Reformulating the Buddhist Free Will Problem: Why There can be no Definitive Solution

In recent years, scholars have become increasingly interested in reconstructing a Buddhist stance on the free will problem. Since then, Buddhism has been variously described as implicitly hard determinist, paleo-compatibilist, neo-compatibilist and libertarian. Some scholars, however, question the legitimacy of Buddhist free will theorizing, arguing that Buddhism does not share sufficiently many presuppositions required to articulate the problem. This paper argues that, though Buddhist and Western versions of the free will problem are not perfectly isomorphic, a problem analogous to that expressed in Western philosophy emerges within the Buddhist framework. This analogous problem concerns the difficulty of explaining karmic responsibility in a world governed by dependent origination. This paper seeks to reconstruct an approach to free will consistent with Madhyamaka philosophy and, in so doing, to demonstrate that the mutual exclusivity of positions such as hard determinism and libertarianism is, from the Madhyamaka perspective, merely superficial. By building on the perspectivalist theory advanced by Daniel Breyer, it is clear that a Madhyamaka stance on free will demands the wholesale abandonment of perspectives, such that the idea of any one solution as definitive is disavowed. Taken to its logical conclusion, therefore, perspectivalism entails the relative truth of perspectivalism itself.


Introduction
In a rare instance of scholarly consensus, it is generally agreed that proponents of classical Buddhism did not recognise what, in Western philosophy, is known as the free will problem. Discussion of the Buddhist stance on free will has therefore emerged only relatively recently, during the late nineteenth century (e.g. Rhys Davids 1898). Though classical sources do not explicitly discuss free will, many texts display an interest in the implications of Buddhist causal theory for moral responsibility. For example, in the Bodhicaryāvatāra (hereafter BCA), Ś āntideva explores the relationship between the theory of dependent origination (pratītyasamutpāda) and ethical accountability, especially in chapter six. The Sāmaññaphala sutta also indicates that the historical Buddha's contemporaries were interested in whether people could be morally responsible for their actions. 1 Accordingly, reconstructivists appeal to these and similar texts in their efforts to defend competing interpretations of the Buddhist view on free and responsible action in a world governed by impersonal causal operations (e.g. Goodman 2009;Federman 2010).
To avoid the dangers of exegetical interpolation and/ or the superimposition of Western conceptual paradigms onto classical Buddhist texts, however, reconstructivists must be mindful of the distinctive cultural context out of which such works emerged (Tuck 1990, pp. 8-16). Indeed, much Buddhist literature is characterized by argumentative strategies and methodological practices different from those typically found in the Western corpus. Further, as is widely acknowledged, traditional Buddhists have not examined the implications of causal determinism or indeterminism for responsible agency. As one of the first scholars to discuss Buddhism and free will, Horner has commented that the will is simply "assumed to be free" (Horner 1938, p. xxiii).
Nevertheless, this paper maintains that a problem sufficiently analogous to the Western free will problem does arise in the Buddhist context. This analogous problem emerges from a tension between the theories of karma and dependent origination. In contradistinction from those who claim that a free will problem cannot be expressed in Buddhist terms (e.g. Conze 1963;Rahula 1959;Garfield 2017), this paper argues that evaluating the conceptual coherence of Buddhism's commitment to karmic responsibility and dependent origination is a philosophically important undertaking.
Early Buddhist teachings present karma as both a causal and a moral force. In its basic formulation the teaching stipulates that people are responsible for their thoughts and deeds-hence the Buddha's dictum "karma is intention." 2 How best to interpret the teaching of karma has been a long-standing source of contention amongst Buddhists. As McDermott has discussed, early Buddhists perceived the tension between karma's ethical and soteriological functions (McDermott 1973, p. 344). Moreover, different sects also appear to have assigned different degrees of importance to the role of intention (McDermott 1975, p. 429). Despite competing conceptions of karma within Buddhism, the teaching essentially affirms that all intentional acts entail consequences, which will rebound upon their agent at some point in the future. 3 Like the teaching of karma, that of dependent origination is multi-dimensional and can be expressed with varying degrees of sophistication. What precisely the teaching involves has similarly been a matter of intense dispute between the different schools of Buddhism (evidenced by fierce debates between Abhidharma and Madhyamaka theorists over the ontological status of dependent entities). Despite this, the teaching of dependent origination has rightly been described as the "common denominator" of all forms of Buddhism (Boisvert 1995, p. 6).
According to early Buddhism, all events are causally conditioned by prior events. The nature of all saṃsāric experience is dependent, such that: "When this exists, that comes to be; with the arising of this, that arises. When this does not exist, that does not come to be; with the cessation of this, that ceases." 4 Madhyamaka has a distinctive interpretation of this theory, which Nāgārjuna enigmatically communicates at Mūlamadhyamakakārikā 24.18 (hereafter MMK). 5 For Madhyamaka, whatever arises in dependence on something else is necessarily lacking in inherent reality (svabhāva) and, for precisely this reason, is capable of undergoing transformation. The spiritual resonances of dependency are held by Mādhyamikas to be profound, since it is by virtue of being dependently arisen that the teachings of the Buddha are soteriologically effective. 6 The various ways in which Buddhists conceive of dependency relations informs the expression and alleviation of the tension between pratītyasamutpāda and karma. However, all Buddhists concur that what is dependently originated cannot be the 3 The Buddhist teaching of no self (anātman) precludes the idea of an agent as a substantial or metaphysically independent locus of activity. Nevertheless, by invoking a distinction between what is ultimately the case and what is merely conventionally so, Buddhists profess the reality of persons/ agents at a conventional level. They are thus able to secure the idea of identity over time without which the idea of karmic consequences rebounding upon the person who performed the initial action would be nonsensical. 4 Majjhima Nikāya 79, Cūlasakuludāyi sutta. Pāli Text Society (PTS), Chalmers (1896), p. 31. "Imasmiṃ sati, idaṃ hoti; imass' uppādā idaṃ upapajjati; imasmiṃ asati, idaṃ na hoti; imassa nirodhā imaṃ nirujjhatīti." (Trans.) Bhikkhu Ñ āṅamoli and Bhikkhu Bodhi (1995), p. 655. 5 The Madhyamaka conception of dependent origination is discussed more fully in subsequent sections of this article, as are the various interpretations of this pivotal, and much contested, verse of the MMK. 6 Such is Nāgārjuna's view, conveyed at MMK 24.14. sarvaṃ ca yujyate tasya śūnyatā yasya yujyate/ sarvaṃ na yujyate tasya śūnyaṃ yasya na yujyate// "All is possible when emptiness is possible. Nothing is possible when emptiness is impossible." (Trans.) Siderits, M. Katsura, S. (2013), p. 276. Candrakīrti's comments on this verse are also revealing-for him the efficacy of the three jewels is secured by their being empty of inherent reality. He maintains that only what is empty is productive of change and capable of achieving results. Thus he writes: "tataśca trīṇyapi ratnāni yujyante" (de La Vallée Poussin, p. 501, line 4). self (ātman) and so is impersonal. Taken together, therefore, the teachings of karma and dependent origination stimulate questions about the possibility of free will and moral responsibility. How can people be morally responsible for their thoughts and deeds if what they think and do is causally conditioned by prior events, at least some of which must count as beyond their control?
Throughout this paper, the phrase the Buddhist free will problem denotes the conceptual tension alluded to between karma and dependent origination. This phrase suggests neither that Western and Buddhist versions of the problem are perfectly isomorphic, nor that strategies effective for dispelling or resolving the problem in one context will necessarily be so in the other. Here it is argued that by posing the free will problem in terms of karma and dependent origination (a teaching which, for Madhyamaka, is not conceptually coextensive with causal determinism), new ways of approaching a perennial problem are revealed.

