Is Role Play an Effective Training Method?

The Question Buyers Actually Ask

Anyone budgeting for training has a fair question: does role play work, or does it just feel as if it should?

The honest answer has two halves, and most providers only give the first. Role play does something that slides and video do not. It also fails in predictable ways, and a buyer who knows them spends the budget better.

What the Evidence Supports

Sidestream's own evidence is described in full, limits included, on our research page. In 2020 co-founder Ben Laumann ran a randomised study at University College London. Fifty-seven participants learned the same feedback model, Situation, Behaviour, Impact, in one of three ways, each capped at 20 minutes: an online coaching session with a role-play coach, an instructional slideshow, or an instructional video.

Each then delivered a simulated feedback conversation, which was recorded and rated blind by two independent assessors. Rehearsal produced significantly better delivery than either passive format, with a medium-to-large effect against slides (Cohen's d = 0.75) and a large effect against video (d = 1.08).

Slides and video did not differ significantly from each other. In a different setting, a 2025 McGill randomised trial on a surgical simulator found that medical students coached by a human outperformed those trained by an AI tutor alone; it measured a manual skill, not a conversation.

Evidence on AI conversation partners is early and mixed: a 2026 pilot trial in clinical communication (Wu and colleagues, medRxiv, n = 58, a preprint not yet peer reviewed) reported skill gains with AI simulated patients that were non-inferior to practice with human simulated patients. It tested one skill in medical students, not leadership conversations under pressure.

Two things matter for a buyer. The outcome was a behaviour people performed, judged by assessors who did not know which group anyone was in, not a quiz or a feedback form. And swapping one passive format for another made no significant difference, so there is no evidence that replacing a slide deck with a video buys a better result. The learning experiences that use this method at scale are listed under our immersive events.

Two much larger reviews are consistent with that, though neither is a study of role play with actors. Taylor, Russ-Eft and Chan's meta-analytic review of 117 studies of behaviour modelling training (Journal of Applied Psychology, 2005), in which people watch a behaviour, practise it and get feedback on it, found that effects on declarative knowledge decayed over time while effects on skills and job behaviour remained stable or even increased, and that skill development was greatest when training time was longest.

Lacerenza and colleagues' meta-analysis of 335 independent samples (Journal of Applied Psychology, 2017) reviewed leadership training in general, found that it improved reactions, learning, transfer and results, and found practice, feedback and spaced sessions among the design features that made it work better. Rehearsal with feedback is the ingredient role play supplies. Where participants bring their own live situation instead of a script, we call it real play; the difference is explained in real play vs role play.

Why Confidence Is the Wrong Measure

The UCL study also asked participants to rate their own confidence on a scale of 1 to 7 three times: before the intervention, after it, and after the simulated conversation.

In the slideshow group the means were 4.46, 5.00 and 5.39. In the video group they were 4.69, 5.13 and 5.38. In the practice group they were 4.27, 5.33 and 5.53 (confidence ratings n = 44) (Laumann (2020), UCL, unpublished research paper, Figure 4). Confidence rose in every group, the modality had no significant influence on it (p = .958), and the interaction between time and modality was not significant either (p = .541).

So the slideshow and video groups left feeling more capable, by a margin not significantly different from that of the group whose delivery was rated significantly better, and any evaluation that asks how people feel at the end of the day reports a success for all three.

Role play is only an effective training method if somebody watches the behaviour, in the room and later at work, which is Kirkpatrick Level 3 rather than Level 1. The gap between felt and rated competence is the territory of the Dunning-Kruger effect, and a satisfaction form cannot see it.

Where Role Play Does Not Help

An honest answer starts with the limits of our own evidence. The UCL study compared three ways of teaching against each other, not against no training, because it had no control condition. It scored delivery once, immediately afterwards, so it says nothing about retention. It tested one skill in one setting. And its practice condition was an online coaching session of at most 20 minutes, not Sidestream's programme, so we do not present it as a measurement of what we deliver.

Beyond the study, there are at least five situations in which role play is the wrong tool.

• When the goal is information. Both passive groups in the UCL study learned the feedback model well enough to attempt the conversation. For a policy, a process or a set of facts, a document or a short video is cheaper and adequate.
• When the workplace punishes the new behaviour. Taylor, Russ-Eft and Chan found that transfer to the job was greatest when trainees' superiors were also trained and when rewards and sanctions were built into the work environment. A manager who rehearses a candid conversation and returns to a boss who prefers quiet is less likely to keep it up. The room can build the skill; it cannot make the organisation want it.
• When it runs once. Lacerenza and colleagues' moderator analyses supported spaced sessions. A single afternoon of role play is a memorable event; whether it holds depends on whether the behaviour is attempted again, with feedback, at intervals.
• When the other person is a colleague. Peers playing each other are sympathetic, have a working relationship to protect and are waiting for their own turn, so the exchange rarely becomes real enough to respond to. Sidestream uses professional actors rather than peers for this reason, as set out in role-play training with professional actors. The UCL practice condition used a role-play coach rather than a peer, but the study did not compare the two.
• When it is designed to expose rather than to practise. Role play that humiliates builds avoidance, not skill. The discomfort that teaches is the discomfort of effort, and separating the two is the subject of why role play feels uncomfortable.

