Why Mixed Practice Feels Worse but Works Better: The Painting-Style Studies

Mixed practice feels worse than grouped practice, but in the painting-style experiments it produced better results, and most learners did not notice. In a replication of a classic study, 75% of mixed-study paintings were recognized correctly versus 66% of grouped ones. Yet 71% of participants said grouping had helped them learn more.

This is the same effect covered in interleaved practice for math, seen in a different task, and it explains why many students quit mixing too early.

1. What the painting studies tested

The experiments were designed by Nate Kornell and Robert Bjork in 2008, as described in a 2012 review by psychologist Doug Rohrer. College students learned to tell artists apart by viewing landscape paintings, each shown with the artist's name. In one version, students saw six paintings by each artist. The "blocked" group saw all six by Artist A, then all six by Artist B, and so on. The "interleaved" group saw one painting by A, one by B, and cycled through the artists six times.

Afterward, everyone saw new paintings and had to pick the artist. Rohrer reports interleaving improved test scores, 59% versus 36% (Cohen's d = 1.28). That is 23 percentage points, or about 1.6 times as many correct answers (59 / 36 = 1.64).

The task is called category induction. Students cannot memorize the study paintings, because the test paintings are new. They must pick up what makes each artist's style distinctive.

2. The replication

Kornell and Bjork's second experiment used a recognition test: new paintings by studied artists mixed with paintings by artists never studied. A 2014 paper by Verkoeijen and Bouwmeester repeated it with 123 online participants. The results, from the first block of the test:

• Mixed ("spaced") paintings: 75% recognized (SD 0.21).
• Grouped ("massed") paintings: 66% recognized (SD 0.22).
• Difference: 9 percentage points, Cohen's d = 0.37, p = 0.00005.
• False alarms were similar (50% vs 48%), so the gain was in hits on studied paintings, not in a general tendency to say "seen it" (the paper reports no significant difference between the two kinds of distractors).

The original study had found 77% versus 67%, a 10-point gap with d = 0.41. The two experiments combined gave an effect size of 0.39 (95% confidence interval 0.24 to 0.53). The authors call the replication successful.

3. The part that surprises people

After the test, Verkoeijen and Bouwmeester asked which method helped most. Of 123 participants:

• 87 (71%) said grouped study helped most.
• 21 (17%) said both were equal.
• 15 (12%) said mixed study helped most.

So only about one learner in eight judged correctly. In the original study, the authors report that 80% had said grouping was better. Rohrer cites a related result: among participants who actually benefited from interleaving, only 25% believed it was more helpful.

A reasonable reading, which Rohrer states directly, is that grouped practice "provides scaffolding." When the same artist appears six times in a row, the task is easier and you feel fluent. Mixed practice makes more errors and goes slower, so it feels like you are learning less. Rohrer calls the false sense of understanding an "illusion of knowing."

4. Is it just spacing in disguise?

Mixing artists also spaces out each artist's paintings. Rohrer addresses this: Kang and Pashler (2012) used painting identification with three artists and compared an interleaved order against a blocked order where each painting was followed by an unrelated cartoon, so the time gap between paintings by the same artist was equal. Interleaving still won, 68% versus 61% (d = 0.78). That is a 7-point gap, smaller than the 23 points above, but it was not caused by spacing alone.

Rohrer's most plausible explanation is comparison. Seeing a painting by one artist next to a painting by another lets you notice which features separate them. Fittingly, he also reports that showing two birds from different families side by side helped bird identification.

5. How to use this

1. Use mixing when categories look alike. Rohrer says the benefit is probably limited to difficult discriminations. Artist styles, bird families, easily confused terms such as mitosis and meiosis, or math problem types that need different strategies are good fits. For dissimilar material, he expects smaller effects.
2. Put different categories next to each other. The comparison seems to matter.
3. Do not judge by how it feels. In both studies learners' impressions pointed the opposite way from their test results, so feeling less fluent while mixing does not mean it is working worse.
4. Try a hybrid. Rohrer notes that studies of combining a short block with later mixed practice are missing, but suggests it may be the most acceptable approach: learn a new concept in a small block, then mix it with earlier material.

6. Limits

Rohrer cautions that most of this work is laboratory research. Many studies had one learning session and a test less than an hour later, so he calls for classroom studies before recommending interleaving without qualification. The replication used an online sample of adults rather than students, and its exploratory data showed the advantage varied across test blocks, with the mixed condition doing worse in blocks 2 and 4. The first-block result is the one used for the main comparison, following the original study.

7. Key takeaways

• Mixing artists' paintings beat grouping them, 59% vs 36% in one study and 75% vs 66% in a replication.
• Interleaving still helped when spacing was held equal (68% vs 61%).
• About 71% of learners in the replication believed grouped study worked better, so your own impression is a poor guide.
• The effect appears strongest for hard-to-tell-apart categories.