Popular Science Monthly/Volume 80/May 1912/Assortative Mating in Man

by James Arthur Harris

Layout 4

ASSORTATIVE MATING IN MAN

By Dr. J. ARTHUR HARRIS

CARNEGIE INSTITUTION OF WASHINGTON

I. Introductory

"THE supreme misfortune is when theory outstrips performance." wrote Leonardo da Vinci.

He gave us, as far as I remember, no illustration of his epigram, but one is at hand in the modern attitude towards the Darwinian principles of natural and sexual selection. Exalted as they were on comparative evidence alone, and by post-Darwinian enthusiasts who were not only fertile, but liberal to a fault, in assumption, their debasement was inevitable. We are now in the period of reaction when men disparage selection, or dismiss it entirely as an evolutionary factor. Against this unreasonable extreme of opinion these essays^[1] are directed. They are simple reviews, pretending merely to set forth honestly the results secured by biometricians in their studies of these exceedingly difficult biological problems.

Their purpose is, I admit, in reality two-fold. Not only are they a direct plea for a more open-minded—a stringently critical rather than a dogmatic—attitude towards the Darwinian factors, but an indirect appeal for a wider recognition of the biometric methods which make possible the measurement of the intensity of the primary factors of organic evolution.

The strongest arguments are those of quantitatively expressed facts. The best way of overcoming the prejudices and other obstacles against which the biometrician works is to allow these facts to speak for themselves, if possible in terms comprehensible to the layman.

Let us turn, therefore, to the available facts.

 

1. The Problem of Intra-racial Sexual Selection

There may be forces other than propinquity tending to fuse different races which through migration or otherwise come to occupy the same territory. Sociologists have, however, long emphasized intragroup marriages as one of the forces which tend to keep them distinct.^[2]

This tendency to mate within the tribe is a factor of first-rate biological and sociological importance, serving as it does to maintain racial boundaries. But if sexual selection be a real factor in the evolution of races—in the differentiation of groups as well as in their maintenance—its action must be sought within the race, an intra-racial sexual selection must be demonstrated.

Sexual selection in man may, as Pearson long ago pointed out,^[3] be of two kinds, preferential mating and assortative mating. By preferential mating one understands that certain classes of women are discriminated against by the average man, or by men as a class, or vice versa. By assortative mating one means that in the population of men and women who do marry, there is a definite tendency for certain classes of men to marry particular classes of women, and conversely.^[4] An almost prophetic quotation from Pearson may render the distinction clear.

For example, preferential mating might lead in a highly social community to the rejection of consumptive mates, while assortative mating might through localization or community of habit, lead to considerable consumptive correlation. Thus sexual selection as a whole may influence in diverse ways the inheritance of the consumptive taint.

Another illustration faces us in the problem of deaf mutism. Normal individuals discriminate against deaf mutes, for obvious reasons. There is a stringent segregation of the class, resulting from educational and social conditions, and as a result there is for the people, as a whole, a strong assortative mating, hearing people as a class marrying hearing, and deaf marrying deaf.

The scope of this review is limited to a discussion of the quantitative results which have been attained for assortative mating.^[5]

It is needless to say that a subject so fascinating to man as

anything pertaining to human mating has been the subject of wide speculation and assertion since the time of da Vinci.

Schopenhauer states that every person requires from the individual of the opposite sex a one-sidedness which is the opposite of his or her own. The most manly man will seek the most womanly woman, and conversely. Weak or little men have a decided inclination for strong or big women, and strong or big women for weak or little men. Blondes prefer dark persons or brunettes; snub-nosed, hook-nosed; persons with excessively thin long bodies and limbs those who are stumpy and short, and so on! Analogous superstitions are wide spread,^[6] though differing in form. Westermarck,^[7] in summarizing the views of various writers adds, "If contrasts instinctively seek each other, this may partly account for the readiness with which love awakens love."

Some have even ventured the opinion that where the husband and wife are unlike, the offspring are more numerous, or stronger! Again there is the popular superstition that after a long life together husband and wife come to resemble each other physically.^[8]

Of course conclusions the opposite of all of these are not wanting.^[9]

Such is the state of knowledge to which the unaided observation of a complex phenomenon can lead us—a snarl of contradictions. As far as we know, the only method of disentangling it and arriving at some certainty is the analysis of large bodies of observations by means of refined statistical methods.

 

2. The Measurement of Assortative Mating'

The precise meaning of the term "assortative mating" may perhaps be made a little clearer in the process of explaining how the similarity or dissimilarity of husband and wife is measured.

