The question
What has to happen between an error occurring and behaviour improving, and why does detection so often fail to produce learning?
Definition
Error monitoring is the process through which a system registers that an action, outcome or developing response does not match a relevant standard; behavioural adjustment is what happens when that information alters what the person does next.
You are typing and press the wrong key. Before you have fully considered what happened, you delete the character, retype it and perhaps slow down for the next few words.
This small sequence contains several distinct events. An action departed from what was intended. The departure was detected. Behaviour changed. Yet none of this guarantees that the change was useful or that the same error will be avoided later.
Error monitoring is the process through which a system registers that an action, outcome or developing response does not match a relevant standard. Behavioural adjustment is what happens when that information alters what the person does next.
The relationship appears straightforward: detect the error, correct the behaviour. In practice, the sequence is less reliable. Errors can be detected without being corrected. Behaviour can change without improving. A correction can solve the immediate problem without producing lasting learning.
Understanding error monitoring therefore requires more than asking whether someone noticed a mistake. It requires examining what standard defined the error, what information became available and how that information affected subsequent action.
An error requires a standard
An event is not an error merely because it produces an undesirable outcome. It becomes an error when an action or result is evaluated against a standard it fails to meet.
That standard might be:
- an intended action;
- a task rule;
- a predicted outcome;
- an external instruction;
- an established procedure;
- a broader goal.
If you intend to type change but type chance, the intended word provides the standard. In a laboratory task, the standard might be a rule requiring one button for a red symbol and another for a blue one. In conversation, a response may be judged against the intended meaning or effect.
Different standards can also produce different evaluations of the same behaviour. An action may comply perfectly with an immediate instruction yet work against the broader goal the instruction was supposed to serve. It is then correct relative to one standard and mistaken relative to another.
Error classification therefore depends on which standard is controlling the evaluation. Without that distinction, disagreement about whether an error occurred may actually be disagreement about what the action was meant to accomplish.
Making an error is not the same as detecting it
People do not always become consciously aware of their errors. A response can deviate from a task rule without the person explicitly recognising the deviation. Conversely, a person can experience uncertainty or a sense that something went wrong even when the response was technically correct.
Research using speeded-response tasks has identified neural activity associated with errors shortly after an incorrect action. The error-related negativity, often abbreviated ERN or Ne, can appear rapidly and does not require a lengthy conscious evaluation. Other activity, including the error positivity or Pe, has been associated more consistently with conscious error awareness and later evaluation (Gehring et al., 1993; Nieuwenhuis et al., 2001).
These findings do not establish a simple division in which one signal always represents unconscious detection and another always represents conscious awareness. They do, however, support an important distinction: some error-related processing can begin before a person has formed a complete conscious account of what happened.
Error monitoring operates on information about mismatches, conflict and outcomes. That information may come from several sources.
A response may feel different from the action that was intended. Competing responses may become active at the same time. Feedback may reveal that the outcome was wrong. The result may violate a prediction. Another person may identify the discrepancy.
These sources do not always agree. Internal monitoring may indicate that something went wrong before external feedback arrives. External feedback may identify a problem that the person did not detect. Either source may also be unreliable if it evaluates behaviour against an unsuitable standard.
Detection is therefore not a single, infallible signal. It is an evaluation produced from available information.
An error can change the next response
After detecting an error, people often behave differently on the next attempt. They may respond more slowly, check additional information, increase attention to the task or change the response itself.
This is behavioural adjustment: a measurable change following the error.
Research on cognitive control has often interpreted such adjustments as evidence that the system is responding to conflict or poor performance. In conflict-monitoring accounts, information about response conflict signals a need for greater control, which can influence later processing and action (Botvinick et al., 2001).
The adjustment does not have to be deliberate. A person may become more cautious without consciously deciding to do so. Attention may narrow. The threshold for responding may rise. A previously ignored cue may receive more weight.
Behaviour can also change before the person can explain why. This is one reason that error monitoring should not be reduced to verbal self-awareness. The ability to describe an error is one form of evidence, but subsequent behaviour may reveal that the event affected the system even when an explicit explanation is absent.
At the same time, behavioural change alone does not prove that the error was correctly understood. A person might slow down because the event was surprising, because confidence fell or because attention was temporarily disrupted. The change shows that the error mattered. It does not yet show how it was interpreted.
