The distinction between formal and casual English as reflected in the lexicon, in phonology, and in syntax is studied. Formality is treated as separate from other categorizations of language such as geographical origin of the speaker, social class of the participants, or specific context of discourse. The study was restricted mainly to the use of two informats. A categorization of lexical entries and rules into three stylistic levels--formal, netural, and casual--is rejected as insufficiently detailed. Sentences exhibiting stylistically discordant elements are then used to get at fine distinctions in level. A more complex gradation model by which each linguistic element is assigned a value between -10 and +10 is outlined. The degree of stylistic deviance of a sentence is then calculated as the difference between the values of the most extreme elements in it. Within the framework given here, the linguist must determine which phonological and syntactic rules and lexical entries are stylistically marked and assign them values in a way that predicts the correct ordering of both elements and discords. Ways in which the more complex model could fail to be adequate are also discussed. (PP)