When the Time Seems Ripe: Eugenics, the Annals, and the Subtle Persistence of Typological Thinking


Corresponding author: Kenneth M. Weiss, Ph.D., Department of Anthropology, Penn State University, 409 Carpenter Building, University Park, PA 16803. Tel: 1-814-865-0989; Fax: 1-814-865-1474; E-mail: kenweiss@psu.edu


This journal began in 1925 as the Annals of Eugenics. Much has changed since then. The original Editors’ primary eugenic objective was not achieved, and eugenics justifiably became notorious for racism and gross abuse of human rights. But one founding aim was to publish advances in statistical genetics, and that objective prospered in the journal's pages from its beginning to the present day. The online availability of the original issues will be useful to those interested in the history of both eugenics and human genetics and will provide a reminder of how the careless use of genetical concepts can go astray.

“The time seems fully ripe for the issue of a journal which shall devote its pages wholly to the scientific treatment of racial problems in man.” So began this journal in October 1925, under a very different name from the copy you are now reading. Titled the Annals of Eugenics, it was founded to serve a specific scientific need. The journal issues from the early years have long been restricted to musty shelves of major research libraries. However, the Editor thought the time was now ripe to make them available to a readership that may include geneticists, historians, sociologists, and philosophers of science. Given the history of eugenics and the social abuse it led to, this was a sensitive decision. But it was a correct one, and Dr. Ruiz-Linares is to be commended for deciding to go ahead.

The journal was launched under the editorship of Karl Pearson (1847–1936), assisted by Ethel M. Elderton (1878–1954). Pearson had drifted among various areas in science and the humanities, but ended up a protégé of Charles Darwin's cousin Francis Galton. He became the leader of academic statistics, and was the first holder of a chair of Eugenics, which Galton had endowed at the University of London (now University College London). Ms. Elderton had also worked with Galton, served as Secretary of the UK Eugenics Record Office as of 1906, and collaborated in research with Pearson. While he was clearly the senior prominent figure, Elderton was well respected for her own research (David, 1954). She was awarded the Weldon Medal in 1919 (named after the biological collaborator who had motivated Pearson to study biostatistics) and received her doctoral degree in 1931.

The Editors wrote the Foreword for the first issue of Annals, which is well worth reading, and highly illuminating. They outlined the nature of their task and of the field at the time.

What Was Eugenics, and Why a New Journal?

Human beings are imperfect, and if you were well off, you had a clear view of individuals in your own country who were not up to standard and perhaps even threatening in their numbers. Across the waters were the less successful countries of the world. Even the short reach of recorded history shows the universal rise and fall of cultural as well as technical prominence, and hence the vanity of nationalistic hubris. But if you were under the Darwinian Zeitgeist and believed that inequality is inborn and that what predominates today must be inherently better than what had gone before, you might naturally feel, with confident superiority, that science could assist Nature to improve humankind. This is how Pearson and Elderton saw the world.

Not long before, August Weismann had shown (e.g., Weismann, 1893) that the germ line and somatic cells were isolated from each other, proving that Lamarckian inheritance was impossible. This led Pearson and Elderton to argue that social idealists who advocated education and other social measures for the long-term betterment of the lower classes were naively misguided. Rather than yielding lasting improvement, such measures could lead only to the endless expense of maintaining the inferior members of society, because the beneficial effects of education and sound nutrition could never pass directly to the germ line.

Nonetheless, the Editors asserted that it was soft headed to believe in the “impracticability of fundamentally improving the race of man.” Such improvements could in fact be achieved by modifying the germ line itself, and in that noble cause arose the Eugenics movement, a term Galton had coined in his book Inquiries into Human Faculty and its Development (Galton, 1883).

