Introduction

Motivation is a fundamental aspect of psychotherapy since asking for help does not guarantee that a client will initiate the necessary steps for change to occur. For this reason, various lines of research have studied the way in which therapists can increase the motivation of their consultants. The Transtheoretical model of change (^{Prochaska & DiClemente, 1983)} and the Motivational interview (^{Miller & Rollnick, 1991)} are considered two of the greatest models of therapeutic change. However, several authors have been critical of some of their proposals (e.g., ^{Littell & Girvin, 2002)}, particularly, with the conception of motivation that they underpin (^{Froján-Parga et al., 2010)}.

From these approaches, motivation is considered something that is located within the person. In that sense, the therapist's goal is to assess the client's degree of motivation, so that they will be able to increase it. Additionally, a distinction is made between extrinsic and intrinsic motivation. Whilst the former stands for the motivation to do something in order to meet some external demands, the latter does it for the motivation to do something for its interest (^{Deci & Ryan, 1975)}. It is assumed that intrinsic motivation is the most beneficial for change in comparison to external motivation, that even is considered that leads to engaging in problematic behaviors (^{Clanton-Harpine, 2015)}. From a behavioral perspective, this way of understanding motivation entails several conceptual problems that might affect the control that therapists have over clients’ behavior during the therapeutic process.

The first critic refers to the ontology of motivation. Motivation is not something that can be seen or touched, either is something that one can have, it is a verbal construct that we use for reporting changes in a specific behavior (^{Skinner, 1957)}. The act of treating hypothetical constructs as if they were real and concrete events is called reification, and we make it when talking about motivation. Thus, the “assessment” or the “exercising” of motivation would be a verbal expression and not something that therapists actually do. The second critic refers to the distinction between the inner and the external world. ^{Skinner (1953)}, among other authors (^{Palmer, 2009}; ^{Baum, 2011)}, pointed out that there is no empirical evidence to make that qualitative distinction since the same processes could be applied to private and overt events. The differentiation between intrinsic and extrinsic motivation is an example of this categorization error. For that matter, addressing mechanisms of change would be a more fruitful way of approaching the study of motivation.

An alternative proposal to the traditional conception of motivation was formulated by ^{Michael (1982)}. This author revived the term establishing operation originally proposed by ^{Keller and Schoenfeld (1950)} and defined it as an environmental event that affects an organism's behavior by increasing its frequency of occurrence and the value of the reinforcers associated with it. Deprivation is a paradigmatic example of establishing operation. Nowadays, the concept of motivating operation (MO) is preferred, since it helps to identify distinct motivating operations that previously have been underemphasized and to clarify their effects (^{Laraway et al., 2003)}. This reconceptualization conceives motivation as an observable and manipulable process and not as an internal hypothetical construct that must be inferred.

There are many studies that show the crucial role that motivating operations play in behavior change, and this importance has been demonstrated for different behaviors, populations, and settings. Incorporating MOs into the analysis of problematic behavior has changed interventions by focusing more on modifying the antecedent variables and, thus, preventing the appearance of problem behaviors. This can be clearly observed in how interventions for people with disabilities have been developing during the last years (i.e., ^{Langthorne et al., 2007)}. For example, self-injury in children with autism can be decreased by reducing the deprivation of stimulation, like providing noncontingent access to preferred toys (^{DeLeon, et al. 2000)}. This growing interest led different authors to study the effect of MOs on consumer behavior, as ^{Fagerstrøm and Ghinea (2011)} did for online purchases. They examined the impact of previous customers’ online ratings and reviews on online shopping, showing how others’ opinions influence us in buying or not a product. These results are decisive for the intervention with adults, and therefore for the present study, since the type of stimuli analyzed and conceptualized as MO was verbal (^{Fagerstrøm & Arntzen, 2013)}.

In clinical intervention with adults, therapists usually change the likelihood of clients’ behavior through their verbalizations. Such verbalizations acquire the motivating function as a result of a learning process, and that is why they are called conditioned motivating operations (^{Michael, 2007)}. In this respect, previous studies in the analysis of therapeutic verbal interaction showed that therapists alter the function of their clients’ behaviors through the anticipation of the appetitive and aversive consequences associated with it (^{de Pascual, 2015)}. These sorts of verbalizations (i.e., “if you keep avoiding spiders you will never stop fearing them”) might constitute rules, that is, verbal descriptions of a contingency (^{Vaughan, 1989)}. Ultimately, this is important for two reasons: (1) It means that we can give a certain function to these kinds of verbalizations, that is, rules as MOs (^{Schlinger, 1990)}, and (2) it shows that clients’ behaviors can be modified not only by direct contingencies (discrimination and reinforcement) but also by the presentation of described contingencies (^{Vargas-de la Cruz et al., 2017)}.