Suppose a most highly refined socialistic community should set about to equalize as nearly as possible not only men's labor and their recompense, but the quality of their wives. It would never do to allow individuals to select their own partners—superior cunning might result in some having mates above the average desirability, which would be socially unfair!

The method adopted would be to write the names of an equal number of men and women officially condemned to matrimony on cards, and to place those for men in one lottery wheel and those for women in another. The drawing of a pair of cards, one from each wheel, would then replace the "present wasteful system" of "competitive" courtship. If the cards were thoroughly shuffled and the drawings perfectly at random, we should expect only chance resemblances between husband and wife for age, stature, eye and hair color, temper and so on; in the long run, a wife would resemble her husband no more than the husband of some other woman. In this case, the mathematician can give us a coefficient of resemblance, or of assortative mating, which we write as zero. The other extreme would be the state of affairs in which men of a certain type (that is to say men differing from the general average by a definite amount) always chose wives of a definite type; the resemblance would then be perfect and the correlation, as we call it, would be expressed by a coefficient of 1.

Fortunately, the meaning of correlation can be illustrated by a character for which the reader knows that there is a high degree of assortative mating. The table^[10] shows the age of bride and groom in 2,500 Chicago couples. The swarm of figures showing the frequencies of different combinations spreads diagonally across the table, demonstrating that while men or women of any class marry consorts of varying ages, there is a pronounced tendency for individuals of the same relative age^[11] to mate. From such a table the statistician calculates the equation to a straight line (or to a curve of a higher order) which describes approximately the change in the average age of brides associated with increase in the age of the grooms. The diagram (Fig. 1) shows that the agreement of the theoretical line with the empirical means is very close indeed.^[12] Or he may express the closeness of correlation quite independently of the absolute values of the two characters on the scale of

Fig. 1. Age of Groom.

 

TABLE I

−1 to + 1. In this case, the correlation is +.75 or about three fourths of the distance up the scale of 0 to 1, from no resemblance to perfect identity.^[13]

To determine whether men and women tend to parity or disparity in matrimonial choice, we must, therefore, take a large number of mated pairs at random from the general population, sort them into groups according to some characteristic—quantitatively measurable whenever possible—and determine by means of the statistician's coefficient of correlation whether generally similar or dissimilar groups of men and women tend to mate, and how strong this tendency to parity or disparity is. Throughout this paper the intensity of assortative mating will be expressed by the coefficient of correlation. The reader will have to bear in mind merely that positive coefficients indicate a similarity and negative coefficients dissimilarity in husbands and wives as compared with random pairs of men and women from the population,

 

II. Assortative Mating for Physical Characters

1. Stature

The psychological basis of the popular notion that men and women seek disparity rather than parity in the stature of their mates is not far to seek. On the streets the linear wife and spherical husband, or the reverse combination, appeal to our sense of humor while the multitude of similarities pass unnoticed. Yet when lumped on the statistical scales the modal multitude may outweigh the extreme combinations whose incongruity provoke a smile as they pass to the front circle after the curtain has gone up.

With the rough statistical methods then available, Galton^[14] was unable to detect any tendency towards marriage selection with respect to stature, but Pearson^[15] on the same data as early as 1897 suspected homogamy. The results from his own more extensive family records are shown in Table II.^[16] For convenience the figures for forearm and span are also given.

TABLE II

	Husband's Character	Wife's Character	Assortative Mating (Correlation and Probable Error)
Direct correlations	Stature	Stature	+.2804 ±.0189
	Span	Span	+.1989 ±.0204
	Forearm	Forearm	+.1977 ±.0205
Cross correlations	Stature	Span	+.1820 ±.0201
	Stature	Forearm	+.1403 ±.0204
	Span	Stature'	+.2023 ±.0199
	Span	Forearm	+.1533 ±.0203
	Forearm	Stature	+.1784 ±.0201
	Forearm	Span	+1545 ±.0203

 

Fig. 2. Scale of Husband's Stature.

We find a resemblance r = + .280.^[17] Thus there is a very pronounced similarity in stature between husband and wife.^[18] This is made clear by diagram 2, which shows the empirical and the (smoothed) mean statures of wives of husbands of different heights.

 

2. Other Bodily Characters

Data for physical characters other than stature are few;^[19] the only important ones are forearm and span in Pearson's investigation.^[20] For the direct^[21] comparisons the resemblance is quite material, amounting to about + .200. Cross correlations are in general somewhat lower. Possibly the assortative mating is primarily for stature, while the resemblance for forearm and span arise merely because these are closely associated with stature. This point is not as certain, however, and deserves more attention.

For forty-eight families of East European (Russian) Jews living in New York City, Boas^[22] found an assortative mating for cephalic index (relative head breadth) of r = + .15 ± .10.

