The Berkeley Science Review: Articles

A Tough Nut to Crack
How campus squirrels find their loot (view PDF)
by James Walker

Anyone walking around campus is sure to notice the antics of our resident fox squirrels, harassing passersby for a handout or digging up nuts from beneath the oaks. However, most may not appreciate the amazing ability of these cute, fuzzy animals to remember where they stored all these nuts, a skill finely honed over millions of years of evolution. Cognitive psychology doctoral student Anna Waisman not only appreciates this ability, but is also studying how the squirrels do it, shedding light on the workings and development of human memory in the process.

Fox squirrels are scatter hoarders—that is, they place each nut they find before the onset of winter in its own well-hidden hole. Each squirrel has thousands of holes scattered in its particular territory. During winter, when the vegetation has changed and snow may cover the ground, the squirrels must rely on their memory to direct them to the exact spot where they buried a nut months before. It was discovered that this ability is based on memory of visual landmarks, which Waisman says "opens up lots of questions as to which [landmarks] are important."

Previous research on this question of spatial memory has fallen into two categories. The first is concerned with the hierarchy of landmarks ("cues"), which is based on how stable the landmarks are over time and how far they are from the hidden object. At the top are global cues, the most consistent objects in an environment— specific trees or buildings, for instance. At the bottom are feature cues, the potentially variable characteristics of the hiding place itself, like colors and smells. Local cues fall somewhere in between. In a typical experiment, squirrels are trained to find a nut in one of several canisters arranged on a table. In this scenario, global cues are everything outside of the table, local cues are how canisters are arranged on top of the table, and feature cues are characteristics of the canisters themselves. The lab of Lucia Jacobs, where Waisman works, showed that squirrels almost always depend on the global cues first, then local cues. Finally, if no other landmarks are available, they will use feature cues to make a decision.

The second category of spatial memory research involves competition between cues of the same type. For example, if two local cues suggest the nut is in spot A, but another local cue suggests the nut is in spot B, which spot does the squirrel choose? While we might expect squirrels to trust trees as cues more than, say, trashcans or light posts, it seems that most of the time the squirrel chooses the location where the majority of cues are in its favor, no matter what those cues are.

Waisman's research, as she explains, "marries the two fields." Her aim was to determine which hierarchy cues would win in the competition test. That is, if the squirrel has three kinds of cues (a global, a local, and a feature), but one of them contradicts the other two, where does the squirrel go? To perform the experiments, Waisman trained squirrels to find a nut in one of four canisters, specifically arranged on a table relative to the different kinds of cues. After the squir- rel had learned where the nut was hidden, she changed one or more of the possible cues and watched where the squirrel went to find the nut. Previous research suggested that the squirrels would follow the global cue all the time, but this turned out not to be the case. In general, Waisman found that the squirrels were able to do the math and usually picked the location that had two cues in its favor— regardless of the type of cue. This wasn't true all the time, however, which made Waisman wonder if there might be more going on.

Based on other research in this area, she thinks the squirrels might be weighting different cues according to their relative usefulness. For example, global cues might have a higher weight (40) than local (30) or feature (20) cues. In her original experiments, this weighting would mean that any combination of two cues would beat any single cue. But the squirrels would hesitate or choose the wrong canister more often when the weighting was close (global vs. local and feature, 40 vs. 50) relative to when it wasn't (feature vs. global and local, 20 vs. 70). Waisman's current efforts are focused on statistical modeling of her data to determine if this theory agrees with her observations. She is also designing experiments to measure the relative weights of each cue. If her hypothesis is correct, then the use of spatial cues would be much more complicated than first thought, and might reflect general decision-making processes in the brain. Researchers would then have much more to study, including what attributes influence the weighting, and whether the weighting is innate or learned.

In the future, Waisman hopes to widen the scope of her research by studying how humans, especially children, use spatial cues to make decisions. Like squirrels, preschoolers appear to prefer global cues in hierarchy tests, but this preference goes away as cognitive development proceeds. No one has yet tested competition between hierarchies in young children, an experiment that Waisman is excited to try.

In the end, the ways humans and squirrels think about spatial cues may not be so different. So the next time you walk through campus, go ahead and give that begging squirrel a treat—the study it just participated in might one day help you remember where you put your keys.

James Walker is a graduate student in molecular and cell biology.

Comments on this article? Drop us a line at with 'letter to the editor' in the subject!