There are only a few publications that describe how to use RFA. Nevertheless, they show that the use of RFA as a research methodology is not fixed or strictly manualized, and there are numerous ways to carry out RFA (Cotton, 2010; Chenail, 1995; Keeney, 1990; Keeney et al., 2015; Pelak, 2016; Peng, 2014). This flexibility is intentional, reflecting the radical concept of the “self-inclusion of the researcher” (Keeney et al., 2015). In this text we describe RFA in eight general overlapping steps, similar to those used in discursive psychology (Keeney et al., 2015; Lester et al., 2018; Wiggins, 2017): (1) Research objective, (2) Data collection, (3) Data processing, (4) Transcription, (5) Data analysis, (6) Presentation of outcomes and the visualization forms of the RFA, (7) Additional methods of analysis, and (8) Quality management of the research.

Research Objective

RFA is generally utilized to analyze the performative level of language use in conversations. Thus, any type of conversation can be subjected to RFA (Keeney et al., 2012). However, it cannot be used to directly study perceptions or any other presupposed “internal entities” that belong to the individual participants of a conversation, nor can it classify these “entities” (Chenail & Duffy, 2011; Keeney et al., 2015). Typically, research studies using RFA as their methodology frame their objectives in relatively broad terms, developing these with the help of additional research questions (Keeney et al., 2015; Lester et al., 2018). In psychotherapy, change process studies employing RFA, the research question is typically centered on the shifts in therapeutic discourse, demonstrating how the change occurred within the session. More specific research questions can be formulated, reformulated, or added based on the actual work with the data. This approach allows for increased research flexibility and the inductive focus typically associated with qualitative research strategies (Lester et al., 2018).

Data Collection

The collection of data is directly linked to the research objective. Generally speaking, RFA aims to work with naturally occurring data that would exist even if not collected for research purposes (Potter & Hepburn, 2005; Wiggins, 2017). A typical and entirely “naturalistic” source of data for RFA comes from recordings of therapeutic sessions (Chenail, 1995; Lester et al., 2018). Ideally, for RFA, video recordings are preferred as they allow researchers to capture both verbal and non-verbal aspects of the conversation. However, it is also possible to use audio recordings alone to gain valuable insight into the changes in therapeutic discourse (Chenail, 1995; Rudes et al., 1997).

Data Processing

This phase is closely related to the previous one. During the data processing phase, the data is stored, shared, organized, transcribed, and annotated (Lester et al., 2018; Miles & Huberman, 1994). The data processing phase needs to be conducted in a transparent, structured, legal, and ethical manner, which includes the protection of client confidentiality. The performance-oriented nature of the RFA requires the researchers to constantly oscillate between the recordings, the transcriptions, and the analytic procedures (Chenail, 1995; Keeney et al., 2015). Therefore, establishing the data processing system is a crucial step towards potential future success.

Transcription

As previously stated, the recordings need to be transcribed. Transcription is a crucial element of the analysis process (Chenail, 1995; Wiggins, 2017). It lays a “solid ground for the analysis” (Nikander, 2007), enables genuine empirical exploration (Keeney & Keeney, 2012b), and enhances the transparency of the study (Nikander, 2007). However, it is important to note that the transcripts cannot replace the primary database (O’Connell & Kowal, 1995). The analysis should always simultaneously incorporate the transcripts and recordings (Lester et al., 2018).

Transcription might be included in the previous phase (data processing), but due to its significance and time-consuming nature, it is often treated as a separate research phase (Lester et al., 2018). When transcribing the recordings, researchers can either follow one of the established transcription systems, such as the Jefferson system (Jefferson, 2004), or they can develop their own system, which then becomes a crucial tool in familiarizing themselves with the data (Lester et al., 2018). When deciding which transcription system to use, it is vital to consider the intended use of the transcript in the research study and to remember who will be the target audience of the resulting text (Mergenthaler & Stinson, 1992; O’Connell & Kowal, 1995). It is also essential to balance focusing on details and the number of transcription symbols and comments while considering the transcript’s feasibility, comprehensibility, and “intellectual elegance” (Mergenthaler & Stinson, 1992).

The RFA studies most commonly use simple transcriptions with only a minimal number of comments and symbols (Cotton, 2010; Chenail & Duffy, 2011; Keeney et al., 2015; Peng, 2014). Following the recommendations of O’Connell and Kowal (1995) and Keeney et al. (2015), and in light of our own experience, we recommend the subsequent system (Table 1).

