Grapheme-colour synaesthesia, (i.e., the automatic and consistent association of colours with letters and/or numbers), is estimated to be prevalent in just over 1% of the adult population. As yet, few published studies have assessed this relatively common phenomenon in children, nor do we understand its developmental patterns. Experiment 1 (immediate-retest condition) tested 383 primary school children (aged 6-7) for grapheme-colour synaesthesia by asking participants to choose the ‘best’ colour for each of 36 graphemes, and then performing an immediate surprise retest. Data was combined with that from Simner, Creed and Faulkes (2005) in order to provide a large sample size of 618 children, from which to gain a prevalence estimate. An initial estimate of grapheme-colour synaesthesia was found to be 4.2%. However, this value confounds true synaesthetes with children who performed well by superior memory alone. Hence Experiment 2 (1-year retest) retested 59 participants (40 controls, 19 potential synaesthetes) who originally took part in a childhood prevalence study by Simner, Creed et al. (2005). Findings showed that five of the potential synaesthetes (26.3%) continued to consistently associate colours to letters and numbers one year later. Applying this proportion to the data of Experiment 1 suggests that 1.1% of the 618 children tested (age 6-7) were true synaesthetes. Finally, the scores in immediate-retest consistency at Time 1 and Time 2 (1 year apart) were also compared, in order to assess developmental patterns of the condition. There was an average increase in consistency by true synaesthetes of 2.6 graphemes, compared to controls, who showed a smaller average increase of 1.1 graphemes, and to superior memory performers who showed an average decrease of 3.5 graphemes. This suggests that child synaesthetes have a distinct developmental pattern.