BACKGROUND: The present study describes a treatment approach that was applied to improve word retrieval for an individual with chronic Broca's aphasia. The procedure combined elements of loose training with aspects of structured treatment. Treatment involved presentation of semantic cues according to semantic feature analysis [SFA] ( Boyle and Coelho, 1995) , as well as a forward chaining technique as in response elaboration training [RET] ( Kearns, 1985) . The technique was intended to elicit the targeted response by activating a semantic network without inhibiting related or creative responses. AIMS: The purpose was to investigate whether the combined SFA-RET treatment approach would improve naming accuracy of object noun pictures. This study also focused on an issue related to degradation in word retrieval, but instead of frequency of occurrence the focus was familiarity. Familiarity was based on to how often the individual with aphasia encountered the object nouns in her everyday life. METHODS AND PROCEDURES: The individual studied, LP, was a 57-year-old, right-handed female who was 8 years post-onset of a left cerebrovascular accident and demonstrated a moderate to severe Broca's aphasia. LP's naming scores were consistent with a severe impairment and the majority of her paraphasic errors were semantic in nature. Treatment materials consisted of 30 photographs of object nouns, of which 15 represented "high-familiarity" objects and 15 represented "low-familiarity". The 30 pictures were divided into three sets of 10 pictures. Each set consisted of five high- and five low-familiarity objects. Two sets of pictures were designated as treatment stimuli and the third as the control set. The treatment programme was carried out in an ABA single subject design. After the baseline phase the individual with aphasia was treated for 3 hours per week for 6 weeks and was seen for three follow-up sessions 6 weeks later. OUTCOMES AND RESULTS: Results indicated that the combined SFA-RET treatment resulted in improved naming of the treatment pictures as well as the untreated control pictures. In addition, during the treatment phase naming accuracy and consistency were greater for the high-familiarity than the low-familiarity objects. During the follow-up phase it was noted that the treatment effect was maintained at a higher level for the treatment pictures than for the control pictures. A similar trend was observed for the high- versus the low-familiarity words. CONCLUSIONS: These findings indicate that the combined treatment approach described was effective in improving this individual's word retrieval of object nouns. However, it is unclear whether the combined approach was necessary or what the individual contributions of each approach were to the final outcome. Finally, it appeared that LP was able to generate semantic features associated with the high-familiarity words and name the stimulus pictures more easily than the low-familiarity words, perhaps because of her regular contact with those objects in her daily life. This finding supports the contention that personalising treatment stimuli can be an important adjunct to any treatment task.