Respondents in self-administered, paper-based surveys answer open-ended frequency questions differently depending on the size of the response field provided. Longer fields increase the proportion of alphanumeric elaborations or answers explicitly designated as an estimate ("about 20,'' "60-70''). By contrast, shorter fields yield higher proportions of pure digit responses. It is assumed that respondents interpret the size of the response field as an instruction regarding the format of the expected answer. In this article, the authors assess the effect of the length of the response field in a web survey compared to a paper-and-pencil questionnaire. Results indicate that the effect of the length of the response field is more pronounced in the paper-and-pencil condition compared to the web survey condition. This raises the question as to what extent the underlying visual design language differs across self-administered survey modes.