Candrakīrti on the Use and Misuse of the Chariot Argument

The publication in 2015 (ed. Li) of Chap. 6 of the rediscovered Sanskrit text of Candrakīrti’s Madhyamakāvatāra (MA) allows us to witness more directly Candrakīrti’s careful and deliberate critique of the ‘chariot argument’ for the merely conventional existence of the self in Indian Abhidharmic thought. I argue that in MA 6.140–141, Candrakīrti alludes to the use of the chariot argument in the Milindapañha as negating only the view of a permanent self (compared to an elephant), rather than negating ego-identification (compared to a snake in its hole). In contrast to this misuse of the chariot argument, in MA 6.150–165 Candrakīrti uses the chariot argument as an allegory to enable the meditator to refute the basis of ego-identification in seven ways. Candrakīrti’s use of the chariot argument does not establish any theory about the self or not-self, but acts as a guide to meditation as part of philosophy as a spiritual practice with the goal of liberation.


Introduction
This article has its starting point in an observation that I made while studying the stanzas of Chap. 6 so far made available (Li, 2015) from the recently rediscovered Sanskrit text of Candrakīrti's Madhyamakāvatāra. 1 The Sanskrit contains what I believe is an allusion to the Milindapañha, which translations into Tibetan appear not to have preserved. In stanza 6.140, Candrakīrti puts forward a subtle challenge to the claim made by other Buddhists about their understanding of non-self (anātman): When there is an understanding of non-self, a permanent self is rejected, but we do not accept that this is the basis of ego-identification. If someone says that they have uprooted that from their own philosophical view by knowing the non-existence of the [permanent] self -that is very brilliantly said. 2 In this stanza, Candrakīrti explores a distinction already made in previous stanzas of Chap. 6 between a permanent self (nitya-ātman), which in his view is entirely nonexistent, and the self which is the object of ego-identification (ahaṃkāra), which is conceived dependent on the constituents (skandhas). This ordinary, everyday self is the unexamined assumption made by almost everyone, except those who have realised the truth of emptiness, about what and who they are; it is the referent of words like 'I', 'me' and 'mine', that we use without thinking deeply about them. This unexamined assumption is also called 'naive realism' (satkāya) and it becomes formulated unwittingly as a 'view that reifies the self' (satkāyadṛṣṭi). This ordinary self, so Candrakīrti explains, should not be confused with the conception of a permanent self, which is usually a religious or philosophical belief about a soul or unchanging essence of the person. According to the stanza above, Candrakīrti's view is that it is one thing for a Buddhist to give up a belief in a permanent self, but quite another to identify and let go of that more subtle and pervasive idea of self to which words like 'I', 'me' and 'mine' refer. And liberation, the goal of Buddhist practice, is the result of giving up ego-identification, not just removing the false view of a permanent self.
But who are the other Buddhists that Candrakīrti has in mind here, who claim to say that they have uprooted ego-identification by knowing the non-existence of the permanent self? It is not possible to work out who they might be from the surviving Tibetan translation of the Madhyamakāvatāra, even though the Tibetan communicates Candrakīrti's philosophical message accurately enough. However, the last line of the rediscovered Sanskrit text concludes with an apparently ironic gesture of mock celebration, the specific wording of which is obscured in the Tibetan: 'that is very brilliantly said' (ucyate ati iva citram). What Candrakīrti means is that for someone to say that they have successfully identified and let go of ego-identification, simply by removing the false view of a permanent self, is something 'brilliantly said' in a completely ironic sense: it will not work. 3 Candrakīrti is not categorically denying that removing the false view of a permanent self could undermine ego-identification, but in subsequent stanzas he will put forward what he believes to be a much more successful method for removing it.
My observation is that the words 'that is very brilliantly said' (ucyate ati iva citram) echoes some words in the Milindapañha ('The Questions of King Milinda'). In a well-known early section of this text, as preserved in Pāli, the Buddhist monk Nāgasena engages in debate with King Milinda about the meaning of the Buddha's not-self teaching. Nāgasena deploys the chariot argument: 4 just as what is called a 'chariot' is merely a name, a conventional designation, for what exists dependent on its wheels, axle, body, pole and so on, likewise what is called a 'self' is merely a name, a conventional designation for what exists dependent on the constituents (khandhas), namely, physical form, feeling, perception, formations and consciousness. For good measure, Nāgasena cites the canonical source for the chariot argument, a short discourse in which an arahant nun explains the not-self teaching to Māra: 'Just as, from an arrangement of parts (aṅgasambhārā), There is that for which we have the word "chariot", Likewise when the constituents (khandhā) exist, There is what we call a "being" (satta).' 5 King Milinda is defeated in debate, and he replies approvingly: Marvellous, Nāgasena, good sir; wonderful, Nāgasena, good sir! The confident responses to the questions you were asked are brilliant (aticitra). If the Buddha were here, he would applaud you. Excellent, excellent, Nāgasena. The confident responses to the questions you were asked are brilliant (aticitra).' 6 I propose that Candrakīrti's ironic praise that an argument was 'very brilliantly said' (ucyate ati iva citram) is an echo of King Milinda's use of the word aticitra ('brilliant') to describe Nāgasena's words. Candrakīrti's allusion to the Milindapañha therefore indicates some philosophical disapproval of the argument about the self that Nāgasena employs. I will argue that this disapproval is a criticism of its practical implications.
