that is to say, parental intemperance is not associated with relatively worse general health in the children but a coefficient even of '07 with a probable error of '03 is not really significant, and we should say that as far as these statistics go there is no general association between parental alcoholism and defective health in the children. The nature of the interrelationship is far more subtle than we think many temperance advocates have realised. The fact, as shown in these figures, that the children of the intemperate are healthier than the children of the sober is probably due to the more virile and physically fit members of the community being liable to alcoholic temptation, and is as such an indirect effect of heredity and not a result of alcohol. The greater percentage of phthisical and epileptic children in the families of the sober is again probably due to the same source; i.e. these pathological conditions arise from inherited constitutions, and the parents of phthisical and epileptic children being themselves of a feebler constitution than average parents are less liable to alcoholism. It may further be noted that in the case of children dying young, while for sons the parents of both sexes are more sober than the average, for girls they are more alcoholic than the average; this appears to mark the result as due to an environmental rather than a toxic influence, and corresponds to what we have noticed in the relation of weight and height of children to parental alcoholism. Taking health as a whole we are compelled to say, that—excepting that phthisis and epilepsy occur less frequently with alcoholic parents-there is no significant association between parental alcoholism and defective health in the offspring. The differences are far too slight to permit of safe conclusions being drawn, and there is no intense and close relation between alcoholism and defective health or pathological condition in the offspring.

(ii) Edinburgh. We may now consider how far the data obtained from the Manchester special schools receive confirmation from other sources, and we turn to the data collected by the Edinburgh C.O.S. to find independently the relation between the drinking of the parent and the health of the children. In the Edinburgh report the general health is not stated for each child but the diseases from which each child suffers are given and upon this we are able to divide the children into various classes, i.e. healthy, suffering with glands (tubercular glands were kept separate), epileptic, phthisical, weak chests, rickets, weak hearts, eczema, etc.; there were not sufficient cases to keep all these categories separate and we have had to content ourselves in the first place with making three divisions, i.e. (1) healthy, (2) suffering from glands, (3) suffering from other diseases. The percentages are given in the four tables below of children divided into these three classes; the drinking of the parents is also divided into three classes: Parent (1) sober, (2) drinks, (3) drinks in bouts.

Drinking in bouts seems to be much less frequent among mothers than fathers and the numbers in that class are rather too small to give regular results. Two facts seem to be clear from these tables: (1) the larger percentage of children suffering with glands when the parents drink in bouts and (2) the larger percentage of children suffering from "other diseases" when the parents are sober. The associations worked out by contingency are Father and son 11, Mother and son 12, Father and daughter,

05, Mother and daughter, 14. Here again it is extremely difficult to tell what the sign should be and in order to ascertain it we divided the "other diseases" into three classes, those suffering from (1) heart, (2) chest and bronchial troubles, (3) other diseases which include rickets, eczema, curvature, delicacy, etc. and, in order to have a sufficient number of cases on which to base means, the two sexes were put together and the correlation ratios worked out for father and child and mother and child by the recently published method* The correlation ratios were 14 in each case but an examination of the means shows a curious difference. In the case of the mothers the means show a general downward tendency which indicates an increasing number of children suffering from heart, chest and other diseases in the sober class of mothers; but when we examine the curve for fathers we see an upward slope of the curve into the drink class except for "other diseases" when there is a very decided drop into the

sober class of fathers. It is very difficult to trace definite relationships here; they are too slight in character to be really significant on the basis of the numbers dealt with; they seem to be different for parents of the two sexes and again for offspring of the two sexes. The only definite conclusion that can be drawn is that there is no close and simple relation between parental alcoholism and defective health in the children, which can be ascertained from a sample of moderate size of the general population. There may exist slight and complex relations as indicated by the fact that "other diseases" show a significant tendency to be associated with parental sobriety rather than with parental alcoholism.

(6) Parental Alcoholism and Intelligence of Children. See Tables XXV to XXXV inclusive in the Appendix. First we will consider the results obtained from the Manchester special schools. In this case we could only divide the children into

* See footnote p. 9.

