This paper consists of three parts. Section 1 considers how difficult it is to analyze natural languages by computer, and, therefore how difficult it is to evaluate automatically the student's responses to some types of questions in computer-aided instruction systems for language teaching. Section 2 discusses drawbacks of conventional computer aided instruction systems for language teaching and of conventional language laboratories, and presents a picture of an idealized computer-aided instruction system as an extension of classroom instruction--a long range goal which cannot be technologically and financially achieved in the immediate future. Section 3 outlines a more modest goal that is within the bounds of the present hardware and software developments in computer sciences--that is, a system for aiding textbook authors in compiling and revising language textbooks, and for aiding classroom instructors in using textbooks in a way more suitable for their students' needs. Added to such a system is the capability for printing non-standard characters on computers, which will make it possible to produce language textbooks printed in the orthographics of the languages. (Author/FWB)