Lexical Subjects

The occurrence, frequency, and function of lexical subjects have not attracted a great deal of attention in the linguistic literature, largely because few have noticed anything strange about such subjects. The Subject-Verb-Object (SVO) word order of English has often led linguists to assume that transitive sentences with lexically headed noun phrases in both subject and object positions are the norm in English. Thus, Sapir (1921) describes (1) as a "typical English sentence" (see Lambrecht 1987 for a discussion of Sapir's example):

Most native speakers would, however, find it very hard to place this sentence in context. This is in part due to the use of the simple present tense (which in current English is not usable for reports of events ongoing at the time of speaking) but in larger part due to the use of the indefinite lexical subject a farmer. While a number of discourse linguistics has noticed that lexical subjects are rare in English conversation (Givon 1983, Du Bois 1987, Chafe 1994, Lambrecht 1994), the Lexical Subjects project is the first large-scale, corpus-based study of lexical subjects in naturally occurring English discourse. Our research suggests that when speakers in conversation use lexical subjects they do so as part of an optimization strategy. This strategy allows for constrained violations of what Lambrecht (1994) refers to as the Principle of Separation of Role and Reference (PSRR): Do not introduce an entity and predicate something of it in the same clause. Such violations subserve effort conservation (Horn 1984), since the use of a lexical, as against pronominal, subject allows the referent-introduction and property-predication functions, which are otherwise distributed across two clauses, to be compressed into one clause. Violations of the PSRR are tolerated insofar as the discourse-new entity in subject position is anchored (e.g., by a frame-based or kinship relation) to something in the prior context.

This study of the occurrence, frequency, and function of lexical subjects was generously supported by the National Science Foundation (POWRE 9805829), awarded to Laura A. Michaelis of the University of Colorado at Boulder. The database used in this study is the Switchboard corpus of American English telephone conversations.