In order to make my case, I will rehearse the use of the chariot argument in canonical sources and in the Milindapañha, where it is used as an analogical argument that shows that the self exists only as a conventional designation dependent on the constit-3 Perhaps the phrase ati iva citram is deliberately ambiguous, meaning 'over-brilliant' (not so brilliant at all) as well as 'very brilliant'. The irony might be in translation rather than in the original. 4 See below for detailed discussion with references. 5 The Pāli version of this discourse is at S 5: 10 pts I 34-5, where the bhikkhuni's name is Vajirā. I will discuss other versions of this discourse below. 6 Milindapañha (Miln) pts 28 acchariyaṃ bhante nāgasena, abbhutaṃ bhante nāgasena, aticitrāni pañhapaṭibhānāni vissajjitāni. yadi buddho tiṭṭheyya sādhukāraṃ dadeyya. sādhu sādhu nāgasena, aticitrāni pañhapaṭibhānāni vissajjitāni.
uents. This will allow us to appreciate what Candrakīrti thought counted as a misuse of the argument and why, and it will also allow us to better appreciate Candrakīrti's own use of the chariot argument in his analysis of the self in Madhyamakāvatāra 6.150-165.

The Canonical Version of the Chariot Argument
The Pāli version of the Milindapañha, quoted above, cites a Pāli canonical version of the chariot argument. In this version, the arahant bhikkhunī Vajirā responds to the character of Māra, a personification of evil, who has appeared in order to sow doubt and distraction in her meditation. He asks her: 'By whom was this being made? Where is this being's maker? Where did this being arise? Where will this being cease?' 7 Such metaphysical questions concerning a 'being' (satta) seem to represent the existential self-questioning of a thoughtful human being who wonders about their origins and destiny. We should note that Māra's questions do not necessarily imply a belief in or concern about a permanent metaphysical self, such as the Upaniṣadic ātman, but rather simply take for granted a belief in the reality of the person we ordinarily take ourselves to be.
The canonical discourse in which the chariot argument appears has also been preserved in parallel versions translated into Chinese and Tibetan, as well as partially in Sanskrit. 8 In these, the main character is a bhikṣunī called Śailā rather than Vajirā. 9 I quote Śailā's reply to Māra from the Sanskrit, preserved in Vasubandhu's Abhidharmakośabhāṣya, as Candrakīrti is more likely to have been familiar with this version, or one like it, rather than with the Pāli: 'Māra, why do you think about a "being"? You have a speculative view. For this heap of fabrications is empty. No being is found here. 'For just as because of a collection of parts what is referred to as a "chariot" is so named, in this way we speak of a being as a convention dependent on the constituents.' 10 In whatever exact form of words, this argument has been called upon throughout the history of Buddhist philosophy. Matthew Kapstein comments: 7 Saṃyutta Nikāya (S) 5: 10 pts I 34: kenāyaṃ pakato satto kuvaṃ sattassa kārako | kuvaṃ satto samuppanno kuvaṃ satto nirujjhati ||. 8 Recently studied in depth by Dhammadinnā (2020). 9 The change of name is explained (with further references) by Dhammadinnā (2020: 3). 10 From Ch.9, the Pudgalapratiṣedhakaraṇa, 'Treatise on the Refutation of the Person': manyase ki nu sattve 'ti māra dṛṣṭigataṃ hi te | śūnyaḥ saṃskārapuñjo 'yaṃ na hi sattvo 'tra vidyate || yathaiva hy aṅgasaṃbhārāt saṃjñā ratha iti smṛtā | evaṃ skandhān upādāya saṃvṛtyā sattva ucyate || (Prahlad Pradhan, 1975: 466 1-4 ).
[W]e find condensed here as nowhere else the fundamental themes that are interwoven throughout the Buddhist enquiry into the reality of the self and the nature of persons. Throughout these developments Vajirā's [or Śailā's] verses remained the epitome of the founder's teaching on the matter of the reality of the self. (Kapstein, 2001: 78-9) It is therefore worth dwelling on the nature of the argument made in Vajirā/Śailā's verses, as they became the scriptural support for later arguments, which interpreted them in different ways. In order to appreciate these different interpretations, I propose a re-casting of the argument in the form of an argument by analogy, in order to make more easily intelligible what is being argued: Premise 1: A chariot depends on a collection of parts, namely, wheels, axle, body, pole, etc.
Premise 2: A human being, by analogy, also depends on a collection of parts, namely, the constituents (skandhas).

Premise 3:
It is a matter of convention to use the word 'chariot' to refer to this vehicle which depends on a collection of parts.

Conclusion:
It is a matter of convention, by analogy, to use the word 'being' to refer to a person who depends on the constituents.
What follows from Bhikṣunī Śaila's argument is that it is perfectly reasonable to use the word 'chariot' as a matter of convention to refer to a vehicle that exists depending on its parts. Likewise, it is reasonable to use the word 'being', with the proviso that the reference for this word is dependent on the constituents. Therefore, in response to Māra, Śailā can be understood to say that it is not reasonable to ask metaphysical questions about this 'being', as if the word could refer to anything real, apart from the constituents. That would be as nonsensical as asking about the origins and maker of a 'chariot' that existed apart from its wheels, axle, body, pole and so on.
