Abstract
Higher nervous activity, which MeSH classifies under mental processes, is Pavlov's name for the integrative work of the cerebral cortex, which his laboratory studied through the conditioned reflex. Later research found that conditioning tracks the contingency between a signal and an outcome, not the count of pairings, although whether contingency is necessary or sufficient has been disputed. Learning came to be described as driven by error, the gap between what was predicted and what happened, and the neural carriers of conditioning have been traced in fear, eyeblink and reward circuits, although what dopamine signals remains disputed. Extinction turned out to be mainly new, context-dependent learning rather than erasure, a result that now shapes exposure therapy for anxiety. Three interactive demonstrations let readers run an error-correction rule, vary the contingency between a signal and its outcome, and trace a generalization gradient.
Keywords: higher nervous activity, conditioned reflex, Pavlovian conditioning, prediction error, extinction
Pavlov treated the conditioned reflex as an experimental window on the brain's highest functions. In a lecture that opened his account of the work, he argued that the cerebral hemispheres earn their place by reacting to signals, events that can stand in for other events, so that a dog comes to salivate to a cue that merely announces food (Pavlov, 1927/2010). The lecture offered no statistics; it summarised some twenty-five years of laboratory work with dogs by more than a hundred collaborators, recorded by an objective method built on the salivary reflex (Pavlov, 1927/2010).
The phrase higher nervous activity came from that school and outlived it. It stayed current in the Eastern European literature that the MeSH note describes; one example is Sokolov's 1963 review in the Annual Review of Physiology, whose title sets higher nervous functions beside the orienting reflex (Sokolov, 1963). Pavlov's life and psychological work are surveyed in a later overview in the American Psychologist (Windholz, 1997). This article follows what became of the programme: the vocabulary of excitation and inhibition, the finding that contingency, not just pairing, matters to conditioning, the error-correction rules that formalised that finding, the circuits that carry conditioned learning, the dispute over dopamine, generalization, and extinction and its clinical uses.
Key Takeaways
- Higher nervous activity is Pavlov's term for the integrative work of the cerebral cortex, studied through signal reflexes, learned responses to cues that announce other events; his doctrine of excitation and inhibition now has historical standing.
- Contingency, how much more likely an outcome is with a signal than without it, is widely accepted as a determinant of conditioning, though whether it is necessary is disputed; when shock was equally likely with or without a tone, no conditioning appeared.
- Error correction, learning in proportion to the gap between the outcome and what was predicted, accounts for blocking, the failure to learn about a new cue when an old one already predicts the outcome; the Rescorla-Wagner model that formalised it is judged a useful heuristic and has since been revised.
- Conditioned learning runs through identified circuits: the amygdala, a temporal-lobe structure, for fear, and the cerebellum, at the back of the brain, for eyeblink conditioning.
- Prediction error, the mismatch between the reward expected and the reward received, was linked to dopamine neurons, but a 2022 mouse study, which tested only the nucleus accumbens core, found that dopamine release fitted a retrospective causal-learning signal better, so the link is contested.
- Extinction, the fading of a learned response once the signal stops predicting the outcome, is mainly new learning tied to context rather than erasure; it underpins exposure therapy, although several strategies built on it have contradictory or unstable evidence.
What Higher Nervous Activity Is
A term used in Eastern European research literature on brain and behavior physiology for cortical functions. It refers to the highest level of integrative function of the brain, centered in the CEREBRAL CORTEX, regulating language, thought, and behavior via sensory, motor, and cognitive processes.
Scope note for the MeSH descriptor Higher Nervous Activity (D006613), U.S. National Library of Medicine.
The term is a historical label, and the MeSH note says as much by tying it to one regional literature. It names a level of function rather than a single process, which is why MeSH files it under mental processes and also under the physiology of the nervous system.
Pavlov built the concept from the reflex. His lecture set fixed, inborn reactions such as instincts beside a second class, the signal reflexes, which form during an animal's life when an indifferent event comes to announce a significant one; he also described an investigatory reaction to novelty, which he called the What-is-it? reflex (Pavlov, 1927/2010). The signal reflexes were, for him, the characteristic work of the cerebral hemispheres, and he took the cerebral cortex to be the seat of the highest nervous activity (Pavlov, 1927/2010). He judged the existing classification of instincts thoroughly inadequate and called for a complete list of the inborn reflexes with an adequate classification (Pavlov, 1927/2010). The procedure he used to study signal reflexes is now called classical conditioning: a conditioned stimulus is presented with an unconditioned stimulus until the first alone evokes a response.
Excitation, Inhibition and Nervous-System Types
In the opening lecture of his account of higher nervous activity, Pavlov sorted reflexes into excitatory and inhibitory kinds and called the mechanism of inhibitions within the central nervous system quite obscure (Pavlov, 1927/2010). The doctrine now has historical standing. Its words survive, but current neural accounts no longer treat conditioning as a general cortical balance: they place it in identified systems, with the amygdala serving fear conditioning and the cerebellum serving eyeblink conditioning, and they find that the plasticity mechanisms in these systems differ considerably (Fanselow & Poulos, 2005).
