From Wu to Gu

A West-Germanic tendency

Although the title may suggest a blogpost about Chinese dialects, the subject remains in the sphere of Proto-Germanic and its illustrious ancestor. For those of you who were hoping for a whole article on Chinese tonogenesis; tough luck. But let’s get to it, shall we? In the West-Germanic languages we find an interesting tendency to alternate the sequence /-wu-/ with the sequence /-gu-/. This wugu-rule[1] has intrigued scholars for over a century now and no one has been able to formulate a strict soundlaw to explain the data. Consider these examples:

Goth *junda : OE geóguþ “youth”

Goth sauil : OE sugil “sun”

OIcel. sýr : OE sugu “sow”

Goth niun : OE nigon “nine”

Explanations have been sought in the vocalization of the syllabic resonant to *uR, whereafter either an already existing /ṷ/ or a secondarily arisen /ṷ/ intensified to /g/. The rise of the secondary /ṷ/ can be explained as a hiatus filling consonant bridging the two /u/’s. However, considering that in almost all the cases a laryngeal is present before the relevant resonant we may suspect influence of a laryngeal causing a doubling of the morae to /ṷu/, i.e. *uHR > *uṷuR (Müller 2007). It may be that only this sequence was originally eligible for the subsequent wugu-rule. Therefore we could formulate the rule as follows:

Late PIE *uHR > Pgmc *uṷuR > Proto-Westgermanic *uguR

 In this way it wouldn’t be so puzzling that OHG doesn’t participate in the case of “nine” with its reflex niun. In the case of “nine” the Ingvaeonic dialects may have had an own independent transition of /ṷ/ to /g/. This becomes even more plausible when we take the words into account that  Kluge (1913) added to our list of wugu-alternations. He listed the different reflexes of the words for “eyebrow/bridge” (OE bruggia : ON brú) and “mosquito” (OE muggia : ON < *muwi) as portraying a similar phenomenon. The word for “eyebrow” has a reasonable IE etymology[2], the word for “mosquito” however, is problematic from an IE point of view. The different supposed cognates, both the Germanic and the IE cognates, show alternations in the root-extensions and suffixes ánd seem to be build on an onomatopaeic root which defies a proper PIE root structure. Before continuing a caveat is needed. As the few and badly accepted “supposed” instances of Cowgill’s law indicate, for a real soundlaw one needs a lot more examples. In this case we don’t have them, so that makes this essay a fun excursion but no more than that. Let’s go through the list of the examples which seem to fit this explanation.

PIE *(h1)neṷn[3] > Early Pgmc *neṷun > Late Pgmc *niṷun

Germanic cognates:

Goth.      niun                                       Ofris       nigun, niugun                    

OIcel.     níu                                         OHG       niun

OE         nigon                                   

Interesting in this example is that it may provide a clue to the dating of the phenomenon. The fact that in al the reflexes Proto-Germanic *eu had already changed into *iu after u,i,j,w, indicates that the transition of /ṷu/ to /gu/ must postdate this soundlaw. Therefore the change has taken place relatively late in the chronology of Germanic soundlaws.

PIE *h2iuHn-ti[4]> Pgmc juṷunþi

Germanic cognates:

Goth       *junda ( .jundai)      OLG        juguð

OE          geóguþ                           OHG       jugund

Gothic *junda, only attested in the dative singular jundai points to a simplification of the sequence /uṷu/ to /ū/. A problem in this reconstruction is the place of the PIE accent. The effect of Verner’s law indicates that the accent must have preceded the nt-suffix and therefore would have lain on a zerograde. Müller (2007)  and others consider an early Proto-Germanic movement of the accent to the right before the shift to the first syllable. This would put this accentshift really early in the Germanic chronology because the communis opinio is that the Germanic auslautgesetze came to be because of the weakening of the unstressed syllables. Also it requires some sort of vocalization vowel already before the syllabic resonant, for the zerograde to be able to bear stress. I choose to remain agnostic on the subject, but without corroboration with more data it remains an ad-hoc solution.

