Language-specific and language-general aspects of speech rhythm were explored in a repetitive speech production task, whereby native English and Japanese speakers repeated phrases in time with an auditory metronome. Subjects were instructed to repeat each phrase (e.g., English ``Give the dog a bone'') once per metronome beep, aligning the beginning of the phrase with each beep. The rate of repetition was systematically manipulated by varying the metronome rate from slow to fast across trials. Rhythmic properties of the utterances were investigated by marking the vowel onset of each syllable, and expressing it in terms of phase with respect to the beginning of the current repetition and that of the following repetition. Preliminary results indicate that, while there was considerable variability across speakers and repetitions, there was a language-general tendency for prominent syllables to cluster around rhythmically motivated phases, such as 1/2 or 2/3. For a subset of the data, abrupt jumps from one stable phase to another were also found, suggesting quasi-categorical modes in rhythmic speech production. The data further address language-specific issues such as the nature of stress-timing in English, and the role of supra-moraic, ``foot''-level units in Japanese rhythm.