Speaker
Description
A central function of cognition is to detect relevant sensory signals to adapt behavior.
To this end, a specialized neural system in insects learns associations between relevant sensory inputs and beneficial changes in behavior, the mushroom body (MB). However, whereas the prevalent idea is that learning in the MB mainly occurs during the presence of explicit reinforcement, such as sucrose rewards or painful experiences, we here summarize recent experimental evidence that underscores the role of previous sensory experiences in shaping learning and adaptive behavior.
Specifically, we highlight that plasticity in the MB calyx, a large expansion layer that forms sensory representations, drives a rapid and lasting depression of sensory responses to repeated, predictable sensory stimuli. In contrast, first evidence suggests that it potentiates sensory responses to novel, unpredicted stimuli. We show that these plasticity mechanisms can be explained with the theory of dendritic predictive learning, where plastic inhibitory feedback suppresses predicted inputs, and modulates excitatory plasticity to exclusively enhance synaptic weights to unpredicted inputs. Moreover, we show that this plasticity is in line with observations in the cerebellum granule layer, where an expansion layer with a strikingly similar interneuron circuit and dendro-synaptic organization transforms inputs to drive adaptive behavior and motor learning.
Functionally, dendritic predictive learning enhances adaptive behavioral responses to salient and novel stimuli, and facilitates the detection of relevant inputs in the presence of distracting background stimuli. Thus, these results highlight a general plasticity mechanism that supports a simple form of curiosity in insects and beyond.
