The problem of informant accuracy is examined in light of principles of memory organization from cognitive psychology. These principles turn out to be powerful, not only in explaining overall patterns of informant error, but in predicting details about the types of errors made. Predictions are made in terms both of different kinds of informants and different kinds of objects. All the predictions are strongly supported by the data. Finally, in the light of these results, two strategies are developed. The “best” informants, it seems, can be used to reveal long-range stable patterns of events, and the “worst” informants can be used to reveal the details of a particular event of special interest.