Project Summary: Project 4 Model-driven therapies are only as effective as the models that drive them. The goal of Project 4 is to establish a synergistic interaction between clinical and theoretical progress in language science. Data collected as part of Projects 1 and 2 will serve as input to basic science research that will lead to new knowledge regarding the computational and neural foundation of speech production. And feeding back in the other direction, basic science models of speech production will be explored as possible diagnostic and treatment outcome measure tools, opening the door to a new era in model-driven aphasia treatment. Aim 1: Map the neural organization of computational subcomponents of naming. Naming is a multistep process. A complete understanding of its neural basis, and how it can breakdown in aphasia, depends on our ability to map the computational subcomponents of the process. Using both VLSM (error type maps) and VLPM (parameter maps), in this aim we investigate the neural organization of the subcomponents of naming in chronic and acute stroke. Aim 2: Improve and extend the model and the neurobiological mappings. While existing computational models are successful in accounting for the range of naming response types that have been fed into the models, aphasic production deficits are rather more complex than standard operational definitions admit. For this reason, and in keeping with the overall goal of the P50 to bring together theory and practice, Aim 2 seeks to push the computational models more toward clinical reality. We will use our recently-developed SLAM model to simulate and map speech repetition behavior and we will map the neural basis of a common behavior in both naming and connected speech context, self-correction. Aim 3: Assess the utility of model-driven diagnostics. The premise of Project 1 is that identifying functional deficits to the dorsal or ventral streams— i.e., dorsal stream aphasia (DSA) vs. ventral stream aphasia (VSA) vs. dual stream aphasia (DuSA)—provides useful information for aphasia treatment. In Project 1, DSA, VSA, and DuSA are classified on the basis of lesion location, which is a good (arguably the best available) proxy for the functional deficit that is targeted by the subsequent treatment. The possibility we will assess in this aim is whether we can do better at classifying patients pretreatment by measuring their functional deficit “directly” using functional/computational models rather than relying on a lesion-based proxy. This investigation could result in the development of new, clinically available assessment tools for aphasia.