supposed returns for non-drinkers and moderate drinkers is necessary to show that they are not entitled to serious consideration.

The grave question as to the reliability of life insurance returns as means of determining the relative death-rates of abstainers and non-abstainers from alcoholic beverages goes much deeper than this. The classification of policyholders as to their respective drinking habits, when based on their own biased statements while trying to qualify as acceptable life insurance risks-of course, as a rule, many years before their deaths figure in the classification of the companies' mortality—at best would be an exceedingly dubious proceeding. But that is by no means the weakest link in the chain of alleged evidence as to the relative death-rates of drinkers and non-drinkers. What conceivable warrant is there for rating in one group all applicants for life insurance who cannot truthfully and categorically assert that they are total abstainers? By any such extremely loose classification the man who actually drinks maybe two or three glasses of wine or beer in the course of an entire year, the man who takes a cocktail or so practically every day in the year, and the man who regularly takes two or three drinks of whisky or other distilled liquor each day, are all counted in the single group of non-abstainers. Almost countless other sub-groupings on the basis of what they drink, how much they drink, and how drinking affects them individually, might doubtless be made in the large body of men comprehensively branded as drinking men, or non-abstainers, by the few life insurance companies which undertake so to classify their policyholders. And by adding up all the deaths in this group, in all probability including more or less men who are comparatively heavy drinkers, as well as the men at the other extreme who practically if not literally are total abstainers, an alleged death-rate for non-abstainers is obtained!

Can any such process be regarded as rational, let alone scientific? The fact that precisely that sort of process has been followed in certain quarters for many years proves nothing, for many foolish things have become habitual. Would it not be equally fair to classify policyholders by their answers to such absurd questions as, for instance, "Do you drink coffee?" "Do you eat Welsh rarebits?" "Do you take supper after going to the theatre?" Some of the men who had to plead guilty to these implied charges might confine their indulgence in coffee to the demi-tasse after the banquet which they semi-occasionally attended, might not care for Welsh

rarebits and not eat more than one a year, or might similarly deviate from the straight and narrow path of gastronomic proprieties by going to supper after the theatre once or twice a year, whereas others who had answered "yes" might be habitually addicted to these hazards of indigestion and belated sleep. Would any sane man consider it fair to group them all in together, and attempt to fix a conglomerate, quasi-composite death-rate for the coffee-drinkers, rarebit-eaters, and late-supper contingent on this basis?

I can see no more sense in pretending to work out the deathrate of non-abstainers on a similar basis; i. e., their inability, if strictful truthful, to set themselves up as total abstainers. In brief, the non-abstainers so classified are about as heterogeneous a body of men, and about as far from a homogeneous body of men, as could well be conceived. And surely there can be no difference of opinion as to the fundamental necessity of at least reasonably homogeneous groups if any classification of the slightest scientific value is to be made. The real issue is by no means the accuracy of mortality statistics thus compiled, but is, it seems to me, what percentage of the apparent excess of mortality in the case of nonabstaining as compared with abstaining policyholders is properly chargeable to their use of alcohol, what percentage to their other habits, what percentage to their occupations, and so on and so on almost ad infinitum? There is the real question, as I see it.

MANY FACTORS CONTRIBUTE TO FIX CLASS MORTALITY

To illustrate the difficulties of credibly and scientifically grouping the death-rates of various classes, suppose we look at the thoroughly reliable mortality statistics of England and Wales, admirably classified by occupations for many years by the Registrar-General's office in its decennial supplements. By the latest of these supplements issued in 1908 we find that the mortality of clergymen, priests and ministers at ages 35-45 in England and Wales in 1900-1902 was 4.06 per 1,000, whereas that of general shop. keepers at the same ages was 19.71 per 1,000, or nearly five times as high. Would it be safe to assume on the strength of these figures that the shopkeeper's occupation was nearly five times as hazardous as that of the clergy? Dozens of similarly striking contrasts might be cited, but perhaps no further evidence is needed to make it clear that almost countless factors contribute to fix the mortality rate of any particular group of lives, and that it would be absurd to

attempt to explain the difference between the mortality rates of any two groups by citing any single factor present in one case and supposedly absent in the case of the other group. That is precisely what is done by the vast majority of people who attempt to prove by the comparatively limited experience of the few English companies with separate classes of abstainers and non-abstainers the supposed difference in the probable longevity of people who do and people who do not use alcohol.

