This chapter explicates the requirement in person–reference for balancing demands for recognition, minimalization, explicitness and indirection. This is illustrated with reference to data from repair of failures of person–reference within a particular linguistic/cultural context, namely casual interaction among Rossel Islanders. Rossel Island (PNG) offers a ‘natural experiment’ for studying aspects of person reference, because of a number of special properties: 1. It is a closed universe of 4000 souls, sharing one kinship network, so in principle anyone could be recognizable from a reference. As a result no (complex) descriptions (cf. ‘ the author of Waverly’) are employed. 2. Names, however, are never uniquely referring, since they are drawn from a fixed pool. They are only used for about 25% of initial references, another 25% of initial references being done by kinship triangulation (‘that man’s father–in–law’). Nearly 50% of initial references are semantically underspecified or vague (e.g. ‘that girl’). 3. There are systematic motivations for oblique reference, e.g. kinship–based taboos and other constraints, which partly account for the underspecified references. The ‘natural experiment’ thus reveals some gneral lessons about how person–reference requires optimizing multiple conflicting constraints. Comparison with Sacks and Schegloff’s (1979) treatment of English person reference suggests a way to tease apart the universal and the culturally–particular.