An examination of the problem of understanding language by an antifact, from a four-week symposium held at RAND in the summer of 1963 (one of a series of symposia on mathematical biology). Understanding has both linguistic and nonlinguistic aspects. An organism (natural or artificial) may be said to understand adequately some feature of its environment if its internal organizing system can take adequate account of it (without necessarily being able to express the understanding linguistically). In the linguistic context, a new set of capacities is needed: (1) the development and use of internal models as such; (2) the distinction between linguistic and direct experiential input; and (3) the capacity to learn.For understanding, linguistic or not, the information system must have something equivalent to an internal model of what is understood. To understand the difference between a description and a mere symptom of a state of affairs, the information system must have something equivalent to an internal model of the communication process. Four levels of demands for language-handling ability are listed and explained.