Learning

Change in performance or behaviour through experience, distinguished from immediate sensory responses, fatigue and maturation.

A sensitive period is an interval of development during which a particular kind of experience has an unusually strong effect on subsequent behaviour or nervous-system organization. It is specific to a function and experimental outcome: susceptibility of visual acuity and susceptibility of a recorded neuron's response need not have identical schedules. Such timing is one aspect of neuroplasticity.

A revealing experiment varied both the age at which a kitten's eyelids were sutured shut on one side and the duration of closure. The investigators later compared vision through the two eyes using graded behavioural tasks. Starting deprivation at different ages produced different lasting effects, and longer early deprivation affected more demanding tasks. This separates the timing of experience from merely its total absence.^[ref]

The same paper reported some improvement when subsequent experience forced use of the previously deprived eye, including in one older animal. A sensitive period can therefore coexist with later improvement. Its experimental meaning depends on the measured behaviour, the timing and duration of deprivation, and the opportunity for recovery.^[ref]

Expertise is reliably superior performance in a particular domain of activity, acquired through substantial learning and practice. It involves effective recognition and use of task-relevant structure, rather than simply knowing many disconnected facts or being good at every unfamiliar task.

A classic experiment asked chess players of different strengths to reconstruct positions after a five-second view. Skilled players performed much better for meaningful game positions; the advantage greatly diminished when the same pieces were arranged randomly. William G. Chase and Herbert A. Simon interpreted reconstruction patterns in terms of learned groups of related pieces.^[ref]

The contrast is illuminating because the number of visible pieces need not change. What changes is whether prior learning supplies useful structure. Measuring expertise therefore requires specifying the task and the conditions under which an advantage is demonstrated.

Learning relationships between sensory events and one’s own actions. Correlated observation and execution can shape a mapping without a special-purpose innate matching mechanism.

Training in which one observed action is consistently paired with a different executed action. Subsequent changes in matching-related responses test plasticity of the sensorimotor mapping.

A learned relationship between a visual event and a motor representation. A matching relationship and a deliberately mismatching relationship can both be learned.

Acquiring information about particular encountered items. It can coexist with abstraction, but improvement on practised items does not establish generalisation to new combinations.

Acquisition of ordered relations among events or actions. A sequence-learning deficit can reflect problems with perception, motor execution, memory or learning itself; control tasks determine which interpretation is warranted.

Experience-dependent changes supporting acquisition or improvement of movement skills. Baseline execution, rate of improvement, retention and transfer should be distinguished when testing a learning impairment.

Planning and executing learned mouth and facial movements. Speech and nonspeech versions can constrain whether an impairment concerns articulatory sequencing, broader motor planning or additional language processes.

An increase in preference associated with prior exposure. Preference and explicit recognition memory can dissociate. This does not by itself determine whether evaluation of the stimulus's significance occurred.

A proposed biological bias that makes some associations easier to acquire than others under comparable experience. Prepared learning does not mean that the resulting response is already present before learning, nor that it is unchangeable.

Learning through observation of another organism rather than only through direct exposure to the relevant consequences. Observational conditioning is one form.

Learning relations among events, actions or outcomes. Classical conditioning and instrumental conditioning provide complementary preparations for testing what information was acquired and how it controls responding.

Learning after a previously established stimulus–outcome or action–outcome relation is reversed. Acquisition, retention and perseveration controls help distinguish flexibility from general learning or performance deficits.

Experience-dependent improvement in a perceptual task. Training can alter sensory representations, attention, decisions or motor responses. Transfer, retention and matched training controls help identify what changed.

Learning that suppresses or competes with an established response or expectation without necessarily erasing the original association. An exposure-based treatment can therefore need practice across contexts and a plan for retrieving the new learning after fear returns.^[ref]

A reduction in a response following repeated stimulation, distinguished from sensory adaptation or motor fatigue. A decrease in anxiety during exposure can occur, but neither defines all therapeutic learning nor guarantees its transfer to another context.

Reduction in responding from one session to another. Comparisons should account for changes in task difficulty and context rather than interpreting lower distress during an easier task as equivalent learning.

Reduction in responding during a session. Its measurement depends on the response and time window; a distress trajectory is not automatically an index of lasting clinical change.

A learning capacity applicable across more than one content domain. Similar performance with speech and tone sequences argues against exclusive linguistic applicability of the measured process, without showing that every representation or neural implementation is identical.^[ref]

Learning from patterns of occurrence, covariation or sequential dependence in experience. A behavioural result can constrain which regularities were learned without uniquely identifying whether the mechanism used associations, stored chunks or hypothesis testing. Distributional information can be exploited by a general algorithm operating over specialised representations.

