# Popular Science Monthly/Volume 84/February 1914/Current Progress in the Study of Natural Selection

 CURRENT PROGRESS IN THE STUDY OF NATURAL SELECTION
By Dr. J. ARTHUR HARRIS

CARNEGIE INSTITUTION OF WASHINGTON

I. Introductory Remarks

IN papers on "The Measurement of Natural Selection" and "On Assortative Mating in Man," which have appeared in these pages,[1] I have endeavored to show by a review of the quantitative work already done that natural selection and sexual selection are not subjects for idle speculation and polemics, or even for inductions from comparative evidence, but that, like the other factors of organic evolution, they are open to direct quantitative investigation.

It is perhaps not too soon to list up for the convenience of those who desire to take a broad view of evolutionary research, the studies in natural selection which have appeared since the first of these papers was written.

In doing this the ideals of the earlier papers will be carefully maintained. That is, only questions of observed facts and the methods of analyzing them will be taken into account. Theories will be ignored. Again, both positive and negative results will be given impartially, for in the real advancement of science both are of importance in the direction of research and in the formulation of laws.

It will be conducive to clearness to recognize that two fairly distinct problems confront the student of natural selection. The first is to determine whether in any given case the death rate is random or selective. The second is to ascertain what physical, physiological or psychological characteristics make for fitness or unfitness for survival. The attack upon the second problem presupposes the successful solution of the first, for if there be no evidence of the selective nature of the death rate, it is obviously idle to test the selective value of individual characteristics. It is equally clear that any study which stops short of the second of these tasks is in a high degree unsatisfactory. From the standpoint of evolutionary science it is desirable that the significance for survival in various environments of each type of variation in structure or function should be worked out. But this is a task of the highest difficulty and will probably never be accomplished for more than a few selected cases. In these exceedingly difficult fields practicability must be a primary consideration. In many cases, it may be quite out of the question to do more than determine whether the death rate is selective or random, and in such cases these first steps may be of very high importance indeed. Again, it may be feasible to plunge at once into the second problem by testing the value of character after character in the battle for life. In this case both phases are simultaneously taken up. These points will be made clear by illustrations.

II. Further Attempts to Ascertain whether the Death Rate is Selective and to Determine the Intensity of Selection

The first problem to be taken up is therefore that of the existence or non-existence of selective mortality. In a considerable range of living forms it is desirable to know whether natural selection is operative, even though it is for the time being out of the question to say how it is operative, i. e., what particular characteristics make for incapacity or for fitness.

A. The Simple Demonstration of the Existence of a Selective Mortality Studies on Plants.—A first illustration of the importance of the simple determination of the selective or non-selective nature of the death rate is to be seen in the cases of the northward extension of cereals or other cultivated plants. At present, very little is definitely known concerning the factors actually involved. It has been frequently assumed that natural selection through the agency of cold or of the shortness of the growing season has been one of these. This view seems to be supported by Waldron's[2] work on alfalfa. He shows that some strains are more resistant to cold than others, and that in the north the less resistant are eliminated. This is all that is necessary to bring about adaptation—which already exists in some strains. Another most interesting piece of work differing widely in material and detail, but depending upon the same kind of reasoning, is that of Montgomery.[3] Our common cereals have been cultivated for hundreds or thousands of years with practically no attention to selection or grading until recent times. He suggests that under the system of planting two or three times as many seeds as can possibly come to maturity, a slow development has taken place through a continuous natural selection with the survival of the strongest.

He has several interesting results for competition, but his most conclusive experiments for selection are those with maize.[4] Planting continuously at the rate of one, three and five seeds per hill[5] for six years, he finds the yields given in the accompanying diagram, 1. Apparently the seed grown under competition yields higher than that grown under thin planting.[6]

Diagram 1. Yield of Maize from Various Types of Seed.

The Selective Death Rate in Man.—Beyond all doubt the most important work on the question of the existence of a selective death rate has been done on man. This is true not merely from the standpoint of the critical nature of the investigations and the soundness of the conclusions, but from the sociological importance of the findings as well.

The pioneer work of Pearson and Beeton[7] already mentioned in these pages has recently been supplemented by the studies of Ploetz.[8] However conclusive these studies may be, it is most important to have light on this question from another angle. Precisely the information needed should be obtainable in the following manner.

If natural selection be a reality, then (other factors being rendered constant) the survivors of an infantile population subjected to environmental conditions making for a high death rate should in later years show a lower mortality than the survivors of a population subjected to less stringent conditions of life.

The Possible Selective Element in Infant Mortality.—That the preservation of the weaker children may result in a population of adults below the maximum physical fitness is an idea as old as the study of Spartan history. The wide acceptance of the Darwinian theory and the modern reduction in the death rate—accomplished largely by the saving in the early months of life—have combined to give it considerable prominence in recent years.

To solve the problem one must find a series of districts[9] differing as much as possible in the mortality of the early months of life, and determine whether those which have a lower mortality in infancy have a higher proportion of men unfit for military service or a higher adult mortality.

