Morphological processing is gradient not discrete in L1 and L2 English masked priming

Abstract

In recent years, evidence has emerged that readers may have access to the meaning of complex words even in the early stages of processing, suggesting that phenomena previously attributed to morphological decomposition may actually emerge from an interplay between formal and semantic effects. The present study adds to this line of work by deploying a forward masked priming experiment with both L1 (Experiment 1) and L2 (Experiment 2) speakers of English. Following recent research trends, we view morphological processing as a gradient process emerging over time. In order to model this, we used a large within-item stimulus design combined with advanced statistical methods such as generalised mixed models (GAMM) and quantile regression (QGAM). L1 GAMM analyses only showed priming for true morpho-semantic relations (the identity ‘bull’, inflected ‘bulls’ and derived conditions ‘bullish’), with no priming observed in the case of other relations (the pseudo-complex ‘bully’ or the stem-embedded ‘bullet’ conditions). Furthermore, with respect to the time-course of effects, we found significant differences between conditions were present from very early on as revealed by the QGAM analyses. In contrast, L2 speakers showed significant facilitation across all five conditions compared to the baseline condition, including the stem-embedded condition, suggesting early L2 processing is only dependant on the form.

How trial-to-trial learning shapes mappings in the mental lexicon: Modelling lexical decision with linear discriminative learning

“…Furthermore, cognitive differences may affect lexical processing (e.g. Fischer-Baum et al, 2018 , Kuperman and Van Dyke, 2011 , Lõo et al, 2019 , Milin, Divjak et al, 2017 , Perfetti et al, 2005 ). As a consequence, the regression weights of lexical-distributional predictors can vary significantly between participants (example in Baayen, 2014 ).…”

“…Further studies confirm the influence of individual differences (e.g. Fischer-Baum et al, 2018 , Perfetti et al, 2005 ), but note that connecting differences in morphological processing to individual psychological measures is not straightforward ( Lõo, Toth, Karaca, & Järvikivi, 2019 ). In the present work we explore individual differences in lexical processing in considerable detail by investigating the random effect structure of a linear mixed model, in the hope of being able to provide an algorithmic characterization of these differences.…”