Scorekeeping in a Therapeutic Language Game

Abstract

In ‘Scorekeeping in a Language Game’, David Lewis famously compares conversations to playing baseball. Just like baseball, conversations have a score which, together with rules for correct play, determines which utterances are acceptable or even true in the course of a conversation. For all similarities, however, there is a crucial difference between conversations and baseball games. Unlike the score of a baseball game, conversational score adjusts in such a way that the utterances made in the course of a conversation count as correct play. This is also known as accommodation. Starting from this scorekeeping approach to language use, the overall aim of the present paper is to provide a better understanding of how the methods and interventions of talking therapies work from a linguistic point of view. According to the scorekeeping model, the methods and interventions of talking therapies are effective by changing the score of the therapeutic conversation, in particular in the form of accommodation. This has significant implications for the therapeutic practice, as it highlights the importance of training therapists in the linguistic aspects of therapeutic methods, in particular in the use of accommodation.

1. Introduction

According to a definition developed by Norcross, which is also advocated by the American Psychological Association, ‘psychotherapy is the informed and intentional application of clinical methods and interpersonal stances derived from established psychological principles for the purpose of assisting people to modify their behaviours, cognitions, emotions, and/or other personal characteristics in directions that the participants deem desirable’ (Campbell, Norcross, Vasquez, and Kaslow, 2013). On this conception, the effectiveness of a therapeutic method/approach can be measured by the degree to which the participants have been able to modify their behaviours, cognitions, and/or emotions in the desired ways. Here the gold standard are randomised controlled trials where one group, the experimental group, receives the intervention being assessed, while the other, usually called the control group, receives alternative treatments or no intervention. However, since supposedly different forms of psychotherapy have shown similar effectiveness (Brown and Lent, 2008), there is considerable debate about which factors are causally responsible for therapeutic change. For instance, there are those who advocate that the specific methods used, also known as empirically supported treatments (ESTs), are primarily causally responsible for the outcome (e.g., Chambless and Crits-Christoph, 2006; Baker, McFall, and Shoham, 2008; Barlow, 2004; Chambless and Hollon, 1998; Siev, Huppert, and Chambless, 2009). Advocates of a common factor approach (CF), on the other hand, maintain that the similar effectiveness of different forms of psychotherapy is best explained by factors that are shared by the different treatments, such as the therapeutic relationship and empathy (e.g., Watts et al., 2013; Smith and Glass, 1977; Stiles, Barkham, Mellor-Clark, and Connell, 2008; Stiles, Barkham, Twigg, Mellor-Clark, and Cooper, 2006; Wampold et al., 1997). Finally, a third group argues for a mixed approach, according to which both common factors and specific methods are causally responsible for therapeutic change. For example, according to Norcross and Lambert's (2019) estimation of the percentage of psychotherapy outcome variance as a function of therapeutic factors, common factors account for roughly 30% of success, whereas ESTs explain approximately 15% of the variance. The other two factors being extra-therapeutic change (40%) and expectancy (placebo effect) (15%).

Surprisingly, however, neither CF nor mixed approaches mention a striking commonality of a large number of therapeutic approaches, i.e., that the therapy is done verbally, in the form of a conversation between therapist and client. This commonality raises the question of what role language plays in the effectiveness of the methods and interventions of such talking therapies.¹ An approach to language use that has proven very fruitful in investigating the social aspects of language, in particular in the context of hate speech and pornography, is David Lewis’ (1979, 1980) scorekeeping model. According to this theory, conversations have a score which, together with rules for correct play, determines which utterances are acceptable or even true in the course of a conversation. The paradigmatic example are presuppositions. For instance, an utterance of ‘Fred's children are asleep’ is only acceptable if its presupposition that Fred has children is already entailed by the shared beliefs (common ground) of the conversational parties, making the latter an important component of the conversational score. However, conversations not only have rules stipulating how the score determines what can be said in the course of a conversation. They also have rules stipulating how what is said determines the score. A special type of such rules are so-called rules of accommodation. Accordingly, a conversational score (usually) adjusts in such a way that the utterances made in the course of a conversation count as correct play. For example, if it is not already part of the common ground that Fred has children, an utterance of ‘Fred's children are asleep’ (usually) results in the common ground adjusting in such a way that the utterance is acceptable, i.e., the conversational parties start to believe that Fred has children. Other components of the score that are subject to accommodation are boundaries between relevant possibilities and ignored ones, boundaries between the permissible and the impermissible, and the like.

