In several papers Budzynska and Reed have argued that inferences should be 'anchored' to relations between utterances rather than to utterances themselves; then, by appeal to what they call 'dialogue glue', these relations are somehow reified as 'implicit' speech-acts. In this paper I will argue that this is a mistake caused by confusion over different ways an illocution can be relational and that there can be no such thing as implicit speech-acts as they describe them, and so the speech-act that they claim is performed implicitly between utterances is actually performed explicitly, if indirectly, at the time of the original utterance. It is here that the inference should be anchored. They also argue that when an arguer's credibility is attacked it is the illocution that is undermined rather than the inference, credibility being linked to the illocution as one of its conditions of satisfaction. When understood in a certain way that I will briefly explain in the paper, I will argue that this is true, that the illocution is undermined, and that this reveals something very interesting about the nature of many of the ad fallacies, but speech-act theory on its own does not support this since credibility is not required for the illocution to be successful. Lastly, Budzynska claims that a speaker cannot testify to his own credibility and that this, rather than being an argumentative circularity, is circularity in the assertion itself, since it mentions one of its own conditions of satisfaction. She calls this 'circular assertion'. While I accept the claim that there is this kind of non-argumentative circularity, I do not find it to be as problematic as Budzynska seems to.