We investigated the impact of filler quality and presence on confidence, response latency, and propensity to respond ‘don't know’ in eyewitness line‐ups and showups. More specifically, we tested the hypothesis that confident, fast witnesses would be accurate in fair line‐ups and showups, but the inclusion of duds (poor fillers) would break down these relationships in a biased line‐up. Participants viewed a mock crime video, made a timed identification decision, and gave a confidence judgment. As predicted, biased line‐up witnesses were fast and confident, regardless of accuracy, and rarely responded ‘don't know’. In addition, we found that witnesses who are the fastest and most confident were equally accurate in fair line‐ups and showups, and both were better than biased line‐ups. These findings suggest that biased line‐ups should not be used (although, unfortunately, they frequently are); in fact, it may be better to conduct a showup than a biased line‐up. Copyright © 2016 John Wiley & Sons, Ltd.