This study investigates the influence of a grammatical reconstruction task on talk and subsequent individual writing. In a South Korean university, two classes alternated learning conditions, between 'read-summarize-write' and 'read-reconstruct-write'. In the summarize condition, small groups collaborated to create a summary of the reading; in the reconstruction condition, small groups collaborated on a summary reconstruction task. This was an activity in which the teacher first wrote a summary of the reading and then removed all the grammar words, punctuation, and capitalized letters, leaving only the content words. Students had to reconstruct the original summary. In the last phase of the lesson, students wrote an essay on the same topic as the reading. The in-class talk was analysed for differences and the complexity, accuracy, and fluency of the writing were calculated. Results showed the collaborative summary prompted discussion about the ideational content of the summary, whereas the reconstruction task promoted talk about which grammatical forms to use. Subsequent individual writing was impacted; accuracy improved after the reconstruction task with no loss of fluency. Fluency increased in both conditions when compared to pre-tests, while syntactic and lexical complexity of the writing was unaffected by condition.