Left-hemisphere stroke patients suffering from language and speech disorders are often able to sing entire pieces of text fluently. This finding has inspired a number of melody-based rehabilitation programs – most notable among them a treatment known as Melodic Intonation Therapy – as well as two fundamental research questions. When the experimental design focuses on one point in time (cross section), one may determine whether or not singing has an immediate effect on syllable production in patients with language and speech disorders. When the design focuses on changes over several points in time (longitudinal section), one may gain insight as to whether or not singing has a long-term effect on language and speech recovery. The current work addresses both of these questions with two separate experiments that investigate the interplay of melody, rhythm and lyric type in 32 patients with non-fluent aphasia and apraxia of speech (Stahl et al., 2011; Stahl et al., 2013). Taken together, the experiments deliver three main results. First, singing and rhythmic pacing proved to be equally effective in facilitating immediate syllable production and long-term language and speech recovery. Controlling for various influences such as prosody, syllable duration and phonetic complexity, the data did not reveal any advantage of singing over rhythmic speech. This result was independent of lesion size and lesion location in the patients. Second, patients with extensive left-sided basal ganglia lesions produced more correct syllables when their speech was paced by rhythmic drumbeats. This observation is consistent with the idea that regular auditory cues may partially compensate for corticostriatal damage and thereby improve speech-motor planning (Grahn & Watson, 2013). Third, conversational speech formulas and well-known song lyrics yielded higher rates of correct syllable production than novel word sequences – whether patients were singing or speaking rhythmically. Moreover, the training of conversational speech formulas led to long-term progress in the production of formulaic language. The patients were able to establish their own individual formulaic repertoire in order to communicate some basic needs in daily life. This finding may result from the fact that, according to present knowledge, the production of formulaic language engages bilateral neural networks including right frontotemporal areas, the right basal ganglia and, possibly, the right cerebellum (Hughlings-Jackson, 1878; Speedie et al., 1993; Ackermann et al., 1998; Van Lancker Sidtis & Postman, 2006; Sidtis et al., 2009). In other words, left-hemisphere stroke patients may use right-hemisphere language resources to produce conversational speech formulas, whether they are sung or rhythmically spoken. In conclusion, the current experiments highlight the critical role of rhythm and conversational speech formulas in melody-based aphasia therapy.