1. The WordReference Forums have moved to new forum software. (Details)

Lexical vs. Grammatical Comparison

Discussion in 'Etymology, History of languages and Linguistics (EHL)' started by dihydrogen monoxide, Jul 16, 2008.

  1. dihydrogen monoxide Senior Member

    Slovene, Serbo-Croat
    Since I don't think Kanes will provide you with that, I thought I'd give it a chance. Perhaps you've heard about Ludolf's rule. He was a 17th century German linguist, he was mainly interested and worked with Ethiopian and Amharic languages. He was the first one to say that languages should be compared according to grammar not according to words. I'll give you an article about it and perhaps if you have a comparative linguistics dictionary I think you'll find it there too.

    Thread split from here.
    Last edited: Jul 16, 2008
  2. Hulalessar Senior Member

    English - England
    Whilst it may be the case that structure is going to be more important in determining whether two languages are genetically related than shared vocabulary, the fact that two languages have similar structures does not mean that they are related. So, whilst linguists warn against reading too much into shared vocabulary they in fact always start their comparison of languages by drawing up wordlists and looking for correspondences.
  3. dihydrogen monoxide Senior Member

    Slovene, Serbo-Croat
    But words aren't enough. My point if it's just the words English is Romance language but if you start comparing the structure it's not Romance at all. That's just an example, I don't want to start this again. Indeed they do that's why we have Swadesh list.
    Words come and go but structure stays the same. Structure is the most stable. You could have an example where certain words just disappear. But then they look into structure where they find genetic relationships. I'm just saying vocabulary is not enough to determine the genetic relationship between languages.
  4. Kevin Beach

    Kevin Beach Senior Member

    Well, structure is not more stable than vocabulary in English.

    Structure changed with the ending of inflexions. Meanings that could be discerned from word endings, wherever a word appeared in a sentence, had to be expressed by word order instead.

    Yet our core vocabulary is still Anglo-Saxon.
  5. dihydrogen monoxide Senior Member

    Slovene, Serbo-Croat
    I was talking in general not English. English was just an example that vocabulary is not enough.
    Structure can change but not as rapidly as vocabulary, so it is more reliable.
  6. Hulalessar Senior Member

    English - England
    Simeon Potter in Language in the Modern World lists the characteristics which demonstrate the degree of relatedness in the following order:

    1. Morphology
    2. Syntax
    3. Phonology
    4. Lexicon
    That lexicon on its own is an unreliable test is, as suggested above, easily shown by the case of English. At the taxonomic level of Germanic versus Romance we know that English is Germanic, though the lexicon is predominantly Romance. At a higher level we also know that the Germanic and Romance languages are genetically related – they are both Indo-European.

    Whilst lexicon is the least reliable test it does not follow that (a) nothing is to be gained by comparing the lexicons of two different languages or (b) if you find that two languages have a similar morphology (or syntax or phonology) that they must be related. A list of languages of the same morphological type would be interesting, but unscientific, just like a list of plants with red leaves. All the characteristics need to be taken into account to some extent. See here for a description of the comparative method:


    It is instructive to note that the book referred to above (first published in 1960) asserts the existence of the Ural-Altaic family. It seemed obvious at one time that Hungarian and Turkish were genetically related because of notable features they shared. Now, however, the general consensus is that there is no demonstrable genetic relationship between Ural and Altaic languages. See here:


    Imagine a linguist with no knowledge of any Indo-European language. If he were presented with Irish and Bengali would he ever be able to show by the comparative method that they were related?
    Last edited: Jul 17, 2008
  7. sokol

    sokol Senior Member

    Vienna, Austria; raised in Upper Austria
    Austrian (as opposed to Australian)
    The most important thing here in this discussion is to keep two things apart:

    1) Genetic continuity and relationship

    We do know for a fact that English, whatever its current status is concerning morphology and syntax, indeed is genetically related to Middle English then Old English then Germanic dialects and then Proto-Indo-European: these links are quite good documented and reconstructed.
    Therefore we know that English is related to German even though structurally the two are quite distant now, further we do know that quite some Romance influence has played a part in shaping modern English, and so on.

    And the most important thing about this is here in this discussion: there is absouletly no point in arguing against English being a 'Germanic' or even 'PIE language': we know that English is that, genetically - while if this were also the case structurally now is a completely different question.
    I chose English here because most people would know some English and therefore can follow my argumentation (or at least, I hope so). The exact same could be said about Bulgarian being Slavic (and PIE), Basque being an isolate, and Finnish-Ugrian not being related genetically to Turkic languages (as mentioned by Hulalessar - referring to recent interpretations of both language groups which once were seen as Finno-Turkic group).

    So, English genetically is Germanic and PIE - I hope you can agree with me on that.

    2) Structural similarity and relationship

    Structural similarity now occurs with many languages who obviously have not been in contact - or at least not in recent times. There even exist structure group names for agglutinative languages, synthetic languages, isolating languages etc. - even subdivisions which only refer to a certain aspect of grammar like ergative-absolutive languages.
    This now has nothing to do with any genetic relationship except for the general human relationship which however is a point here: many structural linguists try to find a basis 'human' grammar - it did not begin with Noam Chomsky's Generative Grammar, but this probably is the best known branch of these universalist grammarians which already has split into a great many sub-branches.

    So a universal grammar would tell us nothing about wether grammatical relationships are genetical or not (contact-induced, that is - and not generally human).
    Sprachbund linguistics (in German: Areallinguistik) is concerned with that: in this case we are talking about grammatical structures shared by speakers of languages which are not genetically related, or at least not very closely related (like Slavic: Bulgarian + Macedonian & Romanian & Albanian & partly Greek - all of these IE languages, but none except the two Slavic ones closely related, and all being part of the Balkan sprachbund).

