Abstract
<p>The increasing use of web-based platforms for evaluating quantitative predictions in psycholinguistics raises an important methodological question: are effect sizes estimated on these platforms comparable to those obtained in laboratory settings? To address this question, we replicated an existing self-paced reading study designed to measure garden-path and syntactic adaptation effects across four platforms: SONA-lab, SONA-online, Prolific, and MTurk. We found that garden-path effects were consistently larger in the laboratory than on all three web-based platforms. Exploratory analyses of filler items revealed a similar attenuation of word frequency and word position effects in the web-based platforms. Together, these findings suggest that testing platform can systematically influence estimated effect sizes and should therefore be considered when evaluating quantitative predictions in psycholinguistics.</p>