This neat diagram was brought up on english.SE from wikipedia, based on research by Finkenstaedt, Thomas and Joseph M. Williams describing where words come from.

enter image description here

On the wikipedia page it also states that latin also includes specialized medical terms, which kind of defeats the purpose in regards to giving an impression of word origins. Is there any source that describes the exact same thing, but multiplies each word 'value' by the amount it is used? Or put differently, taking the entire theoretical corpus of English literature and for each word counts it's origin (germanic, latin/french, unknown, etc.), regardless of whether it already got counted it before.

  • 1
    @jlawler Well, I pressume that the research takes a bit of a simplistic approach to any of those situations ignoring specific uses or multi-step etymologies, because otherwise parsing the entire language is impossible. And just to be fair and clear, I am not looking for a pie chart or super fancy details, just some sensible statistics expressing how much of normal English is based on latin/French and how much on Germanic languages, as my expectations didn't match up with the stats from that research (germanic influence being so minimal :S ). Commented Jun 11, 2015 at 22:09
  • 1
    Easy homework assignment I used to give my freshman etymology class: pick a favorite paragraph from your favorite book and look up the etymology of every word in it. Any decent dictionary will point to either an English source or a French/Latin source. It usually works about 50-50. Most of the Latin/French words are content words, and so are quite a few Germanic ones; but all the grammatical words (articles, prepositions, auxiliaries, etc.) are Germanic. English is a black-sheep Germanic language, sneaking into alleys worldwide and stealing words. They call it borrowing, but it's mugging.
    – jlawler
    Commented Jun 11, 2015 at 23:05
  • 4
    If the interest is purely in knowing the ultimate origin of the words, it makes sense to group French and Latin together. When looking at register, though, it seems to me that a lot of originally French words are indistinguishable in formality from Germanic ones. It seems odd to group common words like "flower", "beast," "around", and "across" with Latinate words like "flora", "fauna", "via" and "ergo" rather than with Germanic words like "bloom", "deer", "through" and "about". Commented Jun 12, 2015 at 0:01
  • 1
    @LuísHenrique: Like this, for example.
    – jlawler
    Commented Sep 19, 2016 at 14:00
  • 1
    @jlawler - Yes, like that. Commented Sep 19, 2016 at 17:02

3 Answers 3


You're looking for the Google books n-gram viewer. :-)

Note that you can specify start year and end year.

See also:

If you are technical, you can use it at a scale large enough to answer your question with Google BigQuery and the public tri-gram sample data.

However, it is also an imperfect data set, and in any case there is no true count of a word's use, it all depends what you include and how you weight it.

You may get more concrete answers on https://opendata.stackexchange.com/ or http://datascience.stackexchange.com.


If you want to establish the origins of a language, there are two main things which have to be taken into consideration: basic vocabulary, and grammar. The former is what would weigh the most, were one to engage in evaluating the average occurrence of each word in the literary corpus of a given language, as you suggested in the question. In the case of English, this is overwhelmingly Germanic. The fact that in the total vocabulary of a certain language at a given time in history there may indeed be more words of foreign origin should not surprise you. For instance, were someone to have created a similar pie-chart for the Romanian language spoken (more than) one or two centuries ago, the majority of its total number of words would have been of Slavic origin, despite the fact that Romanian is a Romance language, descending from Latin, which can be quite easily noticed by evaluating its basic vocabulary, and grammar.


The real difficulty here is to create or acquire a representative corpus of "English". There are many variables to control: American vs. British (or even other, what about Indian English or Hongkong English?) usage, spoken vs. written language, time (do you want a specific time slice, or should it comprise all the years since, say, 1800?), register and genre, and more.

Once you have a satisfactory representative corpus, the rest is easy: pick a sample of enough (but not too many) words at random, determine their etymology, and count.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.