2. Vasubandhu on Existence, Truth, and Language

The ‘No Self’ Thesis – though universally shared amongst Buddhists – is fleshed out in different ways by different factions of Buddhists, partly depending on the conception of substantial or fundamental existence that they work with. In the sixth chapter (kośasthāna) of the Treasury of Abhidharma (Abhidharmakośa), Vasubandhu distinguishes two modes of existence (sattā): fundamental (pāramārthika) and conventional (sāṃvṛtika). Roughly speaking, a fundamentally or substantially existent object is a basic constituent of reality: its nature and existence is (in some sense) independent of other entities. By contrast, a conventionally existent object isn't fundamentally existent, but rather something whose existence we accept because it is practically useful for us to do so. Here, I spell out this distinction, and explore what this view entails about the truth-aptness of our talk about ordinary objects.

In v. 6.4 of the Treasury of Abhidharma (Abhidharmakośa), Vasubandhu fleshes out the distinction between the two modes of existence as follows: ‘That object, which is such that there is no awareness of it when it is divided or when there is an exclusion of other entities [from it] by means of awareness, is conventionally existent (saṃvṛtisat). For example, a pot [and water]. What is fundamentally existent is different’ (AK 6.4).Footnote ² On Vasubandhu's view, a conventionally existent object is something that ceases to be an object of awareness (buddhi),Footnote ³ i.e., an object of an experience or a thought, either (a) when it is divided or (b) when we mentally exclude other entities from it.

Vasubandhu unpacks the idea in his commentary (AKBh 334.4–6). For Vasubandhu, any fundamentally existent object (paramārthasat) is independent in two ways. First, it is mereologically independent of other things. So, it cannot be destroyed in virtue of being physically divided into parts. But an ordinary object like a pot depends on its parts for its existence. If I physically break the pot into the two pot-halves that it is made of, it will no longer exist. Then, we cannot correctly apply the concept ‘pot’ to the pot-halves that are left over: for example, we will no longer look at the pot-halves and think, ‘That is a pot’. That is the sense in which the awareness of a pot ceases when it is physically broken down into parts. Second, any fundamentally existent object is conceptually independent of other things. Take water. We can mentally analyse the nature of water into other things: qualities such as visible form (rūpa), liquidity (sneha), and so on. Once we have mentally excluded these qualities from something, nothing remains of the intrinsic nature of the water. We cannot imagine something that is devoid of the characteristic qualities of water and still think of it as water. That is the sense in which the awareness of water ceases when the characteristic qualities of water, e.g., its visible form and so on, are mentally excluded from it.

For Abhidharma Buddhists like Vasubandhu, the only fundamental constituents of reality are certain entities – called the dharmas – that are both mereologically and conceptually independent of other things. Typically, these Ābhidharmikas divide the dharmas into five kinds: (a) matter (rūpa), (b) mental events (citta), (c) associated mental factors (caitta), (d) conditioned factors that are dissociated from the mental (cittaviprayuktasaṃskāra), and (e) causally unconditioned dharmas. However, amongst the Ābhidharmikas, Sautrāntika philosophers like Vasubandhu think that categories (d) and (e) are empty. For instance, Vasubandhu explicitly argues that all dharmas are causally conditioned and momentary: they arise from a collection of causal conditions and cannot persist through time (AKBh 92.4–5 ad AK 2.55, AKBh 193.2–194.14 ad AK 4.2–3). What, then, are these causally conditioned, momentary dharmas according to Vasubandhu? According to one natural interpretation, such dharmas are tropes or property-particulars.Footnote ⁴ Vasubandhu takes matter to be of two kinds: (a) the material elements (mahābhūta) like the earth element, the water element, and so on, and (b) derived matter (upādāyarūpa or bhautika), i.e., physical qualities that are causally dependent on the material elements. The material elements are simply property-like entities: the earth element is solidity, the water element is fluidity, the fire element is heat, and the wind element is motion. Similarly, derived matter includes perceptible qualities like visible form (rūpa), odour (gandha), and flavour (rasa), which are derived from the material elements. By contrast, mental events are conscious or unconscious states of awareness, while the associated mental factors are those mental tropes that invariably accompany mental events, i.e., concepts or mental labels (saṃjñā), hedonic states (vedanā) like pleasure and pain, and conditioned forces (saṃskāra) like intention, mental habits, and so on.

