Motivating Operations and Discriminative Stimuli: Distinguishable but Interactive Variables (2024)

Abstract

The motivating operations concept has been of considerable interest and practical value to behavior analysts, including practitioners. Nonetheless, the concept has generated substantial controversy and has significant limitations. To address some of these limitations, we suggest that it would be wise to redefine motivating operations, to deemphasize the importance that has historically been placed on subtypes of conditioned motivating operations, to emphasize how motivating operations and discriminative stimuli interact, and to further examine the kinds of environmental changes that alter the reinforcing value of particular kinds of stimuli. These suggestions are detailed elsewhere and summarized in this article.

Keywords: motivating operations, establishing operations, reinforcer effectiveness, stimulus control

The motivating operations (MOs) concept has been of considerable interest and practical value to behavior analysts, including practitioners (e.g., Carbone, 2013; Langthorne & McGill, 2009; Laraway, Snycerski, Olson, Becker, & Poling, 2014; McGill, 1999; Michael & Miguel, 2019). The concept is deemed of sufficient importance by the Behavior Analyst Certification Board (BACB) that the Fourth Edition of the BACB Task List states that behavior analysts should be able to define and provide examples of conditioned and unconditioned MOs (BACB, 2012). Nonetheless, the concept has generated considerable controversy and is not universally applauded by behavior analysts (see, e.g., Catania, 1993; Lotfizadeh, Edwards, & Poling, 2014; Whelan & Barnes-Holmes, 2010). In an attempt to strengthen the concept, we recently proposed substantial changes in how it should be defined and used (Edwards, Lotfizadeh, & Poling, 2019a, 2019b). The analysis we provided in those articles is detailed and somewhat complex, but our major points are simple and potentially important to practitioners. We summarize them in this article. Although the MO concept is relevant to punishment, prior discussions have primarily focused on reinforcement, as do we.

Redefining MOs

The MO concept evolved from Michael’s (e.g., 1982, 1993) extensive treatment of “establishing operations” (EOs). In 2003, he and his colleagues proposed using the term MOs as a replacement for EOs. They indicated that MOs refer to changes in the environment that have two effects (Laraway, Snycerski, Michael, & Poling, 2003). In their words, “In sum, MOs have two defining effects. They alter (a) the effectiveness of reinforcers or punishers (the value-altering effect) and (b) the frequency of operant response classes related to those consequences (the behavior-altering effect)” (p. 212). As we explained elsewhere (Edwards et al., 2019a, 2019b), there is solid evidence that the behavior-altering effect usually, if not always, involves a change in the evocative effects of discriminative and other antecedent stimuli relevant to the consequences of interest. Additionally, nothing of value is gained by considering value-altering and behavior-altering effects as fundamentally different. Therefore, we abolished the distinction and emphasized the importance of antecedent stimuli in our revised definition of MOs: “Motivating operations modulate the reinforcing or punishing effectiveness of particular kinds of events and the control of behavior by antecedent stimuli relevant to those events” (Edwards et al., 2019b, p. 57). These changes are evident in behavioral dimensions that index reinforcing effectiveness (e.g., preference, progressive-ratio breaking points) and antecedent stimulus control (e.g., generalization gradients, probability of responding to an established discriminative stimulus [S^D]). The convention is to refer to MOs that increase reinforcing effectiveness relative to baseline as EOs and to refer to those that reduce it as abolishing operations (AOs; Laraway et al., 2003).

S^Ds Mediate the Effects of MOs

Defining MOs as we do is useful in predicting the specific responses that will occur when a particular MO is in effect. Consider a situation in which lunchtime is approaching and a 7-year-old bilingual child has been playing outside and has not eaten since breakfast, 5 hr earlier. Time passing without eating is likely to increase the reinforcing value of food—that is, to serve as an EO—and on that basis, one can predict that the probability of the child engaging in behaviors that have produced preferred food in the past increases as the time since breakfast grows.