The Pavlovian school also proposed that individuals differ in basic properties of the nervous system. That idea had a long afterlife in temperament research. The Strelau Temperament Inventory, in its revised 166-item form answered on a 4-point Likert scale, scores people on how mobile their nervous processes are and on how strong their excitatory and their inhibitory processes are, the nervous-system properties as Pavlov understood them (Strelau et al., 1990). Its developers judged the scales' psychometric properties satisfactory in general and recommended the 4-point versions, which had the highest reliability (Strelau et al., 1990). Its developers assumed that the inventory measures the nervous-system properties Pavlov described, although it does so through answers to questionnaire items (Strelau et al., 1990).
Figure 1
A Conditioned Reflex and the Error That Drives It
From Pairing to Contingency
A major change in the study of conditioning was the recognition that pairing alone is not enough. Rescorla argued in 1967 that the usual control groups for Pavlovian experiments were flawed, because they changed the relation between signal and outcome as well as the number of pairings; he proposed a truly random control in which the outcome is equally likely with or without the signal, and he distinguished excitatory from inhibitory contingencies (Rescorla, 1967). The argument was theoretical and was illustrated by only two experimental results (Rescorla, 1967).
A direct test followed. Across two fear-conditioning experiments, Rescorla varied the probability of shock both during a tone and in its absence, using four different probabilities, and measured conditioned suppression (Rescorla, 1968). When shock was as likely without the tone as with it, the tone produced no conditioning even though tone and shock were often paired, and shocks delivered without the tone weakened conditioning (Rescorla, 1968). Twenty years on, Rescorla argued that the traditional picture of conditioning as a reflex stamped in by pairing was inadequate on three counts, and that conditioning is better understood as the learning of relations among events, through which an animal represents its environment (Rescorla, 1988). He added that psychologists had poorly appreciated the change (Rescorla, 1988). On that view, conditioning is one case of association learning rather than a special reflex mechanism. The contingency principle has had critics. In a theoretical review, Papini and Bitterman held that a contingency between signal and outcome is not required for conditioning and cannot by itself produce it, and that too much weight had been placed on the early experiments with random training; they acknowledged, though, that contingency is widely accepted as an empirical generalization (Papini & Bitterman, 1990).
Contingency also bears on how signals shape voluntary action. Rescorla and Solomon's analysis held that Pavlovian conditioning mediates instrumental behaviour, the province of operant conditioning, and separated two processes of acquisition (Rescorla & Solomon, 1967). They judged the evidence from measuring Pavlovian and instrumental responses at the same time to give the mediational hypotheses only weak support at best, whereas interaction studies showed strong Pavlovian control of instrumental responding (Rescorla & Solomon, 1967).
The demonstration below sets the chance of an outcome with and without a signal; it is illustrative, not data.
Pairing is not enough: contingency
Illustrative only: ΔP, the difference between the two probabilities, is a simple index of contingency; the demonstration does not simulate an experiment or show data.
Error-Correction Learning
The Rescorla-Wagner model, which Rescorla and Wagner set out in a 1972 chapter on the theory of Pavlovian conditioning (Rescorla & Wagner, 1972), turned contingency into a learning rule. On each trial, the change in a cue's associative strength is proportional to the difference between the outcome that occurs and the combined prediction of every cue present. Because the error is shared, a cue added to one that already predicts the outcome finds little error left to absorb, which is the model's explanation of blocking. Kamin described blocking in a 1969 chapter on predictability, surprise and attention in conditioning (Kamin, 1969). The model's standing is revised. A review devoted to assessing it, written about 25 years after it appeared, set it beside the Widrow-Hoff delta rule and connectionist networks and concluded that it is a useful heuristic rather than a correct description, with its predictive failures traceable to its assumptions and with little value as a benchmark for other proposals (Miller et al., 1995). HeiDI keeps a pooled error term but adds reciprocal associations, from signal to outcome and from outcome back to signal, to explain why animals differ in whether they approach a cue or the place where food arrives (Honey et al., 2020). The version of that work examined for this article was its bioRxiv preprint.
Other proposals let the learning rate itself change. Mackintosh represented attention as a learning rate specific to each stimulus, its associability, and of two assumptions made by theories of selective attention he judged the first justified by the data (Mackintosh, 1975). Pearce and Hall proposed instead that a conditioned stimulus loses associability once the outcome it precedes is already predicted, while the effectiveness of the reinforcer stays constant (Pearce & Hall, 1980). Drawing on a selective review, Le Pelley proposed a hybrid model, combining changes in a cue's associability with changes in error, that accounts for the effects of a cue's prior training on later learning better than any previous single-process theory (Le Pelley, 2004). Pearce later offered a configural proposal about generalization, in which the whole array of stimuli on a trial is learned as a unit and responding to a new array depends on its similarity to the trained one; he judged earlier treatments inadequate on the basis of a selective review (Pearce, 1987).
