Research figure: B. F. Skinner · Written by Artificial Sapiens
From the Law of Effect to Operant Learning
Edward L. Thorndike and B. F. Skinner established two historically connected but conceptually distinct experimental approaches to learning from consequences. Thorndike investigated how animals gradually acquired effective responses in problem situations and explained the results through strengthened connections between a situation and a response. Skinner subsequently developed a free-operant laboratory system in which response classes, their rates, and their relations to environmental consequences could be analyzed continuously. Their shared scientific question was how consequences change subsequent action; their answers differed in the behavior selected for measurement, the theoretical entities invoked, and the experimental arrangements used to establish functional relationships. The foundational historical evidence comes from Thorndike's 1898 monograph, his more explicit 1911 account of the law of effect, Skinner's 1937 definition of an operant, and his 1938 book The Behavior of Organisms.
This history matters because the familiar summary that “reward makes behavior more likely” compresses several distinguishable scientific propositions. A consequence can strengthen an association between a stimulus and a response, change the measured frequency of a response class, modify the expected value of an outcome, alter choices among alternatives, or affect the probability that an action will be repeated in a particular context. Those formulations may predict similar surface behavior while proposing different explanatory structures. The direct comparison clarifies what Thorndike observed, what Skinner inherited and transformed, what subsequent experimentalists established, and why contemporary learning science continues to distinguish habits, instrumental actions, and outcome-sensitive choices. A recent historical reexamination by Michael Domjan (2026) emphasizes the importance of preserving the actual stimulus–response formulation of Thorndike's law.
The Historical Sequence: 1898, 1911, 1932, 1937, and 1938
Thorndike's Animal Intelligence: An Experimental Study of the Associative Processes in Animals appeared in 1898 as a Psychological Review monograph supplement, reporting controlled studies with cats, dogs, and chicks. The innovative empirical program preceded John B. Watson's 1913 behaviorist manifesto and preceded Skinner's professional research. It combined deliberately constructed apparatus with repeated quantitative observations, replacing reliance on isolated anecdotes about animal cleverness with observable learning curves. The original bibliographic record confirms the 1898 publication, while Lattal's historical analysis evaluates the monograph's enduring methodological significance.
Thorndike collected and reconsidered earlier investigations in Animal Intelligence: Experimental Studies, published in 1911. The 1911 volume reproduced the earlier monograph, incorporated additional research, and presented the explicit law of effect and law of exercise in its chapter on laws and hypotheses of behavior. Distinguishing the first experimental publication from the subsequently articulated general law is essential for faithful historical attribution. In 1932, The Fundamentals of Learning reexamined earlier assumptions about the effects of satisfying and annoying consequences, and Thorndike's mature position no longer treated punishment as the simple symmetrical inverse of reward. The 1932 book is independently documented by its bibliographic record, and the evolution of the theory is traced in Domjan's 2026 historical review.
Skinner's relevant conceptual development took place during the 1930s. His 1937 reply to Jerzy Konorski and Stefan Miller introduced an explicit terminological distinction between respondent and operant behavior, and his 1938 book organized years of rat experiments into a distinctive experimental analysis. Later publications refined the program: Science and Human Behavior appeared in 1953, while Charles B. Ferster and Skinner jointly published Schedules of Reinforcement in 1957. A scientific chronology should therefore recognize Thorndike as the historical originator of the law-of-effect tradition, Skinner as the architect of his own operant analysis, and Ferster as Skinner's named collaborator in the canonical 1957 schedule study. The B. F. Skinner Foundation bibliography verifies the book authorships and publication dates.
Edward Thorndike's Experimental Problem: How Does an Animal Escape?
Puzzle Boxes, Subjects, and Experimental Conditions
Thorndike constructed several puzzle boxes whose doors would open only when an animal performed a specified action, such as manipulating a latch, pulling a loop, or pressing a mechanism. A food-motivated cat placed in such a box could initially scratch, turn, push, or otherwise act with little organization. Eventually a movement released the door; the animal escaped and obtained access to food outside the enclosure. The researcher then returned the subject to the apparatus and repeated the procedure. Some variants used different species and mechanisms, and the original monograph also discussed observational learning, instinctive actions, and possible reasoning. Thus the well-known cat-and-latch illustration represents one family of procedures rather than a single standardized experiment whose parameters were identical across subjects. Thorndike's own experimental account and Chance's retrospective document the variety of apparatuses and the research questions involved.