Bhikṣunī Śailā's argument assumes an acceptance of the premise that the human being can be analysed into the constituents (skandhas), namely, physical form (rūpa), feeling (vedanā), perception (saṃjñā), formations (saṃskārā) and consciousness (vijñāna). Bhikṣunī Śailā says that the 'being' which Māra speaks of, an empty heap of fabrications (śūnyaḥ saṃskārapuñjo), is dependent on the constituents. By implication, these constituents are themselves dependently-arisen, without a maker, and of a nature to arise and cease. This is an argument for Buddhists, against the tendency to reify the self; it is not an argument aimed at non-Buddhist conceptions of the self. The non-Buddhist who believed that there was a soul or essence of the person separate from the constituents would be unlikely to accept the conclusion of the argument, since they would be unlikely to accept the truth of premise 2, as I have given it above. The chariot argument therefore appears to have been devised to appeal to those who reject the idea of a permanent self, but who may nevertheless be assailed by metaphysical questions about the self. Just as Śailā's stanzas cause Māra to go off dejectedly, so the chariot argument helps in re-establishing an understanding of the teaching in the face of doubt. This would suggest that the canonical version of the chariot argument was not intended to refute the concept of a permanent, metaphysical self, but rather to undermine the kind of residual belief in an ordinary but real self or person that Māra relied on to assail Śailā.
There is, nevertheless, a bigger context for the chariot argument in ancient Indian philosophy and beyond. In the Kaṭha Upaniṣad, the self (ātman) is compared to the rider in a chariot (ratha): Know the self as a rider in a chariot, and the body, as simply the chariot. Know the intellect as the charioteer, and the mind, as simply the reins.
The senses, they say, are the horses, and sense objects are the paths around them… 11 Though the Kaṭha Upaniṣad cannot be dated with precision (Cohen, 2018a), its comparison of the self with a 'rider in a chariot' (rathin) was probably an idea in circulation among the ascetics and brahmans of ancient India, such that the Buddhist chariot argument is a conscious response to it. 12 If this is the case, the chariot argument might have been understood to offer an implicit challenge to the Upaniṣadic idea of a metaphysical self (ātman), 13 as well as undermining the everyday belief in a self.

The Chariot Argument in the Milindapañha
I now turn once again to the Milindapañha, in order to show more precisely how it handles the chariot argument, and why Candrakīrti objects to that handling. If my observation is correct, that Candrakīrti's ironic praise, 'that is very brilliantly said' (ucyate ati iva citram), is an allusion to King Milinda's words to Nāgasena, that 'the confident responses to the questions you were asked are brilliant (aticitra)', then Candrakīrti would have been familiar with a version of the Milindapañha like the one that has survived in Pāli, though more likely it would have been a version in Sanskrit. 14 11 Kaṭha Upaniṣad 3.3-4 (Olivelle, 1998): ātmānaṃ rathinaṃ viddhi śarīraṃ ratham eva tu | buddhiṃ tu sārathiṃ viddhi manaḥ pragraham eva ca. || indriyāṇi hayān āhūr viśayāṃs teṣū gocaran |. 12 Discussed further by (Cohen, 2018b), (Schiltz, 2018), and specifically in this context in Dhammadinnā (2020). Many miles away and a few centuries later, the Greek philosopher Plato also compared the soul (ψυχη) to a chariot, notably in Phaedrus 246a, in a remarkable parallel to the passage in the Kaṭha Upaniṣad. Whether this parallel is the result of the diffusion of ideas, presumably from India to Greece, or the result of the independent development of ideas in the two cultures is an open question; on which see (Mignone, 2016) and (Schlieter, 2016). 13 This point is made in more detail by Dhammadinnā (2020: 17).
14 On the possibility of the Milindapañha having been composed in or translated into Sanskrit, see Levman (2021: 112). Levman and Anālayo (2021) have independently explored how the surviving versions of the King Milinda opens the debate by asking Nāgasena his name. 15 He replies by saying that, although he is known by the name Nāgasena, this is but an expression (sankhā), a designation (samaññā) a concept (paññati), a common usage (vohāra), a mere name (nāmamattaṃ), 'for here no person is found' (na h' ettha puggalo upalabbhati). Nāgasena sets the stage for the rehearsal of an argument for the merely conventional or nominal existence of the self or person, as a name or designation, while in an ultimate sense, no self or person exists. King Milinda, in the context of debate, now tries two lines of argument against Nāgasena. Firstly, King Milinda argues that if no person is found, then there can be no agent, and no-one who practices ethics or who experiences the results of actions. Then King Milinda goes on to ask if Nāgasena is identical with any part of his body, or any of the constituents of feeling, perception, formations and consciousness. Since Nāgasena answers 'no' to all these questions, King Milinda accuses Nāgasena of speaking falsely, for there simply is no Nāgasena. 16 Garfield (2015: 107) makes the observation that in this section of the debate, King Milinda gives voice to a distorted version of a Buddhist analysis of the person, 'a kind of parody of an Abhidharma reductive analysis'. Noting that the self is not found in any part of the human body or experience, he concludes that there is no person at all. Now, in reply, Nāgasena goes on to present a more nuanced view. He asks the King, regarding the chariot in which he had travelled to the debate, whether any part of the chariot, such as wheels, axle, body, pole and so on, is the chariot. Since King Milinda says 'no' to each question, Nāgasena replies that the King has spoken falsely, for there simply is no chariot. This is a reductio ad absurdum of King Milinda's previous line of questioning. The King now thinks again, and admits that the chariot exists as an expression, a designation, a concept, a common usage, a name, dependent on its various parts. In conclusion, Nāgasena can now point out that 'Nāgasena' also exists as a expression, a designation, a concept, a common usage, a name, dependent on his various parts, even though 'in the ultimate sense' (paramatthato) no person exists.
