Incoming sensory information is affected by hierarchical probabilistic predictions based on context and previous experience. Generating predictions from the highest level of the hierarchy (top-down processing) is thought to require consciousness, while lower-level predictions proceed pre(un)-consciously. Predictions are particularly important for successful language comprehension. In a semantic priming paradigm, the processing speed of target words is increased by presentation of preceding semantically-related words (primes) –an automatic/low-level process. However, by manipulating the overall proportion of semantically-related or unrelated word-pairs across a task (relatedness proportion paradigm), we also create a global context for strategic/top-down processing. Each trial includes a visual target word 1240ms after a visual prime word presented in one of two colours that cue the participant about the probability of the target being related (78%/22%). We expect to observe priming in both contexts, reflecting lower-level ‘local’ predictions/spreading activation. Moreover, we expect to observe an interaction between the relatedness of the target and the validity of the prime, reflecting strategic predictions based on ‘global’ context. Moreover, we will present evidence for the neural correlates of these predictions from EEG recordings in healthy participants. Our task may be beneficial for improving the accuracy of diagnosis of patients with disorders of consciousness, as it differentiates automatic and strategic processing - in a similar way to the local-global paradigm - but in the context of the clinically-relevant domain of residual language processing. 21/01/2019: We uploaded a file named Prereg_form_References_Amendment as we found some mistakes in the references and citations. This document contains the corrections for these mistakes.