Previous studies (Dias&Harris, 1988, 1990; Markovits&Vachon, 1989) have established that young children can correctly respond to certain conditional reasoning problems with empirically false premises, when these are presented in a fantasy context. Such results have been interpreted as indicating that young children possess logical reasoning competence which must be protected from interference with their empirical knowledge. However, Markovits (1993) has proposed a model of reasoning in which context effects are considered in the light of both developmental and information retrieval processes. This predicts that presenting empirically false premises in a fantasy context should improve performance on the valid logical formmodus ponensamong young adolescents. It also predicts that presenting empirically true premises in a fantasy context should decrease performance on the invalid logical form affirmation of the consequent. A total of 89 12‐years‐old and 117 14‐year‐olds were given two reasoning problems involving either true or false premises. These problems were embedded in either a fantasy or a realistic context. Results were consistent with the predictions made by th