Correcting the Text of the Sarvadarśanasaṃgraha

Attempts have been made to correct the text of the Sarvadarśanasaṃgraha on the basis of the texts that its author used—and sometimes refers to by name—while composing his work. This procedure is promising in texts like the Sarvadarśanasaṃgraha, which makes abundant use of other works, and might in principle give results that are independent of, and prior to, the detailed study of its manuscripts. A closer investigation shows that this procedure is not without risks, and may occasionally give rise to unjustified “corrections”. The article shows that quotations in the Sarvadarśanasaṃgraha deviate from their source-texts in numerous cases. It further illustrates that the archetype underlying the manuscripts used for the available editions on occasion demonstrably differs from what must have been the text’s autograph. Other cases demonstrate that already the autograph sometimes deviated from its source-texts. The article concludes that careless “correcting” of the text may have serious consequences and can stand in the way of its correct interpretation.

There is no critical edition of the Sarvadarśanasam : graha. The New Catalogus Catalogorum (Dash 2015, p. 119) enumerates its surviving manuscripts, but it is not known whether any of these manuscripts were used in the existing editions. Of the existing editions, only three, as far as I can see, are based on manuscript evidence. All the other editions appear to be based on one or the other of these three editions.
These are the following: There is no guarantee that readings that we find in all these three editions are identical with what the author of the text committed to writing more than six centuries ago. Strictly speaking, we do not know whether readings shared by all surviving manuscripts are identical with what the author wrote. That is to say, there is no guarantee that the archetype of all surviving manuscripts is identical with the author's autograph; the same is true, a fortiori, for the "archetype" of the existing editions (which I will henceforth refer to as "the archetype of the Sarvadarśanasam : graha"). What we do know is that the manuscripts used for the editions represent two stages in the development of the text: Most manuscripts contain only 15 chapters, 1 whereas some have an additional chapter on Ś an . kara's philosophy that does not, with the exception of some transitional remarks, refer back to earlier chapters (Bronkhorst forthcoming). One way to obtain a text that is as reliable as possible would be to make a critical edition that takes all manuscript readings into account. 2 There is conceivably also another way, which does not replace the need for a critical edition but may in certain cases provide us with even better, i.e. more original, readings than a critical edition. The Sarvadarśanasam : graha makes extensive use of other texts, hereafter called its source-texts, from which it sometimes copies, with or without acknowledgment. In the best of circumstances, the identification of explicit or implicit source-texts may make it possible to correct the text of the Sarvadarśanasam : graha.
As so often, circumstances are not always perfect. In principle, we can be sure that the Sarvadarśanasam : graha quotes in cases where it explicitly mentions the source-text (or its author). There are numerous such instances, but many of those "literal" quotations deviate in minor or major ways from their source-texts. The following two examples will illustrate this.
