The cognitive resource and foreknowledge dependence of auditory perceptual inference

Abstract

Auditory perceptual inference engages learning of complex statistical information about the environment. Inferences assist us to simplify perception highlighting what can be predicted on the basis of prior learning (through the formation of internal "prediction" models) and what might be new, potentially necessitating an investment of resources to remodel predictions. In the present study, we tested the hypothesis that sound sequences with multiple levels of predictability may rely on cognitive resources and be cognitively penetrable to a greater extent than was previously shown by studies presenting simpler sound sequences. Auditory-evoked potentials (AEPs) were recorded from 117 participants. All participants heard the exact same sound sequence but under different conditions: 51 while watching a DVD movie and 66 while performing a cognitively demanding task. Participants were asked to ignore the sounds and focus their attention on the movie/task. However, prior to commencing the experiment we manipulated what participants knew about the sound sequence by providing explicit sequence information to 15 and 34 of the participants in the DVD and cognitive-task conditions, respectively, and no information to the others. The results demonstrated that although local pattern violations elicited distinctive AEP responses (namely, mismatch negativity), the way the amplitude of this response was modulated by sequence learning over time was dependent upon both task and explicit sequence knowledge. The implications are discussed with reference to how the division of available attention resources between the primary task and concurrent sound impacts what is learned.

Setting precedent: Initial feature variability affects the subsequent precision of regularly varying sound contexts

“…Thus, our previous notion of the nature of the first impression bias is expanded by the current results. Whereas the previous two-tone sequence studies Frost et al, 2018Frost et al, , 2016Mullens et al, 2016;Todd et al, 2011Todd et al, , 2013 showed an effect of the initial stimulus probability on processing the same sounds later in the sequence even after their probabilities have changed, the results obtained from the current three-tone sequences suggest that also the initial variability of the sound features can impact on how the same sounds are processed later in the sequence, even after their probabilities have varied. This means that primacy bias effects rely on a representation encoding more than just the initial probabilities; this representation also includes some estimate of featural variabilities.…”

“…Remarkably, this limitation disappears if the changeover between languages is cued by a period of silence and if participants receive explicit instruction on the existence of a second structure. This is precisely what happened in Frost et al () when participants heard the sequence depicted in Figure a and were told beforehand about the sequence structure. Although participants were asked to watch a movie and ignore the sounds just as in the version of the task with no explicit knowledge of the sequence, but in this case, the MMN elicited by deviants was large across the entire period of both block types, therefore, not showing the pattern of lower precision for the alternate block.…”

“…The comparison to the ART neural network raises the possibility that the removal of order effects on learning when participants have explicit knowledge about the likelihood (or rather certainty) that a future change will occur, could operate via a release from such attentional biases (implicit in the absence of a task in Frost et al’s () binary state multi‐timescale sequences, and explicit through a task in Gebhart et al () artificial language learning setup). In the present study, the three‐tone/three‐block sequences appear to anchor precision estimates to the featural variability present at sequence onset, which is arguably a distributional property.…”

“…Naïve listeners must structure models based on initial experience and task‐based difference across studies may indicate that the availability of attentional resources determines the period of time over which the “first impression” will be set. This period would be very short when attention is required in the visual modality for a demanding concurrent task, perhaps only allowing only a brief auditory sampling window capturing which sound is common and which rare with this first impression setting different precisions for future encounters of the two tones throughout (Frost et al, ). In comparison, when hearing the sequences under low concurrent task demand (watching a movie) the first impression might instead extend for as long as the duration of the first block of sound explaining why block differences in learning rates occur and become re‐set when block length is violated (Fitzgerald & Todd, ; Mullens et al, ; Todd, Heathcote, et al, 2014).…”

Hierarchical learning of statistical regularities over multiple timescales of sound sequence processing: A dynamic causal modelling study

“…In previous studies, differential modulations of MMN amplitude to the first and second deviant tone have persisted across as many as four repetitions of the same sound sequences (36), and are shown only to be altered when superordinate patterning is abolished or violated (35, 65) or if the participant has prior experience with the sounds (48) or prior knowledge of the sequence structure (66). The apparent failure to override a first-impression, even after four repetitions of the same sound sequence, implies that the system maintains differential precision weightings for the two deviants unless a substantive level of surprise is encountered (36).…”