Testability

Beginner · 11 min

Three conditions

Testing a claim requires three things. Clear definitions: every key word in the claim must have an agreed meaning. A measurable outcome: what will be counted or observed must be known in advance. A criterion set in advance: which result would support the claim and which would weaken it must be written down before the test begins. The third condition is the one most often skipped — yet if the criterion is set afterwards, almost any result can be bent to fit the claim.

You can think of this trio as a contract: whoever makes the claim and whoever tests it agree on the rules of the game from the start. If the rules are not written down first, then whatever the outcome, one side will say “but that doesn't count”.

Let's transform a claim together

Take a popular example: “Electronic devices break down more often during Mercury retrograde.” As it stands, the sentence is blurry — but it can be made testable:

1. Definitions: Which devices count? What does “break down” mean — a fault needing repair, or a temporary freeze? Which date ranges are the retrograde periods?
2. Measurement: For instance, a repair shop's fault records; retrograde periods are compared with non-retrograde periods of equal length.
3. Criterion: How large a difference would count as “breaking down more often”? That threshold is written down in advance.

Setting this up moves the claim into testable territory; from then on, the verdict belongs to the data, not to opinions. When both the claim's supporter and its critic accept the same criterion, the debate finally becomes productive.

Notice that none of these steps requires a special laboratory; the only thing needed is the discipline of writing the rules before the result.

Why an after-the-fact explanation is not a test

Checking the calendar after your computer breaks and saying “it was retrograde anyway” is not a test, because the explanation was sought after the result was known. The thousands of devices that worked fine that day never entered the count. The power of a test comes from writing down a prediction whose outcome is not yet known. This is why researchers register their hypotheses before collecting data; in everyday life, the same principle can be summed up as “predict first, observe second”.

The same principle applies in the positive direction too: if your phone never broke down during a retrograde, that alone does not refute the claim either. Individual anecdotes, whichever way they point, are no substitute for a comparison set up in advance.

Not every question is a matter for testing

Questions such as “What is the meaning of my life?” or “Is this music beautiful?” are not matters of measurement — and they do not need to be. Testability determines not the value of a sentence but the tool appropriate to it. What deserves attention is when a statement closed to measurement is presented in the language of certain, measured results. Seen this way, even the sentence “science cannot explain everything” is really a statement about categories: science deals with questions open to measurement, and the rest is the legitimate territory of other kinds of conversation. In the next lesson, falsifiability, we take this distinction one step further.