The evolution of motivational interviewing

Abstract

This review traces the development of motivational interviewing (MI) from its happenstance beginnings and the first description published in this journal in 1983, to its continuing evolution as a method that is now in widespread practice in many professions, nations and languages. The efficacy of MI has been documented in hundreds of controlled clinical trials, and extensive process research sheds light on why and how it works. Developing proficiency in MI is facilitated by feedback and coaching based on observed practice after initial training. The author reflects on parallels between MI core processes and the characteristics found in 70 years of psychotherapy research to distinguish more effective therapists. This suggests that MI offers an evidence-based therapeutic style for delivering other treatments more effectively. The most common use of MI now is indeed in combination with other treatment methods such as cognitive behaviour therapies.

Introduction: happenstance beginnings

An inherent part of history is that as events are occurring, we do not know where they will lead. Seemingly chance encounters can have substantial enduring effects on the course of a life, a field, or a community (Bandura, 1982). Motivational interviewing began through such a convergence of happenstance events (Moyers, 2004).

Eugene, Oregon

In the 1970s when this journal was still called Behavioural Psychotherapy, I received my pre-doctoral training in the clinical psychology program at the University of Oregon, which was unabashedly behavioural in orientation. Even cognitive therapy was viewed with scepticism. Before learning and practising behaviour therapy we were required to complete a year-long course on how to talk with clients. As it happened that year, none of the behavioural faculty chose to teach the course, so from the counselling psychology faculty they engaged Susan Gilmore who had been mentored by Leona Tyler and thus was an academic grandchild of Carl Rogers. She taught us fundamentals of therapeutic relationship from a client-centred perspective, with particular focus on the skill of accurate empathy (Gilmore, 1973).

Milwaukee, Wisconsin

Immediately following this course, my introduction to addiction treatment came in 1973 as a summer psychology intern at the Veterans’ Affairs Medical Center in Milwaukee. The internship director invited me to tour the treatment programs and choose where I would like to learn. On the in-patient alcohol unit the psychologist director, Robert Hall, asked me what I knew about alcoholism. ‘Nothing’, I confessed. But what had I learned in my graduate courses? Actually, I didn’t recall the topic ever coming up. ‘Well’, he advised, ‘this is one of the most common diagnoses you will encounter in practice. Come and learn’. I spent the summer there and knowing nothing, I used my rudimentary listening skills to learn from the patients who taught me about alcoholism. The experience stirred me to head upstream in search of ways to intervene earlier before alcohol causes the depth of suffering that I witnessed there. It shaped my dissertation (Miller, 1978), and addiction treatment became a passionate focus of my career.

Albuquerque, New Mexico

Joining the clinical faculty at the University of New Mexico (UNM) in 1976, I began conducting clinical trials of behaviour therapies for alcohol use disorders. Having no extramural funding, I trained our own students to be the counsellors, teaching them both behaviour therapy and empathic listening. In one of these trials three supervisors independently rated nine therapists on both their adherence to behavioural procedures and their skill in accurate empathy using a scale developed by Carl Rogers’ research group (Truax and Carkhuff, 1967). To my surprise, therapist empathy while delivering behaviour therapy accounted for two-thirds of the variance in clients’ drinking outcomes (r=.82 at 6 months), an effect that persisted for two years after treatment (Miller and Baca, 1983; Miller et al., 1980). In contrast, the various behavioural treatment procedures we were comparing did not differ from each other in efficacy. There seemed to be something important about the therapeutic relationship.

Hjellestad, Norway

These data followed me on a sabbatical leave in 1982 at the Hjellestad Clinic near Bergen, Norway, where I was invited to lecture on the behavioural treatment of alcohol problems. The clinic’s medical director asked whether I might also meet regularly with their psychotherapists, which I gladly did. I again offered training in both client-centred and behavioural methods, and they wanted to observe my own clinical style. They role-played clients whom they were finding challenging and I did my best to respond, at which point they did something that my American students rarely did: they interrupted me in the middle of demonstrating. ‘What are you thinking right now?’, ‘Why did you ask that particular question?’, ‘Of all the things the client said, how did you decide to reflect that?’. Their thoughtful questions evoked from me some decision rules that I seemed to be using but of which I had not been consciously aware. Together we were following in the footsteps of Carl Rogers whose client-centred approach arose not from a pre-existing theory but from the careful observation and analysis of clinical practice (Kirschenbaum, 2009; Miller and Moyers, 2017).

