Research figure: B. F. Skinner · Written by Artificial Sapiens
The Three-Term Contingency: Behavior in a World of Changing Conditions
B. F. Skinner's three-term contingency is a functional account of operant behavior in which an antecedent stimulus establishes the occasion for a response, and the consequences of that response alter its future probability under similar conditions. The relation is conventionally represented as discriminative stimulus (SD) → response (R) → consequence (C). The three elements form a history-dependent system: a stimulus acquires discriminative control when reinforcement has been differentially available for responding in its presence. The antecedent does not mechanically compel the response, and the consequence acquires its function through observed effects on behavior. The defining scientific question is how the probability, frequency, distribution, and form of an operant change as a function of the conditions that precede it and the outcomes that follow it. Skinner, 1953; Dinsmoor, 1995a.
Skinner's framework gave researchers a way to analyze orderly relations among context, action, and consequence without relying on a simple stimulus–response reflex model. A green light in an experimental chamber may signal that a pigeon's key peck will produce food, while a different signal marks a period in which the same response has no programmed food consequence. After differential experience, pecking becomes more frequent in one condition than the other. The green light thereby functions as a discriminative stimulus, and the resulting difference in responding is stimulus control. The color has no inherent reinforcing authority; its function belongs to a learned history and a particular behavioral arrangement. Skinner, 1934; Skinner, 1953.
This chapter examines the historical Skinnerian account, the experimental procedures that established discrimination and stimulus generalization as measurable phenomena, subsequent developments by other investigators, and what contemporary learning science adds to the analysis. Its subject is the functional architecture of three-term contingencies and the acquisition of stimulus control, rather than a general introduction to all operant conditioning techniques.
Historical Origins: From the Law of Effect to Experimental Discrimination
Thorndike, Watson, and Skinner's distinctive problem
Edward L. Thorndike's law of effect supplied a crucial predecessor: the consequences of actions affect their subsequent occurrence. John B. Watson's 1913 behaviorist manifesto established a public program for the objective study of behavior, and Pavlovian research provided sophisticated methods for examining stimulus relations. Skinner's particular contribution was an experimentally tractable analysis of operants, responses that could occur without a directly identifiable eliciting stimulus and whose rates could be brought under control through programmed consequences. He replaced the search for a single initiating stimulus with the analysis of an organism's changing behavior in a structured environment. Skinner Foundation, Full Bibliography; Dinsmoor, 1995a.
This shift mattered because the same physical movement could be produced in different circumstances, whereas the scientific unit of analysis was a response class defined by its function in relation to the environment. A lever press, a key peck, or a spoken answer becomes interpretable through its antecedent conditions and its consequences. Skinner's insistence that an operant need not be elicited by a preceding stimulus opened a further research problem: how do antecedent stimuli nevertheless acquire systematic control over operant responding? The development of discriminative control provided the answer within his framework. Skinner, 1935a; Dinsmoor, 1995a.
Early discrimination experiments in the 1930s
In his 1934 Proceedings of the National Academy of Sciences paper, “A Discrimination without Previous Conditioning,” Skinner investigated the establishment of differential responding as part of early operant research. The paper is an important primary source for the experimental origins of stimulus control; it predates his major 1938 synthesis and the later classroom shorthand of the antecedent–behavior–consequence diagram. Its importance lies in demonstrating that discrimination could be studied through the organization of reinforcement conditions and observed changes in responding, not simply assumed to be a fixed perceptual capacity. Skinner, 1934.
His 1935 research “A Discrimination Based upon a Change in the Properties of a Stimulus” extended attention to changing stimulus properties as variables relevant to differential responding. In the same year, “Two Types of Conditioned Reflex and a Pseudo Type” contributed to clarifying the contrast between respondent and operant processes. In 1936 Skinner investigated the reinforcing effect of a differentiating stimulus. These publications show a developing experimental program rather than a single moment at which a complete three-term theory suddenly appeared. Skinner, 1935a; Skinner, 1935b; Skinner Foundation, Full Bibliography.
The 1938, 1953, and 1969 formulations
The Behavior of Organisms: An Experimental Analysis (1938) organized Skinner's early laboratory program around operant behavior, conditioning, extinction, and stimulus control. Science and Human Behavior (1953) developed the larger explanatory significance of operant discrimination and explicitly discussed the three-term contingency using everyday examples such as traffic signals, telephone ringing, verbal responses, and classroom questions. Contingencies of Reinforcement: A Theoretical Analysis (1969) elaborated the conceptual priority of contingency relations in explaining behavioral selection and complex environments. These books represent successive stages of conceptual development; the 1969 work was a theoretical extension, not the first discovery of discrimination. Original publication years and authorship are confirmed by the B. F. Skinner Foundation's bibliography. Skinner Foundation, Books; Skinner Foundation, Full Bibliography.
