User:Burgundaz/Germanic sound change theories

The voiced bilabial nasal in Germanic seems to have developed an allophonic voiced bilabial fricative or nasalized voiced bilabial fricative. Compare the vascillation in the suffix(es) -umni-/-ubni-, also attested in Latin and Greek renderings of the Germanic ethnonym Dulgubnii and Doulgoumnioi, respectively, probably representing Germanic *Dulgumnijōz, /ˈdul.ɣuβ̃.ni.jɔːz/, < **dulgumniją/ō, "owing, debt".

Examples:

• *eb̨naz, /ˈeβ̃.nɑz/ < pre-Gmc *h₁émnos
• *fastub̨niją/jō, /ˈɸɑs.tuβ̃.ni.j-/
• *lauhub̨nī, but non-nominative *lauhmunjō-
• *steb̨nō, /ˈsteβ̃.nɔː/ < pre-Gmc *stémneh₂

Positive examples of -hw-, -kw-, -gw- yielding -f-, -p-, -b-:

• *gʷʰedʰye- > *bʰedʰye- > *bidjaną
• *gʷʰern- > *bʰern- > oblique stem *bern- of *berô (the expected strong stem *werô was analogically leveled)
• *gʷʰonyeh₂ > *bʰonyeh₂ > *banjō
• *kʷéltus > *péltus > *felþuz (possibly; this allows us to ignore the extremely problematic *pelth₂-, which is the only instance of Schwebeablaut of the root *pleth₂-, ie. *flataz)
• *kʷetwṓr > *petwṓr > *fedwōr
• *likʷréh₂ > * lipréh2 > *librō
• *skʷetlóm > spetlóm > *spedlą > *spellą (compare Proto-Celtic *skʷetlom)

• Notes:
  • *kʷéltus - The original meaning of "cultivated/agricultural field", therefore "turned field", would make it an attractive cognate to Latin cultus, "cultivation, tilling". A derivation from *pleth₂- is impossible. Compare also *hwelą, "turning (thing)" for the vocalism.

Presumably this allows us to posit a conditional change where labiovelar plosives become bilabial plosives when the syllable closes in a dental or alveolar (coronal) obstruent/sonorant, and the following syllable begins with a sonorant. *banô, "murderer", probably represents a later Germanic derivation from an unattested strong verb *banjaną, or maybe *binjaną in view of Greek θείνω.

Negative examples of -hw-, -kw-, -gw- yielding -f-, -p-, -b-:

*wulbī seems to offer itself as the probative example, where pre-Germanic:

... yielded:

... and further Germanic:

Where the root was not delabialized the labiovelar was allowed to develop into a bilabial plosive, as also *wulfaz < *wĺ̥kʷos, thus Old High German wulpa.

Further examples

*hnīwaną ("to bow down"):

• • *hnaigijaną ("to make bow (down)"): like *saigijaną from *sīhwaną, the Sievers variant -ij- must also have delabialized, being perceived as one phoneme -j-. Otherwise, these two causatives are hard to explain. Gothic hnaiwjan takes its consonantism from the strong verb, as usual: Gothic almost always replaces a Verner's voiced consonant with a regular Grimm's form where a strong verb or other perceived "base" form is present from which to regularize.

• *sīhwaną ("to sink, drain, trickle")
  • *saigijaną ("to cause to sink, drain"): the consonantism is secondary, as we would expect **saiwijaną from the suffix *-éyeti, perhaps this provides proof that not just -j- but -ij- itself functioned as a delabializer, as a similar situation is found in *hnīwaną, causative *hnaigijaną
  • *saigiz ("able/fit to drain, trickle, fall") < (nom.) *soykʷ-ís: gave **saiwiz) but non-nom. *soykʷ-yó- gave *saigij- - ON seigr ("stubborn, viscid;"); OE on-sǣge, OHG ana-seigi ("falling upon")
  • *saiwiz ("lake, sea") < *soykʷ-ís: nominal i-stems decline differently to adjectival i-stems, there weren't as many forms with a palatal glide to delabialize
  • *sīganą ("to sink, drain, trickle"): secondary verb formed from the causative *saigijaną, which was perceptually no longer identifiable as belonging to *sīhwaną

