Chinese anaphors can be divided into two semantic types: One (the X-benshen 'X-self' anaphor, in which X is either ziji 'self' or pronoun-ziji 'pronoun-self') requires pure identity with its antecedent; the other (the non-X-benshen 'non-X-self' anaphor) allows near identity with its antecedent. Pure identity shown by the anaphor ziji-benshen 'self-self' can be derivable from the semantic composition of the near reflexive function of the morpheme ziji 'self', the focus function of the morpheme benshen 'self' and the operator status of ziji-benshen 'self-self' while near identity (or near reflexivity) shown by the non-X-benshen 'non-X-self' anaphor is due to its being "a pronoun in coreference." In other words, in Chinese pure identity is near reflexivity plus a focus marker which picks out the best representation of the antecedent that happens to be the actual person. Thus, the X-benshen 'X-self' anaphor should not be considered a pure anaphor without content. Typologically, there are two ways for human languages to get pure identity: One is by using an anaphor without content; the other is by using a focus marker that picks out the best representation of the antecedent which happens to be the actual person. The typological difference in establishing pure identity provides an answer for the long standing question of why the notion of coargument-hood is often adopted by linguists in defining binding conditions in languages which show an antilocality effect but seldom in Chinese. The distribution and coindexation of all Chinese anaphors, either the X-benshen 'X-self' or the non-X-benshen 'non-X-self' anaphor, are determined by one single syntactic condition, namely, the traditional binding theory, regardless of whether they are inside or outside the coargument domain. In contrast, in languages which show an antilocality effect, the binding theory allows both pure and near anaphors within the coargument domain but Condition R filters out one of them in the absence of lexical/morphological reflexivity; however, outside the coargument domain, the binding theory itself governs the distribution and coindexation of all anaphors.