Oct. 31, 1889] 



NATURE 



643 



•where a mulliplicity of unknown or unanalyzed causes ( " acci- 

 dents," pp. 19, 55) determines some mathematically measurable 

 quality. Eye colour, artistic faculty, and temper, to which Mr. 

 Galton applies his method, are not mathematically measurable 

 qualities, and his results, self-consistent though they are, are 

 scientifically untrustworthy because of variation in the standard 

 •of popular judgment in such matters. Stature, however, comes 

 properly within the range of his method, and Mr. Galton has 

 made the most of this point. 



The method employed by Mr. Galton may be styled a 

 mathematical interpretation of the law of uniformity of nature 

 as apparent in the tendency of progression and regression to an 

 :^average for all new examples of measurable qualities. (By the 

 way, it is curious that Mr. Galton does not employ the term 

 progreision, instead of making regression cover all movements 

 both upward and downward to an average. Rise to an average 

 is certainly progress.) The comparatively easy method of 

 -average in dealing with such problems as heredity in the lump 

 is hardly, what Mr. Galton claims it to be, the pioneer method 

 in the "science of man" (p. 62). Indeed, the method of 

 ■chance is infertile in both biology and .•■ociology so far as it 

 mistakes statements of average tendency as in any wise compar- 

 able in value to the particular predictions given by inductive 

 inquiry. The continued use of such terms as accident and 

 chance to cover a multitude of undetermined particular causes 

 may be directly hurtful so far as this tends to slur over the 

 patient investigation of special cases. The greatest so-called 

 -accidental variations, or "sports," have, of course, their reason 

 in a peculiar conjuncture of influences, and the exact determina- 

 tion of these, specially with reference to stable stocks, would be 

 of the greatest scientific and practical value. The method of 

 chance can never be an aid to progress, for it always fails in 

 particular predictions, and reduces our outlook to the level 

 plain of averages. 



In a book on natural inheritance we should expect some 

 thorough treatment of the relation of heredity to other factors, 

 and a clear exposition of how it can be isolated from them in 

 its effect, especially upon stature. But the statistics employed 

 are avowedly statistics of stature and not of heredity : how, then, 

 are the results made to stand not merely for stature but also for 

 heredity ? Mr. Galton concludes that his results for stature are 

 really laws of heredity because of the peculiar relations of the 

 ratios obtained (p. 132), and also because the results are con- 

 firmed by general deductions from the laws of chance (p. 102). 

 But as to this first point we must regard it as assumed rather 

 than proved that kinship is to be measured by the comparison 

 of ratios of deviations amongst kin. If we knew from other 

 investigations or a priori that the influence of heredity on 

 regression is in the numerical order given, then Mr. Galton's 

 results would be merely confirmatory evidence. The laws of 

 heredity must be based on the facts of heredity, or it must be 

 •clearly shown by the method of elimination that given results 

 <:an only be ascribed to heredity. Does Mr. Galton accomplish 

 this ? He slurs over other influences than heredity {e.g. educa- 

 tion, p. 156), or he hastily concludes them to be in harmony 

 with heredity {e.g. natural selection, p. 119). He also does 

 not satisfy us on the equal influences of parents in heredity (p. 

 98), which is a fundamental assumption for his process. That 

 the average regression of the son to the general average of 

 stature is by one-third parental deviation (p. 104) does not, 

 on the face of it, prove anything with regard to transmission of 

 stature. I cannot see that Mr. Galton has clearly shown this 

 ratio to be more than a law of stature as determined by all in- 

 fluences and not by heredity alone. To make the ratios 

 obtained a "measure of family likeness" (p. 132) is certainly 

 unproved till it is shown that heredity alone enters into the data 

 upon which the ratios are founded. It is plain that in any case, 

 whether the cause be heredity alone or heredity plus many other 

 influences, certain definite ratios will be obtained for father, son, 

 brother, &c. 



As to the way in which an abstract calculation of the laws 

 of chance confirms Mr. Galton's statistics, it is enough to 

 observe that no evidence is adduced why the results attained 

 should not stand for the multiple "accidents" of environment, 

 nourishment, occupation, heredity, &c., rather than to "acci- 

 dents " of heredity alone. Mr. Galton fails to prove that his 

 ratios are not the mathematical expression for the operation of 

 the law of frequency of error as applied to the chance operation 

 of heredity plus other agencies, rather than the formula for 

 heredity simple and unadulterated. 



