Body-build and its inheritance

TABLE OF CONTENTS. — BODY-BUILD AND ITS INHERITANCE

by

Charles Benedict Davenport
Director of Department of Genetics, Carnegie Institution of Washington

PART I. BUILD, ITS DEFINITION AND ITS ONTOGENY. — A. GENERAL CONSIDERATIONS. — Types of Variation in Build. — The Measurement of Build. — Sexual Differences in Build. — Racial Differences in Build. — Geographical Differences in Build.

If a hundred men of about the same stature be compared, it is seen that they vary greatly in weight. At the same time they vary in form, and especially in bulk. This variation is popularly recognized by the variety of terms applied to build. It may be of interest to pause a moment to consider popular terminology relating to build. We have, first, terms expressing a marked deviation below the normal build. We speak of persons as “slender,” “thin,” “gaunt,” “slim,” “slight,” “spare,” “lank,” and “spindling.” These terms are not exactly synonyms. “Lank” implies angularity; “gaunt” connotes the ravage of disease; “thin” connotes a loss of weight; “slight” connotes lightness and smallness of bone; “spindling” is used especially of youth in the period of rapid growth preceding adolescence; “slim” has a faint connotation of insufficiency; “slender” best expresses the idea which we shall want to use in this work where we have a relatively small interest in stature and where we wish to avoid connotation of disease, developmental changes, etc. In other languages there exists a series of terms which similarly differ slightly in connotation. Thus, in French, there is “maigre,” in German “mager,” which often connote a loss of weight through disease, “dünn,” which connotes loss of a former more nearly average weight, and “schlank,” which is close to the English “slender.” On the other hand, the English language contains a variety of terms applicable to deviation in build above the average. Thus we have the words “stout,” “portly,” “fleshy,” “corpulent,” “thick-set,” “obese,” “chubby,” and “fat.” The word “stout” usually carries a connotation of vigor. The term “portly” connotes large size with a tendency to excessively great circumference. “Fleshy” is nearly synonymous with “portly” but has less connotation of majesty of size. “Corpulent” usually carries a connotation of abdominal enlargement. “Thick-set” implies a large bony frame. “Obese” frequently connotes excessive, strictly pathological, increase of build. “Chubby” is applied especially to infants. “Fat” connotes excessive production of fat in the body, as opposed to an unusually large muscular development. Perhaps of all of these terms “fleshy” is as satisfactory as any as an expression for build without connotation in respect to degree or source of great weight, whether due to fat, muscle, or bone. In the German language there are the terms, “plump,” and “schwerfällig,” which serve to express large build. In the French language there are the terms, “gros,” “obese,” and “embonpoint,” which is near to the English equivalent “stout” or “fleshy.” In the present work the term “fleshy” is used, despite its slight suggestion of muscular development merely, largely because it begins with a different letter from slender. The word “slender” will be used for the other extreme of build. It has been found convenient to indicate those terms by their initial letters “S” and “F” respectively.

Our main problem is, in how far does this difference in build between slender and fleshy persons depend on constitutional factors?

Two types of variation in build have to be distinguished: (a) the ontogenetic change in normal build during development with increasing stature, and (b) the change in weight in adults of relatively invariable stature. In type a, stature and other proportions are rapidly changing, but in type b, stature remains constant, and, throughout the race, stature does not differ as much in mature persons as it does from birth to maturity. These two types follow different laws and must be studied by different methods. Consequently they are considered in distinct parts of the present paper.

It is now necessary to consider how build may best be expressed quantitatively. The subject of the best index of build has been much discussed, but without sufficiently differentiating between the two types of variation in build, the ontogenetic and the adult. One of the latest authors to consider the matter is Bardeen (1920, p. 486), who mentions the desirability of recording the volume of the body as a whole, notes its impracticability, and concludes that we may estimate volume from weight. It may, however, be doubted if volume is really involved in the popular notion of build. At least, it is equally probable that the idea of build, as popularly conceived, is a relation of transverse to vertical diameters. When I look at a man, or a photograph of one as in plate 1, and think, “he is slender,” it is because I make a mental comparison of his breadth (of shoulders or chest) with his height and find that his breadth in comparison with that of most men I know of that height is small; or if he is stout the diameter of the chest is large in relation to stature (plate 2). It seems probable that breadth in relation to height gives the best expression of the popular idea of build. By the use of this relation, build can be easily expressed for any age, since chest circumference (which bears a nearly constant relation to chest diameter) has been recorded for many persons of all ages.

Ontogenetic.—Since in so many children and young people the stature and chest circumference have been measured, it is possible to use these data in finding the law of normal ontogenetic changes in build from birth to maturity. This ratio, chest-girth ÷ stature, has thus been used in discussing this law, as more fully described in section B. This relation can be used for tracing the change in build of the same developing child or for tracing the average change of build.

Adult.—In the study of adult changes of build we start with the condition that in the individual the stature is fixed. Consequently, in an individual whose weight is changing, the relation of the build at a years is to that at another period n years later as chest-girth at a years : chest-girth at a + n years. Thus, in the adult period, changes of build are proportional to chest-girth. In different persons, of differing stature, the stature has to be taken into account, and the differences in build are measured by the relation of relative chest-girth as in children. Unfortunately, in our study of heredity in adult build, we usually do not know the chest-girth of the different members of the family, but only their stature and weight. Our problem is, then, to find a relation between chest-girth and weight that will enable us to infer the one from the other. This problem will be further considered in a later section (p. 24).

