Humans often need to make inferences without knowing what individual objects are present, or whether an object they're currently observing is the same as one they observed before. Consider the basic task of telling whether the object that emerges from one side of a tunnel is the same one that entered the other side a moment earlier, or the more elaborate problem of figuring out whether the "Michelle Stone" you met at a conference is the same as the "M. Stone" cited in a paper you're reading. This lecture will extend the idea of relational probability models to handle such scenarios. We will discuss a modeling language called Bayesian Logic, or BLOG, that can define probability distributions over "possible worlds" with varying sets of objects. A BLOG model describes a hypothetical generating process that builds up a world step by step: some steps specify relations among objects, and others determine the existence of objects in the world. We will also show how to do inference on a BLOG model by running a Markov chain over possible worlds with varying relational structures.