Because clinicians often do not recognize that patients have pain and patients do not spontaneously communicate their pain, clinicians may fail to prescribe or administer adequate pain medications. One method of improving clinicians' assessments of pain is to coach patients to communicate their pain in ways that clinicians recognize. The aims of our pilot study were to (a) examine the feasibility of implementing a randomized clinical trial of a COACHING protocol in 18 outpatients with lung cancer pain and (b) estimate the effects of COACHING on nurses' knowledge of patients' pain location, intensity, quality, and pattern. The expectation was that COACHING would show a trend toward reducing the discrepancy between patients self-report of sensory pain and sensory pain data known to nurse clinicians. Patients were randomly assigned to be COACHED or NOT-COACHED (usual care) and pretest-posttest measures with the McGill Pain Questionnaire (MPQ) and Visual Analogue Scale (VAS) were taken from nurses and patients. Patients- and nurse-completed MPQs and VASs were compared for agreement. Improvement in percent agreement occurred consistently more often (pretest to posttest) between patient self-report of sensory pain and nurses' pain assessments in the COACHED group than in the NOT-COACHED group. Pilot study findings demonstrated feasibility of implementing the COACHING protocol and suggest that COACHING may be effective in reducing discrepancies between patients' self-reports and nurses' assessments of sensory pain. Design modifications are recommended for implementation of future studies.