Visual Perception Theories and Models

Attention, part 1

• January 23, 2025 presentation
• Select attention to speaker's face, not computer

Visual Perception

• Constructivist and direct models of visual perception
• Sensory information used to drive perception
• Direct perception of the world via sensory input
• Damage to visual system
  • Blindsight from primary visual cortical damage
  • Damage to dorsal pathway: problems with spatial perception and movement
  • Damage to ventral pathway: problems with object recognition
  • Prosopagnosia: inability to recognize faces
• Theories of object recognition

Theories of Visual Object Recognition

• Perception involves processing basic visual features and integrating into a probe pattern
• Combination of features depends on feature detection processes
• Pattern recognition compares the pattern to existing patterns in memory (like a barcode)
• Probe is compared to patterns in long term memory
• Highest similarity between probe and memory pattern determines recognition
• Template matching theory
• Prototype theory

Template (exemplar) Matching Theory

• Every object has a template in long-term memory
• Too simplistic and computationally demanding
• Cannot explain identification or classification with shifts in perspective (shifts in perspective/recognition/new objects being classified)

Prototype Theory

• Prototype is the average representation of an object concept
• Recognition determined by matching to a prototype, not a literal match
• A "good enough" match leads to identification
• Allows for more flexible object identification (visual probe matching to prototype image)

Context Matters for Object Recognition

• Scene consistency effect
• Accuracy (% correct) differing based on consistent/inconsistent scenes
• ERPs per condition (brain wave activity related to scene consistency/inconsistency)

What is Attention?

• William James (1890) definition
• Withdrawal from some things to effectively deal with others
• Opposite of confused, dazed, scatterbrained state (distraction)
• Understood in terms of what it does, not what it is

Agenda

• Types of attention
• Neural mechanisms of attention
• Top-down attention
• Selective attention models

Types of Attention

• Top-down attention: observer-guided, controlled.
• Bottom-up attention: stimulus-guided, automatic.
• Arousal
• Alertness and Awareness
• Autonomic Nervous System
• Reticular Activating System
• An inverted U-shaped function relates different levels of arousal and attention

Attentional Processes

• Top-down attention: Observer-guided, controlled, frontal-parietal brain regions (IPS; FEF)
• Bottom-up attention: Stimulus-guided, automatic, temporoparietal junction, and ventral frontal cortex
• Arousal, alertness, awareness, autonomic NS, reticular activating system

Neural mechanisms of attention

• Network of frontal and parietal lobes
• Intraparietal sulcus/Intraparietal lobule (IPS/IPL)
• Frontal eye fields (FEF)
• Temporoparietal junction (TPJ)
• Ventral frontal cortex (VFC)

Attentional shift

• Shift between attending to image and sound
• Brain regions during attentional shift (a, b)

Divisions of attention

• Endogenous attention: top-down processing, goal directed (Intraparietal sulcus/IPS and FEF)
• Exogenous attention: bottom-up processing, environmental stimuli (Temporo-parietal junction (TPJ) and VFC)

Spatial Neglect

• Damage to right hemisphere (ventral parietal cortex)
• Deficits in spatial attention, representation of the contralateral field (of view), and cannot attend to that information

Top-Down Attention

• Sustained attention: Maintaining focus for long periods; vigilance
• Divided attention: Shifting focus between tasks; multitasking
• Selective attention: Focusing on one input, ignoring others

Selective Attention

• Limited information processing resources
• Prioritization of information based on goals
• Early selection filter models
• Attenuator model
• Late selection filter models
• Load theory

Theories of Selective Attention

• Filtering of information during processing
• Early selection models (e.g., attenuator model)
• Late selection models
• Load theory

Broadbent's Early Selection Filter Model

• Filters information at a perceptual level, before meaning analysis
• Sensory buffer–filter–perceptual analysis–semantic analysis (STM)–responses

Broadbent's Early Selection Filter Model

• Information selection via perception (spatial location, frequency)
• Processed for meaning, entering awareness
• Information not selected by filter decays and isn't processed for meaning

Dichotic Listening Tasks

• Presenting two messages simultaneously to each ear
• Participants better recall messages ear by ear, not simultaneously
• Conclusion: Information selected for attention at the perception stage

Dichotic Listening: Shadowing Tasks

• People remember sensory features (noise or speaker gender) but not message content of unattended ear
• Evidence suggests unattended information not processed for meaning, but at perceptual level.

Evidence against Early Selection

• Unattended information can "break through" in certain situations
• Hearing your name in a different conversation
• Participants' heightened physiological responses (e.g., skin conductance) to words in an unattended ear when the words predicted a shock

Treisman's Attenuator Model

• An early filter that dials down unattended material, letting some aspects still be processed for meaning (but weaker).

Late Selection Filter Models

• We process information to the meaning level, then select information for further processing
• Filter is between perceptually analyzing and semantic (short term memory) analyzing.

Stroop Task

• Name the color of the ink
• Different colors presented in different colors
• Congruent trial (ink and color match) vs incongruent trial (ink and color don't match)

Controlled and Automatic Tasks

• Controlled tasks: require effort and voluntary attention (e.g., Stroop task: naming color of ink)
• Automatic tasks: highly familiar and well-practiced, don't require voluntary attention (e.g., Stroop: reading color names)*

Stroop Task and Late Selection Filter Model

• For interference to occur, written color names (unattended information) must be processed for meaning
• The filter in this case is between the perceptual and semantic aspects of information, then the response

Removing Automatic Processing

• Hypnotized participants to think color names were meaningless
• Removes automatic processing of word meaning (e.g., color)
• Result: No Stroop interference effect

The Load Theory

• Attentional filtering can occur at various stages of processing
• Placement of the filter depends on the demands of the current task
• Low-load tasks: non-attended info processed later (at semantic level)
• High-load tasks: non-attended info processed earlier (perception level)

The Load Theory (2)

• Difficult task (high load): processes all info percept only
• Easier task (low load): processes all info to meaning stage

When Am I Most Distracted by the Cog Dog?

• The load theory applies to this question: low load on easy watched movie vs. higher load on new or complex watched movie
• Jurassic Park (low load) vs. Harry Potter(high load)

Two Ways to Define Load

• Central resource capacity view: one resource pool
• Multiple resource capacity view: multiple resources, dependent on the matching between relevant and irrelevant information
• Load is reached sooner if relevant and irrelevant info is from the same modality

Central Resource Capacity

• Driving simulator task under two conditions: low (no radio) and high (radio)
• Higher load means the performance decreases because of more sources of interference/distractions.

According to Multiple Resource Capacity?

• Driving and following directions is more challenging/distracting.
• Multiple resource capacity: it is more challenging to pay attention if directions are present (on the radio) alongside concentrating on driver's task (car control and driving).