James Gavin ✉ (Department of Applied Human Sciences, Concordia University, Montreal, Canada) Nicolò Francesco Bernardi (Department of Applied Human Sciences, Concordia University, Montreal, Canada) Elizabeth Thomas (Department of Applied Human Sciences, Concordia University, Montreal, Canada) Julie Anne Chacra (Department of Applied Human Sciences, Concordia University, Montreal, Canada)
Professional coaching associations advocate for appropriate patterns of verbal interventions by coaches. This investigation portrays what coaches say based on analyses of 48 sets of 3 consecutive coaching sessions obtained from professional coaches. 37 ways in which coaches verbally intervened with clients were sub-grouped into four categories with average frequencies estimated as follows: 20% - active listening; 33% - inquiring; 33% - influencing; 14% - miscellany. Analysis across three sessions argues for consistency in coaching interventions rather than for varied patterns. For the most part, intervention profiles align with guidance from professional coaching associations for effective coaching practice.
process research, verbal behaviour, coach behaviour, mixed methods, coaching interventions
Accepted for publication: 03 January 2025 Published online: 03 February 2025
© the Author(s) Published by Oxford Brookes University
The field of coaching seems challenged in offering a coherent and consistent description of what constitutes a coaching experience, despite the efforts of professional coaching associations to prescribe exactly what it is and what it is not (International Coaching Federation, 2024a, 2024b, 2024c). Novel frameworks for coaching practice emerge with some regularity (Stelter, 2013, 2014, 2018). Such diversity was signalled in an early review of the nascent coaching field (Brock, 2008, 2014) and thoroughly detailed in more recent typologies of the traditions and genres of coaching practice (Bachkirova, Cox, & Clutterbuck, 2018; Cox, Bachkirova, & Clutterbuck, 2014). Ambiguity about what constitutes a coaching experience is understandable, especially given the dearth of coaching process research.
This study began with a broad interest in what happens in a coaching session, and settled into a focus on the interventions that coaches make with their clients through their verbal behaviours. The question of what coaches say seems fundamental to our understanding of the field (Brock, 2014; Graf & Dionne, 2021a, 2021b; Gavin, 2022). Verbal behaviour of coaches was more explicitly defined as all intentional vocal expressions that could be detected in session recordings. Expressions by clients were only considered as context for understanding the meaning or function of coaches’ verbal behaviours. Extant literature on coaching process provided guidance for categorizing verbal behaviours. Intentions were to obtain frequency estimates of coaches’ verbal expressions within a session and to reference findings against normative descriptions of coaching discourse. Guidance regarding such descriptions was largely derived from perspectives offered by the International Coaching Federation (2024c; Passmore & Sinclair, 2020), which advocates for certain types of coaching interventions while cautioning against others. Admittedly, much is omitted from the limited lens of this study, but we believed it to be a useful first step toward appreciating what occurs in coaching experiences. This report will describe the development of our analytic framework, findings from the application of this framework to recorded coaching conversations, interpretations of findings, and questions arising to inform future research directions.
The call for more coaching process research has been persistent, yet the challenges of conducting such research and replicating it to achieve coherent understandings are daunting (Fillery-Travis & Cox, 2014; Fillery-Travis & Lane, 2020; Grant & O’Connor, 2018). According to Ianiro-Dahm and Kauffeld (2022), coaching process can include anything occurring “between the first and last contact between coach and client” (p.40). Process might reference observable and unobservable elements, meaning-making, tasks and actions, and a myriad of nonverbal communications during sessions. As well, it can point to phenomena happening for coach or client before and after any experience or session (Erdös & Ramseyer, 2021; Ianiro-Dahm & Kauffeld, 2022; Myers & Bachkirova, 2018). In the following sections, we highlight specific streams of literature that clearly pertain to our focus and approach.
Myers and Bachkirova (2018) grouped coaching process investigations into three categories of hypothesis testing, event-focused, and classificatory; this study did not fit easily into any of these. However, Hardy and Llewelyn’s (2015) proposal of psychotherapy research categories, as described by Myers’ (2017), offered a more apt possibility. These authors identified a category called descriptive studies that “aim to provide a clear account of behaviours and processes which can be observed to occur during therapy, or in the accounts, beliefs, feelings and behaviours of participants, and which do not yet have a theoretical base” (p. 186). Our study was atheoretical. It focused on behaviours, on what could be heard during coaching sessions. Even so, a degree of interpretation guided largely by practice-based perspectives was required in our process. The next section will further explore coaching behaviours and identify sources for our meaning-making.
