Sharing the spotlight: Uncovering common attentional dynamics across species
Fig 2
Attentional metrics and models.
A) Five variables (middle) were derived from the VR paths: three concerning task accuracy (hit rate, precision, and bias) and two concerning speed (running speed and reaction time, RT). The correlation matrix of these variables (left) confirmed that the variables concerning the classes accuracy and speed were highly correlated within each class, but largely independent between the two classes, though precision showed only a weak association with either. The variables were therefore entered into two separate HMMs, the conjunction of which partitioned the data into four categories (right). B) Plotting a representative variable from each of the two classes (hit rate and reaction time) against each other and colouring the data by the four states, confirmed a correct partitioning of the data by the HMM conjunction, for humans (left), monkeys (middle), and mice (right). Legend of state colours: see C. C) The contribution of the five variables towards the partitioning into four states by the conjuncted HMMs was very similar between humans (left) and monkeys (middle). In mice (right), the precision variable had slightly different contributions towards the four states, but the overall pattern was similar to that of humans and monkeys. For this analysis, reaction time, bias and precision were inverted (marked as inv. in the labels) so that higher values in all metrics represent higher performance. All metrics were normalised to each species’ average values.