The Editors saw a need for a new journal to serve a new science of human improvement. That science had to be professional in contrast to the “dilettante” work that was too prominent in anthropology at the time. It must also be mathematical because “The whole development of Mendelism in recent years has been in the direction of a multiplicity of factors, even for apparently simple characters,” so that one “must ultimately abbreviate his analysis by continuous methods of algebraic representation”– that is, rather than by discrete qualitative analysis alone. But investigations in this direction had to overcome the resistance that many in biology had towards mathematics, which they viewed as irrelevant if not outright threatening. Mathematical biology needed its own publication outlet to work independently, because no appropriate journals existed, and the protective reactionaries who controlled established journals might prohibit the “unclean novelty,” or even misappropriate its ideas as if they were already well known. Indeed, the Royal Society had recently proscribed biometry from its publications. This may seem strange in today's era of quantitative biology (although many students when confronted with such massive challenges as Mendelian segregation proportions or Hardy–Weinberg still protest that they chose to study biology because they did not like maths!).

Eugenics was intimately related to Darwinian selection-driven evolution, and thus had a natural concern with variation in biological quality among individuals and families. But Pearson and Elderton were determined to keep focused on a larger scale: “no other topics than the problems of race in man will be dealt with.” They insisted on “National” eugenics, a word “essential to the idea Galton had in mind … the statecraft which would elevate a whole nation and make it fittest for its work in the world.” The blunt truth was that “all men are not born equal in mind or body,” or else we would face no need for eugenics: “Even now we are progressing slowly towards tests for occupational fitness, and eventually that fitness should be intensified by marriage within the caste.”

In an evolutionary view so naïve and confused that it's hard to believe intelligent persons could hold it even in the 1920s, Pearson and Elderton wrote that “In precisely the same way there is a relative fitness of nations, their racial history … [that] … fit them best for definite forms of work.” In their own manifestly dilettantish confounding of short-term ethnocentric observation and long-term evolutionary consequences, and in group rather than Darwinian interindividual selection terms as well, the Editors noted that the eugenic problem for the British, with their long coastline, is to breed sailors, while Germany needs soldiers to protect its many landed frontiers. Their attention then predictably passes southward, where “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.”

With this perspective, the Editors stressed that, as with “Racial Hygiene” in Germany at that time, their objective was the study of “agencies under social control that may improve or impair the racial qualities of future generations, physically or mentally.” We added the italics to show that these authors clearly have policy in mind, not just voluntary action by individuals. Following closely and predictably on the heels of this declaration is a statement typical of the period, that “One of the great problems of Eugenics is concerned with the limits of immigration.” And typically as well for its time, one group they singled out was the Jews.

Britain was not alone in such views. Immigration was a persistent concern of the powers that be in the United States as well, but in somewhat different ways, and it remained a major issue well into the 20th Century (indeed, it still is). To make those decisions, categorical thinking was fundamental. Initially, only “white” persons were eligible for birthright citizenship. Naturalized citizenship was limited to “white” persons as well (Naturalization Act of 1790). Even with post-Civil War reforms that extended citizenship to persons of African descent, Native Americans, and Asians remained ineligible for citizenship. Native Americans were not eligible for birthright citizenship until 1924, and generally birthright and naturalised citizenship decisions were not entirely disconnected from race until 1940 and 1952, respectively. Moreover, Congress restricted immigration by countries of origin (beginning with the Chinese Exclusion Act in 1882 and continuing with, e.g., the Immigration Act of 1924 that established a permanent quota system). Whenever an individual's eligibility was challenged, courts had to decide how a particular individual fit into the legal framework. For example, courts had to determine whether persons like Asians or those of Mexican ancestry were “white.” To do so, the courts consulted scientific literature. By the 1920s, the US Supreme Court changed the definitions from more technically definable traits such as skin colour or standard race terms such as “Caucasian” to popular or common-knowledge categories by which, for example, “white” largely referred to Northern and Western Europe (Dr Jennifer Wagner, JD, personal communication, and see Lopez, 2006, Motomura, 2006, and “Enforcing forced definitions of race” by Jennifer Wagner, JD, at http://ecodevoevo.blogspot.com, 2010).

The Editors warn that proper social policy to address these problems “will aim at the betterment of future generations rather than at the increased comfort of the individual.” But they quote Galton that such policy must wait until “the desired fullness of information shall have been acquired,” and then “will be the fit moment to proclaim a ‘Jehad’ or Holy War against customs and prejudices that impair the physical and moral qualities of our race.”

