Sporadic Sound Change

English - United States
Most sound changes as a language evolves are regular - they affect all instances of a sound in a given context.

"Sporadic" or "irregular" sound changes are far less common, but can still occur. For example, the /k/ in Middle English "Maked" was elided, hence modern "made" - this did not affect other intances of intervocalic /k/, even when both vowels were the same. (i.e. "baked" has not become "bade")

On the other hand, some words will refuse to undergo certain sound changes - for example, English "grass" (Old English "græs") should have metathesized to something like **gars /gɑːz/. (Old English /r/ metathesized with adjacent vowels, such as Old "þridda" > Modern "third")

I'm interested to see what (if any) examples exist in other languages.
 
  • I'm interested to see what (if any) examples exist in other languages.
    One typologically common (and probably the only truly irregular) example is various simplifications in frequent words, and Russian has these in abundance, especially (but certainly not exclusively) in unstressed positions. For instance, unstressed человек ([ʨɪɫɐ'vʲek] "human, person") will be pronounced roughly as [ʨɪk] in fast casual Moscow speech, with all the consonants in between being irregularly elided; когда ([kɐ'gda] "when") becomes [kɐ'da] in casual speech (but cf. всегда [fsʲɪ'gda] "always" which never undergoes the same change even despite sharing the same etymological root), etc.

    Morphological levelling is another widespread source of irregularities. The most iconic example in Russian is probably конечно, the pronunciation of which actually depends on the meaning: in the meaning "of course, surely" it demonstrates the supposedly regular shift /*-ʧn-/ > /*-ʃn-/ and is normally pronounced [kɐ'nʲɛʂnə], but as the singular neuter short adjective with the meaning "finite" it has been reconstructed back to [kɐ'nʲɛʨnə] (as the connection to конец [kɐ'nʲɛʦ] "end" becomes totally transparent here); in fact, morphological levelling has destroyed most of the results of this shift in Russian.

    And then, of course, there are results of old dialectal interactions, often obscured. For example, early Proto-Slavic /*vĭrvĭka/ "(little) rope" (/-ĭka/ subsequently shifted to /-ĭʦa/ as the result of the third palatalization but then was restored back to /-ĭka/ by morphological analogy in Old East Slavic) should have produced *вервка ['vʲɛrfkə] in Russian, but here Old Novgorodian with its /CĭrC/ > /CĭrĭC/ shift intervened, so instead we (unexpectedly) have верёвка [vʲɪ'ɾʲɔfkə] "rope".
     
    Last edited:
    Aquam > OF /'ɛwə/ instead of the expected /'ɛvə/ is an obvious one in French. /'ɛwə/ yielded modern "eau" instead of the expected /ɛ:v/, which can be found in place and people name such as Entrèves, Les-Deux-Èves or suffixed forms like aquarium > évier or Aquianum > Évian. It probably is a north-eastern form that spread beyond its normal area since /gw/ > /w/ is the normal outcome there.

    The contracted determiner del yielding du (/dɛl/ > /dy/) is also puzzling since normally stressed /ɛl/ becomes /o/ and unstressed /ɛl/+consonant becomes /ø/. In the same register, the pluraldemonstrative celz /(t)sɛl(t)s/ yielded ces /sɛ(z)/ when a determiner but ceux /sø/ as a pronoun.

    We've talked before in this forum about catum becoming gato in many Romance varieties, which also affected a few other /k/ initial words.

    My variety of Walloon developed tremulare > */trɛmler/ into /trijane/ instead of the expected /trɛne/ (compare similare > */sɛmler/ > /ʃɛne/ or in simul > */ɛnsɛmlə/ > /ɛʃɛn/). The root /trijan/ might be the result of a diphthongisation of /ɛ/ into something like */iə̯/ > /ija/ (stabilised ə turns into /a/ regularly) but that still looks ad hoc.

    A few resulting from aborted shifts in French: /r/ > /z/ in the 17th century turning chaire into chaise (chaire survives in a limited number of meanings, mostly episcopal or academic chairs) and be-/béricles (viewing glasses) into besicles/bésicles (compare Dutch brillen and Wall /bɛʀik/) ; /ke/ and /kje/, /kɛ/ and /kjɛ/ almost merging in the 19th century leading to a modern variation in the pronunciation of word with the root quiet- such as inquiéter as /ɛ̃kjɛte/ or /ɛ̃kɛte/ ; loss of coda /r/ in medieval Occitan leading to the mistaken restoration of /r/ in Massilia > Marseille.
     
    I don't have a source for this, but this looks like the normal lenition of open syllable medial /a/ in Old Latin, also seen in afficiō (ad+faciō) or inimicus (in+amīcos). Old Latin had initial stress, so the /a/ of the first syllable in Massalia/faciō/amīcos was preserved as is, but unstressed /a/ raised to /e/ or /i/ (probably [ɨ] at first) depending on syllable structure before stress shifted.