Now, the child goes into the house and smells the delicious odor of potatoes frying in oil. The child sees a familiar adult in the kitchen, hovering over a deep fryer, and approaches. If one grandparent who speaks only English and calls fried potatoes by their American name is cooking, the child asks, “May I have some French fries, please?” But if the grandparent is English speaking and from the United Kingdom, where “French fries” are “chips,” the child substitutes the latter term for the former one in making the request (i.e., in manding). Similarly, if the grandparent speaks only Spanish, the child says, “¿Puedo tener algunas papas fritas, por favor?” If no one is cooking, the child says nothing and makes a peanut butter sandwich.

Although hypothetical, this example is perfectly plausible and a good illustration of how MOs and S^Ds work together to control behavior. Time passing without food is an MO that increases the likelihood of the occurrence of responses that have historically produced food, such as asking an adult for a clearly available and highly desirable comestible. If only one response topography has done so, such as a rat’s pressing of a lever in a test chamber, then one can safely predict that such responding will occur. When multiple response topographies have been successful and are hence members of the same operant response class, the specific one that will occur depends on the S^D that is present at the time of interest. Our hypothetical child speaks Spanish to the grandparent who has reinforced only Spanish-language requests in the past, and speaks English to the other two, who have reinforced only English-language requests. Similarly, the child asks one of them for “chips” and the other for “French fries” because doing so has paid off in the past.

In another example, the class of responses that might be emitted when “knowing the time” would function as a reinforcer will vary depending on who is present. For example, in the presence of a friend, “What time is it?” may be evoked rather than “Excuse me, do you happen to know the time?” which is commonly reinforced and, hence, evoked in the presence of a stranger. Additionally, depending on the characteristics of the stranger or friend, the mand may not be emitted at all. Joint control of behavior by MOs and S^Ds is the rule, not the exception, in everyday life and in clinical situations.

In the clinic and other settings, functional analysis (FA) is a useful tool for developing behavior-change interventions, but a number of “idiosyncratic” variables can influence the outcomes of an FA and contribute to unclear or erroneous results (Schlichenmeyer, Roscoe, Rooker, Wheeler, & Dube, 2013). One such variable is the presence of people who have historically provided the kind of reinforcer being evaluated in an FA condition. Assume that a child’s screaming is the target response and that attention is the reinforcer of interest. A parent has regularly provided attention when this response occurred, but a behavior-analytic therapist has not. Regardless of the momentarily reinforcing value of attention, the response is more likely to occur when the parent (an S^D) is present than when the therapist is present. In fact, the response may not consistently occur if the parent who functions as an S^D is not present, leading to the conclusion that the target response is not attention-maintained, which is incorrect. Practitioners conducting or interpreting FAs should always be aware of the potential interaction of MOs and S^Ds, but this interaction is easily overlooked. A potential benefit of defining MOs as we propose is that the definition explicitly focuses attention on antecedent stimuli, which should reduce the likelihood of overlooking their importance when conducting FAs, and in general. Given this focus of attention, the FA obviously would be conducted twice, once with the parent present and once with the therapist present. Absent this focus, either might well seem to suffice, which could lead to incomplete, hence potentially problematic, findings.

S^Ds Require EOs

Of course, a given S^D evokes the kind of behavior that was previously reinforced in its presence only if an effective EO for the kind of reinforcement previously arranged is currently in effect. The child in our first hypothetical example is highly unlikely to ask either grandparent for French fries soon after eating a big meal but highly likely to do so 8 hr later if no food is consumed in the interim. Michael (1982) clearly recognized the importance of MOs in determining the behavioral effects of S^Ds in his seminal article differentiating the two. This recognition is evident in his excellent definition of the S^D:

It is a stimulus change which, (1) given the momentary effectiveness of some particular type of reinforcement (2) increases the frequency of a particular type of response (3) because that stimulus change has been correlated with an increase in the frequency with which that type of response has been followed by that type of reinforcement. (p. 149)

The “momentary effectiveness of some particular type of reinforcement,” the first component of his tripartite definition, depends on the presence of an EO for that type of reinforcement. In our experience, perhaps because Michael and other theoreticians made it clear, behavior analysts at all levels of experience and training typically recognize that EOs modulate the effects of S^Ds.

