Search

Wolves and Wolfpacks: The Chase is On

You’re out of breath. Your heart is pounding as you make your way through the crowd. You’re being chased, and you know it. But you’re surrounded by hundreds of bystanders, and not sure where your pursuer is. Then suddenly, you notice someone, and know he is the one chasing you without a second thought. How did your brain determine that?

Brian Scholl, Professor of Psychology at Yale University, is attempting to answer that question by studying the cognitive mechanisms responsible for the detection of chasing. This research is part of the Yale Perception and Cognition Lab’s broader investigation of how humans perceive animacy — the ability of objects to have motivations or goals and to act accordingly.

Psychologists first recognized animacy as a distinct property of visual experience during the early 20th century. An important early development in the field came in 1944, when Fritz Heider and Marianne Simmel showed that observers attributed animacy, to simple geometric figures. Since then, multiple research groups have found that animacy perception persists when observers have explicit knowledge that objects are not animate, and that animacy perception occurs across cultures and even in infants.

Scholl became interested in animacy research when he asked himself the simple question, “What is it that I see, and what is it that I’m thinking about?” He realized that objects’ animacy stood out with as much immediacy as their color or shape, which led him to wonder whether animacy might be processed at a fundamental level in the brain, instead of at the higher levels on which most previous research had been focused.

A still frame from one of the videos used in Heider and Simmel’s 1944 experiment, in which the red triangle appears to be chasing the other two shapes. Courtesy of Professor Scholl.

Quantifying Animacy

When Scholl and graduate student Tao Gao began to study the perception of animacy several years ago, they faced a lack of quantitative methods to measure animacy perception. As Scholl puts it, animacy perception has been “fascinating psychologists … for decades as demonstration, and we’ve been in search of a way to turn it into rigorous science.”

The lack of quantitative methods resulted from two main methodological issues. First, most of the animations used in animacy studies were scripted manually and included multiple types of implied behavior, making the influence of any single feature difficult to isolate. Second, the most common measurement of animacy perception was a subjective questionnaire. The combination of these two issues made it difficult for researchers to distinguish animacy perception in the visual system from higher-level inferences.

To overcome these challenges, Scholl and Gao developed two models to measure one kind of animacy perception, chasing detection. Both involve three types of simple shapes moving on a two-dimensional screen: one “sheep,” one “wolf,” and multiple “distractors” identical in appearance, but not behavior, to the wolf. The behaviors of both the distractors and the wolf are generated by mathematical algorithms, allowing systematic control of the differences between them.
The first experiment (“Find the Chase”) generates the sheep’s movements algorithmically and asks observers to identify whether any chasing behavior is present, and if so, to identify the sheep and the wolf. The second (“Don’t Get Caught”) requires the observer to control the sheep and attempt to avoid the wolf for a fixed duration. In both experiments, observer performance can be objectively quantified by the number of correct detections and the number of escapes in the second, respectively.

Cues for Chasing

Using these new methods, Scholl and Gao examined different features of wolf motion, attempting to determine which were important for chase detection. One important cue that they identified was the maximum deviation of the wolf from the line between it and the sheep, which they called “chasing subtlety.” At a chasing subtlety of zero degrees, detecting a chase initially seemed very difficult, but the wolf and sheep quickly became obvious, allowing observers to detect chases in nearly 90 percent of “Find the Chase” trials and to escape 60 percent of “Don’t Get Caught” trials. In constrast, at a chasing subtlety of 60 degrees, the wolf and sheep failed to stand out and performance decreased drastically, with chase detection falling to 60 percent and escape rate to 25 percent.

As chasing subtlety increases, observers’ ability to detect a chase falls dramatically, becoming no better than chance by 120 degrees of subtlety. Courtesy of Professor Scholl.
At low chasing subtlety, the wolf is “obvious” — easy to detect and evade. At high chasing subtlety, the wolf is “incompetent” — hard to detect, but not likely to catch the sheep. The wolf is most dangerous at a medium chasing subtlety, where it remains hard to detect, but can gradually move towards the sheep. Courtesy of Professor Scholl.

