Humans use internal models to make predictions. When the environment changes such that these predictions are suddenly violated, there is model reset, arousal, and potentially learning of a new model. These are accompanied by locus coeruleus activity/norepinephrine release, pupil dilation responses (PDRs), and a neural signal in MEG/EEG equivalent to the MMN. While MMN does not seem to require attention, it is still unclear whether PDRs require attention. To examine this, we generated auditory patterns consisting of regular or random sequences of tones. Regular sequences allow the construction of an internal model. When a regular sequence changes to a random sequence or to another regular sequence, then model reset is induced, as reflected in PDRs. In previous studies, although the sequence structure was task-irrelevant, participants’ attention was focused by requiring them to detect a short auditory gap. We examined the consequence of diverting their attention to vision by requiring them to report a change in a fixation cross. Although given auditory attention, PDRs were induced by model reset (as in Zhao et al., 2019, Nature Communications; Basgol et al., 2022, ECVP-conference), they were absent when participants had to perform the visual task. These results suggest that attention might be necessary for PDRs to signal model reset due to environmental changes.

Supported by German Research Foundation (DFG): SFB 1233, Robust Vision: Inference Principles and Neural Mechanisms, TP 05, No 276693517, Max Planck Society and Humboldt Foundation (PD), and Machine Learning Cluster of Excellence, EXC 2064/1 No 390727645 (VF).