The Tsimshianic Language Family.

Marie-Lucie Tarpent, who used to comment frequently here (come back, Marie-Lucie!), gave a lecture for the Sealaska Heritage Institute called The Tsimshianic Language Family, its Ancestry, and Distant Relatives (YouTube; there doesn’t appear to be a transcript, alas). You can read a brief description here, and of course there’s a Wikipedia article. It’s fun to hear her voice and to hear the Tsimshian spoken.


  1. jack morava says:

    Kudos for Marie-Lucie, and to Edward Sapir and Penutian.

  2. Stu Clayton says:

    In the family division section of the WiPe article, I read this:

    # The map is backwards for the Nisga’a and Gitxsan. The boundaries are also wrong. Someone fix the map. #

    I bet she wrote that. Good ol’ Marie-Lucie !

  3. Trond Engen says:

    I watched the whole lecture yesterday. She’s not fond of being the center of attention, and I could feel her pain with both that and the format in the opening minutes. But she got past that and gave a solid lecture. And of course, very nice to both see her face and hear her voice.

    She once told that she has lots of unpublished material that she should get out into the world, All field linguists have, I suppose, but it can’t be easy when she’s retired and without much institutional support. I don’t know what she plans now, but let’s hope that this was part of it.

  4. Back in November 2011 (How time flies! I wonder how my younger self back then would have reacted if a time traveler had given him an outline of what 2020/early 2021 would be like?) I had asked “So: when can I order my copy of AN INTRODUCTION TO PROTO-PENUTIAN: PHONOLOGY, VOCABULARY AND GRAMMAR (By Marie-Lucie and associates)?”. I hope this presentation indicates that such an Introduction is on its way!

  5. Thanks for that! I am glad she’s still working on Penutian. It’s a shame that a century after the Penutian hypothesis has been proposed, it is still in limbo because so few have done rigorous work on it, while too many in the past have made the field harder by sloppy work that needs to be cleared up.

  6. there doesn’t appear to be a transcript alas

    A browser extension can download Youtube’s autogenerated Closed Captions, but since those mangle all the proper nouns it’s not very useful here.

  7. marie-lucie says:

    Dear Hatters,

    I am glad to be back in your company.

    Some time in December I got a message from Alaska saying they wanted me to speak for an hour about the Tsimshianic languages and their relationship with the Penutian group. It would have been nice to get a free trip to Alaska at another time! but now it would have to be through Zoom, a name I had only seen written. Well, you saw the result! Fortunately my daughter, who lives in California, knew how to handle it. She even sent me the camera and its support (?) and taught me the basics. The organizers of the event also set up a “practice run” before it. But I was still very stressed at the beginning, and I did not know if I could use up an hour. In fact I spoke for much longer, even though I omitted a number of relevant topics. Then I had a shock the next day or two when I found that several persons I knew (some of them native friends) had listened to the whole thing! Thanks Mr Hat for giving me this opportunity to renew my Hattic friendhips.

    A few details and corrections:

    – Wikipedia: I don’t often look at pages that mention me or my research topics, but they all need corrections, sometimes severe ones. Indeed the map of the languages is wrong. I showed a map on the written pages I used as a base of my talk, but not this one.

    – Language names: What you heard me say was in the Nisga’a language, not in “Tsimshian” except when I was comparing the two.

    – Penutian: This group and most linguists’ attitude to it remind me of the story of the blind men and the elephant. Few people have worked on the whole group, and the ones who have done so, or even who mention it casually, expect to find the vocabulary and features they are used to. But Sapir called the group a “phylum”, not a family, and he did not think that the rsimilarities he noticed (actually very few) were such as to lend themselvesto the usual historical methods. In fact the historical methods should be 1) compare morphologies and 2) compare vocabularies , rather than 1) compare vocabularies, 2) check the morphologies to make sure they are compatible. It is much easier to make egregious mistakes with vocabulary than morphology.

    Further plans: I have decided to write a book. I have done many conference presentations and i need to put them together in a coherent whole, as well as adding other topics. It is unlikely to be what Etienne suggests! Again, this group is not a single family (like Tsimshianic or Miwokan in which the languages are very obviouly related, and Proto-language reconstruction relatively easy), but still there is a lot that can be said, and sometimes even reconstructed, even without being very ambitious.

    Thank you all for your encouragements!

  8. Welcome back, marie-lucie! We missed you!

  9. What he said, and I’m sure I’m not the only one excited to hear about the book!

  10. Marie-Lucie: Welcome home! I just finished watching your talk today. I wish there had been more time for Q & A at the end.

    Hattics: The flyer listing the nine lectures of this series is here, and a YouTube playlist containing eight of the nine talks here. (I assume there are copyright issues with the “Evolution of Northwest Coast Art” talk.)

    I look forward to watching Ed Vajda on Tlingit with reference to D-Y, Jeff Leer on the Na-Dene languages with special reference to Tlingit, and (time permitting) Marianne Ignace on Haida. There is also a palaeogenomics talk for those of us who enjoy that sort of thing.

  11. Welcome, marie-lucie! I hope you can drop by more often.

    I have a couple of questions for you, which I think would be appropriate to ask here. I recall that you gave a talk once about excluding Kalapuya from Penutian. Do you have anything written about it at an accessible place? Or, can you say something about it here?

    Your point about morphology first and vocabulary second is well-taken. However, one advantage to working with vocabulary is that it provides ample examples for establishing sound correspondences. Morphological material is much sparser, even in such amply synthetic languages. Languages this distantly related tend to have developed some non-trivial sound correspondences. What is your experience in that regard?

  12. marie-lucie says:

    JC, thank you. I took more time than I expected, but no one told me and I did not have a watch, so that may be why there was no more time for Q ans A. But ask away here!

    Y, no time right now but I will reply later.

    Meanwhile, anyone can access these articles, which are still valid apart from the occasional mistake. Please do not quote without attribution!

    – 1997 : Tsimshianic and Penutian: problems, methods, results and implications. IJAL 63:1.65-112

    – 2002a : Tsimshianic l- initial plurals: Relics of an ancient Penutian pattern. Report 11, Survey of California and other Indian languages (edited). 88-108. (from 2000 meeting, Berkeley)

    – 2002b : A pan-Penutian database for comparison and reconstruction: its organization, uses and current results. Report 12, Survey … 119-136. (from 2002 meeting, Berkeley)

  13. Athel Cornish-Bowden says:

    Glad to have you back, Marie-Lucie. I was thinking about you just the day before Hat’s post, and wondering if we’d see you again.