What Effective Role Play Looks Like in Practice

The conditions the evidence points to are the ones Sidestream builds around. The other person in the scene is a professional actor working from a written brief, who holds the position when the participant gets it wrong. Scenarios are built from the organisation's own recurring conversations rather than a generic script.

Alongside the scripted scenes, a participant with a real conversation coming up can bring that one instead and brief the actor themselves, a variant we call real play. Each conversation is attempted, stopped, coached and attempted again, because the first attempt shows the default and only the second shows whether something else is available under pressure.

A typical format is a half-day of four hours with 5 to 30 participants and one or two scripted scenarios; the new manager programme runs over two days, and the same rehearsal can be the working part of a corporate away day, with venue, lunch and social programme left to the client.

Work is priced per engagement, because the effort sits in the scenario design and the casting. Where the target is collective (conduct, inclusion or speak-up), forum theatre puts the whole room in the scene; where it is one person's capability in one exchange, individual role play with an actor does the work.

We look at what people do, in the room and in the follow-up simulation, not what they tick on a form. An optional follow-up simulation around 90 days after delivery gives each participant a fresh scenario with a coach they have not met and personal feedback, as described for conflict resolution training; how an organisation can measure for itself, baseline first and re-read the same indicators afterwards, is set out in how to measure behaviour change.

The wider design sits on the COM-B model, which is how a behavioural target becomes a scenario, and the method as a whole is set out in our approach.

How to Judge Any Provider's Answer

Whoever you ask, five questions separate an evidence-based answer from a confident one, and a provider who cannot answer the last of them is selling role play as a cure rather than a method.

• Who plays the other person, and are they able to hold the role under pressure?
• Is each scenario attempted more than once, with feedback in between?
• Are the scenarios built from conversations your people actually have?
• What is measured afterwards, and is it observed behaviour or a satisfaction score?
• What does the provider say the method does not do?

Is Role Play an Effective Training Method? In Short

For behaviour under pressure, yes. In a randomised UCL study in 2020 with 57 participants, those who rehearsed with a role-play coach delivered feedback rated significantly better, blind, than those given slides or a video (d = 0.75 and d = 1.08), and the two passive formats did not differ significantly from each other. Two meta-analyses in the Journal of Applied Psychology point the same way: the skills built by behaviour modelling training held or grew over time, and leadership training worked better with practice, feedback and spaced sessions.

Role play is the wrong tool for conveying information, it fails where the workplace punishes the new behaviour or where it runs once, and it cannot be judged by confidence scores, which rose in every condition of the UCL study regardless of performance. Rehearse the behaviour, with someone who can hold the other side, and watch what people do afterwards. Bring the conversation your managers avoid to a free 30-minute call and we will say honestly whether rehearsing it is the right tool. Get in touch today. We are Sidestream.

Frequently Asked Questions

Is role play more effective than e-learning or video?

For behaviour, on the evidence we have, yes. In the randomised UCL study behind Sidestream's method, rehearsal with a role-play coach produced significantly better blind-rated feedback delivery than a slideshow or a video, and the two passive formats did not differ significantly from each other; the design and effect sizes are on our research page.

Is one role-play session enough?

Rarely. Lacerenza and colleagues' 2017 meta-analysis of 335 samples found spaced training sessions among the design features that made leadership training more effective, and Taylor, Russ-Eft and Chan's 2005 review found skill development greatest when training time was longest. A single session shows people their default and gives them one rehearsal of an alternative. Whether it holds depends on the behaviour being attempted again, with feedback, after the session. That is why Sidestream's manager programme runs over two days rather than one, and why the optional follow-up simulation comes around ninety days later rather than on the day.

Does role play work for every skill?

No. It is built for behaviour performed under pressure: a difficult conversation, a challenge, a decision with someone pushing back. For a policy or a process, a document or a short video is adequate and cheaper. The UCL study tested one skill in one setting, so it does not establish that rehearsal beats passive formats for every skill, in every population, at every scale.

How do you see whether the behaviour has stuck?

We look at what people do, in the room and in the follow-up simulation, not what they tick on a form. An optional follow-up simulation around 90 days after delivery: each participant books an hour over Zoom, steps into a fresh scenario with a coach they have not met, built around the situations practised in the room, and gets personal feedback. It is a chance to practise again and to see how much has stuck, not an assessment. Confidence scores are not used as evidence, because in the UCL study they rose in every condition, with no significant difference between them, even though blind-rated delivery differed significantly.

Does role play have to use professional actors?

The UCL study does not settle that question: its practice condition was a coaching session with a role-play coach, and it did not compare actors with peers. Sidestream uses professional actors because peers playing each other are sympathetic, have a working relationship to protect and are waiting for their own turn, so the exchange rarely becomes real enough to respond to. An actor working from a brief holds the position when the participant gets it wrong.