Sex influences the body so profoundly that we have, first of all, to consider its influence on build, either in early or in adult stages. At birth there is, on the average, a difference between the sexes in weight. The male is about 2.5 per cent heavier than the female (3,310 : 3,230 grams in German children, Daffner, 1902, p. 125, quoting Hecker; 3,606 : 3,485 grams in American children, Benedict and Talbot, 1915). But this does not imply that the boy baby is the chubbier, since the boy baby is longer by about 1.7 per cent than the girl baby. There is no obvious difference in the chubbiness of the sexes at birth. Likewise, in later infancy no obvious difference has been detected, though no thoroughgoing studies have been made on this subject. In childhood and youth children of different sex differ in build on the average, but this is because the form of the ontogenetic curve of build is very different. After maturity and cessation of growth, there is a marked difference in form between the sexes. The female has more subcutaneous fat and appears plumper. If the criterion of weight be applied, the complication arises that the specific gravity of the female seems to be less than that of the male (Bardeen, 1920, p. 488, following Meeh, 1895). As found by Medico-Actuarial Mortality investigations of the Associated Life Insurance Medical Directors of the Actuarial Society of America, 1912, volume I, p. 251: “The difference in weight between men and women of the same height is slight under the age of 20, but above that age young men are distinctly heavier than young women, the difference becoming less marked as they grow older. The tall women are markedly lighter than men of the same height.” The main reason for the greater weight of young men is their relatively much greater chest circumference.

Table 1.—The distribution of frequencies of the various classes of build of offspring derived from the various types of matings.

Summation Table, Males only. Based on Table 11.

Table 2.—The distribution of frequencies of the various classes of build of offspring derived from the various types of matings.

Summation Table, Females only. Based on Table 11a.

Proceeding by a method to be developed later, I have divided adult build into five classes, and counted for the various matings the number of children falling into each. (See tables 1 and 2.) The average index of build, obtained from the individual indices of build, is 2.52 (metric system) or 35.81 ± 0.12 (English system) for males, and 2.43 (34.54 ± 0.13) for females, a difference of only 3.7 per cent in favor of the males, a difference that for our purposes can be neglected. A separation of the sexes in our studies will therefore be, ordinarily, not attempted, and this has the advantage in giving us larger frequencies in our tables.

That there are marked racial differences in build is notorious. The slender Scotchman (plate 1, fig. 3) is in striking contrast with the South Italian, Greek, or Russian Jew (plate 2, fig. 3). The Eskimo are noted for their fleshiness, but Arctic conditions seem to favor large build. (See Davenport and Love, 1921, p. 165.) Martin (1914, p. 248) gives an average body-build of 1.42 (2.3 our system) for South Russian Jews and 1.07 (1.7 our system) for Bushmen. Probably the Nilotic negroes are the slenderest race on earth (Martin, 191, p. 254; for photograph see Martin, p. 263, or Davenport, 1917, p. 347). The racial differences in body-build are so great that, when feasible, race should be taken into account in studies on body-build.

Apart from race, it seems probable that climate influences build. The races that live in the north polar region are of stout build, but this may be a racial trait. On the other hand, the whites who come to live near the pole are heavier than those who live near the equator. This may be due to relative freedom in the subpolar area from certain diseases which reduce weight. There may, however, be a physiological response of the body to the long, cold winters. Whatever the explanation, men from Alaska were found, at mobilization of the United States Army in 1917-18, to have a much higher index of build than those from any other region, i. e., 2.28 (32.41);⁠[1] North Dakota and South Dakota came next, with indices of 2.24 (31.85) and 2.23 (31.73) respectively. These were followed by Montana, Minnesota, and Wisconsin. Contrariwise, the recruits from the Gulf States had a low index of build. Whatever the determining causes, geographical differences in build do exist. Consequently, in studies of build, it is desirable to consider the residences of the persons studied.

B. THE ONTOGENY OF BUILD. — Materials and Methods. — The Ontogenetic Curve of Build. — Later Ontogenetic Changes in Individual Build. — Relation of Juvenile to Adult Build.

From the time of fertilization of the egg it proceeds on its course of development directed by its “hereditary factors,” but modifiable by the conditions offered by the environment. Thus is determined the weight of the baby at birth, and thus the build of the later stages to maturity are directed. It has seemed useful to measure the changes in build from birth to maturity, using the method of measuring build during development described in an earlier section (p. 4).

The materials available for such a study are considerable and may be drawn from a large number of different nationalities. A selection had to be made from this material; there were chosen the best available series from those nationalities most representative of the population of the United States, considering the diversity of its European origin. Two sets of data were utilized: first, collections of measurements made on males from 1 year to 21 years; second, some collections from children of both sexes from birth to 1 year of age.

The first collection of data is gathered from nine sources, listed below. The data used are given in detail in table 3, which shows the number of individuals measured by each observer at each age, the average relative chest-girth at that age, and the average, for each age, of the different averages of the relative chest-girths formed by each observer. These averages are not weighted, since the racial stocks measured by the different observers differed. The average of each observer has therefore to be considered as a unit of nearly equal value with that of any other (table 3).

Table 3.—Average relative chest-girth (chest girth + stature) at different ages from birth to maturity, males.

[According to the observations of 9 sets of observers. The main entries of the table are given in 4 places of decimals (points omitted). Immediately to the left of each main entry is given the number of persons measured to obtain that average. The column on the extreme right gives the (unweighted) average of the averages standing on the same line.]