Such attempts have been made, for instance, by Rahts, Kruse, Gruber, Koeppe, Prinzing, Elben and others. The indeterminateness of these studies is apparent not only from the discordant conclusions but also from the obvious inadequacy of the statistical technique.[10]

To Yule[11] and Snow[12] belongs the credit of having first applied the modern statistics to this problem. Yule's data and methods, however valuable they may be from the standpoint of the relationships between the mortality of early and later life in a series of districts, seem quite inadequate to the solution of the problem of the selective or non-selective nature of infantile mortality.[13]

A first great merit of Snow's laborious study is that he fully recognizes the multiplicity of disturbing factors and has attempted in as far as possible to correct for them. His method may be described as follows:

The mortality of children born in any year, say 1903, in as large a series of districts as possible[14] is followed year by year. Working with this series which is homogeneous with respect to year of birth,[15] the problem to determine whether, allowing for environmental influence, the death rate of, e. g., the first year, or of the first two years, has any influence upon that of subsequent periods of life.

Now this factor, which for convenience we have designated as environmental,[16] is of great importance. The death rate differs widely from district to district, and in response to many factors.[17] Thus absolutely districts having a low mortality for the first year of life might have a low mortality for the second to the fifth years of life; districts having a high death rate for the first twelve months of life would also have a high death rate for the thirteenth to the sixtieth month of life, since many of the causes operating in the two cases (bringing about high or low death rates) affect both age periods.

Thus a high mortality of infancy does not necessarily imply a low mortality of childhood or a high military efficiency. This is true because any influence of selection is largely obscured by such factors as ethnic or social composition or physical environment in the various districts. Before one can say anything at all concerning the possible selective nature of the death rate at the beginning of life he must make due allowances for these factors.

Snow attempts to correct for this environmental factor by using the deaths other than those for the infants born in the particular year under consideration as a measure of the stringency of its influence. The precise manner in which this is done need not concern us here, nor is it necessary to explain in detail the various ways in which the mortality of the first years of life was split up into earlier and later periods in order to ascertain what influence, if any, excessive mortality in the earlier period has upon the chances of survival in the later. Indeed, to discuss adequately all of the difficulties encountered and the highly complex methods by which they were largely overcome by Mr. Snow would treble the space which may be devoted to his research, and transform a review intended for the layman into a discussion comprehensible only to the trained statistician.

For present purposes, it is sufficient to say that (correction being made for environment), those districts in which the mortality for the first period was high had in general a low mortality in the second period.[18] Thus in the long run a high mortality in childhood follows a low mortality in infancy; low mortality in childhood follows high mortality in infancy—remembering always the correction for environmental factors which may hide the action of selection.

Natural selection, in the form of a selective death rate, is strongly operative in man in the early years of life.[19]

M. Greenwood and J. W. Bevan, "An Examination of Some Factors Influencing the Rate of Infant Mortality," Jour. Hyg., 12: 5-45, 1912, find some evidence of the selective nature of infantile mortality in the Bavarian data of Groth and Martin, but they justly emphasize the great difficulty of the task and the need of wider and better statistics.

B. The Selective Value of Particular Characters

Whenever possible students of natural selection have plunged at once into the problem of the way in which elimination takes place. When only the normal mortality is found for individuals with a given character, or intensely of the given character, selection is there inoperative; when a higher mortality is demonstrated, selection is tending to weed it out; when a lower mortality exists, natural selection is allowing it to gain ground in the struggle for existence. In the following paragraphs the results secured in this field are set forth.

Seed Weight and Mortality.—Among flowering plants, the highest death rate occurs in the seedling stage, just as among animals the force of natural selection is well nigh spent by the time a given generation reaches maturity.

It seems most important, therefore, to inquire what influence, if any, the characteristics of the seed or of the plant from which it was harvested have upon its viability. Closely correlated with, but quite distinct from, this problem is that of ascertaining what weight the morphological or physiological characteristics of the young seedling has in determining its chances of survival. It is only recently that these promising fields have been entered.

Consider first the visible characteristics of the seed itself. Montgomery[20] in addition to his studies on competition in cereals, has investigated the survival of plants from small or undeveloped as compared with that of large plump seeds of wheat and oats when planted in competition.[21] He finds that when each kind is planted alone a slightly higher percentage of plants is harvested from the large, well developed seeds. Thus there is a considerable difference in the original quality of the seed. When planted in (inter-varietal) competition there is apparently a still further advantage in the large seeds. But it appears to be very slight indeed.

It seems that there are almost as many weaklings susceptible to the effect of competition among the plants from large seed as among those from small seed.

As far as I am aware[22] the only comparable studies have been made on garden beans.

[23] In these experiments the seeds were all carefully examined to make sure that they were perfectly developed[24] and differed only in size.[25] The biometric constants of 28 series which developed into mature plants have been compared with those for the original series of seeds from which the plantings were drawn.[26]

Diagram 2. Differences in Mean Weight of General Population of Seeds and of those which Produce Plants. All differences are reduced to a percentage basis.

