Thorndike's law of effect and its inconsistent description over the years

Abstract

The law of effect, originally proposed by E. L. Thorndike in a book that was published in 1911, had a major influence on the development of studies of instrumental or operant conditioning in the twentieth century. It remains a core concept in psychology and is invariably covered in undergraduate and graduate courses. Thorndike's law of effect also remains a topic of inquiry in contemporary efforts to understand the nature of habitual responses as contrasted with goal‐directed behavior. The mechanisms of the law of effect continue to be studied by behavioral and social psychologists as well as neuroscientists. However, many statements of the law of effect deviate in significant ways from Thorndike's original formulation. The present article reviews the history of Thorndike's law of effect and traces how the law has been, and continues to be, misrepresented. The hope is that familiarity with common misrepresentations of the law of effect will encourage greater consistency and clarity in discussions of the law and better appreciation of its role in contemporary research.

Edward L. Thorndike was one of the major figures in the development of modern psychology. Soon after the publication of his PhD dissertation (Thorndike, 1898) Thorndike became a member of the faculty of the Teachers College at Columbia University where he remained until his retirement in 1940. By the end of his career, Thorndike wrote more than 500 publications, including numerous articles and books on educational psychology and related topics (Derenne, 2022). Much of that work is no longer of contemporary interest. However Thorndike's research involving learning in various nonhuman species continues to be cited in contemporary accounts of conditioning and learning. Thorndike is ranked #9 in an assessment of the 100 most eminent psychologists of the twentieth century (Haggbloom et al., 2002) primarily because of frequent references to his law of effect.¹

Introductory psychology textbooks and more advanced books on conditioning and learning inevitably include descriptions of Thorndike's puzzle box experiments with kittens and other animals, along with references to the “law of effect,” which Thorndike proposed to explain his results. Descriptions of the procedures and results of the puzzle box experiments have remained fairly consistent and accurate over the more than 100 years since the original report of those studies was published (Chance, 1999). However, the same has not been the case for what people consider to be Thorndike's “law of effect.” My goal in this article is to call attention to persistent and common misrepresentations of Thorndike's “law of effect” and to trace the emergence of these misrepresentations. I will end with some remarks about why it remains important to be accurate in how we discuss the law of effect and the relevance of Thorndike's formulation in contemporary research. To recognize misrepresentations of the law of effect, let's begin with a reexamination of Thorndike's originally writings.

THORNDIKE'S STATEMENT OF THE LAW OF EFFECT

Thorndike's first detailed description of his puzzle box experiments was in a monograph supplement that he wrote for the Psychological Review (Thorndike, 1898). This was a detailed report, running 109 pages, in which he described the rationale and basic methodology of the puzzle box experiments and reviewed data obtained with kittens, dogs, and chicks. The basic procedure was to place an animal in an enclosure (a puzzle box) with a piece of food available if the animal managed to successfully operate a latch to get out. Different puzzle boxes required different responses before the animal would be released. Each placement in the box (and subsequent release) was considered a training trial. The general finding was that animals were quicker to escape from the box with successive training trials. The essential features of this description of the puzzle box experimental method and the results obtained have not changed over the years and remain faithful to Thorndike's original writings.

In discussing what animals were learning in the puzzle box procedure in the 1898 monograph, Thorndike described the formation of associations but in this initial report he did not expand on the concept of an association or discuss what he later proposed as the law of effect. However, one can see precursors to the law of effect in the 1898 monograph on page 8, where Thorndike wrote that “the connection of a certain act with a certain situation and resultant pleasure” is what is responsible for the decreased escape latencies that occurred with repeated training trials. This basic idea became Thorndike's law of effect.

Thorndike's first definitive statement of the law appeared in his book, Animal Intelligence: Experimental Studies (Thorndike, 1911). Thorndike thought about behavior or a particular response R as occurring because of an antecedent stimulus or “sense impression,” which we may designate as S. In describing the success of one of his cats in escaping a puzzle box, Thorndike wrote that “the cat hits upon the successful movement, and gradually associates it with the sense‐impression of the interior of the box until the connection is perfect, so that it performs the act as soon as confronted with the sense‐impression” (p. 38). Notice that unlike Skinner (1938) who talked about strengthening operant responses, Thorndike did not describe the strengthening of a response but the learning or strengthening of an S‐R association.

