<p>What is said by a sentence seems to vary systematically with the question the sentence is used to answer. To explain this, Schoubye and Stokke have developed an account according to which what is said by a sentence is always an answer to the question under discussion, namely, the weakest answer that entails or is entailed by the sentence’s minimal content. This paper presents a new challenge for their account; the challenge consists in avoiding implausibly weak predictions. The paper also explores different approaches to solving the challenge. It argues that the challenge should be solved by modifying Schoubye and Stokke’s account, and it proposes and defends a new, modified view that solves the challenge. According to the proposed modification, what is said by a sentence is an answer to the question under discussion that is in a certain sense as <i>close</i> to the minimal content as possible. The paper also discusses whether the key to solving the challenge lies in adopting a specific view of what propositions count as answers to a question under discussion; that way, the paper also sheds light on which conception of answerhood is best suited for the purposes of the theory.</p>