“…Generalized additive mixed models (GAMMs; Wood, 2004Wood, , 2006a are an extension of GAMs as mixed models, in which random effects are estimated from a GAM by computing the variances of the so-called 'wiggly' components of the smooth terms (i.e., the degree of smoothness of the terms). GAMMs have previously been used to investigate speech production over time (Baayen, Vasishth, Kliegl, & Bates, 2017;Kirkham, Nance, Littlewood, Lightfoot, & Groarke, 2019;Mielke, Carignan, & Thomas, 2017;Sóskuthy, 2017;Wieling et al, 2016;Winter & Wieling, 2016) and space (Barlaz et al, 2018;Wieling, 2018), to observe the effects of word frequency and lexical proficiency on articulation (Tomaschek, Tucker, Fasiolo, & Baayen, 2018), and to model spatio-temporal relations in flesh-point kinematics (Tomaschek, Arnold, Bröker, & Baayen, 2018). One distinct advantage of employing GAMMs for speech articulation research is that they can capture the interaction effects of two different continuous variables (such as time and space), using tensor product interaction, which allows the smooth coefficients for one variable to vary in a non-linear fashion depending on the value of the other variable (Wieling, 2018, p. 102).…”