The Editors intend the journal to give the new science the breathing room in which to generate the necessary basic knowledge. And they envision leading the way to the day when every university “will have its professor and laboratory of Eugenics.” An ambitious agenda indeed!

From This Beginning, a Surprise

The first Annals issue was promptly reviewed in Science (Holmes, 1926). The reviewer approvingly singled out the Editors’ Foreword, and gave considerable attention to Pearson's lead article on the problems with Jewish immigrants to Britain. “There is no evidence in the memoir of any prejudices against the Jew,” wrote the reviewer, in a remarkable reflection of prevailing attitudes. The journal also received a laudatory review by the prominent American anthropologist E. A. Hooton (Hooton, 1926b).

The Tables of Contents for the first 2 years (shown in Fig. 1) are revealing, in the context of the Editors’ ambitious prospectus. Surprisingly, other than Pearson's own series of papers on the “problem” of alien immigration to Britain, throughout his years as Editor there were few if any papers directly concerning national eugenics. Instead, the widely ranging contents largely consisted of studies of human variation, mostly in families, often concerning inheritance patterns of disease. Such papers are found today in many journals, including the one you are now reading. There are studies of environmental effects, possible heritable susceptibility to infection, and even the traits and affinities of human fossils. Other than Pearson's own series on immigrants, one might never guess that “no other topics than the problems of race in man” were to be published in the journal! It was a policy that, in the extreme, was honoured in the breach. Instead, what Pearson and Elderton had launched was a successful outlet for quantitative analysis of inheritance in humans.

Figure 1.

Annals of Eugenics, Table of Contents for volumes I and II.

In 1934, the Editorship passed to R. A. Fisher, another founder of modern statistics, but the nature of the journal remained unchanged. There was little if any obvious racism in the paper titles, even in the run up to, and through the war years. In that sense, the Annals was essentially never actually about national eugenics or race, in spite of the rationale in its stirring founding editorial.

Why Was This?

Nation and race represented sampling frames too weak and vague for the scientific tools available to the eugenicists at the time. Instead, the genetic basis of human traits had to be studied in situations where their transmission among relatives of known relationship could be directly observed. Statistical predictions could be derived from Mendel's laws of inheritance, even with nothing known about their actual causal genes – or even what genes were (Waters, 1994, 2004).

However, as the original Editors noted, the multifactorial nature of human traits was recognised by the time of the first Annals edition, and that presented much greater inferential challenges. There were scant methods for explicit genetics of traits that did not segregate in qualitative Mendelian ways, and all that could be observed were quantitative correlations among relatives. Recognition of the importance of such correlations for understanding heredity of complex traits went back to Galton, and in his landmark 1918 paper Fisher (Fisher, 1918) had shown that Mendelian transmission was still the underlying key. But even by 1926, quantitative genetics was still implicit genetics: one could not specify how many genes were acting, much less their individual allelic effects or transmission. Genetic effects could only be quantified in aggregate, and even the aggregate transmission of those effects had to be observed directly among relatives of known degree – not just by their presence in a nation or race.

In this important discovery phase of human genetics, mathematical methods for inferring genetic causation under a wide variety of conditions were developed and applied in the Annals’ pages. Indeed, quantitative genetics is alive and well today, with the same basic theoretical basis as it had in the journal's early years. However, subsequent high-throughput genotyping and computing technologies have made it possible to identify individual contributing genes, which can then be followed up in confirmatory molecular detail.

Family studies are still done as they were then, and disease is still a topic of major interest. However, the major gene-identification approach now, the genome-wide association study, is of a very different design. Directly measured polymorphic markers, densely spaced across the genome, are searched to find markers whose variation is statistically associated with quantitative or qualitative phenotypes, as for example in case–control comparisons. These are still family studies, but with a remarkable difference: genetic analysis can now be done on a population or national level, where parent-offspring transmission is not directly observed. Instead, individuals with shared phenotypes are assumed to be distant relatives from implicit pedigrees.

