< Back to Article

A Generative Model for Measuring Latent Timing Structure in Motor Sequences

Figure 5

The influence of additional global factors on model fit and song timing.

A, Goodness of model fit with additional global factors included (black) vs. 1 global factor (white) for the 5 birds for which BIC was lowest for factor. Numbers above black bars indicate the number of factors associated with the lowest BIC for that bird. B, Distribution of global parameters for the 2nd factor (sign-normalized, see Methods), separated by whether the song segment was a syllable or silent gap. Across birds syllable and gap parameters tended to be of opposite sign. C, Timing covariance matrices generated for the 2nd global factor in 3 representative birds. D, Average changes in the 2nd global variable in the 3 birds that showed significant circadian variation for that factor. The circadian pattern showed that syllables tended to elongate at the expense of gaps over the afternoon. E, Example of day-to-day drift in the 2nd global variable for one bird over a 1-month period. As in figure 4, hour-of-day and daily averages in D and E have been adjusted for unequal sampling of factor combinations while dotted lines in E represent adjusted means ± standard error.

doi: https://doi.org/10.1371/journal.pone.0037616.g005