Absolute denial macro can be an artifact of being Bayesian-rational, and being absolutely convinced (P=1, or ridiculously close to it) about something that just happens to be false. If you use your brain's natural ability to generate most plausible hypotheses consistent with data, and P(arm is not paralyzed)=1, then P(it's daughter's arm) > P(arm is paralyzed), so this hypothesis wins. If it's disproved, you just go for the next hypothesis, and you have plenty of them before you have to go for one with P=0 that happens to be true.