The phonemic effects of word length, consonant-vowel structure, syllable structure, and meaning on word segmentation were investigated in two experiments with young children. The decentration hypothesis, which predicts that children who habitually direct their attention to word meaning would concentrate better at analyzing a spoken form without meaning (a pseudo-word) and would perform better on pseudo-words than on existing, structurally similar words, was tested. Results provide strong evidence that an onset-rime distinction is relevant for the process of segmentation, and word meaning seemed to have no influence, suggesting that the decentration hypothesis can be abandoned as an explanation for segmentation difficulties. It is concluded that effects of length and syllabic boundary can be explained by the disruptive effect of consonant clusters, which are not only difficult to segment by themselves but also adversely affect the processing of segments earlier in the word, and that a simple, strictly serial model for segmentation is not adequate. The results also indicate that an articulatory rather than phonological code is the object of segmentation. (MSE)