Metaphor understanding in Computational Linguistics has largely been focused on the development of stand-alone prototypes for which only small-scale evaluations are carried out. This has made difficult the inclusion of metaphor in the development of natural language processing applications. However, dealing with metaphor properly is ultimately crucial for any automated language technology that is to be truly human-friendly or able to properly appreciate utterances by humans. This paper proposes to bring metaphor into the Recognizing Textual Entailment task. By doing so, the coverage of textual entailment systems would be broadened and metaphor research would benefit from the textual entailment evaluation framework.