Pruning of memories by context-based prediction error

Abstract

Significance Forgetting is often considered to be bad, but selective forgetting of unreliable information can have the positive side effect of reducing mental clutter, thereby making it easier to access our most important memories. Prior studies of forgetting have focused on passive mechanisms (decay, interference) or on effortful inhibition by cognitive control. Here we report the discovery of an active mechanism for forgetting that weakens memories selectively and without burdening the conscious mind. Specifically, we show that the brain automatically generates predictions about which items should appear in familiar contexts; if these items fail to appear, their memories are weakened. This process is adaptive, because such memories may have been encoded incorrectly or may represent unstable aspects of the world.

Statistical learning of language: Theory, validity, and predictions of a statistical learning account of language acquisition

“…Although this process is dependent on attention and working memory to select the to-be-bound elements (e.g., Baker, Olson, & Behrmann, 2004), the process of binding may also involve the hippocampus (a structure we have suggested is also involved in integration), given its role in memory formation. Neurological data are at least partially consistent with this account, as at least some extraction tasks have been shown to involve hippocampal activation, which is more consistent with long-term memory processes (Kim, Lewis-Peacock, Norman, & Turk-Browne, 2014; see also Schapiro, Gregory, Landau, McCloskey, & Turk-Browne, 2014 for evidence that medial temporal lobe damage including the hippocampus disrupts statistical segmentation, an extraction task; but see Knowlton, Ramus, & Squire, 1992 for evidence of intact Artificial Grammar Learning in amnesiac patients). Similarly, attentional processes (which we hypothesize to be related to extraction) have been shown to stabilize extracted representations in the hippocampus (Aly & Turk-Browne, 2015) …”

Attention promotes episodic encoding by stabilizing hippocampal representations

“…We wanted to compare such interregional correlations separately for subsequent hits and misses; however, because each participant provided relatively few hits (art: mean = 6.6 trials, SD = 2.7; room: mean = 5.0, SD = 2.4), within-participant correlational analyses would be underpowered. We thus pooled data across all participants and performed a supersubject analysis with random-effects bootstrapping (22).…”

“…To increase reliance on the kind of detailed episodic memory supported by the hippocampus, the memory test included a highly similar lure for each encoded item (20,21): a novel painting from the same artist or the same layout from a novel perspective, for the art and room blocks, respectively. We used memory performance on this task to sort the fMRI data from phase 2, which allowed us to relate attentional states during encoding to subsequent memory (22,23).…”

Insight Reconfigures Hippocampal-Prefrontal Memories

“…One possibility is that anterior medial temporal lobe triggers segregation of irrelevant events by signaling the mismatch between the integrated narrative representation and temporally proximal unrelated events [7,19,42]. This mismatch signal might support the reconfiguration of event representations in memory by inducing neural plasticity [20,43] and might relate to findings that the connectivity between anterior hippocampus and the mPFC increases for schema-incongruent representations [7]. Another possibility is that the mismatch signaling may be a response to a newly established narrative context or schema in the mPFC.…”

“…Nevertheless, the emergence of this new context may be a consequence of the co-occurrence (i.e., shared temporal context [30]) of individual event features from events A and B during the event L or due to retrieval of both events A and B during event L. After the insight phase, the non-linked events would become incongruent with this newly established narrative context and thus elicit a mismatch signal ( [7,19,20,39], cf. [43]), which would work against similarity that is simply based on temporal proximity of events.…”