In this paper, we develop a memory model that predicts retrieval characteristics of real-world facts. First, we show how ACT-R’s memory model can be used to predict people’s knowledge about real-world objects. The model assumes the probability of retrieving a chunk of information about an object and the time to retrieve this information depend on the pattern of prior environmental exposure to the object. Second, we use frequencies of information appearing on the Internet as a proxy for what information people would encounter in their natural environment, outside the laboratory. In two Experiments, we use this model to account for subjects’ associative knowledge about real-world objects as well as the associated retrieval latencies. Third, in a computer simulation, we explore how such model predictions can be used to understand the workings and performance of decision strategies that operate on the contents of declarative memory.