The unit of repetition was the discrete trial: a placement inside the box followed by release when the criterion act occurred. Thorndike's principal quantitative indicator was escape latency, the elapsed time between placement and successful escape. He plotted escape times over successive trials, producing curves that made improvements and irregularities visible. If a cat escaped after a shorter interval on later trials, this supported the conclusion that its behavior had changed with experience. But the measured decline by itself did not directly reveal a neural connection, an experienced satisfaction, or a mental calculation. The experimental observation was a change in performance under repeated arrangements; the explanation required an additional inferential step. Thorndike's original work should be read with this distinction in view.
What the Escape Curves Demonstrated
The repeated trials documented acquisition of more efficient escape responses in a constrained task. Thorndike interpreted the generally gradual improvement, together with the sometimes irregular pattern of errors and successful movements, as evidence against the idea that animals typically solved such tasks through a sudden, humanlike apprehension of a mechanical principle. His interpretation emphasized the selection and retention of successful responses through their consequences. He also recognized species-specific patterns and the effects of prior activity; an animal's available repertoire determined which potentially successful movements could occur. These claims were consequential because they turned animal problem solving into a measurable learning problem rather than an argument conducted primarily through anthropomorphic stories.
The familiar label “trial-and-error learning” captures an important dimension of the procedures, but it can conceal their experimental specificity. An error was not necessarily separately penalized; many ineffective movements simply failed to open the door. A successful movement ended the confinement and led to food access. Those two aspects of consequence—the end of a situation and access to an attractive outcome—were intertwined in many trials. Motivational state, fatigue, exploration, differences among boxes, and the mechanics of returning animals to the apparatus could all affect the recorded latency. Chance (1999) and Lattal (1998) regard the methodological contribution as fundamental while recognizing that the records require interpretation within the actual apparatus and procedure: Chance; Lattal.
Thorndike's Law of Effect: Its Original Theoretical Claim
The Situation–Response Bond
In its developed 1911 presentation, the law of effect proposed that, among responses made to the same situation, those accompanied or soon followed by satisfying consequences would become more firmly connected with that situation and more likely to recur when the situation returned. The 1911 wording also proposed that discomfort would weaken such connections. The theory therefore concerned the modification of a situation–response connection; satisfaction and discomfort characterized consequences, while changes in connection strength explained altered future behavior. Thorndike's position was part of an associationist psychological tradition, and the law made an explicitly historical claim about how prior experiences shaped later performance. The language of “satisfaction” was historically integral to the account, although its operational and mechanistic meaning remained a matter of subsequent debate. The original formulation is available in Animal Intelligence, the chapter on laws and hypotheses.
This focus on a bond matters theoretically. A simplified paraphrase that reinforcement “strengthens a response” omits the specific connection to the situation that Thorndike believed was altered by experience. A cat could learn to pull a loop in one apparatus yet behave differently when the apparatus or environmental cues changed. Whether performance transfers is partly a question about which stimulus elements are shared across contexts. In Thorndike's framework, learned performance is organized by the relation between a particular situation and a particular response. Domjan's 2026 analysis traces how many modern descriptions lose this key element and thereby silently replace the original statement with a different version of consequence-based learning.
The Law of Exercise and Its Later Reassessment
Thorndike's 1911 account also advanced the law of exercise, according to which repetition and use strengthen connections. Read historically, law of effect and law of exercise were related but distinguishable explanatory claims: consequences could select connections, and practice could modify their strength. Subsequent research complicated the assumption that simple repetition, without informative feedback or consequential change, reliably produces better learning. Thorndike's later work revised elements of his earlier formulations rather than preserving the 1911 dual laws unchanged. The evolution is important for avoiding the retrospective portrayal of Thorndike's theory as a single immutable doctrine established at one moment. Its main editions and revisions can be checked against the 1911 primary text, The Fundamentals of Learning (1932), and Domjan (2026).
The Asymmetry of Reward and Punishment
A consequential qualification emerged from Thorndike's later experimental work. The 1911 law had been expressed as an approximate symmetry: satisfying states strengthened bonds and annoying states weakened them. Results from subsequent human learning studies persuaded Thorndike that corrective feedback or punishment could have effects unlike the strengthening produced by satisfying consequences. His mature statements gave greater weight to the strengthening effects of positive consequences. This historical change should not be transformed into the blanket modern assertion that punishment cannot suppress behavior. The effects of aversive events depend on the procedure, the response measured, the available alternatives, context, and the persistence of change after the contingency ends. The historically accurate claim is that Thorndike revised the theoretical symmetry he initially proposed; it is not that every aversive contingency was experimentally shown to be ineffective. See The Fundamentals of Learning (1932) and Domjan (2026).
