This study describes the realization of onset and coda clusters in a 4-year old child acquiring American English, and with a higher than usual level of unintelligible speech. It reviews previous studies that have tested cluster realization against markedness and, in particular, the sonority hypothesis. This latter predicts steep rises in sonority at onsets and more gradual falls in sonority at codas of syllables. The data collected for this study cover four different onset cluster types and three different coda cluster types, and are derived from spontaneous speech produced during clinical sessions. An analysis of the patterns of cluster realizations shows that the participant's cluster realizations do not always follow the sonority hypothesis, and various factors are discussed that could be motivating the child's choice of realization.