Importance of electrophysiological signal features assessed by classification trees

Sustained activity in prefrontal cortex is associated with the maintenance of information during short-term memory (STM). We have used impurity reduction criteria of classification trees to investigate how the behavioral performance of a monkey during STM is reflected in the information content of three features of recorded signals: rates of individual neurons, oscillations in the LFP, and oscillations in the spiking activity. The LFP power in all bands, but in the αα and ββ bands in particular, is more informative than the firing rate of neurons and the spike power with respect to the monkey's performance.