Objective: The feeling of loneliness is inherent in the human condition, and is considered a public health issue that is directly related to welfare. The aim of this study was to identify the sociodemographic predictor variables of loneliness in the elderly. Methodology: Cross-sectional study. The sample consisted of 869 adults aged 55 and older who attended senior citizen groups in Medellin in 2010. Results: The model explained the 21.6% of variance in the scale scores for loneliness. The main predictor variables for family loneliness was the number of people whom they live with; for marital and social loneliness was marital status; and for crisis of adaptation was socioeconomic status. Conclusions: The demographic variable that best predicts loneliness is marital status, appearing in different types of loneliness and global loneliness.