Model fits to Mandarin Chinese
	0.61 free parameters per syllable, 13 Hz RMS error.