The English verb fail, when followed by to + infinitive, has two senses. One presupposes active attempt on the part of the subject, as in (1); here fail is a negative implicative verb in the sense of Karttunen (1971). The other, as in (2a), lacks this presupposition of active attempt and comes close in meaning to simple negation, cf. (2b):

(1)       The mountaineers (tried and) failed to reach the summit.
(2)       a.        It fails to surprise me that he won an Oscar.
            b.        It does not surprise me that he won an Oscar.

The paper will consider the best analysis of the second sense of fail to, using the framework of Functional Discourse Grammar, and will be based on an examination of all 12,670 occurrences of fail to (and its conjugated forms) and failure to (the corresponding nominalization) in the 100m-word British National Corpus. 

The paper will first examine the evidence that fail to in the second sense is a subject-raising verb similar to seem. This analysis will be put aside, however, in favour of an analysis of the relevant use of fail to as a periphrastic negative not involving a lexical verb. Close analysis of the corpus will reveal that fail to, in both its uses, is subject to a number of collocational restrictions pertaining to such semantic concepts as dynamicity and telicity and to the rather more impressionistic (but cognitively real) notion of semantic prosody. This will lead into a case study of the 177 instances in the corpus fail to be, which reveals several additional facts that are in need of clarification. 

The FDG-theoretical section of the paper will pursue the notion that the first sense of fail reflects a lexical verb with a Configurational Property as its complement. The second, periphrastic sense involves a Verb Word fail at the Morphosyntactic Level that corresponds to a Negative operator at the Representational Level. Drawing on facts about subordination and secondary predication, the claim will be defended that it is again a Configurational Property to which this Negative operator applies.

The paper moves on to consider the use of fail to in litotes (approx. 470 occurrences in the corpus) as entailing two loci for the Negative operator. The nominalization failure, in its periphrastic use, is then argued to arise from the application of the Negative operator to a Lexical Property. The paper ends with some reflections on the possible motivations for the grammaticalization of the lexical verb fail as a periphrastic negative, with reference to Givón’s (2005: 167-168) view of negation as not purely semantic but also as having pragmatic aspects. This will bring the Interpersonal Level into play and the relation between the Grammatical Component and the Conceptual and Contextual Components of the overall model of the natural language user.