Capture-recapture methods were originally developed for use in fisheries and wildlife biology for estimating the size of a closed (fixed) population. Later, they were adapted for use in connection with human populations. Applications to epidemiology came more recently, especially those involving multiple-record systems. Over the past 25 years, efforts have been made to link this methodology to the more traditional literature on loglinear models for contingency tables. This talk focuses on methods associated with data from multiple sources or lists and illustrates how loglinear and related models (e.g., the Rasch model from educational testing) can allow for estimation in the presence of both dependence among lists and heterogeneity among individuals or units in the population. We apply the methods to a number of examples including (1) the ascertainment of diabetes, (2) adjusting the U.S. decennial census counts, and (3) estimating the size of the WWW. The latter problem requires some special statistical features for scaling up analyses.