We devised criterial recollection tests to investigate why testing memory for pictures elicits lower false recognition than testing memory for words. Subjects studied unrelated black words paired either with the same word in red font, a corresponding picture, or both. They then took three memory tests, always using black words: a recognition test (say ''yes'' to all studied items), a red word-test, and a picture-test (say ''yes'' only if you recollect a red word or a picture, respectively). Regardless of whether pictures were more or less familiar than red words, false recognition was lowest on the picture test. These results cannot be explained easily by familiarity or strength-based criterion shifts. Instead, they suggest that subjects expected more detailed recollections for pictures, thereby facilitating a diagnostic monitoring process (the ''distinctiveness heuristic''). This recollective difference also influenced source monitoring errors (an ''it-had-to-be-a-word'' effect), again suggesting that detailed recollective expectations influence monitoring processes.