Memory Guided Prediction

The ability to remember prior experience is a core aspect of cognition. Memory can be comprised of information about individual items or contexts, but information about the temporal relationships between entities in the environment can provide not only a record of what happened in the past but also facilitate predictions about what might happen in the future. To address how temporal relationships are learned and used to facilitate predictions, the current work includes three investigations, two experiments using functional magnetic resonance imaging (fMRI) and one experiment using behavior. The reported fMRI experiments tested neural predictions from a recent model positing the existence of dissociable posterior-medial (PM) and anterior-temporal (AT) cortical networks that are thought to represent spatio-temporal relationships and item information, respectively, in the service of memory-guided behavior (Ranganath and Ritchey, 2012). In Chapter 2, representational similarity analysis was used to test for evidence of PM network involvement in supporting representations of shared temporal structure across event sequences. In Chapter 3, computational model-based fMRI was employed to assess evidence for the involvement of the AT network in representing the shared meaning of objects in a task where shared meaning could be discerned by learning to predict outcomes associated with sequentially presented pairs of object cues. In Chapter 4, we tested for behavioral evidence of increased flexibility of predictions about temporally associated stimuli following a period of memory trace stabilization. Together, the results across these investigations extend our understanding of the neural and behavioral correlates of learning and using temporal information to guide expectations and predictions about upcoming information.

Theoretical reflections on memory and prediction, linking these concepts to the role of the cerebellum in higher cognition. What is memory? What is memory for? Where is memory in the brain? Although memory is probably the most studied function in cognition, these fundamental questions remain challenging. We can try to answer the question of memory's purpose by defining the function of memory as remembering the past. And yet this definition is not consistent with the many errors that characterize our memory, or with the phylogenetic and ontogenetic origin of memory. In this book, Tomaso Vecchi and Daniele Gatti argue that the purpose of memory is not to remember the past but to predict the future. Vecchi and Gatti link memory and prediction to the role of the cerebellum in higher cognition, relying on recent empirical data to support theoretical reflections. They propose a new model of memory functions that comprises a system devoted to prediction, based in the cerebellum and mediated by the hippocampus, and a parallel system with a major role for cortical structures and mediated by the amygdala. Although memory is often conceived as a kind of storehouse, this storehouse is constantly changing, integrating new information in a continual process of modification. In order to explain these characteristics, Vecchi and Gatti argue, we must change our interpretation of the nature and functions of the memory system.

"An argument to view memory as predicting the future, rather than merely archiving the past. Based on experimental evidence from psychology and neuroscience"--

For a variety of reasons, cognitive scientists tend to divide the study of the brain into separate domains. Some of the most studied domains are episodic memory, visual perception, and prediction. However, the underlying processes that give rise to one's memories, perceptions, and predictions do not operate independently. Instead each of these processes influences the others. Previous research has demonstrated that the memory system is critical for the generation of predictions (Buckner, 2010; Corbit & Balleine, 2000; Johnson, Meer, & Redish, 2007; Lisman & Redish, 2009). Other researchers have argued that portions of the memory system are actually a part of the visual perception system (Baxter, 2009; Bussey & Saksida, 2007). Here we present research that furthers our understanding of how these processes interact. First we demonstrate, through a series of behavioral experiments, that predictions influence memory. Specifically, in high-predictability environments, we find enhanced memory for items that could have been predicted to recur, but did not. We then present a computational model of recognition memory that attempts to formally explain this pattern using a prediction-based learning mechanism. Finally, we present neural evidence that the information content of the perceptual system is contingent on subsequent memory. Taken together, these results demonstrate how the perception of present stimuli and the prediction of future stimuli interact with the memory system.

Re-examines frequency, entrenchment and salience, three foundational concepts in usage-based linguistics, through the prism of learning, memory, and attention.

Adaptive decision-making is governed by at least two types of memory processes. On the one hand, learned predictions through integrating multiple experiences, and on the other hand, one-shot episodic memories. These two processes interact, and predictions - particularly prediction errors - influence how episodic memories are encoded. However, studies using computational models disagree on the exact shape of this relationship, with some findings showing an effect of signed prediction errors and others showing an effect of unsigned prediction errors on episodic memory. We argue that the choice-confirmation bias, which reflects stronger learning from choice-confirming compared to disconfirming outcomes, could explain these seemingly diverging results. Our perspective implies that the influence of prediction errors on episodic encoding critically depends on whether people can freely choose between options (i.e., instrumental learning tasks) or not (Pavlovian learning tasks). The choice-confirmation bias on memory encoding might have evolved to prioritize memory representations that optimize reward-guided decision-making. We conclude by discussing open issues and implications for future studies.