Two further lines question whether trial-by-trial error correction is the right description at all. In a timing-based alternative, Gallistel and Gibbon argued that conditioning reflects learned temporal intervals and rates of events, read out by decision processes, stressed a timescale invariance they argued the data show, and contrasted this framework with the associative one (Gallistel & Gibbon, 2000). Gershman's Bayesian reformulation joined a Kalman filter with temporal difference learning and reproduced, in simulation, effects that include blocking, overshadowing, latent inhibition and second-order conditioning (Gershman, 2015). The fit was qualitative only, quantitative tests were left for future work, and in its basic form the reformulation cannot explain learning-rate changes that depend on prediction errors, though the paper sketches an extension resembling the Pearce-Hall model (Gershman, 2015).
The demonstration below runs this article's illustrative error-correction rule: a single cue's strength climbs toward the outcome over trials, and a second cue added to a trained one is blocked.
Error correction and blocking
An illustrative error-correction rule, the article’s own, not fitted data: on each trial every cue present changes by the learning rate times the gap between the outcome and the summed strength of the cues present.
Neural Circuits of Conditioning
Where Pavlov spoke of the cortex as a whole, later work found specific circuits. LeDoux's review placed the amygdala at the centre of the circuits that form and store memories of fear, drawing chiefly on animal work that was then extended to humans, and noted that the cellular and molecular picture was only partly understood (LeDoux, 2000). Maren's review mapped the circuits, the synaptic plasticity and the molecules involved in Pavlovian fear conditioning, while stressing that the molecular machinery was only beginning to be understood (Maren, 2001).
Fear is not the only well-mapped case. A review of mammalian associative learning, built on lesions, stimulation and recording, drug manipulations and mutant mice, identified the amygdala as the hub for fear conditioning and the cerebellum as the hub for eyeblink conditioning (Fanselow & Poulos, 2005). Thompson, using reversible inactivation, recording, lesions and pathway tracing, identified an essential memory trace circuit for eyeblink conditioning and localised its basic memory trace to the cerebellum; traces in the hippocampus he regarded as only putative (Thompson, 2005).
How these circuits relate to felt emotion is disputed. LeDoux and Pine proposed a two-system view in which one set of circuits produces behavioural and physiological responses to threat and another produces the conscious feeling of fear, arguing that running the two together had slowed progress and that threat research had so far brought little clinical gain (LeDoux & Pine, 2016). In an earlier paper, Fanselow proposed that how much fear an animal feels determines which of three defensive modes it enters, from pre-encounter defence to circa-strike defence with a post-encounter mode between them, and that each mode is served by its own circuits, among them the amygdala and regions of the periaqueductal gray; the prominence of freezing in this scheme was described for the rat (Fanselow, 1994). Fanselow's interpretation, his own and drawn mainly from rodent and primate evidence, gives learning the larger role: he argued that learned fear, more than innate recognition of threat, organises defensive behaviour across the stages of threat imminence, and he disputed LeDoux's treatment of emotional experience (Fanselow, 2018). The two positions remain unreconciled.
Prediction Error and Dopamine
The error term of learning rules found an apparent neural counterpart in dopamine. Schultz, Dayan and Montague's synthesis proposed that dopamine neurons report reward prediction errors, the difference between reward received and reward expected, a signal they described as apparently coding error (Schultz et al., 1997). Recordings from single dopamine neurons, with cues signalling how probable a reward was, then found that the phasic response grew steadily as reward probability rose; a distinct response to uncertainty built up until the reward was due, and the authors treated any role it plays in attention or risk-taking as speculative (Fiorillo et al., 2003). The reward prediction error hypothesis is now contested. On the supporting side, optogenetic activation of ventral tegmental dopamine neurons at the moment of reward, in 115 female transgenic rats, drove learning about cues, although the effect of stimulation in a blocking procedure fell short of significance over its three test trials (p = 0.055), stimulation was unilateral and the rescue of extinction was only partial (Steinberg et al., 2013). On the other side, fibre photometry of dopamine release in the nucleus accumbens core of mice, across eleven experimental tests in groups of 8 and 7 animals, fitted a retrospective causal-learning signal better than a temporal-difference prediction error (Jeong et al., 2022). That study measured only the accumbens core; in extinction the dopamine cue response stayed significantly positive after behaviour changed abruptly, although its change-trial comparison was marginal (p = 0.0531) (Jeong et al., 2022). Neither result settles the question for dopamine in other regions. A rival account denies dopamine a role in hedonic impact and in reward learning. Combining a literature review with rat experiments, Berridge and Robinson put forward the hypothesis that dopamine may mediate incentive salience, the wanting of a reward, rather than the reward's hedonic impact or learning about it; in rats with dopamine depleted from the nucleus accumbens and neostriatum, taste reactivity measures showed hedonic responses to be preserved (Berridge & Robinson, 1998). In a 2016 review its authors judged that evidence for the account had continued to grow; they again separated wanting from liking and extended the distinction to addiction, which they treated as an excessive amplification of wanting, not necessarily accompanied by greater liking, arising in susceptible individuals (Berridge & Robinson, 2016).
