The ability of musically talented and trained subjects (124 non- absolutepitch- possessors, 11 AP possessors), was tested to determine whether or not brief musical stimuli were in the nominal (i. e., correct) key; no pitch reference was given. As test stimuli, excerpts of J. S. Bach's major preludes of the Well- Tempered Clavier, Vol. 1 (duration 4- 5 sec), were employed. Three categories of test were carried out, i. e., pilot, main, and control experiments. In the pilot and main experiments, test stimuli were generated on a piano, while in the control experiments modified piano sounds and electronic sounds were used. The pilot experiments yielded good replicability of previous findings (Terhardt & Ward, 1982), and showed insensitivity of key- identification performance to the type of experimental paradigm employed. In the main experiment, 78% of the subjects (only 11 being AP possessors) achieved significant identification performance with transpositions of up to 4 semitones. Forty-five percent of subjects were significantly able to tell the nominal key from ± 1-semitone transpositions. The distribution of identification rates (normalized for guessing) extends from 0 to 100% and shows a pronounced minimum at around 75%. This suggests two different modes of key identification, one used by non-AP-possessors, and the other by AP possessors. The control experiments reveal that key- identification performance is not piano-specific. Results of all three test categories indicate that no perceptual cues other than pitch were involved. It is thus concluded that both AP possessors and non-AP-possessors depend on absolute pitch information when identifying musical key; however, they employ different perceptual modes: AP possessors primarily identify individual notes, while non-AP-possessors unconsciously deduce from a series of notes a feeling of key.