Cognitive task analysis (CTA) captures unobservable cognitive processes, decisions and judgments of expert performance. Over 100 different CTA methods are identified in prior literature. However, existing classifications typically sort techniques by process rather than outcome, application or causal mechanism. Therefore, techniques can be misapplied and comparative analysis of methods made difficult. Based on the frequency distribution of CTA methods in 1065 studies, a subsample representing 60% of the most frequently published methods was coded based on elicitation and analysis techniques. Consistency of resulting applications was assessed. Inconsistent matching of CTA methods and subsequent applications indicate CTA is currently more craft than technology. Therefore, there is no robust basis for selecting one method over another for research or practice. Specific challenges include comparing the reliability and validity of individual methods and optimising selection of techniques for intended applications. Developing a causal taxonomy may facilitate such advancements.