Research on peoples arithmetic knowledge (e.g., 2 + 5 = 7) suggests that it is organized and accessed in a way analogous to other forms of conceptual knowledge (e.g., tulips and daisies are flowers). Evidence for this claim comes in part from research that recorded event-related brain potentials (ERPs) and found that incorrect arithmetic facts evoke a response that is analogous to the N400 response evoked by semantically incongruous words in sentences. Some researchers debate this conclusion by pointing out various differences between responses to arithmetic and language stimuli as well as differences among the studies on arithmetic. These differences could be due to variations in methodology, properties of the stimuli (digits vs. words), or properties of the semantic networks in question (language vs. arithmetic). To examine these possibilities, we elicited the N400 effect in arithmetic by closely following the ERP methodology used in language research and varying the presentation format of the arithmetic stimuli (12 + 3 = 15 or Twelve plus three equals fifteen). A comparable N400 effect was elicited by incorrect answers in both presentation format conditions. We conclude that the N400 incongruency effect in arithmetic is analogous to that in language. However, the peak of the arithmetic incongruency effect occurred about 100ms earlier than is typically observed in language. We suggest that this onset difference is due to differences in the size and the constraints of the arithmetic and the language networks.