Recent research has shown that non-Māori-speaking New Zealanders (NMS) have a te reo Māori proto-lexicon: an implicit store of Māori word forms and part-words acquired largely via ambient language exposure (Oh et al., 2020; Panther et al., 2023). While NMS are exposed to both written and spoken Māori, this research has only used orthographic stimuli, assuming that NMS can straightforwardly map between written and spoken forms. This assumption, and NMS’ responses to auditory stimuli, have not been examined directly, however.

To address these gaps, this thesis combines old, new, and previously unexamined data from Māori nonword spelling and Māori nonword wellformedness rating tasks. Using spelling data, it was found that NMS did not robustly map between spoken and written modalities, but transcriptions nevertheless aligned with Māori constraints. Wellformedness data were used in ordinal regression models to investigate NMS’ phonotactic and orthotactic knowledge in both modalities. Ultimately, NMS were sensitive to Māori phonotactics and orthotactics in orthographic stimuli, but for auditory stimuli they were only sensitive to phonotactics.

I argue that NMS automatically combine information from across language domains and modalities in a mutually-reinforcing manner, though this knowledge is more robustly accessed via orthography due to perceptual interference auditorily. These findings contribute to a growing body of research showing that automatic statistical learning mechanisms (Aslin et al., 2017) persist into adulthood. Furthermore, while my findings support the use of orthographic stimuli when measuring implicit knowledge, I emphasise the need to simultaneously account for various types of information when doing so, including phonotactics, orthotactics, and language-specific features.