This paper describes the development of an oral language skills assessment instrument, called the Children's Language Assessment-Situational Tasks (CLA-ST), and its utilization in both formative and summative evaluation of the Tucson Early Education Model (TEEM) in Project Follow Through. First and third graders in Follow Through classrooms implementing TEEM and non-Follow Through comparison counterparts in a Southeastern community served as subjects. Language samples were collected in the fall and in the following spring in sessions designed to elicit language from the children. Pretest and posttest data were independent. Tape recordings made of each session were transcribed and coded for 11 language variables. A 2-factor analysis of variance was done with treatment (TEEM or comparison) and trials (pretest and posttest) as factors. Results indicate that primary children in the TEEM classrooms demonstrate superior language performance on many dimensions of language development. The paper concludes with a discussion of the utility of this kind of evaluation procedure to supplant traditional paper-and-pencil strategies. (Author/SB)