This article is a single-case investigation of phonological naming therapy. The individual involved had fluent jargon speech, with neologisms, verbal paraphasias, and paragrammatisms. The jargon was underpinned by a severe anomia. Content words were rarely accessed either in spontaneous speech or naming. Single word investigations highlighted some preserved skills. Auditory comprehension, at least for concrete words, was relatively intact and although nonwords could not be repeated, words could, and at a level which was far superior to naming. The patient also had some ability to respond to phonological cues. These results suggested that phonological representations were preserved and that there were some intact semantic abilities. It seemed that the naming disorder was primarily due to an inability to access phonology from semantics. Therapy took a phonological approach. The patient was encouraged to reflect upon the syllabic structure and first phoneme of pictured targets. Subsequently, she was required to use this partial phonological knowledge as a self-cue. It was hypothesized that this therapy might equip the subject with a self-cuing naming strategy. Posttherapy investigations of naming demonstrated dramatic improvements, which generalized to untreated items. However, there was little evidence that these were due to a self cuing strategy. Performance on phonological judgment and discrimination assessments, which required conscious phonological reflection, was unchanged, and there were no signs that the patient was self-cuing during naming. Reasons for these paradoxical results are discussed. (JINS, 1998, 4, 675–686.)