Longitudinal Evidence for Attenuated Local-Global Deviance Detection as a Precursor of Working Memory Decline.

From the perspective of predictive coding, normal aging is accompanied by decreased weighting of sensory inputs and increased reliance on predictions, resulting in the attenuation of prediction errors in older age. Recent electroencephalography (EEG) research further revealed that the age-related shift from sensorium to predictions is hierarchy-selective, as older brains show little reduction in lower-level but significant suppression in higher-level prediction errors. Moreover, the disrupted propagation of prediction errors from the lower-level to the higher-level seems to be linked to deficient maintenance of information in working memory. However, it is unclear whether the hierarchical predictive processing continues to decline with advancing age as working memory. Here, we longitudinally followed a sample of 78 participants from three age groups (including seniors, adults, and adolescents) over three years' time. Seniors exhibited largely preserved local processing [consisting of comparable mismatch negativity (MMN), delayed P3a, and comparable reorienting negativity (RON)] but significantly compromised global processing (consisting of suppressed frontocentral negativity and suppressed P3b) in the auditory local-global paradigm. These electrophysiological responses did not change with the passing of time, unlike working memory which deteriorated with advancing age. Correlation analysis further showed that these electrophysiological responses signaling prediction errors are indicative of concurrent working memory. Moreover, there was a correlation between earlier predictive processing and later working memory but not between earlier working memory and later predictive processing. The temporal asymmetry suggested that the hierarchy-selective attenuation of prediction errors is likely a precursor of working memory decline.

The mismatch negativity to abstract relationship of tone pairs is independent of attention.

The mismatch negativity (MMN) implicating a comparison process between the deviant and the memory trace of the standard can be elicited by not only changes in physical features but also violations of abstract patterns. It is considered pre-attentive, yet the use of the passive design makes it difficult to exclude the possibility of attention leak. In contrast to how this issue has been well addressed with the MMN to physical changes, much less research directly investigated the attentional effect on the MMN to abstract relationships. Here we conducted an electroencephalography (EEG) experiment to study whether and how the MMN to abstract relationships is modulated by attention. We adapted the oddball paradigm of Kujala et al. by presenting occasional descending tone pairs among frequent ascending tone pairs, while additionally implementing a novel control of attention. Participants' attention was either directed away from the sounds (with an engaging task of visual target detection, so that the sounds were task-irrelevant) or toward the sounds (with a conventional task of auditory deviant detection, so that the sounds were task-relevant). The MMN to abstract relationships appeared regardless of attention, confirming the pre-attentive assumption. The attention-independence of the frontocentral and supratemporal components of the MMN supported the notion that attention is not required to generate the MMN. At the individual level, a relatively equal number of participants showed attention enhancement and attention suppression. It is unlike the attentional modulation on the P3b, which was robustly elicited in the attended condition only. The concurrent collection of these two neurophysiological markers in both unattended and attended conditions might be potentially suitable for testing clinical populations showing heterogeneous deficits in auditory function independent/dependent of attention.

Load-dependent alpha suppression is related to working memory capacity for numbers.

Alpha suppression is proposed to reflect a surge in cortical excitability to enhance stimulus processing in working memory. The attenuated state of alpha might reflect the prioritisation of behaviourally relevant information, making it a proxy for working memory functioning. Despite the growing interest in utilising the advancement of brain-based measures to evaluate individuals' cognitive processes, there was a lack of consistent evidence on the relationship between alpha suppression and working memory performance. To investigate whether interindividual differences in alpha suppression might be related to variability in working memory capacity, we recorded participants' electroencephalography (EEG) while they performed an arithmetic task of either low or high working memory load. Participants were required to calculate either the product of digits (i.e., low-load condition) or the difference between the product of digits (i.e., high-load condition). We found alpha suppression at parietal regions, which became more prominent as working memory load increased. The pattern was present in approximately 80% of the participants. Importantly, the more the alpha suppressed as working memory load increased, the larger the drops in behavioural performance and the lower the Digit Span score. That is, alpha suppression was more prominent in participants of poor working memory capacity. Our findings suggest that alpha activity, subject to interindividual differences in sensitivity, could serve as a brain-based measure of an individual's working memory functioning.

