Linguistic Miscellany Thread

[Kuchigakatai]

Amharic /sʼ/ is also [tsʼ ~ sʼ] in free variation. And Semitic languages in general, at some point in their development, likely had [tɬ ~ ɬ] and [tɬʼ ~ ɬʼ] as phonemes too (and [s ~ ts], [z ~ dz], [tθʼ ~ θʼ]). I've even seen "[dɮˁ]" sometimes as a possible realization of the Classical Arabic ض, as opposed to "[ɮˁ]", but I don't know on what grounds that might be based on.

And there are Spanish dialects in parts of northern and eastern Mexico, and Chile, where /tʃ/ is [ʃ] or can be [ʃ], possibly in free variation with [tʃ] at least in mid-frequency words.

[Travis B.]

Amharic /sʼ/ is also [tsʼ ~ sʼ] in free variation. And Semitic languages in general, at some point in their development, likely had [tɬ ~ ɬ] and [tɬʼ ~ ɬʼ] as phonemes too (and [s ~ ts], [z ~ dz], [tθʼ ~ θʼ]). I've even seen "[dɮˁ]" sometimes as a possible realization of the Classical Arabic ض, as opposed to "[ɮˁ]", but I don't know on what grounds that might be based on.

And there are Spanish dialects in parts of northern and eastern Mexico, and Chile, where /tʃ/ is [ʃ] or can be [ʃ], possibly in free variation with [tʃ] at least in mid-frequency words.

It is highly suspected that Proto-Semitic *s was really [ts] based on evidence from loans and like, and apparently it's almost certain that PS *s' (I'm having trouble typing an underdot) was really [tsʼ] (or at least glottalized if not ejective), considering its reflexes in Ethio-Semitic and Hebrew. As for PS *š, it is the only PS sibilant that is clearly, unambiguously a fricative, but what it was realized as beyond that is unclear aside from that it was voiceless. It should be noted for the reader that all standard PS transcriptions of sibilants (and PS *ś, which is pretty uncontroversially not a sibilant) are essentially convention and should be taken with a heaping of salt.

[anteallach]

The point being it's certainly possible for a language to have a certain affricate without the corresponding fricative, or the two to be in free variation.

Indeed UPSID claims 120 languages in its sample with "voiced palato-alveolar sibilant affricate" but only 63 with "voiced palato-alveolar sibilant fricative".

[foxcatdog]

Regardless in this particular case unless you're going for a Nahuatl or proto semitic ripoff it believe it would be pertinent to change any *tɬ to *ɬ or something else like *tɬ > *ts.. *s > *ɬ is a good sound change if you want to acquire *ɬ btw followed by depalatisation of any palatal sibilants (attested in Uralic).

[foxcatdog]

Besides this of course non affricate sources of *ɬ include ( i searched Index Diachronica and remembered stuff for you)...
Fortition of *ɬ see Welsh for the initial version as well as before *t but tons of instances of geminate *l becoming *ɬ
*θ to *ɬ (Eastern Muskogean)
Voicelss *l > *ɬ (Though this is trivial)
I'm assuming any sort of fricative + *l cluster could become *ɬ if Stop + *ɬ is the source of your *tɬ
*ɬ > *ʃ is the more well known sound change in arabic but you could possibly reverse it.

Sources of *θ include
*t in between vowels (Trivial lenition)
*ʃ (Gros Ventrina but it is conditioned by a back vowel and also note it already has the phoneme)
Any voiceless affricate (Old Spanish to Spanish, Indo Iranian to Old Persian, Algonquian to Arapaho, Salishan, Iroquian and probably Tosk Albanian (which has the same source of Persian *θ but as a direct sound change but possibly through the same stages)
*sd (so also *st) > *θ medially (Tosk Albanian)
*xt > *θ (Middle Welsh here it also includes an *i)
*t > *θ or *th > *θ if you have to many stop series (Germanic) Also Greek and Latin if you have voiceless or voiced aspirates (at least people have hypothesised the middle stage to initial *θ in Italic would have been its voiced counterpart rather then following the greek example)
*s before *r (Latin)
*t: (Irish but it already has medial *t > *d)

[bradrn]

Besides this of course non affricate sources of *ɬ include ( i searched Index Diachronica and remembered stuff for you)...

I will note, as I always do, that the Index Diachronica is notoriously unreliable; if those changes are all from the ID, probably some of them are correct, but I wouldn’t stake money on it.

