A second opinion

One of the authors recently saw a young man who came in to get a second opinion. In the session, he said he had been “anxious my whole life” and used to think it was “a personality trait” that he would have to live with. Recently, he had met with a doctor who diagnosed him as having “Generalized Anxiety Disorder” and told him there were medications (SSRIs) that would “cure” him and that “always work” and “had no side effects.” The client went on to say that he had doubts about SSRIs because he knew friends taking them who were not so pleased with their therapeutic effectiveness or their side effects; hence his decision to seek a second opinion.

He answered the miracle question with many details about how he would feel, think, act, and interact with others if a miracle that solved the problem had happened while he slept. When asked about instances when parts of the miracle were happening, he identified small parts currently happening and major parts that were happening during entire months two years ago. During the break the therapist prepared the following feedback to him:

He nodded gravely and said “Yeah. It would have been easier to be dim and happy.” The therapist responded with “yeah,” and the two of them burst into laughter.

Solution-focused practitioners will readily recognize the significance of what occurred in this conversation. Like the therapist, when they hear the client’s language shift from “anxious my whole life” and “generalized anxiety disorder” to “it would have been easier to be dim and happy” (while laughing), they know something potentially important and more hopeful is happening:

Structural and post-structural views of language

We have Steve de Shazer in particular to thank for alerting us to the therapeutic significance of clients’ shifting their language about themselves and their situations. By the mid-1980’s, he and his colleagues at the Brief Family Therapy Center had invented techniques (e.g., exception questions, the miracle question, scaling) to construct solutions with clients rather than to solve their problems. At that time, he began to observe that the old ways of talking about therapy no longer worked and, as he stated later, it became necessary to find new ways to describe and analyse what clients and practitioners do in the therapy room (de Shazer, 1991, pp. xiii–ix). He stressed that therapy is accomplished through language interaction, an obvious point that he claimed the field of psychotherapy had largely ignored, and he began drawing on the ideas of several philosophers and scholars of language including Bakhtin (1981), Berger and Luckman (1966), Derrida (1978, 1981), Lyotard (1984) and especially Wittgenstein (1958) to create a more useful description of what is happening in therapy. He built this new description around the distinction between a structural and post-structural view of language (de Shazer & Berg, 1992; de Shazer, 1991, 1994).

The structural view of language (Chomsky, 1968; Saussure, 1959) is that the words used in a conversation (called the surface structure) are representations of underlying and true meanings (deep structure) which are assumed to be discoverable for any word (de Shazer & Berg, 1992; Harland, 1987). In this way of thinking, clients’ words have essential, knowable meanings which therapists can uncover through their expert assessments and evaluations. For example, in a structural view, when a client comes in and says, “I’m depressed,” the therapist should do a professional assessment, asking questions to uncover the existence and degree of a particular clinical condition named “depression”.

In contrast, post-structuralism identifies the meanings of words within particular interactional contexts. So, rather than being seen as stable and beneath the surface, the meanings of words occur at the surface level of conversation and are knowable “through social interaction and negotiation” (de Shazer, 1991, p. 45). In this view, the meanings and descriptions that the client in our introduction attributed to his experiences are seen as shifting from “anxious all the time” and perhaps having “generalized anxiety disorder” to “it would be easier to be dim and happy.” These meanings may shift even further through additional therapeutic dialogue as well as through whatever he does with his new understandings of himself when he leaves the therapy room.

de Shazer called his post-structural view of how words work in therapy interactional constructivism (1991, p. 48). He suggested that “we need to look at how we have ordered the world in our language and how our language … has ordered our world” (1994, p. 9). The implication that we can re-order our world with language was illustrated by de Shazer and Berg (1992; de Shazer, 1991) with a case in which the therapist (Berg) and the couple negotiated the meaning of the wife’s condition and the couple’s problem from an initial description as “nymphomania” (the wife’s word) to “more of a sleep problem for both of us” (the husband’s words, which the wife accepted). The shift in meaning seems to have been useful to the couple because, two weeks later, the woman sent a note to the therapist saying that her “sleep patterns and libido” had returned to normal and more therapy was not needed (de Shazer, 1991, p. 67). It was this case, de Shazer said, that persuaded him and his colleagues that they must develop new ways to describe and analyse what is happening in therapy.