This study focuses exclusively on the coach’s verbal behaviours. While added value would be realized by analysing both sides of the coaching relationship (Gessnitzer & Kauffeld, 2015; Graf & Dionne, 2021a; Heineke, 2013; Ianiro-Dahm & Kauffeld, 2022; Klonek, Will, Ianiro-Dahm, & Kauffeld, 2020), for pragmatic reasons we limited our attention to the coach’s voice alone. In this work, verbal behaviour was considered as intentional vocalizations that could be heard and thus transcribed from recordings. A sound such as a sneeze or cough would not be an intentional communication, whereas laughter or a sound such as “mmm hmm” would be.
Relatively few coaching studies have examined the verbatim discourse within sessions. While a few have ambitiously analysed interactional patterns between coach and client (Gessnitzer & Kauffeld, 2015; Graf, Aksu, & Rettinger, 2010; Klonek et al., 2020), others like ours attended mostly to what the coach was doing (Bachkirova, Sibley, & Myers, 2015; Greif et al., 2010). However, it is important to note that the way in which verbal behaviours have been considered has varied depending on the type of process research undertaken (Myers, 2017; Myers & Bachkirova, 2018).
In the limited number of studies that coded verbatim coach-client behaviours in sessions, we sensed a need like ours to explicitly or implicitly rely on frameworks to devise categories of behaviour for analysis. For instance, in works by Lofthouse and colleagues (Leat, Lofthouse, & Towler, 2012; Lofthouse & Hall, 2015), dimensions emerged from an iterative process of analysing actual coaching conversations, where some elements were proposed by experienced team members and others “became obvious as the transcriptions were analysed” (Lofthouse & Hall, 2015, p. 760). Of particular interest to our work, these researchers identified 17 “conversational functions” labelling elements of the “purposes, processes and outcomes of interaction” (Lofthouse & Hall, 2015, p. 761); examples included questions, challenges, suggestions, defence, and acceptance.
Stein’s (2009) work on process considered segments of a coach’s remarks, which could have been a phrase, a sentence, many sentences, or even multiple exchanges with a client. Behaviour was described in terms of roles that the coach was assumed to be taking in each remark. Stein called these the “hats” that coaches wear, identifying 16 such roles labelled as agenda facilitator, expert, guide, and so on. There is heuristic value in this investigation, though its role labels represent interpretations of behaviour rather than behavioural categories per se.
In a study of coach-client pairs, Ianiro, Schermuly and Kauffeld (2013) focused on behaviours reflecting themes of dominance and affiliation. This study relied on a coding system that categorized both verbal and non-verbal behaviours of participants on the two dimensions of interest. As noted by Myers and Bachkirova (2018), such hypothesis testing investigations intentionally limit the scope of interest within a coaching process.
In Gessnitzer and Kauffeld’s (2015) work, a unique method of interaction analysis captured “coded relationship-relevant behaviors initiated by the coach or the client” (p.177) to gain insights about the working relationship. Their study represents a theory-driven coding system highlighting the working alliance as evidenced in video-recorded sessions. As thorough as the study was, its theoretical focus precluded a more comprehensive analysis of all the different categories of coaches’ verbal behaviours.
While Klonek and colleagues (2020) examined video recordings of career counselling sessions, their findings seemed relevant due to similarities to coaching in approach. Their work relied on the microskill definitions initially advanced by Ivey (1983), coding behaviours according to such micro-behaviours as paraphrasing, questioning and summarizations They were particularly interested in the relation of micro-behaviours to goal and task orientation. The study demonstrated that micro-behaviours could be reliably observed and, perhaps more significantly, that certain micro-behaviours were related to expressions of empathy and developing the working alliance.
Heineke’s (2013) study of coaching discourse provided another approach to coding emerging from an interest in learning and professional development among teachers. The methodology examined exchanges and moves, where an exchange represented a back-and forth process while a move was a singular expression that either initiated or followed an exchange. Pertinent to our approach, Heineke’s work tied coding directly to such functions of the coach’s verbal behaviours as questioning, clarifying, evaluating, or justifying.
