Language comprehension is critical to a wide variety of child outcomes, including academic success and emotional and social well-being. Effective intervention relies on valid, reliable language comprehension data to determine the intensity and techniques that are appropriate for an individual child. The present study investigated language comprehension in a sample of 113 toddlers with significant developmental delays using IRT methods. We found that the aggregate data adequately fit the Rasch model, though each measure also contained items with poor fit. Analyses of the correspondence between item difficulties and participant abilities generally supported the appropriateness of the measures for our sample, and indicated acceptable measurement precision for the majority of participants. Examination of the relative difficulty of items revealed patterns that were largely consistent with the literature on typically developing children, with a few exceptions. Investigation of individual items showing the highest proportions of change in our sample indicated that parent-report items of moderate difficulty were most likely to reflect language comprehension improvement. Our findings inform clinical practice by underscoring the strengths and limitations of currently available measures. They also inform future measure development by emphasizing the benefits of integrating IRT methods in order to maximize both measurement precision and testing efficiency. Finally, they add to knowledge about language comprehension development in atypical populations.