This dissertation is a linguistically-motivated investigation into different areas of language in children with autism spectrum disorders (ASD), compared to typically developing (TD) children. Fine distinctions between linguistic units were used in designing tasks on language production and comprehension in seven experiments. The focus of each chapter of this dissertation was on three main hypotheses respectively, namely (1) the Abstract Representation Difficulty Hypothesis that children with ASD (perhaps limited to the subgroup with co-morbid language impairments) have difficulties activating abstract lexical representations as effectively as TD children, due to their hyperattention to phonetic details of speech, (2) the Pragmatic over Grammatical Deficit Hypothesis that pragmatics is particularly difficult for all the ASD children, while morphological and semantic aspects of language are relatively intact, and (3) the Cognitive Factor Hypothesis that cognitive factors such as nonverbal intelligence quotient (NVIQ) and nonverbal working memory play a greater role in the ASD than the TD performance on linguistic tasks. Chapter 2 investigates the morpho-phonological and semantic aspects of the lexical processing of Thai compound and simplex words. Results suggest that morphological facilitation effects can be obtained independently of phonological and semantic relatedness in the processing of Thai compounds. While children with ASD with lower task performance display hyper-attention to the acoustic differences between primes and targets, children with ASD in the higher performance group have enhanced morphological effects, compared to their TD peers, and the effects appear to be independent of the presence of phonological effects and enhanced semantic effects. The lack of phonological effects in the first set of experiments was explored further in the later experiments. Children with ASD were found to be slower in processing natural-sounding surface phonological forms, suggesting that a deeper processing of neutralized forms than full forms. The similar performance on the next task with the integration of visual information suggests that the slower processing may result from their slower lexical semantic processing. The Abstract Representation Difficulty Hypothesis, thus, holds for a subgroup of children with ASD, while other children with ASD display intact phonological representation, enhanced morphological processing compared to TD controls, and intact but slower lexical processing. Chapter 3 explores the Pragmatic over Grammatical Deficits Hypothesis. Using fine distinctions within the personal reference terms, consistently replicated results suggest that while grammatical person phi-features are intact in children with ASD's representation of pronouns, these children are less sensitive to deictic information in their interpretation of pronouns and tend to avoid using the first-person pronoun, with high deictic level, when they have freedom to choose personal names to refer to themselves. Children with ASD also performed more poorly on the comprehension of unmarked pronouns which requires implicated presupposition, suggesting that even with minimal comparisons among the pronouns, lexically-encoded core grammatical features and pragmatic ones are distinguished in children's language processing. Chapter 3 also adds to the literature on lexical presuppositions, scalar implicature, and implicated presuppositions that not only adolescents, but also children with ASD are age-appropriate in deriving scalar implicatures and that not all kinds of pragmatic inferences are equally challenging for children with ASD. The most indicative difference between the children with ASD and the TD group lies in the children with ASD's heavier reliance on literal, logical meaning when other semantically- and pragmatically-inferred meanings are violated. Chapter 4 partly contributes to the Cognitive Factor Hypothesis, suggesting a possibility that cognitive factors, as opposed to developmental factors, correlates more with children with ASD's performance on linguistic tasks. Additionally, children in both groups displayed correlations in their performance across all of the experiment in the dissertation. Individual language profiles were compiled with the results from the previous chapters. Two subgroups of children with ASD were identified through k-means cluster analysis. The children with ASD in Cluster 1 have globally better performance across experiments than children with ASD in Cluster 2, supporting that ASD children may be able to be classified into subgroups based on their performance on linguistic tasks alone. Even with globally better linguistic task performance, the children with ASD in Cluster 1 still appear to be less sensitive to social-deictic information, confirming that certain types of pragmatics are indeed more challenging than the others. In sum, this dissertation advances our understanding on morphological, semantic, and pragmatic abilities of children with autism through carefully-designed linguistically-motivated experiments.