PSY1004: Cognitive Psychology - Lecture 4: Attention
Fundamentals and Historical Origins of Attention
Attention is the foundational cognitive faculty that allows an organism to focus on specific stimuli, environmental locations, or internal mental representations while managing incoming perceptual inputs.
Core taxonomy of attention:
Selective Attention: The cognitive process of focusing on a single, specific message, object, or location while systematically ignoring or filtering out competing environmental stimuli.
Divided Attention: The distribution of mental resources to attend to two or more distinct tasks or sensory streams concurrently.
Historical emergence in experimental psychology:
Academic focus on attention expanded rapidly during the 1950s.
This research surge was catalyzed by World War II and the rapid postwar development of complex human-machine interfaces, radar consoles, and aviation systems.
Engineers and psychologists were confronted with the operational challenge of optimizing communication protocols and cockpit instrument layouts for pilots who suffered cognitive overload when processing simultaneous, dense streams of auditory and visual data.
Auditory Selective Attention and Shadowing Paradigms
The human sensory environment presents far more information than the nervous system can consciously represent. The cognitive apparatus filters out the vast majority of environmental signals while selecting a subset for further structural, semantic, and mnemonic processing.
The Dichotic Listening Paradigm (Cherry, 1953):
Participants wear stereophonic headphones delivering two separate, simultaneous audio streams: one to the left ear and another to the right ear.
Participants perform a shadowing task, meaning they repeat the speech input from the designated "attended" ear aloud in real time as they hear it.
Shadowing ensures that the participant actively maintains continuous, focused selective attention on the target channel.
Empirical findings from dichotic listening:
Participants are unable to report the semantic content, language, or specific words delivered to the unattended ear.
Unattended auditory signals are not entirely eliminated at the peripheral sensory level. Participants reliably detect:
The acoustic presence or absence of a message.
Changes in the physical gender of the speaker's voice.
Transitions from spoken speech to a pure acoustic tone.
Cocktail Party Effect: The ability to selectively focus on a single conversational partner within a noisy, multi-speaker social setting while perceptual systems monitor background acoustics at an implicit level, allowing significant stimuli (such as the mention of one's own name) to capture conscious awareness.
Models of Selective Attention: Early, Intermediate, and Late Selection
The theoretical debate regarding selective attention centers on the locus of the attentional filter within the information-processing hierarchy: does filtering occur early (prior to semantic analysis) or late (subsequent to meaning extraction)?
Broadbent's Filter Model (Early Selection / Bottleneck Model, 1958):
Information passes through four sequential stages:
Sensory Memory: Holds all incoming raw sensory signals for a fraction of a second, preserving physical characteristics before transferring the complete stream forward.
Filter: Identifies the target message based exclusively on low-level physical attributes (e.g., voice pitch, speaker location, vocal cadence). Only the attended signal passes through; all unattended messages are blocked.
Detector: Processes the selected signal to determine semantic characteristics, meaning, and higher-order syntactic properties.
Short-Term Memory (STM): Receives the processed output of the detector, holding the information for and potentially transferring it into long-term memory storage.
Limitations of Broadbent's early selection model:
Fails to explain the Cocktail Party Phenomenon, wherein an unattended, highly salient stimulus (such as one's own name) readily passes through the cognitive filter into conscious awareness.
Fails to account for cross-ear semantic integration, demonstrated by the Dear Aunt Jane experiment (Gray & Wedderburn, 1960):
Attended ear received: "Dear – 7 – Jane"
Unattended ear received: "9 – Aunt – 6"
Rather than reporting "Dear 7 Jane", participants shadowed "Dear Aunt Jane".
This proves that participants automatically switched attention between physical ears based on linguistic meaning, demonstrating that semantic analysis occurred prior to the final attentional selection.
Treisman's Attenuation Model (Intermediate Selection / "Leaky Filter" Model):
Proposes that selection occurs in two stages, replacing Broadbent's rigid filter with an Attenuator and adding a Dictionary Unit.
The Attenuator:
Analyzes incoming messages across three distinct dimensions: physical characteristics, language structure (syllabification, grammar), and semantic meaning.
The attended stream passes through at full signal strength.
The unattended stream is not deleted; it is attenuated (dampened or weakened in signal intensity).