  14. David Marjanović says:

    ^_^ Yay yay yay! *bounce* *bounce* ^_^

    A lot of video has been accumulating lately that I really need to watch. There’s this, Navalny’s “A palace for Putin”, the impeachment videos, the annual non-meeting of the Palaeontological Association…

  15. Trond Engen says:

    @marie-lucie: Really good to see you. And here too.

    Y: Your point about morphology first and vocabulary second is well-taken. However, one advantage to working with vocabulary is that it provides ample examples for establishing sound correspondences. Morphological material is much sparser, even in such amply synthetic languages. Languages this distantly related tend to have developed some non-trivial sound correspondences. What is your experience in that regard?

    m-l: Y, no time right now but I will reply later.

    We can start with this comment from 2012.

    John C.: I look forward to watching

    So do I. Thanks! But the Youtube link doesn’t work. I don’t know if that’s Akismet’s doing.

  16. Trond Engen says:

    … and go on to this from 2017.

  17. A bibliography of m-l papers available on line (but be sure to note her disclaimers in the following comments). There is as well one of David E’s first posts on Kusaal.

  18. marie-lucie says:

    JC, Trond: Thanks for finding those papers. Some are quite old, and most are simply conference handouts, which were never revised as such except in my grammar of Nisg_a’a and some later papers. Anything before 1995 is not on Penutian. This “bibliography” is far from complete, but most of my work is unpublished and hopefully will find a place in the book to come.

    Y: Kalapuya(n):

    I did not “give a lecture” but first presented a joint paper in 1998 (with Daythal Kendall, who had worked on Takelma and at least looked at Kalapuya)

    Swadesh’s “Takelman” (1965) putting together Takelma and Kalapuyan (mostly the Santiam language, the best attested one) could be a textbook chapter on what can happen when lexical comparison comes first. If you read the section(s) on Kalapuya in one of the Smithsonian series (the encyclopedic, multi-volume Handbook of North American Indians), you read things like (I quote from memory) “it is very obvious that these two languages are very closely related”, within a family that is described in two (not just one) paragraphs; Campbell’s book on American languages approves of Swadesh’s “Takelman”, which Shipley (a Maidu specialist) tried to support by producing “some two dozen sound correspondences and a good number of lexical comparisons” in order to reconstruct “Proto-Takelman” (he had already been “reconstructing Proto-Kalapuya” but found “Proto-Takelman” much more challenging as correspondences were more complex than expected). Not a word about any grammatical resemblances, or even the need for some. Those linguists relied on probably the only writer on Kalapuyan before that time, Frachtenberg (1918), who had done some fieldwork on Kalapuyan and other “Oregon Penutian” languages (except Takelma, which was more than competently described by Sapir 1922). So Frachtenberg is credited for having proven that Takelma and Kalapuyan are closely related. F’s 1918 article has to be read in the context of the time, when Dixon and Kroeber had classified Californian languages, including the group they called “Penutian”. F’s article seems to aim to provide a counterpart to D & K for some of the Oregon languages. But it is extremely frustrating. His claim to fame is the list of 50-odd lexical resemblances between Takelma and Santiam Kalapuya (as a comparison, Sapir’s index to his Takelma Texts lists 800 lexical items, some of them quoted under different forms, as with verb conjugations). Many of F’s correspondences are of the type t : d, p : b, and such, which F finds too close to suggest borrowings! Otherwise the article is poorly organized and one wonders what F really thinks: his only evidence is this list of lexical items, but he keeps saying that one should really look at morphological differences, and that is one feature he criticizes Dixon and Kroeber for in their “Penutian” grouping. But he does not give examples, except vague descriptions which turn out to be misleading if not totally wrong.

    So why did Sapir put both Takelma and Kalapuya (separately) with the rest of “Oregon Penutian”? In his correspondence with Kroeber (published by Victor Golla) Sapir mentioned that he was waiting for more work from Frachtenberg (presumably in Kalapuyan morphology), but it is obvious that he did not think much of F’s work. So both languages were placed in Oregon, but not in the same subgroup.

    Meanwhile, only a few people have even tried to study Kalapuyan (Howard Berman is probably the best known one). The most obvious feature in the relationship between Takelma and Kalapuyan is the fact that both have very complex verbal morphology – but in vastly different ways. Takelma verbs, like those of “Core Penutian” (in California), have an IE-like structure with stem modification, plus suffixes for various types of inflection. Kalapuyan verb structure is more reminiscent of Turkish, except that the verb is at the end of (often) a series of affixes, not the beginning as in Turkish. It is not very difficult to separate the Kalapuyan affixes from the stem, but much harder to identify their meaning and function. Recent works by linguists rarely agree on their interpretation.

    As to whether Kalapuyan should be classified as Penutian, it does seem VERY doubtful. It somewhat recalls its neighbout Molalla, but not very closely. Neither of them look at all like Takelma. But “Takelman” is definitely out.

  19. marie-lucie says:

    Most of them are of the type t : d, p : b, etc, which

  20. Thanks a lot!

    The talk I was referring to was “The classificatory status of Takelma”, at the New York SSILA meeting in 2019, which I saw in the abstract book.

    The parallels in the pattern of verbal stem change between Takelma and the southern languages sound quite interesting. I should read more about that.

  21. marie-lucie says:

    I was not able to go to the New York meeting and had to cancel my participation, ahead of time but probably too late for my abstract to be removed. I had forgotten it, but I still intend to present my views on the subject.

    Sapir, who had studied Takelma very thoroughly with the last speaker, wrote in at least two publications that he had been struck by the language’s similarity with Yokuts (Southern California Penutian), but this does not seem to have been noticed, perhaps because the linguists based in Berkeley did not venture outside the borders of the state. He did not go into details, but nobody seems so have asked him to explain. Frachtenberg won over Sapir!

  22. Does Takelma look to you closer to Yokuts than Utian is? The evidence for Yok-Utian always seemed weak to me.

  23. marie-lucie says:

    I have not tried to compare them, since I have concentrated on the Northern potentially Penutian languages, which are less well-known than the California ones. I did take a look at some of them earlier and I intend to take a second look. If you know some of them, and Yokuts, you would be in a good position to look at Takelma! Sapir’s Takelma grammar is in the 1922 HAIL.

