A multiple baseline design was used to assess the effects of a treatment programme for sound errors with a speaker with moderately severe apraxia of speech (AOS) and Broca's aphasia. Treatment consisted of training correct production of three groups of sounds (i.e. stops, fricatives, and glides/liquids) in sentences containing multiple exemplars of those sounds. The treatment combined modelling, repetition, integral stimulation, visual cueing, and response-contingent feedback and was applied sequentially to the groups of sounds. Acquisition effects of treatment were measured by evaluating production of trained sentences in probes. Response generalization effects were assessed by examining sound production in untrained sentences containing exemplars of trained sounds and untrained sentences containing untrained sounds. Treatment resulted in improved production for trained sound groups, with response generalization closely following acquisition effects. Generalization across sound groups was negligible. Additionally, measures of sentence duration were conducted for sentences produced in two baseline, one mid-treatment, and two end-of-treatment probes. Statistically significant reductions in duration were noted at the completion of treatment in comparison to baseline measures.