This study explores a common assumption made in the cognitive development literature that children will treat gestures as labels for objects. Without doubt, researchers in these experiments intend to use gestures symbolically as labels. The present studies examine whether children interpret these gestures as labels. In Study 1 two-, three-, and four-year olds tested in a training paradigm learned gesture-object pairs for both iconic and arbitrary gestures. Iconic gestures became more accurate with age, while arbitrary gestures did not. Study 2 tested the willingness of children aged 40-60 months to fast map novel nouns, iconic gestures and arbitrary gestures to novel objects. Children used fast mapping to choose objects for novel nouns, but treated gesture as an action associate, looking for an object that could perform the action depicted by the gesture. They were successful with iconic gestures but chose objects randomly for arbitrary gestures and did not fast map. Study 3 tested whether this effect was a result of the framing of the request and found that results did not change regardless of whether the request was framed with a deictic phrase ("this one (gesture)") or an article ("a (gesture)"). Implications for preschool children's understanding of iconicity, and for their default interpretations of gesture are discussed. (Contains 1 table and 6 figures.)