Prediction error and memory across the lifespan.

The influence of Prediction Errors (PEs) on episodic memory has generated growing empirical and theoretical interest. This review explores how the relationship between PE and memory may evolve throughout lifespan. Drawing upon the predictive processing framework and the Predictive, Interactive Multiple Memory System (PIMMS) model in particular, the paper highlights the hierarchical organization of memory systems and the interaction between top-down predictions and bottom-up sensory input, proposing that PEs promote synaptic change and improve encoding and consolidation processes. We discuss the neuroscientific mechanisms underlying PE-driven memory enhancement, focusing on the involvement of the hippocampus, the entorhinal cortex-hippocampus pathway, and the noradrenergic sympathetic system. Recognizing the divergent trajectories of episodic and semantic memory across the lifespan is crucial when examining the effects of PEs on memory. This review underscores the heterogeneity of memory processes and neurocognitive mechanisms underlying PE-driven memory enhancement across age. Future research is suggested to directly compare neural networks involved in learning from PEs across different age groups and to contribute to a deeper understanding of PE-driven learning across age.

No evidence of fast mapping in healthy adults using an implicit memory measure: failures to replicate the lexical competition results of Coutanche and Thompson-Schill (2014).

Fast mapping (FM) is a hypothetical, incidental learning process that allows rapid acquisition of new words. Using an implicit reaction time measure in a FM paradigm, Coutanche and Thompson-Schill (Coutanche, M. N., & Thompson-Schill, S. L. (2014). Fast mapping rapidly integrates information into existing memory networks. Journal of Experimental Psychology: General, 143(6), 2296-2303. https://doi.org/10.1037/xge0000020) showed evidence of lexical competition within 10 min of non-words being learned as names of unknown items, consistent with same-day lexicalisation. Here, Experiment 1 was a methodological replication (N = 28/group) that found no evidence of this RT competition effect. Instead, a post-hoc analysis suggested evidence of semantic priming. Experiment 2 (N = 60/group, online study, pre-registered on OSF) tested whether semantic priming remained when making the stimulus set fully counterbalanced. No evidence for either lexical competition nor semantic priming was detected. Experiment 3 (n = 64, online study, pre-registered on OSF) tested whether referent (a)typicality boosted lexical competition (Coutanche, M. N., & Koch, G. E. (2017). Variation across individuals and items determine learning outcomes from fast mapping. Neuropsychologia, 106, 187-193. https://doi.org/10.1016/j.neuropsychologia.2017.09.029), but again no evidence of lexical competition was observed, and Bayes Factors for the data combined across all three experiments supported the hypothesis that there is no effect of lexical competition under FM conditions. These results, together with our previous work, question whether fast mapping exists in healthy adults, at least using this specific FM paradigm.

Unveiling coding of prediction and sharpening of perceptual features through multivariate pattern analysis.

A recent study by Blank, Alink and Büchel, uses multivariate neuroimaging to investigate how the human brain processes the strength of face-related expectations and explores whether these expectations are represented in the same regions that process facial stimuli. In line with predictive coding theories, their study presents compelling evidence that the brain adjusts its processing based on the certainty of expectations. This occurs exclusively within high-level face-sensitive regions, rather than across the entire processing hierarchy. Here we critically discuss these findings and outline potential directions for future research to better understand how the human brain expects, processes, and perceives images.

Shape of U: The Nonmonotonic Relationship Between Object-Location Memory and Expectedness.

The schema-linked interactions between medial prefrontal and medial temporal lobe (SLIMM) model predicts that memory for object locations is a U-shaped function of the expectancy of those locations. Using immersive virtual reality, we presented participants with 20 objects in locations that varied in their congruency with a kitchen schema. Bayes factors across four experiments (137 adults in total) confirmed the (preregistered) prediction of better memory for highly expected and unexpected locations relative to neutral locations. This U shape was found in location recall and in forced-choice recognition in which the foil locations were matched for expectancy, controlling for the bias toward guessing expected locations. A second prediction was that the two ends of the U shape are associated with different expressions of memory: recollection of unexpected locations and familiarity for expected locations. BFs, propagated across experiments, provided evidence against this second prediction; recollection was associated with both ends of the U shape. These findings further constrain theories about the role of schema in episodic memory.

Modified MRI Anonymization (De-Facing) for Improved MEG Coregistration.