[Rounin Ryuuji]

It's also worth noting that a language with an artistic purpose can have an unusual feature or two, and the creation is hypothetically a snapshot of a language at a given point, so it could be that whatever creation happens to be at the point where the affricate exists, but hasn't (yet) deaffricated, or the corresponding fricative, developing from another source, has not yet emerged. I don't think there's any reason a source of the fricative has to be found if it isn't wanted.

[Zju]

Fair enough. I can handwave it as being a transitional stage of one or two generations before conditional t͡ɬ > ɬ (and before anyone brings it up, I do know all language stages are transitional).
I guess I just wanted more than a few examples of ANADEW.

[Nortaneous]

It's also worth noting that a language with an artistic purpose can have an unusual feature or two, and the creation is hypothetically a snapshot of a language at a given point, so it could be that whatever creation happens to be at the point where the affricate exists, but hasn't (yet) deaffricated, or the corresponding fricative, developing from another source, has not yet emerged. I don't think there's any reason a source of the fricative has to be found if it isn't wanted.

Yeah most language families have at least one bizarre sound change that wouldn't make sense were it not attested. At one point in ZBB lore the conditioned shift of *t > tɬ in Nahuatl was the standard example, but I'd also cite most developments in North Bougainville, which AFAICT has to be reconstructed with 13 consonants, over half of which merged in the only descendant documented to contemporary standards (see here)

[Jonlang]

[ɬ] in Welsh also seems to have come out of some bat-shit nowhere place. If I'm remembering correctly the cluster [sl] > [ɬ] may have been first, and then other changes like [ln] > [ll] > [ɬ], and [lt] > [ɬt], and word-initial [l] > [ɬ].

[anteallach]

[ɬ] in Welsh also seems to have come out of some bat-shit nowhere place. If I'm remembering correctly the cluster [sl] > [ɬ] may have been first, and then other changes like [ln] > [ll] > [ɬ], and [lt] > [ɬt], and word-initial [l] > [ɬ].

I think it's analogous to the fortis laterals in some Goidelic varieties. There, in initial position you usually get the fortis one but under the lenition mutation it becomes lenis (parallel to /ɬ/ becoming /l/ under soft mutation in Welsh) while elsewhere the fortis sounds come from older geminates and are spelt <ll>. Once you think of it as a fortis sound it turning into [ɬ] seems less weird.

[Linguoboy]

Once you think of it as a fortis sound it turning into [ɬ] seems less weird.

Honestly it's a less weird sound change to me than the spontaneous palatalisation of fortis laterals in Romance.

[Travis B.]

Once you think of it as a fortis sound it turning into [ɬ] seems less weird.

Honestly it's a less weird sound change to me than the spontaneous palatalisation of fortis laterals in Romance.

I must agree - I don't get how gemination should turn into palatalization myself...

[Darren]

Once you think of it as a fortis sound it turning into [ɬ] seems less weird.

Honestly it's a less weird sound change to me than the spontaneous palatalisation of fortis laterals in Romance.

I must agree - I don't get how gemination should turn into palatalization myself...

The best explanation for the palatalisation of /lː/ is a combination of articulatory emphasis thanks to its fortis nature, and a functional incentive to distance it from singleton /l/. If you can get your hands on the book Palatals in the History of the Romance Languages [edit: it's actually called Palatal sound change in the romance languages : diachronic and synchronic perspectives], section 4.3.2.2 talks about the changes which /lː/ underwent throughout Romance (i.e. degemination, fortition, retroflexion, palatalisation and rhotacism in different varieties). The author suggests that there may have been a split at some point which led to these diverse results;

Vázquez Obrador [...] for example proposes that inherrent variability in the production of [lː] contributed to have its fate follow two directions; namely (i) an articulatory emphasis on the tongue dorsum against the hard palate, which produced results such as [ʎ] in Spanish and Catalan; and (ii) an articulatory emphasis on the tongue tip and blade, whose retroflex movement produced outcomes such as [ɖ] in Upper Aragonese, Gascon, dialectal Astur-Leonese, and the southern dialects of Italy.

It's also important to note that in Spanish, palatalisation of /lː/ was functionally motivated by the desire to keep it distinct from singular /l/; all other single/geminate pairs in Spanish remain distinct in general (other than /m~mː/); and /lː/ was super common for a geminate. Palatalisation of /lː/ (and likewise /nː/) was a way to degeminate them without losing their distinctiveness, which would've merged too many minimal pairs. In contrast in Portuguese, singleton intervocalic /l/ was lost, so there was no incentive to palatalise the long form and it was just degeminated.

