During phonological development, children frequently produce consonant clusters as consonant singletons, a process commonly referred to as cluster reduction. The principles of sonority may provide a theoretical basis for explaining patterns of cluster reduction evident in children's speech. Two studies were conducted to investigate whether children's word-initial cluster reductions adhered to the sonority hypothesis. Study one involved 16 children with typically developing speech, and study two involved 40 children with impaired speech. The children's consonant cluster productions characterized by a cluster reduction were analysed. When both groups of participants reduced wordinitial clusters to a target consonant, the sonority hypothesis was adhered to; but when the clusters were reduced to a non-target consonant, the sonority hypothesis was violated. Analysis of target and non-target reductions revealed that some reductions of the individual clusters, and those within specific cluster categories, adhered to the sonority hypothesis while others did not. In light of these findings, it is suggested that although sonority is a valuable concept, it may not account for all patterns of cluster reduction evident in children's speech.