Skinner's Reconstruction: Operant Behavior as an Experimental Unit
From Elicited Reflexes to Emitted Operants
Skinner worked within a field shaped by Pavlovian conditioning, reflex theory, Watsonian behaviorism, and Thorndike's instrumental investigations. In the 1937 paper “Two Types of Conditioned Reflex: A Reply to Konorski and Miller,” he distinguished respondent behavior, which is elicited by identifiable antecedent stimuli, from operant behavior, whose occurrence need not be explained by a single immediately eliciting stimulus. His term “operant” designated a functionally defined response unit whose relation to consequences could be studied experimentally. Antecedent conditions could still influence the probability of an operant; the distinction was about how responses were organized and controlled, not a claim that behaving organisms act in an absence of environmental history. The 1937 primary paper and its publisher record establish this conceptual transition.
Thorndike's question about which connection was strengthened became, in Skinner's laboratory program, a question about how a class of behavior changes when particular consequences follow its members. Defining a response class functionally allowed movements with variable physical form to count together when they produced the same environmental effect. The scientist arranged contingent outcomes and measured changes in behavior over time. Reinforcement in this framework became an empirical relation between consequences and subsequent responding, although particular definitions and causal interpretations were themselves debated. Skinner's The Behavior of Organisms (1938) is the major early synthesis of this strategy. This was a reorganization of methods and explanatory priorities, not merely a change in vocabulary for Thorndike's associations.
The Free-Operant Procedure
In Skinner's chamber, a rat could press a lever repeatedly during an extended session, and an apparatus could deliver food following selected responses. For pigeon experiments, a pecking response to a key could serve a corresponding function. Unlike a puzzle-box escape task in which a successful response often ends the trial, the free-operant arrangement permitted repeated responses while the animal remained in the controlled environment. The researcher could arrange reinforcement after every eligible response, after certain numbers of responses, after intervals of time, or under more complex contingencies. Rate, temporal pattern, persistence, and distribution of responses became continuous empirical variables. The difference between a discrete-trial escape latency and an extended-session response rate changed which properties of learning could be observed. Domjan (2026) explicitly identifies this contrast as central to the historical transition.
The experimental advantage was substantial: repeated measurement within a single organism under changing conditions enabled researchers to characterize orderly relations without making every conclusion depend on averaging across large groups. Yet methodological precision did not eliminate every inferential difficulty. Response rates could be affected by hunger, fatigue, competing behavior, motor demands, stimulus discriminability, or the exact temporal relationship between response and consequence. A change in rate identified a functional effect under the arranged conditions; it did not automatically identify an exclusively stimulus–response neural mechanism. The methodological literature on Skinner's instrumentation, including the history of the cumulative recorder, explains why measuring behavior repeatedly and visibly over time was central to this research tradition.
Comparing the Apparatuses and What They Measure
The Puzzle Box Versus the Operant Chamber
The historical comparison is best understood as a change in experimental design. Thorndike's animals faced a problem with a criterion escape action and a discrete endpoint. Skinner's subjects typically remained within the chamber, generating many measurable instances of an operant during a continuous session. Thorndike used escape latency and learning curves to infer acquired efficiency across trials. Skinner used response counts, rates, interresponse timing, and cumulative response records to describe behavior under programmed consequences. Both placed organisms in designed environments with experimentally controlled outcomes. Their dependent measures nevertheless answered different questions, and neither dependent measure should be treated as interchangeable with the other.
Thorndike's curves could reveal acquisition but offered relatively coarse temporal access to the behavior taking place before successful escape. Skinner's apparatus made the timing of individual operants available for further experimental control. Conversely, a continuous lever press under a carefully simplified contingency cannot by itself reproduce the demands of navigating a complex puzzle box, such as selecting among many possible manipulations. The methodological choice traded the richness of a problem episode for precision about recurring response units and reinforcement schedules. Historians Lattal (1998) and Chance (1999) emphasize both the continuity and the development in these techniques.