Social constructionism

de Shazer’s interactional constructivism is very similar in meaning to the term used more broadly in the fields of psychology and sociology, namely social constructionism. In particular, Kenneth Gergen (1985, 1999, 2009) has written extensively about social construction in the field of psychology. Gergen uses the term to refer to the proposal that people, through their social and language interactions, continually create and rework the meanings that influence their lives. He views the collaborative language systems approach (Anderson & Goolishian, 1992; Anderson, 1997; Hoffman, 1990), narrative therapy (White & Epston, 1990), the reflecting team approach (Andersen, 1991), solution-focused brief therapy, and solution-oriented therapy (O’Hanlon & Weiner Davis, 1989) as instances of social constructionism where new and hopeful possibilities are co-constructed between therapist and client in therapeutic dialogue.

According to Gergen (2009), the practices of these social-constructionist therapists are different from others in the field in two respects. First, they show no interest in categorizing personal or interpersonal problems of clients nor in figuring out the causes of problems. None of these are seen as useful ways to promote therapeutic change. A second difference is in the stance of the therapist relative to the client. The therapist is not a separate, neutral assessor of a client’s objectively discernible problem(s). Instead, the therapist is (in Anderson & Goolishian’s, 1992, term) deliberately a “not-knowing,” collaborative partner who continually seeks to be informed by the client’s language and expertise about his or her own situation and who invites the client to participate in a dialogue that co-constructs new meanings that will create the more satisfying and productive life that the client is seeking. A central concept in social constructionism and in these constructionist therapies, then, is the process of co-constructing new meanings in the therapeutic dialogue.

It is noteworthy that, although the process of co-construction is central to social constructionism, it has remained a broad theoretical concept, not specifically linked to dialogue. As we have observed elsewhere (De Jong, Bavelas, & Korman, 2013, p. 19), the presumed outcomes of co-construction are as abstract as the concept of social construction itself. For example, various psychotherapy theories have proposed that co-construction leads to new subjective meanings, understandings, realities of everyday life, knowledge, narrative realities that reflect power relations, the self, and many other broad categories of meaning (Anderson & Goolishian, 1992; Berger & Luckmann, 1966; Gergen, 2009; Hoffman, 1990; White & Epston, 1990). However, these are reified end products without descriptions of the process. In short, the description of co-construction in the literature has remained abstract. To advance this view, we need to study co-constructing as an activity rather than abstracting it as co-construction.

It is the purpose of this paper to make co-constructing concrete and observable, that is, to develop an empirically based understanding of how co-constructing (as a verb) happens in therapeutic dialogues. We do this by first turning to a section of the psycholinguistics literature where research has revealed how participants in face-to-face dialogue collaborate with one another to create shared meanings — just as de Shazer (1991, 1994) and Gergen (1985, 1999, 2009) proposed. After summarizing this research, we propose a micro-model of an observable process by which participants in face-to-face dialogue collaborate to co-construct shared and new meanings. We conclude by addressing the implications of our model for the field of psychotherapy and identify useful directions for empirical research to further document the model.

Experimental evidence for co-constructing in dialogues

The same paradigm shift in two different fields

As outlined above, the traditional and dominant view for how language works in therapy is the structural view in which language transfers meanings from the mind of one person to the mind of another. We have joined with others including Berg, de Shazer, and Gergen who propose an alternative and sharply contrasting view, namely, that people in dialogue, including therapists and clients, co-construct meanings together. This position implies that, in order to understand how therapy works, the focus needs to be on the interactive process of co-constructing.

In experimental psycholinguistics, Clark and his colleagues (e.g., Clark, 1992, 1996) also proposed an alternative view of dialogue. They called the traditional and dominant view an autonomous view, in which speakers choose language that best conveys their meaning and send it to a receiver whose role is simply to comprehend this meaning correctly. In their alternative collaborative theory of dialogue, Clark and his colleagues proposed that the participants in a dialogue collaborate, moment by moment, to create shared meanings. In the collaborative view, meaning is created, modified, and sustained by their mutual actions — a view that is remarkably similar to theories of co-construction (Bavelas, 2011). The next section summarizes some of the key evidence from experimental psycholinguistics showing that a collaborative theory can better predict what happens in dialogue than an autonomous theory can.

Tests of collaborative theory

The autonomous view of language use seems like common sense and therefore has intuitive appeal. It has led to a great deal of research on individuals as speakers and listeners and virtually none on their interaction. In the autonomous view, the interaction is irrelevant because a listener in dialogue is just like any other listener, such as an audience or an overhearer. This line of research focuses on the cognitive processes of a listener who is treated as “mute or invisible” in the interaction (Clark & Wilkes-Gibbs, 1986, p. 3), with no influence on the speaker.

