Current research on the task of determining the characteristics of an anonymous
writer, authorship profiling, does not meet the demands of the forensic context,
because of the lack of transparency of certain computational techniques, their
requirements for large data sets, and, most importantly, since the strength of register variation does not guarantee that findings obtained in other registers apply to forensic registers such as, for example, a threatening letter. The present article demonstrates how previously established findings related to stylistic variation in English for gender, age, and social class also apply to the kinds of texts often analysed by forensic linguists through an experiment involving 96 participants. These results constitute an example of linguistically-motivated profiling research and it is argued that the agenda to move from authorship profiling to forensic authorship profiling should be led by previously established knowledge of language variation.