The Dictionary Unit:
Contains mental representations of words, each possessing an activation threshold required for conscious detection.
Words of high subjective priority or high environmental frequency (e.g., one's own name, words like danger or fire) maintain permanently low activation thresholds, firing even when triggered by attenuated, weak signals.
Common words (e.g., boat) have intermediate activation thresholds.
Rare, uncommon words (e.g., rutabaga) possess elevated activation thresholds and require an unattenuated, full-strength signal to reach awareness.
Late Selection Models (e.g., MacKay, 1973):
Propose that all incoming messages are processed to the level of meaning before perceptual selection occurs.
In MacKay's experiment, participants shadowed ambiguous sentences presented to the attended ear, such as: "They were throwing stones at the bank."
Simultaneously, the unattended ear received a biasing word: either "river" or "money".
When tested on sentence comprehension, participants chose which sentence matched the intended meaning:
Option 1: "They threw stones toward the side of the river yesterday."
Option 2: "They threw stones at the savings and loan association yesterday."
The presentation of the biasing word significantly steered participant interpretations toward either the river or the financial institution, despite participants remaining consciously unaware of the unattended biasing words.
Perceptual Load, Cognitive Capacity, and the Stroop Effect
Lavie's Load Theory of Attention establishes the parameters governing susceptibility to distraction based on two principles:
Processing Capacity: The finite volume of cognitive resources an individual possesses to process sensory and intellectual information at any given moment.
Perceptual Load: The intrinsic difficulty and cognitive demand imposed by a specific task.
Low-Load Tasks: Easy, well-practiced tasks that consume a small fraction of overall processing capacity. The remaining cognitive capacity processes task-irrelevant, unattended environmental distractors.
High-Load Tasks: Difficult, resource-intensive tasks that consume all available cognitive capacity, leaving zero leftover capacity to process peripheral, task-irrelevant stimuli.
Behavioral manifestation:
Operating a motor vehicle along an empty highway represents a low-load task, permitting the cognitive processing of high-volume music or complex audiobooks.
Parallel parking into a tight space amidst traffic elevates perceptual load, prompting drivers to instinctively pause conversation or eliminate auditory inputs to dedicate all capacity to spatial maneuvering.
The Stroop Effect (J. Ridley Stroop, 1935):
Demonstrates the cognitive interference generated when automatic processing conflicts with controlled task demands.
Participants are presented with color words written in incongruent ink colors (e.g., the word "RED" printed in blue ink) and instructed to name the ink color while ignoring the word text.
Reading is a highly overlearned, automatized process executed without conscious effort or volitional control.
Because the visual system cannot inhibit reading the word, the semantic processing of the word name competes directly with the ink-naming task, producing severe response latency and error rates.
Training and Cognitive Automatization:
Extended practice transitions tasks from controlled processing (effortful, slow, capacity-draining) to automatic processing (fast, effortless, minimal capacity use).
Increased automaticity reduces reliance on frontal executive control networks.
Empirical examples:
Habitual video game players exhibit enhanced visual attentional capacity and reduced performance degradation under elevated visual loads.
Polyglots and experienced language learners demonstrate automatic lexical access, reducing the cognitive load of foreign language production.
Overt Attention, Fixations, and Visual Eye Tracking
Overt Attention: The shifting of mental attention accompanied by observable, physical movements of the eyes.
Fixations: Brief visual pauses on specific environmental points of interest during which central, foveal vision extracts high-resolution spatial and color information.
Saccades: Rapid, ballistic, jerky movements of the eyes connecting consecutive fixations, during which visual input is suppressed.
Measurement: Precision infrared eye-tracking apparatus measuring the relative positioning of the pupil center against the corneal reflection of infrared light.
Determinants of visual scanning:
Stimulus Salience (Bottom-Up Control):
Involuntary capture of attention driven by physical stimulus characteristics, including bright colors, sharp contrast boundaries, unique orientations, and physical motion.
Computational models synthesize these physical properties into a Saliency Map, accurately predicting initial saccadic targeting prior to cognitive evaluation.
Scene Schemas (Top-Down Control):
Cognitive representations and semantic expectations regarding the typical composition of visual environments (e.g., expecting a refrigerator, stove, and sink inside a kitchen).
