Th e Etymology of Vedic brav i ‘to say, to speak, to tell’

Th is paper argues that the standard etymology of Vedic brav i ‘to say, to speak, to tell’ from Proto-Indo-Euro-pean * mleu ̯ h 2 - ‘to speak’ (and its connection with Avestan mrao - ‘to say, to speak’) cannot be upheld, since it is based on an irregular consonant change that cannot be independently motivated and explained. As an alternative, two diﬀ erent PIE verbal roots will be proposed, * melH-u - → mleu ̯ H - ‘to say, to speak’ and * bleu ̯ h 2/3 ‘to speak or to call’, that provide phonologically and semantically regular bases for the words involved.


THE SUGGESTIONS FOR THE INITIAL CONSONANT
Setting aside EIEC: 535 and Mallory and Adams 2006: 353-354, who leave the initial consonant unexplained, the most widespread view attributes the change *ml-> brto alleged special circumstances surrounding the verb meaning 'to tell': 'br° beruht wahrscheinlich auf den Sonderbedingungen eines Verbums für "sagen"' (EWAia-II: 236, followed by LIV 2 : 446, see already KEWA-II: 452, followed by Schmidt 1982: 365 n. 21). KEWA-II: 452 refers to Bartholomae 1896: 712 andTurner 1937: 13-14. Bartholomae argues as follows: 'Die ausnahmsweise Verallgemeinerung der Satzanlautsform brfür mrbei ai. brávīmi hängt jedenfalls mit dem besonders häufigen Gebrauch von Formen wie brūhí, brávīmi und ähnlichen im Satzanlaut (auch in eingeschobenen Sätzen, vgl. unser "sag' ich", "sagt er") zusammen. ' However, Bartholomae's argumentation is based on a misunderstanding: the frequent usage of the phrase 'I say' and the like, as a kind of interjection and/or conjunction may indeed influence the phonetic shape of the word, but typically this means shortening the word via vowel-or syllable loss (see Bartholomae's quoted German phrases), weakening the vowels, assimilation and/ or simplification of consonant clusters, and none of these happens here. What does happen here is the substitution of one well-formed initial consonant cluster by another well-formed initial consonant cluster without any shortening, weakening, simplification, assimilation or even dissimilation. This cannot be motivated by the interjection-and conjunction-like usage of the verb 'to say' . Furthermore, no other special usage of the verbs 'to say, to speak, to tell' is known that could explain any phonological irregularity. 6 Mayrhofer's second reference (Turner 1937: 13-14) is false, since Turner did not attribute this change to any special circumstances surrounding the verb 'to say' . Instead, he suggested that this is an anticipated sound change, more precisely, it is an anticipation of the Khowar sound law mr-> br-. However, there is no reason for any Khowarism in Vedic texts.
The other explanations are not convincing either. In view of Schmidt 1875: 283-284, *mr-is a 'schwirige lautverbindung' and thus the change *mr > br is 'lautphysiologisch wol begründet' . However, since Sanskrit does have many words with this allegedly complicated initial cluster (mrityáti 'decays' , mrócati 'goes down' , etc.), this cannot be the reason. In general, since both clusters are well-formed in initial position in Sanskrit, no substitution rule can account for this change.
Osthoff 1881: 55 assumes that *mr-regularly became brin Vedic and thus all cases that have mrare analogical, resulting from postvocalic allomorphs where *-mr-was preserved. However, 6 ' Anonymous reviewer B' suggests as an explanation that mlhas an extremely low type and token frequency in Vedic and that 'this low frequency is best explained by the fact that mlwas highly marked in Vedic (which strongly prefers complex onsets with higher sonority contrasts). ' Since blis 'completely inconspicuous' , the onset mlis 'clearly far worse' . However, the explanation of the low frequency of mlis different: as it is well-known, PIE *l became regularly r in Vedic, all known exceptions are irregular and usually attributed to another dialect (cf. above). In other words, PIE *ml-led to *mr-in Vedic and thus, there was no competition if the onset with blor the onset with mlis worse in this word. Note that in terms of sound laws the explanation provided by the reviewer is ad hoc. this is refuted by the above quoted cases with mr-< *ml-without any possible postvocalic allomorphs where *-mr-could have been preserved (cf. also Kobayashi 2004: 93). 7 Wackernagel 1896: 182 suggested that brmight have come into being only under special phonological conditions (for instance after a pause or after words ending in consonants), i.e. /b/ is the result of paradigmatic levelling and opts for a position after a vowel. Unfortunately, there is no evidence for Wackernagel's proposal. 8 Thus, since all suggestions to explain the change *ml-> brare arbitrary, and this change cannot be independently accounted for (no sound law, no substitution, and no analogy can explain it), the traditional etymology cannot be upheld without ad hoc assumptions.