A clear statement of Thorndike's law of effect was not provided until much later in the 1911 book (p. 244), in a chapter titled “Laws and Hypotheses for Behavior.” Here Thorndike states,

of several responses made to the same situation, those which are accompanied or closely followed by satisfaction to the animal will, other things being equal, be more firmly connected with the situation, so that, when it recurs, they will be more likely to recur; those which are accompanied or closely followed by discomfort to the animal will, other things being equal, have their connections with that situation weakened, so that, when it recurs, they will be less likely to occur. The greater the satisfaction or discomfort, the greater the strengthening or weakening of the bond.

This statement follows his earlier conjecture that what is learned or strengthened is not a response but an S‐R association. This statement also makes clear that the association is strengthened because the response produces a satisfying or reinforcing outcome. The original law of effect also included the converse (weakening of an S‐R association) if the response results in discomfort rather than satisfaction.

As a part of his discussion of “Laws and Hypotheses for Behavior,” Thorndike also described a second law, the law of exercise, which states that the more often a response occurs in a particular situation, the stronger will be the association of the responses with that situation. Thus, Thorndike's law of exercise is also concerned with how S‐R associations are strengthened. Thorndike considered these mechanisms to be fundamental to the understanding of behavior, noting that the laws of effect and exercise “stand out clearly in every series of experiments on animal learning and in the entire history of the management of human affairs” (Thorndike, 1911, p. 244). Thorndike's discussion of the two laws in combination served to underscore his thinking about S‐R association rather than what Skinner (1938) called “response strength.”

In subsequent writings, Thorndike abandoned the idea that an S‐R association is weakened if the instrumental response results in discomfort but his view of the law of effect otherwise remained unchanged. In his book, Human Nature and the Social Order (Thorndike, 1940), which was published shortly after he retired from Columbia University, Thorndike wrote that “The simple biological fact that, in S→R→E, S→R is especially strengthened by the confirming reaction when E is satisfying supplies the practical teleology which the social sciences have assumed or longed for since Aristotle” (p. 17). This statement illustrates his continued commitment to the proposition that a satisfying response outcome strengthens a connection between the response R and the stimuli that were encountered when R occurred. The statement also illustrates Thorndike's continued conviction that this S‐R learning mechanism is basic to our understanding of behavior. Thorndike was clear and consistent in his statement of the law of effect during his long academic career. Therefore, any subsequent altered statements of the law of effect or misrepresentations cannot be attributed to possible mixed messages that Thorndike might have provided.

THE LAW OF EFFECT IN THE MIDDLE OF THE TWENTIETH CENTURY

The law of effect attracted a great deal of scrutiny and discussion in the decades following publication of Thorndike's 1911 book. In addition, there seems to have been little confusion about what Thorndike had in mind when he stated the law of effect. In particular, everyone seemed to accept the fact that Thorndike considered the law of effect to involve S‐R learning.

In a landmark early “general” textbook, Conditioning and Learning, Hilgard and Marquis (1940) introduced Thorndike's law of effect by quoting Thorndike's statement of the law directly from Thorndike's 1911 book. Thus, Hilgard and Marquis did not offer a reinterpretation or their personal version of the law. However, they may have planted the seeds for confusion by discussing Thorndike's law of effect in a section titled “The Principle of Effect,” which considered more generally the fact that the consequence or effect of a response can alter the future likelihood of that response.

Shortly after the seminal volume by Hilgard and Marquis, Clark Hull published his Principles of Behavior (Hull, 1943), which made generous use of the S‐R learning mechanism in Thorndike's law of effect. However, Hull recast Thorndike's notions of what type of response consequence was required to produce S‐R learning. Hull emphasized that for a response outcome to produce learning it had to reduce a drive state. Except for that modification, Hull stayed close to Thorndike's original statement of the law of effect. He noted that Thorndike's law “is based on a very penetrating bit of scientific insight into the dynamics of adaptive situations in general” (p. 81).

Hull's emphasis on drive reduction as a key variable in instrumental learning generated a great deal of research. Some of that line of research was discussed earlier by Hilgard and Marquis (1940). Studies of the importance of drive reduction for instrumental learning were also reviewed in detail by Postman (1947) in his extensive review of research on the law of effect. However, debates about whether drive reduction was necessary for instrumental learning did not call into question Thorndike's original proposition that instrumental learning basically involved the strengthening of an S‐R connection.

