Extending a substitution pattern

In relatively new loans into Finnish (for the last 1500 years, at least; AFAIK similarly in most other Uralic languages), *s+stop clusters are uniformly retained medially (piispa “bishop”, masto “mast”, viski “whisk(e)y”) and simplified to the stop initially (piikki “spike”, tyyli “style”, kuuro “shower”).

This has not always been the case. In sufficiently old Baltic and Germanic loans, a word-initial substitution *st → /s/ appears instead: e.g. Baltic *staibas → seiväs “pole” (interestingly, in the more southern Finnic languages the subtitution is already t-: Estonian teivas, Livonian tāibaz), pre-Germanic *starrā “rigid” → sara “sedge” (= “rigid grass”). This pattern is also found beyond Finnic: e.g. Samic *soampē “ski pole” derives from Proto-Indo-European *stombʰos “pole”. (Several IE branches could have been the actual donor, e.g. Balto-Slavic. The Samic word has also been further loaned to Finnish as sompa “stopper in a ski pole”.)

I’m not aware of particularly old estabilished cases of *sp- → *p-, *sk- → *k-. Or even many medial examples of any sort, beyond the PU-PIE comparison *mośkə- ~ *mozg- “to wash” (for which a loan etymology proposal would be complicated by the unmotivated palatalization on the U side). What if a substitution *sP → *s was originally employed even in more general? There is even a strange absense of *sp in the older stages of Uralic languages, despite ever-present Indo-European neighbors and the existence of *st and *sk in the inherited Uralic vocabulary.

At least one case well analyzable as *sk → *s can be found, I think. This is *sitta “shit”, with an immediate resemblance to its English translation. At the Proto-Germanic level, we find a long-vocalic *skīta-, with the desired *a-stem. The short vowel in the English noun is a late development (cf. German Scheiße(n-), Scandinavian skit(a-)), but for a sufficiently early loan, this is no problem: long vowels likely did not exist in Finnic at this stage, and in particular before open vowels they are considered a quite late introduction. Also the Finnic root-type *i-a is typical of loanwords anyway; no clear Uralic origin has been identified for it (though a few cognate sets do reach all the way to Ob-Ugric).

Aside from the unusual initial correspondence, the loan could be dated even as fairly late within the Proto-Finnic period. However there’s reason to consider it to be on the older side, which also grants some more space between it and the first clear cases of *sk → *k: the existence of a Permic cognate *sit (→ Komi /sit/, Udmurt /siť/). Direct Germanic loans in Permic are of course not an option, but precedents are known for Gmc → P transmission anyway, likely mediated by some early eastward offshoot of Finnic.

A slight problem may be the vowel correspondence F *i ~ P *i. Other known cases of loans from the extinct easternmost varieties Finnic into Permic have rather shown a correspondence *i-a ~ *ɨ (e.g. *liiva ~ *lɨa “sand”, originally from Baltic). Yet, the very oldest IE loans of the *i-a type do not show any such backing: consider the Iranian loans *wiša “green; poison” → Finnic *viha “hate”, *viherä “green” ~ Permic *vež (→ Komi /vež/, Udmurt /vož/) “green” , and *iša “skin” → Finnic *iho “skin” ~ Komi /ež/ “inner side of leather”. Not only do these retain the original front vowel, they appear to even be sufficiently old to have participated in the Permic lowering of *i in open syllables. (These also have reflexes beyond Permic. They’re not relevant here though.) If *sitta were to be of compareable age, *sit would indeed be the expected Permic reflex. Possibly the main problem in this scenario rather becomes if the Germanic word can be projected this far back?

I have another candidate for *sk being reflected as *s too. A root *sumə ~ *sumu “fog” is found in the westmost Uralic: Finnic *sumeda “blurry”, *sumu “fog”, Samic *somō “misty weather”, and Mordvinic *suv “fog”. Of note here is the irregular retention of *u as *u in Mordvinic, which could be explained by this being a loan acquired after the lowering of inherited *u. This same correspondence is found in some other words of limited distribution as well, particularly interestingly Erzya /ruŋgo/ ~ Mokša /roŋga/ “body”, which has been compared with Finnic *ruŋko “stem, body”; which in turn has been explained as a Germanic loan.