The reason this works is evolutionary, though not in the eugenecists’ Darwinian context. Because specific mutations at any given site in the genome are rare, marker alleles that are identical by state (IBS) in different individuals can often also be assumed to be identical by descent (IBD) through the many generations of the unobserved pedigree that connects their copies of the marker to their common ancestor in which the mutation occurred. The same IBS-IBD relationship among copies of marker alleles shared among individuals with the same phenotype, like presence of a disease, is assumed also to apply to an unobserved causal allele. The evolutionary reasoning is that the marker and nearby causal alleles descend together on a clone – a genealogy – of chromosomal segments from the copy on which the marker mutation arose.

By contrast, in Pearson's time, before the nature of DNA as a nucleotide sequence was known, alleles had been thought to arise recurrently and reversibly, with stable forward and backward mutation rates. If that were so, sharing of marker alleles among cases would not imply that they resided IBD on homologous chromosome segments.

If the evolutionary IBS = IBD assumption holds for members of a population, it might seem that Pearson's idea of race-based analysis could be realised after all. However, the notion of race and nation in Pearson's time was implicitly statistical in a way that he would have recognised, but did not state. There is an important evolutionary lesson here, and we can easily see what it is.

The Evolution of a “Race”

We illustrate the evolution of the notion of a pure race using a forward simulation program called ForSim (Lambert et al., 2008). We ran a simple simulation of 5000 individuals for 5000 generations of random mating, a size and time not unlike the history of human habitation of a continent such as Europe (home of one of the traditional major human races). We simulate just three 30 kb genes, in two of which mutations affect a quantitative phenotype that is not subject to natural selection.

At the end of the simulation, we randomly sampled 1000 individuals from the population, and computed all pairwise genotypic distances for the 356 single nucleotide polymorphisms (SNPs) that were present at the end of the simulation run. The sampled individuals were sorted by phenotype, and the resulting colour-coded pairwise distances are presented in heat map format in Figure 2. The diagonal is red because it is the comparison of each individual with itself. But off the diagonal, and except for the expected relationships among close relatives in the simulated population (not visible in printed resolution) there is a dearth of close genotypic identity.

Figure 2.

Pairwise genetic identity heat map of members of a simulated three-gene “race.” One thousand randomly sampled individuals sorted by a simulated selectively neutral phenotype, and compared for all variable sites; genotypic identity at each SNP is 0 if they share no allele, 1/2 if they share 1 allele, and 1 if they share both alleles at that site; identity averaged for all variable sites. Each individual is represented by a row and column, and the pixel at position (x, y) represents the genotypic identity between individual x and y, G(x, y), as coded by the heat bar legend at the left of the figure, from red (high identity) to violet (low identity). Because G(x, y) = G(y, x) the figure is symmetric about the diagonal, which is the genotypic identity of each individual with itself (by definition equal to 1). Simulated recombination was 1Mb/cM, mutation 2.5 × 10−8 per site per generation, in a diploid random-mating population. Simulated by ForSim (Lambert et al., 2008).

This population has evolved in isolation, like classical notions of a pure race, and a simulation of just three genes is much more favourable to finding high genotypic identity than data from the real world of 3 billion base pair genomes. Yet, we do not find it, even between individuals sorted by phenotype. The reason is simple and relates to typological but implicitly statistical notions of racial variation as expressed in the eugenics era.

Statistical Typology Then …

From Linnaeus onwards biology has divided the living world into distinct species, represented by type specimens. Types are essentially distinct, and a key problem of 19th century biology, the “mystery of mysteries” so central to Darwin, Wallace, and many others, was to account for the origin of new and differing types.

The idea of type specimens had long been applied to races as types of humans. In 1926, a year after the Annals was launched, Hooton wrote that humans could be divided by morphological characteristics into pure races and individuals admixed among them (Hooton, 1926a). The world's leading human genetics text in the time of the Annals founding, Human Heredity (Baur et al., 1921), expressed a similar view. However, instead of vague morphological measures, its authors insisted that the defining traits should be traits that are clearly Mendelian, and hence genetic, real, reliable, and inherent. Using such traits they, too, sorted humankind into pure races and individuals admixed among them.

This is classic typological thinking, yet Pearson, Fisher et al. were far from naïve. They may have focused on a few visible or imagined behavioural stereotypes, but they knew that there was variation even within racial types: not all Europeans (not even all Jews!) are morphologically or genetically identical. So what kind of “type” were they thinking about?