Vasubandhu claims that all these fundamentally existent objects are mereologically and conceptually independent of other things (AKBh 334.7–11). Even after a visible material object has been whittled down to its mereologically simple constituents, or other entities like flavour have been mentally excluded from its nature, the awareness of the intrinsic nature of visible form doesn't cease. This is because the existence and the intrinsic nature of visible form are neither mereologically reducible to simpler parts nor mentally analysable into other things. Since visible form is both mereologically and conceptually independent in this way, it can be treated as fundamentally existent. The same, according to Vasubandhu, is true of other physical and mental tropes that he regards as the fundamental constituents of reality.

For Vasubandhu, any object whose existence is practically useful for us to accept but fails to be either mereologically or conceptually independent in this way, can at best be conventionally existent. Most objects that we ordinarily treat as existent – tables and chairs, pots and cloths, chariots and carts – are like this. However, Vasubandhu emphasises that this distinction between the two modes of existence does not mean that we speak falsely when we assert the existence of conventionally existent objects like a pot or water. He says:

On one way of understanding the point, corresponding to the two modes of existence, there are two notions of truth. The content of a state of awareness (or a linguistic utterance) is conventionally true (saṃvṛtisatya) just in case that content presupposes or entails the existence of fundamentally non-existent objects, but still is practically useful for us to accept because that content indirectly reflects or tracks the way fundamentally existent objects are. By contrast, the content of a state of awareness is fundamentally true (paramārthasatya) just in case that content does not presuppose or entail the existence of any fundamentally non-existent objects and accurately reflects how things are with the fundamentally existent objects. The contents of our ordinary states of awareness – our experiences and thoughts about ordinary objects like a pot or water – can only be conventionally true. The states of awareness that represent what is ultimately true are much harder to attain.

After introducing the distinction between the modes of existence, Vasubandhu mentions a view of earlier Buddhist teachers, according to which what is ultimately true can be grasped only by means of two kinds of awareness: an extraordinary state of awareness (lokottarajñāna) and an ordinary state of awareness that is obtained on the basis of that extraordinary awareness (tatpṛṣṭhalabdhajñāna) (AKBh 334.11–13). The first is an epistemically direct (i.e., non-inferential), non-conceptual insight into reality, which arises when a Buddhist practitioner achieves meditative equipoise on the path of insight (darśanamārga) after having listened to Buddhist teachings and rationally reflected and meditated on them. By contrast, the second kind of awareness is an ordinary experience or thought that is influenced by that earlier direct and non-conceptual awareness. In the second kind of awareness, ordinary objects like tables and chairs might appear as real, but the practitioner will no longer assent to that false appearance.

There is an important lesson to be drawn from this discussion about how Vasubandhu understands the relationship between language and reality. For Vasubandhu, the realm of fundamentally existent objects is inaccessible to language. Contents that are conveyed by means of linguistic utterances proliferate distinctions – the distinction between qualities and substances, between properties and their bearers, between actions and their agents, and so on – which do not exist amongst fundamentally existent objects. Yet, this does not mean that language is inherently deceptive. Vasubandhu makes this point while discussing the Buddhist doctrine of dependent arising. Roughly speaking, this doctrine says that everything is causally conditioned: existent objects arise by depending on other existent objects. The Sanskrit term for ‘dependent arising’ is ‘pratītyasamutpāda’: ‘pratītya’ is a gerund that means ‘having depended’ whereas ‘samutpāda’ just means ‘arising’. So, the term literally means ‘arising having depended on something else’. But, as a grammarian opponent notes, this implies that the object which arises, and therefore serves as the agent (kartṛ) with respect to the action or event (kriyā) of arising, does so by initially entering into a state of depending on something else. This is incoherent: since that object could not have existed before it arose, it could not have antecedently depended on something else. While commenting on the grammarians’ distinction between an action (conveyed by the verb in a sentence) and the agent (conveyed by the expression declined in the nominative case), Vasubandhu says:

Even though the literal content of a linguistic utterance may be misleading insofar as it involves distinctions that are fundamentally non-existent, nevertheless, in most cases of successful linguistic usage, it is possible to recover a content that isn't misleading to the same degree: a content that fits the way things are at the level of fundamental reality. Vasubandhu would agree with other Buddhists that the relationship between language and reality is dysfunctional: language, without some independent epistemic work on our part, cannot help us see how the world really is. But it is not so dysfunctional that we cannot recover any fundamentally true content at all from the literal contents of linguistic utterances. This idea – namely, that fundamental truths can somehow be salvaged from misleading literal contents conveyed by language – will be useful for understanding Vasubandhu's view about the first-person pronoun ‘I’.

