So I've talked a little bit about sleep and memory formation prior, and it's actually a fun topic for me. So I'm running with it.
Now, there's a lot more in the field of normal memory, and I'll get to that eventually, but there's always a fun side-field to memory research - FALSE memories. In other words, you just KNOW it happened... except it didn't. You made it up. Except you didn't really make it up intentionally, but for whatever reason, it didn't happen but you think it did. And then your head hurts.
This is false memory, for the most part. It's a big field, and a lot of theories behind it, like the idea that we tend to remember generalizations more than we care to admit, and a lot of the detail-filling-in is reconstruction. This at least the generally accepted form, in brief, as to what happens (although I can't say I experience this a lot, because I just forget everything anyway).
Anyway, the most common paradigm in false memory research is the Deese-Roediger-McDermott paradigm (DRM) - basically, throw a bunch of words at someone with a 'theme' and people will be SO SO SURE that you said the word that fits the theme. For example, in the paper the authors use the wordlist door, glass, pane, shade, ledge, sill, house, open, curtain and then with that, there is a pretty high chance that the participant will report having heard the word window... even though you said no such thing. (In the DRM, these words are the 'critical words'.) The original research with the list showed that a 2-day delay produced HIGHER levels of false recall than true recall.
Payne, J., Schacter, D., Propper, R., Huang, L., Wamsley, E., Tucker, M., Walker, M., & Stickgold, R. (2009). The role of sleep in false memory formation Neurobiology of Learning and Memory, 92 (3), 327-334 DOI: 10.1016/j.nlm.2009.03.007
So, subjects here: 101 WEIRD Merrimack college sophomores, repeated with 84 WEIRD Harvard college sophomores. (Actually, we don't know that they're sophomores... that's me making up information.) Give them eight DRM wordlists. Two long-term groups - study at 9 AM and repeat at 9 PM, and study at 9 PM and repeat at 9 AM - and two short-term groups - study at either 9 AM or 9 PM and repeat 20 minutes later. The latter was a baseline, mainly to see if there were circadian effects (this is HARD to control for well, though, in sleep/memory research, and I don't think anyone has ever done it very effectively because it's practically confounded by nature).
Anyway, what did they get?
The sleep group had better recall. No big surprise - they remembered about six more words on average (out of the 96 total - 8 groups of 12 words each). Both groups had similar intrusions (non-critical words 'remembered'), but the sleep group showed on average one more critical word recalled. Because of the similarity in intrusions, but difference in critical words, the authors suggest that the false critical words aren't simply a by-product of having more overall recalls.
There's two more parts to the experiment, trying ever so more to zoom in on what's going on. A second experiment found a negative correlation with the amount of slow wave sleep and recall rates, but no significant difference in the number of critical words recalled - this experiment was smaller, and thus greater variance, and there's only eight critical words after all. Part 3 was a nap study, and the authors see (significant) greater critical word recall in the sleep groups compared against the no-sleepers, without a significant effect in the true-recalls or intrusions.
Question on part 3 - The authors conclude that slow wave sleep preferentially favors critical recall over true recall. But on what basis? Well, that critical recall between sleep/no-sleep was above significance threshold, tested against the null, but not for true recall. But wouldn't you want to be comparing the differences against each other, and then look for significance? Granted, looking at the graphs, it should pass that test - the recall measures look almost identical for the sleep/no-sleep peeps, but really... methinks you did the wrong test. I enjoy the enthusiasm in running multiple experiments, and I think their data probably do support their tentative conclusion from 3... but they didn't test it right, unless I happened to misread it.
So the authors suggest that a big part of the DRM's success in getting false critical words is that in the delay, there is sleep. Fair enough. And it's actually a good paper to show that sleep and memory work is hard, and results can be confusing and seem to run counter to each other sometimes (although this is often an effect of the constructions we use to 'divide' types of memory.