Localising the sources of MEG/EEG signals often requires a structural MRI to create a head model, while ensuring reproducible scientific results requires sharing data and code. However, sharing structural MRI data often requires the face go be hidden to help protect the identity of the individuals concerned. While automated de-facing methods exist, they tend to remove the whole face, which can impair methods for coregistering the MRI data with the EEG/MEG data. We show that a new, automated de-facing method that retains the nose maintains good MRI-MEG/EEG coregistration. Importantly, behavioural data show that this "face-trimming" method does not increase levels of identification relative to a standard de-facing approach and has less effect on the automated segmentation and surface extraction sometimes used to create head models for MEG/EEG localisation. We suggest that this trimming approach could be employed for future sharing of structural MRI data, at least for those to be used in forward modelling (source reconstruction) of EEG/MEG data.

Explicitly predicting outcomes enhances learning of expectancy-violating information.

Predictive coding models suggest that the brain constantly makes predictions about what will happen next based on past experiences. Learning is triggered by surprising events, i.e., a prediction error. Does it benefit learning when these predictions are made deliberately, so that an individual explicitly commits to an outcome before experiencing it? Across two experiments, we tested whether generating an explicit prediction before seeing numerical facts boosts learning of expectancy-violating information relative to doing so post hoc. Across both experiments, predicting boosted memory for highly unexpected outcomes, leading to a U-shaped relation between expectedness and memory. In the post hoc condition, memory performance decreased with increased unexpectedness. Pupillary data of Experiment 2 further indicated that the pupillary surprise response to highly expectancy-violating outcomes predicted successful learning of these outcomes. Together, these findings suggest that generating an explicit prediction increases learners' stakes in the outcome, which particularly benefits learning of those outcomes that are different than expected.

A predictive account of how novelty influences declarative memory.

A rich body of studies in the human and non-human literature has examined the question how novelty influences memory. For a variety of different stimuli, ranging from simple objects and words to vastly complex scenarios, the literature reports that novelty improves memory in some cases, but impairs memory in other cases. In recent attempts to reconcile these conflicting findings, novelty has been divided into different subtypes, such as relative versus absolute novelty, or stimulus versus contextual novelty. Nevertheless, a single overarching theory of novelty and memory has been difficult to attain, probably due to the complexities in the interactions among stimuli, environmental factors (e.g., spatial and temporal context) and level of prior knowledge (but see Duszkiewicz et al., 2019; Kafkas & Montaldi, 2018b; Schomaker & Meeter, 2015). Here we describe how a predictive coding framework might be able to shed new light on different types of novelty and how they affect declarative memory in humans. More precisely, we consider how prior expectations modulate the influence of novelty on encoding episodes into memory, e.g., in terms of surprise, and how novelty/surprise affect memory for surrounding information. By reviewing a range of behavioural findings and their possible underlying neurobiological mechanisms, we highlight where a predictive coding framework succeeds and where it appears to struggle.

Tau pathology in early Alzheimer's disease is linked to selective disruptions in neurophysiological network dynamics.

Understanding the role of Tau protein aggregation in the pathogenesis of Alzheimer's disease is critical for the development of new Tau-based therapeutic strategies to slow or prevent dementia. We tested the hypothesis that Tau pathology is associated with functional organization of widespread neurophysiological networks. We used electro-magnetoencephalography with [18F]AV-1451 PET scanning to quantify Tau-dependent network changes. Using a graph theoretical approach to brain connectivity, we quantified nodal measures of functional segregation, centrality, and the efficiency of information transfer and tested them against levels of [18F]AV-1451. Higher Tau burden in early Alzheimer's disease was associated with a shift away from the optimal small-world organization and a more fragmented network in the beta and gamma bands, whereby parieto-occipital areas were disconnected from the anterior parts of the network. Similarly, higher Tau burden was associated with decreases in both local and global efficiency, especially in the gamma band. The results support the translational development of neurophysiological "signatures" of Alzheimer's disease, to understand disease mechanisms in humans and facilitate experimental medicine studies.

Response to commentaries on our review of Fast Mapping in adults.

We thank all the commentators for their thoughts on our review of Fast Mapping (FM) in adults, where we questioned the evidence that FM is a distinct learning mechanism, and urged caution over the excitement generated by the original report of FM in adults with amnesia using the fast mapping paradigm (FMP) . While some commentators remain convinced that there is good evidence to support a FM process in adults, most reported a skepticism similar to ours. Here we respond to the main comments, and clarify some of the terms of debate.

Investigating Fast Mapping Task Components: No Evidence for the Role of Semantic Referent nor Semantic Inference in Healthy Adults.