Lutz's interesting data on the inheritance of the method of clasping the hands^[23] show a positive sign in the correlation between the two parents, but it is so very small that no significance is to be attached to it.

 

3. Complexion and Hair and Eye Color

Complexion and hair and eye color are conspicuous with stature in the popular superstition of "the charm of disparity" in human matings.

The calculation of the intensity of relationship in the case of a character which like eye color is not quantitatively measurable presents considerable difficulties, and the results vary with the method employed. From an analysis of Francis Galton's "Family Records," Pearson^[24] has deduced the following values:^[25]

Correlation by four-fold method	+ .10 ± .04
Mean contingency	+ .31
Mean square contingency	+ .26 ± .03

The only series of data for hair color in husband and wife (where one may not suspect artificial selection of cases) is that given by Mr. Galton^[26] for the parents of English scientific men. I have grouped his data into three classes, and calculated the coefficient of mean square contingency, which is + .34. No stress is to be laid on the result, since the number of cases is small and the coarse grouping—which is necessary if the constant is to be calculated at all—unsatisfactory.

 

4. Physical Defects and Pathological Conditions

To Alexander Graham Bell^[27] probably belongs the credit of first laying great stress upon the social consequences of assortative mating. The title of his memoir, "Upon the Formation of a Deaf Variety of the Human Race," sufficiently indicates the seriousness with which he regarded the intermarriage of the deaf.

Probably assortative mating for deafness is more nearly perfect than that for any other known character. The reasons for this are patent. Hearing individuals rarely choose non-hearing mates. When both partners are deaf, on the other hand, they are united by the strong bond of fellowship and sympathy growing out of their similar condition, they communicate with each other with perfect ease and freedom, and the social interests and sympathies outside their own home are the same.

The extensive records given by Fay^[28] rather than those used by Bell in his pioneer study may furnish illustrations.

Table III. shows that in the marriages of the deaf, 72.5 per cent.

 

TABLE III

Marriages of the Deaf	Number	Percentage
Both partners deaf	3,242	72.512
One partner deaf; the other hearing	894	19.995
One partner deaf; the other unreported whether deaf or hearing.	335	7.493
⁠Total	4,471	100.000

have both of the contracting parties deaf as contrasted with 20 per cent. in which one is deaf and the other a hearing person. When we consider that in the general population of the United States there are roundly 1,500 hearing persons to one deaf, and consequently about 1,500 hearing persons to one deaf from whom a given deaf individual might seek to select a life partner, we see to what enormous extent sexual selection is at work for this character.^[29]

Bell laid great emphasis upon the influence of educational segregation, especially upon the use of a sign language, with its subjective influence on thought, in bringing about the intermarriage of the deaf. That this is a factor appears from Fay's elaborate records. He classified 7,277 deaf individuals according to the method of education and found that of those who attended boarding schools for the deaf, 86.2 per cent, married deaf mates, while of those who attended day schools, or both day and boarding schools, for the deaf 77.8 per cent, married deaf consorts.^[30] In contrast are the records of those who attended no school for the deaf: in this class, 62.4 per cent, married deaf individuals. The difference between 62.4 per cent, and 86.2 per cent, probably indicates roughly the influence of scholastic segregation.

Fay also finds that of the pupils who attended exclusively oral schools 78.2 per cent, married deaf partners, while of those who were educated at schools not exclusively oral, or partly at schools exclusively oral and partly at schools not exclusively oral, somewhat over 86 per cent, of marriages were homogamous for deafness. Perhaps these figures indicate a sensible influence of the method of instruction.^[31] Nevertheless, one cannot but be impressed with the intensity of the assortative mating that occurs independent of this factor. With no such isolation 62 per cent, of deaf individuals marry those who are deaf. Considering the intensity of the inheritance of deafness,^[32] we see what grave social results may be expected from this tendency.

Apparently unions where both members are deaf are more happy than those where only one is so afflicted. Table IV. gives the best available

 

TABLE IV

Marriages of the Deaf	Num- ber of Mar- riages	Divorces		Separations		Divorces and Separations
		Num- ber	Per- centage	Num- ber	Per- centage	Num- ber	Per- centage
Both partners deaf	3,242	33	1.018	51	1.573	84	2,591
One partner deaf, the other hearing	894	25	2,796	33	3,691	58	6,488
One partner deaf, the other unreported	335	7	2.090	7	2.090	14	4.179
⁠Total	4,471	65	1.454	91	2.035	156	3.489

records indicating the "success" or "failure" of like and unlike matings. Of course divorce, separation or number of children do not tell the whole tale; they give rather a lower limit to the measure of domestic infelicity.