In light of this, the author also quotes

Holt (2003b) argues that [...] "instead, in the process of loss of length, original energy associated with the articulation of geminates is maintained by spreading out the region of contact of the tongue with the roof of the mouth"

I'd be interested to see if there are any good non-Romance examples of "spontaneous" palatalisation of geminates to compare this to.

[anteallach]

I'd be interested to see if there are any good non-Romance examples of "spontaneous" palatalisation of geminates to compare this to.

I haven't found a very clear description, but there's apparently some palatalisation of geminate coronals in northern Norwegian dialects.

[Jonlang]

[ɬ] in Welsh also seems to have come out of some bat-shit nowhere place. If I'm remembering correctly the cluster [sl] > [ɬ] may have been first, and then other changes like [ln] > [ll] > [ɬ], and [lt] > [ɬt], and word-initial [l] > [ɬ].

I think it's analogous to the fortis laterals in some Goidelic varieties. There, in initial position you usually get the fortis one but under the lenition mutation it becomes lenis (parallel to /ɬ/ becoming /l/ under soft mutation in Welsh) while elsewhere the fortis sounds come from older geminates and are spelt <ll>. Once you think of it as a fortis sound it turning into [ɬ] seems less weird.

This gives me the opportunity to bring up something I have often wondered. Is a shift to a particular phoneme more likely if that phoneme already exists (phonemically) in the language, rather than to, say, a more-likely shift cross-linguistically?

[vlad]

[ɬ] in Welsh also seems to have come out of some bat-shit nowhere place. If I'm remembering correctly the cluster [sl] > [ɬ] may have been first, and then other changes like [ln] > [ll] > [ɬ], and [lt] > [ɬt], and word-initial [l] > [ɬ].

I think it's analogous to the fortis laterals in some Goidelic varieties. There, in initial position you usually get the fortis one but under the lenition mutation it becomes lenis (parallel to /ɬ/ becoming /l/ under soft mutation in Welsh) while elsewhere the fortis sounds come from older geminates and are spelt <ll>. Once you think of it as a fortis sound it turning into [ɬ] seems less weird.

This gives me the opportunity to bring up something I have often wondered. Is a shift to a particular phoneme more likely if that phoneme already exists (phonemically) in the language, rather than to, say, a more-likely shift cross-linguistically?

I've always assumed so.

In English, [tj dj] became [tʃ dʒ] in words like nature, gradual (as well as tune, duke outside North America), but [nj] either remained [nj] (as in annual) or became [n] (as in new). I suspect that if English had already had [ɲ], then the [nj] in words like annual would have become [ɲ].

Then again, [zj] became [ʒ] in words like pleasure despite [ʒ] not previously existing in English. But maybe it's easier to innovate a voiced fricative when the voiceless equivalent already exists (and your language already has a voiceless/voiced contrast for other fricatives).

[bradrn]

Then again, [zj] became [ʒ] in words like pleasure despite [ʒ] not previously existing in English. But maybe it's easier to innovate a voiced fricative when the voiceless equivalent already exists (and your language already has a voiceless/voiced contrast for other fricatives).

I always thought that [ʒ] was a loan from French… is that not the case?

[Rounin Ryuuji]

Then again, [zj] became [ʒ] in words like pleasure despite [ʒ] not previously existing in English. But maybe it's easier to innovate a voiced fricative when the voiceless equivalent already exists (and your language already has a voiceless/voiced contrast for other fricatives).

I always thought that [ʒ] was a loan from French… is that not the case?

I think it probably first showed up in some post-deaffrication borrowings from Continental French. Yod-coalescence is a fairly recent change, if I'm remembering correctly, and it hasn't yet affected all varieties of English.

[WeepingElf]

It is more likely that a language gains a new phoneme from loanwords if that new phoneme fits a gap in the existing phonology rather than necessitating the introduction of a new distinction. Greek and Germanic loanwords in Slavic are a particularly illustrative case. Middle Greek and early Germanic had the spirants /f/, /θ/ and /x/, of which only /x/ existed in Proto-Slavic. What happened was that /f/ was added to the Slavic consonant inventory, as it filled a gap - it was the voiceless counterpart of the already existing /v/, and at the same time to /p/ what /s/ was to /t/ and /x/ to /k/. But /θ/ was not - there was no /ð/ or whatever, and /t/ had already /s/ as its "fricative partner"; it would have been a completely new consonant. Instead, it was rendered as /f/ in loanwords, hence Russian names like Fyodor (Theodore), and two letters for /f/ in the Glagolitic and Cyrillic alphabets corresponding to Greek θ and φ (the former was abolished later).