The often-used acronym ABC—antecedent, behavior, consequence—is a convenient contemporary teaching form of this logic. It should not be mistaken for the totality of Skinner's analysis. The discriminative function of an antecedent depends on prior differential reinforcement, and a full account may also require motivational variables, biological constraints, competing responses, temporal context, and a history extending across many occasions. The functional unit is learned over time, not assembled from three labels at a single instant. Dinsmoor, 1995a; Nevin and Grace, 2000.
The Scientific Structure of a Three-Term Relation
The antecedent: a discriminative stimulus
A discriminative stimulus, commonly written SD, is a stimulus in whose presence a particular operant has had a different history of reinforcement than under relevant alternative conditions. It signals the availability or likely effectiveness of a consequence conditional on responding. A laboratory light can function as SD for key pecking if key pecks have produced food while that light was illuminated. Its role is defined by the contingency and a resulting change in response probability. A signal can be visual, auditory, tactile, social, linguistic, or compound, and the organism may respond selectively to one feature of a complex arrangement. Skinner, 1953; Reynolds, 1961.
An SD differs from a Pavlovian conditioned stimulus that elicits a conditioned response following stimulus–stimulus pairings. In operant discrimination, the antecedent modulates the probability of an emitted response because the response–consequence relation varies across antecedent conditions. Real procedures often contain both Pavlovian and operant processes, however; the distinction is functional, and observing a signal-responsive action by itself does not identify the entire causal mechanism. Modern associative research explicitly considers interactions between stimulus–outcome and action–outcome learning. Rescorla, 1988; de Wit and Dickinson, 2009.
The response: an operant class rather than a single muscle movement
The response term refers to a defined class of behavior. In an experiment, a force-threshold lever press or a detected key peck permits reliable counting; in a classroom, the response might be writing an answer, selecting an option, or asking for clarification. The investigator must specify what qualifies as an occurrence. An operant is defined in relation to its environmental effects and experimental contingencies, so mere similarity of physical movements cannot establish a shared function. A pupil raising a hand to request help and the same pupil raising a hand to stretch are topographically similar but may belong to different functional classes. Skinner, 1938; Skinner, 1953.
Response measurement may include rate per unit of opportunity, latency after stimulus onset, duration, allocation between alternatives, accuracy across trials, or resistance to change. Rate of response was especially important to Skinner's experimental analysis, but it is not sufficient for every scientific question. Discrimination performance requires comparing responding across antecedent conditions while controlling unequal exposure, relevant motivational changes, and reinforcement density. A high overall response rate can coexist with poor stimulus discrimination. Dinsmoor, 1995a; Nevin and Grace, 2000.
The consequence: reinforcement and the selection of future behavior
A consequence is a change in the environment following a response. When its contingent delivery increases the subsequent probability of that response under relevant conditions, the operation has a reinforcing effect. Food delivered after a pigeon's key peck is a common laboratory example, yet whether an event functions as reinforcement is established experimentally. A consequence judged pleasant by an observer is not necessarily a reinforcer for a given organism, behavior, or context. Nor does every temporal succession of a response and an event demonstrate a response–consequence contingency. Researchers distinguish response-dependent delivery from response-independent events and must examine what changes when that dependency is manipulated. Skinner, 1938; Staddon and Simmelhag, 1971.
The tripartite relation is therefore conditional: when stimulus S is present, response R changes the likelihood of consequence C, and the experienced arrangement alters future responding in S. Experimental evidence comes from the difference made by manipulating the conditions. It is scientifically more informative to specify a reinforcement probability given both stimulus and response than to simply claim that a light “causes” an animal to peck. One schematic contrast is P(C | R, SD) versus P(C | R, SΔ), where SΔ identifies a condition in which the specified response no longer produces that programmed reinforcer. These probabilities summarize the programmed arrangement; they do not alone specify the neural or cognitive computations that support learning.
The role of learning history
A stimulus has discriminative control because of differential experience. A traffic signal acquires practical significance through training, law, social consequences, and the driver's learning history; the same red light means different things to a driver, pedestrian, and child who has never encountered traffic rules. In the laboratory, the control relation can be studied by reversing which stimulus signals reinforcement and observing corresponding changes in response allocation. Reversal demonstrates that performance tracks organized contingencies rather than an unchanging preference for one signal. Repeated reversals may also alter learning speed and the range of stimuli that exert control. Dinsmoor, 1995b.
