Work

Hypopriors in autism spectrum disorder

Year: 2020
Type: dissertation
Abstract: Perception has traditionally been conceived as a bottom-up process: a stimulus from the environment enters the brain through sensory organs, after which it is refined until it is at some stage perceived (Purves et al., 2008;Smith S.M., 2009).Helmholtz (1867) turned this idea on its head: rather than sensory processing flowing one way from the senses to consciousness, he posited that humans use their perceptual system to test pre-existing hypotheses about the state of the external world.Here, the purpose of the senses is to sample data from the environment, guided by prior knowledge.This framework entails two key components: first, the hypothesis about the world.This component can be called a prior (Pellicano & Burr, 2012;Yuille & Kersten, 2006), prediction (K.Friston, 2005;Rao & Ballard, 1999), or a generative model (Clark, 2013).I will use the term 'prior' .The second component are the data sampled by the senses.This, too, is known under different names, such as likelihood (Pellicano & Burr, 2012;Yuille & Kersten, 2006) or sensory evidence (Clark, 2013).I will use the latter term.When perception is discussed as a type of Bayesian inference, the prior and sensory evidence are probability distributions that jointly determine what we perceive.Let us try to imagine this in practice.Many animals use camouflage to protect themselves from predators.Among these are the family of phylliidae, or leaf insects, who take the appearance of leaves (Fig. 1.1A).The sensory evidence provided by these insects is a flat probability distribution: its shape matches everything in between a leave and an insect equally well.Thus, without additional information (prior), you would be equally likely to see this as a leave or an insect (Fig. 1.1B).Having priors changes the equation.When visiting a zoo and having read about phylliidae, we might expect to see them (Fig. 1.1C).This prior shifts the resulting posterior distribution towards 'insect' rather than 'leaf' (Fig. 1.1D).Since the posterior represents what we perceive, we perceive the animal as what it really is.Elsewhere than at the zoo though, we will not have an expectation for phylliidae.Our prior would be wide and imprecise (Fig 1.1C).This imprecise prior would do little to resolve the ambiguity of the leaf insect (Fig. 1.1D).Hence, it would successfully camouflage itself as a leaf for us.The fact that the sensory response is influenced by prior experience has been observed in several neuroimaging studies.For example, repeated stimulation with the same stimulus evokes a weakened response in areas that are tuned to that stimulus (Grill-Spector et al., 1999;Grill-Spector & Malach, 2001;Krekelberg, Boynton, & van Wezel, 2006).This phenomenon is called repetition suppression (RS), or sometimes fMRI adaptation (e.g.Larsson & Smith, 2012).RS is not limited to one modality or stimulus class.In human vision, RS is for instance reliably found with stimuli such as houses (Grotheer, Hermann, Vidnyánszky, & Kovács, 2014),
Language: English
Open Access status: green
OpenAlex ID: W3004886123
Cites: 181
Cited by:
Related to: 9