Brief description of behavioural data collected from YouTube, and associations between them.

When videos had no sound-track or the soundtrack was replaced by music, missing values were recorded for data reliant on sound; similarly missing values were recorded for videos without relevant comments or where the behaviour could not be clearly seen. The proportion of tail-chasing videos (excluding those with missing values) showing each characteristic is displayed, along with any significant associations with relevant predictors, for which the odds ratios, degrees of freedom, and P-values are displayed. *This odds ratio was calculated from a model using ‘Habitual’ vs other frequencies as a binary variable.