Despite that, an important question is left: how these verbalizations can have an impact on clients’ behavior? In other words, what are the mechanisms that explain why a behavior is altered through language? Up to date, the explanations based on operant conditioning have led the field in the study of language from a behavioral approach. Relational Frame Theory (RFT) constitutes an example of this (^{Hayes et al., 2001)}. RFT is based on the studies of equivalence relations (^{Sidman, 2000)} and it aims to explain how the function of stimuli is transformed through language. In very abridged form, the principal idea is that subjects would learn to relate stimuli in different ways through multiple exemplar training and, based on these different established relations (which are called frames), the function can be transformed. For example, if a person is reinforced for responding that “A is more than B” and later B acquires an aversive function, he or she would react more aversively to A than to B based on the previous relation or frame. However, this proposal is not exempt from critics. ^{Tonneau (2001)}, in his review of equivalence relations, claims that the role of pavlovian conditioning has not received enough attention in the explanation of function altering.

During the last few years, some authors have studied the role of respondent procedures in the emergence of language functions (^{Delgado & Rodríguez, 2020}; ^{Tonneau et al., 2006};). That has opened the door to reconsidering the role of classical conditioning in the explanation of language. The analysis of the explanatory potential of classical conditioning about language began to be studied before the mid-20th century, in the context of mediated generalization studies (^{Eilifsen & Arntzen, 2021)}. For some authors, classical conditioning would be the process behind the referential capacity of words, that is their capacity to elicit the same reactions as their referents. Different studies revealed how the pairing of words with other words could change the way people responded to them without a process of reinforcement. In this line, ^{Staats & Staats (1957)} conducted experiments in which a nonsense syllable (i.e., “YOF”) was paired with positive words (i.e., “Beauty” and “Healthy”) and they demonstrated that the syllable was able to elicit positive responses by its own afterward.

In the light of the foregoing, the present study aimed to analyze the verbal sequences that might function as MOs in the therapeutic process. The main goal was to analyze how therapists use them and explore the role that they might play in the clinical change. To do that, the latest advances in therapeutic verbal interaction were considered. The ACOVEO System (^{de Pascual-Verdú et al., 2019)}, a coding system that describes the function that a verbalization or set of verbalizations can have according to their possible role within the functional chain, notes this progress.

In previous versions of this coding system (e.g., SISC-INTER-CVT), certain categories were based on morphological descriptions, while others were based on functional descriptions (^{Ruiz-Sancho et al., 2015)}. In this updated version all the categories are functional. That is the case with the clinical information that therapists provide to clients. The SISC-INTER-CVT made a distinction between the informative and the motivational categories, however, the ACOVEO System considers the clinical information as a MO, since both are considered rules. This is very important for the study of MOs because they can be extended to other verbalizations apart from the anticipation of behavior consequences. This, and the consideration of the contribution of pavlovian conditioning to language were the prompts to start the study of pairings in verbal interaction. In the ACOVEO System, verbal pairings are considered the third type of MOs, since they are also rules that pursue changes in behavior by altering the appetitive and aversive function of certain stimuli. Besides, they constitute the connection between the current study of language and the traditional pavlovian research. In this work, the anticipation of consequences and the pairings earn significant attention since they are defined by the establishment of a relationship between the client's behavior and appetitive or aversive events- whereas the clinical information verbalizations establish other kinds of contingencies - and the elicit function has been proposed as the key aspect of motivation (^{Dougher & Hackbert, 2000)}

In line with the aforementioned, the following hypotheses are proposed according to the use and the role of the different motivating operations:

Hypothesis 1: Focusing on their shared function, it is expected that the verbalizations with motivating operation function will be emitted together by the therapists in their speech. That is, verbal motivating operations will be preceded and followed more by other verbal motivating operations than other types of verbalizations of the therapists.

Hypothesis 2: Regarding the type of contingencies that verbal motivating operations specify, it is foreseeable to find different relations between them and the clients’ verbalizations. Specifically, when the clinical information is combined with either the anticipation of consequences or the pairings it is expected to find a greater association with clients’ agreement than when it is emitted alone.