Works by de Haan and associates (de Haan, 2008a, 2008b; de Haan, Bertie, Day, & Sills, 2010; de Haan & Niess, 2012, 2015) analysed coaches’ descriptions of critical moments in coaching rather than verbatim dialogue. Critical moments were seen as occurring in “an exciting, tense, or significant moment” (de Haan, 2008a, p. 92) with a client in a coaching relationship. A framework resulting from their explorations suggested 18 different representations of anxiety in a coaching relationship. A useful by-product of this group’s work was the creation of a “Coaching behaviours questionnaire” (de Haan & Burger, 2014) that listed 60 actions that coaches might express in different moments of their sessions. These actions crafted from practitioners’ reports seemed reasonably descriptive of what coaches do in sessions.
Coaching literature largely shows that process researchers create idiosyncratic systems to understand coach-client interactions based on theory- or practice-based ideas about what they interpret as happening in the moment. Though our coding was influenced by a wide range of perspectives, we relied most on four sources or frameworks that were largely descriptive or practice-based:
Our codes were intended to point to specific functions yet be broad enough to capture a coherent array of verbal expressions. For instance, reflections in Ivey’s works (Ivey, 1983; Ivey et al., 2023) were subdivided into ones regarding content and those mirroring emotions. In a similar manner, when considering questions in our data, codes reflected more than the fact that a coach asked a question by suggesting the question’s function, for instance, whether it was intended to explore outcomes, learning, relationships, or feelings.
Another consideration was how to segment the coach’s verbalizations for analysis. As noted, Stein (2009) looked at remarks, Heineke (2013) considered exchanges and moves, and Klonek and colleagues (2020) analysed units, defined as “the smallest meaningful communicative statement that could be understood by the other member of the interaction” (p.369). In their study, only one code was assigned to each unit, which we generally thought appropriate recognizing, however, that a single response might be comprised of multiple units.
Transcripts of coaching sessions provided the raw data for a hybrid qualitative and quantitative approach to the analysis. The first part of the data analysis involved an examination of transcripts following a procedure informed by discourse analysis (Hardy, Phillips, & Harley, 2004; Jones, 2012; Phillips & Hardy, 2002; Wood & Kroger, 2000), such that following the creation of a codebook, each verbal expression by the coach was coded with respect to the function it appeared to play in the context of a coaching session. This corpus of coded interactions was then used as input to a quantitative examination of frequencies and changes over time.
The study involved 48 coach-client pairs. Virtually all participants were active professional coaches participating in a continuing development program endorsed by professional coaching associations (e.g., ICF, EMCC) called ReciproCoach (https://reciprocoach.com). Coaches coached other coaches who had also volunteered for the experience. In the ReciproCoach program, each registrant was assigned a coach who would coach them for up to 6 sessions, as well as being assigned a different coach who would serve as their client for up to 6 sessions. This experience was unmonitored.
The study was launched after a full review and approval by the Human Research Ethics Committee at Concordia University in Montreal. Participants reviewed and signed a consent form that fully described participation requirements and the management of confidentiality. They were not compensated financially.
Note: PCC = ICF professional certified coach; ACC = ICF associate certified coach
As shown in Table 1, most coaches (88%) and clients (73%) identified as female, and all coaches and most clients (90%) were over 35 years of age. The gender distribution was more female than statistics indicate about the profession (ICF, 2020). In addition, all coaches who answered the question about credentials (85%) had achieved associate (ACC) or professional (PCC) coach certification from ICF.
Participants submitted recordings from their second, third and fourth of six coaching sessions representing the contractual ReciproCoach relationship. The decision to exclude the first session was based on the presumed agenda of first sessions such that significant time is typically spent in relationship building and identifying elements of the overarching coaching objective (Gavin, 2022; Ianiro-Dahm & Kauffeld, 2022). Both coach and client completed questionnaires following each of three consecutive sessions, though these post-session questionnaires were not considered in this set of analyses.
Participants submitted a digital audio file for each session. To maintain confidentiality, all files were identified by a pseudonym chosen by participants. Recordings were initially submitted to Temi, a web-based transcription service (https://www.temi.com), and then reviewed by team members to accurately represent every intentional audible expression by coach and client. Each transcript was reviewed twice by team members as they listened to recordings.
From the sample of 48 coach-client pairs, we obtained a total of 47 recordings for session 1, 48 for session 2 and 47 for session 3, thus missing one submitted recording from sessions 1 and 3 (from a different coach-client pair).
Over a four-month period, team members met weekly for 2–3-hour sessions to review a random set of transcripts and suggest labels for coding categories. As this process continued, definitions with examples for each code were created. Following the initial four months, biweekly sessions continued for another three months wherein new transcripts were coded and modifications to the original codes were made. By the eighth month, a final codebook was realized, and the formal coding process began. All previously coded transcripts were recoded based on the final code definitions.
