I've found 'pass-through' functions to be useful in the survival library. For example cluster <- function(x) x This allows the coxph routine to use the specials attribute of the terms structure to spot variables that should be treated specially, with needing to create a special class. I'm updating the expected survival routines to be more open minded wrt dates, i.e., no longer wedded to my old 'date' class, and the following would work very nicely ratetable <- function(...) data.frame(...) Then later steps in the processing can make sure that variables in a rate table which are known to be dates, like 'year' in the US death tables, can take special care when matching them to the user's data. Nice idea, except that it doesn't work: model.frame appears to be unhappy with any multi-column object except a matrix. Any suggestions to work around this? Terry Therneau