THIS PAPER PRESENTS A FRAMEWORK FOR DESCRIBING, IN TERMS OF A MANAGEABLE NUMBER OF CATEGORIES, THE WIDE VARIETY OF DATA WHICH A SURVEY OF EXPERT OPINION SHOWED TO BE IMPORTANT TO TEACHER EVALUATORS. IT PROCEEDS FROM THE ASSUMPTIONS THAT A TEACHER MAY BE SAID TO BE "GOOD" ONLY WHEN HE SATISFIED SOMEONE'S EXPECTATIONS, THAT PEOPLE DIFFER IN WHAT THEY EXPECT FROM TEACHERS, AND THAT A SCHEME FOR EVALUATING TEACHERS AND FOR PREDICTING TEACHER EFFECTIVENESS MUST TAKE THOSE DIFFERENCES INTO ACCOUNT. THREE POSTULATED CATEGORIES OF EXPECTATIONS RELATING TO INSTRUCTIONAL VARIABLES ARE EMPLOYED--(1) THE KINDS OF INSTRUCTIONAL OBJECTIVES ENDORSED BY THE TEACHER, (2) THE KINDS OF METHODS OF INSTRUCTION HE EMPLOYS, AND (3) THE KINDS OF RELATIONS THE TEACHER MAINTAINS WITH HIS PUPILS. AN INSTRUMENT INTENDED TO TEST OUT THIS MODEL IS DESCRIBED, AND THE RESPONSES OF MEMBERS OF A NUMBER OF PROFESSIONAL GROUPS ARE REPORTED. THE FINDINGS SUPPORT THE CONCLUSION THAT THE SUBJECTS DO, IN FACT, DIFFER IN THEIR EDUCATIONAL BELIEFS ALONG THE DIMENSIONS BUILT INTO THE MODEL. THE EVIDENCE ALSO INDICATES THAT THERE IS A STRONG AFFECTIVE COMPONENT TO THESE BELIEFS AND THAT MANY PEOPLE WILL BECOME EITHER ANGRY OR ANXIOUS IF CONFRONTED WITH EVIDENCE THAT OTHERS WHOM THEY RESPECT SEE THE "GOOD TEACHER" DIFFERENTLY THAN THEY DO. (HW)