A case for case studies in L3 research: a simulation of L3 VOT production

Gastvortrag

Kyle Parrish (Goethe Universität Frankfurt)

In the last few decades, there has been increasing discussion surrounding third language (L3) acquisition, in which a range of studies have been carried out across linguistic domains, including (morpho)syntax and phonetics and phonology. One key debate in the field surrounds whether the L3 is influenced in a wholesale fashion by either an L1 or L2 (Rothman, 2015), or whether it can be simultaneously influenced by both of these known languages (Westergaard et al., 2017). While both of these views have received empirical support, it is unclear whether studies supporting whole language transfer have had sufficient statistical power to detect (likely) relevant effects, given their relatively low numbers of participants. The statistical power of a study refers to the probability that an effect (which actually exists) is detected. Any given effect (such as a mean difference) is more difficult to detect when there is greater variation in a data set. This variation can come from at least three sources: the variation caused by sampling of different experimental items, the variation caused by sampling random participants from an underlying population, and random error. Using simulation, and VOT as a hypothetical outcome variable, the present work systematically manipulates the number and variation in participants and items, while taking random error into account, in order to make specific recommendations for necessary and numbers of participants and stimuli on the basis the proportion of the expected variation relative to a given effect. Ultimately, It is argued that low sample or case studies could be an advantageous avenue for L3 studies to consider, since it removes the factor of individual variation from the equation.

• Termin in lokalen Kalender importieren