PIE (heteroclitic[5]) nom. *seh2u(e)l, obl. *suh2n-/ *suh2l- > Pgmc *saṷel, *sūn/*suṷul

Germanic cognates:

Goth       sauil                                      OE          sygil,[6] sigil

Goth       sugil                                      OE          sunne

Goth       sunno                                    OS          sunno

OIcel.     sól                                          OHG       sunna

OIcel.     sunna

The PIE word for “sun” shows the only heteroclitic paradigm alternating with l/n, the subsequent paradigm splits accounting for the difference between the roots of the daughter languages ending in /-l/ and roots ending in /–n/.  The word shows different ablaut pattern (both acrostatic and holokinetic?) in the different daughter languages, but laryngeal metathesis that placed the laryngeal after the /u/ may have happened even in the PIE stage (Sanskrit sūryas < *suh2lios)[7]. In Old English the reflexes showing umlaut may be due to a reinterpretation of the suffix as ressembling the diminutive –ila. Interesting for our story is the supposed Gothic attestation of sugil, showing wugu-alternation where we don’t expect it. However, we must take the philological context in which the attestation is found into account. In the socalled Alcuiner Handschrift (cod. Salisb. 140) , the manuscript tradition of which goes back to the sixth century, we find a few Gothic glosses and after an excerpt of Alcuin’s de ortographia an Anglosaxon and Gothic list of runes with their corresponding names. It is very plausible that somewhere in the manuscript tradition an Anglosaxon scribe couldn’t fathom that the Gothic name for the S-rune was so different from the Anglosaxon name and changed the Gothic word to a form more similar to the Anglosaxon word. A misreading is also possible. Therefore I believe that the Gothic attestation isn’t genuine.

PIE nom. *suHs, acc. *suHm[8] > Pgmc sūz, suṷum

Germanic cognates:

OIcel.     sýr                                          OS      

OE         sugu                                      OHG      

OS         suga

PIE *suHs is possibly the only non-cryptonymic word for “pig” we have attested that can be reconstructed for the ancestral language. It is possible that in the North-West IE languages the word shifted its meaning to “female pig” (maybe because the non-IE substrate language had an own word for male pig that they adapted), because in Latin, sūs also means “female pig.” The Icelandic word shows umlaut because of a phenomenon called z/R-fronting. It is clear that the variants continuing the oblique root are responsible for the subsequent wugu-rule. 



Beekes, Etymological dictionary of Greek, 2 vols (Leiden 2009).

Beekes, Comparative Indo-European Linguistics; an introduction (Leiden 1995).

Benjamin W. Fortson IV, Indo-European Language and Culture; an introduction (second edition: 2010).

H. Frisk, Griechisches Etymologisches Wörterbuch, 2 vols (Heidelberg 1960-1970).

A. Kloekhorst, Etymological dictionary of the Hittite inherited lexicon (Leiden 2008).

Friedrich Kluge, Nominale Stammbildungslehre der altgermanischen Dialecte (1886).

Friedrich Kluge, Urgermanisch; Vorgeschichte der altgermanischen Dialekte (1913).

Stefan Müller, Zum Germanischen aus larungaltheoretischer Sicht; mit einer Einführung in die Grundlagen der Laryngaltheorie (2007).

E. Prokosch, A Comparative Germanic Grammar (Philadelphia 1939).

Wilhelm Streitberg, Urgermanische Grammatik (Heidelberg 1943).

M. de Vaan, An etymological dictionary of Latin (Leiden 2008).

[1] During a course at the Leiden University Summerschool for Linguistics 2009 the subject of the wugu alternation was brought to the fore, whereafter the present students perpetuated the phrase “wugu-rule”, because it sounds awesome.

[2] The PIE ablaut allows the same uRC > uṷu

[3] I included bracketed *h1 in this reconstruction, because it seems to be the only explanation for Greek ἐννέα which must be a blend of Proto-Greek *ἐνϝα, which led to the Homeric ordinal ἔννατος, and Proto-Greek*νεϝα.

[4] Initial  *h2 is often reconstructed to connect it to *h2ei “vital lifeforece”. The socalled posessive “Hoffman suffix” should, according to Hamp be reconstructed as *-h3n-/*-h3en-.

[5] Apart from the obvious –l/-n paradigm split, also the Avestan declination gives an argument for an original heteroclitic declination; consider Old Avestan huuarɘ, gen. xyɘng-.

[6] In Anglo-Saxon dialectal spelling variation one also finds segel, sægl.

[7] A cognate to the Proto-Indo-Iranian form is also to be found in the modern Iranian dialects such as Sogdian xwyr and Ossetic хур.

[8] If Kloekhorst is right in connecting the word to PIE *suh1/3– “to fill”, you can fill in two candidates for possible laryngeals; do so at your own discretion!

Leave a Reply

Your email address will not be published. Required fields are marked *