Chapter 12-on Jaimini's philosophy-claims to quote the following passage from : 3 atra kusumāñjalāv udayanena jhaṭ iti pracurapravr : tteḥ prāmāṇyaniścayādhīnatvābhāvam āpādayatā praṇyagādi/ pravr : ttir hīcchām apeks : ate/ tatprācuryam : cecchāprācuryam/ icchā ces : ṭasādhanatājñānam/ tac ces : ṭajātīyatvalin . gānubhavam/ so 'pīndriyārthasam : nikars : am/ prāmāṇyagrahaṇam : tu na kvacid upayujyata iti/ Udayana has stated the following in his Kusumāñjali, while putting forward that much activity that takes place instantly does not depend on certain knowledge of authoritativeness: "For activity requires desire. And abundance of activity requires abundance of desire. And desire requires knowledge that something is the means to attain the desired goal. And that knowledge requires an experience of an inferential sign (lin . ga) that something is of the same kind as the desired goal. That experience, in its turn, requires contact (sam : nikars : a) between sense organ and object. Grasping authoritativeness, however, plays no role anywhere." The Kusumāñjali under verse 2.1 (p. 229) contains the passage that is here no doubt referred to: (yad api jhaṭ iti pracuratarasamarthapravr : ttyanyathānupapattyā svataḥ prāmāṇyam ucyate, tadapi nāsti/ anyathaivopapatteḥ/ jhaṭ iti pravr : ttir hi jhaṭ iti tatkāraṇopanipātam antareṇānupapadyamānā tam āks : ipet/ pracurapravr : ttir api svakāraṇaprācuryam /) icchā ca pravr : tteḥ kāraṇam/ tatkāraṇam apīs : ṭābhyupāyatājñānam/ tad api tajjātīyatvalin . gānubhavaprabhavam/ so 'pīndriyasannikars : ādijanmā/ na tu prāmāṇyagrahasya kvacid apy upayogaḥ/ It is impossible to believe that the author of the Sarvadarśanasam : graha believed that he quoted literally from the Kusumāñjali. And yet, he presents this as a quotation, thus confirming our conclusion that quotations in the Sarvadarśanasam : graha must be treated with caution, and do not in all cases justify "corrections" in the light of the source-texts. Before we proceed, it is necessary to take the following points into consideration: -It is not always obvious that the Sarvadarśanasam : graha quotes directly from the source-text. In certain cases it may quote through the intermediary of other texts. As already pointed out by de la Vallée Poussin (1902, p. 391), this seems particularly clear in the chapter on Buddhism, which appears to derive at least some of its Buddhist quotations from Vācaspati's Bhāmatī and other Brahmanical texts. It is conceivable, but hard to prove, that the same happened in other chapters. -We have no guarantee that the text of the Sarvadarśanasam : graha that we find in its editions (or even in its manuscripts) is identical with the text committed to writing by its author, i.e. with its autograph. We have no guarantee either that the existing editions (and indeed, the surviving manuscripts) of source-texts are in all details identical with their autographs. In comparing passages quoted in the Sarvadarśanasam : graha with their source-texts, we compare two uncertain readings. 4 -We have no guarantee that the author of the Sarvadarśanasam : graha intended in all cases to quote a passage from a source-text verbatim. This uncertainty is particularly pronounced in cases where the Sarvadarśanasam : graha does not name the source-text or its author.
-Even in cases where the author of the Sarvadarśanasam : graha intended to quote a passage verbatim, we do not know what reading he found in the manuscript(s) of the source-text available to him.
An example that illustrates these uncertainties occurs in the chapter on Pratyabhijñā (ch. 8). We read here (ed. Abhyankar,: tathopadis : ṭam : śivadr : s : ṭau paramagurubhir bhagavatsomānandanāthapādaiḥekavāram : pramāṇena śāstrād vā guruvākyataḥ/ jñāte śivatve sarvasthe pratipattyā dr : ḍhātmanā// karaṇena nāsti kr : tyam : kvāpi bhāvanayāpi vā/ jñāne suvarṇe karaṇam : bhāvanām : vā parityajet// iti/ 5 The venerable Somānanda, the supreme guru, has taught this in his Śivadr : s : ṭi: "Once it is known thanks to a means of knowledge (pramāṇa), with firm understanding (pratipatti), whether from books or from the words of a guru, that the Ś iva-nature is present in all, nothing remains to be done by means of instruments of knowledge or even (api) mental cultivation (bhāvanā). When knowledge is gold, one should abandon instruments and mental cultivation." Three of the four quoted lines (i.e. ekavāram : … bhāvanayāpi vā) do indeed occur in the edition of the Śivadr : s : ṭi: seventh Ā hnika, v. 5cd-6ab). The final line (jñāne … parityajet), though clearly included in the quotation attributed to the Śivadr : s : ṭi, does not occur in the available edition of that text, which has, at this place: jñāte 'pi tarubhūmyādidārḍhyān na karaṇādikam. It is possible, though far from certain, that the Sarvadarśanasam : graha here quotes an earlier version of the Śivadr : s : ṭi. There is, to my knowledge, no way at present to resolve this issue. Another example occurs in the chapter on Yoga, which quotes Yogasūtra 2.5 in the following form (ed. Abhyankar, p. 361 l. 15.298-299): pariṇāmatāpasam : skāraduḥkhair guṇavr : ttyavirodhāc ca duḥkham eva sarvam : vivekinaḥ This corresponds to the form the sūtra has in the critical edition of the Yogaśāstra, with the exception of the form guṇavr : ttyavirodhāc which, in that critical edition, has guṇavr : ttivirodhāc; this reading is confirmed in the Yogabhās : ya and in Vācaspati's Tattvavaiśāradī. The negative form°vr : ttyaviro°in the Sarvadarśanasam : graha is apparently supported by all manuscripts used in the preparation of the Bhandarkar and Ā nandāśrama editions; only the Bibliotheca Indica edition has guṇavr : ttinirodhāc. It is tempting to conclude that the "incorrect" reading°avirodhāc was not part of the autograph and must be corrected. However, Vijñānabhiks : u's comments on this sūtra defend the reading°avirodhāc, suggesting that this reading was current in at least certain manuscripts of the Yogaśāstra.
I will below present examples that illustrate the following: I. The archetype of the Sarvadarśanasam : graha is different from its autograph. II. Its archetype is identical with its autograph but different from the source-text.

I. Archetype different from autograph
Ia. A case where the editions of the Sarvadarśanasam : graha all go back to an erroneous reading occurs in the chapter on Vaiśes : ika (ch. 10: Aulūkyadarśana). In this chapter the Sarvadarśanasam : graha often makes use of the Padārthadharmasam : graha, better known by the name Praśastapādabhās : ya. While discussing the kind of division born from division (vibhāgajavibhāga) that is called "born from a division between cause and non-cause" (kāraṇākāraṇavibhāgaja), the Sarvadarśanasam : graha has the following line in all its editions (ed. Abhyankar p. 228, l. 10.137-138): haste karmotpannam avayavāntarād vibhāgam : kurvad ākāśādideśebhyo vibhāgān ārabhate/ An activity that has arisen in a hand, while making a division from another part of the body, brings about divisions from positions of ether etc. 6 The corresponding line in the Padārthadharmasam : graha reads (WI § 189, p. 32): yadā haste karmotpannam avayavāntarād vibhāgam akurvad ākāśādideśebhyo vibhāgān ārabhya The reading akurvad is confirmed by all editions and commentaries of the Padārthadharmasam : graha, and this is not surprising: only this reading makes sense in the Vaiśes : ika scheme of things. When one moves one's hand, no division between the hand and other parts of the body appears, whereas a division from positions of ether does.
Since it is hard to imagine that the author of the Sarvadarśanasam : graha introduced this change on purpose, we must conclude that a mistake entered the manuscript-tradition at an early date (unless we assume that the author's mastery of Vaiśes : ika left to be desired, an option that cannot be totally discarded). Ib. In its chapter on Pratyabhijñā (ch. 8) the Sarvadarśanasam : graha quotes a verse from the conclusion (upasam : hāra) of the chapter on action (kriyādhikāra) of Utpaladeva's Īśvarapratyabhijñākārikā, as follows (ed. Abhyankar,p. 197, no variants in the different editions): upasam : hāre 'piittham : tathā ghaṭapaṭādyākārajagadātmanā/ tis : ṭhāsor evam icchaiva hetukartr : kr : tā kriyā// iti/ This verse is Īśvarapratyabhijñākārikā 2.4.21, which however reads somewhat differently in the critical edition (Torella 1994, p. 61): ittham : tathā ghaṭapatādyābhāsajagadātmanā/ tis : ṭhāsor evam icchaiva hetutā kartr : tā kriyā// The critical edition notes no variants, except tadā in one manuscript for tathā. Torella (1994, p. 187) translates: Therefore causality, agency, action are nothing but the will of him who wishes to appear in the form of the universe, in the various manifestations of jar, cloth and so on.
