ABSTRACT
Mixed models (aka multilevel models) are incredibly important when dealing with situations where there are clusters of non-independent data points, such as is the case with almost all linguistic experiments (“repeated measures experiments”, in particular in psycholinguistics, sociolinguistics, and phonetics). This chapter provides a conceptual introduction to mixed models from the perspective of the “independence assumption” of regression modeling, according to which each data point has to be independent. When this is violated, mixed models can be used to resolve the violation and inform the model about sources of heterogeneity in the data. The reader is introduced to the distinction between fixed and random effects, as well as to the distinction between random intercepts and random slopes. The chapter also introduces the R syntax used to specify mixed models in the widely used lme4 package.