From those discussions I wrote an informal discussion paper describing an approach called motivational interviewing (MI) and I mailed it to several respected colleagues for comment, one of whom was Ray Hodgson in Wales. To my surprise, he replied that he wanted to publish it as an article in Behavioural Psychotherapy, which he edited at the time. I protested that I had no data, but he persisted and it became the first published description of MI (Miller, 1983). Unbeknownst to me, one of the people whom Ray chose to review the manuscript was a young colleague named Steve Rollnick.

The drinker’s check-up

Returning to New Mexico and newly interested in motivation for change (Miller, 1985) I began considering how this approach might be turned into a treatment method. To reach problem drinkers early, before they developed serious consequences, we decided to advertise a free ‘drinker’s check-up’ for anyone wondering whether their alcohol use might be harming them (Miller and Sovereign, 1989). We developed a thorough check-up using measures sensitive to early adverse effects of alcohol use, providing individual feedback of these assessment results using an MI style. Many drinkers responded, and relative to a random waiting list control group, their alcohol use decreased significantly following the check-up (Miller et al., 1988; Miller et al., 1993). This MI-based check-up model has since been adapted to address marijuana use (Stephens et al., 2007; Walker et al., 2015), marital distress (Morrill et al., 2011), and children’s behaviour problems in families (Dishion et al., 2008; Dishion et al., 2014; Shaw et al., 2006). A computer-based adaptation of the check-up also proved to be effective (Hester et al., 2005; Walters et al., 2007b).

Sydney, Australia

For a second sabbatical leave in 1989–90 I was invited to the National Drug and Alcohol Research Centre at the University of New South Wales. I found myself officed next to an affable South African from Wales named Steve Rollnick who was there to do a PhD thesis on brief MI with heavy drinkers in a hospital setting. ‘Are you the Miller who wrote that article on motivational interviewing?’ he asked. He had been teaching MI in the United Kingdom where it was already becoming a popular method in addiction treatment. ‘I don’t even know if I’m doing it right’, he said. ‘You need to write more about it’. Through a year of collaboration we developed the first edition of the MI text on ‘preparing people to change addictive behavior’ (Miller and Rollnick, 1991). Steve brought fresh perspective and depth to the psychotherapeutic method, adding the concept of resolving ambivalence and using his practical experience from teaching MI. It has been our experience ever since that through conversation we arrive at clearer understanding and expression than either of us would have found alone. He soon began applying MI more broadly in health care, where ambivalence about behaviour change is a common obstacle (Rollnick et al., 1992; Rollnick et al., 1999).

Transtheoretical stages of change

In the early 1980s while MI was emerging, a transtheoretical model of change was also being developed by Prochaska and DiClemente (1984). MI had synergy with their model by offering an early evidence-based treatment method for working with clients who were not yet ready for change; i.e. in the pre-contemplation, contemplation and preparation stages (Prochaska and DiClemente, 1992; Prochaska et al., 1994). Resonance between MI and the stages of change was explored in writings from both research groups (DiClemente et al., 2017; DiClemente and Velasquez, 2002; Miller and Rollnick, 1991) and MI found its way into Project MATCH.

Project MATCH

It has long been known that seemingly different psychotherapies typically yield similar outcomes when compared head-to-head in clinical trials (Imel et al., 2008; Rosenzweig, 1936; Wampold and Imel, 2015a). It is possible, however, that particular therapies may be most beneficial for different kinds of people, a client–treatment interaction that can be masked when averaging outcomes of therapies. This idea of matching treatments to clients was the central purpose of Project MATCH, the largest randomized clinical trial ever conducted with psychotherapies for alcohol use disorders (Babor and Del Boca, 2003). In funding the trial, the U.S. National Institute on Alcohol Abuse and Alcoholism (NIAAA) did not specify the treatments to be studied. Instead, nine clinical sites including UNM were selected to participate based on applications describing how they might conduct the trial. Then in 1990 the nine principal investigators, a coordinating centre, and NIAAA staff began negotiating the eventual design of the multisite clinical trial (Project MATCH Research Group, 1993).