It is at this point in the debate that Nāgasena cites the canonical story of the bhikkhunī Vajirā and her use of the allegory of the chariot to overcome Māra. King Milinda then congratulates him, saying that 'the confident responses to the questions you were asked are brilliant (aticitra)'. But how brilliant are they really? If my observation concerning an allusion to Milindapañha in Madhyamakāvatāra 6.140 is correct, then Candrakīrti does not object to the chariot argument as such, but rather to the use to which it is put in this context. In the Milindapañha, Nāgasena uses it as a canonical reference that is supposed to confirm his argument that the self or person exists as a matter of convention, but does not exist in the ultimate sense. While Candrakīrti would not find this objectionable, let us look again at Candrakīrti's argument in MA 6.140 to see the problem that he notices: Milindapañha relate to an earlier, simpler, question-and-answer or teaching instruction. 15 Here I summarise Miln pts 25-28.
When there is an understanding of non-self, a permanent self is rejected, but we do not accept that this [permanent self] is the basis of ego-identification. If someone says that they have uprooted [ego-identification] from their own philosophical view by knowing the non-existence of the [permanent] self -that is very brilliantly said.
His criticism of Nāgasena is that the argument that the self or person does not exist in the ultimate sense can only succeed in showing the non-existence of a permanent self. By showing this he has not properly conveyed the import of the Buddha's teaching of non-self. Someone who understands non-self certainly rejects the existence of a permanent self, but this in itself is insufficient to uproot the ego-identification that is the basis of the ordinary sense of self. By affirming the conventional existence of the self or person, Nāgasena has failed to identify the way that the ordinary sense of self depends upon an identification with the five constituents.
In the following stanza of Madhyamakāvatāra, Candrakīrti gives a vivid illustration of what he believes to be the mistake in Nāgasena's argument: [To suppose that] seeing a snake which has gone into a hole in one's own home one could remove one's terror by saying, 'there's no elephant in there!' and also abandon the fear that is because of the snakeso much indeed for our opponent's naivety. 17 In this illustration, the snake is the sense of self, and the elephant is the belief in a permanent self. Assuming that Candrakīrti's opponent here is indeed someone who holds the view represented by Nāgasena, Nāgasena's argument that in the ultimate sense no person is to be found is like saying there is no elephant in the house as a way to remove the terror of having a snake making its home there. Such an argument certainly appears to display 'naivety' (ārjavatā) in the sense of being sincere but misguided.
This may seem like a nit-picking criticism of Nāgasena. But, independently of Candrakīrti, and many centuries later, the English bhikkhu, Ven. Ñāṇavīra, made the same point about the teaching of the Milindapañha. In a discussion of the idea of paramattha sacca, of 'truth in the highest sense', Ñāṇavīra evokes the problem with Nāgasena's argument by noticing that, should one accept his argument and thereby suppose that one properly understands the non-self teaching, then: 'The unwary thinker comes to believe that he understands what, in fact, he does not understand, and thereby effectively blocks his own progress' (Ñāṇavīra, 2010: 40). In Candrakīrti's terms, the thinker may believe that they understand the non-self teaching, but in fact they only understand the non-existence of the permanent self, and not the ego-identification which is what one actually needs to understand to make progress.
More recently, Dhammadinnā (2020) has also identified the misuse of the chariot argument in the Milindapañha as the result of a misunderstanding of how Bhikṣuni Śailā employs it in its canonical context. There, a chariot is an example of a concept that refers to a complex assembly of parts and yet is easily recognised as a perception (saṃjñā) in experience. Perceptions such as this arise as a feature of the way that experience becomes fabricated or constructed, the process of which is a common theme of early Buddhist discourses. Dhammadinnā goes on: [T]he point of bhikṣuṇī Śailā's illustration is not a denial of the conditioned existence of a chariot or a self, but a calling into question the shift from concept to ontology evident in Māra's pressing her on the characteristics of a 'being'. Māra ignores, or possibly denies, the dependently arisen nature of the very construction of experience. The position Śailā takes does not entail a proposal of nominalism (be it in the form of a rejection of abstract objects or in the form of a rejection of universals). The use of the notions or terms 'chariot' and 'being' are not problematized as such. Nor does Śailā make a metaphysical affirmation of their conventional existence but ultimate or absolute non-existence, unlike the type of arguments that developed in later Buddhist tradition. (Dhammadinnā, 2020: 13) The later Buddhist arguments to which Dhammadinnā here refers take their cue from the Milindapañha. The Pāli version of this text appears to have created a presentation of the non-self teaching, based on a misuse of the chariot argument, that influenced later Abhidharma presentations. When Candrakīrti criticises the Milindapañha (or so I claim) for mistakenly teaching the undermining of the belief in a permanent self as a way to uproot ego-identification, his criticism extends to those Abhidharma presentations of the non-self teaching that repeat the Milindapañha's approach.