The dispute is part of a wider conversation about reward processing. A 2019 review in Nature Reviews Neuroscience, titled Believing in dopamine, takes up the relation between dopamine and beliefs (Gershman & Uchida, 2019). Prediction-based ideas about perception are treated separately under predictive coding.
Generalization
A conditioned response spreads to stimuli that resemble the trained one, a phenomenon called stimulus generalization. Shepard proposed a universal law of generalization: the probability of generalizing from one stimulus to another decays exponentially with their distance in psychological space, a regularity he held to be true to a good approximation and derived from the geometry of natural kinds (Shepard, 1987).
Human fear generalization may need more than perceptual distance. Dunsmoor and Murphy argued that people generalize fear through concepts, categories and inductive inference, so that models based on perceptual similarity alone are insufficient and animal models are a limited guide (Dunsmoor & Murphy, 2015). The authors set out questions for future research, and whether fear generalizes at the basic level of categories is not known (Dunsmoor & Murphy, 2015).
The demonstration below plots responding to test tones as they move away from a trained tone; the gradient is an illustrative exponential, not data.
Generalization from a trained tone
An illustrative exponential gradient on a linear frequency axis, not data: real gradients depend on the species, the stimulus dimension and the training procedure.
Extinction and Clinical Reach
When a signal stops predicting its outcome, the conditioned response fades. The inhibitory learning account of extinction holds that this fading reflects new learning that depends on context, laid over the original association rather than erasing it (Bouton, 2004). That review drew on the behavioural literature only and called for models that handle inhibition better (Bouton, 2004). The account remains supported: a 2021 review of Pavlovian and instrumental extinction again treated extinction as new inhibitory learning, while noting that inhibition may differ between the two kinds of learning and that an integrated explanation spanning behaviour and biology is still lacking (Bouton et al., 2021). Fear-conditioning experiments in rats add a complication: extinction delivered inside the reconsolidation window produced a lasting reduction of fear, with much lower renewal, reinstatement and spontaneous recovery than usually follow extinction (Monfils et al., 2009). Pooling sixty-three comparisons from animal and human studies, a later meta-analysis found that extinction given after retrieval outperformed standard extinction overall, but the size of that advantage varied with the paradigm, the effect for fear in animal work was nonsignificant, and the evidence remains preclinical (Kredlow et al., 2016). The Monfils result is hard to square with the view that extinction never touches the original memory, so the inhibitory account leaves room for some unlearning.
Relapse is the practical signature of this view. Extinguished behaviour can return when the context changes (renewal), with the passage of time (spontaneous recovery) or after the outcome is presented on its own (reinstatement) (Bouton et al., 2021). A review of animal and human work concluded that extinction is learned readily but retrieved with difficulty and that conditioned fears recover easily, noting that exposure treatment is generally very effective but relapse is not uncommon (Vervliet et al., 2013). Translational work has linked the rodent infralimbic cortex to the human ventromedial prefrontal cortex as a homologous site for recalling extinction; the same review reported a link to the dorsal anterior cingulate cortex that has a contrary report against it, and described itself as not comprehensive (Milad & Quirk, 2012).
Craske and colleagues drew eight strategies for exposure from the inhibitory learning account and illustrated them with five clinical cases, while reporting that findings on safety signals were contradictory, effects of multiple contexts unstable and a retrieval cue's effect very weak in one study (Craske et al., 2014). A later review judged extinction a strong translational bridge to anxiety but found the meta-analyses on d-cycloserine mixed and said that many strategies still need clinical validation (Craske et al., 2018). For laboratory studies of human fear conditioning, representatives of fourteen major European laboratories set out methodological guidance on design and analysis, prompted by concern over replicability, offering general recommendations wherever possible (Lonsdorf et al., 2017).
Worked Example
The following calculation is this article's own illustration, not data from any study. It applies the error-correction rule that Miller and colleagues assessed (Miller et al., 1995).
On each trial, every cue present changes in strength by a learning rate times the gap between the outcome and the summed strength of all cues present. The outcome is coded 1 when it occurs. With a learning rate of 0.3 and a single cue A paired with the outcome from a starting strength of 0, the first trial adds 0.3 × (1 − 0) = 0.300. The second adds 0.3 × (1 − 0.300) = 0.210, giving 0.510. Continuing, A reaches 0.657, 0.760 and 0.832 by the fifth trial: each gain is smaller because less error remains.
Blocking follows from the same arithmetic. Suppose A has been trained to 0.95 and a new cue B is now presented together with A, followed by the outcome. The summed prediction is 0.95, so only 0.05 of error remains, and B gains 0.3 × 0.05 = 0.015 on the first compound trial. A gains the same amount, the combined prediction climbs to 0.98, and later trials leave B near zero. Had A and B both been new, each would have gained 0.300 on that first trial. The pretrained cue has used up the error, so B is barely learned about. The Rescorla-Wagner demonstration above uses this rule.
