The intervallic rivalry model of key identification is outlined and evaluated in two experiments that use a completion judgment task. Experiment 1 replicates an earlier experiment by Cuddy and Badertscher (1987), in which the rare-interval hypothesis of the intervallic rivalry model was considered. In the present study, listeners with different levels of musical training rated probe tones in the context of three different melodic patterns: arpeggiated major triads, ascending major scales, and arpeggiated diminished triads. Results of Experiment 1 indicated that in both the C major triadic and the C major scalar contexts, listeners gave higher completion ratings to all three probes that were members of the presented C major triad than to the other probes, with the exception of F, thereby producing a jagged (multipeaked) profile. For the diminished triadic context, listeners rated the single probe C, that which corresponds to the tonal center in major mode for that group of three tones, as the best completion. Experiment 2 tested the temporal-order hypothesis of the intervallic rivalry model by reordering tones in all three contexts. Again jagged tone profiles appeared with major triadic and major scalar contexts, although in the former the tone F, a perfect fifth below the root of the presented C major triad, received the best completion rating. A single-peaked function appeared with probes in the diminished triadic context, where the major-mode tonic garnered the highest rating found in all conditions of both experiments. Data are interpreted as support for both the rare-interval hypothesis and the temporal-order hypothesis derived from the intervallic rivalry model of key discovery. Complementary findings consistent with the tonal hierarchy model are also discussed.