For general health in husband and wife, classifying as very robust, robust, normally healthy and delicate. Miss Elderton^[33] calculated from Pearson's "Family Records" a relationship of + .27.

For freedom from constitutional diseases—i. e., freedom from any specific pathological taint without regard to the strength or delicacy of constitution—Goring^[34] found these resemblances for families of criminals:

Very poor and destitute	+ .17
Well to do and prosperous poor	+ .08
All	+ .11

The possibility of infection reinforcing constitutional likeness in consorts is presented by a disease like tuberculosis. From the evidence now available^[35] there can be little doubt that if one member (husband or wife) of a pair be tuberculous the other is more likely to be affected than if the first be sound. In short, there is a correlation for tuberculosis between spouses which has sometimes been called "marital infection."

But such a term implies entirely too much. The correlations are not so high but that one may suspect them to be due, in considerable part, to assortative mating for the physical and psychical characteristics which underlie, or at least accompany, the predisposition to tuberculosis. Those who have analyzed the problem most minutely are inclined to attach importance to both factors, but to lay especial stress upon assortative mating.

 

5. The Influence of Numerous Local Races

The reader who is a keen traveler will probably suggest that the general population of England, whence most of the data have been drawn, is made up of local races differentiated with respect to physical characters,^[36] and that marriages tend to be contracted between neighbors—i. e., within the local race.

Pearson has emphasized the possibility of this factor^[37] but with justice points out that his data for stature were taken largely from the professional classes—chiefly residents of London and other larger towns. While these marry in their own "sets," these can hardly be regarded as "local races." The records from cemeteries, to be discussed in the following section, were taken purposely from narrowly limited districts; tables were formed for each locality separately. Certainly such correlations can not be attributed to the heterogeneity arising from the mixing of differentiated samples. In other cases, the possible influence of local races has been well excluded.

Even if the demonstrated resemblances were due merely to the tendency to marry within the local race, and indicated no conscious or unconscious sexual selection on the part of individuals of the same race, they would nevertheless be of great interest as showing quantitatively the force of one of the factors tending to maintain racial boundaries.

 

III. Assortative Mating for Duration of Life

That human matings should so depend upon the visible physical and intangible, but none the less real, psychical characteristics as to give rise to a measurable bodily and mental similarity of spouses, while contrary to popular belief, seems not unreasonable. That this resemblance should extend to duration of life—a character quite unknowable at the time of marriage—appears at first thought highly improbable.

But duration of life is not a simple attribute. It is rather a conveniently measurable epitome of many physical and physiological traits, as well as of environmental conditions. Both its inheritance and the evidences which it furnishes of the action of natural selection in man probably depend upon its being the resultant of a complex of factors. May not assortative mating for appreciable personal characters result in an assortative mating for duration of life?

The answer given by the correlation between the age at death of husband and wife (Table V.) is affirmative—clear and emphatic.

 

TABLE V

Source of Material	Number of Pairs		Coefficient of Correlation Measuring
Wensleydale and district cemeteries	876		+ .2200 ± .0244
Oxfords'hire cemeteries	890		+ .2500 ± .0211
Society of Friends' records	1,000		+ .1999 ± .0212
			⁠————
⁠Mean			⁠.2233

Warren, Pearson, Lutz, Lee and others^[38] drew their records from the tombstones of rural English churchyards and from the archives of the Society of Friends—two quite dissimilar sources. Yet the results are in remarkable agreement; considering the errors common to statistical constants based on numbers of a thousand or less, one can not assert that they differ at all. The close agreement with stature, forearm, span and other physical characteristics is clear.

A criticism which may occur to the reader is that these resemblances are purely spurious, and due to the fact that husband and wife are not likely to be buried in the same ground unless they die within a short period of each other. This is precisely what would occur in districts with a shifting population. This very difficulty was anticipated. All urban churchyards were excluded because of the heterogeneous and transitory nature of the population. "In most rural districts, on the other hand, with a stable population, there is a very strong feeling—amounting in the case of the Yorkshire Dales almost to a superstition—that husband and wife must share the same grave." In view of the careful selection of localities and the agreement of results secured from Quaker archives with those from the gravestones, it seems clear that the resemblance can not be dismissed as purely spurious.

Duration of life is doubtless dependent upon environment as well as upon constitution. Slight differences in the healthfulness of the several parishes of a district might promote or oppose fullness of years in men and women alike. If this were true, the lumping together of the records from a number of churchyards would result in a correlation for duration of life, whether there be any real assortative mating or not. If this criticism be valid, random pairs of husbands and wives from the same parishes lumped together to form a table for the whole district, should show correlations as high as those for actual married pairs. The correlation really found was sensibly zero, demonstrating that local environmental conditions can not explain the results.