Owing to the complex nature of the curve of build during the first year of development, special collections of data were made for this portion of the curve (table 4). They are listed as follows:

From the foregoing collections of data, tables 3, 4, and 5 were drawn up, giving for each age the mean relative chest-girth found by each observer. Also, the number of individuals upon which the measurement is based, necessary for weighting the mean. The mean of all measurements was weighted in table 4, but not in table 3 for the reasons stated on page 8.

Table 4.—Average relative chest-girth of infants from birth to 13 months.

[Obtained from 6 series of observations, described on pages 12, 13. The main entries of the table are given in 4 places of decimals (point omitted). Immediately to the left of each main entry is given the number of persons measured to obtain that average. The column at the extreme right gives the weighted averages of the averages standing in the same line.]

It will be observed that the means of build differ, for the constant age, more than do the differences between successive ages. That is to say, the absolute difference between races at any given age is greater than the average differences between one year and the next.

The mean relative chest circumferences for the different ages are plotted in figure 2.

The resulting curve of build from birth to maturity is given in figure 1. This curve shows that build in males is greatest at birth, or at about 1 month, and that it then steadily diminishes to 12 years of age. It then slowly increases to maturity, and indeed to 30 years, after which it increases with extreme slowness, probably to 55 years.

From table 4a and the curve of build, figure 2, we see that the infant at birth has a chest-girth which is somewhat over two-thirds of stature (0.6685). It falls rapidly to one month after birth, and then begins to rise. This minimum is connected with the loss of weight which follows birth, and is due to the difficulty of making adjustments to the new conditions which the organism has to meet. This adjustment is soon thereafter completed and within 2 months chest-girth and weight are not only recovered, but have caught up with increasing length of body. At 6 weeks the body has reached its maximum post-natal chubbiness. Another temporary loss in chubbiness occurs at about 8 months, due perhaps to the cutting of the incisors. The decline which subsequently ensues is prevailingly due to the rapid growth of the legs without corresponding increase in transverse diameter (plates 3 and 4). At about 12 years the boy enters the awkward age when, his legs and arms having grown from 5 to 10 cm. in a single year, he has not yet acquired muscular control of them (plate 7, fig. 2). At the same time his trunk retains childish proportions (plate 5, fig. 12). A little earlier than the boy, the girl enters the “Backfisch” stage of similar slenderness and awkwardness (plate 7, fig. 3, from Stratz, 1922, p. 258). It is commonly believed that this rapid growth of the appendages and their long bones is controlled by the secretions of the pituitary gland. Gradually in the boy, at 14 years, as the gonads begin to ripen, the growth of the legs is retarded and the transverse chest diameter begins to increase rapidly; the lad is quickly transformed into a broad-shouldered, broad-chested, stocky man (plate 3, fig. 3, plate 5, fig. 8) and accordingly the curve of build rises much more in the young man than in the young woman (plate 7, figs. 4 and 5). Consequently the adult female build, as measured by relative chest circumference, is relatively small. At about 19 or 20 years’ growth in stature has usually practically ceased. Meanwhile chest-girth increases slowly by the enlargement of trunk muscles and deposition of fat. Weight is, indeed, stated to increase, on the average, until the age of 55 years (tables 6 and 7).

Table 4a.—Weight, chest-girth, length of body, and relative chest-girth at birth and daily thereafter at 8ʰ30ᵐ for 10 days for 11 babies.

[Data furnished by Bret Ratner, M.D., from Manhattan Maternity Hospital and Dispensary, New York City. Weight in grams; chest-girth and length in millimeters. Relative chest-girth in ten-thousandths. Final average relative chest girth is multiplied by 1.037 to adjust to Table 4.]

It may be well to consider briefly, necessarily somewhat speculatively, the significance of this ontogenetic change in build. First of all, the high relative chest circumference of infancy is due chiefly to the extremely short legs of the infant. Leg-length constitutes only about 40 per cent of stature at birth, whereas it comes to constitute 53 to 55 per cent at maturity. Consequently, the relative leg-length at infancy is only a trifle more than 70 per cent of the adult relative leg-length. Likewise the leg-length of the gorilla is about 70 per cent that of a European man (German, cf. Martin, 1914, p. 308). Consequently the relative leg-length of the infant, at birth, is to that of the adult as the relative leg-length of an anthropoid ape is to the relative leg-length of man. Incidentally, it may be added that the relative length of the arm of the infant is only slightly (8 per cent) less than that of the adult, while in the chimpanzee the relative arm-length is 18 per cent greater than that of man.

We have seen that the relative leg-length of the boy increases to about 12 years and thereafter diminishes to the period of completed growth. The white man has his greatest leg-length shortly before adolescence. We find that among the races of mankind it is the long-legged negro tribes of East Africa, especially the upper Nile (Johnston, 1906. II, p. 932), which are among the lowest races of mankind, and which seem to represent in the adult the physical stage through which the white boy passes at 12 years. Other African tribes have, indeed, relatively shorter legs and thus show a persisting adult stage that is either slightly younger or slightly older than that of the 12-year-old white boy.

Figure 1 (to return to it after this digression) seems thus adequately to measure the varying build of humans from birth to past maturity. It may, indeed, be regarded as composed of two parts: first, that from birth to the cessation of growth in stature, and second, that beyond cessation of growth in stature.

The curve of build thus obtained is, apart from the first period of adjustment that follows birth, a smoothly flowing one, that might indeed be expressed by a formula

y = ax + bx² + cx³ + dx⁴, etc.