The Misuse of the Chariot Argument
The preceding study of how in the Sanskrit stanzas of the Madhyamakāvatāra we can identify Candrakīrti's allusion to the Milindapañha comprise the original contribution I had wished to make in this article. Candrakīrti argues that the author of the Milindapañha misuses the chariot argument as it is presented in canonical sources. This misuse is symptomatic of the Abhidharma reductionist account of the Buddhist teaching of non-self. The Milindapañha, as preserved in Pāli, and hence we assume available to Candrakīrti in a similar version, already shows signs of being expanded or revised to include the Abhidharma conception of the non-existence of the person in the ultimate sense. This is evident in Nāgasena's argument to King Milinda that the self or person exists as an expression, a designation, a concept, a common usage, a name, dependent on the five constituents (khandhas), even though in the ultimate sense no person exists. The Abhidharma approach can broadly be categorised as reductionist and eliminativist because it claims that the self can be reduced to its component parts through philosophical analysis in order to eliminate it. 18 Let us review how Candrakīrti believes the Abhidharma account is a misuse of the chariot argument. The Abhidharmikas present the chariot argument as illustrating how the self does not exist in an ultimate sense, but is merely a convention. In the Milindapañha, the argument is presented in the context of debate with a non-Buddhist, and might therefore be justified in terms of the public presentation of Buddhist ideas. Whereas in MA 6.140 Candrakīrti makes an ironic or ambiguous comment about Nāgasena's victory in debate, in MA 6.141 he implies that the Abhidharma approach is naive, which is to say, ineffective at bringing about the liberating insight that is its supposed aim. It is ineffective because realizing the non-existence of the permanent self is not sufficient to uproot the true cause of suffering and saṃsāra, which is belief in the referent of ego-identification at the root of the ordinary and everyday sense of self.
It is possible to re-cast the Abhidharma account of the chariot argument as an analogical argument as follows: Premise 1: A chariot depends on a collection of parts, namely, wheels, axle, body, pole, etc.
Premise 2: A human being, by analogy, also depends on a collection of parts, namely, the constituents (skandhas).

Premise 3:
It is a matter of convention to use the word 'chariot' to refer to a vehicle which does not ultimately exist, because it depends on a collection of parts.

Conclusion:
It is also a human convention, by analogy, to use the word 'being' to refer to a person who does not ultimately exist, because the self depends on the constituents. This way of presenting the Abhidharmikas' version of the chariot argument helps us to see how they have made a subtle change to its form. The addition of the 'does not ultimately exist' conveys the reductionism they have introduced. The parts of which the chariot is made are themselves made up of parts, and likewise the constituents can be analysed into processes consisting of dharmas arising and passing away. The process of reduction stops at what ultimately exists, namely, dharmas. The ultimately existent dharmas have become necessary for the cogency of the argument. It is insight into the nature of the dharmas that allows the Abhidharmika to eliminate the self. Candrakīrti disagrees with this soteriology, which depends on a commitment to the Abhidharma metaphysics of dharmas.

Candrakīrti's use of the Chariot Argument
I will now summarise how, in Madhyamakāvatāra 6.150-165, Candrakīrti goes on to revisit the allegory of the chariot, to show how it should be used in a properly liberating way, without depending on the metaphysics of the dharmas. This proper use of the chariot argument begins after first clarifying the need for a refutation of the self, which Candrakīrti states in MA 6.120: Clearly seeing through wisdom both afflictions and faults that are wholly produced from views that reify the self (satkāyadṛṣṭi), and understanding the object of this [view] to be the self (ātman), the meditator makes a refutation of the self (ātman). 19 It should be obvious by now that the refutation of the self is not going to be simply the proof that a permanent self is non-existent. In fact, Candrakīrti uses another argument from analogy to show that the refutation of the permanent self is not what is needed: the idea of a permanent self is like the birth of a son from an infertile woman (MA 6.122), which does not exist even conventionally. He categorises such a self as one which is conceived as existing apart from the constituents (skandhas). But, he argues, even some animals have a sense of self (one thinks of higher mammals like dogs and elephants), which Candrakīrti calls ego-identification (ahaṃkāra), though animals certainly do not have a philosophical view of a permanent self (MA 6.125). Therefore, the view of a permanent self is not necessary or sufficient for ego-identification, the ordinary everyday experience of a self or person.
