Automatic processing in psychology refers to cognitive activities that are relatively fast and require few cognitive resources. It runs below conscious awareness.
This makes it different from focused, effortful attention, and it is common when carrying out familiar, highly practiced tasks.
These tasks or mental processes have become automatic through repetition and learning, meaning they no longer require focused attention to be successfully completed.
We do this without conscious intention. Automatic processing simply takes over once a task becomes familiar enough.
For example, these include carrying out familiar tasks like reading, driving, cycling, playing a game, or observing things. They do not need much cognitive effort and are relatively easy to perform.
Key Features
Controlled processing is the slow, effortful, conscious mode we use for anything new. Consistent practice moves a task along a continuum from controlled to automatic (Schneider & Shiffrin, 1977). Six characteristics mark the automatic end:
- Unconscious: Automatic processes need no conscious awareness or control. We have repeated these tasks so often that they are habitual, so we do them without actively thinking.
- Effortless: Automatic processes need minimal mental effort. They are well-learned tasks we have practised extensively, so they demand fewer cognitive resources than unfamiliar ones.
- Fast: Automatic processes are quick and efficient. They enable us to respond rapidly to familiar stimuli or carry out routine tasks at high speed.
- Simultaneous Processing: Automatic processes run alongside other tasks with little interference, because they need fewer limited resources than controlled processes. We can walk (motor), talk (language) and recognize a familiar face (perceptual) at once.
- Not Easily Modified: Once learned, an automatic process is hard to change or unlearn. That helps when the response is appropriate, but hurts when circumstances change.
- Errors and Biases: Automatic processing can sometimes lead to errors or biases, especially when the task or situation changes unexpectedly or heuristics (mental shortcuts) lead to incorrect judgments or decisions.
Automatic vs Controlled Processing
The two modes differ on eight dimensions.
| Feature | Controlled processing | Automatic processing |
|---|---|---|
| Demand on attention | Heavy; capacity-limited | Little or none; unaffected by load |
| Speed | Slow | Fast |
| Conscious awareness | Present; we know we are doing it | Largely absent; runs below awareness |
| Control | Deliberately initiated and steered | Triggered by the stimulus; hard to start or stop at will |
| Effort | Effortful, mentally taxing | Effortless |
| Serial vs parallel | Serial: one step at a time | Parallel: several can run at once |
| Flexibility | Flexible; adapts to novelty | Inflexible; disrupted when conditions change |
| Development | The default for anything new | Acquired gradually through consistent practice |
Controlled processing is serial because it is capacity-limited. Comparing items one comparison at a time means more items take longer.
Automatic processing is parallel because it is not capacity-limited. An over-learned detection process can inspect all the items at once, so adding more barely slows it (Schneider & Shiffrin, 1977).
The Landmark Search Study
Here is the classic evidence.
- Aim: Schneider and Shiffrin (1977) tested whether one search task could separate a slow, capacity-limited controlled process from a fast, capacity-free automatic one.
- Method: Participants memorised a few target letters or digits, then scanned a brief display for a match. Under consistent mapping, targets and distractors never swapped roles; under varied mapping, they switched from trial to trial.
- Results: Varied mapping: decision time rose steeply with load, the mark of a controlled, one-at-a-time search. Consistent mapping: time stayed fast and almost flat, the signature of automatic, parallel detection.
- Conclusion: Practice on a consistent mapping builds genuinely automatic detection; an inconsistent mapping never does, however much it is practised (Schneider & Shiffrin, 1977).
This gave psychology both the theoretical distinction and a method for testing which mode is in play.
How Practice Builds Automaticity
Automaticity is not a fixed property of a task. A process acquires it through consistent practice, where the same stimulus always calls for the same response (Shiffrin & Schneider, 1977).
Where the mapping keeps changing, no amount of repetition helps. There is no stable routine to consolidate.
Skill learning follows a familiar arc:
- Cognitive phase: Performance is slow, effortful and error-prone, and depends wholly on controlled processing.
- Autonomous phase: The component operations run automatically, freeing attention for strategy, expression and anticipation.
