The argument in this paper is that there exist a very large number of well-formed sentences which do not seem natural to a sensitive native speaker; therefore these sentences must violate some restrictions which are not among the criteria for well-formedness. It is important to examine what the further restrictions might be, for at least three reasons: a- there is no reason to believe that the restrictions are any less central in language structure than those for wellformedness; b - decisions about well-formedness are normally made on sentences in isolation, by people whose intuitions are shaped by experience of continuous text. As text study grows in importance, the concept of well-formedness of sentences in text may prove to be of value. This concept I want to label naturalness, for the time being; c. the concept of naturalness may be particularly useful to the learner of a language.