The results of these studies can be most conveniently presented graphically. To demonstrate more clearly to the eye the existence of a

differential mortality with respect to seed weight, I have combined in Diagrams 2-3 the data from the field trials already published with those from a series of experiments on germination in sand. The lumping of the two sets of experiments which differ in some slight but apparently significant details to give sufficient series to make smooth graphs is justified by the fact that individually they lead to essentially the same conclusions and that the data and minute comparisons are to be presented in full detail eventually.[27]

Diagram 2 shows the relative differences in type (mean seed weight)[28] between the total samples of seeds weighed and those which produced plants.[29]

Here the heavy vertical bar represents zero difference between the average weight of the total population of seeds and those which actually produce plants. The broken lines and circles to the right show on the scale at the bottom, where each unit represents one tenth per cent, the number and amount of the positive differences, that is to say of those in which the seeds which survived are heavier. The heavy lines and solid dots to the left of the zero bar indicate the number of experiments giving negative differences—i. e., in which the mean weight of the series of seeds which produced plants was less than that of the general population—and the amount of the difference in relative weight.

Judging the areas of light and dark shading, by the eye alone, one would conclude that the surviving seeds are slightly heavier than the population from which they were drawn. But the deviation from the equality of division which would be expected if there were no relationship between the weight of the seed and its capacity for survival is only 4 ± 2.98 cases, and little significance can be attached to it. For the whole 78 series the mean difference is less than five tenths of one per cent.[30]

Thus if we confine our attention to the mean, there is apparently no selective elimination whatever, for within the limits of experimental error there is no certain change in the mean value of the character considered. But an entirely new and different light is thrown upon the whole question when the variabilities are examined. These are distinctly less on the average for the series which develop to maturity.

This is brought out with great clearness by Diagram 3.

Diagram 3. Differences in Variabilities of General Population of Seeds and of those which Produced Plants. The figure to the left shows the ratio of differences in standard deviation to their probable errors (1 space on scale =.2). The figure to the right shows differences in coefficients of variation (1 space on scale =.11 per cent.). The vertical lines give the points of zero difference.

In general form these figures are similar to the one representing the means, but an additional point is brought out by the one for the standard deviations, to the left. In this case the length of the bars indicates not absolute nor relative values of the variabilities, but the trustworthiness of the constants. Here the base scale is in terms of the ratio

${\displaystyle {\frac {\text{Difference}}{\text{Probable Error of Difference}}},}$

each unit being equivalent to .20.

Instead of the light and dark line areas being approximately equal they are widely different. In only 22 cases is the variability of the series of seeds which survive greater than that of the original distribution, while in 56 cases it is less. This is a deviation from equality of 17 ± 2.98, or over five times its probable error. One might have to toss coins a long time to get 22 heads and 56 tails in a series of 78 throws!

Furthermore, the individual constants indicating a lowering of variability by selection are statistically much more trustworthy than those suggesting a decrease. To carry much weight a difference between two constants should be at least two and one half times its probable error. But only 5 of the broken lines reach a straight edge laid 2.5 above (to the right of) the zero bar, while 22 extend beyond the same limit on the other side.

The coefficients of variation are already in relative terms—variability expressed in percentages of the mean—hence they may be represented directly by the direction and length of the lines in the figure to the right. For this graph, one unit on the base scale means one tenth of one per cent. The results amply confirm those secured by the preceding method.

These graphs are deduced from experiments involving tens of thousands of individually weighed seeds. Their evidence for a selective mortality can not, therefore, be lightly set aside. That the average weight remains unchanged while the variability is decreased can only mean that there is an elimination from both the upper and lower extremes of variation, that is, of both large and small seeds.[31]

Nevertheless, too great caution can not be used in the interpretation of the result. Purposely, the materials selected for study were most varied, and while the validity of the general average result can not be seriously questioned, there remains the problem of determining whether, and if so to what extent, the selective mortality may not fluctuate widely with different varieties and conditions.

The whole problem of the underlying physical and chemical causes of the differential mortality remains to be investigated. Finally, the possible relationship of the selective mortality to organic evolution can not be discussed until we have further evidences along several different lines.

Potential Characteristics and Seed Mortality.—When we turn to the question of the possible influence of the characteristics of the adult plant innate but invisible in the seed upon the chances of survival, the data are very scarce indeed.

Montgomery's studies of the rise in productiveness in corn as a result of increased competition may perhaps be of interest here. The most important case is undoubtedly that of single and double stocks. Double stocks are completely sterile, forming neither ovules nor pollen. They must, therefore, be propagated exclusively by seeds from singles. In the effort to place on the market seed which will produce the highest possible proportion of doubles, the closest attention has been given to all factors—shape and color of seed, position of the seed in the pod, position of the pod on the plant, etc.

At the beginning of the last century, the belief was current[32] that a larger proportion of doubles can be obtained from old than from recently harvested seed. Apparently, the original idea was that the transformation took place in the harvested seed, but Goebel suggested that its foundation may lie in a differential viability, the seeds which would have produced singles losing more and more their power of germination as time goes on.