A beginner driver must attend consciously to the clutch, mirrors and gears, and any distraction undoes them. An experienced driver does all of this automatically, with attention to spare for conversation.
An expert typist attends to the text, not the keys.
Brain imaging shows the same shift.
Early, controlled performance recruits the prefrontal and parietal cortex heavily. That activation falls as the skill automatizes, and the work moves toward more efficient subcortical circuitry such as the basal ganglia (Poldrack et al., 2005; Schneider & Chein, 2003).
Automatized skills draw on procedural memory, the implicit memory for how to do things. Because automaticity accrues only where practice has been consistent, expertise is domain-specific: a chess master’s pattern recognition does not transfer to an unrelated game.
Automatic Processing Examples
There are many examples of automatic processing that humans undergo in our daily life, such as:
- Reading and understanding words in your native language.
- Recognizing the faces of familiar people.
- Performing routine motor tasks like walking or tying shoelaces.
- Driving a familiar route without actively thinking about where to turn.
- Quickly deciding the meaning of a simple sentence.
Other examples include:
Stroop Effect
The Stroop Effect is a psychological phenomenon that demonstrates automatic processing. It is one of psychology’s most famous demonstrations.
In a standard Stroop task, participants see color words, such as “red,” “blue,” and “green,” printed in colored ink. The catch: the ink colour does not match the word’s meaning, so “red” might appear in blue ink.
They must name the ink colour, not read the word.
This is hard because reading is deeply automatic, practiced since childhood. Naming colors is different: it stays more controlled and deliberate.
When people see a colored word, their automatic processing system reads it without conscious effort.
Naming the ink colour instead requires controlled processing, the slower and more effortful mode of conscious thought, so they must suppress that automatic response.
The clash slows response times and raises the error rate.
In fact, naming the ink colour of a mismatched word takes around 74% longer than naming plain coloured squares (Stroop, 1935).
The interference ran almost entirely one way. Reading the words was barely affected by the colour of their ink (Stroop, 1935). That asymmetry marks reading as the automatic process.
The Stroop Effect is thus a clear example of automatic processing: reading intrudes on the controlled task of naming colors. It also shows how hard an automatic response is to inhibit once triggered.
Heuristics
Heuristics are mental shortcuts or “rules of thumb” that we use to make decisions or judgments quickly and efficiently.
They are often used when we need to make a decision but don’t have the time, resources, or information to go through a full, logical decision-making process.
- Availability Heuristic: We judge how likely an event is by how easily examples come to mind. Easily recalled plane-crash news, for instance, can inflate the perceived risk of air travel.
- Halo Effect: A cognitive bias in which our impression of someone in one area shapes our view of them in others. Attractive people, for instance, may be unconsciously judged kinder or more intelligent.
- Anchoring Heuristic: We rely too heavily on the first piece of information (the “anchor”) we receive. A very high first price at a dealership, for instance, can shape how we judge later cars.
Motor Skills
Motor skills are a form of automaticity. They can be defined as a state where other ongoing tasks do not significantly impact the performance of a primary task.
This means the primary task can be carried out without conscious effort or attention. The brain becomes accustomed to it through repeated exposure and practice (Poldrack et al., 2005).
Walking, jumping, and riding a bicycle are common examples. You may notice that breaking up your walking steps is very difficult; the brain does not consciously plan each step.
Instead, hours and hours of practice and repetition wire the task into our central nervous system, making the processes automatic.
Implicit Biases
Implicit bias (unconscious bias) is a form of bias that occurs automatically and unintentionally yet still impacts our judgments, decisions, and behaviors.
Implicit biases are an example of system 1 thinking, so we are unaware they exist (Greenwald & Krieger, 2006).
It can target racial groups or any group. It is said to be shaped by the learned associations between certain qualities and social categories and is a part of implicit social cognition.
This is one example. It shapes perceptions, attitudes, and stereotypic thoughts without conscious acknowledgment or intention. This makes implicit bias part of automatic processing (Greenwald & Krieger, 2006).
Automatic activation can shape overt behaviour, not just judgement.
