Conjunction errors occur when participants incorrectly identify as "old" novel test stimuli created by recombining parts of two study stimuli (parent items). Prior studies have reported that the conjunction error rate is higher when parent items are studied together than when they are studied apart (a parent proximity effect). In several experiments we attempted to obtain parent proximity effects with naturalistic faces or pairs of symbol strings. We also varied the type of facial conjunction (Experiments 1, 3A-3C), the presentation rate during study (Experiment 5), and the study list length across experiments (long lists in Experiments 1, 3A, 3B, and 3C; short lists in Experiments 2, 4, and 5). Conjunction effects, but not parent proximity effects, occurred in each experiment. The results are consistent with familiarity-based explanations of the conjunction effect but fail to support a feature bundling hypothesis.