Innateness and (Bayesian) Visual Perception

Vocabulary flashcards on nativism, Bayesian inference in visual perception, natural constraints, unconscious inference, and developmental neurobiology based on the provided text.

Visual Perception

The process of recovering useful information about the structure of the world based on the shifting patterns of light that enter the eyes.

Inverse Problem

The fundamental challenge in vision where shifting light patterns on the retina are insufficient by themselves to uniquely fix the structure of the 3D external world via deductive inference.

Underdetermination

The condition in visual processing where a single 2D retinal pattern could correspond to an infinite variety of 3D objects in the world of different shapes, sizes, reflectances, or lightings.

Surface Lightness Equation

The mathematical relationship $R \times L = 12$ (where $R$ is surface reflectance and $L$ is illumination), illustrating that recovering reflectance alone requires the visual system to solve an underdetermined problem by 'unmultiplying'.

Coincidence Avoidance

A fundamental general principle of visual processing stating that the visual system inherently avoids interpreting visual data as the result of accidental configurations or unlikely coincidences.

Helmholtz's Principle of Unconscious Inference

The principle stating that what is perceived are essentially those objects and events that under normal conditions would be most likely to produce the received sensory stimulation.

Bayesian Inference

A formal method of optimal reasoning under uncertainty that specifies how to rationally choose from a set of mutually exclusive hypotheses ($H$) based on a given pattern of data ($D$).

Likelihood Function

In Bayesian visual perception, the expression $p(\text{Image} \mid \text{Scene})$, which models the probability of obtaining the observed retinal image assuming a particular candidate scene in the world is true.

Bayesian Prior

In Bayesian perception, the expression $p(\text{Scene})$ or $p(H)$, which models the visual system's degree of prior belief or natural constraints about world structure independent of current sensory data.

Posterior Distribution

The resulting distribution in Bayesian perception, $p(\text{Scene} \mid \text{Image}) \propto p(\text{Image} \mid \text{Scene}) p(\text{Scene})$, which specifies the relative degree of belief for each scene hypothesis after combining the likelihood function and prior.

Overhead Illumination Assumption

An innate or learned visual prior assuming a single light source originates from overhead in a head-centered reference frame, allowing the brain to compute 3D shape from shading gradients.

Overhead Illumination Illusion in Real Images

The phenomenon where photographs of real surface relief (such as moon craters) invert between convex bumps and concave dents when rotated 180^\n\n degrees, demonstrating the head-centered overhead illumination prior.

Hershberger's Chicken Experiment (1970)

An experiment showing that chicks raised in environments illuminated exclusively from below still assumed light came from overhead when discriminating bumps from dents, demonstrating an innate perceptual parameter.

Opportunistic Learning

A perceptual process where new priors are acquired or updated specifically during rare moments when the visual system detects conditions of relatively low ambiguity in the environment.

Spontaneous Endogenous Activity

Internally generated neural signals (such as independent retinal oscillations) that form and structure innate neural circuits prior to or in the absence of external environmental stimulation.

Informational Encapsulation

A defining property of perceptual modules wherein internal processes operate on a restricted range of inputs, though the module itself can still develop internally and update its constraints over time.