Bottom-Up Processing Psych 240

Bottom-up processing

refers to processing that originates in sensory areas and proceeds upwards towards areas responsible for higher cognition

pattern recognition

Matching what we are currently perceiving versus what we have stored in our memory to determine what we are seeing

shape constancy

the tendency to interpret the shape of an object as being constant, even when its shape changes on the retina

Template theory

We match up our perceptual representation with a template in our memory; We try to figure out which memory our perceptual representation best matches up with

Template theory Problems

1. Not all versions of a stimulus are the same out in the world, but we need to be able to recognize all versions regardless of size, position, color, ect → won't match our exact memory template exactly

2. Obstructed objects → we often do not get to see the exact pattern that we are trying to recognize

3. Visual System feature Decomposition →We do not see a whole template most of the time, we mostly see unique features. Cells in temporal lobe are not "tuned" to specific stimuli (no grandmother cell)

Feature theory

Perceptual Representation is broken down into a bunch of individual features and we match up those unique features with stimuli we have stored in our memory

evidence for features

Physiology (recordings from neurons)

Stabilized Retinal Images --> We would lose the ability to perceive individual features one by one if our eyes had not shaked → so we are computing individual features

Visual Search

visual search (feature theory)

To identify a stimulus we search for the unique/ distinct features that cause it to stand out from other stimuli we have stored in our memory. The stimulus is easier to recognize when it is competing against stimuli that are vastly different

Pandemonium model

- we have individual detectors for individual features of which count the number of times we see the feature in a stimulus; decoding specific features → feature demons

- Each stimulus/ pattern stored in memory signals to the feature demon when the stimulus you are trying to recognize may be a match; Shout when receive contain combinations of features → Cognitive demons

- Trying to decide which cognitive demon is the most accurate → Decision Demon

caricatures (feature theory)

Evidence for feature theory —> people are actually better at recognizing a caricature then they are an actual photograph because it emphasize the person’s distinctive features that we are able to remember the most

problems with feature theory

Does not tell us about spatial relationships between features → different arrangements of the same features produce different objects

Structural descriptions

We try to recognize objects based on their component parts and the structural relationships among those parts

RBC theory

1. Detect elementary features, edges

2. Find the non-accidental properties

3. Using the non-accidental properties, determine the component geons

4. Put the geons together to create a structural description and match to memory

geons

the primitive atoms of the structure of objects which can be put together in different ways which are stored in our memory

non-accidental properties

- Not all features are equal; some features are an accident of our viewpoint as opposed to others being more robust

- Straightness vs Curvedness, Cotermination vs no Cotermination, parallel vs antiparallel, symmetry vs asymmetry

- These non-accidental properties can be used to recognize the geons

evidence for RBC

1. Partial or degraded objects

- When we delete the accidental properties but can still see the non, the object is easier to recognize as opposed to the opposite

- When we don't have the non-accidental properties, we can't recognize the geon, without the geon we cannot recognize the structure

2. Object complexity

- Having a more complex perceptual object makes it easier to recognize

3. Unusual orientations

- When you look at a component at an angle of which you can see all the components, it is easier to determine what the object is

problems with RBC theory

1. Facial recognition

- if this theory were true then we would not be able to distinguish unique faces as the structures and the geons of every face are relatively the same

2. Brain evidence?

3. Context effects → Based on being purely bottom up theory, but doesn't take into account our biases and expectations of what to see