Another prominent contributor to debates about the nature of instrumental learning was B. F. Skinner, whose first major book, The Behavior of Organisms, was published in 1938 (Skinner, 1938). It was in this book that Skinner described the development of his free‐operant method for studying instrumental behavior and the development of lever pressing in rats as a model system for the study of instrumental learning. The free‐operant method contrasted with the discrete trials that were arranged by the experimenter in Thorndike's puzzle box experiments. With the free‐operant method, the rat was “free” to repeat the instrumental response at any time during an extended experimental session. This made the rate of repetitions of the response much more interesting than the latency of the first response when the rat was placed in the experimental chamber. Although Skinner mentioned Thorndike's law of effect in The Behavior of Organisms, he did not say much about it in this book. Ironically, Skinner discussed the work of Pavlov in much greater detail in The Behavior of Organisms than the work of Thorndike.

Skinner's free‐operant method and focus on response rates and patterns encouraged an entirely novel approach to the study of instrumental behavior. This new psychology of learning was the basis for a textbook, Principles of Psychology by Keller and Schoenfeld (1950). Interestingly, like Thorndike, Keller and Schoenfeld were at Columbia University, but they were in the Department of Psychology rather than the Teachers College. Their book included a detailed description of Thorndike's puzzle box experiments as well as a discussion of Thorndike's law of effect, including the familiar quotation of the law from Thorndike's 1911 book. Keller and Schoenfeld emphasized that both Skinner and Thorndike recognized that a positive or reinforcing response outcome can increase the likelihood of repetitions of that response. The fact that Thorndike differed from Skinner in emphasizing S‐R learning did not seem as important to Keller and Schoenfeld. However, it is important for the present discussion to note that Keller and Schoenfeld did not omit Thorndike's proposal that the mechanism responsible for increased operant responding with positive reinforcement was the learning of an S‐R association.

Soon after Keller and Schoenfeld's Principles of Psychology, Skinner's book Science and Human Behavior became available (Skinner, 1953). Skinner dedicated the book to Fred Keller, who had become a close friend. Science and Human Behavior became widely disseminated and helped shape the thinking of new generations of scientists who joined the new specialty in the study of learning known as the experimental analysis of behavior. The book had breathtaking scope, covering not only conditioning and learning but also thinking, social behavior, government, law, religion, and culture, among other topics.

Science and Human Behavior included a description of Thorndike's puzzle box experiments. Skinner's description of the experiments matched the conventional account that a young cat placed in a puzzle box would learn to repeat whatever behavior was required to get out of the box and get a piece of food. Skinner noted that Thorndike “could describe the results simply by saying that part of the cat's behavior was ‘stamped in’ because it was followed by opening of the door.” Skinner then went on to say that “the fact that behavior is stamped in when followed by consequences Thorndike called “The Law of Effect” (p. 60, capitalization as in Skinner, 1953). Notice that this characterization of Thorndike's law of effect ignores the formation of an S‐R association that was such a critical factor not just in Thorndike's thinking but in subsequent research and theory involving the law of effect that I described earlier. Interestingly, Skinner's account of Thorndike's law of effect also departs significantly from the account provided by his friend Fred Keller, who stayed close to Thorndike's original statement of the law in the Principles of Psychology that Keller coauthored with Schoenfeld (Keller & Schoenfeld, 1950).

In addition to providing an incomplete account of Thorndike's law of effect, Skinner criticized Thorndike for measuring the latency of escaping from a puzzle box as the primary behavioral index of learning. It seems to me, however, that with the discrete trials that were necessitated by the puzzle box methodology, Thorndike had few options to document learning without using something like response latency. Thorndike's method was not amenable to measurements of response rate, which Skinner argued was more suitable and which became the standard response measure in the experimental analysis of behavior.

MISREPRESENTATION OF THORNDIKE'S LAW OF EFFECT AFTER 1953

Skinner's discussion of Thorndike's law of effect in Science and Human Behavior seemed to be a watershed that disrupted the consensus that previously existed in what investigators considered to be the law of effect. In his landmark revision of Conditioning and Learning by Hilgard and Marquis, Kimble tried to make sense of the confusion by describing several versions of the law of effect (Kimble, 1961). He suggested that Thorndike's statement is the “weak law of effect” because it did not elaborate what constituted a satisfying response outcome or reinforcer. In contrast, Kimble characterized Hull's version as a stronger form of the law of effect because it included the assumption that a reinforcing response outcome would reduce a drive state. Kimble also referred to what he called the “empirical law of effect,” which just described the fact that response consequences change the future probability of behavior.

