> Yes.  Do you have a good reference for that?

Last I needed it, I looked it up in Encyclopedia of Statistical
Sciences (Johnson & Kotz). It's not really all that hot, because it
has the same sensitivity to nonnormality as Bartletts test. The entry
for "heterogeneity" refers to a paper by Harrison&McCabe, JASA 74
(1979) pp494-499. Our "recent issues" of JASA start at 1980, so I
didn't look it up...  

> >> var.test(x, y = NULL, group, v = 1, method,
> >> alternative = "two.sided", conf.level = 0.95)
> >> 
> >> (`v' is the null variance in the 1-sample case and the ratio of the vars
> >> in the 2-sample one ...)
> 
> > Looking good.
> 
> :-)

One thing that could be useful here and in other places would be to
allow the y~g notation that is used in stripplot() et al. (i.e.
"if(is.formula(x))...")
