Making sense of Mendelian genes

Evelyn Fox Keller's The century of the gene (2000) is the most famous obituary of the entity forever associated with Gregor Mendel (Keller 2000). But what kind of a life did the Mendelian gene ever really have, anyway? Haven't thoughtful biologists long regarded the idea of it as an oversimplification? Indeed, from early days, all kinds of complexity have been allowed for. At the same time, the power of Mendelian explanations to cut through complexity and expose underlying simplicity has been relentlessly showcased. In the spirit of Keller's subsequent book, Making sense of life (2002), I want in what follows both to exhibit this ‘ambi-valence’ (Keller 2002, 130) and to suggest how crucial it has been in making the Mendelian gene – and the determinism it underwrites – so long-lived. Although my analysis of what has made gene talk so durable will not coincide with hers, our accounts are complementary, to such an extent that the curriculum reform experiment I shall describe at the end could well be considered applied Kellerism.

The Mendelian gene

Gene concepts are legion. 1 By ‘the Mendelian gene,’ I mean the entity that does the explaining in elementary genetics, e.g. in explaining why two blue-eyed parents can have only blue-eyed children. On the standard Mendelian explanation, there is a gene for eye colour, and it comes in two versions or ‘alleles.’ There is the brown-eye allele, which is dominant. And there is the blue-eye allele, which is recessive. The father's sperm and the mother's egg each carry only one allele. If the zygote formed from the union of these gametes brings together a brown-eye allele with another brown-eye allele, then the child will have brown eyes. If the zygote brings together a brown-eye allele and a blue-eye allele, then the child will also have brown eyes, because brown-eye is dominant. Only if the zygote brings together a blue-eye allele with another blue-eye allele will the child have blue eyes. A corollary is that, since blue-eyed parents can only ever contribute blue-eye alleles, two blue-eyed parents will have only blue-eyed children.

Note two features of the entity invoked in this example. First, the gene is either in the form of the dominant allele, responsible for the dominant character-version, brown-eye; or it is in the form of the recessive allele, responsible for the recessive character-version, blue-eye. The Mendelian gene is, in short, binary. Second, it is completely determinative, in the sense that the allelic pairs explain without remainder. There is no need to consider anything other than which allele ends up with which allele.

This definition emerged between 1900 and 1910, largely in the work of William Bateson and his research school at Cambridge. Although Mendel's 1866 paper ‘Experiments on plant hybrids’ furnished inspiration (Mendel 1866), the Mendelian gene is, it should be emphasized, in no straightforward way Mendel's. Yes, Mendel deliberately started his work with garden-pea varieties selected because they showed binary characters – seed colour as either yellow or green, with no intermediates; seed shape as round or wrinkled, with no intermediates, and so on – which were unaffected by context. And yes, he categorized each member of each character pair as ‘dominant’ or ‘recessive,’ depending on whether it was visible in the hybrid. But he did all of that as part of an inquiry not into inheritance but into hybrids, in particular the fate of hybrid characters in plants where such characters are not perpetuated intact. Furthermore, in explaining the patterns he derived, Mendel did not postulate the existence of discrete, paired, gene-like entities, but only character-making material, which mixed and unmixed down the generations. In Robert Olby's famous phrase, Mendel was no Mendelian (Olby 1979). 2

The Mendelian gene as an idea (not an object)

To leave the matter of definition there, however, is to invite confusion over what, exactly, we are talking about when we talk about the life and death of the Mendelian gene. A distinction of Ian Hacking's comes in handy here, between objects and our ideas about them (Hacking 1999, 21–2).

