Two experiments explored rapid extraction of gist from a visual text and its influence on word recognition. In both, a short text (sentence) containing a target word was presented for 200 ms and was followed by a target recognition task. Results showed that participants recognized contextually anomalous word targets less frequently than contextually consistent counterparts (Experiment 1). This context effect was obtained when sentences contained the same semantic content but with disrupted syntactic structure (Experiment 2). Results demonstrate that words in a briefly presented visual sentence are processed in parallel and that rapid extraction of sentence gist relies on a primitive representation of sentence context (termed protocontext) that is semantically activated by the simultaneous presentation of multiple words (i.e., a sentence) before syntactic processing.