In contrast, Schober and Clark (1989) pointed out that the listeners in a dialogue are fundamentally different from other kinds of listeners outside of dialogue. The listener in a dialogue is an addressee, that is, the unique individual whom the speaker is addressing and for whom the speaker is shaping what he or she says. The addressee has the right — and the responsibility — to indicate understanding and to assist when necessary. In doing so, the addressee has considerable influence both on what the speaker says and how it is said: “Speakers and their addressees go beyond … autonomous actions and collaborate with each other moment by moment to try to ensure that what is said is also understood” (Schober & Clark, 1989, p. 211). The following experiments have focused on the influence of the addressee on the dialogue.

Clark and Wilkes-Gibbs (1986)

In this experiment, the speaker and addressee, who were separated by a partition, had the same set of 12 cards. Each card showed an abstract geometric figure (“Tangram figure”). As can be seen in Figure 1, these figures have no standard names and therefore can be described in a wide variety of ways. The speaker’s cards were laid out in a certain order, and the addressee’s cards were in a random order. The task was for the speaker to tell the addressee the correct order in which to place them. They did this six times, in a different order each time.

The autonomous and collaborative models differ sharply in their predictions on how the pairs would accomplish their task. Recall that, in the autonomous model, speakers choose language that best conveys their meaning, and the receiver’s role is to comprehend this meaning correctly. So the speaker would be in charge and would choose the best way to describe each figure, unilaterally providing a term or phrase that the two of them could continue to use over the six trials whenever they referred to that particular figure. However, it turned out that speakers did not determine the names the pairs used for these figures. Instead, over the course of the six trials with the same figures in different orders, the speaker and listener collaboratively developed ways of referring to each figure. Often, it was the addressee who initiated a reference that they both adopted, as in this example (adapted from Schober & Clark, 1989, pp. 216–217):

Example 1. Trial 1, describing #12 in Figure 1

• 1. Speaker: “Then number 12, is (laughs) looks like a, a dancer or something really weird. Um, and, has a square head, and um, there’s like, there’s uh — the kinda this um,”
• 2. Addressee: “Which way is the head tilted?”
• 3. Speaker: “The head is, eh — towards the left, and then th — an arm could be like up towards the right?”
• 4. Addressee: “Mm-hm.”
• 5. Speaker: “And, it’s —”
• 6. Addressee: [overlapping] “an —, a big fat leg? You know that one?”
• 7. Speaker: [overlapping] “Yeah, a big fat leg.”
• 8. Addressee: “and a little leg.”
• 9. Speaker: “Right.”
• 10. Addressee: “Okay.”
• 11. Speaker: “Okay?”
• 12. Addressee: “Yeah.”

They went on to refer to this figure in several subsequent trials, and by the last trial, they had co-created a reference to this figure that was brief and required only one turn each; it incorporated features that each of them had suggested:

Example 1a. Trial 6, the same pair are describing the same figure

• 1. Speaker: “Um, 12, the dancer with the big fat leg?”
• 2. Addressee: “Okay.”

Clark and Wilkes-Gibbs found that the average number of turns and average length of each turn decreased significantly over the six trials using the same figures. This effect is consistent with the pairs having collaborated to establish a shared vocabulary that permitted them to refer unambiguously to each figure. However, although we can see their collaboration in examples such as above, the numbers themselves do not confirm that there was a collaborative process; it could have been that speakers simply got better at providing more succinct information as they went along. The next experiment addressed that possibility.

Schober and Clark (1989)

This study used the same task with an experimental design that tested collaborative theory more directly. They created two conditions that were identical in the information that the speaker provided but that differed in whether collaboration was possible. For each speaker-addressee pair, there was a third person, also behind a partition, who could overhear everything the speaker and addressee said. The difference in this study was that this third person could not interact with the speaker and could not speak up at all. These instructions created two kinds of listeners to the same speaker, an addressee who was free to engage the speaker in dialogue and an overhearer who could not. The overhearer could not clarify his or her understanding, suggest terms, ask questions, or even indicate when the speaker could go on to the next figure. Thus, in each triad, the overhearer had all of the same information as the addressee but did not have the benefit of interacting with the speaker. The autonomous model predicts that only the quality of the speaker’s information would matter. However, it turned out that the ability to collaborate also mattered: The results showed that the addressees did significantly better at getting the figures in the right order than the overhearers did with the same information.

