The purpose of this study was to identify acoustic features related to the perceptual identification of speech sound errors and thus to provide an acoustic basis for interpreting the speech motor deficits in Cerebral Palsied (CP) children as well as tracking their changes for future treatment efficacy studies. Six Mandarin-speaking children with Spastic Cerebral Palsy (3 diplegia and 3 quadriplegia) read a list of 140 bi-syllabic words including minimal pairs of all Mandarin phonemes. Speech sound productions consistently judged by two listeners as “correct” or “incorrect” were compared on acoustic measures, including segmental lengths, vowel formant frequencies, and consonant spectral measures. The frequency of speech errors was, in average, highest for consonants (21.2%) and minimal for both vowels (1.8%) and tones (0.6%). Consonants most often misarticulated were fricatives and affricates, especially those involving retroflex. The centre of gravity and the spread of the consonant noise frequencies were generally decreased, reflecting a more posterior tongue placement and a lower degree of diffusion for frication noise. Vowels associated with “incorrect” consonant productions were found to show decreased durations and a more compressed vowel space, indicative of an articulatory “undershoot” phenomenon. The tone effect on the segmental length in the CP speech samples remained similar to what would be expected from normal speakers, with Tone 1 resulting in longer segmental lengths than Tone 4 and Neutral Tone. The length of consonants was generally prolonged in “incorrect” production. The causal link between these speech symptoms is most likely to be a restricted motor range related to increased muscle tones. No consistent difference on the pattern of speech errors was found between the diplegia and quadriplegia groups, suggesting that assessment of specific speech motor constraints might be more relevant than traditional medical classification in predicting the frequency and type of speech errors. The general patterns of tone, vowel, and consonant production in the CP children will be discussed in relation to previous findings, along with interpretations of the acoustic features found to be useful in revealing the underlying physiological constraints responsible for speech deficits in CP children.