Thank you everyone for your answers. I think I've come to the conclusion that, as many people have said, 'usually but not always' is the way forward. Since I asked the question I've looked at a lot more original images of Parish Registers, and I'm finding that the context in which any one particular entry is found is actually the most helpful clue. Looking at how the priest or whoever made the record recorded all the deaths helps enormously. For instance, if you've got 'X, weaver', 'Y, Farmer', 'A, wife of B', 'Ann X', and then you get 'D, son of E', then it seems to be safe to assume the last one is a child. To make sure, I've been trying to match up probable child with a baptism... and it does turn out (mostly) that I can find a likely one in the preceding years. A lot of work, and impossible if you don't have sight of original docs, but worth the effort to be as sure as one can be!
Once again, thanks for your answers, folks!