Buddhism and Free Will: A Reconstructive Approach
The reconstruction of a Buddhist stance on the free will problem is held by certain scholars to be a futile, if not positively reprehensible, enterprise (e.g. Garfield 2017;Flanagan 2017). Objectors resist efforts first to raise and then resolve problems not traditionally recognized by Buddhists, claiming that this involves the superimposition of culturally specific paradigms onto conceptually incommensurate systems. Repetti has responded to several charges raised against Buddhist free will reconstruction, so his defence will not be fully rehearsed here (Repetti 2017a, pp. 22-31). However, since the credibility of the reconstruction offered in this paper depends on the legitimacy of Buddhist free will theorizing in general, it is necessary to answer some of the most serious and/or sustained criticisms. Normally these assume one of the following forms: (1) the gravity of the problem is denied; (2) the problem is held to be the product of anachronistic-and by implication, viciousreasoning; (3) the problem is considered irrelevant to Buddhism, since its primary interests are soteriological.
Objections of type (1) suggest that the free will problem is merely pseudo in the Buddhist context (Conze 1963, p. 104). If there is ultimately no self (anātman), then why speculate about which properties-such as a free will-it would be endowed with? Sometimes, however, the profundity of a problem can only be determined through the process of its attempted resolution. The brevity with which Conze and others, such as Rahula (Rahula 1959, pp. 54-55), dismiss Buddhist free will theorizing as conceptually impossible is troubling, for, if anātman disqualifies Buddhists from this debate, it presumably also disqualifies them from every debate in which the notion of self features. 7 As the history of Indian philosophy attests, however, Buddhists have successfully contributed to such debates through challenging the idea that our ordinary linguistic practices reflect ultimately existent referents. Essentially, the objector considers belief in selfhood to be a prerequisite of free will theorizing whilst the reconstructivist holds that this is not so.
More recently, scholars such as Adam (Adam 2011) have explored the possibility of Buddhist free will theorizing in greater depth and with more initial optimism, yet have still concluded that, in the absence of self, there can be no free will and, hence, no problem. One response available to the reconstructivist is to invoke the Buddhist distinction between ultimate truth (paramārtha-satya) and conventional truth (saṃvṛti-satya). This distinction permits Buddhists to discuss persons, as conventionally established entities. Defenders of certain versions of compatibilism maintain that this is the only proper framework for Buddhist free will theorizing (e.g. Meyers 2013, p. 43;Siderits 2008). From the perspective of conventional truth, then, it is philosophically problematic to hold people morally responsible for their thoughts and deeds because these are the product of an interacting, interdependent and impersonal network of causes and conditions.
As a concomitant aspect of objection (1), reconstructivists are sometimes charged with making a category mistake. This criticism is encapsulated by Garfield's assertion that it is "impossible to formulate the free will thesis in a Buddhist framework" (Garfield 2017, p. 50). Similarly, Flanagan argues that the Western version of the free will problem emerges from a "very particular, parochial language game" which is non-transferable to the Buddhist context (Flanagan 2017, p. 61). This objection stems from the belief that anātman renders speculation on free will pointless. However, many contemporary Western discussions on free will proceed without recourse to the idea of the self as a necessarily existent, substantial, and immutable entity (e.g. Dennett 2003, p. 1). If modern Western philosophers can generate a free will problem without recourse to the concept of ātman, why should the idea of a Buddhist free will problem be impossible (Federman 2010, p. 8)? Garfield and Flanagan are right to observe that Buddhism cannot accommodate the free will problem as formulated in medieval Western philosophy, but to claim that Buddhism cannot face any analogous problem-one borne out of tension between two of its own doctrines-is to claim too much.
Objections of type (2) come from the widely acknowledged absence of free will theorizing in classical Buddhist sources. Gier and Kjellberg have suggested that one reason for this absence could be Buddhism's preoccupation with freedom from saṃsāric suffering, rather than with the exercise of unenlightened freedoms (Gier and Kjellberg 2004, p. 278). Objections (2) and (3) are therefore closely connected. Reconstructivists accept that Buddhist free will theorizing involves anachronism-it is because Buddhists have not typically discussed this topic that reconstruction is required. The key question, though, is whether a-historical methodologies can deliver enhanced appreciation of Buddhist thought. Reconstructivists maintain that they can and, moreover, that anachronism is not per se vicious. The idea of a nonvicious anachronism might strike some as a contradiction in terms. However, as Ganeri and Pollock have both intimated, the idea of anachronism as malignant requires a particular conception of history, which also seems to be absent from classical Indian sources (Ganeri 2011, pp. 102-103;Pollock 1989, pp. 603-605).
It is conceivable, therefore, that some instances of anachronistic reasoning will prove philosophically fruitful. The dividing line falls where Buddhist conceptual resources are invoked to settle questions with which Buddhists have not engaged as a matter of principle, and where they are invoked in the resolution of problems which as a matter of historical fact Buddhists did not address. 8 Responses to objection (1) indicate that there is no reason why in principle Buddhists should refrain from discussing free will. A further response to the charge of anachronism is available to contemporary Buddhist practitioners who might argue that, though they draw on ancient sources for inspiration, theirs is a living tradition. In consequence, the reconstructivist strives not only to articulate a position theoretically consistent with the Buddhist texts but also to provide a response with which contemporary Buddhists can resonate.
The fact that modern Buddhists may be perturbed by the prima facie incompatibility of karmic responsibility and dependent origination itself constitutes a rebuttal of objection type (3). From Gowans's conjecture that the Buddha did not discuss free will because of its irrelevance to the spiritual development of his interlocutors, it does not follow that free will theorizing remains superfluous (Gowans 2017). Indeed, a crucial feature of the Buddhist pedagogical technique of skill-in-means (upāya) is that the relevance and benefit of a teaching is determined on an individual basis. What counts as soteriologically irrelevant-as nothing other than metaphysical speculation for its own sake-in one historical context may not necessarily be so in another (Repetti 2017a, p. 24).
Finally, before turning to a consideration of how Madhyamaka might approach the Buddhist free will problem, a word on reductionism is in order. Western philosophy has, throughout its history, offered a vast array of responses to the free will problem. Since Buddhist philosophy is not monolithic, efforts to establish the Buddhist position on free will are probably misguided. A fully worked out Theravāda position would be different from a similarly robust Tibetan account (e.g. Harvey 2017; McRea 2017). Each Buddhist school nuances its presentation of metaphysics, epistemology and semantics to build a coherent worldview. Accordingly, whilst the various schools might agree on the core aspects of karma and dependent origination, it seems impossible to provide a pan-Buddhist stance on free will. The reconstructivist may therefore either attempt a broad-brush response to the problem or provide a tradition-specific response.
In anticipation that the latter approach will yield both more interesting and more exacting results, this paper sets out to articulate a response viable for the Madhyamaka system. Madhyamaka's distinctive approach to the doctrine of two truths, coupled with its refined analysis of what it means for something to be dependently originated, facilitates new ways of thinking about the nature of-and potential responses to-the Buddhist free will problem. Since Madhyamaka, like Buddhism as a whole, is divisible into sub-schools, it could be objected that efforts to express the Madhyamaka view are equally misguided. In one sense this is true, not least because a key part of the Madhyamaka strategy is to challenge notions of fixity or absoluteness wherever they arise. 9 However, there is considerable convergence of opinion amongst Mādhyamikas interested in the explication of Nāgārjuna's Madhyamaka. This paper therefore primarily focuses on the implications of Prāsaṅgika views for free will and draws particularly on the works of such figures as Ā ryadeva, Candrakīrti, Ś āntideva and Prajñākaramati.

Anticipating a Madhyamaka Response
A comparison of the early and more recent discussions on Buddhism and free will reveals the extent of the progress so far. This is clear, for example, when we contrast Keith's assessment that in Pāli Buddhism "the issue is solved by the simple process of ignoring it" (Keith 1923, p. 116) with the array of competing theories now available. Attendant upon the proliferation of interest in this topic has been the defence of every position on the spectrum between hard determinism and libertarianism as that implicitly endorsed by Buddhism (see Repetti's 2017 volume for an overview). The differences between the various schools of Buddhism notwithstanding, the idea of such profound internal disagreement on so central a question as whether people are morally responsible is dubious. It is natural to wonder, therefore, if divergent reconstructions reflect the views of particular reconstructivists more than they reflect the implied Buddhist position. Part of what drives the insatiable appetite for competing theories on Buddhism and free will, however, is the misguided assumption that reconstructivism can deliver a definitive solution to the problem.
The error encoded in this way of thinking is given sharper focus in Madhyamaka, where the very notion of an absolutely right perspective is rejected wholesale. Even Mādhyamikas who, like Candrakīrti, endorse the distinction between non-definitive (neyārtha) and so-called definitive (nītārtha) teachings would be inclined to see this distinction as itself nothing more than a pedagogical device. 10 If, as some scholars suggest (e.g. Burton 1999, p. 97), the neyārtha/ nītārtha distinction parallels the conventional/ ultimate distinction, then, since the latter eventually breaks down in the Madhyamaka system, teachings can only be respectively better or worse, rather than definitively correct. Nāgārjuna intimates the parallelism of these distinctions in verses 56-57 of his Acintyastava, suggesting that teachings on emptiness reflect ultimate truth whilst those on other topics require interpretation. 11

