Recent research has indicated that young children's responses to requests containing the prepositions in, on and under are strongly influenced by the context in which they are made, but that they probably understand the terms in and on before the term under. However, these studies employed contextual manipulations that bias the child towards one particular placement. In this study, the contextual manipulation involved a pre-test activity aimed at equating the likelihood of all three placements. Children from 1 year 6 months to 3 years old showed markedly better comprehension of the terms than children who were not exposed to such pre-test activity. This superiority cannot be put down to contextual bias towards a particular placement, and hence performance under these conditions may give a better measure of actual comprehension. It is suggested that the conditions in real life that force a clear distinction between spatial prepositions are those similar to the present experimental manipulation, in which one object commonly enters into more than one relationship with another.