Saunders[33] seems to have put the empirical conclusion and Goebel's interpretation on a scientific basis. In actual experiments which need not be detailed, she found that as the percentage of germination decreased by keeping the seeds for long periods of time, the proportion of doubles increased. She also found that when, through unfavorable conditions, the seeds were of a poor quality and a high percentage failed to germinate, the proportion of doubles was greater. Thus if the pods of 1909 be classified in two groups according to whether they produced less than fifty per cent, or fifty per cent, or more seedlings, we find for two strains:

 Percentage Doubles in: Variety Low Germination High Germination Marine blue 72.0 51.5 Light purple 59.5 52.5

While Miss Saunders results seem fairly conclusive the difficulties of the problem are, as she points out, considerable. A careful experimental investigation on a large scale of the viability of the seed in double stocks and in other ever-sporting varieties would be of the greatest interest.[34]

Seedling Characters and Survival.—Having shown that the measurable characteristics of the apparently normal seed, or the invisible potentialities of its embryo, may be of importance in determining its viability, i. e., that they may be of "selective value," one next inquires whether in the young plantlet some variations tend to be weeded out.

We are indebted to Baur for a neat demonstration.[35] That plants with inadequate photosynthetic apparatus would be incapable of survival seems on a priori grounds quite probable.[36] He found that in the snap dragon, Antirrhinum majus, a variegated or "aurea" form could not be bred true, but on self fertilization gave a progeny of "aurea" and green plants. The inbred green plants produced only green offspring, while the variegated individuals again gave two types of offspring in the ratio of two variegated to one green.

Diagram 4. Mortality in Sixteen Lots of Typical and Atypical Bean Seedlings.

Evidence which need not be detailed here pointed to the conclusion that the variegated was the hybrid form, segregating on self-fertilization according to the simple Mendelian formula into one fourth green, one half variegated (heterozygous) and one fourth lacking chlorophyll—and consequently unfit for survival. Subsequent studies proved that the seeds incapable of forming chlorophyll are actually formed, but that they die in the early stages of germination, or before. Confirmatory results were obtained with a geranium, Pelargonum zonale "verona" where the seedlings lacking chlorophyll died at an early stage.

The preliminary results of a series of investigations on the

structural characteristics of seedlings in their relation to survival are at hand for garden beans, Phaseolus vulgaris.[37] From a lot of about 238,000 seedlings germinated in the greenhouse in sand, somewhat over 9,000 abnormal and normal plantlets were transferred to the field under as nearly as possible identical conditions. The diagram shows most clearly that the death rate, though very low, is unquestionably selective. Solid dots and lines represent the mortality of the atypical individuals in sixteen arbitrary but logical classes.[38] The actual numbers of deaths in these classes is not large. The death rates consequently fluctuate widely. Yet in every case but one the mortality of the atypical is higher than that of the typical individuals. The solid bar gives the rate for all atypical seedlings while the broken line smoothes the circles connected by broken lines in the same way.

Seedlings are thrown only into two classes, typical and atypical. The latter is highly heterogeneous, comprising a very wide range of structural variations. Possibly some of these are more fit for survival than are the normal individuals, while others are far less so. Only the collection of far wider series of data will settle the question. On an average the variations from type are clearly inferior. This is precisely the condition which one would expect if natural selection has been a factor of weight in the development of the structural characteristics of the seedling, for the most fit type would be the one preserved.

It is important to remember that this selective mortality is found in seedlings germinated under as favorable conditions of substratum and temperature as we could give them, and then transplanted to fairly good garden conditions. In nature, a considerable part of the seedling death rate doubtless occurs in the early stages of germination where the nascent root and shoot are subjected to a substratum far less favorable to growth than those of the seed pan. Again, the transplanted seedlings were practically free from the inter-specific and intra-specific competition which must be intense in nature. The detection of a conspicuous selective death rate under such optimum conditions can leave little doubt as to the force of natural selection under the severe conditions in which plants grow in nature.

Pigmentation in Man in Relation to Selection.[39]—The problem of the relationship between the pigmentation of the hair and eyes of the individual and his mental characteristics, his bent towards criminality, his health and his capacity for survival has received the widest discussion.[40]

The relation of pigment to selection has been discussed chiefly from two points of view—that of urban selection[41] and that of susceptibility

As early as 1904 Pearson,[42] working with Pfitzner's data for Lower Elsass, suggested that the high correlation between age and pigmentation in the case of post mortem cases is more nearly explained by a selective death rate of the lighter types than by the assumption of a darkening with age alone. In the same year appeared a most suggestive paper by Strumball,[43] who attempted by the comparison of hospital censuses with the general English population to ascertain whether susceptibility to various diseases is dependent upon the anthropometric characteristics of the individuals affected. He concluded that blond features are associated with acute rheumatism, heart disease, tonsilitis and osteo-arthritis, and that the brunette traits are associated with nervous diseases, tuberculosis and malignant diseases.[44]

MacDonald[45] finds that for scarlet fever, diphtheria, measles and whooping cough among Glasgow school children recuperative power is

associated with hair color and eye color in such a way that the darker classes have the greater recuperative power.[46]

Pearson[47] has also demonstrated correlation of ${\displaystyle r=.19}$ between health and hair color and ${\displaystyle r=.07}$ between health and eye color for data relating to 2,317 boys. Similar results were obtained for girls.

