To determine what counts as success in a French first grade, an ethnographer investigated what happens when the teacher "applies" her categories of "things students do" to particular situations. The category called "searching" served as an example. To produce a body of data, the teacher commented on videotapes of ordinary events in her classroom. However, perceptual cues which inspired her to say that a child was "searching" could not always be found within the videotaped incidents. The model of an informant who matches cues from the environment to a cognitive map did not work. Instead, the teacher's comments about the tapes had to be understood as speech acts in which she was accounting for previously stated theories about students. (Author)