Conditioning and aversive learning: meaning and application

A brand fixed a problem, yet years later its name still feels unpleasant, even to people who never owned the product. This article explains how conditioning describes such associations and how learning from co-occurrence differs from learning from consequences. You also read what 'negative' and 'social' conditioning are and are not, and how to use them responsibly in brand research.

Conditioning is learning from experience: people and animals learn which events go together (classical conditioning) and what consequences their behaviour has (operant conditioning).

A special form is evaluative conditioning: a brand or object is liked more or less because it was perceived together with something pleasant or unpleasant. 'Negative conditioning' is not an established scientific term. Neurofactor uses it only as an umbrella term for associations that arose from unfavourable experiences. Repetition alone is not enough: what matters is whether one thing predicts the other.

Classical conditioning: learning from co-occurrence

In the early twentieth century, physiologist Ivan Pavlov described how dogs began to salivate at a signal previously given just before food, such as a ticking metronome. He distinguished innate 'unconditioned' from learned 'conditioned' reflexes. One condition he named: the signal must overlap in time with the unconditioned stimulus and should begin just before it (Pavlov, 1927).

Operant conditioning: learning from consequences

From animal experiments, Edward Thorndike formulated the 'law of effect': behaviour followed by satisfaction becomes more firmly connected to the situation, behaviour followed by discomfort less so (Thorndike, 1911). B.F. Skinner developed this into operant conditioning. He distinguished positive reinforcers, whose presentation strengthens behaviour, from negative reinforcers, whose removal does. Punishment he defined separately: removing something positive or presenting something aversive (Skinner, 1953).

Contingency, not just repetition

Robert Rescorla showed that, in the modern view, co-occurrence in time is neither necessary nor sufficient for conditioning. If co-occurrence stays constant but the outcome often also occurs without the signal, little is learned. He described the learning organism as an 'information seeker' that learns relations between events to predict its environment (Rescorla, 1988). The difference between expectation and outcome, the prediction error, plays a central role in later learning models; that concept has its own knowledge base article.

For brands, evaluative conditioning matters most. Jan De Houwer describes it as a change in the liking of a stimulus that results from pairing it with another stimulus. He points out that the term can refer to a procedure, an effect or an explanatory process (De Houwer, 2007). What an association is, you can read in Association: meaning and application.

A meta-analysis of 214 studies found a mean effect of d = 0.52 (95% confidence interval 0.47–0.58): by common rules of thumb, a medium-sized effect on liking. About 70 per cent of the variation between studies reflected real differences in conditions. Effects were stronger with awareness of the pairing, with visible rather than subliminal stimuli, when measured directly after learning rather than after extinction, and with self-report rather than implicit measures (Hofmann et al., 2010). That is an experimental average, not an expected campaign return.

Brand research shows that the procedure matters. Sweldens, Van Osselaer and Janiszewski distinguish an indirect route, in which a memory link between brand and stimulus forms, from a direct transfer of affect. The direct route proved more resilient, for example when a well-known endorser later falls from grace (Sweldens et al., 2010).

In 1982, Gorn showed a pen with music participants liked or disliked and found effects on pen choice (Gorn, 1982). Vermeulen and colleagues repeated it three times: no effect with the original music, a significantly smaller effect with contemporary music, and a null effect overall (Vermeulen et al., 2014).

In learning psychology, 'negative conditioning' is not an established technical term. It usually refers to one of four different things:

• Negative reinforcement: behaviour becomes stronger because it removes something unpleasant (Skinner, 1953).
• Punishment: behaviour is followed by something unpleasant or by the loss of something pleasant (Skinner, 1953).
• Aversive conditioning: a signal comes to predict an unpleasant event and therefore itself triggers a defensive or fear response. Fear conditioning is a widely studied form of this (Olsson & Phelps, 2007).
• Negative evaluative conditioning: a brand or object comes to be liked less because it was perceived together with something negative (Hofmann et al., 2010).

Neurofactor therefore uses 'negative conditioning' only as an umbrella term and editorial working definition: associations with a brand, product or choice that arose from unfavourable experiences. The underlying learning process must be investigated case by case.