Discriminative Stimuli, S-Delta, and the Limits of the ABC Diagram
In a basic discrimination procedure, reinforcement is arranged for a response in the presence of SD, while the same response is not reinforced in the presence of SΔ. The typical result is greater responding during SD than SΔ. These are functional designations attached to a specified response and reinforcer, not intrinsic properties of a light, tone, or object. SΔ means that a particular reinforcement relation is absent or reduced in the arranged comparison; it does not mean the setting lacks all reinforcement, nor that every response is biologically impossible. Different operants can have different SDs within the same setting. Dinsmoor, 1995a.
The distinction also clarifies punishment and extinction. The absence of a programmed reinforcer in SΔ is an extinction-like arrangement for the response of interest; aversive outcomes, punishment, and escape contingencies require separate analysis. Researchers must distinguish stimulus control established through differentially available reinforcement from suppression resulting from punishment, fear, fatigue, or restricted access to the response device. Similar surface patterns of “doing less” can have materially different functional histories.
A complete antecedent analysis often extends beyond SD. Motivating operations can change the momentary effectiveness of a consequence and the likelihood of behavior that has previously produced it. Hunger alters the value of food relative to satiation; deprivation, fatigue, competing options, and the current cost of responding matter as well. Such variables help explain why the same trained stimulus produces different levels of responding at different times. The three-term framework remains a functional organizing tool while the actual explanation requires the additional relevant variables. Nevin and Grace, 2000.
How Stimulus Control Is Established Experimentally
Differential reinforcement and alternating stimulus conditions
The standard experimental strategy is to define a measurable response, present two or more distinctive antecedent conditions, and arrange different consequence relations across those conditions. In a multiple schedule, successive signaled components may arrange reinforcement in one context and extinction or a different schedule in another. An experimenter records each component's response rate, latency, and distribution. Acquisition is indicated by an orderly divergence between conditions, evaluated over repeated sessions and individual organisms. A convincing demonstration requires stable apparatus operation, calibrated stimulus properties, reliable reinforcement delivery, and enough baseline and training observations to distinguish learning from momentary fluctuation. Skinner, 1938; Dinsmoor, 1995a.
A simplified hypothetical pigeon study illustrates the logic. During a blue keylight, key pecks produce grain according to a specified schedule; during a yellow keylight, pecks do not deliver grain. The sequence and durations of components are programmed and counterbalanced as practical, while response rates are recorded separately. A growing difference between blue and yellow components demonstrates discriminated responding. A reversal, in which yellow becomes the reinforcement-correlated signal, provides a further test of acquired stimulus control. This example is a methodological illustration; its colors, settings, and outcome pattern are not a report of a particular historic Skinner experiment.
Stimulus dimensions, salience, and experimental controls
Stimulus control depends on what the organism can discriminate, attend to, and learn from. Stimulus disparity is the physical separation between relevant cues, while salience concerns the prominence of a cue against background stimulation. A very subtle change in illumination may be harder to use than a conspicuous difference in color or location. James A. Dinsmoor's two-part 1995 review carefully analyzes how discrimination performance varies with such factors and the contingencies that establish control. Crucially, equivalent responding to stimuli does not prove identical sensory processing, and failure to discriminate in one training arrangement cannot establish an absolute sensory incapacity. Dinsmoor, 1995a; Dinsmoor, 1995b.
Experimental design must also rule out inadvertent cues. Food-hopper sounds, experimenter behavior, temporal order, ambient illumination, and spatial location may become controlling stimuli even when an investigator intends to study only a colored key. Randomization, masking, response-independent probes, apparatus calibration, and replication with changed stimulus arrangements help separate programmed control from incidental control. The operative stimulus is the stimulus dimension that actually governs responding, which may differ from the dimension named in the protocol.
Acquisition, baseline, reversal, and generalization tests
An acquisition curve tracks changes as differential training proceeds. A baseline condition permits assessment of preexisting preference, response tendencies, and uncontrolled reinforcement. A reversal test changes the stimulus–reinforcer mapping to establish whether responding follows the new relation. A generalization test presents a range of stimuli after training, often with reinforcement withheld during brief probes to avoid altering the gradient being measured. Each test answers a different question. Steep gradients indicate selective control along the tested dimension; transfer across novel stimuli concerns generalization; rapid reversal concerns adaptation to a changed contingency. None, by itself, identifies all mechanisms of attention, expectancy, or reinforcement learning. Guttman and Kalish, 1956; Dinsmoor, 1995a.
