Traditionally, the Shatapatha is considered in its mss. to have underscore ("_") markings for udaatta. But, as noted, among others, by Hoffman [1] the marking denotes anudaatta. Much of the confusion arises because the mss. show markings of only one kind, that of the understroke. The challenge has been to derive the traditional threefold accentuation scheme of udaatta, anudaatta, svarita, from this rarified notation. Caland, Minard, and Weber (cf. below) hold the former position (i.e. markings are, in various explanatory justifications, udaatta) and are corrected, somewhat, by Hoffman, whom I quote in full:

"The accent mark used in the ShB is, as it is well known, a horizontal stroke below the letter. Leumann and Wackernagel believe this to be the designation of the main accent, in every case. But then, this results in very curious deviations from the Vedic norm of accentuation, both for single words as well as for sentences (for example: manúSya- instead of manuSyá-, or yátrá kva ca for yátra kvà ca ). In order to arrive at the normal accentuation, Weber (ShB p. xii sq.), Caland (ShBK introd. p. 7 sqq.), and Minard (Trois ènigm. p. 7 sqq.) therefore suppose, that accent mark does not only indicate the main tone (udaatta) of the same syllable but also svarita of the following one. This results in the only important peculiarity of ShB accent: when accented and unaccented vowels collide (as for instance in Rigvedic abhinihita sandhi, too), the result is svarita (á + a> aà [i.e. svarita- long a], á + í> è, etc.), except when the prepositions aa and prá (prá + ihi>préhi ) and when composition (citrá + uti> citróti ) are involved. Although the accent system supposed by Weber, Caland, and Minard, in practical application allows to arrive at correct results (manuSyá-, yatra kvà ca ), it is hardly probable that one accent mark should have had two different functions. If it is taken as granted that ShB had another peculiarity, viz. that in all instances an udaatta syllable is followed by a svarita- i.e. also in the sequence udaatta : svarita : udaatta (instead of RV udaatta : anudaatta : udaatta), then the accent mark only announces svarita of the following syllable (thus, apparently already Weber, VS specimen II 6), and it does not designate udaatta itself. Thus, it becomes explainable, why in an uninterrupted sequence of udaattas apparently only the last one is marked: it is only the last one that is followed by svarita. On the other hand, it is in concurrence with the present system of notation, that in a sequence of svarita's not only the last one, but each one is announced by the accent mark. in transliteration, one therefore would have to write yatrà kvà ca, which however, must be interpreted and transcribed as yátra kvà ca. Only this accentuation, which appart from very few exemptions, coincides with the rest of the Vedic norm, is a phonematic one. Even if one should believe in accent shifts or in a special nature of the tone preceding a svarita, these phenomena would be phonologically irrelevant, because an accentuation yátrá kva ca (according to Leumann-Wackernagel) could only exist, just because the phonematic word accents had been yátra and kvà in the ShB too. Thus, the phonematic accent of a word can, in almost all cases, be ascertained by a comparison of its occurrences in different accentuations according to its position /among other words/ [sic] in the ShB itself."

As might be expected, such a lengthy explanation could only leave room for further clarification--in particular as regards svarita. We turn now to a discussion by George Cardona of ShB accent, according to the bhaaSika accentuation system. Fundamental to understanding the accentuation of the ShB begins with the observation that this later accent system omitted the combined udaatta/anudaatta tone of svarita.[2]

Resolving the issue is a careful outline of the BhaaSikasuutra 1.13 presented in the cited article. The syllables wherein certain sounds conjoin are called bhaaSika and are thus udaatta. For instance, udaatta vowels followed by anudaatta vowels are anudaatta (BhaaSikasuutra 1.5: udaattaanudaattau bhaaSikas tatsandhiH).[3] VS 25.24 . . . váruno aryamaáyuúr (grave "'" here marking udaatta) becomes marked as va-runorya-maayur in ShB 13.5.1.18, (anudaatta marked with "-"):

