Mother Tongue 22 Holst Lead Article

Click here to view this article in the published journal format.



Heinz-Jürgen Pinnow has passed away. He was born on January 22nd, 1925 in Danzig,
studied and became a professor in Berlin, and died on the Frisian island of Sylt on July 1st, 2016.

Heinz-Jürgen Pinnow started out with work on languages of India and adjacent areas.
Later he moved on to Native American languages. Here he has made far-reaching contributions to the question of Na-Dene, in fact so many as probably no other scholar, considering the fact that he wrote several monographs on the topic. Pinnow was an Honorary Member of the SSILA (Society for the Study of the Indigenous Languages of the Americas). He was a polyglot, very gifted in didactic matters (which can be gathered from many of his works) and extraordinarily well-versed in historical linguistics. Later on in his life, he called himself only Jürgen Pinnow.

A “Festschrift” for Pinnow with contributions from renowned Na-Dene scholars exists: Dürr / Renner / Oleschinski (1995). It says a lot about Pinnow’s life, but it must be
remarked that also an autobiography, Pinnow (2009), is available which informs us that
some statements in Dürr / Renner / Oleschinski (1995) are incorrect according to Pinnow
himself (Pinnow 2009: 22–25). Dürr / Renner / Oleschinski (1995) contains a curriculum
vitae authored by Pinnow himself which will therefore be reliable (Pinnow 1995).

As a reaction to Pinnow’s death, having been a colleague and friend, I wrote a paper
about him and his work: Holst (2017). It appeared in the periodical Amerindian Research,
which, despite its English name, is a German periodical publishing in German. There had
earlier, in 1992, been a mistaken obituary on Pinnow already in Mother Tongue Newsletter 17 (briefly mentioned at Holst 2017: 110). This time, unfortunately, the news is true.

Holst (2017) is a combination and mixture of an obituary, some personal recollections,
a homage, and putting forward some new thoughts on research, for instance on Na-Dene
and on the attempts to combine American language families, most of all Na-Dene and Eskimo-Aleut, with languages to the west of Bering Strait. One of my concerns in Holst
(2017) was to point out that Pinnow was an underestimated scholar, that attention should be drawn to his legacy, and that we should carefully study what he left us. I give justifications in more detail there. The fact that Pinnow mostly wrote in German should not keep anyone off from studying his works. It should, on the contrary, be an encouragement to improve one’s reading abilities in this language, if not yet done. In my view, Holst (2017) exemplifies why after all we write obituaries: it is not only about remembering a dear person, but also to continue, in whatever way, from where this person regrettably had to stop.

The question may be asked how the paper presented here relates to the paper in German, i.e. Holst (2017). There is some overlap, but in fact only extremely little so: what follows here is an entirely different paper.

In the following, two topics intimately connected with Pinnow’s research will be discussed, and new thoughts will be brought into the debate. Section 2 deals with Na-Dene. Section 3 deals with possible language relationships beyond Na-Dene.

Na-Dene (henceforth ND), having been treated time and again, probably requires merely a very brief introduction here. In North America there is the large Athabaskan group of languages (Navajo etc.), and there are three single languages on the West Coast and nearby islands, from north to south Eyak, Tlingit, and Haida. Na-Dene is the genealogical unit that unites these languages.

The ND hypothesis was launched by Sapir (1915). It received support from Hymes
(1955, 1956) with a method he called positional analysis, showing that the morphemes in the verb occur in a very similar order in the languages concerned, which speaks against
chance and for their relationship (see also Pinnow 1976: 47–49). Heinz-Jürgen Pinnow
became the leading figure in the second half of the 20th century to contribute evidence to
Na-Dene; he wrote various monographs as well as articles on these languages.

Greenberg (1987: 321–330) declared himself a supporter of ND. Enrico (2004) contributed new evidence. A couple of other scholars worked on ND as well. As to me personally (J. H. Holst), I underlined the existence of ND on several occasions in my own works; I have not, however, contributed any new evidence anywhere so far.

