The original article can be viewed online (pdf) starting from here.

As I had presumed the main effect of treatment (predicting chaos vs. predicting a random control condition) was not significant in the article. Therefore, the chaos group did not significantly better than the control. However, since the chaos group's results were getting worse over days and the control group's results were getting slightly better over time, the linear component of the interaction was significant. The statistical analysis is full of numerical corrections about which not enough details are given to allow scrutiny.

Of this particular article and its results we will not hear more in the future.

Wolfgang