Intention-based and sensory-based predictions.

We inhabit a continuously changing world, where the ability to anticipate future states of the environment is critical for adaptation. Anticipation can be achieved by learning about the causal or temporal relationship between sensory events, as well as by learning to act on the environment to produce an intended effect. Together, sensory-based and intention-based predictions provide the flexibility needed to successfully adapt. Yet it is currently unknown whether the two sources of information are processed independently to form separate predictions, or are combined into a common prediction. To investigate this, we ran an experiment in which the final tone of two possible four-tone sequences could be predicted from the preceding tones in the sequence and/or from the participants' intention to trigger that final tone. This tone could be congruent with both sensory-based and intention-based predictions, incongruent with both, or congruent with one while incongruent with the other. Trials where predictions were incongruent with each other yielded similar prediction error responses irrespectively of the violated prediction, indicating that both predictions were formulated and coexisted simultaneously. The violation of intention-based predictions yielded late additional error responses, suggesting that those violations underwent further differential processing which the violations of sensory-based predictions did not receive.

Visual Predictions Operate on Different Timescales.

Humans live in a volatile environment, subject to changes occurring at different timescales. The ability to adjust internal predictions accordingly is critical for perception and action. We studied this ability with two EEG experiments in which participants were presented with sequences of four Gabor patches, simulating a rotation, and instructed to respond to the last stimulus (target) to indicate whether or not it continued the direction of the first three stimuli. Each experiment included a short-term learning phase in which the probabilities of these two options were very different (p = .2 vs. p = .8, Rules A and B, respectively), followed by a neutral test phase in which both probabilities were equal. In addition, in one of the experiments, prior to the short-term phase, participants performed a much longer long-term learning phase where the relative probabilities of the rules predicting targets were opposite to those of the short-term phase. Analyses of the RTs and P3 amplitudes showed that, in the neutral test phase, participants initially predicted targets according to the probabilities learned in the short-term phase. However, whereas participants not pre-exposed to the long-term learning phase gradually adjusted their predictions to the neutral probabilities, for those who performed the long-term phase, the short-term associations were spontaneously replaced by those learned in that phase. This indicates that the long-term associations remained intact whereas the short-term associations were learned, transiently used, and abandoned when the context changed. The spontaneous recovery suggests independent storage and control of long-term and short-term associations.

Attention modulates repetition effects in a context of low periodicity.

Stimulus repetition can result in a reduction in neural responses (i.e., repetition suppression) in neuroimaging studies. Predictive coding models of perception postulate that this phenomenon largely reflects the top-down attenuation of prediction errors. Electroencephalography research further demonstrated that repetition effects consist of sequentially ordered attention-independent and attention-dependent components in a context of high periodicity. However, the statistical structure of our auditory environment is richer than that of a fixed pattern. It remains unclear if the attentional modulation of repetition effects can be generalised to a setting which better represents the nature of our auditory environment. Here we used electroencephalography to investigate whether the attention-independent and attention-dependent components of repetition effects previously described in the auditory modality remain in a context of low periodicity where temporary disruption might be absent/present. Participants were presented with repetition trains of various lengths, with/without temporary disruptions. We found attention-independent and attention-dependent repetition effects on, respectively, the P2 and P3a event-related potential components. This pattern of results is in line with previous research, confirming that the attenuation of prediction errors upon stimulus repetition is first registered regardless of attentional state before further attenuation of attended but not unattended prediction errors takes place. However, unlike previous reports, these effects manifested on later components. This divergence from previous studies is discussed in terms of the possible contribution of contextual factors.

Both contextual regularity and selective attention affect the reduction of precision-weighted prediction errors but in distinct manners.

