42 • CHAPTER 2 Cognitive Neuroscience
picture. Participants were asked to think of properties of the object as they looked at
the picture. For example, when looking at the drill they might think about drilling holes
in a board. Each picture was presented for 3 seconds, followed by a 7-second rest inter-
val. While the participants viewed the pictures, the activity of their cortex was being
recorded by the fMRI scanner.
The key to the success of this experiment was the computer program, which ana-
lyzed the responses of the brain voxel by voxel, where a voxel is a small cube-shaped
area of the brain about 2 or 3 mm on a side. (The size of the voxel depends on the reso-
lution of the fMRI scanner. Scanners are being developed that will be able to resolve
volumes smaller than 2 or 3 mm on a side.) By determining which voxels were activated
by each picture and how strongly they were activated, the computer created a response
profi le, or “neural signature,” for each object, which included many areas of the brain.
Eventually, after collecting patterns from a dozen participants, the computer deter-
mined the neural pattern associated with each class of objects (tool vs. dwelling) and
with each individual object (hammer, apartment, or screwdriver, for example).
The computer was then tested by having it analyze a person’s brain activity as he
or she was viewing an object. Based on the pattern, the computer predicted what the
person was seeing. When the computer’s task was simply to indicate whether the person
was looking at a tool or a dwelling, the accuracy for 4 of the 12 participants was 97
percent; for the entire group of 12 participants, it was 87 percent (chance performance
being 50 percent because there were two possible answers). The average accuracy for
identifying specifi c objects was 78 percent (chance being 10 percent, because there were
10 different objects).
This is impressive performance, but what is even more impressive is that the
computer made accurate predictions even for people whose data had not been pre-
viously analyzed. Imagine what this means. You walk into the brain imaging facil-
ity for the first time, are placed in the scanner, and view a picture of an apartment
building. The computer analyzes your brain activity and concludes that you are
looking at a “dwelling,” and also predicts “apartment building.” Average accuracy
for determining the category (“dwelling”) is 82 percent. This ability to determine
what a particular person is seeing based on the data from other people is pos-
sible because patterns of brain activation are similar for different people. In other
words, different people have similar neural signatures for specific types of objects.
This commonality among people is illustrated in ● Figure 2.24, which shows the● FIGURE 2.23 Stimuli for the Shinkareva et al. (2008) experiment. Participants viewed a series of pictures for 3 seconds
each, with 7 seconds between pictures, while their brain activity was being measured in an fMRI scanner. (Source: S. V. Shinkareva
et al., “Using fMRI Brain Activation to Identify Cognitive States Associated with Perception of Tools and Dwellings,” PLoS One, Figure 1, p. 2, 2008.)3s
7s
X
X
X
Copyright 2011 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part.