There have also been, and there still are, doubters as to ND. That Eyak is related to
Athabaskan is generally accepted since the work of Michael E. Krauss. Doubts exist, however, concerning Tlingit (though they have decreased), and fierce opposition still exists in
some quarters to the inclusion of Haida. From my own study of the languages and the literature on them, however, my impression, and my conviction, is that the ND family is real (including both Tlingit and Haida). The insight into this fact is in my opinion open to any trained historical linguist by studying the question – especially, among other sources, by studying Pinnow’s works. For more details justifying this view see Holst (2017: 111–113). The fact that Campbell (1997), for instance, rejects ND, is in my opinion a pity, since many readers unfamiliar with the ND question will turn to works such as Campbell (1997) – and in many other issues they will find authoritative information there, since Campbell is otherwise extraordinarily well-informed about the indigenous languages of the Americas. The non-acceptance of ND by some leading experts on Native American languages – another example is Mithun (1999) – is also remarkable when compared with the fate of Algic (Wiyot + Yurok + Algonquian): in this case Sapir’s starting-point (Sapir 1913) has stood the test of time and the relationship is nowadays accepted by all scholars. The same sort of development should have happened, in my view, to Na-Dene.

The situation just outlined with the ND controversy is probably too well-known to be
laid out in more detail here. It is an interesting question whether this situation will ever
change, and if so, when and into which direction. This is the point where I would like to
add some reflections.

They evolve from an idea expressed in a somewhat hidden place in a book on EskimoAleut: Holst (2005: 230). What I wrote there was that possibly Haida is not opposed to the rest of the ND family but may rather form a branch together with Tlingit. Thus, I drew the attention to subgrouping, and I challenged the established subgrouping of ND. The very short passage adds that checking this issue may give new impetus to ND studies.

It is rewarding to elaborate on this point. The “traditional”, if one may say so, family
tree of ND is (tree A):

Haida Tlingit Eyak Athabaskan]

Based on the idea that actually Haida and Tlingit may form a branch together, the tree Holst (2005: 230) brings into play is (tree B):

Haida Tlingit Eyak Athabaskan]

Haida is usually described as being the most deviating language within ND. However, this fact does not automatically mean that Haida has to be the language that split off first (as tree A would suggest). The special character of Haida may also have come about due to extensive change. Irrespective of whether language contact – a frequent cause in such cases – is at work or some other factor: a language can step out of line from a family, change more rapidly and more profoundly than others and thereby loosen its ties with its relatives. In such situations it may become difficult to determine the position of the language within its family correctly.

Once some sensitivity for this issue has developed, the insight occurs that it is also
possible to readdress Eyak. As mentioned, Krauss demonstrated that Eyak and Athabaskan are related. This does not automatically entail, however, that Eyak must be the closest relative of Athabaskan – something that scholarship so far has tacitly taken for granted. Consider relationships of four languages (of whatever family) with a tree which looks like this:

A B C D]

If now a scholar provides proof (or evidence) that B is related to D, this may be an excellent advance in historical linguistics, but interestingly it does not guarantee that he has hit a correct subgroup. In this hypothetical case correct subgroups would only be A and B, or C and D. Applying this to ND, it means that Eyak’s position could theoretically be elsewhere. Eyak could be part of the other branch (i.e. the one of Tlingit and Haida), or it could be somewhere in the tree outside both branches. The following diagram shows one of the possible positions, next to Tlingit (tree C):

Haida Tlingit Eyak Athabaskan]

There would be several other possible positions for Eyak which will not be illustrated with trees here.

Future research could continue investigating these questions with concrete data. Some
hints must suffice here. There are nowadays Swadesh lists for some ND languages on the internet. A cursory examination of them shows that the traditional tree A is not necessarily confirmed. The data may also speak for tree B, tree C, or another solution. Of course it is not only the vocabulary which is relevant in such issues (and Swadesh lists are moreover only short excerpts of the vocabularies). Considering the prefixes for person presented in a table by Pinnow (2006: 52), it becomes clear that sometimes Tlingit and Eyak exhibit related morphemes which the other languages do not share. This may be meaningful. It should be stressed that these are preliminary observations which may not be decisive. It is only that it is necessary to report them, for further research to explore the matter further, which otherwise could not be taken up.

