Acquiring a second language speech contrast that does not exist in the native language is often difficult. High variability phonetic training (HVPT) is a wellestablished method used to train learners on specific non-native phoneme contrasts: it critically uses high variability (HV) input after earlier attempts using low variability (LV) input had proved unsuccessful. HVPT has since been successfully applied in many different adult studies. However, there is no consensus on the effect of input variation on children’s learning of non-native phoneme contrasts. This thesis aims to further investigate the effect of input variability on phonetic training for children, and examining whether they show the same HV benefit which has been argued to hold for adults. In the first set of studies, native English speaking adults and children were taught Dutch vowels in a single computerised training session, during which they received either HV or LV input. Additionally, the traditional HVPT paradigm was adapted to see if mapping vowels to orthography-like symbols representing phoneme categories was more or less effective than a vocabulary training method without such representations. Learning was stronger with training most akin to vocabulary learning, particularly for children, suggesting a benefit for a more meaningful learning context. Crucially, there was no evidence of a HV benefit for either children or adults. The second study was a two-week training study in which Dutch children of two age groups were trained on Standard Southern British English vowel contrasts. Since picture-based training had proved beneficial, this study combined both orthography and pictures in training. Potential effects of HV or LV input in training were investigated using a pre/post-test design. Older children outperformed younger children throughout, and again no evidence for a variability benefit was found. This indicates children might not benefit from high input variability.