Exact definition of Cognates?

[marconatrix]

I thought I fully understood the concept of cognate words, morphemes etc. as being the descendants of the 'same' original item coming down through different branches of a language family to present and past attested forms. However a list of 'cognates' I was asked to comment on recently, mostly concerning the Celtic languages and their relationship with English, Norse, Latin etc. has got me a bit confused.

For example, Old Irish borrowed a fair number of words from British many of which in turn had been borrowed by the Ancient Brits from Latin during the Roman Occupation, or later in connection with christianity. The loans to OI then passed down to the modern three Gaelic languages, while the British word descended to Welsh/Cornish/Breton. So for example, the Latin word vespera 'evening' was borrowed by British and comes through in Welsh as gosber, Breton gousper (mainly in connection with the 'vespers' evening service), and was passed on to OI giving e.g. Scots Gaelic feasgar, Manx fastar which are the normal everyday words for 'evening'. Question : which of these words are cognates?

If a word common to all the modern Celtic languages could be shown to go back to Common Celtic, but nevertheless could then be shown to have been an early loan to CC (for example from Germanic), would those words be technically cognates or loans?

The English word 'potato' (from wherever) appears to have been borrowed independently into five or the six modern Celtic langs. (and probably independently in at least two Mod. Irish dialects). Are these words 'cognates' with English 'potato' or with one another, or neither?

In the last case I'd say definitely 'no', apart from anything else the sound changes are pretty random. But in the earlier examples there are usually quite predictable sound changes, e.g. Latin /-kt-/ > British /-xt-/ > OI /-xt-/ > SG /-xk/, and on a different branch > Welsh /-jT/ > Breton /-jz/.

To add to the confusion, a British word might be borrowed into Irish at two different periods, and either undergo a whole list of Irish sound changes (if borrowed early) or a parallel but different list of British developments (if borrowed later), and sometimes the same word went through both routes given two different words in Irish. How do you deal with that when compiling a cognates list?

Or a borrowing from the other branch can sit alongside the same word that's been inherited down the regular path.

So you see, I'm a bit confused ...

[Shm Jay]

If a word common to all the modern Celtic languages could be shown to go back to Common Celtic, but nevertheless could then be shown to have been an early loan to CC (for example from Germanic), would those words be technically cognates or loans?

Cognates. The loan is to Common Celtic.

[zompist]

So for example, the Latin word vespera 'evening' was borrowed by British and comes through in Welsh as gosber, Breton gousper (mainly in connection with the 'vespers' evening service), and was passed on to OI giving e.g. Scots Gaelic feasgar, Manx fastar which are the normal everyday words for 'evening'. Question : which of these words are cognates?

The Welsh and Breton terms are cognates of the British word; the Manx and Scots Gaelic words are cognates of the OI word.

But the British and OI words aren't cognate, because they're not inherited from proto-Celtic.

[Bristel]

I think potato is borrowed from Taino batata, does that change anything?

[Niedokonany]

I've sometimes found these terms lacking: words borrowed by two languages from a common source word are ...., a loanword and its source are ...., an inherited word and a loanword of a source word that's cognate to the inherited word are .... Somebody should definitely coin these in English so we could steal the goodies, sparing our derivational morphology.

[Terra]

So for example, the Latin word vespera 'evening' was borrowed by British and comes through in Welsh as gosber, Breton gousper (mainly in connection with the 'vespers' evening service), and was passed on to OI giving e.g. Scots Gaelic feasgar, Manx fastar which are the normal everyday words for 'evening'.

To make a small tangent, what's the approximate range of years that the /v/ -> /g/ sound change happened? And taking a stab in the dark, was there ever a /k/ -> /g/ sound change as well?

[The Unseen]

Latin "v" is /w/ (or /v\/, but no big diff)

there was a common change of initial [w] -> [gw], then [gw] -> [g].

I don't know if that's what happened in the Celtic langs, but it happened in French.

[Terra]

Latin "v" is /w/ (or /v\/, but no big diff)

I know.

there was a common change of initial [w] -> [gw], then [gw] -> [g].