Thus, the truth seems to be that we do not know the family tree of ND at all. It is time
to realize this crucial fact. A problem in research so far was that a certain tree, called tree
A here, has always been taken for granted – by supporters and by some skeptics alike. The issue should be investigated and the data should be approached in a neutral way. It is possible, and even probable, that new insights arise. This gives new fuel to all debates on ND. Even if it should turn out that the idea to question the established subgrouping ought to be abandoned, there will still be a gain.

It would be possible to use mathematics to study word lists, e.g. Swadesh lists, in order
to see what this yields. However, one must not expect too much from such an investigation. Cognates on the Swadesh lists of the ND languages, if taken as raw material without any deeper understanding of these languages, are only very few. There are nowadays many scholars who may be mathematicians but who unfortunately lack any expertise of the languages whose data they absorb. I hasten to add that the fact that cognates on Swadesh lists are few in the case of ND does by no means signify that these languages are unrelated. Languages can very well be related without this being visible with rather primitive surveys of limited and unanalyzed or underanalyzed data. Moreover, mathematical assessments of word lists, of whichever family, do not reveal a large array of facts. They do not reveal, for example, where shared sound laws exist, as well as all issues with grammar (which are highly important, of course, in relationship issues).

I expressed the idea that the family tree of ND may be incorrect to Pinnow in a telephone conversation some years ago. To report the events in the conversation truthfully: he
tended not to believe in the worthiness of the idea and defended tree A. However, this does not deter me from mentioning the issue here. The impression I get from various ND data has made me return to the idea repeatedly. Even if the basic idea – to question the family tree – should turn out to be incorrect, it may very well be that interesting insights develop along the way. This is not to say, however, that the idea is likely to be incorrect.

These issues lead, finally, to some general thoughts about subgrouping. Interestingly,
there are problems, and often serious ones, with the subgrouping of astonishingly many
language families. This affects every continent. Turkic, for instance, is a language family
for which subgrouping is largely unclear, and there are further examples wherever one
looks. Uralic has only few problems, but some remain (e.g. the position of Mari, also
known as Cheremis). Even for Indo-European we do not know the subgrouping; in recent
decades some progress has been made, but many claims with insufficient backing have
seen the light as well. For each family mentioned, more research is needed. The question
can even be asked the other way round: for which families or branches do we know the
subgrouping with certainty? An example is Kartvelian, but this group is rather small with
its four languages, and it is thus not so surprising that the correct subgrouping has been
known since Deeters (1930: 2). On the uncertain subgrouping for most of the world’s language families see also Holst (2005: 143).

All of this makes the insight arise that frequently subgrouping is crucial in historical
linguistics today. A question is often how to get ahead in historical linguistics. There are
moments in which the impression prevails that no real progress can be made any longer. Then the possibility to (re)investigate subgrouping can frequently be a helpful idea, but it
is often not made use of.

There are several reasons why correct subgrouping is so important. When language
families diverge, sound laws, grammatical changes and other innovations often affect intermediate proto-languages, and it is therefore essential to know which these are. The possibility to identify innovations and retentions is dependent on whether one grasps the structure of the family well.

In the case of Na-Dene, the ideas put forward here could lead either to the old tree, but
with more confirmation then, or to a new tree. In the latter case, the possibility then opens up to reconstruct the proto-languages of subgroups – or, to be more exact, it is more likely that certain features of them will be reconstructed, i.e. not the entire linguistic systems. This is the new impetus for ND.

Pinnow was not only concerned with establishing Na-Dene. He was also interested in the question what ND in turn may be related to; he sought to “look beyond Na-Dene”. This was part of his general interest in distant language relationships. On the one hand Pinnow was very well aware of the fact that attempting to go beyond the established language families often leads into the realm of speculation, on the other hand he was such a knowledgeable scholar that in some cases he was able to adduce tantalizing evidence that is worth being studied by us.

Bengtson (1999: 173) describes this research interest of Pinnow’s as follows: “Pinnow
is a Long Ranger. That is, he allows himself to think and hypothesize about distant relationships between the traditionally accepted language families. He thinks there is evidence for remote relationships between Na-Dene and certain other language families.”

