Attention
Attention = the cognitive processes that determine which information receives further processing.
Attention has two main properties:
- Selectivity: only some available information is selected for further processing.
- Control: attention can be deliberately directed toward information that is relevant to our goals.
Top-Down vs Bottom-Up Attention
| Type | Definition | Example |
|---|---|---|
| Top-down (endogenous) | Attention is deliberately directed according to goals | Looking for your friend in a crowd |
| Bottom-up (exogenous) | Attention is automatically captured by a stimulus | Turning toward a sudden loud noise |
NOTE: Both processes can operate at the same time.
Auditory Selective Attention
Dichotic Listening
A dichotic listening task presents a different auditory message to each ear.
Participants are usually instructed to:
- Attend to one ear.
- Shadow the attended message by repeating it aloud.
- Ignore the message presented to the other ear.
These tasks investigate how much information from the unattended channel is still processed.
Cherry (1953) Example
Method
Participants shadowed one auditory message while ignoring another.
Findings
Participants noticed physical properties of the unattended message, such as:
- speaker’s gender
- changes in pitch/loudness
However, they generally failed to notice semantic properties, including:
- changes in language
- speech being played backwards
Conclusion
Unattended information appears to receive some physical processing, but relatively little processing of its meaning.
→ Supports the idea of early attentional selection.
Moray (1959): Cocktail Party Effect
Moray found that approximately one-third of participants detected their own name when it appeared in the unattended channel.
Cocktail Party Effect
Personally significant information can capture attention even when it occurs in an unattended stream.
Importance
This challenges a strict early-selection account.
If unattended information receives no semantic processing, how could the system recognise that the unattended word was the person’s name?
Bottleneck Models of Attention
A bottleneck is a stage where limited processing capacity prevents all incoming information from being processed further.
The three major models disagree about where this bottleneck occurs.
| Model | Selection | What happens to unattended information? |
|---|---|---|
| Broadbent | Early | Filtered out before semantic processing |
| Treisman | Early/intermediate | Weakened, but can still be processed |
| Deutsch & Deutsch | Late | Fully processed for meaning before selection |
Broadbent’s Filter Model
Main Idea
Attention operates as an early, all-or-none filter.
Sensory input → sensory register → selective filter → further processing → STM
The filter selects information according to physical characteristics.
Attended information
Passes through the filter → receives further processing.
Unattended information
Filtered out → does not receive semantic processing.
Evidence
Explains Cherry (1953):
Physical properties detected → meaning generally not detected.
Limitation
Cannot easily explain Moray’s cocktail party effect, because recognising your own name requires some processing beyond basic physical features.
Treisman’s Attenuation Model
Treisman proposed that unattended information is attenuated rather than completely blocked.
Attenuation
Attenuation = reducing the strength of unattended information.
So:
- Attended information → strong signal
- Unattended information → weakened signal
Both can still receive some processing.
Thresholds
Information differs in how much activation it requires to reach awareness.
Low-threshold information is particularly easy to activate.
Examples include:
- your own name
- personally significant information
- information relevant to the current context
Therefore:
Unattended message → attenuated → personally important word has low threshold → reaches awareness
Explains
The cocktail party effect.
Deutsch & Deutsch: Late Selection
Main Idea
All incoming information receives semantic processing.
Selection occurs after meaning has been processed, when information is selected for awareness or response.
Sensory input → semantic processing → selection → response
Therefore, unattended information can be processed for meaning even if it does not ultimately reach awareness.
Testing Attenuation vs Late Selection
Treisman & Riley (1969)
Participants:
- shadowed one auditory message
- listened for a target word
- tapped when they detected the target
Targets could occur in either the shadowed or non-shadowed ear.
Predictions
Attenuation model:
Detection should be better in the attended ear because unattended information is weakened.
Late-selection model:
Detection should be similar because both channels receive semantic processing.
Results
| Ear | Target Detection |
|---|---|
| Shadowed | 87% |
| Non-shadowed | 8% |
Conclusion
Supports attenuation because target detection was dramatically poorer in the unattended channel.
Limitation
The shadowed ear was more important to the task, which could itself explain why targets were detected more successfully there.
Flexible Bottleneck Theory
Rather than selection always occurring at one fixed stage:
The attentional bottleneck can shift depending on task demands.
Focused attention
Attention concentrated on one information source.
→ selection tends to occur earlier
→ unattended information receives less semantic processing.
Divided attention
Attention must be distributed across multiple information sources.
→ selection occurs later
→ more information receives semantic processing.
Johnston & Wilson (1980)
Aim
Test whether the location of the attentional bottleneck changes depending on whether attention is focused or divided.
Task
Participants detected target words belonging to a category such as:
Musical instruments
Some targets were ambiguous.
Example:
ORGAN
Another word was simultaneously presented to the opposite ear.
Context Conditions
| Context | Example | Interpretation encouraged |
|---|---|---|
| Appropriate | church + ORGAN | Musical instrument |
| Neutral | paper + ORGAN | Neither meaning |
| Inappropriate | kidney + ORGAN | Body organ |
If the word in the other ear is processed semantically, its meaning should influence whether organ is recognised as a musical instrument.
