It has been suggested that when additive particles are obligatory (e.g. "Aisha sang. Bertrand #(also) sang"), it is to avoid an unwanted exhaustivity effect (Bade 2016, a.o.). This poster considers how exactly this comes about, both in terms of ensuring the sentence bearing the additive ("S2") is compatible with the previous discourse ("S1"), and that S1 is compatible with S2. I follow Aravind & Hackl's (2017) suggestion that S1 is not strengthened to be contradictory of S2 due to domain restriction, and give new evidence from negative sentences to confirm this hypothesis. However, I also show that this domain restriction cannot be triggered by the domain of the QUD (pace Aravind & Hackl 2017), so that it must instead be brought about by the additive itself.