Once upon a time, doctors made house calls and eye contact. Chatting at patients’ bedsides or with their families at kitchen tables, doctors assessed both patient and context. They understood the sensible counsel of postbellum physician William Osler: Listen, and the patient will tell you the diagnosis. So how can 21st-century physicians hope to interpret their patients’ illness narratives, when, in the typical encounter, the doctor interrupts 18 seconds after the patient begins speaking, and within 20 seconds has formed some opinion of what is wrong?

Jerome Groopman, a Harvard professor of medicine and New Yorker staff writer, became upset that the medical students, interns, and residents he was training did not seem to be “thinking deeply about their patients’ problems.” He asked astute diagnosticians around the country how they approached and cracked difficult diagnoses and what happened when they failed. Misdiagnosis is not an insignificant problem: Groopman cites a finding that between one in six and one in seven patients is incorrectly assessed. Most medical errors, he discovered, arose from all-too-human “mistakes in thinking,” not technical glitches.