Recognition judgments, such as those involved in eyewitness identification, are often depicted as relatively pure products of memory processes and criterion setting, such as is depicted in the WITNESS model of eyewitness identification behavior. The presence of extra-memorial factors, such as social influence, however, necessitates a negotiation between memorial and extra-memorial information. Participants (N = 450) viewed a short video depicting a person planting a bomb down an airshaft. Before attempting lineup identifications, participants in Experiment 1 learned that an alleged co-witness made a "not there," plausible, or implausible identification decision with high or low confidence. Co-witness information dramatically influenced identification decisions in the direction of the co-witness's decision. Particularly important was the finding that there was a significant effect of co-witness information even for participants who learned that the co-witness identified an implausible lineup member. This latter finding cannot be explained by standard two-parameter models of recognition (such as the WITNESS model). Participants' confidence in their identification decisions tended to match that of their co-witness, regardless of whether the co-witness gave a "not there," plausible, or implausible identification decision. However, this confidence-matching effect was greater for participants who made the same decision as the co-witness than it was for participants who made a different decision than the co-witness. Experiment 2 participants (N=323) viewed the same crime video and then either viewed the lineup first or received the co-witness information (i.e., a "not there" or plausible identification decision) first. As in Experiment 1, co-witness information influenced identification decisions in the direction of the co-witness's identification, but the extent of the influence did not vary as a function of the presentation order of a lineup and co-witness information. When asked later how they would have responded if they had never received the co-witness information, participants were able to somewhat correctly imagine the identification decision they would have made without the co-witness information, but this correction always underestimated the co-witness's influence. The results suggest a need for a third parameter, in addition to memory strength and decision criteria, when attempting to predict and explain eyewitness identification behavior in a real-world setting.