About this topic I'd suggest you read "Designing with the mind in mind" by Jeff Johnson.
It's a must-have book and Johnson dedicates chapter 7 to how memory works.
In particular, he distinguishes between:
- Working memory (a.k.a. short-term memory) - "which covers situations in which information is retained for intervals ranging from a fraction of a second to a few minutes".
- Long-term memory - "which covers situations in which information is reatined over longer periods (e.g., hours, days, years, even lifetimes)".
How do these 2 memory systems work?
Working memory and long-term memory are not separate memory stores, as some theories have considered them in the past.
Researchers believe now that they're functions of a single memory system - one that is more closely linked with perception than previuosly though (Jonides et al., 2008).
What is the capacity of working memory?
Johnson writes:
Many college-educated people have read about "the magical number seven, plus or minus two", proposed by te cognitive psychologist George Miller in 1956 as the limit on the number of simultaneous unrelated items in human working memory (Miller, 1956).
[...] Later research in the 1960s and 1970s found Miller's estimate to be too high.
[...] When the experiments were revised [...] the average capacity of working memory was shown to be more like four plus or minus one - that is, three to five items (Broadbent, 1975; Mastin, 2010).
Anyway, I think that your question is more focused on how recognition and recall work rather than on "memory duration".
Recognition is easy, because - in Johnson's words:
The human brain was "designed", through millions of years of natural selection and evolution, to recognise things quickly.
By contrast,
Recalling memories - that is, retrieving them without perceptual support - is long-term memory reactivating old neural patterns without immediate similar perceptions and is, therefore, hard.
How long should we be shown an image to register it?
Johnson writes:
If you're shown an image for 5-10 milliseconds, you won't be aware of seeing it, but low-level parts of your visual system will register it. One effect of such exposure to an image is that your familiarity with it will increase: if you see it again later, it will seem familiar.
What do long-term memory, working memory, recognition and recall have to do wuth your question?
To summarise:
- While a user is interacting with your website for the first time (s)he relies first on recognition and working memory to interpret the UI: navigation, buttons, etc. Recognition is a fast process and the user can concentrate on his goal rather than on the interface. If the UI is "intuitive" (= recognisable) the user will be able to focus on his goal(s).
During the visit, every element the user focuses on for more than 5-10 millisecond becomes somewhat familiar to the user. The longer the exposure, the more aware the user is of his own familiarity with the object.
At the same, some parts of the UI might be not immediately recognisable. For instance, when you use a non-standard UI. That's when the user needs to recall, relying on the long-term memory. It's an harder task, takes longer and prevent the user to fully concentrate on his goal(s).
Anyway, every time the user comes back to the website repeating the same operations, he will become more and more familiar with some recurring paths. These paths slowly become unconscious and possibly he will also be able to follow the different steps in his mind thanks to what Don Norman calls "procedural memory" ("The design of everyday things", pag. 47).
How many times does the user need to visit the website to become familiar with it and even stop reading instructions, labels, etc? It's hard to tell.
It depends mainly on: how often does he visit the website, how standard is the interface, how complex is the operation he needs to do.
Therefore, my advice is to be as standard as possible.
As Steve Krug's writes in "Don't make me think", the less we make the users think (on their first visit and on next ones), the better.
There would be many many more things to say (your question is really interesting and opens many discussions), but I think it's enough for the moment :)
Bottom line:
Memory is a complex subject and still a research topic.
Using standard interfaces and relying on recognition rather than recall makes a better user experience.
We want the users to focus their (very) limited working memory (remember: three to five items) on their goal(s) rather than understanding the interface.