In a landmark study, participants primed with rudeness interrupted an experimenter more readily than those primed with politeness. Others, primed with an elderly stereotype, walked more slowly down a corridor afterwards. Neither group was aware that their behaviour had been affected (Bargh et al., 1996).
The walking result is now contested. A later, tighter rerun found no walking-speed effect once the experimenters timing it were blind to condition (Doyen et al., 2012). The Evidence Check below weighs it.
Cognitive Distortions
Cognitive distortions are irrational, inflated, or inaccurate thoughts and beliefs that distort our perception of reality, often contributing to mental health disorders such as anxiety and depression.
Cognitive distortions are automatic. They typically operate outside our conscious awareness. They are habitual ways of responding to certain situations or stimuli based on our past experiences, beliefs, and conditioning. They often influence our perceptions and decision-making processes without us realizing it.
Here are a few examples of common cognitive distortions:
- All-or-Nothing Thinking: Viewing things in black-and-white categories with no middle ground (e.g., “If I’m not perfect, I’m a failure”).
- Overgeneralization: Applying the outcome of one event to all similar events (e.g., “I failed this test, so I’m bad at all tests”).
- Catastrophizing: Expecting the worst-case scenario to happen.
- Personalization: Assuming excessive personal responsibility for events outside of your control (e.g., “It’s my fault that my friend is upset”).
- Mental Filter: Focusing exclusively on the negative aspects of a situation while ignoring the positive.
Priming
Another example of automatic processing is priming. Exposure to one stimulus quietly speeds up how we respond to a related stimulus later, without our awareness.
Priming is often described as implicit, or non-declarative, memory. People are unaware that an earlier stimulus is shaping their current behaviour or response.
For instance, someone exposed to the word “yellow” will respond more quickly to “banana” than to an unrelated word like “television.” Yellow and banana are closely linked in memory, so the second word is processed faster once the first has primed it (Higgins & Bargh, 1987).
Evidence: The Cost-Benefit Priming Study
- Aim: Posner and Snyder (1975) tested whether the speed-up from a prime reflects a fast, automatic spread of activation or a slower, conscious strategy.
- Method: Participants judged targets, such as whether two letters matched, after a prime. Valid trials (prime predicted the target) and invalid trials (prime pointed elsewhere) were compared with a neutral baseline, at varied intervals.
- Results: At short intervals, priming gave a speed benefit with little or no cost when misleading. At longer intervals, a cost on invalid trials also appeared, the sign of a slower, conscious strategy.
- Conclusion: Priming works two ways. Fast, automatic activation brings benefits without costs; a slower conscious strategy adds costs, drawing on limited attentional capacity (Posner & Snyder, 1975).
The method isolates both influences within one task. Its limit is that the clean split holds best for simple, well-associated material and short intervals.
Advantages
Automatic processing has clear advantages. It is functional and robust under high-stress levels.
Functions Under Stress
It can be hard to “think straight” under stress. Automatic processing comes in handy here, since it does not require as much mental capacity or guidance, and people can act on instinct instead.
As people gain more experience, automatic processing helps them feel comfortable in their environment. It also makes people robust under stress or distraction. Repeated exposure to events and outcomes helps people anticipate probable outcomes in familiar settings, without draining much mental energy.
Facial recognition and surrounding observation are also automatic.
Our eyes and ears take in information automatically, without conscious awareness. Sight works the same way. We do not think about how we see, or about walking, breathing, and blinking; the brain simply handles all of this automatically.
Parallel Processing
Automatic processing also allows many processes to occur in parallel. Riding a bike is a good example.
When people ride a bike, they don’t think about how their legs pedal or the mechanics of the bike. They just “know” how to do it.
This frees people to stay aware of the surrounding environment while biking and focus on more important things. It increases productivity and helps people focus on urgent needs (Schneider & Chein, 2003).
There is a reason for this. Automatic processing is not capacity-limited, unlike controlled processing, so an over-learned process does not compete for the same limited attentional resources that a controlled task needs (Schneider & Shiffrin, 1977).
This is also why an experienced driver can hold a conversation while driving, something a learner cannot do.
Not every pair of tasks can run together. What matters is how each one is processed:
- Two controlled tasks: They compete for the same limited capacity and interfere badly.
