Essay

Why scientific evidence changes institutions slowly

A field can be shown to rest on weak foundations and carry on much as before. Why does decisive evidence have so little immediate effect?

A persistent belief about science holds that decisive evidence has decisive consequences. Demonstrate that a prominent theory is false, expose widespread methodological failures, or even uncover outright fraud across an influential field, and the expectation is that researchers will abandon old conclusions and rebuild on firmer ground. Reality is less accommodating. Entire bodies of scientific literature can be shown to rest on weak foundations without producing changes proportional to the force of the criticism. Evidence changes minds more readily than it changes institutions.

Part of the difficulty lies in mistaking science for a single decision-making body. There is no central authority that can declare a discipline corrected and move everyone forward. Science consists of universities, journals, funding agencies, professional societies, laboratories, companies, regulators, and thousands of individual careers. Every published finding is connected to grants, promotions, doctoral theses, textbooks, conference presentations, and reputations accumulated over years or decades. Challenging a foundational claim therefore threatens an idea and the network of incentives built around it.

The replication crisis in psychology illustrates the pattern. During the past fifteen years, researchers attempting to reproduce influential experiments found that many could not be replicated with similar methods. Some celebrated effects appeared substantially weaker than originally reported; others disappeared entirely. The resulting reforms were genuine and important. Larger sample sizes became more common. Pre-registration spread rapidly. Journals increasingly accepted registered reports in which studies were evaluated before results were known. Data and code became more openly available. Even so, many of the original claims remain in textbooks, popular writing, and secondary research. The evidence changed methodology more quickly than it changed accumulated belief.

Nutrition science presents a similar story. Decades of dietary advice often rested on observational studies capable of detecting correlations but poorly equipped to establish causation. Researchers recognized publication bias, multiple comparisons, unreliable dietary questionnaires, and the tendency of statistically significant results to receive disproportionate attention. Critics demonstrated that many confident nutritional claims rested on fragile evidence. Yet public guidance shifted only gradually, partly because replacing old recommendations requires stronger evidence than creating them. Institutions are better at producing hypotheses than retiring them.

Medicine offers the clearest examples because patients provide an unforgiving test of scientific ideas. Numerous treatments once regarded as standard practice later proved ineffective or harmful when subjected to large randomized trials. Hormone replacement therapy was once widely believed to protect against cardiovascular disease. Certain antiarrhythmic drugs administered after heart attacks seemed physiologically sensible before careful trials revealed increased mortality. Arthroscopic surgery for some degenerative knee conditions survived years of routine use before trials showed little advantage over sham surgery or over physical therapy alone. None of these reversals occurred overnight. Physicians, hospitals, insurers, and guideline committees all required time to absorb conclusions that directly contradicted established practice.

Actual fraud complicates matters further but does not alter the pattern. Scientific fraud is less common than scientific error, yet its consequences can be profound when fraudulent papers become influential. Fabricated data may inspire hundreds of follow-up studies, shape funding priorities, or influence clinical practice before detection. Retraction removes a paper from the formal literature but rarely erases its intellectual descendants. Retracted studies continue to receive citations years later, often because authors rely on earlier reviews rather than examining primary sources. Once ideas become embedded within a research tradition, removing the original support does not automatically dismantle the structure built on it.

Thomas Kuhn’s account of scientific revolutions remains valuable because it emphasized institutional behavior rather than abstract rationality. Established paradigms do not disappear the moment contradictory evidence appears. Researchers initially reinterpret anomalies, question experimental methods, introduce auxiliary hypotheses, or treat conflicting findings as exceptional cases. Only when alternative frameworks become sufficiently compelling—and often when younger scientists build careers around them—does a genuine transition occur. Kuhn exaggerated the discontinuity of these revolutions, but he correctly recognized that science advances through communities of people rather than through logic alone.

None of this implies that science is uniquely corrupt or intellectually dishonest. The same dynamics appear in every complex institution. Businesses continue investing in failing products because abandoning them requires admitting previous mistakes. Governments preserve ineffective programs because constituencies have formed around them. Universities continue teaching superseded theories because curricula change slowly. Human beings rarely update beliefs in isolation from the social structures that reward those beliefs.

What is unusual about science is that correction happens at all. Most institutions have mechanisms for preserving existing authority. Science also contains mechanisms that eventually undermine it. Independent replication, competing laboratories, technological advances, improved statistical methods, and practical application continually expose weaknesses that rhetoric alone cannot conceal. A false theory may survive criticism for years, but it struggles to produce reliable engineering, effective medicines, or accurate predictions indefinitely.

The distinction between eventual correction and immediate response matters. People often describe science as self-correcting, a phrase that invites misunderstanding. It suggests an automatic process by which false ideas are rapidly discarded once contrary evidence appears. The actual process resembles evolution more than computation. Competing ideas persist, institutions adapt unevenly, and selection occurs through repeated confrontation with reality rather than through single decisive moments. Errors accumulate, criticism accumulates, reforms accumulate, and only after sufficient pressure does the balance shift.

Evidence must interact with incentives, generational turnover, methodological reform, funding priorities, and repeated practical failures before it produces durable transformation. Exposing an entire field’s weaknesses can therefore seem strangely inconsequential in the short term, because institutions have an inertia that evidence alone cannot overcome. The correction eventually comes, if it comes at all, through the slow convergence of intellectual argument and social change. Scientific knowledge advances by discovering better explanations, but scientific communities advance only when enough people find it worthwhile to abandon the old ones.

Sources: