This study examined 31 schizophrenic children to determine if they used discourse devices that make speech coherence differently from sex and mental age matched normal children. It also investigated whether the discourse deficits of the schizophrenic children were related to clinical measures of formal thought disorder. Using Halliday and Hassan's analysis of cohesion, the authors found that schizophrenic children underutilize some discourse devices and overutilize others. Several of their discourse deficits were similar to those described in schizophrenic adults. The schizophrenic children, however, also had additional discourse deficits, which probably reflect developmental delays. The authors also demonstrated that the schizophrenic children with loose associations had different discourse deficits and discourse/IQ correlates than schizophrenic children without loose associations. The schizophrenic children receiving neuroleptic medication had lower loose associations scores than the unmedicated subjects. The authors discuss the possible confounding effect of medication and loose associations, as well as the developmental, cognitive, and clinical implications of the study's findings.