The verse as quoted in the Sarvadarśanasam : graha is harder to translate. Cowell and Gough (1882, p. 133) propose: The mere will of God, when he wills to become the world under its forms of jar, of cloth, and other objects, is his activity worked out by motive and agent.
The translation "activity worked out by motive and agent" for hetukartr : kr : tā kriyā hardly makes sense, and we are justified in considering that the editions of the Sarvadarśanasam : graha do not preserve the original reading of this verse. And yet, it is hard to believe that the author of the Sarvadarśanasam : graha quoted a nonsensical verse. The conclusion must, once again, be that the archetype of the Sarvadarśanasam : graha differs at this place from its autograph. Ic. Consider now the following lines from the chapter on Nyāya (ch. 11; l. 11.200-203): īśvarasya jagatsarjanam : na yujyate/ tad uktam : bhaṭṭācāryaiḥprayojanam anuddiśya na mando 'pi pravartate/ jagac ca sr : jatas tasya kim : nāma na kr : tam : bhavet// It is not right to claim that God created the world. This has been stated by the teacher Bhaṫt˙a: "Not even a dim-witted person acts without a purpose. What has not been made by Him who creates the world?" The two half-verses here quoted have been taken from Kumārila Bhat˙ṫa's Ślokavārttika (Sambandhāks : epaparihāra vv. 55ab and 54cd respectively), but the second line is rather different in the one edition of that text accessible to me, as we will see below.
But let us first look at the text as we find it in the Sarvadarśanasam : graha. The lines are quoted to support the view that God did not create the world. The first quoted line does support this, for it could reasonably be argued that creating the world serves no purpose to Him all of whose desires are fulfilled. The second line, on the other hand, makes no sense in this context. This does not change if we accept the reading of the Bibliotheca Indica edition and supported by several manuscripts used for the Ā nandāśrama edition: jagac cāsr : jatas tasya … … by Him who does not create the world?
The edition of the Ślokavārttika has a different reading for this line: jagac cāsr : jatas tasya kim : nāmes : ṭam : na sidhyati What object of desire is not attained by Him even without creating the world? and this makes perfect sense. Since it is hard to believe that the author of the Sarvadarśanasam : graha quoted a nonsensical line, there are good grounds to believe that he quoted it as we find it in the edition of the Ślokavārttika. Clearly the archetype of the editions of the Sarvadarśanasam : graha contained an error, supporting the view that this archetype was different from the autograph of this text.
These examples give us reasons to think that the readings provided by the editions of the Sarvadarśanasam : graha do not always coincide with the readings of its autograph and can in certain cases be corrected with the help of the source-texts. Some scholars have raised this possibility into a principle. Uma Shankar Sharma stated already in 1964 that "the text of the Sarvadarśanasam : graha is … defective because the quotations of other works occurring in the present work sometimes present different readings when compared with the original text" (p. 22). Others have used this principle to correct the text.
