In his 1918 article, “Le Royaume de Çrivijaya,” George Cœdès examined information from three types of sources—inscriptions, Arabic texts, and Chinese texts—to make the argument that there had historically existed a polity based at Palembang on the island of Sumatra called “Srivijaya.”
As we have seen, his understanding of Chinese texts was deeply flawed. This is because the scholars who had worked with Chinese texts prior to 1918 had introduced numerous errors.
There was a similar problem with the information in Arabic texts, and while my focus is on examining the issue of how Chinese texts have been misunderstood and misused in the scholarship on Srivijaya, it is important to get a brief sense of the issues with Arabic texts as well.
Arabic Confusion
In his 1918 article, Cœdès states that “It is known that, on the basis of the Chinese forms (Fo-che [Foshi], Che-li-fo-che [Shilifoshi], Fo-ts’i [Foqi], San-fo-ts’i [Sanfoqi]) and the Arabic form (Sribuza), the name of the kingdom of Palembang has been restored as Śrībhoja.” (23)
His citation for this statement is “Méthode, no. 297” (23, fn. 1). That is a reference to Stanislas Julien’s 1861 work, Méthode pour déchiffrer et transcrire les noms sanscrits qui se rencontrent dans les livres chinois, in which Julien documented how Sanskrit terms were rendered in Chinese.
The entry number that Cœdès was referring to is No. 299, not 297, and Julien gave an example there of how in the name Shilifoshi, the character, fo 佛, was used to transcribe the Sanskrit “bhô” as follows: “Fo, pour bhô, dans Çrîbhôdja. Nan-haï-k’i-koueï-neï-fa-tch’ouen, liv. I, fol. 3.” (103)
Julien made no reference to Foshi, Foqi, Sanfoqi or Sribuza, only to Shilifoshi. Cœdès’s statement is therefore inaccurate. What Cœdès did here was to give coherence to a topic that was far from coherent, namely the question of what certain Arabic placenames referred to, and which Chinese names they (possibly) corresponded to.
In 1876, Willem P. Groeneveldt argued that the Chinese name, Sanfoqi, was referred to by Arab travelers as “Sarbaza,” and he cited for that term the 1845 work, Relation des voyages faits par les Arabes et les Persans dans l’Inde et à la Chine dans le IX siècle, traduite par Reinaud, p. 93 (187).
In 1883, a translation of the Arabic work, Kitāb ʿAjāʾib al-Hind, a collection of tales, travel reports, and maritime anecdotes from the tenth century was published as Livre des Merveilles de l’Inde, and it mentioned a place in Southeast Asia called “Serboza” that was a vassal of “Zabedj” (Zabag).
Samuel Beal has a note in that work where he states that “With respect to the island called Sarbaza dependent on the Mahârâja of Zabej, I will express no opinion, as I am not an Arabic scholar. But from a kind communication made to me by Col. Yule, I should think that Sarbaza and S’ribhôja were identical.” (253)
So, in these two comments we see Groeneveldt equate Sanfoqi with Sarbara/Serboza and Beal equate the earlier Shilifoshi with Sarbaza/Serboza.
Then in 1894, Édouard Chavannes suggested that Shilifoshi might have been Zabag (37, fn. 3). Further, he also believed that Shilifoshi was probably in southern Sumatra because it was close to Moluoyu and Chavannes followed the comment in Henry Yule’s translation of Marco Polo’s account that mentioned a “Malayo” on Sumatra that sixteenth-century Portuguese admiral Afonso de Albuquerque had recorded, etc.
