It is interesting that in the linked document exercises 4 and 10 ask you to implement addition for those kinds of numerals. Can you think of a way doing this without checking for zero? (I know that sometimes there are unsolvable exercises in the literature, but the author gives this routinely as an assignment it seems [1].)

[1] http://wcl.cs.rpi.edu/pdcs/slides/Chapter2-LambdaCalculus.pp...

I sent an email to the professor that taught the course in 2011 asking for clarification...