In J. Acoust. Soc. Am. Suppl. I69, S103 (1981) a computer program was presented, which works as a musical version of a text‐to‐speech‐conversion program. Thus, the input is the melody in musical notation, and the output is the melody played. The “pronunciation rules” in this program, which serve the purpose of improving the musical quality of the performance, have now been further developed. New rules have been formulated which work with a time window corresponding to one or several bars. In a formal test the musical effect of these rules have been approved by a group of professional musicians. The purpose and means of the rules are discussed. Some rules apparently serve the purpose of emphasizing unexpected notes or marking structural components of the melody such as chord shifts and phrases. The ways in which this is signaled in the performance often show a striking similarity with rules for speech suggesting that the rules refer to the listener's acquaintance with speech. Other rules seem to allude to the listener's previous experience of motion. In this way, the project seems to shed some light on some basic requirements for music communication.