Research Article: Being right matters: Model-compliant events in predictive processing

Date Published: June 13, 2019

Publisher: Public Library of Science

Author(s): Daniel S. Kluger, Laura Quante, Axel Kohler, Ricarda I. Schubotz, Caroline Di Bernardi Luft.


While prediction errors (PE) have been established to drive learning through adaptation of internal models, the role of model-compliant events in predictive processing is less clear. Checkpoints (CP) were recently introduced as points in time where expected sensory input resolved ambiguity regarding the validity of the internal model. Conceivably, these events serve as on-line reference points for model evaluation, particularly in uncertain contexts. Evidence from fMRI has shown functional similarities of CP and PE to be independent of event-related surprise, raising the important question of how these event classes relate to one another. Consequently, the aim of the present study was to characterise the functional relationship of checkpoints and prediction errors in a serial pattern detection task using electroencephalography (EEG). Specifically, we first hypothesised a joint P3b component of both event classes to index recourse to the internal model (compared to non-informative standards, STD). Second, we assumed the mismatch signal of PE to be reflected in an N400 component when compared to CP. Event-related findings supported these hypotheses. We suggest that while model adaptation is instigated by prediction errors, checkpoints are similarly used for model evaluation. Intriguingly, behavioural subgroup analyses showed that the exploitation of potentially informative reference points may depend on initial cue learning: Strict reliance on cue-based predictions may result in less attentive processing of these reference points, thus impeding upregulation of response gain that would prompt flexible model adaptation. Overall, present results highlight the role of checkpoints as model-compliant, informative reference points and stimulate important research questions about their processing as function of learning und uncertainty.

Partial Text

Predicting upcoming events constitutes one of the fundamental qualities of brain function. Based on internal models shaped by previous experience, top-down predictions are compared to bottom-up sensory signals [1]. Redundant components of perceptual information are disregarded whereas surprising expectancy violations are propagated upward in the processing hierarchy [2–3]. Model adaptation in consequence of such prediction errors (PE) has been proposed to be the foundation of associative learning mechanisms [4–5], as unexpected events are particularly informative with regard to their current context. Importantly, probabilistically occurring expected events have also been suggested to inform the internal model [6]: While PE instigate model adaptation, expected events verify model-based predictions. These verifications are particularly informative when we face uncertain environments. A recent fMRI study [7] found that in uncertain environments, so-called checkpoints (CP) emerged as points in time where distinctive processing of expected events pointed to a context-sensitive adaptation in predictive processing. While the entire stimulus sequence could be predicted reliably in stable environments, unstable environments prompted stepwise predictions. This way, CP were used to verify the internal model in order to predict the next section accordingly. Thus, while model adaptation is induced by prediction errors, context-dependent model evaluation does not seem to require expectancy violations. Instead, selected time points carry information about the on-line validity of the internal model, raising the intriguing question of how checkpoints and prediction errors functionally relate to one another.

The study was conducted according to the principles expressed in the declaration of Helsinki and approved by the Local Ethics Committee of the University of Münster (Department of Psychology).

Predicting events of everyday life, our internal model of the world is constantly compared to sensory input we perceive. Prediction errors induced by unexpected events are deemed particularly informative in that they instigate learning through model updating. We show here that information is equally sampled from expected events at particularly relevant checkpoints, suggesting that under uncertainty, model-affirmative events similarly prompt recourse to the internal model. Both checkpoints and prediction errors showed a significant P3b component when compared to sequential standards, indexing the relative (im)probability of CP and PE occurrence. Conversely, the direct comparison of CP and PE revealed a significant N400 component as the mismatch correlate elicited solely by prediction errors. Combined with findings from behavioural and functional microstate analyses, checkpoint characteristics highlight the significance of informative reference points for abstract predictive processing, raising intriguing questions for future research.

Checkpoints are probabilistic, cue-compliant events informing predictive processing. Their functional profile closely resembles that of canonical prediction errors, indicating similar roles of the two event classes in abstract prediction. Both types of events presumably serve as reference points providing behaviourally relevant information, the central distinction being whether the respective outcome violates the internal model (PE) or not (CP). We suggest that despite the expected input observed at checkpoints, information at these particular positions is exploited on-line in order to adapt behaviour. Intriguing questions remain with regard to underlying network dynamics and their potential modulation as a function of uncertainty.




Leave a Reply

Your email address will not be published.