In sociology and in the wider academia, caste is regarded as the basic unit of the Indian society. Yet, this academic notion of caste, I argue, had emerged only in the late 19^th century in the course of the census operations in colonial India. Hence, the pervasive scholarly practice of translating varna and jati, indigenous social forms whose existence can indeed be traced from the earliest times, as caste, is erroneous; it has led to inconsistencies in the writings on the subject. To illustrate, I analyse the writings of the eminent sociologist M.N. Srinivas in this article. Srinivas’s explanation of caste was caught up in an inherent contradiction that led him to make inconsistent inferences. Srinivas was not unaware of the problem. However, the idea of ‘system’ that he had imbibed from the discipline of social anthropology prevented him from reviewing his historically received notion of caste. Hence, those inconsistencies continued.