An experiment requiring the learning of 'new' (nonsense) words replacing the words 'eager' and 'easy' in the linguistic structure 'John is eager/easy to please' showed that only adults and higher-IQ children could perform better than chance on the new words. Contrary to prediction, children showed no differential learning ability on two types of words to be learned. However, an analysis of the strategies used gave evidence of a language-specific learning ability in children but not in adults. It is proposed that the inability of lower-IQ children to learn the words, as well as the lack of differential learning of the two types by higher-IQ children, was due to the nature of the task—being more like that used in concept formation experiments than like natural language.