A new tool for equating lexical stimuli across experimental conditions

Abstract

In cognitive psychology and psycholinguistics, lexical characteristics can drive large effects, which can create confounds when word stimuli are intended to be unrelated to the effect of interest. Thus, it is critical to control for these potential confounds. As an alternative to randomly assigning word bank items to stimulus lists, we present LIBRA (Lexical Item Balancing & Resampling Algorithm), a MATLAB-based toolbox for quickly generating stimulus lists of user-determined length and number that can be closely equated on any number of lexical properties. The toolbox comprises two scripts: a genetic algorithm that performs the inter-list balancing, and a tool for filtering/trimming long omnibus word lists based on simple criteria, prior to balancing. Relying on randomized procedures often results in substantially unbalanced experimental conditions, but our method guarantees that the lists used for each experimental condition contain no meaningful differences. Thus, the lexical characteristics of the specific words used will add an absolute minimum of bias/noise to the experiment in which they are applied. Our toolbox balances word lists for arbitrary lexical properties to control confounds in cognitive psychology research. Our toolbox performs more efficiently than pure randomization or balancing manually. A graphical user interface is provided for ease of use.