A conclusion’s scope is bounded by the sample

What it is

How far a finding can travel is set first by which people, tasks, devices, and period actually entered the corpus. Sample-bounded inference writes the subject of a conclusion as the kind of observation the sample supplies, not as “users” or “the product.” Sample here is not headcount: who was recruited, what they were asked to do, and on which interface jointly draw the line where the claim still stands.

Why it happens

Inference jumps from observed instances to unobserved ones. How far the jump can go depends on whether the unobserved cases are of the same kind on attributes the question cares about. If the question is first-time account opening, a sample of long-time customers cannot carry a claim about newcomers. If the question is one-handed use on a small screen, desktop sessions on a large monitor do not supply it. Adding people without changing composition describes the same kind more precisely; it does not open a new range. Putting sample composition into the grammatical subject blocks silent subject substitution.

Studying it

Place a sample-boundary table beside each major claim: inclusion, exclusion, how completers differed from the plan, task set, devices, and period. Run a subject-replacement test: after swapping “users” for the actual sample description, is the sentence still one the researcher would defend? Where a sampling frame or product-log baseline exists, compare the sample with the baseline on variables tied to the outcome, and narrow the claim accordingly rather than asserting representativeness after the fact.

Where it stops holding

Bounding by sample is not a ban on saying anything about people outside the room. When a mechanism is clear and a new setting shares the premises that mechanism needs, a conditional transfer can be argued—but that is a second argument, not a license the sample issues by itself. An experiment may identify a treatment effect inside a convenience sample; who the effect holds for is still sample-limited. A finely written sample description is not by itself a demonstration of external validity.

Applying it

• Give every conclusion a restricted subject: “Among these participants, on these tasks, on this interface, we observed…”
• Ban unmodified “users always…” from decision materials; add the sample description or delete the sentence.
• Attach recruitment and the task list to the findings page, not only to a methods appendix.
• Ask someone who did not run the study to read the conclusion sentences and say whom they think the claim covers; rewrite if that set is larger than the sample.

Related

• Same group: Q4.10.2 A change of context can void a conclusion · Q4.10.3 Reports must state boundary conditions · Q4.10.4 Qualitative work seeks analytic generalization, not statistical representativeness · Q4.10.5 A single-case finding is not a general law until replicated · Q4.10.6 Recheck premises before transferring across cultures or platforms · Q4.10.7 Packaging small-sample findings as universal truths is a common overreach
• Adjacent: Q1.04 Sampling and representativeness · Q1.08 Sample size
• Search terms: sample-bounded inference · external validity · target population