Advanced Principles of Psychology - Lecture 9

Physicalist Approach

  • Examines physical image features associated with 3D perception.
  • Employs a cue-based approach.
  • Utilizes Linear Weighted Averaging (LWA).

Depth Cues

  • Discovered by Artists (Pictorial Cues):
    • Contour
    • Shading
    • Interposition
    • Relative size
    • Height in the field
    • Texture
    • Perspective
    • Blur
  • Discovered by Scientists:
    • Binocular disparity
    • Motion Parallax

Depth Estimation using Linear Weighted Averaging (LWA)

  • Different cues are weighted based on reliability.
  • Examples:
    • Shading (S) - weight wS
    • Texture (T) - weight wT
    • Motion parallax (M) - weight wM
    • Binocular Disparity (D) - weight wD

Perceptual Judgments and Physicalist Model

  • Perceptual Judgments: Depth, Distance, Size, Shape, Orientation.
  • Physicalist model: 3D space and object perception
    • Size = [x<em>1,y</em>1,z<em>1],r</em>1,r1[x<em>1, y</em>1, z<em>1], r</em>1, r_1
    • Distances = [Z,z<em>1,z</em>2,z3][Z, z<em>1, z</em>2, z_3]
    • Orientations = [90,90,90],θ[90, 90, 90], θ
    • Geometry = Sphere, cube, etc.
  • Physical description of the external world considers depth, distance, size, shape, and orientation.
  • Depth & Distance cues (physical description of the retinal image and extra-retinal information).
  • Linear Weighted Averaging (“Mixing Model”).

Problems with Linear Weighted Averaging (LWA)

  • Binocular Viewing of a Picture:
    • Disparities within the picture are negligible.
    • Binocular disparities are present due to the picture in real space.
    • Disparity cue indicates flatness (depth = 0).
    • LWA predicts flattening or distortion of 3D space and object shape compared to pictorial cues.

LWA Model Prediction

  • Binocular viewing of a picture:
    • Binocular Disparity: No depth, Depth = 0
    • Perspective: Depth present, Depth > 0
    • Shading: Depth present, Depth > 0
  • LWA predicts that the 3D space and object shape when viewing pictures with both eyes should appear flattened or distorted compared to what the pictorial cues are specifying

Empirical Evidence and LWA

  • No empirical studies have shown an increase in perceived depth magnitude under monocular viewing of pictures compared to binocular viewing.

Monocular-Aperture Viewing

  • Lab results indicate differences in perceived slant with Mono-Aperture viewing vs. binocular viewing.

Further Problems with LWA

  • Predictions for perceived depth magnitude comparing binocular and monocular (or monocular aperture) viewing do not match psychophysical results.
  • Unable to explain why picture perception involves duality (seeing a 3D space in the picture and perceiving the picture itself as a flat surface).

Visual Duality in Picture Perception

Monocular Stereopsis

  • Historical context: Brunelleschi (1448), da Vinci (1560s), Wheatstone (1838), Claparède (1904), Ames (1925), Schlossberg (1945), Michotte (1948), Koenderink et al. (1994).
  • Monocular viewing can enhance the perception of depth and 3-dimensionality in images.

Phenomenology of Visual Realness

  • Sense of “real” separation in depth.
  • Sense of things “coming out of the picture.”
  • Feeling of tangibility (ability to reach out and touch things).
  • LWA does not explain this crucial aspect of 3D space and object phenomenology.

Perceptual vs. Physical Realms

  • Distinction between the physical description of a stimulus and the perceptual experience.
  • Need for an integrated model of 3D object and space perception that accounts for phenomenology.

Phenomenological Model

  • Analytic description of phenomenology connects to an abstract perceptual model.

Phenomenology of Egocentric Distance

  • Impression of how far away things are in terms of reachability.

Phenomenology of Depth

  • Impression that things are separated in depth from each other.

Phenomenology and Pictures

  • In pictures, we have a phenomenology of 3D shape and spatial layout, but not of visual scale.
  • We do not have a phenomenology of the actual distances between things and their actual size (inferred cognitively).
  • Distinction between exocentric distances (scaled depth) and relative depth (unscaled depth).

