Speech sound acquisition is important for the development of spoken language. One of the longstanding debates in the field of Linguistics is whether speech sound development is driven by exposure to complexity in input or by exposure to simple speech input. Traditional theories (e.g., behaviorist theories, scaffolding theories, connectionist theories, dynamic systems theory) suggest that speech sound development is facilitated by exposure to simple speech sounds first. On the other hand, linguistic-based complexity theories (e.g., generative phonology, natural phonology, optimality theory) suggest that it is the exposure to complex input first that necessitates speech sound development. However, still there is no consensus on whether it is complex input or simple input that promotes acquisition. In the current dissertation, I address this research question via four different approaches that include Metaanalysis, Maximum Entropy grammar modeling (MaxEnt), Artificial language training, and Speech therapy..

In the meta-analysis study (Chapter 2), data were extracted, processed and combined from 15 selected studies from the literature on treatment of speech sound disorders. Meta-analysis suggested that treatment with complex speech sounds not only improves the production of complex sounds but also generalizes to production of untreated simple speech sounds. In comparison, treatment with simple speech sounds only leads to improvement in the production of simple speech sounds but does not generalize to untreated complex sounds.

In the MaxEnt study (Chapter 3), 3-6-year-old typically developing children (n=30) and those with speech sound disorders (n=31) were compared on their trajectory of phonological development, analyzed using MaxEnt. Based on the patterns of development, it was found that within the group of speech sound disorders, there existed "deviant" and "delayed" subtypes. Children with "delayed" profiles showed a trajectory of development similar to typically developing children while those with "deviant" profiles showed a different trajectory compared to typically developing children. As one of the pre-requisites for success of complexity therapy in speech sound disorders is the similarity of trajectory of development with that of typically developing children. These findings shed light on the use of complex versus simple input in the therapy of different patterns of phonological development.

In the artificial language training study (Chapter 4), Cantonese-speaking adults trained with complex stimuli (n=48) were compared (using behavioral and electrophysiological measures) with those trained with simple stimuli (n=48) in a pseudoword-picture association learning paradigm. We found that the subjects who were trained on complex stimuli showed improvement in complex stimuli and generalization to simple stimuli while those trained on simple stimuli showed improvement in simple stimuli without showing any generalization to complex stimuli.

In the speech therapy study (Chapter 5), children with speech sound disorders from Cantonese-speaking homes were provided therapy with either complex (n=3) or simple (n=2) speech sounds. It was found that the children who were provided therapy with complex speech sounds improved in complex speech sounds and generalized to untreated simple speech sounds. On the other hand, the children provided therapy with simple speech sounds improved in simple speech sounds but did not generalize to untreated complex speech sounds.

These four studies, corroboratively, indicate that complexity induces widespread generalization to elements with equivalent and/or lesser complexity across the phonological system. Overall, from the findings of this dissertation, it can be concluded that complex input is potentially more efficient than simple input in promoting behavioral and neural changes in the phonological system. The current findings have implications for education and rehabilitation. Traditionally, it is believed that exposure to simple stimuli first followed by a sequential increase in complexity leads to the acquisition of new concepts. In light of the current findings, one can speculate the use of complex stimuli first for a time-effective learning process.