abstract Xenia Schmaiz

Xenia Schmaiz (Università degli Studi di Padova, Italy)

Establishing invariance in psycholinguistics: The case of the letter bigram frequency effect

To create a model of a phenomenon, one needs to establish a set of benchmark effects that the model should account for, but it is equally important to determine aspects which do not have an influence. Recent research has stressed the need to consider the absence of effects to provide a complete picture, and the inadequacy of traditionally used p-values to establish invariance (Dienes, 2014; Rouder et al., 2009). Here, we discuss statistical and methodological approaches which could be used in psycholinguistic research to establish whether or not a variable has an effect on word reading latencies. We focus on the effect of letter bigram frequency: previous studies have shown mixed results about the presence or direction of these effects. To reach a conclusion, ideally, converging evidence should be provided from different approaches, including Bayes Factor analyses, Structural Equation Models, and factorial designs and covariate analyses.