    So /'mas:alia/ > */'mas:ɨlia/ > /ma'sːilia/
     
    In Catalan, specially in the central variety, schwa (unstressed a/e) in syllables such as B.R/P.R/V.R/F.R/T.R is often elided in the casual speech. But this may happen in some words while -I don't know why- it just never does in others.

    berenar 'snack between lunch and dinner' [b(ə)ɾəˈna]​
    però 'but' [p(ə)ɾɔ]​
    veritat 'truth' [b(ə)ɾiˈtat] or [v(ə)ɾiˈtat]​
    feredat 'horror' [f(ə)ɾə'ðat]​
    taronja 'orange' [t(ə)ˈɾɔɲʒə]​
    Teresa 'Theresa' [t(ə)ˈɾɛ.zə]​
    but​
    barallar 'to quarrel' [bəɾəˈʎa]​
    perill 'danger' [pəˈɾiʎ]​
    verinós 'poisonous' [bəɾi'nos] or [vəɾi'nos]​
    ferida 'wound' [fə'ɾiðə]​
    tarannà 'way of being, mindset' [təɾən'na]​
     
    On the other hand, some words will refuse to undergo certain sound changes - for example, English "grass" (Old English "græs") should have metathesized to something like **gars /gɑːz/. (Old English /r/ metathesized with adjacent vowels, such as Old "þridda" > Modern "third")
    Dutch has the same sound change.

    Kerstmis < Christmas (Christus)
    (kik)vors < German Frosch
    pers < press
    borst < breast
    dertig < thirty (Dutch drie)
    worstelen < wrestle

    But grass in Dutch is simply gras.
     
    In Catalan, specially in the central variety, schwa (unstressed a/e) in syllables such as B.R/P.R/V.R/F.R/T.R is often elided in the casual speech. But this may happen in some words while -I don't know why- it just never does in others.
    Curious! Galician tends to do the same, so, for example pero 'but' was usually written pró, during the 18 and 19 centuries; Tereixa 'Therese' gave origin to treixadura, a local grape; and, while one must write ferida 'wound', many would rather pronounce frida ['fɾiðɐ].

    Other possible weird rules in Galician:
    a) result of Old Galician-Portuguese -ãar in verbs.
    The expected result would have been -ar, but in fact Galician (at least wester and central Galician) inserted a group -nd- in the hiatus:

    Latin sanare 'to heal' > OGP sãar > Galician sandar / Portuguese sarar (expected: **sar)
    Latin applanare 'to level; flatten' > OGP *achãar > West and Central Galician achandar / East Galician achaiar / Portuguese achanar (expected: **achar)
    Latin *rapinare 'to plunder' > OGP *rebãar > Galician rebandar 'to slice' (expected: **rebar)

    b) -ste/i- > -sche/i- > -che/i- versus -ste/i- > -stre/i-:

    Latin cōgitāstī 'you thought' > Medieval Galician cuydasche 'you thought' (13th century, Cantigas de Santa María) > Mo Galician coidaches 'you thought' (Portuguese cuidaste).

    You can find this same evolution in toponyms, as in Moeche, from the early medieval genitive formula Villa Modesti 'Modestus' villa' [attes.: Moesche 1323; Sancto Iohanne de Moestii 1159; Villam Modesti 977]. On the other hand there are fields called Moeste and Moestre elsewhere in Galicia.

    The same happens with place names derived of a local early medieval name, Nausti (with derivative Naustila, m, so probably an East Germanic name):

    - Noche, Vilalba, Lugo: ssan Martino de Noste 1282 < pontem de Nousti 1124
    - Vilanustre, Rianxo, A Coruña: *villa Nausti
    - Carnostre (now, a field), Sober, Lugo: *casa Nausti ('Nausti's house')
    - Nouche (a field), O Pino, A Coruña: *Nausti
    - Lustiás, Taboada, Lugo: Nustiaas 1402 < *villa Nautilanis 'Naustila's villa'.

    - Faluche, Chantada, Lugo: Salusche 1473 < *villa Sallusti 'Sallustius' villa'
    - Sabachán (a field), Santiago de Compostela, A Coruña: *Sebastiani 'Sebastianus' villa'
    - Fachán, Rianxo, A Coruña: *Fastilani 'Fastila's villa'

    I don't know a geographic or linguistic context for words taking one way or the other...
     
    Last edited:
    One typologically common (and probably the only truly irregular) example is various simplifications in frequent words, and Russian has these in abundance
    When in my first Russian lesson I learned that "hello" was "Здравствуйте" I wondered if I had made a mistake choosing Russian. The teacher did though go on to say that it was often reduced to "Здрасте".
     
    When in my first Russian lesson I learned that "hello" was "Здравствуйте" I wondered if I had made a mistake choosing Russian. The teacher did though go on to say that it was often reduced to "Здрасте".
    Well, the first "v" here is always silent anyway. On the other hand, there's nothing unfeasible in the [fstv] cluster (cf. words like шефствовать) if you remember that the Russian labiodentals are lax and heavily retracted (up to the extent when you can actually call them "bilabiodentals"; it's no coincidence that they're occasionally transcribed simply as bilabial fricatives or approximants) and that the Russian /s/ and /t/ are normally dental and dentialveolar respectively; as a result, [fstv] will be produced by very small articulatory movements.
     
    an example of sporadic sound change is the English verb to speak. In Old English it was "sprecan", normally the consonant cluster spr- is not simplified. Compare the related Dutch word 'spreken' and German 'sprechen'.
     
    Last edited:
    an example of sporadic sound change is the English verb to speak. In Old English it was "sprecen", normally the consonant cluster spr- is not simplified. Compare the related Dutch word 'spreken' and German 'sprechen'.
    Finding this proposed loss of r from the stable cluster spr unparalleled, Hill instead sets up a different root, Proto-West Germanic *spekan (“to negotiate”) from Proto-Indo-European *bʰégʾ-e- (“to distribute”) with *s-mobile, which collapsed in meaning with *sprekan ("to speak" < "to crackle, prattle") and so came to be seen as a free variant thereof.
     
    Back
    Top