Study after study in cognitive psychology demonstrates the power of memory retrieval: We tend to have better memory for information that we have tried to recall (or retrieve) [1], even if that recall is unsuccessful [2]. This effect has been observed across varied learning scenarios and content types; it even generalizes from the lab to the real world [3,4]. One can easily envision how such a robust phenomenon, known as the “testing effect,” could be harnessed by educators to boost learning in the classroom [5].

But how does this happen, and why? Retrieval of new memories is supported by a part of the brain known as the hippocampus [6], a structure made famous for its critical role in memory formation [7]. Practicing retrieval may be advantageous because it strengthens the connections among brain regions representing memory contents [9,11,12]. For instance, remembering “where” would strengthen the connections among regions used for perceiving spatial information. Reactivation of the same circuit—and therefore, a subjective experience of remembering “where”—is then easier at a later point in time. This strengthening of connections across the brain may help the memory more quickly stabilize and become independent of the hippocampus (i.e., consolidate).

However, recent perspectives from neuroscience and psychology propose that strengthening is just one of the many outcomes that can follow memory retrieval. More generally, retrieval represents an opportunity for memory modification. One example of such modification has been termed “integration,” in which relevant memories are retrieved and updated to account for new information in the environment [13]. The integration process may be critical for connecting related experiences that occur across different days or learning environments. The result? Memories that are flexible, that anticipate novel situations, and that can help us accomplish a host of complex cognitive tasks [14,15]—from imagining new futures to reasoning about the world.

The neural underpinnings of integration

Empirical research shows that, across species, new information that can be connected to existing knowledge is “special.” It is behaviorally privileged: It can be learned more readily and stabilized more quickly [16]. It is also treated differently by the brain: It engages a specific neural mechanism, carried out via coordinated interactions of the brain’s memory and control systems [15,17,18]. In particular, the hippocampus is thought to drive retrieval of related memories during a new experience. Medial aspects of prefrontal cortex—a part of the brain associated with control and higher order cognitive function—may guide this retrieval process, all the while building up a “memory model” or “schema” that generalizes across multiple, similar events [15,18,19]. This model may be, in essence, the beginnings of complex knowledge: It abstracts over individual experiences, summarizing their common threads. During a new experience, memory models can be deployed to help us understand, remember, or even “fill in” missing information [20]. 

In the lab, we study integration using stimuli ranging from simple item-item associations in humans (e.g., remember that these two objects go together) [21,22,23] to spatial layouts in rodents (e.g., remember the location of this scent in a given environment) [16,24]. Such simple, controlled learning experiences combined with brain imaging techniques have made possible a host of discoveries in recent years. For instance, we now know that the hippocampus and medial prefrontal cortex are more functionally coupled—or, “talking” to one another more—when memories are being updated [21,22,25,26]. Under conditions favoring integration, medial prefrontal cortex may also take over some of the functionality normally carried out by the hippocampus—namely, it might actually form new memories—as it shows greater-than-normal engagement during these new experiences [21,27]. Parts of the hippocampus and medial prefrontal cortex also represent two related memories as highly similar, “distorting” experience to emphasize the interrelationships among different events [23]. (Paradoxically, this is not so for much of the brain, which instead highlights the discrepancies across similar memories.)

What relevance do these findings have for learning science? Does the integration mechanism “scale up” to things like remembering a new fact related to a previous course lecture? It seems to: One study demonstrated that undergraduates show neural signatures of integration when learning new facts [28]. Critically, this was the case when the new information was related to the students’ major area of study, but not when it was from an unstudied domain. These brain signatures also predicted later classroom performance, suggesting that integration mechanisms are:

1. engaged for the kind of material learned in the classroom, and
2. beneficial to learning as assessed in a university setting.

Despite these promising links, however, much about how memory integration plays out in the real-world classroom remains to be seen. 

Integration beyond the lab

We now know that memories are built by integrating information from multiple events, and that the memory system can support flexible behaviors not traditionally conceptualized as “memory tasks” per se. However, this complexity has been embraced only relatively recently by experimental psychology and neuroscience. And while interest in these ideas is growing, the bulk of the work remains to be done. For instance, virtually all of the human research on integration has been conducted in young adults—college students, generally speaking. What about younger learners [29]? We know that the brain—including hippocampus [30,31] and prefrontal cortex [8]—continues to mature into young adulthood. We also know that kids (and perhaps even teens!) form qualitatively different kinds of memories than adults [10]. Taken together, these bits of evidence underscore an important point: Kids may not yet be capable of integrating memories like adults. Instead, they may go about producing integration-like behaviors through a completely distinct, alternate neural mechanism. If this is the case, what are the implications of engaging this alternate system—under what learning conditions does it thrive, and when does it fail? Neurally, when does an adolescent become indistinguishable from an adult? The short answer: We don’t know. Future work on this topic will aim to flesh out this mechanism, and understand how we form complex knowledge—leading to a better understanding of how we learn, even early in development.