I'm not sure if I'm atypical, but I do remember having Hugh Woodin give a talk about set theory that I, as a non-specialist (heck, as a computer scientist moonlighting in math grad courses to keep my brain on!) could follow reasonably easily. I still don't know what projective geometry is, but it wasn't necessary to follow his argument.
There are some mathematicians who make big efforts to make their talks accessible to a wider audience. Inexplicably, at least in the mathematics department where I did my PhD, those kinds of talks were actually looked down on by many people.
Same reason most postmodernism and literary theory uses obscurantist language. People value more what they had to work harder for, whether it's understanding or a mug. Making something comprehensible to a broad audience doesn't do anything for the people whose opinion really matters to you, your colleauges, and it makes their work less prestigious by association, because its theoretically comprehensible.
[Rant about Edward Said, and how he said some worthwhile stuff, then made it virtually incomprehensible thought better of]
