Andreas Widmann

Theories of word production rely heavily on semantic context effects. For example, when pictures are named repeatedly and sorted by semantic category, it takes longer to name them than when they are named in an intermixed manner. Wöhner et al. (2021) found that semantic interference in blocked-cyclic naming was approximately three times greater for sounds (e.g., barking) than for pictures (e.g., a dog). This study explored possible reasons for this differential interference pattern. When the stimuli were presented briefly, the differential interference was replicated, ruling out the possibility that it results from a large overlap between stimulus processing and phonological response preparation. When the sounds and pictures were replaced with spoken and written words that the participants translated into their second language, the size of the interference was similar, showing that the differential pattern does not result from different stimulus modalities (auditory vs. visual). When participants typed their responses, the differential interference was replicated again, showing that it does not depend on a certain response modality (spoken vs. written). Additionally, pupil dilation data suggest that sounds, compared with pictures, may activate additional representations that contribute to the interference. Overall, the data imply that semantic interference in word production is shaped by early, stimulus-specific, perceptual and semantic “lead-in” processes that precede lexical selection. More attention should be given to these early processes in word production research.

The unpredicted omission of expected sound, that is “silence”, elicits omission N1, N2 and P3 (early and late oN1, oN2, oP3) event-related potential (ERP) and pupil dilation responses (PDR), which are interpreted as prediction error signals located along different processing levels. This study aims to investigate whether these responses to unpredicted silence can also be elicited by the unpredicted continuation of sound, that is, the “absence” of expected silence. A prediction error response to the absence of silence would indicate that silence was predicted including the levels of processing on which a prediction of silence was established. Participants (27 women, 10 men) pressed a button every 1-2s while exposed to continuous Brownian noise. Button presses predictably (88% of trials), randomly (50% of trials) or never (0% of trials; motor-control) interrupted the noise. EEG and pupil diameter data were recorded. We found that the unpredicted absence of a silent interval in the noise elicited late oN1 (~150ms) and oN2, oP3 and PDR omission prediction error responses. However, we did not find an early oN1 (~100ms). Predictive representations of silence can be established at the level of categorical sensory processing (late oN1) and at the levels of later, higher cognitive processing and attention (oN2, oP3, and PDR), but not at the level of feature-based, early cortical sensory processing (early oN1), possibly because the absence of features cannot easily be encoded. These results suggest that silence is represented not only as the absence of sound but also as a predictable event in itself.

The human brain anticipates the sensory consequences of an action and generates a prediction error (PE) signal when the intended action effect does not occur. This study investigated auditory event‐related potential PE markers based on whether participants intended to perform a specific action or produce a specific action effect. Participants were instructed to press a left or right button to produce low‐ or high‐pitched tones, following a visual pattern. The instructions, actions and tone sequences were identical for all participants. The visual patterns differed in two groups: In the action‐effect intention group, the visual pattern consisted of ‘notes’ (indicating low/high pitch). In the action intention group, the visual pattern consisted of ‘letters’ (indicating left/right button‐press). In both groups, a button‐press occasionally failed to produce the associated tone (incongruent sounds). The key finding was that these incongruent sounds elicited an enhanced auditory N1 component compared to congruent sounds only in the ‘notes’ group. We propose that participants in the ‘notes’ group selected their actions based on the intended action effect, which induced a predictive sensory representation of the expected tone. A violation of this prediction resulted in the early PE, reflected in the auditory N1. Later PE responses, specifically the N2b and P3 components, were observed in both groups. This suggests that action‐effect associations were represented, and their violation was processed at a conceptual level even in the ‘letters’ group. These results support theories postulating that event representations integrate features of stimuli, actions and their associated outcomes.