- One automatic, one controlled: The automatic task costs so little that it can be paired with a demanding one.
- Practice: Dual-task interference falls as one task automatizes (Shiffrin & Schneider, 1977).
Texting while driving in heavy traffic shows the limit. Both tasks need controlled processing, so performance degrades. What looks like multitasking is usually rapid switching, or one automatic task paired with one controlled task. Safe dual-tasking needs at least one task to be genuinely automatic.
Disadvantages
The convenience of automatic processing comes with disadvantages too.
Prone to Errors
Automatic processing is more prone to mistakes than controlled processing. This is because an automatic response fires whenever its usual trigger appears, whether or not that is still the right response, rather than being chosen deliberately.
Driving shows this well.
People rely on automatic processing for routine actions like changing lanes or turning. A tired or distracted driver’s automatic response can still lead to mistakes that cost lives.
Road changes are a good example. Suppose a driver is used to turning left at a junction, but the layout changes so only a right turn is allowed. A tired or distracted driver’s automatic response may still turn left, leading to a potentially fatal collision.
This is a case of an automatic habit misfiring when the situation has changed.
Hard to Modify
Automatic processing is also less easy to change once established. It takes time to wire a process into the brain, so once it is wired, an unhelpful habit or mistake can be hard to undo.
This inflexibility is the flip side of automaticity’s efficiency. A response that fires without being chosen cannot easily be redirected when the old routine no longer fits.
This cost is not just anecdotal. A classic reversal experiment measured it.
- Aim: Shiffrin and Schneider (1977) tested whether automatic detection develops gradually with consistent practice, and what it costs when the learned mapping is reversed.
- Method: Participants practised a consistent-mapping search task for about 2,100 trials. The mapping was then reversed: former targets became distractors, for a further 2,100 trials.
- Results: Performance improved steadily over training. After the reversal it collapsed, taking nearly 1,000 trials to return to the level shown at the very start of training.
- Conclusion: Automaticity is built gradually by consistent practice and, once built, is highly inflexible. It cannot be readily suppressed or re-mapped when conditions change (Shiffrin & Schneider, 1977).
The result is often cited as the experimental basis for habit interference and everyday action slips, where a learned routine fires in the wrong context.
For example, someone used to typing with one finger may find it difficult to switch to the correct technique later. Nail-biting and procrastination are similar: both are hard to break once the brain has wired them in as automatic responses.
System 1 Thinking
In cognitive psychology, automatic processing maps onto Daniel Kahneman’s concept of System 1 thinking, from his book Thinking, Fast and Slow (Kahneman, 2011). System 1 is fast, effortless and largely unconscious, like automatic processing itself.
What Is System 1?
System 1 is the brain’s fast, emotional, unconscious thinking mode. It requires little effort, but it is often error-prone. Most everyday activities, such as driving, talking, and cleaning, make heavy use of System 1.
System 1 thinking is our intuitive system: fast, automatic, immediate, and effortless. It often involves mental shortcuts and is driven more by the subconscious mind.
System 2 thinking is the opposite. It is slower, more deliberate, more effortful, and used for complex decision-making, problem-solving, and reasoning.

When We Rely on Each System
We rely on System 2 thinking for unfamiliar situations, since it is less prone to error. System 1 relies on mental shortcuts called heuristics, which can lead to illogical decisions.
System 2 is slow and cognitively taxing. We cannot rely on it all the time, so System 1 handles most of our daily tasks and decisions instead.
System 1 also includes the “fight or flight” response and other instinctive reactions critical for survival. These responses are fast and automatic, letting us react quickly to threats.
Individual Differences in Overriding Intuition
The balance between the two systems is not set by the task alone. People differ reliably in whether they override an automatic answer.
Stanovich and West (2000) studied problems where a fast, intuitive answer conflicts with the correct but more effortful one. People higher in cognitive ability, and in a disposition toward effortful, open-minded thinking, were more likely to engage deliberate processing.
Others defaulted to the quick answer. That held even when the answer was wrong.