Potential counter-examples

• *awjō: often derived from the same root as *ahwō < *h₂ékʷ-eh₂, as if from *h₂ekʷ-yéh₂, however in terms of colexification it is more common to derive "island" from "land, earth" than "water". Therefore, since the semantic connection is weak, there is no reason to assume it is a counter-example: we would expect *h₂ekʷ-yéh₂ to regularly give **agjō (which might or might not be found in the name for the Eider river, ON Egða, LA Agidora, but this would require further research). Instead, *awjō can be connected with *auraz ("(wet) earth, mud", indicating a shared root *He/ow(H)- and a pre-form *He/ow(H)-yeh₂.

Examples include:

• *newaklahaz < *newaklaz < *newaknaz < *newogn(H)ós, "new born"; cf. Gothic niuklahs, further Ancient Greek νεογνός
• Possibly *fimflaz < *fimfnaz < *pémpnos, "monster" (< reduplicated *pen-pn-?); cf. OE fīfel, ON fimbul < *fimml(l) < *fimfl-
• Gothic dauþubleis "sentenced to death" < adj. *dauþublijaz < *dauþub̨nijaz (/ˈdɑu̯.θuβ̃.ni.jɑz/, see sound change above), derived from nominal *dauþub̨niją/ō, "putting to death" < *dauþijaną, "to put to death; to kill"; cf. ON dauðyfli, "lifeless thing; carcass", which was probably derived from the adjective, as the dissimilation was triggered by the weak n-stems of the adjective, ie. *dauþub̨nijaniz

Observed by Dimitri Pisarev in "A new sound law for Proto-Germanic?" (2017), Germanic seems to have experienced a sound change where single short -a- was assimilated to -au- pretonically before a labialized element, -p-, -b-, -w-, -kw-, or -gw-.

Pisarev's examples:

• *augan-, "eye" < *akʷán- < *h₃ekʷón-
• *hafudą, obl. *haubiþa-, "head" < *kapít- against nom.acc. *káput-
• *hauhaz, "high", *haugaz, "mound", from earlier *kákʷos and *kakʷós
  • The daughter languages mostly leveled the vocalism of *hahwaz to *hauhaz, except for Old Norse hár, which cannot continue *hauhaz, although *hauhaz is continued by expected hór
    • Also, compare ON hæð < *hahwiþō against hœð < *hauhiþō, both "height"
  • Cognate with Latin cacūmen, "peak, top", which together with Germanic points to earlier *kakʷos or *kH̥kʷos; this foregoes the awkward connection with a root *kewk- meaning "bent, twisted"
• *laubaz > Dutch loof, "tired, weak", from earlier *labá- against normal *lafaz, "loose, weak", Dutch laf; Kroonen adduces MHG labschīze, "loose shit, diarrhea", as evidence for the voicing, however this sound change is dependent on immediate pretonic stress, whereas compounds shift stress as part of composition, it would be expected that
• *laumaz > Dutch loom, "sluggish", from earlier *lamá- against normal Germanic *lamaz, "lame"
• *laubaz, "leaf" < *laupás < *lapás, cognate with Lithuanian lapas, "leaf, sheet, metal plate", literally "broad material"; also cognate to OE læppa, "cloth, fabric, robe", German Lappen, "cloth", from earlier *lap-nós after Kluge's law.
• *raubaz, "robbery, seizing", OE rēaf, OHG roub; a tomós-type noun, derived from earlier *rapós, pre-Germanic *h₁rh̥₁p-ó-; further cognate with marginal Germanic verb *rafjaną, "to seize take", and Latin rapere, "to seize, take away". The derivation of *rauba- and *raubōną from *reufaną, "to break (off), burst" has always seemed awkward to etymologists, it can now be shown they are not related, and the Germanic forms make much more sense now.

