Several studies have reported improved naming using the semantic feature analysis (SFA) approach in individuals with aphasia. Whether the SFA can be modified and still produce naming improvements in aphasia is unknown. The present study was designed to address this question by using a modified version of the SFA approach. Three, rather than the typical six, features were used, and written along with verbal responses were allowed in an individual with both aphasia and apraxia of speech. A single-subject multiple-baseline design across behaviors was used to treat naming of single objects across three different semantic categories in a 72-year-old individual with aphasia and apraxia of speech. Stimulus generalization of training was measured by using photographs of trained items presented in natural contexts. Training of the three different categories resulted in improved naming. At a 6-week follow-up session, naming remained above pre-treatment levels but declines were noted compared to treatment levels. Generalization to the same trained items presented in different contexts was also demonstrated although declines in performance were also noted over time. Results of the study provide qualified support for the use of three features in promoting long-term improvement of naming in an individual with both aphasia and apraxia of speech. Future SFA studies should focus on whether it is the number or types of features used, aphasia severity, or length of treatment that are critical factors in rehabilitating naming deficits in aphasia.