The idea is a somewhat elusive one that can be related to our simulation exercise. A race can be characterised as a variable type in the following statistical way. In genetic terms, a “race” R would be defined by a vector of allele frequencies at a set of defining loci, say R = (p1, p2, p3, …) for a selected allele at locus 1, 2, 3 … . Each individual in a “pure” race is a random draw from this vector of allele frequencies. A race is thus a population in multilocus Hardy–Weinberg genotype proportions based on its defining allele frequency type vector. For example, the chance that a member of the race would be a homozygote for all the defining alleles (AABBCC …) would be p12 p22 p32 … Another member of the same race could have, say, the alternative allele at each locus (aabbcc …), with probability (1-p1)2(1-p2)2(1-p3)2 … If even a modest number of loci are considered in the defining set, as in our simulation, the probability of any given multilocus genotype is so small that effectively every nontwin member of the race is genetically different, yet all are drawn from within the same type specification. This is not the usual notion, but it is a way of formalising what was in essence said by the early geneticists, and it is what is graphically reflected in the simulations in Figure 2.

Expressed this way, a race's type vector points to the centroid of its expected genotype distribution as might be represented in a multilocus Punnett square. The quantitative genetic difference between races is related to the Euclidian distance between their type vectors. These considerations explain how races could be distinct yet have the obvious overlaps that make defining races a challenge, as Darwin had long ago clearly understood (Darwin, 1871). While there will not be high genotype identity between individuals within one group, pairwise identities of individuals between groups will, as expected, be even less, as can easily be shown by simulation (Weiss & Lambert, 2010).

Individual, family, nation, and race mix and mingle in a complex and probabilistic way, since they imply that the notion of pure type, or race, must be defined externally and applied by assumption. The reason is that frequencies are statistics that can only be estimated once a frame of reference – in this case, a population – is specified. At least as important is that the eugenicists’ idea that individuals are either members of a pure race or are admixed between races, entirely depends on the actual rather than simulated existence of those parental types to be admixed from.

… and Now

One might fancy that typological days are safely locked away in the cobwebs of history, but the same typological thinking is still here, all around us (Weiss & Long, 2009). In one of its more rigorous forms it is called “structure” analysis, after the first of several programs that perform it (Pritchard et al., 2000).

Structure programs divide sampled individuals into statistically homogeneous types (in the above Hardy–Weinberg sense of sampling from a type vector) and other individuals who are admixed among them. The analysis is a sophisticated statistical genetic partitioning, based on specific assumptions about the nature of human evolutionary history. It has become a routinely used approach by scientists who would not dream of using words like “race” and have no discriminatory or eugenic intent, but who may not be aware of the history of such concepts.

Ironically, as is well known, even with low intrapopulation genotypic identity, if many loci are genotyped it is easy to place individuals in their respective populations (Witherspoon et al., 2007; Nievergelt et al., 2008; Weiss & Long, 2009; Weiss and Lambert, 2010). This is simply a reflection of the distance between the population type vectors. The low level of intrapopulation identity is a surprise only if one thinks as classical typologists seem to have done, in terms that only include a few stereotypical traits.

Structure analysis reflects categorical assumptions about population history and the distribution of contemporary genetic variation, and is not in itself directly concerned with phenotypes, although one important purpose of its development was to detect population stratification that might generate false positive results in disease association studies (Pritchard et al., 2000; Voight & Pritchard, 2005). Structure analysis apportions ancestry in an assumed genetic landscape, just as classical eugenics was based on assumed phenotypic landscape comprised exclusively of distinct parental types and their hybrids.

However, as our simulation illustrated, there is little genotypic identity within a “pure” group, even when, as here, only three genes affect the trait (and, though not shown, natural selection makes little qualitative difference). This is true even among phenotypically similar individuals (Fig. 2), and shows why genome-wide mapping studies typically identify few contributing genes but leave most of the familial aggregation (e.g., as measured by heritability) unaccounted for. The reason is that if the contributing causal genotypes involve more than a few genes, their allelic heterogeneity is substantial: there are many genotypically different ways to be diabetic just as there are many different ways to be European. This presents a challenge to the idea of finding the genetic basis of disease – a main target of the early years of the Annals, and today's goal of personalized genomic medicine. Even classically Mendelian diseases are genotypically diverse (and most persons affected by classic recessive disorders like phenylketonuria are heterozygotes), a modern echo of Pearson and Elderton's “multiplicity of factors, even for apparently simple characters.”