Starting from this scorekeeping approach to language use, the overall aim of the present paper is to provide a better understanding of how the methods and interventions of talking therapies work from a linguistic point of view. According to the scorekeeping model, the methods and interventions of talking therapies are effective by changing the score of the therapeutic conversation, in particular in the form of accommodation. This can affect different components of a conversational score. For example, the effectiveness of a therapeutic method can consist in a change in belief or an expansion of the possibility/permissibility range (to name just a few), all components of the conversational score. In addition, therapeutic methods and interventions can be effective by making certain utterances of the client possible: utterances about certain events or one's attitudes and feelings (Pennebaker, Kiecolt-Glaser, and Glaser, 1988; Lieberman et al., 2007). Using the scorekeeping model, this will be explained by the fact that the therapeutic method or intervention changes the conversational score so that new utterances count as correct play in the course of the therapeutic conversation. In this way, the scorekeeping approach to language use promises a completely new understanding of talking therapy and its effectiveness, one that has significant implications for the therapeutic practice, as it highlights the importance of training therapists in the linguistic aspects of therapeutic methods and interventions, in particular in the use of accommodation.

Note that the aim is not to provide a common factor explanation of the effectiveness of talking therapies. The results of the present paper should rather be incorporated into a mixed approach, which also considers other therapeutic factors, such as the therapeutic relationship and the use of ESTs. However, most of these factors cannot be assessed independently of the fact that talking therapy is done verbally. For instance, the therapeutic relationship is to a large degree a result of the conversation between therapist and client. Similarly, the specific methods and interventions of the different approaches of talking therapy are delivered verbally, which is why, in the terminology of the scorekeeping model, they are just different ways of changing the score of the therapeutic conversation. Or to say it with Carr: ‘therapists must engage in specific forms of therapy for common factors to have a medium through which to operate’ (Carr, 2008, p. 53). Hence, the fact that talking therapies are done verbally, in the form of a conversation, is not only a common factor of these therapies, but it is also a basic factor which significantly affects most of the other therapeutic factors. This emphasises again the importance of investigating the linguistic aspects of psychotherapy by resorting to insights from philosophy of language and linguistics. The present paper is intended to act as a pioneer in this new field of research.

In the next section, I will discuss the scorekeeping approach to language use in more detail. Starting from this, in Section 3, I will develop an explanation of how the methods and interventions of talking therapies work from a linguistic point of view. This will be illustrated using the example of systemic questions and the use of transference and countertransference in psychodynamic approaches. Concluding, I will briefly address the question how the therapeutic relationship affects accommodation, pointing towards a completely new understanding of both the effectiveness of the therapeutic relationship and the linguistic phenomenon of accommodation.

2. The Scorekeeping Approach to Language Use

In his seminal paper ‘Scorekeeping in a Language Game’, Lewis famously compares conversations to playing baseball. For Lewis, just like baseball, conversations have a score which, together with rules for correct play, determines which utterances are acceptable or even true in the course of a conversation. While the score of a baseball game is a septuple of numbers ⟨r_v, r_h, h, i, s, b, o⟩ consisting of the runs of the visiting team, the runs of the home team, the half, the inning, the strikes, the balls, and the outs, the score of a conversation consists of the underlying circumstances that make a conversation possible. These circumstances contain, among other things, a set of presupposed propositions, a boundary between relevant possibilities and ignored ones, a boundary between the permissible and the impermissible, standards of precision for the uses of vague terms, and the like. For example, an utterance of ‘Italy is boot-shaped’ is true if and only if the standards of precision for ‘boot-shaped’ in place are such that Italy counts as boot-shaped. Lewis compares this to the fact that in a baseball game it is correct play that the batter advances to first base if the score counts four balls.

