ABSTRACT
When designing content analyses, social scientists construct coding rules to define the measurement categories that will be used in their investigation of media message patterns. The goal of coding rules is to fully specify the task of determining which messages fit into which content categories. When the coding task is fully specified, coders are able to make decisions that are consistent throughout the sample and over time. This results in high reliability of a dataset that can then be considered to be systematically generated and hence scientific. However, when the coding rules do not work well, the degree of consistency in classifying messages across coders, across types of content, and over time is low; thus reliability and hence validity are low, and the resulting data must be regarded as having little scientific worth. Constructing good coding rules that fully specify the coding task is fundamental to a scientific content analysis.
