Science provides a particularly rigorous model of inquiry. The scientific method is a systematic cycle: making observations, forming hypotheses, conducting experiments, analyzing results, and drawing conclusions. A hypothesis is a testable proposed explanation, while a theory is a well-substantiated explanation supported by extensive evidence. According to Popper, the criterion of falsifiability distinguishes science from non-science: a theory is scientific only if it makes predictions that could, in principle, be shown false. Operational definitions specify precisely how a concept is to be measured, allowing others to apply them consistently. A measure is valid to the extent that it assesses what it claims to assess, and reliable to the extent that it produces consistent results across instances. Accuracy refers to closeness to the true value, while precision refers to repeatability of measurements; the two can come apart.
Scientific work is sustained by communal practices designed to catch error. Peer review subjects scholarly work to expert evaluation before publication. Reproducibility means that others can obtain the same results using the same data and methods, while replication refers to obtaining similar results via a new study. The replication crisis is the troubling finding that many published results cannot be reliably replicated, prompting reforms in research methods. Good science also disciplines itself through standards such as Sagan's standard—that extraordinary claims require extraordinary evidence—and through skeptical maxims like Hitchens's razor: what can be asserted without evidence can be dismissed without evidence.
The burden of proof is the obligation to provide evidence for a claim, typically on the asserter; burden of disproof, by contrast, is usually inappropriate because it asks doubters to prove a negative. Russell's teapot analogy illustrates that the unfalsifiability of a claim does not make it credible. Several patterns distort scientific and quasi-scientific reasoning. Pseudoscience presents claims as scientific while lacking proper methodology or evidence. Conspiracy thinking explains events as the result of secret, malevolent plots. Cargo cult science, a phrase from Feynman, mimics the form of science without its substance or rigor. Data dredging, also called p-hacking, repeatedly tests data until a statistically significant result emerges by chance. Overfitting builds explanations too tightly tailored to limited evidence, capturing noise rather than signal. The principle of parsimony, also known as Occam's razor, recommends preferring simpler explanations when all else is equal. Together, these principles and pitfalls form a toolkit for evaluating claims that purport to be empirical, and the principle of charity—interpreting others' arguments in their strongest reasonable form, or steelmanning them—encourages fair engagement with opposing views.