Constructing the hierarchy of predictive auditory sequences in the marmoset brain

Abstract

Our brains constantly generate predictions of sensory input that are compared with actual inputs, propagate the prediction-errors through a hierarchy of brain regions, and subsequently update the internal predictions of the world. However, the essential feature of predictive coding, the notion of hierarchical depth and its neural mechanisms, remains largely unexplored. Here, we investigated the hierarchical depth of predictive auditory processing by combining functional magnetic resonance imaging (fMRI) and high-density whole-brain electrocorticography (ECoG) in marmoset monkeys during an auditory local-global paradigm in which the temporal regularities of the stimuli were designed at two hierarchical levels. The prediction-errors and prediction updates were examined as neural responses to auditory mismatches and omissions. Using fMRI, we identified a hierarchical gradient along the auditory pathway: midbrain and sensory regions represented local, short-time-scale predictive processing followed by associative auditory regions, whereas anterior temporal and prefrontal areas represented global, long-time-scale sequence processing. The complementary ECoG recordings confirmed the activations at cortical surface areas and further differentiated the signals of prediction-error and update, which were transmitted via putatively bottom-up γ and top-down β oscillations, respectively. Furthermore, omission responses caused by absence of input, reflecting solely the two levels of prediction signals that are unique to the hierarchical predictive coding framework, demonstrated the hierarchical predictions in the auditory, temporal, and prefrontal areas. Thus, our findings support the hierarchical predictive coding framework, and outline how neural circuits and spatiotemporal dynamics are used to represent and arrange a hierarchical structure of auditory sequences in the marmoset brain.

A vocalization-processing network in marmosets

“…Consistent with the well-known auditory pathway in marmosets (34) and a recent fMRI study in marmosets (35), we found auditory responses in the inferior colliculus and medial geniculate nucleus. Interestingly, the inferior colliculus, medial geniculate nucleus, pulvinar, amygdala, caudate and putamen also showed larger activations for vocalizations.…”

The neural dynamics of feedforward and feedback interactions in predictive processing

“…Consistent with this interpretation, several studies have shown transient and spectrally broad-band (i.e. extending well above 100Hz without a narrow-band peak) increases in gamma LFP power for unpredictable as compared to predictable stimuli (Chao et al, 2018;Todorovic et al, 2011;Jiang et al, 2022;Bastos et al, 2020) (see Figure 4A). A recent study has directly dissociated narrow from broad-band spectral energy in marmoset A1 (Figure 2A).…”

“…The idea that alpha and beta rhythms are involved in FB communication of predictions is primarily based on Grangercausality analyses of field potential data (see next section), rather than their modulation by behavioral or stimulus conditions. Several (but not all, Canales-Johnson et al (2021a); Todorovic et al (2011); Arnal et al (2011)) studies have observed a suppression of alpha/beta-band power for unpredicted stimuli (Bastos et al, 2020;Chao et al, 2018;Jiang et al, 2022;Von Stein et al, 2000) (Figure 4), although these studies did not explicitly detect rhythms and distinguish these from aperi-odic processes (BOX 2). This finding has been interpreted as evidence for a role in FB signalling.…”

Distributed representations of prediction error signals across the cortical hierarchy are synergistic

“…Cortical mechanisms of auditory prediction error (PE) have been extensively studied using evoked-related potentials (ERP) and spectral analyses (Canales-Johnson et al, 2021; Blenkmann et al, 2019; Chao et al, 2018; Jiang et al, 2022; Parras et al, 2017). A well-studied ERP marker of auditory PE is the mismatch negativity (MMN), an event-related potential (ERP) that peaks around 150–250 ms after the onset of an infrequent acoustic stimulus (Parras et al, 2017).…”