This paper reports a study of VanPatten's processing instruction (PI) and Ray's TPRS. High school students in a beginning Spanish course were divided into three groups (PI, TPRS, and control) and instructed in forms using the Spanish verb gustar. Treatment included sentence-level and discourse-level input, and tests included interpretation and production measures in a pretest, an immediate posttest, and a delayed posttest given two and a half months following treatment. The PI group made the greatest gains in production measures and in a grammaticality judgment test, and the TPRS group made the greatest gains in written fluency. The PI group's statistical gains in production measures held through the delayed posttest.