Two experiments studied orientation effects on cognitive mapping. Experiment 1 used a 4 x 3 within-subjects factorial design, with 12 participants, following the procedure of Levine, Jankovic, and Palij (1982). Participants were taught a map by finger tracing, then were tested on either an unaltered map, a 180°-rotated version, or an unaltered version viewed from a new 180º perspective. Results replicated the Levine et al. data, but the map-rotation condition showed more errors on reverse than forward-movement types. Experiment 2, with 36 participants, added a pre/posttest, training period, and modified procedure. Experiment 1 was replicated, and the modified procedure improved performance when participants rotated the map. Training was ineffective, but some practice improvements on the posttest were found. The present studies' anomaly with reverse-movement types offers only limited support for Levine et al.'s principle of equiavailability.