Visual and Cross-Modal Attention

• Attention is a key topic in visual and cross-modal contexts
• Feature Integration Theory is a model for understanding how we combine visual features
• Paradigms for studying attention include visual search, distractor interference, and illusory conjunctions
• Guided search helps explain how attention works in vision
• Early versus late selection in vision is another important area of study

The Binding Problem

• Features of stimuli are extracted from visual input
• Populations of neurons respond to color(V4), motion(V5), and orientation(V3)
• These individual attributes are experienced as coherent objects/environments
• A key question is how these disparate neural codes relate to meaningful object perception

Spotlight Analogy

• Anne Treisman suggests that a spotlight scans a master map of locations
• It selects features active in specialized feature maps
• It suppresses features to prevent erroneous binding

Feature Integration Theory (FIT)

• Spatial attention integrates visual features e.g. color and shape
• These features were coded separately in early vision

Predictions of FIT

• Features are perceived in parallel pre-attentively without limitations
• Attention does not enhance performance
• Features are extracted but not bound to a point unless under focus
• No location information is retained without attention
• Conjunctions of features require focal attention affecting performance
• Features can be erroneously combined when disrupted
• Illusory conjunctions can occur for unattended stimuli

Testing FIT

• Illusory conjunction paradigms, visual search paradigms and distractor interference paradigms can be used to test FIT

Illusory Conjunction Task

• Participants remember numbers at the beginning and end of a display
• They then report as much as possible about the letters and colors in the middle

Error Analysis

• Outcomes can be correct responses, feature errors or conjunction errors
• If conjunction errors are more frequent than feature errors, it supports illusory conjunctions

Illusory Conjunction Support

• Attention is consumed by the primary number task
• Unattended features lack direct attention
• Feature maps indicate "something green", "something X shaped"
• They will be incorrectly combined to produce illusory conjunctions

Theoretical Criticism

• It is unclear if combination failures are due to inattention, memory failure or other factors

Visual Search Task

• Visual search is a classic attention task
• The participant searches for a target and indicates presence or absence as fast as possible

Reaction Time Analysis

• Reaction times function of set size are used in analysis
• Steeper slopes indicate inefficient serial processing
• Shallower slopes indicate efficient parallel processing

Conjunction vs Feature Search

• Feature searches are 'pre-attentive - dont require spatial attention
• Conjunction searches require focal attention

Visual Search & FIT

• FIT supports the idea that visual search can be efficient
• Participants can Search for a unique color
• Or for unique orientation

Combining Features

• More complex, participants Search for a unique combination of orientation and color.
• This requires focussed attention

Feature Search Evidence

• ''Feature search'' is pre-attentive
• ''Conjunction search'' requires focal attention.

Parallel & Serial Slopes Explanation

• Serial slope indicates conjunction search
• Each item must be attended sequentially
• Parallel slope indicates feature search
• All stimuli are processed at once

Neural Imaging

• Attention to color/shape/speed activates V4, IT, and MT/V5
• There are two types of tasks to test this: Selective and Divided attention
• Selective attention tasks perform better than Divided attention tasks

Feature & Conjunction Task Analysis

• Attending to colour increases dorsolateral occipital cortex activity
• Attending to shape increases superior temporal sulcus activity
• Attending to speed increases intraparietal sulcus activity
• Greater activity when a feature is attended over divided attention
• Conjunction search modulates activity in the parietal lobe but Feature search does not

Motion Task

• Searching for a moving tilted line supports this idea

Guided Search

• Searches are more efficient when searching for a black circle
• Salient features form perceptual groups
• Attention is restricted to those groups, working in parallel
• It accounts for efficient conjunction target searches
• It uses pre-attentive information to limit items

Distractor Interference Task

• Participants respond to the central stimulus and press one key for target 1 and another for target 2
• Instructions state NOT to respond to stimuli or

FIT Further Support

• Slower times occur with incongruent vs. congruent conditions
• Distractor effect is the same, so features are not bound into objects
• Attention is necessary for binding features into coherent objects

Summary Of Conclusions

• Illusory conjunction studies show attention binds features
• Memory failures can complicate findings
• Conjunction search is serial, may be efficient with guided search
• Distractor interference and response competition support FIT
• Convergent evidence from different paradigms provides strong FIT support
• Key concepts include feature integration theory, visual search tasks, illusory conjunctions etc

