sound, graphics, and analysis. In particular situations there can be some divergence in the needs of these three areas. For example, the realization of notes inegales would favor single dots in print and double dots in sound. A B^b-trumpet part in a C-Major score would be printed at written pitch in a graphics applications and played back at concert pitch in a sound application. We accommodate such divergences with what amounts to a double representation scheme and make MuseData files available in five formats MIDI for sound applications, DARMS and SCORE for printing applications, and Kern and our own format for research and analysis.

The string search, familiar to us from our experiences with text processing, offers the possibility of retrieving precise information from comprehensive 'readings' of large corpora of music to a degree that has never before been possible. A patient reader might hope to find all the instances of a particular series of words in a corpus of written works merely by reading through it attentively a few times, but a patient musician is unlikely to find with the same success a particular sequence of sound events from a corpus of musical works. The rapidity with which such searches can be accomplished is quite amazing. At present our Bach database contains approximately 70% of the music published by the Bach Gesellschaft. When search algorithms are applied to compressed data, we are able to search this corpus in about 5 seconds.

Certain writers of the past were greatly concerned with tracing Bach's use of the musical acronym for his surname. A string search of the Bach Database has informed us that one unheralded use of it occurs in the bass part of the opening movement of the second Brandenburg Concerto (see Ex. 1).

This use is apparently overlooked because of the tied notes and the harmonic rather than melodic function. What the computer is unable to tell us is whether such an occurrence was intentional. The underlying harmonic movement of this passage is very typical of Bach's time and it would therefore seem that this figuration could simply have come about as a byproduct of harmonic convention. Particularly since the spelling is 'transposed' in the first instance and the passage consists of sixth ones, the case seems dubious. Yet the potential availability of the passage makes all the more striking the fact that it does not occur in any other conspicuous but unexpected instances among the other works in the Bach database.

Another seemingly simple task in stylometrics is the development of lexicons of particular musical features. In an unpublished study of the figuration process in chorale preludes, I attempted to study the methodology of figure-for-interval substitution by compiling a list of all the sixteenth-note patterns used in duple-meter works from such keyboard corpora as the Inventions, Sinfonias, Well-Tempered Clavier, French Suites, and English Suites.

I wished each four-note sequence to be classified according to the melodic interval that was hypothetically formed by the first note of the set and the first accented note following the set. Data for works in major and minor keys were kept separate.