Flexible long-term memory representations support adaptive behaviour

Memory allows us to draw on experiences from minutes, days, or weeks ago to guide adaptive behaviour. For example, by drawing on memory of the subway system, we can anticipate the order of subway stops and flexibly adjust our plans when the route changes. In this talk, I will describe a body of research that explores how we  flexibly form and use structured knowledge about the world to guide behaviour—particularly, how we make predictions about future events. First, I will present two studies that demonstrate how we encode memories that prioritize relevant aspects of our environment. I will show how both anticipatory attention before an event begins and reward after an event ends can modulate memory, illustrating how factors surrounding encoding adaptively shape memory for important information. Next, I will describe how we retrieve memories to support future-oriented behaviour. I will discuss empirical work showing that temporal structure is flexibly represented and updated in the hippocampus to guide future predictions, and that these temporal representations are strengthened over time to build durable models of the world. Together, these studies reveal how we form memories that flexibly prioritize relevant aspects of our environment, and how we adaptively use those memories to make predictions, shedding light on how memory systems are optimized to support adaptive behaviour.

Alternate locations:

Online: https://utoronto.zoom.us/j/88523051719