The Yajnadevam paper is a recent research publication that presents a novel and pioneer approach to deciphering the Sindhu-Sarasvati Valley Script (SSVS), which has famously evaded the grasp of multitudes of previous researchers. The vast variety of signage, rarity of long inscriptions, and the lack of a Rosetta Stone—a multiscript inscription containing the same information to facilitate referral—have made this field nigh unassailable for many a decade. The Yajnadevam paper approaches the question from a cryptanalytic perspective, taking an information theory model to tackle an epigraphic problem. The paper seems to have attracted considerable attention on social media and academic discourse alike for primarily three reasons:
1. An acceptance of Sanskrit as the language of the Sindhu-Sarasvati Valley Script,
2. A clear deviation from subjective readings espoused by previous attempts, and
3. A purported scheme for reading the script like any other Sanskrit representation.
Regarding the first point of interest, those who are looking for new evidence in the paper for Sanskrit being the script’s language may be slightly disappointed; indeed, this idea forms the premise and not the conclusion of the Yajnadevam paper. The author provides no new evidence towards identifying Sanskrit in the SSVS, but cites a solid bank of existing research literature to support this fact. Specifically, the work of Michel Danino in disproving the equid argument in support of a non-Vedic Harappan identity is referred to copiously, as is the epigraphic evidence of SSVS signs appearing in early Brahmi inscriptions at Vaishali and Keezhadi – both in Sanskrit contexts. From a direct epigraphic approach, too, the author draws upon the work of Steven Bonta: A Partial Decipherment of the Indus Valley Script: Proposed Phonetic and Logographic Values for Selected Indus Signs and Readings of Indus Texts (2024), where the samāsa structure inherent in SSVS signs is sufficiently exposited. Countering the argument of identifying the script with some form of hypothesised Proto-Dravidian language (only 800 words of which have been tentatively reconstructed to date), Bonta remarks that the frequent recurrence of compound graphical units cannot be interpreted as affixes (pratyayas) of the Proto-Dravidian language since the compounded units have also been documented with independent appearances – an impossibility for purely grammatical devices like affixes; the paper concisely documents these findings as proof for its Sanskrit approach. The dictionary method also yields poor results for Tamil (a stand-in for the purely fictional Proto-Dravidian language), running into dead ends with the relatively frequent double- and triple-repetitions of the same sign in SSVS (a phenomenon tentatively explained by the Sanskrit liṭ conjugation, although this deserves more thought). As a final point, the paper presents the fact that the Proto-Dravidian language contains no words independent of Sanskrit or its Prakrit dialects for such basic Sindhu-Sarasvati technologies as bricks, barley, and cities; it is impossible for such basic words to simply drop out of a language’s lexicon. The Yajnadevam paper cites all these facts as the first step of its methodology—which I shall summarise shortly—and rejects the Proto-Dravidian claim, accepting the Sanskrit case. Although the paper sheds no new light on this topic, its documentation of the major arguments for Sanskrit is truly commendable, and interested readers will surely find that particular section to be a useful guide to pursue the actual research articles which exposit the relevant facts.
As for the second point of interest regarding the paper’s novel objective approach, this is the real strength of the Yajnadevam paper, and we must thus summarise the methodology in as much detail as possible. In essence, the author pursues a cryptanalytic approach by treating the whole SSVS signary as a cryptogram and using a five-step process created first by Claude Shannon in his A Mathematical Theory of Cryptography. The first step is to determine the language. We have already examined the paper’s determination of Sanskrit as the script’s language, and can therefore move on to the second step, which is to compile a cryptogram from the available signary. This is a simple exercise of creating a database that is readily available. The third step is the first exercise in decipherment where a single sign is assigned a certain value. The paper chooses the ‘jar’ sign, which is assigned the value /an/. The reasoning behind this assignment is that the sign appears with high frequency at the end of strings, marking probably an anusvāra or visarga. The visarga case being eliminated due to the ‘jar’ sign’s presence in the middle of strings too (as an anunāsika stand-in), the ‘jar’ is deciphered to be /an/. What follows is a bootstrapping process – the signary is searched for instances where an unknown sign appears in a string with the only other member being a known sign (in this case the ‘jar’ sign). The dictionary of the language—Sanskrit in this case—is consulted for all words of that particular structure, and a set of possible words is created for that particular pattern – say jar-x-jar where x is the unknown. The process is repeated until all observed combinations of jar and x are collected, and the dictionary process followed to compile possible word-sets for each combination. The set intersection of the valid assignments of x is thus narrowed down until it is a unit – a single assignment for x. This x is then added to the set of known signs, and the process is repeated again. The five-step process is thus:
1. Determine a language [Sanskrit]
2. Compile all signs into a cryptogram
3. Decipher a single sign [‘jar’, value /an/]
4. Use known sign-set to narrow down possible assignments to unknown signs, add the new sign to the known sign set after assignment set becomes a unit
a. This algorithm has been encoded using a regular expression (RegEx) scheme of computer science