The curve of build of figure 1 seems to be new in this form. It is, however, of the same general shape as the curve of varying “Körperfülle” drawn by Martin (1914, p. 246) from Quetelet’s data and based on the relation of weight ÷ stature³. The numbers of children considered in Martin’s figure are inadequate and there is hardly sufficient justification for the use of this formula to express changes of build during ontogeny.

The ontogenetic curve of build may serve as a graphic representation of the stages of development as listed by Stratz (1922). These stages are shown in table 5.

Table 5.—Growth-periods in children.

[Modified from C. W. Stratz, 1922, p. 87.]

Any investigation of the heredity of build is apt to meet with the difficulty that only the present weights and statures of children, parents, and grandparents are known; that the persons are of different ages and hence the data concerning them are not precisely comparable. Even if we exclude from consideration children of the rapid developmental period, under 18 years of age, still we have the problem of individual change of build in mature life. Since stature is practically immutable at this period, change in build may be measured by change in weight. That change in weight does occur is notorious. Figures 3 to 6 illustrate graphically the changes in weight in various individuals from records taken almost at random. In figure 3 the curves of increasing weight run rapidly upward; in figure 4 they go up and then fall again; in figure 6 they are more nearly horizontal.

On the whole, the weight tends to increase with age. This is shown by the tables of the “Medico-Actuarial Mortality Investigation” (vol. 1, pp. 38 and 67). The tables are reproduced in our tables 6 and 7. They show that, on the average, weight tends to increase with age up to 50 or 51 years. The fact that there is, on the average, an increase in build, makes necessary an adjustment in some cases of the youthful indices of children to make them comparable with the adult indices of their parents.

The adjustment was made to the weight of the child by finding in the appropriate column of stature and line of age the expected weight and adding the difference between this weight and the weight in the same column opposite the mid-parental age to the given weight of the child.

Table 6.—Graded average weight of men of different statures at various ages.

[Copied from Table IV of “Medico-actuarial mortality investigations.”]

Table 7.—-Graded average weight of women of different statures at various ages.

[Copied from Table IX of “Medico-actuarial mortality investigations.”]

This adjustment was not always necessary, first, because in not a few cases the children were of an adult age and the data for the build of the parents, now dead, were given for the same age; second, examination of such developmental curves as are shown in figure 6 strongly indicates that when the parents are of slender build, and the children are of slender build at the age of 25 years younger than the parents, the increase in weight of the children will not usually follow the actuarial tables. The actuarial tables, indeed, represent only the average increase, and this average is made of a population that does not increase its weight at all when it grows older (fig. 6) and a population that increases its weight at a much faster rate than indicated in the actuarial tables (fig. 3). Consequently, no adjustment is made to the given weights of the children whose parents are slender. In the case where one parent was slender and one fat, the adjustment is still made in the build of all the children. This procedure, undoubtedly, introduces an error which, so far as I see, can not be avoided. Baldwin (1921, p. 74) states that “as a general rule heavy children remain relatively heavy during the period studied”; and his developmental curves of individual children show that the same persistence of the juvenile build is commonly true for children of slight weight. It is striking how often the children of heavy parents will be heavy even in youth, and conversely, children of exceptionally great weight are apt to retain, and add to, their build.

Chambers (1850, pp. 139-143) cites the following cases of early familial obesity: A boy of 3 years weighs 39.5 kg. (87 pounds); his 6 sisters and brothers are obese. A girl of 5 years 5 months weighs 89 kg. (196 pounds); obesity on both sides of the house. A boy of 16 weighs 114 kg. (252 pounds); there is collateral heredity.

A striking case of early obesity persisting to maturity is that of Miss Allen (plate 8).

On the other hand, it is often strikingly true that in families with a tendency to fleshy build some of the children will remain slender until 20 or 25 years and then begin to grow fat. The metabolic changes that induce fatness first appear in later life.

PART II. MASS STUDIES IN HEREDITY OF ADULT BUILD. — METHODS AND MATERIALS. — THE ADULT INDEX OF BUILD. — CLASSIFICATION OF BUILD. — SIGNIFICANCE OF VARIATIONS IN BUILD. — DISEASES IN RELATION TO BUILD. — MASS STUDY OF VARIATION AND HEREDITY IN BUILD. — REGRESSION OF PROGENY TOWARD MEDIOCRITY. — HYPOTHESIS. — MATE SELECTION IN BUILD. — THE BASAL TABLES.

It is a matter of common observation that in some families the parents and children are all slender; in others, there may be many examples of obesity. Worthington (1877, p. 50) cites a number of examples from C. Bouchard. A woman of 45 years weighs 107 kg. (236 pounds); her obesity began shortly after marriage; her father is very obese and her mother obese. A woman of 49 years, whose father is a Turk and whose mother is French, weighs 117 kg. or about 258 pounds; her mother was obese. A woman of 115 kg. or about 250 pounds has an obese mother and two sisters who were obese in infancy; also a gouty mother’s father and father’s father.

The following, from Chambers (1850), show obesity “on both sides of the house”: Male of 28 years, 120 kg. (266 pounds); woman of 48 years, 127 kg. (280 pounds); woman of 52 years, 98.4 kg. (217 pounds); man of 57 years, 227 kg. (500 pounds); woman of 58 years, 104 kg. (231 pounds); woman of 68 years, 118 kg. (260 pounds); woman of 70 years, 107 kg. (238 pounds). In many other cases cited by Chambers, one parent of the obese patient was obese. Howard (1908, p. 54) cites the case of a 7-year-old girl, 45.5 inches (115.6 cm.) tall, who weighed 40 kg. (88 pounds), had a pendulous abdomen, and was feeble-minded. Her sibs were not abnormal and her parents were of average build. One of her great uncles weighs 127 kg. (280 pounds), an uncle, at 40 years, about 109 kg., and an aunt of 31 years, 95 kg. (210 pounds). This case is instructive because of the skipping of a generation.

