Background: The aim of lexical retrieval treatment for people with anomia is not just to improve accessibility of lexical items for confrontation naming but to carry over this improvement to communicative situations. However, there is no consensus on what measures are the most suitable to evaluate whether such changes have occurred. Anomia is one of the core presenting symptoms for people with primary progressive aphasia (PPA), yet while there is increasing evidence for the efficacy of word retrieval treatments in PPA, there is minimal information about how improvements in picture naming transfer to connected speech. Aims: This paper evaluates a word retrieval treatment targeting personally relevant words conducted with two people with PPA, focusing on patterns of generalisation. Methods & Procedures: The study is a single-blind single-case experimental design. Participants and primary communication partners identified words related to two topics that are personally relevant for conversation. Participants completed three baseline and one or two post-test measures of their ability to produce these words in picture naming and a structured interview. Therapy activities lasting 10–15 min/day were carried out over a 2-week period at home, with participants viewing a picture of each target word on a computer screen together with its written and spoken name and repeating/reading the name. Half of the words from one topic were treated; a matched set of words from the treated topic and words from the untreated topic served as controls to allow us to assess generalisation within and across topic. Outcomes & Results: At post-test, the participants’ naming improved for the treated items, showing generalisation to different pictures of these items. There was neither generalisation to untreated items, nor to retrieval of the same words in a structured interview. Conclusions: Further research is required to understand the limits and the potential of word retrieval treatments to improve conversation in both acute-onset and progressive aphasia.