Another prominent author who discussed several versions of the law of effect was the behavior analyst A. C. Catania. Catania's textbook Learning first appeared in 1979. I examined the 4th edition which appeared on the 100th anniversary of Thorndike's dissertation (Catania, 1998). Catania commented that the law of effect “went through many revisions,” but he did not describe what those revisions were except to describe a strong and a weak version of the law of effect. To muddy the waters, Catania's strong and weak versions of the law of effect do not match those of Kimble (1961). For Catania, the “strong” version of the law included both positive reinforcement and punishment, whereas the “weak” version just included positive reinforcement. Unlike Kimble, Catania did not discuss the formation of an S‐R association as a major part of the law of effect. Catania also did not carry forward Kimble's suggestion of an “empirical” law of effect.

Most of the misrepresentations of Thorndike's law of effect involve versions that are similar to what Kimble called the empirical law of effect. That is, these descriptions often omit the S‐R learning component that was prominent in Thorndike's original writing and just note that responses that result in a positive outcome are repeated and responses that result in a negative outcome are not repeated. However, the negative side of the law of effect is not always included in the description. In what follows, I provide examples of how various textbook authors have presented the law of effect. This is not a “random” sample of examples but examples from highly reputable authors and prestigious and authoritative books. I will start with textbooks focused on learning and then broaden the examples to include neuroscience.

Not long after Kimble's revision of the Hilgard and Marquis book, Ferster and Perrot published a book titled Behavior Principles in which they sought to systematize what was known at that time about learning from a behavior analytic perspective (Ferster & Perrott, 1968). The first author of this collaboration was the same person who collaborated with Skinner on the massive tome Schedules of Reinforcement (Ferster & Skinner, 1957). Like Skinner's Science and Human Behavior, the Ferster and Perrott book was dedicated to Fred Keller, who, as I mentioned earlier, stayed faithful to Thorndike's original statement of the law of effect. However, it is hard to know what Ferster and Perrot thought about the law of effect because, to my surprise, they made no mention of it in their book.

A number of textbook authors have stayed close to Thorndike's original writing and included the S‐R association as a critical component of the law of effect. In all of the textbooks that I have written (seven editions of the Principles of Learning and Behavior and five editions of the Essentials of Conditioning and Learning), I have always written about the law of effect as involving the establishment of an S‐R association (Domjan, 1993, 1996, 1998, 2000, 2003, 2005, 2010, 2015, 2018; Domjan & Burkhard, 1982, 1986; Domjan & Delamater, 2023). Because others described the law of effect differently, I went back to Thorndike's original text several times over the course of my career to make sure that my presentation of the law was historically correct. Other textbook writers who have also been consistent in their adherence to Thorndike's original writing include Rachlin (1976) and Frieman and Reilly (2016). I found Rachlin's approach to Thorndike remarkable because Rachlin got his PhD at Harvard and was a prominent contributor to the experimental analysis of behavior that Skinner founded.

Just 2 years after the publication of Rachlin's book, Barry Schwartz published the first edition of his textbook Psychology of Learning and Behavior. I examined the 3rd edition of this book, which appeared 10 years later (Schwartz, 1989). Schwartz described Thorndike's puzzle box experiments and was highly respectful of the law of effect, writing that “the principle that has formed the cornerstone of virtually all conceptual and empirical work on the control of behavior by its consequences is the law of effect” (p. 142). However, Schwartz never included the learning of an S‐R association in his remarks about the law of effect.

James Mazur's textbook, Learning and Behavior, first appeared in 1990 and is now in its 8th edition (Mazur, 2017). In this volume, Mazur describes Thorndike's puzzle box experiments and then states the law of effect correctly as involving the establishment of an S‐R association. He also states that another name for the law of effect is “the principle of positive reinforcement” (p. 115), which does not include S‐R learning. In the Glossary at the end of the book (p. 367), Mazur states that the law of effect is “Thorndike's version of the principle of reinforcement” and fails to mention anything about an S‐R association. Thus, Mazur leaves students with an inconsistent account of the law of effect.

