The Non-omission of Nonfinite Be

It has long been known that children learning English optionally omit finite forms of the verb be (both copula and auxiliary). What makes be omissions possible? A classic answer is that be is semantically empty, hence a good choice to omit under performance-related pressures. What would this hypothesis lead us to expect about the infinitive form of be? In terms of semantic vacuity, nonfinite be is an even better choice for omission than finite be , since it does not carry tense or agreement information—most such deletions would be completely recoverable. Thus, the semantic vacuity hypothesis would lead us to expect omission of nonfinite be to be at least as frequent as omission of finite be . This is contrasted with the suggestion that be omission be incorporated into the Root/Optional Infinitive phenomenon. The latter finiteness hypothesis makes different predictions from the semantic vacuity hypothesis with respect to the relative rate of nonfinite be omission. The finding is that, in each relevant transcript, omission of finite be is attested, use of nonfinite be is attested, but there are no instances of omission of nonfinite be , contra the prediction of the semantic vacuity hypothesis. I develop an analysis within the Agreement/Tense Omission Model of the underspecification of Infl (cf. Schütze 1997). I claim that finite forms of be in (adult and child) English are fused V+I heads, in the sense of Halle & Marantz’s (1993) Distributed Morphology. Their locus for vocabulary insertion has values for person/number, tense, and lexical category. The fused vocabulary items cannot be inserted in a syntactic structure in which INFL features have been underspecified. Overt be arises only when both AgrS and Tense are fully specified. Null be , i.e. Ø, the default member of the paradigm, arises from underspecification of Tense and/or underspecification of AgrS.