Fast mapping (FM) is an incidental learning process that is hypothesized to allow rapid, cortical-based memory formation, independent of the normal, hippocampally dependent episodic memory system. It is believed to underlie the rapid vocabulary learning in infants that occurs separately from intentional memorisation strategies. Interest in adult FM learning was stimulated by a report in which adults with amnesia following hippocampal damage showed a normal ability to learn new object-name associations after an incidental FM task, despite their impaired memory under a conventional intentional memorization task. This remarkable finding has important implications for memory rehabilitation, and has led to a number of neuropsychological and neuroimaging studies in other patients and controls. Given this growing interest in adult FM, we conducted four behavioural experiments with healthy adults (N = 24 young or older adults in Experiments 1-3 using within-participant designs; N = 195 young adults in Experiment 4 using a between-participant design) that attempted to dissect which component(s) of the FM task are important for memory. Two key components of the FM task have been claimed to support FM learning: (1) provision of a known semantic referent and (2) requirement that the new association be inferred. Experiment 1 provided no evidence that removing the semantic referent impaired memory performance, while Experiment 2 provided no evidence that removing the semantic inference impaired performance. Experiment 3 was a replication of Experiment 2 with older participants, based on the hypothesis (from studies of amnesic individuals) that FM would be more effective following the hippocampal atrophy typical of increasing age, but again found no evidence that semantic inference is beneficial. Given potential concerns about contamination between tasks when each participant performed multiple variants of the FM task, we ran a final between-participant design in which each participant only ever did one condition. Despite 80% power and despite being able to detect better memory following intentional memorization in the explicit encoding (EE) control condition than in each of the FM conditions, we again found no evidence of differences between any FM conditions. We conclude that there is no evidence that the components hypothesized to be critical for FM are relevant to healthy adults.

Knowledge is power: Prior knowledge aids memory for both congruent and incongruent events, but in different ways.

Events that conform to our expectations, that is, are congruent with our world knowledge or schemas, are better remembered than unrelated events. Yet events that conflict with schemas can also be remembered better. We examined this apparent paradox in 4 experiments, in which schemas were established by training ordinal relationships between randomly paired objects, whereas event memory was tested for the number of objects on each trial. Better memory was found for both congruent and incongruent trials, relative to unrelated trials, producing memory performance that was a "U-shaped" function of congruency. The congruency advantage but not incongruency advantage was mediated by postencoding processes, whereas the incongruency advantage, but not congruency advantage, emerged even if the information probed by the memory test was irrelevant to the schema. Schemas therefore augment event memory in multiple ways, depending on the match between novel and existing information. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Little evidence for Fast Mapping (FM) in adults: A review and discussion.

Conventional memory theory proposes that the hippocampus is initially responsible for encoding new information, before this responsibility is gradually transferred to the neocortex. Therefore, a report in 2011 by Sharon et al. of hippocampal-independent learning in humans was notable. These authors reported normal learning of new object-name associations under a Fast Mapping (FM) procedure in adults with hippocampal damage, who were amnesic according to more conventional explicit memorisation procedures. FM is an incidental learning paradigm, inspired by vocabulary acquisition in children, which is hypothesised to allow rapid, cortical-based memory formation. In the years since the original report, there has been, understandably, a growing interest in adult FM, not only because of its theoretical importance, but also because of its potential to help rehabilitate individuals with memory problems. We review the FM literature in individuals with amnesia and in healthy adults, using both explicit and implicit memory measures. Contrary to other recent reviews, we conclude that the evidence for FM in adults is weak, and restraint is needed before assuming the phenomenon exists.

The missing link? Testing a schema account of unitization.

Unitization refers to the creation of a new unit from previously distinct items. The concept of unitization has been used to explain how novel pairings between items can be remembered without requiring recollection, by virtue of new, item-like representations that enable familiarity-based retrieval. We tested an alternative account of unitization - a schema account - which suggests that associations between items can be rapidly assimilated into a schema. We used a common operationalization of "unitization" as the difference between two unrelated words being linked by a definition, relative to two words being linked by a sentence, during an initial study phase. During the following relearning phase, a studied word was re-paired with a new word, either related or unrelated to the original associate from study. In a final test phase, memory for the relearned associations was tested. We hypothesized that, if unitized representations act like schemas, then we would observe some generalization to related words, such that memory would be better in the definition than sentence condition for related words, but not for unrelated words. Contrary to the schema hypothesis, evidence favored the null hypothesis of no difference between definition and sentence conditions for related words (Experiment 1), even when each cue was associated with multiple associates, indicating that the associations can be generalized (Experiment 2), or when the schematic information was explicitly re-activated during Relearning (Experiment 3). These results suggest that unitized associations do not generalize to accommodate new information, and therefore provide evidence against the schema account.

No effect of hippocampal lesions on stimulus-response bindings.