Stimulus Generalization: Why Learned Control Extends Beyond Training
Stimulus generalization occurs when training with one stimulus changes responding to other stimuli that share features or relations with it. The central empirical measure is a generalization gradient: a function relating a stimulus dimension, such as wavelength, pitch, or spatial location, to a behavioral measure such as response rate. The sharper the gradient around a training value, the more selectively responding is allocated along that dimension under the conditions tested. A flat gradient suggests weak differential control by that dimension, though contextual and procedural explanations still matter. Guttman and Kalish, 1956.
Norman Guttman and Harry I. Kalish's 1956 study, “Discriminability and Stimulus Generalization,” helped establish the rigorous experimental analysis of generalization. Their research investigated gradients after operant training, connecting the physical dimensions of stimuli with systematic changes in responding. Later research tested whether prior discrimination training sharpened or displaced those gradients. The resulting literature showed that stimulus control is graded, modifiable, and sensitive to the training history, rather than a binary property in which a response either belongs exclusively to a single cue or never appears elsewhere. Guttman and Kalish, 1956; Hanson, 1959.
Peak shift and the effects of discrimination training
Harley M. Hanson's 1959 study examined generalization gradients after discrimination training with wavelength stimuli in pigeons. Under specific differential-training arrangements, the highest response rate on a post-training test could appear at a stimulus displaced away from the nonreinforced training stimulus rather than at the original positive stimulus. This phenomenon is known as peak shift. Its empirical significance is that learning about one stimulus changes responses to nearby stimuli, with competing influences of excitation, inhibition, comparison, and perceptual representation. The magnitude and even presence of peak shift depend on the design; it is a class of findings, not a universal prediction for every discrimination. Hanson, 1959; Dinsmoor, 1995a.
Auditory generalization and cross-modal scope
Herbert M. Jenkins and Robert H. Harrison's 1960 work examined the effects of discrimination training on auditory generalization. It helped demonstrate that systematic generalization gradients could be investigated with sound frequency, extending the research beyond visual keylight color. The details of any gradient depend on stimulus detectability and training history. Evidence from different sensory dimensions supports the general scientific value of stimulus-control analysis while also warning against treating all modalities as interchangeable. Jenkins and Harrison, 1960; Dinsmoor, 1995b.
Selective Attention and Compound Stimuli
Most real environments present several features at once. A colored illuminated key has wavelength, brightness, position, background contrast, and temporal pattern; a spoken instruction contains content, loudness, speaker identity, emotional tone, and contextual expectations. Differential reinforcement can bring responding under the control of one feature even when several are correlated with the outcome. Determining which feature controls behavior requires experimental probes that separate those dimensions. Reynolds, 1961.
George S. Reynolds's 1961 “Attention in the Pigeon” offered a clear demonstration of selective stimulus control. After training with compound stimuli, different pigeons showed behavioral control by different components of the trained arrangement. Its significance was methodological as well as theoretical: success on a task using a compound signal does not tell an investigator which element of that signal the animal used. Subsequent attention research extended this problem with models of cue competition, predictive validity, and changes in associability. Reynolds, 1961; Mackintosh, 1975.
Nicholas Mackintosh's 1975 theory proposed changes in how readily stimuli enter into associations depending on their predictive history. It belongs to a distinct later associative tradition and is not a theory authored by Skinner. Its relevance is that the same acquired discrimination can be described at different explanatory levels: a functional relation between conditions and behavior, and a model of attentional or associative computations proposed to generate that relation. These approaches motivate different experiments and can sometimes make divergent predictions. Mackintosh, 1975.
Discrimination Learning With and Without Errors
A common misunderstanding is that an organism must repeatedly make incorrect responses for stimulus discrimination to be learned. Herbert S. Terrace's 1963 experiments showed that pigeons could acquire certain discriminations with very few or no recorded responses to the negative stimulus when training began with readily distinguishable cues and stimulus differences were changed progressively. His study “Discrimination Learning With and Without ‘Errors’” established an influential experimental basis for errorless-discrimination procedures. The procedures were developed by Terrace and must be credited to him rather than retroactively attributed to Skinner. Terrace, 1963a.
Terrace's companion study on transferring discrimination across two stimulus continua showed that fading and systematic stimulus arrangements could influence whether training errors occurred. These studies refine the picture of stimulus control: acquisition depends on the sequence, discriminability, and history of stimulus presentations as well as on the ultimate difference between positive and negative cues. Errorless training is therefore an empirically defined class of procedures, not a guarantee that every skill, learner, or setting can be taught without mistakes. Terrace, 1963b.
In education, a related idea is to establish clear cues and then fade prompts so that responding comes under the control of naturally relevant signals. Successful transfer depends on measuring independence in the final context. Responses made only when a teacher points, verbally reminds, or physically guides do not yet demonstrate autonomous control by the intended instructional stimulus. Ethical application requires attention to choice, accessibility, meaningful goals, individual needs, and the difference between teaching a useful skill and imposing arbitrary conformity.
