Visual Perception & Object Recognition Theories

Visual Perception and Damage

• Visual perception is based on constructivist and direct models.
• Information beyond sensory input drives what we perceive.
• The world can be directly perceived via sensory information.
• Primary visual cortical damage can cause blindsight.
• Damage to the dorsal pathway causes spatial perception and movement issues.
• Damage to the ventral pathway causes issues with object recognition.
• Prosopagnosia can also arise following visual insult.

Theories of Visual Object Recognition

• Perception processes basic visual features of an input, integrating them into a probe pattern.
• Perception depends on feature detection processes.
• Pattern recognition compares a pattern to those stored in memory and is like recognizing a barcode.
• A probe matches to long-term memory patterns.
• The highest similarity between the probe and a memory pattern triggers recognition.
• Template matching theory and prototype theory describe what a probe is being compared to in long-term memory.

Template Matching Theory

• Every object gets matched to a 'template' in long-term memory.
• It is too simplistic and computationally demanding for human perception.
• It cannot explain identification, which is recognizing objects despite shifts in perspective.
• It also does not address classification, when new objects gets recognized as members of a known category.

Prototype Theory

• Prototypes are an average representation of an object concept. Recognition is determined not by literal match but a 'good enough' match (resemblance) to a prototype.
• Allows for ‘flexible' object identification.

Context Matters

• Scene consistency affects object recognition.

Attention

• Defined by William James (1890), attention involves withdrawing from some things to deal effectively with others, with distraction as its opposite state.
• Attention is best understood in terms of what it does, rather than what it is.

Types of Attention

• Top-down attention is observer-guided and controlled.
• It involves frontal-parietal brain regions like the IPS and FEF.
• Bottom-up attention is stimuli-guided and automatic, involving areas like the TPJ and VFC.
• Arousal is linked to alertness and awareness via the Autonomic Nervous System and Reticular Activating System.

Attentional Processes

• Bottom-up attention is determined by physical stimuli and salience.
• Top-down attention is driven by a goal or target in mind.

Neural Mechanisms of Attention

• Involves networks across frontal and parietal lobes.
• The frontal eye fields (FEF) are active when engaging in top-down tasks.
• The intraparietal sulcus/intraparietal lobule (Ips/IPL) controls and prepares attention, setting goals.
• The temporoparietal junction (TPJ) drives automatic attentional orienting.
• The ventral frontal cortex (VFC) is also involved in automatic attentional orienting.
• Attentional shift is a task between attending to an image vs sound.

Divisions of Attention

• Endogenous attention is when an individual chooses what to pay attention to (goals and intention). It involves top-down processing, intraparietal sulcus (IPS), and FEF.
• Exogenous attention is when stimuli in the environment drive attention. It involves bottom-up processing, temporo-parietal junction (TPJ) and VFC.

Spatial Neglect

• If there is right hemisphere (ventral parietal cortex) damage there can be spatial neglect.
• This causes deficits in spatial attention and egocentric representations in contralateral field of view.
• Individuals cannot attend to information present to the opposite side of lesion.
• Spatial neglect results in unawareness of the left side of the world.
• Symptoms include reading only words on the right side, eating from one side of the plate, and describing only half of imaginations and memories.
• Deficit presents across sensory modalities (not just vision).
• If blindfolded, individuals have problems searching for objects on one side of the table.
• Severity is modulated by behavioral interventions.
• E.g. Training to increase alertness, asking to perform hand movements to neglected side.

Top-Down Attention

• Sustained attention: Maintaining focus on one input for a long period, also known as vigilance.
• Divided attention: Shifting attentional focus between tasks, multitasking.
• Selective attention: Focusing on one input while ignoring others.

Selective Attention

• Selective attention occurs due to limited information processing resources need to be prioritized.
• Prioritization depends on the goal and what you want to attend to.
• Theories on selective attention include early selection filter models, the attenuator model, late selection filter models, and load theory.

Theories of Selective Attention

• Information filters out when processing information.
• Broadbent describes Early selection models.
• Treisman describes the Attenuator model.
• Deutsch & Deutsch and Norman offer Late selection models.
• Lavie has introduced Load theory.

Broadbent's Early Selection Filter Model

• Information is filtered at the level of perception, before it is processed for meaning (semantic analysis).
• Information is selected via perception (spatial location, frequency of sound).
• Selected information processes for meaning and enters awareness.
• Information not selected by the filter decays and is not processed for meaning.
• Dichotic listening tasks support early selection model.
• Participants better recall ear by ear than simultaneously presented messages.
• Information is selected for attention at perception.
• In dichotic listening and shadowing tasks,people do not remember the content of an unattended message, but they notice some sensory features (new noise; gender of the speaker).
• This suggests that unattended information is not processed for meaning, but is processed at the level of perception.
• Early selection model can be challenged when un-attended information "breaks through".
• This occurs if you hear your name if spoken in a non-attended-to conversation.
• The early selection model is also challenged by shock presented with a word in pre-shadowing task:
• Participants presented with a word (e.g., apple) paired with a shock.
• Shadowing task with the 'shocked' word in the unattended ear
• Increased skin conductance when the 'shocked' word was presented in the unattended ear and detected.

Treisman's Attenuator Model

• An early filter dials down the influence of unattended material, and aspects of unattended stimuli are processed for meaning.

Late Selection Filter Models

• Input is processed to the level of meaning before deciding what to process further.
• The Stroop task highlights late selection models.
• Tasks with controlled effort require voluntary top-down attention like naming the color of the 'ink' in the Stroop task.
• Automatic tasks are those that are highly familiar and well-practiced and do not require voluntary top-down attention, like reading color names in the Stroop task.
• The interference effect results in a slower naming time.
• The congruent trial ink and color match with automatic activation of the visual cortex.
• The incongruent trial ink and color do not match, and naming is slower.
• To be able to do the naming, the subject needs to access meaning of color.
• Interference on the Stroop task means the written color name (unattended information) has to be processed for meaning.
• Removing the automatic processing of color leads to no Stroop interference effect.

Load Theory

• Attentional filtering (selection) can occur at different points of processing.
• Filter placement depends on how much of your resources are required for your current task.
  • If low resource load, non-attended information goes to a later stage.
  • If high resource load, non-attended information goes to an early stage.
• A difficult task with a high load - We only process all information (relevant and irrelevant) at the the level of perception.
• When our attention is selected: early selection takes place and there is focused attention.
• An easier task with a low load - We process all information (relevant and irrelevant) to the level of meaning.
• Our attention is selected later, and process irrelevant information is given meaning.

Defining Load

• Central resource capacity view: there is one resource pool for all attention resources.
• Multiple resource capacity view: the attentional load depends on the match between relevant and irrelevant information.
• For example, attentional capacity is reached sooner if relevant and irrelevant information are presented in the same modality.