What they may overlook, however, is that S^Ds mediate the effects of MOs. S^Ds strongly influence which member of an operant response class will emerge when an EO for a particular kind of reinforcement is in effect, and MOs strongly influence the likelihood that a given S^D will engender responding. A child that historically has garnered attention from one adult by raising a hand and from another adult by screaming will do neither if an EO for attention is not currently operative. If it is, whether the child screams or raises a hand depends on which adult is present. We hope that our reconceptualization of the MO concept helps to emphasize the ubiquity of such joint control of behavior.

Of course, there are some situations where MOs appear to have purely evocative effects (Laraway & Snycerski, 2019; Lechago, 2019; Miguel, 2019). But, even in such cases, antecedent stimuli may play an important role. A case in point is Skinner’s (1957) example of a man dying of thirst and gasping “water” in the absence of any relevant stimuli. Here, the poor unfortunate might well be imagining a person who had provided water in the past (i.e., an S^D). That S^D, a private event, evoked the response. Such an analysis is difficult to test, but that does not render it false or useless. Mediation of MO effects by antecedent stimuli appears to us to be the rule, not the exception.

Although this phenomenon has been largely ignored in discussions of MOs, practitioners should recognize that MOs can affect stimulus generalization around relevant S^Ds. A substantial number of basic research studies indicates that MOs affect both the shape of generalization gradients and the range of stimuli that evoke a response historically controlled by a particular S^D (see Lotfizadeh, Edwards, Redner, & Poling, 2012). In our example of the hungry child, at a given level of food deprivation, the likelihood that a strange adult will evoke a Spanish request for food is likely to vary directly with the degree of physical similarity between that person and the Spanish-speaking grandparent; this is a generalization gradient. As the duration of food deprivation increases, the likelihood that a given stranger will evoke the response increases. So, too, does the range of people who will evoke a request. If the child is very hungry, people quite unlike the grandparent may well be asked for food, although this would not occur with mild food deprivation.

Classifying MOs

The conventional MO concept (Laraway et al., 2003; Michael, 2007; Michael & Miguel, 2019) emphasizes that the reinforcing effectiveness of particular kinds of stimuli is not fixed but instead depends on the historical and current presence of other environmental variables (i.e., operations), and that is perhaps its greatest strength. Michael (e.g., 2007) distinguished two general kinds of environmental alterations that can serve as MOs; he termed them “unconditioned” and “conditioned” MOs (UMOs and CMOs, respectively). Deprivation and satiation are common examples of UMOs, but other operations, including drug administration, temperature changes, and contracting illnesses, would also be classified as UMOs because their effectiveness does not depend on a learning history. The effectiveness of CMOs, however, depends on a relevant learning history. Three types of CMOs have been proposed: surrogate (S), reflexive (R), and transitive (T). Michael and others (e.g., Langthorne & McGill, 2009; Laraway et al., 2003; McGill, 1999; Michael, 2000, 2007; Michael & Miguel, 2019) described CMO subtypes in detail and made the claim that recognizing and distinguishing among them is of conceptual and practical importance. We respectfully disagree (Edwards et al., 2019a, 2019b).

A CMO-S is a stimulus change that acquires its ability to alter reinforcer effectiveness by being respondently paired with (i.e., made predictive of) an established MO. For example, if 11:55 on the clock is frequently paired with a long period of time without food, then, presumably, the appearance of this stimulus (11:55 on the clock) would come to have the same function as the period of food deprivation. Although there is some evidence that CMO-Ss can be established (see Laraway et al., 2014), the conditions under which this occurs are limited, and to our knowledge, there is no published demonstration of the use of a CMO-S to improve behavior in an applied setting.

A CMO-R is described as making “its own offset a reinforcer (e.g., due to its predictive relationship with aversive stimuli)” (Laraway et al., 2014, p. 606). As an example, Langthorne and McGill (2009) described a child for which “seeing his or her mother frown” is “correlated with the subsequent onset of an aversive stimulus, such as being scolded and thus the ‘worsening’ of his or her condition” (p. 26). They went on to explain that “the onset of the mother’s frown may establish its own offset as an effective form of reinforcement and evoke behaviors that have been associated with its removal in the past” (p. 26).

