Google owes its phenomenal success to its powerful search engine, which assigns priority in search results to web pages that are the most densely linked to other pages. The assumption is that if a lot of people link to a particular site, than it must be particularly interesting or useful. For example, if you search on Google for "memory fitness," the fourth link from the top of the first page of "hits" takes you to our recent profile of memory coach Cynthia Green. From this you can infer that lots of people find that article interesting or useful.

The brain also employs strategies for Googling your myriad memories to find a specific piece of information, but the details of the process have been somewhat unclear. Now, three psychologists at the University of Pennsylvania in Philadelphia believe they have observed an important aspect of the brain's search engine: When recalling a specific piece of information, the brain first searches for cues based on the general category to which that piece of information belongs.

In the experiment, volunteers memorized lists of items from three categories: celebrity faces, famous places, and common objects. As the volunteers memorized the lists, the researchers scanned their brains with a functional MRI (fMRI) machine. They fed the fMRI results to a special computer program designed to recognize the distinct patterns of activity associated with the different categories.

Later, the volunteers recalled as many of the memorized items as they could as their brains were scanned again. Using the patterns of brain activity obtained from the first set of scans, the computer program was able to predict the category of things the people were trying to remember-say, faces-up to five seconds before the person thought, "Jack Nicholson."

Team leader Sean M. Polyn, Ph.D. says he isn't planning to develop a mind-reading machine, but perhaps the fMRI experiments could help him to create a computer model of how people search through memory. This could have practical applications, such as teaching machines to more effectively search large databases of information.

"There's something amazing about how flexible the human memory system is," Polyn says, "and if we could extract the basic algorithm used by the human mind, then one would imagine it would be very useful." But, he adds, "I don't know if we'd be able to beat Google."

Resources

• "Category-specific cortical activity precedes retrieval during memory search," by Sean M. Polyn and others. (Science, December 23, 2005, Volume 310, Number 5756, pp. 1963-1966.)