Table 1
Associative Strength Under the Illustrative Error-Correction Rule (Rate 0.3, Outcome Value 1)
| Trial | Strength before | Prediction error | Change (rate 0.3) | Strength after |
|---|---|---|---|---|
| A alone, trial 1 | 0.000 | 1.000 | 0.300 | 0.300 |
| A alone, trial 2 | 0.300 | 0.700 | 0.210 | 0.510 |
| A alone, trial 3 | 0.510 | 0.490 | 0.147 | 0.657 |
| A alone, trial 4 | 0.657 | 0.343 | 0.103 | 0.760 |
| A alone, trial 5 | 0.760 | 0.240 | 0.072 | 0.832 |
| A + B, first compound trial (A pretrained to 0.950) | B: 0.000 | 0.050 | B: 0.015 | B: 0.015 |
Note. Values come from this article's own illustrative calculation, not from data. Each change equals the rate times the prediction error, the gap between the outcome and the summed strength of the cues present; in the last row, cue A already predicts most of the outcome, so cue B learns little.
Discussion
Some conclusions are well established. In Rescorla's experiments, conditioning tracked the contingency between a signal and its outcome rather than the bare number of pairings (Rescorla, 1968), although whether contingency is necessary or sufficient for conditioning has been disputed (Papini & Bitterman, 1990). Conditioned fear depends on amygdala circuits and eyeblink conditioning on the cerebellum (Fanselow & Poulos, 2005). Extinction depends mainly on new inhibitory learning, expressed chiefly in the context where it was acquired, rather than on erasure, although some unlearning is not ruled out (Bouton et al., 2021). Each of these replaced a piece of Pavlov's doctrine with something more specific, which is why higher nervous activity now reads as the name of a research programme rather than a live explanation.
Three questions remain open. The first is what dopamine signals: the error-coding interpretation has causal support from optogenetics in female rats (Steinberg et al., 2013), but a retrospective causal-learning signal fitted accumbens core release better in mice (Jeong et al., 2022). The second is whether trial-based error correction describes learning at all, or whether intervals and rates are what animals learn (Gallistel & Gibbon, 2000). The third is how threat circuits relate to conscious fear, where LeDoux and Pine separate the two (LeDoux & Pine, 2016) and Fanselow stresses learned defence (Fanselow, 2018). Settling the dopamine question would need recordings across several dopamine targets with tasks designed to pull the competing signals apart; settling the timing question would need experiments in which interval learning and error correction make opposite predictions.
Current Directions
The dopamine signal is being re-examined. In mice, accumbens core dopamine release behaved more like a retrospective causal-learning signal than a temporal-difference error, a finding confined to that one region (Jeong et al., 2022). Learning theory is being widened to include associations that run from outcome back to signal, with the aim of explaining individual differences in sign-tracking and goal-tracking (Honey et al., 2020).
Extinction research continues to look for the integrated biobehavioural explanation that a 2021 review found missing (Bouton et al., 2021), while clinical work weighs which extinction-based strategies survive testing in therapy (Craske et al., 2018). The relation between threat circuits and felt fear is argued from both directions (LeDoux & Pine, 2016; Fanselow, 2018). Human fear-conditioning studies now have methodological guidance on design and analysis, drawn up by fourteen European laboratories as a basis for a common framework and meant to improve replicability (Lonsdorf et al., 2017).
Common Misconceptions
- Conditioning is simply the result of pairing a signal with an outcome many times.
- Pairing is not sufficient. When shock was as likely without a tone as with it, the tone produced no conditioning despite frequent pairings (Rescorla, 1968). The belief persists because, in ordinary procedures, pairing and contingency rise together.
- Extinction erases what was originally learned.
- On the inhibitory learning account, extinction is new learning that depends on context, laid over the original association rather than replacing it (Bouton, 2004). The response does disappear in the extinction setting, which is why erasure seems plausible, but it can return when the context changes.
- Higher nervous activity is an outdated Soviet label with no modern descendants.
- The label is historical, but the programme it named continues as the neuroscience of associative learning, which has located fear conditioning in the amygdala and eyeblink conditioning in the cerebellum (Fanselow & Poulos, 2005).
- Dopamine is the brain's pleasure or reward chemical.
- The influential proposal was that dopamine neurons signal the error in predicting reward, the difference between reward received and reward expected (Schultz et al., 1997). Even that proposal is now contested, as the section on prediction error describes. A further account ties dopamine to wanting a reward rather than liking it, which would make it something other than a plain pleasure signal (Berridge & Robinson, 1998).
Glossary
- Associability.
- The rate at which a particular stimulus enters into associations; in some proposals it changes with experience, for example falling once the outcome is already predicted.
- Blocking.
- The failure to learn about a new cue when it is added to a cue that already predicts the outcome.
- Conditioned inhibition.
- Learning that a signal predicts the absence of an outcome, so that the signal comes to suppress the conditioned response.
- Conditioned reflex.