Respondent Conditioning, Operant Discrimination, and Occasion Setting
Skinner's contrast between respondent and operant behavior served to clarify two experimentally distinguishable relations. In a respondent arrangement, stimuli acquire the capacity to elicit responses through relations among events. In an operant arrangement, the occurrence of a response affects its consequences and future probability, with antecedent conditions modulating that relation. A bell followed by food may acquire conditioned eliciting properties, while a bell that indicates when an action produces food can acquire discriminative properties. These functions can coexist in a single stimulus. Skinner, 1935a; Rescorla, 1988.
Modern learning research adds the idea that stimuli can signal when another relation is in effect—often discussed under occasion setting or contextual modulation. This resembles the conditional role of SD at a functional level but arises in partly different experimental literatures, especially Pavlovian conditioning. There is no scientific gain in equating every conditional association with a Skinnerian discriminative stimulus. Careful experiments distinguish what predicts an outcome, what predicts an action's effectiveness, and how context selects among learned relations. Bouton, 2004; Rescorla, 1988.
Extinction, Context, and Changing Stimulus Control
When a response that previously produced reinforcement no longer does so, responding often declines under the altered contingency. But extinction performance cannot be reduced to permanent deletion of a learned relation. Mark Bouton's synthesis of extinction research shows that learning during extinction is often especially context-dependent, and responding may return when context changes or time passes. The cited review focuses substantially on Pavlovian extinction; its lessons about contextual control illuminate related questions in operant learning without turning those procedures into identical paradigms. Bouton, 2004.
This perspective changes how a three-term analysis is applied. A behavior may disappear in one signaled context while remaining probable in another. Apparent learning failure may reflect changed cues; apparent treatment success may be restricted to a training environment. An investigator therefore tests not only whether responding changed at the end of training but also whether behavior generalizes to the circumstances that matter and remains functional when consequences and contexts vary.
Conditional Discrimination and More Complex Stimulus Relations
Ordinary discrimination can involve one signal that correlates with reinforcement for a specific response. Conditional discrimination adds a further relation: which of several responses is appropriate depends on the conjunction of contextual and sample stimuli. In a matching-to-sample task, the significance of comparison stimuli changes with the sample; the correct selection in one trial may be incorrect in another. Such designs became prominent in later experimental behavior analysis, including research on equivalence classes, relational learning, and language-relevant performances. Their experimental development cannot be treated as if every later procedure were already contained in Skinner's 1930s studies.
The broader lesson is that a simple SD–R–C diagram can function as a starting unit without being the complete architecture of complex human cognition. Natural verbal behavior can involve layered cues, conditioned reinforcement, multiple simultaneous contingencies, and rules describing distant consequences. Skinner's own Science and Human Behavior used verbal and educational examples to illustrate discriminative control, but precise explanation of novel language use and relational cognition requires additional evidence and models. Skinner, 1953; de Wit and Dickinson, 2009.
Reinforcement Schedules and the Stability of Discriminated Responding
Discriminative control operates alongside schedules of reinforcement. Under fixed-ratio, variable-ratio, fixed-interval, or variable-interval conditions, responses and reinforcers are distributed differently in time and across opportunities. The frequency of reinforcement can influence response rate, but contextual reinforcement histories can also affect how resistant responding is to disruption. The canonical 1957 Schedules of Reinforcement is a coauthored work by Charles B. Ferster and B. F. Skinner. The central experimental program associated with that book must credit both researchers. Ferster and Skinner, 1957.
John A. Nevin and Randolph C. Grace later developed behavioral momentum theory, examining resistance to change and preference in relation to reinforcement history. Their work extends the study of discriminated operants by separating baseline response rate from persistence when conditions are disrupted. This is a later research development with its own evidence and theoretical commitments; it cannot be inferred merely from listing SD, R, and C. Nevin and Grace, 2000.
Alternative Accounts and Modern Models of Learning
Functional behavior analysis and associative representation
The three-term contingency states a functional relation among discriminative circumstances, responses, and consequences. Modern associative learning models ask what representations and learning rules enable an organism to use those circumstances. A stimulus may come to predict an outcome independently of action; a response may become associated with its specific outcome; learned cues may bias action selection. A functional contingency can be experimentally demonstrated without selecting one unique representational theory. Conversely, an associative model may explain mechanisms not spelled out in the three-term description. Rescorla, 1988; de Wit and Dickinson, 2009.
The advantage of keeping explanatory levels clear is practical. Manipulating the actual response–outcome contingency tests sensitivity to instrumental causation; changing the desirability of an outcome tests its present motivational value; probing new stimuli tests generalization; varying temporal intervals tests timing and prediction. These procedures can yield different results even when a simple observation classifies all the behavior as reinforced.