In the class of obese cases known as adiposis dolorosa, heredity is usually obvious. Price (1909) cites a case of an obese woman of 48 years and weighing 140 kg. (310 pounds) who belongs to a fraternity of 7; 1 was a miscarriage, 2 died young of accident, 1 died at 22 of typho-pneumonia, 1 died young of scarlet fever, and 1 brother is large and rheumatic. The father seems to have been of average build and the mother is stated to have been “very thin.” Of her sibs, 6 were fleshy or very fleshy, 1 medium, and 1 slender; the children of these fleshy sibs of the mother are “all stout.”

Lyon (1910, p. 68) discusses heredity in adiposis dolorosa and lipomatosis and cites a considerable number of cases of family recurrence in his cases and others. Thus he twice treated a father and his son for multiple fatty tumors; also twice a mother and daughter. Lyon’s obese case No. 5 was like her 3 sisters and 1 daughter; a son of her father’s brother showed similar fatty deposits. 10 other instances of family recurrence of abnormal fat deposit are cited.

Maranon and Bonilla (1920) cite the case of a girl of 18 years who was slender, like her parents, until after an attack of syphilis, when she came to weigh 157 kg. or 350 pounds, while her height was 160 cm., her chest-girth 130 cm., and that of her abdomen 150 cm. or 90 per cent of her height. She had a very large brother, and both mother’s parents were obese, though the parents were not known to be so.

Such examples might be multiplied indefinitely.

Our problem is not what are all the causes of this diversity of build, but rather in how far do genetical factors play a part in this diversity. We are not oblivious to the fact that there are many factors responsible for the result—deviation from the average build. These we shall consider in detail in a later section, and the consideration will help us to see the limits to the action of the genetical factors. Before going on to that, we shall have to consider more in detail the nature of the facts for which an explanation has to be sought.

The method of analyzing the genetic factors in build is that of tabulating the distribution of aberrant builds in the family network. There is required, first, a large mass of family data which includes many extreme or aberrant types of build, and which is as reliable and as accurately quantitative as possible; secondly, this has to be subjected to the ordinary methods of genetic analysis.

The available material has consisted of data on stature and weight given in the Records of Family Traits which constitute a fair sample of the population; and of quantitative data on special schedules giving stature and weight of a fraternity, its parents, uncles and aunts, and grandparents. These special schedules had been mailed to an address list of overweight and underweight persons obtained through the kind cooperation of Mr. Arthur Hunter. Those who returned the schedules showed an especial appreciation of the requirements of our study. A third source was the A file of the Eugenics Record Office, where are gathered miscellaneous pedigrees of families showing aberrancy in build. A fourth and especially valuable source was the field work of Miss Louise A. Nelson, of the Eugenics Record Office; this started with selected, usually obese, cases.

After the data had been assembled and tabulated, a certain amount of correspondence and personal visitation was undertaken in order to secure a confirmation or revision of the records in hand. In some cases this brought to light errors in the records, in others, useful details. Naturally, it was not possible to secure a revision of all of the data used, but an attempt was made to select only records that had been compiled with care and conscientiousness, and these traits in the compiler reveal themselves pretty clearly to a person who has examined thousands of these records, just as carelessness is revealed also by slipshod speech or posture.

For our study we desire the data of stature and weight for children, parents, and grandparents. With some exceptions only those families are studied in which all these data are accurately given. Also, only children who are above 18 years of age can be utilized, because stature changes so rapidly until that age. However, since it is build and not stature we are studying, the fact of increase of stature from 19 to 21 years of age affects the result very little. Finally, in a certain proportion of the cases the stature and weight of all of the grandparents are not given quantitatively. Such families are utilized, nevertheless, with such quantitative data as may have been afforded.

The data were taken from the Records of Family Traits by Miss Miriam Kortright, who long assisted in our statistical work. The computations of index of build were made by Miss Kortright and Mr. William Kraus, Miss Laura Craytor, and Miss Margaret Andrus, who checked one another’s work. The tabulation and seriations of the indices were done by Misses Margaret Babcock and Katharine Belzer.

In an earlier section of this paper the question of the best index of build has been discussed generally. It was pointed out that many regard it as a truism that build is a relation of volume to stature. Since the volume of a person’s body is rarely known, and it is difficult to determine it, weight has been substituted for volume. However, this substitution assumes that specific gravity is the same for slender and for fat persons; but this is not at all the case. The specific gravity of a fat person is about that of water (0.978 to 1.079 in 4 children 7 to 13 years of age, Meeh, 1879, and 1.014 in a 61-year-old man of 98 kg. weight, Mies, 1899); of a thin person it may be 5 to 8 per cent above that of water (1.049 to 1.082 for thin convicts, Mies, 1899). This variable specific gravity complicates the attempt to infer volume from weight. In view of these difficulties it were better to measure build by a relation of chest diameter (or circumference) to stature. But this ratio can not be used in our studies, since our data, for the most part, give only weight and stature and not chest-girth. It remains thus to determine the closest relation between weight and chest-girth. This determination I have attempted to make for 100 young men, 20 to 25 years of age, measured at Harvard University where they were students. It will hardly be worth while to reproduce the detailed tables of measurements and ratios, but they will be found summarized in table 8. In this table is given the frequency of occurrence of each of the different ratios (or rather classes of ratios) found using chest-girth in first, second, and third powers as a divisor. If weight varied exactly with the chest-girth, then the ratio of the former to the latter should remain constant. Such a strict relation is hardly to be expected and, of course, is not found. The ratios obtained show a certain variability about the mean condition, and this variability is measured by the standard deviation. Similarly, if each weight be divided in turn by the second and third powers of stature, and the corresponding variability of the ratios be considered, we shall have a method of deciding whether weight varies more closely with the first, second, or third power of chest-girth, and which of those powers gives in its fluctuations the best measure of the corresponding fluctuations in weight.

