Normal children between 3 and 6 yrs old generated word and sentence messages for use in a pediatric speech intelligibility (PSI) test. Word materials did not differ as a function of chronological age (CA), vocabulary skills, or receptive language (RL) ability. Elicited sentence materials, however, did reflect differences in CA, vocabulary skills, and RL ability. To represent differences in the children's responses, two different types of test sentences, subsequently referred to as format I and format II sentences, were formed. Performance for format I and format II sentences was significantly different in normal children between 3 and 6 yrs old. Performance differences were related to CA and RL ability. However, performance for format I sentences in children with relatively low RL age and performance for format II sentences in children with relatively high RL age was equivalent. This observation yielded an algorithm that determines the sentence format as a function of a child's RL age in order to yield “language equivalent” norms for speech audiometry. In contrast to performance for PSI sentence materials, performance for PSI word materials was not influenced by differences in RL skill. Reliability coefficients for PSI word and sentence test-retest measures ranged from about 0.85 to 0.95 in normal-hearing and hearing-impaired children. Illustrative PSI results in children with confirmed central auditory disorders are presented.