This paper is based in part on evidence and insights found in various studies. These studies had as their objects to prove (or at least to demonstrate the plausibility of) either an analysis of a language or a theory of grammar. Their evidence, in addition to that which is new here, demonstrates the difficulty of using superficial relational information for semantic interpretation. It may be noted that, some years ago, Jespersen pointed out many such difficulties, though not with the orientation of relating syntax to semantics.