This article argues based on some concrete empirical evidence that what is called “(social) commitment” is grammaticalized in human language at the latest stage in the evolution of syntax. It is further argued that as a result of this grammaticalization process, our ancestors acquired a way of making their linguistic communication sufficiently trustable/reliable, by encoding the signaler’s liability to the truthfulness of what is communicated. That is, the presence of commitment as a concrete grammatical element provided our species with a way of (indirectly) solving the problem of dishonesty of linguistic signals. The proposal is made in such a way that its validity can be tested by experimental means, and hence it is hoped that the model presented in this article facilitates important collaborative works among theoretical linguists, (evolutionary) biologists, and other experimentalists. Overall, the idea laid out here aims to bridge the gap between formal linguistics and language evolution studies.