The target of a refutation of the self is therefore ego-identification. According to different schools of Buddhist thought in Candrakīrti's day, this self is one that is the same as one or all of the constituents, in different ways. It is not necessary here to rehearse Candrakīrti's arguments (in MA 6.126-137) against different Buddhist philosophical schools, 20 except that in MA 6.135 Candrakīrti invokes the chariot argument in the voice of a Buddhist opponent, who cites the canonical discourse about the chariot in order to argue that the self is the same as the constituents considered as a heap: '[…] it should be [argued that] the self is comparable to a chariot, the idea of a chariot belonging to its parts existing in a heap.' But, in a discourse it is said to take the constituents as its basis, therefore the simple combination of the constituents is not the self. 21 That is to say, while one Buddhist philosopher uses the chariot argument in this way, Candrakīrti cites another canonical discourse, in which the self is said to exist by appropriating or taking the constituents as its basis. 22 Canonical discourses, when taken out of context, can be misunderstood and used to defend what Candrakīrti 19 MA 6.120: satkāyadṛṣṭiprabhavān aśeṣān kleśāṃś ca. doṣāṃś ca. dhiyā vipaśyan | ātmānam asyā viṣayaṃ ca. buddhvā yogī karoty ātmaniṣedham eva ||. 20 This is done in depth by Duerlinger (2013). 21 MA 6.135: kūṭasthānāṃ syād rathatvaṃ tadānīṃ tasyāṅgānāṃ tulya ātmā rathena | skandhāṃś copādāya sūtre niruktas tasmān nātmā skandhasaṃhātamātram ||. 22 In his own bhāṣya on the Madhyamakāvatāra, Candrakīrti cites a discourse now preserved in Tibetan translation, entitled Pitāputrasamāgamasūtra (Liland, 2019: 308), in support of his own stanza. However, another relevant discourse is found in Pāli at S 22: 83, with a close parallel at SĀ 261 preserved in Chinese translation at T II 066a05. These discourses report Ānanda teaching the monks what he has learned from would regard as incorrect philosophical positions. But he argues that those discourses are correct in which the Buddha taught that the self exists taking the constituents as its basis: The sage taught that the self is dependent on (pratītya) the six elements -earth, water, fire, wind, consciousness and spaceand that it is dependent on the bases of contactthe six senses starting from vision. 23 He explained it as taking as its basis (upādāya) phenomena, mind and mental events, and therefore its reality is not from them, nor is [its reality] the state of mere combination. Therefore the conviction of ego-identification is not in reference to them. 24 The keyword here is upādāya, which I have translated 'taking as its basis'. 25 The word upādāya suggests that the sense of self is based on or depends on the constituents (there would be no sense of self without the constituents) and that the sense of self takes or appropriates them (one's sense of self makes the constituents one's own). One says or thinks "I am" taking the constituents as a basis. For this reason, the self is neither other than, nor the same as, the constituents, but nor is it from them, nor is it their combination, nor is it in reference to them.
This is bewildering at first, because Candrakīrti leaves nothing for us to conceive the self as or in terms of, while apparently not denying it. By contrast, the reductive Abhidharma argument allowed us to conceive of the self as not really real, but as instead nothing more than a name for the constituents. Candrakīrti's conception of the self is non-reductive, and he also claims that it is what the Buddha taught. 26 Therefore, the starting point for understanding Candrakīrti's account of the self is to notice the ordinary and everyday sense of self that is present as the presumed subject of experience for anyone who has not yet uprooted it through liberating insight into emptiness. Since the self is always present as the "I" of experience, no analysis or conceptualisation is necessary to become aware of the self, only the willingness to recognise what is already there. Furthermore, philosophical reflection about the self his teacher, Puṇṇa Mantāniputta, that saying or thinking "I am" is takes as its basis (upādāya) the five constituents. 23 (Edgerton, 1953) s.v. upādāya. cannot by itself bring about insight into selflessness, so philosophical reflection is not undertaken for that reason. Rather, one engages in philosophical meditation in order to undermine the unthinking belief that the self exists in the way that it appears to exist, which is as an enduring personal identity worth holding on to. 27 Undermining this belief brings forth the profoundly transformative inner experience of the selflessness or emptiness of the person, which liberates one from wrong views. In MA 6.145, Candrakīrti drops the triṣṭubh metre of the other stanzas and breaks into a more complex śakvari metre, as if to sing: Like those towering peaks are these long-enduring and immovable views that reify the self (satkāyadṛṣṭi). The self is torn apart by the thunderbolt of awakening to selflessness (nairātmya). The mountain of philosophical views (dṛṣṭi) goes to oblivion too. 28 Now, in MA 6.150, Candrakīrti presents a summary of his account of the ordinary sense of self, the snake to be followed down into its hole, which is the basis of egoidentification, which can be conceptualised as the "I am" which depends on the appropriation of the constituents: Therefore the basis of ego-identification is not a real thing, nor is it other than the constituents, nor is it the same as the constituents, nor is it the support of the constituents, nor does it possess them, but it is established as taking the constituents as its basis. 29 Candrakīrti now introduces the allegory of the chariot as as a way to illustrate this account of the self, and which he takes to be the correct use of the chariot argument.
Candrakīrti does not cite the allegory of the chariot from a canonical source to prove an analysis he has already made, but rather borrows its argument in order to use it in a correct way. In order to understand how he now uses the argument, I will once again re-cast it as an analogical argument, but this time with the inclusion of Candrakīrti's distinctive kind of realism about the self: Premise 1: A chariot depends on a collection of parts, namely, wheels, axle, body, pole, etc.