Objects, on Hacking's distinction, are items in the world. They can come into being and pass away, either for science or full stop. In Versailles in November 2018, for example, the platinum–iridium standard kilogram was voted into retirement, to be replaced not by a new hunk of metal but by a value derived from Planck's constant. The standard kilogram ceased to exist for metrology, if not for physics. (‘Le grand K,’ and its official copies, will remain in their vault, to see whether they lose mass over time.) 3 Since 1980, there has, according to medical officialdom, been no smallpox virus in the wild, thanks to a worldwide eradication program. The smallpox virus, we hope, ceased to exist full stop, outside of a handful of secure government labs at any rate. 4 Where objects are in the world, our ideas about objects are in our heads. Ideas too, of course, can come into being and pass away. The history of science is notoriously full of discarded ideas about objects. The idea of the vortex atom – the atom understood as a whirl of ether – is now a curiosity of Victorian physics. The idea of the ether itself, as something existing in the world, petered out in the early twentieth century. 5

Why adopt Hacking's distinction? Because it steers us away from ambiguous talk, about ‘the death of the vortex atom,’ say, or ‘the death of the ether.’ On our best current physics, the vortex atom and the ether never existed, and so, strictly speaking, cannot have ceased to be. Rather, what ceased to be were beliefs in the existence of these objects. To disambiguate in this way can itself give life to a new line of questioning about the interaction between existing beliefs and the world itself (Hacking 1999, 27–8 & 31–2). The historiography of genetics furnishes a shining instance in Robert Kohler's Lords of the fly (1994), about T. H. Morgan's fly room at Columbia (Kohler 1994). Although initially skeptical about the existence of Mendelian ‘factors,’ Morgan came to preside over the research program that, through fruit-fly crossing experiments, secured the identity of those factors as nothing other than bits of chromosome. (Another Hacking borrowing: if you can map them, then they are real.) At a distance, one might have guessed that what Morgan and his students did was to collect fruitflies and then check, first without and then with the aid of microscopes, whether the patterns of inheritance exhibited fit with Mendelian expectations. Not at all. Wild Drosophila melanogaster does not ‘Mendelize.’ Cleanly binary characters are conspicuous by their absence, so there is little scope for spotting such tell-tale Mendelian signs as uniform dominance in offspring, or the 3-to-1 dominant-to-recessive ratio in the offspring of offspring (Kohler 1994, 28–9). Morgan and his students had to breed Mendelizing fruitflies into existence, by breeding out all of the pattern-wrecking variability (Kohler 1994, ch. 3). The chromosomal Mendelism that the Morgan group set out in The mechanism of Mendelian heredity (1915) co-evolved along with the lab-bound fly lineages to which that theory now answered. 6

Kohler's book should be a permanent reminder that there is a marvellously interactionist history of the Mendelian gene to be told, rich with objects as well as with ideas about objects (Radick 2003, 197–8). Here, however, I will touch on such interactions only incidentally. My concern is mainly with the life and death of the idea of the Mendelian gene. I wish to show how a new interpretation of the nature of the life of this idea throws light on its remarkable ability to elude death.

Two ways of understanding the idea of the Mendelian gene

Although the term ‘gene’ is not used in The mechanism of Mendelian heredity, the Morgan group's discussion there of how Mendelian talk should and should not be construed – and how much scope there is for misconstrual, even among experts – remains peerless. The book's retention throughout of the older term ‘factor’ was certainly idiosyncratic. As the bibliography testifies, ‘gene’ had spread rapidly among geneticists after its introduction in 1909 by Wilhelm Johannsen. Morgan's own co-authors, Alfred Sturtevant, Calvin Bridges, and Hermann Muller, had used ‘gene’ in the titles of cited papers. But The mechanism of Mendelian heredity, aiming at a more general biological readership, carried a message for which ‘factor’ was indeed a better fit than ‘gene.’ Over and over again, the reader is reminded that whenever chromosomally borne hereditary factors are identified as ‘for’ a visible character (e.g. eye colour), or for a version of a character (e.g. red eye), that identification is limited to situations where the many other factors – hereditary but also developmental and environmental – influencing that character or character-version are held constant. Absent such constancy of background, all bets are off.

