Hanley and Kay (1997) reported the case of a patient (PS) who showed effects of imageability on tests of auditory repetition, but whose errors were phonological rather than semantic. They argued that this pattern of performance could be explained in terms of a partial impairment to both lexical and nonlexical repetition routes so long as some interaction between the two routes was allowed, consistent with Hillis and Caramazza's (1991) "summation" hypothesis. The present paper investigates the performance of a new patient (MF) who also makes a large number of phonological errors when repeating words of low imageability. MF performs at a similar level to PS on tests of picture naming, but is less impaired than PS when repeating words and nonwords. It is argued that the pattern of performance that MF demonstrates on these and on a wide range of other tests of word production and comprehension can be readily accommodated in terms of the dual-route account of impaired auditory repetition that Hanley and Kay (1997) put forward. However, we argue that it is difficult to explain these patients' performance in terms of models of auditory word repetition that do not incorporate a separate nonlexical repetition route (Dell, Schwartz, Martin, Saffran, & Gagnon, 1997; Foygel & Dell, 2000).