{"id":6678409,"date":"2026-09-06T08:00:00","date_gmt":"2026-09-06T13:00:00","guid":{"rendered":"https:\/\/www.learningandthebrain.com\/blog\/?p=6678409"},"modified":"2026-08-11T12:11:14","modified_gmt":"2026-08-11T17:11:14","slug":"retrieval-and-generation-does-the-order-matter","status":"publish","type":"post","link":"https:\/\/www.learningandthebrain.com\/blog\/retrieval-and-generation-does-the-order-matter\/","title":{"rendered":"Retrieval and Generation: Does the Order Matter?"},"content":{"rendered":"\n

For almost three decades now, cognitive scientists — led by Robert Bjork and Elizabeth Ligon Bjork — have championed desirable difficulties<\/strong>. The simple definition: any study strategy that<\/p>\n\n\n\n

    \n
  1. Requires students to think harder<\/em>, and<\/li>\n\n\n\n
  2. Ultimately results in success. (For this reason, a difficulty that overloads working memory would certainly be undesirable<\/em>.)<\/li>\n<\/ol>\n\n\n\n

    For obvious reasons, retrieval practice<\/strong> fits comfortably within this category. After all:<\/p>\n\n\n\n

      \n
    1. Students have to think harder when they retrieve than when they review, and<\/li>\n\n\n\n
    2. In an impressive body of research, students remember more after retrieval practice than after rereading notes or textbook pages.<\/li>\n<\/ol>\n\n\n\n

      So too, a well-designed generative learning<\/strong> activity could be desirably difficult. When students summarize or draw or re-enact or explain, they have to think harder…and typically end up learning more.<\/p>\n\n\n\n

      Because both of these desirable difficulties help students learn, we can reasonably ask: does the order matter<\/em>? Should students retrieve before they generate? Generate before they retrieve? Or, perhaps either approach works equally well? Perhaps the answers to those questions will be different depending on the timing<\/em> of the retrieval and\/or generation?<\/p>\n\n\n\n

      A recent study<\/a> looks at this very question.<\/p>\n\n\n\n

      Intro to Cognitive Psychology<\/h2>\n\n\n\n

      In this study, roughly 200 German university students read a short passage defining four concepts from social psychology and cognitive psychology: “the availability heuristic,” “the mere exposure effect,” and so forth.<\/p>\n\n\n\n

      One group of students first generated<\/strong> examples of these concepts — they could look at the text if they wanted — and then revised their examples based on feedback. Next they did three cycles of retrieval <\/strong>exercises.<\/p>\n\n\n\n

      A second group reversed that order: they started with that substantial retrieval exercise, and only then did the multi-step generation task.<\/p>\n\n\n\n

      (I’m leaving out another group, because that third variable adds more complexity than a blog post warrants.)<\/p>\n\n\n\n

      Helpfully, researchers had two other groups follow the same steps — except that they did the retrieving and generating two days after <\/em>the initial reading.<\/p>\n\n\n\n

      And, fully a week later, the students came back for a final test to see if they could recall <\/em>the definitions and apply <\/em>them to new situations.<\/p>\n\n\n\n

      \"\"<\/figure>\n\n\n\n

      Given all these steps, researchers can start to answer these questions:<\/p>\n\n\n\n

      A. Does the order matter? Should retrieval come before generation, or vice versa?<\/p>\n\n\n\n

      B. Does the timing matter? Is one order better when retrieval\/generation happen right away, while the other order is better after two days?<\/p>\n\n\n\n

      Those are intriguing questions, and we really want to have the answers…<\/p>\n\n\n\n

      The (Initial) Answers<\/h2>\n\n\n\n

      “No,” and “no.”<\/p>\n\n\n\n

      That is: no, the order didn’t ultimately matter. Students who retrieved before generating<\/em> scored as high on the post-tests as those who generated before retrieving<\/em>.<\/p>\n\n\n\n

      And: no, the two-day delay didn’t change that result. Whether retrieval preceded generation or generation preceded retrieval, both groups scored roughly the same on the test they took one week later.<\/p>\n\n\n\n

      By the way: researchers pre-tested the students to see if they already knew about “the availability heuristic” and the other terms. Students all scored very low on this pre-test — so “prior knowledge” probably didn’t drive these results.<\/p>\n\n\n\n

      Tentative Questions<\/h2>\n\n\n\n

      On the one hand, this study starts to answer an important question. I frequently discuss both retrieval practice and generative learning with teachers, so it’s helpful to have a research-based answer about the appropriate order.<\/p>\n\n\n\n

      On the other hand, I have two reasons to wonder about this study’s conclusions.<\/p>\n\n\n\n

      First<\/strong>: as noted above, researchers pre-tested the students to see what they knew about those four terms. (Research methodology indeed encourages such pre-tests.) <\/p>\n\n\n\n

      However — and this is a big however — we have a growing pool of research<\/a> suggesting that pretesting itself<\/strong> enhances learning. In other words, the (methodologically appropriate) pretests might themselves have helped students learn all those definitions. The stated results — “the order doesn’t matter” — could be true after <\/em>pretesting, but not necessarily without <\/em>pretesting.<\/p>\n\n\n\n

      Second<\/strong>: <\/p>\n\n\n\n

      More hypothetically, I wonder about a potential distinction between two kinds of generation:<\/p>\n\n\n\n