We contrast two types of experiments: pull-out experiments, in which students are pulled out of their normal educational environment, and curriculum-embedded experiments, in which student experience experimental conditions as a normal consequence of proceeding through a curriculum.
We argue that, for practical reasons, curriculum-embedded experiments are preferred for large-scale experimentation in schools, and such experiments may also avoid issues with ecological validity.
The choice to use a curriculum-embedded approach affects methods used for randomization, subject screening and the way that student experience after the experiment concludes is affected by the experiment, even for students who did not themselves participate.