A reasonable loan original is here provided by Germanic *skum- ~ *skūm- “darkness” (cf. German Schummer, Scandinavian skum). A similar assumption of vowel shortening upon loaning as before may be necessary, depending on which Germanic variant is older. The final labial vowel (found in both Finnic and in Samic, and I wonder if even the irregular /v/ in Mordvinic could reflect the same) must be analyzed as a suffix added on the Uralic side.

I have not identified any cases of *sp → *s. This also seems somehow less likely: [p], as a labial consonant, is articulated more separately from [s] than [t] and [k] are. In fact I even have some proposals waiting for old IE loans with a different treatment of *sp. But once again, that shall be left to another post.

Advertisements
Tagged with: , , , , ,
Posted in Uncategorized
13 comments on “Extending a substitution pattern
  1. M says:

    “darkness” and “fog” are clearly not the same thing, so I have to (respectfully) question the grounds on which you call this Germanic-western Uralic comparison “reasonable”.

    Are there any known parallels to the development *sk- > *s- that don’t involve intermediate stage of *sk- > *š-?

    Regards,
    M.

    • Juho says:

      I’m aware this is a weaker proposal than the previous one. However, the immediate development does not have to be “darkness” → “fog”. It seems that despite being the widest-spread, “fog” is not the original sense of the Uralic etymon: nouns ending in labial vowels in Finnic/Samic are most of the time derivativs. Finnic *sumeda (with the default adjectival ending *-da) seems more likely to be the original root, and “darkness” → “blurriness” is less of a stretch. Note that this is mostly used of vision, not of blurriness in shapes on things. It might even be relevant that this word has uses in Finnish expressions such as maailma sumenee = maailma pimenee “the world is blurring/darkening out” (i.e. “I’m passing out”).

      Phonetically, I do not assume any phonetical evolution of *sk to *s here but simply adaptation of *sk as *s in the first place, entirely parallel to the case of *st as *s. I do not know parallels from other languages entirely — those would be nice, of course, but if we accept *st → *s, I don’t see *sk → *s as any more of an issue.

      • M says:

        Hello,

        I’m sorry for the delay in responding. (I was travelling for some time with only my smartphone, and it wasn’t clear that my response to your post had even gone through.)

        Phonetically, I do not assume any phonetical evolution of *sk to *s here but simply adaptation of *sk as *s in the first place, entirely parallel to the case of *st as *s.

        What is the basis for assuming this kind of development of *st to *s? It seems to me that there are other possibilities: e.g., metathesis of *st to *ts and subsequent simplification to *s, which seems to have happened (partially) in Celtic. Of course (as I realized shortly after posting my last response), the same sort of metathesis could be used to get from *sk > *ks > *s, and this metathesis also seems to have occurred partially/sporadically in IE branches such as Celtic.

        So, my main objection to your proposal is the lack of a semantic match, not phonetic criteria.

        I do not know parallels from other languages entirely — those would be nice, of course, but if we accept *st → *s, I don’t see *sk → *s as any more of an issue.

        Without any intermediary stages, changes such is *st > *s, and *sk > *s seem to require the instantaneous deletion of an intervening sound (rather than the deletion of a peripheral sound or sounds, as in the case of the C1C2- > C2- onset simplification). While it’s not impossible to imagine such deletion occurring in the process of loaning, I wonder whether it’s probable enough to be treated as the “null” hypothesis in this case?

        • Juho says:

          Hello again!

          At least your suggested metathesis followed by loss of the initial consonant is clearly not possible, since Proto-Finnic (ditto Proto-Samic, Proto-Uralic etc.) did not allow initial consonant clusters. If there are other options yet, I’m not sure what these would be.

          There did occur an affricate *c [t͡s] in both Proto-Finnic and Proto-Samic though, but this would not end up as /s/ uniformly in the modern languages (indeed, in Samic this should be uniformly retained).