But within the same general environment, members of a family are exposed to a set of conditions peculiar to themselves. In a city block or country parish food, temperance, sanitation, risk of zymotic disease, and physical and mental habits differ greatly from family to family. In addition to the physical and social environment common to man and wife, there is also the fact that the death of one member of the pair has a profound effect on the other. Financial and associated domestic changes are often due directly to this cause, to say nothing of the overstrain of care during long illness or the shock of sudden death. May not the similarity in the duration of life of husband and wife be a consequence of domestic environment? This point was investigated by methods rather too complicated for explanation here, but which indicate that the sameness of home conditions can not account for the relationship.

Until further evidence is available, we must, therefore, conclude that there is a real, though assuredly unconscious and quite indirect, assortative mating for duration of life. Nor is this to be marvelled at, considering the results already noted for normal and especially for abnormal bodily characters.

 

IV. Assortative Mating for Psychical Characters

That psychical characteristics should play some part in human matings seems a priori highly probable.^[39] Actual facts are, however, few.^[40] Galton concluded^[41] that even good and bad temper made very little difference in marriage selection, but he pointed out many difficulties of obtaining trustworthy evidence.

Elderton^[42] found for "Intelligence, Temper, Temperament"^[43] and success in career the values given in the accompanying table.^[44]

Mental Character		Relationship
	Intelligence	.33
	Temper	.18
	Temperament, excitable	.11
	⁠Sympathetic	.15
	⁠Reserved	.27
	Success in career	.48

 

For insanity, working with Pearson's "Family Records" and using two different methods of classifying the "normal," "insane," "nervous" and "doubtful" entries so as to get the upper and lower limits for assortative mating. Miss Elderton^[45] finds + .244 as the lower and + .361 as the upper limit, say roughly an intensity of .30 ± .05. Goring^[46] considers that assortative mating is a factor of greater importance in the upper than in the lower social classes. In his records for insanity in criminals, he finds an assortative mating +.35 for the "well-to-do and prosperous poor" while it is probably absent in the "very poor and destitute."^[47]

 

V. Preferential Mating and Assortative Mating for Social Attributes

To mark off sharply social attributes from those which are physical and psychical is as impossible as it is idle. Certain traits dependent on wealth, family history, education or opportunity may be, for convenience merely, designated as social. To what extent do they influence mating?

Their potency is greatest in caste, royalty and peerage. Even in countries which pride themselves on the absence of social strata, wealth, family pride and feuds, religion and education, play their part in limiting the range of choice in marriage selection. But nowhere is mating within the class universal. The much-multiplied American dollar plays havoc with continental pedigrees. The pure breeding of the English nobility is a pretence; "the lawyer, the farmer, the silk mercer lies perdu under the coronet, and winks to the antiquary to say nothing." Some day the weight of these social forces will be determined, but the proper kinds of facts are not yet available.

Alcoholism is one of those interesting cases in which direct personal or social influence may supplement and reinforce the resemblance possibly due to assortative mating. Goring,^[48] dividing his material for English criminals^[49] into three classes for social status, finds these coefficients of resemblance:

Very poor and destitute	+ .44
Prosperous poor	+ .58
Well-to-do	+ .69
All	+ .70

For criminality, Goring's results will strike many as surprisingly low. They are:

Very poor and destitute.	+ .18
Well-to-do and prosperous poor	not calculated
All	+ .20

An interesting point of an entirely different nature has been raised by Heron^[50] who in a study of the distribution of sex in human families finds, "that in the free mating of man, families with a preponderance of female or male elements are not drawn upon equally with families in which the sexes are more equally balanced."

 

VI. Homogamy and Fertility

From the notion that in marriage "opposite poles attract" the step to the conclusion that dissimilar are more fertile than similar unions is so easy that it has sometimes been taken, though without any valid evidence in justification.

Fay^[51] considers that marriages of the deaf are possibly slightly less fertile than those of hearing persons. When both partners are deaf the percentage of sterile marriages seems higher and the mean number of children in fertile mariageslower than in unions in which one member of the pair is a hearing person.^[52]

Homogamy for stature^[53] and for eye color Pearson (Phil. Trans. Roy. Soc., A, Vol. 195, pp. 148-150, 1900; Proc. Roy. Soc. Lond., Vol. 66, p. 323, footnote) has considered Galton 's data without arriving at final conclusions. De Candolle (fide Westermarck, "History of Human Marriage," p. 335) states that the number of children is considerably smaller when the parents have the same color of eye than when they were contrasted. Wittrock (Ymer, Vol. 5, pp. vii-ix, 1885) was unable, on Swedish materials, to detect any difference in fertility between the two classes of marriages. have been considered in relation to fertility.

