29 August 2014

Placing the Heart in Indo-European

In Sanskrit the main word for 'heart' is hṛd or the suffixed form hṛdaya. However most of us are also familiar with the word for faith, śraddhā, which we think means 'placing the heart'. Here the word for heart is śrad. Most sources suggest that the two words hṛd and śrad are in fact two forms of a single word that has undergone a series of phonetic transformations. However some sources suggest that there are two distinct roots. This word makes for an interesting case study in comparative linguistics and shows the kind of evidence that is used to reconstruct Proto-Indo-European. In this case the reconstructed root is written different ways: k̑ered- : k̑erd-, k̑ērd-, k̑r̥d-, k̑red-.

We need to note some conventions. Where a form is reconstructed and/or not actually attested it is frequently suffixed with an asterix: *kred. In order to distinguish phonemes or sounds from letters they are written between two forward slashes. Thus come has initial c but is pronounced /k/. Similarly c can be pronounced also as /s/. Linguists make a distinction between /ḱ/ or /k̑/ the palato-velar and /k/ the plain velar. The difference is the k sound in "scum" and "come". If you pay close attention to where your tongue is in your mouth as you pronounce these words you'll notice the tongue makes contact with the roof of the mouth further forward in the word "scum" and is the /ḱ/ sound. In English the two are allophones, meaning we don't hear the difference and don't notate it differently. This seems to be true of other Indo-European languages also, though the distinction is important in Proto-Indo-European (PIE) the putative mother language of all the present day and dead IE languages.

I'll begin with a survey of the various cognates in other Indo-European languages by family grouping. In Indo-Iranian we have several related forms:

Vedic: śradśraddhā

It's quite regular for Avesta to have /z/ where Sanskrit has /ś/. In Sanskrit dhā is plain root meaning 'place/put'. Perhaps because of the conservative nature of religion we can see this form throughout the Indo-European language family. We need to ask about the two distinct forms in Vedic. There are two possible explanations for this situation.
  1. There was a progressive change /ḱ/ > /ś/ > /h/
  2. The word came into Vedic twice: PIE /ḱ/ > /ś/ and /ḱ/ > /h/.
There is a regular change from PIE /k/ to Vedic /ś/ so we can quite easily explain kred > śrad. What we need to explain then is either  /ś/ > /h/  or /ḱ/ > /h/.


As we move from west we find three other language family has the change from palatal stop to sibilant. The Armenian form is sirt. It's very easy for a voiced consonant /d/ to change to an unvoiced consonant /t/ with the same articulation (compare Latin pater - German fader - English father). Slavic languages follow a similar pattern: heart = Church Slavonic srĭdĭce; Russian serdtse; Polish serce; Slovak srdce. And finally the Baltic languages: Old Prussian seyr; Lithuanian šerdìs; Latvian ser̃de. We can see that in most of these words a vowel is interposed between the initial /s/ and /r/. Here, then, the initial consonant change is /ḱ/ > /s/ 

The Anatolian, Helenic and Italic families preserved the /k/ though this is often spelled 'c'. Thus we see Hittite kar-ti-ya-aš 'heart'; Greek kardia (καρδία); Latin cordis (and credo 'trust, believe'). The Latin gives rise to Romance Language forms: French cœur; Italian cuore; Spanish corazón; Portuguese coração; Romanian cord. Note the dropping of the final stop in French and Italian. In Spain /d/ becomes /z/ and in Portuguese /d/ > /sh/. One can see how this might have come about in the sequence: /d/ > /dz/ > /z/ > /sh/. Initial /ḱ/ is preserved, though it may become the plain velar /k/. The notation is ambiguous.

Celtic similarly preserved initial /k/: Old Irish cretim; Cornish créz; Welsh craidd. Though the more common word for 'heart' in Welsh seems to be the unrelated calon.

Germanic languages change initial /k/ to /h/ which is interesting because this is just the change that we are looking for. There's no question of any communication between Germanic and Sanskrit, it's just a case of parallel evolution, but it's helpful to know that one of the transformations that an initial /ḱ/ can undergo is change to /h/. The Proto-Germanic form is *herton- (OEtD) The Germanic family is divided geographically. West Germany covers what's now Germany and Holland. East is represented by a single dead language, i.e. Gothic. North is all of the Scandinavian languages. Usually English has been considered to be part of the West Germanic sub-family, related to Old Saxon. However more recently a case has been made to consider it part of the Northern sub-family along with Scandinavia. The Germanic speakers—Saxons, Angles and Jutes—who settled in Britain did come from what is now the state of Schleswig-Holstein in the North of Germany, which abuts Denmark. And of course there was a significant overlay of Danish onto Old English as well.