Extinction: unlearning is new learning

If the unpleasant event no longer follows, the response usually weakens: extinction. Mark Bouton summarised that extinction does not destroy the original learning; it is new, strongly context-dependent learning. The old response can return in a different setting (renewal), after time has passed (spontaneous recovery) or after a new unpleasant experience (reinstatement) (Bouton, 2004). In evaluative conditioning, too, effects after extinction were on average smaller than directly after learning (Hofmann et al., 2010).

Bandura, Ross and Ross showed that children adopted new behaviour simply by observing an adult, without anyone being rewarded (Bandura, Ross & Ross, 1961). This is called observational or social learning.

This has also been studied for fear. Olsson and Phelps describe how people can learn fear by watching others undergo an unpleasant experience (observational) or simply by being told that something is dangerous (instructed). In their view this largely fits a conditioning model, with extra involvement of systems for social behaviour, language and explicit memory (Olsson & Phelps, 2007). In an fMRI study, the amygdala was active both while participants watched someone receive a shock and later when they expected one themselves (Olsson, Nearing & Phelps, 2007).

Neurofactor uses 'social conditioning' as a working term for adopting associations and evaluations through others, without personal experience. That stories, news or reviews shape brand associations this way is a plausible translation; the studies examined fear of simple stimuli in the lab, not brands.

1. Context colours the brand. Where and with what a brand appears can influence its evaluation; how strongly varies.
2. Repetition without information has limited effect. Repeatedly pairing a brand with something pleasant adds little if people's own experience predicts otherwise.
3. Silence does not remove a negative association. New, reliable experiences produce new learning; the old association can return in another context.
4. Awareness is not a side issue. In the meta-analysis, effects were stronger when people noticed the pairing. Whether evaluative conditioning works entirely without awareness is disputed (De Houwer, 2007). So do not promise unconscious steering.
5. No manipulation. Neurofactor uses knowledge of conditioning to understand where associations come from and to test honest communication, not to steer people with hidden stimuli or fear.

How to research associations and their association strength is explained in How do you measure associations?.

1. Map the associations people have with the brand or choice, positive and negative.
2. Trace their origin: personal experience, other people's stories, news coverage or context such as advertising.
3. Name the suspected learning process using the terms in the table.
4. Choose an intervention that provides new, reliable information, not just repetition.
5. Define outcome measures, control group and measurement points in advance, including a later measurement in a different context.
6. Draw conclusions about behaviour only from behavioural data.

This example is fictional. A few years ago, a bicycle brand recalled one model because of a component. The problem is solved, but potential buyers still often link the name to 'recall' and 'unsafe', even those who never owned one of its bikes.

• Choice question: should the brand let the incident rest, pair the name with atmospheric images, or offer concrete safety information and test rides?
• Available information: qualitative conversations; no data on how common the association is or its origin.
• Suitable approach: measure the associations in three groups: owners who experienced the recall, people who only heard about it and people without prior knowledge. Then compare an atmospheric campaign, a campaign with verifiable safety information and a test-ride offer with a control group. Measure immediately and after a few weeks. There are no results; this is a research plan.
• Possible interpretation: if 'unsafe' mainly appears among those who only heard about it, social transmission probably plays a role. If the association drops right after a test ride but returns later, that fits context-dependent extinction.
• Next step: test the best approach more widely on consideration and purchase.

• Thinking repetition is enough. Co-occurrence without predictive value does little.
• Mixing up terms. Negative reinforcement is not punishment, and 'negative conditioning' is not a technical term.
• Translating laboratory effects to campaigns. An average from experiments predicts nothing about your brand, and a well-known advertising study proved hard to replicate.
• Treating extinction as erasure. An old association can return.

Conditioning describes how people learn from co-occurrence and consequences, and how liking can shift through context or others. Negative reinforcement, punishment, aversive and negative evaluative conditioning are different processes; at Neurofactor, 'negative conditioning' is only an umbrella term. To reduce an aftertaste, first find out where it comes from, offer new and reliable experiences, and test whether the change lasts.