But as yet the data are far too meager for such complex problems. The whole problem of the relationship between homogamy^[54] and fertility is open to investigation.

 

VII. Significance of the Results

The statistical facts reviewed in this essay make it highly probable that a great variety of physical and mental characters influence human matings, and in such a way that, on the average, similar individuals tend to marry.

These results will probably be received with much scepticism. The "charm of disparity," the "selection of opposites," has been so long asserted that the notion will not readily be given up. Concretions of vague impressions compacted into popular superstition are not soon broken up by the hammer of logical deduction from scientific measurement. This scepticism of preconception can, however, be ignored; in time it must give way to orderly arranged facts. Yet the scientist should not forget that when cracked open, the nodules of popular belief are often found to contain a scrap of truth—and in his turn should avoid dogmatism.

Purely biological phenomena are far more complex than the majority of naturalists have realized. When social factors are superimposed the difficulties of research become almost unsurmountable. Pearson has warned us that "in many factors there may actually be two opposed currents, one giving a tendency for like to mate with like and the other marked by the fascination of extremes." Goring's studies of criminals vindicate one's a priori conviction that assortative mating may be influenced by social conditions. Human society differs so profoundly from place to place and time to time that, however great the temptation to generalization may be, it would be folly to press the conclusions far beyond the data which they represent.

Moreover, if the biometric results reviewed in the preceding pages must be fitted post haste into some evolutionary scheme, or find an immediate practical social application, each reader must be responsible for his own. The difficulties of interpretation are even greater than the dangers of generalization. To-day, an unfortunately insistent demand that every datum must count for or against some current theory has largely replaced the Darwinian spirit of collecting facts in the hope that when sound and sufficiently numerous, reasonable theories may be fitted to them. To-day "a fact is not a fact until it fits a theory." Personally, I feel as thoroughly satisfied that the time is not yet ripe for interpretation as I am completely convinced that in the differentiation and painstaking measurement of the intensity of the individual possible factors we have begun to move in the right direction in the attack upon the phalanx of problems which we designate as organic evolution. The immense value of these pioneer studies by Pearson and his associates lies in the fact that they represent the definite and substantial beginning of quantitative research which is so large a part of the solution of a problem.

