Ever since the publication of The Bell Curve by Herrnstein and Murray, a much-cited and highly controversial book on the subject of intelligence, the question of how race and intelligence interact has been simmering away. It looked for a while as though it had largely blown over, becoming fodder only for the odd dinner-party conversation and white supremacist rally. However, a couple of weeks ago James Watson, one of the discoverers of DNA's helical structure, created a storm of publicity with some extremely poorly-judged comments:
"All our social policies are based on the fact that their [Africans'] intelligence is the same as ours – whereas all the testing says not really.
As if that wasn't enough to send the world's media into a frenzy, he followed it up by claiming that although it would be nice to think that people of different races are equally intelligent, "people who have to deal with black employees find this not true." Now, before we go any further it's important to note that Watson apologised just a few days later:
"I cannot understand how I could have said what I am quoted as having said. I can certainly understand why people, reading those words, have reacted in the ways they have. [...] To all those who have drawn the inference from my words that Africa, as a continent, is somehow genetically inferior, I can only apologize unreservedly. That is not what I meant. More importantly from my point of view, there is no scientific basis for such a belief."
A good retraction, certainly, and a necessary one, but the damage was already done. No-one remembers the mitigating language of this retraction, and the media feeding frenzy had already begun, leading to people suggesting that perhaps his first set of statements had actually been true. So what, in this horribly difficult topic, is true? (I knew that psychology degree would be useful for something.)
Most of the controversy around The Bell Curve centred on just one fact that its authors highlighted, and I'll reproduce it here. Note the language very carefully - this fact doesn't necessarily mean anything more than it directly says.
In general, Asians tend to score more highly on IQ tests than Caucasians, who in turn tend to score more highly on IQ tests than Africans.
However much anyone dislikes it, that's a pretty sturdy, reproducible scientific fact. However, the problems arise when people assume that its ramifications are much larger than they actually are. Let's go through the conclusions that we can draw from this one fact - and those that we can't.
IQ might not mean anything in the first place.
Critics of intelligence testing have maintained for years that IQ is nothing more than a measure of how well people do on intelligence tests. It's certainly the case that general intelligence, or g, is a controversial concept. It is produced by running a procedure called factor analysis on the scores obtained by individuals on various different types of test; although this procedure does tend to show one main factor driving performance in all of these disparate areas, my experience of factor analysis is that it is almost infinitely malleable and subject to interpretation. If you know what you're doing with the data and can choose suitable settings for the analysis, you can make it say practically anything.
That said, it's not ridiculous to suggest that people who are good at one cognitive task may well be good at others, and it's not impossible to put a single number on this ability. However, because this number will necessarily be a summary of various different results, all of which have their own margins for error, I would be extremely wary of any study which claims to show that tiny differences in IQ are in any way important. An employer who chose to hire someone because his IQ was one point higher than that of another candidate, regardless of any other information, would rightly be regarded as an idiot; why, then, do people suggest that IQ can, by itself, let you draw sweeping conclusions about entire populations?
The differences that we're talking about are minuscule.
One of the difficulties that statisticians find when trying to summarise information about populations is that people vary immensely; the "bell curve" referred to by Herrnstein and Murray is the Normal distribution, a probability distribution that describes natural variation in a population. Now, when these distributions for the IQ of different races are plotted on the same graph, they overlap hugely. Yes, the peaks are at different positions; however, the bulk of the area under each curve is shared with at least one other curve. This means that it's completely impossible to determine someone's race from their IQ, and more importantly, it means that the variation in IQ within each race is considerably greater than the variation between the races. Your chances of accurately estimating someone's IQ from their race are consequently not great.
"But Phil," some of you will (probably) say, "you've admitted that there is a difference between the races! Surely that has practical application?" No, it doesn't, because no-one ever deals with entire populations at a time. We deal only with individuals, and as it's pointless to try to estimate an individual's intelligence from the colour of his skin, the entire question seems moot.
"Race" doesn't necessarily mean very much.
How do you determine someone's race? By the race of their parents. Already we're veering dangerously close to circular reasoning, as the race of the parents is determined by that of their parents, and so on and so on. In the end, either we're all Neanderthals or similar, or we have to determine a cut-off point - and the fact that this is never determined in this type of experiment means that there is great difficulty in being certain that the race definitions of participants mean anything at all. Researchers tend to get around the problem by getting participants to report the race that they consider themselves to be, which means that there's even more fuzziness in the results.
It is incredibly difficult to create a culturally unbiased intelligence test.
Assuming for a moment that researchers did manage to come up with fully acceptable definitions of "intelligence" and "race", how would you go about testing that intelligence? If you use different tests for different groups of people, it's very difficult to draw any conclusions about the comparison of the results; if you use the same test, it will mean nothing to people who aren't familiar with the subject of the questions. Even if researchers use non-verbal tests, like Raven's Progressive Matrices (incidentally the only intelligence test that would also make an awesome name for a band), there is no guarantee that the test paradigm is going to translate across cultures.
To give you an idea of what I mean, imagine that you're a member of a small Ethiopian tribe and have never had to take a school test in your life. If someone shows you a picture with a number of patterns and asks you to select the "correct" pattern to fill in the gap, is it really going to immediately occur to you to complete it in line with logical syllogisms? Because the test can only record "success" or "failure" for each attempt, a "failure" could mean either "was incapable of making the required logical inference" or "was unfamiliar with the concept of abstract logical inferences", and this confusion enormously complicates the conclusions.
The correlation between IQ and success in later life is not as conclusive as it appears.
There is indeed a correlation between IQ and job success, earnings and so forth. However, the first thing to remember is that correlation does not necessarily imply causation. There may be other factors - good nutrition or ability to cope with stress, say - which moderate both IQ and other successes. Secondly, the correlation is not perfect; having a high IQ is by no means a guarantee of later success. And thirdly, the definition used for "success" is itself culturally biased towards the type of society where prosperity is the ultimate goal. Use "satisfaction" or "happiness" and things might be very different - take Ghana, for example, which is apparently in the top 40 countries for happiness but barely makes the top 150 for GDP per capita.
Possibly the most depressing thing about the storm over Watson's comments is that many people wanted to close down the debate entirely, suggesting that this entire topic is somehow off limits. Obviously I don't think that's correct; censoring debate not only means that important topics never get thought through, it also makes them the sole preserve of those who would use them for their own ends.
Although I tend to shy away from qualifying anything I say on this blog with "this is just my opinion" - obviously it's my opinion, otherwise I wouldn't be writing it - I'm going to make an exception in this case because this is a hugely complex topic. Hopefully I've contributed to it in some way, but clearly I haven't said everything that there is to say, and nearly everything that I've written could be challenged by other research. Please do make the effort to look into it further; important topics require careful consideration, and the best outcome we can hope for is that, rather than hiding from the subject, more people will do just that.
No comments:
Post a Comment