It's interesting how the PLT team dealt with this teaching problem in "How to Design Programs". The first part basically shows how to write simple Java-like code with Scheme syntax, using symbols and fixed records, and including tests as part of a complete program. Then in the later sections, lists and the more dynamic handling of data are introduced gradually. Aren't PLT Scheme structures way more complicated than lists? Yes, but since simple errors show up sooner with a struct, it seems to be a better way to teach brand-new programming students how to work with compound data.

>Minor errors that in bulkier languages are caught by the compiler, become logic errors in a Lisp

doesn't that apply to any less static -> more static typed language comparison?