Generalized Additive Mixed Models for intraspeaker variation

Intraspeaker sociolinguistic variation is typically characterized by repetitiveness in what choices speakers make from moment to moment, but there are multiple possible sources of such repetitiveness. We distinguish two types of temporal clustering: sequential dependence and baseline deflection. We argue that because we have independent reasons from sociolinguistics and psycholinguistics to believe both types are at play in speech production, it is desirable to adopt quantitative models that can simultaneously estimate these distinct sources of temporal clustering. We propose the use of Generalized Additive Mixed Models (GAMMs) for this purpose and illustrate with a case study of DH-stopping in Philadelphia English sociolinguistic interviews. We advocate for the adoption of GAMMs to advance the use of naturalistic data for studying psycholinguistic questions about intraspeaker variation.