The parabigeminal nucleus as a recursive estimator

The parabigeminal nucleus (PBN) is known to estimate the retinal position error (RPE) of an intended target. Recently it has been discovered that PBN activity continues to encode the extrapolated RPE of a ¿virtual¿ target, although less vigorously as compared with an actual target. Besides target movement, PBN activity also responds to eye movements generated by the animal itself that change the RPE of the virtual target. These phenomena imply the existence of an internal model within the PBN. We hypothesize that PBN performs recursive estimation akin to a Kalman filter, and manifests the characteristics of the computation in its spiking activities. This hypothesis is tested with a point process generalized linear model (GLM) of the PBN spike train. The results suggest the hypothesis is viable, and attribute the origination of PBN´s less vigorous responses to virtual target to the uncertainty of the estimation. Furthermore, this study provides a generalizable means for the neural computations that are essential for proper control of behavior and decision making to be embodied in spike train data.