How does a scientific account earn trust?

What would make us reconsider?

Imagine a greenhouse team that blames its young plants’ slow growth on a shortage of light. The team moves some trays nearer a window, and they grow faster. Has it established its explanation, or only found a result that several explanations could fit? The window is also warmer.

With this lesson the course turns from periods to problems: each lesson from here on brings approaches met earlier to bear on one particular question. Hume questioned induction, the reasoning from observed cases to unobserved ones, and pragmatist inquiry asked how beliefs can be tested and corrected. Here we compare selected twentieth-century arguments by Karl Popper and Thomas Kuhn, then turn to Heather Douglas’s later work on values. They describe scientific judgment in different ways, and none of them offers a recipe for every science.

A claim that can lose

Popper asks what distinguishes an empirical scientific theory from a claim that can accommodate anything. His proposed criterion is falsifiability: the theory rules out some possible observations. Being falsifiable does not mean being false; it means running the risk of a conflict with evidence. A theory that passes a demanding test earns what Popper calls corroboration, though passing does not prove it true. [1]

For a stronger test, the team could use lamps to vary the light while keeping every tray at the same temperature. Before looking at the results, it should say what difference its explanation predicts and what outcome would count against that prediction. Faster growth beside a window tells us less, because both light and warmth have changed.

Popper objects to collecting agreeable examples while making every unwelcome outcome fit. If the plants grow quickly, light helped; if they grow slowly, the light must have helped invisibly. An explanation protected in this way can no longer be tested. Even so, testability alone does not make an account trustworthy, and Popper does not treat every claim outside science as meaningless. [1]

When a result goes wrong

Suppose the predicted difference fails to appear. The test relied on more than a sentence about light: it also assumed comparable plants, working lamps and trustworthy measurements. If those assumptions and the explanation together implied the prediction, the failure shows that something in the combination needs attention. It does not, by itself, say which part.

This is the kind of inference that logical analysis helps us examine. Checking a faulty lamp can improve a test; inventing an undetectable fault whenever a result disappoints only shields the explanation. What separates the two is whether the proposed repair can itself be investigated. One surprising result therefore calls for further work rather than an automatic verdict on a whole theory.

Why scientists keep working

Kuhn’s The Structure of Scientific Revolutions, first published in 1962, asks how scientific communities develop and change. Normal science works within established achievements, using them to solve further problems. A paradigm, in one important sense, is an exemplary achievement that guides such work; Kuhn also used the word more broadly for shared commitments. Either way, it is more specific than a personal outlook. [2] [3]

An anomaly is a finding or difficulty that the established approach struggles to accommodate. In Kuhn’s account, scientists commonly work on such difficulties without abandoning their framework. A crisis may eventually bring an alternative into competition with it, and judging between them means asking how each relates to the evidence and to the other. Merely noticing an anomaly does not do that comparing. [2]

Our greenhouse team might revise a measurement or an assumption while keeping its wider understanding of plant growth. That shows the difference between solving a difficulty and replacing a framework, though it is hardly a historical scientific revolution in miniature.

Change without arbitrariness

Popper and Kuhn disagree about what makes inquiry productive. Popper emphasizes attempts to refute a theory; Kuhn emphasizes the sustained work that shared commitments make possible. Does Kuhn, then, make choosing between frameworks a matter of allegiance? As Alexander Bird explains, Kuhn identified shared standards: accuracy, consistency, scope, simplicity and fruitfulness for further inquiry. Scientists can disagree about how to apply them or how much each matters. An account that fits more observations may be harder to reconcile with another accepted explanation, and no agreed formula need settle that tension. [3]

The absence of a formula does not make the reasons disappear. Kuhn’s later clarification treats shared values as able to guide inquiry even when they do not produce unanimous choices. Whether his account adequately explains scientific rationality remains disputed. Historical change alone, though, establishes neither that later theories must be true nor that any story is as good as another. [2] [3]

Values enter the judgment

Accuracy is a value concerned with knowledge. Ethical concerns raise a different question: what happens if a scientific judgment is wrong? Douglas draws attention to inductive risk, the possibility of error in accepting or rejecting a claim. When errors have foreseeable practical consequences, she argues, their seriousness should affect how much uncertainty scientists accept. Values enter through that judgment about risk; wanting a conclusion cannot supply evidence that it is true. [4]

Suppose the team will use these results to decide whether the plants can withstand an expensive move to a cooler site. Judging them ready too soon could lose a season’s plants; wrongly judging them unready could waste heating and delay planting. Those consequences give the team reason to scrutinize the evidence and to explain its uncertainty. They cannot change what the trays did.