In 1904, Paul Pelliot equated Shilifoshi with Serboza (Sarbaza). He argued further that it was a vassal of Zabedj (Zabag), which he also argued was Yavadvïpa, or the island of Java, and that Serboza was located in southern Sumatra near Jambi or Palembang. (338-39)
Finally, in 1909, Colonial G. E. Gerini argued that Zabag was a kingdom on Sumatra (557, 580) and that Serboza/Sarbaza was the same as the Chinese Xianliu, Sanfoqi, Shilifoshi, and Jiugang (from which, through certain romanizations and misunderstandings, he got “Old Lagoon” or Thalāj), and that all of these terms referred to Palembang. (566-67)
Near the end of this article, Cœdès cites Chavannes and Gerini as possibly correct that his proposed kingdom of Srivijaya at Palembang was known to Arab geographers as “Zabag.” (26) As such, in his 1918 article, Cœdès presented conflicting information about what the Arabic names, Serboza/Sarbaza and Zabedj/Zabag, referred to.
When we combine that with all of the flawed assumptions about the placenames in Chinese sources that Cœdès relied on, we can see that his article contains numerous unresolved issues that needed to be addressed.
Śailendra Scholarship
In the first few decades following the publication his article in 1918, the main scholars who responded were either scholars who worked on Javanese and Malay inscriptions and languages (J. Ph. Vogel, N. J. Krom, W. F. Stutterheim, J. L. Moens, Ph. S. van Ronkel), or Sanskrit texts (R. C. Majumdar, K. A. Nilakanta Sastri).
One of the inscriptions that Cœdès examined, the Ligor inscription, mentioned a king from the “Śailendra” line or dynasty. Scholars had already discovered that name in an inscription on Java, and as such, immediately following the publication of Cœdès’s article, some scholars (Vogel and Krom) started to incorporate the information about the Ligor inscription and what Cœdès had written about Srivijaya into their scholarship.
Ultimately, this led to a debate that continues to the present about what exactly the Śailendra was, and where it was from, and how it was related to Srivijaya.
In general, these scholars accepted whatever Cœdès and others claimed was in the Arabic and Chinese sources. Further, in 1922, Gabriel Ferrand collected together many of the texts related to this topic and published them in an article in the Journal Asiatique entitled “L’empire sumatranais de Çrivijaya.” This study was also published as a book. For Chinese sources that had not yet been translated into French, Ferrand translated from English translations, like those of Groeneveldt.
Like Gerini, Ferrand equated Shilifoshi, Sanfoqi, Zābag, and what he wrote as “Sribuza.” He envisioned these names as indicating a Sumatran empire based at Palembang. (110-111)
One exception to this general trend, was R. C. Majumdar, who in a 1933 article, utterly demolished much of the Arabic and Chinese textual scaffolding that had been used to support the “Srivijaya is a kingdom at Palembang that was referred to in Chinese sources as (Shili)Foshi and Sanfoqi” narrative.
That said, Majumdar’s main thesis was still about the Śailendra. In particular, he proposed that the Śailendra and Srivijaya were originally separate entities, and that the Śailendra originated in India and later established their primary seat of power on the Malay Peninsula, an area which Srivijaya had previously extended its authority to. Nonetheless, in making this argument he brought up many points about how the information in Arabic and Chinese texts does not support the idea that Srivijaya was located at Palembang.
Majumdar’s Critique
Majumdar provides a clear and concise summary of the scholarship from the time of Cœdès’s 1918 article to 1933 that is worth citing in full here:
M. Cœdès quite naturally deduced from the Ligor inscription that the authority of Śrīvijaya had, by the end of the eighth century A.D., extended as far as the northern part of the Malay Peninsula. He further advanced the hypothesis that the king of the Sailendra dynasty mentioned on face B of the Ligor inscription was identical with the king of Śrīvijaya referred to on face A of the same stele.
M. Cœdès therefore logically concluded that there existed a Sailendra empire having Palembang as its capital and comprising Sumatra and the Malay Peninsula. He also regarded as probable the hypothesis of Chavannes and Gerini, namely that this empire was identical with the one known to us from the Arabs under the name Zābag. Ferrand went a step further and declared that this identity was beyond doubt by identifying Zābag with San-fo-ts’i [Sanfoqi]. Thus, in the eyes of scholars, the Sailendra dynasty of Palembang became the ruling house of a powerful Pacific empire, whose eloquent descriptions have been left to us by so many Arab authors.
