A number of large-scale trials have established that language therapy with acquired aphasic patients can result in significant improvement. However, such trials use a variety of different treatments with patients with qualitatively varying disorders. The group results give no information about the treatments that were effective for particular types of problem. More recent studies of treatment have examined the effects of more closely defined treatments for more closely defined disorders. Treatment based on the facilitation of word retrieval show quite long-lasting effects from limited amounts of treatment, when the treatment gives either semantic or phonological information about the word, but the improvements are mostly limited to the items involved in treatment. The establishment of strategies for word retrieval based on patients' retained abilities results in more generalized improvement. The need for studies that relate analysis of a patient's disorder more closely to the process of treatment is discussed.