Fixations are guided by internal goals and schemata (e.g., looking at a kitchen counter when preparing breakfast versus looking near door hooks when searching for car keys).
Observers fixate significantly longer on schema-incongruent objects placed inside a visual scene (e.g., a printer placed on a kitchen stove).
Task-Driven Motor Coupling:
Eye movements in natural behaviors are closely tied to motor execution.
Fixations precede corresponding manual motor actions (such as reaching, grasping, or cutting) by a fraction of a second, demonstrating "just-in-time" perceptual sampling.
Covert Attention and Spatial Cueing Paradigms
Covert Attention: The allocation of attentional processing to a specific spatial location or stimulus in the absence of physical eye movements ("attending out of the corner of the eye").
The Posner Spatial Precueing Paradigm (Posner, 1980):
Investigates how shifting attention across visual coordinates alters perceptual processing speed using Donders' subtraction logic.
Participants maintain central fixation on a display featuring peripheral target boxes.
A precue directs attention toward one of the locations:
Valid Trials: The cue accurately signals the exact location where the target stimulus will subsequently appear.
Invalid Trials: The cue directs attention toward the opposite side of the display, away from where the target stimulus appears.
Results demonstrate that reaction times are significantly faster on validly cued trials compared to invalidly cued trials, even when the eyes remain immobilized on the central fixation point.
Cues operate through two distinct mechanisms:
Endogenous Cues: Central, symbolic cues (e.g., an arrow) requiring top-down cognitive interpretation to voluntarily orient spatial attention.
Exogenous Cues: Peripheral, physical events (e.g., a sudden brightening of a target box) that reflexively capture attention via bottom-up mechanisms.
Practical application in prestidigitation and magic:
Magicians exploit the dissociation between overt gaze and covert attention.
Practitioners deploy misdirection using exogenous cues (e.g., sudden flashes, expressive gestures) and social cues (e.g., direct eye contact) to draw audience attention away from the mechanical execution of sleights.
Divided Attention, Dual-Task Performance, and Distracted Driving
The feasibility of divided attention depends on task difficulty, structural overlap between stimulus streams, and the degree of task automatization.
Experimental Investigation of Divided Attention (Schneider & Shiffrin, 1977):
Participants engaged in a dual task requiring simultaneous retention of memory targets and visual monitoring of rapid stimulus frames.
Structure:
Memory Set: One to four target characters (e.g., the digit 3) held in working memory.
Test Frames: An array of presented at rapid speeds, containing distractors (letters), random patterns, or the target digit.
Findings and the development of automaticity:
Early in training, participants exhibited low baseline accuracy (\text{--}) and slow reaction times, relying on conscious, controlled serial scanning.
After approximately , performance improved systematically to over \ accuracy.
At this threshold, participants reported that target detection occurred automatically, without conscious effort or awareness.
Automatic Processing: Processing that occurs involuntarily, without conscious intent, while consuming minimal cognitive processing resources.
Divided attention collapses when both tasks are difficult, unstructured, or draw upon identical cognitive modalities.
Real-World Cognitive Distraction: Distracted Driving (Strayer & Johnston, 2001):
While driving along familiar routes under ideal environmental conditions can feel automatic, it remains vulnerable to attentional competition.
Experimental evaluation of cell phone usage during driving:
Drivers conversing on cell phones missed twice as many red traffic lights as undistracted drivers.
Response latencies to brake application increased significantly.
Hands-free phone units produced identical decrements in driving performance compared to handheld mobile devices.
This demonstrates that phone-induced driving impairment stems from central cognitive distraction rather than peripheral biomechanical demands (holding the handset).
Attentional Failures: Inattentional Blindness and Change Blindness
Inattentional Blindness:
The failure to consciously perceive an unexpected, clearly visible stimulus within the visual field when attention is actively directed to another task or object array.
Mack & Rock paradigm:
Participants perform a visual discrimination task, judging which arm of a briefly presented cross is longer.
On critical trials (e.g., Trial 6), an unexpected geometric figure (such as a small square) is presented adjacent to the central cross.
The vast majority of participants fail to detect the geometric figure despite its clear visibility and proximity to fixation.
Sustained inattentional blindness (Simons & Chabris, 1999):
Observers watch a video of two teams passing basketballs (one team wearing white shirts, the other black shirts).