    There is absolutely no doubt about the Balkanic features of the sprachbund members being closely related (or at least I would say that this already has been established in science; everything is relative ...) - many of them being grammatical features.
    But this does not mean that Romanian no more would be a Romance language, or that Greek and Albanian both would not be any more isolates within the IE family, or that Bulgarian or Macedonian would not be Slavic any more: this would be nonsense, there's absolutely no reason why someone should try to proove that. (Except ideological ones, of course, which I don't and won't discuss here.)

    Therefore please let us keep in mind that there are two levels:
    - genetic relationship and
    - linguistic area relationship
    Both exist on both lexical and grammatical level, but as with genetically related languages a great many words are shared in any case and as with genetically not or not closely related languages words are different or at least have changed their 'looks' over the millenia so that they look like they were different it is only natural that with genetically not (not closely) related languages the most salient similarities seem to occur on the grammatical level if there exists language contact and if we do observe contact phenomena as we know them from sprachbund areas.

    There are also examples of such old sprachbund areas outside Europe, of course; JJ Gumperz did extensive research in India and described some cases of close structural similarity between a Dravidian and an Indoaric language on the subcontinent.
    (Unfortunately I can't quite remember which of his publications is about this case; I am rather sure that it was 'Convergence and Creolization: A Case from the Indo-Dravidian Border' (1971), but it also could have been 'Some Desiderata in South Asian Area Linguistics' (1968); unfortunately I have none of these articles at hand right now.)

    So the point here is, to not let the discussion drift off topic:
    - Genetical continuity is genetical relationship, even if a language has drifted apart from its relatives, even if it has done to a degree like English has, for Germanic languages, or Persian for Aryan ones. The fact that grammatic categories point rather to other language groups does not tell us that there would be no genetic relation.
    - Grammatical similarity alone does not constitute any relationship at all (any genetic relationship) except the general human one. Grammatical similarity as genetic similarity exists in sprachbund areas but does not necessarily constitute a 'breach of continuity', that is a language being part of a sprachbund still can be part of its genetic family.

    A real breach of continuity only occurs with pidginisation being followed by creolisation, but that is a completely different chapter of linguists.
    Last edited: Jul 17, 2008
  8. Athaulf

    Athaulf Senior Member

    Toronto, Canada
    Croatian/Bosnia, Croatia
    What does it mean "predominantly"? Of course, if you assign equal weight to the very basic vocabulary and to words like "contesseration" or "sciolism", then yes, English is overwhelmingly Latin/Romance. But the basic vocabulary of the spoken language is still dominated by Germanic roots, especially the basic function words and closed-class words. Needless to say, these words exhibit regular sound correspondences with other Germanic languages that would leave no doubt about their relatedness with English.

    Furthermore, there is another factor that you omit from consideration, which is possibly the most powerful proof of relatedness, and which could be considered as borderline between lexical and grammatical comparison: suppletion. I remember reading once something a linguist wrote (sorry, I forgot the exact reference -- it was an informal internet post, not a scholarly work): if English and German were poorly attested ancient languages about which we knew almost nothing except the irregular comparisons good/better/best and gut/besser/best, it would be a satisfactory proof that they are related.
  9. Athaulf

    Athaulf Senior Member

    Toronto, Canada
    Croatian/Bosnia, Croatia
    An even more interesting question: how far would linguists ever get with Indo-European comparative linguistics if they could work only with modern languages, without the thousands of years old records of Latin, Greek, Sanskrit, Hittite, etc.? Individual subfamilies that split within the last ~2000 years, such as Germanic, Slavic, or Romance, would still be obvious, but how long would it take for someone to figure out their mutual relationship (if anyone ever would)? And would anyone ever figure out their relation with isolates within the IE family, such as Albanian and Armenian?
  10. sokol

    sokol Senior Member

    Vienna, Austria; raised in Upper Austria
    Austrian (as opposed to Australian)
    I'd like to answer that one concerning Armenian - modern Armenian, without Old Armenian support, and only modern IE languages to compare with.

    First of all Armenian phonetics and phonology does not look IE at all - but Caucasian (there's kind of a phonetic and phonological Caucasian sprachbund - different languages not related to each other at all, or not closely related, with similar phonetics and phonology).
    So there would have to be a linguist even thinking about linking Armenian to Indoeuropean in the first place.
    If such a linguist then would try to proove some links he would have a very hard time at the beginning and probably never convince the sceptics even though there would be quite some similarities - but if you don't have any ancient linguistic documents linguists very quickly counter with arguments (and rightly so) like that similarities might just be pure chance.

    So it wouldn't be easy at all.
    This by the way is one of the reasons why it is so difficult to proof language families with other language groups: because for most languages we don't have documents in ancient languages to proof genetic relationships, or if we have them then sometimes they are in pictographic script like Chinese where we can only reconstruct the ancient pronunciation - to a degree, and not for all pictograms.
  11. Frank06

    Frank06 Senior Member

    Nederlands / Dutch (Belgium)
    Just a note. On the basis of the lexicon, Armenian was originally thought to be an Iranian dialect. It's only later that linguists classified it as a separate branch, and mainly because they started to concentrate upon much more important things than the lexicon.


  12. dihydrogen monoxide Senior Member

    Slovene, Serbo-Croat
    That reminds me, it was the same with Hittite, they didn't think it was an Indo-European language.

Share This Page