We examine data from the rural Arusha region in Tanzania in which households are asked to recall the illness episodes of randomly chosen other households in their village. We interviewed 502 randomly selected households from 22 villages in 20 wards of Arusha. We analyze the probability that a household can recall another illness episode as a function of the characteristics of the illness, the location and type of health care chosen and the outcome experienced. We found that households are more likely to recall severe illnesses, illnesses for which good quality care is important, illnesses that resulted in visits to hospitals and illnesses when the patient was not cured. In addition, households are more likely to recall illnesses that resulted in a visit to a facility where the average tenure of clinicians is less than two years. The results suggest that households deliberately collect information in order to learn about clinicians and facilities in their local area. We show evidence that households use this information when they choose whether to visit new health care providers. In particular, households are less likely to visit a new provider when they hear of bad outcomes and more likely to do so when they hear of good outcomes.