Hélène Brunner, in her study of the chapter on the Śaivadarśana (1981), takes the position that in quoting verses from known source-texts, the author of the Sarvadarśanasam : graha did not wish to deviate from their original reading, so that we are entitled to correct the text in cases where the quoted verses differ from their source-texts. We are not, however, entitled to do so in the case of prose passages (Brunner 1981, p. 107): Les śloka cités par [Mādhava] proviennent, à une exception près, de textes dont on possède des éditions, ou plusieurs mss.; et beaucoup d'entre eux sont couramment cités dans la littérature śivaïte. A part un ou deux détails que nous signalons, leur forme est bien assurée et on peut les corriger sans hésitation; il ne s'agit pas de suggérer pour eux des lectures nouvelles issues d'un cerveau imaginatif, mais de rétablir celles qui sont attestées partout. Il en va autrement pour la partie en prose, c'est-à-dire l'exposé de [Mādhava], dont la forme correcte ne peut être rétablie par simple comparaison avec les passages qui l'inspirent, puisque justement [Mādhava] modifie ceux-ci, peu ou prou. 7 Raffaele Torella's article "Due capitoli del Sarvadarśanasam : graha: Ś aivadarśana e Pratyabhijñādarśana" (1980) follows by and large the same method. It proposes numerous emendations of the text of those two chapters, which it justifies with the observation that these chapters are largely based on a small number of known texts. Chapter six, on the philosophy of the followers of Ś iva (śaivadarśana), Torella (1980, p. 363) states, is like a collage of passages taken from two works: Aghoraśiva's commentary on Bhojarāja's Tattvaprakāśa and Nārāyaṅakaṅṫha's commentary on the Mr : gendrāgama, called Mr : gendravr : tti. Similarly, the seventh chapter, on the philosophy of recognition (pratyabhijñādarśana), makes extensive use of Abhinavagupta's Īśvarapratyabhijñāvimarśinī. 8 The approach adopted by Brunner and Torella is understandable and no doubt justifiable in certain cases. However, there appear to be cases where their approach does not work.
12 Torella (1980, p. 389): Mi sembra però più probabile che ciò sia da imputare all'intervento successivo di un copista (il compilatore del ms. capostipite di quelli che ci sono pervenuti), il quale, trovandosi davanti al composto ormai corrotto in pr[ā]v[r : ]tīśo balam (così effettivamente è riportato nel ms. Kh), abbia ritenuto, pensando che corrotto fosse il commento, di dover adeguare quest'ultimo a quello. Altra ipotesi è che corrotto in questo sense fosse già il testo della Mr : V cui l'autore del SDS attingeva. les traducteurs et le commentateur moderne. Et c'est elle qui est à l'origine de la fâcheuse transposition d'une ligne un peu plus loin." In this case, then, we can only "correct" the wording of a quoted verse on condition that we change the following prose as well. Such a correction can only be justified by evoking various actors (presumably copyists) who actively and knowingly interfered with the text. This activity must then have taken place before the archetype of the Sarvadarśanasam : graha editions, and presumably at a time close to the composition of this text. 13 But obviously, Occam's razor prefers Torella's less preferred hypothesis, viz. that the author of the Sarvadarśanasam : graha himself changed the wording of this verse, or that the text of the Mr : gendravr : tti used by him contained this corruption. Either way, a translation of the Sarvadarśanasam : graha must translate, or try to translate, what its author wrote, not what he should have written according to modern scholars.
In this particular case, Torella, unlike Brunner, is willing to consider that a "corruption" goes back all the way to the author of the Sarvadarśanasam : graha. He does so again on p. 388, where he observes that the word māyā (l. 7.181) should be mahāmāyā, then adds that the author of the Sarvadarśanasam : graha himself may have introduced the change out of ignorance. 14 Brunner (p. 135) is less tolerant, and replaces "incorrect" māyā with "correct" mahāmāyā, without further comments.
However, "correcting" the verse obliges her also to change the preceding prose, which contains the compound tattatkarmāśayavaśād. Brunner "corrects" this (note 64) into tat-tat-karmāśayādhivāsita-bhoktr : , because "Le terme bhoktr : qui apparaît dans la version correcte du śloka suivant, doit nécessairement apparaître ici." But the term bhoktr : does not occur in the verse as we find it in the editions of the Sarvadarśanasam : graha. We are once again in a situation where we must either accept that an early copyist did not just make a copying mistake but reworked the text, or we accept that the author of the Sarvadarśanasam : graha did so himself. As it is, the readings as we find them in the editions of the Sarvadarśanasam : graha make perfect sense. Occam's razor obliges us, once again, to attribute those readings to its author.
