to futher comment on super's resoponce, we are obviouly only interested in the case where a person answers yes to the first two questions. this essentially limits the possibilies to being only yes to the third question or no, with 50/50 chances. but let's say i changed the questions to be... do you have two children? are both childern boys? are both children girls? we know the answer is yes to the first question and no to the second question. we can have one coin toss per question, but count an answer of yes, yes, yes to be a miss-toss. now if you ran the simulation, it would be 1/3. the questions however are very similar.
two questions would be are both boys? is one a boy and one a girl?
all question answered leaves a a definitve answer
if one is a boy and one is a girl then the child is a boy, if both are girls the child is a girl
two outcomes of even probability = 1/2