George Walford: How Exact is Eysenck?

People who try to understand why society behaves as it does are accustomed to being asked: “Why do you not pay more attention to Eysenck’s work?” The idea behind the question is usually that we ought to pay attention to Dr. Eysenck, more than to other psychologists who concern themselves with political and social behaviour, because his work is exact and quantitative. It is thought to approach the solid reliability of what are sometimes called “the exact sciences,” and therefore to be entitled to share in their prestige. This present article considers this belief. It draws attention to parts of the chapter “Politics and Personality” in Dr. Eysenck’s book Sense and Nonsense in Psychology (the Pelican reprint of 1978). It is an attempt to determine whether, at least in this one chapter, Dr. Eysenck’s work is as valuable a contribution to our knowledge as it is sometimes thought to be.

The first point to be made is that Dr. Eysenck does not always write with strict quantitative accuracy. On the first page we find these phrases: “widely held,” “many people,” “frequently,” “some,” “usually,” “at certain times,” quite a large number.” None of these references have been exactly quantified; Dr. Eysenck is (very sensibly) content to leave then numerically imprecise.

Dr. Eysenck himself is concerned to establish that his work is scientific. He connects “factual empirical research” with “the Social Sciences” (his capitals) and contrasts this approach with that of “politicians and others who have an axe to grind.” The politicians are busy grinding their axes and enjoying “their intoxicating draught of power” when:

the social scientist appears and says: ‘Here we have two opposite sets of hypotheses; let us not waste time in talking about which is nearer the truth, but let us go and carry out an experiment to see which hypothesis is, in fact, nearer the truth’ (Emphasis in original).

Few of us think highly of politicians, and the prospect of a firm, clear, precise, scientific decision which will enable us to put them in their places is attractive. It may even give us a feeling of power. The question is: Does Dr. Eysenck supply this requirement?

Dr. Eysenck distinguishes four classes in society, and four status-groups which roughly correspond to them. He says:

it seems almost inevitable that political groups should arise to represent these respective interests, and, indeed, as is well known, groups of parties have arisen in all the democratic countries which represent these differences of interest.

“It seems almost inevitable.” “As is well known…” I ask any reader, who may have been thinking that Dr. Eysenck’s work is to be valued for its scientific rigour, to note that of practically every belief which science has shown to be fallacious, it could have been said that “it seemed almost inevitable” and “it was well known.” Dr. Eysenck is not, in this passage, using scientific modes of speech or basing his argument upon scientifically established grounds. The question at issue is whether the parties, or groups of parties, found in democratic countries, do represent the difference of interest between classes or status-groups. He says it is “well known” that they do. Similarly, those who condemned Galileo held it to be “well known” that the sun moves around the earth.

Having distinguished four classes and four status-groups Dr. Eysenck says:

there can be little doubt that in a representative sample of the population there is a close relationship between voting behaviour and social status. There is a similar close relationship between social class and voting behaviour.

Newton, Boyle and Einstein did not say “there can be little doubt” about gravitation or the expansion of gases, or relativity, but let that pass. Dr. Eysenck says there is a close relationship between social class and social status on the one hand and voting behaviour on the other, and he distinguishes four classes and four status-groups which roughly correspond to them. But a few pages farther on, when he comes to specify the political groups for which votes are cast, there are five of them, namely: Fascists, Conservatives, Liberals, Socialists, Communists. Four classes (or status-groups) and five political groups which, Dr. Eysenck asserts, “arise to represent” their interests. We are offered no explanation of the discrepancy.

And why does Dr. Eysenck ignore the Anarchists? This would give six political groups which “arise to represent” the interests of four classes or status-groups. Is it that his sampling procedure did not reveal the existence of a significant number of Anarchists? If so, then his sampling procedure is defective. Before dismissing Anarchists as insignificant because, of their small numbers it would be well to check with Special branch.

So far we have been concerned with Dr. Eysenck’s verbal arguments, but his work is valued for its quantitative precision. Let us now meet him at his strong point, let us look at some of his figures.

So far Dr. Eysenck holds that the parties of the right further the interests of the high-status groups and therefore attract their support. The parties of the left, he holds, further the interests of the low-status groups and therefore attract the support of these groups. He assures us that this bifurcation “represents an inevitable consequence of the lay of learning and can be directly deduced from it.” He adds: “If our generalisation is true, and if the law is as stated, then surely all working-class people should vote Labour, and all middle-class people Conservative.”

In order to confirm this expectation – which is, remember, an “inevitable consequence of the law of learning” – we turn to the table on page 276. This shows two low-status groups. One is the Average-Minus, and of these 47% support Labour – that is to say, 55%, more than half, do not support the party which Dr. Eysenck’s law says they will support. The other low-status group in this table is the Very Poor. Of these, 52% support Labour – that is, 43%, almost half, do not support the party which Dr. Eysenck’s law says they will support.

Perhaps I should emphasise that these are not my figures but Dr. Eysenck’s. And, they do not confirm his “law.” They are not significantly different from the 50-50 to be expected from a random choice between two alternatives.

Dr. Eysenck’s figures do not confirm that his “law of learning” governs political behaviour, but he goes on to speak as if they had done so. On a later page he says: “We can predict that working-class groups should be predominantly radical in their sympathies.” Yes, we can, provided we are willing to accept a correlation between prediction and event, hardly better than that given by chance.

Dr. Eysenck does say “the relationship is not perfect.” One has to agree, even if finding the comment unduly mild. But one also has to ask whether such results support the belief that Dr. Eysenck’s work is entitled to respect for its quantitative exactitude, whether such results entitle the work to be described as “Social Science.” To use this phrase is to claim a share in the respect accorded to the physical sciences, but it was not by putting forward theories with this lack of predictive power, or “laws” that receive so little confirmation from measurement, that those sciences won their prestige.

from Ideological Commentary 8, November 1980.