Behavioral Analysis of Mice in a Virtual Navigation Decision Task

Abstract

Mice were trained on an accumulation of evidence task in a virtual reality, in which they had to integrate evidence of left and right cues presented in the cue zone of a straight runway. After a short memory zone, in which cues were absent, they had to turn into either a left or right branch at the end of the maze depending on which side had previously more cues. Mice were trained in two different versions of the task. The version with the shorter cue zone as well as shorter memory zone and total length enabled more mice to accomplish the task by advancing to the goal maze. A logistic regression model predicted how mice weighed cues in different sections of the cue zone was and showed that mice do not integrate over all cues equally. The mouse data was then analyzed with regard to the Signal Detection Theory Model, previously developed by Scott and Constantinople et al. [26]. The Signal Detection Theory assumes a Gaussian probability distribution for what the mouse perceives the number of cues to be [10]. The standard deviation of this Gaussian was found to scale linearly to the number of cues for rats [26]. A similar analysis was conducted for mouse data and the standard deviation was furthermore found to scale linearly to the number of cues for mice. . . .