Smoking; drawing by David Levine

When tobacco was first introduced into Europe and Asia, potentates saw harm in it and some enacted the death penalty for those of their subjects who persisted in smoking. But only a proportion of the culprits were in fact put to death for the widespread crime, and the potentates soon substituted taxes for excommunications and executions. After three hundred years the death penalty is still linked to smoking, though not now by due process of law, and taxes on tobacco are still levied by governments: neither the risk of a shortened life, nor expense, nor alternative pleasures have eliminated smoking, nor is it less profitable for the revenue. Why the habit should have spread so fast and so far, and proved so tenacious, is a matter for conjecture. Other problems which it offers have not been left to conjecture, but made the subject of very active research. These are the problems raised by the association of cigarette smoking with disease, or, more specifically, by the association of cigarette smoking with cancer of the lung. Two impressive reports, from the Royal College of Physicians in England and the Surgeon General’s Advisory Committee in the United States, documented and reviewed the present state of knowledge about this threatening aspect of public health. Both reports concluded that cigarette smoking is an important cause of lung cancer, far outweighing all other factors.

Although the consensus of informed opinion is powerfully to this effect, there are dissenters whose criticisms have demanded close scrutiny—people like Dr. Berkson of the Mayo Clinic who have been voicing their misgivings for a good many years, and others more recent, like Professor Eysenck, the author of this book.

Professor Eysenck is a psychologist well known for his many valuable contributions to the study of personality in its varied aspects, and for his lively polemical writing, as spirited and forthright in defense as in attack, and firmly based on the principle of tit for tat. His powers of lucid exposition have been evident in a large number of publications, describing the psychological experiments and arguments on which his views rest. In this book, characteristically readable, he lays an axe to the generally accepted conclusion about the role of cigarette smoking, and he links the association between smoking and cancer to certain features of his study of personality. His survey and his assertions are broadly on the following lines.

Epidemiological methods of investigation can suggest causal relation but cannot confirm or ensure it: Only experiment can do that. The association between smoking and lung cancer is not specific; other diseases also are related to cigarette smoking. The populations on whom the observations have been made were not selected in such a way that they can each be accepted as a representative sample, nor did a sufficiently large proportion of the people to be studied agree to take part in the inquiries. Measurement of the amount of smoking was unduly crude. The samples of population studied were self-selected. These and other objections, mainly derived from Berkson and another bio-statistician, Yerushalmy, are given in considerable detail, but they are followed by chapters expounding the more novel explanation which Professor Eysenck offers to account for the association between cigarette smoking and mortality from lung cancer, which he does not dispute.

His thesis is in keeping with the suggestion advanced by the late R. A. Fisher, that the practice of cigarette smoking and the development of cancer of the lung might be largely influenced by a common cause, the genotype or hereditary make-up of the individual. Fisher considered that the genotype exercises a considerable influence on smoking and on the particular habit of smoking adopted; groups classified according to their smoking habits would furthermore differ in their genetic predisposition to develop cancer. Eysenck carries this line of thought further.

He had found that, by appropriate inquiries (chiefly questionnaires, experimental tests, and study of twins) we can recognize two dimensions of personality—extraversion and neuroticism—which are determined to some extent by heredity. In large surveys carried out under his supervision by Mass Observation Ltd. on behalf of the Tobacco Manufacturers’ Standing Committee, he found that cigarette smoking was related to extraversion and pipe-smoking to introversion; neuroticism was not related to smoking. The relation between extraversion and cigarette smoking was stronger in those who smoked heavily than in the light smokers and the ex-smokers. Therefore, he argues, a constitutional factor may be presumed which is responsible for the extraverted personality and the predisposition to cancer: This makes it unnecessary, in his view, to attribute to cigarette smoking the main causal role in lung cancer. He accounts for the rapid growth in the incidence of the disease in recent years by incriminating air pollution. In brief, he holds it is justifiable, on the available evidence, to say that persons of an extraverted temperament are, because of their hereditary endowment, more likely than introverts to smoke cigarettes and more likely to develop cancer: hence the strong association.


Eysenck is too familiar with the uncertainties of psychological and genetic inquiries into causation to make dogmatic assertions for or against the hypotheses he examines (though the blurb on the jacket intemperately declares that he “rejects as unproved, exaggerated, or full of paradoxes, riddles, and contradictions the assertion that smoking causes lung cancer”). His chief concern is first to demonstrate to the reader the weakness of the accepted view, and then to indicate the superior cogency of his claim for a common inherited predisposition. He reiterates his destructive arguments lucidly and for the most part with studied moderation; occasionally, however, partisan fervor gets the better of him, and the advocate elbows the judge out of the way. In the end, however, he gives us his pacific credo about the smoking hypothesis and his own: “I believe (although I cannot at the moment prove) that both the hypotheses under discussion will be found to be true”—but he promptly adds that the direct deleterious effects of smoking are, he believes, “less important than suggested by the calculations in the British and the American Report.”