West North
Old Frisian herte/hirteOld Icelandichjarta
Dutch hart Swedish hjärta
Old Saxon herta Danish hjerte
Old High German  herza Old English heorte
German herz Mid. English hert
English heart

With the Norman Invasion English picked up the Latin derived words for heart as well, as can be seen in words such as accord, cordial, courage, credible, credit, creed, grant, miscreant, and quarry (i.e. prey). 

What comparative linguistics does is to look at all the forms and look for logical transformations that might account for all the forms. Such changes much be checked against a range of words with the same sounds. It's only when patterns emerge across a wide range of words that one can describe regular changes (what we might once have called formulating laws). The more obvious examples help to explain the less obvious.

*kred vs *kerd

I recently read that treating kred and kerd as the same root might be incorrect.
Outside of the verbal system we find another word that curiously seems to display such a gradation and that is *ḱerd- 'heart', while in Sanskrit we find hṛd- and in Avestan we find zərəd- which both seem to go back to *ǵʰrd-, and then there's the Sanskrit śrad- (notice the schwebe ablaut!) which in combination with dhā- 'to give' give a lovely indo-european expression also found in Latin Credere 'to believe'. This form seems to go back to *ḱred-.
A schwebe ablaut is a full-grade vowel that is not always in the same position within the root. We don't really talk about vowel grades in English though we use them, e.g. in the verb sing sung sang and the related noun song the changed vowel gives us grammatical information. Ablaut is a very important part of Sanskrit morphology, for example those derived from √dṛś exhibit the various vowel strengths: dṛṣti, darśana, draṣṭṛ, adrākṣit. But note that darśana and draṣṭṛ invert the order of the vowels in the stronger grades (what the Sanskrit grammarians called guṇa and vṛddhi). Modern grammarians make guṇa the normal grade and talk about a weaker () and a stronger (ār/rā) grade. So in √dṛś the vowel grades from weakest to strongest are: , ar/ra, and ār/rā. With one sees that strengthening in Sanskrit is like adding ă (short a) before the root vowel and applying the rules of sandhi. With other vowels the changes are a bit less obvious.

Another example is √bhū 'to be'. The vowel grades for ū are: ū, o, & au, where o ≈ ă+u; and au ≈ ā+u. [note au is a diphthong]. √bhū forms a present stem by strengthening the root and adding a.
  • root:  bhū
  • guṇa: bh[ăū] (= bho)
  • + a:    bhă-ūa (which sandhi resolves to) bhava-
  • conjugations: bhavāmi, bhavasi, bhavati etc.
In the past particle, the root stays in the weaker grade so: bhūta. And in the strongest grade we find the noun bhauta 'related to living beings'. These processes were first described by Pāṇini in perhaps the 4th century BCE. Through study of Sanskrit grammar the principles were rediscovered by the first European comparative linguists. The term ablaut (German 'off sound') for this phenomenon was coined by Jacob Grimm of "the Brothers Grimm" in the late 19th century.

Edward Sapir notes a difference in the Tocharian word for heart:
The Tocharian word [käryā] does not represent IE *ḱṛd-yā́ (i.e. *erd-yā́) but *ḱred-yā́ (reduced from the basic *ḱred- seen in Sanskrit śrad and Latin crēdō < * krede-dō).
Selected Writings of Edward Sapir. University of California Press, 1968 p.227

So Sapir is also distinguishing *kred and *kṛd (> kerd). Jonathan Slocum's PIE Etyma List records two roots: kered- and kred-, but they both mean heart and the list of cognates or reflexes does not distinguish between them.

Sanskrit roots with initial /h/ are very often degraded from /dh/ or /gh/. Sometimes this becomes obvious. For example in the root √han 'to strike' The present tense 3rd person singular is hanti, but the 3rd person plural is ghnanti. Similarly we find the perfect jaghāna, a past participle ghāta (also hata). (cf my comments on the word saṅgha). So we can deduce from this that √han must originally be from *√ghan. In other roots the archaic forms don't survive. So on face value the idea that hṛd might represent an archaic *ǵʰrd- is not outrageous. However I can't find a root *ǵʰrd- in any PIE etyma list I have access to. Nor does any root that I can find seem to fit the bill.

Words related to either hṛd or śrad are few and far between in Sanskrit. If hṛd were a separate root we might expect a word hrada, and we do find such a word, but it means "a deep pool" and it seems to be connected to the root √hlād or √hlad 'refresh', which has only sporadic use. In India there is a regular confusion of l and r. In fact on the Asoka pillars the word for king is lāja not rāja. Both words have a specific domain beyond which they have little use: hṛd, hṛdaya 'heart'; hṛdya 'in or of the heart; charming etc'. Versus śraddhā 'faith', śrāddha 'faithful, funeral rite'. We do see some variants on √dhā used with śradśrad-dadhānaśrad-dhayitaśrad-dhitaśrad-dheya but the sense stays the same. 


