lecture 7 - attention as glue Flashcards
attention as glue
TREISMAN’ S “GLUE”
(FEATURE INTEGRATION
THEORY)
-what did she test and how did she create the theory
As a PhD student, tested Broadbent’s model, looking at attention and speech perception. Her “attenuation model” proposed that unattended information was
weaker but processed.
- Went on to develop Feature Integration Theory, supported by many innovative behavioural experiments
two stages of treisman’s ‘glue’
feature integration theory
Step 1: Specialized and separate visual brain areas process basic stimulus features (e.g. motion, color, orientation, brightness)
- Step 2: Attention directed to regions of the visual field “glues” features together into identifiable
objects that now have continuity over space and time.
-attentional spotlight selects a region of space, and the features in that area are bound together into a unified object
free floating features, and then when attention focuses on a region, the features in that area are bound together into an object
what is the binding problem
-Objects have different “features” (size, color, shape, motion, sound, smell…)
- Features are processed and
represented by different specialized
networks in the brain. - How do you know which features
belong to which object? This is the
binding problem
considers that features of an object need to be bound together by some neuronal mechanism across a population of neurons, so that the object can be perceived as a whole.
attention solves the binding problem
-this theory suggests that everything before you have attention is unbound, but when you have attention things are bound together
FIT: features are represented on
separate maps. Attention to a spatial location binds all the features in that location into a single object.
evidence that attention solves binding problem : illusory conjunctions
-the theory suggests in the absence of attention , you have a representation of features but not conjunctions, so predicts when things aren’t attending there can be mis binding of features
- Failure to correctly bind features can lead to “illusory conjunctions”
evidence that attention solves binding problem : illusory conjunctions
-how is it carried out
-what questions asked
- experiment done by Diverting attention away from the shapes by asking participants to report the digits (in a picture of shapes and digits)
then asked
What were the two digits presented?
What shapes do you remember seeing:
* Was there a blue triangle? (yes)
* Was there a yellow circle? (nothing yellow, so should not hallucinate yellow)
* Was there an orange square? (orange circle and green square) illusuro conjuction condition
* Was there a green diamond? (green square)
-questions are specific to fit exp
result of illusory conjunction experiment
- Result: illusory conjunctions experienced and reported
-people mis bind the colours to the shapes (since they pay attention to the numbers instead)
(Treisman & Schmidt, 1982).
EVIDENCE FOR FIT II:
-OBJECT SEGREGATION
Certain properties are processed simultaneously and automatically across the visual field. This “Feature Extraction” stage defines object boundaries
If an object can be defined based on a feature, it will pop out of the
background
An object defined by a combination of multiple features does not pop
out automatically. It requires serial, effortful ATTENTION to be found.
(eg you can find lines going left direction in lines going right) (but seeing an arrow up the way in many down the way is hard)
look at slide 8
EVIDENCE FOR FIT III:
SERIAL VS. POP-OUT SEARCH
-serial search
-search slope, slope ratio
visual search task
-have a target (your searching for a red vertical bar)
-it is present in one photo and not present in the other
-your task is to search the display and say whether the target is present or absent
-in this case the target your looking for is defined based on a conjuction of two basic features ( have to look for an orientation and colour)
-since this requires attention to bind the tow features, triesman argues you are serially placing your attention on each of the items in the display to decide if its a target or not
Serial search: Attention needs to serially inspect each item to determine if it is the target.
Search Slope: The increase in reaction time associated with each additional distractor added to the display
should have 2,1 slope ratio: target present reaction time is always about half of your target absent reaction time
EVIDENCE FOR FIT III:
SERIAL VS. POP-OUT SEARCH
-pop out search
Pop-out search (AKA “feature search”): Target can be defined based on a single, unique feature. These can be detected without need for attention.
Pop-out search is defined as that which has a search slope that is close to 0
what are pre attentive features
features that guide attention don’t need attention to be detected
you cant have object identity in the absence of ______
attention
attention and the conjunction of features
-why don’t we mis bind more often if everything is unbound
-theory proposes we have unbound features for most of the objects, so why dont we mis bind more often?
well: Constrained by prior knowledge (e.g. no blue bananas
or fuzzy bicycles, Treisman and Schmidt, 1982)
-Attention “opens a file” for a particular object (Kahneman, Treisman and Gibbs, 1992), like creating a file for storing related documents on your desktop
- Without attention, collections of features are interchangeable (resulting in inattentional and change blindness, Simons & Chabris, 1999; Rensink, 1992)
what exactly is a feature
Definitely
Colour
Motion
Orientation
Size
probably
depth
luminance
closure
curvature
maybe
lighting
glossiness
number
aspect ratio
doubtful
novelty
letter identity
non feature
conjuctions
category
identity
3D shape (objects that require attention to extract- broadbents theory)
what is the guided search model
Initially proposed in 1989, now on version 6.0 (Wolfe, 2021)
-proposes that search can be serial and parallel (pop out)
- Targets have “guiding features” that can be used to limit the search set, rapidly and in parallel
- Once limited to a subset, attention is guided to each candidate until the target is found
- The model includes a lot of other aspects of visual search (visual acuity, decision-making, memory)
- It’s a model of search, not a model of attention, but has largely subsumed FIT.