Classic automaticity research manipulates the task. It varies mapping consistency or prime timing, and treats the automatic-to-controlled balance as a property of the task and its practice history.
Holding the task constant, people still differ in how readily they recruit controlled override. A purely task-driven account has no place for that variance.
The finding links automatic processing to the wider dual-process literature on thinking dispositions and rationality (Evans & Stanovich, 2013).
Evidence Check: Does the Theory Hold Up?
Verdict: Supported with qualifications.
Interference with colour naming, the main behavioural sign of automatic reading, has been reproduced in a very large independent sample, though in a related task. The behavioural-priming illustration did not reproduce.
No independent large-sample rerun of the defining consistent-mapping claim was found. So the verdict covers the reading signature, not the two-process theory as a whole.
Reading Interferes With Colour Naming
Level of evidence: Experiment; partly replicated.
Reading, the claim goes, is automatic for a literate adult. The meaning of a familiar word is extracted involuntarily and slows the naming of a conflicting ink colour (Stroop, 1935).
No meta-analysis of the interference effect in healthy adults was found. The meta-analyses located pool clinical comparisons, such as children with attention-deficit disorders, and do not test whether reading intrudes involuntarily.
The strongest evidence is therefore experimental.
Many Labs 3 reran the effect in a preregistered, computerised, partial replication:
- Task: Participants categorised the font colour of the words red, blue and green.
- Sample: 20 university participant pools and an online sample, N = 3,279, more than the 2,696 in the pools alone.
- Result: Responses were slower on incongruent words (meaning conflicts with ink colour) than on congruent words (M = 0.27, SD = 0.31, d = .88, 95% CI [.84, .92]).
The authors describe this as replicating the basic effect (Ebersole et al., 2016).
The slowdown from an unwanted word is real and large. That is what automatic reading predicts.
A rival view that treated reading as fully controlled would predict no intrusion from a task-irrelevant word, so the result favours automaticity over that rival.
A graded account, in which attention modulates how far the word is processed, predicts the same slowdown. Kahneman and Chajczyk (1983) found the interference reduced when attention to the word was diluted, which favours that account over “no attentional demand”.
The replication therefore secures a Stroop-type congruency effect. Conflicting words slow responses compared with matching ones. It does not secure the strong “involuntary and attention-free” reading of it.
The replication is strong: preregistered, very large and independent of the original authors. Its limit is that it compared conflicting with matching words, not with solid colour patches as in 1935, so it mixes interference with facilitation.
Primed Stereotypes and Walking Speed
Level of evidence: Experiment; failed to replicate.
The claim is that participants primed with an elderly stereotype walked more slowly down a corridor, unaware that their behaviour was affected. The 1996 study by Bargh and colleagues is offered as an illustration of automatic social cognition shaping behaviour.
The original used a scrambled-sentence task to prime the stereotype, with walking speed as the outcome. Doyen et al. (2012) reran it in two experiments:
- Experiment 1: 120 undergraduate students (against 60 in the original studies 2a and 2b combined) were randomly and blindly assigned to a prime or no-prime task. Walking time was recorded by infrared sensors, with experimenters blind to condition.
- Experiment 1 result: There was no difference in walking time between the groups (F(1, 119) < 1, eta-squared = .01).
- Experiment 2: 50 new participants were tested by experimenters told to expect either slower or faster walking. A slowing appeared, with objective timing, only in the group tested by experimenters expecting slower walking.
A search of works citing the original found commentary and reviews. It found no further direct rerun.
The original result did not reappear under tighter measurement in a sample twice its size. That is a failure to find the effect, not proof of its absence in all conditions, because one laboratory in another country cannot exclude moderators.
Experiment 2 is consistent with experimenter expectation contributing. But it was small, and that account is the authors’ suggestion, not a settled cause.
Debriefing also suggested some awareness of the primes. That bears on the “unaware” part of the claim.
Either way, the finding cannot support the claim that unconsciously activated stereotypes reliably change behaviour.
The rerun improved on the original with a larger sample, automated timing and blind experimenters. Its limits are one laboratory, French-speaking Belgian undergraduates and no reported preregistration, so Experiment 1 is the part to trust most.
