The invention relates to a sound data processing device and method. For each of a plurality of performance parts, a database stores therein a plurality of part performance data. The part performance data for each of the parts includes a sound generation pattern and tone data corresponding to the sound generation pattern. A query pattern indicative of a sound generation pattern to be made an object of search is input by a user. A search is made through the database for part performance data including a sound generation pattern matching the query pattern. In response to a user's operation, one part performance data is identified from among searched-out results from the database, and the sound generation pattern of the identified part performance data is instructed as a new query pattern. Then, a further search is made through the database for part performance data including a sound generation pattern matching the new query pattern. In accordance with a user's operation, one part performance data is identified from among searched-out results, and the identified part performance data is edited. The thus-edited data can be registered into the database as new part performance data.