Thus in this e-text by Lehmann, based upon Weber, we have anudaatta marked in 3 ways: with grave "'" over most vowels, with "!" when with R (for internet/font reasons), and sometimes by ampersand "&" as above. Following Weber's markings, then, the ampersand--per BhaaSikasuutra 1.5ff--can denote an udaatta-anudaatta conjoint as with bhaasaá + antarikSam in ShB 9.2.3.34, herein marked as bhaasaa&ntárikSam-- "bhaasaa&" marks the conjoined udaatta 'aá' and anudaatta "a," marked in VS 17.22 as bhaasaántárikSam. Note, the second syllable in antarikSam is udaatta in VS, and anudaatta in ShB.[4]It frequently becomes the case, then, that some prior presence of udaatta is usually indicated by the ampersand. The historical factors involved in this change have been detailed at some length along linguistic and historical lines by both Cardona and Witzel. In Tracing the Vedic Dialects, Witzel cites the abundant variance and subsequent historic and geographic implications in pronunciation dialects throughout the Vedic period. Relevant here, for instance, is the distinction between Eastern and Western dialectical pronunciation, (often a cause for scorn as in the JB viz. Eastern accents). [5] In "The BhaaSika System," Cardona notes that these changes were not unanticipated or unknown in the aSTAdhyaayii and the praatishaakhyas.[6]

In addition, then, the BhaaSikasuutra indicates that the anudaatta preceding an udaatta is, of course, a low tone. When this low tone precedes another low tone, a super-low tone is indicated.[7] The historical and linguistic issues go beyond the immediate purposes here with regard to deciphering the basic notation as found in the e-text, but the articles cited--and their related bibliographies--are essential reading to precede any solid ShB inquiry.

In this e-text of the ShBM, then, the resulting "key" to accent follow's Weber's markings by placing an ampersand "&" in the conjoining of certain udaatta and anudaata syllables. Thus in the following passage (click on it to go to the passage):

Thus the accent for mahato&kthéna is a conjoined syllable with anudaatta on -é-.[8] NOTE: here also is case of where, for reasons of ITRANS representation of vocalic R as an upper case letter, the grave stroke does not compose and, accordingly, the anudaatta is denoted by "!" Additionally, then note the following example with the ampersand:

In the article cited above re. Hoffman, Witzel discusses in some detail issues of variance in tonal pronunciation according to a given shaakha.[9] In addition there is the tendency in modern pronunciation to render the svarita as the high tone.[10] These variances in tone are regional as well.[11]

Again, the text herein--according to these rules--should be interpreted as follows re. accent. anudaatta is marked with an accent grave " ' ", with the exeptions--as above--wherein udaatta and anudaatta have conjoined, or other variance in the BhaaSika rules. The results of Hoffman, et al may yield frequently accurate results, but they do not account for the ampersand and, fundamentally, they assume the presence of a non-existent accent mark: the svarita..

[1]Review of Wackernagel-Delbrunner, Altindische grammatik (1896-1954): "Aufsaeze sur Indologie und Iranistik," (Wiesbaden, 1975), cited and translated by Witzel, "On some unknown systems of marking the Vedic accents," Vishveshvaranand Indological Journal, 12, 1974, p. 475, n. 17.

[2]George Cardona, "The BhaaSika Accentuation System," Studien zur Indologie und Iranistik, Band 18, 1993, p. 1.

[3] ibid, p. 2.

[4] ibid, pp. 2f.

[5]Summary in e-mail, Michael Witzel [witzel@husc3.harvard.edu], "Re: Last Posting," personal e-mail message, November 9, 1996.

[6] Cardona, pp. 13f.

[7] ibid, p. 16.

[8]The explanation of this, and general direction to Hoffman and other attendant concerns on accent, come thanks to Michael Witzel, [witzel@husc3.harvard.edu], "Re: Sorry, Eudora : SB accent," personal e-mail message, 1 September, 1996.

[9]Witzel, 1974, pp. 496ff.

[10]ibid, p. 474.

[11]Michael Witzel, "Tracing the Vedic Dialects."