- Pavlov's signal reflex: a response acquired during an animal's life to a cue that announces a significant event.
- Conditioned stimulus.
- An initially neutral signal, such as a tone, that comes to evoke a response through its relation to an unconditioned stimulus.
- Contingency.
- The degree to which an outcome is more, or less, likely in the presence of a signal than in its absence.
- Extinction.
- The fading of a conditioned response when the signal is presented without its outcome; it is thought to reflect mainly new, context-dependent learning rather than erasure.
- Generalization gradient.
- The decline in responding as test stimuli differ more from the trained stimulus, proposed to be approximately exponential in psychological distance.
- Higher nervous activity.
- Pavlov's term, kept in Eastern European literature and in MeSH, for the integrative functions of the cerebral cortex as studied through conditioned reflexes.
- Prediction error.
- The difference between the outcome that occurs and the outcome predicted; error-correction rules learn in proportion to it, and dopamine neurons have been proposed, contestedly, to signal it for reward.
- Reinstatement.
- The return of an extinguished response after the outcome is presented on its own.
- Renewal.
- The return of an extinguished response when the context differs from the one in which extinction took place.
- Spontaneous recovery.
- The return of an extinguished response after time has passed since extinction.
- Unconditioned stimulus.
- An event, such as food or shock, that evokes a response without prior learning.
Key Researchers
Mark E. Bouton
(University of Vermont). Psychologist who studies extinction as new learning that depends on context, including renewal and relapse. Faculty page · Google Scholar
Michelle G. Craske
(University of California, Los Angeles). Clinical psychologist who developed an inhibitory learning approach to exposure therapy for anxiety. Faculty page · Wikipedia · Wikidata
Michael S. Fanselow
(University of California, Los Angeles). Neuroscientist known for threat-imminence theory of defensive behaviour and for work on the neural basis of fear conditioning. ORCID · Faculty page · Wikipedia · Wikidata
Joseph E. LeDoux
(New York University). Neuroscientist who traced the amygdala circuits of fear conditioning and argues that defensive circuits and conscious fear should be kept apart. Wikipedia · Wikidata · Faculty page
Stephen Maren
(University of Illinois Urbana-Champaign). Neuroscientist who studies the neural mechanisms of Pavlovian fear conditioning and its extinction. ORCID · Faculty page · Wikipedia · Wikidata
Ivan P. Pavlov
(Institute of Experimental Medicine, St Petersburg; 1849–1936). Physiologist whose laboratory studies of the conditioned reflex in dogs founded the study of higher nervous activity. Wikipedia · Wikidata
Robert A. Rescorla
(University of Pennsylvania; 1940–2020). Psychologist who showed that conditioning tracks the contingency between a signal and its outcome, and co-authored the Rescorla-Wagner model. Wikipedia · Wikidata · Faculty page
Wolfram Schultz
(University of Cambridge). Neuroscientist whose recordings from dopamine neurons tied their firing to errors in the prediction of reward. ORCID · Faculty page · Wikipedia · Wikidata
Frequently Asked Questions
What is higher nervous activity?
It is Pavlov's term for the integrative work of the cerebral hemispheres, which he saw as reacting to signals that stand for other events and studied through conditioned reflexes in dogs (Pavlov, 1927/2010).
Is higher nervous activity still an active research topic?
The label is historical, but its subject continues as the neuroscience of associative learning, which assigns fear conditioning to the amygdala and eyeblink conditioning to the cerebellum (Fanselow & Poulos, 2005).
Does pairing a signal with an outcome guarantee conditioning?
No. In fear-conditioning experiments, a tone produced no conditioning when shock was equally likely with or without it, even though tone and shock were often paired (Rescorla, 1968).
How is the Rescorla-Wagner model regarded today?
A major review judged it a useful heuristic rather than a correct description of learning and traced its predictive failures to its assumptions (Miller et al., 1995).
Do dopamine neurons signal reward prediction errors?
The idea is contested. In mice, dopamine release in the nucleus accumbens core fitted a retrospective causal-learning signal better than a temporal-difference error, although only that region was tested (Jeong et al., 2022).
Which brain structure is central to conditioned fear?
Reviews of fear conditioning place the amygdala at the centre of the circuits that form fear memories, an inference drawn mainly from animal work and extended to humans (LeDoux, 2000).
Does extinction erase a learned fear?
Mostly not. On the inhibitory learning account, extinction is new learning that depends on context and overlays the original association, which is why fear can return when the context changes (Bouton, 2004).
How is extinction research used in therapy?
It informs exposure therapy for anxiety, from which several strategies have been derived, although evidence for some of them, such as safety signals and multiple contexts, is contradictory or unstable (Craske et al., 2014).
