Help Privacy Policy Disclaimer
  Advanced SearchBrowse





Causal inference in multisensory perception and the brain


Rohe,  T
Department Human Perception, Cognition and Action, Max Planck Institute for Biological Cybernetics, Max Planck Society;
Research Group Cognitive Neuroimaging, Max Planck Institute for Biological Cybernetics, Max Planck Society;

Fulltext (restricted access)
There are currently no full texts shared for your IP range.
Fulltext (public)

(Publisher version), 5MB

Supplementary Material (public)
There is no public supplementary material available

Rohe, T. (2014). Causal inference in multisensory perception and the brain. PhD Thesis, Eberhard-Karls-Universität, Tübingen, Germany.

Cite as: https://hdl.handle.net/11858/00-001M-0000-0027-80D4-1
To build coherent and veridical multisensory representations of the environment, human observers consider the causal structure of multisensory signals: If they infer a common source of the signals, observers integrate them weighted by their reliability. Otherwise, they segregate the signals. Generally, observers infer a common source if the signals correspond structurally and spatiotemporally. In six projects, the current PhD thesis investigated this causal inference model with the help of audiovisual spatial signals presented to human observers in a ventriloquist paradigm. A first psychophysical study showed that sensory reliability determines causal inference via two mechanisms: Sensory reliability modulates how observers infer the causal structure from spatial signal disparity. Further, sensory reliability determines the weight of audiovisual signals if observers integrate the signals under assumption of a common source. Using multivariate decoding of fMRI signals, three PhD projects revealed that auditory and visual cortical hierarchies jointly implement causal inference. Specific regions of the hierarchies represented constituent spatial estimates of the causal inference model. In line with this model, anterior regions of intraparietal sulcus (IPS) represent audiovisual signals dependent on visual reliability, task-relevance, and spatial disparity of the signals. However, even in case of small signal discrepancies suggesting a common source, reliability-weighting in IPS was suboptimal as compared to a Maximum Estimation Likelihood model. By temporally manipulating visual reliability, the fifth PhD project demonstrated that human observers learn sensory reliability from current and past signals in order to weight audiovisual signals, consistent with a Bayesian learner. Finally, the sixth project showed that if visual flashes were rendered unaware by continuous flash suppression, the visual bias of the perceived auditory location was strongly reduced but still significant. The reduced ventriloquist effect was presumably mediated by the drop of visual reliability accompanying perceptual unawareness. In conclusion, the PhD thesis suggests that human observers integrate multisensory signals according to their causal structure and temporal regularity: They integrate the signals if a common source is likely by weighting them proportional to the reliability which they learnt from the signals’ history. Crucially, specific regions of cortical hierarchies jointly implement these multisensory processes.