Much of personality research attempts to identify causal links between personality traits and various types of outcomes. I argue that causal interpretations require traits to be seen as existentially and holistically real and the associations to be independent of specific ways of operationalizing the traits. Among other things, this means that, to the extents that causality is to be ascribed to such holistic traits, items and facets of those traits should be similarly associated with specific outcomes, except for variability in the degrees to which they reflect the traits (i.e. factor loadings). I argue that, before drawing causal inferences about personality trait–outcome associations, the presence of this condition should be routinely tested by, for example, systematically comparing the outcome associations of individual items or facets, or sampling different indicators for measuring the same purported traits. Existing evidence suggests that observed associations between personality traits and outcomes at least sometimes depend on which particular items or facets have been included in trait operationalizations, calling trait‐level causal interpretations into question. However, this has rarely been considered in the literature. I argue that when outcome associations are specific to facets, they should not be generalized to traits. Furthermore, when the associations are specific to particular items, they should not even be generalized to facets. Copyright © 2016 European Association of Personality Psychology