This paper discusses the notion of grammatical well-formedness in the light of certain optimality approaches to syntactic phenomena (e.g.,Pesetsky 1998; Grimshaw 1997; Grimshaw & Samek-Lodovici 1995; Costa 1998). Such approaches adhere to assumptions that lead to the following theorem: a linguistic representation may violate a rammatical constraint and still be well-formed if and only if all other alternative candidates also violate some grammatical constraint. The point the paper makes is: if well-formedness is the theoretical correlate of full acceptability, this theorem is in trouble. The arguments come from the analysis of two marked constructions of English: logophoric reflexives (Reinhart & Reuland, 1993) and peculiar passives (Davison, 1980). The paper argues that these phenomena arise as a result of a Gricean implicature triggered by violations of grammatical constraints, and that conversational implicatures cannot be characterized as the result of competition among grammatical constraints. This paper discusses the notion of grammatical well-formedness in the light of certain optimality approaches to syntactic phenomena (e.g.,Pesetsky 1998; Grimshaw 1997; Grimshaw & Samek-Lodovici 1995; Costa 1998). Such approaches adhere to assumptions that lead to the following theorem: a linguistic representation may violate a rammatical constraint and still be well-formed if and only if all other alternative candidates also violate some grammatical constraint. The point the paper makes is: if well-formedness is the theoretical correlate of full acceptability, this theorem is in trouble. The arguments come from the analysis of two marked constructions of English: logophoric reflexives (Reinhart & Reuland, 1993) and peculiar passives (Davison, 1980). The paper argues that these phenomena arise as a result of a Gricean implicature triggered by violations of grammatical constraints, and that conversational implicatures cannot be characterized as the result of competition among grammatical constraints.