After a first chapter introducing the pairing and segregation of chromosomes as underpinning the pairing and segregation of Mendelian factors, the second chapter, misleadingly entitled ‘Types of Mendelian heredity,’ drives home how phenomenally complex the relationship between a factor and an associated character can be. It is not merely, as Bateson had stressed, that simple ‘dominance’/‘recessive’ relationships cannot be counted on, or that a single factor can affect multiple characters, or that a single character can be affected by multiple factors. A change in temperature, for example, can change the character associated with a given genetic constitution, as Erwin Baur had found in work with primroses. Raised at around 20° C, the white primrose, P. alba, produces white flowers and the red primrose, P. rubra, produces red flowers. But when the temperature rises to 30° C, P. rubra produces not red flowers but white ones. The lesson drawn is as much linguistic as biological:

Strictly speaking, we should say, not as we generally do for brevity's sake, that the difference between the two races is that one has white, the other red flowers, but we should say rather that P. rubra reacts at 20° by producing red, at 30° by forming white flowers; P. alba, on the other hand, reacts both at 20° and at 30° by producing white flowers. The constant difference between these races is not their color, but in the possibility of producing specific colors at certain temperatures. (Morgan et al. 1915, 38–9)

Brilliantly, these complications come across not as problems for chromosomal Mendelism but as taken-for-granted presuppositions. It is precisely because everything is interacting with everything else, all the time, with the most complex consequences, that we need to study heredity in the Mendelian manner, using organisms purged of variability and raised in controlled environments. Anyone who understands that, as Morgan and company put it at the end of the chapter, ‘every character is the realized result of a reaction of hereditary factors with each other and with their environment,’ will grasp the wisdom of a method of inquiry which imposes constancy on the factorial and environmental background in order to investigate how differences at a single locus of a single chromosome can affect a body (Morgan et al. 1915, 46).

Of course, there is a catch: the need not to forget all the contrived homogeneity behind Mendelian ‘factor for’ and ‘unit character’ talk and, conversely, the need to remember the status of that talk as abbreviating much more cumbersome talk about what was observed when, against a backdrop of factorial, developmental, and environmental constancy in a particular set of organisms, a particular change took place at one locus. Concern that new Mendelians, like some old ones, might need extra coaching in order to keep the limitations in mind surfaces throughout the rest of the book. A discussion of the ‘sex factor’ known to be carried in the X chromosome, for example, goes out of its way to disabuse the reader of the notion that the factor is, in any non-simple way, for sex:

As in the case of sex limited characters, so in the case of sex itself there must be many factors in the fertilized egg that are as essential to the development of sex as are the sex factors themselves, but as they are distributed to all individuals alike, they are not thought of as differentiators, but as forming the chemical background on which the single or the double amount of the sex factor gives its result. It is quite conceivable that one or more of these other factors might so change that the sex differentiators would become inoperative or even change so that those other factors themselves become the differentiators that determine sex.

The book's concluding chapter, entitled ‘The factorial hypothesis,’ reinforces all these lessons and then some. Using the factor for red eye in flies as an example, it begins by setting out two ways of expressing the relations between factor and character. On the first way, a factor is a segregating, intra-gametic something whose effects, upon combination and recombination, explain the 3-to-1 ratio of red-eyed flies to white-eyed flies in the second hybrid generation. On the second way, there are gametes which produce red eyes and gametes which produces white eyes, and these gametes differ by just one factor-difference. We go on to learn that, like pretty much every character, eye colour in flies is under the influence of multiple loci (25 were known by 1915). Can we nevertheless talk of a mutation associated with, say, pink eye as ‘the cause of pink’? Yes, because ‘we use cause here in the sense in which science has always used the expression, namely, to mean that a particular system differs from another system only in one special factor’ (Morgan et al. 1915, 208–9). And again, the chemical ramifications of a change at a single locus may affect multiples characters, not just the character where the effect is so conspicuous as to tempt ill-advised talk of a ‘unit character.’ (‘So much misunderstanding has arisen among geneticists themselves through the careless use of the term “unit character” that the term deserves the disrepute into which it is falling’ (Morgan et al. 1915, 210).)