Douglas’s argument reaches beyond the later use of a result into the choice of methods and of interpretations. It is a defended position about responsible inquiry, not an uncontested rule for weighing every risk. [4]

Try the prediction

The team announces: “The trays near the window grew faster, so light alone explains their growth. If another trial fails, its equipment must be defective. Besides, theories change, so any explanation is acceptable.” Identify the three gaps in this reasoning. Then explain what the possible loss of the plants should change, and what it should not. This exercise is optional.

Read a possible response

The first result cannot tell light from warmth. A stronger comparison would separate the two and state in advance what outcome would count against the explanation. The second claim turns a possible equipment problem into a guaranteed excuse. Inspecting a lamp is worthwhile because it could reveal either a fault or evidence that the lamp worked.

The third claim confuses changing explanations with having no grounds for choosing between them. We can still compare how well rival accounts handle the evidence, where they conflict and what further work they make possible, even when the judgment remains difficult.

The possible loss gives the team reason to decide how much uncertainty it is willing to accept, and to explain that choice. It gives no reason to hide a result or to call an uncertain finding certain. Who should bear the costs is a further question, which the experiment alone cannot settle.

Carry this forward

Ask what an account risks getting wrong, how its failures are investigated and how its judgments of uncertainty are defended. These questions help separate earned confidence from mere insistence. They also leave a question open: when inquiry depends on other people, whose contributions are heard and taken seriously? The next lesson, Who is heard, and who is treated as credible?, examines credibility and testimony.

Sources and limits

[1] Karl R. Popper, “Science as Falsification”, a teaching excerpt identified as drawn from Conjectures and Refutations (1963), for falsifiability, corroboration and the objection to explanations that fit every outcome. Read the opening discussion through the seven numbered conclusions, section II through its concluding account of demarcation, and note 3 on setting criteria for a test. The excerpt’s display title is retained here. Neither a printed edition nor the accuracy of the transcription was independently checked. Popper’s own examples involving psychoanalysis, Marxism and the eclipse are not assessed here or adopted as this lesson’s cases.

[2] Thomas S. Kuhn, The Structure of Scientific Revolutions, second edition, enlarged (University of Chicago Press, 1970; first edition 1962), for normal science, paradigms, anomalies and shared values.

• Normal science and paradigms: chapter II, pp. 10–11.
• Anomaly and crisis: chapter VIII, pp. 77–80.
• Shared values: the postscript’s discussion, from “A third sort of element” on p. 184 through the paragraph ending “success of its enterprise” on p. 186, including notes 9–10.

These are the printed page labels, not PDF page numbers. The linked file is a digital transcription; images of its pp. 77 and 185 were inspected, but no original printing was. No reading of a full chapter or the whole book is claimed.

[3] Alexander Bird, “Thomas Kuhn”, Stanford Encyclopedia of Philosophy, substantive revision 2 September 2025, for later scholarly qualification of paradigms, comparison and theory choice. Read §§2–3, the opening of §4 and §4.1 (the section sign, §, marks a numbered section). Kuhn’s later essays discussed there were not separately read. This is one specialist’s interpretation; it does not settle the debates it describes.

[4] Heather Douglas, “Inductive Risk and Values in Science”, Philosophy of Science 67.4 (2000), pp. 559–579, for inductive risk and the place of values in accepting a claim. Read the abstract and §§1–3, pp. 559–565, and §7, pp. 577–578; p. 564 was also inspected as an image. Sections 4–6 contain empirical case studies that were not reviewed as a whole, and their scientific findings are not evaluated here. The works by Hempel, Jeffrey, McMullin and Longino that Douglas discusses were not independently read.

Sources accessed 20 September 2026. The greenhouse, trays, cooler site and exercise are original teaching material. The comparison selects arguments; it is not a history of all scientific methods. No whole-book reading, independent specialist review or learner testing is claimed.