(1) mentally defective and (2) normal. There were a small number of children who were very dull, without apparently being actually defective, but these we included in the mentally defective category as they were certainly not normal, but the number of them being only 1 per cent. of the whole number of children the transference from one group to the other would make practically no difference to the correlation coefficient. The percentages tables are given on p. 14, and the correlation coefficients worked by the fourfold method, the only available method in this case, are given below each table.

FIG. 2. PARENTAL ALCOHOLISM AND HEALTH OF CHILDREN (EDINBURGH).

Of the four tables three show a negative correlation, that is to say a correlation between temperate parents and mentally defective children; the correlation coefficients are too small to be of great significance but they show very clearly that the statement that intemperance in the parent causes mental defect in the children must be received with caution; it is not the case among the defectives in Manchester. When the father is temperate we find 41 per cent. of his sons and 31 per cent. of his daughters

Son

Son

are mentally defective, which must be compared with 34 per cent. and 30 per cent. when he is intemperate. When the mother is temperate we find 39 per cent. of her sons and 30 per cent. of her daughters are mentally defective, which must be compared with 40 per cent. and 24 per cent. when the mother is intemperate. Here again we must repeat that we do not suppose temperance to be a cause of mental defect any more than we supposed it to be a cause of phthisis or epilepsy. The small association, if it be significant, is probably a secondary effect of an hereditary influence, the mentally defective children coming from a feebler stock, which has not the desire or possibly the capacity for alcohol of a stock of a more vigorous physique.

The above data are of course selected, they only show the relationship of alcohol and mental defect within families, one member at least of which is mentally defective. We shall next deal with the non-selected data obtained from the Edinburgh school. As before we will first give the table in the form of percentages, and the association found by the two methods of contingency and of the correlation ratio is placed below each table.

66

In examining the tables it must be remembered that children with "excellent" intelligence are few in number and we shall consequently find some irregularity in this class, and as those parents who drink in bouts are fewer in number than either sober" or "drinkers" so we shall find more irregularity in the percentages in the bout class. It is probably due to the smaller numbers that we find in the first table, 12 per cent. "excellent" boys when the father drinks in bouts, while there are only 3 per parents are concerned we find that excellent intelligence in their children is evenly divided; there is a slight excess of excellent intelligence among the sons or temperate fathers but in the other three cases we find a slight excess of excellent intelligence among the children of drinking parents. When we examine the percentages of children with good intelligence we find a slight excess of good intelligence among the sons of drinking parents and a slight excess of good intelligence among the daughters

cent. of "excellent" girls in the second table. As far as the sober and dry

Son

Son

of non-drinking parents. In the category of dull and defective intelligence we find an excess among the sons of sober fathers and in the other three tables we get the excess of dullards in the children of drinking parents. The differences throughout are slight and irregular and the correlation coefficients are only just significant. In this case as in the case of health we worked the tables a second way and found and plotted the mean deviations in order to find what sign to attach to these values—drinkers and bout drinkers were classed together. The values for ʼn are given below those found by contingency, and it will be seen that the correlations are scarcely significant.

η

It is almost impossible to determine whether the slight relationship noted above is to be considered positive or negative, i.e. whether temperance or intemperance of parents is associated with good intelligence in the offspring. We can bring this home to the reader in two ways. First the material was divided into four fourfold tables, the parents were divided into sober and drinking, the offspring into two groups excellent with good, and medium with dull. It was found that alcoholism of the parent went with the better intelligence of the offspring in the cases of father and son and mother and daughter, while in the cases of father and daughter and mother and son the reverse was the case. Even, if any weight could be given to the extremely small correlations the difference of sign in the four correlations precludes our asserting any marked and simple relationship. Secondly the accompanying diagrams indicating the average alcoholic tendency of the parent of each class of child, obtained by Pearson's new correlation ratio method, show how small is the trace of any significant relation between parental alcoholism and filial intelligence—the deviations bear wholly the