Predictive coding conceptualizes attention as a weighting of prediction error signals. However, empirical findings on how attention influences common markers of prediction error have been inconsistent, likely because these markers are typically derived from stimulus‐evoked responses. To avoid stimulus‐related confounds and isolate effects related purely to prediction, we investigated how attention modulates brain responses to unexpected stimulus omissions. Using visual–auditory couplings where the auditory stimulus was occasionally omitted, we recorded EEG responses that revealed a multistage omission response—from early sensory to later higher‐level prediction error activity. Voluntary attention was manipulated along two dimensions: (1) toward the visual or auditory modality, and (2) toward the moment of stimulus presentation or sustained over time. Early sensory prediction error, reflected by the omission N1, was unaffected by any manipulation of attention. In contrast, later high‐level prediction error processing, reflected by omission P3 responses, was strongly affected by directing attention: robust responses were elicited when attention was directed to the auditory modality—where the prediction had been violated—but these were markedly reduced or absent when attention was directed to the visual modality. These results suggest an attentional system that does not affect low‐level sensory prediction error but is capable of influencing distinct stages in the processing hierarchy in service of task performance. This first investigation of how attention affects different stages of omission activity suggests that voluntary attention may modulate prediction error processing via specific neurotransmitter systems and demonstrates this approach's potential for reliably studying precision‐weighting in the brain.

To interact efficiently with our environment, our brain predicts the sensory effects of our actions and compares them with the actual outcomes. This allows us to adapt our actions when predictions and sensory outcomes mismatch. While this process is generally well understood for action‐sound predictions, it is an open question whether these predictions can flexibly switch in frequently changing environments, as they occur in real life. To investigate the flexibility of top‐down predictions, we asked participants ( N = 41) to press one of two buttons, a left‐hand and a right‐hand button, and switch hands autonomously. One button frequently produced a sound (80%) and rarely no sound. The other button frequently generated no sound (80%) and rarely produced a sound. In a third, separate condition, each button produced a sound in 50% of the trials. Unexpected sounds and unexpected sound omissions elicited a series of error‐related brain responses in the electroencephalogram (EEG) at different levels of auditory processing, including a mismatch negativity (MMN) and the P3 complex for unexpected sounds, and the oN1, oN2, and oP3 complex for unexpected omissions. Moreover, unexpected sounds elicited an equivalent MMN, regardless of whether silence was expected (80%) or no reliable expectation was possible (50%), while later P3 components showed different amplitudes. Our results demonstrate flexible action‐sound predictions at sensory and higher cortical levels. Furthermore, they indicate that predicted silence does not have an explicit sensory representation at lower levels but emerges at later stages, when higher‐level information has been integrated.

Predictive coding conceptualizes attention as a weighting of prediction error signals. However, empirical findings on how attention influences common markers of prediction error have been inconsistent, likely because these markers are typically derived from stimulus-evoked responses. To avoid stimulus-related confounds and isolate effects related purely to prediction, we investigated how attention modulates brain responses to unexpected stimulus omissions. Using visual-auditory couplings where the auditory stimulus was occasionally omitted, we recorded EEG responses that revealed a multi-stage omission response – from early sensory to later higher-level prediction error activity. Voluntary attention was manipulated along two dimensions: (1) toward the visual or auditory modality, and (2) toward the moment of stimulus presentation or sustained over time. Early sensory prediction error, reflected by the omission N1, was unaffected by any manipulation of attention. In contrast, later high-level prediction error processing, reflected by omission P3 responses, was strongly affected by directing attention: robust responses were elicited when attention was directed to the auditory modality – where the prediction had been violated – but these were markedly reduced or absent when attention was directed to the visual modality. These results suggest an attentional system that does not affect low-level sensory prediction error, but is capable of influencing distinct stages in the processing hierarchy in service of task performance. This first investigation of how attention affects different stages of omission activity suggests that voluntary attention may modulate prediction error processing via specific neurotransmitter systems and demonstrates this approach’s potential for reliably studying precision-weighting in the brain.

To interact efficiently with our environment, our brain predicts the sensory effects of our actions and compares them with the actual outcomes. This allows us to adapt our actions when predictions and sensory outcomes mismatch. While this process is generally well understood for action-sound predictions, it is an open question whether these predictions can flexibly switch in frequently changing environments, as they occur in real life. To investigate the flexibility of top-down predictions, we asked participants (N = 41) to press one of two buttons, a left-hand and a right-hand button, and switch hands autonomously. One button frequently produced a sound (80%) and rarely no sound. The other button frequently generated no sound (80%) and rarely produced a sound. In a third, separate condition, each button produced a sound in 50% of the trials. Unexpected sounds and unexpected sound omissions elicited a series of error-related brain responses in the electroencephalogram (EEG) at different levels of auditory processing, including a mismatch negativity (MMN) and the P3 complex for unexpected sounds, and the oN1, oN2, and oP3 complex for unexpected omissions. Moreover, unexpected sounds elicited an equivalent MMN, regardless of whether silence was expected (80%) or no reliable expectation was possible (50%), while later P3 components showed different amplitudes. Our results demonstrate flexible action-sound predictions at sensory and higher cortical levels. Furthermore, they indicate that predicted silence does not have an explicit sensory representation at lower levels but emerges at later stages, when higher-level information has been integrated.