Observers are instructed to silently count the passes made by the white team (a high-load primary task).
A person dressed in a gorilla suit walks across the center of the court, pauses, thumps their chest, and exits over a 9-second span.
Roughly half of observers fail to notice the gorilla, demonstrating that conscious visual perception requires attentional resource allocation rather than ocular exposure.
Change Blindness:
The difficulty or inability to detect structural changes between two alternating, slightly different visual scenes when separated by a brief visual blank, flicker, or saccadic interruption.
Changes that are obvious once pointed out remain undetected across multiple viewing cycles.
Demonstrates that the cognitive system does not maintain a comprehensive, photorealistic internal representation of the environment; rather, it retains sparse representations dependent on focal attention.
Mechanistic tie to Perceptual Load:
Low-load visual tasks leave residual capacity that spills over to process task-irrelevant environmental anomalies.
High-load tasks consume total perceptual capacity, producing functional blindness to unattended stimuli.
Feature Integration Theory and the Binding Problem
The Binding Problem:
Visual neuroscience demonstrates that separate, specialized cortical structures analyze distinct physical dimensions of an object in parallel (e.g., visual area V4 processes color, while area MT/V5 processes motion and spatial orientation).
The binding problem asks how the brain recombines these disparate, anatomically segregated physical features into a unified, coherent perception of an object.
Feature Integration Theory (FIT - Anne Treisman):
Resolves the binding problem via a two-stage information-processing model.
Stage 1: Preattentive Stage:
Occurs early in the processing stream, prior to conscious awareness.
Visual stimuli are automatically parsed into individual elemental features across independent feature maps (e.g., orientation, color, motion, curvature, depth).
Features exist independently, in a "free-floating" state.
Requires no conscious effort or attentional allocation.
Stage 2: Focused Attention Stage:
Attention acts as the spatial "glue" integrating independent feature maps.
The Attentional Spotlight focuses on a specific spatial coordinate on the Master Map of Locations.
All isolated features mapped to that specific spatial coordinate are bound together, synthesizing a unified object representation that proceeds to object perception and recognition.
Empirical Evidence for Feature Integration Theory:
Illusory Conjunctions:
Experimental demonstrations where features from separate visual objects are incorrectly combined.
In Treisman & Schmidt's paradigm, displays featuring two black numbers flanking four colored geometric objects (e.g., a green circle, red square, blue triangle) are presented for approximately , followed by a visual mask.
Participants prioritize reporting the flanking numbers (consuming attentional resources).
When subsequently asked to identify the shapes, participants reliably report illusory conjunctions (e.g., reporting a green square or a red circle).
This confirms that during the preattentive stage, features float independently and are prone to miscombination when focused attention is withheld.
Balint's Syndrome:
A neuropsychological condition resulting from bilateral lesions to the posterior parietal and occipitoparietal cortices.
Characterized by simultanagnosia: the clinical inability to attend to and perceive more than one visual object at a time.
Because patients cannot allocate focal spatial attention across multiple stimuli, they exhibit an exceptionally high incidence of illusory conjunctions, even with long display exposures.
The deficit is object-based rather than purely spatial: if separate colored dots are connected by a contour line, patients perceive them as a single object, correctly integrating their constituent features.
Visual Search Paradigms in FIT:
Feature Search (Pop-Out Search):
Searching for a target defined by a single primitive feature (e.g., finding a horizontal line among vertical lines, or a red square among green squares).
Operates preattentively and in parallel; reaction times remain constant regardless of the total number of distractor elements (set size).
Conjunction Search:
Searching for a target defined by the integration of two or more distinct features (e.g., finding a red circle among red squares and green circles).
Requires serial scanning by the focused attentional spotlight; reaction times increase linearly as a function of display set size.
Interaction with Top-Down Processing:
Although FIT describes a predominantly bottom-up architecture, top-down semantic expectations can override illusory combinations.
When participants are provided meaningful context (e.g., informed they will see a carrot, a lake, and a tire), they rarely make illusory conjunctions (e.g., reporting an orange lake), because top-down knowledge guides appropriate feature binding.
