Conventional wisdom holds that verbs are more difficult to borrow than nouns. Recent studies have supported this claim, inferring it from the fact that synchronically almost every language studied contains a larger proportion of identifiable borrowings among nouns than among verbs. In this paper, I demonstrate that, while true, there is a logical fallacy in this inference. Using a large diachronic corpus of Russian texts, I show that verbs have lower turnover rates and, consequently, longer life expectancies than nouns, i.e., they are generally more difficult to replace. I argue that this fact alone could theoretically result in the synchronically observed disparities. The hypothesis of cross-linguistically lower verbal turnover rates, which I propose based on these findings, is difficult to verify directly on a large sample of languages. However, it makes a non-trivial prediction, which can be tested more easily. It predicts that if a contact situation lasted for a while, but ceased to exist several centuries ago, the proportion of verbs borrowed during that period and surviving to the present day may equal or exceed the proportion of such borrowings among nouns. The data found in the World Loanword Database (Haspelmath & Tadmor 2009) are consistent with this prediction, thus providing evidence in favor of the hypothesis.