just because the IQ measures the kind of skills one needs to pass GCSEs. To some extent, the causal chain here can make this statement a tautology.

The reason being that in the UK we use IQ tests, or other tests (it doesn’t really matter which) to decide on the kind of education one gets. The 11+ is used to separate kids off to grammar school (like I was), or test results are used when ‘setting’ classes. Even if this doesn’t happen, teachers have the previous results and can use these to decide at what level to teach, where help is needed, and so on. Therefore, the education one receives at high school is dependent on previous results. So if two children of equal intelligence scored differently on a test at 11 (one has a bad night’s sleep), the type of education after can be different, so causing later results to be different. The cause of good GCSEs isn’t the intelligence, but the 11-year-old’s test result, so any correlation with the test is a mere tautology.

There’s only one way to test this… a randomised controlled trial. Get 1000 kids and IQ test them. Then give them identical educations… 500 in a school where the teachers don’t know the 11+ result and 500 in a school where they do. (The kids can’t know either). Can I get funding for this?

Leave a comment

Filed under Statistics and simplicity