The notion of "fluency" is most often associated with spoken-language phenomena such as stuttering. The present article investigates the relevance of considering fluency in writing. The basic argument for raising this question is empirical--it follows from a focus on difficulties in written and spoken language as manifestations of different problems which should be investigated separately on the basis of their symptoms. Key-logging instruments provide new possibilities for the study of writing. The obvious use of this new technology is to study writing as it unfolds in real time, instead of focusing only on aspects of the end product. A more sophisticated application is to exploit the key-logging instrument in order to test basic assumptions of contemporary theories of spelling. The present study is a dictation task involving words and "non-words", intended to investigate spelling in nine-year-old pupils with regard to their mastery of the doubling of consonants in Norwegian. In this study, we report on differences with regard to temporal measures between a group of strong writers and a group of poor ones. On the basis of these pupils' writing behavior, the relevance of the concept of "fluency" in writing is highlighted. The interpretation of the results questions basic assumptions of the cognitive hypothesis about spelling; the article concludes by hypothesizing a different conception of spelling.