Recent work has demonstrated that retrieval practice, or practice retrieving the names for depicted objects from long-term memory, confers more persistent benefits to naming performance in aphasia compared to word repetition practice (i.e., errorless learning). Though naming impairment in people with aphasia is a well-researched topic due to its prevalence across the aphasia population, word comprehension deficits, or problems reliably retrieving the meanings of words, are studied far less. The present study examined the effects of two factors on the efficacy of word comprehension treatment in a group study of people with aphasia: (a) retrieval practice versus errorless learning; (b) modality of training (production versus comprehension). Training effects were assessed at one day and one week following either training modality. Mixed regression results across the group revealed better word comprehension performance following comprehension-based training when compared to production-based training. Retrieval practice did not outperform non-retrieval based practice in either the comprehension or production modality in terms of post-training performance. However, retrieval practice in the production modality was associated with superior retention of gains between two post-training test timepoints, compared to non-retrieval-based practice and untreated control words. Implications for treatment are discussed.