That are a number of axioms in copyright, by which I mean foundational principles. One is that infringement of the reproduction right occurs when there has been unauthorized copying of a material portion of a protected work. There are many different formulations of this principle, sometimes expressed as tests, and the principle is frequently broken down into smaller elements, like access and copying: there must be access to the work, copying from it, and then the copying must be substantial. Sometimes these divisions have importance for things like the use of expert witnesses: such witnesses are permitted generally to establish copying, but are not permitted at the substantial similarity stage. That at least has been the custom since Arnstein v. Porter, although it was breached in Computer Associates v. Altai, which permitted experts at the substantial similarity stage, or at least what purported to be that stage.
Sometimes, substantial similarity gets a confusing double duty, as when it is said there has to be access and substantial similarity. The late Professor Alan Latman proposed "probative similarity" to avoid the confusion inherent in such a formulation: there would be access, probative similarity to establish copying, and substantial similarity for the ultimate infringement. That's something of an improvement, but not a lot. All one really needs are access (since you can't copy something you don't have access to), copying (of anything, even unprotectible material like a phony address), and substantial similarity, this last term being merely a metaphor for the legal conclusion that plaintiff wins because defendant took too much protectible stuff.
The substantial similarity test has also been refined when plaintiff's work has a considerable amount of unprotectible material. Then, instead of comparing the two works from the perspective of the ordinary observer (the copyright equivalent of the reasonable person in general tort law), the "more discerning observer" is used, by which the same trier of fact manages the amazing trick of factoring out unprotectible material while still comparing the protectible elements as a whole. The purpose of this heightened scrutiny is two-fold: to avoid giving the copyright owner indirect control over unprotectible elements, while not ignoring any originality in a unified whole that nevertheless has unprotectible elements.
Having attempted to be a more discerning observer (and, I would like to think with success) in comparing two expensive rugs, I understand theoretically why it should be done, and perhaps how. But, is it really possible and if so how can be validate the exercise? Should such copyright determinations take into account social science research into how people perceive simlarities and dissimilarities? Do people generally perceive similarities more readily or dissimilarities? Are people influenced by how long they look at the items? What if they are told to focus on or not to focus on certain things? Are there some types of similarities/dissimilarities that people perceive more easily than others? And, finally (or first) what does it mean to even say something is "similar"?
Professor Arthur Markman at the University of Texas has been doing very interesting work on similarity and cognition, both through his writings and through his Similarity and Cognition lab. Right now, copyright law exists at the crude level of Potter Stewart's "I know it when I see it." He may have known it, but are the rest of us sure, especially when wearing our more discerning observer goggles? Is it possible to put copyright on an empirically more sound basis or is the Potter Stewart approach good enough for government work, as the saying goes?