A study examined the utility of latent partition analysis for describing the structure of emotional responses in reading. Subjects in the first experiment, 40 undergraduate university students, read a 2,100-word story and then immediately reported any scenes or events that evoked an emotional response and any mental images they recalled from reading the story. Sixty-nine affect terms were then identified for the latent partition analysis. Subjects in the second experiment, 53 undergraduates who had not read the story, were presented with a deck of 69 randomly ordered cards, each containing one of the emotional response terms reported by the students from experiment one. The sorters were then told to divide the cards into different piles representing categories of terms that were similar in meaning. The word sorts were subjected to a factor analysis. Results indicated that the factors identified through latent partition analysis by an independent group of sorters proved useful in interpreting readers' emotional responses to a story. Findings suggest that readers call on their background knowledge of emotional responses during reading and that free reports can provide a rich source of information about readers' emotional experiences that can be reliably coded and meaningfully interpreted. (Four tables of data are included.) (RS)