  24. marie-lucie says:

    My 2002a paper mentioned above is not only about a Tsimshianic “irregular” pattern but about similar formations in several Penutian languages, with data not only from Takelma but also Yokuts and some Miwok. It is possible that Costanoan data would fit in also. There are some Kalapuyan ones too, which can be explained by borrowing.

  25. marie-lucie says:

    My emphasis on morphology may give the impression that I downplay lexical items. Not at all, but lexical items are not just roots but complete words, with their added non-lexical morphemes. This is not only theoretical, but it is in the nature of many words. So, in studying the occurrence and the meaning of a morpheme or an inflectional category, the instances of it are found in the relevant lexical items. As an example, the interest of the l-initial verb pattern is not just the fact that it occurs in several Penutian languages, including the geographically isolated Tsimshianic ones, but that the items in which it occurs appear to be in cognate verbs belonging to those languages. Same words, same structure, same morphemes, same meanings, occurring in languages thousands of miles apart, are unlikely to have arisen totally independently or to reflect only “massive borrowing”, which is typical of very close contact and assumed to be the cause of numerous Penutian similarities.

  26. An excellent concise explanation of an important concept: “lexical items are not just roots but complete words” indeed.

  27. marie-lucie says:

    Merci LH! Too many times, people relying on vocabulary only use “citation forms”, like dictionary entries, rather than (where possible) the full paradigm of an inflected word.

  28. Again, marie-lucie, is there enough material in the grammatical morphemes, across the different Penutian languages you’ve worked with, to establish robust sound laws?

  29. marie-lucie says:

    Y, I think there is a misunderstanding: I am not trying to establish sound laws using grammatical morphemes, I am first trying to establish which languages or families are related to which within the “phylum”.

    Sound laws can be established from comparison of vocabulary (lexical) items which are likely to be cognate. A lot of cognate vocabulary items also have cognate components (roots and affixes), while borrowed items may look similar but do not have cognate components, or only older ones (for instance, if some modern affixes have been added demonstrably recently to an old borrowing). Grammatical morphemes especially are much fewer than lexical ones, so they can be found as components in many lexical ones if the items are related. So in order to find cognates among the lexical items, we do need to consider the non-lexical morphemes that are part of their constitution, not just their lexically meaningful roots and the meaning of the total words. Once we are pretty sure that the relevant morphemes are cognate, we can look at the sound correspondences. In practice of course, we don’t necessarily consider these aspects separately, at different times, but we should keep their working hierarchy in mind.

    Of course, when the vocabularies of two or more languages, as well as the their grammatical systems, are so similar that that the question of language vs dialect can arise (as in some Tsimshianic), the first step (determining relatedness) can well be skipped, and the correspondences worked out in detail right away, but for more distant relationship possibilities the first step is crucial. A few years ago Victor Golla wrote some thing like: “A relationship is either obvious, or it is forever unknowable”. Indo-Europeanists obviously never received such advice! Neither should Penutianists.

  30. Right, but for cognate grammatical morphemes to be easily recognizable, the sound correspondences should be relatively trivial. So, in IE, it’s easy to compare the -Vs noun suffixes in Latin, Greek and Baltic. On the other hand, the Old Norse nominal suffix -r is also cognate with Latin -us, but that isn’t evident unless you knew that Norse r < PGmc *z < PIE *s, and I presume those sound laws were first established using the comparison of lexical items, not grammatical morphemes.

    So, in Penutian, do all the grammatical parallels you see show more-or-less trivial sound correspondences? And if some show less obvious ones, how were those sound correspondences established?

  31. On the other hand, the Old Norse nominal suffix -r is also cognate with Latin -us, but that isn’t evident unless you knew that Norse r < PGmc *z < PIE *s, and I presume those sound laws were first established using the comparison of lexical items, not grammatical morphemes.
    Not necessarily in this case – remember that Gothic has -(V)s here corresponding to Nordic -r; taken together with the -s shown in other IE languages, it must have been quite obvious what the development was. If you look at an early work like Bopp’s, he discusses /s/ – > /r/ in West Germanic using examples both from lexical items and grammatical endings, so there was no clear hierarchy “lexicon first, morphemes later” in establishing sound laws – where the correspondences were obvious, both were used.

  32. marie-lucie says:

    Thank you Hans. Actually, s to r is also established in Latin intervocalically, as in “flos, floris” ‘flower’ and many others, something well-known to the Neo-Grammarians. I think that “lexicon first” is recent. In courses (or even chapters) on Indo-European meant for beginning students, the focus is usually placed on lexical comparison and the importance of sound correspondences. But the early scholars in the field were writing “Comparative Grammars”, not “Comparative Dictionaries”, to start with. After that, not before, thousands of lexical items needed to be compared, and their proto-forms reconstructed (and the job is far from finished).

    Anyway, Y, I mentioned three papers of mine above, which should be easily available (IJAL and California Survey). These three present both the kinds of data and methodology that I have been successfully using since 1995.

  33. David Marjanović says:

    On the other hand, the Old Norse nominal suffix -r is also cognate with Latin -us

    It gets worse if you don’t have conveniently old documents. Of Modern Icelandic -ur, only the -r is cognate with the Latin -s – the Icelandic -u- is epenthetic, while the Latin one changed from the inherited -o- not long before Classical times (and in fact later after qu).

  34. David: Oh, it’s worse. Because if you only have Modern Icelandic and Modern Romance languages, you have nothing to compare Icelandic /r/ to: nowhere in Modern Romance will you find a reflex of Latin nominative singular /s/ still productively used as a nominative singular case marker.

    Y, Hans, Marie-Lucie: The relationship between Old Norse /r/ and Latin /us/ is a little less straightforward than what you sketch, and in an alternate universe where no pre-Modern Germanic written data are available I suspect historical linguists would not be able to conclusively prove the common origin of Old Norse /r/ and Latin /us/. Their common ancestral morpheme, Proto-Indo-European *-/os/, yielded two allomorphs, *-/as/ and *-/az/, in Proto-Germanic. This is a result of Verner’s law, whereby (inter alia) in pre-Proto-Germanic times word-internal post-vocalic */s/ is voiced to */z/ except if the preceding vowel was stressed in Indo-European. In the transition from pre-Proto-Germanic to Proto-Germanic the inherited Indo-European stress was lost, replaced by fixed root-initial stress, and as a result the distribution of *-/as/ versus *-/az/ was synchronically unpredictable in Proto-Germanic.

