This document presents a study of natural language understanding of computer programs. In the study, the performance of IRUS, a natural language query system designed to interface with a database, was compared with the performance of preschool and early elementary school children in answering questions about a specific database. Questions used in a test of the IRUS system were classified into semantic and syntactic categories, and a set of specifications for measuring human ability to understand them was developed. Similar questions were then developed and asked of preschool and early elementary children. The children either read the questions themselves or were asked them by an adult, and responded orally or by pointing to a picture on the computer screen, the database developed for the test. Students were asked to explain their responses to validate their understanding of the questions. Characteristics of the students' interactions with the database were recorded and analyzed, and are reported. Further research on higher and lower levels of performance and refinements of the test are planned. (MSE)