The use of mixed effects models in quantifying the dynamics of speech

Abstract

Mixed-effects models have been used often in quantifying the variability in data from experiments in speech and language. In most of these experiments, the dependent variable is measured at some landmark of a kinematic or decision process. However mixed-effects models are increasingly being used to quantify the variability of dependent variables that vary in time, such as articulator movements, formant transitions, and eye tracking data. This presentation will first provide a tutorial introduction to the use of the mixed-effects model, especially the growth-curve variant, for quantifying variability where time is an essential independent variable. It will then be argued that the model coefficients can be interpreted as dynamic coefficients of differential equations that describe the dynamics of the underlying processes. © 2013 Acoustical Society of America.