Summary

Understanding what sentence fragments requires a detailed investigation of their properties. Relevant studies have classified fragments into four types according to the use of the copular verb shi ‘be’ with fragments: the occurrence of the sentence-final particle ne, syntactic categories allowed, island effects, P-stranding, and connectivity effects. It is shown that the distinction of fragments into different types is not necessary. The wider range of data considered fails to convincingly support the need for distinction, and a unified analysis of all fragments should be pursued. Three logical possibilities for a unified approach to sentence fragments are evaluated: (a) fragment as a result of deleting all but the fragment of a sentence (movement + deletion approach), (b) fragment base-generated as [pro + copular verb + fragment], and (c) fragment base-generated as a fragment. The first two options face challenges. The last option, even though seemingly a more adequate analysis for the range of facts, requires connectivity effects to be analyzed in semantic terms.