I will talk about recent and ongoing research in our lab on the ways in which auditory and written input interact in second language lexical-phonological acquisition. In German, final obstruents are devoiced such that underlying voicing contrasts are (mostly) neutralized. For example, /rad/ ‘wheel’ and /rat/ ‘advice’, spelled <Rad> and <Rat>, are both pronounced [rat]). In a previous study, colleagues and I have found that written forms in the input to native English learners can interfere with their acquisition of target-like pronunciation of underlyingly voiced final obstruents (e.g., the letter <d> in <Rad> caused learners to misremember the surface voicing of the final obstruent). However, these learners were not exposed to the alternations; that is, they did not also learn suffixed (plural-like) forms like [raden] and [raten] in which the underlying voicing contrast is maintained. More recently, we have investigated whether exposure to both the singular and plural forms of new words affects acquisition of final devoicing, and whether it interacts with the previously-observed interference effect of written input. We taught native English speakers with no prior German learning experience a set of singular and plural(-like) German nonwords with and without alternating surface forms, (e.g., [trop, troben] spelled <trob, troben> and [krat, kraten] spelled <krat, kraten>) along with pictured ‘meanings’ (e.g., butterfly, boot) in two conditions: with and without accompanying written forms. In a subsequent picture naming task, we measured the proportion of the time that learners produced underlyingly voiced final obstruents as voiceless. We predicted that all participants would produce underlyingly voiced obstruents as voiceless more often in singular than plural forms, but that the participants who were exposed to written input would be more likely than those who were not to (inappropriately) produce word-final obstruents as voiced. Both hypotheses were supported, suggesting that the availability of the surface alternation in the auditory input led participants to learn the final devoicing rule to some extent, and further that the availability of potentially misleading orthographic input interferes with the acquisition of the German pattern of surface voicing even when auditory evidence of the alternation is available. In a follow-up experiment we are currently investigating the acquisition of the process of final devoicing by determining whether participants devoice finally when they only have access to the plural forms of new words. I will discuss the findings of these studies as they relate to the effects of various sources of input in second language acquisition.