References
Berridge, K. C., & Robinson, T. E. (2016). Liking, wanting, and the incentive-sensitization theory of addiction. American Psychologist, 71(8), 670–679. https://doi.org/10.1037/amp0000059
Berridge, K. C., & Robinson, T. E. (1998). What is the role of dopamine in reward: Hedonic impact, reward learning, or incentive salience? Brain Research Reviews, 28(3), 309–369. https://doi.org/10.1016/s0165-0173(98)00019-8
Bouton, M. E. (2004). Context and behavioral processes in extinction. Learning & Memory, 11(5), 485–494. https://doi.org/10.1101/lm.78804
Bouton, M. E., Maren, S., & McNally, G. P. (2021). Behavioral and neurobiological mechanisms of Pavlovian and instrumental extinction learning. Physiological Reviews, 101(2), 611–681. https://doi.org/10.1152/physrev.00016.2020
Craske, M. G., Hermans, D., & Vervliet, B. (2018). State-of-the-art and future directions for extinction as a translational model for fear and anxiety. Philosophical Transactions of the Royal Society B: Biological Sciences, 373(1742), 20170025. https://doi.org/10.1098/rstb.2017.0025
Craske, M. G., Treanor, M., Conway, C. C., Zbozinek, T., & Vervliet, B. (2014). Maximizing exposure therapy: An inhibitory learning approach. Behaviour Research and Therapy, 58, 10–23. https://doi.org/10.1016/j.brat.2014.04.006
Dunsmoor, J. E., & Murphy, G. L. (2015). Categories, concepts, and conditioning: How humans generalize fear. Trends in Cognitive Sciences, 19(2), 73–77. https://doi.org/10.1016/j.tics.2014.12.003
Fanselow, M. S., & Poulos, A. M. (2005). The neuroscience of mammalian associative learning. Annual Review of Psychology, 56, 207–234. https://doi.org/10.1146/annurev.psych.56.091103.070213
Fanselow, M. S. (1994). Neural organization of the defensive behavior system responsible for fear. Psychonomic Bulletin & Review, 1(4), 429–438. https://doi.org/10.3758/bf03210947
Fanselow, M. S. (2018). The role of learning in threat imminence and defensive behaviors. Current Opinion in Behavioral Sciences, 24, 44–49. https://doi.org/10.1016/j.cobeha.2018.03.003
Fiorillo, C. D., Tobler, P. N., & Schultz, W. (2003). Discrete coding of reward probability and uncertainty by dopamine neurons. Science, 299(5614), 1898–1902. https://doi.org/10.1126/science.1077349
Gallistel, C. R., & Gibbon, J. (2000). Time, rate, and conditioning. Psychological Review, 107(2), 289–344. https://doi.org/10.1037/0033-295x.107.2.289
Gershman, S. J., & Uchida, N. (2019). Believing in dopamine. Nature Reviews Neuroscience, 20(11), 703–714. https://doi.org/10.1038/s41583-019-0220-7
Gershman, S. J. (2015). A unifying probabilistic view of associative learning. PLoS Computational Biology, 11(11), e1004567. https://doi.org/10.1371/journal.pcbi.1004567
Honey, R. C., Dwyer, D. M., & Iliescu, A. F. (2020). HeiDI: A model for Pavlovian learning and performance with reciprocal associations. Psychological Review, 127(5), 829–852. https://doi.org/10.1037/rev0000196
Jeong, H., Taylor, A., Floeder, J. R., Lohmann, M., Mihalas, S., Wu, B., Zhou, M., Burke, D. A., & Namboodiri, V. M. K. (2022). Mesolimbic dopamine release conveys causal associations. Science, 378(6626), eabq6740. https://doi.org/10.1126/science.abq6740
Kamin, L. J. (1969). Predictability, surprise, attention, and conditioning. In B. A. Campbell & R. M. Church (Eds.), Punishment and aversive behavior (pp. 279–296). Appleton-Century-Crofts.