We first decided on the therapies to be tested. We determined that the design could accommodate three different treatment methods, and we wanted them to be quite different in theoretical orientation. Given the prior work of the investigators it was a foregone conclusion that one of the treatments would be cognitive behavioural (Kadden et al., 1992). Because a 12-step approach was prevalent in U.S. community treatment programs, we developed 12-step facilitation therapy as a second method to be compared (Nowinski and Baker, 1998). What should be the third? We considered a variety of evidence-based treatments (Institute of Medicine, 1990; Miller and Hester, 1986) and chose MI as being quite different conceptually from the other two. The research team was uneasy, however, that the usual 1–2 session duration of MI could be insufficient, given that the other two treatments would consist of 12 sessions. Could MI be made longer? We settled on a 4-session format of the drinker’s check-up and called it motivational enhancement therapy (MET; Miller et al., 1992). As usual, the three treatments proved to be equally effective (Project MATCH Research Group, 1997).

The research group also developed detailed a priori predictions about which clients would fare better in each treatment, resulting in 18 matching hypotheses to be tested (Longabaugh and Wirtz, 2001). Most of these predictions were not confirmed; in fact seven yielded interactions in a direction opposite from what had been predicted. Bottom line: we are not very good at predicting which psychotherapy will be best for whom. One prediction that was confirmed is that angrier clients were more likely to benefit from MET (Waldron et al., 2001). Subsequent analyses also found that MET was more effective for females (but not males) with lower initial motivation (Witkiewitz et al., 2010). In any event, Project MATCH did increase awareness of MI as a bona fide way of treating alcohol use disorders.

The spirit of MI

In our early MI training workshops Steve Rollnick and I focused on treatment techniques, giving particular emphasis to those for responding to resistance, a common interest of trainees. As we watched people practising what we had taught them, however, we were dismayed. It was as if they were trying to use the techniques on clients, trying to hoodwink them into changing. The words were there, but the music was wrong. The fault was ours, not theirs; clearly we had omitted something important in our training.

In this journal we described the missing element as the underlying spirit with which MI is practised (Rollnick and Miller, 1995). Carl Rogers had much earlier described important attitudes underlying a client-centred approach: empathic understanding, unconditional positive regard, and genuineness (Rogers, 1957; Rogers, 1959). Such attitudes are expressed in clinician behaviour, of course; indeed they must be communicated in order to benefit clients. Both internal (attitude) and external (behaviour) components matter in therapeutic relationships (Miller and Moyers, 2021).

Our understanding of these attitudinal components of MI has evolved over time. In our second edition we briefly described three elements of the spirit of MI as collaboration, evocation and autonomy (Miller and Rollnick, 2002). Our third edition (Miller and Rollnick, 2013) contained a full chapter on the underlying spirit, describing partnership, acceptance and evocation, and adding compassion – a conscious commitment to the well-being and best interests of the client. We added this fourth component because some of the same techniques involved in MI can be used in self-serving ways, prompting us also to offer some guidelines on the ethical use of MI (Miller, 1994).

Fostering change

The third edition of Motivational Interviewing included two important developments that are reflected in its subtitle: Helping People Change. Although our early expectation had been that MI would primarily be used in the contemplation (ambivalence) stage, practitioners found that they were not ‘letting go of’ MI once readiness for change emerged. Whereas our first two editions had focused on preparing people for change, it became clear that MI can be a more general way of working with clients (Rogers, 1980b) and useful throughout a helping process across a broad range of concerns.

A second development was to broaden the concept of change beyond behaviour. Although it is simplest to describe MI in relation to increasing or decreasing a particular behaviour like exercise or smoking, the same method can also be applied in promoting complex change that is not about specific behaviour except in the broadest sense of the term. An example that we used is forgiveness, a clearly important clinical issue (Worthington, 2003; Worthington, 2005). A choice to forgive may not result in overt behaviour change; the forgiven person may have been dead for decades, and the change is more cognitive and affective in focus (Wagner and Ingersoll, 2008). We were also considering how MI can be applied in primary and secondary prevention outside the realm of treatment (Dishion et al., 2014; Miller et al., 2000).

