All browsers have a default encoding. Some also have an auto-detect. The first non-ASCII character on the present page was one that can't be interpreted in UTF-8 but will work in Latin-1, so that's what I saw, overriding the browser default. Graeme presumably saw the UTF-8 "I can't display this character" question mark inside a black diamond.
Some exceedingly old browsers can't read the charset declaration even if it's present. But that takes you back a very long way indeed. (MSIE 5, say ;))