Kredlow, M. A., Unger, L. D., & Otto, M. W. (2016). Harnessing reconsolidation to weaken fear and appetitive memories: A meta-analysis of post-retrieval extinction effects. Psychological Bulletin, 142(3), 314–336. https://doi.org/10.1037/bul0000034
LeDoux, J. E., & Pine, D. S. (2016). Using neuroscience to help understand fear and anxiety: A two-system framework. American Journal of Psychiatry, 173(11), 1083–1093. https://doi.org/10.1176/appi.ajp.2016.16030353
LeDoux, J. E. (2000). Emotion circuits in the brain. Annual Review of Neuroscience, 23, 155–184. https://doi.org/10.1146/annurev.neuro.23.1.155
Le Pelley, M. E. (2004). The role of associative history in models of associative learning: A selective review and a hybrid model. The Quarterly Journal of Experimental Psychology Section B, 57(3b), 193–243. https://doi.org/10.1080/02724990344000141
Lonsdorf, T. B., Menz, M. M., Andreatta, M., Fullana, M. A., Golkar, A., Haaker, J., Heitland, I., Hermann, A., Kuhn, M., Kruse, O., Meir Drexler, S., Meulders, A., Nees, F., Pittig, A., Richter, J., Römer, S., Shiban, Y., Schmitz, A., Straube, B., . . . Merz, C. J. (2017). Don't fear 'fear conditioning': Methodological considerations for the design and analysis of studies on human fear acquisition, extinction, and return of fear. Neuroscience & Biobehavioral Reviews, 77, 247–285. https://doi.org/10.1016/j.neubiorev.2017.02.026
Mackintosh, N. J. (1975). A theory of attention: Variations in the associability of stimuli with reinforcement. Psychological Review, 82(4), 276–298. https://doi.org/10.1037/h0076778
Maren, S. (2001). Neurobiology of Pavlovian fear conditioning. Annual Review of Neuroscience, 24, 897–931. https://doi.org/10.1146/annurev.neuro.24.1.897
Milad, M. R., & Quirk, G. J. (2012). Fear extinction as a model for translational neuroscience: Ten years of progress. Annual Review of Psychology, 63, 129–151. https://doi.org/10.1146/annurev.psych.121208.131631
Miller, R. R., Barnet, R. C., & Grahame, N. J. (1995). Assessment of the Rescorla-Wagner model. Psychological Bulletin, 117(3), 363–386. https://doi.org/10.1037/0033-2909.117.3.363
Monfils, M.-H., Cowansage, K. K., Klann, E., & LeDoux, J. E. (2009). Extinction-reconsolidation boundaries: Key to persistent attenuation of fear memories. Science, 324(5929), 951–955. https://doi.org/10.1126/science.1167975
Papini, M. R., & Bitterman, M. E. (1990). The role of contingency in classical conditioning. Psychological Review, 97(3), 396–403. https://doi.org/10.1037/0033-295x.97.3.396
Pavlov, I. P. (2010). Conditioned reflexes: An investigation of the physiological activity of the cerebral cortex (G. V. Anrep, Trans.). Annals of Neurosciences, 17(3), 136–141. (Original work published 1927) https://doi.org/10.5214/ans.0972-7531.1017309
Pearce, J. M. (1987). A model for stimulus generalization in Pavlovian conditioning. Psychological Review, 94(1), 61–73. https://doi.org/10.1037/0033-295x.94.1.61
Pearce, J. M., & Hall, G. (1980). A model for Pavlovian learning: Variations in the effectiveness of conditioned but not of unconditioned stimuli. Psychological Review, 87(6), 532–552. https://doi.org/10.1037/0033-295x.87.6.532
Rescorla, R. A., & Solomon, R. L. (1967). Two-process learning theory: Relationships between Pavlovian conditioning and instrumental learning. Psychological Review, 74(3), 151–182. https://doi.org/10.1037/h0024475
Rescorla, R. A., & Wagner, A. R. (1972). A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement. In A. H. Black & W. F. Prokasy (Eds.), Classical conditioning II: Current research and theory (pp. 64–99). Appleton-Century-Crofts.
Rescorla, R. A. (1967). Pavlovian conditioning and its proper control procedures. Psychological Review, 74(1), 71–80. https://doi.org/10.1037/h0024109
Rescorla, R. A. (1988). Pavlovian conditioning: It's not what you think it is. American Psychologist, 43(3), 151–160. https://doi.org/10.1037/0003-066x.43.3.151
Rescorla, R. A. (1968). Probability of shock in the presence and absence of CS in fear conditioning. Journal of Comparative and Physiological Psychology, 66(1), 1–5. https://doi.org/10.1037/h0025984
Schultz, W., Dayan, P., & Montague, P. R. (1997). A neural substrate of prediction and reward. Science, 275(5306), 1593–1599. https://doi.org/10.1126/science.275.5306.1593
Shepard, R. N. (1987). Toward a universal law of generalization for psychological science. Science, 237(4820), 1317–1323. https://doi.org/10.1126/science.3629243
Sokolov, E. N. (1963). Higher nervous functions: The orienting reflex. Annual Review of Physiology, 25, 545–580. https://doi.org/10.1146/annurev.ph.25.030163.002553
Steinberg, E. E., Keiflin, R., Boivin, J. R., Witten, I. B., Deisseroth, K., & Janak, P. H. (2013). A causal link between prediction errors, dopamine neurons and learning. Nature Neuroscience, 16(7), 966–973. https://doi.org/10.1038/nn.3413
Strelau, J., Angleitner, A., Bantelmann, J., & Ruch, W. (1990). The Strelau Temperament Inventory–Revised (STI-R): Theoretical considerations and scale development. European Journal of Personality, 4(3), 209–235. https://doi.org/10.1002/per.2410040304
Thompson, R. F. (2005). In search of memory traces. Annual Review of Psychology, 56, 1–23. https://doi.org/10.1146/annurev.psych.56.091103.070239
Vervliet, B., Craske, M. G., & Hermans, D. (2013). Fear extinction and relapse: State of the art. Annual Review of Clinical Psychology, 9, 215–248. https://doi.org/10.1146/annurev-clinpsy-050212-185542
Windholz, G. (1997). Ivan P. Pavlov: An overview of his life and psychological work. American Psychologist, 52(9), 941–946. https://doi.org/10.1037/0003-066x.52.9.941