The tendency of decreasing religiosity is explained by the theory of secularization through differentiation and pluralization. Using the ess 2002–2016, the impacts of both on church attendance and self-ascribed religiosity are tested, controlling for determinants of religiosity—that is, for belonging (cohort and denomination) and choice (education, urban residence, marriage, parenthood, and employment)—with multi-level models separating between- from within-country effects. Without controls, time negatively affects religiosity: there is a secularization tendency. But controlling for cohort and denomination annihilates this effect and strongly reduces individual-level as well as country-level error variances. Effects of belonging are stronger than those of choice, cohort succession has a negative effect, and religiosity differs between denominations. Differentiation and pluralization have only a few effects between countries and only one within countries such that secularization theory is not confirmed.