The validity of interpreting teachers' perceptions of their performance as an index of their actual performance was examined. Two matching instruments were constructed; each contained 72 items in 6 categories of skills. One assessed teachers' perceptions of their competence on behaviorally stated generic teaching skills, and the other assessed teachers' actual performance on the same skills. The items represented verbal information, concept identification, or problem-solving skills that teachers need to perform the skill objectives; and skills that could realistically be assessed using pencil and paper questions. One hundred seventy-five classroom teachers were paid to participate in the study. Results showed that teachers' perception scores were significantly higher than their actual performance scores in all six content areas. Using teachers' perception scores to predict actual performance scores on teaching skills appeared to be an invalid practice; this finding held true for three different types of questions: recall of verbal information, concept identification, and problem-solving questions. It was also recommended that instructional needs of teacher education programs be determined by the teachers' actual performance rather than their perceived skills. (MV)