Two experiments were conducted to test the hypotheses that the ability to draw inferences from a text about an unstated agent is based on the quality of information available to the reader, and that the quality of the information is itself based on the unique nature of cues embedded in the text and on the relevance of prior knowledge held by the reader. In the first experiment, 20 second and 20 fourth grade students listened to stories that contained a helpful (unique) or unhelpful (nonunique) clue and then drew inferences. In the second experiment, 28 third and 28 fourth grade students listened to stories with clues that varied as in the first experiment. In addition, half of the students first received a relevant prior knowledge and half received an irrelevant prior knowledge treatment. Both experiments confirmed that the unique nature of an embedded clue was directly related to the ease of drawing an inference. In addition, relevant prior knowledge was shown to enhance inference making in an additive, linear fashion. No grade differences were found. (Stories used in the experiments are included.) (FL)