In a field study, we examined choice blindness for eyewitnesses' facial recognition decisions. Seventy‐one pedestrians were engaged in a conversation by two experimenters who pretended to be tourists in the center of a European city. After a short interval, pedestrians were asked to identify the two experimenters from separate simultaneous six‐person photo lineups. Following each of the two forced‐choice recognition decisions, they were confronted with their selection and asked to motivate their decision. However, for one of the recognition decisions, the chosen lineup member was exchanged with a previously unidentified member. Blindness for this identity manipulation occurred at the rate of 40.8%. Furthermore, the detection rate varied as a function of similarity (high vs. low) between the original choice and the manipulated outcome. Finally, choice manipulations undermined the confidence–accuracy relation for detectors to a greater degree than for blind participants. Stimulus ambiguity is discussed as a moderator of choice blindness. Copyright © 2013 John Wiley & Sons, Ltd.