The Response Variable and the Unit of Analysis
Consider an illustrative comparison, not an account of a particular archival data set. If a cat's escape time falls from a long initial delay to a short delay across sessions, the result tells us that the entire successful performance has become faster. If a rat's lever-pressing rate increases after contingent food deliveries, the result tells us that the defined operant occurs more frequently over time. The first observation bundles exploration, selection, and task completion into a latency; the second repeatedly samples a target response in an ongoing environmental arrangement. Translating one into the other requires additional assumptions about the behavioral process and the available opportunities to respond.
The difference also affects causal inference. A researcher can establish that manipulating a reinforcement contingency alters a measured response under specified experimental control. That is a strong claim about a relation within the test system. An account of a hidden association, expectation, incentive value, or neural mechanism requires additional evidence. Thorndike's bond and Skinner's operant strength represent different levels and traditions of explanation, while contemporary research often combines behavioral measurement with independent manipulations that distinguish competing mechanisms. The most defensible scientific lineage is one of cumulative experimental questions with successive changes in the operational definitions used to answer them.
Theories of Learning: Thorndike's Connectionism and Skinner's Radical Behaviorism
Thorndike's associationist theory treated the learned bond between situation and response as a central explanatory construct. Skinner's radical behaviorism prioritized functional relations among behavioral events, antecedent conditions, consequences, and organism histories, and it resisted explanatory appeals to hypothetical intervening structures when these did not improve experimental prediction and control. In his 1950 article “Are Theories of Learning Necessary?” Skinner criticized a class of explanations that displaced the explanation of measured behavior into events postulated at another observational level. The article is a methodological and philosophical argument about how to construct a science, not proof that internal physiological or cognitive events cannot be investigated scientifically. See Skinner (1950) and the full primary text.
Their philosophies share a commitment to historical learning from environmental interaction, but they organize causation differently. Thorndike sought to explain a performance tendency through a modified association. Skinner sought experimentally reliable relations among response classes and environmental contingencies. Radical behaviorism also extended its behavioral analysis to private events such as thinking and feeling, without granting introspective reports privileged explanatory authority. These distinctions prevent the common mistake of classifying every consequence-based model as the same form of behaviorism. Edward Thorndike also preceded Watson's formal behaviorist program; historical priority, scientific influence, and membership in a later philosophical school are separate questions.
A second distinction concerns the descriptive term “reinforcement.” A rewarding event is not universally reinforcing regardless of context. Its empirical effects depend on deprivation or satiation, the response–outcome contingency, the organism's learning history, and competing reinforcers. Describing a reinforcer simply as something pleasant ignores the operational issue of whether responding changes. At the same time, defining reinforcement solely through an observed increase in responding risks becoming circular unless the experiment distinguishes the arranged consequence from the measured change and tests the relevant contingencies independently. Later analysis revisited this problem in Baum's discussion of contingency and allocation and related theoretical work.
Antecedents, Consequences, and the Structure of Control
Thorndike's formulation already contained a contextual element: a response becomes more firmly connected with the situation in which it occurs. Skinner's mature operant analysis made stimulus control and consequential contingencies central to experimental arrangements, often summarized by a discriminative antecedent, a response, and a reinforcing or otherwise relevant consequence. A discriminative stimulus signals that a particular response is more or less likely to contact a consequence under current conditions; it need not mechanically elicit the response like an unconditioned reflex. These concepts describe experimentally testable conditional relations and provide a more precise vocabulary for asking why a learned act occurs in one environment and not another.
The relationship between context and consequence complicates simplistic statements such as “behavior followed by reward always increases.” Timing matters, as do competing responses, the relevance of the outcome, and previous exposure to contingencies. Two events occurring close together may covary because of a third variable, while a genuinely effective instrumental contingency can depend on how an action changes the probability of the outcome compared with what happens when the action is withheld. Modern instrumental research therefore investigates both contiguity and contingency. The Balleine and Dickinson (1998) review00033-1) analyzes action–outcome contingency and incentive learning as separable determinants of goal-directed performance.
Reinforcement Schedules: A New Experimental Domain
The capacity to measure many successive operants allowed researchers to compare how behavior developed under distinct reinforcement schedules. Continuous reinforcement, fixed and variable ratios, and fixed and variable intervals create different relations between opportunities to respond and programmed consequences. Across controlled conditions, these schedules produce characteristic distributions and temporal organizations of behavior, although actual patterns also depend on species, history, magnitude of outcome, response effort, and experimental details. The classic extended analysis was the 1957 book Schedules of Reinforcement, coauthored by Charles B. Ferster and B. F. Skinner. Both authors must be named in the history of that work; its findings belong to a collaborative experimental research program. Publication details are documented in the Skinner Foundation bibliography.
