This paper presents some reasons for reconsidering the use of discrete structure-point tests of language proficiency, and suggests an alternative basis for designing proficiency tests. Discrete point tests are one of the primary tools of the audio-lingual method of teaching a foreign language and are based on certain assumptions, including the following: (1) there are a given number of specific structure points, the mastery of which constitutes "knowing" a language; (2) the surface structure of a language can be systematically described and its elements listed and compared with any other language similarly described; and (3) the mastery of a language may be divided into the mastery of a number of separate skills: listening, speaking, reading and writing. Objections to all of these assumptions are outlined, and discrete point tests are compared with integrative tests. The latter are based on the premise that "knowing" a language should be determined by the students' ability to operate in a specified sociolinguistic situation with ease or effect. Cloze tests, dictation, and tests involving error recognition all test integrative skills. The construction and use of these tests are outlined. (CLK)