Lecture Outline - Early vs Late Selection

• Early vs late selection studies vision
• Attention and consciousness
• Change blindness/inattentional blindness is also important

Early Selection Model

• Stimuli are filtered before semantic analysis
• Channel 1 is unattended in this model

Late Selection Model

• Stimuli reach semantic analysis before filtering
• Channel 1 is unattended in this model

Model Implications

• Both spotlight and FIT imply early selection/attenuation

Spotlight Effects

• There is no distractor interference outside the spotlight

FIT Effects

• Unattended perception has feature registration

Selective Reading Task (Neisser, 1969)

• Participants are asked to read only red words
• Very little information is retained from unattended green text

Video Task (Neisser & Becklen, 1975)

• Contents of an ignored video clip are not able to be reported

Recognition Task(Rock & Guttman, 1981)

• Participants make aesthetic red figure judgements
• Ignored green shapes can not be recognised

Change & Inattentional Blindness

• Inattentional Blindness: Failure to notice the presence of an unexpected item e.g. gorilla
• Change Blindness: Failure to notice a transition, like a color changing curtain

Flicker Paradigm Task Setup

• Two images alternate with a brief blank screen
• Participants identify and respond to observed changes

Flicker Paradigm Outcomes

• It is difficult to identify changes
• For marginal-interest items, it takes 17 alternations and 11 seconds

Visual Transients

• Fast changes in retinal image luminance or color, or object appearance, disappearance, or motion, attract attention

Working Memory Load

• People must encode one image to working memory to compare, this is often impaired in CB paradigms

Late Selection Support

• Prior experiments support early selection from direct measures
• Late selection support uses indirect measures

Stroop Task

• Participants ignore word meaning and name font color
• Ignored words are perceived, supporting Late Selection

Stroop Variations

• Stroop's weak support is that the dimension is part of object
• Separating word & color supports Late selection (Gatti & Egeth, 1978)

Response Competition Task

• Participants indicate the central letter (H or X)
• Eriksen's paradigm has separate distractors from target
• Congruent trials are faster and Incongruent are slower
• Response competition effects show distractors are perceived which supports Late Selection

Late Selection

• Conclusion based on indirect interference measures

Early Selection

• Conclusion based on retrospective direct measures
• Rapid forgetting may be a factor

Selective Attention Flanker Task

• H or X in the center of screen
• Unattended distractors do not produce response competition effects
• Response competition paradigms here are in favor of early selection

Alternative Theory

• Inhibition theory of selective attention suggests ignored distractors are suppressed
• This is in agreement with late selection evidence from negative priming paradigms

Two Theories

• "Distractor interference demonstrates distractor perception late selection
• BUT no distractor interference does not necessarily mean no distractor perception i.e. Early selection

Negative Priming Paradigm Task

• The image has a solid line - is this image an animal, furniture or instrument?
• NP occurs with SupImposed letters/shapes i.e. evidence for late selection is weak

Negative Priming Example

• The Flanker task provides NP evidence
• These responses show they are slower which supports late selection

Early Selection Summary

• Early Selection may result from failures of memory
• It's unlikely to forget the gorilla in Simons & Chabris' (1999)

Alternative Online Perception Model

• Also supports early selection and tests online perception (no memory)

Attentional Blink Task

• Present one key when you see a white letter when you see an X
• Press button when see an X
• Vary the time between the two

Attentional Blink Outcome

• Attentional blink is failure to detect the probe due to target attention
• Attention is needed + takes time to process another piece of information

Perceptual Load Theory

• Nilli Lavie 1995, early and late selection resolved by that

Load Examples

• High load = early selection, low selection = late selection results

Perceptual Load Theory Explained

• Perception has limited capacity, as with early selection models
• Perception proceeds automatically within its capacity, as with late selection models

Visual Search Example

• Processing may depend on the task load
• Under low load, you have spare capacity, so non-selective perception occurs which is late selection
• But under high load,full capacity is High-load, results in selective perception which is early section

Perceptual Load - Response example

• You search for X or Z of a certain position but you ignore the other peripheries
• Distractor interference is higher in low selections

Brain Load Example

• High task load can reduce brain response to unattended stimuli
• Task: Detect the case of a word low or detect number of syllables which will be high
• Background dots can be moving or not, these are never relevant

V5-MT Motion Response

• Perceptual load modulates the response to motion
• Motion occurs in the low-load condition