Principles to processes

In the first two editions we described some general principles of MI, whereas in the third edition we jettisoned these in favour of four component tasks or processes of MI: engaging, focusing, evoking and planning. These may sound linear and sequential: first you engage, next you identify shared goals, then you evoke the client’s motivation to pursue those goals, and finally you develop a plan to reach them. In practice the processes are more iterative and flexible. The ‘how to’ discussion of planning can intermix with the ‘whether’ of evoking, and both can emerge quite early in consultation. Goals may change, and sometimes one needs to re-engage. The empathic clinical skills of engaging undergird all four processes, and evoking is still used to develop plans for change.

The language of change

Paul Amrhein’s psycholinguistic contributions to MI emerged between our second and third editions, informing how we understood and described change talk. Paul’s perspective was that our concept of change talk (as well as sustain talk) was too broad, mixing together a variety of speech acts that to a psycholinguist are quite different. There are verbs signalling desire for change (e.g. I want, wish, would like). Other words describe perceived ability or self-efficacy to change (e.g. I can, could, am able). Both desire and ability speech differ from giving reasons for change that have an if-then structure (e.g. If I do, then …; If I don’t, then…). Furthermore, there are words that bespeak a need or urgency for change without necessarily stating reasons, desire, or ability (e.g. I have to, need to, must). In English these four – Desire, Ability, Reasons and Need – formed the convenient acronym DARN.

Paul distinguished these preparatory (DARN) forms of speech from commitment language of implied promise (e.g. I will, I promise, I guarantee; Amrhein et al., 2003). There are also near linguistic neighbours that are not actually promissory, but nevertheless signal an inclination toward action (e.g. I am willing, will consider, probably will). These are insufficient when taking an oath (e.g. Will you tell the truth, the whole truth, and nothing but the truth?) or entering into a contract, but they still sound more promising than DARN speech and we termed these activation language. Yet another form of change talk emerged from coding MI sessions: taking steps (Moyers et al., 2007). Such statements indicate that the person has already done something that favours change; e.g. I filled the prescription, disposed of all the alcohol, got the book you recommended. Together Commitment, Activation and Taking Steps form the English acronym CATS and are what we call mobilizing change talk.

The very same forms of speech can augur against change as well, and we termed these sustain talk. Such statements bespeak desire, ability, reasons or need for and even a commitment to the status quo. In practice, change talk and sustain talk often come intermixed in the same sentence or paragraph (Miller, 2022), and the ratio of change talk to sustain talk is an operational definition of ambivalence or decisional balance (Prochaska et al., 1994). When directly compared, sustain talk tends to be a better predictor of outcome than change talk alone, and the ratio of change talk to sustain talk is related to the probability of subsequent change (Magill et al., 2018). Sustain talk may be a particularly important predictor with certain populations, such as clients mandated or coerced into treatment. MI-consistent and -inconsistent counsellor responses do influence the balance of client change talk and sustain talk in correlational (Fischer and Moyers, 2014; Magill et al., 2018), sequential (Drage et al., 2019; Moyers and Martin, 2006; Walthers et al., 2019), and experimental studies (Glynn and Moyers, 2010; Miller et al., 1993).

These findings caused us to question the popular clinical technique of constructing a decisional balance by exploring the client’s pros and cons of change. If one equally evokes and reinforces client change talk and sustain talk, the expected outcome would be ambivalence and no change. Experimental trials show that doing a decisional balance intervention with people who are ambivalent tends to decrease their commitment to change (Miller and Rose, 2015). Although MI is sometimes confused with decisional balance, MI involves preferentially evoking and reinforcing client change talk.

Resistance

Another significant development in our third edition was deconstruction of the concept of resistance. The word itself seems to blame the client for not changing via intentional obstruction or pathological processes. Theresa Moyers correctly pointed out to us that most of what we had been calling ‘resistance’ was simply sustain talk – just one side of a person’s ambivalence. Sustain talk is interpersonal behaviour that can be influenced by counsellor responses and is not pathological in itself (Moyers et al., 2017). This raised an interesting question: If we were to subtract sustain talk from what therapists mean by ‘resistance’, what is left? We called the remainder discord. Whereas sustain talk is about the change being considered, discord is about the therapeutic relationship itself and often contains the word ‘you’ (e.g. You’re not listening to me; You can’t tell me what to do).