Structure analysis relies on a kind of “as-if” fiction that treats actual individuals as if they were the product of a type-and-hybrid history (Weiss & Lambert, 2010). As-if fictions are useful and well known in evolutionary theory. Effective population size (Ne) is an example; nobody thinks that humans literally evolved in a random-mating population of 10,000 individuals, but the use of Ne is very helpful in approximate reconstructions of some aspects of the past that would be unwieldy if not impossible given the unknown population structure of our ancestral populations over thousands of past generations.

However, as-if fictions can be materially misleading if presented as literal truths, as structure analysis papers often seem to be (Weiss & Long, 2009; Weiss & Lambert, 2010). For example, structure-analysis papers portray individuals as admixed between contemporary parental populations, when clearly the individuals are of comparable adult age and sampled geographically far apart, and although each sampled person only has two literal parents, he can be assigned admixture fractions from multiple parental populations. So samples of individuals are assumed to represent parental populations in the past – but when, where, and how far back?

Of course, in some situations, such as the Americas, which have recently been settled by peoples from diverse continents, the assumptions, if only approximate, are at least intuitively and historically sensible.

A statistical definition may seem to rescue typological thinking in a scientifically acceptable way, but structure analysis does not in fact isolate genotypic or phenotypic homogeneity, even in the type-vector sense. This can be seen by the fact that when a parental population identified in a global structure analysis is then analysed in detail by itself, it is found to have comparable internal parental-admixture structure. Structure analysis provides appealing graphical displays that digest the hilly distribution of genetic variation across a study's sampled geographic space, but in most cases it is historical fiction and there are other equally effective, more evolutionarily sound, noncategorical ways to portray the geographic patterns of human variation (Weiss & Long, 2009).

We stress again that the authors of structure analysis rarely if ever use the word “race,” and we are not suggesting that their analysis has any social racist or eugenic intent whatsoever. But what we know about population and evolutionary history shows that typological thinking is at best inaccurate and can be misinterpreted.

Beware of Wolves in GORE-TEX

The history of eugenic thinking has been much studied (Kevles, 1995; Carlson, 2001) and need not be repeated here. This journal's founding Foreword reflected the assumptions of its time and context. The eugenicists were the most prominent and respected scientists, most of them trying honourably to understand Nature and to improve society (Pearson was an active socialist, e.g.). But whereas religion had earlier provided the framework for value judgments about humankind, evolution took the stage with values that seemed natural in the most objective and ultimate possible sense. In effect a new belief system replaced the old.

The papers in the Annals’ early years were concerned with ordinary aspects of inheritance in humans. What they were reporting was science of high quality. But the science was used by scientists, physicians, and policy makers, to impose a particular Darwinian view of social responsibility that led them to incarcerate, sterilize, experiment on people, or even to exterminate them.

The idea of reproductive strategies to improve society was not new. Plato's Republic suggests selective breeding of the elite and Malthus bemoaned the futility of attempts to help the poor, as they will always outrun the available food supply. Darwin's ideas and their eugenic interpretation reflected contemporary debates about the role of natural law in the dynamics of society, history, and inheritance. From Galton to Pearson and beyond, geneticists could imagine a better society and thus could believe in the power of science to achieve it.

The societal issues are complex. Selective abortion is legally sanctioned and widely practiced today, sometimes as a result of fetal genotyping. Genetic counselling, and measures such as in vitro fertilization (IVF) can prevent or remove conceptuses carrying deleterious genotypes, genetically based therapies hopefully can avoid much grief and suffering, and the future will see greatly augmented ability to engineer improved genotypes in IVF fetuses (e.g., at least in the sense of replacing deleterious alleles). And we all want to help our children choose what we would consider desirable mates (even if our children usually do not consider this to be “help”). Indeed, parental reasoning can be subtle: deaf parents sometimes want to do genetic screening to ensure that they will bear deaf children (Middleton et al., 1998; Templeton, 2007; Enns et al., 2009). These are, however, individual voluntary rather than institutional or governmentally coercive decisions of who should exist and who should not.