It is natural to look to Asia when looking for relatives of ND – also because the urheimat of ND was in Alaska, thus so to speak at the entry of the continent. In one of our
telephone conversations Pinnow said to me several years ago that there are two language families in Asia he would like to see combined with Na-Dene: they were on the one hand Yenisseian, on the other hand Sino-Tibetan. I instantly agreed since I had arrived independently (but in part by reading his works and those of other authors) that these were the two reasonable families. We both agreed on the phone too that other hypotheses were not as promising: one should add these two families in Asia but at present go no further than that. Note that there is a difference here to Dene-Sino-Caucasian, or shorter, Dene-Caucasian, a more inclusive grouping advocated by other researchers. I particularly would like to deny an inclusion of Burushaski, on the basis of lack of sound evidence. My (rather extensive) work on Burushaski has led me to this position, see e.g. Holst (2014: 16). Moreover, see Tiffou (1995) in the same vein.

As to Yenisseian, Vajda (2010) has done work to combine it with Na-Dene which
raised attention. There is no unanimous agreement on whether Vajda has managed what he intended. Something that strikes any expert on ND with Vajda’s work is that he understands only Eyak-Athabaskan plus Tlingit by the term “Na-Dene”; he does not incorporate Haida. This is somewhat weird and not really pleasant for those who studied the evidence published so far, nor for those who have even worked themselves on ND. As far as I can see, there are two possible reasons for Vajda’s decision. Firstly, Vajda was certainly aware of the fact that influential scholars do not count Haida as Na-Dene and continue to express doubts on its inclusion. Incorporating Haida would have meant for him to get involved into this discussion, which is something he possibly wanted to avoid, since scholars who propose language relationships are often in a situation where they have to defend themselves and thus are not eager to be saddled with additional problems. However, if this applies, it should be remarked that one should stick to one’s convictions. Secondly, and alternatively, it may have been that Vajda did not fully realize that Haida does in fact constitute part of Na-Dene. This, however, would not speak for the work of a scholar who claims to be able to point out evidence for a much more distant relationship, that between Na-Dene and Yenisseian. Either way, including Haida may very well be relevant for Vajda’s case. There is also work attempting to link Na-Dene and Yenisseian, less widely known than Vajda’s, by Ket specialist Heinrich Werner (Werner 2004).

As to Sino-Tibetan, Pinnow contributed very interesting evidence himself. The hypothesis goes back to Sapir, as is well known. Later, Robert Shafer and others added to the evidence. Pinnow (1976: 94–105) has extraordinarily interesting data and observations on this matter.

A detail on the term Sino-Dene, which is sometimes used for this combination, should be intercalated here. I called it (Holst 2017: 113) an infrequent term and a spontaneous creation by Bengtson (1999). The term is indeed not so frequent. However, as kindly pointed out to me by John Bengtson (p. c.), the claim that the term is his spontaneous creation does not hold: the term is actually older, and it was already used, for instance, by Golla (1991: 139) in an editorial note to Sapir’s work. See also Bengtson (1994).

In my personal notes, a possible unit of Na-Dene, Yenisseian and Sino-Tibetan bears the name Lakitisch, thus in English Lakitic. I coined the term from a shared word for “hand” which is lak or similar in these families:

Na-Dene: Haida s-tla, s-tláay (s- is a prefix, lacking in the plural), Mattole la?

Yenisseian: Ket l’aŋat, pl. l’aŋen’ (‘ palatalization, velar nasal ŋ possibly < velar stop)

