There is a current of opinion both among the general public and among psychologists (who can’t so easily be forgiven) that Eysenck is not a serious psychologist. It seems to be felt that his psychological findings are coloured by right-wing views which necessarily disconfirm them, that the sins of Burt (of whom Eysenck was a student) have somehow been visited upon the son, and that his prodigious output demonstrates mere facility rather than being a just return for Herculean labours.
I must make my position clear. In my view, Eysenck’s factor-analytic work in the field of personality, on which much of the work reported in The Causes and Effects of Smoking is based, is among the best of its kind in the world: it is scientific research of irreproachable quality, technically brilliant and easily replicable in other laboratories including my own. Thus any criticisms of the book which I shall make do not stem from a general feeling that Eysenck’s work must be poor or that his politics are unpleasant: they arise from an objective examination of the findings, regardless of their provenance. Ad hominem arguments are out of place in science. It will also be clear that, despite my recognition of the quality of Eysenck’s work, I am by no means in agreement with his conclusion.
The book falls into two parts, of which the first is concerned with a study of the largely medical evidence linking various diseases – the best-known are lung cancer and coronary heart disease – with smoking, mainly cigarette-smoking. Eysenck demonstrates that much of this evidence is of poor quality and not such that any strong inferences can be made from it. However, as is by now well-known, Eysenck’s survey is somewhat vitiated by the fact that he did not examine the review paper of this evidence by Peto and Doll, published in 1976, which went some way towards answering his points. It would have been valuable to see Eysenck’s analysis of this paper. Nevertheless, the point remains that the power of much of the evidence linking smoking to various diseases has been exaggerated.
Eysenck’s argument in this first part of the book is not that cigarette-smoking causes or does not cause lung cancer, coronary heart disease and other smoking-related disorders.What he is basically claiming is that the kind of evidence that has been produced in favour of smoking causing such diseases simply does not establish the case it is supposed to have established. The reason, of course, is that in terms of scientific evidence, it is far harder to demonstrate (clearly) a causal link than to fail to reject a null hypothesis that there is no link between two events. Eysenck has long proposed, as part of his general personality theory, that the correlations between lung cancer and smoking, to take a particular case, and Eysenck stresses that the link between smoking and each of the putatively related diseases needs to be treated separately, are due to a further factor: the link between smoking and constitution both psychological and physical. That this constitution is linked to lung cancer is the explanation which, Eysenck argues, best fits the facts. In practice, of course, it does mean that smoking appears to be (indeed is) related to lung cancer: but theoretically the implications of this new proposition are very different. Thus the constitution that predisposes us to smoke, predisposes us to cancer. Thereare constitutional and genetic factors which predispose people to disease and to smoking, which lead people to adopt life-styles leading to less resistance to disease. Smoking is only one indicator of such a life-style: others are drinking, philandering and other forms of sensation-seeking behaviour. Much of the second part of the book is devoted to reports of research bearing on these points: the genetics of smoking, the genetics of personality and the link between smoking and personality.
This second section, on the genetics of smoking, has been written by Eysenck and Eaves, who is an acknowledged expert in the statistical methods which are employed here and which are necessary for the study of the genetic determination of human variables. These are known as biometric genetics and have been highly developed by Mather and colleagues, of whom Eaves was one, at the University of Birmingham. Biometric genetics is a complex and specialised subject. The majority of psychologists are, if not unaware of it, quite unable to use it – my students, even when mathematically inclined, find it hard to grasp. It seems to be applied appropriately here, and there appear to be no major errors.
The conclusions are striking. A genetic model for smoking behaviour based upon the biometric analysis of twins can distinguish those who have never smoked from those who have persisted; between these categories are those people who have managed to relinquish the habit, so easy to relinquish, indeed, that Mark Twain did it frequently. That this predisposition to smoke is inheritable, rather than environmentally determined, is supported by the fact that the inclusion of a family-environmental effect in the model produces no changes in fit. One somewhat complicating conclusion has to be drawn from the results: namely, that the genetic dimension which differentiates how many cigarettes are smoked is quite different from the dimension differentiating smokers and non-smokers.