This body of work moved beyond the original puzzle-box question of whether a response would be selected at all. It examined the shape of ongoing behavior under recurrent, deliberately programmed contingencies. A rate that appears high on one schedule may not indicate that the animal has acquired more knowledge, a stronger conscious desire, or a better abstract understanding of the task. It may reflect the schedule's control over response timing and the opportunity to obtain reinforcement. Later quantitative analyses, including Herrnstein's 1970 law-of-effect formulation, studied how behavior is allocated among alternatives in relation to obtained reinforcement. That development was the work of later researchers, not a result contained in Thorndike's original law or a theorem personally discovered by Skinner.
Was Skinner Simply Repeating Thorndike?
The intellectual continuity is genuine: consequences alter subsequent performance, and animal experiments can establish this relation under measured conditions. Thorndike made a foundational methodological and conceptual contribution through the puzzle boxes, quantitative learning curves, and law of effect. Skinner acknowledged the earlier law while constructing an independent experimental and philosophical system. Skinner's replacement of escape latency with free-operant response rate, his functional definition of the operant, and his investigation of reinforcement schedules changed the structure of the inquiry. An account that makes Skinner the inventor of learning through consequences erases Thorndike; an account that reduces Skinner's laboratory program to another name for the puzzle box erases its technical and methodological advances.
The continuity also contains selective reinterpretation. As Domjan (2026) documents, Skinner's description of the law of effect sometimes condensed Thorndike's association-strengthening account into a formulation about consequences “stamping in” behavior, without preserving the original theoretical emphasis on the stimulus–response bond. That condensation became influential in textbooks and later discussions. Reconstructing historical arguments requires attention to the texts and the explanatory commitments of each author, rather than assuming that a modern introductory definition can be projected backward unchanged. The resulting genealogy is more precise: Thorndike established an experimentally grounded connectionist law; Skinner organized a free-operant functional analysis inspired partly by that tradition.
Can the Law of Effect Explain All Instrumental Learning?
Contemporary experiments distinguish learning about action–outcome relations from the establishment of habits that may persist even when the current value of an outcome changes. In an outcome-devaluation procedure, researchers first establish an action that produces a particular outcome, then alter that outcome's value and test whether the action changes. A reduction in action after devaluation can provide evidence that current outcome value contributes to behavioral control. Continued performance under some conditions can indicate that stimulus–response processes or other influences exert greater control. The interpretation depends on the procedure, especially whether the test prevents new reinforcement and whether the organism actually learned and registered the devaluation. Balleine and Dickinson (1998)00033-1) is a foundational analysis of this experimental distinction.
Action–outcome contingency degradation provides a complementary test: if the outcome occurs regardless of the action, sensitivity to the changed causal relation can be examined. Goal-directed action is often described as depending on knowledge of what an action produces and its current incentive value, whereas a habit is often described through learned stimulus–response control. These are experimentally motivated models with their own limitations, not timeless labels for two sharply separated kinds of behavior in every real-world setting. The history of the law of effect is directly relevant because Thorndike's original stimulus–response bonds anticipated one influential strand of habit theory, while research on action–outcome knowledge demonstrated forms of instrumental control requiring additional explanatory dimensions. For a comparative account of human and animal evidence, see Balleine and O'Doherty (2010).
Neural Systems and Mechanistic Explanation
Work in behavioral neuroscience has associated goal-directed and habitual control with partly distinguishable corticostriatal circuits, while emphasizing the interaction between systems rather than an inflexible one-to-one map from brain region to psychological function. Experiments combine behavioral manipulations with neural recordings, lesions, pharmacological interventions, or imaging to test mechanistic claims. Such research offers evidence at a level of explanation neither Thorndike's original bond concept nor Skinner's free-operant measure could directly supply. It extends historical questions about the control of action into contemporary comparative neuroscience. The synthesis by Balleine and O'Doherty (2010) reviews evidence from both humans and rodents, including the roles of cortical and striatal circuits.
