[mattkrause]

That statement is not particularly precise and tacitly claims a lot of things. It’s best to think of it as shorthand for something like “When the test, which is thought to measure creativity, is run with music (vs. without music), the subjects’ performance decreases more than the sampling variability we’d expect when repeatedly running one of the conditions in the experiment over and over.”

This is essentially a Frequentist claim, which is a bit different from your Bayesian belief updating about music’s effects. If, when, and how you can from one to the other has been a major debate in statistics and science.