It’s a beautiful day outside. While you recline comfortably under a tree, you see a small creature in the distance. It has four legs, a long tail, brown fur, and dark eyes that stare directly at you. If you had no prior experience with a creature like this, how would you encode and later remember this particular episode? Would you even remember? Luckily you know what it is. You can rely on your accumulated knowledge to identify the creature before you as a squirrel; adding structure and clarity to an event that might have otherwise faded from mind. The issue of how episodic memory and prior knowledge interact remains mysterious, though a new perspective shines a much-needed light on the subject.

The authors of a recent paper published in Psychonomic Bulletin & Review encourage researchers to adopt a multi-method approach in the hopes that we can move closer to answering the longstanding question: “How do people use their prior knowledge and expectations to help encode, store, and retrieve information from memory?”

In their review, Carla Macias and Kimele Persaud (a former Psychonomic Society Digital Associate Editor) first take a bird’s eye view of the field—highlighting the strengths and limitations of three of the most common approaches used to study the intersection of episodic memory and prior knowledge: Adult behavioral studies, developmental studies, and computational modeling. While each approach plays an essential role in scientific discovery and theory building, they also have notable limitations. These limitations can be overcome, the authors argue, by leveraging the strengths of all three at once.

The authors focus on a segment of the field, one centered on understanding how memory differs for items congruent with prior expectations (e.g., a yellow banana) and those incongruent with prior expectations (e.g., a blue banana). Macias and Persaud highlight a sea of mixed results. Some studies find better memory for incongruent items in adults, while others find the opposite. Young kids are primed to remember novelty, but older kids are better at remembering congruent items. Meanwhile, computational models struggle to formalize expectation-incongruent items and to accurately capture how prior knowledge changes over the lifespan.

How do we make progress amidst a turbulent ocean of mixed findings and incomplete models? One path forward is to take a critical look at our methods to identify their strengths and, more importantly, their weaknesses. Then, we can work towards a combinatorial approach that will result in studies greater than the sum of their individual methods.

Adult behavioral studies are the bread and butter of cognitive science. They are the workhorses, gathering the evidence needed to test and revise the existing theories. However, due to massive variability across studies, it can be difficult, if not impossible, to draw concrete conclusions that account for all existing evidence. This issue is compounded when canonical memory tasks make it challenging to untangle which types of information participants draw from to perform the task.

This concern is perhaps best illustrated when researchers struggle to parse whether participants’ responses are task-related memory retrieval or are educated guesses based on prior experience. For example, if the task requires participants to memorize a list of items found in kitchens, do they really have to memorize the list to perform well on the task, or can they draw on their prior knowledge of kitchens to guide their guesses? Current behavioral methods available for circumventing this issue require researchers to work from aggregate responses, meaning that individual trial-level data is lost. Computational models may help tease out this difference—while also preserving the individual-level data.

Developmental studies, meanwhile, focus on characterizing the relationship between prior knowledge and episodic memory as it changes across the lifespan. Previous studies have highlighted that the influence of prior knowledge on memory begins early in childhood and remains important throughout adulthood and older ages; lending insight into how the memory system matures and adapts to various constraints over time.

However, developmental data is both expensive to collect (both financially and temporally) and is often very noisy. To reduce these issues, researchers can augment their developmental studies with computational models to help parse the signal from the noise and use insights from adult behavioral studies to precisely probe theoretical predictions at different stages of development.

Finally, computational models are extremely powerful tools that help scientists define the goals and constraints of the memory system in line with current theories. They have helped elucidate the nature of the interaction between prior knowledge and memory representations and are instrumental in adjudicating competing theories of episodic memory.

Unfortunately, many computational models are limited in what types of data they can handle as well as the types of people they can describe. Most models are best able to capture young adult behavior, as that is the data that is most readily available. This is an issue, as only a fraction of a person’s life is spent in young adulthood, and we know from developmental studies that the influence of prior knowledge on memory changes as a function of age.

As teased above, researchers can benefit from weaving together multiple approaches to reduce the shortcomings of any single method. Beyond just filling in the gaps, integrating all three approaches can help the field answer questions that remain at the intersection of each approach. The power of approach combination is plain to see in recent work that tested predictions of three prominent computational models of episodic memory across the lifespan—a study that was only possible through the combination of adult behavioral historical findings, developmental-informed study design, and the formalization of computational modeling.

It is only through a combination of approaches, not isolation, that a holistic and detailed understanding of cognition—and episodic memory—can be achieved. By integrating the strengths of various methods, we can answer questions that have thus far been beyond our reach.

As elegantly put by the authors,

“[an] integrated approach has great potential for offering novel insights into the relationship between prior knowledge and episodic memory, and cognition more broadly.”

Featured Psychonomic Society article

Macias, C., & Persaud, K. (2024). From silos to synergy: Integrating approaches to investigate the role of prior knowledge and expectations on episodic memory. Psychonomic Bulletin & Review. https://doi.org/10.3758/s13423-024-02505-4