In this study, we examined whether increasing the proportion of false information suggested by a confederate would influence the magnitude of socially introduced false memories in the social contagion paradigm Roediger, Meade, & Bergman (Psychonomic Bulletin & Review 8:365-371, 2001). One participant and one confederate collaboratively recalled items from previously studied household scenes. During collaboration, the confederate interjected 0 %, 33 %, 66 %, or 100 % false items. On subsequent individual-recall tests across three experiments, participants were just as likely to incorporate misleading suggestions from a partner who was mostly accurate (33 % incorrect) as they were from a partner who was not at all accurate (100 % incorrect). Even when participants witnessed firsthand that their partner had a very poor memory on a related memory task, they were still as likely to incorporate the confederate's entirely misleading suggestions on subsequent recall and recognition tests (Exp. 2). Only when participants witnessed firsthand that their partner had a very poor memory on a practice test of the experimental task itself were they able to reduce false memory, and this reduction occurred selectively on a subsequent individual recognition test (Exp. 3). These data demonstrate that participants do not always consider their partners' memory ability when working on collaborative memory tasks.