In this paper we report the findings of an experiment to test whether training via processing instruction transfers to output tasks. Two groups of Turkish L1-English L2 learners participated: one that received processing instruction on passive structures and a control group that did not receive any instruction on passives. A pre-test/multiple posttest design was used with three measures: ability to interpret passives correctly, ability to produce passives at the sentence level, ability to reconstruct a three sentence passage in which passives were embedded. Results showed that the processing instruction group made significant gains on all three measures. The control group made no gains. Our conclusion is that even though processing instruction is input oriented, its effects are not limited to input-oriented tasks. (C) 2011 Elsevier Ltd. All rights reserved.