logistic regression (binary data, upper asymptote not eq. 1.0)

[awp@mathstat.yorku.ca]

Using S+ v.5 on UNIX, I would like to perform a logistic regression of binary 
data. these raw data (ones and zeros) relate to a probability (p) which 
depends on the explanatory variable (x). Without going into too much detail, 
the evidence is that p(0)=0 but the p(infinity) is much less than 1.0 (say, 
0.4). Can/should I constrain the fit to have an upper asymptote that is not 
1.0 ? If this is a valid method, how is it done? Many thanks, Andrew Park, 
Dept. Math & Stat, York Uni, Toronto, Canada.