In this article, two approaches to schwa representation are evaluated using data from four German children (aged 1;3 to 3;1). The first, the No Mora account, claims that schwa is segmentally but not prosodically specified, whereas the second, the Projection account, proposes that schwa is prosodically but only minimally segmentally specified, as [-cons]. The data reveal several stages in acquisition: an early phase characterised by vowel reduplication or deletion; a middle phase characterised by production of a full vowel (and consonant); and a final phase characterised by production of a central vowel, schwa, or syllabic consonant. In addition, the data show that children produce syllabic consonants earlier than schwa, and that they often produce filler schwas even when they substitute target schwas with full vowels. It is argued that these findings, in particular the different timeline observed for schwa versus syllabic consonants and the presence of filler schwas, can be best accounted for by the Projection account.