But, on the other hand, there are contradictory evidences. For instance, the conclusion reached by Saunders[48] from his study of pigmentation and susceptibility to diseases in Birmingham school children is that pigmentation is not a factor in selection.

He also finds that relationships between pigmentation and stature and weight, if they exist, are of so delicate a nature that much more refined data than those furnished by the ordinary anthropometric surveys or school medical officer's reports are necessary for their detection.[49]

The discrepancy between his results and those of MacDonald is possibly due to differences in the nature of the populations dealt with. Perhaps, too, data derived from the official examination of school children are less reliable than the hospital returns.

Woodruff has attacked the problem of the relationship of pigmentation to selection from an entirely different, and most important, side.[50] He seeks to determine the relationship between skin color and survival in tropical sunlight. He concludes that the lack of pigmentation is an immense barrier against the penetration of the blond races into the tropics.[51]

Summarizing in a word the results of these studies on man, we may say that the death rate is unquestionably selective. There are still those who assert that while natural selection applies to the lower organisms its force is nil in civilized society. Against such a view the evidence of biometric workers seems fairly conclusive. But concernng the way in which this selective death rate occurs we know lamentably little. Indeed, the whole problem of the basis of natural selection in man is open to investigation. The biometric work which has been done shows how complex the whole problem is, and how idle to attempt its solution by any means but the analysis of large masses of carefully collected data by refined statistical methods.

III. Supplementary Tests of Fitness

The capacity of an individual for survival is doubtless dependent upon the fitness of its several organs for performing their respective functions, or upon the nicety of their coordination. At present, the ultimate goal of investigations of natural selection would seem to be the determination of the significance for survival of each deviation from type of as many organs or characteristics as practicable. Upon the evidences afforded by a comprehensive series of investigations of this kind must depend our final views concerning the significance of natural selection as a factor in organic evolution.

Fitness may be tested in various ways. A series of individuals may be actually subjected to a struggle for existence—be "exposed to risk," to use an actuarial term—and the difference between the series of individuals which survive and those which perish measured in terms of biometric constants. This is essentially the course followed in the studies reviewed in the preceding paragraphs. It is from the standpoint of the evolutionist the most direct method.

Fitness may, however, be tested in some favorable cases in which the individual lays down a series of organs (with measurable characteristics) only a portion of which may develop to maturity. Here one may find that the elimination of organs within the individual is not random, but selective. A comparison of the characteristics of the organs which fail with those which complete their development may furnish information as to the characteristics which make for fitness or unfitness for survival.

Again, physiological criteria—e. g., efficiency in the maturing of ovules into seeds, or in the formation of well-developed seeds—may be found.

Obviously, it will be of great advantage if direct demonstrations of the action of natural selection can be supplemented (or in some cases it may be preceded) by evidences of an entirely different sort.

Such supplementary evidences have so far been sought only in the case of the organization of the plant ovary. Studies of the selective elimination of ovaries have been reviewed in the earlier paper on the measurement of natural selection.[52] Since then considerable side light has been thrown upon the problem of the intra-individual selective elimination of organs by two studies of a purely physiological character.

One of the characters dealt with in studies of the development of the ovary is the "odd" or "even" number of ovules which it produces. This is essentially a criterion of the bilateral asymmetry of the plate of carpellary tissue giving rise to a locule. In large series of pods of garden beans it has been shown[53] that pods with an "odd" number of ovules—that is, those which have the ovules unequally divided between the two carpellary margins, and are consequently bilaterally asymmetrical—are less capable of maturing their ovules into seeds than are those with an "even" number. Again,[54] all the available data indicate that the weight of the seeds is lower in pods with an "odd" than in those with an "even" number of ovules.

The interest of these results is heightened by the fact that the type of structure which in Staphylea shows an inferior capacity for development, in Phaseolus shows (by two different tests) a physiological inefficiency. As soon as proper materials and technique are available it will be of importance to consider asymmetry in its relation to the capacity for survival of the individual.

IV. Concluding Remarks

A summary, properly so called, of the materials of this paper is precluded by the fact that the various sections are in themselves summary reviews of researches carried out upon the most diverse materials. But all these studies have this in common: they are attempts to determine by quantitative methods whether natural selection be a reality, and if so, to measure its intensity. In conclusion, stress may be laid upon two points.

The first of these is a matter of fact. Evidences of the occurrence of natural selection for many characteristics are rapidly accumulating. That mortality is not random, but differential, and that the intensity of the selective death rate is a problem open to quantitative treatment, are propositions supported by large bodies of sound scientific evidence. Nevertheless, neither the complexity of the phenomena nor the difficulties of the collection or of the analysis of the data can be underestimated. As yet, only the surface has been touched. The results are all subject to such revision as may be rendered necessary by wider data and narrower analyses.

