Answering the Call for a Standard Reliability Measure for Coding Data

In content analysis and similar methods, data are typically generated by trained human observers who record or transcribe textual, pictorial, or audible matter in terms suitable for analysis. Conclusions from such data can be trusted only after demonstrating their reliability. Unfortunately, the content analysis literature is full of proposals for so-called reliability coefficients, leaving investigators easily confused, not knowing which to choose. After describing the criteria for a good measure of reliability, we propose Krippendorff's alpha as the standard reliability measure. It is general in that it can be used regardless of the number of observers, levels of measurement, sample sizes, and presence or…