A group of higher than average ability learners of the phonetic characteristics of a foreign language attempted to imitate stimuli from Japanese in tasks involving long and complex phrases and shorter minimal pair contrasts. Three information modalities were used (aural, visual, orthographic), introduced progressively, in order to test the subjects' use of different input sources. Their performance is described and the conclusion is drawn that these speakers function largely within an aural modality, but do assimilate external, primarily visual, information, as reflected in rapid improvement in the rhythmic and timing characteristics of the imitations.