Neyārtha and Nītārtha: A Provisional Distinction
The neyārtha/ nītārtha distinction is also important to Buddhism's overarching pedagogical technique of upāya. In the Prasannapadā (hereafter PsP), Candrakīrti appeals to the provisional and definitive qualities of different teachings to explain how seemingly incomplete instructions do not thereby lack authenticity. Commenting on MMK 18.5, Candrakīrti suggests that previous teachings are incomplete-such as the teaching on the insubstantiality of persons. The Mahāyāna teachings expand on earlier themes, taking discussions on the three characteristics of saṃsāra to their logical conclusions and demonstrating the emptiness of all categories (niḥsvabhāva). 12 Candrakīrti discusses the graded teachings of the Buddha further in his commentary on the next verse, MMK 18.6, where he presents the theses of self, no self and neither self nor no self as successively liberating. 13 Near the end of an aspirant's spiritual transformation, however, the validity of the neyārtha/ nītārtha distinction must itself be recognized as provisional. Without this, the aspirant risks attachment to the nītārtha teaching of emptiness, which, as Nāgārjuna warns, is like grasping a venomous snake by the tail. Hence, Candrakīrti asserts that, fundamentally, the Buddha did not teach anything at all (neither self nor non-self) because the enlightened perspective disavows both referents and terms of reference equally.
Nevertheless, the pedagogical utility of this distinction is not compromised by its ultimate-level inapplicability. Rather, Buddhist texts are replete with the idea that instructional discrepancies reflect the intended audiences of different texts. The workings of upāya are thoroughly demonstrated in the Saddharmapuṇḍarīka, a key Mahāyāna source, in which the legitimacy of temporarily sustaining illusions is 11 Nāgārjuna-Acintyastava 56-57.
dharmayautukam ākhyātaṃ buddhānāṃ śāsanāmṛtam/ nītārtham iti nirdiṣṭaṃ dharmāṇāṃ śūnyataiva hi// 56 yā tūtpādanirodhādisattvajīvādideśanā/ neyārthā ca tvayā nātha bhāṣitā saṃvṛtiś ca sā// 57 "It has been declared that the ambrosia of the teaching of the buddhas, called 'the gift of the doctrine,' is of well-established meaning: (it is) only the voidness of the dharmas.
affirmed on the basis of right motivation. The text itself questions the moral integrity of the father who in the famous fire parable represents the Buddha, for enticing his children out of the burning house (saṃsāra) with the promise of desirable-yet perishable-goods. Are the enlightened sometimes justified in sustaining illusions? According to this text intention is pivotal in distinguishing between instances of deceit and skilfulness, for only in the latter cases is the liberation of others the main motive (Brill's Encyclopedia of Buddhism Vol. 1, 2015, p. 149). 14 Mādhyamikas therefore accept the necessity of sometimes employing categories which, from an enlightened perspective, are redundant, but which nevertheless attenuate aspirants' suffering, leading them towards an understanding of emptiness. 15 The relevance of this insight to the reconstruction of Madhyamaka's stance on free will should become apparent in the subsequent sections of this paper.

Madhyamaka's Abandonment of Views
As indicated, Nāgārjuna emphasizes the danger associated with taking any proposition as expressive of ultimate or definitive truth. He warns at MMK 24.11 that (if poorly understood) even the theory of emptiness can become an object of attachment, capable of destroying "the slow-witted, like a serpent wrongly held or a spell wrongly executed." 16 The danger of misunderstanding emptiness is a dominant theme in Madhyamaka literature. This idea is attested by Ā ryadeva, who invokes the idea of Buddhist teachings as therapeutic. In order for emptiness to work as a medicine it must be correctly prescribed: he says, "someone who desires merit should not speak about emptiness at all times… medicine that is unsuitably applied becomes poison." 17 Candrakīrti's remarks on the immediately preceding verse are 14 The famous fire parable occurs in chapter three of the Saddharmapuṇḍarīka. Immediately after the father lures his children from the burning house on the pretext of giving them wonderful gifts of animaldrawn carriages, the question of whether he has behaved deceitfully arises: "tat kiṃ manyase Śāriputra mā haiva tasya puruṣasya mṛṣāvādaḥ syād yena teṣāṃ dārakāṇāṃ pūrvaṃ trīṇi yānāny upadarśayitvā paścāt sarveṣāṃ mahāyānāny eva dattāny udārayānāny eva dattāni" (Kern and Nanjio 1970, p. 76). 15 Such is precisely Ś āntideva's point at BCA 9.76. yadi sattvo na vidyeta kasyopari kṛpeti cet/ kāryārtham abhyupetena yo mohena prakalpitaḥ// "[Objection] Whose is the task to be done, if there is no being?
[Mādhyamika] True. Moreover, the effort is made in delusion, but, in order to bring about an end to suffering, the delusion of what has to be done is not prevented." (Trans.) Crosby and Skilton (1996) also illuminating, as he asserts that attachment to the Buddhist teachings (dharma) fetters people to saṃsāra. 18 Madhyamaka's resistance to the idea of definitive solutions to philosophical problems is expressed at Vigrahavyāvartanī 29 (hereafter VV), where Nāgārjuna affirms that he holds no thesis (pratijñā) of his own. He suggests that any criticism brought against Madhyamaka is aiming at the wrong target if the opponent seeks to undermine the foundations of Madhyamaka reasoning-since Mādhyamikas themselves reject foundationalism. Both Ruegg and Westerhoff interpret Nāgārjuna's statement at VV 29 as requiring qualification to avoid the charge of selfcontradiction (Ruegg 1986, p. 232;Westerhoff 2009, p. 25). Since the no-thesis claim seems itself to be a thesis, and given that Candrakīrti describes some of Nāgārjuna's points as pratijñās, for VV 29 to make sense the rejected pratijñās must be postulated as inherently real (i.e. they are niḥsvabhāva). 19 For all these reasons, therefore, the idea of a definitive solution to the Buddhist free will problem would be anathema to Mādhyamikas. This, however, does not preclude the successful reconstruction of a philosophically, psychologically and soteriologically integrated Madhyamaka stance on free will. Some modern exegetes judge that Svātantrika Mādhyamikas alone accept logical entailment at the conventional level (e.g. Siderits 2010; Tillemans 2010). This judgement arises from: (a) Candrakīrti's vehement attack on Bhāvaviveka's use of the qualifier 'ultimately' in denying the reality of causal relations; 20 and (b) Candrakīrti's disparaging remarks about conventional truth, which are well-documented in contemporary scholarship (see The Cowherds 2011). 21

