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

We propose a novel statistical method for detecting and comparing the onsets of experimental effects in visual world eye-tracking. Existing approaches do not allow researchers to statistically compare onsets of effects between experimental conditions or participant groups, despite their relevance to psycholinguistic research. A previous solution used a bootstrap-based method to estimate onsets of effects and their confidence intervals (Stone et al., 2020). However, the statistical properties of this method, including the empirical coverage of its confidence intervals, were never assessed. Our proposed method uses generalized additive mixed models (GAMMs) and posterior simulations to integrate the modelling of fixation curves over time with onset detection. We demonstrate the applicability of the GAMM-based method by reanalysing two visual world datasets. In addition, we formally evaluate the GAMM-based method and the bootstrap-based method in two recovery simulation studies. We show that the previously proposed bootstrap-based method tends to yield delayed estimates, exhibits poor coverage properties, and inflated type I error rates. In contrast, the new GAMM-based method produces estimates with low bias and well-calibrated confidence intervals. Since determining temporal onsets is informative to models of language processing, the GAMM-based method constitutes a valuable addition to the psycholinguistic analytical toolkit. Data and code are publicly available in an OSF repository at https://osf.io/9ethq