The second of these is a question of interpretation. The demonstration of the existence of a selective death rate in a given case is by no means equivalent to proof that evolutionary change is taking place in the character under consideration. Natural selection may only maintain a characteristic at the stage already attained. Or the force of natural selection may be offset by that of some other factor. Or, again, the variations dealt with may be of a kind not inherited; and without inheritance selection is powerless to effect any change. Indeed, first-hand experience in quantitative work on organic evolution must convince any one that the problem of the methods by which it has taken place is far more recondite than biologists have been wont to consider it. This great complexity demands an attitude of extreme caution in generalization. For the present, we must be content to attempt to measure one possible factor after another in as wide a series of organisms as possible. Having done this, we may hope in time to form a fairly trustworthy conception of the resultant of these forces as they may be combined in nature.

Ges., 25: 442–454, 1907. Also, "Die Aurea-Sippe von Antirrhinum majus," Zeitschr. f. Ind. Abst. u. Vererbungsl., 1: 124–125, 1909. Cf. also Bateson, "Mend. Princip. Her.," p. 253, 1909.

1. Pop. Sci. Mo., 78: 521-528, 1911; loc. cit., 80: 476-492, 1912.
2. L. R. Waldron, "Hardiness in Successive Alfalfa Generations" Amer. Nat., 46: 463-469, 1912.
3. E. G. Montgomery, "Competition in Cereals," Bull. Neb. Agr. Exp. Sta., 127, 1912.
4. E. G. Montgomery, loc. cit. Also "Thick and Thin Planting for Growing Seed Corn," Bull. Neb. Agr. Exp. Sta., 112: 28-30, 1909.
5. When planted at the rate of one per hill about 25 good ears weighing 12 ounces or more are produced to every 100 plants. With 3 plants per hill there are only about 10 good ears, with 5 plants, only about five. A plant capable of producing a good ear with four others in the same hill must be unusually vigorous, but in thin planting it is not possible to tell which of the plants would have been capable of reaching the high standard under keen competition.
6. The details given in Montgomery 's paper are entirely too meager for a problem of such great complexity.
7. Pop. Sci. Mo., 78: 533-534, 1911.
8. A. Ploetz, "Lebensdaur der Eltern und Kindersterblichkeit. Ein Beitrag zum Studien der Konstitutionsvererbung und der natürlichen Auslese unter den Menschen," Arch. Bass.-u. Gesellschaftsbiol., 6: 33-43, 1909.
9. Snow (see below) is quite right in insisting that the question as to what proportion of the general death rate is selective should be answered on national mortality statistics. From the point of view of evolution, or of sociology, such data are of far more value than the more complete records which can be secured in individual pedigrees, for to be of evolutionary or of national social importance the intensity of the selective death rate must be measured on a perfectly general population.
10. Examples are given in subsequent footnotes.
11. G. U. Yule, "On the Possible Selective Influence of Mortality in Infancy on Mortality in the Next Four Years of Life," Supplement to the Report of the Medical Officer of the Local Government Board (Great Britain), 1910, Cd. 5,263.
12. E. C. Snow, "The Intensity of Natural Selection in Man," Drapers' Co. Res. Mem. (Univ. Coll., Lond.), Stud. Nat. Det. 7, p. 43, London, Dulan & Co., 1911.
13. This is clear from the criticisms brought forward by Snow. Practically as much has been admitted elsewhere; see Jour. Roy. Stat. Soc, 75: 133-135. In his study Yule shows a caution in interpretation of results which is not as evident in the main body of the medical officer's report.
14. It is very important that the subdivisions be as numerous and as homogeneous within themselves as is consistent with data for trustworthy death rates. For, obviously, the death rate in one district in a given year may be abnormally high (or low) because of purely local and transitory conditions. These are precisely the factors which make for a high or low selective death rate. By lumping a number of districts together one may cancel out the very terms he is seeking to investigate! The value of some of the published work is nullified by the neglect of this point.
15. Obvious as the importance of this point is, it has been entirely overlooked or disregarded by some. To determine that the death rate of children to 1 year of age and that of those 1 to 5 years of age in a series of districts are correlated for a given year, say 1905, proves nothing at all concerning a selective death rate. The infants exposed to conditions (in the various districts) resulting in high and low death rates for their first year of life in 1905 are being compared with those exposed to the action of selection under what may have been widely different conditions in 1904, 1903, 1902 and 1901. The whole factor of the variation in mortality from year to year due to epidemics, meteorological conditions, economic changes, etc., is thus left entirely out of account.
16. Local influence might have been a better term, since racial composition as well as environment may play a part.
17. The fact that the death rate is to so great extent within the control of the sanitary and charity boards is sufficient general evidence for this statement. A quantitative demonstration is seen in the fact that a correlation is found between, e. g., the birth rate and infantile mortality; also between artificial feeding rate and infantile mortality. See Greenwood and Bevan, Jour. Hyg., 12: 5-45, 1912.