Both of these components – sustain talk and discord – predict poor outcome if they persist or predominate. Both are interpersonal behaviour and highly responsive to counsellor style. ‘Resistance’ can be dialled up or down by the counsellor within a session, as demonstrated in A-B-A-B experimental studies (Glynn and Moyers, 2010; Patterson and Forgatch, 1985).

Research on motivational interviewing

Clinical trial outcomes

As of 2022, more than 1900 controlled clinical trials involving MI have been described across a broad range of issues, and there are more than 200 published meta-analyses and systematic reviews of MI research. Although MI is now commonly found on lists of ‘evidence-based treatments’, a question like ‘Does it work?’ is far too simplistic. In terms of human change and growth, a better question would be ‘How much of what kind and quality of MI delivered by whom to whom in what context will improve what outcomes how much and for how long?’. Between two-thirds and three-quarters of clinical trials have reported significant benefits of MI, indicating that something about it does seem to be working.

At the same time, outcomes of MI are highly variable. The efficacy of MI differs across studies, by sites within multisite trials, and by counsellors within sites. Average effect sizes are in the small to medium range, again with substantial variability. Effects also vary across target problems and outcome measures. Such heterogeneity is not unique to MI – similar outcome variability is often found in psychotherapy and pharmacotherapy research – but it is certainly characteristic of MI studies.

There are also some patterns of findings. When smaller ‘doses’ of MI (for example, fewer sessions) are compared with other more extended interventions, outcomes are often comparable (e.g. Project MATCH Research Group, 1997; UKATT Research Team, 2005). This renders MI a reasonable baseline treatment with which to compare other psychotherapy or pharmacotherapy. However, when MI is added to other evidence-based interventions, outcomes are often improved or longer-lasting (Hettema et al., 2005; Westra et al., 2016). Efficacy can also vary over time. The earliest trials of MI for smoking cessation showed little promise; later studies, however, yielded sufficiently positive results to regard MI as an evidence-based component of treatment for tobacco smoking (Heckman et al., 2010; Hettema and Hendricks, 2010). Although earlier studies tested MI alone, most clinical trials now use MI as one component of treatments being tested. Particularly common have been combinations of MI with cognitive behavioural therapies (Barrett et al., 2018; Marker and Norton, 2018; Naar and Safren, 2017; Thurstone et al., 2017; Westra et al., 2016).

A further question is for whom MI does or doesn’t help. The only known contraindication to date is based on reports that MI can delay or diminish improvement for people who are already high in readiness or confidence for change (e.g. Kuerbis et al., 2018; Peters et al., 2019; Rohsenow et al., 2004). The mismatch here may be using the wrong MI process; evoking is unlikely to benefit people who are already motivated for change. The planning process and proceeding directly to action seem more appropriate for such clients. MI typically benefits adults and older teens, but efficacy is unknown with younger children, for whom MI with their parents or caregivers can be effective (e.g. Arrow et al., 2013; Cheung et al., 2022; Weinstein, 2007).

Process research

MI process measures usually predict client outcomes. Various reliable measures are available to rate the fidelity/quality of MI from recordings or transcripts of sessions (Lane et al., 2005; Madson and Campbell, 2006; Moyers et al., 2003; Moyers et al., 2005). Providers’ skillfulness in delivering MI accounts in part for variability in findings, with higher-quality MI yielding more positive outcomes (Miller and Rollnick, 2014).

Miller and Rose (2009) distinguished between relational and technical components of MI. The relational aspects derive primarily from client-centred counselling (Truax and Carkhuff, 1967) with particular emphasis on the skill of accurate empathy (Bohart et al., 2002; Elliott et al., 2018). Research on the technical element of MI has focused primarily on evoking change talk, the client’s own expressed motivations for change (e.g. Barnett et al., 2014; Bertholet et al., 2010; Moyers et al., 2007). As mentioned earlier, client sustain talk (voiced arguments against change) appears to be a stronger (inverse) predictor of client outcomes than change talk alone, with the ratio of change talk to sustain talk capturing variance from both (Magill et al., 2014; Magill et al., 2018; Magill et al., 2019). Therapist skill in MI influences the ratio of change talk to sustain talk which mediates client outcomes (Glynn and Moyers, 2010; Moyers et al., 2017; Moyers et al., 2009). In a randomized clinical trial comparing client-centred counselling with and without technical components of MI, the relational elements alone did not improve outcomes relative to a control condition whereas adding the technical components did (Sellman et al., 2001). In a similar experimental trial, Morgenstern and colleagues (2012) found that MI including the technical components yielded change more rapidly than did the relational elements alone, although by the end of treatment the outcomes were comparable in this and a subsequent trial (Morgenstern et al., 2017).

