Classical Conditioning

A cue appears before an event. Later, the cue itself changes responding.

The familiar explanation is that the two were repeatedly paired until the cue began to trigger the response originally produced by the event. That is a useful introduction to classical conditioning, but it is not an adequate account of the phenomenon.

Repeated pairing matters, but it does not tell us everything about what is learned. A cue can acquire significance because of what it signals, and the response to that cue does not have to reproduce the response to the event itself.

Classical conditioning, also called Pavlovian conditioning, is a form of associative learning in which experience with predictive relationships among events changes the significance of cues and how an organism later responds to them.

Here, predictive does not imply conscious prediction or commit us to one theory of how the relationship is represented. It means that experience changes what the occurrence of one event indicates about another.

Classical conditioning changes what a cue predicts and therefore how that cue can influence later responding.

The basic architecture of classical conditioning

Four terms are traditionally used to describe classical conditioning.

The conditioned stimulus (CS) is a cue whose functional significance changes through learned relations with other events.

The unconditioned stimulus (US)—often described here as the outcome—is the event relative to which that cue learning is being studied. It can produce relevant responding without the particular cue–outcome relation having been learned.

The unconditioned response (UR) is responding to that outcome that does not depend on the specific conditioned cue–outcome relation.

The conditioned response (CR) is responding influenced by the learned relation involving the conditioned cue.

These terms describe roles within a learning relation rather than immutable properties of stimuli or responses.

Introductory examples often begin with a neutral cue that is repeatedly followed by something already significant, such as food. That is an important case, but complete neutrality is not a universal requirement. A cue may already attract attention, evoke behaviour or have a previous learning history before new experience changes its significance. Pavlovian learning can also involve relations among events that were not established through simple direct pairing of a neutral cue with the final outcome.

The central question is therefore not whether a stimulus was perfectly neutral before conditioning. It is how experience changed what the cue signals and how responding to it subsequently changed.

Figure — Classical conditioning as predictive event learning

CONDITIONING EXPERIENCE — cue (CS) → outcome / event (US)

↓ LEARNED PREDICTIVE RELATION — what the cue has come to signal

↓ interacting with context, current state and other learning

↓ CONDITIONED RESPONDING (CR) — anticipatory · physiological · attentional · behavioural

Alongside: OUTCOME / EVENT (US) → UNCONDITIONED RESPONDING (UR)

The conditioned response reflects learning about the predictive cue but need not reproduce the unconditioned response to the outcome. The figure is explanatory rather than a literal serial mechanism.

Pairing alone is not enough

Events that occur near one another in time can become associated, so timing matters. But repeated temporal pairing does not by itself determine classical conditioning.

Consider a cue that is repeatedly followed by an outcome. If the outcome rarely occurs when the cue is absent, the cue provides substantial information about what is likely to happen next. If the same outcome occurs just as frequently without the cue, those cue–outcome pairings provide much less additional information.

Classic contingency experiments demonstrated this clearly: what happened when the cue was absent affected subsequent conditioned responding to it.

This reveals an important distinction between contiguity and contingency.

Contiguity concerns whether events occur close together. Contingency concerns the relationship between them: whether the occurrence of one event changes what can be expected about another.

A cue can therefore be repeatedly paired with an outcome without becoming equally informative in every arrangement. Timing, prior learning and other conditions also matter, but the broader lesson is clear: repeated pairing alone does not adequately describe the learning relation.

Prediction should not be confused with causation. A warning signal can reliably precede an aversive event without producing that event. A place can signal access to food without causing the food to exist.

A conditioned cue can predict an event without causing it.

Acquisition is not the same as response growth

The development of conditioned learning across experience is called acquisition. Researchers commonly study acquisition by measuring changes in conditioned responding.

Those measurements are indispensable, but learning and its expression are not identical.

A conditioned response may vary with the response being measured, the current context, physiological conditions and other aspects of the situation. A larger response can therefore provide evidence about learning without serving as a transparent measure of everything that has been acquired.

This is the same distinction established earlier between learning and performance. Classical conditioning gives it a specific form: experience can change the significance of a cue, while the observable expression of that change remains dependent on the conditions under which it is measured.

The conditioned response does not have to copy the unconditioned response

The familiar diagram of conditioning can encourage a simple inference: the outcome produces an unconditioned response; after conditioning the cue produces a conditioned response; therefore the conditioned response is a learned version of the unconditioned one.

That conclusion does not hold generally.

Different cues predicting the same outcome can produce different forms of conditioned responding. A cue predicting food may produce orientation or approach before food arrives. A cue predicting an aversive event can alter defensive or physiological responding before the event occurs. Conditioning can also affect attention and autonomic activity.

