Many languages have devices that affect the default discourse effects of certain speech acts. In this paper, I discuss the effects of the Mandarin discourse particle "ba" on questions and assertions. This particle appears to weaken the forcefulness of assertions, but it has been claimed to have the opposite effect on questions, making them more forceful: "you must answer this question". Zhu (1999) accounts for this contrast by postulating a lexical ambiguity: ba1 in declaratives, which weakens an assertion; ba2 in interrogatives, which strengthens a question. I will discuss new data showing that ba-assertions are not always ‘weak’ and ba-questions are not stronger than unmarked interrogatives in the same context, which is not explained by current theories. I present a unified account of "ba" in a commitment-based discourse model (Farkas and Bruce 2010). I propose that, uniformly, "ba" signals that the speaker does not undertake any new commitments (so its only effect is to steer the conversation) and that the ‘weak’ and ‘strong’ contrast comes from the requirements that ba imposes on the previous discourse.