The experimental and field studies surveyed by Latane and Nida (1981) establish an inverse relationship between the probability that a person is helped and the size of the group of potential helpers. Harrington (2001) attempts to account for this phenomenon using a 'rational choice' model in which agents play Nash strategies. In Harrington's model the probability that anyone helps a person in trouble decreases as the number of potential helpers increases. Also, the probability that a victim receives help is bounded below and away from zero. This second implication of the model is somewhat at variance with the analysis of Hochman and Rogers (1969), Bergstrom (1970), Nakayama (1980) and Arrow (1981) where, if there are two rich people (potential helpers) and one poor person (victim), then 'helping' has the characteristics of a pure public good, with predictable consequences for the equilibrium level of help offered. The principle purpose of this paper is to extend the model of Harrington (2001) to include the Hochman and Rogers-Bergstrom-Nakayama-Arrow result as a special case.