The cue is therefore not necessarily reproducing what the outcome itself does. It can alter responding in relation to what the cue has come to signal.

The conditioned response does not have to be a copy of the unconditioned response.

This is one reason classical conditioning should not be reduced to reflex learning. Some historically important demonstrations involved reflex-like responses, but reflexes do not define the phenomenon. Pavlovian learning can influence orientation, approach, defensive behaviour, physiological activity and other forms of responding.

Because a conditioned cue precedes an expected event, some conditioned responses are anticipatory or preparatory. This does not require every CR to be an optimal preparation for the outcome. It shows that learned responding can be organized around an event before that event occurs.

Classical conditioning is defined by the learned event relation, not by whether the resulting response looks like a reflex.

A response does not tell us everything that was learned

Suppose a cue changes behaviour after conditioning. We now have evidence that prior experience altered the cue's significance.

We do not yet know everything about the information controlling that response.

It is useful to distinguish three aspects of a conditioning experiment: the cue–outcome relation arranged in experience, the relation acquired through that experience, and the conditioned response through which learning is expressed.

These are analytically distinct, not three literal stages through which every instance of conditioning must pass.

The distinction becomes important in outcome-devaluation experiments. Imagine that a cue has been conditioned to predict a particular food. If the value of that food is subsequently reduced, responding to the predictive cue can also decline.

Such findings are difficult to explain using only a fixed connection between the cue and a particular response. In at least some conditioning arrangements, responding depends on information about the specific outcome the cue predicts.

That conclusion has limits. It does not establish that all Pavlovian learning contains the same kind of information or operates through one representational mechanism. It shows why conditioned behaviour cannot always be explained as a simple stimulus–response bond.

A conditioned response can show that a cue acquired significance without revealing everything that was learned.

Conditioning and conscious awareness

Humans can sometimes report explicitly that one cue predicts another event. Researchers can therefore compare conditioned responding with what participants consciously know about the contingency.

The relationship is not uniform across every conditioning procedure or response measure. The evidence does not justify either extreme: classical conditioning cannot simply be defined as conscious learning, but neither should it be characterized universally as unconscious.

Explicit expectancy is one way researchers can assess what participants know about a contingency. It is not the definition of classical conditioning itself.

Classical conditioning is not passive learning

Classical and operant learning are both forms of associative learning, but they organize different kinds of relationships.

In classical conditioning, the central relation is between events: cue → outcome.

In operant learning, the central relation concerns an organism's action and its consequence: action → outcome.

This difference is sometimes described as passive versus active learning. That distinction is misleading. Pavlovian cues can produce active orientation, approach, defensive movement and other organized behaviour.

The classical–operant distinction concerns the relationship being learned, not whether the learner is passive or active.

The distinction is analytical rather than absolute separation. Pavlovian cues can influence actions shaped through operant learning, and multiple forms of learning can contribute to the same behaviour. Operant Learning develops action–outcome learning in its own right.

What classical conditioning can explain

Classical conditioning provides a mechanism through which events acquire new significance because of their learned relationships with other events. Sounds, places, bodily sensations and social or environmental cues can all potentially alter later responding because of what prior experience has made them signal.

That does not mean every response occurring after a cue is classically conditioned.

A cue-linked response might instead reflect instruction, operant learning, an existing response tendency or another learning history. Observing that a cue precedes behaviour therefore does not establish classical conditioning as its cause. Evidence about the relevant learning history is required.

Conditioned responding can also vary across contexts and current conditions. How learned responding generalises across cues and circumstances is developed later in Generalisation and Discrimination.

Another question remains unresolved here. What happens when a conditioned cue repeatedly occurs without the outcome it previously predicted?

A reduction in conditioned responding would be observable, but—as this Concept has established—response change alone cannot tell us everything about the fate of prior learning. That problem belongs to Extinction Is New Learning.

Learned relationships change what cues mean for behaviour

Classical conditioning is larger than the familiar image of a bell producing saliva. Its central phenomenon is a change produced by experience: relationships among events alter the significance of cues, and those cues can then influence later responding.

Temporal pairing can contribute to that process, but pairing alone is not a sufficient description. Predictive contingency matters. A conditioned cue can signal an event without causing it, and the response to that cue need not reproduce the response to the outcome.

Most importantly, three things should remain distinct: the relationship arranged in experience, what is learned from that relationship, and how the learning is expressed.

That distinction prevents conditioned behaviour from being mistaken for a transparent readout of an underlying association.

It also opens the two branches that follow. Operant Learning asks what changes when actions become related to their consequences. Extinction Is New Learning asks what becomes of conditioned responding when the predictive relationship itself changes.