The resultant set of 37 codes captured 99.6% of coaches’ verbal utterances. The 37 codes were grouped into four clusters: (1) Active listening, largely evidenced by reflections of content and feelings, was represented by four codes. (2) A large cluster of 17 different types of questions, based on the apparent function of the question, was referred to as inquiring. For instance, a question might pertain to the content of what a client just expressed, to the goal the client desired, or to values underlying the client’s remarks. (3) A third cluster of 11 codes was described as influencing in that the coach was thought to be offering interpretations, metaphors, information, advice, or perhaps feedback. (4) A fourth cluster was called miscellany for lack of a more descriptive term; it consisted of five codes. These included interventions that seemed to represent casual or social conversation (“chit chat”) and coaches’ externalization of internal thought processes (e.g., “I was wondering to myself as I was driving around yesterday afternoon and listening to some music that I like….”). Other codes were created for verbalizations where the coach seemed to direct the client’s agenda (e.g., coach tells the client what the agenda should be), when the coach ignored what the client explicitly expressed, and when they interjected in a manner that seemed disconnected or suggesting a loss of self-regulation (e.g., an unrelated self-disclosure or blurting an irrelevant comment). Interventions coded in the fourth cluster may have a higher degree of subjectivity in their determination. Code names are provided in Table 2 while brief definitions are described elsewhere.
Each time a coach intervened verbally, whether minimally or in a lengthy remark, at least one code was assigned to the intervention. In simple comments, such as “Okay” or “Makes sense,” a single code would be attributed to the remark, and it would be counted as an instance of that code. Longer interventions might be comprised of multiple units and result in different codes for each unit. In cases where a coach might repeat the same words representing the same type of intervention (e.g., “Okay, okay”), only one code would be counted in the appropriate category. In cases where a verbal remark might be equally representative of two related categories, only one category would be scored. Exceptions were made for categories 36 and 37, where the content of the coach’s verbalization would be coded as well as the fact that this intervention might be considered dysregulated (code 36) or have ignored the client’s focus or request (code 37).
Two team members independently coded 10 randomly-chosen transcripts, and Intra-class correlation coefficients estimated their interrater reliability (Fisher, 1954; Liljequist et al., 2019; Shrout & Fleiss, 1979). The intra-class coefficient across all cases was .963, while reliability varied somewhat across aggregated scores for the four clusters of codes as follows: Active listening – rxx = .969; inquiring – rxx = .786; influencing – rxx = .863; miscellany – rxx = .959.
Data analysis was performed in SPSS version 29. Codes were analysed using three different formats for the purpose of different analyses: a) raw count scores, b) within-participant percentage scores, and c) frequency/minute scores. Raw count scores were used to compare the total number of interventions across the three sessions. This comparison was statistically treated as a mixed effects analysis of variance with frequency scores as the dependent variable and time as a repeated measure factor. Within-participant percentage scores were used to rank the frequency of interventions as well as to compare relative changes in the frequency of occurrence of a certain strategy over time. Statistical comparisons for percentage scores were obtained by means of non-parametric related-samples sign tests. Frequency/minute scores were used to further validate the analysis on overall raw count scores as well as to assess changes over time of specific interventions. The comparison in frequency/minute scores across the three sessions was statistically treated as a mixed effects analysis of variance with frequency/min scores as the dependent variable and time as a repeated measure factor.
Findings will be reviewed in two sections: The first considers the frequency of each intervention by coaches in sessions, while the second examines changes in coach interventions across three consecutive sessions (2nd to 4th).
Table 3 shows the percentage for each intervention made by coaches in our sample, combined across the three coaching sessions (means and standard deviations are computed as inter-coach scores). It was created after removing the category of minimal encouragers from the analysis, since these interventions alone constituted 46.1% (s.d.=13.9%) of all interventions. Removing this category from Table 3 calculations allowed differences among other interventions to be more evident. Moreover, minimal encouragers typically represented few words or almost guttural responses (mmm hmm), compared to other interventions which might take form over one or more sentences.