Exocentric Distances

  • The observer sees the objects located at specific distances from each other (units are arbitrary).
  • The observer apprehends the actual size of objects and the configuration (i.e. apprehends the scale).

Relative Depth Structure

  • Two configurations can have the same “shape” and “layout” but a different scale.
  • The ratios of distances between objects are the same.
  • Pictures appear to present a perception (phenomenology) of purely relative depth structure.

Conjecture

  • When looking at a picture normally with 2 eyes we are perceiving only relative depth structure (unscaled).
  • In certain stimulus conditions, we also perceive exocentric distances (scaled structure).
  • In these instances, we experience the phenomenology of stereopsis Conditions: Real world & stereoscopic (disparity), motion parallax, monocular-aperture viewing, synoptic viewing, etc.

Implications of Exocentric Distances

  • Are required to interact with objects and the environment.
  • The phenomenology of “realness” (stereopsis, “seeing in 3D”) can be thought of as the visual impression of the potential for interaction with the objects.

Dissociation

  • Dissociation between the phenomenology of egocentric distance and exocentric distance.

Blind Walking Paradigm and Egocentric Distance

  • Subjects walk blindfolded to a remembered distance to measure perceived distance.
  • Subjects are surprisingly accurate at this task.

The Loomis et al., 1992 Study

  • Blind walking to two different distances (separately, A and B).
  • Then the observer judges the distance between the two points using a probe.
  • Walking distances to A and B are accurate, but the judged distance between the two rods is incorrect (too low)

Phenomenology related to The Loomis et al., 1992 study

  • The vertical stripes are equally spaced in the scene yet they appear to be unequally spaced (the farter interval appears shorter than the nearer one.

Implications of Loomis et al., 1992 Study & Phenomenology

  • Suggests that there is a separate encoding of distance perception for ambulatory distances

Theoretical Proposal (Vishwanath, 2014, 2023)

  • 3D space and object perception entails three distinct forms of spatial phenomenology:
    1. The phenomenological impression of the distance to objects (egocentric distance)
    2. The phenomenological impression of 3D object shape and layout (relative (unscaled) depth structure)
    3. The phenomenological impression of the actual distances within and between objects and object size [exocentric (scaled) distances]
  • Each type of phenomenology is caused by a different and separate neural encoding in the brain and is optimised by evolution for distinct areas of space.

Evolutionary Perspective (Vishwanath, 2014, 2023)

  • Egocentric distance perception is the evolutionary oldest spatial capacity and is a separate system associated with locomotion and reaching.
  • Capacity to perceive 3D object shape and layout evolved before the capacity to perceive exocentric distances.
  • Capacity to perceive exocentric (scaled) distances is evolutionarily the most recent and is associated with grasping and manipulation and only optimised for near space (reach space).

Encodings & Phenomenology

  • Egocentric Distance: Distance to objects
  • Relative (Unscaled) Depth: 3D shape and layout
  • Scaled (Exocentric) Depth: Object size and separation between objects (stereopsis, realness).

Relative Depth

  • Perceiving relative depth implies that the distance to the objects is unknown, i.e. X is unknown

Exocentric distances

  • Perceiving exocentric distances requires implicit knowledge of the egocentric distance to the object or configuration

Depth and Distance Cues

  • Depth Cues:
    • Contour
    • Shading
    • Interposition
    • Relative size
    • Height in the field
    • Texture
    • Perspective
    • Blur
    • Binocular disparity
    • Motion Parallax
  • Distance Cues (Near Space):
    • Accommodation
    • Convergence
    • Depth of Focus Blur
  • Distance Cues (Far Space):
    • Declination from eye level
    • Ground-plane information

Distance Cues (Extra-Retinal)

  • Lens accommodation (0.5m)
  • Ocular convergence angle (2m)
  • Declination from eye level (5m-20m)

Integration of Cues

  • Near-distance cues scale depth cues to derive exocentric distance.

Sensory Input, Encoding, and Phenomenology

  • Sensory Input: Depth Cues, Near Distance Cues, Far Distance Cues
  • Encoding: Relative (Unscaled) Depth, Exocentric (Scaled) Distances, Distance
  • Phenomenology: 3D shape and layout, Object size and separation between objects (stereopsis), Distance to objects