Table 8.—Variability of weight in relation to the second and third powers of chest-girth as found in 100 Harvard students.

[f, frequency of occurrence of the given index-class.]

A comparison of the standard deviations gives the following results for man:

From these results the conclusion is drawn that since the variability (standard deviation) of weight ÷ chest-girth² is least, the square of the chest-girth varies more closely with weight than either the first or third power of chest-girth; consequently the square of chest-girth is the best measure of weight of the three.

By hypothesis, the chest-girth in persons of the same build varies very closely or exactly with stature; consequently we could substitute in the foregoing ratios for chest its average equivalent, stature/K, in which K is nearly 2, more precisely 1.9. In any case it is thus clearly deducible that a better index of build is got by dividing weight by the square of stature than by its cube, as has been so often done. Accordingly, the ratio of weight to stature² has been adopted in this paper as the standard index of build. The correlation between this index of build and relative chest-girth is found by calculation to be about 0.45.

In any scale of index of build it is, of course, desirable to use the metric system. Unfortunately, most of our data are in English units, so that our indices were first obtained by the use of these units. We have in many cases transmuted the English into the equivalent metric measures. We have, however, retained the original English index, since a large portion of the more cultured part of the world uses that system in daily life. A table to facilitate transmutation is also given in the Appendix, table XVIII. To facilitate the determination of the index of build when stature and weight (in English or metric units) are known, table XVI has been prepared (pages 169, 170).

To avoid decimals, the ratio, weight in pounds ÷ (stature in inches)² is multiplied in this book by 1,000; this gives a series of ratios running from 20 to 60 and over. To avoid confusion with the English system, the metric equivalents are taken as the ratio of weight in grams ÷ (stature in centimeters)². This gives a series of index numbers of the order 1.5 to 4.0; in this case, at least, one decimal is always expressed. The small integral figure and the decimal at once indicate that the index is from metric units. Since the index of build has often been expressed as the ratio of weight to, respectively, stature, stature²⸳⁵, and stature³, table XVII has been prepared to permit these ratios to be transmuted into weight ÷ stature², English system.

For the purposes of analysis, it was found necessary to make a small number of classes of build. To decide upon the limits of these classes, a polygon of frequency of all indices of build was made, as shown in figure 7. It appeared plain at the outset that it is desirable to plot the data in this polygon by using as abscissæ the logarithms of the index of build rather than the absolute indices, since the range of weight above the mode is, for obvious reasons, very much greater than below the mode. Taking mean weight at 68 kg., or 150 pounds, the minimum weight is about 20 kg. (45 pounds), or 25 kg. (55 pounds) below the mean, and the maximum weight is about 150 kg. (330 pounds), or 182 kg. (400 pounds) above the mean. That is, the range of weight classes is three times as great above as below the mean. Plotting data in logarithmic fashion, as shown in figure 7, it appears that the modal index of build is 2.3 (33). The range is from 1.4 (20) to 4.5 (64). Using the logarithms of abscissæ, the curve is more nearly a symmetrical one. It is more irregular above than below the mode, because the classes are more numerous and the frequency of each class smaller. The presence of two modes is suggestive of the hypothesis that the medium class and probably the fleshy classes are not strictly homogeneous, but, on the contrary, comprise groups of individuals whose build is due to dissimilar factors, or sets of factors.

To derive the desired classes from figure 7, the polygon was somewhat arbitrarily divided into five parts, as indicated. Taking 33.5 as a starting-point, an equal logarithmic distance was laid off, above and below this point, on the base-line. This was taken as the range of middle class. An equal logarithmic range was accorded the classes next above and below the median respectively. All of the remainders were thrown into the extreme classes to which are given, thus, a somewhat greater range than the interior classes. This seemed desirable, since their frequencies were so low. The adjusted classes finally adopted are as shown in table 9.

Table 9.—The five standard classes of build; limits and middle points of each.

What is the meaning of the great variations of build described in the preceding paragraph? What is known in this matter may here be briefly summarized that it may be held in mind in considering the numerous cases to which we shall have occasion to refer.

In general, it may be stated that variations in build are due to endogenous causes and exogenous causes. In this book we shall have occasion to examine especially the former—the constitutional or hereditary factors. These include idiosyncrasies of metabolism, in part controlled by peculiarities in the functioning of the endocrine glands; in part, probably, by even finer protoplasmic differences. Thus it is known that the thyroid gland greatly influences metabolism; its activity in growing children tends to produce tall and slender form. On the other hand, deficiency in its activity in childhood leads to the type of obesity known as cretinism, and in middle life to myxedema. The secretions of the pituitary gland cooperate with the thyroid in stimulating growth, especially in the preadolescent stage. When pituitary secretions are deficient there frequently results, it is believed, the adiposogenital syndrome, in which great masses of fat are deposited on breasts, abdomen, hips, and buttocks, and the gonads remain infantile. A case that quite certainly belongs to this category is shown by Beck (1922, p. 881) and reproduced in plate 8, figures 3, 4, 5; 3 of this man’s 4 sibs are fleshy and have scant beards. This is quite like our standard very fleshy case (plate 2, fig. 5). See, also, the extreme cases falling under this category described by Lyon, 1910.

