|MadSci Network: NeuroScience
Human memory is now thought to be a large number of separate systems. You may be familiar with terms such as long-term memory, a hypothesized memory system that holds facts like your current phone number and mother's maiden name. There are plenty of others: the auditory loop which stores recently perceived sounds, the visual-spatial sketchpad for "mental imagery," muscle memory, etc.. The brief answer to your question is that some of these memory systems clearly require maintenance. For other memory systems, especially those storing the most stable long term memories, no one knows.
Here's a somewhat longer answer:
One of the better understood memory systems, the auditory loop, acts like a two-second long strip of audio tape. In one experiment, psychologists tested the "digit span" --- i.e. the longest sequence of random digits a person can remember --- of people from different countries. It turned out that Chinese-speakers could remember the most, then English speakers, then Welsh-speakers. They then compared the average digit-span to the average length of time needed to speak the name of a digit in each language. It turned out that cultural differences in digit span come from differences in length of digit words. Since the Chinese had the shortest words for digits, their memory went the farthest.
Other experiments show the limits on people's ability to "protect" old memory in the auditory loop from getting replaced (overwritten) by new memories. In particular, people must repeat the words to themselves (subvocally), an activity called "maintenance rehearsal." This requires a lot of effort and is quite suseptible to disruption. In this case, even though we don't know how memory works at the neurological level, we know that maintenance is needed to keep the memory usable.
One case where maintenance probably does not occur (except in the sense that the usual life processes are keeping the organism from decaying) is in memories used to process phonemes --- the smallest meaningful sound units in human speech. Infant brains learn to recognize the phonemes in their parents' language by selectively killing off cells in a certain part of the brain, allowing the remaining cells to become useful. Think of it as "sculpting" phoneme memories out of "unshaped" brain material. Here, it is the permanent connectivity of the remaining neurons that produces the memory. Maintenance is unnecessary.
In other cases, our ignorance of how brains store memories makes it difficult to guess whether a memory requires maintenance or not. For example, amnesiacs taught how to solve a puzzle called the "Tower of Hanoi" were unable to remember the experience of being taught --- in fact, they sincerely denied ever seeing the puzzle (or the experimenters) before --- but retained most of their skill at solving the puzzle. This seems to indicate that memory for "experiences" is stored separately from memory for "procedures." Along with some other evidence, it seems that procedure memory is more stable than experience memory.
Does this mean that procedure memory works like phoneme memory and doesn't require maintenance? Or that experience memory is like the auditory loop? This is exactly the sort of thing that no one knows. That being said, it's unlikely that human memory systems of any kind need maintenance in exactly the same sense as computer RAM --- i.e. the maintenance of current through a circuit. There is no bioelectic circuit to run current through. Rather, the electric nature of the neurosystem comes from interactions between individual neurons. Maintaining these electrical properties requires only keeping the cells alive and letting them stay connected to their neighbor cells.
As for cryogenic storage, you may want to check out a Scientific American article from a few years back. Mammals with built-in cryogenic sleep cycles (hibernation) have a substance in and between their cells that acts as anti-freeze. This prevents the water in their body from turning into ice which would destroy cells and, importantly for brains, disrupt the connections between cells. With anti-freeze present, the memories most likely stay intact. (I've never heard that bears have to relearn their way around the forest every Spring, although I suppose this is possible.) But the presence or lack of electricity makes no difference for preserving memory during cryogenic stasis. The problem stems from what you already know about ice: when water freezes, it expands and breaks things. So when it expands in the brain, permanent brain damage is the result.
Try the links in the MadSci Library for more information on NeuroScience.