Slower is not necessarily better
One of the most frequently studied post-error adjustments is post-error slowing. In many speeded tasks, responses immediately following an error are slower than responses following a correct action.
A common interpretation is that the person becomes more cautious. By delaying the next response, the system may allow more information to accumulate and reduce the risk of another impulsive error.
Sometimes that interpretation is supported by improved accuracy. But slowing does not consistently produce better performance. A slower response can reflect strategic caution, disrupted attention, surprise or temporary disengagement from the task (Danielmeier & Ullsperger, 2011; Notebaert et al., 2009).
An influential integrated account proposes that errors can first trigger a broad interruption of ongoing activity and then, where conditions permit, support more selective task-related adjustment (Wessel, 2018). This should not be treated as a universal sequence for every error. It helps explain why an immediate reaction may be disruptive even when later processing becomes adaptive.
The ambiguity becomes especially important in complex environments. Lewis and Gutzwiller (2023) found that after an error in a multitasking environment, participants could remain excessively focused on the task in which the error occurred and miss a response opportunity in a competing task. The adjustment was related to the error, but it did not improve overall performance.
Post-error slowing is therefore evidence of behavioural change, not a direct measure of successful control. Its meaning depends on what changed alongside the delay.
Did accuracy improve? Was more relevant information considered? Did attention become appropriately focused or excessively narrow? Did performance improve only on one task while deteriorating elsewhere?
Without those comparisons, slower behaviour cannot be classified confidently as better behaviour.
Correction is not the same as learning
An immediate correction can restore performance without creating a durable change.
You may notice a typing mistake, correct it and make the same mistake again a minute later. A driver may take the wrong turn, reverse course and repeat the error on the next journey. A person may recognise that a familiar reaction was unhelpful only after it has already occurred many times.
Correction concerns the present discrepancy. Learning concerns whether relevant processing or behaviour changes beyond that moment.
Evidence of durable learning might include:
- greater accuracy on later attempts;
- earlier detection of the developing error;
- improved discrimination between similar situations;
- more appropriate action under changed conditions;
- reduced dependence on external feedback.
Repeated errors do not necessarily mean that monitoring is entirely absent. A person may detect the result but fail to identify its cause. The relevant cue may not be available early enough to guide action. The correct alternative may be known but difficult to execute under pressure. Feedback may arrive after the same response pattern has already been initiated again.
A correction can also be too narrow. If someone responds to one missed deadline by adding more reminders, the immediate problem may disappear. But if the underlying difficulty is an unrealistic workload, reminders may not change the condition producing the error.
For monitoring to support durable improvement, information about the discrepancy must influence something relevant: what is attended to, how the situation is interpreted, which response is selected or how the next attempt is organised.
Even then, improvement should not be assumed. A person can learn the wrong lesson from an error. One failed attempt may lead to excessive caution, avoidance or the abandonment of an otherwise workable strategy. Behaviour has changed, but the adjustment has been calibrated poorly.
Why error monitoring matters for change
Error monitoring gives a system access to information it could not obtain from successful performance alone. A mismatch reveals something about the relationship between an intended action, the conditions in which it was attempted and the outcome that followed.
That information can interrupt automatic continuation. It can make a relevant cue more noticeable, create uncertainty about the current response or increase the probability that another action will be considered.
But the error does not dictate its own solution.
The same detected discrepancy can lead to careful correction, indiscriminate slowing, distraction, defensive explanation, withdrawal or no usable adjustment at all. Which outcome follows depends on the information available, the standard being applied and the actions the person can access in that moment.
This places error monitoring between performance and change. It is not identical to either.
Without monitoring, behaviour may continue despite repeated mismatch. With monitoring, the mismatch becomes available to influence what happens next. Whether that influence becomes useful adjustment depends on how accurately the error is interpreted and whether the resulting response serves the relevant goal.
The practical significance of an error is therefore not exhausted by the fact that it occurred. The more important sequence is:
What departed from the relevant standard? Was the departure detected? What changed afterwards? Did that change improve performance in the appropriate context? Did the improvement persist?
A mistake can produce a correction without learning. It can produce adjustment without improvement. It can also provide the information from which better behaviour becomes possible.
Error monitoring opens that possibility. It does not guarantee the result.