The question the study asked
Organisations spend heavily on communication training and mostly deliver it passively: a slide deck, a video, a module to complete. Educational research had long argued that active practice outperforms passive exposure, but the evidence for workplace communication specifically was thin, and almost none of it measured what people could actually do afterwards rather than what they said they had learned. Sidestream's Research Lead and Programme Coordinator is Patrick Smith.
The study set out to test one narrow, answerable version of that question. Take a single, well-defined communication skill. Teach it three different ways. Then have people perform it, and have the performance judged by assessors who do not know which group anyone was in.
The skill chosen was the SBI feedback model, short for Situation, Behaviour, Impact. It was picked because it decomposes into three observable components, which makes delivery assessable rather than a matter of taste: did the person name the specific situation, describe the observed behaviour, and state the impact?
How it was run
Fifty-seven participants were recruited through communications and coaching networks and through professional contacts, and assigned to one of three conditions in a one-way between-subjects design. Every participant received the same information about the SBI model. Only the delivery differed.
All three groups then completed the same task: a simulated feedback conversation with a fictitious colleague, based on a workplace scenario set during the 2020 shift to remote working. Each attempt was recorded.
The recordings were rated blind by two independent external assessors on a 1 to 7 scale for each of the three SBI components. Inter-rater reliability was established across 55 measurements before the scores were used. Group differences were tested with post hoc Tukey-HSD comparisons and a Bonferroni correction, so the result is not an artefact of running several comparisons at once.
- Practice. A live coaching session with a role-play coach, in which the participant rehearsed the conversation.
- Slideshow. An instructional slide deck covering the same content.
- Video. An instructional video covering the same content.
What it found
Participants who had rehearsed delivered significantly more effective feedback than either passive group. There was no significant difference between the slideshow and the video: the study found no evidence that swapping one passive format for another made a difference.
A different question was tested in a 2025 McGill randomised trial on a surgical simulator: medical students coached by a human educator using the AI's performance data learned better than those trained by the AI tutor alone. That trial measured a manual skill, not a conversation. Whether role play works beyond this one study is weighed in our evidence review on role play.
Effect sizes were substantial. Cohen's d was 1.08 between practice and video, which is a large effect, and 0.75 between practice and slideshow, which is medium to large. For context, a d of 0.8 is conventionally read as a large effect in behavioural research.
Rehearsal against the two passive formats, as Cohen's d. Randomised 2020 UCL study by co-founder Ben Laumann, n = 57: delivery of the SBI feedback model, rated blind by two independent assessors. No percentage is quoted because the study reports effect sizes, not percentages.
Source: Laumann, B. (2020), Active vs. passive L&D: differences between modalities for effective L&D in communicational competence, PSYCH0056, University College London, unpublished. Cohen's d = 1.08 (practice vs video) and 0.75 (practice vs slideshow); slideshow vs video not significant, p = .838.
Rehearsal scored higher than either passive format. The bars overlap: individual scores varied, and the difference lies in the group means. Slideshow and video did not differ significantly. The rehearsal condition was an online coaching session of at most 20 minutes with a role-play coach, so the study tests the principle behind the method, not today's programme.
Source: Laumann, B. (2020), Active vs. passive L&D: differences between modalities for effective L&D in communicational competence, PSYCH0056, University College London, unpublished. Randomised, n = 57. Cohen's d = 1.08 (practice vs video) and 0.75 (practice vs slideshow), Tukey-HSD with Bonferroni correction; slideshow vs video not significant, p = .838.
Show the values as a table
| Condition | Mean score (1 to 7) | SD |
|---|---|---|
| Practice with a role-play coach | 5.25 | 1.15 |
| Instructional slideshow | 4.26 | 1.31 |
| Instructional video | 3.82 | 1.12 |
The finding that mattered more
Participants also rated their own confidence, before the intervention, after it, and after performing the task. Confidence rose significantly in all three groups, regardless of which modality they had received and regardless of how they actually performed.
That is the uncomfortable result, and it is the one with the most direct commercial consequence. After a slide deck, people felt more capable, but they did not perform as well as those who had rehearsed. Any evaluation that asks participants how confident or satisfied they feel will report a success in every condition, including the two that produced significantly weaker delivery.
Confidence rose in all three groups. Only performance separated them.
Source: Laumann (2020), UCL, randomised, n = 57; confidence on a 1 to 7 scale at three points, performance rated blind by two assessors.
