The present study had three aims: firstly, we wished to examine the nature of deaf children’s semantic representations on ‘familiar’ and ‘less familiar’ vocabulary items (determined by naming ability). Four tasks were used in assessment: definition and picture-drawing to establish production ability; picture-matching and meaning recognition, focused on receptive ability. Secondly, we hoped to ascertain whether a strong understanding of the vocabulary item in the child’s preferred modality (BSL/ English) led to more complete learning of the corresponding word in their less preferred modality (written English/BSL respectively). Four post-training tasks assessed knowledge of the new form: a snap-game (immediately post-training), a naming task (writing word/producing sign), a sign/word array and a picture array. The final aim was to determine whether certain pairings implemented in teaching the items in the new modality would make the mapping between word form and its meaning most explicit (sign-word; picture-word for training in English/picture-sign; word-sign for teaching in BSL). A case-study approach was adopted to allow for a more in-depth assessment of semantic representations. It was found that naming ability on the pre-test was generally associated with level of semantic knowledge. Knowledge of objects’ physical properties was demonstrated to be a core feature underlying both ‘familiar’ and ‘less familiar’ vocabulary items. No effect of familiarity was seen with respect to facilitating more complete learning in the less preferred modality. All children demonstrated at least partial learning of the new forms, however it was not possible to deduce whether one training method was of greater benefit: this varied with respect to the individual. Our findings will be discussed with reference to relevant future directions: in particular, it is recommended this study be replicated on a larger-scale, to ensure any trends highlighted were not unique to these children.