Knowledge acquisition is governed by striatal prediction errors

Alex Pine,Noa Sadeh,Avi Mendelsohn,Aya Ben-Yakov,Yadin Dudai

doi:10.1038/s41467-018-03992-5

Abstract

Discrepancies between expectations and outcomes, or prediction errors, are central to trial-and-error learning based on reward and punishment, and their neurobiological basis is well characterized. It is not known, however, whether the same principles apply to declarative memory systems, such as those supporting semantic learning. Here, we demonstrate with fMRI that the brain parametrically encodes the degree to which new factual information violates expectations based on prior knowledge and beliefs—most prominently in the ventral striatum, and cortical regions supporting declarative memory encoding. These semantic prediction errors determine the extent to which information is incorporated into long-term memory, such that learning is superior when incoming information counters strong incorrect recollections, thereby eliciting large prediction errors. Paradoxically, by the same account, strong accurate recollections are more amenable to being supplanted by misinformation, engendering false memories. These findings highlight a commonality in brain mechanisms and computational rules that govern declarative and nondeclarative learning, traditionally deemed dissociable.

Highlights

Discrepancies between expectations and outcomes, or prediction errors, are central to trial-and-error learning based on reward and punishment, and their neurobiological basis is well characterized
In the face of reward and punishment, has been shown to accord with normative computational models, such as the Rescorla–Wagner model of Pavlovian conditioning[4]. Axiomatic to these models is the generation of outcome-based prediction errors (PEs) that update ensuing expectations in proportion to their magnitude. They have been well documented in the brain[5,6], where it has been posited that the phasic activity of dopaminergic neurons in the ventral tegmental area (VTA) encodes PEs for reward, firing in response to unpredicted reward and pausing in response to unexpected omission of reward[7]
Enhanced medial temporal lobe (MTL) activity in functional magnetic resonance imaging (fMRI) is evoked by simple perceptual stimuli deemed surprising, or which violate expectations based on prior statistical regularities/associations, such as a change in the temporal order of stimuli presented in a repeating sequence[20,21,22,23,24,25]

Summary

Introduction

Discrepancies between expectations and outcomes, or prediction errors, are central to trial-and-error learning based on reward and punishment, and their neurobiological basis is well characterized. We demonstrate with fMRI that the brain parametrically encodes the degree to which new factual information violates expectations based on prior knowledge and beliefs—most prominently in the ventral striatum, and cortical regions supporting declarative memory encoding These semantic prediction errors determine the extent to which information is incorporated into long-term memory, such that learning is superior when incoming information counters strong incorrect recollections, thereby eliciting large prediction errors. Rather than enhancing memory, repetition bears no consequence and can even be detrimental to long-term retention[16,17,18] Such anomalies would be explicable if prediction-error-based updating is globally applicable to memory, whereby the surprise of new information in light of what we already know and believe should serve as a prominent driving force of learning. Pharmacological intervention in humans, with L-dopa and monoaminergic stimulants which enhance dopamine transmission, has been shown to improve declarative memory[35,36,37]

Objectives

Methods

Results

Conclusion