Method

Sample

In this study, recordings of 40 clinical sessions, involving 9 adults with different problematics, were observed. Six cognitive-behavioral therapists with different years of clinical experience participated in the study. All of them rendered service at ITEMA, a private clinical center in Madrid (Spain). Psychological interventions were individual and lasted approximately one hour per session. Therapists delivered their usual therapy and the observations were collected eventually, so neither the therapists nor the clients knew the specific objectives, the hypotheses, or the design of the study. All participants obtained written informed consent and approval from the Ethics Committee of the Autonomous University of Madrid. The specific characteristics of the sample are described in Table 1.

Table 1: Characteristics of clients and therapists. 

^Note.W= Woman M= Man

Design and Variables

The study design was prospective observational. The therapist and the client's verbal behavior were the variables considered in this study. The predictor variables of the study were the categories of the therapist coded as Motivating operation with clinical information (MO Information), Motivating operation with consequences (MO Consequences) and Motivating operation with pairing (MO Pairing). For the first hypothesis, the criterion variables were the categories of the therapist coded as Discriminative stimulus, Instructional discriminative stimulus, Reinforcer, Punishment and the previous three MOs. For the second hypothesis, the criterion variables were the categories of the client coded as Agreement and Disagreement. Two different observational category systems were employed for their codification. The ACOVEO System (^{de Pascual-Verdú et al., 2019)} was used for the identification of the therapist's verbalizations and the SISC-CVC (^{Ruiz-Sancho et al., 2015)}, for the client's verbalizations. A detailed description of both observational category systems is presented in Table 2. Additionally, a third variable was considered to group the different sessions. In order to have a representative sample of the therapeutic process, sessions were selected according to the clinically relevant activities undertaken by the therapists. These were Evaluation, Explanation, Treatment I, Treatment II and Consolidation (see Table 3 for their definition). Sessions were already assigned to one of these clinically relevant activities in a previous study that conducted discriminant analyses (see ^{Froján-Parga et al., 2011)}.

Table 2: Descriptions of coding categories of ACOVEO System for therapists and SISC-CVC for clients. 

Table 3: Descriptions of the clinically relevant activities undertaken by the therapist. 

Results

Descriptive Analyses

The percentage of appearance of the different categories of the therapist's verbal behavior is presented in Table 4. It is observed that the most frequently coded category is Antecedent stimuli, followed by Motivating operations and Consequent stimuli, respectively, in all the clinically relevant activities except in Treatment II where Consequent stimuli percentage is higher than Motivating operations. Regarding the motivating operations themselves, the one that represents the highest percentage is the category of MO Information, followed by the MO Consequences and, finally, the category of MO Pairing, in all the clinically relevant activities, except in Evaluation where MO Pairing percentage is greater than MO Consequences.

The analyses show that the relations between categories remain constant along the therapy, although they also show that the percentages of each category change. Antecedent stimuli are higher in Evaluation than in Explanation, Treatment and Consolitation in general, while Consequent stimuli are higher in Treatment and Consolidation in general than Evaluation and Explanation. Motivating operations are higher in Explanation, Consolidation and Treatment than in Evaluation, but attending to each motivating operation, MO Information is greater in Evaluation, Explanation and Consolidation than Treatment I and II, while MO Consequences and MO Pairing are higher in Treatment I and II. This might reflect the particular role of each type of therapist's verbalizations that will be discussed below.

Table 4: Descriptive statistics for the variables of the therapist's verbal behavior for each clinically relevant activity. 

^Note.Eval = Evaluation, Expl = Explanation, Treat = Treatment, Consol = Consolidation.

Sequential Analyses

To test hypothesis 1, the degree of association between the therapist's behaviors categorized as MOs and all the categories of the therapist's verbal behavior was calculated based on the -1 and +1 delays. The results obtained are collected in Table 5.

Table 5: Relationship between motivation operations and therapist's verbal behavior for -1 and +1 delays. 

^Note.MO Information= Motivation operation with clinical information; MO Consequences = Motivation operation with consequences; MO Pairing = Mo-tivation operation with pairing; DS = Discriminative stimulus; IDS = Instructional discriminative stimulus; REI = Reinforcer; PUN = Punishment

Q = Q de Yule

r = Adjusted residuals

^*p < .05

As can be observed, these results support hypothesis 1. There is a positive and significant relationship between the different categories of MOs among themselves more than with other categories. Regarding the category of MO Information, there is a strong positive association between it and the categories of MO Consequences and MO Pairing for -1 and +1 delays, but not with itself. Furthermore, this category also shows a positive association when the delay is -1 with the Reinforcer and Punishment categories. This positive association can also be observed between this category and the Instructional discriminative stimulus category, for both delays.