Degrees of conventional truth
Writers like Tillemans (2010) rightly point out that if Prāsaṅgika Mādhyamikas conceive of conventional truth in terms of mere opinion, theirs is indeed a dismal 18 Candrakīrti-Catuḥśatakavṛtti 8.17.
evaṃ pratijñātrayam api yojyam/ "The remaining triad of propositions (pratijñā) ['from other,' 'from both' and 'without cause'] is to be connected in that same manner." (Trans.) MacDonald (2015), p. 50. 20 Commenting on MMK 1.1, Candrakīrti attacks Bhāvaviveka's use of the word 'ultimately' in ruling out the possibility of self-causation. In the first place, such a theory is not even established according to conventional-worldly-usage and, secondly, such a qualification amounts to an assertion at the conventional level. This is anathema to Madhyamaka's exclusive use of reductio ad absurdum methods. (MacDonald pp. 100-101). 21 For example, commenting on MMK 24.8, Candrakīrti offers three definitions of conventional truth as: (1) that which conceals, i.e. ignorance; (2) that which depends on something else, i.e. is relational and; (3) that which has worldly import and is suitable for transactional usage (Nagao 1991, pp. 14-15). Candrakīrti's remarks in chapter 6 of his Madhyamakāvatāra are also sometimes read as entailing epistemic pessimism. Whilst scholars have strong views about how to understand Candrakīrti's attitude towards conventional truth, the texts themselves attest to ambiguity. epistemic enterprise. 22 However, other scholars present more charitable and optimistic readings of Prāsaṅgika understandings of conventional truth-which, in the absence of the ultimate, transpires to be the only kind of truth (Salvini 2008, p. 156). These more charitable readings of Candrakīrti reflect the ambiguity in his remarks and allow us to make sense of the important role he ascribes to skill-inmeans. If Madhyamaka concedes conventional truths to be the only kind, must they also view all conventions as equally sophisticated? To use Garfield's example (Garfield 2015, p. 28), botanists and farmers alike know that certain plants yield certain crops. Identifying the plant's genetic code gives the botanist deeper understanding of why this happens without it following that the botanist ascertains an objective correlation between truth and world. Though Mādhyamikas would reject the possibility of a definitive solution to the Buddhist free will problem, they might concede that some reconstructions are soteriologically more sophisticated than others.
Though certain scholars consider the idea of a spectrum of truth within the domain of convention to be an exclusively Svātantrika view, this is not so. The Madhyamakahṛdayakārikā 3.12-13 makes clear Bhāvaviveka's endorsement of a hierarchy of conventions. For him, the ultimate truth of emptiness is only realized via ascending through successive levels of convention. 23 However, in classifying Indian Mādhyamikas as either Svātantrikas or Prāsaṅgikas, it is important to remember that this way of systematizing Madhyamaka reflects the retrospective doxography of later Tibetans (Dreyfus and McClintock 2003, p. 3). It is therefore not surprising to find that Prāsaṅgikas sometimes defend views typically associated with Svātantrika. Although Candrakīrti rejects conventional svabhāva (which he regards as a contradiction in terms), he can also be read as rejecting a plateau of conventions. All Mādhyamikas accept that the correct apprehension of the conventional is the sine qua non of liberation, but, if conventional truths are the only type of truths, Prāsaṅgikas must distinguish between the understanding of worldly people and the spiritually adept. 24 This is achieved through accommodating a hierarchy of conventional insights. At Madhyamakāvatāra 6.27-29 (hereafter MAV), Candrakīrti contrasts the stainless mental faculties of those who are 22  tattva-prāsāda-śikharārohaṇaṃ na hi yujyate/ tathya-saṃvṛti-sopānam antareṇa yatas tatḥā// pūrvaṃ saṃvṛti-satyena pravivikta-matir bhavet/ tato dharma-sva-sāmānya-lakṣaṇe suviniścitaḥ// "Also, without the stairs of the tathya-saṃvṛti, the ascension to the top of the palace of ultimate reality (tattva) is lost. For this very reason, firstly, the saṃvṛti-satya should be discerned by the intelligence. Thence, one proceeds to investigate thoroughly the unique and universal characteristics of dharmas." (Trans.) , p. 67. 24 Nāgārjuna-MMK 24.10.
vyavahāram anāśritya paramārtha na deśyate/ paramārtham anāgamya nirvāṇaṃ nādhigamyate// "The ultimate truth is not taught independently of customary ways of talking and thinking. Not having acquired the ultimate truth, nirvāṇa is not attained." (Trans.) Siderits and Katsura (2013), p. 273. enlightened with the defiled faculties of the unenlightened, saying that the distinction between the two truths should be understood analogously. 25 However, since for him no ultimate truths can be instantiated, a reasoned interpretation of his position is that he permits a gradation of conventional truth. This explains the importance of pursuing a middle path-where one's mode of apprehending the world is gradually transformed. Other Prāsaṅgikas likewise seem untroubled by the possibility of levels of conventional truth. Prajñākaramati seemingly interprets the BCA 9.4 in this way, endorsing the idea of increasingly refined conventions for those on the path to full Buddhahood. 26 Prajñākaramati even quotes Candrakīrti's MAV to adduce support for the idea that yogins may be refuted by those still further along the path (i.e. superior yogins). 27 Textual evidence supports the view that some Mādhyamikas accept increasingly sophisticated expressions of conventional truth even though, as a matter of philosophical principle, they deny that any statement expresses ultimate meaning. The inner logic of this worldview demands that, since spectrums too have their upper and lower limits, Mādhyamikas embrace the notion of a limitless continuum of degrees of truth, which is always relative to an individual's current level of insight.
To summarize the findings so far, if Madhyamaka's arguments on the pervasiveness of emptiness succeed, there are reasons to interpret truth as relative. The dynamic relationship between a truth and the one for whom it is a truth precludes the possibility of arriving at a definitive solution to the free will problem. Though no statement counts as unqualifiedly true, the idea of soteriologically successful practice grounds Madhyamaka notions of truth. In addition to distinguishing conventional and ultimate truth, Madhyamaka distinguishes conventional truth and falsity, or veridical and illusory experience within the domain of 25 Candrakīrti-MAV 6.29.
"Under the influence of ophthalmia one forms a false image of hairs and so forth, while an unimpaired eye spontaneously perceives what is real. [The distinction between the two truths] must be understood in an analogous fashion." (Trans.) Huntington (1989), pp. 160-161. 26 Ś āntideva-BCA 9.4. bādhyante dhīviśeṣeṇa yogino 'py uttarottaraiḥ / dṛṣṭāntenobhayeṣṭena kāryārtham avicārataḥ// "Even the views of the spiritually developed are invalidated by the superior understanding of those at successively higher levels, by means of an analogy which is accepted by both parties, irrespective of what they intend to prove." (Trans.) Crosby and Skilton (1996), p. 115. Oldmeadow translates Prajñākaramati's remarks on this verse as follows: "Yogins are also refuted by yogins one after another… By successively greater ones who have obtained the eminence of acquiring superior qualities existing in greater and greater degrees is the meaning." (Oldmeadow 1994, p. 33). 27 Prajñākaramati quotes Candrakīrti's MAV (6. 27) in order to argue that there are different levels of insight amongst yogins who are not yet fully awakened. This verse is sometimes appealed to by scholars who view Candrakīrti as epistemically unambitious (Tillemans 2010). However, Prajñākaramati-a fellow Prāsaṅgika-does not necessarily read the verse in this way (Oldmeadow 1994, p. 33). On the other hand, Nagao takes Prajñākaramati as following Candrakīrti in his assessment that the realm of conventions is strictly the realm of "delusions" (Nagao 1991, p. 16). In summary, scholars on both sides have ample textual backing for whichever interpretation they want to pursue. convention. In any ultimate sense, however, Madhyamaka considers all statements as meaningless.
Only under the rubric of a hierarchy of truth can the mutual exclusivity of the resolutions proposed so far to the Buddhist free will problem be impugned. If the reconstructivist takes seriously the Madhyamaka insistence that all views are to be abandoned, then eventual abandonment of perspectives on free will is also necessary -including the perspectivalist theory examined below. The soteriological goal of Buddhism informs its philosophy to the core, so that freedom, rather than merely theorizing about freedom, is the objective. Abandonment of views is identified early on in the Buddhist tradition (Collins 1982, pp. 120-123) as necessary for the attainment of spiritual freedom but resonates with particular force in the Madhyamaka tradition (Westerhoff 2009). If adopting any metaphysical perspective as definitive is, as Nāgārjuna maintains, an impediment to the cultivation of the Buddhist path, then, presumably, all such perspectives must be relinquished. Madhyamaka would therefore regard resolute commitment to any stance on free will as cognitively unwarranted and morally debilitating. Through abandoning the idea that beliefs reflect objective reality, people may realize that apparently rival theories in fact express different degrees of insight. It follows that these 'conflicting' perspectives on free will are, in a more profound sense, harmonious. The task of harmonizing such theories as hard determinism and compatibilism, however, constitutes a psychological as well as a philosophical challenge. The best hope of meeting it lies in harnessing the force of a hierarchical notion of truth.

The Buddhist Free will Problem: Karma and Dependent Origination
Having argued that the Buddhist free will problem emerges from the strained relationship between the teachings of karma and dependent origination, it is necessary to explore these theories more fully. As Buddhist traditions themselves attest, teachings on karma can be interpreted in a multiplicity of ways. In its basic form, the theory states that people are morally responsible for their intentional states (including mental attitudes, verbal expressions and physical actions). At some point in the future a person's karma will come to fruition, so that the consequences of action rebound upon their agent. Modern Buddhists sometimes try to interpret this teaching in metaphorical and/ or psychological terms (e.g. Batchelor 2017, p. 118). Goodman demonstrates that the idea of karma as a psychological process is often implicit, and sometimes explicit, in classical Buddhist texts (Goodman 2017b, pp. 135-136). Similarly, Hiem identifies instances of terminological overlap in the use of cetanā and saṅkhāra, suggesting that early Buddhists recognized the power of intention in shaping experience (Hiem 2014, p. 43 (2012), pp. 536-537. "kāye vā bhikkhave sati kāyasañcetanāhetu uppajjati ajjhattaṃ sukhadukkhaṃ…kāyasaṅkhāram abhisaṅkharoti…"/ "when there is the body, then because of bodily volition pleasure and pain arise internally…one performs that bodily volitional activity…" idea by emphasizing the conceptually constructed nature of all phenomena. Central to all of the many different ways in which Buddhists have analysed karma, however, are the notions of moral choice and responsibility.
Teachings on dependent origination maintain that all events arise dependent upon prior conditions, which are impersonal. The impersonal nature of whatever is dependently arisen (including human action) is assured by (i) Buddhism's denial of any ātman and (ii) the tradition's assessment that dependency relations unfold at both the level of individual dharmas, i.e. on a micro scale, and on a cosmological level of world cycles, i.e. on a macro scale (Masefield 1983). As with karma theory, there are less and more sophisticated ways of understanding dependent origination. What even basic formulations indicate, however, is that there is something seemingly problematic about holding people morally responsible for their karma in a world governed-at every level-by the impersonal process of dependent origination.

