Prior work has suggested that open- and closed-class words elicit negative components in the event-related potential (ERP) that differ in timing and scalp distribution. We tested this hypothesis against the possibility that the word-class effects are attributable to quantitative differences in word length and frequency. Event-related brain potentials (ERPs) were recorded from 13 scalp sites while participants read normal or scrambled prose. ERPs were averaged as a function of word class (open vs. closed) and grammatical category (articles, nouns, verbs, etc.). Regression analyses indicated that the latency of an early-occurring negative component was highly correlated with the mean length and normative frequency of words in each grammatical category. Stronger correlations were observed in the scrambled prose condition than in the normal prose condition. Differences in the scalp distributions of these negativities were found to be a function of grammatical category rather than word class. These results are taken to be inconsistent with the claim that open- and closed-class words elicit qualitatively distinct negativities.