I have been studying fair use seriously for over 25 years. I have read every English and U.S. fair use/fair dealing decision I could find, many of them many times over; I have read a voluminous amount of commentary on the subject; read and re-read the legislative history of Section 107 and spoke to those involved in that process; testified before Congress on the subject, and as a Congressional staffer, I was heavily involved in the 1992 amendments on fair use and wrote the House Judiciary Committee report on the subject; as a Policy Planning Advisor to the Register of Copyrights I went to Brussels to discuss with European Union officials how fair use works and how it might work with software reverse engineering; as a law professor I spent years teaching the subject; I wrote a treatise on fair use and many law review articles, gave many speeches on the subject and participated in many panels; for 20 years I have been discussing the subject with friends who are court of appeals judges and who have issued some of the great fair use opinions; in private practice, I have argued fair use cases at both the trial and appellate level, and counseled clients on whether uses are permissible.
Yet, despite all this concentrated attention, I doubt I understand fair use much better than when I was a law student. No doubt this mostly reflects my own shortcomings; those with greater abilities or self-confidence may feel differently. But aside from my own intellectual infirmities, there are reasons anyone might feel uneasy about coming to grips with the subject: it is ad hoc, fact-oriented, allegedly completely equitable in nature, and dependent on a shadowy weighing of vague factors, to say nothing of the luck of the draw in the decision maker. On top of that, there is the long common law history but since 1978, a statutory recognition in Section 107 (often misdescribed as a codification). The effect of this statutory recognition has been entirely negative, frequently disastrously so, as in the Sony presumptions and a tendency toward mechanistic, fair use by the numbers adjudicating. The Supreme Court's 2 Live Crew case helped loosen things up, but any student of fair use has to be aware of the history of the doctrine, and as well as the troughs and valleys of its treatment by the Supreme Court and by the lower courts.
Professor Barton Beebe has undertaken to understand how fair use has been decided on the ground, in a article to be published soon in the Penn Law Review, but available on his website, here. Entitled "An Empirical Study of U.S. Copyright Fair Use Opinions, 1978-2005," Professor Beebe canvasses how courts have treated each factor, the phenomenon of stampeding, and reversal rates. It is an important work that should be carefully studied. Whether it will help advocates in actual cases is a different issue, and is complicated by the strong divergence between the rhetoric employed by courts in reaching results and how they reach those results. Even when courts say, for example, that they are engaging in an analysis of whether defendant's use is "transformative," what any particular judge means by transformative may vary greatly. In the end, like beauty, fair use tends to be in the eye of the beholder.