First, though, you have to suffer one bit of theory — namely the definition of mental models. A mental model is what the user believes about the system at hand.
Note the two important elements of this definition:
A mental model is based on belief, not facts: that is, it's a model of what users know (or think they know) about a system such as your website. Hopefully, users' thinking is closely related to reality because they base their predictions about the system on their mental models and thus plan their future actions based on how that model predicts the appropriate course. It's a prime goal for designers to make the user interface communicate the system's basic nature well enough that users form reasonably accurate (and thus useful) mental models.
Individual users each have their own mental model. A mental model is internal to each user's brain, and different users might construct different mental models of the same user interface. Further, one of usability's big dilemmas is the common gap between designers' and users' mental models. Because designers know too much, they form wonderful mental models of their own creations, leading them to believe that each feature is easy to understand. Users' mental models of the UI are likely to be somewhat more deficient, making it more likely for people to make mistakes and find the design much more difficult to use.
Finally, mental models are in flux exactly because they're embedded in a brain rather than fixed in an external medium. Additional experience with the system can obviously change the model, but users might also update their mental models based on stimuli from elsewhere, such as talking to other users or even applying lessons from other systems.
Remember Jakob's Law of the Internet User Experience: Users spend most of their time on websites other than yours. Thus a big part of customers' mental models of your site will be influenced by information gleaned from other sites. People expect websites to act alike.
Many of the usability problems we observe in studies stem from users having mixed-up mental models that confuse different parts of the system.
For example, the word "Google" is usually the top query at other search engines, and words like "Yahoo" and "Bing" score high on Google. Why, oh why, do people search for a website if they already know its name? Why not just type, say, www.bing.com into the URL field?
The reason is that many users have never formed an accurate model of how the "type-in boxes" on their screen function. When they type stuff into a box, they sometimes get where they want to go. What to type where and exactly how each type-in box functions, however, are often beyond their ken.