Premise 3:
It is a matter of convention to use the word 'chariot' to refer to this vehicle which exists conventionally, and which depends on a collection of parts.

Conclusion:
It is also a human convention, by analogy, to use the word 'being' to refer to a person who exists conventionally, and who depends on the constituents.
The addition of the phrase exists conventionally now throws us back to consider how exactly a chariot exists conventionally, and hence how a person likewise exists conventionally. The answer to both of these questions is that, despite the fact that chariot and self exist conventionally, philosophical analysis reveals that there is no way to conceptualise the relationship between chariot or person and their parts, except as a convention. In MA 6.151, Candrakīrti systematically analyses the possibilities for conceptualising the relationship of a chariot to its parts: The idea here is strictly to use the example of a chariot as an analogy, to better analyse the relationship of the self or person to its constituents. Each of the seven possibilities offer a way to come to understand how it cannot explain the relationship of the chariot to its parts. 31 Of particular interest in this regard is possibility [7], that a chariot is a functional arrangement of its parts. This has probably always seemed like the strongest candidate for explaining the relationship of a chariot to its parts. In MA 6.153-4 Candrakīrti offers an ingenious argument against it. He imagines an opponent who wishes to argue that a chariot is a functional arrangement of its parts, and infers from this argument that that person must hold that there was a pre-existing idea of a chariot, which was the basis of the subsequent arrangement of its parts into a functioning chariot: According to you, the idea of a chariot pre-existed as an arrangement among its parts understood individually. But if that is the case, just as among those separate parts, in the same way a chariot does not exist even now. 32 That is to say, before the parts of the chariot were assembled into a functional arrangement, the chariot was an abstraction ('chariotness', rathatā), an idea. But, when the parts are put together into a functional arrangement, has anything about the parts 30 MA 6.151: svāṅgebhya iṣṭo na ratho yathānyo na cāpy ananyo na ca. nāma tadvān | nāṅgeṣu nāṅgāny api tatra nāpi saṃghātamātraṃ na ca. sanniveśaḥ ||. 31 On the background to these seven possibilities, see Kapstein (2001: 102). 32 MA 6.153: saṃsthānam aṅgeṣu yathā purābhūt pratyekaśas te rathatāṃ gateṣu | tathaiva cen nāsti ratho 'dhunāpi viśliṣṭabhūteṣu yathaiva teṣu ||. changed? Someone might argue that, indeed, something has changed -the parts are now in a functional arrangement. To which Candrakīrti replies: If right now there is an arrangement of wheels and so forth that is distinct [from the arrangement] during the space of time one had the idea of a chariot, one should indeed perceive that this is the case. But this [distinction] does not exist. Therefore the chariot is not simply an arrangement. 33 Candrakīrti claims that the opponent must suppose that there is some difference between the parts before they are arranged in a functional way and once they are so arranged, so that this difference is the difference between the mere idea of a chariot and there actually being a chariot. One might add: at what exact moment in the construction process do the parts, which are not yet a chariot, become the parts of a chariot? The answer is, that there is no such moment. 34 In this way and in others, Candrakīrti undermines through analysis any attempt to conceptualise the relationship of a chariot to its parts. And yet: Although it has been proven in seven ways not to exist in reality nor in ordinary experience, nevertheless, in ordinary experience, without investigation, it is conventionally spoken of as depending on its parts. 35 That is to say, if we do not analyse it in a philosophical way, it is perfectly reasonable to speak about the existence of a chariot: 'It has parts, it has components, it is a [grammatical] agentjust that is a chariot' -this is human communication.
It is what has proven to be an appropriate usage among people. One should not destroy the world's proven conventions. 36 Talk of a chariot is totally unproblematic in ordinary life, indeed necessary, even though such talk does not bear any serious philosophical analysis.
