<p>Bipartite reciprocal phrases are common in Indo-European, Indo-Aryan, and Semitic languages. When occurring with a case particle (K) or a preposition (P), K/P intervenes between the two units of the reciprocal phrase, producing an otherwise exceptional K/P-medial KP/PP. While diachronic descriptions successfully trace the origin of the two units to separate constituents that have gradually become closer, they fall short of explaining the stability of the K/P-medial outcome. Based on a detailed study of the Hebrew reciprocal construction, I argue that its two components are generated as sisters but cannot persist as such because they yield an illicit point of symmetry – an unlabeled phrase. The first member thus raises past K/P, breaking the symmetry, thereby allowing the complement of K/P to be labeled. The analysis is supported by data from intra- and crosslinguistic variation and predicts systematic correlations between the degree of symmetry between the two units of the reciprocal phrase and their separability.</p>