18. All of Snow's results are not concordant. There are good reasons for believing that some of the series of public statistics analyzed by him are inadequate for so complex and delicate a biological problem as that of selective mortality. Those series of data which biologically and statistically may be regarded as most suitable and trustworthy evidence the most strongly in favor of the selective nature of infantile mortality. One's confidence in Snow's own interpretation of his results is strengthened by the fact that he has laid all his evidence—that which goes against his own general conclusions as well as that which supports his view—before his reader, believing it to be "more in accord with scientific spirit that the reader should be allowed to draw his own conclusions from the whole of the research, and to form his own opinion on the value of the material used and of the results deduced from it." It is a great pity that such merit should be so distinctive as to require comment, but to-day there is a most unfortunate tendency, among biologists at least, to pigeonhole the contra and publish the pro. Thus current and popular theories are often for a time bolstered up, when if all of the facts were brought forward their standing would be much less secure.
19. The reader who goes thoroughly into these problems will read an editorial criticism of Snow's paper in Jour. Boy. Stat. Soc, 75: 133-135, 1911; also the reply by Snow in Biometrika, 8: 456-460, 1912, where the criticisms seem to be fully met.
20. E. G. Montgomery, "Competition in Cereals," Bull. Neb. Agr. Exp. Sta., 127, 1912.
21. Unfortunately, an intra-varietal competition test for seeds of the two kinds could not be made. The large and small seeds were alternated in the row. To distinguish the two at harvest time it was necessary for them to be of different varieties. Inter-varietal competition probably introduces some factors not present when all the individuals are of the same strain.
22. The literature of seed testing is very large and much attention has been given to the produce of large and small seeds. Practically all the work has been done on too small a scale to be conclusive. Possibly among these writings some records of the viability of seeds of different sizes may be found.
23. J. Arthur Harris, "On Differential Mortality with Respect to Seed Weight Occurring in Field Cultures of Phaseolus vulgaris," Amer. Nat., 46: 212-225, 1912.
24. Unfortunately many students of seed weight in its relation to viability or productiveness have not distinguished between small but perfectly developed seeds and those which are blighted or shriveled and immature. It is not at all unlikely that very different results will be secured from the two sorts.
25. Weighings were made of each seed in units of .025 grams, that is, 0.000-0.025 grams = 1 unit, 0.025-0.050 = 2 units, etc.
26. Since the seeds were taken quite at random any stringent selective mortality will be seen in the differences between the constants of the original bulk of seeds weighed and those of the sub-samples planted which actually developed to maturity. The method might not be adequate for a very low selective death rate. In any case it must be expected to give irregular results. A much more satisfactory method is to draw the comparisons between the constants of the seeds which died and the constants of those which developed. Appropriate data for field culture series are being collected. Large greenhouse cultures in sand in which the comparisons can be made between the seeds actually developing and those failing to develop fully substantiate the conclusions drawn from field tests.
27. J. Arthur Harris, "Supplementary Studies of the Differential Mortality with Respect to Seed Weight in the Germination of Garden Beans." To be published shortly.
28. Several varieties of beans grown under diverse cultural conditions are involved. The varieties with the largest seeds are about three times as heavy as the smallest. To express the differences in absolute weights has its advantages, but when the number of series involved is too large for individual labeling in the graph, it is best to reduce values to a relative (percentage) basis by multiplying the difference by 100 and dividing by the mean for the general population.
29. Here lies one of the objections to combining the two series of experiments. In the field culture the eliminated seeds were those which failed to produce fertile plants. In the sand cultures the fate of a seed could be followed only to germination. Some of the seedlings were abnormal, but to avoid all possibility of criticism every seed which germinated at all was included in the viable class. Doubtless in field cultures some of these would have perished before producing seeds. By retaining all these we are possibly making out a poorer case for differential mortality than we might by considering a part of the abnormal seedlings incapable of survival to maturity under field conditions.
30. The mean difference in weight is more nearly zero in the 28 field experiments than in the 50 made in the greenhouse. There may be valid biological reasons for this, but they can not be discussed here.
31. Montgomery, supra, found little difference in the capacity of large and small seeds for producing mature plants in the cereals. Had he worked with the whole range in size and taken into account variability as well as type he might have found stronger evidence for selective mortality than he did.
32. See Goebel, Pringheim's Jahrb. Wiss. Bot., 17: 285, 1886, for references.
33. E. R. Saunders, "Further Studies on the Inheritance of Doubleness and Other Characters in Stocks," Appendix I., Journ. Gen., 1: 361-367, 1911.
34. 34 Experimental work along the lines suggested by de Vries's discussion, "Species and Varieties," 2d ed., pp. 329-339, would be most important.
35. E. Baur, "Untersuchungen liber die Erblichkeitsverhältniss einer nur in Bastardform lebensfähigen Sippe von Antirrhinum majus," Ber. Deutch. Bot.
36. De Vries ("The Mutation Theory," I., 229–230, 347–353, 1909; "Species and Varieties," 2d ed., pp. 537–538, 553) finds that albida mutants of Œnothera are very weak—exceedingly difficult to raise when appearing in the seed pan, and never found in nature.

Among large numbers of bean seedlings which I have grown in the greenhouse those with white, yellow or variegated primordial leaves have occasionally appeared. It has never been possible to grow these for any considerable time.

