Dave Nelson wrote:

IOW, you should not plan your inventory of cars visible ON THE LAYOUT according to the distribution hypothesis, but your COMPLETE roster. And you must vary what's on the layout over time.

Here I think Dave makes a vital point. The cars you see when you visit a layout multiple times should not be the same cars every time. Of course the most eye-catching cars are the real offenders: the CofG "football" scheme, that huge transformer on an FD flat car, the Chateau Martin burgundy wine car, even the yellow MKT box car. But the unusual and small railroads are in this category too, even though to a lesser extent: the LS&I, the Rutland, West India Fruit, Buffalo Creek, Portland Terminal. They are absolutely entitled to be there, but just not every time or even every other time.
For me, the essential part of the Gilbert/Nelson hypothesis is that free-running cars from farther-away railroads are NOT less likely than nearer ones. The frequency-inverse-with-distance rule is an incorrect idea that dates back at least to Boomer Pete's book in the 1940s, and it needs a decent burial.

Tony Thompson Editor, Signature Press, Berkeley, CA
2906 Forest Ave., Berkeley, CA 94705 www.signaturepress.com
(510) 540-6538; fax, (510) 540-1937; e-mail, thompson@...
Publishers of books on railroad history