Sino-Tibetan: Burmese lak, Tibetan lag-pa

All these words mean “hand”. Compare also interrelated words for “five”, which exist in Na-Dene and Sino-Tibetan, such as Tibetan lŋa. Of course this is only a single etymon which is not probative for the genealogical unit, nor is there a need that it delimits the unit correctly. No families outside Lakitic with a similar word for “hand” are known to me. In Burushaski, for instance, “hand” is Yasin -rén, Hunza -ríiŋ. The Hunza form has repeatedly, but erroneously, been compared to Yenisseian (apparently going back to Toporov
1971: 114f.), ignoring the Yasin dialect. However, Hunza -ríiŋ is actually historically a
plural form, to be segmented -rí-iŋ, in which -iŋ is a plural suffix; the Hunza singular -ríiŋ
corresponds to the Yasin plural -réiŋ. In Hunza the singular must have been *-rín. The
reconstruction for Proto-Burushaski must refer to this *-rín and to the -rén of Yasin, and
will then be *-rín, since i is the older vowel with the correspondence Yasin e / Hunza i (Holst 2014: 70–81). For details on this analysis of “hand”, which is unavoidable, see Berger (2008: 97), Holst (2014: 100).

Na-Dene, Yenisseian and Sino-Tibetan all have vowel alternations in roots, often
called ablaut. This is a typological trait. Further research will have to show whether the
patterns exhibit similarities and whether there are differences to ablaut in, for instance,
Indo-European and Kartvelian.

There is one major problem when wanting to evaluate Sino-Tibetan for a relationship
hypothesis: it’s huge. I would like to dwell on this point for a while and study some of its
implications. (Some points are essentially similar for all large language families.)
Sino-Tibetan comprises more than 300, or possibly more than 400, languages. This
means that it is an enormous task to reconstruct the proto-language. A reconstruction
should, if possible, be consistent with all daughter languages. It is true that some languages are in a special situation. Tibetan has an old attestation, is highly important and conservative in many respects – but despite all this one must not rely too much on Tibetan. If one took a form from Tibetan and acted as if it was Proto-Sino-Tibetan, this would be what R. L. Trask has called “reaching down”. Another language with some older documentation is Burmese. Chinese has the oldest attestation but the script does not do researchers a favour. There are competing reconstructions of the phonology of earliest Chinese, and anyone interested is forced to take a stand or to develop still another view. Knowledge of Japanese and Korean, due to their massive loan word layers from Chinese which provide various insights, is an advantage. It is also possible to do fieldwork on various Sino-Tibetan languages (though there are some places where governments do not allow you to travel). Subgrouping of Sino-Tibetan would profit from further research. It will then be possible to reconstruct the proto-languages of subgroups, and later compare such reconstructions with each other. One may wonder whether there are any shortcuts towards a more or less reliable Proto-Sino-Tibetan. However, these may not exist.

Thus, unfortunately, Sino-Tibetan provides more work than what is possible in a lifetime, and a scholar may get stuck within this family and die before he is able to, or ventures
to, “look beyond it”. Consolation may come from the fact that there are often typologically
interesting structures in this family, e.g. cross-referencing systems of person and number,
to name but one field of inquiry. Scholars engaged with distant language relationships often feel that it is more thrilling to do their type of work than to enhance understanding of existing families. In such a situation, a hint to typological beauty makes sense.

There is more to add here. Work within established families is usually more reliable,
and a person can draw much satisfaction from the fact that his (or her) results are rather
reliable, instead of being shaky. An advantage of reliable results is also that it is less likely that colleagues will have different opinions on them. Recognition and praise are easier to obtain here.

This all is of course not to say that one should refrain from any work across the
established language families. Especially when good progress can be made, such work
should be done.

All potential discussions will often be connected to discussions of methods. There is
nothing wrong with this, but this is a necessity, since with flawed ideas about methods one is likely to arrive at incorrect results. A question that is never raised, however, is: Where do methods come from? Were they already there before the Big Bang? This is a highly interesting question. Some scholars act as if methods were fixed and would then just have to be followed, and results will then be arrived at (or the result is that no results can be gained on a particular question). This is not quite the way things are. Rather, it can sometimes be an impediment to view methods as being chiseled in stone. Methods can be subjected themselves to study, to doubts and to scholarly discussion. There are no limits for intellectual freedom. I do defend much of the paradigm of historical-comparative linguistics as it has developed over time, I do subscribe to most of what textbooks such as Campbell (1998) proclaim, and I encourage everyone to take the established methods, with their long history, quite seriously. Sometimes, however, it is a tiny being ahead in openmindedness that enables one to see, or to hypothesize, a point which others may be unable to reach.