          The general picture is that when a language that disallows consonant clusters loans some words that have clusters in the source language, there are broadly speaking two options: dropping parts of the cluster (e.g. Finnish koulu “school”) or adding prothetic vowels (Hungarian iskola, Indonesian sekolah). The latter does not work as an option: PF, PS, PU etc. also featured prominent initial stress, and without exception have retained all word-initial syllables. Hence we must pick direct substitution. There is also no way seiväs, sitta could derive from *teiväs, **kitta (etc.) So we end up with the admittedly typologically rare substitution type *C₁C₂ → *C₁ not as a null hypothesis, but thru the elimination of the other options.

  2. David Marjanović says:

    Possibly the main problem in this scenario rather becomes if the Germanic word can be projected this far back?

    I just noticed something. OK, it’s past 2 am… but…

    First of all, additional evidence for the */iː/ in Proto-Germanic: Irish-and-other English shite, Low German /ʃiːt/.

    But, in southern German, in addition to die Scheiße, there’s der Scheiß. So far, so good, apocope turning into gender reassignment surgery is normal. However, in the actual Bavarian-Austrian dialects, that latter form shows up as /ʃas/ (no phonemic vowel length, stressed vowels are long like in Russian). This /a/ is the reflex of Middle/Old High German ei and Proto-Germanic */ai/, not */iː/.

    So… do we perhaps need to project some kind of ablaut variants all the way to Proto-Germanic or at least Proto-West-Germanic?

    Or should I just go to bed? :-)

  3. M says:

    I’m responding here to Juho’s last comment (from 2013-08-03); WordPress doesn’t show a “Reply” option directly below the comment, so this comment may be disconnected from the chain.

    At least your suggested metathesis followed by loss of the initial consonant is clearly not possible, since Proto-Finnic (ditto Proto-Samic, Proto-Uralic etc.) did not allow initial consonant clusters.

    I don’t see a reason for projecting this as an absolute rule throughout the entire history of Uralic languages. The fact that there are no ancestral Uralic words with initial consonant clusters doesn’t imply that at no moment in the (pre-modern) history of any Uralic language could any speaker have pronounced a word with an initial consonant cluster, particularly if said word came from a foreign language that *did* allow initial consonant clusters.

    It could be that the metatheses *st > *ts and *sk > *ks (which, again, are attested in other languages) are statistically less probable here than instantaneous deletion, but I’m not sure about this.

    There did occur an affricate *c [t͡s] in both Proto-Finnic and Proto-Samic though, but this would not end up as /s/ uniformly in the modern languages (indeed, in Samic this should be uniformly retained).

    *ts- from metathesized IE *st- need not have been pronounced in an identical way to the PF and PS affricate *c.

    The general picture is that when a language that disallows consonant clusters loans some words that have clusters in the source language, there are broadly speaking two options: dropping parts of the cluster (e.g. Finnish koulu “school”) or adding prothetic vowels (Hungarian iskola, Indonesian sekolah).

    Only if this disallowance is a mathematically rigid rule: if not, options such as metathesis and subsequent simplification of the cluster are also possible.

    Incidentally, didn’t some (or all) Sami languages start to retain initial clusters such as *sC- in Germanic words at a fairly early date, rather than deleting the initial consonant?

    • Juho says:

      The fact that there are no ancestral Uralic words with initial consonant clusters doesn’t imply that at no moment in the (pre-modern) history of any Uralic language could any speaker have pronounced a word with an initial consonant cluster, particularly if said word came from a foreign language that *did* allow initial consonant clusters.

      No, probably not. It’s likely there have always been some speakers who also knew an IE language and could also manage its clusters. This analysis just fine-tunes any sound substitutions in loanwords as having occurred when the monoglot and less skilled polyglot Uralians also adopted the word, though. U/IE bilinguals were not a consistent separate speech community in which words could be first loaned, then subjected to sound changes, then loaned further onto the main speakerbase.

