Ideas & History · Science, gender and public power
Ruth Hubbard: what an experiment can fail to see
Repeating an experiment can expose an error. It can also repeat the assumptions that made the error difficult to see. Ruth Hubbard’s challenge is to distinguish evidence from the social story built around it.
A reproducible observation and a justified explanation are different achievements. Researchers may agree about what an instrument records while disagreeing about what it measures. They may also reproduce the same mistake if they share an assumption that nobody tests. Ruth Hubbard’s feminist criticism of science begins at this awkward point: agreement within a profession is valuable, but it is not a guarantee that the profession has asked the right question.
Objectivity has a history
In an essay collected in Profitable Promises in 1995, Hubbard asked whether a science restructured by feminism would alter gravity. She distinguished the existence of the world from historically formed accounts of it. Apples would still fall; that did not make every feature of scientific explanation inevitable. She accepted that carefully defined experiments could reduce bias. Her objection was that this local discipline could leave commitments shared across a scientific community unexamined. [2]
The distinction is more demanding than saying that everyone has opinions. A source of error must be located: in the question, the category, the measurement, the comparison or the inference. Once located, it can be investigated. Otherwise a criticism of scientific authority risks becoming an unlimited suspicion that explains everything after the event and predicts nothing in advance. The political task is to make an institution more capable of detecting its errors, including errors congenial to its critics.
Hubbard also resisted an account of feminist insight based only on individual empathy. Discussing women’s reinterpretations of primate behaviour, she asked how professional and cultural conditions shaped observation. Her argument concerned the organisation of inquiry, not an assurance that women possessed an automatically correct view of nature. [2]
That distinction changes what inclusion requires. Appointing a more diverse research team can expand the questions people recognise, but it cannot ensure that an unfashionable question receives time, equipment or a hearing. A junior researcher may notice a problem and still lack the authority to reopen a settled result. A serious institutional response must therefore protect criticism and make the chain from observation to conclusion inspectable. Representation matters partly because it can change what becomes contestable, rather than because demographic categories confer scientific infallibility.
Counting offspring, inferring behaviour
A later dispute over Angus Bateman’s 1948 fruit-fly experiment makes the problem concrete. Bateman used visible inherited mutations to identify offspring’s parents. Offspring displaying markers from both parents allowed him to infer which pairs had reproduced. He recognised that these counts could miss matings and could not distinguish repeated inseminations by the same pair. Nevertheless, he drew broad conclusions about differences between the sexes, linking greater male reproductive variation to stronger selection through competition for mates. [3]
His paper moved from reproductive measurements to a general account of male eagerness and female passivity or discrimination. That movement matters because a count of descendants and a description of motivation answer different questions. Even a correctly measured association requires further reasoning before it can explain the behaviour that produced it. [3]
In 2012, Patricia Adair Gowaty, Yong-Kyu Kim and Wyatt Anderson repeated the marker-based approach. Under the relevant inheritance assumptions, a quarter of offspring should have displayed both parents’ markers. Only about 15 per cent did so across their replicated populations. The mutations affected survival, making some reproductive pairings disappear from the inferred record. Similar deficits occurred when the researchers confined adults to monogamous pairs, where competition among alternative mates could not explain them. [4]
The conclusion requires care. The authors identified an unreliable method for inferring mating success; their repetition could not establish whether sexual selection was operating in those populations. They also acknowledged uncertainty about changes in mutant lines across six decades. A failed measurement does not demonstrate the absence of the phenomenon it was intended to measure. [4]
This is a strong scientific criticism because it specifies a failure that could be checked. It does not depend on declaring Bateman malicious or on substituting one political identity for another. The relevant disagreement concerns how an observable sign becomes evidence for an unobserved event. That is also why the result cannot bear the opposite ideological conclusion that sex differences are always imaginary. An argument against an inference must retain its proper target. Widening it beyond the evidence would reproduce the error being criticised.
