Harnad, S. (1999) Why I believe intelligence can be measured by IQ tests, and how universities could make use of them. Times Higher Education Supplement. Friday February 12 1999.

Why I believe intelligence can be measured by IQ tests (and how universities could make use of them)

Stevan Harnad
Department of Electronics and Computer Science
University of Southampton
Highfield, Southampton

This is not even a question of belief. It's an empirical question: something that you can test and check. Psychometric tests are deliberately constructed to correlate with certain things that people can do in the real world, such as perceive and produce pitch and rhythm in music, navigate in space, solve algebra problems, understand complicated texts. Among other things, tests can be designed to predict academic performance. The predictive power of such tests can be validated by throwing out test-items that don't correlate and constructing items that do until the test has a high enough correlation with what it is trying to measure.

Those who question whether such tests are effective usually haven't given it a great deal of thought; they doubt IQ tests "in principle," usually for (noble) ideological reasons -- they prefer a world (as do I) in which aptitudes are all identical and only individual effort determines success in every field of endeavour -- or they doubt tests on the basis of personal experience (invariably negative personal experience) of how unfair and unrepresentative their own test results have been. But those who know (not believe, know) that tests do work base this on their predictive power at the population level, not from indivividual experience. positive or negative. They have analysed the scores of large numbers of people and confirmed the strong correlation between the scores on those tests and how people actually perform in the world. Tests will sometimes under- (or over-) predict: statistics are like that; they are not infallible in every individual case. But if they are robust, they should do a good job overall for the population as a whole. The rest is down to what you want to use the results for. If you want to select candidates who are likeliest to pass a demanding course, you will want to be guided predictive tests.

There are legitimate differences in judgment over the degree to which aptitude tests measure a fixed inborn potential or the outcome of experience and effort. Scores are probably a reflection of both; what the relative proportions are is controversial, but by the time people reach the age of, say, 16-17, their aptitudes, whether genetic or experiential, have largely "solidified": they no longer change very much (some say they never did). So a second controversial question, and a much more practical one, is: Once it's solidified like that, at a population level, what is the best way to channel it?

I favour more people getting into university: The larger the proportion of the population gaining a higher education, the more everyone gains, individual and population. But to assume that everyone's aptitude is identical is a mistake. If, in an effort to increase admissions, universities keep their level of instruction fixed but lower their entry requirements, there are only two possible outcomes; either a larger proportion of the student population is going to fail out (which defeats the purpose of admitting more) or content requirements must be lowered to match the lower entrance requirements. But then the result is more higher education degrees, but not necessarily more higher education overall (which again defeats the purpose of the expansion).

There is another way, not an unfamiliar one, as it was applied in this country (too early, at age 11, and too rigidly) in the form of "streaming." Different curriculum levels (between or within universities) would be designed to match differences in incoming aptitude. Our tests are probably too crude for anything finer-grained than a 3-4 tier system, with the middle 2nd and 3rd the biggest, in keeping with the "bell" shape of all normal population distributions (e.g., for height, most of us are near the average, high normal or low normal, fewer are nearer the extremes).

The only alternative to having the universities conform to the incoming curve ito deform the universities so they all regress on the population average. I believe everyone loses in that case.