field in which the federal government has exerted almost no influence, has been as great as the extension of the registration of deaths.

Regarding divorce it need only be said that this is the one branch of vital statistics in which the United States has made contributions of capital importance, when judged by international standards. The most important sources of statistical information about divorce throughout the world are the two reports of the federal government covering together the forty years between 1867 and 1906.

The outlook for American vital statistics then appears hopeful. The next generation may and should do much to elevate it to the level of the best work done in older and more densely settled countries.

The recent rapid development of the public health movement has reinforced the demand for competent demographers and so for an adequate training in vital statistics. Our states and cities might do much more for public health than they are doing if they could find more readily men able to show statistically the need for and the success of remedial measures. The medical schools to which we naturally look give little, if any, training in demography and cannot, because their curricula are badly overcrowded. A suggestion may be found in British experience. In that country a degree of B.S. in public health, or a diploma in public health, is given to a doctor who pursues after graduation a prescribed course in public health, including vital statistics. This example could be followed by our medical schools only if the position of health officer or registrar were to be filled by the best-qualified person, whatever his residence at the time of appointment, and carried a salary on which the appointee could live.

It would be possible to review the other important branches of American statistical work and to show that in most, if not in all, of them interest has been growing during the last few years. The development of demands for an effective and detailed supervision of public-service corporations, for a dispassionate and expert investigation of the industrial and commercial conditions with which a federal tariff law has to deal, for carefully planned

budgets, federal, state, and municipal, may be mentioned merely as examples of other directions in which the call for better statistics is now imperative. To develop that argument this afternoon would require too much time. Allow me, then, if you please, to assume that such a growth has occurred, is likely to continue, and deserves encouragement and support.

How, then, may it be expedited? If American statistics are to progress more rapidly in the future than they have in the past, there is urgent need for a larger number of men qualified by nature and training for a statistical career. Many of those now engaged in it have obtained all their knowledge in the office. Some have been drafted in from other occupations; others have risen through the successive grades of the service; few knew anything of statistics before their appointment to a statistical position.

The work of a statistician is not yet established in this country as a profession and hardly as a career. Evidence of this may be drawn from the position of Superintendent or Director of the Census, no doubt our most conspicuous statistical office. Since 1850 there have been eight superintendents or directors, the average length of whose service has been four and one-half years. I have compared the careers of the seven persons most conspicuously identified with recent census work in England, France, Germany, Prussia, Italy, Austria, and Russia. The average duration of their official life and their work in statistics was twentysix years, nearly six times that in this country. The most notable exception to the rule that the official lifetime of an American statistician is very short is in the career of our late honored president, Carroll D. Wright, who was in the harness at Boston or Washington for thirty-two years. Vires adquirit eundo. May there be many such instances in the years before us!

How should statisticians be trained? Some say the office is the only good training school; others demand a preliminary course of study at an educational institution. No doubt each method can produce good men, but the best results in most cases are secured by a combination of the two sorts of training.

There are parts of statistics which can be taught in a univer

sity far better than in an office. Such are its history and theory, the presentation of its main results as a coherent system of facts and principles, the comparison and criticism of methods and possibly in some cases the interpretation of conclusions.

There is another and less obvious aid derived from the academic teaching of statistics. An office with a large clerical force is seldom so organized as to enable its heads to select quickly and accurately the clerks who best deserve promotion to responsible positions. Most new clerks are assigned duties of a routine character which do not quickly reveal a man's quality. It is one important function of a university to evaluate its students, not merely or mainly by the crude test of marks, but rather by developing in the teachers a sympathetic and yet critical estimate of each student's power and promise. This estimate helps to direct into the statistical field some who have a native capacity for and interest in it and helps also to hasten the passage of such students through the deadening early stages of office life.

During the last few years, notwithstanding an increase in the public appreciation and demand for trained statisticians, there has been little addition to the amount or improvement in the quality of statistical education. If the fact were otherwise, we might feel much more confident of the outlook for statistics. In my judgment the colleges and universities have not kept pace with the popular readjustment. No doubt the short supply of trained men has retarded the progress of the work.

Another serious obstacle to the rapid development of statistical work has been the lack of any national statistical center, such as each great European country possesses at its capital. Years ago that position belonged probably to Boston; now it does not. If federal work in statistics continues to grow as it has done of recent years, Washington will soon take, if indeed it has not already taken, the leading place. Whether or not our development shall make any one place, like Washington or New York, pre-eminent in the variety, importance, and quality of its statistical work, there is surely need for continuing the education of statisticians in official life by providing a forum where they may meet and by mutual conference and friendly criticism may con

tribute to the progress of each other's work. European experience has found the great advantage of such conferences and since 1853, when the first international statistical congress assembled at Brussels, the number and variety of such meetings have slowly increased, barring the slight check after the war of 187071. In the variety of its conditions and its problems this country is almost continental, and periodic conferences of statisticians within its limits are needed now and the need is likely to become imperative. How the need shall be met and whether this association finds in it any duty and opportunity are questions I submit to its consideration.

Closely correlated with this but perhaps even stronger is the need for a center at which the producers and the critical consumers of statistics may meet and fraternize. There has long been some danger that these two groups would not work harmoniously in the common interest. Probably most of us have seen instances of the kind. While the danger is less than it was ten years or more ago, yet so long as the two groups keep at arm's length and fail to exchange experiences and opinions, the danger may revive, either with groups or with individuals. Here too may be a field. for this association.

Our society is now seventy years of age and, with the exception of the Royal Statistical Society, it is, I believe, the oldest statistical organization in the world. The subjects with which it is especially concerned have developed but slowly in the United States, owing to conditions which are fast disappearing. Never have the growth of interest in statistics and the demand for thorough and dispassionate statistical analysis and interpretation been so rapid. The most imperative need is for the recognition of statistics as a career or profession and for facilities whereby it may be adequately taught and effectively acquired. Notwithstanding this reservation, the outlook for American statistics is bright and encouraging.

THE SOCIAL MARKING SYSTEM

PROFESSOR FRANKLIN H. GIDDINGS
Columbia University

In their quantitative aspect many data of sociology are positions in a scheme, rather than distances from zero. Tables of births, deaths, and migrations, like tables of height and of weight, give us measures from zero. Degrees of difference or of resemblance which we observe among our acquaintances and others, the affiliations of nationalities and of races, of religious beliefs, and of political interests, the values that we assign to ability and to conduct, and the social ranks that make up differentiated communities, are merely positions in a scheme.

Sir Francis Galton long ago showed that when positions in a scheme are successive, and may be successively numbered, and when the instances in which given phenomena fall into successively numbered positions, can be counted, and their frequencies set down, we can legitimately subject the numerical data so assembled to familiar methods of statistical analysis. We can plot their rough curves of magnitude and of frequency; we can ascertain their medians, quartiles, and probable errors; we can determine their modes. Whether we can also make significant use of their standard deviations and coefficients of variation, is a question that need not now be considered.

In order to assemble and to sort such data, however, we must have a marking-scale. Also we must have a concrete knowledge of the facts to be sorted, like that on which we rely when we assign numerical grades to examination papers, or to tests of conduct. The marking-scale must not be arbitrary, a mere ingenious invention. It must be a simple and natural expression of observed relations. If possible, its elements should be derived from those common-sense gradings which the world uses in rough and ready fashion for the purposes of everyday life, and which therefore have the warrant of experience.