week 13 (recognition + attention) Flashcards
(24 cards)
question: which gestalt rule dominates?
- committee model
⤷ loudest member wins but not permanently - ambiguous figures -> many diff. interpretations
- accidental viewpoints -> producing regularity that’s not real
⤷ ex. illusion of holding the tower of pisa
question: how to separate figure from ground in figure ground? (5)
- surroundedness
⤷ if one region is entirely surrounded by another, surrounded area = figure - size
⤷ smaller region = figure - symmetry
⤷ symmetrical region = figure - parallelism
⤷ parallel contours = figure - relative motion
⤷ how details move relative to an edge helps differentiate
explain: bayesian approach for making inferences about object recognition
- bayes theorem = allows calculation for prob. that the world is in a particular state given a particular observation
explain: changes in cells of what pathway (diff. between V2, V4)
- V2 = sensitive to contours and borders
- V4 = sensitive to more complex attributes
**cells along what pathway have a progressive change
⤷ anterior = more sensitive to complex stim.
explain: subtraction method
- compare activity measured in 2 conditions
⤷ one with and one without mental process of interest - look at diff. between images to see diff. in regions of brain
explain: fMRI experiment for object-decoding
- train participant by presenting many images
⤷ record brain responses - test by presenting new image
⤷ researches see brain activity and try to guess what the object is - shows that info about object identity is in that region
explain: computation model for visual processing
- pandemonium model
⤷ demons = neurons - sig. detected
- feature demons extract features
- cognitive demons look for specific features
- decision demons pool info and choose the loudest one to make the decision
explain: geon concept (limitations?)
- suggested geons = basic building blocks of objects
LIMITATIONS
- doesn’t explain how we can recognize specific individual objects we are familiar with
⤷ ex. if cup = cylinder + handle why can’t cup be a geon?
- some objects can be hard to break into geons (complex)
- doesn’t explain recog. after brain injuries
question: how can computers be taught to recognize things?
- DNN extract features
- info = pooled
- operations create new image for DNN to extract
- keep going to make more categories
explain: face inversion effect
- harder to recog. which faces have been modified if face is inverted
- harder to ID faces when inverted
question: do we use feature-by-feature analysis for faces?
- no
⤷ too slow - use holistic processing
⤷ process the face as a single thing
⤷ similar to gestalt
**holistic needs upright, normal contrast images
explain: prosopagnosia
- person cannot ID faces
- can be congenital or acquired
⤷ acquired = damage to temporal lobe (esp. FFA)
define: attention
- mechanism to restrict input to a subset of things
- dev. to deal with all the input
⤷ bc perceiving it all would be too much
**no single locus in the nervous system
⤷ has broad areas instead
name + explain: types of attention (5)
-
external or internal
- internal = thoughts
- external = ext. stim. (ex. music) -
overt or covert
- O = orienting senses to thing you’re focusing on
- C = secretly paying attention -
divided
- focus on more than 1 thing
- bad at this -
sustained
- long time focused on one thing -
selective
- paying att. to one among many stim.
explain: visual search experiments
- measure time it takes for observer to say yes or no
⤷ when asked if target is present
question: what impacts the efficiency of a visual search experiment?
- increasing set size = decreases efficiency
- changing task (easiest to hardest)
⤷ feature search > conjunction search > spatial configuration
define: salient feature
- stands out visually from neighbour
- ex. colour, orientation
explain: serial self-terminating search
- searching objects one by one
- terminate when find the target
- inefficient when target and distracts share a lot of features
⤷ less salient features
compare: guided search vs conjunction vs serial
GUIDED
- attention restricted to a subset of possible items on based on info about basic features
CONJUNCTION
- search for target defined by the presence of 2+ attributes
SERIAL
- checking each item until find target
guided faster than conjunction faster than serial
question: what are ways to improve efficiency of vis. search? (2)
- priming
⤷ stim. that makes it easier to resp. to a subsequent stim.
⤷ ex. finding one monkey makes finding the other monkeys easier - scene-based guidance
⤷ info of understanding scene helps find specific objects
⤷ ex. using anchoring objects
explain: binding problem in visual searches
- hard to tie diff. attributes of vis. stim.
⤷ handled by diff. circuits - possible explanation = feature integration theory (treisman)
⤷ basic features can be processed together but other properties need attention
⤷ ex. binding needs attention
define: illusory conjunction
- incorrect grouping of features from 2 or more objects
- both features present but not together
question: what happens to vis attention when lesions in parietal lobe? (2)
**parietal lobe = association cortex
- contralateral neglect
- inability to attend in contralesional visual field
⤷ ex. lesion in R -> ignore L field - extinction
- inability to exp. a stim. to one side of the point of fixation in the presence of another stim.
⤷ ex. lesion in R -> can’t attend to smth on the L if there’
- can’t attend contralaterally if there something in the ipsilateral field
question: what happens if both parietal cortexes are damaged?
- “blind”
⤷ can’t attend to anything
⤷ parietal = association cortex - can still see but can’t pay attention to anything -> can’t perceive anything
