More on Pearson and Eugenics
This is a follow-up to the previous post about Karl Pearson’s endorsement of Eugenics and Herr Hitler. The first paper published in the Annals of Eugenics was by Pearson and other, entitled “The Problem of Alien Immigration into Great Britain, illustrated by an Examination of Russian and Polish Jewish Children.” Below is a review of this article and of the journal itself by E.A. Hooton (kindly provided to me by Murray Jorgensen). It goes without saying I hope that my posting this material is not an endorsement of it. Indeed, from the distance of 82 years and viewing the journal through an anthropological prism, Karl Pearson himself appears to be an alien species.
E.A. Hootons’ review: Every anthropologist will welcome the issue of a journal devoted to the study of racial problems when the editors are Professor Karl Pearson and Ethel M. Elderton. Such a journal will undoubtedly take its place beside Biometrika as an indispensable source of information and inspiration for all workers who endeavor to apply exact mathematical methods of treatment to anthropological data. (CL: Now there’s a tough hard-hitting opening paragraph if ever I read one).
The first number of Annals of Eugenics begins with a trenchant editorial explaining the need for such a journal and outlining its purposes. Existing journals and text-books which deal with eugenics are dismissed with contempt. The future of the science of eugenics is stated rather than predicted. The science of Eugenics is in fact only highly developed and applied anthropology, and the day will inevitably come when every university of standing will have its professor and laboratory of Eugenics.
The attitude of the editors upon the subject of race differences may be gathered from the following excerpt:
No more than there is equality between man and man of the same nation is there equality between race and race . . . .Many races have hardly yet found their true place and function in the community of nations. Science will not flinch from the conclusion, if such be inevitable, that some of these races scarce serve in the modern world any other purpose than to provide material for the history of man.
About one-half of the present issue of Annals of Eugenics is devoted to the first installment of a monograph entitled “The Problem of Alien Immigration into Great Britain, illustrated by an Examination of Russian and Polish Jewish Children” by Karl Pearson and Margaret Moul. About 600 Jewish boys and nearly as many girls constitute the sample upon which the work is based. The data include anthropometric measurements, medical information, sociological facts, and intelligence rating. Non-Jewish school children in similar districts are used for comparative data. Coefficients of correlation and of mean square contingency are employed in measuring the relationships of the different variables.
In physique the Jewish children are not found to be superior to the non-Jewish children, but inferior to them in the great bulk of categories dealt with. In cleanliness of clothing the Jewish children fall far short of the native Gentile population. (CL: Gee. Do you reckon they might have been poorer being recent immigrants? And how do you objectively measure cleanliness of clothes? Oh, yeah. Yamulka’s sure make a kid look dirty to certain eyes.)
In the estimation of the intelligence of children a qualitative scale devised by Professor Pearson was employed, the children being graded by their teachers. (CL: So he devised his own qualititative scale. Did the general knowledge questions include Hitler’s birthdate? And the tests are graded by their teachers. No potential for bias there.) The measurements on this scale are said to be highly correlated with Binet-Simon tests, with certain physical and psychical characters, and give the values to be expected for fraternal and parental hereditary coefficients. It is concluded from the studies of intelligence that for practical purposes of prognosis there does not exist in the present material any correlation of the slightest consequence between the intelligence of the child and its physique, its health, its parents’ care or the economic and sanitary conditions of its home. . . . .Intelligence as distinct from mere knowledge stands out as a congenital character. (CL: So Karl’s IQ test is unrelated even to health. Psychometricians spent the next 50 years trying to find such a pure test of IQ. Note to self: Find a record of this test, tweak the words, call it my own and win Nobel prize).
The Jewish girls have less intelligence than the Gentile girls in any type of Council School. The Jewish boys are not so good as the Gentile boys of the medium or average schools, but better than the boys of the poor type of school. (CL: Implicitly then Karl’s IQ test is strongly related to schooling. Strange considering he has just claimed his test measures a congenital characteristic.) Taken on the average, and regarding both sexes, this alien Jewish population is somewhat inferior physically and mentally to the native population. The contention of the authors is that in a crowded country only immigrants surpassing the average of the native population in physical and mental qualities should be admitted.
You might argue that Pearson is not here to defend himself. I don’t image the Jewish kids were given much chance to defend themselves either.
You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.
October 22nd, 2007 at 1:12 pm
Well so far this thread seems to have only exercised Chris and me! I do think that it holds a lot of lessons for statisticians, though. It seems to me that the Pearson foray into social analysis from perspectives common enough in his time but now seen in hindsight to be ghastly is one object lesson of the political and ethical dangers of this kind of statistical work. It is also another reason, apart from a common antipathy to numerical methods, that scholars in the social sciences and humanities are so averse to nontrivial uses of statistics.
Anyway I have been trying to do a big cleanup of my home office and I have reminded myself that there is quite a lot written in the history of science about Karl Pearson and his co-workers. Eileen Magnello is someone who has written quite a bit about Pearson. Donald Mackenzie is another name. I have just discovered a paper on the web by the latter that seems interesting in its blend of technical statistics and political analysis:
Donald MacKenzie. Statistical Theory and Social Interests: A Case-Study Social Studies of Science (SAGE, London and Beverly Hills), Vol.8 (1978), 35-83.