The study discusses this through the lens of the Dunning-Kruger effect: the weaker the underlying competence, the less reliable a person's estimate of it. It is the reason Sidestream measures at Kirkpatrick Level 3, observed behaviour, rather than by satisfaction scores.
What this study does not show
Being straight about the limits is part of citing it honestly. This is one study with 57 participants, conducted in 2020 for a taught module at UCL and supervised, not a peer-reviewed publication. It has not been independently replicated.
There was no control condition, and no measurement before the intervention. The study compares three ways of teaching against each other, not against no training at all, and it scored delivery once, in the final simulation. So it cannot tell you how much anyone improved, or whether the two passive formats improved anyone at all. It can tell you which of the three produced the better performance on the day, which is the question it set out to answer.
It tested one skill, the SBI feedback model, in one setting, a remote video call. It does not establish that rehearsal beats passive formats for every skill, in every population, at every scale. It measured delivery immediately after the intervention, so it says nothing about retention weeks later.
What it does establish, within those limits, is a clear and blind-rated difference on a task people actually performed, in the direction the wider educational literature predicts. That is the claim we make, and no more than that. How we corrected our own description of this study is set out in why we stopped publishing every day.
How it shaped the method
Three things in how Sidestream works come directly from this result. Participants perform rather than watch, because rehearsal produced significantly better delivery than either watching condition, and the study found no significant difference between the two watching conditions.
The other person in the scene is played by someone trained to hold the position, because the effect depends on the rehearsal being real enough to respond to. And outcomes are measured as observed behaviour, because self-reported confidence rose in every condition, with no significant difference between them, including in the two whose delivery was rated significantly weaker.
The wider design sits on the COM-B model from Michie, van Stralen and West, which is what turns a behavioural target into an intervention. This study is the evidence for the format; COM-B is the framework for choosing what to rehearse. More on both in our approach and in what immersive theatre training is.
The Study Behind the Method: In Short
A randomised study at UCL in 2020 with 57 participants tested three ways of teaching the same feedback model. Rehearsal with a role-play coach produced significantly better delivery than an instructional slideshow or video, rated blind by two independent assessors, with effect sizes of d = 0.75 and d = 1.08. Slideshow and video did not differ significantly from each other. Self-reported confidence rose in every condition regardless of performance, which is why Sidestream measures observed behaviour rather than satisfaction.
Frequently Asked Questions
Who conducted the study?
Ben Laumann, co-founder of Sidestream, in 2020, as a piece of research for the PSYCH0056 Business Psychology Seminars module at University College London, supervised by Dr Dimitrios Tsivrikos. Ben holds an MSc in Industrial, Organisational and Business Psychology from UCL and an MPhil in Innovation, Strategy and Organisation from the University of Cambridge, and is engaged in doctoral research in management and organisational behaviour at Bocconi.
How many people took part?
Fifty-seven participants, recruited through communications and coaching networks and professional contacts, assigned across three conditions. Inter-rater reliability for the blind assessment was established across 55 measurements.
What exactly was measured?
Delivery of the SBI feedback model in a simulated feedback conversation. Each recording was scored on a 1 to 7 scale for each of the three components, situation, behaviour and impact, by two independent external assessors who did not know which condition the participant had been in.
Is the study published or peer-reviewed?
No. It was written and supervised as university coursework, not submitted to a journal, and it has not been independently replicated. We cite it as what it is: a single randomised study with a blind-rated behavioural outcome, consistent with the wider educational literature on active practice.
Why do you not quote a percentage?
Because the study reports effect sizes, not percentages, and a percentage would depend on which comparison and which denominator you pick. Divide the difference between the two group means by either mean and you get roughly 19 to 23 per cent; other defensible denominators put it outside that range in both directions, and the gap against video is larger again. Cohen's d is the figure the analysis actually produced, so that is the figure we use.
Does this mean slides and video are useless?
No. They are efficient ways to convey information, and both groups did learn the model well enough to attempt the task. What they did not do is produce capability at the level rehearsal did, and they did not differ significantly from each other. If the goal is knowledge, passive formats are fine. If the goal is behaviour under pressure, they are not sufficient.
Source: Laumann, B. Active vs. passive L&D: differences between modalities for effective L&D in communicational competence. PSYCH0056 Business Psychology Seminars, University College London, supervised by Dr Dimitrios Tsivrikos, 2020. Unpublished. The full paper is available on request from info.sidestream@gmail.com.