5. Repeat until all signs are given consistent assignments.
The novelty of this approach is specifically its mathematical and objective nature. As the author states, the field of SSVS interpretation is rife with subjective readings, superimposing assumptions about context onto signs that are themselves assumed to be logograms. Though this may yield fast results depending on the imaginative power of the interpreter, it does not make for sound decipherment. The author gives the example of two signs—‘overlapping rings’ and ‘triple drop’—which were interpreted by Parpola as ‘bangles’ and ‘hearth’ respectively without much concrete reasoning, culminating in an even more inscrutable combined reading of ‘pregnancy bangles.’ The Yajnadevam paper avoids such confusions and unfalsifiable readings. The unique advantage of the second step of Yajnadevam’s process is that the signs are analytically divorced from their contexts, enabling independent readings and avoiding extrapolation. Shannon’s algorithm is also language-independent; no assumptions about language will guide the decryption method, and the language is referred to only in the dictionary process in the last leg of the algorithm. This is vastly different from the ē rēbus approach of previous researchers. The flaws of that approach are self-evident; the author himself refers to Iravatham Mahadevan’s inexplicable reading of the indented-rhombus sign followed by two standing tailed birds as ‘rooster’s city,’ trying to tie the so-called Dravidian civilisation to the southern city of Uraiyūr (உறையூர்). Such assumptions have no place in Shannon’s argument, making the method that much more sound. This key perspective shift towards utilising well-established decryption schemes of information theory is cause enough to credibly consider the Yajnadevam paper in the leading edge of Sindhu-Sarasvati Valley Script decipherment research.
As for the third point of interest, it seems unlikely that the Yajnadevam scheme can be used to simply read Sindhu-Sarasvati inscriptions like Devanāgarī, Gupta, or Sharada. As the author himself has already stated, the early and still developing nature of the script means that a lot of the representational features that make a script effective in communicating a language maybe absent, partially formed, or even locally eliminated for extraneous purposes. Combined with the predominantly short nature of SSV inscriptions, this makes for highly volatile readings which—for the most part I daresay—make no sense. Readers eager to acquire zoomed pictures of SSV texts and read off Sanskrit phrases with Yajnadevam’s syllabary in one hand should rein their fantasies in. The strength of the paper is not that it is a panacea to the SSVS problem; rather, it is a new path, a new method, of approaching the problem, which is more scientific and more logically consistent than previous models.
That being said, we must still voice some concerns and remark on some possible improvements. Primarily, the sheer variety of glyphs listed under each Sanskrit letter in the results section of the paper cannot but raise scepticism and suspicions about coercive classification. Perhaps, in the long time period through which Sindhu-Sarasvati seals are spread, there are some patterns of lettering flux; creating a table of correlations between every glyph with every other glyph may assist in dividing the 417 (as stated in the paper) signs into syllabaries of different points in time. This would retain the assignments made by the decryption algorithm while adding a temporal dimension to better historically understand each sign. Second, there seem to be gaping holes—pits even—in the translation and interpretation of Sanskrit expressions taken from the seal inscriptions and even more so in their connection to Vedic segments. The author seems to be given to convenient translations wherever they look like they fit; for instance, tana (तन) is translated as ‘child’ or ‘offspring’ in M-459A but is translated as ‘roarer’ from tan in M-359; the reason for this choice is not elaborated other than the fact that it fits better with some Vedic passage. This is not very scientific and, indeed, reduces the paper’s credibility. It makes very little sense to first utilise a commendable, unyielding, and sterile scientific approach to reading the SSV signs and then becoming completely subjective in interpreting the decryption in an attempt to refer to the Veda as much as possible. Thirdly, the author has provided no proof of concept. Testing the algorithm to assign phonetic values to characters from a finite non-SSV inscription set, drawn perhaps from the copious corpus of Epigraphia Indica, would have sufficiently demonstrated the efficacy of the method. With no such cross-referrable results, there is a very real scope for the findings to be found entirely dubitable. Fourthly, Shannon’s method performs best on a large corpus of text. The Sindhu-Sarasvati signs predominantly appear as short inscriptions on small seals—a fact admitted by the author—and even the longer inscriptions are no more than a small sentence in length. This can cause serious issues with the measurements of representational entropy and redundancy which are so crucial to a successful application of Shannon’s decryption. Especially given the author’s own declaration of Sanskrit redundancy at ~0.7, the small size of inscriptions may not properly reflect this, acting against the paper’s overall point. Of course, taking up a deciphered epigraphic dataset with high similitude to the Sindhu-Sarasvati signary and testing the algorithm on it, as we have stated before, would provide a far more accurate picture about the credibility of these findings. With these four main lacunae being fulfilled, I have no doubt that the place of the Yajnadevam paper in the field of SSVS study will be concretely understood.
In all, the Yajnadevam paper is by no means a lightweight. It demonstrates quite an intellectual feat in applying concrete information theory techniques to a sorely misunderstood and subjectivity-ridden field. Its staunch support with suitable citations to the case for Sanskrit is a much-needed voice. Its approach is novel and seems to be on the right path, applying techniques used for Minoan Linear-B to the SSV script, but is still in real need of refinement and completion. The premature release and publication, however, may adversely affect the expansion of this new study, which would be detrimental to the cause of earnest research. With some further work in the areas defined in the previous paragraph, the Yajnadevam paper has the potential to afford us a clearer-than-ever understanding of the writings of the Sindhu-Sarasvati Valley. The paper is an earnest and valuable addition to scientific literature that strives to dispel dubious befuddlement regarding the hoary past of the Indian people, and merits close examination by readers and researchers alike. Competent and earnest students of history, culture, archaeology, epigraphy, as well as information theory and computer science will find parts of this paper highly interesting from the perspectives of their own fields, and—I hope—will be motivated to engage in their own considerations of the problem and take forward this study of the Sindhu-Sarasvati Valley script that is so imperative to setting right the record of the history of Sanātana-dharma and Vedic culture.
Author's Note of Acknowledgemet: My sincere thanks to Dr. Pratap Simha for his support in brainstorming and critical review of my analysis; his insight, especially regarding the crucial nature of cross-reference with existing inscriptional literature, added new and vital depth to the article. I am greatly in his debt for his help in this regard.