"This project aims at elucidating fundamental neural computations underlying language processing. While we have gained enormous insights into the localization of language in the brain and the temporal sequence of language processes (e.g., syntactic vs. semantic), we know very little about the actual computations underlying language processing. I propose that the framework of predictive coding can fill this gap. Predictive coding is a fundamental theory of sensory processing in the brain that has recently gained enormous attention in psychology and neuroscience. While models of language assume a bottom-up driven processing stream from sensory areas through different linguistic processing stages (e.g., phonetic, syntactic) towards semantic interpretation, predictive coding postulates that high-level brain systems actively construct models of the external world and pass resulting predictions about upcoming events to lower sensory systems. Only if predictions are violated, a prediction error is signalled in a bottom-up fashion to higher areas where internal models are adjusted to minimize prediction error. Here, I postulate that language-processing is the result of multiple overlapping predictions from different sources of linguistic information, if available. I propose a research program that (a) explores the presence of neurophysiological markers of predictive coding during language processing, (b) tests quantitative hypotheses from predictive coding concerning strength and precision of prediction error signals, for established language effects, and (c) explores the domain generality of identified mechanisms. To this end, established electrophysiological and brain activation markers of sentence processing will be combined with advanced model-based data analysis tools. Finally, a comprehensive functional architecture of language shall be established that incorporates dynamically reconfigurable feedforward and feedback information flow in the language system."