    Different daughter languages of Proto-Germanic dealt with the situation differently: pre-Old Norse generalized *-/az/ at the expense of *-/as/, pre-Gothic did the exact opposite (i.e. generalized *-/as/ at the expense of *-/az/), and finally West Germanic generally lost both endings, although the situation there is unclear (if -*/az/ was generalized at the expense of -*/as/, as in pre-Old Norse, then this was possibly a phonologically regular change). As a result of this, without older Germanic data, it would be difficult if not impossible to prove the Proto-Germanic origin of Icelandic nominative singular -/r/.

    I see no reason to assume the diachrony of Penutian morphemes into its daughter languages will be any less complex.

  35. PlasticPaddy says:


  36. David Marjanović says:

    Oh, it’s worse.

    Yes. I mean, you’d more likely have only written Latin than only written Old Norse, but of course the Penutian situation is as you say.

    pre-Gothic did the exact opposite

    That’s an unnecessary assumption, because Gothic had exceptionless devoicing of all word-final voiced fricatives: b > f, d > þ, z > s, g > …still g, because h apparently wasn’t [x] anymore (most likely it was a whiff of a [χ], not raspy as [χ] usually is elsewhere). It seems to be generally thought today that Verner voicing applied word-finally regardless of stress, so that North and West Germanic inherited that situation, but there may not be enough evidence to really test this.

    The Old English -as and Old Saxon -os plurals, by the way, have been explained as coming not directly from Verner-proof plurals in *-/os/, but from plurals suffixed with *=/ses/ “themselves” that sometimes show up as -ā́sas in Vedic. The unstressed version, Vedic -āsas, is then the origin of the Old Frisian -ar.

    (…That paper starts more than halfway down the text. The rest consists of the foreword of a Festschrift, then its table of contents, then another foreword.)

    East Germanic devoiced *-/z/; North Germanic turned it into *-/r/ like every other *-/z/; West Germanic lost it, except that southern WGmc kept it in monosyllabic words and then turned it into *-/r/ – northern and southern WGmc meet in Dutch according to this both monumental and introductory textbook on the phonological and morphological history of the Dutch language.

  37. PlasticPaddy: I had written-

    “nowhere in Modern Romance will you find a reflex of Latin nominative singular /s/ still productively used as a nominative singular case marker”

    -and while diachronically, Modern French “fils” /fis/ indeed derives from Latin “filius”, with final /s/ still pronounced, synchronically, there are no grounds for analyzing /fis/ as anything other than a monomorphemic noun: its final /s/ is neither a case nor a number marker. In an alternate Universe with no pre-Modern Romance and no pre-Modern Germanic data available, only a phenomenally bold (or utterly crazy) scholar would propose that this final /s/ and the Icelandic case-marker -/(u)r/ have a common ancestor: there would certainly not be enough evidence to sway the undecided.

    David: thanks for the references. Whatever one’s beliefs on these and sundry issues of Germanic diachrony, however, it is clear that all Modern West Germanic varieties, just like all Modern Romance varieties, fail to exhibit a clear-cut cognate of Icelandic nominative singular -/(u)r/.

  38. Y, Hans, Marie-Lucie: The relationship between Old Norse /r/ and Latin /us/ is a little less straightforward than what you sketch, and in an alternate universe where no pre-Modern Germanic written data are available I suspect historical linguists would not be able to conclusively prove the common origin of Old Norse /r/ and Latin /us/.
    But luckily, Indo-European had the good luck of the old stages being amply attested, so that many endings could be seen to be obviously related, otherwise it might never have taken off the way it did. I have only skimmed Bopp, and he seems to have taken Gothic as a stand-in for Proto-Germanic; sometimes not yet knowing all the details may be actually helpful, because it allows one to see the bigger picture. IE was established because so many endings obviously looked related, and that laid the foundations for sorting out the details in all their complexity.
    If all we had was (say) English, spoken French, modern Celtic, Albanian and Hindi, we might still be able to postulate an IE language family, but we wouldn’t be able to reconstruct much of the morphology, and the relationship would be much less obvious.

  39. marie-lucie says:

    Etienne: “I see no reason to assume the diachrony of Penutian morphemes into its daughter languages will be any less complex” (than that of Germanic morphemes).

    I am not at the stage where the fine points of diachornic development can be discussed. First things first: 1. Can the Tsimshianic languages be considered related to at least some of the Penutian languages? – Yes. 2. Can the other Penutian languages be considered related to each other? – For some of them, yes, but others need more work in order to provide a definitive answer, either allowing promising work to continue in one area, or definitively cancelling work that leads nowhere.

  40. I am quite ignorant about the early history of IE studies. At what point did IE morphology begin to be reconstructed? How much proto-morphology had been sketched out by the time the first ideas of the proto-phonology (Grimm’s Law?) had been articulated?

  41. ə de vivre says:

    definitively cancelling work that leads nowhere.

    Is there a lack of this kind of research? (Well, relative to research in historical linguistics in general…) I can imagine there being a hesitancy to take on long-term projects that might only turn up a decisive lack of evidence.

  42. Marie-Lucie: So what exactly will your forthcoming book be about? Tsimshianic and its proven relatives within “Penutian” (Giving the arguments and data in support of your “Yes” to question 1)? The non-Tsimshianic languages within Penutian that are demonstrably related to one another (Giving the data and arguments in support of your “some of them” answer to question 2)? Or will it be about both questions?

  43. This is probably a naive question since most of the linguistics here is over my head, but why is it irrelevant that the s in fils marks gender? I don’t think of it as just part of the word. I see fils and fille as paired. Would French speakers not see it that way? And even if they don’t, wouldn’t we likely be able to understand the relationship of fille and fils, and perhaps suspect the s might have been a grammatical relict, without any historical language, as long as Spanish was also available to us?

    Generalizing and proposing a relationship to Icelandic on that basis would still be a crazy stretch.

    Also why did French use/retain a nominative form in this one word?

  44. marie-lucie says:

    Etienne, I don’t have much in the way of true academic publications (not having submitted many, which was a mistake) but I have presented many papers on these and related topics at conferences, and I need to organize them in a coherent manner. These two major topics, and more, will be included.