IIc. The Sarvadarśanasam : graha cites a line from the Kiraṇāgama, as follows (ed. Abhyankar, p. 180 l. 7.80-81): tad uktam : śrīmatkiraṇe: śuddhe 'dhvani śivaḥ kartā prokto 'nanto 'hite prabhuḥ// iti/ This has been stated in the Kiraṇāgama, as follows: "Ś iva has been stated to be the agent on the pure path, on the improper path it is Ananta." This corresponds to Kiraṇāgama, vidyāpāda 3.27cd (Vivanti 1975, p. 14), with this difference that the edition of the Kiraṇāgama has 'site ('black') instead of 'hite ('improper'). Brunner (1981, p. 121) and Torella (1980, p. 387) "correct" the verse, but are then confronted with a difficulty in the immediately preceding sentence, which expresses essentially the same meaning, but has kr : cchrādhvavis : aye "in the area of the evil path", which supposedly corresponds to the "corrected" expression asite 'dhvani "on the black path". They now feel free to "correct" this to kr : s : ṇādhvavis : aye "in the area of the black path", even though they know that this modification is not, apparently, supported by any of the source-texts. This form is found in one of the manuscripts used by the editors of the Ā nandāśrama edition, but even this manuscript had ahite rather than asite (as far as we can tell), which suggests that it is no more than a corruption inspired by the opposition with śuddhādhvavis : aye "in the area of the pure path" earlier in the same sentence. Brunner and Torella's "correction" would imply that a corruption from asite to ahite has subsequently motivated an early copyist to change kr : s : ṇa into kr : cchra, because kr : cchra 'evil' and ahita 'improper' have overlapping meanings. This sequence of assumptions can be avoided if we accept that the author of the Sarvadarśanasam : graha consciously introduced both the words ahita and kr : cchra.
IId. The first chapter of the Sarvadarśanasam : graha contains two verses that are quoted twice over, but not in identical form. On p. 5 ll. 1.50-51 it quotes the following proverb (ābhāṇaka): agnihotram : trayo vedās tridaṇḍam : bhasmaguṇṭhanam/ buddhipaurus : ahīnānām : jīviketi br : haspatiḥ// The Agnihotra sacrifice, the three Vedas, the triple stave of the religious ascetic, covering oneself in ashes, these constitute the livelihood of those devoid of intelligence and exertion. This is what Br : haspati says.
On p. 13 ll. 1.112-113 it quotes the same verse in this form: agnihotram : trayo vedās tridaṇḍam : bhasmaguṇṭhanam/ buddhipaurus : ahīnānām : jīvikā dhātr : nirmitā// The Agnihotra sacrifice, the three Vedas, the triple stave of the religious ascetic, covering oneself in ashes, these constitute the livelihood, made by the creator (dhātr : nirmitā), of those devoid of intelligence and exertion. Bhattacharya (2011, pp. 207-211) argues that the reading ending in … jīviketi br : haspatiḥ is original, while the ending … jīvikā dhātr : nirmitā is a modification introduced by the author of the Sarvadarśanasam : graha in order to avoid mentioning Br : haspati twice over in short succession. Indeed, in all parallel instances (ten out of eleven) the reading is jīviketi br : haspatiḥ (p. 73) The same chapter quotes another verse twice over, first on p. 2 ll. 1.17-18: yāvajjīvam : sukham : jīven nāsti mr : tyor agocaraḥ/ bhasmībhūtasya dehasya punar āgamanam : kutaḥ// 15 One should live happily as long as life lasts, for nothing is beyond the reach of death. Why should we believe that a body that has been reduced to ashes will come back into this world?
and then again on p. 14 ll. 1.122-123: yāvaj jīvet sukham : jīved r : ṇam : kr : tvā ghr : tam : pibet/ bhasmībhūtasya dehasya punar āgamanam : kutaḥ// One should live happily as long as life lasts; having incurred a debt one should drink ghee. Why should we believe that a body that has been reduced to ashes will come back into this world?