Isaacs and Clark (1987)

This experiment also showed that the addressee had an important influence even when, unlike the Tangram figures, there was a correct name for the pictures the speaker was describing. That is, they did not have to come up with a new name. As in the two experiments just described, the speaker had to tell the addressee the correct order of a set of pictures, but these were postcards of well-known landmarks in New York City (e.g., Rockefeller Center, the Central Park Lake). There are, of course, proper names for these landmarks, but Isaacs and Clark (1987) arranged that not everyone knew them, as follows: They pre-tested potential participants for their knowledge of New York City and identified “experts,” who had lived there and knew the city well versus “novices” who had not been to New York City and did not know the proper names of the landmarks. Then Isaacs and Clark created four possible speaker-addressee pairings: expert to expert, novice to novice, expert to novice, and novice to expert. Not surprisingly, the pairs in which both the speaker and addressee were experts started with the proper names and continued to use them. The pairs with two novices were like the pairs describing Tangrams; they worked out a way to describe a salient feature in each postcard (e.g., “the tall building with the triangular top”) and used their agreed-upon description.

One might suppose that, in the mixed pairs, an expert talking to a novice would introduce the correct terms, and the pair would use those thereafter, but this is not what happened. The results showed that, surprisingly, the speakers’ expert knowledge of the correct term did not determine how the pair described the pictures. For example, expert speakers quickly learned that their novice addressees did not recognize the proper names, and their use of these names declined significantly over the trials as they changed to creating collaborative descriptions.

Example 2. Speaker knows New York City, but addressee does not

• 1. Speaker/expert: “Tenth is the Cidicorp — Citicorp Building?”
• 2. Addressee/novice: “Is that with the slanted top?”
• 3. Speaker/expert: “Yes.”
• 4. Addressee/novice: “Okay.” (adapted from Isaacs & Clark, 1987, p. 28)

However, when the speakers were novices, they also significantly increased their use of proper names — because they were learning them from their expert addressee, who often supplied them as an afterthought; for example,

Example 3. Speaker does not know New York City, but addressee does

• 1. Speaker/novice: “Fourteen is the fountain with the arch in the background.”
• 2. Addressee/expert: “Right, Washington Square, good.” (p. 33)

In both the expert-to-novice and novice-to-expert pairs, the speaker, who was the one who knew the correct order, adapted to the level of expertise of his or her addressee. Experts talking to novices used descriptions that were not “correct” but worked, and novices talking to experts learned some of the proper names. It is tempting to apply these results to the way expertise operates in different therapeutic practices: Does a client learn to talk about his or her life in language that works for the therapist or does the therapist learn to talk about the client’s life in the client’s language? Or a bit of both?

Bavelas, Coates and Johnson (2000)

This experiment showed the importance of a responsive, interacting addressee in a different, more typical kind of dialogue. Each narrator told a personal close-call story (where something bad could have happened, but in the end everything turned out all right) to an addressee in a face-to-face dialogue. They were strangers to each other, so the addressee could not have known the story. In the experimental condition, the addressee had to focus on an irrelevant aspect of the speaker’s narrative (e.g., counting the number of words the narrator used that started with the letter t). These addressees made significantly fewer listener responses (e.g., nodding, wincing, commenting on the story), and — most important — their narrators told their stories significantly less well than did the narrators whose addressees were listening normally. For example, the narrators with distracted addressees tended to be more dysfluent, and more likely to end abruptly or to over-explain. Thus, even though none of the addressees could contribute to the content of story, the unresponsive “t-counter” still made a difference to the quality of their narrator’s story-telling.

In summary, each of these experiments supported a collaborative rather than an autonomous view of language. In particular, it was not the case that the speaker imposed the language the pair would use. Instead, the addressee played an essential role in helping to shape language that would work for both of them (Clark & Wilkes-Gibbs, 1986; Isaacs & Clark, 1987). Listeners who could not collaborate did worse (Schober & Clark, 1989), as did speakers with unresponsive addressees (Bavelas et al., 2000). If, as we propose, collaborating equals co-constructing, these results also support a co-constructionist, post-structural view of dialogue.

