For some 40 years, neuroscientists have believed that the brain forms memories by using a "sketch pad" to quickly record experiences and information learned throughout the day.

Stenographic duties, under this model, fall to the hippocampus, the two slightly curved sections of the brain located under the temporal lobe that are implicated in episodic memory. During sleep, the thinking goes, neurons in the hippocampus fire, driving a transfer of its information to the neocortex, the top layer of the cerebrum that serves as the brain's hard disk, or permanent storage bin. This model seemed to explain why people with hippocampus damage could recall old memories but could not create new ones.

Over the past decade or so, however, a number of new findings have cast doubt on this proposition. And now there's a new theory that's calling it into question even more; this one is based on new research by scientists at Brown University and the Max Planck Institute for Medical Research in Heidelberg, Germany, who studied communication between the neocortex and three parts of the hippocampus in anesthetized mice. "We can't just talk about the hippocampus as one monolithic block driving the neocortex," says Brown neuroscientist Mayank Mehta about the group's findings. "What seems to be [happening] is that all the neuron types in the hippocampus are showing some echo or antiecho of the neocortex. None of them seem to be driving neocortex."

According to Mehta this is the first time that scientists have observed the connection between the neocortex and hippocampus. In mice anesthetized to mimic deep sleep, the research team monitored neuronal activity in the neocortex as well as neurons in three regions of the hippocampus: the dentate gyrus (or input station), the CA1 (or output channel) and the CA3, an area consisting of a maze of internal links.

Their findings: a synchronous firing of neurons in the neocortex was followed by a diverse set of responses in the hippocampus. Neurons in the dentate gyrus activated shortly after the neocortex's signals were sent as if an echo, cells in the CA3 region "echoed" weakly and neurons in the CA1 actually became calmer when the neocortex fired.

During sleep, instead of replaying a neuronal activity pattern of a learned behavior to the neocortex, the hippocampus responds to activity in the neocortex. Mehta and his colleagues surmise, "that [the] synchronous activation of [the] hippocampus by the [neo]cortex erases whatever you have learned recently from hippocampus."

In other words, rather than memories being transferred to the neocortex during sleep, the authors speculate that memories are stored in both the neocortex and the hippocampus. Then, during sleep, the hippocampus, acting as a temporary storage system, is cleared for another day of learning, while the memories are retained in the neocortex, which provides permanent storage much like a computer hard disk.

Bruce McNaughton, director of the Division of Neural Systems, Memory and Aging at the University of Arizona, believes the new work adds another piece to the puzzle, but does not believe it is a breakthrough. "The bottom line here is that this is a very very complicated system," he says, adding that he expects it to take another 20 years before the science community fully understands exactly how memories are formed and stored. "One has to be very careful, in interpreting the results done under anesthesia," he warns, "because it's totally not the same brain."

Mehta acknowledges that this is a working hypothesis, and not what is happening. "We have been fooled so many times by this circuit," he says, "it would be foolish to think we have figured it out."