A difficult result for every simple story
Later evidence complicates any easy verdict. A 2016 meta-analysis by Tim Janicke and colleagues combined 72 studies covering 66 animal species, accounting for shared evolutionary history and repeated estimates from the same species. It excluded Bateman’s original experiment. On average, males showed greater variation in reproductive success and a steeper relationship between reproductive success and number of mates. The overall sex difference in variation in mating success was not statistically significant. [5]
The authors interpreted the findings as support for stronger sexual selection on males across the sampled taxa, while recognising exceptions and variation within species. They also distinguished variance measures, which indicate opportunities for selection, from stronger claims about selection itself. Their conclusion concerned patterns in polygamous animals, not a universal account of human conduct. [5]
The scope remains contested. In a review published online in December 2025, Joan Roughgarden argued that the sampled species and reproductive statistics could not establish the sweeping behavioural account implied by the study’s title. Her objection targets the passage from a comparative pattern to a general theory of sex roles. It does not erase the reported averages. [6]
These disagreements should be kept distinct. Whether a marker works, whether an association recurs across species, and whether that association explains behaviour are separate empirical and theoretical questions. Answering one does not dispose of the others. A responsible reader can accept evidence for an average difference while challenging the proposed mechanism or the breadth of the generalisation. Equally, a critic must acknowledge evidence that survives a better design. Political seriousness demands this precision because arguments that rely on convenient ignorance are vulnerable to the next competent reader.
There is a further boundary between explanation and entitlement. An account of how a trait evolved cannot, by itself, establish who should govern a university, perform unpaid care or control a household’s income. Such decisions require arguments about rights, needs, institutions and available alternatives. Defending equality should not depend on proving that every measurable difference is absent. That would make a political right hostage to a changing empirical literature and concede the mistaken premise that a description of nature supplies a social command.
A democratic science must remain corrigible
Hubbard’s essay ended by demanding democratic participation in the choice of research priorities and stronger relationships between universities and communities, including trade unions. She wanted scientific skills to help people investigate problems arising in their own lives. [2]
The serious objection is that research directed by immediate public preferences could neglect discoveries whose uses are unforeseeable. It could also replace professional orthodoxy with political orthodoxy. Those risks require institutional safeguards, not a return to unanswerable expertise. A defensible arrangement would make priorities and funding contestable while protecting researchers’ ability to report unwelcome results. It would allow disagreement about a project’s purpose without authorising anyone to decide its findings in advance. Public accountability and intellectual freedom should constrain one another productively.
For a worker or community organisation, participation becomes meaningful when it can alter the question, inspect the evidence and challenge an interpretation. Supplying samples or appearing at a launch is not enough. For researchers, the corresponding obligation is to explain where a finding stops and an extrapolation begins. Hubbard’s most durable challenge lies there: make scientific authority answerable at the points where assumptions enter, and preserve the evidence’s capacity to surprise everyone involved.
Sources and further reading
- Harvard Faculty of Arts and Sciences, Ruth Hubbard Wald memorial minute (opens in a new tab) Adopted 7 March 2017; published 8 March.
- Ruth Hubbard, In a Science Restructured Along Feminist Lines, Would the Laws of Gravity No Longer Hold? (opens in a new tab) From Profitable Promises (1995), preserved by Barnard’s Scholar & Feminist Online.
- A. J. Bateman, Intra-sexual selection in Drosophila (opens in a new tab) Heredity 2 (1948), pp. 349–368.
- Patricia Adair Gowaty, Yong-Kyu Kim and Wyatt W. Anderson, No evidence of sexual selection in a repetition of Bateman’s classic study of Drosophila melanogaster (opens in a new tab) PNAS 109 (2012), pp. 11740–11745.
- Tim Janicke and colleagues, Darwinian sex roles confirmed across the animal kingdom (opens in a new tab) Science Advances 2 (2016), e1500983.
- Joan Roughgarden, Explaining sex: contrast between social and sexual selection (opens in a new tab) Published online 31 December 2025; Biology of Sex Differences 17 (2026), article 12.
