Less Is More: Better Memory With Fewer Neurons

Ever heard that "you're born with all the brain cells you'll ever have"? It turns out that could be a good thing - if it were true. A new study shows that at least in some circumstances, neurogenesis actually impairs memory performance.

To understand why this might be the case, consider that adults are constantly generating new neurons in a long-term memory structure - the hippocampus. This region requires a large number of neurons to store episodic memories accumulated over a lifetime (and understandably so!). Similarly, to be able to store experiences that may have occurred very quickly, neurons in the hippocampus learn very fast. Finally, to reduce interference from other similar memories, neural activity there is very sparse and highly non-overlapping (so that the hippocampus doesn't accidentally confuse two similar experiences).

These characteristics make this region particularly useful for quickly learning relationships between stimuli. Accordingly, damage to the hippocampus, and disruption of neurogenesis in particular, will usually result in impaired learning. However, Saxe et al. showed that disrupting neurogenesis actually improves performance on some of the more complex learning tasks we know of!

Saxe et al. placed mice in the middle of an asterisk-shaped "radial arm maze" with 8 arms; at the end of each arm was a morsel of food. Then, one arm was opened. After mice had retrieved the food from the end of that arm, the same thing was repeated with another arm. Finally, after mice had retrieved food from the second arm, the mice were given a 30- second delay, and then given the option to return to the first arm or to an adjacent arm. In this case, returning to the first arm would reflect a memory failure - that arm had already been cleared of food.

Mice without neurogenesis (as dirsupted by X-Rays and a genetic manipulation) actually did better on this task than normal mice - they were more likely to go to the adjacent arm in search of food, rather than return to the original arm from which they had already cleared the food!

Mice with and without neurogenesis performed equally on a task where the delay was reduced to 5-seconds. A similar pattern emerged on another version of the task in which mice were only tested on the first arm vs. an adjacent arm, demonstrating that the memory advantage of disrupting neurogenesis is not due simply to less interference from the memory of the second arm.

To explain this thoroughly confusing result, the authors suggested two explanations:

1) Mice are less likely to suffer from the "interfering memory" of visiting the first arm because having less neurogenesis results in worse short-term memory - there is no memory to cause interference! However, this explanation seems unlikely both because of previous results (showing no short-term memory defiicts resulting from hippocampal irradiation) and because short-term memory could just as easily help performance as hurt it in the way they imply.

The second explanation is far more interesting, though somewhat counterintuitive:

2) Less neurogenesis has no effect on short-term memory, but reduces interference by making overlap between cells less likely.

"But wait, didn't you say the purpose of neurogenesis was to create less overlap, by increasing the total number of available neurons?"

Ultimately, new neurons do result in less overlap between representations, but young neurons are quite different from older neurons in that they are far more excitable. In other words, very young neurons will fire to almost anything, and will cause increased interference from previous memories.

So less really can be more: an abundance of new neurons can result in overexcitability in the hippocampus, a region that requires very sparse, low-level patterns of activity in order to store memories without the threat of interference. Over the long term, the young neurons become less excitable and do ultimately increase the capacity of memory (indeed, they may do so even earlier for highly distinct memories). An interesting question for future research will be the kind of cues that are required to recruit non-overlapping regions of the hippocampus to make a memory "highly distinct" and thus resistant to the initially deleterious effects of hippocampal neurogenesis.

Related Posts:
The Myth of Infantile Amnesia
Overgrowth, Pruning and Infantile Amnesia
Modeling Neurogenesis
Neurogenesis in Kids, Adults, and SIlicon
Structural Signatures of Intelligence