    ə de vivre, by “work that leads nowhere” I mean things like Swadesh’s “Takelman”, a blend of Takelma and Kalapuyan which is quite undefensible (see above) but appears to have been uncritically accepted and still has a prominent place in some reference works (eg Smithsonian HAIL, or Lyle Campbell’s 1997 book on Amerindian languages).

  45. For an overview of rhotacism, see:

  46. marie-lucie says:

    Ryan: “fils” [fis]

    Yes, ‘”fils” and “fille” are a pair, although an unusual one.

    According to the TLFI (Trésor de la langue française informatisé), the history of “fils” is quite complex. The -s of “fils” was no longer pronounced in final position as early as the 13th century (e.g. in masculine names like Georges, Jacques, James) but it was kept in this word (otherwise pronounced [fi]) when used as a term of adress and also in the Christian religious context. The “l” was also lost before a consonant, as also in “faulte” (now “faute”, but English “fault”) and similar words. But the final, non-plural [s] was restored much later, as also in “ours” ‘bear’. The [l] was not, in order to avoid confusion with “fil” ‘thread’. (More details and quotes in the FILS entry, #4)

  47. Wow, that’s quite a complex little word!

  48. Thanks, Marie-Lucie! I really like that. It paints a story for me, perhaps false, but consider: parents using an archaic pronunciation with their sons to exaggerate and perhaps make fun of their own status as ancients, (alongside the religious usage, where archaisms are common, perhaps reinforcing the sense) till it stuck, and lingered long after anyone remembered it was an archaism. “Listen, sone.”

    I also put something together that I read about Old French last night while trying to understand the etymology, that I’m sure M-L knows, and perhaps many of you, but I thought I’d share. That the Fitz that came into English was simply carrying an Old French pronunciation of fiz. Even if whatever etymology first taught me about Fitz mentioned fiz, I would have assumed a buzzed pronunciation of the z, and believed that the pronunciation changed when it came to Britain.

  49. marie-lucie says:

    I knew that Fitz was a version of “fils”, but not how it came about, especially the -t-.

  50. Lars Mathiesen says:

    I once saw a claim that fils was singled out on an OF equivalent of Appendix Probi, peeving over the loss of nom.sing. -s in general — so people kept it in that word alone because it was now special. But I’ll take TLFI over hearsay.

  51. @Y: the early history of IE studies
    In the earliest phase, there basically was no reconstruction in the way we have it now. There were comparisons of sounds, words, and morphemes, so one would find descriptions like “sound / word / ending x in Latin is the equivalent of sound / word / ending y in Sanscrit and of z in Greek”, etc. The underlying assumption was that Sanscrit was the closest to PIE, and if you take an early attempt to reconstruct PIE like Schleicher’s fable, it still looks like slightly cleaned-up Sanscrit. Grimm’s law wasn’t really about PIE phonology, it was an equivalence rule for comparison (Germanic /f, h/x, þ/ correspond to /p, k, t/ in the other main PIE languages). Only in the 2nd half of the 19th century, around Schleicher’s time, it became clear that Sanscrit also had undergone developments. The early IEanists had seen that in the European IE languages, /e, o, a/ corresponded to Sanscrit /a/, but assumed that Sanscrit as the oldest language (at that time, it was assumed to be 4,000 years old or older) represented the original situation and the systems of the European languages were due to spontaneous splits. Here the neo-grammarians came in, postulating the regularity of sound changes (so no spontaneous splits), and demonstrating that the phonology of Sanscrit showed traces of previous /e/ and /o/. But even in the magnum opus of neo-grammarian IE learning, Brugmann’s “Grundriss”, there are no asterisked words or morphology, only tables of comparison. He showed what morphology went back to PIE, but didn’t try to reconstruct how it looked (or rather sounded) like in PIE. That became the usual approach only a bit later.

  52. David Marjanović says:

    I knew that Fitz was a version of “fils”, but not how it came about, especially the -t-.

    [ls] > [lts] > [ts], right?

  53. marie-lucie says:

    DM, probably!

  54. Ryan: Native speakers of French certainly perceive that “fils” /fis/ and “fille” /fij/ are related, but there is no other masculine/feminine noun or adjective pair with final /s/ for the masculine and /j/ for the feminine (in fact, some adjectives have final /s/ in the feminine, opposed to a masculine zero marker: cf. masculine “roux” /ru/, feminine “rousse” /rus/ ‘red-headed’: this makes it even more difficult to treat the final /s/ of “fils” as a specifically masculine ending). As result, both “fils” and “fille” are (synchronically!) monomorphemic.

    David, Marie-Lucie: Not quite! The actual sound law is that Proto-Romance /ɲ/ and /ʎ/ (which, in Proto-Romance, had to be followed by a vowel), when brought into contact with a following /s/ through the loss, in the transition from Proto-Romance to (pre) Old French, of an unstressed vowel, yields /ɲts/ and /ʎts/. But /ls/ clusters are unaffected by the rule.

  55. Above: “As a result”, not “As result”.

    So from Latin to Old French the sequence of sound changes must have been: Classical Latin /fi:lius/, Proto-Italo-Western Romance */fiʎos/ (N.B. Both forms were stressed on the initial syllable), (prehistoric) pre-Old French */fiʎs/, Early Old French /fiʎts/.

    The accusative /fi:lium/ yielded Proto-Italo-Western */fiʎo/ (N.B. Both forms were likewise stressed on the first syllable), which one could argue does not need to be starred, since one of its descendant forms, Modern Italian “figlio”, is segmentally identical to its Proto-Italo-Romance ancestral form. With regular sound changes this */fiʎo/ form yielded the Spanish word “hijo”, Portuguese “filho” (/fiʎu/ today, but still realized as /fiʎo/ in Old Portuguese: hence the spelling of the final vowel today) and Catalan “fill”.

    So: I am sure you feel you have had your fill of Romance linguistics for today, but is the filiation of all these words clear now?

  56. Man, I love seeing the details spelled out like that.

  57. Why bare Fitz without de or of after it?

  58. David Eddyshaw says:

    Old French quite often used the oblique case “bare” as a genitive after another noun, as in la mere Aymon “the mother of Aymes.” (Example from Einhorn’s Old French, p17.)

  59. marie-lucie says:

    So the name of the epic “Les quatre fils Aymon” means ‘Les quatre fils d’Aymes”, “Aymes’s four sons”?