Predictive coding model of perception postulates that the primary objective of the brain is to infer the causes of sensory inputs by reducing prediction errors (i.e., the discrepancy between expected and actual information). Moreover, prediction errors are weighted by their precision (i.e., inverse variance), which quantifies the degree of certainty about the variables. There is accumulating evidence that the reduction of precision-weighted prediction errors can be affected by contextual regularity (as an external factor) and selective attention (as an internal factor). However, it is unclear whether the two factors function together or separately. Here we used electroencephalography (EEG) to examine the putative interaction of contextual regularity and selective attention on this reduction process. Participants were presented with pairs of regular and irregular quartets in attended and unattended conditions. We found that contextual regularity and selective attention independently modulated the N1/MMN where the repetition effect was absent. On the P2, the two factors respectively interacted with the repetition effect without interacting with each other. The results showed that contextual regularity and selective attention likely affect the reduction of precision-weighted prediction errors in distinct manners. While contextual regularity finetunes our efficiency at reducing precision-weighted prediction errors, selective attention seems to modulate the reduction process following the Matthew effect of accumulated advantage.

Human Brain Ages With Hierarchy-Selective Attenuation of Prediction Errors.

From the perspective of predictive coding, our brain embodies a hierarchical generative model to realize perception, which proactively predicts the statistical structure of sensory inputs. How are these predictive processes modified as we age? Recent research suggested that aging leads to decreased weighting of sensory inputs and increased reliance on predictions. Here we investigated whether this age-related shift from sensorium to predictions occurs at all levels of hierarchical message passing. We recorded the electroencephalography responses with an auditory local-global paradigm in a cohort of 108 healthy participants from 3 groups: seniors, adults, and adolescents. The detection of local deviancy seems largely preserved in older individuals at earlier latency (including the mismatch negativity followed by the P3a but not the reorienting negativity). In contrast, the detection of global deviancy is clearly compromised in older individuals, as they showed worse task performance and attenuated P3b. Our findings demonstrate that older brains show little decline in sensory (i.e., first-order) prediction errors but significant diminution in contextual (i.e., second-order) prediction errors. Age-related deficient maintenance of auditory information in working memory might affect whether and how lower-level prediction errors propagate to the higher level.

Context-dependent minimisation of prediction errors involves temporal-frontal activation.

According to the predictive coding model of perception, the brain constantly generates predictions of the upcoming sensory inputs. Perception is realised through a hierarchical generative model which aims at minimising the discrepancy between predictions and the incoming sensory inputs (i.e., prediction errors). Notably, prediction errors are weighted depending on precision of prior information. However, it remains unclear whether and how the brain monitors prior precision when minimising prediction errors in different contexts. The current study used magnetoencephalography (MEG) to address this question. We presented participants with repetition of two non-predicted probes embedded in context of high and low precision, namely mispredicted and unpredicted probes. Non-parametric permutation statistics showed that the minimisation of precision-weighted prediction errors started to dissociate on early components of the auditory responses (including the P1m and N1m), indicating that the brain can differentiate between these scenarios at an early stage of the auditory processing stream. Permutation statistics conducted on the depth-weighted statistical parametric maps (dSPM) source solutions of the repetition difference waves between the two non-predicted probes further revealed a cluster extending from the frontal areas to the posterior temporal areas in the left hemisphere. Overall, the results suggested that context precision not only changes the weighting of prediction errors but also modulates the dynamics of how prediction errors are minimised upon the learning of statistical regularities (achieved by stimulus repetition), which likely involves differential activation at temporal-frontal regions.

Dynamics of brain activation during learning of syllable-symbol paired associations.