Neural accounts must not be retroactively attributed to nineteenth- or early-twentieth-century experiments. A modern mechanism may explain patterns that earlier researchers described correctly without being part of those researchers' original theories. Equally, discovering an identifiable neural correlate does not render behavioral contingencies irrelevant; an adequate explanation can connect neural dynamics, experience, action selection, and environmental consequences. Historical behavioral data and modern neuroscience operate at different evidential levels. Their integration requires experiments that make distinct predictions and isolate the proposed causes rather than replacing one set of explanatory words with another.
Why Learning Curves and Response Rates Require Caution
One recurrent problem in both research traditions is the relation between performance and learning. A shortening escape latency or increase in lever pressing is a behavioral observation under particular circumstances. The organism may have acquired a new association, become more motivated, developed a more efficient action sequence, reduced exploratory activity, or shifted attention among alternatives. Several of these changes can occur together. Separating them requires control conditions, carefully selected outcome manipulations, repeated sessions, and tests designed to distinguish models. The early experiments remain foundational precisely because they made important variables observable, but their interpretation should reflect the inferential reach of the measurements actually recorded.
Single-organism designs can reveal strong within-subject experimental relations when baseline and intervention phases are adequately controlled and replicated. Generalization across individuals, tasks, species, and social settings is a further question. Conversely, averaging across subjects can conceal heterogeneous trajectories. An accurate history of experimental behavior analysis therefore considers both methodological advances and sampling or generalization limitations. The idea that a highly controlled animal demonstration directly establishes universal laws of human educational or clinical practice exceeds the evidence of the original experiment. Applied claims require contemporary task-relevant empirical studies, appropriate comparators, and independent assessment of outcomes that matter to participants.
Historiography and the Problem of Simplified Textbooks
Textbook genealogies commonly describe three steps: Thorndike discovered the law of effect, Watson founded behaviorism, and Skinner developed operant conditioning. Each statement identifies an important historical contribution, but the story becomes misleading when those distinct programs are merged. Thorndike was an experimental comparative psychologist and connectionist whose early work preceded Watson's behaviorist manifesto. Watson publicly advanced a particular early program of behaviorism. Skinner later developed radical behaviorism and a specific free-operant laboratory technology. The traditions influenced one another while retaining different commitments and methodological practices. Lattal (1998), Chance (1999), and Domjan (2026) provide useful historically grounded corrections.
Another recurrent distortion portrays Thorndike's law as a timeless, exact formula identical to all later reinforcement principles. The original law specified connections between situations and responses and evolved over several decades. Skinner's research emphasized experimentally observed response probabilities and rates; later researchers developed quantitative relations, theories of choice, associative models, and neurological accounts. An accurate interpretation keeps historical attribution and present scientific validity in separate columns. A historical article can establish that a scientist proposed an idea and that an apparatus produced recorded observations; present empirical research is needed to assess the generality, mechanisms, and limits of the principle today.
Education, Behavior Change, and the Limits of Translation
Thorndike's work became influential in educational psychology through its emphasis on practice, consequences, and measurable learning. Skinner later applied behavioral principles to instruction, including programmed teaching and the analysis of reinforcement contingencies in classrooms. These historical programs helped shape questions about feedback, skill acquisition, task sequencing, and environmental design. But the presence of positive feedback in an effective educational intervention does not by itself show that the intervention is an application of Skinner's entire theoretical philosophy, and outcomes may also depend on cognition, explanation, prior knowledge, social participation, and institutional context. The existing encyclopedia study Learning Theories in Education examines broader theoretical traditions and retains the canonical educational topic.
Likewise, behavioral methods employed in habit formation, rehabilitation, organizational practice, or clinical services must be evaluated by evidence relevant to the specific population and intervention. Modern applied behavior analysis was developed by multiple researchers, including Donald Baer, Montrose Wolf, and Todd Risley; it should not be credited to Skinner alone. Historical learning theory supplies a conceptual ancestry, not proof of efficacy, safety, or ethical acceptability for every intervention later labeled “behavioral.” Assessment of autonomy, consent, meaningful outcomes, unwanted effects, and the availability of alternatives is especially important when people are subject to intensive behavioral programs. The current encyclopedia article Behavior Change Techniques: What Research Supports treats broad interventions as their own evidence question.
Ethical and Methodological Boundaries
The foundational experiments involved food motivation, confinement, and controlled manipulation of consequences in nonhuman animals. Their scientific importance should be understood alongside the ethical responsibilities now associated with animal care, minimization of distress, and justification of experimental procedures. Retrospective evaluation also needs attention to the norms, instruments, and research questions of the time. Neither researcher operated under contemporary animal-welfare review procedures, and descriptions of historical experiments should not be repackaged as advice to reproduce the conditions without modern oversight. The question of whether an experiment identifies a reliable learning relation is distinct from the ethical question of whether that experiment should be performed in a given setting.