For all similarities, however, there is a crucial difference between conversations and baseball games, viz. regarding the rules that determine the kinematics of the score. Unlike baseball, conversations have so-called rules of accommodation. For example, if in a baseball game the batter walks to first base after only three balls, his behaviour does not make it the case that there are four balls and his behaviour is correct. Conversations are different. Lewis illustrates this using the example of presuppositions. Even if the score of a conversation does not already include the proposition that Fred has children, an utterance of ‘Fred's children are asleep’ results in the score adjusting in such a way that the utterance counts as correct play (unless somebody objects ‘Hey, wait a minute, I didn't know that Fred has children’). A similar rule has been discussed by Karttunen (1974) and Stalnaker (1974). However, Lewis points out that accommodation rules also exist in connection with other components of the conversational score. For instance, an utterance of ‘Italy is boot-shaped’ can lead to the standards of precision for ‘boot-shaped’ adjusting in such a way that the utterance comes out true. Similarly, in a master/slave game where two people are both willing that one of them should be under the control of the other, an utterance of the master that such-and-such courses of action are permissible/impermissible expands/restricts the permissibility range if this is required to make the utterance of the master true.

Against the scorekeeping approach, Abbott (2008) objects that accommodation is much like a magical process. Gauker (1998, 2008) expresses similar concerns. However, starting from Stalnaker's (1998, 2002) conception of a score as common ground, Thomason (1990) and von Fintel (2000, 2008) could very much demystify Lewis’ accommodation rules, at least those for presuppositions. According to Thomason and von Fintel, following an utterance of ‘Fred's children are asleep’, the conversational parties adjust their beliefs because they recognise: (i) that the speaker wants to add to the common ground that Fred's children are asleep, and (ii) that this is only possible if it is already entailed by the common ground that Fred has children. Von Fintel mentions two cases where the hearers are disposed to cooperate with the speaker in this way: (i) the hearers may be genuinely agnostic as to the truth of the relevant proposition, assume that the speaker knows about its truth, and trust the speaker not to speak inappropriately or falsely; (ii) the hearers may not want to challenge the speaker about the presupposed proposition because it is irrelevant to their concerns and the smoothness of the conversation is important enough to them to warrant a little leeway. Hence, for Thomason and von Fintel, accommodation is not something magical, but something utterly pragmatic.

A significant development and application of the scorekeeping model has been in connection with the philosophical discussion of hate speech, i.e., speech that promotes hatred toward an individual or group on the basis of their race, gender, nationality, or the like. Philosophers such as Langton and West (1999), Langton (2012), West (2016), McGowan (2003, 2004, 2009, 2019), and Popa-Wyatt and Wyatt (2018) have argued that accommodation plays a crucial role in connection with the attitudinal, emotional, and behavioural changes that come with hate speech. For example, if a speaker utters the sentence ‘Even George can read’, the hearers tend to adjust their beliefs so that the utterance counts as correct play, i.e., they start to believe that George has intellectual deficits. Langton (2012) extends this explanation to desires and feelings. In this way, the scorekeeping model provides an explanation of how speech, and in particular hate speech, brings attitudinal and emotional changes about, even if the speaker has no intellectual or social authority whatsoever (Langton, 2018). This does not only hold for conversational contributions using presupposition accommodation, but also for utterances that lead to the adjustment of other components of the conversational score, such as the standards of precision for vague terms. For instance, if a speaker utters the sentence ‘Italy is boot-shaped’, the standards of precision for ‘boot-shaped’ tend to adjust so that the utterance comes out true and, thereby, add that Italy is boot-shaped to the common ground of the conversational parties.

If successful, such conversational contributions even change what future moves are permissible, or even possible, in the course of the conversation. For instance, if successful, an utterance of ‘Even George can read’ may license making snide remarks about George or jokes at his expense. Therefore, McGowan (2003, 2004, 2009, 2019) calls conversational contributions, especially those using accommodation rules, covert exercitives. Exercitives are speech acts that enact rules (or permissibility facts). For example, if a lawmaker declares that it is not allowed to smoke in the vicinity of a pregnant woman, this utterance (said under the appropriate circumstances) makes it the case that it is not allowed to smoke in the vicinity of a pregnant woman. Usually, such speech acts require the speaker to have a certain kind of authority in order to be successful. However, McGowan notes that conversational contributions can change what future moves are permissible in the context of a conversation without satisfying the felicity conditions of standard exercitives, such as authority, speaker intention, and hearer recognition. As we have seen with the example of ‘Even George can read’, this can be exploited in socially relevant contexts such as hate speech.