Human coding of MI sessions is a time-intensive and expensive process, particularly with sequential analyses that require reliability at the level of individual responses. A recent development is automated computer coding based on voice recognition and machine learning technology to detect particular responses. An obvious advantage of this approach is that thousands of interviews can be coded within the same amount of time required for human coding of a few samples. Automated coding converges reasonably well with human expert coding at a session level for responses like questions, reflections, and giving information (Klonek et al., 2015). Agreement is more difficult at the level of individual utterances but will likely improve with technology (Atkins et al., 2014; Tanana et al., 2016). Automated coding can be used to provide more economical feedback to interviewers on MI practice (Imel et al., 2019).

Teaching motivational interviewing

Training research

Enough studies have evaluated the effectiveness of MI training methods to warrant reviews and meta-analyses (Madson et al., 2009; Schwalbe et al., 2014). Clinicians who are professionally trained in MI show significant improvement in practice behaviour (Barwick et al., 2012; Schwalbe et al., 2014). Workshop training alone without follow-up typically yields short-lived gains in MI skill that are too small to have much impact on client outcomes (Hall et al., 2016; Miller and Mount, 2001). Adding follow-up feedback and coaching based on observed practice can significantly improve the acquisition and retention of MI skills (Croffoot et al., 2010; Fu et al., 2015; Miller et al., 2004).

Training of trainers

Within 2 years of the first edition of Motivational Interviewing we were receiving enough requests for training that Steve Rollnick and I decided to begin offering an annual workshop to prepare MI trainers. For the initial 1993 workshop in New Mexico, we accepted 40 participants, several of whom would become prominent MI researchers and trainers. After 4 years, some who had completed this workshop asked whether they might gather informally to exchange ideas about how to teach MI, and the first such meeting convened in parallel with our 1997 workshop in Malta. Annual meetings began and led to the eventual formation of the MI Network of Trainers (MINT; www.motivationalinterviewing.org) that has continued offering workshops and prepared more than 3000 trainers in dozens of nations and languages. The COVID-19 pandemic required MINT to offer virtual training for trainer events in 2020 and 2021, and in the process discover the feasibility of this training format. Although there are pros and cons to virtual training, it does facilitate international participation without requiring travel.

Therapist effects

It is clear that client outcomes vary with the therapist providing MI (Project MATCH Research Group, 1998). This is not unique to MI, of course. Outcomes normally differ among therapists delivering treatment at the same agency, or even those following the same manual-guided therapy in a clinical trial (Anderson et al., 2009; Crits-Christoph et al., 1991; Erekson et al., 2020; Wampold and Bolt, 2006; Wampold and Imel, 2015b). Component skills of MI such as accurate empathy themselves predict improved outcomes in research on psychotherapy and counseling (Bohart et al., 2002; Elliott et al., 2018; Truax and Carkhuff, 1967). Nor are such effects limited to psychotherapy; student outcomes vary by their teachers (Chetty et al., 2014), surgical outcomes by physicians (Mehta et al., 2017), and athletes’ outcomes by coaches (Jowett, 2017).

In 2019 I set out with my colleague Theresa Moyers to review 70 years of psychotherapy research to discover characteristics of therapists whose clients have better (and worse) treatment outcomes. What is it that more effective therapists actually do regardless of their theoretical orientation? This is one of the most studied questions in psychotherapy research, dating to the very beginnings of clinical science in psychology (Kirschenbaum, 2009; Miller and Moyers, 2017; Truax and Carkhuff, 1967). We identified eight characteristics of more effective psychotherapists (Miller and Moyers, 2021). The strongest effects are found for accurate empathy – the effort and ability to understand what clients are telling you and to reflect that understanding back to them. Three other therapist characteristics highlighted by Carl Rogers (1959, 1980b) are also linked to better treatment outcomes: expressing positive regard, therapist genuineness or authenticity in the helping relationship, and unconditional acceptance of people as they are. Better client outcomes have also been linked to therapists having a clear focus or goals, conveying and evoking hope, evoking clients’ own motivation and resources, and offering information and advice.

