The research on error correction in second language (L2) classrooms tends to provoke argument among both language teachers and L2 acquisition researchers. Many studies have been conducted about the effectiveness of oral corrective feedback (CF) types, but collecting information about teachers' cognitions and feedback practices is not adequately considered. The present study addresses the relationship between teachers̕ stated cognitions about oral corrective feedback and their actual classroom performance. The research was carried out in four EFL classrooms in an institute in Mazandaran province and the data was collected from four EFL teachers in these classrooms through observation, audio -recording, and a questionnaire, supplemented by stimulated recall session. Moreover, each of the classes was observed and audio-recorded about 210 minutes in two sessions. Then, the audio-recordings were transcribed, coded, analyzed and rechecked by the researcher. After that the teachers were required to complete a questionnaire and the ones whose questionnaire data was not clear, participated in a stimulated recall session. Finally, different aspects of oral CF used in the classrooms were compared with teachers stated cognitions in the questionnaire and stimulated recall session. The findings indicated that almost all of the teachers were not completely conscious of the amount and types of feedback they provided in their classroom. Even though all of the teachers stated that giving corrective feedback to the students is a significant task and their students were satisfied to be corrected, they concerned about negative affective responses. This might be the reason of provision of recast as the most frequent type of correction. Further the finding presented discrepancies between teachers' cognitions and their practices in EFL classrooms.
