A static expression like a book will never generate another expression... it's a book, and nothing more. Dynamic expressions can lead much more interesting lives, where expressions can themselves generate more expressions, without the human intervention that's implied whenever one static expression is used in the creation of another static expression.
This happens all the time in the real world, but I think one of the clearest examples of this is the Gallery of Random Art page. The author of the Random Art page, Andrej Bauer, created the Random Art program. Considering the sorry state of post-modern Art, this qualifies as Art... in my humble opinion much more so then many other things labelled "art". Yet nobody sees the Random Art program, only the random art results. Some of these results are quite good, as shown in the archive on that page, although most are not; if they were a more normal desktop resolution many of them would make great desktop backgrounds.
So, here's the question: Who owns the copyright on those expressions? Technically, since Andrej Bauer wrote the program, he is the only human candidate to hold the copyright, so one may probably safely assume they default to him. But in a very real sense he did not create the thousands of art pieces output by the program. Furthermore, you used to be able to get a screen saver for Windows NT that also generates random art. If you grab one of the results of that program's execution, who would hold the copyright? In this case, modern law would say that the owner of the computer running the program holds the copyright. Yet in a way, these expressions are springing forth from nothing at all, with no distinct author. This is an extreme case, where the user has absolutely no input into the process at all.
Perhaps one could argue the works are not copyrightable, as there is no creativity in the pictures, only the program. (Of course the program's copyright is held by the author.) But those are awfully complex pictures to say that they have no more creativity to them then the phone book, and had a human an absolutely identical expression, we would say they are creative works deserving of copyright protection.
The ability of an expression to generate another expression makes it really hard to draw the line of where one begins and another ends, and if we can't even define what an expression is without ambiguity, the whole copyright system comes crashing down.
I picked Random Art as an example because of its extreme nature; the Random Art program essentially accepts nothing as input and creates output. In the real world, most expressions that generate expressions, such as Microsoft's Hotmail programs, take other expressions and do something with them.
A Deeper Philosophical Issue
I'd like to highlight one aspect of the above that is for the next generation to work out. What is creativity? As computers continue getting more powerful, it will get increasingly difficult to determine by examination whether a work was created by a human or a computer. Can the computer's work be said to pass the creativity test for copyright? If so, then why doesn't the computer hold the copyright?
Consider the Random Art program. Like I said, if a human were creating those works none of us would think twice about granting the human copyright over the works. When we refer to a work as "creative", are we referring to an intrinsic property of the work, such that no matter how it is created it is "creative", or a property conferred upon the work by how it is made? I can imagine arguments in support of both answers.
One assumption that copyright law is based on is the assumption that only humans are doing the creating. It won't affect my analysis, which is intended for the current time frame, not the future, but it will be an issue soon enough, and is interesting to ponder.