The purpose of this study was to assess the role that phonological, orthographic, and contextual sources of information play in a group of adults who were learning to read compared to adult skilled readers. Participants read short paragraphs that contained a correct homophone, an incorrect homophone, or a spelling control. Target words were orthographically similar or dissimilar, and they appeared in context that predicted the target or was neutral with respect to the target. The pattern of data obtained for skilled readers was consistent with past work (Rayner et al., "Journal of Experimental Psychology: Learning, Memory and Cognition" 24(2), 476-497, 1998). Skilled readers showed no reading time differences between the correct homophone and the incorrect homophone, as long as the two were orthographically similar, but reading times on these words were faster than the spelling control condition. The pattern of data for the adults who were learning to read was different. These readers were actually better at noticing that an incorrect version of the homophone was present. Importantly, we did find consistent significant differences between the incorrect homophone condition and the spelling control condition. This suggests the adults who were learning to read use phonological codes during word recognition, but they do so less efficiently than skilled readers.