Percentage scores based on all but the minimal encourager category are listed from most to least frequent. To aid understanding, we used a qualitative criterion to segment the list into four layers (top, middle, bottom, fractional). The top layer (12.3%-5.2%) contains the highest frequency interventions; the middle layer (3.8%-2.0%) lists those occurring relatively less frequently; the bottom layer (1.6%-0.5%) represents quite infrequent behaviours; the fractional layer (0.3%-0.1%) includes rare interventions. These layers were created by examining the gradient of differentials between the mean occurrence of one code and that of the next less frequent code. Points of discontinuity were judged to occur when a code was followed by a notably less frequent code; these are highlighted as shaded cells in the “Differentials” column.
Beyond tallying and ranking intervention frequencies, we examined whether the rates of coach interventions changed over the course of three consecutive coaching sessions within the same coach-client pair.
First, we investigated whether coaches tended to intervene at different rates over time. The raw counts for all types of interventions were summed, thus creating an “overall” category, which was statistically compared over the three time points in a mixed effects analysis of variance. Findings presented in Table 4 show that coaches made fewer interventions as work with clients progressed from the 2nd to the 4th session. Importantly, this finding was not merely a by-product of shorter sessions since the average duration of sessions did not vary significantly across the three sessions.
Notes: Scores for session 2-4 are reported as raw counts. ME = Minimal Encourager. Since MEs had both high overall frequency and high variation between coaches, Total interventions and Interventions / minute scores were computed with and without MEs to represent how this unique category might confound findings.
A closer look at change involved investigations of each code. Statistically significant changes were found for several interventions, in most cases in the direction of decreasing frequency over the course of three coaching sessions. The only exception was for Language Exploration, which doubled in frequency between the first and third session. Table 5 displays a summary of these findings. Frequency/minute scores were used instead of raw count scores to prevent the analysis from being confounded by differences in session duration.
Notes: Scores for session 2-4 are reported as interventions / minute. * = Statistically significant finding. ⁺ = Marginally statistically significant finding.
Finally, we investigated the pattern of change over time for each intervention relative to the others, rather than in absolute terms. If the frequency of an intervention remained stable over time in absolute terms, against a backdrop of overall decrease in interventions, that stability would in fact represent a relative increase in reliance on that strategy by the coach. To examine this pattern, we computed percentage scores for each intervention with respect to the total number of interventions in that session. We then used nonparametric statistics to compare the percentage frequency of each intervention between 2nd and 4th sessions (Table 6). Looking at the data in this way, a previously unseen finding emerged: a statistically significant increase in the relative frequency of the exploration of importance and meaning.
Notes: Scores for session 1 and 3 are reported as percentages for the frequency of that intervention from the total of all interventions in the same session. * = Statistically significant. ⁺ = Marginally statistically significant.
Considerations of our results rest upon the unique ground of our approach. Similar to other process studies where verbatim remarks were coded according to unique methodologies (Ianiro-Dahm et al., 2015; Myers & Bachkirova, 2018; Stein, 2009), this work presents the reader with 37 coding categories largely unrelated to theoretical perspectives. As best possible, we labelled what we thought the coach was doing in the moment. We relied on a variety of frameworks to guide our coding though none of these adhered strongly to any particular theoretical perspective. As an analysis of coaching interventions, our results allow for a characterization of single session patterns as well as comparisons across sessions. The novelty of our coding framework and longitudinal approach precludes direct comparisons with other process studies.
Percentages estimated for different codes of intervention (Table 3) afford a reasonable narrative about what coaches do. The top layer of codes accounts for almost 60% of what coaches verbalize in sessions excluding their use of minimal encouragers. One might think of these as favoured interventions that coaches rely on multiple times in a session. The portrait is one of coaches listening actively, verifying what they hear, exploring content, inquiring about goals, being friendly (“chit-chat”), affirming the client, offering advice or information, and occasionally interjecting inappropriately.
A disquieting note in this top layer might be found in the high percentage of input from coaches in the form of advice and information (M=11.5%; s.d.=4.6%). The International Coaching Federation’s (2024a; 2024b; 2024c) guidance for coaches allows for occasional offers of this nature by coaches, yet we might query whether this level of frequency leans more toward a consulting than coaching approach for some coaches in our sample.
The middle layer in Table 3 accounts for another 25% of interventions, each of which might occur a few times during a session, ranging between frequencies of 2% and 4%. This layer largely portrays coaches as engaged in inquiry and influencing behaviours. All types of inquiries summate to 33.5% of coaches’ interventions and inquiries in this layer (13.5%) are focused on the structure and process itself, action and design matters, values and beliefs, emotions, and commitments and accountability. These kinds of questions seem appropriate to a coaching role. Influencing interventions in this layer indicate behaviours of challenging clients, interpreting and reframing perspectives, and using humour; these sum to almost 12%.