The example of the lawmaker also shows that some utterances depend for their appropriateness on the social role of the speaker. Therefore, Popa-Wyatt and Wyatt (2018) add so-called discourse roles to the conversational score. These roles determine what can be said by the conversational parties and are typically inherited from their social roles and the norms that come with them. An utterance of ‘I hereby declare you husband and wife’, for instance, is only appropriate if the speaker has a social role that allows them to marry people. In addition, the appropriateness of an utterance can also depend on the discourse roles of the hearers. For example, Popa-Wyatt and Wyatt point out that an utterance of ‘Johnny is a faggot’ is only appropriate if Johnny has been assigned a certain subordinating discourse role. If this discourse role has not already been assigned to Johnny, the conversational score will straightaway adjust so that the utterance counts as correct play. This, in turn, will license further derogatory remarks about Johnny, emphasising again how accommodation can be exploited in hate speech.

3. Scorekeeping in a Therapeutic Language Game

According to the scorekeeping approach to language use and its different developments, the methods and interventions of talking therapies are nothing other than covert exercitives which change the score of the therapeutic conversation. This suggests that the effectiveness of these methods and interventions is largely due to this change in the score. After all, many components of a conversational score are highly relevant when it comes to the effectiveness of talking therapies, such as the shared beliefs of the conversational parties, their discourse roles, and the possibility/permissibility range, to name just a few. Hence, by changing the score of the therapeutic conversation, the therapist can change the client's beliefs, expand their possibility and permissibility range, and the like. Furthermore, by changing the score of the therapeutic conversation, the therapist can make certain utterances of the client possible, utterances about certain events or one's attitudes and feelings. As Pennebaker, Kiecolt-Glaser, and Glaser (1988) and Lieberman et al. (2007) point out, this is often what makes a therapeutic method or intervention effective.

According to the scorekeeping model, there are two ways in which the therapist can change the score of the therapeutic conversation. First, the therapist can change the score of the therapeutic conversation with their own conversational contributions, in particular those using rules of accommodation. Since such utterances, in turn, change what future moves are permissible, or even possible, in the course of a conversation, in this way the therapist also indirectly affects future scores of the therapeutic conversation. In addition, the therapist can affect the score of the therapeutic conversation by blocking the conversational moves of the client. In this case, the therapist makes a conversational move that prevents an utterance of the client from changing the score in a certain direction. Again, such blockings do not only prevent the conversational contributions of the client from changing the score directly, e.g., by using a rule of accommodation. They also prevent certain conversational contributions that the therapist would like to make later in the therapy from becoming unacceptable. Let me briefly illustrate the different ways in which the therapist can affect the score of the therapeutic conversation using the example of systemic questions.

With the use of a systemic question the therapist usually wants to provide the client with an alternative, and possibly less problematic, description of reality (von Schlippe and Schweitzer 2016, pp. 249–51). For example, if a mother utters the sentence ‘My son is bad’, the therapist could respond with the systemic question ‘What is your son doing that you call “bad”?’. With this question the therapist (implicitly) provides the mother with a less problematic description of reality, according to which a behaviour of the son is perceived as bad, instead of the son having the property of being bad. If the mother responds with an utterance of ‘He throws things around’, then she has accepted the implicit offer of the therapist.

Systemic questions are often ascribed an almost hypnotic effect, since clients tend to tacitly accept their implicit offers. This is why Schmidt (1985) describes systemic questions as a kind of hypnotherapy. It is very likely that the almost hypnotic effect of systemic questions can be explained by the fact that their implicit offers are presuppositions that are accommodated by the client. In addition, the scorekeeping model provides an explanation of how, exactly, systemic questions open up alternative, and possibly less problematic, descriptions of reality on the part of the client. By changing the conversational score, the systemic question ‘What is your son doing that you call “bad”?’ makes new utterances possible in the context of the therapeutic conversation, i.e., utterances that presuppose that a behaviour of the son is perceived as bad (e.g., ‘He throws things around’). Finally, the use of the systemic question also blocks the mother's conversational contribution ‘My son is bad’, preventing the score from changing in such a way that it includes that the son has the property of being bad.²

Note that in the example of systemic questions the use of accommodation directly leads to the intended effect of the therapist's intervention, i.e., providing the client with a less problematic description of reality. Accordingly, the effectiveness of the intervention depends to a large extent on the use of accommodation. After all, an explicit utterance of the therapist's implicit offer (‘Your son isn't bad, but a behaviour of your son is perceived by you as being bad’) wouldn't have the desired effect. However, the causal connection between the use of covert exercitives and accommodation, on the one hand, and the intended effect of the respective therapeutic method or intervention, on the other hand, is not always that obvious. For example, if Popa-Wyatt and Wyatt (2018) are correct about the accommodation of discourse roles, then it is very likely that the effectiveness of transference and countertransference depends to a large extent on accommodation rules for the discourse roles of therapist and client. According to this explanation, in transference, the utterances of the client (and the therapist) result in the score adjusting in such a way that the therapist suddenly takes on the discourse role of someone from the client's past, making it possible to process unresolved conflicts. Similarly, the client could take on the discourse role of one of their past selves. In such cases the use of accommodation does not directly result in the intended effect of the examined method, i.e., to process unresolved conflicts, but it is nevertheless decisive for its effectiveness.