      I.e. while you could imagine an individual or small group loaning, let’s say Germanic *staθa- “town, place” (whence Finnish satama “harbor”, etc.) first as *stata, idiosyncratically metathesizing this to *tsata, and this then being simplified to *sata by other speakers, this kind of an explanation cannot account for the systematic correspondence between IE *st- : (West) Uralic *s- that appears in the oldest layer of loanwords.

      Attempting to account for *st- → *s- by means of a sound change *st → *ts also runs into some chronological problems, since these words were adopted over a whole stretch of time. The example of seiväs ~ teivas shows that at a point, pre-Finnish still had the old substitution pattern, while pre-Estonian had switched to the newer one. So it seems under your explanation we would have to posit a new, separate, North Finnic-only metathesis to explain this case, while for cases like satama ~ sadam, we’d posit a common Finnic instance of the change? And then a third metathesis at some point in pre-Samic to explain words like North Sami soabbi, since the Finnic loans in consideration are younger than the separation from Samic? Sound laws start getting multiplicated at an alarming rate here.

      Unlike sound changes (of this kind, at least) however, phonotactic limitations by default tend to persist over a language’s history.

      It could be that the metatheses *st > *ts and *sk > *ks (which, again, are attested in other languages) are statistically less probable here than instantaneous deletion, but I’m not sure about this.

      What precedents exactly do you have in mind for these changes, anyway? Are these metatheses in the initial position, or perhaps rather medially or finally? I can easily think of examples of the latter in both directions (English vacillations like ask ~ aks; in Permic *ks → *sk regularly; in Ob-Ugric, perhaps even all of East Uralic, *sk → *ɣs regularly) — but not of metathesis in word-initial clusters. In particular PIE *st- seems to have an excellent track record in being retained in its descendants. Perhaps better than plain *s-, and maybe even plain *t-?

      Incidentally, didn’t some (or all) Sami languages start to retain initial clusters such as *sC- in Germanic words at a fairly early date, rather than deleting the initial consonant?

      So they did. Also *Cr-, *Cl-. A few such words are even presented as a part of the common Proto-Samic stock in J. Lehtiranta’s Yhteissaamelainen sanasto (1989) (all reconstructed with a single consonant though), but since differing substitution strategies appear between the languages (e.g. South Sami kraesie vs. North Sami rássi “grass, flower), they’ve probably rather been adopted in parallel.

      Retention of such initial clusters seems to be the most frequent in South and Ume Sami, slightly less so in Pite/Lule/North Sami, rarer yet in Inari and Skolt Sami, and pretty much unheard of in Kola Samic. Incidentally, the same applies to the adoption of initial /h/ in loanwords. This pattern is of course not much of a surprize, when Germanic influence was coming in mainly from SW anyway.

  4. M says:

    No, probably not. It’s likely there have always been some speakers who also knew an IE language and could also manage its clusters.

    I don’t think that Finnic speakers need have been fluent in an IE language in order to perceive the initial consonant clusters in IE words, and to attempt to pronounce them as such (i.e., as clusters). Metathesis of *st-/*sk- is almost an example of failing to manage these clusters, as one might expect non-fluent speakers to do.

    I.e. while you could imagine an individual or small group loaning, let’s say Germanic *staθa- “town, place” (whence Finnish satama “harbor”, etc.) first as *stata, idiosyncratically metathesizing this to *tsata, and this then being simplified to *sata by other speakers, this kind of an explanation cannot account for the systematic correspondence between IE *st- : (West) Uralic *s- that appears in the oldest layer of loanwords.

