Logistic regression modelling for first and second language perception data
Logistic regression analysis has, for some time, been successfully applied to L1 speech perception data, but has not been widely applied in L2 speech perception research. This chapter is a tutorial which makes use of simple data sets to introduce logistic regression analysis as applied to categorical response data from L1 and L2 speech perception experiments. Data are taken from an experiment on L1 Spanish vowel perception by Álvarez González, and experiments on L1 and L2 English vowel perception by Escudero & Boersma, and Morrison. Model fitting is demonstrated as a technique to determine which acoustic cues are attended to by listeners. Logistic regression coefficients are used to quantify how listeners use those acoustic cues, to produce graphical representations of their use of acoustic cues, and as statistics in secondary analyses used to determine whether there are significant differences in the perception of stimuli by L1 versus L2 groups of listeners.