The authors address some of the methodological problems found in earlier investigations of the relationship between social networks and mortality. Of particular concern was the use of rudimentary measures of social interaction in previous work. Utilizing nineteen social network variables, the authors examined 155 elderly residents of midtown Manhattan single-room occupancy hotels. On three-year follow-up, twenty-eight persons had died. Discriminant function analysis indicated that ten of nineteen network variables were relatively strong discriminators between survivors and non-survivors (i.e., standardized coefficient of .20 or greater). The network variables cut across all categories of social interaction, thereby revealing why, despite the few network variables used in previous studies, nearly all had been significant.