Distinct neural representations of content and ordinal structure in auditory sequence memory

Abstract

Two forms of information - frequency (content) and ordinal position (structure) - have to be stored when retaining a sequence of auditory tones in working memory (WM). However, the neural representations and coding characteristics of content and structure, particularly during WM maintenance, remain elusive. Here, in two electroencephalography (EEG) studies, by transiently perturbing the 'activity-silent' WM retention state and decoding the reactivated WM information, we demonstrate that content and structure are stored in a dissociative manner with distinct characteristics throughout WM process. First, each tone in the sequence is associated with two codes in parallel, characterizing its frequency and ordinal position, respectively. Second, during retention, a structural retrocue successfully reactivates structure but not content, whereas a following white noise triggers content but not structure. Third, structure representation remains stable whereas content code undergoes a dynamic transformation through memory progress. Finally, the noise-triggered content reactivations during retention correlate with subsequent WM behavior. Overall, our results support distinct content and structure representations in auditory WM and provide a novel approach to access the silently stored WM information in the human brain. The dissociation of content and structure could facilitate efficient memory formation via generalizing stable structure to new auditory contents.

Reactivating ordinal position information from auditory sequence memory in human brains

“…Alternatively, it might follow a dynamic trajectory in neural space but repeat the same trajectory when reactivated ("stable-dynamic"). The latter alternative would also predict encoding-to-maintaining representational generalization for structure, as demonstrated in previous studies (Fan et al, 2021;Kalm et al, 2017). Unraveling the coding dynamics is also critical for our understanding of ordinal position representation in WM.…”

“…A time-resolved RSA-based decoding analysis (Fan et al, 2021; Kriegeskorte et al, 2008) was conducted on EEG activities to evaluate the neural representation of tone frequency and ordinal position, respectively, throughout the encoding and maintain phases (see details in Materials and Methods). The rationale of the RSA analysis is that the neural representational dissimilarity is proportional to the factor-of-interest dissimilarity, based on which the regression between the design dissimilarity matrix and the neural dissimilarity matrix was calculated to represent the decoding strength (beta).…”

“…We finally asked whether the neural code for ordinal position is the same or different between encoding and retention stages, by employing a cross-time encoding-tomaintaining generalization analysis (King & Dehaene, 2014). Although previous studies convergingly support stable representations for structure (Al Roumi et al, 2021;Fan et al, 2021;Kalm et al, 2017), the stable codes might arise from different possibilities, as illustrated in Figure 4A. For example, it might be specific temporal segment, in short or long duration, during encoding (y-axis, vertical span from A to D) that carries the structure code, which would be reactivated during retention in a sustained way (x-axis, horizontal span from A to B), referred to as "Stable-sustained representation" (Figure 4A, left).…”

“…This issue poses a tremendous challenge to human studies using non-invasive recordings since the to-be-memorized information couldn’t be easily or reliably accessed during retention (Lundqvist et al, 2018; Sreenivasan et al, 2014), somewhat commensurate with the ‘activity-silent’ WM view (Stokes, 2015). Recently, an impulse-response approach (Fan et al, 2021; Wolff et al, 2017) has been developed to successfully reactivate WM information during the delay period, thereby the underlying neural mechanism could be studied. For instance, our previous work (Fan et al, 2021) showed that when participants retained a list of pure tones, an auditory white noise that was neither memory-related nor task-relevant could trigger neural representations of memorized tone pitch, which further predicts WM behavior.…”

“…Recently, an impulse-response approach (Fan et al, 2021; Wolff et al, 2017) has been developed to successfully reactivate WM information during the delay period, thereby the underlying neural mechanism could be studied. For instance, our previous work (Fan et al, 2021) showed that when participants retained a list of pure tones, an auditory white noise that was neither memory-related nor task-relevant could trigger neural representations of memorized tone pitch, which further predicts WM behavior. The white-noise impulse is posited to act as a triggering event to transiently perturb the WM system so that the retained information could be reactivated (Manohar et al, 2019).…”