Lesions of the pineal gland (Beck, 1922, p. 909) and of the gonads are stated in some cases to induce obesity. Certain it is that, on the other hand, the activity of the gonads tends to slow up growth in stature and to increase the chest circumference (plate 6), and this change is more marked in the male than the female.

The exogenous causes of build are striking, so much so that many physiologists seem to accept the hypothesis that they are of sole importance, that excess of fat is due merely to excess of calories ingested over those concerned in bodily activity. While no one will deny the possibility of starving most fleshy persons thin or of increasing the weight of most adults by an excess of food, yet it is also obvious that two persons of the same stature and fed equal amounts of similar food may come to differ enormously, due to internal conditions, sometimes of glandular origin and sometimes dependent upon, or at least associated with, disease.

Table 10.—Incidence of disease in relation to build.

As just stated, it is frequently true that build is influenced permanently by disease. To test the influence of disease on, or association of diseases with, different types of build, a tabulation was made of the diseases recorded (in the Records of Family Traits, Eugenics Record Office) as occurring during youth and during middle age in 10,000 fairly well described persons. This constituted the control. Then there was determined for our groups of very slender, slender, fleshy, and very fleshy, the incidence of disease. The ratio of the percentage incidence of the latter to the former was then calculated. In table 10 is given in sum many of the results found for the principal diseases. This table may now be briefly discussed.

Persons of very slender build are characterized in youth by an excess of respiratory diseases—influenza, tuberculosis, and colds. In middle age they show an excess of melancholia, nervousness, and tuberculosis.

In 737 persons of slender build there are found in youth many diseases in excess of normal incidence. These comprise diseases of the respiratory tract—tuberculosis, tonsillitis, bronchitis, pneumonia; some nervous diseases, “nervous breakdown”; various general infections, such as “fevers,” appendicitis, anemia, “rheumatism,” diphtheria, scarlet and typhoid fevers. One might conclude that slender youth are relatively nonresistant to infections. In slender persons there is found in middle age an excess of tuberculosis and pneumonia, much appendicitis and intestinal trouble, and (as also in youth) anemia.

These associations of slender build and disease are not always easy to interpret. The common idea of the tubercular diathesis comprises slender form. On the other hand, a person who has, or has recovered from, active pulmonary tuberculosis is apt to remain underweight, partly because the respiratory apparatus is damaged. The association of “nervousness” with slenderness is probably due to the double effect of some glandular dystrophy, as, for example, of the thyroid gland. Hyperthyroid individuals are usually tall, slender, and “nervous” or irritable.

In 103 persons of very fleshy build, the only outstanding disease of youth is pneumonia. In middle age occur “kidney trouble,” “dropsy” (which often accompanies chronic nephritis), and apoplexy, which is sometimes caused by extra pressure on the blood-vessels resulting from impeded elimination from the kidney or to a diabetic tendency which puts extra work on the vessels. “Heart disease” is also commoner than usual.

In 542 persons of fleshy build, “bladder trouble” (probably including diabetes) and kidney trouble are exceptionally frequent, also arterio-sclerosis and its accompaniments, apoplexy and paralysis. Hernia is frequent, as are various diseases of the digestive tract, such as appendicitis, hemorrhoids, liver trouble, and gallstones. These are doubtless not the cause of, but a consequence or concomitant of, overweight. Sibilant bronchitis, lithiasis (uric and biliare), and diabetes mellitus are mentioned, in addition to the above, by Heckel (1920, p. 31) as especially apt to be associated with obesity.

Table 11.—Distribution of progeny of the different matings according to index of build; males and females tabulated separately.

[From Appendix tables, omitting starred families.]

Table 11a.—Distribution of progeny of the different matings according to index of build; males and females tabulated separately—Continued.

[From Appendix tables, omitting starred families.]

The one clear conclusion from this study is that no single disease and no special single collection of diseases is exclusively responsible for exceptionally slender or exceptionally fleshy build. The variations in build are due primarily rather to various idiosyncrasies of development and metabolism which have largely an hereditary basis, upon which may be superimposed modifications by various types of disease.

Table 12.—Distribution of progeny of the different matings, according to index of build; sexes tabulated together, based on Appendix tables, omitting starred families.

Having considered the classification and something of the causes of variation in build, we have now to consider the relation between the build of the parents and that of the progeny. This is the mass treatment of the data of “heredity” which was the prevailing method 25 years ago and earlier. It is still a useful method in the case of traits due to multiple factors, such as the present one.

The distribution of build in the children of the different matings is given in tables 11, 11a, and 12. There are 15 possible different combinations of matings of the five grades. The first of these (VS × VS) is not represented in our data, and the fifth, VS × VF, is represented by only one mating and no column is devoted to it. The frequencies are given separately for male and female offspring (table 11), and again for both sexes together (table 12). Table 13 shows that there is a considerable correlation between the average build of the parentage and that of the progeny. From the matings of the fleshier parents the progeny are fleshier; from those of slender parents, slenderer. This relation may conceivably be due to family tradition handed down from parents to children. We shall see later that this hypothesis meets with formidable difficulties to acceptance. The most reasonable hypothesis is that there are, above all, hereditary family tendencies that help determine build.

