In the cancer population, the term breakthrough pain typically refers to a transitory flare of pain in the setting of chronic pain managed with opioid drugs. The prevalence and characteristics of this phenomenon have not been defined, and its impact on patient care is unknown. We developed operational definitions for breakthrough pain and its major characteristics, and applied these in a prospective survey of patients with cancer pain. Data were collected during a 3 month period from consecutive patients who reported moderate pain or less for more than 12 h daily and stable opioid dosing for a minimum of 2 consecutive days. Of 63 patients surveyed, 41 (64%) reported breakthrough pain, transient flares of severe or excruciating pain. Fifty-one different pains were described (median 4 pains/day; range 1–3600). Pain characteristics were extremely varied. Twenty-two (43%) pains were paroxysmal in onset; the remainder were more gradual. The duration varied from seconds to hours (median/range: 30 min/1–240 min), and 21 (41%) were both paroxysmal and brief (lancinating pain). Fifteen (29%) of the pains were related to the fixed opioid dose, occurring solely at the end of the dosing interval. Twenty-eight (55%) of the pains were precipitated; of these, 22 were caused by an action of the patient (incident pain), and 6 were associated with a non-volitional precipitant, such as flatulence. The pathophysiology of the pain was believed to be somatic in 17 (33%), visceral in 10 (20%), neuropathic in 14 (27%), and mixed in 10 (20%). Pain was related to the tumor in 42 (82%), the effects of therapy in 7 (14%), and neither in 2 (4%). Diverse interventions were employed to manage these pains, with variable efficacy. These data clarify the spectrum of breakthrough pains and indicate their importance in cancer pain management.