Looking back on this 2-year project, I was struck that at least seven of these eight characteristics of effective therapists have been central components of MI. Table 1 traces these parallels. Only genuineness has been given little consideration thus far in MI, and we mean to correct this in our fourth edition. This causes me to wonder whether what we have actually been studying, developing and teaching for four decades is the dynamics of therapeutic relationship. By far the usual use of MI is not instead of but in combination with other evidence-based treatment methods. MI has always been about relationship and is not allied with a comprehensive theory of behaviour or psychopathology (Atkinson and Woods, 2017; Egizio et al., 2019). It can be and has been used in tandem with a broad range of other services including cognitive behaviour therapy (Naar and Safren, 2017), diabetes education (Steinberg and Miller, 2015), sports coaching (Rollnick et al., 2019), classroom teaching (Reinke et al., 2011; Rollnick et al., 2016), social work (Forrester et al., 2021; Hohman, 2021), pastoral care (Miller and Jackson, 2010), preventive dentistry (Carlisle, 2014), and working with offenders (Bradford and Nandi, 2012; Stinson and Clark, 2017). To the practical know-how of such professions (e.g. in behavioural treatment manuals), MI can add evidence-based guidance for how to be in relationship with those whom you are serving. MI may also offer observable and learnable practices to try when implementing an aspirational humanistic approach such as servant leadership, where practical how-to guidelines have been sparse (Organ, 2021).

Table 1. Parallels of motivational interviewing with characteristics of more effective psychotherapists

Continuing evolution

When someone asks me what I think will be happening with MI in 10 years I am reluctant to say. My reticence is related to how much evolution has occurred in the decade prior to each edition of our Motivational Interviewing text. As we completed each edition, we could not have accurately anticipated what we would be saying in the next. It has never been a matter of making minor tweaks to the prior edition. Each time we have almost completely rewritten the book, and that is happening again as we prepare a fourth edition anticipated for release in 2023.

Nevertheless, some future directions do seem to be emerging. Whereas MI began by focusing on addictive behaviour, it has now been applied and tested across a wide range of behavioural changes. Furthermore, we now understand MI as useful beyond specific behaviours, in helping people change and grow. The concept of ambivalence suggests pros and cons of a particular change, whereas human growth is more about a multiplicity of choices and the development of values. Applications are growing in education (Rollnick et al., 2016), life and health coaching (Clifford and Curtis, 2016; Lawson and Wolever, 2009), sports (Rollnick et al., 2019), organizational leadership (Marshall and Nielsen, 2020; Organ, 2021; Wilcox et al., 2017), probation and parole (Stinson and Clark, 2017; Walters et al., 2007a). MI also appears to cross cultures well, now being taught and practised in many languages and nations on six continents (Self et al., in press).

It is also clear that MI can be provided through media well beyond face-to-face conversations. It can be delivered remotely via telephone and other audio and virtual connections (e.g. Boccio et al., 2017; Cranney et al., 2019; Gates et al., 2012). Interactive journalling is a psychoeducational format that can be MI-informed (Miller, 2014). More recently, MI-based text messaging is being evaluated (Mason et al., 2016; Nelson et al., 2021).

Across four decades of research, Carl Rogers and his colleagues developed person-centred counselling methods with wide applicability (Kirschenbaum, 2009). At first, his insights were presented as a rival therapeutic approach to compete with psychodynamic and behavioural treatment methods. His final book, however, described a person-centred approach as ‘a way of being’ with broad applicability in education, peacemaking, politics and social change (Rogers, 1980b). He was concerned about vital relational aspects such as empathy being misunderstood and reduced to techniques like reflective listening (Rogers, 1980a).

Our own journey with MI has been similar. We early discovered the hazards of teaching techniques that could be used on or done to people, and soon we were emphasizing the mindset and heartset with which it is practised for and with people. Our understanding of evoking has expanded to empowerment – helping people to realize and use what they already have, their own wisdom and strengths. As discussed earlier, MI appears to be applicable far beyond the world of counselling and psychotherapy. We hope that MI will help to further humanize services for those who receive and those who provide them (Douaihy and Driscoll, 2018; Gordon and Edwards, 1997; Rakel, 2018; Roberts, 2001).

