[Google AI said:] “Yes, scientific theories are based on hypotheses. But a theory is a well-substantiated explanation that has been extensively tested and supported by a vast amount of evidence, while a hypothesis is a tentative, testable explanation that serves as a starting point for an investigation. Essentially, a hypothesis is the precursor to a theory; it's an educated guess that, if repeatedly supported by experiments, can evolve into a full-fledged theory.”
As I wrote elsewhere (
2025, Nov 22), describing the differences between law, theory, and hypothesis:
In science, a law is a generalization about empirical data that seeks to describe the regular and consistent patterns and relationships that are found. But these are descriptions, which are not explanations. That is the role of a theory, a concise unifying conceptual structure that ties together and explains observed and predicted empirical phenomena (causes, forces, etc.) and their relationships. Theories also encompass and integrate many different hypotheses, which are limited explanations of more narrow sets of phenomena. [1] Here is an example that should help to illustrate the difference between each term and their relationships to one another: A hypothesis about graviton particles might serve a general theory of gravitation proposed to explain the law of gravity.
And that was not informed by Google AI, it was informed by: Ernan McMullin, “Hypothesis,” in Wilbur Applebaum, ed.,
Encyclopedia of the Scientific Revolution From Copernicus to Newton (New York: Garland, 2000), 315–318; “Theory,”
ibid., 641–643; Helen Hattab, “Laws of Nature,”
ibid., 354–357; Carl G. Hempel,
Philosophy of Natural Science (Prentice-Hall, 1966);
New Encyclopedia Britannica, 15th ed., s.v. “Theory”;
APA Dictionary of Psychology, s.v. “Hypothesis,” “Theory,” and “Law.”
Google AI was wrong—unsurprisingly. Scientific theories consist of hypotheses but are not structurally dependent on them. Hypotheses like “gravity is mediated by gravitons” are theory‑guided attempts to further specify the mechanism or to extend the theory (into a quantum domain in this case), and can be rejected or revised while the core gravitational theory remains in place.
A theory does not depend upon but rather exists independent of any particular hypothesis. A theory can survive the falsification of any number of its auxiliary hypotheses. That is precisely because its structure is not reducible to them. A hypothesis is the basis not of a theory but of an
argument. A clue is found in its etymology. By the time the Greek word (
hypóthesis) enters late Latin (
hypothesis) and then Early Modern English, it retains this sense of “a proposition assumed for the sake of inquiry.”
Google AI is also wrong about theories being “extensively tested and supported by a vast amount of evidence.” A theory remains a theory regardless of the amount of testing it has endured. (But the degree of testing does determine its epistemic authority and value, just not its classification.) Cell theory? Extensively tested. String theory? Barely tested at all. Theories are explanatory frameworks that attempt to make sense of the data.
And it was wrong when it said that “a hypothesis is the precursor to a theory.” A hypothesis is a specific, testable proposal addressing a narrow question. A theory is a broad explanatory framework that integrates many hypotheses, datasets, and lines of evidence. They are not the same thing at different stages of development. A hypothesis is not a fledgling theory waiting to mature. A hypothesis is a narrow, limited explanation; a theory is a broad, comprehensive explanation.
A hypothesis is also not an educated “guess.” It is a well-structured, testable proposition derived from existing knowledge and serving a larger explanatory framework (theory), like a hypothesis about a disease treatment within the larger germ theory. The clue is in the very term itself, hypothesis, wherein the prefix hypo- means under. Calling it a guess encourages the false impression that science begins in speculation rather than disciplined inference.
Perplexity AI provided an accurate explanation: “A theory is a broad, structured explanatory framework that unifies many phenomena (e.g., general relativity as a theory of gravitation). From such a theory, scientists derive specific, testable hypotheses—particular claims or predictions that can be checked by observation or experiment (e.g., ‘light passing near the sun will be deflected by such‑and‑such amount under these conditions’).”
