Research

Current Research

Our lab investigates human information processing heuristics that shape how we perceive, encode, remember, and pay attention to information in the surrounding environment. Check out our Publications page for more information about our current research, and our Participation page for details on how to participate in our psychophysical, eye tracking and EEG experiments.

Ensemble Encoding in Human Vision

Just from looking at the images of the two trees on the left, you can quickly tell that they are from different seasons from the average hue of the leaves without needing to look at any of the individual leaves. From the different spatial variance on the right, you can rapidly judge the top forest as being more "natural" compared to the more "artificial" forest in the bottom image.

Humans are heuristic, not algorithmic. Unlike a computer which can quickly render each pixel of detail, we have an extremely limited capacity to encode only a fraction of what is visible in each glance. Even then, the resultant internal representations are far from faithful to the true state of the external environment. Yet somehow, we get the gist, and perceive the world as stable and complete. How does the brain fool us into this illusion of perceived order amidst continuous sensory chaos?

Although we are poor at encoding and retrieving the specific details of more than a few individual items, the visual system can effortlessly and rapidly extract statistical summary or “ensemble” representations of entire sets of items without needing to maintain a detailed representation of each individual item. In other words, even though we feel as if we are somehow taking faithful “mental snapshots” of the visible environment, we often “miss the trees for the forest.”

When we try to memorize all the dots in the set on the left and then judge whether a test dot presented afterwards was present, we can't remember all of the dots and usually to have to guess. HOWEVER, even though we can't remember all of the dots, we much better at saying whether a test dot represented the mean size of the entire set. In fact, ensemble encoding is so pervasive over features (e.g., orientation; economic value), modalities (e.g., auditory), and statics (e.g., numerosity), there has never been an empirical demonstration that it can be prevented.

Ensemble Bias in Human Vision

In addition to allowing us to rapidly interpret complex information, our fundamental and often unrealized reliance on ensemble representations introduces profound distortions in the way we perceive and remember important individual objects. Unlike a computer which can quickly render each pixel of detail, we almost never "see" the exact physical properties of single objects in isolation, but instead evaluate their size, color, contrast, orientation, and even semantic meaning with respect to the contest in which they are organized. We are just beginning to uncover the profound consequences of ensemble encoding for how we evaluate individual items.

Left: We have learned so much about how ensemble properties allow us to make rapid sense of complex information, including insight from cross disciplinary applied work in information visualization. But, humans otherwise rarely go around explicitly calculating ensemble properties; we are usually interested in a few important objects from moment-to-moment, depending on the task at hand. Yet, the implicit, warping effects ensemble representations have on the way we perceive and remember individual objects are far less understood.

Right: Sometimes individual items are evaluated as more similar to their ensemble properties. Regression (top): The same sized dot is remembered as bigger than it really was when it was presented in the larger average size blue set of dots on the left versus the smaller average size set of red dots on the right. Other times, individual items can be biased away from their ensemble properties. Repulsion (bottom): Even though the center dot on the right is the same size as the center dot on the left, the right dot in the smaller average size set often appears larger than the left dot in the larger average size set.

To learn more about our work on ensemble encoding:

Check out our Publications page, especially our our new book: The Pervasiveness of Ensemble Perception

Google Sites

Report abuse