This paper explicitly addresses the ‘elicited versus natural data debate’ in cross-cultural speech act realisation research through critical discussion of an empirical study of comparable request sequences by Spanish and British undergraduates to one of their lecturers. Elicited (discourse completion tests) and natural data (unsolicited emails) were used and, not unexpectedly, produced significantly different results for each language community. That these differences related to crucial aspects of the interpretation and performance of requesting behaviour — such as organisation, density and politeness choices — leads us to argue that the goals of cross-cultural speech act realisation research can be best pursued through the analysis of natural data. The latter, however, should not be regarded as some methodological panacea but needs to be exposed to the same intellectual rigour that elicited data have been.