If being scolded constitutes an aversive stimulus—that is, something that the child will respond to escape or avoid—then, by virtue of being predictive of scolding, frowning probably would be established as a conditioned aversive stimulus. The operation through which this occurs is respondent conditioning. Functionally defined, a conditioned aversive stimulus is a stimulus whose onset functions as a positive punisher and whose offset functions as a negative reinforcer. Nothing is gained by calling the onset of a conditioned aversive stimulus a CMO-R. For an organism to escape from a stimulus, that stimulus must be presented. This is a logical necessity. The presentation of the stimulus in no way alters the reinforcing effectiveness of the offset of the stimulus but simply makes the offset possible.

In Langthorne and McGill’s example, the onset of frowning is a requirement for the offset of frowning to occur; the onset of frowning in no way alters the reinforcing effectiveness of the offset of frowning, and, therefore, it cannot be an MO as typically defined (e.g., Laraway et al., 2003). Describing it and similar stimuli as MOs is confusing and directs attention away from the actual operation, respondent conditioning, which establishes them as conditioned aversive stimuli. In our opinion, CMO-R is a term and a concept devoid of value. That said, conditioned aversive stimuli clearly play an important role in the everyday practice of applied behavior analysis, as several studies construed in the context of CMO-Rs demonstrated (for examples, see Carbone, Morgenstern, Zecchin-Tirri, & Kolberg, 2010). Practitioners should be aware of the procedures used in these studies and use them to benefit clients when appropriate, regardless of what they are called.

The CMO-T, sometimes termed the “blocked response” CMO (e.g., Michael, 1988), occurs when the presentation of one stimulus alters the conditioned reinforcing effectiveness of another stimulus, typically in the context of an individual performing a behavior chain. Michael (1982) illustrated this relation with the now well-cited example of an electrician working with an apprentice to remove a piece of equipment from a wall. When the electrician encounters a slotted screw that must be removed to proceed with the job, the reinforcing effectiveness of a slotted screwdriver is established, and the electrician “turns to the apprentice and says ‘screw driver’” (p. 153). This response (saying “screwdriver”) indicates that sight of the slotted screw is a CMO-T that established slotted screwdrivers as positive reinforcers.

Calling the slotted screw a CMO-T does not help to explain the necessary and sufficient conditions for it to alter the reinforcing effectiveness of slotted screwdrivers or to cause the electrician to ask the apprentice for one. Both depend on long and complex histories of reinforcement. As we argued elsewhere (Edwards et al., 2019a, 2019b), the enduring effects of such histories can be explained in terms of behavioral chains and the ordinary behavioral principles (e.g., conditioned reinforcement and conditional discriminations) involved in performing them. Simply put, the CMO-T concept is unnecessary.

Consider the “interrupted chain procedure,” which has been used effectively in mand training (e.g., Albert, Carbone, Murray, Hagerty, & Sweeney-Kerwin, 2012). This procedure involves removing an operandum that is required to advance toward the terminal reinforcer in an operant chain (e.g., the spoon is hidden when the trainee is preparing hot cocoa), and an appropriate mand is required to produce access to the operandum. Rather than describing this procedure as a CMO-T, which does not clarify the conditions responsible for the procedure’s effectiveness, we can conceptualize the procedure as a “chain extension procedure,” in which an additional link is inserted into the chain. The new link consists of a compound S^D (e.g., the open jar of cocoa and the researcher who has previously provided access to spoons following an appropriate mand), a response (e.g., “Spoon, please.”), and a conditioned reinforcer (e.g., the sight of the spoon), which is also the S^D for the next response in the chain.

