6 7 Ó Springer Science+Business Media Dordrecht 20178 Abstract Much previous research has conceptualized pauses during writing as 9 indicators of the engagement of higher-level cognitive processes. In the present 10 study 101 university students composed narrative or argumentative essays, while 11 their key logging was recorded. We investigated the relation between pauses within 12 three time intervals (300-999, 1000-1999, and [2000 ms), at different text 13 boundaries (i.e., between words, sentences, and paragraphs), genre (i.e., narrative 14 vs. argumentative), and transcription fluency (i.e., typing speed). Moreover, we 15 investigated the relation between pauses and various lexical characteristics of essays 16 (e.g., word frequency, sentence length) controlling for transcription fluency and 17 genre. In addition to replicating a number of previously reported pause effects in 18 composition, we also show that pauses are related to various aspects of writing, 19 regardless of transcription fluency and genre. Critically our results show that the 20 majority of pause effects in written composition are modulated by pause location. 21 For example, increased pause rates at word boundaries predicted word frequency, 22 while pause rates at sentence boundaries predicted sentence length, suggesting 23 different levels of processing at these text boundaries. Lastly, we report some 24 inconsistencies when using various definitions of pauses. We discuss potential 25 mechanisms underlying effects of pauses at different text boundaries on writing. 26 27 Keywords Writing Á Pauses Á Computational linguistics 28 29 30
A1Electronic supplementary material The online version of this article (