Typically, children learning languages with opaque orthographies (e.g. English) are found to use direct word recognition while children learning transparent orthographies (e.g. Welsh) are likely to use phonological decoding skills when reading (after Wimmer and Goswarmi, 1994). However, very little is known about reading strategies in biliterate adults who learned contrasting orthographies at an early age. The present study aims to address this issue by testing whether adult Welsh-English bilinguals who learned two languages differing in their orthographic transparency use different reading strategies to monolingual English speakers. The experiment compared reading latency and production accuracy in 6 English monolingual and 6 early simultaneous Welsh-English bilingual participants. We also explored the effects of word length, gender and participant reading habits. Participants were asked to read aloud a low frequency English word list and a nonsense word list designed around contrastive phonemes in Welsh and English (e.g. <u, f, dd, th, y, ll, r, ch>). We coded productions as ‘Welsh expected’ (based on direct grapheme-phoneme correspondence), ‘English expected’ (based on pronunciation of similar English words) and ‘unexpected’. While there was no between-group difference in reading latency, preliminary analyses of the pronunciation accuracy show that the groups appear to be using different reading strategies in the non-word reading task. While the monolinguals produced English expected outcomes in 86% of productions, the bilinguals produced a Welsh expected outcome 48% of the time, and an English outcome 38% of the time. We analyse this as suggesting that the Welsh speakers rely on a combination of direct word recognition and phonological decoding while the monolingual English group rely primarily on direct word recognition. This paper provides insight into reading skills in Welsh-English bilingual adults and addresses questions about reading skills in biliterate adults in general.