Critical Evaluation
The distinction is among the more secure ideas in cognitive psychology. Independent methods converge on two dissociable modes.
Load-slope searches, reversal-cost studies, cost-benefit priming and the Stroop effect each show it (Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977; Posner & Snyder, 1975; Stroop, 1935).
Each predicts its own signature: steep versus flat load functions, benefits without costs, and one-way interference. The distinction is still not beyond challenge. Several criticisms are worth weighing:
- Not Fully Attention-Free: Allegedly automatic tasks still interfere with other tasks, and the Stroop effect itself can be reduced by manipulating attention.
- A Matter of Degree: The features said to define automaticity do not reliably occur together, so most processes are a blend rather than purely one or the other.
- Descriptive, Not Explanatory: Saying practice creates automaticity does not explain how a slow, serial process becomes fast and parallel.
- Not Strictly Unavoidable: Automatic influences often depend on task relevance and attention, so they are less obligatory than first claimed.
Not Fully Attention-Free
Hampson (1989) found that tasks assumed to be automatic still interfered with a second task. This challenged the classic claim that automatic processing consumes no attentional capacity at all. The Stroop effect adds a second complication.
If reading were purely automatic, attention should not matter to how strongly it interferes with colour naming. But it does. When the conflicting word sits away from the colour patch it names, Stroop interference drops substantially (Kahneman & Chajczyk, 1983).
Even Schneider and Shiffrin’s own consistent-mapping search was not perfectly flat at the highest loads tested. Automatic search, it turns out, still carries some small cost.
None of this overturns the automatic/controlled distinction. It does soften the strongest version of the claim: automaticity makes few demands on attention, not none at all.
A Matter of Degree
Posner and Snyder (1975) proposed three criteria for an automatic process: it occurs without intention, without conscious awareness and without interfering with other ongoing mental activity. Later analysis showed they are harder to apply than they sound.
Moors and De Houwer (2006) took apart the features researchers had used to define automaticity: goal-independence, unconsciousness, efficiency, and speed. They found these features do not reliably travel together. A process can be fast without being unconscious. It can be efficient without being fully uncontrollable.
So automaticity turns out to be graded rather than all-or-none.
Most real mental processes are a blend of automatic and controlled elements. The honest claim about any one process is usually comparative: process X is more automatic than process Y, not simply “automatic” full stop (Moors & De Houwer, 2006).
This componential view has since become the mainstream position (Moors, 2016). It reframes the whole debate. Instead of sorting processes into two bins, researchers now ask, feature by feature, in what respects and to what degree a given process is automatic.
Descriptive, Not Explanatory
This is the deepest criticism of the three.
Schneider and Shiffrin’s own account has a well-known gap. It shows THAT practice produces automatic detection, but not HOW a slow, serial, capacity-limited process turns into a fast, parallel, capacity-free one (Shiffrin & Schneider, 1977).
Logan’s (1988) instance theory is the leading attempt to fill that gap. It proposes that automaticity is really a memory phenomenon: with repetition, a stored past answer is retrieved rather than freshly computed.
With repetition, a stimulus stops being computed step by step. Instead, it comes to retrieve a stored past solution directly from memory, in a single step.
Logan summed this up neatly: “only the knowledge base changes with practice” (Logan, 1988). The theory explains automaticity’s speed, its low attentional demand, and its unconscious character all at once.
But it has a weakness. Single-step retrieval is both Logan’s definition of automaticity and his explanation of it, which verges on circular.
Not Strictly Unavoidable
Schneider and Shiffrin (1977) treated automatic processing as strictly unavoidable. Once the trigger appears, the process runs. Later evidence suggests it is more conditional.
Logan (1988) gave participants 512 training trials with words that always appeared in the same colour, in a task that required the colour to be processed. At transfer, the words were paired with new colours.
When the transfer task still required colour processing, reversing the trained colour disrupted performance. When it did not, the same reversal caused no disruption.
So stored information shaped performance only when it was relevant to the task (Logan, 1988). Kahneman and Chajczyk (1983) found a similar conditionality: when the conflicting word was separated from the colour, Stroop interference fell.