Focused Attention Condition
Participants knew which ear contained the target.
Therefore, they could focus attention on that ear.
Result: Context had little/no effect on target detection.
Interpretation
The unattended word was not processed deeply enough for its meaning to influence target recognition.
→ Early selection
Divided Attention Condition
Participants did not know which ear contained the target.
They therefore had to monitor both ears.
Result:
Appropriate context > Neutral context > Inappropriate context
Interpretation
The meaning of the other word affected interpretation of the target.
Therefore, both auditory streams must have received semantic processing.
→ Later selection
Overall Conclusion
Focused attention → earlier bottleneck
Divided attention → later bottleneck
→ Supports the flexible bottleneck theory.
Divided Attention
Divided attention = directing attention toward multiple sources of information/tasks simultaneously.
Dual-task performance = performance when completing two tasks at the same time.
Because cognitive processing capacity is limited, tasks can interfere with one another.
Factors Affecting Dual-Task Interference
1. Stimulus Modality
More interference occurs when two tasks require information from the same sensory modality.
Example:
Driving + reading a text → both heavily visual.
2. Response Modality
More interference occurs when tasks require the same response system.
Example:
Steering + typing → both require the hands.
3. Cognitive Demand
Tasks can interfere even when they use different sensory and response modalities if both require substantial cognitive processing.
Therefore, even hands-free phone conversations can interfere with driving performance.
Practice and Automaticity
With extensive practice, tasks can require progressively less attention.
Novel task → controlled processing → repeated practice → automatic processing
This can improve dual-task performance because fewer attentional resources are required for the practised task.
Controlled vs Automatic Processing
| Controlled Processing | Automatic Processing |
|---|---|
| Slow | Fast |
| Attention-demanding | Requires little attention |
| Conscious | Can occur without awareness |
| Flexible | Relatively inflexible |
| Common for novel tasks | Develops through extensive practice |
| Can usually be intentionally controlled | Difficult to suppress |
Shiffrin & Schneider: Developing Automaticity
Shiffrin and Schneider investigated how extensive practice can produce automatic processing using a memory-search task.
Basic Task
Participants:
- Memorised a set of target items.
- Viewed a display containing several items.
- Determined whether a target was present.
- Responded as quickly and accurately as possible.
Consistent Mapping (CM)
Targets and distractors never switched roles.
Example:
- Targets = numbers
- Distractors = consonants
Because the relationship remains consistent across trials, extensive practice allows processing to become automatic.
Varied Mapping (VM)
Items can switch roles across trials.
An item that is a target on one trial may later become a distractor.
Therefore, participants continually need to determine:
What are the targets on this trial?
This prevents a stable automatic association from developing.
→ Processing remains controlled.
Set-Size Effect
Set size = number of items that must be searched/processed.
Varied Mapping
As set size increases:
Reaction time increases.
This suggests serial processing:
Items are searched individually → more items = longer search.
Consistent Mapping After Extensive Practice
Reaction time becomes relatively unaffected by set size.
The RT function becomes approximately flat.
This suggests parallel automatic processing:
Multiple items can be processed simultaneously.
Key Pattern
| Condition | Processing | Set-Size Effect |
|---|---|---|
| Varied mapping | Controlled / serial | RT ↑ as set size ↑ |
| Consistent mapping + practice | Automatic / parallel | Little/no effect of set size |
Characteristics of Automatic Processing
Automatic processes are:
Fast
Require relatively little processing time.
Attention-free
Require few attentional resources.
Inflexible
Once learned, automatic associations can be difficult to change.
Reversing previously learned target/distractor relationships therefore causes substantial interference.
Unavoidable
Automatic processes can occur even when they are irrelevant to the current goal.
Stroop Effect
The Stroop task demonstrates the unavoidability of automatic processing.
Participants identify the ink colour while ignoring the written word.
Example:
BLUE written in red ink
Required response:
Red
However, reading the word occurs automatically.
Incongruent Trials
Word meaning ≠ ink colour
→ automatic word reading interferes with controlled colour naming
→ slower responses
Conclusion
Well-practised processes such as reading become difficult to suppress, even when they interfere with the current task.
Logan’s Instance Theory of Automaticity
Instance Theory explains how practice produces automaticity.
Instance
An instance is a stored memory representation of a previous encounter with a task or stimulus.
Each time a task is performed, another instance is stored.
Early Practice: Algorithmic Processing
Initially, a person uses a rule or algorithm to solve the problem.
Example:
3 × 4
New learner:
3 + 3 + 3 + 3 = 12
This is:
- slow
- controlled
- attention-demanding
Extensive Practice: Memory Retrieval
Repeated experiences create many stored instances.
Eventually, seeing:
3 × 4
allows the answer:
12
to be retrieved directly from memory.
The calculation itself is no longer necessary.
Development of Automaticity
Novel task
↓
Algorithm/rule used
↓
Experience stored as an instance
↓
Practice creates more instances
↓
Direct memory retrieval becomes more likely
↓
Automatic processing
Main Idea
Practice shifts performance from algorithmic processing to direct memory retrieval.