Goal-directed action and habit
Anthony Dickinson's 1985 analysis of actions and habits proposed experimental distinctions based on sensitivity to current outcome value and instrumental relations. Subsequent work by Sanne de Wit and Dickinson reviewed associative models of goal-directed behavior linking animal learning and human action. Under some conditions, behavior is guided by prospective outcome value; after other histories, a more habitual response tendency may persist despite changes in outcome value. These models enrich the analysis of stimulus control and introduce tests that go beyond observing cue-dependent responding. Dickinson, 1985; de Wit and Dickinson, 2009.
Cue-controlled responding is not synonymous with habit. An animal or person can respond selectively to an antecedent while also evaluating the likely outcome. Likewise, the presence of reinforcement history does not demonstrate that a response is automatic or insensitive to goals. Outcome-devaluation tests, contingency-degradation procedures, and transfer tests distinguish these possibilities better than an isolated ABC observation. This distinction is important whenever a training result is invoked to explain human intention, autonomy, or complex decision-making.
Behavioral neuroscience and computational learning
Contemporary neuroscience studies neural systems involved in action selection, reward prediction, attention, and habits. Computational reinforcement-learning frameworks examine how expected outcomes are updated and how states and actions guide choice. These approaches can model aspects of cue-dependent instrumental learning and help identify mechanisms supporting stimulus control. They are historically independent developments and must be evaluated according to their own experiments and predictive successes. No direct inference from an SD–R–C diagram establishes a specific brain region, dopamine signal, or algorithm in an individual participant. de Wit and Dickinson, 2009; de Wit et al., 2009.
The resulting modern picture is plural: functional environmental analysis, associative theories, attentional models, decision theory, and neuroscience describe related phenomena at different levels. Their compatibility in particular cases is an empirical issue to be tested rather than assumed.
Causality, Experimental Limits, and Replication
A strong contingency claim requires a manipulated relation, a measurable response, appropriate control conditions, and repeated observation. Simply observing that a child obeyed after a request, a customer clicked after a notification, or a pigeon pecked before food appeared does not establish the functional role of the cue. The investigator must compare alternatives: what happens when the cue is absent, when responses no longer produce the consequence, when reinforcement is delivered independently of responding, or when a relevant stimulus feature is altered? Temporal contiguity alone is weak evidence for causation.
John E. R. Staddon and Virginia L. Simmelhag's 1971 reexamination of Skinner's 1948 “superstition” experiment is instructive. Their observations supported alternative interpretations involving interim and terminal activities rather than treating every response near a food delivery as evidence of accidental reinforcement. Though the experiment was not designed as a three-term-contingency primer, it provides a methodological warning about assigning response–consequence functions without adequate controls. Its findings belong to Staddon and Simmelhag and should be distinguished from Skinner's earlier interpretation. Staddon and Simmelhag, 1971.
External validity must also be established. Pigeons' discriminations among keylights illuminate general mechanisms, but human learning occurs amid language, social institutions, unequal histories, many competing reinforcers, and agency. Evidence for a laboratory gradient does not provide a numerical prediction about behavior in classrooms, workplaces, treatment settings, or digital platforms. Generalization across species, tasks, stimulus dimensions, and social contexts requires its own investigations.
Applications in Education and Behavioral Assessment
In education, the three-term framework can help specify when a learned response is expected, whether the instruction is detectable and understood, what response is measured, and what consequence follows. A student who correctly answers a question only when a tutor gives an additional prompt has acquired stimulus control partly by that prompt. Assessment should therefore examine independent responding in the intended circumstances, transfer to unfamiliar examples, and durable performance after assistance is withdrawn. These are measurable design questions, not promises that any one teaching method will work universally.
In behavioral assessment, an antecedent–behavior–consequence record may generate hypotheses about behavior–environment relations. The occurrence of a behavior just before attention, task removal, or access to an object is descriptive information; a functional relation requires further converging evidence and, when safe and appropriate, systematic assessment. A descriptive ABC log is consequently a starting point for inquiry rather than a confirmed diagnosis of a person's motivation, personality, or clinical disorder. Any intervention should be evaluated through contemporary evidence and ethical safeguards, particularly with children and people who rely on caregivers.
The conceptual lineage of functional assessment and applied behavior analysis draws partly on Skinner's experimental program. Modern applied behavior analysis was developed by later researchers, notably Donald M. Baer, Montrose M. Wolf, and Todd R. Risley, together with many subsequent investigators. Specific therapeutic efficacy, comparative benefit, adverse outcomes, and consent standards must be judged from contemporary intervention studies and professional guidance, rather than granted by the historical success of a laboratory discrimination experiment.
