Mice in the Manhattan Maze

Overview

What is intelligence, and how does the brain create it? The neural basis of intelligent behavior remains one of the biggest mysteries in science, and evolution leaves clues in model organisms. This project seeks to uncover in laboratory mice conserved principles underlying complex cognition. The Manhattan Maze is a reconfigurable 3D labyrinth that allows systematic task designs in a vast space of 2¹²¹ possible maps. Naive wildtype mice improved within two rewards, approached efficient paths within about 20 rewards, retained a 9-turn route overnight, and learned faster in new configurations. Much of this few-shot improvement follows from a rule-based forward bias that emerged even before the first reward. In a maze with loops, where that rule is less useful, mice learned within a few rewards to prefer one bottleneck corridor far from the reward location, while their choices elsewhere stayed flexible. I also tested acortical mice, a structural mutant born without the hippocampus and most of the neocortex. They were impaired during initial exploration, where repetitive scanning made them about 3-fold slower to obtain the first few rewards, but past that stage their learning, retention over weeks to months, and generalization to new mazes were largely preserved. Learning this fast is hard to reconcile with trial and error, in which value propagates backwards from the reward. A neuromorphic circuit model instead accounts for the bottleneck choice: it builds a map of the environment without reward, and needs neither cortical nor hippocampal circuit motifs. In this task, then, acortical mice do not need cortex or hippocampus for learning and memory, and structure learned before the first reward, rather than the reward itself, may be what supports few-shot learning.