Neural Networks and Functional Architecture of Attention
Attention is coordinated across large-scale intrinsic functional brain networks showing antagonistic, reciprocal dynamics:
The Frontoparietal Attention Network (Task-Positive Network):
Comprises the frontal eye fields (FEF), intraparietal sulcus (IPS), superior parietal lobule, and dorsolateral prefrontal cortex (DLPFC).
Displays increased metabolic and blood-oxygen-level-dependent (BOLD) activation during externally directed, cognitively demanding tasks requiring focused, selective attention.
The Default Mode Network (DMN / Task-Negative Network):
Centered on the medial prefrontal cortex (mPFC), posterior cingulate cortex (PCC)/precuneus, and inferior parietal lobule.
Shows maximal metabolic activity during wakeful rest, task-unrelated thought, internal mind-wandering, daydreaming, and episodic memory retrieval.
Dynamic Antagonism:
The task-positive Attention Network and task-negative Default Mode Network maintain an inverse, anticorrelated functional relationship.
Successful attentional performance requires suppression of the DMN; transient intrusions of DMN activity during sustained cognitive tasks produce attentional lapses, delayed reaction times, and behavioral errors.
Comparative Cognition and Animal Models in Attention Research
Human cognitive architecture shares deep phylogenetic continuities with non-human animals, including core capabilities across sensory perception, selective orienting, communication, spatial mapping, action planning, memory, and executive decision-making.
Animal model paradigms:
Behavioral Analogues:
Tolman's rodent cross-maze experiments demonstrated that rats navigate via internal cognitive maps (spatial memory representations) rather than rigid stimulus-response reflexes.
- Selective rearing paradigms (e.g., Blakemore & Cooper's kittens exposed exclusively to horizontal or vertical visual orientations) proved the developmental plasticity and experience-dependent tuning of primary visual feature detectors.
Behavioral Neuroscience and Electrophysiology:
Animal models allow invasive methodologies unavailable in human research, enabling causal investigations into brain systems:
In vivo patch-clamp electrophysiology: Measuring micro-level resting membrane potentials (), synaptic currents, and action potential firing thresholds in targeted neurons.
- *Ablation and Lesion Studies:* Primate lesion studies demonstrated the functional dissociation between visual processing streams:
- Bilateral ablation of the temporal lobe selectively impaired the "what" pathway (object discrimination).
- Bilateral ablation of the parietal lobe selectively impaired the "where/how" pathway (landmark discrimination and spatial localization).
Ethics of Animal Cognition Research:
Research programs balance scientific insight against animal harm.
Implementation of the 3Rs Framework:
Replacement: Substituting animal models with computational simulations, cell cultures, or human psychophysics where feasible.
Reduction: Employing rigorous experimental designs to minimize the absolute number of animals used.
Refinement: Refining procedural and husbandry practices to eliminate animal pain, distress, and discomfort.
Ethical compliance is monitored by institutional oversight bodies (e.g., Institutional Animal Care and Use Committees [IACUC]).
Questions & Discussion
Inquiry 1: Resolution of Filter Models
Question: How does the brain switch between early and late selection depending on task context?
Response: The auditory system does not use a rigid, single-locus physical filter. Early selection predominates under conditions of high perceptual load and demanding sensory environments, attenuating unattended sensory streams at peripheral stages. When perceptual load is low, spare cognitive capacity allows unattended messages to receive semantic evaluation, manifesting as late-selection phenomena.
Inquiry 2: Everyday Automaticity and Executive Function
Question: What everyday skills transition to complete automaticity, and can automaticity be reversed?
Response: Highly practiced sensory-motor skills—such as reading text, tying shoelaces, touch typing, and shifting gears in a manual transmission—become fully automatized, functioning with minimal executive control. However, when unfamiliar or unexpected circumstances arise (e.g., driving on an icy road surface or reading inverted typography), the cognitive system must suspend automatic processing and re-engage controlled, high-load executive processing.
Inquiry 3: Ethical Boundaries of Animal Models
Question: What ethical responsibilities exist when modeling human cognitive disorders in non-human primates and rodents?
Response: Investigators and institutional review bodies must balance the translational value of neuroscience discoveries against potential distress to animal subjects. Under the 3Rs framework, non-invasive neuroimaging and high-density EEG in humans are prioritized, reserving invasive single-unit recordings or optogenetic manipulations for hypotheses that cannot be validated through other methodologies.