When considering the history of reinforcement responsible for this outcome, we note that the relevant behaviors (e.g., holding spoons and making scooping motions) have only been reinforced with progression through the chain (e.g., getting cocoa powder into a mug) when the relevant stimuli (e.g., the jar of cocoa) have been present. Such an analysis is both simple and straightforward. Of course, what practitioners really need to understand are the kinds of environmental changes, or procedures, they can arrange to engender the appropriate behavior in their clients in particular circ*mstances. Using technical terms of any sort to refer to those procedures is beneficial only to the extent that doing so leads to more effective everyday or scientific practices. We believe that the terms associated with CMO subtypes have failed to do so in both cases.

Moreover, the term conditioned MO is itself problematic. According to Michael and Miguel (2019), “Motivating variables that alter the reinforcing effectiveness of other stimuli, objects, or events as a result of the organism’s learning history are called conditioned motivating operations (CMOs)” (p. 383). This means that a CMO is effective only when the organism in question has an appropriate learning, or conditioning, history. It is the behavior of the individual that is conditioned, not the operation. The ability of that operation to serve as an MO depends, or is conditional, on the conditioning of the individual. Michael and others well versed in the MO concept know this well, but their choice of terms tempts confusion.

Wanted: Some Good MOs

When we reviewed how authors who published in the Journal of Organizational Behavior Management (JOBM) used terms relevant to the MO concept (Lotfizadeh et al., 2014), two things became abundantly clear. One is that they very rarely used those terms in a manner consistent with the analysis of MOs proposed by Michael and his colleagues (e.g., Laraway et al., 2003; Michael, 2007). Rather than referring to observable changes in the environment that altered the reinforcing (or punishing) effectiveness of other stimuli, JOBM authors typically wrote about the motivational effects of abstruse states or characteristics, such as “organizational culture,” “management style,” or “personality.” The other is that JOBM authors almost never referred to one of the CMO subtypes proposed by Michael.

Of course, the fields of organizational behavior management and applied behavior analysis diverged years ago, to the point that they share a common foundation in the science of behavior but little else. A perusal of the articles included in a comprehensive review of the MO literature by Laraway et al. (2014) indicates that mainstream behavior analysts rarely misuse terms relevant to the MO concept. But those articles also indicated that mainstream behavior analysts consistently ignore variables that serve as MOs for most of the stimuli that serve as (typically, conditioned) reinforcers for significant behaviors of humans, especially verbal humans functioning in complex social environments, which is most people, most of the time. For example, a common way to change the reinforcing effectiveness of a stimulus ignored by the MO concept is to make it a member of a new equivalence class (or relational frame)—one that differs in reinforcing effectiveness from the stimulus in question.

Assume that you are with friends playing five-card stud poker, with one wild card picked by the dealer. The person next to you is dealing and says, “Deuces [cards with the number 2] are wild [i.e., treated as any one of the 13 cards of the same suit, at the player’s discretion].” That statement is highly likely to substantially increase the reinforcing value of deuces, until the time that another card is designated wild. If the dealer’s statement increased the reinforcing effectiveness of deuces, but not of other cards, then by definition it served as an EO. Moreover, this function was obviously dependent on a complicated history involving both you and your friend, which means that it is a conditional (or, to use the currently popular term, conditioned) MO. But what kind of CMO is it? It does not appear to us that it meets the definition of any of the three subtypes discussed by Michael and his colleagues (e.g., Laraway et al., 2003; Michael & Miguel, 2019). Although Belfiore, Kitchen, and Lee (2016) proposed that rules can function as CMO-Ts, we do not find their analysis persuasive and, furthermore, fail to see how applying the label “CMO-T” would help to clarify the function of rules beyond the description provided by Schlinger and Blakely (1987; rules are described as function-altering, contingency-specifying stimuli). Further research is needed to isolate and classify the kinds of environmental changes that alter reinforcer effectiveness, and we encourage work in this area. Isolating and defining MO subtypes is useful to practitioners so long as the categories in their totality are meaningful, account for most of the relevant manipulations, and provide a basis for designing effective behavior-change interventions.

