Orthography is omnipresent in everyday life, and an increasing research pool investigates its effect on L2 speakers’ speech (Bassetti et al., 2015). Orthography has facilitative effects on word learning and phonological accuracy (Bürki et al., 2019), but may be detrimental on phonetic aspects of speech production and perception (see Bassetti et al., 2015 for an overview). The focus of past research has been on learners who acquired their L2 from oral and written input simultaneously, and it remains unclear whether orthography likewise impacts bilinguals who acquired both languages before the onset of literacy. The present study is set out to test whether the orthography of a dominant language impacts on speech perception and production in Spanish-Basque childhood-bilinguals in the Spanish Basque Country. Spanish and Basque both have the grapheme <z>, which represents the voiceless interdental fricative /θ/ in Spanish, but the voiceless lamino-dental fricative /s̻/ in Basque. In this study, we tested whether this conflicting grapheme-to-phoneme link leads Spanish-Basque childhood bilinguals to accept mispronunciations of Basque /s̻/ as Spanish /θ/, and whether they produce similar mispronunciations. In addition, we tested whether the order of acquisition is associated with participants’ acceptance of orthographically-guided mispronunciations. Sixty bilinguals (30 L2-Basque, *M*age=22 years; 30 L1-Basque, *M*age=23 years) participated in a lexical decision task (LDT) and two speech production tasks. The LDT consisted of auditorily-presented words and pseudowords in Basque. The pseudowords were created by replacing the critical phoneme /s̻/ with the interdental fricative /θ/ (orthographic condition) or the voiceless velar fricative /x/ (control condition). The results reveal a significant effect of Condition (*β*=1.022,* SE*=0.340,* z*=3.011,* p*=.003), showing that participants in both groups are less likely to detect mispronunciations in the orthographic condition than in the control condition. To rule out that these results are limited to the possibility that participants are frequently exposed to orthographically-guided mispronunciations, participants read words including the critical grapheme in Basque and Spanish. The results show that participants in both groups produce Basque /s̻/ and Spanish /θ/ distinctly (*β*=-703.050,*SE*=69.700, * t*=-10.086,* p*<.001), which makes it unlikely that the results of the LDT are driven by inaccurate input. To conclude, even highly proficient bilinguals who acquired both languages before the onset of literacy are affected by orthography in their speech perception. Interestingly, the effect of Spanish orthography on Basque speech perception was present in L1-Basque and L2-Basque speakers alike. While the effect was robust in perception, it was not observed in production, which is in contrast to previous findings on *late* L2-learners (Bassetti et al., 2015; Bürki et al., 2019). Taken together, these findings suggest that orthography plays a crucial role in the speech system of highly proficient bilinguals, though it does not automatically lead to non-native productions in childhood-bilinguals. To understand the bilingual speech system, it is crucial to incorporate orthography into models on bilinguals’ speech perception and production. Bassetti, B., Escudero, P., & Hayes-Harb, R. (2015). Second language phonology at the interface between acoustic and orthographic input. *Applied Psycholinguistics, 36*, 1–6. Bürki, A., Welby, P., Clément, M., & Spinelli, E. (2019). Orthography and second language word learning: Moving beyond “friend or foe?”. *Journal of the Acoustical Society of America, 145*, EL265.