We see that, according to the scorekeeping model, therapeutic methods and interventions can affect different components of a conversational score. Hence, in order to get a better understanding of how individual methods and interventions work from a linguistic point of view, we have to investigate which components of the conversational score are affected by the respective method or intervention. What all the different methods and interventions of talking therapies have in common, according to the scorekeeping model, is the fact that they are covert exercitives, and, as such, they change the score of the therapeutic conversation, possibly using the linguistic phenomenon of accommodation. This has significant implications for the therapeutic practice, as it highlights the importance of training therapists in performing and blocking covert exercitives using rules of accommodation. The latter would also require that the therapists be trained to recognise the performance of potentially problematic exercitives on the part of the client. For example, Sbisà (1996) carried out an empirical study that shows that speakers can be trained to identify potentially problematic presuppositions of an utterance. Similarly, Langton (2018, p. 149) notes that speakers can be trained to block problematic presuppositions via explication, paraphrasing, a raised eyebrow, and the like.

4. Concluding Remarks

Investigating the linguistic aspects of the different methods and interventions of talking therapies starting from the scorekeeping model constitutes a research field in its own right. The primary aim of this paper was to provide a general understanding of how the methods and interventions of talking therapies work from a linguistic point of view. Accordingly, the effectiveness of the methods and interventions of the various forms of talking therapy depends in large part on the performance of conversational exercitives, in particular those using rules of accommodation. This was illustrated using the examples of systemic questions and transference/countertransference. Concluding, I will briefly address the question of how the use of covert exercitives and accommodation interacts with another important factor for the effectiveness of talking therapy, i.e., the therapeutic relationship.

The only way to prevent accommodation is by blocking. For example, if the therapist uses the systemic question ‘What is your son doing that you call “bad”?’, the mother could respond with an utterance of ‘What do you mean by “call ‘bad’”? He is bad!’. In this way, she would prevent the score of the therapeutic conversation from adjusting in such a way that it includes that a behaviour of the son is perceived as bad. Hence, if the effectiveness of the methods and interventions of talking therapies depends in large part on the performance of covert exercitives using rules of accommodation, the effectiveness of these methods and interventions can be expected to strongly correlate with the clients’ blocking behaviour. Therefore, it is important to investigate what influences the clients’ blocking behaviour.

If the mother blocks the presupposition of the therapist's utterance in the described way, it won't count as correct play. This, in turn, would amount to an interruption of the conversation. It is very likely that this is the main reason why listeners tend not to block the conversational moves of the speaker, because they usually have a strong interest in keeping the conversation going. The reasons for this can vary. However, it is plausible that listeners have a strong interest in keeping a conversation going if they want to build or maintain a relationship with the speaker. This suggests that the degree of blocking decreases if the importance of the relationship to the conversational parties increases. Together with the claim that the effectiveness of the methods and interventions of talking therapy depends in large part on the use of covert exercitives, in particular those using rules of accommodation, this promises a completely new explanation of why the therapeutic relationship has such a large influence on the effectiveness of talking therapies. By reducing the blockings on the part of the client, the therapeutic relationship increases the effectiveness of the therapeutic methods and interventions.

The influence of the relationship of the conversational parties on their blocking behaviour has to be tested empirically, which goes beyond the scope of this paper. A confirmed correlation between blocking and certain aspects of the relationship of the conversational parties would not only provide a better understanding of how the therapeutic relationship affects the effectiveness of the methods and interventions of talking therapies, but it would also contribute to the demystification of the linguistic phenomenon of accommodation, by providing an explanation of why speakers tend to adjust their attitudes and feelings so that the utterances made in the course of a conversation count as correct play. This illustrates again how research at the intersection of psychotherapy research and philosophy of language can be highly beneficial for both research areas.