It has been claimed that children use a nonlinguistic strategy in the interpretation of reflexives, despite their apparent knowledge of Condition A of the Binding Theory. This claim, that children equate the presence of a word containing self in a sentence with an action on oneself, has been a part of a general account of children's performance on structures governed by the Binding Conditions (Grimshaw and Rosen (1990)). We conducted an experiment to test this claim, in which the use of the strategy would result in erroneous performance. This experiment, done in Hebrew through a truth-value judgment task, presented the children with sentences containing reflexives and pictures with corresponding reflexive actions, where the antecedent did not match the one in the sentence. It also presented the same pictures along with sentences that contained reflexive predicates and no reflexive. It was found that children of the relevant age group (5- to 6-year-olds) made no errors, thus providing strong evidence against the use of the strategy. We also found that younger children made some errors but that these were equally spread over the Condition A sentences (i.e., with reflexives) and the sentences with reflexive predicates, indicating that the errors were unrelated to knowledge of the Binding Theory in its standard formulation. We argue that the results for the older children support an alternative view of their abilities in binding, proposed by Grodzinsky and Reinhart (1993). We also note that the findings from the younger children provide suggestive evidence favoring the formulation of Condition A of the Binding Theory over reflexivity.