Teachers should be evaluated chiefly by the results they produce. Those results will not be properly detected through the use of norm-referenced achievement tests (NRT) due to the following major deficits of NRT: (1) the descriptions of what is measured by NRT are far too loose; (2) evaluative expectations are unclear; (3) the necessity for NRT to produce sufficient variance in examinee responses makes it less likely that such tests will measure pupil mastery of key knowledge and skills; and (4) most NRT do not differentiate among students chiefly on the basis of what those students have learned in school--rather their differentiations are based heavily on the nature of the semantic experience that students bring to school. Pupil test performance should be used as a major factor in any teacher evaluation system, however. Those tests should be well constructed criterion-referenced tests (CRT) which will provide teachers with a clear set of instructional targets and test items which can detect the effects of skillful instruction, communication of evaluative criteria, absence of the automatic excision of items on which pupils perform well, and a vehicle which can measure affective gains. (RL)