Responsibility for Infinitely Many Intentions
In brief, can people be morally responsible for their thoughts, words and deeds if these arise from impersonal causes and conditions, which stretch back infinitely? According to Galen Strawson, moral responsibility is impossible under such circumstances. Strawson maintains that the truth or falsity of causal determinism is irrelevant to the question of moral responsibility because there is a 'basic' argument undermining the possibility of moral responsibility, rendering the very notion absurd. The thrust of the argument is that, for a person to count as deeply morally responsible, they would have to be sui generis (Strawson 1994, p. 5). However, since self-creation is impossible, Strawson argues that nobody can ever be responsible. Deep responsibility requires the ability to select one's initial mental state, from which all subsequent responsible decisions would follow. Strawson's argument is potentially damaging to accounts of karmic responsibility, especially because Buddhists believe that all beings have had infinitely many past lives. How can a person be responsible for their present mental state if they never had the opportunity to select their first mental state, or, indeed, if the very idea of a first such state is denied?
Repetti answers the 'basic' argument by suggesting that deep moral responsibility is neither necessary nor possible. If moral responsibility requires self-creation, free will theorists should revisit-and deflate-this unworkable concept (Repetti 2015, pp. 88-89). Whereas Strawson presents freedom as something individuals either have essentially or do not have at all, Repetti argues that freedom admits of degrees and that by undertaking meditational practices, individuals may come to acquire it. Such an interpretation of what it means to be free accords well with the Buddhist emphasis on the need to restructure cognitive reflexes and refine moral behaviour in accordance with the apprehension of no self/ pervasive emptiness. Though arguments against the existence of a substantial self would lead Buddhists to forfeit the possibility of deep responsibility, the question remains as to whether a person could ever freely choose to adopt a programme of meditation.
Despite the problem alluded to here, traditional Buddhists may find the idea of an inherent tension between such core concepts strange, especially given that both teachings perform the indispensable function of explaining the quality and structure of experience. The prima facie conflict between these teachings is obscured by the fact that, traditionally, dependent origination is conceived of as encompassing karma. By what logic can specific instances of the causal mechanism-such as the karmic mechanism-conceptually conflict with the chain of dependency relations taken as a whole?
Buddhists generally assume that karmic operations enable spiritual attainments, allowing for the cultivation of merit. On this view, the workings of karma provide opportunities for moral and spiritual development. At MMK 23, Nāgārjuna argues that whatever is dependently arisen lacks inherent reality and is therefore malleable. Like everything else, intentional/ karmic states are dependently arisen and, as such, are impermanent. The soteriological upshot of this is that moral defilements are not inherently real and can be successfully expunged from the mind through overhauling false conceptions. 29 Nevertheless, if a person's intentional states are the product (1) of their psychological state and (2) the external conditions of their life, then the Buddhist free will problem is unavoidable.

Meditation and Moral Motivation
Mindfulness may be a helpful tool enabling people to achieve greater control over their mental states, so that they are not merely enslaved by their passions (kleśas). Insofar as meditation promotes reflective self-regulation, it is perhaps the key to acquiring freedom from the destructive habits so characteristic of saṃsāra. Buddhists hold that meditational insights confer soteriological advantages when applied to the task of restructuring cognitive reflexes and undermining our ordinary propensity for reification. Yet the question remains: can people choose to cultivate specific mental states or is that cultivation rather something which happens to them? Buddhist scriptures explain that not everything that happens to a person is the result of their previous karma. The Moḷiyasīvaka sutta, for example, indicates that nonkarmic causal factors often shape experiences, informing both disposition and even moral capacity. 30 If the belief in infinitely many sentient beings is taken seriously, it 29 Nāgārjuna-MMK 23.24.
yadi bhūtāḥ svabhāvena kleśāḥ kecid dhi kasyacit/ kathaṃ nāma prahīyeran kaḥ svabhāvaṃ prahāsyati// "If someone had some defilements that were intrinsically real, how would they be abandoned? Who abandons intrinsic essence?" (Trans.) Siderits and Katsura (2013) Bhikkhu Bodhi (2000), p. 1279. "Tatra Sīvaka ye te samaṅabrāhmaṅā evaṁ vādino evaṁ diṫṫhino yaṁ kiñcāyam purisapuggalo paṫisamvedeti sukhaṁ vā dukkhaṁ vā adukkhamasukhaṁ vā sabbantaṁ tam pubbe katahetūti yaṁ ca sāmaṁ ñātaṁ taṁ ca atidhāvanti yaṁ ca loke saccasammataṁ taṁ ca atidhāvanti. Tasmā tesam samaṅabrāhmaṅānaṁ micchāti vadāmi." / "Now when those ascetics and bhamins hold such a doctrine and view as this, 'whatever a person experiences, whether it be pleasant or painful or neither-pleasantnor-painful, all that is caused by what was done in the past,' they overshoot what one knows by oneself seems inevitable that sometimes our personal experience will be shaped by events/ actions for which others are morally responsible. 31 Appreciation of the karmic process as a continually unfolding one further intensifies the tension at the heart of the Buddhist free will problem. Functioning simultaneously as both cause and effect, the theory of karma is supposed to explain present experience on the basis of past conduct whilst leaving scope for free action in the present moment. However, if past conduct informs present experience, can people act freely or are they compelled by their previous karma to act as they now do? Moreover, if enlightenment is a matter of degree, can people steeped in wrong views take responsibility for what they think and do? As one scholar asks, "If we are the sum of our yesterdays, how can we effect our tomorrows?" This captures the heart of the problem (Creel 1986, p. 3). Whilst Buddhism can mostly assure practitioners of the compatibility of karmic conditioning and moral responsibility, at the later stages of a person's spiritual maturation karmic operations cease to facilitate liberation and in fact begin to hinder that process. As Lusthaus's three-fold analysis of karma implies, initial belief in this process stimulates moral motivation. However, there comes a point at which liberation, rather than merit, is the goal and, then, Lusthaus argues, Buddhism "envisions a soteric project in which karma is the villain" (Lusthaus 2003, p. 168). This, presumably, is what Candrakīrti alludes to at MAV 6.42 when he avers that "a person will find liberation [only] when his mind is free from [reified concepts concerning] what is favourable and unfavourable." 32 In order to examine and alleviate the tension between karma and dependent origination, a more detailed presentation of dependency relations is required. Many scholars reconstruct a Buddhist response to the free will problem on the estimation that dependent origination and scientific causal determinism are commensurate. This is not the most effective stratagem. The advantages of assuming the parity of these ideas are outweighed by the limitations of working with the restricted account of dependent origination it affords. This is because the causal dimension of dependent origination constitutes just one aspect of dependency, such that it is mistaken to overlook its mereological and conceptually imputative dimensions. This last is especially informative in reconstructing a Madhyamaka response to the Buddhist free will problem. Hence, to provide a satisfying and accurate account of the relationship between karma and dependent origination care must be taken to avoid confusing what is only part of a theory with the whole.
Footnote 30 continued and they overshoot was is considered to be true in the world. Therefore, I say that it is wrong on the part of those ascetics and Brahmins." 31 Some scholars consider the Moḷiyasīvaka sutta as decisively settling the historical Buddha's own views about free will, insofar as karma is presented as steering a middle course "between determinism and randomness" (Gombrich 2009, p. 19). If unenlightened people cannot disentangle karmic and non-karmic causality without risking inducing insanity, then, from a pragmatic perspective it makes sense to believe in one's own capacity to shape the future. However, the philosophical question of whether this is possible remains. 32 Candrakīrti-MAV 6. 41. (Trans.) Huntington (1989), p. 162.

Dependency Relations in Madhyamaka
Hitherto, reconstructions of Buddhism's stance on free will have largely been driven by questions regarding the similarities and differences between causal determinism and dependent origination (Siderits 2008;Goodman 2009;Federman 2010). Scholarly consensus on this matter remains a distant prospect, and, although a final verdict might aid the debate, it cannot settle it. This is because different schools of Buddhism conceive of dependent origination in their own specific ways. To understand Madhyamaka's view of the connection between karma and dependent origination, scholars should start with an analysis of their causal relationship, but they should not end there. Madhyamaka presents dependency relations as being of three, increasingly sophisticated, kinds: causal, mereological and conceptually imputative. Failure to consider how this last dimension of dependency informs Madhyamaka's expression of the Buddhist free will problem unduly limits the responses available to it. By examining the increasingly complex ways in which dependent origination has been understood, we can discern the trajectory of the concept as it has progressed through incrementally more abstract iterations.
For example, early expressions of this thesis are couched in predominantly causal language and are intended to dislodge the idea of permanence. Mereological dependency-i.e. the relations holding between parts and wholes-acquires a different gloss in Madhyamaka thought to that presented in earlier forms of Buddhism. The idea of wholes as dependent on parts is fairly straightforward. However, Mādhyamikas develop this idea (arguing that parts and wholes are, both existentially and conceptually, mutually dependent) in order to bolster their challenge against the coherence and instantiation of a unified, substantial, self. 33 Madhyamaka and Yogācāra philosophers then extend and subsume the causal and mereological aspects of dependency beneath conceptually imputed dependency relations. Mādhyamikas think the general understanding of causal and mereological relations is fundamentally misguided and rests on the presumption that entities possess inherent reality (svabhāva). Madhyamaka claims that elimination of this wrong view affords insight into the conceptually constructed nature of experienceultimately everything is empty.

