The conjoint-recognition model (CRM) implements fuzzy-trace theory's opponent process conception of false memory. Within the family of measurement models that separate the memory effects of recollection and familiarity, CRM is the only one that accomplishes this for false as well as true memory. We assembled a corpus of 537 sets of conjoint-recognition data, with estimates of CRM's parameters plus goodness-of-fit statistics being available for all the data sets. This corpus was used to conduct a meta-analysis of CRM's underlying process assumptions by pitting two theoretical interpretations of the model against each other: (a) the original interpretation, which assumes that its retrieval parameters tap a single recollection process (verbatim retrieval) and a single familiarity process (gist retrieval), and (b) the dual-recollection interpretation, which assumes that its parameters also tap a second recollection process (context retrieval). The two interpretations generate a series of differential predictions that fall into three groups--namely, predictions about invariant relations among parameters, about the structure of CRM's parameter space, and about the location of individual parameters within the space. When these predictions were evaluated with the corpus, the results converged on the dual-recollection interpretation. The results also resolved a long-standing uncertainty about whether the familiarity process for true memory is semantically or perceptually driven.