Six normal young Japanese children and six children with Williams syndrome (WMS) were presented a nonsense label to identify an unfamiliar target solid object that was either rigid (e.g., steel) or flexible (e.g., rubber). The presentation was accompanied with or without a gesture. When the gesture was performed, it emphasized either the shape and function (e.g., a rolling gesture), or the material (e.g., squeezing) of the target. Thereafter the children were presented with a pair of unfamiliar objects, one of them matched with the target in shape and function the other in material. They were asked, given the same label, to choose an object that matched the target in shape and function, or in material. When a gesture accompanied the preceding presentation of the target, the normal children were likely to choose that object which matched the target in shape and function if the gesture had emphasized the shape and function of the target, but tended to choose the object that matched the target in material if the gesture had emphasized the target material. When no gesture was provided, they chose either of the two objects at random. In contrast, the choices of the children with WMS were not influenced by the type of preceding gesture, but were determined randomly regardless of the presence of the gesture. These findings are discussed in terms of the possible impairement of working memory in children with WMS, which would make their language acquisition unique.