Perception: the process of taking sensory input and interpreting it meaningfully. Perception can
be subdivided into visual perception, auditory perception, olfactory perception, haptic (touch)
perception, and gustatory (taste) perception.
Distal Stimulus: an object, event, or pattern as it exists in the world.
Proximal Stimulus: reception of information and its registration by a sense organ.
Retina: A layer of visual receptor cells at the rear of the eyeball. The retinal image is created
here; it is two dimensional, its size depends on your distance from the object, and the image is
upside down and reversed from left/right.
Percepts (your interpretation of a stimulus) is different than the proximal stimulus.
o Size Constancy- the phenomenon that one's perception of an object remains
constant even as the retinal image of the object changes size (for example, because
the object has moved closer or farther away from the perceiver).
o Pattern recognition- the recognition of a particular object, event etc. as belonging to
a class of objects, events, and so on. The formation of most percepts involves some
classification so recognition of patterns is essential.
Gestalt Approaches to Perception
Form Perception- the process by which the brain differentiates objects (figures) from
their backgrounds (also called the ground). Most of us perform this cognitive task quickly
and take it for granted, believing that we perceive objects and backgrounds simply
because there really are objects and backgrounds and all we do is see them.
Robert Solso- intersection of cognitive psychology and the visual arts, presents a "blow
up" of the hidden "bust of Voltaire" in the bottom half of the figure.
Gregory (1972)- subjective contours- illusory outline created by certain visual cues that
lead to erroneous form perception. The existence of this phenomenon suggests that
perception is an active constructive process. He believed it was they mind's way of
Gestalt psychologists are particularly concerned with how people apprehend whole
objects, concepts, or units.
General Law: Law of Pragnaz- mind will recognize the simplest interpretation of a stable
o Principle of Proximity- objects close together will be grouped together.
o Principle of Similarity- similar objects will be grouped together.
o Principle of Closure - when looking at subjective contours, the mind has a tendency
to fill in gaps.
Minimal Model Theory is an attempt to formalize the above law.
Data driven processes
Perceiver starts with environment and combines information to get a percept.
Ex. Looking around the room and noticing that you are inside.
Think of how an ATM can read cheques; information is compared to templates or
o The problem with this idea and humans is that we face numerous variations of
symbols everyday, such as different styles of handwriting, yet we can still recognize
and understand the symbols.
Viewing a scene/object based on its individual features; the recognition of the whole
model relies on the recognition of all its parts.
o Ex. Lettvin et al- attached electrodes to retina cells in frogs and found that certain
stimuli fired the cells more frequently.
Edge-detectors notice the boundary between light and dark.
A detector exists for moving edges
"Bug-detectors" notice small marks or imperfections.
Horizontal-line and Vertical line detectors were discovered by Hubel and
Wiesel in cats and monkeys.
Irving Biederman (1987)- theory of object perception in which people view objects in
segments of geometric components called geons of which there were 36. From
these geons, we could construct mental representations of many objects just as we
can create any English word from a list of Phonemes (basic units of sound).
Known as Recognition of Components
People can complete in-complete objects given that their vertexes are present.
Tarr and Bulthoff (1995)- present a complex competing idea.
o Evidence for featural processing--> flashing letter on a computer; more errors made
between similar looking letters such as C and G.
Eleanor Gibson (1969)- tabulated features of capital letters for the Roman alphabet
o Neisser (1963)- visual search task with arrays of letters, in which they had to respond
if they detected a specific letter. When presented among letters with straight and
angular lines, Z's were harder to detected than Q's were.
Similar findings have been found with articulatory features and auditory
perception of syllables.
o Selfridge (1959)- Pandemonium Model- for the perception of letters; consists of a
number of "demons" which act as feature detectors. The demons will scream based
on what they perceive. The softness or loudness is how demons communicate their
level of confidence that the feature is indeed present.
Image Demon-->Feature Demon-->Cognitive Demon--> Decision Demon
Prototype Matching- matching an input to a stored representation of information called a
prototype which is an idealized presentation of some class of objects or events (ex. Dog-->
Sensory device registers new stimulus and compares it to prototypes for an
o Different in that this model doesn't require an object to contain any one specific
feature or set of features to be recognized. o Posner and Keele (1968)- series of dot patterns of 9 dots in a 30x30 grid to form a
letter, a triangle, or a random pattern. Dots were moved to distort the image. They
were shown brand new grid and could classify old stimuli 87% of the time, 67% of
new stimuli, and 85% of prototypes that they had never seen before.
In first task, people formed mental representations of each class of items.
o Cabeza, Bruce, Kato, and Oda (1999)- photographs of faces altered by displacing
features up or down a few pixels. Participants were more likely to "recognize"
prototype faces they had never actually seen than to recognize other less
prototypical new faces.
Problems with Bottom-up Models:
o Context and Expectation Effects- the context (information surrounding the target)
sets up certain expectations in the mind of the cognitive processor.
Theory or conceptually driven processes
Processing, expectations, theories, or concepts guide the selection and combination of
information in the pattern recognition process.
David Marr (1982)- proposed model of interaction between top-down and bottom-up
o Proceeds in many different special purpose computational mechanisms (ex. For
colour or motion). They work autonomously form the other modules and real-world
knowledge meaning they are bottom-up processes.
o Visual perception involves 3 mental representations:
Primal sketch- relative brightness/darkness in 2D image and geometric
2 1/2 D sketch- Shading, texture, edges, etc. to determine depth.
3D Sketch- recognition of objects and understanding meaning of visual scene.
o How perception changes with practice.
J.J. Gibson and E. J. Gibson (1955)
Participants shown an original card then were expected to discern others that