  60. David Eddyshaw says:


    Quite a number of Old French masculine proper names decline like that: others include Charles, Charlon and Hugues, Hugon; feminine proper names of this kind go like Marie, Mariain or Eve, Evain (cf the common noun that declines nominative pute, oblique putain.)

    With the collapse of the case system, some names got fixed in the oblique form, or ended up with doublet forms.

  61. Couldn’t Aymon be appositive in all these cases, like, say, les frères Aymon, functioning something like a family name?

  62. David Eddyshaw says:

    Well, it isn’t in that case, at any rate: the epic is about the four sons of Duke Aymes/Aymon.

    Old French certainly can do apposition: Por Aymon, le baron “for Aymes, the baron” (nominative: Aymes, li ber.)

    I don’t know enough Old French to know if you can pluralise proper names in apposition like that, but if you could, the nominative plural would certainly be Aymon. But it’s not a surname in these examples, in any case: it’s the old boy’s given name.

    Nominative la mere Aymon can’t be appositional, of course: quite apart from being the wrong gender, Aymon is also in the wrong case.

  63. Is the example in Einhorn made up? If not, does it refer to the Duke’s mother, or to Ma Aymon?

  64. David Eddyshaw says:

    I don’t know if Einhorn made up that example; but I don’t think there’s any controversy about the oblique case being usable as a genitive in that way. Old Occitan is the same: q’el vol la terra mos enfans “that he wants my children’s land” (Paden, p285.) In both languages, the construction is commonest with proper names which immediately follow the possessum.

  65. The Four FitzAymes.

    Or would it be the Four FitzAymons? Are there any Fitz names that show case? And what case was used?

    I’m fascinated by this site, which gives medieval names organized by country and language, giving the case in which the name was written and the date, but either it doesn’t have a relevant citation, or the answer is that they weren’t declined.

  66. Hmm. Old English had a different case system. It looks like the genitive for Gerald might have been Geraldes or Gerwaldes, based on the Oswald citation at the medieval names site. I don’t find any evidence that Fitzgerald ever took such form.

  67. David Eddyshaw says:

    Not a problem so long as the entire name is French en bloc, which (after all) it has to be, on account of the order of the elements: otherwise it would have been Gerald(s)Fitz. Same principle as O’Connor (not *O’Connors), Macdonald (not *Macdonalds) and Bowen (not *Bowens.) To say nothing of arap Moi.

  68. Early examples are Syre Roberd le Fyz Haim, iesus fiz mari and, in another vein, Fiz a putaines (all from around 1300).

  69. David Eddyshaw says:

    Fiz a putaines looks like a different construction: “son to whores.”

    The case system was getting a bit wobbly by 1300 even in the Frensh of Paris; much earlier in Stratford atte Bowe.

    Sire/seigneur is another doublet arising from the loss of the case system, incidentally: the forms go back to the nominative and oblique, respectively, of the same Old French noun.

  70. David Marjanović says:

    Man, I love seeing the details spelled out like that.

    Me too!

    nominative pute, oblique putain

    *lightbulb moment*

    Attack of the N-Stems! Another Standard Average European feature, it seems.

    Fiz a putaines looks like a different construction: “son to whores.”

    That construction has become quite popular lately.

  71. David Eddyshaw: the distinction in Standard French between “de” (of) versus “à” (to) is somewhat artificial, inasmuch as “à” remains widely used in colloquial spoken French as a genitive-marking preposition, and this usage goes back to Vulgar Latin. So “Fiz a putaines” almost certainly meant “son of whores”.

    And I agree with David Marjanović: not only is that slur becoming more common, but seen from this side of the Atlantic it seems much more widespread in France than here. This perception of mine, assuming it corresponds to the truth, may be due to the fact that in France the slur seems typical of Southern rather than Northern France (Francophone Canada had next to no founding migrants from the “langue d’oc” area) and was apparently common if not ubiquitous in Occitan and Gascon (“Hilh de pute” in Gascon, with “hilh” yet another reflex of Proto-Italo-Western Romance */fiʎo/). Apparently in France this expression is now losing ground to expressions such as “Ta mère!”, where the not-very-virtuous status of the mother is implied rather than said outright. Its spread in France is apparently due to North African immigrants loan-translating an Arabic equivalent, if this article is to be believed:

    David Marjanović again: I would call the N-stems an Old/Medieval Germanic-Old French Sprachbund feature, rather than a European one: It is widely accepted that Frankish loans played a major role in strengthening this declension type in the transition from Vulgar Latin/Proto-Italo-Western to Old French.

  72. David Marjanović says:

    And I agree with David Marjanović: not only is that slur becoming more common

    Two misunderstandings, I’m afraid.

    I meant à for standard de was becoming more common; probably you’re right instead, and that’s just the recency illusion (and second- or third-hand at that).

    Putain is probably becoming more common, but probably not as a slur.

    The most common usage these days seems to be ô, putain as an expression of surprise, exactly parallel to Polish o, kurwa (and not the vocative kurwo for what that’s worth), a synonym of o Jezu (…that’s the vocative, though). Some people repeat it 5 or 10 times when they can’t believe what they’re discovering, the u reduced to a voiceless fricative.

    The second most common usage must be as an expletive, commonly more or less stressed on the first syllable, unlike the reduction of that syllable in the other use. It’s often combined with good old merde. (…The Polish equivalent there is kurrrrrrrrrrrwa, often combined with mać “mother” in suggestive but ungrammatical apposition.)

    Meanwhile, I’m assured the Occitan word for “dammit” is what is probably spelled maquerèu and, though literally “mackerel”, means “pimp”.

    rather than a European one

    Same thing, in that northern French and southern German are the core of the “Standard Average European” sprachbund (very brief introduction here). That’s interesting about the Frankish loans, though – it makes sense, but I would not have guessed.

  73. David Eddyshaw says:

    The Hausa for “Damn it!” (actually, it’s ruder than that, probably more like “Fuck it!”) is Shege! “Bastard!”

  74. marie-lucie says:

    Although I don’t spend much time in France (especially this year), I agree with Etienne about recent developments. ‘À” for standard “de” is quite old, but considered low-class, and “ta mère”, apparently short for “Nique ta mère”, I figured had to come from Arabic, a culture where you most insult a man by insulting his mother or sister. As for the Occitan fish word, it is either a cognate or (more likely) a borrowing of French “le maquereau” ‘mackerel’, which is indeed a slang word for ‘pimp’.

