Abstract: Language-specific rhythmic categories play a vital role in the processing of spoken language, which involves both speech perception and speech production. Some researchers argue that the MORA is the basic rhythmic or timing unit of spoken language processing in Japanese. For example, Ito and Tatsumi (1997), on the basis of their developmental studies, contend that Japanese children are capable of segmenting words into mora units before they master Japanese kana reading. However, Inagaki, Hatano, and Otake (2000) claim in their developmental study that Japanese children do not appear to be especially sensitive to the mora before they become kana literate. Thus, there is contradictory evidence regarding the intrinsic importance of the mora in Japanese, indicating that we do not yet fully understand how mechanisms for Japanese spoken language processing are acquired. In previous psycholinguistic research, mora-oriented phenomena in speech perception and speech production have been investigated through the analysis of speech errors, stuttering, and segmentation and blend experiments. While Kubozono (2003) has examined Japanese children's speech errors naturalistically, no experimental study has been conducted to investigate such speech errors and what they can tell us about the significance of the mora in Japanese children's phonological representations. My experimental study investigates the developmental aspects of speech production through an examination of Japanese children's slips of the tongue (SOT). The present study consisted of two tongue-twister (TT) experiments designed to elicit SOTs from 3- to 8-year-old Japanese children and from Japanese adults. A total of 668 phonological SOTs obtained from 59 subjects were analyzed and categorized as involving the following phonological units: features, consonants, vowels, non-syllabic moras, syllabic moras, rhymes, or syllables. The results of the experiments showed that both child and adult subjects made far more errors involving segments than errors involving any other phonological unit, a finding consistent with SOT studies in other languages. Hence, whereas many researchers who have conducted perception and metalinguistic experiments have shown that the mora plays a significant role in Japanese speech perception and phonological representation, the results of my experiments demonstrate that the segment plays a significant role in speech production and, thus, that the mora is not the only basic unit of spoken language processing in Japanese. Given the fact that segmental errors are common in many other languages, my findings also suggest that there might be some sort of universal underlying cognitive mechanism for speech production planning.