Part of this genetic variation in smoking habits is mediated through inherited differences which affect personality. Results are reported from an investigation into the relationship between smoking and the three variables regarded by Eysenck as the most important dimensions of individual differences, Psychoticism, Extraversion and Neuroticism. In fact, psychoticism, tough-mindedness, lack of empathy, a delight in bizarre and strange sensations, was related to smoking, although it explained only a little of the variation in smoking. It should be noted that this finding is not quite in line with previous work, since usually extraversion is related to smoking. P., E. and N. are themselves largely genetically determined, so that these results help to build up the thesis that there is a predispositional factor for both smoking and its related diseases.
Eysenck is duly cautious about the conclusions to be drawn from all this work. To repeat, he argues that a simple causal link between smoking and disease has not been demonstrated. A constitutional hypothesis of a common predisposition fits the data better, especially since genetic factors play a part in smoking and in the personality variables related to smoking. However, the work he cites in his attempt to demonstrate the influence of personality in smoking, and hence the nature of the third factor, the predisposition to smoke and to become ill, is not entirely convincing.
The mediating variables in Eysenck’s account are personality variables. In any precise scientific hypothesis, the variables must be quantifiable and, quite rightly, Eysenck began with the three most clearly defined factors, Extraversion, Neuroticism and Psychoticism. However, his studies indicate that only one of these, P., relates to the smoking habit, and this only slightly. This supports the view which many less rigorously scientific psychologists have held: that Eysenck’s three factors represent a gross oversimplification of the complexity of personality.
For example, one well-known personality theory, the psychoanalytic account (a red rag to Eysenck), postulates that smoking is related to orality (excessive concern with stimulation of the mouth), resulting from early child-rearing practices, weaning, feeding and the like. Vague though this is, I have myself developed psychometric tests of oral characteristics (quite as reliable as those used by Eysenck and constructed in similar fashion), and shown these to be related to heavy cigarette-smoking. This confirms the claim that personality is indeed implicated in smoking, but aspects of personality that Eysenck would not admit to be worthy of study. This example indicates Eysenck’s rather narrow approach to the personality of smoking.
A further point needs to be raised about the logic of Eysenck’s arguments. Even it he is correct and there is a predisposition both to smoke and to get cancer, it is still not necessarily true to argue that smoking does not cause cancer in some cases. There is little doubt that smoke contains carcinogenic substances. The work with beagles which Eysenck does not discuss shows this. Hence it is likely that to smoke exacerbates the predisposition to cancer. Since the predisposition to smoke is not so compelling that individuals with such a constitution must smoke and that those without such a predisposition must not, it follows that smoking may increase the likelihood of cancer in those with the suitable predisposition. On Eysenck’s own argument, this must be the case, since otherwise all smokers would get lung cancer. In other words, the predisposition to smoke and the predisposition to get lung cancer are not perfectly correlated. Thus in certain cases (those with the cancerous constitution) cigarette-smoking may cause cancer, although in the majority of individuals it does not do so.
In summary, this book has cast legitimate doubt on the certainty with which, in some quarters of the medical profession, it is claimed that smoking causes a variety of diseases. It certainly demonstrates, albeit not strongly, that genetic and constitutional factors are implicated in smoking and that the hypothesis that there is a factor predisposing us to smoke and to become ill is viable, although the nature of this factor is still not known. It certainly does not suggest that cigarette-smoking is safe or healthy. Indeed, the conclusion to be drawn from the work is that until the predisposition to these smoking-related diseases can be identified, it is wise not to smoke. However, it also suggests, by implication, that could this predisposition be identified, thus isolating those at risk from smoking, then smoking could be enjoyed by those not at risk without the feeling that they were committing suicide.
Send Letters To:
London Review of Books,
28 Little Russell Street
London, WC1A 2HN
Please include name, address, and a telephone number.