Outliers in reaction time data : Methodological considerations and practical suggestions

Reaction time is used to measure various types of human performance such as perception, memory, and problem solving. Many constructs, from unconscious prejudice to intelligence to personality, can also be measured by use of reaction time. It is, therefore, fundamentally important to remove the influence of spurious long reaction time in a positively skewed distribution, which reaction time data tends to follow. The present article evaluated methods for dealing with reaction time outliers. These methods were categorized into three types : Sample selection, transformation, and whole distribution analysis. In this article, I summarized pros and cons of these methods and made suggestions for a practical reaction time analysis