Funded by
What we finally did, can be aligned along four lines of research:
...Here, we sharpened our own thinking on prediction. This resulted in several conceptual, theoretical, and methodological papers. One achievement is the development of a conceptual model, in which István Winkler and me join two separate research fields, namely auditory scene analysis and auditory deviance/target detection Winkler & Schröger, 2015). Noting that both functions relate incoming information to what is already known about the environment, we argue that object representation and deviance detection can be described as relying on a common generative model of the auditory environment, called the Auditory Event Representation System (AERS). This framework was connected to a computational model of auditory stream segregation, called CHAINS, which has been developed by Sue Denham (UK) and colleagues (Schröger et al., 2014). In CHAINS, the auditory sensory event representation of each incoming sound is considered for being the continuation of likely combinations of the preceding sounds in the sequence, thus providing alternative interpretations of the auditory input. Detecting repeating patterns allows predicting upcoming sound events, thus providing a test and potential support for the corresponding interpretation. Alternative interpretations continuously compete for perceptual dominance. In a further theoretical approach we discussed a common framework for attention and prediction, two areas which were rarely considered together (Schröger et al., 2015). We suggest that “prediction” relates to predictions sent down from predictive models housed in higher levels of the processing hierarchy to lower levels and “attention” refers to gain modulation of the prediction error signal sent up to the higher level. As predictions encode contents and confidence in the sensory data, and as gain can be modulated by the intention of the listener and by the predictability of the input, various possibilities for interactions between attention and prediction were unfolded. From this perspective, the traditional distinction between bottom-up/exogenous and top-down/endogenous driven attention was revisited and the classic concepts of attentional gain and attentional trace were integrated.
...Here, the predictions we considered involved some sort of action intended to generate a sound. This approach receives more and more attention in many labs not only in the field of perception but also in social neuroscience and in animal research. In one approach, in collaboration with Marc Schönwiesner and Julian Keil (Montreal), we manipulated the pre-reflective sense of agency using single pulse transcranial magnetic stimulation over primary motor cortex (Timm et al., 2014, JN). In this way we could generate an involuntary motor act (a finger movement), that, in turn, generated a sound by pressing a button. Interestingly, this involuntary motor act was nevertheless the result of activity in the participant’s motor cortex, and we can thus speculate that the efference copy of the command would be issued as well. Thus, all other things being equal to a voluntary button press, with this procedure the participant would not feel as the agent of the self-initiated sound. Results show no N1-attenuation for self-initiated sounds under these circumstances, thus we can speculate that sensory attenuation of auditory brain responses to self-initiated sounds depends on a pre-reflective sense of agency. In another important series of studies (with Iria SanMiguel as driving force), we combined the basic ingredients of paradigms tapping into regularity extraction in the auditory domain and paradigms tapping into motor-driven prediction effects in auditory processing (SanMiguel et al., 2013). Participants repeatedly initiated a sound by pressing a button, and on a low percentage of the button presses, the sound was unexpectedly omitted. With this procedure we induced a very strong and simple auditory prediction that was precisely time-locked to the motor act. Thus, by omitting the expected sound, we could gain access to the underlying neural activity coding for the prediction. The findings obtained were quite remarkable: when the self-initiated sound was omitted an auditory-like response (as revealed by event-related potentials and functional imaging measures) was elicited. In cooperation with Sonja Kotz (Maastricht) we could show that patients suffering from lesions in the cerebellum show particular deficits in predictive processing (Knolle et al., 2013, Cortex); in cooperation with Juanita Todd (Newcastle, Australia), we investigated predictive processing in Schizophrenia (in prep.).
...Here, auditory predictions were investigated resulting from within the auditory modality. Prediction refers to the modulation of brain res¬ponses to stimuli whose onset/offset obeys conditional inferences. In one experi¬ment, we studied under which conditions auditory stimulus predictability trans¬lates into prediction in the absence of attention to stimuli (pre-attentive setting). Recent theories distinguish between temporal regularities – which trace the tem¬poral relation of successive events - and formal regularities – which refer to the identity of an event. We thus tested the effects of temporal and formal regularities in modeling the onset of predictable and unpredictable deviant sine tone repeti¬tions in fast tone sequences. The results show that temporal and formal regula¬rities interact in suppressing activation for repeated deviants (Tavano et al., 2014, EJN). In another set of experiments, we (with Alexandra Bendixen as the leading figure) investigated how predictive processing may subserve one of the auditory system’s central tasks, namely auditory scene analysis: the disentangling of concurrent auditory signals emitted from different sources, and the connection of consecutive signals emitted by the same source into a stream of meaningful information (e.g., a series of footsteps, or a human utterance). We clarified one important prerequisite for a functional role of predictive processing in auditory scene analysis by demonstrating that predictive relations between events can be extracted from non-adjacent events, i.e. across intervening signals (Bendixen et al., 2012, Front. Psychol.). In terms of auditory scene analysis, this means that the auditory system can use its predictive power to disentangle a mixture of sounds.
...Here, we considered situations in which the source of the auditory prediction involves not only the auditory but also another sensory (i.e., the visual) modality. It turned out that auditory predictions can be set up by mere visual information but it must be beneficial for the subject to predict sound from vision. Moreover, as it is easy to construct independent experimental “worlds” from auditory-auditory and visuo-auditory links, we could investigate the interesting cases of how the auditory system handles contradictory and redundant predictions. It turned out that the auditory system maintains and tolerates independently represented predictions even if they are redundant or contradictory (Pieszek et al., 2013, PlosOne). Another experiment was based on recent research on dyslexia suggesting dysfunctional orthographic-phonemic binding as one cause of the reading impairment. In cooperation with a Finnish team around Mari Tervaniemi and Teija Kujala (Helsinki, Finland), we tested dyslexic children in a paradigm requiring the establishment of visuo-auditory links that are not speech-related. We hypothesized that a more general visuo-auditory deficit of mapping an incoming sound to a sound predicted on the basis of the visual input may be involved (Widmann et al., 2012, Front. Psychol.). Indeed, we were able to show that dyslexic children differed from non-dyslexic children in the particular ERP components and in oscillatory brain responses. This suggests a widespread impairment in predictive visuo-auditory processing and integration in dyslexic children, which contributes to the reading impairment in dyslexia.