3. Reductionism about ‘Self’

It's now time to return to The ‘No Self’ Thesis. In the ninth chapter of his Commentary – sometimes called ‘The Ascertainment of the Person’ (Pudgalaviniścaya)Footnote ⁵ – Vasubandhu argues that a person is nothing but a causally connected stream of momentary physical and mental tropes called the aggregates (skandha). There are five aggregates: matter (rūpa) and mental events (citta) along with associated mental factors such as hedonic states (vedanā), mental labels (saṃjñā), and conditioned forces (saṃskāra) like intentions and mental habits. These correspond to parts of the body as well as the physical and mental events that accompany it. A causally connected stream of these tropes is mereologically dependent on its constituents, and it is temporally extended. So, on Vasubandhu's view, it cannot be fundamentally existent. ‘Therefore’, Vasubandhu concludes, ‘a person exists as a matter of conventional designation (prajñaptisat) just like a heap or a stream’. (AKBh 467.12–13) Where does this leave us with regard to the self?

Vasubandhu wants to reject two distinct views about the expression ‘self’. Call the first substantialism, the view that the expression ‘self’ refers to a substantially or fundamentally existent object. Call the second eliminativism, the view that the expression ‘self’ refers neither to any substantially or fundamentally existent object nor to any collection of substantially or fundamentally existent objects. The first view was defended by the Brahmanical thinkers, such as the Nyāya and Vaiśeṣika philosophers, who took the self to be a fundamentally real inner constituent of a person, which serves as the subject of mental states and as an agent of actions. By contrast, the second view was defended by other Buddhists – Mādhyamikas like Nāgārjuna (2^nd century CE) and Candrakīrti (6^th century CE) – who argued that the self is neither any fundamentally existent object that is distinct from aggregates, nor reducible to either the aggregates or a collection of aggregates. Vasubandhu wants to defend:

In a remarkable passage in the third chapter of his Commentary, Vasubandhu distinguishes his position from that of the Brahmanical thinkers (while implicitly distancing himself from other Buddhists):

Under the first conception, the self is an inner agent of activity (antarvyāpārapuruṣa), which exists independently of the aggregates and acquires new aggregates when it is reborn and throws away the old aggregates when it dies. But there is a second conception of the self on which there is no such inner agent of activity. On this view, the term ‘self'’ is just figuratively or non-literally applied to the aggregates themselves. Vasubandhu emphasizes that the momentary aggregates cannot literally transmigrate (AKBh 129.14ff ad AK 3.18b). But we can figuratively say that a stream of aggregates, conditioned by afflictions, enters a womb, just as we can figuratively say that a lamp moves to a different place.

Elsewhere, Vasubandhu explains what this misleading view about the self (ātmadṛṣṭi) is:

On Vasubandhu's view, there is in fact no fundamentally existent constituent of a person that can serve as a subject of mental states at different times, or as an agent of actions, or as something over which we have perfect control. So, when we talk as if there is such a self, we must be ascribing these uninstantiated properties associated with the self to the impermanent aggregates themselves. In doing so, we are treating the aggregates as what they are not: as a subject, as an agent, as something under our control. This view that there is a substantial self (satkāyadṛṣti) forms the basis of different extreme philosophical views about the self like eternalism (śāśvatadṛṣti), i.e, the Brahmanical view that the self is a permanent substantial entity that can survive death and be reborn, and annihilationism (ucchedavāda), i.e., the Cārvāka or Lokāyata view that the self is just the living body and therefore endures through our lifetime but then is destroyed at death. And these extreme views about the self, in turn, give rise to certain afflictions (kleśa), e.g., one's love for oneself and what belongs to oneself (ātmātmīyasneha), that ultimately are the cause of all our physical and mental suffering. According to Vasubandhu, since these afflictions are based on an illusory awareness of the aggregates as what they are not, they can be abandoned through insight (darśanaheya), i.e., by seeing the aggregates as what they are.

The Buddhist view that Vasubandhu prefers is the source of that insight. On this conception of the self, the term ‘self’ refers to a causally connected stream of aggregates. It steers clear of both the extremes of eternalism and annihilationism. Since the stream of aggregates consists of aggregates that are destroyed at every moment, there is no permanent fundamentally existent self. Yet, since a stream of aggregates can survive death and be reborn in virtue of new causal connections with new sets of aggregates, the self isn't destroyed at death.

Why should we accept this reductionist view of the self? Vasubandhu offers the following argument:

The argument here should be construed as an argument from the absence of evidence.

