In a large-scale field experiment, data were collected on proximity, demographic and attitudinal similarities, and frequency of interaction of strangers and acquaintances in a situation of free choice. The results of the log linear analysis show that while previous acquaintance and proximity have a strong effect on frequency of interaction, demographic and attitudinal similarities have only a small effect. All these factors together do not account for the observed frequencies of interaction. Future models of early periods of interaction must utilize other factors in addition to those common in research on interpersonal attraction.