Testing as learning

Testing can be learning. Across multiple meta-analyses, trying to retrieve studied material produced reliably better later retention than spending equivalent time restudying it. The advantage appears in laboratory research and in a large body of classroom studies.¹² This is one of the strong foundations of memory research, so it does not need to be described as merely promising. That conclusion guides practice.

The figure shows why retrieval counts as learning: initial testing without feedback produced better delayed recall than repeated study after two days and one week.³

<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 590 320" role="img" aria-labelledby="testing-title" style="font-family: system-ui, sans-serif">
  <title id="testing-title">Recall of prose passages by college students after repeated study or initial testing without feedback</title>
  <line x1="70" y1="250" x2="555" y2="250" stroke="currentColor" stroke-width="2"/>
  <text x="20" y="145" font-size="16" transform="rotate(-90 20 145)">Free-recall performance</text>

  <rect x="325" y="20" width="18" height="18" fill="currentColor" opacity="0.25"/>
  <text x="351" y="35" font-size="14">Repeated study</text>
  <rect x="455" y="20" width="18" height="18" fill="none" stroke="currentColor" stroke-width="3"/>
  <text x="481" y="35" font-size="14">Initial testing</text>

  <rect x="105" y="88" width="42" height="162" fill="currentColor" opacity="0.25"/>
  <rect x="151" y="100" width="42" height="150" fill="none" stroke="currentColor" stroke-width="3"/>
  <text x="111" y="78" font-size="14">81%</text>
  <text x="157" y="90" font-size="14">75%</text>
  <text x="107" y="277" font-size="15">Five minutes</text>

  <rect x="271" y="142" width="42" height="108" fill="currentColor" opacity="0.25"/>
  <rect x="317" y="114" width="42" height="136" fill="none" stroke="currentColor" stroke-width="3"/>
  <text x="277" y="132" font-size="14">54%</text>
  <text x="323" y="104" font-size="14">68%</text>
  <text x="286" y="277" font-size="15">Two days</text>

  <rect x="437" y="166" width="42" height="84" fill="currentColor" opacity="0.25"/>
  <rect x="483" y="138" width="42" height="112" fill="none" stroke="currentColor" stroke-width="3"/>
  <text x="443" y="156" font-size="14">42%</text>
  <text x="489" y="128" font-size="14">56%</text>
  <text x="451" y="277" font-size="15">One week</text>

  <text x="161" y="307" font-size="14">Delay before the free-recall test</text>
</svg>

Figure: Testing lagged repeated study after five minutes, then led it after two days and one week. The reversal matters. Restudying can produce better performance after only a few minutes, while retrieval produces better retention after a meaningful delay. Meta-analysis likewise finds a larger testing benefit after delays of at least a day than after shorter delays. Tests requiring recall also produce larger benefits than tests requiring only recognition.¹

With a deck and twenty minutes, spend part of the session attempting answers before revealing them. Do not turn every prompt into another reading opportunity. A word-list experiment found that removing correctly recalled items from further study did not reduce one-week recall, but removing them from further retrieval did.⁴ The retrieval attempt was the consequential event.

Boundary. Testing does not guarantee better performance after only a few minutes, and test formats are not interchangeable. The evidence favors initial recall over recognition, but it does not establish that every difficult question is useful. The supported judgement is narrower: for later retention, genuine attempts to retrieve are generally more effective than equivalent restudy.