I don't know if that's what happened in the Celtic langs, but it happened in French.

Hence my question. I thought perhaps the French sound change was influenced by the Celtic one (or vice-versa). Also, French had a sporadic sound change that changed some /k/ to /g/, but not all. I can look up/recall some words that underwent both of these sound changes if anyone wants.

[marconatrix]

I doubt very much if the French change was influenced by Brittonic. The change happened in British after the bulk of Latin loans were taken on board, so by that time any Celtic languages on the continent would have been dead or nearly so. The change also affected native words. In non leniting initial positions /w/ > /gw/. A later change sometimes removed the /w/ before rounded vowels. This was regular in Cornish, in Welsh /wu/ > /u/ (which is why the Welsh can use 'w' to write /u/). W. gosper is not quite regular. I would guess that at some stage it was gwsper /gusper/ < /gwusper/ < /gwesper/. All these changes are almost if not quite regular, but that's often the case with Latin loans, especially Christian words which were usually borrowed a bit later than other Latin loans.

In Irish initial /w-/ (> /B/ ?) > /f-/ is regular and happens between the proto-OI ogham inscriptions, and the earliest ms OI. The substitution of /k/ (< /q/) for /p/ (or in this word /sk/ for /sp/ OI fescor) is characteristic of the earliest layer of loans from British to Irish. At this stage Irish had no /p/ phoneme. The change of /sk/ > /st/ in Manx is regular, and relatively recent. (I was starting to wonder if the Irish word might be direct from PIE, but as far as I can tell PIE /-sp-/ would have become /-ss-/ in Celtic).

As for time scale in the British languages we're talking about at least Romano-British > Old WCB, say 600 years minimum, with the changes fully completed by the time of Middle Welsh etc. c1,000 years.

[Tropylium]

I've sometimes found these terms lacking: words borrowed by two languages from a common source word are ...., a loanword and its source are ...., an inherited word and a loanword of a source word that's cognate to the inherited word are .... Somebody should definitely coin these in English so we could steal the goodies, sparing our derivational morphology.

After the excellent blog, I propose "bradshaw" for the last of these (incl. cases when both words are loans).

[Circeus]

So for example, the Latin word vespera 'evening' was borrowed by British and comes through in Welsh as gosber, Breton gousper (mainly in connection with the 'vespers' evening service), and was passed on to OI giving e.g. Scots Gaelic feasgar, Manx fastar which are the normal everyday words for 'evening'. Question : which of these words are cognates?

The Welsh and Breton terms are cognates of the British word; the Manx and Scots Gaelic words are cognates of the OI word.

But the British and OI words aren't cognate, because they're not inherited from proto-Celtic.

Actually, I beg to differ here. I dunno if it's only me, but cognates only applies to word having a sister/cousin/uncle relation, not to descendants. The Manx, Scots, Welsh and Breton words are cognate to each others: they are descended from the same word that existed in a single language at some point (in this case the borrowed word in British, if I understood the explanations right). Words descended from a borrowing can still be cognate: Fr. "coupe" and Ge. "Kopf" are cognates: both are descended from the Latin (There can be separate cognates via PIE, of course).

But then that's just my interpretation

[zompist]

Yeah, it'd be better to say the Welsh and Breton words are reflexes of the British word.

[marconatrix]

The thing is (and I'm sure there are many parallel cases) there was quite of flow of words from Latin > British (and on downstream to the daughter languages) and then many of the same words, plus some native British ones > Old Irish (and so down to the modern Gaelic languages), and the adaptations that took place with each borrowing were reasonably regular. So these paths are just as familiar, and just as important as the path from PIE > Common Celtic > British/Old Irish.

Normally, I think, we think of loans as rather random, irregular and unpredictable events that follow no rules (e.g. the 'potato' words borrowed independently into modern Celtic langs from English). So how does one describe these heavy 'cross flows' of words travelling in parallel from lang A > lang B within a restricted time period and showing 'cognate-like' sound correspondences?

Would not a similar state of affairs apply to e.g. Arabic loans to Swahilii, or Arabic loans via Persian to Urdu, or Germanic words in Fennic etc.