Although tense errors are one major source of grammatical errors in learner English, there has been almost no work on their detection. Tense error detection seems extremely difficult considering that its determination greatly relies on intention and context. Despite the difficulties, this paper shows that tense error can be efficiently detected by exploiting a linguistic property of English verbs called stativity. Our proposed method predicts the stativity of the verbs and then detects tense errors based on the prediction. Experiments show that it achieves an F-measure of 0.571 and outperforms methods implemented for comparison.