A considerable amount of work has been done in the field of verifying that computer programs fit their specifications. However, providing correct formal specifications is itself a fairly difficult and tedious task. Typically, people accept informal and incomplete descriptions of programs. These are used for "finding your way around" a large program, perhaps to identify troublesome parts. If formal correctness techniques are used at all, it will only be once suspect areas have been identified informally. This report presents a computer system that compares a LISP program with a description of the program written in a subset of English. The system interprets the English text; compares its interpretation with the code, using techniques concerned with superficial characteristics of the interpretation; investigates its comparison in detail using more formal techniques; and finally produces a documented version of the program, a brief report on how the specifications are implemented, and a database of detailed assertions about how the program and the specifications correspond.