A pilot study of a larger project on second language comprehension under two input conditions is reported. The first condition is characterized by the availability of samples of target input that have been modified a priori toward greater semantic redundancy and transparency and less complex syntax. The second condition is characterized by the availability of opportunities for non-native speakers (NNS) to interact with the native speaker, bringing about modification and restructuring of the interaction in order to arrive at mutual understanding. The pilot study, although limited to nine NNS of English, indicates that interaction generates a larger quantity and greater redundancy of input, which both help make a linguistically complex version of directions more comprehensible than those given without interaction, as a premodified text. Questions remain regarding the mechanism by which these input modifications are brought about during the course of interaction. It appears that interactional adjustments such as comprehension and confirmation checks and clarification requests may be the means by which input is repeated or reworded until understanding is reached. It is hoped that these findings will contribute to second language acquisition theory and provide a framework for the development of learning materials and instructional techniques. Contains 15 references. (LB)