Ethical Questions: Control, Choice, and the Design of Environments
Stimulus control is pervasive in ordinary social life: signage structures safe traffic behavior, accessible design makes relevant actions easier to identify, and feedback changes how people learn. Ethical problems arise when control procedures are designed without transparency, consent, opportunity for dissent, or attention to unequal power. The same ability to make a behavior more probable can support a learner's independent access to information or serve institutional objectives that conflict with the learner's interests. Scientific description of what changes responding cannot alone decide which consequences ought to be arranged.
Ethical educational and clinical practice requires meaningful goals, respectful communication, support for autonomy, proportionality, monitoring for distress, and revision when interventions create undesirable effects. A carefully measured increase in compliance does not automatically constitute improved well-being. Neither the three-term contingency nor any single laboratory tradition licenses broad psychological claims about the totality of a person's values, relationships, or self-directed choices.
What the Evidence Establishes
The empirical existence of operant discrimination and stimulus generalization is well established across a large experimental literature. Differential contingencies can establish measurable differences in responding; generalization gradients can be measured; manipulating stimulus dimensions, training sequences, and contexts changes performance. The work of Skinner, Guttman and Kalish, Hanson, Jenkins and Harrison, Reynolds, Terrace, and subsequent researchers demonstrates the value of controlled experiments for isolating these effects. Dinsmoor, 1995a; Dinsmoor, 1995b.
More specific explanatory claims require qualification. A measured discrimination does not uniquely establish how attention operates, which associations have formed, or what neural systems produced learning. A functional account may coexist with multiple internal-process models. The generalization of laboratory findings to complex human skills is an empirical problem; evidence for one cue-training procedure cannot be transferred automatically to clinical effectiveness. Contemporary goal-directed and habitual-action research likewise shows why current outcome value and action–outcome knowledge sometimes matter beyond the history summarized by a simple discriminative relation. Mackintosh, 1975; de Wit and Dickinson, 2009.
Skinner's enduring contribution in this area is the experimentally disciplined question of how an antecedent acquires control of an operant through its history of consequences. Answering that question requires identifying the function of a stimulus, measuring a defined response, comparing different consequence arrangements, and studying transfer across contexts. The three-term contingency remains a durable framework for organizing those questions, while the detailed mechanisms and applications remain open to experimental refinement.
Frequently Asked Questions
What are the three terms in Skinner's contingency?
They are a discriminative antecedent stimulus, an operant response, and its consequence. The scientifically relevant relation is learned across repeated occasions: reinforcement available after the response in a particular context changes how likely the response becomes when that context reappears.
What does SD mean in operant conditioning?
SD denotes a discriminative stimulus, an antecedent whose presence has correlated with a specific response–reinforcer relation. Its function is defined through differential reinforcement history and measured behavior, rather than through its appearance or a person's verbal interpretation alone.
What is S-delta?
SΔ denotes an alternative stimulus condition in which the particular response does not produce the specified programmed reinforcer, or does so less effectively under the experimental comparison. It is a relation-specific designation and does not imply that the entire environment lacks consequences.
Is a discriminative stimulus the same as a conditioned stimulus?
The distinction concerns function. In respondent conditioning a conditioned stimulus can elicit a learned response through stimulus relations. In operant discrimination an SD changes the probability of a response given its history of response–consequence relations. The same physical cue may acquire both types of function.
What is stimulus generalization?
It is the extension of trained stimulus control to other cues that share relevant features or relations. Scientists measure it with generalization tests and gradients, which show how response probability varies over a dimension such as color, pitch, or location.
Can discrimination be learned without errors?
For some tasks, yes. Terrace's 1963 pigeon experiments showed that certain discriminations could be trained with few or no responses to negative stimuli when stimulus differences were carefully arranged and gradually changed. Those results do not promise error-free learning for every skill or person.
Does an ABC chart prove the function of a behavior?
No. An ABC record describes what was observed before and after behavior and can guide hypotheses. Functional conclusions depend on additional measurements, appropriate comparisons, and, where justified, experimental manipulations of the relevant relations.
Does the three-term contingency explain all human learning?
It supplies a powerful functional unit for discriminated operant learning. Full explanations of language, decision-making, social behavior, and clinical outcomes may require additional contextual, cognitive, developmental, neural, and cultural evidence.
Related Articles
B. F. Skinner — A comprehensive scientific account of Skinner's research program, major publications, experimental methods, and intellectual legacy.
Learning Theories in Education: Behaviorism, Cognitivism, Constructivism, and Sociocultural Learning — A broader comparison of learning theories and their implications for educational design.