1. A first essay, "The Measurement of Natural Selection," has appeared in these pages, Popular Science Monthly, Vol. 78, pp. 521-538, 1911.
2. For instance, Ripley ("Races of Europe," p. 49) says: "However strenuously the biologist may deny validity to the element of artificial selection among the lower animals, it certainly plays a large part in influencing sexual choice among primitive men and more subtly among us in civilization. Just as soon as a social group recognizes the possession of certain physical traits peculiar to itself—that is, as soon as it evolves what Giddings has aptly termed a 'consciousness of kind'—its constant endeavor thenceforth is to afford the fullest expression to that ideal."
   Westermarck ("History of Human Marriage," pp. 362-373, 1901) gives a terse summary of the social prejudices, tribal practises and laws concerning marriage between different castes, tribes or elans. Others are to be found in various sociological works.
3. Pearson, K., Phil. Trans. Roy. Soc. Lond., A, Vol. 187, pp. 257-258, 1897; also "Grammar of Science," 2d ed., pp. 421-437, 1900.
4. Thus in man sexual selection is a somewhat more complicated process than it has been assumed to be in the mass of writings on the lower animals.
5. The quantitative data bearing directly on the problem of preferential mating are few. See K. Pearson, Phil. Trails. Roy. Soc. Lond., A, Vol. 187, p. 258, 1897; "Grammar of Science," 2d ed., pp. 425-428, 1900; Biometrika, Vol. 2, pp. 270-272, 1903.
6. The way in which mere impressions may become stamped with authority by the approval of careless writers is illustrated by the following quotation, from a standard authority on sociology.
   "It is almost proverbial that tall men choose short wives, and the union of tall women with short men is only a little less common. Thin men and plump girls fall in love, as do fat men and slender women. Blondes and brunettes rush irresistibly together."
7. '"History of Human Marriage," pp. 353-354, 1901.
8. Fol ("La Resemblance éntre Epoux," Rév. Scientifique, Vol. 47, pp. 4749, 1891) has tried to investigate this by means of photographs of newly married and aged couples, and while he concludes that there is a considerable resemblance between husband and wife, it is no more intense in aged than in newly-married couples.
9. For example, Francis Galton, whose data and methods were not yet adequate for dealing with so complicated a problem, wrote ("Natural Inheritance," p. 85, 1894 ed.) with a caution which led him into error: "Whatever may be the sexual preferences for similarity, or for contrast, I find little indication in the average results obtained from a fairly large number of cases, of any single measurable personal peculiarity, whether it be stature, temper, eye-color, or artistic tastes, in influencing marriage selection to a notable degree. Nor is this extraordinary, for though people may fall in love for trifles, marriage is a serious act, usually determined by the concurrence of numerous motives. Therefore we could hardly expect either shortness or tallness, darkness or lightness in complexion, or any other single quality, to have in the long run a large separate influence."
10. From a paper by F. E. Lutz, "Assortative Mating in Man," Science, N. S., Vol. 22, pp. 249-250, 1905.
11. Since women marry younger than men, relative, not absolute, age must be specified.
12. Such differences as occur are probably due to errors of random sampling.
13. This value is perhaps a little too high. However much popular opinion may overestimate women 's reticence concerning their ages, statisticians know that even the correct age of marriage of both men and women is hard to obtain. Especial difficulty is to be expected near the extremes of the series. Those embarrassed by years may declare themselves of legal age, or even deduct a few years. Those who are not yet of legal age may falsify to obtain a license. Aa Lutz aptly remarks, these figures, "instead of telling the exact truth, show us the state of things modified somewhat by man's idea of how he thinks they had better be."
14. Galton, F., "Natural Inheritance," p. 206.
15. Pearson, K., Phil. Trans. Roy. Soc. Lond., A, Vol. 187, p. 273, 1897; also, "Grammar of Science," 2d ed., pp. 429-431, 1900.
16. Pearson, K., and A. Lee, "On the Laws of Inheritance in Man. I., Inheritance of Physical Characters," Biometrika, Vol. 2, pp. 372-377, 1903.
17. From Mr. Galton's "Family Records," Pearson found the resemblance for man and wife to be +.09 ±.05, but between fathers and mothers of adult children, as in the case of his own material, it was +.18 ±.02. Considering the smallness of Mr. Galton's pioneer series, the results are not in bad agreement.
18. Gallon's series of data ("English Men of Science," pp. 22-24, 1895) for the parents of English scientific men is too small (embracing only 62 cases) to give conclusive results for assortative mating. It indicates, however, no tendency towards contrast.
19. I have omitted Galton's ("English Men of Science," p. 21, 1895) data for the "figure" of 71 pairs of parents of English scientific men, since the numbers are too few to calculate the contingencies on the basis of his grouping, and the classes are too indefinite for satisfactory combination.
20. Length of forearm is the distance from the bony projection of the elbow, with the arm folded as much as possible, to the tip of the middle finger. The span is the greatest possible distance between the tips of the middle fingers of the outstretched hands.
21. By direct correlation we mean the degrees of resemblance calculated for the same organs in husband and wife, by cross correlation those in which the characters are different. The entries in the table makes this quite clear.
22. Boas, F., "Heredity in Head Form," Amer. Anthrop., N. S., Vol. 5, p. 532, 1903.
23. Lutz, F. E., Amer. Nat., Vol. 42, pp. 195-196, 1908.