The mismatch negativity (MMN) is a well‐studied event‐related potential (ERP) component in the EEG reflecting deviance detection in the auditory modality. It taps into the basic functioning of auditory regularity processing. The auditory multi‐feature paradigm is widely used in sensitive and special populations to measure the MMN simultaneously for different sound features in a short amount of time. It is consensus in the field that both adaptation and genuine deviance detection contribute to the “classic” MMN computed as deviant minus standard ERP difference. However, no attempts have yet been made to disentangle adaptation from the “genuine” MMN in the multi‐feature paradigm. Here, we propose a cascadic control condition for the auditory multi‐feature paradigm that controls for adaptation and physical differences between standard and deviant sounds. Using this new paradigm, we measured the genuine MMN, computed as deviant minus control ERP difference, for frequency, location, intensity, and duration deviants. The genuine MMN amplitudes for frequency and location were found substantially smaller than in traditional paradigms. No genuine intensity MMN and only a later and smaller genuine duration MMN were found. The results suggest stronger contributions of adaptation than in the traditional oddball paradigm. Controlling for adaptation is particularly relevant in research concerning predictive processing and the use of the MMN as a biomarker related to impaired NMDA receptor synaptic transmission as observed in schizophrenia. The presented multi‐feature cascadic control condition enables the measurement of the genuine MMN, which presumably reflects higher‐order cortical computations, such as predictive processing, still in a short amount of time.

The mismatch negativity (MMN) is a well-studied event-related potential (ERP) component in the EEG reflecting deviance detection in the auditory modality. It taps into the basic functioning of auditory regularity processing. The auditory multi-feature paradigm is widely used in sensitive and special populations to measure MMN simultaneously for different sound features in a short amount of time. It is consensus in the field that both adaptation and genuine deviance detection contribute to the “classic” MMN computed as deviant minus standard ERP difference. However, no attempts have yet been made to disentangle adaptation from “genuine” MMN in the multi-feature paradigm. Here, we propose a cascadic control condition for the auditory multi-feature paradigm that controls for adaptation and physical differences between standard and deviant sounds. Using this new paradigm, we measured genuine MMN, computed as deviant minus control ERP difference, for frequency, location, intensity, and duration deviants. Genuine MMN amplitudes for frequency and location were found substantially smaller than in traditional paradigms. No genuine intensity MMN and only a later and smaller genuine duration MMN were found. The results suggest stronger contributions of adaptation than in the traditional oddball paradigm. Controlling for adaptation is particularly relevant in research concerning predictive processing and the use of MMN as a biomarker related to impaired NMDA receptor synaptic transmission as observed in schizophrenia. The presented multi-feature cascadic control condition enables the measurement of genuine MMN, which presumably reflects higher-order cortical computations, such as predictive processing, still in a short amount of time.

When an action does not yield the sensory action effect associated with it, prediction error (PE) signals are generated. The present study investigates auditory event-related potential (ERP) markers of PE for violations of strong action-effect couplings as a function of whether the participants’ intention focuses on a particular action or a particular action effect. Participants produced high- and low-pitched tones by left and right button presses according to a pattern of visual stimuli. While the instruction as well as the action and tone sequences produced by the participants were identical, the type of intention induced by the visual symbols differed (in a between-subject design): In the action-effect intention group, the pattern consisted of “notes” (for high/low pitch), in the action intention group it consisted of “letters” (for left/right button-press). Occasionally, a button-press did not produce the associated sensory action effect. These incongruent sounds elicited an enhanced auditory N1 compared to congruent sounds in the “notes” but not the “letters” group which we interpret as PE signal. Actions selected by the intended action effect in the “notes” group presumably induced a predictive sensory representation of the action outcome, its violation resulting in an early PE at the level of the auditory N1. N2b and P3 were elicited in both groups. This suggests that the respective action-effect couplings were represented, and their violation processed at a conceptual level also in the “letters” group. These results support theories postulating that event representations bind together features of stimuli, actions, and associated outcomes.