Table 4.

Average information content and classification performance of models trained and evaluated on test sets with rhythms from Germany and China.

	Training set	Test set
		Identical priors		Empirical priors
		German	Chinese	German	Chinese
Information content	German	1.21	1.63	1.34	1.72
	Chinese	1.32	1.49	1.70	1.49
Classification	German	0.84	0.80	0.73	0.72
	Chinese	0.59	0.77	0.47	0.75

Results are reported for two different conditions. One in which training sets contain identical distributions of metrical categories, and one in which training sets contain empirical distributions of metrical categories.