The present investigation replicates O’Connor and Rosenblood’s (1996) experience sampling study (ESM) of the homeostatic regulation of social interaction and addresses the statistical limitations of the original study. The original study developed the homeostatic Social Affiliation Model (SAM). SAM proposes that future interaction is a function of past desire to be alone, social interaction state, and social satiation. Using ESM, the original study supported the proposed model. There were several statistical problems with the original study. ESM involves intensive longitudinal sampling, which generates multiple observations nested within individual. When responses are nested within person, it violates the assumption of independence of observations. By disaggregating observations from participants, the original study violated this assumption and failed to estimate within-person effects of desire to interact on future social interaction. Multilevel modeling (MLM) accounts for the non-independence of observations and can estimate within-person effects. The present replication study used surveyed community (N = 62) and student (N = 54) participants at five time points for five consecutive days. MLM results (n = 2,747) indicated that desire to be alone reduces future likelihood of social interaction, replicating the original study. Individuals’ optimal social interaction state changed from no-contact desired to contact desired over the day, but results did not support the original study’s claim regarding social satiation. Future research will require new measures of satiation, threshold, and desire to examine equilibrium effects proposed SAM. To demonstrate homoeostatic regulation the time window of social satiation and appropriate rates of experience sampling to must be established.