Consider another example. You are at home eating a good meal, having a glass of wine, and listening to music. Your phone rings. You answer and receive unexpected and tragic news: A close friend has just died. With the receipt of that message, the reinforcing effectiveness of your food, your drink, and your music declines precipitously. Because it depends on a learning history, the bad news you received is by definition a CMO for each of those reinforcers. But calling it that explains nothing about the necessary and sufficient conditions for a message to serve as an AO. Naming a phenomenon does not explain it, but it is tempting to assume that it does. This issue is, of course, not unique to MOs, but it is worth emphasizing nonetheless.

As we have discussed elsewhere (Lotfizadeh et al., 2014), in early work that led eventually to relational frame theory, Hayes and Wilson (1993) described verbal statements that alter the reinforcing (or punishing) effectiveness of other stimuli as augmentals. They distinguished two types of augmentals: motivative and formative. Motivative augmentals alter the effectiveness of established reinforcers, and formative augmentals establish new reinforcers. The bad news that reduced the reinforcing effectiveness of several kinds of stimuli in our last example is a motivative augmental. The dealer’s statement that established deuces as reinforcers in an earlier example is a formative augmental. Invaluable conceptual and experimental work has been conducted from the perspective of relational frame theory, but a recent review of the relevant literature suggested that there is very little empirical research on augmentals (Kissi et al., 2017). Further research on augmentals might go far in explaining how verbal statements modulate the behavioral effects of a wide range of consequences. Moreover, no published conceptual analysis of the relation between the MO and motivative augmental concepts has appeared, but a good one would be interesting and potentially important.

Concluding Comments

One simple and important fact underlies the MO concept: Some kinds of changes in an organism’s environment subsequently influence how likely it is that behaviors that historically produced a particular kind of reinforcement will occur, as well as influence the strength of those behaviors if they do occur. The MO concept calls attention to that fact, which behavior-analytic practitioners ignore at their peril. Over time, however, the concept has grown cumbersome, with a proliferation of terms (Catania & St. Peter, 2019) and a level of detail that makes it difficult to understand and, arguably, reduces its value to practitioners. The opportunity cost of mastering the MO concept as it is presented in, say, Michael and Miguel’s (2019) chapter in Cooper, Heron, and Heward’s (2019) influential “white book” is substantial. For us, the trip is not worth the fare.

A ubiquitous problem faced by practitioners is isolating changes in the environment that serve as powerful reinforcers for the behavior of their clients. MOs by definition alter the effectiveness of reinforcers, so their potential clinical significance is readily apparent. To be maximally effective, behavior-analytic practitioners should be familiar with interventions, such as the interrupted chain procedure, that have been studied in the context of MOs and are effective in generating “contrived” reinforcers or in altering the effectiveness of established reinforcers in useful ways. But we do not believe that learning the details of the MO concept in its present form will help them do so.

Parsimony and pragmatism are the philosophical cornerstones of applied behavior analysis (Cooper et al., 2019). We have provided a parsimonious account of MOs. Whether that account is pragmatically justified—that is, whether it helps practitioners better serve their clients—remains to be seen. Our attempt to simplify, and thereby improve, the MO concept is just that: an attempt. Our initial article (Edwards et al., 2019a) is accompanied by nine commentaries (Carbone, 2019; Catania & St. Peter, 2019; Killeen, 2019; Laraway & Snycerski, 2019; Lechago, 2019; Miguel, 2019; Petursdottir, 2019; Pilgrim, 2019; Rehfeldt, 2019) and our response to those commentaries (Edwards et al., 2019b). Although, in our opinion, none of the criticisms contained in the commentaries negated the value of our reconceptualization of the MO concept, a perusal of their contents suggests that many behavior analysts will find fault with some of our conclusions; some will reject all of them. That is fine, but we hope that careful consideration, not blind precedent, guides their judgment. We also hope that critics offer suggestions that help behavior analysts to better understand the variables that modulate the effectiveness of reinforcers, and to manipulate those variables to the betterment of our species and our world. Finally, we hope that members of the BACB seriously consider what is gained by including the MO subtypes on its Task List, and that all behavior analysts disabuse themselves of the notion that our field has developed an adequate explanation of motivation, even in the limited sense of the variables that modulate reinforcement.

Compliance with Ethical Standards

Footnotes

References