as the untoward condition has been removed, the child's misbehaviour has diminished and his outbreaks have ceased. In other instances some salient quality of the child's mind, existing from birth or inherited from his parents, at once explains the misconduct : a strong sex instinct, a weak and suggestible temper, or a general deficiency of intelligence. But in many cases to discover any one predominating factor is a more doubtful and precarious business. Here, to reduce the effect of personal preconceptions, each individual child has been discussed with other investigators who had equally a first-hand knowledge of his case; and an opinion has been passed only when an agreement has been attained. In a few rare instances, two or more factors seem to have exercised an influence that was almost equal; neither element alone, it would appear, could have precipitated the delinquency: it is the mutual reaction of the two which, by a kind of psychological chemistry, has generated the ultimate explosion. In such an event I have allowed the same weight to both cooperating factors by counting the equivalent of half a case to each. Last of all, there remains a distinct proportion, which, whether from the nature of the circumstances, or from the incompleteness of the analysis, have baffled all efforts at assigning any paramount factor. These last have been recorded under a separate heading of their own. In the rest of the cases the contributory causes have been separated into major factors and minor.

A summary of the numerous conditions, classified under fifteen heads, is shown in Table II. Major factors seemed discernible in about 95 per cent. of the cases, leaving about 5 per cent. (rather less in the case of the girls) with the major factor undetected or unassigned. In addition, subordinate factors1 were recorded about 900 times per hundred casesrather more with the girls, rather less with the boys. On an average, therefore, each delinquent child is the product of nine or ten adverse circumstances, one as a rule predominating, and all conspiring to draw him into crime.

The type of condition noted, however, is by no means peculiar to delinquent families. The same circumstances were observed in the nondelinquent cases 300 times per cent.2. Thus, in children of the same social class similar conditions may coexist-on an average, about 3 per case

1 Many of these are, of course, but aspects or consequences of other factors; thus the death of the father may lead to poverty, weak discipline, re-marriage of the mother, and a step-father complex with two or three different elements, all separately enumerated in the tables. In this section of the text, hereditary conditions, existing in the parents and relatives, have not been reckoned as additional to the corresponding condition in the child. * This figure does not include complexes, which were left unexplored in the controlgroup.

2 Hereditary conditions, enumerated under I as occurring in the family history, have not been included in the grand totals, since presumably they have already been reckoned, as occurring in the children themselves, in headings II, III and IV.

without plunging the child into a criminal career. It must, as a rule, therefore, be either the number of factors or the particular combination of them, that renders delinquency a probable result.

ERRORS OF SAMPLING.

With groups and group-differences so small as those here studied, it is needful to bear always in mind the degree of error which the narrow range of cases inevitably permits. Consider, as an illustrative instance, the effect of a sexually immoral home upon boys and girls respectively. Among the delinquent girls there were six living under such conditions; among the delinquent boys there were only three--half the number in a group nearly twice the size. Is this difference significant? May we validly deduce from it that an immoral mother contaminates her daughters more than she corrupts her sons? Or may the slight divergence of figures be, after all, nothing but a chance fluctuation due to the small numbers thus compared? The point can be settled by a simple statistical check. Computed by the customary formula1, the standard error of sampling for the difference observed proves to be 3-5. The difference between the two percentages is itself 2.4~ 8.2 5.8-less than twice the sampling error. Accordingly, it is highly possible that a difference relatively so slight might have arisen by pure accident. When, however, we turn to the larger groups, and compare the percentage for the entire set of delinquents (4.6) with that for the entire set of non-delinquents (0.25), the sampling error, in virtue of the larger numbers, now sinks to 1.5; the difference observed is almost three times this figure. Here, therefore, there is little danger that we may be dealing with some accidental discrepancy; and it becomes legitimate to infer that an immoral home definitely favours crime.

=

I shall not burden the reader with a "probable error" for each isolated figure. Where the statistical precisian requires it, the margin for inaccuracy due to sampling can be roughly gauged from the abbreviated table below (Table III). It will be seen that, in comparing the main groups of delinquents and non-delinquents, the differences in the body of the table can seldom be significant, unless the one percentage is three or four times the size of the other; in the totals, an addition of about half as much again may be suggestive. Thus, pairs like 0-5 and 5.0, 2 and 8, 6 and 14, 24 and 36, begin to indicate a genuine difference. In comparing the smaller groups, however-for example, the delinquent boys with the delinquent girls-little weight can be attached to the numerical differences, without further argument in their support.

1 G. Udny Yule, Introduction to the Theory of Statistics (2nd ed.), p. 269.

In this brief review, with data so slight and cases so meagre, I shall not venture to examine in concrete detail the mode and the direction in which every traceable condition may be presumed to have exercised its influence. My purpose is simply to submit a first preliminary survey; to emphasise the complexity of the problem and the variety of the issues; and to criticise one or two of the sweeping generalisations still current, for which the data here brought forward may be expected to supply some rough and tentative test.

The final upshot of my analysis is shown in numerical form in Table IV. The figures measure the degree of association between juvenile delinquency on the one hand, and the various types of condition observed, on the other1. With the loose data inevitable in sociological enquiries, statistical coefficients must not be too zealously pressed. Broadly speaking, however, the averages suggest the following deductions. All the conditions enumerated in the table are positively correlated with delinquency; but none to a very high degree. Extrinsic or environmental conditions, and intrinsic or psychological conditions, have about equal influence. Of psychological conditions the intellectual are rather less important than the emotional. Of environmental conditions those obtaining outside the home are far less important than those obtaining within it; and within it, material conditions such as poverty, are far less important than moral conditions such as ill discipline and vice. Physical conditions have roughly but half the weight of psychological and en

1 For the value in psychology of the coefficient of association, and for the formula used, I may refer to Appendix II of my book on Mental and Scholastic Tests. As with the coefficient of correlation, unity indicates complete interdependence; and zero, absence of all connection.

TABLE IV. Coefficients of Association between Delinquency
and Conditions Observed.

* In calculating this coefficient the percentage used for the control-group is that for the amount of poverty obtaining in the general London population, not the 200 nondelinquents. See Table VI, footnote (1).

vironmental; and, while hereditary factors might seem to bear the same importance as these two latter, it is not certain, without further data and discussion, that the influence of an inheritable condition works through inheritance only, and not also through the environmental situations and psychological reactions which it may incidentally set up.

In the following pages I shall briefly comment on the effect of these several factors, one by one, and examine the evidence of the detailed tables as succinctly as I can.

I. HEREDITARY FACTORS.

Whether there is such a being as a born delinquent, and whether criminality as such may be inherited, have for long been favourite topics of dispute among criminologists. For the psychologist they seem already to have passed into the region of those speculative questions which the