The evidence base, stated honestly.
We would rather state the limits ourselves than be caught by them. Here is what the evidence supports, what is contested, and what we no longer use.
What is well evidenced.
-
Well evidenced
Doing beats listening, at scale, in the most conservative subjects there are. Examination performance rose by 0.47 standard deviations, and students under traditional lecturing were about 1.5 times more likely to fail.
Freeman et al. 2014, PNAS, a meta analysis of 225 studies in undergraduate science, engineering and mathematics.
-
Well evidenced
Generating beats receiving. A learner who produces the answer retains more than a learner who is given it.
Bertsch et al. 2007, Memory and Cognition, 86 studies. McCurdy et al. 2020, Psychonomic Bulletin and Review, 126 articles and 310 experiments. A 2023 multilevel meta analysis in Educational Psychology Review puts the benefit at g = 0.41 for learning from texts.
-
Well evidenced
Enacting beats reading. Material that is performed is remembered better than material that is only read.
Roberts, MacLeod and Fernandes 2022, Psychological Bulletin.
-
Well evidenced
The body is part of the thinking, not the vehicle carrying it. Children told to gesture during a lesson learn more, and a learner's gestures predict understanding before their speech shows it.
Goldin-Meadow 2011, WIREs Cognitive Science.
-
Well evidenced
Psychological safety is real, measurable and tied to learning behaviour. This is our headline citation, ahead of Google's Project Aristotle, which is powerful in a room but was never peer reviewed.
Frazier et al. 2017, Personnel Psychology, a meta analysis of 136 samples covering more than 22,000 individuals.
-
Well evidenced
Singing together bonds strangers faster than the alternatives. Across seven months of adult education classes, singing groups bonded significantly faster than creative writing or craft groups. The comparison group is the point: this was singing against other shared activity, not singing against nothing.
Pearce, Launay and Dunbar 2015, Royal Society Open Science.
-
Well evidenced
The bonding effect survives scale. It holds in large groups as well as small ones.
Weinstein et al. 2016, Evolution and Human Behavior, community choirs of 20 to 80 alongside a 232 person choir.
-
Well evidenced
Participatory music making moves biology, not only mood.
Fancourt, Perkins and colleagues 2016, PLOS ONE.
-
Well evidenced
Art making lowers a stress marker regardless of skill. Forty five minutes of art making produced a significant drop in salivary cortisol in around 75 percent of participants.
Kaimal et al. 2016, Drexel.
-
Well evidenced
Improvisation training raises tolerance of uncertainty. Note the honesty of the finding: the gain in uncertainty tolerance was robust, the gain in divergent thinking was partial.
Felsman, Gunawardena and Seifert 2020, Thinking Skills and Creativity.
-
Well evidenced
Recovery is a legitimate category of its own, separate from learning.
Sonnentag and Fritz 2007, and a 2022 meta analysis covering 316 samples and 99,329 participants.
-
Well evidenced
Arts based leadership development has a body of evidence and knows its own weakness. Its own conclusion is that the field needs longitudinal and quantitative work.
Behavioral Sciences 2024, a metasummary of 31 empirical studies.
What we have retired.
Earlier Artfulness material used the following. It no longer does, and we would rather say so on this page than have it found.
-
Retired
Polyvagal theory. Earlier material leaned on it for the safety claim. It is now seriously contested in the physiological literature, and nothing here rests on it any more.
-
Contested
Synchrony causing cooperation. The famous study has failed replication under blinded conditions, and the meta analyses find real but small to moderate effects with unclear moderators. We use synchrony to describe what happens in the room, not as proof of what follows from it.
-
Never used
Learning styles. Never used, and often confused with the experiential cycle, which we keep as lineage rather than as proof.
What we do not claim.
The full list of eight limits, in our words, is on one page so that it can be quoted whole.
That a session raises revenue, innovation output or retention on its own.
That one day permanently improves anyone's cognition. The defensible claim for a day is that it is a concentrated workout of attention, working memory and error monitoring, not a lasting change in them.
That this is therapy, or treatment for anyone who is unwell. It is built for the large group who are neither ill nor thriving.
That a day fixes an organisation whose problems are structural. Workload, job design and management practice are out of reach of a song. What the day restores is a capacity in the people, which is a smaller and different claim.
What we measure.
| Measure | How and when |
|---|---|
| Psychological safety | The seven item team scale, anonymous, before the session and again at thirty days. It is free and takes four minutes. |
| Named behaviours | Rated by managers or peers rather than by us, thirty to sixty days later. |
| Salivary cortisol | Before and after, when a client wants the stress claim read physiologically. |
| Thirty day application | Rather than same day recall, because an end of day feedback form measures the wrong thing. |
Attendance should be invited rather than mandated. An activity whose mechanism is voluntary engagement is damaged by compulsion.
The fastest way to understand this is to be in a room where it happens.
Eight people, one room, no preparation.