There are weaknesses in his grounds of objection to the accepted view. The chief of these lies in the differing rigor with which the evidence is scrutinized. A striking example of this is afforded by Eysenck’s reference to Poche in the controversy that followed publication of Eysenck’s book: It is worth detailing at a little length. Professor Poche and his two colleagues examined the smoking history of 1229 people in the North Rhineland and Westphalia who had histologically verified bronchial cancers. They concluded that there was no relation between cigarette smoking and the development of the cancer. Professor Eysenck calls this “an extremely careful study” and says that it “was methodologically far superior to the studies usually cited.” In fact the defects of the study, for which it has been arraigned by several authorities on medical statistics, are gross. The selection of cases studied was biased in that it depended on whether doctors had filled out answers to a questionnaire about the patients’ habits; the participation of the pathological institutes which supplied the material was uneven and arbitrary; and Poche himself characterizes the selection as casual (Zufallsauswahl). The specific questions about smoking were very crude: All the information about this that the doctors were asked to provide was a bare statement whether the patient was a smoker and, if so, how many cigarettes, cigars, or pipes he smoked per day and per week. No intimation was given as to what period the question referred to: It might have been the last few months before the patient came into hospital with advanced disease or it might have been his last few years. The avowed purpose of the investigators was to determine the complex of possible part-causes of carcinoma, but they did not include a control group of non-affected persons such as is needed for a retrospective study of causes; instead they compared patients who had a particular type of cancer known to be mainly responsible for the increase in prevalence of cancer of the lung, with others who had a different form of bronchial new growth. Poche and his colleagues accepted the diagnoses made in the participating institutes of pathology; but from internal evidence it is plain that the institutes differed very sharply in their criteria and diagnostic practice, so much so that it is impossible to accept the diagnoses in the absence of any organized exchange of histological preparations and without dual or multiple assessment by experienced pathologists in order to check “observer-error.” Professor Koller, of Mainz, who has hauled Poche and his colleagues drastically over the coals, points to several other departures from sound statistical or clinical procedure, and declares roundly that Poche’s failure to discover any significant evidence of a relation between cigarette smoking and cancer is of no scientific importance, because of the crass methodological defects. Nevertheless Professor Eysenck acclaims it as methodologically far superior to the studies usually cited, i.e., those by Bradford Hill and Doll, Hammond and Horn, Dorn, and others, who earn his strictures on one count or another.

Similarly he devotes considerable space to the findings and speculations of Beffinger, and declares that “the neglect of Beffinger’s hypothesis by the authorities is curious”: The explanation is that Beffinger’s hypothesis met the fate of Belshazzar—weighed in the balance and found wanting.

Eysenck reviews the evidence provided by studies of twins that the extravert type of personality is inherited, and that the tendency to smoking likewise has a hereditary determinant. He does not refer, however, to the mainly negative findings in what should be the third limb of his argument—twin studies of the inheritance of cancer, such as those by Clemmesen and his colleagues in Copenhagen. Nor does he give the statistical data on which he no doubt bases his rejection of the view that extraverts develop lung cancer more readily than introverts simply because they smoke more: “a very small proportion of the relationship may possibly be due to this factor, but it cannot by any stretch of the imagination be held responsible for the total effect.” Whose imagination? And what evidence?


Arguments about the etiology of particular diseases used to be centered on “the cause,” as if only one cause was operative. Putting aside philosophic or semantic issues, it is now commonly recognized that multiple causes are responsible, though in very different degrees: Delirium tremens, for example, and cirrhosis of the liver have alcoholism as a cause, indeed as the commonest cause, but alcoholism is not a necessary cause nor a sufficient cause of either condition. The Surgeon General’s Advisory Committee underlined the point:

No member of this Committee used the word “cause” in an absolute sense in the area of this study. Although various disciplines and fields of scientific knowledge were represented among the membership, all members shared a common conception of the multiple etiology of biological processes. No member was so naive as to insist upon mono-etiology in pathological processes or in vital phenomena. All were thoroughly aware of the fact that there are series of events in occurrences and developments in these fields, and that the end results are the net effect of many actions and counteractions.

By a regrettable slip Professor Eysenck in quoting the Royal College of Physicians’ conclusion that “cigarette smoking is a cause of cancer,” substitutes “the” for “a” This alters the statement significantly and runs counter to the explicit reminder in the College Report: “It is important to recognize that the hypothesis is not that cigarette smoking is the only cause of lung cancer. The fact that the disease does, rarely, occur in non-smokers and the effects of air pollution and various industrial hazards clearly indicate that other factors are concerned.”