Initial stages of reading acquisition require the learning of letter and speech sound combinations. While the long-term effects of audio-visual learning are rather well studied, relatively little is known about the short-term learning effects at the brain level. Here we examined the cortical dynamics of short-term learning using magnetoencephalography (MEG) and electroencephalography (EEG) in two experiments that respectively addressed active and passive learning of the association between shown symbols and heard syllables. In experiment 1, learning was based on feedback provided after each trial. The learning of the audio-visual associations was contrasted with items for which the feedback was meaningless. In experiment 2, learning was based on statistical learning through passive exposure to audio-visual stimuli that were consistently presented with each other and contrasted with audio-visual stimuli that were randomly paired with each other. After 5-10 min of training and exposure, learning-related changes emerged in neural activation around 200 and 350 ms in the two experiments. The MEG results showed activity changes at 350 ms in caudal middle frontal cortex and posterior superior temporal sulcus, and at 500 ms in temporo-occipital cortex. Changes in brain activity coincided with a decrease in reaction times and an increase in accuracy scores. Changes in EEG activity were observed starting at the auditory P2 response followed by later changes after 300 ms. The results show that the short-term learning effects emerge rapidly (manifesting in later stages of audio-visual integration processes) and that these effects are modulated by selective attention processes.

Prior Precision Modulates the Minimization of Auditory Prediction Error.

The predictive coding model of perception proposes that successful representation of the perceptual world depends upon canceling out the discrepancy between prediction and sensory input (i.e., prediction error). Recent studies further suggest a distinction to be made between prediction error triggered by non-predicted stimuli of different prior precision (i.e., inverse variance). However, it is not fully understood how prediction error with different precision levels is minimized in the predictive process. Here, we conducted a magnetoencephalography (MEG) experiment which orthogonally manipulated prime-probe relation (for contextual precision) and stimulus repetition (for perceptual learning which decreases prediction error). We presented participants with cycles of tone quartets which consisted of three prime tones and one probe tone of randomly selected frequencies. Within each cycle, the three prime tones remained identical while the probe tones changed once at some point (e.g., from repetition of 123X to repetition of 123Y). Therefore, the repetition of probe tones can reveal the development of perceptual inferences in low and high precision contexts depending on their position within the cycle. We found that the two conditions resemble each other in terms of N1m modulation (as both were associated with N1m suppression) but differ in terms of N2m modulation. While repeated probe tones in low precision context did not exhibit any modulatory effect, repeated probe tones in high precision context elicited a suppression and rebound of the N2m source power. The differentiation suggested that the minimization of prediction error in low and high precision contexts likely involves distinct mechanisms.

Human susceptibility to social influence and its neural correlates are related to perceived vulnerability to extrinsic morbidity risks.

Humans considerably vary in the degree to which they rely on their peers to make decisions. Why? Theoretical models predict that environmental risks shift the cost-benefit trade-off associated with the exploitation of others' behaviours (public information), yet this idea has received little empirical support. Using computational analyses of behaviour and multivariate decoding of electroencephalographic activity, we test the hypothesis that perceived vulnerability to extrinsic morbidity risks impacts susceptibility to social influence, and investigate whether and how this covariation is reflected in the brain. Data collected from 261 participants tested online revealed that perceived vulnerability to extrinsic morbidity risks is positively associated with susceptibility to follow peers' opinion in the context of a standard face evaluation task. We found similar results on 17 participants tested in the laboratory, and showed that the sensitivity of EEG signals to public information correlates with the participants' degree of vulnerability. We further demonstrated that the combination of perceived vulnerability to extrinsic morbidity with decoding sensitivities better predicted social influence scores than each variable taken in isolation. These findings suggest that susceptibility to social influence is partly calibrated by perceived environmental risks, possibly via a tuning of neural mechanisms involved in the processing of public information.

The processing of mispredicted and unpredicted sensory inputs interact differently with attention.

Prediction and attention are fundamental brain functions in the service of perception. Interestingly, previous investigations found prediction effects independent of attention in some cases but attention-dependent in other cases. The discrepancy might be related to whether the prediction effect was revealed by comparing mispredicted event (where there is incorrect prediction) or unpredicted event (where there is no precise prediction) against predicted event, which are associated with different precision-weighted prediction error. Here we conducted a joint analysis on four published electroencephalography (EEG) datasets which allow for proper dissociation of mispredicted and unpredicted conditions when there was orthogonal manipulation of prediction and attention. We found that the mispredicted-versus-predicted contrast revealed an attention-independent effect of prediction suppression, whereas the unpredicted-versus-predicted contrast revealed a prediction effect that was reversed by attention on auditory N1. The results suggest that mispredicted and unpredicted processing interact with attention in distinct manners.

