A learner can drill the present perfect for forty minutes, get all twenty items right, and misplace it in conversation three days later. The smooth feeling during a blocked drill is real, and it is a weak predictor of what survives. Mixing several kinds of practice in one sitting feels slower and produces more errors in the moment, yet it predicts better recall a week later. The claim has a boundary, and the boundary is the useful part: mixing helps when a learner must choose among similar forms, tenses, or rules, and it adds little when the items share nothing to confuse.
Why blocked practice feels so good
Blocked practice is the default in most self-study. A student drills conditionals on Monday, relative clauses on Tuesday, and collocations on Wednesday. Each session feels productive because the answers arrive quickly. The problem is that the mind never has to decide anything. When every item in a set follows the same rule, the learner can answer by pattern recognition alone. The rule never has to be selected from among alternatives, and selecting is exactly what a real sentence demands.
What three studies found
Rohrer and Taylor tested this with four kinds of mathematics problems. College students practised either in blocks, one problem type at a time, or in a shuffled order that mixed all four types. On a test given one week later, the shuffled group did better, even though the blocked group had been more accurate during practice itself. The 2007 study in Instructional Science names the feeling most learners recognise: blocked practice feels as if it is working, and the feeling misleads.
Categorisation shows the same pattern, and it sits closer to how a grammar rule behaves. Kornell and Bjork had participants study paintings by different artists. Some saw each artist's work grouped together. Others saw the artists interleaved. Asked to classify new paintings by artist, the interleaved group did better, even though they rated the grouped format as more effective for learning. The 2008 paper in Psychological Science expected spacing to hurt induction. It helped.
Verb conjugation gives the cleanest language result. Pan and colleagues taught adults to conjugate Spanish verbs in the preterite and imperfect past tenses across four experiments. In the first two, with a single training session, interleaving was no better than blocked practice, and in one case numerically worse on a two-day delayed test. The advantage appeared in experiments three and four, when training spread across two weekly sessions: the blocked group worked one tense per session, the interleaved group worked both. On a one-week delayed test, interleaving won by a wide margin. The 2019 study in Journal of Educational Psychology is the useful reference, because it states the condition under which the effect appears: spaced sessions, mixed item types, and a delayed test.
Why it fails when applied badly
That detail explains why the technique flops when people apply it badly. Mixing needs more than one session and a reason to discriminate. Added to a single cramming session, it produces little. If a learner shuffles fifteen unrelated words from fifteen topics, there is little shared structure to confuse, and the shuffling becomes noise. The benefit arrives when the choices look alike: for against since, the preterite against the imperfect, make against do, countable against uncountable nouns.
A week's routine
A workable routine for a Taiwanese learner runs like this. Take three grammar points that keep colliding, say the present perfect, the past simple, and for against since. Write twelve sentences that mix all three rather than thirty that drill one. Mark the answers, sort the errors by which point broke, and repeat a fresh mixed set two days later. Vocabulary follows the same logic: study make and do in one pass with their usual partners, so the mind must pick a verb each time. For the scheduling side of the same idea, read our article on letting a word fade before reviewing it, which treats spacing as its own lever.
Mixed practice has an honest cost: it feels worse. Learners rate blocked practice as more effective even after their own test scores show the opposite, a result that appears in both the Kornell and Bjork and the Rohrer and Taylor work. A teacher should say this aloud: the confusion today is the signal that the choice is being practised. A 27-year-old engineer preparing for TOEIC does not need forty minutes of one tense. She needs twenty minutes where every item forces a decision between two tenses she keeps merging.
One counterexample matters. A beginner who has just met a form for the first time gains little from mixing, because there is nothing to choose among yet. Mixing is for the stage after introduction, when forms exist in memory and keep colliding in use. A learner who mixes before that stage gets confusion without the payoff. The order is introduce, then mix, then test after a delay. Most materials stop after the first step.
For tomorrow, take the three forms you swap most often, write eight sentences that require all three, and complete them in one mixed pass. Grade them the same day. Two days later, do a second mixed set with different sentences and compare the error count. The number that matters is the second set, because that is the one that predicts what you keep.