predict.lme() gives missings for new subjects

Hi,

why does predict.lme() give missings in the "predict.Subject" column for Subjects which are not in the original training set?

library(nlme)
example(predict.lme)
## c.f. Subject=="F30" has missing values.

How can I predict/impute values for such "new" Subjects based on the fitted model?

Thanks in advance,

Will