  75. I’ve seen maquereau used as an insult in the Trinidadian Creole English of V. S. Naipaul (The Suffrage of Elvira) along with a few other French loanwords. I was surprised to see them there. I can’t tell from the context if maquereau means specifically ‘pimp’ or something else, maybe ‘dirty old man’.

  76. marie-lucie says:

    If the word is used in a non-French creole context, it may reflect a regional usage, in a French creole, rather than in non-creole French.

  77. The odd thing is that Trinidad is separated from the French islands to the North and from French Guyana to the south by non-French languages.

  78. @Y:
    trinidadian musical & performance traditions and language have a pretty substantial layer of french colonial / creole still in them from the long period when the largest part of trinidad & tobago’s white population was french (under both spanish & british rule*). language-wise, that history’s especially visible at this time of year, when jouvay (or j’ouvert, ultimately < jour ouvert) starts the ritual cycle of the final carnival day, with people playing mas' (< masquerade) at dawn, before the first chantwell (< chantuelle) sings in the calypso tents… calypso (/kaiso/canboulay music [< cannes brulées]) itself was originally sung in a french creole (i assume a relative of other antillean creoles), and older songs often have words or lines in it.

    * oddly enough, some of the early tries at colonizing tobago were by courlanders – but for some pesky accidents of history, we could've had a baltic creole in the caribbean!

  79. The Hausa for “Damn it!” (actually, it’s ruder than that, probably more like “Fuck it!”) is Shege! “Bastard!”

    Is this (ultimately) from Arabic شقي šaqīy “wretched, naughty” and “a wretch, scoundrel, brat”?

  80. , kurwa (and not the vocative kurwo for what that’s worth)

    The vocative would be very strange in that context. My intuitition is that the underlying thought is more like “ta sytuacja – to kurwa”, you aren’t adressing any particular individual.

  81. David Eddyshaw says:

    I don’t think so: both vowels are long, for example, and “illegitimate offspring” seems to be the core meaning, from which the – expletive – sense has developed secondarily, rather than vice versa.

    It certainly isn’t a learned borrowing. Still, Arabic words have got to Hausa by very circuitous routes, sometimes. Kasuwa “market” doesn’t look much like the Arabic su:q, for example, but both the initial ka- and the change q -> w make sense when you know that the word was transmitted via Kanuri.

  82. David Eddyshaw says:

    (Replying to Xerìb: can’t edit from here)

  83. “ta mère”, apparently short for “Nique ta mère”, I figured had to come from Arabic

    This expression has certainly been current in French since the 1980s, when I learned it. Is it possible this was brought back to France by the pieds-noirs in the 1960s?

  84. marie-lucie says:

    I doubt it. I was in France at the time and never heard it, or about it. The pieds-noirs were Christians and Jews, but the expression is associated with young Muslim boys from long-term immigrant families, who often do not speak Arabic.

  85. calypso (/kaiso/canboulay music [< cannes brulées])

    I presume the “cannes brulées” origin only applies to canboulay (a term I was unfamiliar with); calypso is from Efik ka isu ‘go on!’, “a common phrase used in urging sb on or in backing a contestant”; see here for details.

  86. John Emerson says:

    “Yo’ mama” from black American vernacular became common enough in white pop and alternative music pretty early, even before rap, and I know American pop has had wide influence.

  87. David Marjanović says:

    The vocative would be very strange in that context. My intuitition is that the underlying thought is more like “ta sytuacja – to kurwa”, you aren’t adressing any particular individual.

    Yes, sorry I wasn’t clearer – that’s the point I was trying to make by pointing out the nominative.

  88. @David Eddyshaw
    I don’t think so: both vowels are long, for example, and “illegitimate offspring” seems to be the core meaning, from which the – expletive – sense has developed secondarily, rather than vice versa. It certainly isn’t a learned borrowing.

    Thanks for this. I was interested in finding an example of the semantic path “scoundrel, brat” > “illegitimate offspring” because of the problem of the etymology of Turkish piç “bastard”, which some say is just Persian پيچ pēč “twisted, crooked, complicated; difficulty, complication”—the idea being that bastard is a “complication”. However, Turkic piç “bastard” shows up in the Codex Cumanicus, which gives it in the Persian rather than the Turkic column. And the other Persian words in the Codex Cumanicus still have their majhul vowels ē and ō (as I saw just now in the entry just below bastardus, for castratus, whose Persian equivalent is given as begaya, that is بی خایه bē khāya “without testicles”). An alternative etymology for piç (first offered in its broad outlines by Hrachia Adjarian?) can be found here:

  89. French “le maquereau” ‘mackerel’, which is indeed a slang word for ‘pimp’.

    Not a unique extension; Italian has stoccafisso ‘salt cod; pimp, procurer’.

    Lunfardo borrowed both (as macró and cafishio, respectively) for its vast array of terms for the sex trade.

  90. John Emerson says:

    Hm. “Stoccafisso” can mean “lutefisk”, but none of the many lutefisk jokes that I know of involve pimps (etc.)

  91. Time to start putting new jokes out there.

  92. marie-lucie says:

    “French “le maquereau” ‘mackerel’, which is indeed a slang word for ‘pimp’”

    While “la morue” ‘cod, codfish’ can (or could) mean ‘prostitute’.

  93. Stu Clayton says:

    Italian has stoccafisso

    That looks suspiciously like the German Stockfisch. At any rate it wouldn’t surprise me to learn that fisso does not mean “fish” in Italian, except possibly in some ethnic sitcom.

  94. Stu Clayton says:

    Oh dear, I find that Fisch and pesce are cognatives. So why not fisso alla veneziana after all ? It’s a free country.

  95. Heh. stoccafisso sounds like the kind of thing a German would come up with who only has an idea how Italian sounds, but doesn’t really speak it: Per me, un stoccafisso e per mio amigo una Bratawursta!.

  96. Trond Engen says:

    Many years me and some friends were at a jazz festival, staying at the campsite and cooking on a gas stove. One of the guys brought a strongly aromatic tørrfisk (or maybe klippfisk) which he took out from his backpack, chipped off pieces from with a knife, and tried to offer to the rest of us as snacks. He cleverly called it stokkafiså, with tone 2 in both elements and the o reinterpreted as a Rogaland dialect feminine definite, making it parsable as something like “the pole farteress”. Strangely, the fish lasted for days, so this ritual kept repeating itself.