Automatic influences, it seems, are conditional rather than strictly unavoidable. That fits the wider finding that automaticity is a matter of degree.
Norman and Shallice’s Graded Model
Norman and Shallice (1986) proposed graded control rather than two categories. Their model has three levels:
- Schemas: Well-learned routines run off automatically at the base level.
- Contention scheduling: When several schemas fire at once and conflict, this arbitrates between them without deliberate control.
- Supervisory attentional system: This steps in only for novel situations, decision-making and trouble-shooting, supplying flexible control that routine schemas cannot.
The supervisory system is this model’s version of controlled processing. It maps closely onto the central executive of the working-memory model.
The two accounts agree on the core division between an effortless routine layer and an effortful override. Norman and Shallice built their model to explain the same reversal-cost and inflexibility findings Schneider and Shiffrin reported.
They differ in what they trade away. Norman and Shallice give up Schneider and Shiffrin’s crisp dichotomy and heavy experimental grounding. In return, their architecture better fits partial and graded automaticity, but it is less precisely specified and less extensively tested.
Contemporary Research
Recent work asks a sharper question.
Once an automatic association is learned, can it be deliberately changed?
Forscher and colleagues (2019) tackled it at scale.
- Aim: To establish, across the whole experimental literature, how far implicit associations can be changed and whether that changes explicit attitudes or behaviour.
- Method: A meta-analysis of 492 studies with more than 87,000 participants, covering every major procedure for shifting implicit measures such as the Implicit Association Test. Each study was coded for intervention type and any change in explicit attitudes or behaviour.
- Results: Implicit associations could be shifted, but effects were generally weak and most procedures produced little durable change. Changes in implicit measures did not reliably translate into changes in explicit attitudes or behaviour.
- Conclusion: Automatic associations are malleable briefly but resistant to lasting change. Manipulating them is not a reliable route to changing behaviour (Forscher et al., 2019).
This vindicates an old claim: automatic processes, once established, are hard to modify (Schneider & Shiffrin, 1977).
A synthesis of nearly 500 studies sits near the top of the evidence hierarchy. Its scope is narrower: the implicit-measures literature. Its verdict concerns automatic social associations, not perceptual-motor automaticity.
Even so, it reframed bias-reduction work by showing that the automatic layer is far less tractable than early enthusiasm assumed.
Key Takeaways
- Fast & Effortless: Automatic processing is quick, needs little mental effort, and runs largely outside conscious awareness.
- Built by Practice: Repetition and consistent practice turn an effortful, controlled task into an automatic one.
- Parallel Processing: Because it costs so little, automatic processing can run alongside other tasks at the same time.
- Hard to Modify: Once learned, an automatic response is difficult to modify, which is why it can misfire when circumstances change.
- Not Fully Attention-Free: Modern research shows automatic processing is not entirely independent of attention; it makes few demands on attention, not none.
- System 1: Automatic processing maps onto Kahneman’s fast, intuitive System 1, contrasted with the slower, deliberate System 2.
- Resistant to Intervention: A large 2019 meta-analysis found automatic associations, once formed, resist lasting change even under deliberate intervention.
- Verdict: Supported with qualifications – colour-naming interference reproduced at scale, but the priming illustration did not.