In the discussion before the Association of Medical Directors, as shown by the extracts on previous pages of this paper, Dr. Emery said: "I would like to make a suggestion for statistics in the future, and that is to find out how much a man eats. for oftentimes it is much more important. A man who is an overeater is not likely to live as long as some excessive drinkers." Have any of the discussions of the apparent death-rates of abstaining and non-abstaining groups in English life companies maintaining such groups of policyholders seriously taken into account the question as to the eating habits of the two groups? Not in so far as I can recollect. Nevertheless, the bad physiological effects of over-eating in the way of undue strain on the digestive organs and the clogging of the whole system, and the resultant tendency toward arterio-sclerosis, apoplexy, and other degenerative diseases are so well known as to call for no enumeration here. I have had many eminent physicians tell me they had no question that more people were killed by over-eating than by over-drinking. But practically no consideration seemed to be paid to this important factor by the people who triumphantly cite the apparent excess of the death-rate of non-abstainers over that of total abstainers.

While it is scarcely safe to trust to generalizations as to the types and general habits of men who do and men who do not drink, it possibly might be guardedly said that as a rule that section of the modern community which would commonly be regarded as the drinking class leads more active and strenuous lives, keeps later hours, and in both their work and play is subject to more strain than are those who never indulge in any form of alcohol as a means of relaxation. Perhaps it would not be going too far to say that, all things considered, the total abstainers and non-abstainers are two entirely different types of men-although of course there are many notable exceptions.. Assuming for the moment that this hypothetical type-differentiation is correct, might it not be ex

pected that there would be more or less difference in the death-rates of the two types of men, even though alcohol were left out of the argument? I think there can be little question on that score. If such is the case, it is manifestly a case of unsound reasoning and unscientific conclusions to set up the difference between the deathrates of abstaining and non-abstaining groups of life insurance policyholders as an accurate measure of the effect of alcohol on human longevity. That is constantly being done, however, in the impassioned arguments of all the would-be reformers who do not themselves approve of the use of alcohol, and are determined to bring the rest of the world around to their way of thinking.

For the reasons already stated, I am personally unable to regard the oft-quoted figures of the handful of small English life insurance companies which have long maintained separate classes of abstaining and non-abstaining policyholders as evidence of any. scientific value to either (1) the respective death-rates of the drinking and non-drinking classes of the community in general, or (2) an accurate measure of the supposed excess of the death-rate of drinking people over that of non-drinking people. If the reasons for this distrust previously cited do not raise a similar question in the minds of all unbiased readers, suppose the figures for the two sections of the oldest and largest of all the English life insurance companies in the field of discussion, the United Kingdom Provident Institution, are brought into the argument and critically analyzed. In his paper presented before the Association of Life Insurance Medical Directors of America, on the subject of "The Use of Alcohol and the Life Insurance Risk," Dr. McMahon included a tabulation of the "mortality experience under ordinary whole life policies" of the "temperance section" and "general section" of the above named English company for the forty-five years, 1866-1910, arranged by five-year periods. In the figures thus presented were included the amounts of expected and actual claims in both sections for each period, but as the amounts of claims have absolutely no bearing on the number of claims, or deaths, and as the averages of claims in the two sections for the forty-five years were substantially identicalrespectively $1,186 and $1,167 for the Temperance Section and the General Section-there would seem to be no reason for clouding the issue with these irrelevant data. Thus edited, the tabulation of the mortality experience of the United Kingdom Provident Institution presented by Dr. McMahon was as follows:

At first glance these figures would seem to show a general uniformity, and both demonstrate and sharply accentuate the difference between the death-rates of abstainers and non-abstainers by an experience of forty-five years with a total of 23,699 deaths or an average of about 527 deaths a year, or approximately 10 deaths a week. A study of the figures in detail, however, reveals some puzzling discrepancies. All of the persons whose deaths are included in the above-tabulated mortality presumably had passed the customary medical examination and had been accepted as good life insurance risks. If the only material difference between the two classes was the fact that one class did not, and the other class did, use alcohol, might it not naturally be expected that the difference between the death rates of the two classes would be a fairly constant difference, at least if computed by five-year periods and any slight annual discrepancies thus eliminated? If the relative mortality of large groups of policyholders were tabulated for five-year periods by either occupations or localities, a substantially constant difference would be expected unless changes in conditions or other abnormal

*This total as printed in the Proceedings of the Association differs by 90 from the sum of figures in the column, indicating a probable typographical error in the column, as the ratio of actual to expected claims, 70.47% tallies with the total given.