Candrakīrti now applies this same line of reasoning analogically to the ordinary sense of self, which is the referent of ego-identification: Through humanity's belief, the self too is thus accepted in respect of being an appropriate usage, taking as its basis 33 MA 6.154: saṃsthānabhedo yadi cādhunāsti cakrādikasyeha rathatvakāle | gṛhyeta nāmaiṣa na caitad asti saṃsthānamātraṃ na ratho 'sti tasmāt ||. 34 This argument is also discussed in Kapstein (2001: 102 the constituents and elements, and so also the sixfold sphere, and it is also an agent [dependent on] the appropriation of action. 37 It has neither impermanence nor permanence, it is not born nor is it destroyed, neither does its eternity and so on exist. Because of not being a real entity it has no reality nor otherwise. 38 The self is that which continually manifests to humanity, in respect of which there is always the conviction of ego-identification, and that in respect of which its comprehension of identifying with what is 'mine' arises from confusion and from a lack of intelligent investigation. 39 Our use of pronouns such as "I" and "me" is like their use in fictional discourse (Garfield, 2015: 114), which, when one analyses them, turn out to have no referents. It is not possible to get satisfying answers to philosophical questions about these pronouns or their referents, and yet ordinary human life takes them for granted. However, the meditator who analyses the ordinary sense of self and realises that it is empty of intrinsic existence is able to attain liberation: Since without agency action does not exist, hence without the self, what belongs to a self does not exist. Hence, seeing self and what belongs to a self as empty that meditator attains liberation. 40 With this in mind, it becomes possible to see how Candrakīrti would interpret the canonical source of the chariot argument. The bhikṣunī Śailā finds herself assailed by Mārā, and a number of challenging philosophical and existential questions arise in her mind concerning her own existence as a self or person. Her reply to Mārā is that of a meditator or yoginī who is well-practised in the kind of analytic reflection that Candrakīrti has outlined: she compares the sense of self to a chariot, that exists conventionally in dependence on its parts. By doing so, she does not engage in ontological speculation of the Abhidharmic sort, but rather she directly evokes the empti-37 MA 6.162: ātmāpy upādātṛtayā tatheṣṭaḥ skandhān upādāya jagatpratītyā | dhātūṃs tathā cāyatanāni ṣaḍḍhā karmāpy upādānam asau ca. kartā ||. 38  ness of the self while appreciating how it exists conventionally, although not in the manner it appears. 41 Rochard (2012: 188-193) concludes her PhD thesis on Candrakīrti's analysis of the self with a discussion of Candrakīrti's use of philosophy as a form of spiritual practice. This is an important theme for distinguishing the aim of Candrakīrti's use of the chariot argument from western philosophical arguments about personal identity, theories of persons, and so on. This gives me a further way to summarise the use and misuse of the chariot argument: its correct use is as an analytic meditation, and this is how we should understand the canonical presentation of the argument by bhikṣunī Śailā:

Conclusion
The investigation undertaken by arhantī Śailā is a prime example of using philosophy as meditation practice. Clearly, while she was meditating, certain thoughts arose -identified in the sūtra as Mara -which she dissolved with meditative analysis. Just as her response is not characterised as presenting a thesis, so Nāgārjuna says, "I have no thesis." 42 This could be taken to imply that it would be a misuse of the chariot argument to try to prove anything, as is the case in the Milindapañha.
However, Candrakīrti's philosophical approach is not completely innocent of the cut-and-thrust of debate. In the stanzas of the Madhyamakāvatāra just prior to the 'refutation of the self' (MA 120-165), Candakīrti explains the purpose and method of the investigation he is about to make: The wise say that the coming to an end of conceptual constructions is the result of investigation. Ordinary people are bound simply by conceptions; the meditator who does not conceptualise attains liberation. 43 Conceptual constructions (kalpanā) would include deep-rooted and habitual ways of conceptualising the ordinary sense of self as existing in just the way it appears. But this is a difficult topic to investigate, and the investigation will include a critical examination of other beliefs, including the beliefs of other Buddhist schools and philosophers: 41 Candrakīrti's use of the chariot argument in his Prasannapadā on Mūlamadhyamakakārikā 24.18 concerns the concept of emptiness (śūnyatā) rather than the self, but rehearses exactly the same approach. Just as one can know a chariot to be dependent on its parts, and not through its intrinsic existence, likewise one can know emptiness as a dependent designation, not through its intrinsic existence. 42 Rochard (2012: 189) The investigation is not carried out from love of debate but the reality talked of in the philosophical texts is for the sake of liberation. If when critically discussing this reality the beliefs of others are refuted -it is not the fault [of the investigation]. 44 Candrakīrti's attitude here seems to be that some engaged philosophical argument is a healthy part of the soteriological aim of the whole enterprise. It could even be said to be necessary, because human beings tend to be attached to views: For compulsion towards one's own philosophical views as well as hatred towards the views of others is just conceptual construction. Therefore through relinquishing compulsion and aversion the one who investigates quickly attains liberation. 45 Therefore, the investigation of the self that Candrakīrti undertakes is not only a philosophical meditation. This very meditation necessarily also involves the refutation of rival views, and perhaps even involves debate. However, the purpose of this theoretical activity remains liberation, not the establishment of a view or theory as correct. 46 It is in this context of critical philosophical debate and analysis, aiming at liberation, that Candrakīrti makes reference, in Madhyamakāvatāra, to the Milindapañha, in order to dispute its use of the chariot argument, and to put forward his own use of it in a way that will lead to liberation. 47 Funding No funds, grants, or other support was received.

Competing interests
The author has no relevant financial or non-financial interests to disclose.
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/. 44 MA 6.118: na vādalobhād vihito vicāras tattvaṃ tu śāstre kathitaṃ vimuktyai | vyākhyāyamāne yadi nāma tattve bhidāṃ gatāny anyamatāny adoṣaḥ ||. 45 MA 6.119: svadṛṣṭirāgo'pi hi kalpanaiva tathānyadṛṣṭāv api yaś ca. roṣaḥ | vidhūya rāgaṃ pratighaṃ ca. tasmād vicārayan kṣipram upaiti muktim ||. 46 All this invites further reflection on ways in which Buddhist philosophy may be interpreted as a 'way of life', as Pierre Hadot has interpreted philosophy in the ancient world; a topic already addressed by Kapstein (2001: 3-20) and Fiordalis (2018). 47 I take this opportunity to thank Dechen Susan Rochard for her astute comments on an earlier draft of this article, which helped me more properly define and refine my argument. I also thank the anonymous reviewer of an earlier draft of this article for some stringent comments that have helped to improve it.