Funded Grants

Developmental origins of hierarchical memory representations

All scientific discoveries are achievements of the human mind, and in some cases cognitive science can reveal parallels between the structure of these discoveries and the thought processes that created them. One example is the hierarchical organization of information into sets and subsets, a foundational concept in many scientific disciplines. For example, classical physics decomposes an object into its component molecules, in turn comprised of atoms, themselves containing subatomic particles. Biological taxonomy divides a family into genera, a genus into species, and a species into individual organisms. One reason hierarchies can play this explicit role in our scientific concepts is that they implicitly structure human thought more generally.

Set-subset relationships are central to our thinking for two reasons. First, nesting each level of representation within a more abstract level efficiently compresses information. Instead of representing many individual species, we can represent a genus, which may then be unpacked into its constituents. Second, different levels of the representational hierarchy support different inferences. An inference about a species (e.g., members of Canis lupus familiaris are domesticated, and so enjoy playing with humans) does not necessarily apply at the more abstract genus level. However, an inference about a genus (e.g., members of Canis eat meat) applies to all species nested within it, including Canis lupus familiaris. Hence, hierarchical structure crucially constrains reasoning about the entities it contains.

We deploy hierarchical structure not only during formal scientific reasoning, but also when facing everyday challenges. For example, adults are widely observed to chunk individual items into sets to expand memory. Such chunking makes the digits 8439910464 easier to remember as the three-chunk phone number 843-991-0464 than as a single 10-digit sequence. The fact that we remember both the number of chunks and the identities of items within each chunk implicates hierarchical representation; at one level are the abstract chunks; at another are the individual numerals.

If representing sets and subsets is fundamental both to formal theory-building and to everyday cognition, then what are the origins of this ability? Surprisingly, almost no evidence bears on this question. One possibility is that the ability to form hierarchical representations is learned, perhaps initially constructed over the course of scientific discovery. Alternatively, it might be a universal faculty of the mind, available throughout the lifespan without instruction.

Our research aims to uncover the ontological foundations of hierarchical representation. We will ask whether untrained infants can chunk individual items into higher-order sets, thereby increasing memory capacity. For example, can infants remember more when items are groupable into meaningful categories (e.g., 2 cats and 2 cars), versus when items are not groupable (e.g., 4 unrelated objects)? What perceptual and conceptual cues support such chunking? How many chunks can infants maintain, and how many items can occupy each chunk? We will address such questions by testing infants in naturalistic behavioral tasks. This developmental approach will allow us to reveal the ways that chunking changes over time and with experience, and thereby to shed light on the origins of a fundamental representational strategy.