Recently I got into a discussion with my friend concerning sizes of lexicons of different languages. He stated something about Japanese having considerably more words than English. (The exact languages don't matter, my point is about assessing the size of a lexicon)

My friend's opinion was based in a big part on the difference between sizes of dictionaries which seem inadequate to me as no dictionary provides the whole lexicon. (AFAIK)

I am very sceptical about any specific assessments of the number of language's words, but I have only my common sense to support that. It appears to me that the difficulty behind that task is based on the impossibility in determining which lexemes to consider "alive", being actively a part of the language at a certain point in time, and which to consider archaic and no longer productive nor being part of the language at this certain point in time. The other difficulties, in my view, would appear in deciding whether a certain recently borrowed word is already a part of the lexicon as a proper loanword or merely a temporal influence that is very likely to fade in usage after a couple of years . One would also have to strictly determine which specific dialect of the language is to be analised. And then it overlaps with the perennial struggle in delimitation within languages and lects.

All these problems seem to be mere manifestations of the most basic and obvious fact of language being in constant change. That inherent feature of natural languages, in my opinion, doesn't allow for any accurate estimations of the size of lexicon.

Can you provide any literature on that topic ? Or maybe there are some loopholes in my thought process or a consensus in linguistics that I am simply not aware of ?

Even if there was a method of producing a safe assessment, would the margin of error of these assessments allow to state that "Language X has more extensive vocabulary than Language Y" ?

  • 1
    "Word" doesn't really mean what you appear to think it does. There are a lot of different kinds of languages, and some have almost infinite capacities for word formation, like this Turkish word (which is also a sentence, like many words in heavily inflected languages). In any event, counting -- and especially comparing -- such incommensurable concepts of "word" does not seem possible. – jlawler Sep 5 '15 at 17:59
  • In this case by word I mean lexeme, I should have cleared that out. – czypsu Sep 5 '15 at 18:17
  • 1
    You might want to compare this: linguistics.stackexchange.com/questions/9674/… – fdb Sep 5 '15 at 19:14
  • 1
    There are still problems with lexemes. These are all theoretical entities, not natural ones. For instance, are the KL- words independent lexemes? And how do you count bear (n), bear (v), bare? Then there's Lushootseed, with CVC roots that can turn into noun or verb at will; polysynthetic languages are all affixes piled on one another. – jlawler Sep 5 '15 at 19:23
  • 3
    Once again I'm surprised that people are down voting a question such as this. This is precisely the kind of thing that it's really valuable to on SE with answers that address it in a serious, careful fashion, so that the many non-specialists who wonder the same thing can both see a good answer and get an idea of the linguistic methodology behind such an answer. +1 to bring it back up! – Gaston Ümlaut Sep 6 '15 at 1:15

There are a number of loopholes. One depends on how you define "is". "Glit" is not a word of English, but it could be. There are 12-syllable words in English but no 30 syllable words -- but there "could" be. Since I imagine there are speakers of English who can say "I'm catching the train to Taumatawhakatangihangakoauauotamateaturipukakapikimaungahoronukupokaiwhenuakitanatahu", then you either have to admit that as a word of English, or say it isn't English because it is etymologically Maori. The underlying premise if a claim about "words of English" is that English is a well-defined entity, and while I think my English is well-defined (though ever-expanding, so well-definition is time-variant) 'cuz it's the language that I speak, there is no such thing as "universal English", and that goes for all languages.

The second regards roots vs. words. If "glit" is a verb, you can form multiple inflected forms from it such as "glits", "glitted", "glitting" and so on. Some people don't want to include inflected forms because they aren't "really" different words, they are just forms of one word. Which then requires you to ask "What do you mean by 'word'?". There are languages with very robust morphology which allow huge numbers of words, and languages with recursive or at least iterative morphology, which have infinite word-formation potential.

In the realm of questions about what constitutes a "word", I believe from his comments here that John Lawler has a position to the effect that Lushootseed utterances are not composed of multiple words, they are composed of multiple morphemes, and "word" is not a useful concept in analyzing the structure of the language (leaving aside a possible and under-investigated stress-related facts). It is certainly not obvious that a distinction between "sentence" and "word" is mandatory for the language.


So if you are only looking at a count of roots, and if you are taking the I-language perspective, then the repository of roots in an individual's mental representation of their language at a given time is finite, thus countable. That is, no lexicon (repository) has an actual infinity of representations. OTOH, there is no specific number of items which defines the largest possible number of entries in a language repository.

  • As I responded under jlawler's comment, by word a mean lexeme/root, that covers the second paragraph of your answer. What I understand from the rest of it agrees with what I think, that assessing the number of words in a language is impossible. I also agree that that kind of thinking is not effective when talking about polysynthetic languages, but for the sake of this question let's focus for example on IE. – czypsu Sep 5 '15 at 18:20

Yes, following the reasoning of The Vastness of Natural Language by Postal and Langendoen, a natural language has an infinite number of words.

I don't recall any good argument from Postal and Langendoen's book. I do recall they say "great grandfather" is a lexical entry and any number of "great"s can be added on to form other words. It seemed to me to be a silly argument, I recall, when I read it. I don't accept their claim.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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