Categorization

background
software(purpose, applet, software)
selected publications

Background

Our world has structure, and we naturally group objects into categories. Sometimes this is good, because it helps us learn language, recognize objects and plan actions. Sometimes this can be bad: if we overgeneralize attributes to all members of a group, stereotypes can form and provide the basis for racism and bias. Regardless of how categories are used, it is important to understand how they form and how you rely on them.

Often the natural world can be organized into dimensions. Color is a good example: it can be organized into the three dimensions of Hue, Saturation, and Brightness. Other dimensions such as size also can be used as dimensions. categorization researchers often describe two dimensions as either separable or integral. To see why this matters, consider the two groups illustrated below. Imagine that within each group you had to categorize the items in the two left-most columns into category A and the two right-most columns into category B, all while ignoring the other dimension. This would be easy to do with theh space on the left, because size can be perceived independently from brightness. However, it would be hard with the group on the right, because the visual system treats saturation and brightness as integral. That is, we find it hard to attend to one dimension while ignoring the other dimension, which can be important. For example, doctors need to attend to relevant symptoms and ignore irrelevant ones, but this is only possible if the dimensions are separable.

Purpose of Our Software

This software allows the user to investigate questions about integrality and separability, or many related questions that concern how we organize and represent preceptual and semantic information. Representation is central to the study Cognitive Science, and categorization is one way to study it.

Applet

Applet

Software

application download
application download (with zipped files and images)
images

The program is very easy to use, but designing an interesting experiment will take some thinking. There are two tabs, Category A and Category B, which are the categories of exemplars used in an experiment. You select images from the drop-down menus, and then you can choose whether to colorize them.

Below is an example of one exemplar item that will show a face. If you check the "colorize it" button, the item will be colorized yellow. You can change the color by clicking on the "change color" button. To select a new picture for Exemplar 1, you would click on the arrow on the right to drop down the menu and choose a new picture.

Once you figure out how to set up exemplars, think about answering an interesting research question. Are certain categories easier to categorize than others? Why do you think this is?
Your goal should be to address some aspect of how we categorize items. What features do we use, and what makes that task easy or hard? You are not limited to the pictures in this application.
You can add .gif or .jpeg files to the TARGETS folder inside the CLASSES folder. Then add the names of your images to the target.txt file. When you add images make them no larger than 100x100 pixels.

Selected Publications on Categorization

1. Models of Visual Search: Finding a Face in the Crowd. Dosher, Barbara A. An Invitation to Cognitive Science (2nd ed), Vol. 4, Methods, Models and Conceptual Issues. Mit Press, 1998.