As with compression and vocoding, synthesized speech can initially be quite difficult to understand. Nusbaum and his colleagues (Fenn, Nusbaum, & Margoliash, 2003; Greenspan, Nusbaum, & Pisoni, 1988) examined perceptual learning by presenting listeners with speech generated by a relatively crude speech synthesizer and looking for improved transcription performance as a function of experience with the synthetic speech.