So there are two ways of understanding ‘factor,’ but they are not equal, since the first, factors-for-unit-characters way needs the second, differences-that-make-a-difference way in order not to be misleading unto falseness. As the Morgan group put it, by way of sharpening the contrast between their Mendelian picture of development-and-inheritance and the Weismannian picture, with its vestigial preformationism:

[t]he factorial hypothesis does not assume that any one factor produces a particular character directly and by itself, but only that a character in one organism may differ from a character in another because the sets of factors in the two organisms have one difference. This point is not likely to be misunderstood by any one who grasps the meaning of the factorial hypothesis. (Morgan et al. 1915, 211–2)

Ambi-valence analyzed

By 1915, then, we find, among elite practitioners of Mendelian genetics, exasperated acknowledgement that the Mendelian gene had all too often been mistaken, even by geneticists, for a simple character-maker rather than a complex difference-maker. To talk of a ‘gene for a character’ was, for the Morgan group, to use shorthand for a causal relationship manifest only when, against a particular genetic, developmental, and environmental background, differences at a chromosomal locus produce differences in a character. Change something in the background, and you may well change, or even eliminate, the effect on the character. When the chromosomal differences produce character differences so cleanly differentiated as to be binary – red or white eyes in flies, blue or brown eyes in humans – that is an artifact, contrived or accidental, of the system, not an intrinsic property of the allele pair.

It was, and remains, hard to make this more complex gloss on the Mendelian gene stick. 7 Why? An important clue lies, I think, with the nature of the ties binding ‘gene for’ talk to the gene-as-difference-maker notion. Those ties are not, on inspection, straightforwardly those of an abbreviation to what it abbreviates. Abbreviations do not, after all, usually cut loose semantically. No one who understands English is in danger of mistaking, say, ‘USA’ for anything other than the United States of America, or ‘USB’ for anything other than the sort of computer port so labelled (though only experts – along with non-experts who read instruction manuals – will know that the label is an acronym of Universal Serial Bus). Furthermore, abbreviations typically arrive on the scene after what they abbreviate. In the case of the Mendelian gene, however – and as Morgan and company well knew – it was the early critics of Mendelism, among them Morgan, who had insisted that, contrary to Mendelian emphases on the factorial and the binary, the same hereditary factor could have different effects under different internal and external conditions. 8 In The mechanism of Mendelian heredity, the Morgan group were fighting a rearguard action in declaring entrenched genes-as-character-makers talk to be shorthand for genes-as-difference-makers talk.

But if the relationship between the two kinds of talk isn't as intimate as that between an abbreviation and what it abbreviates, neither is it as distant as that between two meanings arbitrarily yoked together under a single term. One of Evelyn Fox Keller's most illuminating reflections on the semantic lives of ‘gene,’ and her own analyses of it, occurs in an endnote in Making sense of life:

My principal claim in The century of the gene was that the term gene has now acquired so many different meanings that its continued usefulness is in doubt, whereas here I argue for the productivity of linguistic imprecision. The question thus arises, when is imprecision productive and when counter-productive? My answer is this: imprecision is productive in the absence of literal meanings, and ceases to be productive either when literal meaning is in manifest conflict with implicit meanings (as happened with ‘gene action’ once the chemical identity of the genetic material was agreed upon) or when two or more different literal meanings have been established (as is now the case today for the gene). (Keller 2002, 321, first note 7)

Returning to the character-maker versus difference-maker understandings of the Mendelian gene, they are, it seems to me, neither metaphoric nor in tension with each other. But Keller's larger claim about ‘gene’ as corralling together different conceptions, in a way that brought cognitive benefits as well as costs, seems to me spot-on, as does her related claim that, whatever the benefits in the past, the costs now outweigh them.