Further light on the power of the Sailendras was provided by M. Krom and M. Vogel. These two scholars, working independently of one another and almost simultaneously, brought out the important role played by the Sailendras in Java. The Kalasan and Kĕlurak inscriptions clearly attest Sailendra supremacy in Java in 778 and 782 A.D. On the basis of this fact, M. Krom showed the great influence that the Buddhist Sailendras must have exercised on the art and religion of Java. In short, he expressed the view that it was to the Sailendras that Java owed the introduction of Mahāyāna Buddhism and the construction of such famous monuments as Borobudur, Candi Mendut, and Candi Kalasan. Thus was born the hypothesis of a “Sumatran period” in the history of Java, with its far-reaching consequences for the political and cultural history of Java.
This theory, however, received a fatal blow when Stutterheim astonished the scholarly world by advancing the bold hypothesis that the Sailendra dynasty was of Javanese origin and subsequently conquered Śrīvijaya. Thus, instead of a Sumatran period in the history of Java, we should, according to this author, envisage a Javanese period in the history of Sumatra. (125-26)
Majumdar first addressed Cœdès’s view that the kings of Śrīvijaya were also of the Śailendra line or dynasty.
Cœdès based this idea on his reading of the Ligor inscription. That inscription has two faces. On one side terms that can be understood to mean the “king of Srivijaya” are written (Śrīvijayendrarāja, Śrīvijayeśvarabhūpati, and Śrīvijayanṛpati), while the other side mentions a Rājādhirāja and Prabhu, or lord, of the Śailendra line or dynasty (Śailendravaṃsa) as well as Mahārāja.
Majumdar argued that the two faces of the Ligor inscription did not refer to the same king, and concluded by stating two points, namely, that “1° that the locality in question formed part of the kingdom of Śrīvijaya in 775 A.D.; 2° that it was under the suzerainty of the kings of the Śailendra dynasty at a subsequent period. Nothing proves that the king of Śrīvijaya belonged to the Śailendra dynasty.” (127)
Majumdar noted further that inscriptions from the end of the eighth century mentioning the Śailendra had been found on the Malay Peninsula and Java, but not on Sumatra. He therefore concluded from this that the Śailendra had ruled over the Malay Peninsula and Java at that time.
He then went on to equate this polity with the place that is referred to in Arabic sources as Zābag, the king of which was referred to as the Maharaja.
Here he noted that “it is now generally accepted that the name Zābag, and the other forms of this name used by the Arab authors, designate the country called by the Chinese Che-li-fo-che [Shilifoshi], Fo-che [Foshi] , or San-fo-ts’i [Sanfoqi], that is, Śrīvijaya.” And he then engages in an extensive examination of the evidence, and in the process, undermines the idea that Srivijaya and Zabag were the same. He presents evidence instead to make the claim that Srivijaya (as Sribuza) was a dependency of Zabag. (134)
He then turned to the Chinese sources, and stated the following:
At present, it is generally taken as established that the kingdom called San-fo-ts’i [Sanfoqi] in Chinese works from the tenth to the fourteenth century A.D. is identical with Zābag or Zābaj. But this theory implies hypotheses that, in my view, have been accepted without sufficient proof. These hypotheses are as follows: 1° San-fo-ts’i [Sanfoqi], Che-li-fo-che [Shilifoshi], Zābag, and Sribuza would all be equivalents of Śrīvijaya; and 2° all these states would be identifiable with present-day Palembang.