For now, let's assume that Non-Vacuity is true: the term ‘self’ (as we ordinarily use it) isn't empty, i.e., it does refer to something. What about Knowability? Vasubandhu assumes that, if the term ‘self’ refers to anything, its referent must be something that we can know either through direct apprehension, i.e., some form of perception, or by means of inference. To motivate this premise, Vasubandhu seems to appeal to the principle that any dharma that exists can be known (or knowable) by either direct apprehension (in the absence of obstructions) or inference. This principle – as stated – seems questionable. Why couldn't there be existent dharmas that can neither be directly apprehended nor known by means of inference? A better version of Knowability would be weaker; it would say that we don't have any reason to posit a dharma that cannot be known either by direct apprehension or by inference. That follows from a principle of parsimony: we should not populate our ontology with entities that we have no reason to posit.

Turn now to Absence of Evidence. Under normal conditions, the six intentional objects of the senses – the sensible qualities such as visible form – and our own conscious mental events can be directly apprehended when they aren't obstructed from us. In the case of unobservable objects such as the senses themselves, we can make a causal inference from observed phenomena. For instance, from the absence of perception in blind or deaf people and from the presence of perception in non-blind or non-deaf people, we can infer that perception requires a distinct cause, i.e., a sense, which is missing in blind or deaf people. But a self that is distinct from the five aggregates cannot be directly inferred or inferred in this way. The thought may be spelled out as follows. If we take a human body and pull it apart, we don't perceive any such thing as a self other than the physical elements that constitute it; similarly, when we introspect, we don't come across anything other than the mental factors such as hedonic states, mental labels, etc. Similarly, as Vasubandhu claims later, there are no observed phenomena that can only be causally explained by positing the existence of a self that is distinct from the five aggregates. This supports Absence of Evidence.

Together, these three premises yield the conclusion that the term ‘self’ doesn't refer to anything other than the aggregates themselves. It doesn't yield the conclusion that Vasubandhu wants: the claim that the term ‘self’ refers to a stream of aggregates. This is where the final premise – Role Sufficiency – becomes relevant (though it is left unstated by Vasubandhu). If we could show that a stream of aggregates can perform all or many of the roles – such as being the referent of the first person pronoun ‘I’, being a subject of mental states, being an agent of actions – that are associated with the referent of ‘self’, a stream of aggregates will indeed be one of the most natural candidates for being the referent of the term ‘self’. But, then, given that the term ‘self’ cannot refer to anything other than the aggregates themselves, we will end up with the conclusion that it must refer to a stream of aggregates. This is precisely what Reductionism about ‘Self’ says.

This, then, could be used to derive the conclusion that Vasubandhu ends the passage with: namely, that there is no self. On the one hand, if we are working with the Brahmanical conception of the self, then there cannot be a conventionally or fundamentally existent object that answers to that conception of the self, since the term ‘self’ simply doesn't refer to any inner agent of activity that is distinct from the aggregates or from a stream of aggregates. On the other hand, if we are working with the thinner Abhidharma conception of the self, then there can be a conventionally existent object that answers to that conception of the self, since the term ‘self’ does refer to a stream of aggregates that has conventional existence. But the self still cannot be treated as fundamentally existent, because that stream of aggregates will be either mereologically dependent on its constituents, or temporally extended. In either case, there will be no fundamentally existent self.

Suppose we accept Knowability and Absence of Evidence. What we have not seen so far is any justification for either Non-Vacuity, i.e., the claim that the term ‘self’ refers to something that has three characteristics of the self – namely, being the referent of ‘I’, subjecthood and agency, and persistence through time – or Role Sufficiency, i.e., the claim that these three roles can be played by a stream of aggregates. To motivate Non-Vacuity, Vasubandhu would need an independent theory of linguistic reference, which explains why – despite our false beliefs about the self – our uses of the term ‘self’ do pick out something that plays these roles. To motivate Role Sufficiency, Vasubandhu would have to argue that the aggregates can indeed play these roles. In the next two sections, I consider whether he can show this.

4. The Reductionist Semantics of the First-Person Pronoun

Role Sufficiency commits Vasubandhu to a reductionist view about ‘I’. If a stream of aggregates is sufficient to play the roles associated with the self, and one of those roles is being the referent of ‘I’, then ‘I’ must refer to a stream of aggregates. This yields a kind of reductionism about ‘I’, the view that, in any context of utterance, ‘I’ does not refer to a substantially or fundamentally existent object, but does refer to a collection of substantially or fundamentally existent objects. In the ninth chapter of his Commentary on the Treasury of Abhidharma, Vasubandhu offers two arguments for this view.