TOWARDS A NEW ETYMOLOGY
It is worth noting at this juncture that not only the Vedic verb, but also Tocharian B pälwā-'to complain, bewail one's fate' is irregular from phonological point of view and almost exactly in the same way due to the assumed change '*ml-> pl-' (cf. Adams 2013: 407 quoting mlutk-'±crush' from PIE *mleu̯ -T-, cf. Avestan mruta-'crushed, weak' and also Tocharian B mlut-'pluck' , Adams 2013: 516). While admitting this, Schmidt 1982: 365 n. 21 cautiously explained this change with the 'Sonderbedingungen eines Verbums für "sagen"' , without explaining what these special circumstances could and should be, referring only to Mayrhofer's above quoted dictionary (followed by LIV 2 : 446-447 with a question mark) and as we could see, this assumption does not solve the problem. 9 The Tocharian form regularly continues *p/b/b h luH-, which shows a remarkable coincidence with the expected proto-form of brávīti, i.e. *bl/reu̯ H-. This observation points to the possibility that we can regularly explain all presumably related forms if we assume two different PIE verbal roots: 1) PIE *bleu̯ h 2/3 -, providing a phonologically regular basis for both Vedic brav i and Tocharian B pälwā-, where the history of the Tocharian verb even identifies the laryngeal as *h 2/3 : pälwā-<*pluwā-<*bluh 2/3 -(Malzahn 2010: 720, cf. also LIV 2 : 447 n. 3, but the restriction to *h 2 presented here is problematic, see the discussion in Hackstein 1995: 7 Theoretically, one could even turn the direction of the change the other way round and assume PIE *br-/*bl-behind these words (on the existence of PIE */b/ see below), *bl-could even explain the Vedic and Tocharian forms. However, it cannot explain the Avestan and Slavic words, since Proto-Iranian and Proto-Slavic *br-and *bl-do not change into mr-/mlin Avestan and in the Slavic languages. 8 ' Anonymous reviewer A' cautiously suggested that Vedic b might have resulted from an assimilation due to the 'labial ū/*uH of the following syllable in the zero grade root' . The problem with this idea is that it is not only ad hoc but also phonetically unmotivated (*m is labial, after all) and, as the reviewer also admits, it is 'not evident' that the suggested assimilation would have worked across -l-. OCS boļii 'bigger, better' . 13 Rarity, however, does not mean lack. In other words, if our proposal is correct, we can add yet another reconstruction with PIE /b/. 14 Finally, the Khotanese verb parī-'to order, to deign' must be mentioned. It has been connected with Avestan mraoalready by Konow 1932: 167 (as *pa(ti)-mrau-) and the question is still unsettled (cf. Emmerick 1968: 73;Bailey 1979: 219 ['hardly better' (than the alternative etymology *pa(ti)-rau-d-'to make sounds'), without arguments]; also cited by Cheung 2007: 275 with a question mark, without refs.). Neither *bleu̯ h 2 -nor *mleu̯ Hcan lead directly to the Khotanese verb, since Proto-Iranian *br-(< PIE *b (h) l/r-) is reflected as <br-> (cf. brātar-'brother') and although the fate of *mrV-is not clear, there is no evidence for a Khotanese reflex <par-> (for Khotanese historical phonology see Emmerick 1989: 210-216). The suggested combination with the preverb *pati-does not help either: in case of *pa(ti)-mrauH-the *m would not disappear in this position, since intervocalic *-mr-became -mbr-in Khotanese (haṃbruīttä 'it heals ' < *hamraudati, cf. Av. raoδa-'to grow' , Emmerick 1989: 215). The other option, *pa(ti)-brauH-leads regularly to *pa(ti)-βrV° > *pawrV° > †paurV° ~ †porV° (cf. ora-'sky' < *abra-[see Av. aβra-], hauda / hoda 'seven' [without morpheme boundary]; kṣundau 'husband (acc. sing. masc.)' < kṣundaku [with morpheme boundary], Emmerick 1989: 211, 214, 215), thus again not to the attested form. All in all, this verb cannot be connected to either of the reconstructed roots.

CONCLUSIONS
The standard etymology of Vedic brav i 'to say, to speak, to tell' , a connection with Avestan mrao-'dto' and a derivation from PIE *mleu̯ h 2 -'to speak' is based on an irregular consonant change that cannot be independently motivated and explained. Alternatively, two different PIE verbal roots can be assumed, *melH-u-→ mleu̯ H-'to say, to speak' and *bleu̯ h 2/3 -'to speak or to call' , that provide phonologically and semantically regular bases for the words involved, i.e. Avestan mrao-'to say, to speak' and Proto-Slavic *mъlva 'speech' on the one hand and Vedic brav i 'to say, to speak, to tell' and Tocharian B pälwā-'to complain, to bewail one's fate' , on the other. 13 Mayrhofer 1986: 99-100;Kapović 2017: 16;Byrd 2018Byrd : 2061Byrd -2063Weiss 2020: 37;Fritz andMeier-Brügger 2021: 138. Clackson's statement (2007: 46), 'there are no secure reconstructions which have an initial *b-' is erroneous. ' Anonymous reviewer A' wants to see non-initial cases, too: see, e.g., *pib(h 3 )e/o-'to drink' , *sei ̯ b-'fließen lassen' (LIV 2 : 462-463, 521). 14 One may also object the widespread view that there is a suppletive relationship between Ved. brav i und vac-, mirrored in Avestan mraoand vac-, which could point to the identity of brav i and mrao-. However, even assuming the suppletion and the identity of the Avestan and Vedic suppletions, it does not necessarily mean that the members must be the same etymologically speaking (note the famous example of Latin ferō, ferre, tulī, lātum). Moreover, Casaretto 2006 demonstrated that the suppletive relationship in Vedic postdates RV and AV, and that the Iranian and the Vedic suppletions are different, in other words, the two phenomena have nothing to do with each other and thus have no relevance regarding the etymology of these verbs ('anonymous reviewer A' pointed out that the Avestan and Vedic verbs also have some shared collocations that 'quite strongly argue for a common (…) poetic tradition' . This is correct, but it still does not prove that the suppletive forms must be etymologically identical, as per above). Further morphological parallels, e.g. the t-less 3 rd sg. pres. mid. ending, cannot be used either as a counter-argument, since they are not restricted to these verbs (for t-less ending see e.g. Vedic duhé 'to milk' , vidé 'to find' , śáye 'to lie' , etc.). ABBREVIATIONS EIEC = Mallory, James P. and Douglas Q. Adams (eds.)