As regards the first hypothesis, we have seen above that Zābag is distinct from Sribuza, and that alone would suffice to demonstrate the weakness of the theory in question. Che-li-fo-che [Shilifoshi] and Sribuza are both obvious equivalents of Śrīvijaya, but the same cannot be said of San-fo-ts’i [Sanfoqi] or Zābag. M. Aurousseau does indeed identify San-fo-ts’i [Sanfoqi] with Che-li-fo-che [Shilifoshi], but M. Ferrand thinks it impossible to see in San-fo-ts’i [Sanfoqi] an equivalent of Śrīvijaya, for philological reasons. Moreover, the Chinese themselves never suggested that Che-li-fo-che [Shilifoshi] was the equivalent of San-fo-ts’i [Sanfoqi]. The History of the Ming informs us that San-fo-ts’i [Sanfoqi] was formerly called Kan-da-li (or Kan-t’o-li [Gantuoli]). (134)
[. . . there is more, but it gets too detailed to duplicate here]
Thus, nothing proves the validity of the identifications: 1° of Śrīvijaya with San-fo-ts’i [Sanfoqi]; 2° of San-fo-ts’i [Sanfoqi] with Palembang. Needless to add, the definitive identification of Śrīvijaya with Palembang at present rests on no solid foundation. That this kingdom occupied the southeastern part of Sumatra is a fact attested by Jia Dan, but it is difficult to identify it more precisely. (136)
Majumdar could not read Chinese. He therefore had to trust in what scholars had already written. In fact, and as we have seen in this and the previous two series, there were many errors in what scholars had already written, like the idea that Jia Dan recorded an itinerary that went through the Strait of Malacca, and that from that itinerary, it was possible to locate a place called (Shili)Foshi on the island of Sumatra.
Nonetheless, even just by basing his ideas on what scholars had written, and in contrast to Cœdès, Majumdar could see that the information did not add up, and that the Chinese sources did not contain information that supported the idea that there was a polity at Palembang called “Srivijaya.”
Majumdar’s 1933 article is a devastating critique of the ideas in Cœdès’s 1918 “Le royaume de Çrīvijaya.” Majumdar was very polite, but he was firm, and ultimately, the entire “Srivijaya is a kingdom at Palembang that was referred to in Chinese sources as (Shili)Foshi and Sanfoqi” narrative was comprehensively destroyed in this article.
However, what Majumdar lacked was the linguistic ability in Chinese that would have enabled him to construct a viable alternative narrative. As such, his article pointed out fundamental flaws in Cœdès’s use of Chinese sources, but he himself had no means to read those sources more accurately to produce an alternative understanding of the past.
Cœdès’s Response
A year later, in 1934, Cœdès offered an initial response in the Journal of the Greater India Society. In this article, Cœdès agreed with Majumdar that the two faces of the Ligor Inscription refer to kings of different polities, however, much of the article was spent proposing an idea (that was not adopted by other scholars) that the Śailendra (or the Śailendra concept) were originally from Funan, in Cambodia, and that after the fall of Funan in the seventh century, a Javanese princely family with real or fictitious ties to the Funan kings adopted this title in the 8th century.
Nonetheless, he did address one of Majumdar’s critiques about the Chinese sources as well as Majumdar’s charge that there is no evidence for placing Srivijaya at Palembang. To quote, Cœdès stated that,
I gladly recognize that the identification of Shih-li-fo-She [Shilifoshi] (Śrīvijaya) with San-fo-ts’i [Sanfoqi] is not certain (in spite of the attempts of attempts of Aurousseau to justify the same), nor historically, as the presence of the two names is attested at different dates. . .
. . . I feel obliged to record my protest, when Dr. Majumdar writes, “The definite identification of Śrīvijaya with Palembang does not rest at all upon any solid basis. Nevertheless, it is from Palembang, that comes the inscription relating to the foundation of Śrīvijaya. [7]
[7] This hypothesis of Prof. Ph. J. van Ronkel (Acta Orientalia, II (1924), p. 21) which Dr. Krom (Hindoe-Javaansche Geschiedenis, p. 119) and myself (BEFEO, XXX, p. 37) did not believe to be maintainable, has been rendered possible by the interesting remarks of Dr. R. A. Kern. Enkele aanteekeningen op G. Coedès’ uitgave van de Maleische inschriften van Śrīvijaya. Bidr. 88 (1931), p. 508 (Cf. esp. p. 511).