Grounding sequences as the process of collaborating

The next question is exactly how collaboration works: What are the participants in a dialogue doing when they are co-constructing meanings in their interaction? An interactive view of meaning has old roots, including the writings of George Herbert Mead (1934):

Markova and Linell (e.g., Linell, 2001; Linell & Markova, 1993; Markova, 1990) revived and emphasized Mead’s triad, namely, a minimum unit in which one person initiates, the other person responds, and the first person completes the triad by responding to this response. Like Mead, though, they did not test this proposal in a body of data.

Based on their intensive analysis of a large collection of dialogues, Clark and Schaefer (1987, 1989) also proposed that meaning in dialogue is established collaboratively, through a process they called grounding, where the speaker and addressee work together, moment by moment, to establish that they understand each other well enough to proceed. To ground something is to lay a foundation for it or to set it on a firm basis (OED Online, June 2014). In their ongoing dialogue, speaker and addressee are continuously ensuring a firm basis of mutual understanding.

Clark and his colleagues’ theory of grounding (Clark & Schaefer, 1987, 1989; Clark & Brennan, 1991; Clark, 1996) emphasized two broad phases: the speaker presents something, and the addressee must accept it. However, they also mentioned a three-step sequence at the micro-level, similar to that of Mead and of Linell and Markova:

• a. The speaker says something to the addressee.
• b. The addressee shows the speaker that he has understood.
• c. The speaker confirms that the addressee has understood her correctly. (adapted from Clark & Schaefer, 1987, p. 22)

Clark and Schaefer pointed out that the addressee’s response in the second step is unilateral; it provides feedback to the speaker. For mutual understanding, the speaker must also provide feedback to the addressee, confirming that the addressee has understood correctly. That is, both the addressee and the speaker require evidence that they are understanding each other, which results in familiar sequences such as this one:

Example 4. A British telephone operator giving a caller the number of a restaurant

• a. Operator: “It’s Cambridge 12345.”
• b. Caller: “12345.”
• c. Operator: “That’s right.” (adapted from Clark & Schaefer, 1987, p. 25)

In this example, the operator presented new information, the caller displayed understanding by repeating part of the information, and the operator confirmed that the caller had understood correctly. However, Clark and Schaefer (1987, 1989; Clark, 1996) did not develop this three step process further; although the three steps are apparent in most of their examples, their analysis remained at a two-step level.

Systematic empirical investigation of the three-step model is in its earliest phase. Bavelas, Gerwing, Allison, and Sutton (2011) tested a three-step model on almost 600 presentations in 22 dyads and found that a three-step sequence fit these data virtually perfectly. (See also Roberts & Bavelas, 1996). In addition, based on intensive observation of face-to-face dialogues, we (Bavelas, De Jong, Korman, & Smock Jordan, 2012) have proposed and begun to empirically document a micro-model of three-step grounding sequences:

• a. The speaker presents new information.
• b. The addressee displays that he or she has understood the information (or has not understood or is not certain).
• c. The speaker confirms that the addressee has understood (or not).

When this sequence is completed, then speaker and addressee have grounded on a particular bit of information, that is, they have overtly demonstrated that they have understood each other so far. We have proposed that these sequences are continuous throughout the dialogue and are the building blocks of co-constructing and meaning-making (De Jong et al., 2013).

Empirical documentation of three-step grounding sequences

We will present the following analyses of grounding sequences in table form. (Note that it is often easier to read the transcript column first, then go back to follow the grounding sequences.)

Example 4a. The caller had asked the operator for the phone number of a particular restaurant (adapted from Clark & Schaefer, 1987, p. 25)

• #1 Operator: “It’s Cambridge 12345.”
• #2 Caller: “12345.”
• #3 Operator: “That’s right.”

The next example, from an unpublished therapy video, is equally simple:

Example 5. At the beginning of a college counselling session, the therapist had asked what year the client was in

• #1 Client: “I’m a junior.”
• #2 Therapist: “You’re a junior here.”
• #3 Client: [quick nod]

There are several variations on the basic pattern which still preserve the three-step sequence. In Example 2, above, the addressee’s display also presented new information, which initiated a second, overlapping sequence in which some utterances had two functions, as shown in the following table.

Example 2a. An expert on New York City is talking to a novice and describing the 10th postcard in the series (adapted from Isaacs & Clark, 1987, p. 28)

• #1 Expert: “Tenth is the Cidicorp — Citicorp Building?”
• #2 Novice: “Is that with the slanted top?”
• #3 Expert: “Yes.”
• #4 Novice: “Okay.”