The Emptiness of Causation
Ā bhidharmikas enunciate two different causal models: the four-conditions model (pratyaya), and the six-causes model (hetu). Both models seek to provide a systematic overview of the causal process. However, as Salvini points out, the hetu model is subsumed under the pratyaya model so that the former becomes explanatorily superfluous (Salvini 2014, p. 474). In the pratyaya model, the 33 Candrakīrti-MAV 6. 161.
"When the carriage does not exist, then the 'possessor of the parts' does not exist, and neither do the parts. Just as, for example, when a carriage has burned its parts no longer exist, so when the fire of discrimination (mati) burns the possessor of the parts, the parts themselves [are incinerated]." (Trans.) Huntington (1989), p. 177. relationship between cause and effect depends upon the four conditions of: causality (hetu), support (ālambana), contiguity (samantara) and reason (adhipati). 34 Mādhyamikas superficially subscribe to this model even though they offer a radically different interpretation of the relation between cause and effect. 35 Commentators read MMK 1.3, where Nāgārjuna mentions the four conditions, as expressing the opponent's position. However, it is not so much the model, but the ontological status of the model's elements, that Mādhyamikas reject. Where Abhidharma thinkers maintain that there is an inherently real connection between cause and effect, Mādhyamikas state that this cannot be so: if causes are inherently real (i.e. possessed of svabhāva) how could they produce change without being changed themselves? Equally, if effects are inherently real, the presence of a cause is redundant. In accepting this model, therefore, Mādhyamikas do not commit themselves to the metaphysics of causation which it presupposes. Determining how deeply one must commit to the metaphysics in order to retain the model as a working-conventionally useful-hypothesis has been a major source of disagreement between Mādhyamikas. Arguably, this constitutes the primary point of contention between Candrakīrti and Bhāvaviveka, who approach the possibility of conventional svabhāva very differently. 36 Prāsaṅgikas hold that causation is merely conceptually imputed. Accordingly, causal relationships-of which karmic relations are a variety-cannot be independently established but always, at least partly, reflect our own specific interests and modes of conceptualization. The proliferation of causes and conditions-which Prāsaṅgikas consider as a reflection of the inner workings of mind-generates further reasons for uncertainty as to whether people can ever be morally accountable for their dependently originated intentions.
Mādhyamikas reinterpret the earlier Buddhist analysis of causal relations. They reject the idea of causal powers as productive of inherently existent phenomena whilst nevertheless accepting the mechanism in conventional terms. Given that causal relations comprise just one aspect of dependency, straightforwardly equating dependent origination and causal determinism is unjustified. While the two theories bear some resemblance, the idea of necessitation is not only absent from Buddhist accounts but is also explicitly challenged in chapter one of the MMK. Moreover, Buddhists hold that sentient beings can arrest (perhaps also reverse) the process of dependent origination and thus escape saṃsāra. Nāgārjuna argues that causal relations cannot obtain between entities with inherent reality (svabhāva). This does not lead him to the conclusion that causation never occurs, but rather to the view that causation and change are possible precisely because nothing possesses svabhāva. At MMK 1.1 Nāgārjuna examines four possible ways something with svabhāva could be causally produced and finds them all to be deficient. It would be either: (i) causa sui, which is logically impossible; (ii) dependent upon something else, thence lacking svabhāva; (iii) a combination of both, and so doubly problematic or; (iv) spontaneously produced, in which case not causally so. In attacking this realist conception of causal power, Nāgārjuna seeks to undermine the coherence of inherent reality-a conception he regards as metaphysically flawed as well as morally debilitating. Accordingly, Madhyamaka does not reject causation so much as a particular account of it.
Throughout the MMK, Nāgārjuna argues that identity and permanence are erroneously ascribed on the basis of a false-but engrained-conviction about the presence of svabhāva in that which lacks it. Amongst the conclusions drawn by Madhyamaka, the most important for understanding dependent origination is the idea that the entire phenomenal realm is dependent upon the conceptually constructive workings of the mind. As a therapeutic philosophy, the aim of which is to enlighten and liberate, Madhyamaka's first contribution is a diagnosis: acts of reification, self-grasping, and conceptual imputation confound the mind. Nāgārjuna proposes a remedy at MMK 25.24, "This halting of cognizing everything, the halting of hypostatization, is blissful." 37

Dependence on Conceptual Imputation
To understand dependency in terms of conceptual imputation is to realize that dependent origination is itself a dependent concept. This interpretation of dependent origination emerges from very specific-yet contested-readings of MMK 24.18. This verse, often regarded as "the most celebrated verse of the work" (Siderits and Katsura 2013, p. 277), reads as follows: yaḥ pratītyasamutpādaḥ śūnyatāṃ tāṃ pracakṣmahe/ sā prajñaptir upādāya pratipat saiva madhyamā//. 38 Various translations have been proposed but Berger has classified the most recurrent renditions as being either "nominalist" or "conventionalist" (Berger 2010). On Berger's estimation, neither of these are adequate because they overlook the "grammatical ambiguity" contained in the second half of the verse (which centres on the question of whether to understand prajñaptir upādāya as a technical 37 Nāgārjuna-MMK 25.24. sarvopalambhopaśamaḥ prapañcopaśamaḥ śivaḥ/ (Trans.) Siderits and Katsura (2013), p. 304. It is noteworthy that some modern scholars consider chapter 25 to be the last authentically Nāgārjunian chapter, and hence 25.24 the final verse (Siderits and Katsura 2013, p. 305). If this is credible, then 25.24 should be read as the climactic expression of dependent origination, the explication of which, as the dedicatory verse informs us, is the purpose of the treatise. 38 Nāgārjuna-MMK 24.18.
"Dependent origination we declare to be emptiness. It [emptiness] is a dependent concept; just that is the middle path" (Trans.) Siderits and Katsura (2013), p. 277.
compound meaning a dependent designation, or whether to take upādāya more straightforwardly as an absolutive qualifying prajñapti, thence meaning something like having acquired this notion) (Berger 2010, p. 46). He also alleges that these types of translation inspire readings of Nāgārjuna which he considers philosophically untenable (Berger 2010, p. 40). As the exchange between Berger (2011) and Garfield and Westerhoff (2011) demonstrates, this verse provokes both philological and philosophical disputation. However, as is generally agreed, Candrakīrti reads the verse as implying thoroughgoing linguistic conventionalism. Since the focus of this paper is the implied Prāsaṅgika Madhyamaka stance on free will, it is justifiable to follow Candrakīrti in regarding dependency relations as themselves conceptually dependent.
In his commentary on this verse, Candrakīrti intimates that the terms referring to dependent origination, emptiness, dependent designation and middle way are all synonyms (viśeṣa-saṃjñā). 39 The upshot of this analysis, which embraces the circularity of these terms, is the so-called emptiness of emptiness itself and a denial of absolute objectivity. The emptiness of emptiness is confirmed in the following verse (MMK 24.19) where Nāgārjuna states that nothing is exempt from being niḥsvabhāva. 40 Hence, if emptiness and dependency are synonyms, and emptiness is empty, then dependency too is itself dependent. In later formulations of Mahāyāna Buddhism, dependent origination is interpreted as inter-dependence, such that all phenomena reflect and influence all other phenomena. On this view the idea of an entity as it is in itself becomes untenable. Here, the ontological primacy of specific entities is disavowed, so that identity can be conferred only insofar as the principle of relationality is universally extended (Hershock 2017).
The idea that mind constructs reality has important ramifications for the Buddhist free will debate. If reality is generated rather than given it follows that there may be many different versions of reality, rendering the attainment of definitive or absolute resolutions of any problem impossible. A reading of dependent origination which focuses on the constructive power of mind also has implications for our treatment of the law of karma. In the absence of mind-independent reality, the mental attitude with which people comport themselves has consequences for the nature of reality itself. The stock example used to indicate the power of mind in shaping experience is the glass half empty or half full-where experience is dictated by description. At both the collective and the individual level, karma is thought of as the architect of one's reality. So, as long as dependency relations are taken as exclusively causal, the most a reconstructivist can hope to achieve is a position on the spectrum between hard determinism and libertarianism (Goodman 2009;Breyer 2013 apratītya samutpanno dharmaḥ kaścin na vidyate/ yasmāt tasmād aśūnyo hi dharmaḥ kaścin na vidyate// "There being no dharma whatsoever that is not dependently originated, it follows that there is also no dharma whatsoever that is non-empty." (Trans.) Siderits and Katsura (2013), p. 278. phenomena are understood as dependent upon conceptual imputation however, the basis for the validity of that spectrum itself can be called into question. This creates new reconstructive possibilities and reveals the futility of striving to articulate a definitive solution to debates about Madhyamaka Buddhism and free will.

