Select your localized edition:

Close ×

More Ways to Connect

Discover one of our 28 local entrepreneurial communities »

Be the first to know as we launch in new countries and markets around the globe.

Interested in bringing MIT Technology Review to your local market?

MIT Technology ReviewMIT Technology Review - logo


Unsupported browser: Your browser does not meet modern web standards. See how it scores »

{ action.text }

Understanding Object Recognition

Much of the work in DiCarlo’s lab concerns object recognition, which is what allows us to identify an object (such as a cow) in many different presentations (cows far away, cows viewed from above, cows at dawn, a cow in a truck) without mistaking it for similar objects (like, say, a horse). DiCarlo and graduate student David Cox published research last August in Nature Neuroscience that focused on one of the basic questions about object recognition: how much of our success in recognizing objects depends on hard-wired, innate circuitry, and how much on learned skills?

DiCarlo and Cox conducted each of their experiments on a dozen people, one person at a time. Subjects sat in front of equipment that could both display images of objects and track the direction of the subjects’ gaze. The objects were computer generated and looked vaguely like anthropomorphized animals, but they were designed to be unfamiliar to the subjects. An object would appear in one of three positions on a screen, and the subject would naturally shift his or her gaze toward it. For certain objects, however, the researchers would substitute new objects while the subjects were moving their eyes. For example, let’s say an object that looked kind of squat, with perky ears, was introduced at the right of the screen while the subject was focusing on the center. As the subject’s gaze shifted toward squat and perky, the researchers would replace the object with one that looked slightly thinner, with droopier ears. Since humans are effectively blind during gaze shifts, the subjects did not notice the swap. But their brains did.

After an hour or two of exposure to different objects, some of which were consistently swapped out when they appeared in particular positions, subjects were presented with pairs of the objects in different positions on the screen and asked to compare them. One might expect that the subjects would distinguish the objects without much difficulty. And so they did, except when the objects had been swapped – and were now reappearing in the same positions where the swaps occurred. Subjects tended to confuse those objects: that is, they were more likely to judge that squat and perky at one position and thin and droopy at another were one and the same object. DiCarlo thinks such errors show that the brain’s mechanisms for recognizing the same object in different places depend on normal visual experience across space and time. “The finding suggests that even fundamental properties of object recognition may be developed through visual experience with our world,” he says. DiCarlo and his team are conducting similar experiments in animals to examine the patterns of neuronal activity that underlie object recognition. (A good example of this research was published in the November 4, 2005, issue of Science magazine. DiCarlo and three collaborators recorded and analyzed the activity of hundreds of neurons in macaque brains. They were able to show that highly reliable information about object identity and category was contained in even handfuls of neurons.)

0 comments about this story. Start the discussion »

Reprints and Permissions | Send feedback to the editor

From the Archives


Introducing MIT Technology Review Insider.

Already a Magazine subscriber?

You're automatically an Insider. It's easy to activate or upgrade your account.

Activate Your Account

Become an Insider

It's the new way to subscribe. Get even more of the tech news, research, and discoveries you crave.

Sign Up

Learn More

Find out why MIT Technology Review Insider is for you and explore your options.

Show Me