Another prominent contributor to the study of learning is Mark Bouton, whose highly scholarly book, Learning and Behavior, is now in its 2nd edition (Bouton, 2018). Bouton discussed the law of effect in several places in his 2018 book. His first description (on page 13) was faithful to Thorndike's original writing. But, when Bouton described four different types of instrumental conditioning in the next chapter (reward learning, punishment, omission training, and escape/avoidance), he wrote that all of these illustrate operations of the law of effect, a claim that goes far beyond Thorndike's original writing. Later in the book (p. 264), Bouton introduced Herrnstein's extension of the matching law to single‐response situations, labeling this “the quantitative law of effect.” That phrase implies that Herrnstein's equation expresses Thorndike's law in a quantitative form. But, that is not the case because Herrnstein's formulation ignores S‐R learning (Herrnstein, 1970). In the Glossary entry for the law of effect (p. 472), Bouton also omits mention of S‐R learning and notes that “nowadays, the term refers to the idea that operant or instrumental behaviors are lawfully controlled by their consequences.” Thus, Bouton's references to the law of effect reflect the various ways that term has been used over the years, with little effort to sort out what is attributable to Thorndike and what is not.

Thorndike was a major contributor to the field of comparative psychology (Galef, 1998). The definitive textbook in this area is Mauricio Papini's Comparative Psychology, which is now in its 3rd edition (Papini, 2021). Papini includes discussions of Thorndike's work along with other research on learning. Papini's first mention of the law of effect (on page 183 of the 3rd edition) is faithful to Thorndike's original statement of the law. However, when Papini defines the law of effect in the Glossary (on page 226), he just states the principle of reinforcement without mentioning S‐R learning. Like Bouton, he provides no reason for this inconsistency.

I also examined two books by prominent learning psychologists whose goals were to promote a particular perspective rather than to present a “general” textbook on learning. In the 3rd edition of The New Behaviorism, John Staddon (2021) introduced Thorndike's law of effect with the familiar quotation from Thorndike (1911), which included the S‐R association. However, Staddon then went on to develop a mathematical model that he referred to as the law of effect (LOE) model, with the implication that this was a quantitative version of Thorndike's law. However, Staddon's LOE model omitted any reference to S‐R learning. To make matters a bit more confusing, although Staddon's LOE model is quantitative, it bears little resemblance to what Bouton referred to as “the quantitative law of effect.” De Houwer and Hughes did not develop a new quantitative model but sought to provide a new framework for thinking about conditioning and learning. In their book, The Psychology of Learning (De Houwer & Hughes, 2020), they cited Thorndike (1911) but omitted any reference to Thorndike's S‐R learning mechanism.

Because learning is a major topic in the neurosciences, I examined several editions of the encyclopedic account of neuroscience titled Principles of Neural Science, edited by the Nobel laureate Eric Kandell and his associates. Each edition of this massive volume has 60–70 chapters, including a chapter on learning (or learning and memory). In the 2nd and 3rd editions (Kandel & Schwartz, 1985; Kandel, Schwartz, & Jessell, 1991), the learning chapter was written by Irving Kupfermann, who cited Thorndike (1911) and described the law of effect as the empirical generalization that reinforced but not punished responses tend to be repeated. But, Kupfermann did not say anything about the learning of an S‐R association. In the 4th edition (Kandel, Schwartz, & Jessell, 2000), Kupfermann was joined by Kandel and Iversen in writing the learning chapter but the account of the law of effect did not improve. The same limitations are evident in the latest edition of the Principles of Neural Science (Kandel et al., 2021) in which the learning and memory chapter was written by Shohamy, Schachter, and Wagner.

WHY IS IT IMPORTANT TO BE CORRECT IN DESCRIBING THORNDIKE'S LAW OF EFFECT?

Science is a dynamic enterprise, with constantly evolving experimental findings and ways of thinking about those findings. Thorndike's significant contributions occurred more than 100 years ago. It is to be expected that ideas about what is responsible for instrumental or operant conditioning have changed significantly since then. One would wish for nothing less. So, why should we be concerned about whether Thorndike's law of effect is represented correctly? I think there are two good reasons.

Being accurate should not require justification

First, I would hope that the importance of being accurate in discussing the work of our predecessors is self‐evident and does not require justification. Scientific discourse is a challenging enterprise under the best of circumstances. The discourse becomes seriously muddled if we are not careful in how we express the ideas that we contribute to that discourse.

Students in particular are ill‐served by being presented with distortions of historically important ideas. Unfortunately, as students turn to artificial intelligence programs to assist in their education, misrepresentations of the law of effect may become even more pervasive. One of the exercises I use in my classes is to pose questions about conditioning and learning to Google or ChatGPT and then ask my students to identify mistakes in the answers that these programs provide. Unfortunately, mistakes are pretty common, especially for the law of effect.

