In most organisations, accurate data are some of their most precious assets. Many empirical studies, however, have sl10wn that many organisations' databases contain errors (Neter and Loebbecke 1975; Ki111ley 1979, Ramage et al. 1979; Johnston et al. 1981; Hylas and Ashton 1982; Morey 1982; Ham et al. 1985; Willingham and Wright 1985; Kreutzfeldt and Wallace 1986; Laudon 1986; Icerman and Hillison 1991). Data errors entering the databases can be reduced by implementing various input validation controls. Higher levels of normalisation can also help reduce the number of errors entering the database. Even ill well maintained databases, however, data errors will still accumulate. The accuracy and quality of data in databases can be improved by periodically querying the databases to locate and correct data errors.
High levels of normalisation may facilitate updates, but adversely affect users' ability to perform queries. The results of this study showed that increased levels of normalisation lowered the effectiveness and efficiency of queries, however, unnormalised data structures had an even more adverse effect on user queries. Data structures normalised to first normal form (1NF) appear to be most effective and efficient for user queries, e.g., for finding existing data errors.