Development of the representations and processes used to understand and produce a language. Evidence about recognising a sound pattern, learning its meaning and generalising grammatical structure addresses different components of the problem.

Knowledge and skill used to understand and produce spoken language. Performance in one language or dialect should not be treated as a complete measure of a child's language competence across contexts.

Teaching intended to improve spoken-language knowledge or its use. A programme can affect reading comprehension, but its package effect does not isolate a single language component.^[ref]

The age at which learning of a language begins, distinguished from current proficiency and duration of use. Delayed access to any first language can affect later language processing and is not equivalent to deafness.

Ability to attend to and reason about properties of language itself. Improvement in this relatively language-proximal ability is a different hypothesis from broad transfer to every executive function.

Knowledge and use of two languages, varying in proficiency, acquisition age and social setting. Treating it as a single binary exposure can obscure which experiences a cognitive hypothesis actually concerns.

Use of a signed and a spoken language. Within-person comparisons can reduce some between-group confounding while leaving differences in sensory input, movement and language structure.

A proposed benefit associated with bilingual experience for a stated outcome. Language proficiency, metalinguistic awareness, domain-general task performance and later clinical symptoms are different outcomes and require separate evidence.

Persistent developmental difficulty with language associated with functional impact. It is not interchangeable with motor speech apraxia or stuttering, and a rare single-family genetic finding does not establish a universal cause.

Speech addressed to an infant, often differing in prosody from adult-directed speech. A listening preference is an experimentally measurable tendency whose magnitude can depend on the stimuli, the infant’s experience and the measurement method; it is not a universal individual trait.^[ref]

The position that some language-relevant structure or learning constraints are present prior to the relevant linguistic experience. It is not equivalent to denying all learning; the specificity and content of the proposed initial endowment require evidence.

Word learning is acquisition of associations and rules connecting linguistic forms to meanings and uses. Recognizing a familiar sound, inferring a referent and generalizing a word to new situations are distinct tasks within language acquisition.

Learning word–referent mappings by integrating information across individually ambiguous encounters. Recurrent co-occurrence can favour a mapping even when no single encounter uniquely identifies it. A target-looking preference need not distinguish parallel associative accumulation from testing and revising a candidate hypothesis.^[ref]

Use of distributional information in acquiring speech-sound distinctions, recurring units, dependencies or word–referent correspondences. These levels require different statistics, and success at one does not show that the complete natural-language acquisition problem has been solved.

Learning a relation between elements separated by intervening material. In an artificial $aXb$ system, increasing variability in $X$ can help expose the stable $a$–$b$ relation. Such a relation need not be hierarchical syntax.^[ref]

Identifying candidate units in continuous speech. Sequential predictability, prosody and other cues may support different boundaries; discrimination of a familiarised unit does not alone demonstrate that it has been assigned a meaning.

Experience-dependent changes in grouping or discriminating speech sounds. Comparing exposure distributions while holding the token range fixed can test a contribution from distributional information.^[ref]

Learning involving a relation between an action and its consequences. In a mobile contingency task, an infant’s kicking causes movement of a mobile; later kicking can be used as a retention measure with appropriate baseline and noncontingent controls.

Escape learning is learning an action that terminates an already-present aversive event. If making the response becomes more likely because it ends the event, the contingency is negative reinforcement. It differs from avoidance learning, in which an action prevents or postpones the event before it begins.

For example, compare a response that stops an ongoing unpleasant sound with one that prevents the sound after a warning cue. The physical response could be identical while its relation to the event differs. A yoked control can receive the same event timings without controlling them, separating exposure from controllability. Classic learned-helplessness experiments used such differences in prior controllability to study later escape performance. ^[ref] Performance also depends on sensory, motor and motivational conditions, not learning alone.

A consequence that decreases the future occurrence of a response. It is distinguished from negative reinforcement, which increases a response by removing or preventing an aversive event.

Arranging consequences contingent on clearly specified behaviour to change its future occurrence. The target, measurement, timing and alternative responses determine what is reinforced; reinforcement is not identical to punishment.

Analysis of antecedents and consequences that maintain a specified behaviour. A useful formulation predicts what changes when a relevant contingency is manipulated, rather than merely giving the behaviour a new label.

Learning actions that prevent or escape an anticipated aversive outcome. In anxiety, immediate relief and the absence of disconfirming experience can maintain avoidance even when the feared prediction is inaccurate.

An action reducing contact with an anticipated aversive situation or experience. It can provide immediate relief while limiting corrective learning, but avoidance of genuine danger is not thereby pathological.

A precaution intended to prevent a feared outcome or conceal anxiety. It can maintain fear when safety is attributed to the precaution, when it blocks informative experience, or when it produces misleading symptoms. Whether a behavior has this function depends on context; objectively useful precautions are not inherently pathological.

