Since inconsistency is basically "not the same rule everywhere", it typically shows in (good) documentation. Indeed:
http://www.opengroup.org/onlinepubs/009695399/utilities/xcu_chap02.html#tag_02_06 2.6 Word Expansions Not all expansions are performed on every word, as explained in the following sections. The above sentence leaves open the scary possibility of a large number of combinations and special cases, but of course it is not that badly inconsistent. Focusing back on field splitting, both the opengroup and the (similar) bash manual have the same documentation logic: they apparently never explicit where field splitting is performed (only before/after which expansions it is performed when it is). So to understand where field splitting is performed the reader must assume that it is performed everywhere, except when it is exceptionally missing from an exhaustive list of expansions ("case", assignment,...). Unclear and inconvenient. Of course the best alternative is the one mentioned by Greg: assume that field splitting is consistently performed everywhere and keep quoting everything all the time. It is a good habit anyway and avoids bugs due to refactoring.
