Why retrieval practice doesn’t automatically lead to transfer (and can sometimes get in the way)

If you have been reading this blog for a while or if you are familiar with my books, then you know that I am a big fan of retrieval practice. We’ve known for a while now that pupils and students generally learn more when they actively retrieve information from memory than when they study it over and over again. In combination with spaced practice, you have two of the best-researched learning strategies.

But… does this also mean that students can subsequently apply that knowledge better in new situations? In other words: does retrieval practice also help with transfer? I do think this is a relevant question. After all, we don’t just want students to remember what they’ve learned. We also want them to be able to do something with that knowledge.

A new study by Ramgopal and colleagues, published in Memory, provides an interesting answer to this. Or better and more frustrating: we get several answers. Because the results are a lot more complicated again.

Practice for a semester

The researchers followed 230 first-year psychology students at the University of Southampton. The participants received three practice sessions after their lectures, spread over several days, over a ten-week period. In these practice sessions all students experienced two study methods:

  • Retrieval practice: students answered multiple-choice questions. They ranked four answer options from most to least likely to be correct. Next, they received feedback on the correct answer.
  • Restudy: students were shown the same type of questions, but the correct answer was indicated immediately. This means they had to study the information again, without first retrieving it from memory.

To compare the two approaches, students received different questions! As the exercises were spread over several days, the researchers did not compare spaced with unspaced practice, but rather two learning strategies both offered in a spaced manner.

Three weeks after the last practice session, a final test followed with no fewer than 200 multiple-choice questions. Some questions were identical to the questions practised earlier (smart!). Others were new, but covered related material. Still other questions dealt with material that had not been covered during the practice sessions. 

Retrieval practice works, but…

First, the good news, although by now we can hardly call it surprising. For questions that students had practised previously, retrieval practice worked clearly better than restudying the material. Students were able to answer 80.2% of the questions previously retrieved accurately, compared with 70.1% of the questions they had restudied. This is a gap of approximately 10 percentage points, with an effect size of d = 0.58.

That is a substantial difference, especially since the final test took place three weeks after the last practice session. But… what happened when students were given related but different questions? Now, the difference disappeared almost completely! Students answered 62.4% of the related questions correctly after retrieval practice, compared to an equally high 62.3% after restudy. So, students retained the practised information better thanks to retrieval practice. At the same time, this advantage did not automatically translate into better performance on related questions.

Sometimes remembering well seems to be a bad thing?

The researchers went even a step further. They also examined which errors students made on the related questions. To understand why this matters, we need to briefly consider how those questions were structured.

Suppose that during practice you are asked who founded Structuralism. The correct answer is Wilhelm Wundt. Another possible answer is William James. Then, later, you are asked another question: who is considered the founder of Functionalism? Now, William James is the correct answer, whereas Wundt is one of the distractors. The answer options are therefore the same here, but the correct answer has changed.

And what did the results show? When students gave an incorrect answer to such a related question, they remarkably often chose the answer that was correct during practice. In retrieval practice, this occurred in 62.1% of the errors, compared to 55.4% after restudy. For completely new questions, where no previously correct answer had been practised, that same answer option was chosen in 47.8% of the errors.

Perhaps the previously learned answer had become so familiar here that students sometimes selected it again, even when it no longer fit the question. One could even almost argue that they had memorised the answer a little too well.

But there is another side to the story. When the researchers tried to account for these specific errors in additional analyses, they also found evidence of positive transfer. Actively retrieving knowledge thus seemed to sometimes indeed help students with related questions.

Therefore, the researchers suspect that two processes may be at play here simultaneously. On the one hand, retrieval practice can ensure that students retrieve more relevant knowledge and make connections. On the other hand, familiarity with a previously correct answer can mislead them.

Those two effects could cancel each other out. However, an important nuance is needed here. These latter analyses were exploratory and not preregistered, unlike the main analyses described above. Moreover, ruling out certain errors retrospectively is not conclusive proof of the underlying mechanism! It is therefore a possible explanation for the results, but not a definitively proven process.

And there was other good news

Not only was performance measured, but the participants were also asked how they felt about the practice. On average, students said they felt more motivated, had more mastery over the content, and had better concentration with retrieval practice.

It is very interesting, as retrieving knowledge is harder work than simply re-reading the information. Thus, learning something more difficult does not necessarily need to be less fun.

But we should be careful here too. These were self-reported outcomes, and the evidence for reduced anxiety was less convincing. For instance, the authors have not shown any improvement in test anxiety reduction due to retrieval practice in the long run.

Team nuance

Like any research, this research too has its own limitations. The authors refer to transfer, but have studied only one particular type of it. The new questions were related to the same subject matter, involved the same answers, but had a different correct answer.

This is not the same as when a student uses a mathematical concept in a totally new situation or uses a historical concept to understand a new situation.

Leave a Reply