The ‘gene for’ locution is, I submit, straightforwardly shorthand for genes-as-character-makers talk and only unstraightforwardly shorthand for genes-as-difference-makers talk. In principle, becoming educated in genetics involves the displacing of the character-makers conception by the difference-makers conception, even as ‘gene for’ talk remains constant. In practice, however, the situation is much trickier, and not just because some people go further in their genetics educations than others. The very structure of a typical genetics education endows the character-makers conception with independent life, by investing it with heuristic power. Begin your education in genetics with Mendel's peas, and you will learn not merely about a case where, you are told, binary characters are determined by genes for those characters, and by nothing else. You will learn too that many apparently more complicated cases can be made tractable by treating them in the first instance like Mendel's peas. (And if you don't learn that, you won't pass.) 12 Of course you will go on to learn about all sorts of exceptions to your rule of thumb, and the reasons why those exceptions are the way they are: the effects of other genes, epigenetic modifications, the interplay of development and environment, chance. By the end of your education, you will know, of course, that ‘it's not all in the genes,’ and become annoyed with anyone who suggests that you think otherwise. But the Mendelian, treat-‘em-like-the-peas rule of thumb will remain in place. It will guide your reasoning and even – in the way of heuristics – perhaps your unreasoned reflections and reactions too, with much reinforcement from the wider culture in the form of gene-personifying 23-and-Me ads, ‘gene for’ discovery stories, jargon talk of what is in an organization's DNA, and so on. You will affirm genes-for-characters determinism in your actions and attitudes while rejecting it if asked about it, because you know that it's false. 13

‘Again and again,’ writes Theodore Porter in his recent study of the role of insane asylums in the making of the science of heredity, ‘geneticists acknowledged the inadequacy of single-gene explanations in one breath and then proceeded in the next as if heredity could mean nothing else’ (Porter 2018, 321). That was the situation in 1930. And now, in the age of proliferating –omics, including personalized genomics? Consider Siddhartha Mukherjee's bestselling The gene: An intimate history, published in 2016. 14 Over and over again, Mukherjee emphasizes that context determines what a gene determines, in mini-essays on everything from race and IQ to breast cancer. But he also has a contrary and, as the book proceeds, increasingly thumping message. Underlying all this surface complexity and diversity, there is a deep simplicity, just as Mendel found. If a gene is in version A, you get character A; if the gene is in version B, you get character B. In a minor but telling way this dismissiveness of context, and the jarring incoherence it introduces, show up in his mini-essay on the genetics of schizophrenia. Genes are, we learn, part of the causal mix behind schizophrenia, becoming ‘genes for schizophrenia’ only with the collaboration of environment and chance. Schizophrenia itself is not one but many, and so are the relevant genes: over a hundred genomic regions have been implicated, and, according to Mukherjee, ‘no single gene stands out as the sole driver of the risk’ – a risk which may in any case be of enhanced creativity as much as of diminished sanity. But then, in a long footnote added at proof stage, Mukherjee appears to take it all back, reporting excitedly about a then brand-new study that found a single gene strongly associated with schizophrenia: if you have the trouble-making variant, then you will get schizophrenia, other things being equal; and if you don't have the variant, you won't (Mukherjee 2016, 298–300, 441–6, quotation on 445–6).

What the psychologist Steven Heine calls ‘switch thinking’ is persistent under binary Mendelism (Heine 2017, 27). So is category thinking generally. With humans, racial categories have figured prominently from the start. In a paper on eugenics in 1905, Francis Galton wondered whether ‘a study of Eurasians, that is, of the descendants of Hindoo and English parents, might not be advocated … both on its own merits as a topic of national importance and as a test of the applicability of the Mendelian hypotheses to men’ (Galton 1906, 16). When, three years later, the German anatomist Eugen Fischer joined an expedition to what is today Namibia, it was to check for Mendelian patterns in the families sprung from Boer men and Hottentot women. Fischer went on to become the boss of Nazi racial anthropology, while Mendelism went on to lend crucial legitimatory lustre to Nazi racial laws and policies. 15 Even today, when official biology repudiates the reality of race, the categories – and the homogenizing they encourage (this race goes with these genes, that race with those genes) – cling on, not least in medical training. A child shows up in your clinic with fluid in the lungs. She is not white, so it can't be cystic fibrosis, right? Maybe you slow down and interrogate that conclusion. Or maybe you don't. In Angela Saini's Superior: The return of race science (2019), we learn about a black child whose cystic fibrosis went undiagnosed for years until an X-ray of her chest happened to be seen by a radiologist who didn't know what colour she was (Saini 2019, 259).

