Consider, for example, the common practice in the United States of testing a person's ability to drive an automobile. Intuitively, we would conclude that a valid test of driving ability should include an actual sample of a person's behind-the-wheel behavior. However, in many localities a paper-and-pencil test of road signs and traffic regulations is a sufficient criterion for renewal of a driver's license. Is such a test valid? Observational studies seem to bear out the contention that no subsequent driving test is needed, but it is doubtful that the written test actually predicts the quality of driving ability. It is more likely that simply previous experience in driving is the best predictor of good driving ability. The paper-and-pencil license-renewal test probably has little validity for predicting good driving; what it does measure is knowledge of various regulations, which is only a small part of total driving ability.

In tests of language, validity is supported most convincingly by subsequent personal observation of teachers and peers. The validity of a high score on the final exam of a foreign language course will be substantiated by "actual" proficiency in the language (if the claim is that a high score is indicative of high proficiency). A classroom test designed to assess mastery of a point of grammar in communicative use will have validity if test scores correlate either with observed subsequent behavior or with other communicative measures of the grammar point in question.
How can teachers be somewhat assured that a test, whether it is a standardized test or one which has been constructed for classroom use, is indeed valid? The technical procedures for validating tests are complex and require specialized knowledge. But two major types of validation are important for classroom teachers: content validity and construct validity.

Content Validity
If a test actually samples the class of situations, that is, the universe of subject matter about which conclusions are to be drawn, it is said to have con- tent validity. The test actually involves the testee in a sample of the behavior that is being measured. You can usually determine content validity, observationally, if you can clearly define the achievement that you are measuring. A test of tennis competency that asks someone to run a 100-yard dash lacks con- tent validity. If you are trying to assess a person's ability to speak a second language in a conversational setting, a test that asks the learner to answer paper-and-pencil multiple-choice questions requiring grammatical judgments does not achieve content validity. A test that requires the learner actually to speak within some sort of authentic context does.
A concept that is very closely related to content validity is face validity; which asks the question: does the test, on the "face" of it, appear to test what it is designed to test? Face validity is very important from the learner's perspective. To achieve "peak" performance on a test, a learner needs to be convinced that the test is indeed testing what it claims to test. Once I administered a dictation test and a cloze test (see below, for a discussion of cloze tests) as a placement test for an experimental group of learners of English as a second language. Some learners were upset because such tests, on the face of it, did not appear to them to test their true abilities in English. Face validity is almost always perceived in terms of content: if the test samples the actual content of what the learner has achieved or expects to achieve, then face validity will be perceived.