The structure of /s/-initial clusters is debated in developmental phonology. Pan and Snyder (2004) took the Government Phonology (GP) framework and proposed that production of /s/-initial clusters requires the positive setting of two binary parameters [+/-Branching rhyme (BR)] and [+/-Magic empty nucleus (MEN)] and the initial /s/ is treated as a "coda" preceded by an empty nucleus. It is predicted that /s/-initial clusters should never be acquired prior to branching rhymes (CVCC sequences under the GP definition). This study tests Pan and Snyder's prediction in disordered speech. Their prediction is supported by the data of seven English-speaking children with phonological delays. None of the children acquired /s/-initial clusters before branching rhymes. Implications for phonological intervention are discussed in light of the finding. (Contains 3 tables.)