Date of Award


Document Type


Degree Name




First Advisor

Timothy D. Sweeny, Ph.D.


Ensemble coding, Perception, Segmentation, Vision


The visual system often integrates information that "goes together". Once information has been integrated, summary information (e.g., average emotion or average size) can be extracted; this occurs during ensemble coding. Integration thus allows for fast and efficient generalizations about sets to be made. In contrast, the visual system sometimes segments input that does not go together. For example, the perception of objects can be exaggerated away from natural category boundaries (e.g., a perfect circle is a category boundary; it is neither "flat" nor "tall"). Segmentation allows the visual system to make quick categorical distinctions. Much of the time, integration and segmentation work in parallel, and they have most often been studied in isolation. However, investigating how these two processes operate together, and potentially even conflict, was the purpose of this dissertation. I examined the ensemble coding of aspect ratio, which is a visual feature roughly equivalent to "tallness/flatness". Aspect ratio has a category boundary (e.g., a circle or square), and the perception of aspect ratio tends to be exaggerated -segmented - away from that boundary. Thus, I predicted that observers' ability to integrate aspect ratio information that spanned the category boundary would be disrupted, since in those instances, integration and segmentation would be at odds. To test this prediction, observers were asked about the average aspect ratio of a set of ellipses. In two experiments, observers were less sensitive to the mean of sets that included both tall and flat ellipses, compared to sets that only included tall or flat ellipses. A third experiment confirmed that segmentation perceptually distorted the appearance of ellipses near the category boundary away from that boundary; shapes were perceived to be more extreme than they actually were. Segmentation thus made sets that included both flat and tall ellipses appear more heterogeneous than they really were, which disrupted ensemble coding. In general, these experiments provide a deeper understanding of how the visual system summarizes large sets of information, by investigating how integration interacts with, and even conflicts with, segmentation.

Publication Statement

Copyright is held by the author. User is responsible for all copyright compliance.


Received from ProQuest

Rights holder

Elric Matthew Elias

File size

84 p.

File format





Cognitive psychology