A predictive account of how novelty influences declarative memory

Abstract

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.

Keywords: Memory; Novelty; PIMMS; Prediction error; Predictive coding; Surprise.

Fig. 1

Schematic illustration of prior and likelihood distributions in the “predictive interactive multiple memory signals” framework (PIMMS). Distributions reflect activity in a layer of topographically organised neurons, where the x-axis captures similarity between items in the semantic level or similarity between features in the perceptual layer (depending on the panel). The dotted line represents the prior predictions from the “higher” level, whereas the solid line represents the likelihood distribution, input from the level below (ultimately the sensory input). The PE, which drives learning, is the divergence between these two probability distributions, whose magnitude is illustrated at the bottom right of each panel. Panel A: precise prior and precise likelihood for items in a certain context, but with different modes (context surprise). Panel B: the same as Panel A, except the precise prior and likelihood refer to features of an object in lower perceptual levels (item surprise). Panel C: Flat prior and precise likelihood (context novelty). Panel D: precise prior and flat likelihood (item novelty). Panel E: precise prior and precise likelihood with the same mean (leading to no PE or learning; no surprise or novelty). Panel F: flat prior and flat likelihood, a combination of context novelty and item novelty (or “complete novelty”), but one predicted to show no PE or learning.

Fig. 2

Illustration of key paradigms. Panel A: Design by Tulving and Kroll (1995): pre-familiarised and novel items are presented intermixed at critical study for which recognition memory was later tested. Panel B: Design of von Restorff/distinctiveness paradigm: items of the same type/class are presented in lists together with a conceptually or perceptually deviant item (e.g. different font type/colour), memory for which can compared to an item same position in a control list without deviants. Panel C: Rule based design by Greve et al. (2017): At study new scenes were paired with new words, which had the same valence as expected from a previous familiarisation phase (low PE) or the opposite valence (high PE). A forced-choice memory test matched target and foils to be of same valence and equally familiar. Panel D: Reward PE design by De Loof et al. (2018): one, two or four Swahili words are presented as options from which the rewarded word is selected, which manipulated the size of RPE. Panel E: Design by Reggev et al. (2018): Judging whether a noun-adjective is congruent and subsequently testing memory for the nouns. Panel F: Item novelty assessed by presenting objects vs. non-objects or words vs. non-words in Kroll and Potter (1984). Panel G: Mismatch design from Kumaran and Maguire (2006): sequences of objects were represented twice, wherein the second presentation, the order of objects was either unchanged (S_rep), changed after the first half (S_half) or completely new (S_new).