The shift away from norm-referenced tests to authentic samples of child performance, such as writing and drawing samples, poses serious problems in organizing and analyzing assessments of child performance. This paper proposes a process utilizing the large-scale storage capacity of centralized mainframe computers and the flexibility and efficiency of table-top computers to conveniently access child observation records, efficiently perform relevant statistical analyses, and reliably produce teacher and parent reports of child performance. The process includes a longitudinal database of performance assessments; an organizational strategy relying on computers and relational databases; and the application of the Rasch measurement model. This paper begins by reviewing trends in child evaluation methods, longitudinal databases, and growth models. The bulk of the paper presents the methods and results of a pilot study applying the proposed process to measure the language behavior of a population of urban disadvantaged children in several early childhood programs. Data sources, including field observations, results of standardized tests, teacher assessments of child performance, and performance samples, are listed; and the structure of the Early Childhood Intervention Database is explained. The paper also explains procedures for population sampling, scale construction and vertical linking, and data analyses. Finally, results of the pilot study are presented and discussed. (AC)