In arguing for adherence to Thorndike's original statement of the law of effect, I am not arguing against changing perspectives. That is not just inevitable in science but a sign of progress. However, I think it muddies the waters to imply that more contemporary descriptions of the law of effect are the same as Thorndike's original statement. I think Kimble (1961) was on the right track when he used different labels for different versions of the law of effect. Without different labels, the implication is that all references to the law of effect refer to the same thing. As I hope I have shown, that is simply not true. Science does not, and should not, stand still. I have no argument with having various versions or updates of the law of effect. However, these updates should be referred to with different labels to clarify that they are not the same as Thorndike's original law. A description of how an update builds on or improves Thorndike's original law of effect would be especially helpful to justify and clarify deviations from Thorndike's original conception.

Thorndike's law of effect in contemporary research

The second reason I urge fidelity to Thorndike's original statement of the law of effect is that there is considerable contemporary interest in the role of S‐R associations in the mediation of instrumental or operant behavior. The concept of an S‐R association has become the focus of contemporary analyses of habitual as contrasted with goal‐directed behavior. I find it particularly interesting that these investigations have been pursued not only by behavioral psychologists (Bouton, 2024; Perez & Dickinson, 2020) but also by social psychologists (Wood et al., 2022) and neuroscientists (Watson et al., 2018). These ideas have also been the focus of successful trade books on how habits are learned and how they may be changed (Duhigg, 2012; Wood, 2019).

Although Thorndike's law of effect is often presented as intuitive and obvious, the details are not at all intuitive. According to Thorndike, a satisfying or positive outcome is necessary for producing instrumental learning, but the learned response does not occur because of the anticipation of the positive outcome. The instrumental response does not represent an effort to obtain a goal object. Thorndike's S‐R mechanism tells us that the learned response R occurs because of the presence of environmental cues S which were present when the response was reinforced. The reinforcing outcome or goal object is not part of the S‐R association. The reinforcer is obviously involved in the learning procedure and the learning process. But, according to Thorndike's law of effect, the reinforcer does not become a part of what ends up being learned. What ends up being learned is just an S‐R association. Once the response has been learned, R occurs automatically, out of habit if you will, as a reaction to S irrespective of the reinforcer.

If R occurs just as a reaction to S, irrespective of the reinforcer, then changes in the value of the reinforcer after learning has occurred should not alter the response. This is a major implication of Thorndike's law of effect that was not appreciated in early empirical tests of the law (e.g., Postman, 1947). However, the effect of reinforcer devaluation on instrumental responding has garnered considerable experimental attention in recent decades (Watson et al., 2024). In an early experiment of this sort, Adams (1982) trained rats to press a response lever for sucrose pellets and then devalued the reinforcer by conditioning a taste aversion to the pellets. Tests in extinction showed that the devaluation procedure disrupted lever pressing after limited training but not after extensive training. This finding suggests that with extensive training the operant response came under the control of an S‐R association and became independent of the value of the reinforcer.

Since early studies of reinforcer devaluation like those of Adams (1982), investigators have continued to focus on the amount of training as a factor that shifts the control of operant behavior to an S‐R association (e.g., Labrecque et al., 2024; Pool et al., 2022). However, it has also become clear that various other parameters of a conditioning procedure, such as the schedule and delay of reinforcement, can also determine whether the operant response becomes independent of the current value of the reinforcer (Bouton, 2021; Kosaki & Dickinson, 2010; Urcelay & Jonkman, 2019). In spite of the complexity of the experimental findings, investigators continue to use sensitivity to reinforcer devaluation as a criterion for deciding whether operant behavior is habitual or goal directed (Perez & Dickinson, 2020; Wood et al., 2022). Thus, whether instrumental responding depends on an S‐R association remains a major topic of research more than 100 years after Thorndike's original work. I find this to be one of the more remarkable stories in the history of conditioning and learning, a story that is missed by those who are not familiar with Thorndike's original statement of the law of effect.

CONFLICT OF INTEREST STATEMENT

There is no conflict of interest.

ETHICS APPROVAL

This is an original theoretical article so does not require institutional review.

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. 10.1002/jeab.70073

DATA AVAILABILITY STATEMENT

No original data are reported in this theoretical article.