Relativizing Free will: Extending Perspectivalism
Perhaps exasperated by the never-ending nature of debates about free will, some Western philosophers have begun to take seriously the idea of "free will subjectivism" (Double 2004). To be a free will subjectivist is to maintain that judgments assigning moral responsibility cannot be objectively true but can only ever be relatively so. Proponents of this view contend that it is the natural counterpart to meta-ethical subjectivism, the view that the truth or falsity of moral claims depends upon the framework in which they are embedded. On this view, free will theories, like normative theories, are susceptible to meta-level analysis so that in the absence of objective moral facts there can be no objective fact as to whether a person is morally responsible.
In recent years, Daniel Breyer has advanced perspectivalism as a response to the Buddhist free will problem (Breyer 2013). Perspectivalism is the view that the truth about free will is always relative to the perspective from which that view is expressed. There is much to commend in Breyer's presentation of perspectivalism, as even those of whom Breyer is highly critical admit (Goodman 2017a, p. 39). Despite this, in its present form, perspectivalism fails to communicate the need to move beyond the idea of definitive solutions to the Madhyamaka Buddhist free will problem. Although Breyer adumbrates an anti-realist version of perspectivalism intended as a viable Madhyamaka response, he nevertheless presents this solution as in rivalry with existing approaches (such as paleo/ neo-compatibilism and hard determinism). It is the contention of this paper that perspectivalism is more fruitfully understood when conceived of as a meta-theory about Buddhism and free will. This is because perspectivalism has more to offer as a critique of theories purporting to describe the distribution of free will than as just another such theory.
Given Madhyamaka's distinctive interpretation of the two truths, which sees an eventual collapse of the semantic isolation between conventional and ultimate truth, assertions can only be relatively true or false (Siderits 2015, p. 191). The two truths theory can be understood in an abundance of ways (see The Cowherds 2011). For the present purposes, however, the most important consideration is that Madhyamaka resists all notions of ultimate truth conceived of as essentially separable from conventional truth. This is confirmed at MMK 24.10 by the claim that "the ultimate truth is not taught independently of customary ways of talking and thinking." 41 Madhyamaka considers a proposition's soteriological efficacy to be the hallmark of its status as a conventional truth.
Where soteriological considerations do not apply-as with morally neutral statements-conventional truths reflect common sense or else align with supporting evidence in a nexus of coherence. However, as there is no mind-independent reality, there are no ultimate truths either. Siderits has interpreted the Madhyamaka stance as entailing the seemingly paradoxical conclusion that "the ultimate truth is that there is no ultimate truth" (Siderits 2016, p. 27), a riddle which is resolved through the so-called semantic reading. This stipulates that two distinct referents are denoted by 'ultimate truth,' and that these can be disambiguated to avoid equivocation, where the first usage of 'ultimate truth' reflects that which it is soteriologically helpful to believe, and the second usage of 'ultimate truth' reflects the way the world mind-independently is (Siderits 2007, p. 202).

The Asymmetric Ascription of Free will
If the Madhyamaka stance towards truth is consistent with subjectivism, whatever approach it implicitly has towards free will should be relativized. Significantly, the free will subjectivist argues against the idea that there is any objective fact about whether hard determinism, compatibilism or libertarianism is true. In different contexts, hard determinism and libertarianism may be affirmed without a conflict arising at the meta-level about which is true. The free will subjectivist is therefore in a position preferentially to systematize the lower-level free will theories on the basis of pragmatic considerations, such as which view about free will results in the execution of moral norms which are themselves arrived at through non-objective analysis (Double 2004). Perspectivalism is, in important respects, similar to free will subjectivism. More elucidation is, however, required before perspectivalism can disabuse Buddhist free will theorists of the illusion that a definitive solution to the problem is available.
Perspectivalism stands out amongst reconstructed Buddhist positions on free will as a theory readily available to Madhyamaka. On Breyer's account, the perspectivalist maintains that from the first-person point of view it is right to see oneself as possessing free will and as morally responsible, but from the second and thirdperson points of view it would always be inappropriate to ascribe responsibilityentailing freedom. This conclusion is derived from endorsing the principle of successful practice, according to which assent should only be given to statements when their acceptance is conducive to liberation. Perspectivalism is thus premised on the idea that blaming others impedes spiritual growth whereas holding oneself responsible promotes it.
Breyer appears not to consider the possibility that, in certain circumstances, it might be to the spiritual advantage of others to hold them morally accountable for their conduct. Goodman, on the other hand, has argued that perspectivalism becomes more convincing when we adopt a doubly asymmetric view (as opposed to a merely asymmetric view). On the doubly asymmetric view people should regard themselves as responsible for morally deficient conduct but, in the effort to thwart the manifestation of pride and other unwholesome attitudes, should not take responsibility for their good conduct. Contrastingly, a person should always regard others as the originators of their own good conduct but never regard them as responsible for unwholesome action (Goodman 2017a, p. 37). The assumption underpinning both the asymmetric and doubly asymmetric proposals is that, by adopting such a stance, greater compassion towards other beings is cultivated and hence the soteriological goal is pursued.
The first thing to note about perspectivalism is that it makes no pretensions to universalizability. If two people accept perspectivalism they cannot both be right about the fact that they alone possess free will and are morally responsible, if to be right about something means to hold a view corresponding to the facts. It follows from the Madhyamaka's analysis of the notion of a mind-independent reality as incoherent that there can be no fact of the matter concerning free will or its distribution. Perspectivalism therefore seems to constitute a solution to the free will problem without thereby suggesting that the solution it offers is definitive. Unlike the other theories-such as paleo-compatibilism, hard determinism and neocompatibilism-perspectivalism pays greater attention to aspects of dependent origination other than the causal dimension. Specifically, in taking Madhyamaka's more refined understanding of what it means for something to be dependently originated-i.e. dependent upon the mind's tendency towards reification and the conceptually imputative process-the perspectivalist arrives at a conclusion which recognises itself to be merely conventionally established.
Nevertheless, insofar as it continues to pitch itself against other theories and insofar as it makes no attempt to harmonize existing approaches by structuring them hierarchically, perspectivalism fails fully to capture Madhyamaka's insight into the pervasiveness of dependency. In his presentation of perspectivalism, Breyer implies that this theory is triumphant over competing approaches to the free will problem. The irony is, that in emphasizing Madhyamaka's disavowal of the possibility of an absolutely correct stance on this issue, the claim's counterpart-that there can be no absolutely incorrect stance either-is overlooked.

