It would be interesting to run your query as thus, going back farther than you did initially as well, (I'd go to around 1,000 BC, the time of Homer)

"(year | in | during | of) [n]" AD

and

"(year | in | during | of) [n]" BC

Since you weren't returning how many results were returned, just a ratio, this should be a more accurate representation. It will weigh more fairly on older dates who are usually referenced in a historical context.

Very interesting btw!

Peaks at all the round numbers (1,100, 1,200..., 1980, 1990...)

Peak at 1024 (2^10)

Peak at 1066 http://en.wikipedia.org/wiki/Norman_Conquest , 1812 http://en.wikipedia.org/wiki/War_of_1812 , 1933/1934 (Hitler's rise), 1964 and 1965 (The Beatles?), 1996 (?)

Very neat. How about a vertical axis, with bars being links to query on Google (or Vivisimo, or another clustering engine)?

Michael, I suppose 1996 simply was when the Web took off (?). Yes I noticed the 1024 too, though I wondered what it did in there – after all I tried to catch years mostly using the words "year", "in", and so on...

Alterego, adding the word "AD" is a fantastic suggestion. I will run a new mass-query like the following:

"120 AD"
"121 AD"

Also I will take the suggestion with vertical axis and links into consideration. Let's see how this one will come out.

The post at http://blogoscoped.com/archive/2004_07_26_index.html#109085348363924884 has been updated with a link to a second timeline:

http://blogoscoped.com/google-years/index-2.html