Table 1 Transcription system for RFA

This straightforward transcription system allows us to generate rather intricate and intelligible transcriptions. All essential elements– barring intonation-related emphasis, which would make the transcription appear overly complex– are incorporated as straightforward comments in parentheses. Concurrently, these transcriptions serve as reference tools, supplementing the original recordings. Hence, all comments within them aim to function as indicators of some intriguing and significant occurrences within the conversation– they urge the researcher to cross-reference these substantial moments with the original recording. Owing to the consistent incorporation of time stamps, this form of transcription also facilitates quick navigation through the recording. Further transcription symbols can also be integrated into the transcription system, provided specific facets of the conversation need monitoring within the research.

Data Analysis

Chenail defines RFA as “a type of sequential analysis which helps researchers note semantic shifts in a conversation” (Chenail, 1995, s. 7). Throughout RFA, one can follow the content (what is said), the consequentiality (sequence of what is said and its impact), and the process (the flow from one topic to another) of the conversation, including how individual participants use language to create shifts in the discourse (Edwards & Potter, 1992; Chenail, 1995; Wiggins, 2017). Researchers can incorporate these observations directly into the transcripts using different types or colors of text to distinguish their interpretations from the raw transcripts (Chenail & Duffy, 2011; Keeney et al., 2015; Wiggins, 2017). They can use suitable software (e.g., ELAN) or various methods of visualizing the RFA, which will be mentioned in the following section of the article. Chenail states that there are four basic levels of analysis used in RFA: semantic, sequential, pragmatic, and consensualFootnote 2 (Chenail, 1995; Chenail & Duffy, 2011).

Semantic Analysis

The semantic level of analysis focuses on capturing the elements that carry meaning: frames and their hierarchical organization. Researchers of the RFA study the “anatomy” of a conversation, or its meaningful structure (Keeney, 1990; Rambo et al., 1993). They observe and capture detailed nuances - distinctions and their organization into frames, or higher levels of discourse, referred to by RFA as “galleries”, “wings”, and “museums” (Chenail & Duffy, 2011; Keeney, 1990). These somewhat unusual terms refer to a hierarchy of frames using the metaphor of a visit to a museum. Examining individual frames can be likened to viewing individual paintings arranged in galleries, which are part of larger units—the museum’s wings—and everything is part of the museum as a whole. The simplest way of doing semantic analysis is to capture conversation themes (they often form frames in conversation) and organize similar themes into higher levels of discourse. It is important not to handle frames as fixed units but to track the changes and re-entries of frames throughout the whole conversation.

The key principle of this analysis phase is to stay as close as possible to the original data, for example by consistently using the language of the observed conversation to name frames (Keeney, 1990; Keeney et al., 2015), avoiding generalized interpretations or revised terminology whenever possible. However, when naming galleries, or other higher discourse levels, as this phase involves a certain level of generalization and thematic grouping of related frames, the aforementioned strict rule is not applicable. This issue might have directed the authors of the RFA towards preferring visualizations of individual frames, without illustrating the galleries or other general meta-frames in their later works (Keeney et al., 2015). Conversely, creating galleries, wings, and museums in semantic analysis makes the “map” of the session easier to understand. We can also point out that even in their latest book, the authors of RFA use graphics of meta-frames, which include frames (Keeney et al., 2015, p. 187, 235, 261).

Sequential Analysis

The following sequential level of analysis focuses on the gradual development of interactional frames over time. In other words, it observes how these frames follow one another chronologically (Chenail & Duffy, 2011). As previously stated, the meaning of a frame is defined through its relationship with other frames– those that preceded it and those that will follow (Chenail, 1995; Keeney & Keeney, 2012a). Closely related to the previous one, this level of analysis examines the transitions between frames, noting how some frames function as “openers” or “closers” to others (Chenail, 1995; Keeney, 1990). The sequential analysis also highlights the recursive (circular) relationships between frames. This is evident, for instance, when some frames recur in new contexts and potentially influence and change the relationships between different frames or contexts within the broader context hierarchy (Keeney et al., 2015).