When a scholar’s counter argument comes in the form of a discussion in a footnote about a study that he previously disagreed with. . . that’s when you know that the scholar has no counter argument.
And as for his admission that he couldn’t link Shilifoshi with Sanfoqi? Apparently, no subsequent scholar ever noticed this (and he mentioned it again, as we will see below), because, as far as I’ve seen, all of the scholarship on Srivijaya from the 1960s onward claims that Sanfoqi was also Srivijaya. . .
Cœdès addressed this issue again in 1936, in an article entitled “A propos d’une nouvelle théorie sur le site de Srivijaya,” Journal of the Malayan Branch of the Royal Asiatic Society 14.3 (December 1936): 1–9.
In this article, again, Cœdès had very little to say about Majumdar’s many critiques. He again agreed with Majumdar that the two faces of the Ligor Inscription refer to two different kings. However, much of the article was spent arguing against another article, one by Quaritch Wales, in which Wales contended that Srivijaya was based at Chaiya in southern Thailand.
Further, in this article Cœdès repeated many of the same flawed assumptions about the Chinese sources, such as that “Sanfoqi is located by the Chinese at Palembang,” and that Srivijaya corresponds “phonetically to the (Che-li-)fo-che of the Chinese.” (5)
Then, however, he made the following concluding remarks:
From the objective observation of these facts, it seems difficult not to conclude that the geographical terms Śrīvijaya, (Che-li-)fo-che [Shilifoshi], San-fo-ts’i [Sanfoqi], Zābag, kingdom of the Śailendras, kingdom of Palembang, simultaneously or successively designated a state whose cradle was situated at Palembang, and which, through its expansion northward as far as Kedah (Kaṭāha, Kaḍāram) or the Isthmus of Kra (Kalah), succeeded in commanding the straits. The equation among these terms therefore remains valid, subject to the following remarks, which constitute certain improvements to the thesis presented by me in 1918 and developed by G. Ferrand in 1922.
1° (Che-li-)fo-che [Shilifoshi] and San-fo-ts’i [Sanfoqi] are attested at different dates; the first name designates the Sumatran kingdom at its beginnings, and the second applies to the empire at the moment of its apogee. I recently wrote on this subject that the identification of Che-li-fo-che [Shilifoshi] with San-fo-ts’i [Sanfoqi] was certain neither phonetically nor historically. I would say here more precisely: historically, the two terms succeed one another; phonetically, they are not exactly superimposable.
Clearly Cœdès had no response to Majumdar’s critiques. He could only repeat the mantra that somehow the following places were the same: Śrīvijaya, Shilifoshi, Sanfoqi, Zābag, kingdom of the Śailendras, kingdom of Palembang.
And yet, he could not link Shilifoshi and Sanfoqi historically or phonetically. All Cœdès could say is that they were two terms that existed at different times in history and that they were “not exactly superimposable”!!
In conclusion, there are two key points that I think we can see here. By the end of the 1930s, 1) Cœdès’s “Srivijaya is a kingdom at Palembang that was referred to in Chinese sources as (Shili)Foshi and Sanfoqi” narrative had come under serious attack, and 2) Cœdès had no evidence to counter that attack.
This being the case, what the world of scholarship on Srivijaya desperately needed at that point was someone who could read the Chinese sources and who would follow up on the issues that Majumdar had pointed to.
Tragically, however, what that world of scholarship got at that point. . . was O. W. Wolters. That is the topic we will look at next.
For anyone interested, I’m uploading the articles discussed here. For articles in French, I’m also providing translations that I made of those articles with ChatGPT.
1933 Majumdar English translation
1934 Cœdès Journal of the Greater India Society article
1936 Cœdès Journal of the Malayan Branch of the Royal Asiatic Society article
