I am working on a way for scientists in my field (archaeology) to evaluate how many words they have published per days they have spent doing fieldwork. Some people do lots of fieldwork, then don't publish much, which is irratating. I would like to quantify this:

Words published on excavation A/(days worked in the field in excavation A × number of workers)

A number of adjustments to this ratio are necessary to make it comparable. For example, some archaeologists work longer days and this needs to be accounted for. (For example, I assume 7 hours of work per day. So seven 8-hour workdays would be adjusted into eight 7-hour workdays.)

Another thing that I believe needs to be adjusted is that some languages almost always take more words than others to say the same thing. For example, as a French native speaker, I have long noticed that French usually takes about 10% more words or characters than English. A number of non-peer-reviewed sources online seem to confirm this (e.g. https://www.kwintessential.co.uk/blog/translation/translation-text-expansion-how-it-affects-design-2).

But I could not find a scientific article (a source that I could cite) that gave actual numbers about this phenomenon. Do you know of one?

  • 1
    ...Not that number of words is a reliable measure of findings. It would be much better to go by number of articles published scaled by the seriousness of the publication, or just go by citations or some other common measure. If you use words you measure verbosity and that's easy to game. Jul 31, 2022 at 3:29
  • 1
    Academia is already enough of a hellhole as it is without even more meaningless rankings.
    – Keelan
    Jul 31, 2022 at 11:10
  • @LukeSawczak I agree that this would measure verbosity to a great extent. But I was not thinking of bashing a project with 180 words published per day when the mean of their peers is 360. It’s more for highlighting excavations that have published 12 words per day when the mean of other excavations is 360. You can make your text twice as verbose, but it is harder to make it 30 times as verbose (although not impossible by giving tedious descriptions of individual artefacts).
    – Pertinax
    Jul 31, 2022 at 13:02
  • There is no language-independent definition of "word", so how would such a study be done?
    – Tsundoku
    Jul 31, 2022 at 18:06
  • It is not that some languages use more words. It's this: in English, we have Noun+ Noun, whereas often in French you are obliged to use: noun OF noun. Over an entire text, all those of's (de in French), for example, can make the text longer than the English. Most translators know that in some circumstances (not all), there is a 10% longer text in Romance languages like French, Spanish and Portuguese. The only way to really do this is to get several texts, have them translated into all the languages and then come up with a percentage
    – Lambie
    Aug 2, 2022 at 15:32

3 Answers 3


This is not a thing that has been done in linguistics, although perhaps it could be done in a limited fashion. The two problems that such a study faces is that there is almost no such thing as "the same thing", and you need an operational definition of "word". It you limit yourself to languages like English or French where "sequences surrounded by space" could be the definition of "word", then that takes care of that problem.

The main problem is defining "the same thing", presumably "expressing the same ideas" but how do you count ideas? Bible translations might then be a good testing ground, since there is (supposed to be) a fairly tight connection between an original text and the corresponding translations. In general, you don't find such things as "The James Joyce Style Bible". Including other widely-translated works such as The little prince; The adventures of Pinocchio; Alice in Wonderland etc. might give you a rough indication of what influence choice of language has on number of words in a text.


Pellegrino et al 2011 attempted to come up with a measure of "semantic information density", getting a variety of translators to translate 20 specific texts into various languages, then comparing the total number of syllables in all of the texts (not words, since those are harder to define).

The results are normalized to Vietnamese, and you can find them on page 554:

their results, table 1

However, the number of texts used, and number of translators used, were relatively small; these numbers might be skewed by the particular individuals involved in the translation process. In the era of massive parallel corpora, scaling up these methods might give a more reliable answer.

  • I remember a translation test that asked about how many words it takes to say in French what it takes 100 words to say in English. The answer was something like 105 on average, but who knows where they got that. And of course it depends on genre. For example, I just compared Notre-Dame de Paris in its original form and some English translation I found on Project Gutenberg and got 169k : 181k respectively, meaning the French is actually denser in that case... Jul 31, 2022 at 3:27
  • It also depends on which way it goes. It's not enough to start with only one language. Instead, you need several works going in every direction. And even then, often translations are so rooted in convention that things could be artificially elongated or shortened depending on how it's typically translated. Averaging it all out can make for a good guess, but difficulties will still remain, especially with e.g. outliers.
    – cmw
    Jul 31, 2022 at 17:07
  • Why are words harder to define? Not sure I agree with that.
    – Lambie
    Aug 2, 2022 at 15:36

Additionally you can read these articles. There no any statistics, but they can explain the state of affairs:






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.