With regards to the MO Consequences category, it shows a positive association with all the motivating operations for both delays, but also no correlation with any other category. Something similar happens with the MO Pairing category, it shows a positive correlation with MO Information and MO Consequences, for +1 and -1 delays, but not with itself (the correlation is positive but not significant). This category also shows a positive association with the Reinforcer category when the delay is -1, but it does not show any other association with other categories.

To test hypothesis 2, two new categories were formed. The GSEQ program only analyses sequences of two behaviors, but on this occasion analysis of three behaviors sequences was required. Fortunately, the GSEQ program allows to combine of two different categories, so when two motivating operations categories appeared next to each other the program could recognize both as one category. This was useful because the sequential analyses could be run using these new categories and a more complex verbal interaction between therapist and client could be studied. Taking this into account, the first formed category was the combination of the category MO Information and the category MO Consequences. The second formed category was the combination of the category MO Information and the category MO Pairing.

As can be seen in Table 6, the second hypothesis is partially supported by the data. The category MO Information was positively associated with both client's categories, Agreement, and Disagreement, in a +1 delay. However, when this category was combined with MO Consequences a positive association with Agreement was found, but not with Disagreement. In the case of the combination of MO Information with MO Pairing a similar result was found; there was a positive association between this category and client's agreement, despite it was not statistically significant (but close). No association with the client's disagreement was found.

Table 6: Relationship between motivation operations and client's verbal behavior for +1 delay. 

^Note.MO Information = Motivation operation with clinical information; MO Consequences = Motivation operation with consequences; MO Pairing = Motivation operation with pairing

Q = Q de Yule

r = Adjusted residuals

^*p < .05

With regards to the degree of association, although a higher relation was found between the MO Information and Agreement than in the case of the combined categories, no relation was found between the latter and Disagreement. This could not be said for the former since almost the same degree of association was found for Agreement and Disagreement. In other words, a more direct association was found between the combined categories and the client's agreement. The possible explanations of these results and the implications for therapeutic interventions are discussed in the next section.

Discussion

The aim of the present study was to test the role of different verbalizations with a MO function in therapeutic verbal interaction. The results confirmed the proposed predictions. The descriptive and the sequential analysis allowed to test how the different verbal MOs are used in therapy. MOs represent a great percentage of the verbalizations emitted by the therapists (from 15,4% to 33,3%), which probably underscores the importance of these verbalizations during the therapeutic process. However, this percentage is lower than the percentages found in previous studies (^{Froján-Parga et al., 2011)}. This could be explained by methodological issues. With the previous observational system, the SISC-INTER-CVT (^{Ruiz-Sancho et al., 2015)}, the Informative Function category, and the Motivating Function category (currently MO Information and MO Consequences in the ACOVEO System, respectively) were state categories. While event categories were coded by their frequency of occurrence, state categories were coded by their duration, which makes the comparison of data more difficult.

Regarding the percentage of occurrence of each motivating operation, the most frequently coded category was MO Information followed by MO Consequences and MO Pairing, in that order. These differences in the emission of the motivating operation may also be due to methodological issues. To ensure a high scientific standard, the ACOVEO System has restrictions when coding MO Consequences and MO Pairing. This means that to code one of these verbalizations, the used terms need to be clearly appetitive or aversive for the majority of the verbal community. For example, if the therapist says that “Tennis is easy”, this could not be coded as an appetitive or aversive pairing, since the term “easy” could be neutral, appetitive, or aversive depending on the client and the context. However, the chances of error are reduced if this verbal community criterion is applied. Thus, if the therapist instead says “Tennis is very fun”, this could be coded as a MO Pairing since “very fun” is an appetitive term for the majority of the verbal community. For this kind of matter, inter-judge analysis becomes essential.

According to the first hypothesis, data showed that the association between the motivating operations was higher among them than with the rest of the therapist's verbal behavior. This association could be specially observed when the motivating operations were different, for the three of them. However, different associations were found when the relation of the motivating operations with the same motivating operation was studied. While MO Consequences correlated with itself, MO Information and MO Pairing did not. These results also could be explained because of methodological reasons. Given the description of the category MO Information and the criteria to code a part of therapist speech as such, it is unlikely to register two of these motivating operations in a row. MO Information account not only for one verbalization but for a group of verbalizations of the same topic. It is not very likely that therapists address different topics one after another, and these results show that. In the case of the MO Pairing, a positive association among itself was also found, but it was not statistically significant. As mentioned before, the criteria for coding pairings is very strict, so it might occur that at this first stage of research (and also in a group design) some pairings could not be detected yet and, thus, the correlation between this category with itself it is being underestimated.