  97. The Miracle of the Farts and the Fishes.

  98. He cleverly called it stokkafiså, with tone 2 in both elements and the o reinterpreted as a Rogaland dialect feminine definite, making it parsable as something like “the pole farteress”.

    That’s why language is my friend.

  99. On maquereau, TLFI etymology says:

    Orig. controversée. Selon l’étymol. traditionnelle, emploi fig. de maquereau², ce poisson ayant, selon une croyance pop., pour rôle de rapprocher les harengs mâles des harengs femelles, qu’il accompagne dans leurs migrations

    They go on to explain why that etymology is uncertain. Meanwhile English Wiktionary says, without sources, that it is “borrowed from Dutch makelaar (‘broker’).”

  100. That looks suspiciously like the German Stockfisch

    Treccani says it’s from Old Dutch stocvisch; I couldn’t say whether that’s a calque from German or independently coined

  101. @Alon Lischinsky: I think it has to be a calque, since there is, ab initio, no reason that the first element should be stoc, as opposed to some other descriptor. (See here for information about the origin of the stockfish name.)

  102. PlasticPaddy says:

    “De benaming is afkomstig uit de Middelnederduitse verkeerstaal van het handelsverbond Hanze, en daaraan eveneens ontleend door het Hoogduits, het Engels en de Scandinavische talen.”
    So it comes from “Hanseatic” Middle Low German, thence to German, English (?) and the Scandinavian languages. However another source on the same site says that the Scandinavian forms may be independent, although attested only much later.

  103. John Cowan says:


    Mackin’ is of course what a pimp does. WIkt says it is a clipping of mackerel, transparently a calque. Mack the Knife was not Scottish, at least not in the U.S.

    not-very-virtuous status of the mother is implied rather than said outright

    In NYC Spanish, which is mostly a mixture of Puerto Rican and Dominican, just using su madre in contexts that look like simple reference is in fact so insulting that it is automatically replaced by su señora madre, which is then calqued into English as your lady mother.

  104. marie-lucie says:

    Y, back to Penutian etc.

    In looking through my old conference papers I found some Mutsun data, compared with Miwok ones, but at that time I was not concerned about Takelma and its possible California relatives. I need to find some more Costanoan data to add to the Miwok and Yokuts ones.

    Another thing, I am planning to attend (through Zoom) the WAIL conference in a couple of months. I tried to register right away but they require you to do so through a Google account, which I don’t have, and I was unable to arrange to get one. I can’t be the only one with this problem!

  105. David Marjanović says:

    It’s absolutely bizarre that they require a Google account, but go to and get yourself an account there.

  106. marie-lucie says:

    Thanks David. Perhaps the organizers have google accounts and they assume everyone else does! This is a small local conference, the organizers are not professionals like those of larger conferences.

  107. Owlmirror says:

    A Google account doesn’t necessarily mean a Google e-mail address. You can use a Yahoo e-mail to set up a Google account. I think it works for e-mail addresses for other systems as well.

  108. marie-lucie says:

    Thanks, I have never tried it before.

  109. David Marjanović says:

    Oh yes. Setting up a Google account creates a Google e-mail address, but you never need to check it. 🙂

  110. No, it doesn’t require nor necessarily create a google email. I have a google account for which I log in via my email.

  111. Gary Moore says:

    John Asher Dunn asserted that Tsimshian was an aberrant Indo-European language. Dunn was not correct in one key area: Tsimshian is not an Indo-European language, but Indo-European instead has a broad affinity with languages of North America, especially the proposed Almosan–Keresiouan phylum and apparently with with Penutian as well – Greenburg’s hypothetical Northern North American group. One study found that Indo-European grouped more closely with the Chukotko-Kamchatkan languages than Uralic, and Murray Gell-Mann, Ilia Peiros, and Georgiy Starostin proposed grouping Chukotko-Kamchatkan languages and Nivkh with Almosan. (See “Support for linguistic macrofamilies from weighted sequence alignment” Gerhard Jäger, Department of Linguistics, University of Tübingen All this may sound strange but ancient DNA analysis shows that the people of the Yamna culture thought to be the earliest Indo-European speakers are share a good deal of “Ancient North Eurasian” ancestry with Native Americans and were influenced by a migration from Siberia during the Neolithic. While the widely-accepted urheimat of the Proto-Indo-European language is located in the Pontic Steppe region north of the Black Sea, it appears that the roots of PIE may go back to the northern portion of the Pacific Rim.

  112. John Cowan says:

    John Asher Dunn

    A name we generally Do Not Speak here at the Hattery, for fear of spreading his pseudoscience even further. But you are forgiven … this time. (polishes bat)

  113. David Marjanović says:

    and were influenced by a migration from Siberia during the Neolithic.

    Oh no, Ancient North Eurasian ancestry came to Europe at the end of the Paleolithic.

  114. marie-lucie says:

    “Tsimshian” is NOT an Indo-European language

    Gary Moore, you are right that “Tsimshian” is NOT an Indo-European language. JAD’s Tsimshian work is not very reliable in other aspects either.

    The “Penutian phylum” set up by Sapir (1921) as an extension of Kroeber & Dixon’s group of California languages has been my basic research focus for over 20 years, first starting with Sapir’s inclusion of “Tsimshian”, which was long doubted (including by myself until I became convinced that Sapir was right (Tarpent 1997). My continuing research keeps confirming this conviction.

    It looks like what GreenbErg means by the name “Penutian” is basically Swadesh’s later “Macro-Penutian” grouping, which added some Meso-American languages to the North American group set up by Sapir, whose 1921 grouping is how the word “Penutian” is understood by most specialists at present.

    I have Ruhlen’s two books explaining and adding to GreenbErg’s work and quoting a number of GreenbErg’s “etymologies”. I can provide alternate data, transcriptions and explanations for a number of the “Tsimshian” data among them.

    As for adding IE and some East Asian languages to the variously suggested groups, I am not competent to judge either the linguistic or the paleontological aspects, but the “Chukchee” grammar published by Boas (1922) alongside some North American languages does show a number of general resemblances with the latter.

  115. David Marjanović says:


    Oh dear, I didn’t even notice. I’ve gone native in English. 😮

Speak Your Mind