The current study describes preschool teacher-child interactions during several commonly scheduled classroom activities in which teachers deliver instructions. An observation system was developed that incorporated measurement of evidence-based compliance strategies and included the types of instructions delivered (e.g., integral or deficient directives, embedded directives, "do" or "don't" commands), the children's behavior with respect to the instructions (e.g., compliance, noncompliance, active avoidance, problem behavior), and the differential responses of the teacher to the child's behavior following an instruction (e.g., appropriate or inappropriate provision of attention and escape). After 4 classroom teachers were observed at least five times in each of five target activities, simple and conditional probabilities were calculated. Results indicated that (a) the frequency of instruction and probability of compliance varied as a function of activity type, (b) "do" commands and directive prompts were delivered almost to the exclusion of "don't" commands and nondirective prompts, (c) the likelihood of compliance was highest following an embedded or an integral directive prompt, and (d) although putative social reinforcers were more likely to follow noncompliance than compliance and were highly likely following problem behavior, compliance occurred over twice as much as noncompliance, and problem behavior during instructions was very low. Implications for using descriptive assessments for understanding and improving teacher-child interactions in the preschool classroom are discussed. (Contains 1 figure and 4 tables.)