    I’m not familiar with all the evidence for this correspondence, but I have to admit that I’m (as yet) unconvinced that it’s a perfectly systematic one. Quite a few of the proposals involving IE *st- > Finnic *s- seem to involve ad hoc semantic or morphological arguments:

    satama: Finnic meaning ”harbor” vs. Germanic meaning “place”; some of the Germanic words have developed a meaning “shore”, and some (in German, specifically) have gone further to the meaning “harbor”, but I don’t think it can be assumed without further evidence that this development dates back to Proto-Germanic. Also, how is the final -ma of the Finnic words explained?

    sija: Finnic meaning ”place” vs. Germanic meanings “pig sty”, “barn”, “pen”, etc.

    sietää: Finnish meaning “tolerate” vs. Germanic meaning “stand (in general)”; also, is it clear that the form *ste:- (rather than just *stand-) goes back to Proto-Germanic?

    salpa means ”bolt (on a door)”, whereas the proposed Germanic cognates (ON stolpi , Middle English stulpe, etc.) seem to mean “post”, “stake”, “beam”, etc. Although the meaning “beam” could be seen as very close to “door-bolt”, I’ve only found this meaning listed for one Germanic language (Middle Low German), and I don’t know the further semantic details of the MLG word.

    Sami soabbi means “staff”, as far as I know; the putative IE source word is translated “column, pillar”

    The most convincing examples of *st- > *s- I know of are sara “sedge”, seiväs as you mention below, and possibly sammas. These are only 2-3 data points, and can be explained through the sporadic metathesis of *st- to *ts- in loanwords.

    Attempting to account for *st- → *s- by means of a sound change *st → *ts also runs into some chronological problems, since these words were adopted over a whole stretch of time. The example of seiväs ~ teivas shows that at a point, pre-Finnish still had the old substitution pattern, while pre-Estonian had switched to the newer one. So it seems under your explanation we would have to posit a new, separate, North Finnic-only metathesis to explain this case, while for cases like satama ~ sadam, we’d posit a common Finnic instance of the change? And then a third metathesis at some point in pre-Samic to explain words like North Sami soabbi, since the Finnic loans in consideration are younger than the separation from Samic? Sound laws start getting multiplicated at an alarming rate here.

    Again, I’m not sure how many common Finnic or Fenno-Samic words really did undergo this change – I find many of the often-cited examples to be doubtful.

    Unlike sound changes (of this kind, at least) however, phonotactic limitations by default tend to persist over a language’s history.

    It doesn’t follow, though, that these phonotactic limitations will be compensated for in exactly the same way throughout time, much less in the same typologically unlikely way. (Again, I don’t know how typologically rare changes such as *st-/*sk- > *s- are, but at this point I only know of a few sporadic parallels, such as Greek histe:mi < *sti-sta:mi.)

    What precedents exactly do you have in mind for these changes, anyway? Are these metatheses in the initial position, or perhaps rather medially or finally? I can easily think of examples of the latter in both directions (English vacillations like ask ~ aks; in Permic *ks → *sk regularly; in Ob-Ugric, perhaps even all of East Uralic, *sk → *ɣs regularly) — but not of metathesis in word-initial clusters. In particular PIE *st- seems to have an excellent track record in being retained in its descendants. Perhaps better than plain *s-, and maybe even plain *t-?

    As I said, the Celtic languages show a partial metathesis of initial *st- > *ts- > s: Middle Irish ser, Welsh seren “star” ; Irish serc, W. serch, “love” (cf. Greek stérgein “love”) ; W. sefyll, Cornish sevel “stand” (cf. Engl. stem, Germ. Stamm etc.). This metathesis doesn’t seem to have been fully completed in the British Celtic languages, since one language will often have st- where another has s-: W. seren vs. Cornish and Breton steren “star”, W. safn “mouth, jaws” vs. Breton staoñ “palate”. Sometimes, there is variation between s- and st- within the same language: W. sarn “causeway, paving”, ystarn “saddle” (cf. Lat. stratus “spread out”), Breton dialectal stan/san (= staoñ).

    In Gaulish and early Irish inscriptions, there is some possibility that *ts from initial *st was still an affricate, since we see special characters (“Д in Gaulish, “Z” in Irish) used in positions where *st- would have appeared in IE.

    Word-initial *sk > *ks has also occurred partially in British Celtic, with the further development *ks > *x > xw: Welsh chwedl “tale, account” corresponds to Irish scél (id.) and Welsh chwalu “scatter, spread” corresponds to Cornish scullye and Bret. skula (id.) and possibly to Middle Irish sceillec ”rock”. No metathesis has occurred in words such as W. ysgwyd “shield”, OIr. scíath “shield”.

