This paper investigates the use of ‘role shift’ by interpreters working from spoken English into Australian Sign Language (Auslan). Role shift is a high-order linguistic skill which students typically find difficult to master. The study looks at possible source text (ST) motivations for its use by four skilled signed language interpreters in an English to Auslan interpreting task, with a view to later pedagogical application. Auslan target texts (TT) rendered by the interpreters were mapped against the English ST using ELAN annotation software. Salient features of the role shift generated by the participants are documented, including: incidence of role shift, native/non-native signer advantage, persona adopted, constructed action versus constructed dialogue, and length and intensity of role shift. Examination of ST segments which trigger role shift in the TT reveals that agent-focused active clause constructions in particular require little manipulation and most readily lead to role shift outcomes. Passive constructions, nominalizations and complex/higher register segments, however, are frequently re-structured into simpler active clauses, with role shift incorporated (or not). The data does not support a strict cause-effect relationship between any particular ST feature and the production of role shift in the TTs; rather, it points to the need for interpreters to recognize ready opportunities for inclusion of role shift, and/or to reconfigure the ST content and form, with role shift as a further layer of depiction.