Some examples, like the voicing of MHG labschīze, "loose-shit, diarrhea", point to exemption from this sound change; however, the sound change is dependent on immediate pretonic stress, which can disappear in compounds as the stress shifts.

Further examples:

• *awwidja- > OE ēow(e)de, OHG ouwiti (but also mixed OHG ewiti)
  • *awwidją against Gothic aweþi < *awiþją, rather points to a secondary collectivization suffix *-ją to what must have already been a collective *awudą, obl. *awwida-, "(sheep) herd", as can still be seen in OHG ewit, OE ēowd and in view of Gothic -þ- which must have been original to the paradigm, because Gothic levels out non-Verner's voicing only when it's available in a related word.
• *gōmô, obl. *gaum-, "gum, palate", from earlier nom/acc *gōmur, oblique *gamin-, reflecting *ǵʰéh₂-mr̥, *ǵʰh̥₂-mén-
  • Compare Lithuanian gomurỹs, "palate", and Latvian gãmurs, "larynx, trachea", as evidence for an original heterocltic r/n-stem. This has the benefit of not requiring a separate u-stem for the root, as in Kroonen (2013) *ǵʰéh₂u-mn- against Baltic *ǵʰ(é)h₂-mr-. Germanic leveled out the strong stem with -r- and remade it a complete n-stem. Unfortunately this means *gōmô/*gaumin- can no longer be used as evidence for shortening of *-ōu- before closed syllables, even though the change is still probably correct for other words.

The PIE cluster *-th₂- has been discussed quite frequently, due to the fact that it shows some outcomes of -th- in Sanskrit, and -tʰ- in Ancient Greek, both of which are irregular to those languages, where we would expect *-t(V)- in Sanskrit, and -t(a)- in Ancient Greek. Instead it seems that the cluster resolved itself to an aspirate *-tʰ- within late-PIE already. The material in Germanic seems to have been missed, but certainly it adds further evidence of the change, at least in some formations. In most cases these developments can be explained in other ways, but the material as it is suggests the matter requires some analysis anyway.

The working theory here is that in pre-Germanic, PIE *-th₂- resulted in *-t- in some cases, but not others. Whether this entered the early Germanic phonological system as *-d- (thence > *-t-) or *-tʰ- (thence > *-t-) is undiscernable at this point, but with regards to the Sanskrit and Greek material, it is safer to assume the latter. The aspiration would then have prevented shift to *-þ- or *-d- as is regular in Germanic.

• *flataz, from earlier *ploth₂os
  • *flaþaz or *fladaz is expected. Could be explained as analogical from a related form with Kluge's law, *flatt-, as per Kroonen (2013), but no form of this shape is yet attested, to the author's understanding.
• Second person singular ending *-t in preterite-present verbs, i.e. *kannt, reflecting earlier *-th₂e
  • Can very easily be explained as analogical from roots ending in velars and labials where *-Ht- and *-Pt- are expected. However there are other instances in Germanic where these allophonic clusters occur but do not become the regular outcome with regards to the morphemes involved. It quite likely provides further evidence to the matter at hand.
• Gothic second person dual ending *-ts (< *-ataz or *-atiz), reflecting earlier *-eth₂es
  • Assumed to be the regular reflex of *-þ(V)s with vocalic apocope, according to Ringe (2006). But there are multiple instances of the reflex of such a cluster (in other environments) simply being *-þ(s) in all Germanic languages, not **-t(s).

