This study investigated the internal consistencies and temporal stabilities of different implicit self-esteem measures. Participants (N = 101) responded twice—with a time lag of 4 weeks—to five different tasks: the Implicit Association Test (IAT), the Brief Implicit Association Test (BIAT), the Affective Priming Task (APT), the Identification-Extrinsic Affective Simon Task (ID-EAST) and the Name-Letter Task (NLT). As expected, the highest reliability coefficients were obtained for the self-esteem IAT. Importantly, the internal consistencies and the temporal stabilities of the APT, the ID-EAST, and the NLT were substantially improved by using material, structural, and analytic innovations. In particular, the use of the adaptive response-window procedure for the APT, the computation of error scores for the ID-EAST, and the computation of a double corrected scoring algorithm for the NLT yielded reliability coefficients comparable to those of the established IAT. Implications for the indirect assessment of self-esteem are discussed. Copyright © 2010 John Wiley & Sons, Ltd.