24. Pearson, K., Phil. Trans. Roy. Soc. Lond., A, Vol. 195, pp. 113, 149-150, 1900; "Grammar of Science," 2d ed., pp. 431-437, 1900; Biometrika, Vol. 5, pp. 475, 1907.
25. Yule (Journ. Anth. Inst., Vol. 36, p. 359, 1906) has drawn these results into question on the supposition that they may be due to a personal equation or bias of the observer, which might lead him to classify both members of a pair alike. His criticism is well answered by Pearson (Biometrika, Vol. 5, p. 475, 1907).
26. Galton, F., "English Men of Science," p. 21, 1895.
27. Bell, A. G., Mem. Nat. Acad. Sci., Vol. 2, pp. 179-262, 1883.
28. Fay, E. A., "Marriages of the Deaf in America," Washington (Volta Bureau), 1898.
29. Schuster (Biometrika, Vol. 4, p. 473, 1906) was unable to calculate the precise intensity of the assortative mating coefficients for deafness because of the mathematical difficulties involved, but it is certainly considerably higher than + .90.
30. In the eases where the mode of education is not known, 77.3 per cent. chose deaf partners.
31. If the probable error of the percentages were calculated, their distinctness would appear much more open to question.
32. See besides Fay's analysis of his data, a paper by Schuster, Biometrika, Vol. 4, pp. 465-482, 1906.
33. Elderton, Ethel M., "Stud. Nat. Det.," 3, London, 1908.
34. Goring, C, "Stud. Nat. Det.," 5, London, 1909.
35. The three chief papers are: Pope, E. Gr., "A Second Study of the Statistics of Pulmonary Tuberculosis: Marital Infection," edited and revised by K. Pearson; with an appendix on Assortative Mating by E. M. Elderton, Draper's Co. Res. Mem., "Stud. Nat. Det.," 3, London, Dulau & Co., 1908. Greenwood, M., "The Problem of Marital Infection in Pulmonary Tuberculosis," Proc. Boy. Soc. Med., Epid. Sect., Vol. 2, pp. 257-268, 1909. Goring, C, "On the Inheritance of the Diathesis of Phthisis and Insanity: A Statistical Study based upon the Family History of 1,500 Criminals," Draper's Co. Res. Mem., "Stud. Nat. Det.," 5, London, Dulau & Co., 1909.
36. That such differentiation exists is strikingly shown by anthropological maps, for instance those in Ripley's "Races of Europe," pp. 300-334, 1900.
37. "Pearson, K., Proc. Roy. Soc. Lond., Vol. 66, pp. 28-32; also Biometrika, Vol. 2, pp. 274-275.
38. "Assortative Mating in Man: A Cooperative Study," Biometrika, Vol. 2, pp. 481-498, 1903.
39. Theoretically, assortative mating should be absent in royalty where marriages are contracted by persons other than those most directly concerned, or are arranged in accordance with some political policy. Woods ("Mental and Moral Heredity in Royalty," pp. 272-273) thinks it can not be held to be entirely absent. He correlated the intellectual grades of 229 couples and found r = + .08, approximately, but with a probable error of ± .076.
40. I have given none of the coefficient for psychical characters calculated by Schuster and Elderton (Biometrika, Vol. 5, pp. 460-469, 1907) from data collected by Heymans and Wiersma. These give results which vary widely among themselves and if one takes those which seem the most likely to be trustworthily determined, he opens himself to the criticism of the selection of evidence. Personally, I have grave doubts concerning the value of data on psychical characters collected by the widespread circulation of schedules. The estimates are too much subject to personal equation and family bias. When they are entrusted to especially trained observers who work comparatively, the ease is better.
41. Galton, F., "Good and Bad Temper in English Families," Fortnightly Rev., July, 1887. Reprinted in "Natural Inheritance."
42. Elderton, Ethel M., Draper's Co. Res. Mem., "Stud. Nat. Det," 3, pp. 30-35, 1908.
43. Galton ("English Men of Science," p. 20, 1895) had only 22 cases where the temperamental characteristics of the parents were marked. He considers that there is a tendency for harmonious matings with respect to temperament.
44. The signs, sometimes difficult to determine in the case of non-measurable characters, seem to be positive throughout, but in some cases there may be distinct cross currents, one tending to produce like and the other to give dissimilar unions. The intensity of resemblance for "success in career" is about double that for other characters, and is possibly to a large extent spurious, because subjective. The "success in career" of a wife is probably largely dependent on or judged by the opportunities which her husband's success gives her for displaying her own abilities.
45. Elderton, Ethel M., loc. cit., p. 35.
46. As a check, Goring determined the correlation between phthisis in one and insanity in the other member of a wedded pair. The resemblance was found to be sensibly zero.
47. Goring, Chas., "Stud. Nat. Det.," 5, p. 27.
48. The reader in noting the high values given for alcoholism by Goring will remember that the individuals studied are the parents of criminals, and that there is a known association between criminality and alcoholism. From data collected by Heymans and Wiersma, Schuster and Elderton (Biometrika, Vol. 5, p. 468, 1908) calculated a correlation for tendency towards drink of + .24 to + .36. But there are several reasons for doubting the trustworthiness of these data.
49. Goring, C., "On the Inheritance of the Diathesis of Phthisis and Insanity: A Statistical Study based upon the Family History of 1,500 Criminals," Drapers Co. Res. Mem., "Stud. Nat. Det.," 5, London, 1909, Dulau & Co.
50. Heron, D., "On the Inheritance of the Sex-ratio," Biometrika, Vol. 5, pp. 79-85, 1907.
51. Fay, E. A., "Marriages of the Deaf in America," pp. 16-18, 29-30.
52. Age at marriage, economic status and other factors probably complicate the problem.
53. Pearson, K., "On the Correlation of Fertility with Homogamy," Proc. Roy. Soc. Lond., Vol. 66, pp. 28-32; also Biometrika, Vol. 2, pp. 373-376.
54. Homogamy means merely the mating of physically or psychically similar individuals. Sameness of stock, endogamy, is of course not implied.