Perspectivalism and the Promotion of Compassion
Goodman argues that an anti-realist perspectivalist could justify advocacy of the (doubly) asymmetric view because it affords soteriological benefits (Goodman 2017a, p. 39). However, if soteriological fulfilment is the perspectivalist's only criterion of truth in debates on free will, their eagerness to dismiss alternative positions is misplaced: people of varied psychological dispositions will be motivated in different measures by the plethora of free will theories available. In the interests of philosophical rigour, the methods used to establish anti-realist perspectivalism ought therefore to be applied to its own conclusions. Rather than compete against the other theories, perspectivalism can establish itself as a metatheory able to subsume them.
Through recognising the relative truth of responses already proposed to the Buddhist free will problem, the perspectivalist accesses a new type of resolution; one which conceives of the mutual exclusivity of existing proposals as superficial. Perspectivalism thus does for debates on free will what the Madhyamaka system does for the entire enterprise of metaphysics: challenging the notion of truth as objective and, rather than offering one more view, undermines the basis of views altogether.
Whether perspectivalists can always justify switching between a first-person/ third-person stance on the attribution, or withholding, of responsibility is questionable-for it is uncertain that this will invariably lead to the most soteriologically successful results. However, this is a matter for the empiricist to decide. Nevertheless, the resolute tone of Breyer's call for the adoption of the asymmetric view should inspire caution. Goodman has argued that the application of the doubly asymmetric view should serve as a corrective to self-cherishing but he acknowledges that, for the enlightened, compassionate conduct is governed not by adherence to a specific view but by "spontaneous responsiveness" (Goodman 2017a, p. 38). In some circumstances, soteriological objectives might be furthered most effectively by holding others responsible for their misdeeds, even if the illusion of agency is dispelled from an enlightened perspective.
Even though perspectivalists and traditional exponents of Madhyamaka share the view that reactive attitudes inhibit spiritual progress, we can only speculate as to whether their final extirpation would coincide with the collapse of meaningful human relationships (Strawson 1962, pp. 193-195). It is difficult to know whether Ś āntideva and others are right in their claim that, on balance, reactive attitudes lead to excessively dysfunctional/ sub-optimal behaviour. The number of people who even claim to have relinquished these attitudes is so few that to compare their conduct with that of those who persist in adopting the participant stance is virtually impossible. Whilst some reactive attitudes evidently lead to anti-social conduct, sociologists and psychologists might challenge the idea that all instances of anger are unhelpful or morally unjustifiable (Goldman 2014, p. 3). Ś āntideva may be right in saying that the logical basis for anger towards people is the same as that directed towards sticks or bile, but anger in the first case is not so clearly futile as in the subsequent cases. 42 In spite of the foregoing reflections, the specific content of perspectivalist views on free will distribution is not of primary concern. Instead, the main perspectivalist idea is that we should think of free will distribution in whichever way promotes soteriologically successful practice. Perspectivalists can decide whether to hold themselves and/ or others responsible on the basis of empirical evidence about how far this promotes compassionate responsiveness. The best approach to take, then, is that which most aids spiritual maturation. Accordingly, there is no reason why in principle perspectivalists cannot also endorse other theories about free will if doing so is likely to encourage soteriological progress. 42 Ś āntideva-BCA 6.22.
pittādiṣu na me kopo mahāduḥkhakareṣv api/ sacetaneṣu kiṃ kopaḥ te'pi pratyayakopitāḥ// "I feel no anger towards bile and the like, even though they cause intense suffering Why am I angry with the sentient? They too have reasons for their anger." (Trans.) Crosby and Skilton (1996), p. 52.

The Consistency of Perspectivalism and Seemingly Rival Theories
Breyer prefaces his presentation of perspectivalism with a sustained attack on Buddhist hard determinism, paleo-compatibilism and neo-compatibilism. As noted, Goodman has recently explored the potential benefits a Buddhist hard determinist may derive from sustaining the illusion of free will, which, he thinks, perspectivalism can deliver. According to Goodman's revised position, even if hard determinism is ultimately true, some people will benefit from the illusion of free will and thus would benefit from acceptance of the perspectivalist claim that they themselves possess it. Goodman has also correctly recognized that to take perspectivalism to its logical conclusion involves the "full abandonment of the ascription of responsibility and the illusion of agency" (Goodman 2017a, p. 39). As the Madhyamaka tradition explains, the illusion of agency and selfhood runs so deep that its consequences are experienced both at the cognitive and affective levels.
Eventually, acceptance of perspectivalism requires abandonment of belief in agency yet, presumably, during the time through which this illusion persists the perspectival approach remains useful. Perspectivalism can undermine unwholesome reactive attitudes and so expose as illusory the facticity of free will theorising. In conjunction with each other, these factors of perspectivalism produce a cognitive and affective shift within the aspirant and contribute to the transition from reflection on free will to the attainment of freedom. Perspectivalism is therefore a useful instrument, which, like all instruments, should be discarded once its purpose is fulfilled.
If perspectivalism is better understood as a meta-theory about free will than as a theory describing which conditions must obtain for the ascription of responsibility, it can encompass various approaches to the Buddhist free will problem. In exploring the entailments of Madhyamaka's call to abandon all views and to recognise the emptiness of reality, the perspectivalist should be open to the possibility that apparently rival strategies are in fact consistent. To take Siderits's theory of paleocompatibilism as an example, there is no need to insist, as Breyer does, on the mutual exclusivity of this approach from perspectivalism.

Paleo-compatibilism and Perspectivalism
Paleo-compatibilism and perspectivalism have more in common than Breyer allows. Firstly, both theories implicitly subscribe to free will subjectivism-the view that there is ultimately no fact of the matter as to whether people have free will. The paleo-compatibilist maintains that statements expressive of conventional truth are semantically isolated from those expressive of ultimate truth. Since people (i.e. conceptual constructs with merely conventional reality) are said to either have or not have free will, ascription of free will can only ever be, at most, conventionally accurate. Since statements about people refer to convenient fictions, statements pertaining to free will are meaningless at the ultimate level. A second respect in which paleo-compatibilism and perspectivalism are similar is that both theories advocate the shifting back and forth between perspectives. For the perspectivalist, vacillation between first and second/ third-person points of view is said to inspire compassion, leading to incrementally more expansive freedom. For the paleocompatibilist, alternation between conventional and ultimate perspectives on free will presumably facilitates a more comprehensive understanding of reality.
Where the paleo-compatibilist provides reasons for dispensing with the notion of ultimate level free will, the perspectivalist dispenses with the idea of others as morally responsible. It is unclear why Breyer contends that paleo-compatibilism and perspectivalism cannot be synthesized for, in fact, by pooling the resources of each, a more robust stratagem for spiritual progress becomes available. The conventional level focus common to both theories prevents the occurrence of an absolute contradiction between the positions advanced: context and audience determine the efficacy of the theories. Accordingly, the relationship between paleo-compatibilism and perspectivalism may be conveyed through the image of a Russian nesting doll. The overarching 'nesting' principle relies on the possibility of increasingly greater degrees of skilfulness in expression so that the pedagogical circumstances guide postulations about the relation between perspectivalism and paleo-compatibilism. Thus, as has been suggested above, perspectivalism acts as a meta-theory, applicable in equal measure to itself and to other theories. Breyer, however, goes to some lengths in his attempt to discredit the paleo-compatibilist theory and establish it and his own theory as mutually exclusive.
Breyer's main criticism of paleo-compatibilism is that it too quickly, and on insufficient grounds, equates dependent origination with causal determinism. The force of this objection has already been explored and acknowledged. However, his other objections are less convincing. Suppose that Breyer's argument succeeds in demonstrating the mutual semantic reliance (and hence lack of insularity) of conventional and ultimate truth. In that case, reconstructivists can to appeal to paleo-compatibilism (originally intended for an Abhidharma context) to articulate a conventional-level response suitable for Madhyamaka. If, as Breyer contends, there is a necessary semantic connection between the conventional and ultimate domains of truth, then the prospects of modifying paleo-compatibilism and rendering it palatable to a Madhyamaka audience are improved. To reiterate Nāgārjuna's presentation of this relationship at MMK 24.10 ""the ultimate truth is not taught independently of customary ways of talking and thinking." 43 Rather than view conventional and ultimate truths as pertaining to strictly incommensurate domains, the Madhyamaka is apt to think in terms of less and more sophisticated expressions of truth relative to a spectrum. Consequently, Siderits's assessment of the ultimate truth of determinism and the conventional truth of the instantiation of free will could be interpreted as reflecting truths of different degrees of sophistication. This is particularly so if paleo-compatibilism is understood as primarily interested in situating truths about free will hierarchically and only secondarily interested in establishing how exactly free will is distributed.

Conclusions
Although reflection on Buddhism's implied stance on free will is a recent development, it is already possible to identify certain sub-optimal trends in approaches to this topic. One such trend has been the perfunctory treatment of what is arguably the central tenet of Buddhism: dependent origination. Straightforwardly equating this concept with the Western thesis of causal determinism has probably impeded progress and has almost certainly opened the reconstructive enterprise to attack from those who see the entire project as founded on a category mistake. The work of Breyer has been instrumental in encouraging more reflection on the grounds for assuming the parity of ideas which, even if superficially similar, have arisen in very different cultural contexts and have fulfilled different explanatory functions therein.
This paper has therefore argued that in discussing the Buddhist free will problem it is preferable to address the tension engendered by Buddhism's simultaneous commitment to the doctrines of karma and dependent origination. Reconstructions of a distinctively Madhyamaka stance on free will benefit from appreciation of the multi-dimensional strands of dependency and, in particular, from a consideration of this system's emphasis on the conceptually imputative aspect of dependency. If the Madhyamaka arguments pertaining to the emptiness and conceptual constructedness of all phenomena are convincing, then the search for definitive or absolutely correct responses to the Buddhist free will problem is futile. Finally, this paper has argued that the perspectivalist theory advanced by Breyer and modified by Goodman constitutes a promising response for a system which has rejected metaphysical realism as incoherent. Insofar as perspectivalism presents itself as an alternative free will theory or as a rival to existing theories (such as paleocompatibilism for example) its soteriological utility is limited. Once presented as a meta-theory under which conventionally competing theories are subsumed, perspectivalism's full soteriological potential can be harnessed.