    I’ve heard that initial *sk shows a similarly inconsistent development in Slavic, but I don’t know as much about the details. For example, Slavic skala “rock” seems to correspond to words such as MidIr. sceillec; in the other direction, some have proposed a connection between words such as Polish chłop “peasant” and Gothic skalks “servant”, with possible *sk > *ks > Slavic x-.

    • Anders Jørgensen says:

      “As I said, the Celtic languages show a partial metathesis of initial *st- > *ts- > s: Middle Irish ser, Welsh seren “star” ; Irish serc, W. serch, “love” (cf. Greek stérgein “love”) ; W. sefyll, Cornish sevel “stand” (cf. Engl. stem, Germ. Stamm etc.). This metathesis doesn’t seem to have been fully completed in the British Celtic languages, since one language will often have st- where another has s-: W. seren vs. Cornish and Breton steren “star”, W. safn “mouth, jaws” vs. Breton staoñ “palate”. Sometimes, there is variation between s- and st- within the same language: W. sarn “causeway, paving”, ystarn “saddle” (cf. Lat. stratus “spread out”), Breton dialectal stan/san (= staoñ).”

      It’s a bit far from Uralic, but here goes:
      It has been suggested (e.g. Schrijver SBCHP) that the metathesis (if that really is what it is) was conditioned, with *st being preserved as such after a consonant and a further development to *ss (via *ts?) after a vowel. That would explain the vacillation in British Celtic as due to external sandhi (something we would anyway expect in Celtic).

      In Goidelic, even preserved *st gave *s(s) eventually, hence no trace of old *st.

      “Word-initial *sk > *ks has also occurred partially in British Celtic, with the further development *ks > *x > xw: Welsh chwedl “tale, account” corresponds to Irish scél (id.) and Welsh chwalu “scatter, spread” corresponds to Cornish scullye and Bret. skula (id.) and possibly to Middle Irish sceillec ”rock”. No metathesis has occurred in words such as W. ysgwyd “shield”, OIr. scíath “shield”.”

      Welsh chwalu cannot be related to Cornish scullye and Breton skuilhañ ‘to pour’. The latter two words clearly correspond to Ir. scaílid, scoílid ‘to disperse’ *u: > Bret./Corn. /y/). Welsh chwalu must derive from something like *swal- or the like and cannot be related by established sound changes.

      The development of *sk to *xw is probably to be explained differently from that of *st > *ss. The reliable examples of initial *sk- > *xw- all occur before a following front vowel
      *skitV- > Brit. *xwïd- ‘to vomit’, Ir. sceith
      *skib- > Brit. *xwïB- ‘to move’, Ir. scib-
      *sketlo- > Brit. *xwedl ‘tale’, Ir. scél
      *skendo- > Brit. *xwïnn- ‘to rise, set out’, W cychwyn.

      Whereas when we have a following non-front vowel, *sk is preserved. On the basis of this observation, I have argued that *sk- > *xw- is due to a form of palatalization, a rough parallel being provided by Swedish sk- (and stj-, sj-) becoming /xW/.

      Note also that while *st > *ss is completely carried through in Goidelic, there is no trace of a special development of *sk in this branch. It is always preserved as sc.

      • M says:

        Hi Anders,

        Welsh chwalu cannot be related to Cornish scullye and Breton skuilhañ ‘to pour’. The latter two words clearly correspond to Ir. scaílid, scoílid ‘to disperse’ *u: > Bret./Corn. /y/). Welsh chwalu must derive from something like *swal- or the like and cannot be related by established sound changes.

        What about W. chwalu and Ir. scaílid? Other than the initial consonant clusters, are there any other phonetic difficulties in matching the two?

        If you accept that intervocalic *-sk- underwent metathesis to *-ks- > *-x- at some stage of British (for example, the verbal suffix -ychu in Welsh clafychu, gwledychu etc. recalls IE *-ske/o), then it doesn’t seem too far-fetched to argue that initial *sk- was sporadically metathesized as well, due to the same sandhi phenomena that (possibly) influenced the development of initial *st-.