What is to be done?

The question is, of course, Lenin's. But it is also Keller's, used as the title of her final chapter in The mirage of a space between nature and nurture (2010). There she recommended greater attention to trait malleability, to the functional roles of non-coding as well as of coding DNA, and to inheritance beyond the genome as helpful ways of dispelling that mirage (Keller 2010, ch. 4). Amen, I say. But, given my analysis above of the heuristic power of the character-making understanding of the Mendelian gene, and the insulation of that understanding from ordinary criticism, I think a more fundamental change is called for.

As we saw, the character-making understanding arrived before the difference-making understanding: a chronology, I suggested, which counts against the Morgan group's claim that the former is shorthand for the latter. But what if the order had been reversed? What if, from the start, it had been not the critics of Mendelism – the Oxford biologist W. F. R. Weldon most incisively and implacably, but also Morgan, and others – who stressed the role of context in shaping the effects that a gametically inherited something can have, but the Mendelians themselves? So that every presentation of Mendel's results by his followers began with discussions of, for example, the variability of inherited characters as typically spanning a wide spectrum, except where selection, by nature or by humans, narrows it; the role of nuclear and cytoplasmic contexts (‘ancestry’), inherited within a lineage along with individual factors, in modifying the effects of those factors on character variability; the comparable modifying role of wider physical and chemical environments; and the way that Mendel, for his purposes, deliberately tried to minimize all of those influences, the better to detect the character difference when he introduced a gametic difference. Then, I think, anyone listening or reading would have at least the possibility of emerging from acquaintance with the case of Mendel's peas without taking the character-making understanding, as a permanent, immutable heuristic, with them.

In a limited way, I've actually done the experiment, in collaboration with Dr Jenny Lewis and Dr Annie Jamieson. In the autumn of 2013 at the University of Leeds, we ran an elective introduction-to-genetics module organized as if it had come from the Weldonian past that never was: the past in which Weldon lived to complete and publish the amazing Theory of Inheritance manuscript left unfinished at his death in 1906. In this counterfactual past, there was Mendel, for sure, but no Mendelism. Thus our module opened with just those emphases on contexts and variability just described, and hammered away at them throughout, so that students saw them not as complications to the main story but as the main story in its own right. Before and after teaching, we set the students – who were doing non-science degrees – a questionnaire to determine their levels of genetic determinism. We did the same with a group of biology students taking the standard, start-with-Mendel introductory module in genetics. What we found was that, whereas students taking the Mendelian module were on average just as determinist about genes at the end as they were at the beginning (suggesting that, as some teachers of genetics fear, all their qualifying ‘it's not all in the genes’ pleading goes in one ear and out the other), students taking the Weldonian module were on average less determinist, and to a statistically significant degree. 16

Between the small size of our study (28 students in each group) and the large number of incomparabilities between the two modules as well as between the two student groups, our findings can be no more than suggestive. Happily, however, the Weldonian curriculum is now being put through its paces much more robustly, within a new, three-year study now underway in collaboration with the education psychologist Brian Donovan, the biologist-educationalist Michelle Smith, and over 50 genetics teachers at colleges and universities throughout the USA. 17 The aim is to examine, through a combination of randomized controlled trials and cognitive ‘think alouds’ (in which students talk through their reasoning processes), how different ways of learning about multifactorial genetics affect not only levels of genetic determinism but willingness to challenge racial and gender prejudice, on the hypothesis that the more fully students grasp trait malleability, the less prone they will be to essentialism and the complacency it engenders. So that is to come. For now, it is pleasing to think that an old debate, counterfactually investigated, is helping to create a new kind of scientific education, factually investigated, and maybe even to create students who will someday wonder how anyone could ever have seen a space between nature and nurture.