Behavioral, acoustic, and physiological classification of children with speech disorders of unknown origin

While some hypotheses regarding speech delay of unknown origin in preschool children implicate deficiencies in speech motor control and coordination, current nosologies of speech disorders in children fail to incorporate these elements in modeling the subclassification of this heterogeneous population.;This investigation was designed to address the question of whether meaningful subgroups of children would emerge from a heterogeneous population as a result of convergences of measures from multiple speech production subsystems: behavioral, acoustic, articulatory, and respiratory. Rather than relying on extant diagnostic groupings and inferential comparisons of measures between a priori groups, this study instead adopted techniques of multivariate classification to allow the measures to drive the classification of speakers. Because variability is a ubiquitous characteristic of preschool speech production, the goal of this classification was not only to minimize within-subgroup variability, but to identify the latent parameters across speech subsystems that characterized each resulting subgroup. Close to two-thirds of the participants were identified a priori as having speech delay of unknown origin, so this study was also motivated by the question of whether motor coordinative variables would distinguish subgroups of children with speech delay (SD) of unknown origin.;The existence of such a group would support an expanded nosology of preschool speech sound disorders that includes speech physiology parameters. The present results supported these hypotheses. Specifically, the wide variety of measures that spanned subsystems of the speech production mechanism converged to reveal six latent subgroups in a large, heterogeneous group of preschool children. Moreover, the partitioning of the sample was empirically explicable and was consistent with clinical descriptions. Three of the subgroups had a predominance (> 80%) of members from the SD group and one of the SD subgroups exhibited distinct differences in speech motor control. The participants comprising the remaining three subgroups were predominantly from the a priori Not SD group. The characteristics that differentiated the participants in the three Not SD subgroups inform hypotheses of the development of speech motor control. Using classification tree analysis a decision tree was created that can be used to predict the cluster identity of new cases.