What the experiment showed
Rohrer and Taylor taught students to compute the volumes of four solids and then gave them practice problems either blocked by solid type or shuffled together. During practice the blocked group performed far better, which is what makes blocking so appealing. One week later the shuffled group substantially outperformed them on a test. The mechanism is the part worth understanding: when problems are blocked you already know which formula applies before you read the question, so you never practise the step of working out which method the problem needs. That step is precisely what an exam demands.
It replicates in a real classroom
The 2007 study was a laboratory result, and laboratory results in education frequently thin out in practice. Rohrer, Dedrick and Stershic ran interleaved practice into an actual middle-school mathematics course over several months and found the advantage held on a delayed unseen test. That is a meaningfully stronger claim than the original, because it survived homework, absences, and everything else that makes real teaching different from an experiment.
When it helps and when it does not
Interleaving helps most where problem types are confusable and the difficulty lies in discrimination — mathematics, physics, chemistry, statistics, any subject where the question does not announce its method. It helps less, and can hinder, when you are learning a genuinely new procedure for the first time and have not yet got it working at all. The reasonable sequence is to block briefly while acquiring a method, then interleave it with everything else you know. Expect the mixed practice to feel worse and your accuracy during practice to drop; that is the same fluency illusion that makes rereading feel productive.