What is testability?
Before a claim can be taken seriously it has to answer one question: "how could we check this?" If there is no answer, what we hold is not knowledge but merely a statement.
Short answer
Testability is the possibility of deriving concrete predictions from a claim that can be compared with observation or measurement. A testable claim says in advance under what conditions it will count as true and by what success will be measured; it is therefore open both to confirmation and to refutation.
From claim to prediction
The first step in testing a claim is to derive a prediction from it: "if this claim is true, then under these conditions we should observe that." The claim that "the Moon's pull creates the tides", for example, yields the prediction that the times of high water will shift along with the Moon's position; the records confirm it every day. A claim that yields no prediction cannot make contact with observation; it can neither be supported nor refuted. It hangs in the air. The everyday practice of science is largely this work of translation: turning a general idea into a narrow, precise sentence that says what will be seen in a particular observation.
Operational definition: what will success be measured by?
The unsung hero of testability is defining concepts in measurable form; this is called an operational definition. The claim "this drink gives you vitality" cannot be tested until vitality is defined: will reaction time be measured, self-reported tiredness in a questionnaire, or work performance? The measure must be chosen and the threshold set beforehand; the approach of "whatever comes out, we will count it as success" is not a test but writing the result afterwards. A good test also describes from the outset what failure would look like.
What do untestable claims look like?
Some claims escape testing by their very structure. They use vague language: "you may experience an important change soon" finds a match in every life. They fit every outcome: if the expected event happens the claim is confirmed, and if it does not, "the conditions were not ripe". Or they are closed to measurement in principle: what observation could falsify a claim about "an energy that is invisible and undetectable by any instrument"? Such claims are not necessarily false, but they cannot function as knowledge, because they say nothing distinguishing about the world. You can find the continuation of this subject in our guide to falsifiability.
A worked example: testing the retrograde claim
Take the claim "electronic device faults increase during Mercury retrograde"; this is a claim that can be tested properly. First the operational definition: what counts as a fault, and from which records will it be counted? A service centre's logbook could be chosen, for instance. Then the comparison: the daily number of faults during retrograde periods is set against the number during non-retrograde periods, and if the data are counted blind, that is, if the person counting does not know which day falls in a retrograde, the expectancy effect is excluded as well. When comparisons of this kind are carried out, no increase peculiar to retrograde periods is seen; for the astronomical explanation of apparent backward motion, see our guide to Mercury retrograde. The good thing about the example is this: both the claimant and the sceptic can agree in advance to accept the result of the same test.
The triumph of prediction: Halley's comet
One of the most elegant historical illustrations of the power of testability comes from the sky. Edmond Halley applied Newton's theory of gravitation to past records of comets and proposed that the bright comets seen in 1531, 1607 and 1682 were in fact the same object. What made his claim testable was that he tied it to a risky prediction: this object would complete its roughly 76-year orbit and return around 1758. Halley died before that date; the comet appeared in the last days of 1758 and has borne his name ever since. Had the prediction failed, the theory would have taken a serious wound; because it held, confidence in the theory of gravitation multiplied. This is the reward of a testable claim: when it turns out to be right, an argument is not merely won, a reliable instrument of prediction about the world is gained.
Testability is not a value judgement
That something is untestable does not mean it is meaningless or worthless. The beauty of a poem, the peace a belief gives a person, the value of a memory are not subject to the criterion of testability, and need not be. The criterion applies only to statements claiming factual knowledge about the world. The problem is an untestable statement presented in the guise of scientific knowledge; being able to draw that distinction is the core of information literacy.
Frequently asked questions
Are testable and tested the same thing?
No. Testability is a potential: the claim has been framed in a way that lends itself to testing. Having been tested means that the testing has actually been done. Scientific confidence comes from the two together.
Can claims about the past be tested?
Yes. Even where no experiment can be set up, claims about the past leave traces: fossils, ice layers and stars whose light is only now reaching us are all records. A claim is tested by predicting what will be found in those records.
Is one test enough?
Rarely. A single study can be affected by chance, measurement error or conditions peculiar to the sample. Confidence accumulates through replications by independent teams.
Do I have to test everything myself?
No, and it would not be possible. The practical skill is to ask whether a claim has been framed testably and whether it has passed independent tests; those two questions are the quickest filter in assessing a source.
This guide draws on the core concepts of the literature in the philosophy of science and research methods.