Behavior Change Techniques: What Research Supports — Contemporary evidence and conceptual distinctions in the study of behavior-change interventions.
B. F. Skinner's Shaping Through Successive Approximations — The experimental development of shaping procedures.
B. F. Skinner on Positive Reinforcement, Negative Reinforcement, and Punishment — Functional distinctions among behavioral consequences.
B. F. Skinner and C. B. Ferster's Schedules of Reinforcement — Ferster and Skinner's 1957 reinforcement-schedule studies.
References
Bouton, M. E. (2004). Context and behavioral processes in extinction. Learning and Memory, 11(5), 485–494. https://doi.org/10.1101/lm.78804
de Wit, S., and Dickinson, A. (2009). Associative theories of goal-directed behaviour: A case for animal–human translational models. Psychological Research, 73, 463–476. https://doi.org/10.1007/s00426-009-0230-6
de Wit, S., Corlett, P. R., Aitken, M. R., Dickinson, A., and Fletcher, P. C. (2009). Differential engagement of the ventromedial prefrontal cortex by goal-directed and habitual behavior toward food pictures in humans. Journal of Neuroscience, 29(36), 11330–11338. https://doi.org/10.1523/JNEUROSCI.1639-09.2009
Dickinson, A. (1985). Actions and habits: The development of behavioural autonomy. Philosophical Transactions of the Royal Society B, 308(1135), 67–78. https://doi.org/10.1098/rstb.1985.0010
Dinsmoor, J. A. (1995a). Stimulus control: Part I. The Behavior Analyst, 18(1), 51–68. https://doi.org/10.1007/BF03392691
Dinsmoor, J. A. (1995b). Stimulus control: Part II. The Behavior Analyst, 18(2), 253–269. https://doi.org/10.1007/BF03392712
Ferster, C. B., and Skinner, B. F. (1957). Schedules of Reinforcement. Appleton-Century-Crofts. B. F. Skinner Foundation bibliography
Guttman, N., and Kalish, H. I. (1956). Discriminability and stimulus generalization. Journal of Experimental Psychology, 51(1), 79–88. https://doi.org/10.1037/h0046219
Hanson, H. M. (1959). Effects of discrimination training on stimulus generalization. Journal of Experimental Psychology, 58(5), 321–334. https://doi.org/10.1037/h0042606
Jenkins, H. M., and Harrison, R. H. (1960). Effect of discrimination training on auditory generalization. Journal of Experimental Psychology, 59(4), 246–253. https://doi.org/10.1037/h0041661
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
Nevin, J. A., and Grace, R. C. (2000). Behavioral momentum and the law of effect. Behavioral and Brain Sciences, 23(1), 73–90. https://doi.org/10.1017/S0140525X00002405
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
Reynolds, G. S. (1961). Attention in the pigeon. Journal of the Experimental Analysis of Behavior, 4(3), 203–208. https://doi.org/10.1901/jeab.1961.4-203
Skinner, B. F. (1934). A discrimination without previous conditioning. Proceedings of the National Academy of Sciences, 20(9), 532–536. https://doi.org/10.1073/pnas.20.9.532
Skinner, B. F. (1935a). Two types of conditioned reflex and a pseudo type. The Journal of General Psychology, 12(1), 66–77. https://doi.org/10.1080/00221309.1935.9920088
Skinner, B. F. (1935b). A discrimination based upon a change in the properties of a stimulus. The Journal of General Psychology, 12(2), 313–336. https://doi.org/10.1080/00221309.1935.9920107
Skinner, B. F. (1938). The Behavior of Organisms: An Experimental Analysis. Appleton-Century. B. F. Skinner Foundation bibliography
Skinner, B. F. (1953). Science and Human Behavior. Macmillan. B. F. Skinner Foundation bibliography
Skinner, B. F. (1969). Contingencies of Reinforcement: A Theoretical Analysis. Appleton-Century-Crofts. B. F. Skinner Foundation bibliography
Staddon, J. E. R., and Simmelhag, V. L. (1971). The “superstition” experiment: A reexamination of its implications for the principles of adaptive behavior. Psychological Review, 78(1), 3–43. https://doi.org/10.1037/h0030305
Terrace, H. S. (1963a). Discrimination learning with and without “errors.” Journal of the Experimental Analysis of Behavior, 6(1), 1–27. https://doi.org/10.1901/jeab.1963.6-1
Terrace, H. S. (1963b). Errorless transfer of a discrimination across two continua. Journal of the Experimental Analysis of Behavior, 6(2), 223–232. https://doi.org/10.1901/jeab.1963.6-223