Corrigendum: Both attention and prediction are necessary for adaptive neuronal tuning in sensory processing.

[This corrects the article on p. 152 in vol. 8, PMID: 24723871.].

Temporal expectancies driven by self- and externally generated rhythms.

The dynamic attending theory proposes that rhythms entrain periodic fluctuations of attention which modulate the gain of sensory input. However, temporal expectancies can also be driven by the mere passage of time (foreperiod effect). It is currently unknown how these two types of temporal expectancy relate to each other, i.e. whether they work in parallel and have distinguishable neural signatures. The current research addresses this issue. Participants either tapped a 1Hz rhythm (active task) or were passively presented with the same rhythm using tactile stimulators (passive task). Based on this rhythm an auditory target was then presented early, in synchrony, or late. Behavioural results were in line with the dynamic attending theory as RTs were faster for in- compared to out-of-synchrony targets. Electrophysiological results suggested self-generated and externally induced rhythms to entrain neural oscillations in the delta frequency band. Auditory ERPs showed evidence of two distinct temporal expectancy processes. Both tasks demonstrated a pattern which followed a linear foreperiod effect. In the active task, however, we also observed an ERP effect consistent with the dynamic attending theory. This study shows that temporal expectancies generated by a rhythm and expectancy generated by the mere passage of time can work in parallel and sheds light on how these mechanisms are implemented in the brain.

Category-specific features and valence in action-effect prediction: An EEG study.

Despite extensive research on action-effect anticipation, little attention has been paid to the anticipation of different attributes of an event. An action-effect is not only a sensory event; it is often also an event of emotional value which can be pleasant or aversive. This latter attribute of action-effect prediction is similar to anticipation of reward versus punishment. To date the neural systems controlling sensory and reward anticipation have not been systematically compared. To this end, we designed an experiment to manipulate the sensory content and the emotional valence of the stimuli in an orthogonal fashion. We recorded and compared event-related potentials (ERPs) to the presentation of stimuli instantiating expected or unexpected features. Our results suggest (1) that both features are processed altogether and (2) that the prediction error resulting from the manipulation is reflected in an enhanced N400 component.

The impact of subliminal effect images in voluntary vs. stimulus-driven actions.

According to the ideomotor theory, actions are represented in terms of their sensory effects. In the current study we tested whether subliminal effect images influence action control (1) at early and/or late preparatory stages of (2) voluntary actions or stimulus-driven actions (3) with or without Stimulus-Response (S-R) compatibility. In Experiment 1, participants were presented at random with 50% of S-R compatible stimulus-driven action trials and 50% of voluntary action trials. The actions' effects (i.e. up- or down-pointing arrows) were presented subliminally before each action (i.e. a keypress). In voluntary actions, participants selected more often the action congruent with the prime when it was presented at long intervals before the action. Moreover they responded faster in prime-congruent than in prime-incongruent trials when primes were presented at short intervals before the action. In Experiment 2, participants were only presented with stimulus-driven action trials, with or without S-R compatibility. In stimulus-driven action trials with S-R compatibility (e.g., left-pointing arrow signaling a left keypress), subliminal action-effects did not generate any significant effect on RTs or error rates. On the other hand, in stimulus-driven action trials without S-R compatibility (e.g., letter "H" signaling a left keypress), participants were significantly faster in prime-congruent trials when primes were presented at the shortest time interval before the action. These results suggest that subliminal effect images facilitate voluntary action preparation on an early and a late level. Stimulus-driven action preparation is influenced on a late level only, and only if there is no compatibility between the stimulus and the motor response, that is when the response is not automatically triggered by the common properties existing between the stimulus and the required action.

