A novel method for detecting the onset of experimental effects in visual world eye-tracking

Determining the onset of experimental effects in timeseries data is important in psycholinguistics, because it allows for a more precise evaluation of accounts of the timecourse of different types of information during language processing, as well as their impact on predictive computations. Different methods have been proposed to determine the onset of effects in visual world studies, but to date, their validity has not been formally evaluated. We bridge this gap by examining the coverage of one of these previous methods, the bootstrap-based method proposed by Stone et al. (2021). In two simulation studies, we demonstrate that the bootstrap-based method tends to produce delayed onset estimates, has poor coverage properties, and shows inflated type I error rates. We propose a novel method that addresses these shortcomings, and that allows researchers to more accurately detect the onset of experimental effects, as well as to compare onsets between different experimental conditions and/or participant groups. This method uses generalised additive mixed models (GAMMs) and posterior simulations to seamlessly integrate onset detection with the modelling of fixation curves over time. Our simulations show that the GAMM-based method produces estimates with low bias and well-calibrated confidence intervals. We recommend the adoption of the GAMM-based method in future psycholinguistic research and provide a user-friendly R function to facilitate its use. By enabling more precise measurements of the timing of cognitive processes in the visual world paradigm, our method can contribute to advancing theories of prediction and language comprehension.