Experimenters' Regress

The Experimenters' Regress is an epistemic problem identified by Harry Collins and Trevor Pinch in their study of scientific practice. It describes the circularity that arises when scientists attempt to validate an experimental result: the result is considered correct if the apparatus is functioning properly, but the apparatus is considered to be functioning properly only if it produces the "correct" result. This circularity means that there is no independent criterion for validating frontier experiments — the validation must be achieved through social negotiation and consensus within the scientific community.

The experimenters' regress is not a philosophical puzzle but an empirical observation. Collins and Pinch documented cases in which competent scientists, using identical apparatus, obtained different results and were unable to resolve the discrepancy by purely technical means. The resolution came not from better instruments but from social processes: the closure of debate, the formation of consensus, and the marginalization of dissenting views.

The concept has been influential in Science and Technology Studies and has been applied to controversies in physics, biology, and medicine. It challenges the traditional philosophy of science by showing that the reliability of experimental knowledge is not guaranteed by method alone but by the social institutions that sustain trust.

The Regress in Practice: Cases and Mechanisms

Collins's original case study was the search for gravitational radiation in the 1970s. Physicists at multiple laboratories reported conflicting results: some detected signals, others did not. The controversy could not be resolved by building better detectors because the criteria for what counted as a genuine gravitational wave were themselves contested. Was the signal real, or was it noise? The answer depended on whether you trusted the detector, and whether you trusted the detector depended on whether it had produced results you already believed. The regress was broken not by a crucial experiment but by the gradual withdrawal of dissenting researchers, the redirection of funding, and the establishment of new detection protocols that everyone agreed to accept.

A similar structure appears in the replication of psychological experiments. When a famous priming study fails to replicate, the original authors may argue that the replication was poorly executed — that the experimental conditions, the population, or the subtle interpersonal dynamics were not properly reproduced. The replicators respond that they followed the published protocol. The original authors counter that the protocol omitted tacit knowledge essential to the effect. The dispute cannot be settled by appealing to the protocol because the protocol is precisely what is in question. This is the experimenters' regress operating in a different domain.

Calibration, Standardization, and the Social Resolution of Regress

The experimenters' regress is not an argument that science is arbitrary. It is an argument that the resolution of experimental disputes requires resources that are not themselves experimental. The most important of these resources is calibration — the use of known standards to establish that instruments are working properly. But calibration itself depends on prior knowledge of what the standard should read, which introduces a smaller-scale regress. A voltmeter is calibrated against a standard cell, but the standard cell's voltage is known only through prior measurements with other instruments.

The practical solution to this smaller regress is standardization — the social agreement to treat certain measurements as fixed, not because they are independently known to be correct but because treating them as fixed enables coordinated action. The kilogram, until its 2019 redefinition, was a physical artifact stored in a vault in Paris. Its mass was "known" not because it was measured against a more fundamental standard but because the international scientific community agreed to treat it as the definition of a kilogram. The experimenters' regress is halted not by reaching bedrock but by social convention.

The Regress and the Replication Crisis

The experimenters' regress has gained renewed attention in the context of the replication crisis across the sciences. High-profile failures to replicate published results in psychology, medicine, and economics have prompted soul-searching about methodological standards. But the replication crisis is not merely a problem of sloppy methods. It is a manifestation of the experimenters' regress at scale: when results cannot be replicated, the dispute over whether the original or the replication is correct cannot be settled by running more experiments, because each experiment is itself subject to the regress.

The proposed solutions — preregistration of hypotheses, larger sample sizes, statistical reform — are attempts to change the social conventions that halt the regress. They do not solve the regress technically; they establish new norms for when a result should be considered trustworthy. This is not a criticism of these reforms. It is a recognition that they are social interventions, not technical fixes.

The Regress in Computational Science

A variant of the experimenters' regress has emerged in computational science and simulation-based research. In climate modeling, for example, a model's predictions are validated against historical data — but the model is also tuned to fit that data. If the model fails to match observations, is the model wrong, or are the observations uncertain? The question cannot be answered by running the model again, because the model's parameters have already been adjusted to match the data. The regress is resolved not by finding independent validation data (which rarely exists for long-term climate predictions) but by the emergence of consensus around ensemble methods and multi-model averaging — social procedures for managing uncertainty rather than eliminating it.

The Deeper Structure: Why Regress is Inevitable

The experimenters' regress is not a pathology of bad science. It is a structural feature of any knowledge-producing practice that operates at the frontier of what is known. At the frontier, by definition, there is no independent criterion of correctness. The criterion must be constructed, and construction is a social process. The regress reveals that scientific knowledge is not a pyramid resting on indubitable foundations but a web held in tension by mutual support among its strands. Cut one strand, and the web adjusts. Cut too many, and it collapses.

The experimenters' regress is often treated as a debunking of science — as if showing that science is social were the same as showing that it is false. This is the opposite of the truth. The regress reveals that science is a social achievement of extraordinary difficulty, and that the institutions that sustain it deserve more study, not less. The critics who fear that constructivism undermines science are confusing description with debunking — a confusion that itself reveals how little they understand about what science actually is. But there is a harder truth that even constructivists sometimes avoid: the social resolution of the regress is not neutral. It rewards power, prestige, and funding concentration. The experimenters' regress is not merely an epistemic problem. It is a political one, and the politics of who gets to declare an experiment closed is the politics of who gets to speak for nature.