In the bottom layer are 15 interventions occurring on average between 0.5% and 1.6% of the time and totalling 14.8%. Actions reflected in this layer largely make sense as professionally appropriate coaching behaviours. However, a few merit further comment. Combining reflections of emotion (1.4%) with inquiries about emotions (2.6%) from the middle layer documents that coaches spend relatively little time emphasizing or delving into emotional content and supports arguments against overdirecting coaching discourse toward feelings or emotional matters (Grant, 2001, 2006; Ives & Cox, 2012).
Self-disclosing comments by coaches also deserves further consideration. Should coaches extensively share personal thoughts and feelings, a problematic depiction of coaching as a form of friendship might be created (Kaneh-Shalit, 2017; Neenan & Dryden, 2013). While our data show that coaches engage in a degree of informal conversation (“Chit chat” – 6%), self-disclosures occur far less frequently (0.7%). Moreover, most coded instances of self-disclosure in this study were ones judged to be appropriate to the dialogue and in service of the client’s agenda (Bachkirova, 2016) as contrasted with self-disclosures intended to develop personal intimacy (Cozby, 1973; Jourard, 1971) or thought to be inappropriate (e.g., some instances coded as “dysregulated interjections”).
Data in Table 3 also provide support for the influence of positive psychology principles on coaching practice (Biswas-Diener, 2009; Green & Palmer, 2018; Lomas, 2019). Disconfirming feedback was expressed only a fractional percentage (0.2%) of the time, whereas coaches were estimated to express affirming feedback at an average rate of 6.6%. As well, such findings reflect the ICF’s stance that coaches should show positive support for their clients without being cheerleaders (Gavin, 2022; ICF, 2024c; van Nieuwerburgh, 2017).
The final layer in Table 3 references codes occurring less than 0.5% of the time. It includes instances where coaches express disconfirming evaluative feedback, as well as non-verbal explorations, inquiries into background details about clients and questions about the larger ecosystem in which they exist. The low occurrence of non-verbal inquiries (e.g., “Where are you feeling this in your body?”) might indirectly reflect coaches’ non-therapeutic focus in their work.
As noted, minimal encouragers represented coaches’ most frequent interventions in absolute terms, reaching nearly half of all coaching interventions in sessions (average frequency 46.1%). However, variability on this intervention was high (s.d.=13.1%) such that some coaches expressed minimal encouragers as little as 8% of their totals while others did so as often as 66% of the time. The implications of varying levels of minimal encouragements in coaching deserve further investigation.
Before moving from this part of the discussion, codes grouped under the Miscellany category deserve comment. The extent of chit-chat (M=6.0%; s.d.=4.1%) occurring in sessions may possibly augment trust and relationship development or erode the professional nature of the work (Anderson & Gehart, 2006; Rogers, 1957; Shah et al., 2022). With high variance on this code, some coaches may indeed have weakened the professional nature of their relationship through excessive social discourse. However, other behaviours in the miscellany category could be more concerning. These included: instances where a coach benignly externalized their internal thought process without adding value (1.6%); ignored or failed to capture key elements of a client’s experience (0.6%); or ‘hijacked’ the agenda by redirecting the client from what they were talking about to something the coach thought should be the focus of dialogue (0.5%). Happily, these behaviours had low frequencies and, in this respect, indirectly reflect the largely appropriate and positive behaviours of professional coaches.
Our last mention in the miscellany category concerns “dysregulated interjections,” which had an average frequency of 5.2%. Examples included times when coaches laughed inappropriately, made a judgmental comment, or self-disclosed in ways that seemed disconnected from the client’s remarks. Instances of this code were largely seen as unhelpful rather than harmful, and none were thought to raise significant ethical concern. Variability on this code (SD = 9.1%) indicates that most coaches had none or perhaps one instance of this code, while others had a troubling number. This point also merits further focus.
If coaching intentionally promotes client autonomy (Grant, 2020; Losch, Traut-Mattausch, Mühlberger, & Jonas, 2016), our findings as represented in Table 4 can be interpreted as speaking to this agenda. The trend of declining interventions by coaches over a series of three sessions might suggest a growing self-direction over time in coaching relationships such that clients’ reliance on coaches for guidance appropriately diminishes. As clients learn the process of operationalizing intentions in concrete ways and formulating practical strategies for goal-attainment, they may come to depend less on their coaches and more on themselves to make desired changes.
