What is the theory behind the idea of attention playing a role in 'binding' visual features into object representations?
- Feature Integration Theory (FIT)
- Neurophysiology suggest we process different features initially in specialised sub-areas of visual cortex.
- FIT arguesattention provides the 'glue' which combines separate features into an integrated object perception.
- Stage model of visual info processing
- Initial stage - single features are analysed and coded separately within specialised feature modules
- Only the presence (and not location) is signalled to subsequent stages
- Selective attention operates like a spotlight moving to specific locations, and when attention applied to a location, different features from different modules are combined and entered into object file --> recognition network.
- FIT assumes spotlight applied serially to different locations.
How does the Feature Integration Theory explain the different RT of Feature search (single feature) and conjunction search?
- Single feature search is fast and parallel because it relies only on the first stage where parallel coding of single fatures in feature modules occurs - without processing of location.
- Conjunction search requires features to be combined in order to detect target, requiring serial movement of attentional spotlight in order to integrate different feature modules.
Who should you mention when talking about differences between parallel feature search and conjunctive search?
- Treisman & Gelade (1980)
- RT unaffected by display size in feature search
- RT affected by display size in conjunctive search
- (Also, target absent searches are 2x as steep because all items need to be scanned)
[What does attention select?] Historically, which was studied first? Space or object attention?
Space attention (spotlight and zoom) before object attention
List the 4 studies you can use to attention can select object instead of space?
- Egly, Driver & Rafal (1994): rectangular object - same object vs equal distance
- Driver & Halligan (1991): Unilateral neglect, tilted object
- Mattingley, Davis & Driver (1997): Unilateral extinction (bar behind box)
- Pylyshyn & Storm (1988): Multiple object tracking
Describe Egly, Driver & Rafal (1994) study.
- Endogenous cueing paradigm
- Cue appeared on one end of rectangular object
- Target appear either on other end of same object, or same end of a different object next to the cued object
- Distance between cue and target on both these conditions equal
- RT faster when target was on other side of same object
- Suggests attention selects whole objects -> cueing subject to one part of object makes it easier to detect stimuli on same object.
- Spotlight model will predict RT will be the same in both conditions.
Describe Driver & Halligan (1991) study.
- Left neglect patients
- Nonsense objects presented and patients were asked to detect difference between two shapes
- Two conditions - object upright or object tilted 45 degrees clockwise.
- Irrespective of whether it was titled or not, patients couldn't detect differences on the left side of object - even though when tilted, some of the left side of object fell on the right of the saggital midline (anatomical line running down centre, seprating right and left of visual field) --> right side of visual field/space.
- Left side of object neglected, not left side of visual field --> suggests attention selects objects
Describe Mattingley, Davis & Driver's (1997) study.
- Patient V.R with left-side extinction
- Box presented with two bars
- Can detect is bar presented alone on left side, but missed it when presented simultaneously with right-hand bar.
- BUT: when the bars from both sides are made to look like it was occluded by the box (made to look like one long bar behind box), she was able to detect both sides.
- Suggests attention selects perceived objects, rather than specific location in visual field.
Describe Pylyshyn & Storm's (1988) study.
- Subjects much better able to track multiple moving objects if objects which periodically disappeared momentarily looked like they disappeared because they went behind an object, rather than simply disappearing.
- Even though visual info of target objects were same in both situations, this difference highlights mehcanism whereby attention seems to be tracking and selecting objects.
How would you consolidate Eriksen & Eriksen's study about zoom lens with object-selecting attention studies?
Reason that interfering effect of task-irrelevant flankers appeared after flankers came close could have been because they became so close to central object that it was perceived to be the same object.
Alongside the multiple object tracking task, what other study shows that we can select 4 objects simultaneously by attention?
- Schiffrin & Gardner (1972)
- Four objects on each corner --> asked to decide if F or T was in display
- 2 conditions:
- Successive condition --> Four objects shown in separate flashes so that F or T is only shown with one other distractor
- Simulateneous conditon --> all four objects shown at once, with one of them being F or T
- Performance same in both conditions
- Four objects can be selected --> initial stages of visual processing, up to level of letter recognition, take place without capacity limitation attentional control (CHECK)
Just sum up what I said in the conclusion of 'what attention selects' essay because Kate seemed to like it.
- Need to consider both object and space selecting attention mechanisms as complementary, rather than antagonistic models
- More could be done to see whether they are distinct ssytems that attention switches between depending on circumstances
- Could do variation on Egly, Driver & Rafal's study whereby distance between cue and target in same object will be greater than between same side of two objects. Could determine whether it selects better fro same-object stimuli or smaller-distance stimuli.
- Finally, we must remember that models are all metaphors which are used to describe more intricate physiological/neurological processes that are happening in visual system, so more neurophysiological evidence might be useful.
[When does attention select?] Sum up the early selection theories and the evidence for them?
- Broadbent's Filter Model
- Treisman & Gelade's Filter Integration Theory
- Visual search paradigm: attention can select very quickly for simple, single-feature objects. (Same with ear shadowing paradigm, voice-pitch)
- However, high-level features do not pop out, which suggested that attention does not select on high-level features. High-level features processed after selection by attention.
- Single-cell recording of primate visual cortex - differences between early striate cortex processing (V1) and higher-level extrastriate (V5)
- This WITH Rees (1997): crucially, V1 showed activation but V5 showed no activation for unattended motion, suggesting that early rudimental processing in V1 happens, but attention selects before V5 processing.
- To avoid processing overload
However, what is the visual search evidence that seems to contradict the early-selection theories?
- Some conjunction features can pop out
- eg. Macleod, Dienes, Driver (1991) --> both motion and form
- Seems to suggest conjunction of features are processed prior to selection by attention
Also, what other evidence seems to weaken the early selection theories?
- Attention selecting objects (Driver & Halligan, Egly & colleagues etc) imply that if attention is selecting betwene object representations then some processing of objects must have arisen prior to selection by attention.
- Objects are often not just defined by one simple sensory feature, but conjunction, suggesting it surely requires some high-level processing.
Can you outline the basic differences in reason for selective attention in the early and late theories?
- Early: selection for avoiding processing overload
- Late: selection for action - selection arises only ensure correct object controls action
Kahneman and Treisman argued that there was a pradigm shift in methodology from when late selection theories became more popular. What was this (briefly)?
- That early studies on early selection tended to overload subjects with many irrelevant stimuli
- Whereas later studies (such as Schiffrin's) tended to give more simple stimuli with not many distractors
- Kahneman and Treisman argue these methodological shifts meant that it was difficlt to generalise from all these studies about the location of selection.
- Lavie argues that in this way, perceptual load determines the locus of selection
Describe Duncan's study and why he is for the late selection theory.
- Similar design to Schiffrin & Gardner's experiment
- Performance is not affected by number of non-targets, but is when simultaneous targets must be detected separately, performance suffers
- When 1 targt must be detected amongst 4 letters --> no difference between successive (2 letters at a time) and simultaneous.
- However, when asked if there were 1 or 2 targets present when 2 targets shown, much harder simultaneous.
- CONCLUSION:Multiple objects can be processed to high levels such that single target can be detected amongst them on basis of high-level properties (eg. being a digit)
- However, only one object can be selected at a time to control our resposne.
What is the modern consensus view these days?
- Still a lot of debate but...
- Most argue that attention may select at more than one stage of processing
- Lavie: perceptual load determines locus of selection --> situation can change the location of selection
- Likely that up to 4 objects can be selected by attention for high-level perceptual processing
- But, only one of these is then selected to control our responses at any given moment.