It is such open-mindedness that can often be observed when reading Pinnow – which
brings us back to the scholar we started out with and who should be remembered.


Bengtson, John D. (1994): Edward Sapir and the “Sino-Dene” hypothesis. In: Anthropological Science 102 / 3. 207–230.

Bengtson, John D. (1999): In honor of Jürgen Pinnow: comparanda dene-caucasica. On the occasion of his 75th birthday 22 January 2000. In: Mother Tongue 5. 173–182.

Berger, Hermann (2008): Beiträge zur historischen Laut- und Formenlehre des Burushaski. Wiesbaden: Harrassowitz.

Campbell, Lyle (1997): American Indian languages. The historical linguistics of Native America. New York / Oxford: Oxford University Press.

Campbell, Lyle (1998): Historical linguistics. An introduction. Edinburgh: Edinburgh University Press

Deeters, Gerhard (1930): Das kharthwelische Verbum. Vergleichende Darstellung des Verbalbaus der südkaukasischen Sprachen. Leipzig: Markert & Petters.

Dürr, Michael / Renner, Egon / Oleschinski, Wolfgang (1995): Language and culture in native North America. Studies in honor of Heinz-Jürgen Pinnow. München / Newcastle: Lincom Europa.

Enrico, John (2004): Toward Proto-Na-Dene. In: Anthropological Linguistics 46 / 3. 229–302.

Golla, Victor (ed.) (1991): The collected works of Edward Sapir. Vol. 6: American Indian languages 2. Berlin / New York: Mouton de Gruyter.

Greenberg, Joseph H. (1987): Language in the Americas. Stanford: Stanford University Press.

Holst, Jan Henrik (2005): Einführung in die eskimo-aleutischen Sprachen. Hamburg: Buske.

Holst, Jan Henrik (2014): Advances in Burushaski linguistics. Tübingen: Narr.

Holst, Jan Henrik (2017): Heinz-Jürgen Pinnow (1925 – 2016): Zu Leben und Werk. In: Amerindian Research 12 / 2. 110–114.

Hymes, Dell H. (1955): Positional analysis of categories: a frame for reconstruction. In: Word 11. 10–23.

Hymes, Dell H. (1956): Na-Déné and positional analysis of categories. In: American Anthropologist 58. 624–638.

Mithun, Marianne (1999): The languages of Native North America. Cambridge: Cambridge University Press.

Pinnow, Heinz-Jürgen (1976): Geschichte der Na-Dene-Forschung. Berlin: Gebrüder Mann.

Pinnow, Heinz-Jürgen (1995): Curriculum vitae. In: Dürr / Renner / Oleschinski (1995). 480–482.

Pinnow, Jürgen (2
2006): Sprachhistorische Untersuchung zur Stellung des Haida als Na-DeneSprache. Westerland / Sylt.

Pinnow, Jürgen (2009): Berichte aus dem Leben eines Sprachwissenschaftlers. Westerland / Sylt.

Sapir, Edward (1913): Wiyot and Yurok, Algonkin languages of California. In: American Anthropologist 15. 617–646.

Sapir, Edward (1915): The Na-Dene languages, a preliminary report. In: American Anthropologist 17. 534–558.

Tiffou, Étienne (1995): About Trask: Basque and Dene-Caucasian: A critique from the Basque side. In: Mother Tongue 1. 159–162.

Toporov, Vladimir N. (1971): Burushaski and Yeniseian languages: some parallels. In: Poldauf, Ivan (ed.): Études de la phonologie, typologie et de la linguistique générale. Travaux Linguistiques de Prague 4. Prague: Academia. 107–125.

Vajda, Edward (2010): A Siberian link with Na-Dene languages. In: Kari, James / Potter, Ben A. (eds.): The Dene-Yeniseian connection. Fairbanks: Department of Anthropology / Alaska Native Language Center. 33–99.

Werner, Heinrich (2004): Zur jenissejisch-indianischen Urverwandtschaft. Wiesbaden: Harrassowitz