Premembering Experience: A Hierarchy of Time-Scales for Proactive Attention

Abstract

Memories are about the past, but they serve the future. Memory research often emphasizes the former aspect: focusing on the functions that re-constitute (re-member) experience and elucidating the various types of memories and their interrelations, timescales, and neural bases. Here we highlight the prospective nature of memory in guiding selective attention, focusing on functions that use previous experience to anticipate the relevant events about to unfold-to "premember" experience. Memories of various types and timescales play a fundamental role in guiding perception and performance adaptively, proactively, and dynamically. Consonant with this perspective, memories are often recorded according to expected future demands. Using working memory as an example, we consider how mnemonic content is selected and represented for future use. This perspective moves away from the traditional representational account of memory toward a functional account in which forward-looking memory traces are informationally and computationally tuned for interacting with incoming sensory signals to guide adaptive behavior.

Keywords: attention; decision-making; episodic memory; hippocampus; implicit memory; memory; prefrontal cortex; priming; working memory.

Figure 1

Mutual Interactions between Memory and Attention Attention draws on past experience from multiple timescales to anticipate and prepare for incoming stimulation and guide adaptive action. Conversely, attention is not only forward looking but can select and bias information during encoding and maintenance in memory. These mutual interactions feed a virtuous cycle that tunes our minds to the most relevant features of the environment. In this review, we consider the multiple mnemonic timescales that are important for guiding proactive attention (dark arrows).

Figure 2

Multiple Timescales for Memory-Guided Attention (A) Even at very short time-scales, attention is influenced by preceding stimuli. For example, the classic exogenous cueing paradigm demonstrates how attention lingers at a previously cued location (adapted from Posner and Cohen, 1984). (B) At slightly longer timescales, WM guides visual search (quantified as contralateral delay activity [CDA]). However, as the timescale increases (over trials with repeating template), attentional control is transferred to intermediate memory (i.e., reduced CDA; scale bar represents relative voltage difference over the scalp surface; error bars represent ±1 SEM). Adapted from Carlisle et al. (2011). (C) At even longer timescales, LTM maintains relevant information for guiding attention. In this study, Stokes et al. (2012) found that spatial information stored in LTM can be used to modulate the visual cortex in preparation for a target stimulus (error bars represent ±1 SEM; scale bar represents the relative BOLD response in left and right visual cortex).

Figure 3

Memory Is Prospective, Representing the Information Most Likely to Be Relevant for Behavior (A) Selective encoding. Serences et al. (2009) used fMRI to show that WM maintains sensory information (color or orientation) that is most relevant to behavior. Decoding patterns of activity in early visual cortex, they found that activity in the memory delay carried orientation angle information when orientation was relevant for future decision-making or color hue information when color was relevant (adapted from Serences et al., 2009). (B) Selective maintenance. Wallis et al. (2015) used MEG to investigate selection of items already in WM. Retro-cues resulted in better memory performance (center panel) and contralateral suppression of alpha power in the visual cortex (right panel; adapted from Wallis et al., 2015). Scale bar represents spectral lateralization (contralateral minus ipsilateral) in normalized units (t-value). (C) Memories can also be distorted to guide behavior. During a visual search task in which distractor stimuli are clustered on one side of the parametric feature space (e.g., color in Yu and Geng, 2019), the search template held in WM becomes distorted away from the veridical target to better separate the target from the competing distractors (adapted from Yu and Geng, 2019). (D) Working memories also code for motor plans when motor affordances are available and can optimize performance. In van Ede et al. (2019), visual stimuli were paired with specific motor plans (left panel), resulting in concurrent visual and motor preparation and their characteristic electrophysiological signatures (right panel; adapted from van Ede et al., 2019). Scale bar represents spectral lateralization (contralateral minus ipsilateral) in percept signal change.

Figure 4

Preparing for Specific Items or Specific Tasks (A) Typically, the prospective nature of WM is studied by varying the type of information that will be probed at the end of the trial. For example, if participants are cued that circles are task-relevant, then they preferentially encode and maintain information corresponding to those items. (B) However, WM tasks can also differ in how the items will be used (e.g., match-to-sample versus reproduction task). From a functional perspective, the neural format can also be adaptive for the type of expected future task.

Figure 5

Setting the Initial Condition in WM for Attention Modulation through a Match Filter Model WM establishes a match filter in sensory areas that computes the perceptual similarity between incoming sensory signals and an internal template. A match filter template need not involve pre-activation of the target stimulus. Moreover, the match enhancement effect could serve a general salience cue for capturing attention.