References
Bargh, J. A., Chen, M., & Burrows, L. (1996). Automaticity of social behavior: Direct effects of trait construct and stereotype activation on action. Journal of Personality and Social Psychology, 71(2), 230–244. https://doi.org/10.1037/0022-3514.71.2.230
Doyen, S., Klein, O., Pichon, C.-L., & Cleeremans, A. (2012). Behavioral priming: It’s all in the mind, but whose mind? PLoS ONE, 7(1), e29081. https://doi.org/10.1371/journal.pone.0029081
Ebersole, C. R., Atherton, O. E., Belanger, A. L., Skulborstad, H. M., Allen, J. M., Banks, J. B., Baranski, E., Bernstein, M. J., Bonfiglio, D. B. V., Boucher, L., Brown, E. R., Budiman, N. I., Cairo, A. H., Capaldi, C. A., Chartier, C. R., Chung, J. M., Cicero, D. C., Coleman, J. A., Conway, J. G., … Nosek, B. A. (2016). Many Labs 3: Evaluating participant pool quality across the academic semester via replication. Journal of Experimental Social Psychology, 67, 68–82. https://doi.org/10.1016/j.jesp.2015.10.012
Evans, J. St. B. T., & Stanovich, K. E. (2013). Dual-process theories of higher cognition: Advancing the debate. Perspectives on Psychological Science, 8(3), 223–241. https://doi.org/10.1177/1745691612460685
Forscher, P. S., Lai, C. K., Axt, J. R., Ebersole, C. R., Herman, M., Devine, P. G., & Nosek, B. A. (2019). A meta-analysis of procedures to change implicit measures. Journal of Personality and Social Psychology, 117(3), 522–559. https://doi.org/10.1037/pspa0000160
Greenwald, A. G., & Krieger, L. H. (2006). Implicit bias: Scientific foundations. California Law Review, 94(4), 945–967. https://doi.org/10.2307/20439056
Hampson, P. J. (1989). Aspects of attention and cognitive science. Irish Journal of Psychology, 10(2), 261–275. https://doi.org/10.1080/03033910.1989.10557746
Higgins, E. T., & Bargh, J. A. (1987). Social cognition and social perception. Annual Review of Psychology, 38, 369–425. https://doi.org/10.1146/annurev.ps.38.020187.002101
Kahneman, D. (2011). Thinking, fast and slow. Farrar, Straus and Giroux.
Kahneman, D., & Chajczyk, D. (1983). Tests of the automaticity of reading: Dilution of Stroop effects by color-irrelevant stimuli. Journal of Experimental Psychology: Human Perception and Performance, 9(4), 497–509. https://doi.org/10.1037/0096-1523.9.4.497
Logan, G. D. (1988). Toward an instance theory of automatization. Psychological Review, 95(4), 492–527. https://doi.org/10.1037/0033-295X.95.4.492
Moors, A. (2016). Automaticity: Componential, causal, and mechanistic explanations. Annual Review of Psychology, 67, 263–287. https://doi.org/10.1146/annurev-psych-122414-033550
Moors, A., & De Houwer, J. (2006). Automaticity: A theoretical and conceptual analysis. Psychological Bulletin, 132(2), 297–326. https://doi.org/10.1037/0033-2909.132.2.297
Norman, D. A., & Shallice, T. (1986). Attention to action: Willed and automatic control of behaviour. In R. J. Davidson, G. E. Schwartz, & D. Shapiro (Eds.), Consciousness and self-regulation (Vol. 4, pp. 1–18). Plenum Press. https://doi.org/10.1007/978-1-4757-0629-1_1
Poldrack, R. A., Sabb, F. W., Foerde, K., Tom, S. M., Asarnow, R. F., Bookheimer, S. Y., & Knowlton, B. J. (2005). The neural correlates of motor skill automaticity. The Journal of Neuroscience, 25(22), 5356–5364. https://doi.org/10.1523/JNEUROSCI.3880-04.2005
Posner, M. I., & Snyder, C. R. R. (1975). Attention and cognitive control. In R. L. Solso (Ed.), Information processing and cognition: The Loyola symposium (pp. 55–85). Erlbaum.
Schneider, W., & Chein, J. M. (2003). Controlled and automatic processing: Behavior, theory, and biological mechanisms. Cognitive Science, 27(3), 525–559. https://doi.org/10.1207/s15516709cog2703_8
Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human information processing: I. Detection, search, and attention. Psychological Review, 84(1), 1–66. https://doi.org/10.1037/0033-295X.84.1.1
Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic human information processing: II. Perceptual learning, automatic attending and a general theory. Psychological Review, 84(2), 127–190. https://doi.org/10.1037/0033-295X.84.2.127
Stanovich, K. E., & West, R. F. (2000). Individual differences in reasoning: Implications for the rationality debate. Behavioral and Brain Sciences, 23(5), 645–665. https://doi.org/10.1017/S0140525X00003435
Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18(6), 643–662. https://doi.org/10.1037/h0054651

