A computer program for classifying elements of a language corpus for large-scale analysis is discussed. The approach is based on the assumption that there is a natural unit in speech processing and production, called a tone unit. The program "tags" the five grammatical phrase types (verb, adverb, adjective, noun, and prepositional) to provide a basic syntactic description of each tone unit of a text. In the first part of the program, all complete words are tagged for word class. In the second part of the program, all word-class tags are scanned by tone unit, and labels assigned to grammatical phrases. Third, the tone units are analyzed from right to left. A demonstration of the program design analyzes some verb phrases. (MSE)