Use of a precaution in a way intended to support engagement while preserving useful learning. Its value cannot be settled from its label; prediction testing, attribution of safety, replacement precautions and longer-term transfer should be measured. An undergraduate contamination experiment showed why reduction in symptoms need not be logically impossible in the presence of a safety behavior.^[ref]

Seeking another person's confirmation of safety or certainty. In an anxiety-maintaining cycle, immediate relief can reinforce repeated requests without changing the underlying interpretation. This function should be distinguished from appropriate information seeking or clinical assessment.

Learning from rewarding outcomes to guide subsequent choices. A reduced change in response bias can reflect altered outcome sensitivity, integration over trials, exploration or performance; it does not uniquely diagnose reduced experienced pleasure.

An outcome that is valued or promotes approach under specified conditions. Its hedonic impact, learned predictive value and ability to motivate action can dissociate; no single behavioural measurement identifies all three.

A behavioural phenomenon in which prior exposure to uncontrollable aversive events impairs subsequent escape or coping. The original explanation invoked learning that action and outcome are independent; later accounts emphasize learning control over stress responses. The phenomenon is an animal model relevant to some symptoms, not a complete diagnosis of human depression.

For an action $A$ and an outcome $O$, a descriptive contingency is $\Delta P=P(O\mid A)-P(O\mid \neg A)$. If both probabilities are high, outcomes may be frequent while control is absent. If both are low, outcomes may be rare while control is still absent. For causal control, actions must be experimentally varied or common causes adequately controlled; observational conditional probabilities alone can be confounded.

An increase in a response because it terminates or avoids an aversive outcome. It is not punishment: reinforcement describes the increase in responding, whereas punishment describes a decrease.

A ribbon links an infant’s ankle to a mobile so kicking produces visible movement. Baseline kicking and a later nonreinforced test permit assessment of a learned contingency. A reminder moving the familiar mobile can reactivate later performance; the crucial control asks whether reminder exposure alone teaches the response.^[ref]

An effect of learning or training on performance beyond the practised activity. The similarity of the new task, its stimuli, procedures and representations determines how strong a generalisation claim is warranted.

Improvement in substantially different tasks or domains, such as academic attainment after isolated working memory exercises. Randomised comparisons with credible active controls and prespecified outcome measures make the claim more informative.

Improvement in an untrained activity closely resembling the trained tasks. It can demonstrate more than memorisation of particular items without establishing improvement in distant educational outcomes.

Learning a relation between a cue and an outcome. Prediction-error models explain why a new cue may acquire little value when an established cue already predicts the outcome, the phenomenon of blocking.

Acquisition of a response through observing another individual's reaction to a stimulus. The recipient's later behavior can be tested without the demonstrator present, distinguishing learning from an immediate reaction to the demonstrator.

Transfer of a learned response to stimuli differing from the original cue. Similarity and context determine what is being generalised; a response to a reminder does not require literal identity with the original stimulus.

A model associated with Allan R. Wagner in which stimulus elements occupy an inactive state $I$, primary activation $A_1$, or secondary activation $A_2$. External presentation recruits inactive elements into $A_1$; activity then passes through $A_2$ to inactivity. Learned associates can recruit elements into $A_2$. These are functional states, not established neuronal or receptor identities.

Represent a three-state probability distribution by occupancy fractions $\mathbf{p}=(p_I,p_1,p_2)$ in $I,A_1,A_2$. A nonnegative transfer $f_{ij}\le p_i$ subtracts the same amount from donor $i$ that it adds to recipient $j$. The total therefore remains $p_I+p_1+p_2=1$. If all simultaneous outflows from a state sum to at most its old occupancy, each occupancy remains nonnegative. This elementary conservation argument makes a proposed numerical realization checkable; it does not identify the states with biological compartments.

A development of the model by Allan R. Wagner and Susan E. Brandon distinguishing sensory-specific and emotive components of outcome representations, with potentially different activation time courses and response consequences. It allows one cue to influence multiple response systems differently without requiring contradictory values in one undifferentiated representation.

An extension allowing association changes involving associatively activated, physically absent cues. In the common symmetric formulation, matched $A_1/A_1$ and $A_2/A_2$ states support excitation, while mismatched $A_1/A_2$ and $A_2/A_1$ states support inhibition. The extra rules allow forms of retrospective revaluation; neural reactivation alone does not verify their signs.

A change in responding to an absent target cue after new experience with its previously associated cues. It constrains accounts that permit changes only when the target is physically present.