In any case, these results showed that these verbalizations are emitted together, which might indicate that they are used in specific moments in therapy and for common goals. These data also support the idea of these verbalizations having the same function and thus, they might be analyzed as a chunk. The codification of therapist verbal behavior using chunks has been shown to be useful in previous studies (^{Galván-Domínguez et al., 2020)}, providing a new level of analysis of the verbal interaction that can portray better what occurs during clinical sessions.

Associations of the motivating operations with clients’ verbal behavior were also explored. In this regard, positive relations were found between these sorts of verbalizations and clients’ agreement, specifically when the different motivating operations were combined. Although the category of MO Information showed a great association with the client's agreement, it also showed a great association with disagreement. By contrast, even if the degree of association between the combinations of motivating operations (MO Information + MO Consequences and MO Information + MO Pairing) showed a lower association with the client's agreement than the just mentioned association, they did not show any relation with the client's disagreement.

In the applied field, these results could be useful for therapists when they try to increase or decrease a client's behavior. The different effects that therapist's contingencies descriptions have on client's concurrence, could make therapists focus more on the anticipation of consequences and pairings than they used to do before. Giving general information is important and it is associated with desirable change, but in the light of these results more concrete specifications with an elicit component showed a greater relation with the client's agreement. Furthermore, therapists could benefit from the reconceptualization of motivation, losing the fear of directly motivating their clients. The distinction between extrinsic and intrinsic motivation does not seem to be valid when context accounts for all stimuli, either public or private. Also, the superior status of intrinsic motivation ceases to be useful when therapists know the mechanisms that explain why changes in behavior occur and are maintained out of the session without their presence. In this sense, the directivity of interventions does not necessarily diminish motivation, as stated by some authors (^{Lepper et al. 1973}; ^{Warneken & Tomasello 2014)}. By changing their concept of motivation, therapists could focus on the alteration of clients’ context and not on the evaluation of their (de)motivation. In sum, therapists could gain greater efficacy and control in their treatments.

In the experimental field, these results make it necessary to pay attention to the study of the mechanisms that might explain this differential effect of MOs on clients’ behavior change. From our point of view, studies that test the role of classical conditioning in verbal behavior and thus in verbal change are promising (^{Tonneau, 2004)}. Far from being a simple process, classical conditioning may contribute to the explanation of human complex behavior, for example, extra-therapeutic change. The emergence of behaviors not directly reinforced (e.g., transfer of functions) is key to understanding how the therapist can influence a client's behavior in an out-session context. Although operant processes are essential to understanding clinical change, proposals that dominate the field of human complex behavior research as RFT (^{Barnes-Holmes et al., 2001)} are being questioned because of their enormous departure from their foundational principles (^{Sidman & Tailby, 1982)}, as well as the limited attention they paid to respondent principles (Tonneau, 2001).

Regarding the limitations of the study, the sample is the first thing to comment on. The size of it was smaller than previous studies of the therapeutic process (i.e., ^{Ruiz-Sancho et al., 2015)}, although at the same time it was a bit larger than specific studies of verbal MOs (^{de Pascual-Verdú & Trujillo-Sánchez, 2018)}. The number of sessions selected was because all cases must have sessions with all the clinically relevant activities, and due to different reasons, not all cases recorded had them. This could affect the number of verbalizations coded and thus the significance of some results. Although the cases were not selected attending to therapists or clients’ characteristics, the impossibility of selecting them randomly from the bank could affect the results as well. In addition, recordings come from the same clinic, which means that the generalization of results is limited. Besides, this was not an experimental study, so the results could not be interpreted in terms of the efficacy of an intervention. The direct association between MO with elicit function should not be interpreted as if they cause clients’ agreement. Further studies can experimentally manipulate the verbal behavior of the therapist and control the possible confounding variables. A non-manipulative study like this one does not meet the requirements that the experimental analysis of behavior requires to label a behavior as elicit. Although inter-judge accordance guarantee objectivity, experimental studies are needed to test the function of the stimuli. Despite this, the results are promising concerning the incorporation of mechanisms like pavlovian processes into the study of motivation and the explanation of the clinical change.