Pragmatic Analysis

The pragmatic level of analysis focuses on the realm of “social action” or the “function of language” (Edwards & Potter, 1992; Lester et al., 2018; Wiggins, 2017). It examines how participants in a therapeutic conversation contribute to changing frames. As we previously mentioned, RFA utilizes Field’s “three-act play structure”, which is suitable for scoring any creative activity, including therapeutic conversations (Keeney, 2009; Keeney et al., 2015). The three-act play structure, in the context of therapeutic conversations, allows us to recognize any shift from the original impoverished frame (Act I), through the emergence of new possibilities (Act II), towards the reinforcement and cementing of change into a resourceful experience (Act III). This level of analysis also identifies how changes occur between individual acts, which conversation participants contributed to the change, and the methods they used to do so.

Consensual Analysis

While the authors of the RFA method do not directly mention consensual analysis, it is optimal to implement the use of RFA in research teams (Chenail, 1995; Peng, 2014). The goal is to foster the development of a shared understanding of the studied conversations through mutual exchanges among various team members, referred to as “data sessions” (Hill, 2011; Wiggins, 2017). This is what is known as a consensual level of analysis which introduces new enriching insights and facilitates the exchange of various perspectives, leading to a newer and more complex understanding of the subject matter. This aligns with the epistemological principles of social constructivist theories that are significant to RFA, and significantly, enhances the “trustworthiness” or “validity” of the study (Elliott et al., 1999; Hill, 2011).

All the levels of analysis mentioned above, possibly except the consensual level, align with the primary focal points of sequential analysis in discursive psychology. These are represented by the queries “what,” “when,” and “how” (Wiggins, 2017). Throughout the RFA process, the researcher continually transitions among them.

Presentation of Outcomes and the Visualization Forms of the RFA

In the RFA outcome presentation, the researchers emphasize visualizing the structure of the therapeutic conversation. They pay particular attention to its performative aspects, the circularity in frame constructions and reconstructions, and the moments of creative development and change. Various visualization forms of the RFA correspond to these aims. It could be a simple “chain” of linked frames, aligning frames into meta-frames (referred to as “galleries”, “wings”, etc.), or positioning one frame inside another and exchanging their inside/outside places in the scheme as necessary (for examples, see Cotton, 2010; Chenail, & Duffy, 2011; Keeney, 1990; Keeney et al., 2015; Peng, 2014). A unified presentation system does not exist; therefore, researchers using the RFA are encouraged to create new ways of presenting their results, aligning with the focus of their analysis and research needs (Keeney et al., 2015).

At this juncture, we wish to unveil the form of the RFA presentation that we utilize. This presentation is Figure 1 especially adept at documenting the evolving dynamics of a therapeutic conversation, which is important for change process research. This visualization encapsulates the conversation discourse’s oscillation between individual acts within a “three-act structure” (Keeney et al., 2015).

Fig. 1

Three-act structure (Keeney et al., 2015)

As a case example for the RFA map, we use the session conducted by Insoo Kim Berg, the co-founder of Solution-Focused Brief Therapy (SFBT). In this session, the therapist works with Carl, a 17-year-old who attempted suicide the previous night. Figure 2 displays a detailed map of the first part of the session, which spans from its onset (time in video: 5:31) to utterance number 128 (time in video: 17:33).Footnote 3

Fig. 2

The first column on the left displays the utterance number from the transcript, and it can also include time in the recording for easier searching and enhanced tracking of the session’s duration. In a typical research report, the whole transcript would be accessible for the readers to examine how the researcher constructs individual frames and galleries. Full transcript is not included here due to space limitations, only a brief example is presented (see below). The individual frames and galleries are then depicted in the three columns marked with individual acts. In an RFA map, frames can be listed separately or grouped into more generalized galleries. This example includes both individual frames and their organization into galleries.

The RFA map illustrates how discourse fluctuated between acts over time. As demonstrated by the scheme, changes in discursive performance do not follow a simple, straightforward linear process. In contrast, there are numerous reversals and progressions throughout the session (refer to galleries 4 and 5; also immediately after utterance 128, where the excerpt concludes, the conversation reverted to Act II, taking roughly ten additional minutes to return back to Act III).

This scheme can be executed using a straightforward table function available in nearly every text editor. The use of arrows between frames, ideally color-coded, is also feasible. Frames are displayed in bounded cells, and absent table rows provide the opportunity to insert notes about the shifts between frames. This method of visualizing RFA has a drawback as it does not accurately represent the circular nature of the change process. We attempt to counteract this drawback by numbering the individual frames and noting the frame number in brackets when a frame re-enters into a new one. Particularly significant markers of changes in discourse are when formerly impoverished frames re-enter into resourceful ones (see 8.1., 9.1., 9.2.).

