The article reports on a study performed during 1987 that tested the belief among catalogers that the quality of original cataloging in online computer library catalog (OCLC) and research libraries information networks (RLIN) differs markedly, and determines whether some kinds of errors occurred with greater frequency in member-contributed original cataloging. In the early years of these two major U.S. bibliographic networks, Library of Congress machine-readable cataloging (LC MARC records) formed the nucleus and large majority of records of both databases. Although LC MARC records are not error-free, they were then and are today the highest standard of quality in the U.S., leaving the networks to assume that most of their cataloging was the best quality available. Using MARC records as their authority, both databases required member libraries to follow LC practice. The proportion of records in OCLC and RLIN contributed by member libraries has grown steadily, while the proportion of LC MARC records declined. The two networks took different approaches to quality control. They had different objectives, pricing structures, cataloging needs and capabilities, and historic developments. At the risk of oversimplifying, it is not unfair to say that OCLC emphasized building its database, encouraging members to contribute new cataloging without applying strict controls on its accuracy and fullness, while RLIN assigned a high priority to cataloging quality, encouraging accuracy and fullness in records even if fewer of them were contributed as a result.