PS1061: Sensation and Perception 2014-15
Term 2,    Thursday 11 am - 1 pm (Windsor Auditorium)

Lecture 4: Travelling through Space and Time

Course co-ordinator: Johannes M. Zanker,, (Room W 246)

Lecture Topics

the dimensionality of the world

        Mobius Belt II M.C.Escher, 1963

the world has three spatial and one temporal dimension

the third dimension : depth perception

Bacchus and Ariadne
Titian, 1522
National Gallery, London

the third dimension needs to be reconstructed from the flat images captured by the eyes
since the Renaissance painters studied perceived depth and became practising experts on how to create the illusion of space on a flat canvas

a multitude of cues can be used in real world (and some of them, 'pictorial cues', are used by painters):

depth cues 1: pictorial cues

a wide range of depth information can be directly extracted from a static monocular (& natural) image (Gibson 1979)

depth cues 2: binocular

why do we have two eyes?
apart from extending the visual field, the combination of information from the two eyes allows precise depth measurements through stereopsis


the left eye view and right eye view different images; they can be combined to retrieve stereo and oculomotor signals as depth cues !

and this is how stereopsis works (for the full story, see Julesz 1971, and Marr & Poggio 1979):
consider an observer fixating the horizon (close to infinity) : parallel optical axes of left and right eye

the retinal projection of an object on opposite sides of the fovea (indicated disparity angle delta, delta') indicates its depth relative to the plane of fixation

stereopsis is exploited to produce depth impressions in projected/printed pictures
(Wheatstone stereoscope, red-green anaglyphs, lenticular stereocards)
single image sterograms (Julesz & Miller1962, Tyler & Clarke 1990, Thimbleby et al. 1992) were commercialised by Tom Baccei and Cheri Smith in 1991: magic eye, more magic eye
(try the flying sausage experiment

depth cues 3: motion

when you look out of the side window of a car or a train, you see close objects translating very fast (bushes) and distant objects passing very slow (mountains) or even being stationary (sun)
the inverse relation between angular speed and distance is called motion parallax (Rogers & Graham 1982)

size illusions 1 : size constancy

using distance information, the retinal (angular) size of objects can be ‘corrected’ to make perceived (object) size independent of distance: size constancy

(see Gregory 1998)

conversely, constant angular size (the two kangaroos have the same size in the image) may be interpreted as difference in object size (the closer kangoroo looks smaller than the distant kangoroo): constancy scaling
the size constancy effect is sometimes believed to be the basis of the Ponzo illusion : the closer red bar between the converging rails looks smaller than the more distant red bar...


size illusions 2 : Ames Room

In the Ames Room, even the size of a familiar object (such as a person) is perceived largely distorted, because the misleading geometry generates a incorrect frame of reference

>> top view of room geometry <<

the size illusion in the Ames Room is a case in point for constructivist theories of perception: the knowledge of the rules of perspective and the assumption of rectangular architecture force the visual system to construct the apparent size difference (this way of drawing conclusions about what is seen in the retinal images, is called 'unconscious inference', see Gregory 1998)

(however, think about the following: with the same theory one could also argue the other way round: why are we not using our knowledge about body size to reconstruct veridical, non-rectangular geometry of the room?)

for instructions to build your own Ames room, and more information about this illusion, click here

so how do you explain the 'hollow face illusion'?

this illusion seems to be a combined effect of

  • 3D-shape from shading, due to illumination from one side
  • preference for the cardinal view of a convex face surface

from the Max Planck Institut für biologische Kybernetik in Tübingen, with kind permission by H Bülthoff, (see also Gregory 1998)

for instructions to build your own hollow face illusion, click here

an application: 3D-movies

how can a ‘realistic’ depth sensation be produced in the cinema, on a flat screen ??

        Visit the Science Museum, or IMAX® Cinema for this extraordinary experience

what connects space and time ? motion !

To understand motion we need a bit of minimal physical background, explained by a simple example:
Consider an astronomer who is watching the night sky to find stars...
she might see two stars next to each other : they appear at a spatial distance
(dimension x)
she might see a star appearing, dis-appearing, reappearing: it is visible in temporal intervals
(dimension t)
she might see a star moving : it is changing position in space and time; this is motion (in space-time: x-t-dagram, motion is characterised by orientation, oblique trajectories)

motion perception is a prime example for the study of brain function with a mixture of neuroscientific methods: physiology, psychophysics, and computational modelling all contribute to a comprehensive understanding of the fundamental processing mechanisms

motion: displacement, time, direction

If we look out of window and watch cars passing by – how could we illustrate this observation as still image ?

