Abstract English as a lingua franca is generally viewed as a user language in which the formal characteristics are adapted to meet the communicative needs of its speakers, inherently resulting in linguistic innovation and variation. In the part-of-speech (POS) tagging of the Vienna-Oxford International Corpus of English (VOICE), many cases were encountered whereby ELF users go beyond the commonly assumed boundaries of word class categories in Standard English by the process defined as word class change. Word class change is understood as a change of word class category, or syntagmatic function, without alternation of the morphological form. In this paper, it is demonstrated how forms which were used in a non-codified function were dealt with in the tagging procedure. Reviewing concepts commonly used to explain word class change, namely conversion and multifunctionality, it is argued that a number of presumptions inherent in these concepts cause difficulties when applied to ELF data. In a case study, the most frequent types of word class shifts or variation (terms which are considered more suitable for ELF data) in VOICE are exemplified, analysed, and discussed with regard to the forms which are shifted, the directionality of these shifts, as well as the environment in which the shifts occur. It is shown that word class shifts in ELF interactions follow clear tendencies and thus a reconceptualization of the concepts of conversion and multifunctionality with regard to ELF is proposed. Finally, it is argued that if tagging practices attempt to display, rather than disguise, initial “problems” arising in the tagging process, this can help to gain useful insights into the nature of the data.