Note that in the Bavelas et al. (2012) model, either person can signal a lack of mutual understanding in the second or third step, so grounding sequences also detect and correct errors. That is, the addressee can show that he or she did not understand, and the speaker can indicate that the addressee got it wrong. Example 6 illustrates a more complex pattern in which the participants used grounding sequences to sort out their problem.

Example 6. The telephone operator had requested the name of the person the caller was asking about but did not initially understand what the caller said (adapted from Clark & Schaefer, 1987, p. 20)

• #1 Caller: “Mrs. Lane.”
• #2 Operator: “Sorry, would you say that again please?”
• #3 Caller: “Lane.”
• #4 Operator: [spelling] “M-A-I?”
• #5 Caller: [spelling] “L-A-N-E.”
• #6 Operator: “N for Nellie, A-N-E.”
• #7 Caller: “No, L for London.”
• #8 Operator: “Oh! sorry, Lane, L for Leonard.”
• #9 Caller: “Yes.”

It may seem narrowly technical to put so much emphasis on the three-step micro-sequence, but it has some radical implications. First, it overtly challenges the widely held individualistic and mentalistic views of dialogue in which the basic unit of dialogue is each individual’s speaking turn. In this view, each individual speaker extracts a package of meaning from his or her mind and delivers it (i.e., a one-step, non-interactive view). There is also a two-step view, in which the addressee passively indicates understanding. Our three-step sequence insists that they have not grounded until the speaker gives the addressee confirmation of his or her correct understanding. Thus, the minimum unit of analysis for dialogue is a three-step grounding sequence, one in which the contributions of the addressee are as important as those of the speaker.

The second implication of our model is that dialogue is a micro-process, proceeding in small sequences of actions that are often unnoticed but not trivial. In the above examples, grounding was not an occasional pause to summarize every few minutes. It was constant, and every contribution counted, including repetitions, nodding or saying “Okay”. Second by second, the participants continuously displayed and confirmed their understanding at each step, accumulating a foundation of agreed-upon knowledge.

A third crucial implication of our model is that the participants do not necessarily ground on what the speaker presented. Instead, their grounding sequence may lead the speaker and addressee to accept a modified version. In Example 1, at #6, the addressee interrupted and displayed understanding by providing an entirely new description (“a big fat leg”), which the speaker accepted, and then proceeded to ground on and use later (Example 2a). Similarly, in Isaacs & Clark’s (1987) mixed pairs, the addressee was contributing to the version they grounded on.

Implications for psychotherapy

Although language has always been right out there in the open for observation, as de Shazer and Berg claimed, the field of psychotherapy largely ignores how language works in the interaction between clients and therapist. We share de Shazer and Berg’s curiosity about language and therapy. Specifically, we propose to identify language interactions between clients and therapist that are directly observable. At any given moment, the client presents his or her view of something; the therapist can respond (e.g., by paraphrasing it); and the client can accept, correct, or reject the therapist’s version. At another moment, the therapist presents his or her version of something else, which the client may modify, and the therapist may accept, correct, or reject. These grounding sequences are micro-negotiations that build the shared meanings we call co-constructions.

While writing primarily about the co-constructive nature of post-modern models of therapy, both de Shazer (1991, 1994) and Gergen (1985, 1999, 2009) theorized that, regardless of model employed by the therapist, all therapy conversations are co-constructive. At the empirical level, our analysis of grounding sequences supports this assertion. Below, we present our analysis of the grounding sequences in two contrasting therapy dialogues; one is solution-focused and the other is from a motivational interviewing session.

Example 7. De Shazer asked the client “What brings you in?” (from an unpublished video; also in de Shazer, 1994, p. 247. This excerpt of the subsequent dialogue was 17.3 seconds.)

• #1 Client: “Well, right now I’m dealing with a drinking problem.”
• #2 de Shazer: “Mm-huh”
• #3 Client: “Yeah” (very softly).
• #4 (de Shazer paused while looking down and writing, then says: “OK, and, uh” and pauses again.)
• #5 Client: “Sometimes I drink —”
• #6 de Shazer: “You say ‘right now’” (with emphasis).
• #7 Client: “Well, I’ve been dealing with it —”
• #8 de Shazer [overlapping]: “Mm-hm.”
• #9 Client (continuing): “— but right now I’m just feeling that it’s the time of my life to really get into it, do something about it.”