• *felþą, usually etymologized as *pelth₂-om, with Schwebeablaut of the root.
  • The Schwebeablaut of the root *pelth₂- would be the only instance for this particular PIE root otherwise everywhere attested as *pleth₂-.
  • An alternative etymology would be de-velarization of earlier *kʷelto- (see this section on the development of labiovelars to labials), which would then make it cognate to Latin cultus(< *kʷoltos < *kʷeltos), "tilling, cultivation" or the participle, "tilled, cultivated"; the semantics being "tilled" or "tilling", literally "turning", to "that which is turned" > "(cultivated) field" as opposed to *haiþī, "(uncultivated) land/field". This etymology is more likely and requires less reliance on the unpredictable nature Schwebeablaut.
• *fuldō/ǭ, "earth, land", reflects earlier *pl̥th₂-eh₂, a similar formation is found in Ancient Greek Plátaia(i), Celtic *flitawī/*flitawyā- and Sanskrit pr̥thivī́
  • Immediately the four can't be exact cognates, according to accepted sound shifts; while Celtic, Greek and Sanskrit certainly are formally perfect (*pl̥th̥₂wih₂'), the resulting form of the same in Germanic is expected to have normally given *fulþwī or *fuldwī.
  • Instead, Germanic mechanically reflects *pl̥t(h₂)-éh₂-. In order to potentially explain the lack of a form **fultō: we could assume the requirement of the cluster to be intervocalic (*-Vth₂V-) to operate, but this would render the stative/preterite ending -t(a) unexpected, as it regularly follows a root-final consonant. Alternatively, we could assume dissimilation of -th₂éh₂ to *-téh₂. Lastly, and less appealing, is to assume its not related to the above extra-Germanic comparisons, but rather related to the alternative etymology of *felþą, where it would represent *kʷl̥-téh₂, but this would require the change of labiovelar kʷ- to p- to precede the resolution of the syllabic resonant to -ul-, otherwise -kʷ- would be dalabialized to -k-, and therefore unable to undergo such a shift.

Some words in Proto-Germanic show a short vowel where we would expect a long vowel from extra-Germanic cognates. This is best explained by assuming pretonic shortening, i.e. shortening of a long vowel before a stressed vowel in the next syllable. This is termed "Dybo's law" in Kroonen 2013. The few certain examples that can be found all occur only before a sonorant, so we will assume that this is the required environment, -V̄R(C)V́-.

• *frawaz, "first, foremost" < *proHwós
  • The derived substantive *frawjaz, "lord", must be secondarily derived from the adjective *frawaz, "first, chief", which was no longer an independent morpheme, being found only in compounds as frēa- in OE and frá- in ON. That the noun is secondary is further suggested by the fact that a form going back to *frawô is the only reflex we find in OE, and an alternative form in OS and OHG.
• *wiraz(*weraz with a-mutation) < *wiHrós

In West Germanic, intervocalic labiovelar -w- is shifted to intervocalic velar -g- in the presence of at least one high vowel.

• *awiþahsijǭ, "lizard" > *agiþahsijǭ
• *juwunþī, "youth" > *jugunþi
  • The -w- was preserved in Gothic junds, where -uwu- presumably collapsed to -ū- or -u-. The difference in voicing is due to paradigmatic ablauting nom./acc. *juwunþī, oblique *juwundijō-, earlier *yuHń̥tih₂ and *yuHn̥tyéh₂-.
• *suwilaz > *sugilaz
  • OE siǵel, seǵ(e)l, "sun". A mix of the weak root *su- and stem *-(w)ila-.
• *suwu (accusative of *sūz) > *sugu
  • Remained complementary to West Germanic *sū, "sow"; source of OE suǵu, OF sūg(e), ModD zeug.
  • Less definitive and obvious due to an innovative Old Norse stem *suggva probably from a mixed stem *sūw- > *suww-, seen in Faroese súgv.
• *þrawilaz, "servant, thrall" > *þragilaz
  • Derived from *þrawjaną, "to threaten, punish". Old Norse þræll confirms the form without -g- (otherwise ON **þregill). Usually derived from *þragjaną, "to run", but that root even outside of Germanic shows -gʰ-, which would then make it impossible for Old Norse to go back to supposed **þrahilaz. The ON form could also formally go back to **þranhilaz < *þranhaz, "dirty" with a meaning of "wretch(ed person)", but derivations in *-ilaz are usually deverbal, and it would be unreasonable to separate the North and West Germanic forms. For the semantics, Lithuanian vergas, "slave", derived from PIE *werg-, "to suffer", cf. also Lithuanian vargùs, "difficult, tiresome".