There is a newer version of this record available.

Journal article Open Access

Timescales and functional organization of neural event segmentation in the human brain

Geerligs, Linda; van Gerven, Marcel; Campbell, Karen L.; Güçlü, Umut

During continuous perception, humans automatically segment their experience into discrete events. At the neural level, each brain region is thought to segment ongoing experience at its own preferred timescale. Previous studies have shown that event segmentation follows a temporal hierarchy in the brain, with regions higher in the hierarchy segmenting experience into events with a longer duration. The assumption is that incoming information is processed and then passed on to higher order regions as discrete units (or chunks), but little is known about how event boundaries are shared both within and across networks. To this end, we use Hidden Markov Models (HMMs) and functional connectivity analysis to study the timescales of event segmentation over the entire cortex and investigate how event boundaries are shared across regions and networks. We demonstrate that there is a distinct temporal gradient of information processing over the entire cortex, with particularly fast events in primary sensory regions and long periods of information integration in the precuneus and medial prefrontal cortex. We also show that event boundaries are shared between regions within functional networks, as well as across the temporal hierarchy between distinct networks. Finally, we observed that the traditional default mode and fronto-parietal networks fractionate into subnetworks with fast and slow event timescales, possibly reflecting functional specialization within these networks during naturalistic viewing. Together, these results provide the first complete overview of how event segmentation is organized in the human brain.

Files (3.7 MB)
Name Size
Paper movie HMM_v4.pdf
3.7 MB Download
All versions This version
Views 20166
Downloads 12639
Data volume 495.7 MB144.1 MB
Unique views 16861
Unique downloads 11236


Cite as