The same distinction applies when behavioral ideas enter human contexts. A measured increase in a selected response may represent improvement in a meaningful skill, adaptation to incentives, compliance under unequal power, or some combination of effects. Evaluating outcomes requires understanding whose goals are being served, how participants experience the procedure, what meaningful benefits follow, and whether the effects persist in everyday circumstances. Thorndike's and Skinner's experiments establish historical scientific concepts about learning; an ethically defensible applied program requires independent evidence and person-centered safeguards. This difference is especially important when interpreting experiments involving animals as precedents for clinical or educational claims.
What Remains Established, Limited, and Contested?
The established empirical core is that instrumental performance is systematically affected by its relation to consequences, and that performance can be studied through repeated controlled measurements. Thorndike's experimental research established consequential changes in escape performance; Skinner and subsequent experimentalists extended the analysis to free-operant responding, reinforcement schedules, and behavior allocation. Historical questions of attribution are also well supported: 1898 belongs to Thorndike's monograph, 1911 to his explicit statement of the law, 1937 to Skinner's published operant terminology, and 1938 to his major experimental synthesis. The line of later work is documented in Lattal (1998) and Domjan (2026).
Mechanistic generalizations remain more differentiated. The same observable action can be influenced by contingency, outcome value, environmental cues, response cost, attention, and prior training. Competing models identify different latent structures and can be evaluated by devaluation, degradation, transfer, or other discriminating tests. Neither a universal situation–response bond nor a single response-rate measure exhausts modern accounts of human action. The relation between controlled laboratory findings and complex educational, cultural, or clinical behaviors requires task-specific evaluation. The research reviewed in Balleine and Dickinson (1998)00033-1), Balleine and O'Doherty (2010), and Domjan (2026) supports a historically continuous but theoretically plural field.
Frequently Asked Questions
Who first proposed the law of effect: Thorndike or Skinner?
Edward L. Thorndike originated the law-of-effect tradition. His 1898 monograph reported the pioneering puzzle-box experiments; the familiar explicit statement of the law appeared in his 1911 Animal Intelligence: Experimental Studies. Skinner subsequently reformulated the experimental study of consequences within an operant framework. The distinction between the 1898 evidence and the 1911 articulation is historically important. See Thorndike's original publication and Domjan (2026).
Is the law of effect identical to operant conditioning?
They belong to the same broad history of instrumental learning but are conceptually distinguishable. Thorndike described consequences strengthening situation–response connections; Skinner studied the functional relations between consequences and operant response classes under precisely arranged contingencies. Their laboratory measures also differ: escape latency in repeated puzzle-box trials and rate or temporal organization of responses in free-operant sessions.
What is the difference between a puzzle box and a Skinner box?
A Thorndike puzzle box presents a constrained escape problem, generally measured across discrete trials. A Skinner operant chamber permits repeated responding within an extended controlled session, often recording response rates and their relationships to reinforcement. Each apparatus makes different dependent measures and experimental manipulations practical. Neither apparatus by itself decides all questions about the internal mechanisms of learning.
Did Thorndike change his mind about punishment?
Thorndike's initial 1911 formulation treated satisfying and annoying consequences as producing broadly opposing changes in association strength. Later research led him to weaken the proposed symmetry, particularly with respect to the direct weakening attributed to annoying outcomes. The change is a documented revision of his theory, not a universal claim that punishment never changes behavior. See The Fundamentals of Learning (1932).
Did Skinner deny that organisms have thoughts or feelings?
Skinner's radical behaviorism addressed private events as part of behavior and its conditions. His methodological dispute concerned the adequacy of explanations that invoke unmeasured internal causes without improving experimentally grounded understanding. The separate question of whether cognitive or neural models can make testable and useful predictions is addressed by later research, including contemporary studies of goal-directed and habitual control.
What did later research add to the two traditions?
Subsequent investigators developed formal models of behavior allocation, quantitative reinforcement relations, outcome-devaluation methods, action–outcome learning, and neuroscience of goal-directed action and habits. Herrnstein (1970) contributed a quantitative law-of-effect framework; Balleine and O'Doherty (2010) synthesized evidence on neural systems of action control. These are independent later research achievements, historically related to but not authored by Thorndike or Skinner.