This RFA map enables a detailed examination of the evolution and unfolding of a conversation over time, within the context of a “three-act structure” (from Act I to Act III and vice versa). An analysis of discursive transitions between acts demonstrates that both - the therapist and client - initiated these shifts in both directions. However, the therapist initiated the shifts from Act I to Act II or from Act II to Act III significantly more frequently than the client. Deeper investigation shows that therapist primarily used questions or empathic utterances to facilitate these transitions. Conversely, the main initiator of shifts in the reverse direction, from Act III to Act II or from Act II to Act I, was typically (though not exclusively) the client when providing problem saturated narratives. Following example from the transcript of the session shows the way of distinguishing frames and transitions between them.

Example (de Shazer et al., 2007, pp. 113–115).

37)

CARL: (tears in his eyes) Until I held back my head and said my last prayers or whatever. And then I just started cutting; and you can see right here (shows a neck to therapist).

38)INSOO: Yes. Yeah. Yes, I can see that, yes.

39)CARL:.. and I’m just lucky to be alive today. You know, to speak about this because had I not known there was a sharper knife next to it, but I went for that knife ’cause it was the biggest.

40)INSOO: Yeah.

41)CARL: So, I thought I would have done more damage or something.

42)INSOO: Right.

43)CARL: So I just, I don’t know; I thank God every day, you know. (pause)

44)INSOO: So, what helped? What helped so it didn’t go through? (points to her neck)

45)CARL: It’s, I mean, I don’t know. I mean, when I was—the knife wasn’t—it wasn’t that sharp, it was like a butter knife or something.

46)INSOO: Yeah. (looks surprised)

47)CARL: It was just, you know, and so I made it to where it could cut harder, so I used force actually.

48)INSOO: Yeah.

49)CARL: So. (pause)

50)INSOO: And you’re saying this is not the first time. It has happened other times as well.

51)CARL: It’s happened other times, but this is the worst.

52)INSOO: This is the worst it’s been.

53)CARL: Yeah. Other times it’s been like, burn myself with an iron or something, like right here.

54)INSOO: Yeah. Wow.

55)CARL:.. or something like that.

56)INSOO: Yeah.

57)CARL: I mean, I would get all kind of thoughts in my head...

58)INSOO: Yeah.

59)CARL:.. that I, you know, I just pray to God that I wouldn’t think of them anymore. They just, you know, they just come around whenever, you know, I’m feeling depressed or something, you know, suicidal, you know.

60)INSOO: Okay. So, how did you manage to get to school today?

61)CARL: Well, that’s a long story, too.

62)INSOO: What a night.. I mean it was a terrible night you had.

63)CARL: Way terrible night.

64)INSOO: How did you show up in school on time?

65)CARL: Well, it’s kind of great actually, ’cause while this whole conflict happened, I was at my aunt’s house...

66)INSOO: Yeah.

67)CARL:.. and they wanted me to stay the night. They said, you need to rest or whatever.

68)INSOO: Yes.

69)CARL: I said no, I want to go back home so I can get up and go to school in the morning.

70)INSOO: Oh my goodness.

71)CARL: ’Cause I haven’t missed a day in school yet. At least—well, two days, but those were counted you know. as, you know, I was here.

72)INSOO: Yeah.

73)CARL: As I would go to the doctor or something.

74)INSOO: Yeah.

75)CARL: But I’ve been on a perfect attendance, you know, roll twice already.

76)INSOO: But some terrible night like last night, you still showed up in school.

77)CARL: Yeah.

78)INSOO: And you were helpful to me; you’re trying to be helpful to me today.. this morning. And you even participated in the group with a large group of people.

79)CARL: Large. (smiles)

80)INSOO: A very large group of people. You are very outgoing, very bright, very articulate. I was just absolutely astounded when I saw.. what I heard this morning and now, ’cause I didn’t get a chance to talk to you because we were so busy. And you were just able to sort of set that aside.

81)CARL: Right.

The excerpt begins in frame 3.7, where the client describes the horrific story of last night’s suicide attempt. The therapist (Insoo) offers to move to Act II with a question focused on what helped Carl not to hurt himself (utterance 44). This invitation opens up a new gallery (4) that could potentially open a pathway to the c