This study investigates the consistency and accuracy of person judgments at zero acquaintance in two conditions: Targets were judged repeatedly across a sequence of videotaped situations, either (a) by different observers in each situation, or (b) by the same observer in all situations. Replicating previous research, accuracy increased (inversely accelerated) when including additional situations rated by different observers. However, much of this increase was owed to improved inter-rater reliability (a confound in previous research). Also, cross-situational consistency of judgments was much higher when all ratings came from the same observer. These results underscore the crucial influence of shared meaning on judgment accuracy and consistency.