Training a relation between initially neutral cues before giving one cue biological significance. Responding to the other cue can then depend on inference through the learned structure rather than direct cue–outcome reinforcement.

Reduced learning about an added cue when another cue already predicts the outcome. Control conditions must match exposure and outcome history; blocking in learning is distinct from blocking in experimental design.

Learning through which a cue signals an outcome’s absence or reduction relative to expectation. Summation and retardation tests can help distinguish inhibition from simple low excitatory strength.

Retarded conditioning to a previously nonreinforced cue relative to an appropriately matched novel cue. Changes in attention, retrieval and context can contribute, so it is not uniquely diagnostic of one mechanism.

A response expressed to a cue after learning. It need not be a literal copy of the response directly elicited by the outcome.

An outcome that can elicit a response without the particular cue–outcome learning under study. The term does not imply that every property of the response is immune to previous experience.

A cue whose relation to another event is learned in a conditioning procedure.

Learning through which a cue associated with an aversive event later evokes measured defensive responses. The learned response, the aversive event and subjective experience must be distinguished.

Enhancement of the acoustic startle reflex in the presence of a cue associated with threat. It operationalizes a defensive response rather than directly reading subjective fear.

A defensive response learned through a relation between a cue and an aversive outcome. Laboratory cue responses and the complex fears involved in a clinical anxiety disorder require an explicit translational bridge.

Learning during presentations of a previously predictive cue without the expected outcome. Reduced responding need not mean that the original association has been erased. Context, delay and subsequent experience can alter which association is retrieved.

Retrieval of learning established during extinction, often tested after a delay. Successful immediate extinction need not guarantee later retrieval under different temporal or contextual conditions.^[ref]

Recovery of a conditioned response when the context differs from the context of extinction. It shows that performance after extinction can depend on retrieval conditions.

Pattwell and colleagues compared conditioned responding and extinction across human developmental groups, with parallel mouse investigations. Skin-conductance responses indexed human fear learning. The developmental laboratory pattern motivates clinical hypotheses but does not directly estimate the efficacy of exposure therapy for adolescents.^[ref]

Reappearance of fear after it has decreased. It can reflect a change in retrieval conditions without implying that no useful learning occurred during treatment.

Return of conditioned responding when testing occurs in a context different from the extinction context. It motivates tests of generalisation beyond the treatment setting.

Learning about the consequences of an action, such as lever pressing that produces an outcome. Tests of contingency and outcome devaluation help distinguish action–outcome control from habit-like responding.

A learned tendency to respond that can persist despite a change in the current value of the outcome. Devaluation insensitivity is informative only under conditions in which the outcome change is understood and performance is otherwise possible. It need not imply exceptionally rapid habit acquisition.^[ref]

Action sensitive to both the relationship between an action and its outcome and the outcome's current value. Outcome devaluation and contingency manipulation provide operational tests, subject to task comprehension and performance limitations.

A change in an odor's taste-like qualities or associations after pairing it with a taste. Pre/post odor-only judgments can test transfer beyond the original mixture, while controls are needed for expectation and use of response scales. Such transfer supports learned cross-sensory effects but does not by itself prove a distinct developmental synesthesia mechanism.^[ref]

Changes in predatory responses through experience, including association of prey signals with unprofitability. Experiments must distinguish learning from immediate avoidance and from loss of appetite.

Learning influenced by other individuals or their products. Its mechanism can range from attention to a location to acquisition of more detailed behavioural information.

Reproducing an outcome or environmental result without necessarily copying the observed means. It must be distinguished from imitation when an observer reaches the same goal by another action.

Reproducing an observed action or its organisation. Model-specific matching must be distinguished from increased general activity, emulation of an outcome or a response elicited by many interesting stimuli.

Selective imitation of a demonstrated means in relation to contextual constraints and inferred goals. A behavioural difference can also admit attentional or perceptual explanations; selective copying alone is not a unique signature of explicit belief reasoning.

A compatibility effect in which observed task-irrelevant actions influence action execution. It is not necessarily deliberate copying, wholly uncontrolled behaviour or a pure measure of empathy.

Whether an observed action matches the action a participant must perform, distinguished experimentally from spatial stimulus–response compatibility.

The problem of mapping another person’s perceptible action onto the observer’s own motor commands, despite differences in viewpoint and access to bodily feedback.

Model-specific action matching during the neonatal period. Early positive reports and subsequent disputed replication and reanalysis make this a useful case for distinguishing whether an action occurs, whether it is selectively elicited by its matching model, and whether it supports a broad theory of imitation.^[ref]

Imitation of an observed action after a delay without continuous access to its demonstration. It supplies a nonverbal memory measure, with baseline and control conditions needed to exclude actions likely to occur without the prior demonstration.