How does that 3-pound lump of electrochemical jelly inside your head form, store and recall an apparently infinite number of memories? This is one of the most enduring mysteries of neuroscience and an endless source of fascination for those investigating how the brain works.

But despite years of rigorous scientific investigation, we are still not exactly sure how memories are formed and where in the brain they reside. Research in the 1970s began to unravel the mystery, and eventually led to a theory about how neurons co-operate to store memories. A memory, the theory holds, is a pattern of neuronal connections. But this widely accepted idea is now being challenged by some recent experimental findings, which point to the neurons themselves as the placeholders for memories.

The current view of how memories are formed originates in a hugely influential 1949 book called The Organization of Behavior, by the psychologist Donald Hebb. "When an axon of cell A is near enough to excite cell B and repeatedly or persistently takes part in firing it," he wrote, "some growth process or metabolic change takes place in one or both cells such that A's efficiency, as one of the cells firing B, is increased." In other words, synapses, or connections between neurons, become stronger and easier to form when cells are repeatedly active at the same time.

Hebb's idea was way ahead of its time. It was not until more than 20 years later that physiologists Tim Bliss and Terje Lømo discovered such a mechanism. They found that simultaneous, repetitive electrical stimulation of pairs of neurons in slices of rabbit brain increased the efficiency of the signaling between the cells, strengthening the synapses at which they communicate with each other for many hours after the initial stimulation.

This process, which Bliss and Lømo called long-term potentiation (LTP) has been studied intensively ever since, and today it is widely believed to be the neural basis of learning and memory. Studying the sea slug Aplysia californica, neuroscientist Eric Kandel and his colleagues found that when a short-term memory is formed, there's a transient increase in the strength of existing synapses within a group of neurons. For a long-term memory to form, the expression of genes in neurons changes and results in producing new proteins and permanent synaptic connections. This work led to a Nobel Prize in Physiology and Medicine for Kandel and his colleagues in 2000, and ever since, the general consensus has been that memories reside in the pattern of synaptic connections and are subsequently recalled by reactivation of the same pattern of activity within a network of neurons.

But new research suggests that synapses aren't the center of action when it comes to memory — instead, rather than being held in synaptic constellations, long-term memories may be actually stored in neurons.

David Glanzman of the University of California, Los Angeles and his colleagues dissected sensory and motor neurons from sea slug, and grew them in Petri dishes. Under these conditions, the cells spontaneously form new connections to each other, but the researchers also added the neurotransmitter serotonin to the dishes. This induces a simple form of learning by strengthening the new connections.

When they looked at the cells 48 hours later, they found that the new connections had disappeared; so, too, had other connections that existed before treatment with serotonin. But they also noticed that the cells had formed other new synapses, so that the overall number of connections between the cells remained the same, suggesting that they had retained information about the number of synapses they had formed, and that the precise pattern of connections is not essential for storing the memories.

In another set of experiments, the researchers sensitized the sea slugs to mild electric shocks, so that they reflexively withdraw when exposed to them later. They then treated the creatures with a drug that inhibits protein synthesis. This prevents synapse formation, and should thus eliminate the creature's memory of the shocks.

Remarkably, though, they found that the memories remained intact, or were reactivated in some way— even though the synapses thought to store them had been destroyed, the sea slugs still withdrew when exposed to more electric shocks later on.

These findings, the researchers said, suggest that established memories are not maintained by synapses, but instead form through changes in the cells.

However, the findings don't mean that synapses are not needed at all, Glanzman says. "Think of the cell body or nucleus of neurons as the brain of the concert pianist, and synaptic connections as his or her hands and fingers. Memory is the music. There is no music without the pianist's hands and fingers, but the pianist's knowledge of how to play Chopin, for example, does not reside there."

The idea that the nervous system may be able to regenerate lost synaptic connections is a radical one, Glanzman says. If the findings extend to mammals and humans, one implication is that developing drugs to target synapses and erase traumatic memory (for treating PTSD) would be futile. On the other hand, in people with Alzheimer's disease, which destroys synapses, the lost memories may still exist as long as the neurons are still alive, Glanzman says.

"This is highly provocative work… [that challenges] the idea that synapses store long-term memories," says Steve Ramirez, a memory researcher at the Massachusetts Institute of Technology, "but it's never fully clear that what happens in a dish actually recapitulates what happens in the brain."

It's well known that protein synthesis, the recycling of the synaptic vesicles that store neurotransmitter molecules, and various other cellular processes are involved in memory, and it's possible that some kinds of memories are more dependent on these processes than on events taking place at synapses.

"Memories that involve highly complex learning procedures might differentially recruit cellular and synaptic processes," says Ramirez, "but this is an under-exploded area in our field, and the biological basis of different kinds of memories is a tantalizing beast to study."

Neurons may also have as yet unknown mechanisms by which they can retain information about their connectivity patterns even when their synapses are abolished. Moreover, some kinds of memories may be more reliant on cellular processes than others.

"Memories indeed recruit synapses, but they also alter the biochemical cocktails existing in neurons, the physiological bolts of micro-lightning that neurons fire, and their overall architecture," says Ramirez. "All these processes seem necessary for memory formation, storage, and retrieval, but the question is, which are sufficient to reinstate particular memories?"

"Long-term memory storage may recruit global cellular changes to alter a brain cell in a long-lasting manner," he adds, "whereas short-term memories may recruit reverberant synaptic activity."

Sea slugs are popular with memory researchers because they have a relatively simple nervous system consisting of just 20,000 neurons. Even though they are separated from humans and other mammals by more than 500 million years of evolution, they share the same molecular machinery for memory. But they may nevertheless have memory mechanisms that aren't found in the human brain.

The conflicting results may therefore point to processes that are unique to sea slugs, or more generally to invertebrates. " Aplysia is beautiful because we can really go in and mechanistically study defined circuits with known structure and unparalleled control over the system as a whole," says Ramirez, "but the kinds of memories they store tend to be more reflexive and not the rich repertoire of memories that humans have [so we should] extrapolate with caution."

But Glanzman finds it very unlikely that he has discovered a novel memory mechanism specific to sea slugs. Generally, basic physiological mechanisms are preserved throughout evolution — and there's no reason to think this one is an exception.

"I am an unregenerate Darwinian in this regard. I believe that all of the basic cellular and molecular mechanisms of learning and memory will prove to be general, from worms to man," Glanzman said.

The brain is, as the old cliché goes, the most complex object in the known universe, and our understanding of how it works is at best rudimentary. "The idea that synapses store long-term memories is an overly simplistic model of a highly complex phenomenon," says Ramirez. And so this new controversy may simply reflect our profound ignorance of brain function.