This article deals with the Roman grammarians’ interpretation of impersonal passive constructions and verbal government. A direct object depending on impersonal passive verbs is unattested in Classical Latin, whereas it is possible in Early and Late Latin and is treated as a feature of colloquial language. I suppose that the passage concerning the passive verbs in Ars de nomine et verbo (GL 5. 372. 35–373. 20), written by the 5th-century A. D. grammarian Consentius, provides evidence for such a usage. The examples from Terence (per quem res geretur maxime), Cicero (rem agi), and Virgil (iam tempus agi res), the meaning of which remains unclear for Consentius, are discussed in this passage. Each of them contains a form of the noun res and passive-looking verb forms geretur and agi, which may have either passive or impersonal meaning. In the latter case the noun inevitably becomes a direct object. According to Classical Latin grammar rules, such an impersonal meaning is impossible, and, consequently, the forms of the noun res are subjects of the subordinate and infinitive clauses. The passage discussed in this paper should be included among other 5th-century Latin examples in which direct objects depend on impersonal passive verbs. Besides, the passage is remarkable from the standpoint of ascribing a colloquial feature to the standard classical texts which were read in Roman schools.