37. J. Arthur Harris, "A Simple Demonstration of the Action of Natural Selection," Science, N. S., 36: 714-715, 1912.
38. These comprise about ten "pure lines" each. The fact that the mortality of normals and abnormals tends to rise or fall together has no necessary significance for heredity. It is probably due, largely at least, to the fact that the two kinds of seedlings were under the same environmental conditions.
39. So nearly all the work on pigmentation which falls in the scope of this paper has been done on man that a more general heading seems unnecessary. About the only other case for mammals is that suggested by the anomalous behavior of yellow in breeding experiments with mice. The problem has been discussed by Cuenot, Little, Castle, Morgan, Bateson, Wilson, Durham and others. Apparently, no one has succeeded in finding a mouse pure to yellowness. The suggestion has been made that two gametes both having the determiner for yellow are incapable of uniting in fertilization or that they are not viable if they do unite.
40. To mention even the chief of these papers, the most of which are based on data or methods inadequate for conclusive results, would require too much space.
41. Anthropologists have devoted much attention to the highly complex problem of the difference in pigmentation between urban and the surrounding rural populations. Ripley in his "Races of Europe" gives a good general discussion, to disease.
42. K. Pearson, Biometrika, 3: 464-465.
43. F. C. Strumball, "Physical Characters and Morbid Proclivities," Saint Bartholomew's Hospital Reports, 39: 63-126, 1904.
44. Of course, general suggestions and some statistical work precede Strumball's work. To many of these he refers. More recently, in a discussion on "Heredity and Disease" (Proc. Roy. Soc. Med., 21: 96-98, 1908), he returns to some phases of the question and concludes that the onset of tuberculosis is earlier in blondes, but that the disease is more frequent in dark types.
45. D. MacDonald, "Pigmentation of the Hair and Eyes of Children Suffering from Acute Fevers, its Effect on Susceptibility, Recuperative Power and Race Selection," Biometrika, 8: 13-39, 1911. Here he has brought together a detailed review of the earlier theories and evidences. While his résumé is restricted to writings by those of scientific standing, the diversity of results show that much of what has been authoritatively laid down is nothing more than casual observation and vague suggestion.
46. These conclusions rest solely on the hospital observations. I omit those which involve questions concerning the liability to infection, since they require a knowledge of the distribution of pigmentation in the general population. Such comparisons involve the use of some such basis as the British Association standards or the general anthropometric surveys, which may not be valid for the particular district or social class from which the hospital or asylum inmates are drawn. This was true, for instance, in Strumball's pioneer study, which is highly suggestive rather than conclusive. Again, in attempting to settle the question of differential incidence by an analysis of hospital populations there is the danger of a large personal equation in the appraisal of non-measurable characters. Neither of these difficulties are met when studies of recuperative power are made by a single observer. See also K. Pearson, Biometrika, 8: 39, 1912.
47. Unpublished results quoted by Saunders, Biometrika, 8: 355, 1912.
48. A. M. C. Saunders, "Pigmentation in Relation to Selection and to Anthropometric Characters," Biometrika, 8: 354-369, 1912.
49. Miss Elderton, "On the Relation of Stature and Weight to Pigmentation," Biometrika, 8: 340-353, 1912, concludes from her study of the relationship between hair and eye color and weight and stature: "So far as the material goes we find that types of hair and eye color are not associated to any substantially significant extent with divergencies in height and weight in children between the ages of seven and fourteen, inclusive." It must be noted that Miss Elderton 's problem was in large part undertaken to determine the influence of racial heterogeneity on stature, in its relation to environmental influence.
50. Chas. E. Woodruff, "The Effect of Tropical Light on White Men," 1905. Also, Science, N. S., 31: 620, 1910.
51. One should also read the most interesting chapter on the problem of the white man in the tropics in Ripley's "Races of Europe." There, structural characteristics as well as pigmentation are considered.
52. The soundness of the conclusions of the papers there reviewed has been emphasized by a research which has appeared since that time ("Further Observations on the Selective Elimination of Organs in Staphylea," in Zeitschr. f. Ind. Abst.-u. Vererbungsl., 5: 273-288, 1911). Here it is rendered highly probable that the observed selective mortality of ovaries can not be explained by such simple factors as a correlation between the position of the ovary on the inflorescence, with a heavier but purely random mortality in certain regions of the inflorescence. In another paper on methods ("On the Formation of Condensed Tables when the Number of Possible Combinations is Large," Amer. Nat., 46: 477-486, 1912) evidence is brought forward for an interesting morphogenetic relationship between radial asymmetry of the compound ovary and its locular composition.
53. J. Arthur Harris, "On the Relationship between Bilateral Asymmetry and Fertility and Fecundity," Roux's Archiv. f. Entwichelungsmechanik, 35: 500522, 1912.
54. M J. Arthur Harris, "On the Relationship between the Bilateral Asymmetry of the Unilocular Fruit and the Weight of the Seed which it Produces," Science, N. S., 36: 414-415, 1912.