This paper is freely downloadable but the URL is absurdly long so I will leave the interested reader to Google it. Here is the abstract to the paper:
“This paper examines the controversy that took place between 1900 and 1914 about how best to measure statistical association. The divergent views of the two sides are examined by means of a study of the work of the major participants in the controversy: Karl Pearson (1857-1936) and George Udny Yule (1871-1951). It is argued that the theorizing and scientific judgments of the two sides embodied different ’cognitive interests’: that is to say, differing goals in the development of statistical theory resulted in approaches to the measurement of association that were structured differently. These different cognitive interests arose from the different problem situations of statisticians whose primary commitment was to eugenics research and those who lacked any such strong specific commitment. It is suggested that eugenics embodied the social interests of a specific sector of British society, and not those of other sectors. Thus differing social interests are seen as entering indirectly, through the ’mediation’ of eugenics, into this episode in development of statistical theory in Britain.”
And here is the conclusion:
“I have argued that the two divergent approaches to the measurement of association to be found in the work of Pearson and Yule can be seen as expressing different cognitive interests; that these different cognitive interests arose from the different problem situations of a statistician whose primary commitment was to a research programme in eugenics and a statistician who lacked any such strong specific commitment; and finally, that eugenics itself embodied the social interests of a specific sector of British society, and not those of other sectors. Thus differing social interests can be seen as entering indirectly, through the ’mediation’ of eugenics, into the development of statistical theory in Britain. In the absence of a great deal of further research, particularly on the hypothetical constellation of interests suggested in the previous section, this conclusion must remain tentative. I hope, however, that this paper has shown that ’hard sciences’, such as the mathematical theory of statistics, should not be excluded a priori from analysis in terms of social interests.”
It’s interesting that this paper is from the “Social studies of Science” literature that came under strong criticism following the famous Sokal hoax. I have not had time to read it yet but Mackenzie seems to be technically capable on both sides of the divide.
October 23rd, 2007 at 1:09 pm
“I have argued that the two divergent approaches to the measurement of association to be found in the work of Pearson and Yule can be seen as expressing different cognitive interests.” And here was I thinking that association was an objectively defined parameter of a sampling model. Silly me.
It seems to me that the problems with Pearson’s article was not some quirky measure of association that he was using - it was the inherent biases in how both the response and predictors were measured. I do not personally think that it is unethical to try to measure differences between races. Indeed, there being some obvious physical differences between races (which is really the only definition of race), we can count ourselves fortunate that there do not seem to be practically significant differences in mentality. It is fascinating and scary to think how the 20th century would have progressed had it turned out, for instance, that Africans were distinctly less intelligent than Europeans.
I have always thought that the best scientific evidence for neither race nor even genetics being the major factor in success of humans individually or collectively was the experiment known as the Sydney Penal Colonly. Within 100 years Australia was the richest nation in the world, despite having being founded (as Winston Churchill observed) by “convicts and Irishmen” .
Thanks for your input on this post Murray.
October 24th, 2007 at 7:06 am
Hang on Chris, the Mackenzie article I talked about (which I have now read!) does not speak directly to the Pearson/Moul study about the children of Jewish immigrants. Rather it claims that Pearson saw a lot of technical issues through the perspective of how they affected his groups work in eugenics. Interestingly there is a conflict between Mackenzie and Magnello on this point. In her paper
“The Non-Correlation Of Biometrics And Eugenics: Rival Forms Of Laboratory Work In Karl Pearson’s Career At University College London,Part 1″ Hist. Sci., xxxvii (1999)
she argues that Pearson only reluctantly took on the Eugenic lab and the two labs worked without much interaction.
I am no great supporter of the “science studies” movement in the social sciences, but it seems to me that Mackenzie might be right in thinking Pearson’s motives were Eugenic in his controversy with Yule. When he wants to compare heritability coefficients between variables measured on different scales his tetrachoric correlation, which estimates the correlation of unobserved continuous latent variables serves this purpose much better than Yule’s Q, which is an ad hoc measure of discrete association with no connection to a model parameter. Certainly for Yule association was not “an objectively defined parameter of a sampling model.”
I think that, leaving race aside for a moment, Yule can be seen as a fore-runner to the nonparametric tradition in statistics and Pearson to the model-building tradition. Bringing race back in, it is amusing that Pearson was worried about Jewish immigration weakening the intellect of the British population. I don’t think anyone would now accept that characterization of what happened.
I myself feel closer to the model-building than the nonparametric tradition but it is salutary to be aware of how model-building can lead you into trouble if you allow yourself to smuggle in too many unexamined assumptions.
October 29th, 2008 at 7:21 am
Good post.
December 14th, 2009 at 2:49 pm
[…] correlation coefficient, chi-square test of independence, classification of distributions (and racial eugenics….) · #2: Ronald Fisher (1890-1962); ANOVA, exact tests, permutation tests. More generally […]
February 6th, 2010 at 3:51 pm
This is a very thought provoking post. However, the definition of “race” as most informed scientists acknowledge cannot be clearly or distinctively made.