A number of studies have found that writers produce text in bursts of language. That is, when creating a text, writers produce a few words, pause, produce a few more words, pause, and so on. Chenoweth and Hayes (2003) hypothesized that language bursts occur when writers translate ideas in to new language. This study tested this hypothesis against the following two alternative hypotheses: (a) Language bursts are caused by proposing new ideas rather than by translating ideas in to written language and (b) language bursts depend on the form of the input to the writing process rather than on the translation process. The study employed an editing task in which participants were required to translate a written language input. The alternative hypotheses led to contradictory predictions about writers' performance in this task. The study also explored the impact of working memory restrictions on task performance. (Contains 3 tables, 2 figures, and 3 notes.)