But stature is undoubtedly modified by many prenatal con- 

 ditions which do not come under the head of heredity, and it is 

 certainly modified by climate, nourishment, and occupation. It 

 is quite likely that human dwarfs might be raised upon the same 

 principles as the Japanese dwarf trees. Mr. Galton makes no 

 deduction from his statistics for other influences than heredity, 

 and his results stand as the expression of the law of frequency 

 of error applied to qualities which are the effect of many com- 

 plex causes beside heredity. HiRAM M. Stanley. 



Lake Forest University, October 5. 



Head Measures at Cambridge. 



I AM pleased to be able to say, with reference to criticisms by 

 your correspondents on the trustworthiness of the head measures 

 at Cambridge, and on the deductions made by myself from the 

 results obtained by Dr. Venn after he had discussed the first 

 batch of them, that he is now about to discuss a second batch. 

 The observations that have since accumulated are about equal 

 in number to those already dealt with, and the new results will 

 afford an efficient check upon the value of those already published. 

 I hope also that Dr. Venn may find adequate material to deter- 

 mine the " probable error" of a single head measure, by means 

 of the differences (discussed under obvious restrictions) between 

 the recorded measures of the same persons at different times. 

 We shall then be better able than we are now to estimate the 

 degree of reliance to be placed on the mean value of any given 

 number of head measures. Francis Galton. 



Trimorpbism in Scabiosa succisa. 



This species is usually described as gynodicecious. Hooker 

 ("Student's Flora") thus refers to it. Darwin ("Forms of 

 Flowers") says, "I have observed the existence of two forms 

 in our endemic S. succisa" ; further, "From what Lecoq says 

 ('Geographic Botanique ') S. succisa appears to occur under 

 two forms in France' ; and again, "According to Lecoq, the 

 female flower-heads of 6". succisa are smaller than those of 

 what he calls the male plants, but which are probably herm- 

 aphrodites." Hermann Miiller ("Fertilization of Flowers") 

 also speaks of S. succisa as existing under two forms in 

 Germany. 



I have recently discovered that the species really exists, in this 

 country at least, under three very distinct forms, viz. (i) the 

 original hermaphrodite ; (2) the small female form described 

 by Lecoq ; and (3) a second female form, larger even than 

 the hermaphrodite, and differing from the first in a very remark- 

 able manner. I will describe the three forms in detail. 



No. I (hermaphrodite). Average measurements of 100 capi- 

 tula : diameter at the base \^ inch, height ^l inch. Average 

 number of florets per capitulum 86 (highest iii, lowest 53). 

 Corolla lavender. Filaments incurved in bud, afterwards erect, 

 and twice as long as the corolla-tube. Anthers pink. Style about 

 ^Ti inch long, thin, remarkably erect, purple, glabrous. Plane 

 of stigma at right angles to the style. Development of style 

 does not take place till anthers have fallen, when stigma becomes 

 viscid. 



No. 2 (straight-styled female). Average measurements of 

 50 capitula : diameter f inch, height f inch. Average number 

 of florets 61 (highest 79, lowest 52). Florets very small. 

 Corolla with a deep lilac tinge. Stamens abortive, filaments 

 very short, within the tube. Rudimentary anthers yellow. 

 Style about | inch long, otherwise precisely as in No. i, but 

 development begins as soon as the floret opens. 



No. 3 (bent-style female). Average measurements of 150 

 capitula : diameter ^ inch, height ^\ inch. Average number of 

 florets 58 (highest 79, lowest 22). Florets very large, and more 

 loosely packed on the receptacle. Corolla blue, with a lavender 

 tinge. Stamens as in No. 2. Style about i\ inch long, very 

 stout, much swollen at the base (? honey-gland), usually white, 

 stigma green. Plane of the stigma much inclined. Styles 

 much bent and twisted. The whole surface of the corolla 

 clothed with long stellate hairs. These are thickest on the face 

 of the limb, which in the other forms is quite glabrous. The 

 style is also thickly covered with similar hairs, which are much 

 crowded immediately below the stigma. To these hairs an 

 immense number of pollen grains may be found adhering. The 

 hairs are not fully developed until the stigma is mature. 



Forms in some respects intermediate between Nos. I and 2, 

 and between Nos. I and 3, are occasionally found, and this 



