PSYC 2121 Chapter 4- Recognizing Objects

The binding problem

a question, if visual processing is done like this, how do things get bound together?

agnosia

neuro disorder characterized by the inability to recognized familiar objects

• damage in parietal/temporal/occipital

how many different types of agnosia are there

2

apperceptive and associative

Apperceptive agnosia

you can perceive an object's individual features, but you can't bind them together.

• Can perceive an objects features but not the object in entirety

Ex: patient D.F, drawing a ladel,

Sometimes people can infer based on their knowledge to make conclusions , but they cannot perceive these things as intact objects

Cannot solve their own binding problem

What is impaired in apperceptive agnosia?

Visual perception

What is apperceptive agnosia caused by

dissuade damage in the brain

(parietal, temporal, and/or occipital damage).

Patient DF: had bilateral lesions in …

the occipital cortex

Associative agnosia

cant name objects or link them to their functions, but they can perceive the entire object with bound features

Ex: i see the object and i see it as bound (one) but idk what the object is

Ex: I see a small yellow ball.. But I cannot make the connection to my own knowledge therefore idk what it is

No connection to semantic knowledge.

Associative agnosia is caused by damage to the…

left temporal lobe.

What does recognition involve?

binding features and recognizing the full object (that’s a line with a circular end) and linking that with your knowledge and experience (that must be a spoon)

People with agnosia struggle in these areas

with intact rec, you can…

• You can recognize a huge variety of objects within a category

• You can recognize objects even with incomplete information (“fill in the blanks”)

• You can recognize a huge number of words, regardless of capitalization and font

Sometimes the stimulus input may not vary, but the context varies

what does intact recognition rely on?

both physical features and knowledge (context)

Bottom up processes

directly shared by the stimulus

data driven/ stimulus driven

Top down processes

shared by context or task goals

Experience driven

T/F: Features are rapidly processed in parallel but this happens before binding/recognition, your brain is tuned in to individual features

True

Single feature search-

ooking for one thing, happens in parallel

Ex: look for something red.

Single feature is special

Conjunction search

searching for a multi thing takes longer than searching for a single feature.

• Ex: look for something red AND horizontal

• We are slower at this

how fast is rec?

depends what we are recognizing, the context, etc

Factors that influence recognition speed (4)

1. Familiarity-more common words are easily rec

2. Priming- words that were just seen is more easily rec

3. Context- letters are easier to rec when they are inside words (Word superiority effect)

4. Well-formedness: Letter sequences that conform to typical spelling patters are easier to recognize (HZQY vs. FIKE)

Well-formedness

letter sequences that conform to typical spelling patterns are easier to rec (HZQY vs FIKE)

are errors systematic?

yes. DPUM is likely read as DRUM

Word rec allows us to read by

a network of detectors that is organized into layers.

Moving from the bottom up

As we move up the layers, the networks become concerned with objects on a slightly larger scale – information flows from the bottom up.

Bottom up

Feature detectors ➡letter detectors ➡bigram detectors ➡word detectors

what does the starting activation at teach level depend on

recency

We don't need to start at feature detectors, we could start at bigram

If it has recently fired (saw), it will have a higher activation to start

what does the starting activation at teach level ALSO depend on

frequency

If it fires often, it will have higher activation to start

why is baseline activation important?

it determines how much info we need

what does recency and frequency explain?

well formedness errors

starting activation at each level depends on…

recency and frequency

what takes longer, conjunction search or a single feature search?

conjunction search