Scholl and Gao then decided to study the impact of object orientation on chasing detection. By switching the shapes used to represent wolves and distractors from circles to arrowheads, they found that escape from the wolf was significantly easier when the darts were oriented in the direction of motion, rather than in a perpendicular or random direction. Critically, these results demonstrate that a very specific correlation between the motion and orientation of the perceived chasers enhances the perception of chasing.

Based on the subtlety and orientation results, Scholl and Gao investigated whether the brain detected chasing only by accumulating positive evidence for it or by looking for both positive and negative evidence. In a variation on the “Don’t Get Caught” task, they introduced interruptions into the movement of the wolf, during which it moved randomly, stayed stationary, or oscillated around a single point. Since the three conditions contain identical chasing behaviors, any difference between them must arise from the different types of negative evidence, i.e. non-chasing motion, present in each.

When the interrupting motion was random, the escape rate was high for both very low and very high proportions of interruption, but it was low when interruptions and chasing motion were present in a 1:1 ratio. However, when the wolf was stationary or oscillating, the drop in escape rate was much less pronounced, a finding that only a model involving both positive and negative evidence can explain.

Behavioral and Neural Effects of Chasing

Building on this new knowledge of the visual features that contribute to chasing perception, Scholl and Gao began to investigate how the perception of being chased affects behavior. Instead of using an actual chase scenario, they focused on the related “wolfpack” phenomenon, in which darts oriented toward the sheep create the impression of being chased. In this task, subjects controlled the sheep and were presented with a screen divided into regions, in which darts were either oriented towards or perpendicular to the sheep. Even though the motions of the darts are identical, the two conditions appear and feel very different, and impact behavior in striking ways. When asked to avoid all darts, subjects spent significantly less time in the wolfpack regions. In other words, the mere perception of being chased by objects causes people to avoid those objects.

To test the behavioral effects of perceived chasing, Scholl and Gao used this experimental setup, in which half the screen contained “wolfpack” darts (red regions) and the other half contained perpendicular darts (blue regions). Courtesy of Professor Scholl.

Having observed such a pronounced effect of perceived chasing on behavior, Scholl and Gao formed a collaboration with Gregory McCarthy, Professor of Psychology in Yale’s Human Neuroscience Lab. Using functional magnetic resonance imaging (fMRI), they monitored the brain activity of subjects completing dart-avoidance tests in the wolfpack or perpendicular orientations. These scans revealed three regions whose activity increased in the wolfpack condition relative to the perpendicular condition. The posterior superior temporal sulcus (pSTS), had previously been shown to relate to animacy perception, but the middle temporal (MT) area and the fusiform face area (FFA) had previously been thought to function only in lower-level visual processing. Scholl calls these results “incredibly interesting and even surprising,” suggesting that animacy processing “seems to pervade many more regions of the brain than we might have thought.”

These discoveries point the way toward further investigations on animacy perception in the Perception and Cognition Lab. They demonstrate that seemingly inefficient behaviors, such as not moving directly towards the target and interrupting chasing motion with random motion, strongly disrupt chasing detection. Based on this evidence, Scholl suggests that a “rationality principle,” which states that animate actors will behave in ways that most efficiently accomplish their goals, may form the basis of animacy perception. Future investigation of these principles could reveal fundamentals underlying our perceptions of the world — and certainly make for an exciting scholastic chase.

About the Author
Jonathan Liang is a junior Molecular Biophysics & Biochemistry major in Ezra Stiles College and the Online Editor for the Yale Scientific Magazine. He works in the laboratory of Professor Ronald Breaker, studying the natural diversity of riboswitches.

Acknowledgements
The author would like to thank Professor Scholl for his valuable time and insight into this area of research.

Further Reading
Gao T, Newman GE, Scholl BJ. (2009). The psychophysics of chasing: a case study in the perception of animacy. Cognitive Psychology 59:154-179.
Videos used in these experiments: www.yale.edu/perception/Brian/bjs-demos.html