At #1, the client presented two related pieces of information: “Right now I’m dealing with” and “a drinking problem.” De Shazer could have commented on, repeated, or paraphrased either one. In his explicit display of understanding at #6, he chose the part of #1 in which the client stated that he was dealing with his drinking problem “right now,” which might represent the beginning of a solution. At #7, the client confirmed de Shazer’s display of understanding by starting to build on the theme of dealing with it. Then at #9, the client confirmed de Shazer’s display of understanding by incorporating “Right now” and adding a stronger statement of commitment, namely, that “it’s the time of my life to really get into it, do something about it,” and they grounded on this as well. After #9, the client continued to contribute information on this theme. Another therapist might have chosen to display understanding of #1 with “You said ‘drinking problem,’” which could have led to further details of that theme instead.

The next example shows a different choice by a therapist.

Example 8. In a Motivational Interviewing video, Miller (Lewis & Carlson, 2000) was asking whether the client had an addiction that was becoming a problem. (The following dialogue was 12.7 seconds.)

• #1 Client: “Smoking’s become a problem because I’m starting to play soccer, OK?”
• #2 Miller: “So you can’t breathe.”
• #3 Client: “And so I can’t breathe.”
• #4 Miller: “Yeah.”
• #5 Client: “The kids are younger and younger, OK?”
• #6 Miller: “Yeah.”
• #7 Client: “They’re half my age, twice my size.”
• #8 Miller: [nods]
• #9 Client: “I’m having a problem with it.”

As in Example 7, the client’s initial statement presented two different pieces of information: “Smoking’s become a problem” and “I’m starting to play soccer.” The therapist’s display (“So you can’t breathe”) referred to smoking as a problem that interferes with the client’s soccer performance, and their subsequent dialogue continued to pile on problems. A different therapist could have chosen to display understanding of the other part of what the client said, for example, with “Oh! You’re starting to play soccer?” which could have led to a discussion of a healthy choice.

In addition to documenting how therapies are co-constructive, analysing therapy dialogues with our three-step micro-model of grounding sequences has more specific implications for what is happening between therapists and clients in psychotherapy. One of these is that therapists are more influential than is often assumed. They are continuously contributing to the direction of co-construction by the paraphrases, elaborations, and questions with which they choose to display understanding. For example, in Example 7, at #6, de Shazer chose to focus his display of understanding on what the client was doing in the present rather than elaborating on the problem. Similarly, in Example 8, at #2, Miller chose to display his understanding by giving more details about a problem instead of commenting on a possible healthy choice. Obviously, these choices were consistent with their respective theories.

A second implication for what is happening in psychotherapy is that clients most often cooperate with the therapist’s contributions, specifically by confirming the therapist’s display of understanding. The therapists in Examples 7 and 8 provided displays of understanding that took the dialogue in different directions, and in both cases the clients followed the therapist’s lead. De Shazer’s client confirmed that “right now” was what his initial presentation had meant by beginning to present more detail about his current motivation. Miller’s client confirmed that “smoking’s become a problem” was what his initial presentation had meant by presenting more details about how smoking was creating a problem for him.

Conclusion

The purpose of this paper has been to make co-constructing in therapy concrete and observable. We found a rich resource for our efforts in the psycholinguistics literature that contains persuasive experimental evidence for a collaborative rather than autonomous view of how meaning arises in dialogue. Taking our lead from social scientists and psycholinguists who have suggested that collaboration in dialogue occurs in interactive sequences, we have proposed a three-step, micro-model of grounding sequences as the empirically observable process through which co-constructing meanings occurs in therapeutic dialogues. So far, the dialogues we have analysed have consistently supported this empirical model which, in turn, lends support to de Shazer’s and Gergen’s theoretical ideas about what is happening in psychotherapy interactions between therapist and clients.

We see the conceptualization and initial testing of our micro-model of grounding sequences as the beginning of an important area of scientific study of therapy interactions. While we have presented some initial findings about grounding sequences here, in a future article we plan to present more findings as well as details about the observational rules for microanalysing grounding sequences in psychotherapy dialogues. Beyond our research, there is room for others to take these rules for microanalyzing grounding sequences and apply them in other investigations, such as microanalysing grounding sequences in couples and family work where there are three or more people simultaneously participating in the dialogue. We believe this is a potentially fruitful line of research for all of psychotherapy and one which clearly respects de Shazer’s belief that we will learn more about how psychotherapy works by focusing on what is happening in the interaction of client and therapist rather than on what might be going on in the minds and emotions of clients.