Eysenck’s illustrative analogies are somewhat inaccurate. He tells us, apropos of the limited efficacy of epidemiological methods, that John Snow “finally removed the handle from the Broad Street pump and brought the cholera epidemic to a conclusion.” This well-worn dramatic tale has been exploded: The number of fatal attacks of cholera commencing each day in the area served by the Broad Street pump in 1854 had been falling rapidly, from 143 on September 1st to 12 on September 8th, the day on which the handle was removed from the peccant pump. In another analogy, Eysenck takes the inherited metabolic abnormality, phenylketonuria (consistently mispelled in text and index as “phenylcatonuria”) as an example of how by “understanding the precise way in which heredity works we can arrange a rational method of therapy which makes use of the forces of nature…” But so far from understanding the precise way heredity works in this instance, it is unfortunately still true that there are many obscurities in the matter. As Harris and others have emphasized, we do not know the nature and role of the factors that decide when and where the crucial enzyme is formed, nor how it is that the metabolic upset leads to a failure in mental development.

Professor Eysenck believes that atmospheric pollution is a more important agent upsetting “the delicate balance between carcinogenic and carcinoprotective factors” than it is credited with being in the British and the American Reports. It is true that the high urban mortality from lung cancer at first blush invites such an assumption, but against it are the very high death rate from this disease in Finland (where air pollution is slight and cigarette smoking very prevalent), the fact that the mortality is not higher in people exposed by their work to petrol and diesel fumes, and the finding that when the urban-rural residence factor has been controlled, there remains a large difference in mortality risk from cancer of the lung between smokers and non-smokers. It is not possible to agree with Eysenck that atmospheric pollution is probably a more important factor than cigarette smoking.

There is however much common ground between those who have satisfied themselves that cigarette smoking is the most important cause of cancer of the lung and those who dispute this, with or without an alternative explanation. The strong association between cigarette smoking and lung cancer is hardly any longer in serious dispute; by common consent the causes of lung cancer are multiple, the smoking habit depends in part on personality factors which are probably transmitted, and there is—who questions it?—much need for further research. Whether research should be concentrated particularly on the topics which Professor Eysenck prescribes—those bearing on the interaction between smoking, personality, and constitution—is a proposal likely to be judged according to our opinion of the strength of the case for the genetic hypothesis which he puts forward.

The most telling objections to a genetic explanation of why cigarette smoking and cancer of the lung are so closely associated all over the world, are those brought out by Doll and Bradford Hill in their report on ten year’s observations of British doctors. They found that a fall in mortality rapidly followed the cessation of cigarette smoking, and that it continued to fall step by step the longer smoking had been given up: This they justly urge “can be explained in terms of a diminishing risk from the previously operative environmental agent, but not in terms of genetic selection of those who choose to give up.” They also found that whereas the death rate from cancer of the lung in all men over twenty-five in England and Wales increased by 22 per cent during the ten years of their inquiry, among the male doctors as a whole whom they studied it fell by 7 per cent: This can easily be accounted for in terms of the great change in doctors’ smoking habits, but it runs counter to the implications of the genetic hypothesis. A further difficulty for anyone who regards inherited predisposition as a major factor is the recency of the enormous rise in the incidence of cancer of the lung: It is highly improbable that the genetic endowment of the population has changed correspondingly in the last forty years, but it is certain that the smoking of cigarettes has increased correspondingly during that period.

Eysenck’s quarrel with the medical conclusion on this matter is about the cogency of the evidence pointing to cigarette smoking as the main agent of death from cancer of the lung. The issue is not academic. The medical view is that we know enough to warrant doing all we can to discourage cigarette smoking. This would not preclude other measures directed, say, at reducing atmospheric pollution, nor would it lessen the need for research; but it would brand the main criminal. The genetic standpoint, as advanced by Eysenck, would not oppose such preventive methods, but would play down the role of cigarette smoking. He declares: “What we think we know is largely surmised, and even if most of it were true it would not help us to take any practical steps to overcome the menace of lung cancer…” Apart from some forms of psychological treatment by conditioning which he has espoused or developed, he sees very little prospect of dissuading people from smoking. He refers to “the possible exception of doctors who seem to have stopped smoking to a significant extent.” The phrasing is unduly cautious: Doll and Hill found that in their population of 31,000 male doctors, 19 per cent had given up the habit between 1951 and 1958. The consumption of cigarettes by men in the United Kingdom has declined by about 10 per cent since the Royal College of Physicians published their Report. There is, then, some hope that more vigorous action, combining educational and fiscal influences, might have more preventive effect than Eysenck supposes, especially in certain sections of the population. “One lifetime, nine lifetimes are not long enough for the task of blocking every cranny through which calamity may enter”; but cigarette smoking is, by a great consensus, such a cranny, which could, and should, be blocked. Medical critics consequently find it hard to give Professor Eysenck’s book, which would delay such action, the praise which its clarity and competence might otherwise earn.

This Issue

February 17, 1966