Understanding brain activity associated with syntactic processing is essential for exploring the nature of human language. Recent fMRI studies suggest that the inferior frontal region can process abstract syntactic structure using function words in sentences with no content word information, i.e., jabberwocky sentences; however, it is difficult to explain the timing of structure building. Although event-related potential studies reported an early anterior negative component ~200 ms post-target-word onset that reflects a local phrase-structure building process both in regular and jabberwocky sentences, it is not clear whether such early syntactic processing occurs in building a sentence-level structure. We focused on case markers, which are predictors of sentence structures, and aimed to investigate the activation time of case information during sentence comprehension. We used regular and jabberwocky sentences, both of which included subjective (-ga) or objective (-o) case-marked nouns, transitive verbs, and nouns marked "-de" to denote place, e.g., NP-ga NP-o VP-ta-yo NP-de. The participants were instructed to judge the grammaticality of sentences. Electroencephalograms were recorded using 63 Ag/AgCl electrodes cap (n = 32; age, 18–26 y). Analyses at the average amplitude (trigger onset: 2nd NP, electrodes: Fz, Cz, Pz) were conducted in two time windows: 160–210 ms and 300–450 ms. Three-way ANOVA (sentence type × case order × electrode position) were performed. Analysis of the early-time window showed a three-way interaction (p<0.05), and in the jabberwocky sentence, the double-objective condition showed a more negative deflection at the Fz electrode (p<0.05). In the late-time window, the case order effect was observed, which interacted with the electrode position (p<0.0001). These results indicated that case information could predict sentence structure of the following transitive verb; an early negativity reflected argument-structure related, but not simple word category-related processing. Late negativity is the N400 that resulted from a semantic role mismatch. Our findings suggest that sentence comprehension is based on an early process of abstract sentence-level structure building using case information.