        It may be that most examples of initial *sk > British *xw- involve a front vowel, but if we only have four such examples (and at least one more possible example, chwalu, involving a back vowel), then I don’t think we can rule out other possibilities about the development of this cluster.

        By the way, is Welsh ysgen(n) “dandruff” (cf. German Schinnen, a possible cognate with the same meaning) thought to reflect *sk followed by a front vowel, or is it explained in another way?

      • Anders Jørgensen says:

        “What about W. chwalu and Ir. scaílid? Other than the initial consonant clusters, are there any other phonetic difficulties in matching the two?”

        Irish scaílid has a diphthong -aí/oí- in the first syllable, cf. later spellings with -aoi-. Even if -aí- were the older form (and not -oí-), it would correspond to W -oe- (PCelt. *ai > Brit. *E: > MW oe, MBret. oa). But there’s no way known to me to connect W -a- (as in chwalu) and Irish -aí-.

        Anyway, the word does not appear to be attested in Old Irish texts proper, which means we can’t be sure whether it had /oi/ or /ai/ originally. In light of PCelt. *oi giving SWBrit. /y/, the connection with Corn. scullye, Bret. skuilhañ is impeccable.

        “If you accept that intervocalic *-sk- underwent metathesis to *-ks- > *-x- at some stage of British (for example, the verbal suffix -ychu in Welsh clafychu, gwledychu etc. recalls IE *-ske/o), then it doesn’t seem too far-fetched to argue that initial *sk- was sporadically metathesized as well, due to the same sandhi phenomena that (possibly) influenced the development of initial *st-.”

        I don’t accept a metathesis for medial *sk either. Medial *sk is almost always preserved. The solid evidence for a development to *-x- seems to be:

        *baskjo- > MW beich, MBret. bech ‘burden’
        *taskjo- ‘badger’, attested in the name Tasc(i)ovan(t)- ‘Badger-slayer’ > *teixwan(t) > W name Teuhuant.

        And then there’s the W denominative verbal suffix -ych-, supposedly from *-iske/o-. This is truly odd, since the suffix is only known from W, with no trace in Breton and Cornish, nor in Goidelic. I’ve speculated that W -ych- actually continues compounds with the verbal root *isk- ‘to search, to look for’, which is preserved in Irish escaid ‘to search for lice’.

        In light of *baskjo- and *taskjo-, I have suggested that the conditioning of non-initial *sk > *x was also palatalization, with *sk > *x / __j $. Given the productivity of the verbal suffix *-j-, it is quite possible that we may reconstruct *-isk-j- in the case of the verbal suffix -ych (whatever its ultimate origin).

        “By the way, is Welsh ysgen(n) “dandruff” (cf. German Schinnen, a possible cognate with the same meaning) thought to reflect *sk followed by a front vowel, or is it explained in another way?”

        I suppose ysgenn is a compound, consisting of *exs- ‘away from’ and *kenn ‘skin’.

  5. M says:

    PS — I forgot that you’d mentioned Finnish sara “sedge” in your original post. I think it’s worth mentioning that the North Germanic word group (Icelandic stör, Swedish starr etc.) has the meaning “sedge” just like the Finnish word, which is why I find it a convincing example of *st > *s.

    If we were to assume that sara was loaned into Finnish with the meaning “stiff, rigid”, and then developed to the meaning “sedge” (independently from North Germanic), a Germanic etymology wouldn’t seem so convincing.

  6. David Marjanović says:

    Unlike sound changes (of this kind, at least) however, phonotactic limitations by default tend to persist over a language’s history.

    Except exceptions. French is developing impressive consonant clusters right now (t’veux conduire, ou t’veux qu’j’conduise ?), Malagasy used to be a (C)V language and now has consonant clusters as well, Pre-Proto-Slavic completely got rid of closed syllables at some late point and then the yers fell… oh, and look at this; it blew my mind.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Enter your email address to follow this blog and receive notifications of new posts by email.

%d bloggers like this: