Unpredictability of the “when” influences prediction error processing of the “what” and “where”

Abstract

The capability to establish accurate predictions is an integral part of learning. Whether predictions about different dimensions of a stimulus interact with each other, and whether such an interaction affects learning, has remained elusive. We conducted a statistical learning study with EEG (electroencephalography), where a stream of consecutive sound triplets was presented with deviants that were either: (a) statistical, depending on the triplet ending probability, (b) physical, due to a change in sound location or (c) double deviants, i.e. a combination of the two. We manipulated the predictability of stimulus-onset by using random stimulus-onset asynchronies. Temporal unpredictability due to random onsets reduced the neurophysiological responses to statistical and location deviants, as indexed by the statistical mismatch negativity (sMMN) and the location MMN. Our results demonstrate that the predictability of one stimulus attribute influences the processing of prediction error signals of other stimulus attributes, and thus also learning of those attributes.

Beat processing in newborn infants cannot be explained by statistical learning based on transition probabilities

“…That is, while statistical learning is an essential function for extracting information from the world, the current results suggest that it is complemented by functions honed to extract the temporal structure of (at least) the acoustic environment. Their complementary nature is supported by findings showing that temporal predictability, which is enhanced by isochronous stimulus presentation, appears to improve statistical learning (Tsogli, Jentschke & Koelsch, 2022, Selchenkova, Jones & Tillmann, 2014. In fact, much of the research on statistical learning, especially in neonates, has been based on isochronous stimuli (Bosseler et al, 2016;Teinonen et al, 2009).…”

“…It is possible that isochrony or, more generally, temporal predictability is important for newborns to extract sequential probabilities necessary for statistical learning. Newborns have no problem extracting even higher order regularities in temporally highly predictable stimuli (e.g., Ruusuvirta et al, 2004; Stefanics et al, 2009), and although predictability in time is not necessary for statistical learning in adults, it does enhance the statistical MMN amplitude, an index of statistical learning (Tsogli, Jentschke & Koelsch, 2022). Further, in general, the noisier the regularity, the lower the MMN amplitude (e.g., Winkler et al, 1990).…”

Can Prediction Error Explain Predictability Effects on the N1 during Picture-Word Verification?

“…Indeed, previous research has highlighted the importance of presentation rate in prediction effects during reading (e.g., Dambacher et al, 2012), and recent findings have shown that unpredictability in stimulus presentation timing (e.g., with jittered inter-stimulus intervals) may interfere with predictive processes, as indexed by the mismatch negativity component (Tsogli et al, 2022). This explanation of our results could be tested by study designs examining how the congruency-predictability interaction varies over stimulus onset asynchronies of different durations.…”

Markov chains as a proxy for the predictive memory representations underlying mismatch negativity

“…The predictive memory or generative model contains the respective “predictive regularity representations” ( Schröger et al, 2014 ; Winkler and Schröger, 2015 ). Although this rather Markovian view on the MMN seems to be widely accepted in the community, explicit reference to Markov chains or transition probabilities is rare in MMN research ( Furl et al, 2011 ; Koelsch et al, 2016 ; Mittag et al, 2016 ; Tsogli et al, 2019 , 2022 ; Korka et al, 2021 ).…”