John Howard Brown forwarded a delightful story about path-dependence and railway track gauge that has been passed around the internet, particularly among people in the Department of Defense and elsewhere who bemoan outdated or bureaucratically oppressive "milspecs"--of which gauge is cited as an ancient (and perhaps ridiculous) example.
Well, I've had an opportunity to look into the issue for my dissertation (and forthcoming book), so perhaps I can clarify the record. As much as I like the idea of path-dependence, and although I believe that I have shown its applicability to the modern history of railway gauge, it seems unlikely that we can carry the process back as far as the story suggests.
American railways were not built by ex-patriate Brits, but (almost exclusively) by native American engineers who copied British practice. This practice they copied was that of the Liverpool & Manchester Railway (opened in 1830), built by mine-works engineer George Stephenson. Stephenson indeed copied the gauge with which he had previous experience in the mines, but this was originally the not quite so "exceedingly odd" measure of 4'8". The extra half inch was added during construction of the L&M in order to allow a little more leeway between rails and wheel flanges. There is some evidence that the original rails were often 2" wide, indicating a width of track including the rails of exactly 5'0"--still less an "exceedingly odd" measure.
Mining tramways differed substantially in width, ranging mostly between 3'0" and 4'6" in southern England and Wales. The 4'8" of northern England was an outlier, and it is could be regarded as accidental that Stephenson happened to have had a history with that gauge. It appears true that mining ore carts were about the same width as road wagons, but the width varied by region. It is plausible that the width of wagons was fitted to road ruts, although ruts at narrow city gates might have mattered more than ruts on open roads. The main "evidence" for carrying the story back to Roman chariots, by the way, comes not from any study of the history of road ruts but from consideration of ancient "groove-ways"-- essentially permanent stone "ruts", a practical form of improved road surface at the time. It is true that one or two of these (NOT in Britain) happen to have roughly the same "gauge" as modern railways--within a broad band of wheel widths that would fit the grooves. However, others are of different widths. So part of the story as told may be consistent with the evidence, but it is hardly proved by it. (But if anyone knows of better evidence, particularly any actual research on ruts, let me know.)
By the way--4'8.5" (1435 mm.) is the standard gauge in North America, most of Europe (not in Iberia or former Russian and Soviet empires), and parts of South America, Asia, and Australia. All told, nearly 60 percent of world route length. The L&M railway had a strong demonstration effect in Continental Europe as well as the U.S., and Stephenson-trained engineers also aided the gauge's diffusion in Britain and the Continent-- but not in North America.
Again by the way--some American engineers copied Stephenson's practice only approximately, doubtless in order not to deal with "exceedingly odd" dimensions. Thus they introduced gauges of 4'9", 4'10", and 5'0". The latter two choices led to some difficulties in later integrating the continental railway network.
University of Munich