Reliability in its technical sense means consistency of measurement: whether an instrument gives the same result on repeated application, across items, across occasions, or across raters. That sense has a real literature and named coefficients — Cronbach’s alpha for internal consistency being the most familiar, alongside test–retest and inter-rater measures — and those coefficients have precise formulations and well-documented criticisms, including the widespread misuse of alpha as though it demonstrated unidimensionality. Alongside this sits the everyday sense in which “reliable” means merely trustworthy, which shares nothing with the psychometric sense but the word. The thresholds people quote are conventions inherited from textbooks, not standards, and are treated as conventions here.
Why this entry carries no source list. This lexicon cites an official primary source wherever one exists, and says so plainly where none does. No body governs this term. There is no standards organisation for it, no register, and no text whose authors could claim to have settled its meaning. Methods literature describes practice and argues about it; a citation to one author would dress a position in a live argument as a rule, which would mislead more than silence does. What follows is the convention as it is actually used, with the disagreements named rather than smoothed over.