The auditory N1 suppression rebounds as prediction persists over time.

The predictive coding model of perception proposes that neuronal responses reflect prediction errors. Repeated as well as predicted stimuli trigger suppressed neuronal responses because they are associated with reduced prediction errors. However, many predictable events in our environment are not isolated but sequential, yet there is little empirical evidence documenting how suppressed neuronal responses reflecting reduced prediction errors change in the course of a predictable sequence of events. Here we conceived an auditory electroencephalography (EEG) experiment where prediction persists over series of four tones to allow for the delineation of the dynamics of the suppressed neuronal responses. It is possible that neuronal responses might decrease for the initial predictable stimuli and stay at the same level across the rest of the sequence, suggesting that they reflect the predictability of the stimuli in terms of mere probability. Alternatively, neuronal responses might decrease for the initial predictable stimuli and gradually recover across the rest of the sequence, suggesting that factors other than mere probability have to be considered in order to account for the way prediction is implemented in the brain. We found that initial presentation of the predictable stimuli was associated with suppression of the auditory N1. Further presentation of the predictable stimuli was associated with a rebound of the component's amplitude. Moreover, such pattern was independent of attention. The findings suggest that auditory N1 suppression reflecting reduced prediction errors is a transient phenomenon that can be modulated by multiple factors.

Distinctive Representation of Mispredicted and Unpredicted Prediction Errors in Human Electroencephalography.

The predictive coding model of perception proposes that neuronal responses are modulated by the amount of sensory input that the internal prediction cannot account for (i.e., prediction error). However, there is little consensus on what constitutes nonpredicted stimuli. Conceptually, whereas mispredicted stimuli may induce both prediction error generated by prediction that is not perceived and prediction error generated by sensory input that is not anticipated, unpredicted stimuli involves no top-down, only bottom-up, propagation of information in the system. Here, we examined the possibility that the processing of mispredicted and unpredicted stimuli are dissociable at the neurophysiological level using human electroencephalography. We presented participants with sets of five tones in which the frequency of the fifth tones was predicted, mispredicted, or unpredicted. Participants were required to press a key when they detected a softer fifth tone to maintain their attention. We found that mispredicted and unpredicted stimuli are associated with different amount of cortical activity, probably reflecting differences in prediction error. Moreover, relative to predicted stimuli, the mispredicted prediction error manifested as neuronal enhancement and the unpredicted prediction error manifested as neuronal attenuation on the N1 event-related potential component. These results highlight the importance of differentiating between the two nonpredicted stimuli in theoretical work on predictive coding.

Optimistic, pessimistic, realistic: Event-related potential evidence for how depressive symptoms influences expectation formation in the Human brain.

Recent research suggested a link between the prediction mechanism and depressive symptoms. While healthy people tend to maintain unrealistic optimism in the face of reality challenging their beliefs, depressed people show systematic pessimism. However, it remains unclear at which stage these individual differences in optimism/pessimism arise in the brain. In the current study we designed a simple gambling task with two difficulty levels, the easy game and the hard game. Participants were required to press one of four keys to gain a bonus signalled by a sinusoidal tone. For three of the four keys, the probability of getting a large bonus was 80% in the easy game and 8% in the hard game. In both games, the fourth key, randomly determined in each trial, yielded a large bonus with a probability of 100%. This arrangement allowed us to observe less/more depressed participants׳ optimistic/pessimistic expectations about hitting the key that guarantees a large bonus. The opposite expectation patterns of less/more depressed participants were reflected on the N1 amplitude. Meanwhile, all participants were well aware of the true probability of obtaining certain bonus in each game as reflected on the P3 amplitude. The results suggest that the subjective system (tracking subjective beliefs) and the objective system (tracking objective evidence) are dissociable in the human brain, with the former feeding information into sensory areas and the latter representing prediction errors on a higher level. Moreover, individual differences arise from variability in the former rather than the latter.