According to traditional wisdom, reciprocal predicates can only occur with plural subjects. This is assumed either because the reciprocal predicates in question are constructed by means of a reciprocal anaphor, which is considered as being inherently plural and hence requiring a plural antecedent, or, if there is no binding requirement, the following principle of argument mapping is implicitly assumed: all participants of a reciprocal situation need an overt realization by the same highest syntactic argument. Since a reciprocal relation minimally involves the existence of two participants, and since (in the languages considered so far) the highest syntactic argument is the subject, this mapping principle leads to the idea that the subjects of reciprocal predicates should be confined to plural or conjoined phrases. In this paper, it will be shown that this principle turns out to be unrealistically strong, once real discourse data are considered, in particular from a cross-linguistic perspective. Under certain structural and pragmatic conditions, participants of reciprocal relations may be backgrounded and also suppressed, with the result that, in the second case, they will lack an overt realization altogether. It will be argued that there is a typological correlation between the following three phenomena: discontinuous reciprocals (where one participant is backgrounded and hence realized as an oblique phrase), “true” singular subject reciprocals (where only one participant is realized overtly, while the other is suppressed), and plural subject reciprocals, admitting the interpretation that each individual among the subject’s referents participates in a reciprocal relation with some other (unknown or arbitrary) individual that is, however, suppressed, i.e. not referred to by the subject phrase or any other phrase in the sentence. Data from four languages are presented: Hungarian, German, (Modern) Greek and Serbian/Croatian. In general, a cross-linguistic approach will be favored which considers differences and similarities at all relevant levels of description, e.g. discourse pragmatics, verbal aspect, lexical-semantic fields, interfering effects of ambiguity, etc. in addition to structural constraints in marking reciprocity.