The hippocampus is believed to be important for rapid learning of arbitrary stimulus-response contingencies, or S-R bindings. In support of this, Schnyer et al. (2006) (Experiment 2) measured priming of reaction times (RTs) to categorise visual objects, and found that patients with medial temporal lobe damage, unlike healthy controls, failed to show evidence of reduced priming when response contingencies were reversed between initial and repeated categorisation of objects (a signature of S-R bindings). We ran a similar though extended object classification task on 6 patients who appear to have selective hippocampal lesions, together with 24 age-matched controls. Unlike Schnyer et al. (2006), we found that reversing response contingencies abolished priming in both controls and patients. Bayes Factors provided no reason to believe that response reversal had less effect on patients than controls. We therefore conclude that it is unlikely that the hippocampus is needed for S-R bindings.

Does prediction error drive one-shot declarative learning?

The role of prediction error (PE) in driving learning is well-established in fields such as classical and instrumental conditioning, reward learning and procedural memory; however, its role in human one-shot declarative encoding is less clear. According to one recent hypothesis, PE reflects the divergence between two probability distributions: one reflecting the prior probability (from previous experiences) and the other reflecting the sensory evidence (from the current experience). Assuming unimodal probability distributions, PE can be manipulated in three ways: (1) the distance between the mode of the prior and evidence, (2) the precision of the prior, and (3) the precision of the evidence. We tested these three manipulations across five experiments, in terms of peoples' ability to encode a single presentation of a scene-item pairing as a function of previous exposures to that scene and/or item. Memory was probed by presenting the scene together with three choices for the previously paired item, in which the two foil items were from other pairings within the same condition as the target item. In Experiment 1, we manipulated the evidence to be either consistent or inconsistent with prior expectations, predicting PE to be larger, and hence memory better, when the new pairing was inconsistent. In Experiments 2a-c, we manipulated the precision of the priors, predicting better memory for a new pairing when the (inconsistent) priors were more precise. In Experiment 3, we manipulated both visual noise and prior exposure for unfamiliar faces, before pairing them with scenes, predicting better memory when the sensory evidence was more precise. In all experiments, the PE hypotheses were supported. We discuss alternative explanations of individual experiments, and conclude the Predictive Interactive Multiple Memory Signals (PIMMS) framework provides the most parsimonious account of the full pattern of results.

Assumptions behind scoring source versus item memory: Effects of age, hippocampal lesions and mild memory problems.

Source monitoring paradigms have been used to separate: 1) the probability of recognising an item (Item memory) and 2) the probability of remembering the context in which that item was previously encountered (Source memory), conditional on it being recognised. Multinomial Processing Tree (MPT) models are an effective way to estimate these conditional probabilities. Moreover, MPTs make explicit the assumptions behind different ways to parameterise Item and Source memory. Using data from six independent groups across two different paradigms, we show that one would draw different conclusions about the effects of age, age-related memory problems and hippocampal lesions on Item and Source memory, depending on the use of: 1) standard accuracy calculation vs MPT analysis, and 2) two different MPT models. The MPT results were more consistent than standard accuracy calculations, and furnished additional parameters that can be interpreted in terms of, for example, false recollection or missed encoding. Moreover, a new MPT structure that allowed for separate memory representations (one for item information and one for item-plus-source information; the Source-Item model) fit the data better, and provided a different pattern of significant differences in parameters, than the more conventional MPT structure in which source information is a subset of item information (the Item-Source model). Nonetheless, there is no theory-neutral way of scoring data, and thus proper examination of the assumptions underlying the scoring of source monitoring paradigms is necessary before theoretical conclusions can be drawn.

The effects of hippocampal lesions on MRI measures of structural and functional connectivity.

Focal lesions can affect connectivity between distal brain regions (connectional diaschisis) and impact the graph-theoretic properties of major brain networks (connectomic diaschisis). Given its unique anatomy and diverse range of functions, the hippocampus has been claimed to be a critical "hub" in brain networks. We investigated the effects of hippocampal lesions on structural and functional connectivity in six patients with amnesia, using a range of magnetic resonance imaging (MRI) analyses. Neuropsychological assessment revealed marked episodic memory impairment and generally intact performance across other cognitive domains. The hippocampus was the only brain structure exhibiting reduced grey-matter volume that was consistent across patients, and the fornix was the only major white-matter tract to show altered structural connectivity according to both diffusion metrics. Nonetheless, functional MRI revealed both increases and decreases in functional connectivity. Analysis at the level of regions within the default-mode network revealed reduced functional connectivity, including between nonhippocampal regions (connectional diaschisis). Analysis at the level of functional networks revealed reduced connectivity between thalamic and precuneus networks, but increased connectivity between the default-mode network and frontal executive network. The overall functional connectome showed evidence of increased functional segregation in patients (connectomic diaschisis). Together, these results point to dynamic reorganization following hippocampal lesions, with both decreased and increased functional connectivity involving limbic-diencephalic structures and larger-scale networks. © 2016 The Authors Hippocampus Published by Wiley Periodicals, Inc.