a motion detection model

a motion detector has the task to assess displacement as function of time (the phiscal definition of motion): technically, this is called spatio-temporal correlation (and can be illustrated as orientation filter in space-time) (see Reichardt 1961, Borst & Egelhaaf 1989)

minimum requirements for a computational model (elementary motion detector = EMD):
  • two spatial separate inputs to measure changes across space: D x
  • temporal filters (delay) to measure changes across time: t
  • a comparator (logical operator) to evaluate spatial and temporal changes - coincidence of original signal from one point in space and delyed signal from a neighbouring point in space leads to a positive output signal

apparent and real motion

apparent (PHI) motion : a set of discrete displacements (jumping)

real motion : continuous (smooth) displacement across space & time

apparent motion history

it was know since a long time and used for props at country fairs, etc: the 'zoetrope' (for some activity, see also here) - a pioneer in this area was Eadweard Muybridge who in some way invented cnema (visit the Museum in Kingston!)
it was 'discovered' experimentally by Exner (1877), who demonstrated that  motion is an independent sensation in space and time

motion correspondence

ambiguous motion stimuli can be used to identify motion processing mechanisms : matching across space and time is the basis of motion perception (Ramachandran & Anstis 1986 )

a typical phenomenon: rapid alternations between two dots in opposite corners of a virtual square can be perceived as jumping up & down, or as jumping left & right

vertical proximity or horizontal proximity can resolve this ambiguity
>>> Gestalt laws of perceptual organisation: 'proximity' 'common fate' (Wertheimer 1912)

discontinuity in apparent motion stimuli >> ambiguity >> need to identify objects in successive frames :
solving the 'correspondence problem'

which dot in image 1 corresponds to which dot in image 2 ???

immedeately resolved by motion correspondence: you see a moving circle !!

the appearance of a motion-defined object is sometimes called 'structure from motion' (see Ullman 1979)

perceptual organization: Gestalt

according to Gestalt Theory, perception is not just passive image acquisition, but is an active process to create meaningful percepts:
'laws’of perceptual organization generate ‘good shapes’ (Wertheimer 1912)  - Gestalt Psychology (Koffka 1935)
  •  Praegnanz: of several geometrical possible organisations, the most simple, stable will be pereived (this generates many illusions)
  •  Proximity: tendency to group elements close to each other (e.g. apparent motion, or here horizontal rows of dots)
  •  Similarity: tendency to group elements that are similar (e.g. segregation, here vertical columns of dots with same colour)
  •   Good Continuation: a tendency to generate smooth contours (‘inertia’) leads to simpler interpretations (here two intersecting lines)
  •  Closure: interpretation is dominate by the tendency to complete shapes (here a ring that is occluded by a white shape)

motion aftereffect ...

after you fixate the centre of a rotating spiral (generating a sensation of expansion) for 2 minutes ...

what happens when you look at HM's on the coin ?   >>>>   (static objects seem to contract)

this effect can be understood as the result of adaptation and opponency mechanisms, similar to colour aftereffecrts
classical case for such dynamic afterimages : Waterfall Illusion (for a brief history, see Wade 1994)

apertures 1: the problem

why does the rotating spiral (in the panel shown above) appear to expand ?? it is rotating, not expanding, after all !

consider the motion of an individual contour - it is ambiguous in an aperture (receptive field, indicated by red circle on the right) : from the restricted view you dont't know whether it moves horizontally or vertically - the motion direction of a contour is under-determined within small regions !!
this direction ambiguity is usually referred to as 'aperture effect' (Wallach 1935)

        the most likely solution (direction perpendicular to the line) is perceived

apertures 2: plaids

gratings presented behind an aperture usually lead to an unambiguous percept: perpendicular direction of motion

superimposing 2 component gratings moving in perpendicular directions
what do you see in such a motion plaid ? (Adelson & Movshon 1982)
… it depends !
(in coloured plaids, for instance you can perceive both directions: gratings slide across each other - this is called 'transparency')

apertures 3: the barberspole illusion

Old Woking High Street

the barberspole illusion demonstrates how a particular shape of an aperture can change the perceived direction of motion (Wallach 1935)

it is often regarded as enforcing a particular solution to the aperture problem (directional ambiguity)

the shape of a particular aperture forces the visual system to adopt a particular solution when integrating ambiguous motion signals from the central regions of the aperture and disambiguated motion signals from the boundary regions (Castet & Zanker1999)

the classical barberspole configuration a recent version (JMZ, unpublished)

                            can you see two directions at the same time?

summary: depth and motion

General Reading:

Specific References:

to download a pdf copy of lecture slides, click here

back to course outline
last update 4-02-2015
Johannes M. Zanker