6 Testing Theories and Falsification

Predictions and empirical tests

Scientific theories are tested by deriving observable consequences and comparing their predictions with evidence. A test may challenge a theory, support it, or leave the issue unresolved. No finite set of successful tests usually proves a broad theory true; instead, evidence changes how strongly we are warranted in accepting it.

Falsifiability and corroboration

In Popper’s account, a scientific theory should be falsifiable: there must be conceivable evidence that would count against it. For example, the claim “all swans are white” conflicts with the observation of a single reliably identified non-white swan. A claim compatible with every possible observation makes no risky empirical prediction.

An unexpected result does not automatically settle a theory’s fate. Observations can be mistaken, measurements can fail, and predictions depend on additional assumptions. Theories that withstand serious attempts to refute them are corroborated, not conclusively verified.

Confirmation and evidence

Evidence confirms a hypothesis when it provides support for it, but confirmation is not proof. In a common hypothetico-deductive approach, scientists derive a prediction from a hypothesis together with background assumptions, then ask whether the observed evidence matches that prediction.

A successful prediction can support a hypothesis, especially when the test is demanding and competing explanations make different predictions. For example, a hypothesis about a medicine’s effect might predict better outcomes for treated patients than for a comparable control group. If a carefully designed study finds that difference, the result supports the hypothesis. Researchers would still consider replication, possible biases, and whether another explanation could produce the same result.

Auxiliary assumptions and failed predictions

A theory rarely predicts an observation by itself. Tests commonly rely on auxiliary assumptions about the setup, instruments, initial conditions, and methods used to interpret results. If a prediction fails, the evidence challenges the combined set of assumptions, but it may not show which part is at fault: the central theory might be wrong, an instrument might have malfunctioned, or an assumption about experimental conditions might have been mistaken.

This is the Duhem–Quine problem: hypotheses are tested in groups rather than in isolation. Scientists can respond by checking instruments and procedures, testing auxiliary assumptions independently where possible, repeating experiments, and comparing alternative explanations. Revising an auxiliary assumption can be reasonable when there is independent evidence for the revision. Changing assumptions solely to protect a favored theory from every contrary result risks making the theory immune to testing.

Underdetermination and theory choice

Underdetermination occurs when available observations do not uniquely determine which explanation is correct. The alternatives may make the same predictions so far, or different theories may be made to fit the evidence using different background assumptions. New tests may help distinguish them, but sometimes further evidence is not yet available.

When choosing among theories, scientists can consider how well each fits the evidence, whether it makes new and discriminating predictions, how well-tested its assumptions are, and how coherently it connects with established knowledge. These considerations guide judgment but do not guarantee that one theory is true. A useful theory should remain open to further testing and revision.