According to Bhattacharya (2011, p. 73), the reading r : ṇam : kr : tvā ghr : tam : pibet is spurious. It occurs only once (viz., in the Sarvadarśanasam : graha) in the fourteen instances he found in the literature in which the verse is wholly or partly quoted or adapted.
In these two cases, then, we have reason to think that, on purpose or out of carelessness, verses were quoted in two different forms by the author of the Sarvadarśanasam : graha.
IIe. The chapter on Yoga (pātañjaladarśana) of the Sarvadarśanasam : graha explains a number of Yogasūtras, staying in general close to the Yogabhās : ya. However, when discussing the postures (āsana), it states (in all editions) that there are ten of them, which it enumerates: sthirasukham āsanam : padmāsanabhadrāsanavīrāsanasvastikāsanadaṇḍakāsanasopāśrayaparyan . kakrauñcani s : adanos : ṭranis : adanasamasam : sthānabhedād daśavidham (p. 376, ll. 15.463-464). It appears that the author of the Sarvadarśanasam : graha skipped one, the hastinis : adana, which is yet included in all surviving editions and manuscripts of the Yogaśāstra under sūtra 2.46 (Maas 2018). The mistake is easily understood, since the Yogabhās : ya does not explicitly state that there are eleven postures, even though it enumerates eleven of them. It seems reasonable to conclude that we are here confronted with a simple mistake by the author of the Sarvadarśanasam : graha.
The quoted line is Gaṇakārikā 6ab, 16 which however has siddhiś instead of śuddhiś. Hara (1958, pp. 14-15) therefore "corrects" the text of the Sarvadarśanasam : graha. However, the reading śuddhi is shared by all editions of this text and was therefore presumably part of its archetype. One might conjecture that it is the result of a simple scribal error, but this cannot be the case, for the immediately preceding line reads (p. 162, l. 6.16-17): jñānatapodevanityatvasthitiśuddhibhedāt pañcavidhaḥ once again with śuddhi. If the autograph had siddhi, a conscious scribal modification must be held responsible for the text as we have it. It is less cumbersome to assume that śuddhi was already part of the autograph, which the author of the Sarvadarśanasam : graha found in his source-text or introduced himself. * * * The examples just considered should discourage us from "correcting" the text of the Sarvadarśanasam : graha too hastily. They suggest that its author did not always blindly copy the source-texts, either willingly or because the manuscripts he used were not identical with those used for their modern editions (or indeed out of carelessness). Either way, it makes sense to understand, and translate, even the quoted passages in the Sarvadarśanasam : graha as we find them in its editions, on condition that those readings are intelligible and make sense. Proceeding otherwise may expose us to serious misunderstandings, as I will now show.
This, as pointed out above, is sufficient reason to stick to the reading of the Sarvadarśanasam : graha editions. However, there is more. Torella's emendation contains the word jīvanmukti "liberation while alive". This word is nowhere found in the Sarvadarśanasam : graha, 19 and there are reasons to think that it was avoided on purpose. Claiming liberation while alive for certain members of a school or sect has political implications. It means that that school or sect is superior to others, since it obviously teaches the right path. It seems probable that the author of the Sarvadarśanasam : graha wanted to avoid such issues-in spite of the fact that his uncle (or at any rate someone close to him) had composed the Jīvanmuktiviveka, a text that does not eschew such a claim. Since I have dealt with this issue elsewhere (Bronkhorst forthcoming), I will say no more about it. Let it be sufficient here to state that if we "correct" the Sarvadarśanasam : graha in the light of a source-text we run the risk of introducing a notion that its author had taken care to avoid.
Funding Open access funding provided by University of Lausanne

Compliance with Ethical Standards
Conflict of interest The author states that there is no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creativecommons.org/licenses/by/4.0/.