By espousing pious promises of human betterment, eugenicists reflected the class structure of their time. But the ensuing abuses showed they could be wolves in sheep's clothing. Today, in our high-technology age, the clothing may be GORE-TEX, but similar promises of genomic Utopia are often made. The extent to which these will lead to the kind of eugenics that targeted individuals a century ago is unknown, but the temptation seems ever-present. For example, China has policies related to sterilization that some feel resemble the classical eugenics era (Harper, 2008). And modern technology raises new societal issues about the use of genetic data. Technical sophistication does not prevent us from falling prey to our cultural biases today, any more that it did in Pearson and Elderton's time. Prevention is unambiguously the better part of valour: things often seem harmless at the outset, but we must assume that there are still wolves in the world.

When Fisher inherited the Editorship of the Annals, it was subtitled “Statistical studies in genetics and human inheritance” (Fig. 3). “Racial problems” were not mentioned. Fisher's Foreword as new Editor acknowledges the Eugenics Society and Galton Laboratory's logistical support for continued study “of the genetic situation in man.” The editorial transition occurred in 1934, during the Nazi regime with its rabid racism, but the sense of race was not a major feature stated by Fisher, despite his well-known personal eugenic elitism (e.g., Fisher, 1930), and the journal remained, as it had basically always been, a quality outlet for quantitative method development and application for “the better understanding of heredity in human populations.”

Figure 3.

Advertisement for volume XII of the Annals. Source: public domain.

In 1954 (coincidentally, the year of Dr. Elderton's death), Lionel Penrose (1898–1972) became the new Editor. He was a prominent British medical and psychiatric geneticist, who held the Galton chair from 1945 to 1965 and had published in the Annals. But by that time eugenics had given itself one of the best deserved bad names in the history of science. Based in part on his own work on intelligence and mental illness, Penrose was convinced of the poor scientific foundations of eugenics and the futility of its proposed strategies to regulate reproduction (Pavey, 1998; Watt, 1998a,b; Harper, 2008). Upon becoming Editor, he changed the journal's name to its current Annals of Human Genetics (he also renamed the Galton chair to Chair of Human Genetics). The Annals maintained its high-level content, and its flavour today remains consistent with the legitimate aspects of its flavour even at the time of its founding.

Race remains a concept whose usefulness is still widely debated, and that stubbornly refuses to go away (Kittles & Weiss, 2003; Weiss & Fullerton, 2006). There are still “racial problems” in human society, but they are mainly social rather than genetical problems. However, categorical thinking is still rife and, as before, often entails value judgments about behaviours such as sexual orientation, gender, disability, and so on. Average differences between samples of members of these externally and a priori defined categories are measured and used to imply group differences, thoughtlessly imputing to individuals their purported group averages, even when the overlap is far greater than the mean differences: men are better than women at math, some “ethnic” groups are less intelligent or better marathoners than others. The litany of such rationalised post hoc generalizations is familiar to us all.

It would be naïve to think that governments will no longer yield to the temptations of technology. A booming industry today involves various forms of genetic ancestry testing that assign individuals to one or more parental groups. Indeed, reflecting this state of play, and resembling a central concern of Pearson and Elderton, a recent editorial in Nature dealt with a UK government idea of using DNA testing to establish nationality, in the hope – you guessed it – of stopping undesirable immigration (Anonymous, 2009)!

Yes, the time seems ripe to look at these issues once again.


We thank Anne Buchanan, Jennifer Wagner, Peter Harper, Steve Jones, Daniel Kevles, and Dr Ruiz-Linares for their very useful comments on this manuscript. We considered every suggestion, and hope to be forgiven where we decided to follow our own muse. ForSim simulations supported by a grant to Penn State Evan Pugh Professors fund, and by NIH grants MH063749 and MH084995 in collaboration with Joseph Terwilliger.