The ability of children and adults lo produce symbolic and non-symbolic magnitude estimations was examined and related to chiidren's familiarity with numbers. Six-year-old kindergartners, 7-year-old first graders, 9-year-old third graders and adults made symbolic estimations either by saying number words that matched to numbers of <lots (i.e., perception task) or by generating numbers of dots that matched to given number words (i.e., production task). In the non-symbolic estimation task,

more »

... pants generated the corresponding numbers of dot' i they had seen previously (i.e., reproduction task). In line with the bi-directional mapping hypothesis (Castronovo & Seron, 2007), children and adults made underestimations in the perception task, overestimations in the production task and intermediate estimations in the reproduction task. However, the perfonnance of kinderga11ners and first graders showed significant deviations from the predictions of thc bi-directional mapping hypothesis. They performed significantly poorer in the production task than in the perception task, implying that these tasks are not mirrored processes in young children. ln addition, they made systematic overestimations in the non-symbolic reproduction task, which suggeslc; that a biased mapping occurs here as weil. The results are discussed in view of children ·s familiarity with numbers and potential estimation strategies.