Cognitve Psych: 9/17/26 Lecture

Theory of Object Recognition: Gestalt Principles of Perceptual Organization

The whole is different than the sum of its parts; perception is determined by specific organizing principles

In this slide, on the right image when she plays the moving picture we see a dog running, now in the left image, although we did not observe it before we can now see a dog in the picture

Gestalt principles

• Similarity

• Proximity (Nearness) (things that are grouped together tend to be close together)

• Familiarity

• Symmetry (things that are or can be symmetrical tend to be grouped together)

• Good continuation  (When we connect these elements they tend to form smooth or curving lines. )

-Closure (Where we tend to fill in gaps of ‘unfinished’  shapes)

- Pragnanz (good figure, simplicity)

- Pattern perceived as simply as possible

Gestalt Laws of organization

• Basic organizing principles

- Similarity

- Proximity

- Familiarity 

- Symmetry

- Good continuation

- Closure

- Pragnanz

• However these principles were thought to be mostly intrinsic and bottom up, rather than learning through experience

Gestalt perceptual organization

• Usually accurate information

- Used unconsciously

- Sometimes misleading

Top down Model of Object Recognition: Helmholtz’s Theory of Unconscious inference (about 1860)

• Unconscious inference: Some perception are the results of unconscious assumptions we make about the environment

• Likelihood principle: we perceive the world in the way that is “most likely” based on our past experiences

Top-Down Model of Object Recognition: Bayesian Inference

Prior probability (similar to regularities) + likelihood

Face recognition

• Faces are highly varied and, except in identical twins, no two are alike

• But, significant similarities:

  • Same parts: (eyes, nose, mouth, etc.)

  • Parts always appear in the same configuration (nose below eyes, etc.)

- We seem to be hardwired to recognize faces, so much so that we see them in everyday objects (pareidolia)

- We are best when faces are right side up and looking forward; worst when faces are upside down

The thatcher illusion

• We process right side up faces in a typically gestalt- like manner, using spatial relations among the face parts to construct a holistic view

• However, when the face is in an unfamiliar rotation, holistic processing is disrupted and we switch to noticing individual facial features

Experience-dependent plasticity

Our brains can adapt and tune to our environment!

After training, FFA response almost as strongly to greebles as to faces!

Visual Agnosias: Prosopagnosia

Prosopagnosia: Unable to recognize familiar faces; sometimes referred to as face blindness

Theory of Perception and Action

• Perception and action work together.

• The brain uses two separate pathways/streams to:

  • Identify/recognize an object

  • Guide our actions toward the object

• Example: You see and recognize a cup → reach for it → grasp it.

We perceive objects not only to recognize them, but also to guide how we interact with them.

What vs Where Pathways

• There is pretty good evidence that the spatial information that helps us act (i.e. navigation; landmark; where) is independent of the information that helps us perceive and identify objects (what)

• Perception (ventral) pathway: 

  • From visual cortex to temporal lobe

  • Corresponds to the what pathway

-Action (dorsal) pathway:

- From visual cortex to parietal love

- Corresponds to the where pathway

To identify an object, which of the following brain structures is most active?

Temporal cortex

Perception and Action: Dissociation Logic

• Single dissociation: One function is lost, another remains

  • Example: Monkey A has damage to the temporal lobe. This monkey is no longer able to identify objects (what) but can still identify locations (where/navigation)\

  • Therefore, what and where rely on different mechanisms, but cannot claim that what is independent of where 

- Double dissociation: requires two individuals with different damage and opposite deficits 

- Example: Monkey A with temporal lobe damage has intact ‘where’ but impaired ‘what’; monkey B with parietal lobe damage has intact ‘what’ but impaired ‘where’

- Therefore, what and where streams must have different mechanisms AND operate independently of one another

Role of mirror neurons in perception

• Respond similarly when viewing an action and performing it; might be responsible for the why or intention of actions 

• Higher rate of mirroring if the subject’s intention to perform the action was greater

Neurons in the premotor cortex, originally discovered in the monkey.

They respond when a monkey observes someone else (usually the experimenter) acting and when the monkey itself performs the action. Evidence exists for mirror neurons in humans.