Changes over time in the occurrences of each of the 37 different codes do not suggest a coherent narrative about why certain interventions increased or declined while others did not. As well, only three showed statistically significant changes at traditionally applied levels of significance (Table 5). Except for reflections of content, codes reflecting change had rather low frequencies of occurrence in any given session, raising questions about their overall pertinence to coaching conversations. In a different look at change over time as seen in Table 6, the few significant differences identified were also difficult to interpret in any meaningful way. The most that might be said is that coaches seem less active in reflecting content the more they continue with their clients.
A conservative interpretation of the change data argues more for a stable pattern of coach interventions from session to session rather than for a meaningfully varied one. This view coincides with the prescribed architecture for coaching as articulated by professional coaching bodies (Gavin, 2022; ICF, 2024c; Passmore & Sinclair, 2020; van Nieuwerburgh, 2017). In coach certification assessments, for instance, the International Coaching Federation applies the same rubric for evaluation to coaching sessions irrespective of whether the session might be the first or the last in a series.
Our intention was to gather from as many professional coaches as our budget would allow. We were reasonably successful though all our participants were English speaking, from Western nations, and largely aligned with one of the many professional coaching associations. More critically, the sessions analysed were those of coaches coaching other coaches. How much of an issue this last point is remains unclear.
As much as we were deliberate in our efforts to understand how previous researchers had approached coding verbal interactions in coaching sessions, we nonetheless created a system that was uniquely different, thereby adding to the Tower of Babble sense of codes in coaching process research. Nonetheless, we take pride in what we created and will retain this system for future investigations.
A further limitation is seen in our methods of analysis which precluded any understanding of patterns of intervention and their impacts. The data provide a profile of average frequencies over time, which yields no real insight into unique styles or noteworthy sequences of interventions and their value. Fortunately, we will be able to revisit our data to analyse more complex questions about coaching interventions.
Finally, our data do not speak to relations between codes and outcome measures. Though we obtained post-session perceptions from both coaches and clients, our determination to maintain clear focus on the question, ‘what do coaches say,’ left little space for introducing yet another set of intricate methods and analyses in this writing. We will be reporting these results in a separate monograph.
Presenting a clear and consistent message to the world about what constitutes a coaching experience is essential. Yet, as a relatively new field of practice, this objective may become ever more elusive as new approaches proliferate. Without clearer guidance and boundaries to practice, there is some concern that Brock’s (2008, 20214) observation that “coaching is what coaches do” may characterize the field.
Referencing a chief objective of scientific inquiry, we believe this study offers a description of what coaches do in sessions and documents the consistency of this profile over a limited series of sessions. In identifying what coaches do most, least and with intermediate frequency, it is encouraging that study findings seem highly congruent with guidelines about what coaches are supposed to do according to professional coaching bodies (ICF, 2024a, 2024b, 2024c).
As well, data do not reveal dubious coaching interventions emanating from esoteric models of coaching. As noted, coaching practice will continue to evolve so that multiple profiles of intervention are likely to be identified in the future. Even so, this study affords a reasonable depiction of what coaches do now and which, we believe, is likely to represent a central coaching process for the foreseeable future.
Coaches, coaching bodies and the general public should feel comforted by these results. They strongly counter criticisms of coaching as a kind of “wild west” of professional helping relationships, offering instead an image of a structured and supportive dialogue that veers away from friendly chatter and from deep emotional explorations. Of course, depictions based on averages hide the extremes which in our study showed up in the high usage of minimal encouragers, concerns about the extent of advice-giving, and evidence of high levels of dysregulated behaviours by some of our participants. Questions about ongoing monitoring of practicing coaches, requirements for continuing education and even ones about industry regulation need to be continually revisited.
James (Jim) Gavin is full professor of Applied Human Sciences and Director of the Centre for Human Relations and Community Studies at Concordia University.
Nicolò Francesco Bernardi is a psychologist and professional coach specializing in developing emotional resilience in highly stressful environments ranging from families in crisis to business start-ups.
Elizabeth (Betsy) Thomas founded and led a manufacturing business and, following her MA in Human Systems Intervention, practices as an Organizational Development consultant.
Julie Chacra is a PhD student at Concordia University, interested in youth worker perspectives on family-school relationships and interdisciplinary collaborations in special education processes.