From such a soaring view, MI might be misunderstood as an abstract aspirational philosophy. It never has been. Like the work of Carl Rogers, it has from the very beginning been rooted in the close observation of practice and the testing of specific hypotheses that can be replicated by others. Good practice of MI is specifiable, and at least in psychotherapy research, well-trained observers can reliably agree on the quality of MI being delivered within a particular sample. Over a thousand clinical trials demonstrate its beneficial impact.

Acquiring MI skillfulness does involve practising observable behaviours that predict positive outcomes, and one could likewise misunderstand MI as nothing more than these specific responses. I believe, though, that there is a deeper level of learning that comes with the practice of a person-centred way of helping such as MI. It is as though practising the skills teaches you the relational spirit. Reflective listening brings a deeper understanding of others’ inner worlds. Affirmations become sincere appreciation of people’s strengths. Practising acceptance fosters a calm presence that is at peace with one’s own and others’ shortcomings. Such changes are less readily observable than specific practice behaviours, and perhaps we can be at peace with that as well while MI continues to evolve in unanticipated ways.

Some reflections for clinical science

What scientific lessons might be learned from the evolution of MI? One is surely the value of directly observing and learning from clinical practice. Indeed, science begins with careful observation. The major scientific contributions of Carl Rogers to psychotherapy did not begin from a theory but from paying close attention to clinical practice, moving through abductive reasoning toward provisional hypotheses (Kirschenbaum, 2009; Miller and Moyers, 2017). MI similarly emerged from observing and asking testable questions about clinical experience. This is the empirical context of discovery that informs the twin context of justification in science (Reichenbach, 1938), serving as a counterpoint to the hypothetico-deductive approach and binary significance testing that are often more highly prized in psychology.

It can be valuable, then, to bring a beginner’s mind to the observation of clinical practice, seeking fresh perspectives and hypotheses that may be missed when viewing data through a priori lenses. Be open and curious rather than defensive when your predictions are not confirmed. Unexpected findings have been and will continue to be important in understanding MI.

Don’t stray too far from direct observation of practice. Developing reliable systems for observing therapy facilitated learning both in MI and in the client-centred approach from which it evolved (Truax and Carkhuff, 1967). My first impression was that coding MI interviews would be deadly dull, but in fact it is a way to see therapeutic processes unfolding in real time. Process coding of interviews seems to be a boon in developing skillfulness in MI. Deliberate practice with direct observation and shaping of skills can be used in training and supervision (Rousmaniere et al., 2017; Westra et al., 2021).

Run new ideas by colleagues who differ in cultural background and professional training. MI first emerged from the questions and observations of Norwegian practitioners. Carl Rogers suggested circulating novel insights in a ‘discussion paper’ to invite comments, which is how the first published description of MI happened (Miller, 1983). MI has been enriched by the insights of colleagues with expertise in anthropology (Carr, in press), psycholinguistics (Amrhein, 2004), and evolutionary psychology (de Almeida Neto, 2017).

Introspection has a long history as an investigational method in psychology, particularly in the development of hypotheses and theory (Danziger, 1980; Weger et al., 2019). A ‘think-aloud’ method asks people to articulate their thoughts while engaged in a simulated situation, or while reviewing recorded sessions (Davison et al., 1997). That is precisely what my Norwegian colleagues asked me to do while demonstrating counselling, and it yielded new testable hypotheses that led to the development of MI. I recommend thinking aloud as a tool in the context of discovery when studying psychotherapy.

Finally, I can commend William of Occam’s parsimonious principle that entities should not be multiplied beyond necessity. There are hundreds of brand-name psychotherapies that are allegedly different and advantageous. Yet when compared directly their average outcomes are usually similar, with large differences depending on who provided the treatment. We are also fairly clueless in predicting who will benefit most from allegedly different psychotherapies. In writing our fourth edition of Motivational Interviewing we are seeking simplicity on the far side of complexity. Rather than continuing to proliferate jargon and brand names, perhaps we can pursue common and complementary principles of behaviour change.