Scientific Conclusion
Thorndike made consequence-based learning an object of controlled experimental investigation and proposed an associationist law linking situations, responses, and their outcomes. Skinner developed a different scientific instrumentarium and explanatory framework: the operant response class, free-operant chamber, continuous measurement, contingency analysis, and experimental study of reinforcement schedules. His approach borrowed from a tradition Thorndike had helped establish while reorganizing its concepts, dependent variables, and philosophical goals. Later scientists expanded both lines of inquiry into choice theory, goal-directed instrumental control, habitual behavior, and neuroscience. Understanding the transition from the law of effect to operant learning thus means tracing a sequence of attributable discoveries and theoretical revisions rather than compressing more than a century of psychology into a single slogan.
Related Articles
B. F. Skinner — The central scientific chapter on Skinner's research program, experimental methods, radical behaviorism, writings, and intellectual legacy.
Learning Theories in Education: Behaviorism, Cognitivism, Constructivism, and Sociocultural Learning — A broader comparison of how major learning traditions explain educational change.
Behavior Change Techniques: What Research Supports — Modern evidence and distinctions concerning behavioral interventions.
B. F. Skinner's Operant Conditioning Theory — Operant contingencies and their experimental foundations.
B. F. Skinner and John B. Watson: Two Traditions of Behaviorism — Early and radical behaviorism in historical perspective.
B. F. Skinner's Experimental Analysis of Behavior — Single-organism experimental methods and scientific inference.
References
Balleine, B. W., and Dickinson, A. (1998). Goal-directed instrumental action: Contingency and incentive learning and their cortical substrates. Neuropharmacology, 37(4–5), 407–419. https://doi.org/10.1016/S0028-3908(98)00033-100033-1)
Balleine, B. W., and O'Doherty, J. P. (2010). Human and rodent homologies in action control: Corticostriatal determinants of goal-directed and habitual action. Neuropsychopharmacology, 35, 48–69. https://www.nature.com/articles/npp2009131
Baum, W. M. (2012). Rethinking reinforcement: Allocation, induction, and contingency. Journal of the Experimental Analysis of Behavior. https://pmc.ncbi.nlm.nih.gov/articles/PMC3266735/
Chance, P. (1999). Thorndike's puzzle boxes and the origins of the experimental analysis of behavior. Journal of the Experimental Analysis of Behavior, 72(3), 433–440. https://doi.org/10.1901/jeab.1999.72-433
Domjan, M. (2026). Thorndike's law of effect and its inconsistent description over the years. Journal of the Experimental Analysis of Behavior, 125(1), e70073. https://doi.org/10.1002/jeab.70073
Ferster, C. B., and Skinner, B. F. (1957). Schedules of Reinforcement. Appleton-Century-Crofts. https://www.bfskinner.org/publications/books
Herrnstein, R. J. (1970). On the law of effect. Journal of the Experimental Analysis of Behavior, 13(2), 243–266. https://doi.org/10.1901/jeab.1970.13-243
Lattal, K. A. (1998). A century of effect: Legacies of E. L. Thorndike's Animal Intelligence monograph. Journal of the Experimental Analysis of Behavior, 70(3), 325–336. https://doi.org/10.1901/jeab.1998.70-325
Skinner, B. F. (1937). Two types of conditioned reflex: A reply to Konorski and Miller. The Journal of General Psychology, 16, 272–279. https://doi.org/10.1080/00221309.1937.9917951
Skinner, B. F. (1938). The Behavior of Organisms: An Experimental Analysis. Appleton-Century. https://www.bfskinner.org/publications/books
Skinner, B. F. (1950). Are theories of learning necessary? Psychological Review, 57(4), 193–216. https://doi.org/10.1037/h0054367
Thorndike, E. L. (1898). Animal Intelligence: An Experimental Study of the Associative Processes in Animals. Psychological Review Monograph Supplements. https://doi.org/10.5962/bhl.title.25848
Thorndike, E. L. (1911). Animal Intelligence: Experimental Studies. Macmillan. https://www.gutenberg.org/files/69904/69904-h/69904-h.htm
Thorndike, E. L. (1932). The Fundamentals of Learning. Teachers College, Columbia University. https://books.google.com/books/about/The_fundamentals_of_learning.html?id=8cScAAAAMAAJ