Table 13.—Distribution of progeny of the various matings, according to classes of build, absolute numbers, and proportions, based on Appendix table, including starred families.

Comparing the tables for male and female offspring, it appears, first, that there are, for some reason, more males than females about whom data of build are given, probably because more males than females know their stature and weight, or willingly record it; second, there are relatively more females than males of very slender build (grades 22 to 31); there are recorded relatively more very fleshy males than females (grades of 50 and above); third, there are relatively more recorded daughters than sons derived from one very slender parent, and from the F × F and M × M matings. The male progeny are more variable than the female as 5.325 ± 0.084 is to 5.133 ± 0.089; but the difference is less than three times the probable error, and is, consequently, not very significant.

Considering next the table of total progeny of the various matings, it appears that the average number of children with recorded build from the recorded matings is variable. In descending order the fecundity of the matings is shown in table 14. This table shows that larger families, on the average, were derived from fleshy parents than from slender parents. Thus the F × F matings yield 2.3 times as many children, on the average, per mating as the S × S matings.

Table 14.—Average number of progeny yielded by each type of mating (based on table 12).

Galton pointed out, in the case of stature, that, since correlation between parents and progeny is not perfect, the progeny of selected parents will tend to be less extremely selected and hence more nearly mediocre than their parents. It has, indeed, been shown in my studies on stature (1917, p. 341) that the progeny of tall parents do not show this regression to mediocrity as much as the progeny of short parents. This was regarded as evidence that the gametes of tall parents carried fewer recessive allelomorphs than those of short parents; hence were genetically “purer” and comprise more recessive factors. What is the condition in respect to the varying indices of build?

The answer to this question is given in table 15, which in turn is based on table 12. This table shows for each of the 13 matings the average departure of the parents from mediocre build (which for the parents is 34.86) and the corresponding departure of their offspring from mediocre build (which for the progeny is 35.24). In the right-hand column of the table is given the difference between these two departures, which measures the amount of regression toward mediocrity on the part of the progeny. The results of the last column are shown graphically in figure 8.

Table 15.—Average build and regression from parental average of the progeny of the various types of mating. Also matings arranged in order of regression. Sexes combined (based on table 12).

Matings Arranged in Order of Regression.

Figure 8 shows clearly that, in spite of considerable irregularities, the line of regression descends from the matings of two very fleshy parents at the left, and in general from matings in which the average parental departure from the mean build of parents is positive, to the mating of two slender parents (or, less strikingly the VS × S mating) or in general to the matings in which the average parental departure is extremely negative. This result is most easily explained on the ground that whereas fleshy parents carry all sorts of gametes for build, slender parents carry a preponderance of gametes of their own kind; hence the progeny do not regress so much from the selected parental condition. This suggests that the slender parents are more nearly homozygous than the fleshy parents.

Still another test of the gametic composition of the parents is the variability of their offspring. The facts regarding such variability are given in table 16. From this table it appears that the mating that yields the least variable progeny is that of two slender consorts. The variability in their progeny is measured by 2.41 ± 0.17. The variability of the progeny of the VS × S mating is greater, 4.21 ± 0.61, but on account of the small number of the progeny the probable error is large, and it is possible that this difference in variability between S × S and VS × S progeny is not a significant one. Next to the least variable are the offspring of the M × M mating, 4.06 ± 0.11, and this leads to the conclusion that a large proportion of the M parents are not merely heterozygous, but constitute a “pure race” of medium build. The offspring of the S × F mating have a fairly small variability 4.13 ± 0.16, as befits a first generation (F₁) hybrid. On the other extreme, we have the M × VF mating with a standard deviation of 9.11 ± 0.41. This large standard deviation is due chiefly to the inclusion of one family (Ber-A) which contains 2 progeny of builds 79 and 103, weighing 180 kg. (400 pounds) and 215 kg. (475 pounds) respectively. Otherwise, the variability of this mating is not extreme. It is 5.25 ± 0.24. The next largest variability is from the VF × VF mating, 6.31 ± 0.55, a variability that is due to the absence of any important mode. The progeny of the VS × F mating are highly variable, 6.18 ± 0.59, but this standard deviation has the largest probable error of any except VS × S, so that great stress must not be laid upon its exact position. In general, the progeny of matings with 2 or 1 F or VF parents, belong to the more variable group and those with S (or VS) parents to the less variable group. The meaning of this is clear to the geneticist who has dealt with multiple factors. It indicates that some or all of the factors that make for fleshy build dominate to a greater or less degree over the factors for slenderness. The test of the regression of progeny toward mediocrity and the test of the variability of the progeny of the various matings thus lead to the same result—the factors for fleshiness are imperfectly dominant over those for slenderness, and the latter probably lack some or all of those factors that make for fleshy build.

Table 16.—Progeny of the various types of matings arranged in order of variability or standard deviation (S. D.), together with the probable errors (P. E.) of the means and deviations; also the coefficients of variation (based on table 12).

The foregoing brief studies of the progeny of classes of matings suggest the following hypothesis:

Fleshy build results from the action of several positive (dominant) factors that make for stoutness, while slenderness results from the absence of one or more of such factors, or is due to recessive factors. Fleshy parents may, and frequently do, carry gametes which lack the “fleshy” or carry the “slender” factor, while in slender parents for the most part the gametes carry only the slender factor, hence the gametes of slender parents are more nearly homogeneous. This hypothesis may be further developed as follows:

Assuming that there are two independent factors A and B for build, then these may be found in different zygotes in the following combinations: