The replication crisis in the behavioural sciences has motivated many discussions about the robustness of published findings. In this talk I will present and justify a number of diagnostic criteria that can indicate whether most findings in a given research field are likely to be robust (I.e. likely to provide similar findings if a well-powered direct replication is attempted). These criteria will then be used to assess whether the current state of evidence supports that bilingualism cause general changes to the cognitive system. I will end by suggesting possible ways forward, including practical actions one can take to increase the likelihood that one’s research findings will be replicable.
High variability phonetic training (HVPT) with multiple speaker input (compared to low variability (LV) input with one speaker) has been used successfully to teach adults L2 speech contrasts. However, whether HV is beneficial over LV for children is not as clear. Results of a two-week phonetic training study will be presented in which two groups of Dutch learners of English, aged 7/8 and 11/12, were trained on four British English phoneme contrasts that are notoriously difficult for Dutch learners. Children received either HV or LV input in training. Effects of variability were investigated using a pre/post-test design in which children’s phoneme identification and discrimination abilities as well as their production were tested.