Abstract
Previous studies relate the lack of adequate metadata training with errors often found in metadata records. Due to lack of empirical evidence, such a relationship is not well understood. In this study, metadata records created by individuals who have not received any metadata training are compared to records created by individuals who have received metadata training. The aim is to investigate the relationship that exists between training and errors in metadata records. Using a 2×2×4 mixed experimental group design, 120 participants were asked to use Dublin Core to create records for a web page, an image, a document, and a video. Records were analyzed on the basis of the total number of errors and on the basis of four kinds of errors: incorrect value, incorrect element, inconsistent value, and incomplete information. Results indicate that depending on the kind of error investigated and the resource for which metadata is created, training may or may not have an effect on errors in metadata records. Implications of observations made and limitations of the experimental design are discussed in order to situate the relevance of the present study.