Sound changes cannot happen at random. And yet the change from /k/ to /s/ is counter intuitive. It is logical however. The steps to get the other sounds look like this (as best I can tell):

/k̑/ > /k/ involves a moving the tongue back slightly in the mouth. Allowing some air past the tongue gives /kh/ and dropping the stop altogether leaves us with /h/. The same change starting from /k̑/ > /h/ > /ś/. Voicing /ś/ gives /z/ and moving the tongue a little forward and dropping the aspiration gives /s/. 

We've already seen how ar can alternate with ra with no change in sense. And lastly we have to allow for a vowel to interpose between two consonants. Describing all the vowel changes would extend this essay too much, but one can work through the logic of that as well. We've now described how one root with a weakest grade *kṛd and strong grades kred or kerd, could produce all the many variants by the application of simple rules that are anchored in how the tongue moves in the mouth to produce vocal sounds. With this particular word the sense of it has remained remarkably stable - all the different languages understand this word to mean "heart".

So can we now say any more about Sanskrit hṛd/śrad? Looking at the IE cognates it seems that only Sanskrit exhibits this alternation of ar and ra. So my suspicion is that Sanskrit has retained this feature of PIE rather more prominently than other languages. Since getting from /h/ to /ś/ is a complex process I conclude that it is less likely than the other option: that either hṛd or śrad is a loan word from another branch of the Indo-European family. Since the expected change is /k̑/ > /ś/ and the Avestan has /z/ it looks like hṛd is a loan word from a closely related language that favoured the change /k̑/ > /h/. This doesn't eliminate the possibility that hṛd comes from a root *ǵʰrd-, just that with the resources at my disposal I cannot find such a root. Since the sense of the word changes so little across time and space, however, it seems less likely that there were two roots.

The point here is not to draw strong conclusions, but to think about how language sounds change over time, and how that change tends to be systematic: e.g. all initial /k̑/ change to /ś/. It is interesting that some of the main distinctions between say Vedic and Avestan, or between Vedic and Pāḷi are just these systemic changes in pronunciation.


Note 7 Sept 2014

According to a chart on wikipedia (I know) Avestan /z/ is a reflex of PIE /ǵ/ and the corresponding Sanskrit reflex is /h/. Thus Skt. hṛd and Av zrad might well come from *ǵʰrd-. And śrad < *ḱred is another root. What is the relationship between *ǵʰrd- and *ḱred? How to square this with cognates in other languages that clearly point to heart < *ḱred?

*ḱm̥tóm, 'hundred' > Vedic Sanskrit: śatám & Later Avestan: satəm 
*ǵʰasto- 'hand' > Skt. hástas & Av. zasta, 

Note 13 Sept 2014
Still looking into this in a desultory way. As well as ablaut (the change in vowel strength) we also see changes in the strength of consonants. The change from unvoiced stop // to voiced stop /ǵ/, as well as the change from stop to fricative /ś/ is an example of lenition or "weakening". *gʰan > han is another example of this phenomenon. The opposite process is fortition. Another type of lenition, called debuccalization is the weakening of final /s/ to /ḥ/ (as in manas > manaḥ).

9 Apr 2017

In a recent discussion on this issue on academia.edu, German Dziebel wrote:

We know that Lat cre:do, OIr cretid go back to *k'red- attested in Skrt śraddha- and, with an original root shape and a syllabic r, in Lith sirdis, etc. So, we don't need extra proof that there was a palatovelar there. We know it already. We  also have a voiced variant of it attested in hrd/zered, so we only need to understand how this voiced palatovelar got devoiced. We know that *s+D can give *sT in any IE dialect [Siebs's Law], and in Skrt ś can go back to sk- (śardha next to Lith (s)kerdzius). This is what Lubotsky shows. If Skrt ś can go back to sk-, then it can go back to sk'-, too, if comparanda warrants a PIE palatovelar after s-. This is simple logic. For us, to link hrd and śraddha- we just need to postulate 1) *g'herd- > InIr *j'hrd; and 2) *(s)g'herd-/*(s)g'hred- > *sk'red- > *śrad-, and both hrd and śraddha- receive a systematic explanation. I think where I got you confused is in bringing up Lubotsky who described how sk- > ś- in Skrt before a front vowel. He didn't tackle a situation such as śraddha where k' is inherited from PIE and what we need to prove is not the emergence of palatovelar from *sk- but the presence of s- before a palatovelar because it is that s-mobile that we're attributing the devoicing of the palatovelar to.

Related Posts with Thumbnails