Vasubandhu's arguments occur in the context of a response to an objection raised by a Brahmanical thinker.

The Brahmanical thinker asks how one could even undertake actions (intentionally) if there were no fundamentally existent self. Consider our self-interested actions, actions we undertake out of future-directed self-concern. When I walk out of my home on a rainy day, I typically take my umbrella with me so that I don't get wet later. Or, when I save a part of my salary each month, I do it precisely because I wish to be comfortable when I retire. As Vasubandhu's Brahmanical opponent notes, this kind of future-directed self-concern is driven by first-personal thoughts that take the form, ‘If I were to save money now, I would be happy later’, or ‘If I were to take my umbrella, I wouldn't catch a cold from the rain and suffer later’. These judgements about the consequences of my actions could be accurate only if the two occurrences of ‘I’ – in the antecedent and the consequent of the conditional – were to refer to a self that persists through time, performs actions intentionally, and later experiences the happiness or the suffering that results from those actions. But, if there were no self, then how could we account for the contentfulness of such judgements? And, if those judgements weren't contentful, then how can we rationally plan and (in some cases, successfully) undertake actions on the basis of such judgements?

Vasubandhu's response to this challenge is to say that, even if there were no fundamentally existent self, the first-personal judgements of the form, ‘I would be happy’, or ‘I wouldn't suffer’, needn't be false or without a truth value. For the first-personal ingredient of those judgements – the ‘I’-awareness – could still be directed at the aggregates. In other words, when we use the concept or expression ‘I’ in our thought or discourse, what we pick out are the aggregates. Vasubandhu offers two arguments for the claim that the first-person pronoun ‘I’ refers to the aggregates.

The first of these arguments – which Vasubandhu does not elaborate on – is what we may call the argument from self-love. The rough thought is that our peculiar love or affection for ourselves isn't directed at anything that is distinct from the aggregates, but rather at the aggregates themselves. When I take an umbrella with me, I do so out of concern for my future aggregates, so that this stream of aggregates won't include states of pain in the future. Similarly, when I save money, I do so out of concern for my future aggregates, so that those aggregates would be accompanied by states of pleasure arising from comfort in the future. But, presumably, whatever ‘I’ refers to is the object of this kind of love or affection for oneself. This, in turn, implies that the aggregates themselves – or a stream of those aggregates – is what ‘I’ refers to.

Turn now to Vasubandhu's second argument: it is what we may call an argument from self-ascriptions. It proceeds from the observation that ‘I’-awareness and states of awareness that take the form ‘fair’ can be co-referential (samānādhikaraṇa), i.e., can be directed at the same intentional objects. If I ascribe physical properties to myself by means of ascriptions like, ‘I am fat’, or ‘I am dark’, the predicates ‘am dark’ and ‘am fat’ apply to the same object that ‘I’ refers to. But these properties that I am ascribing cannot be the properties of the self as conceived by the Brahmanical thinkers, a self that is distinct from the physical aggregates. So, if these ascriptions are indeed correct, the referent of ‘I’ cannot be distinct from the physical aggregates.

I don't find myself persuaded by these arguments, and neither should you. Start with the argument from self-love. We can reconstruct it as follows.

Here, the Brahmanical thinkers wouldn't deny P1: typically Nyāya and Mīmāṃsā philosophers are happy to grant that the object of self-love is in fact what ‘I’ refers to. However, such Brahmanical thinkers would resist P2. They could argue that if there is any object of self-love – more specifically, the kind of future-directed self-concern that underlies many of our ordinary actions – then it cannot be the momentary aggregates, but must be a unitary self that performs actions at an earlier time and later comes to experience the results of those actions.

Vasubandhu seems to reject this line of objection elsewhere. In the third chapter of his Commentary on the Treasury of Abhidharma, Vasubandhu considers why bodhisattvas (i.e., morally ideal beings that can achieve enlightenment but delay it so as to alleviate the suffering of others) undertake altruistic actions (AKBh 182.7–15 ad AK 3.94a). His opponent here is a psychological egoist, who thinks that no one acts out of any motive other than self-concern. In response, Vasubandhu offers an account of self-interested actions. Under this account, we ordinarily undertake self-interested actions because we are unfamiliar with the true characteristics of the conditioned factors that constitute us as persons. Persons are nothing over and above streams of aggregates, and these aggregates are causally conditioned. As such, they share a number of general characteristics (sāmānyalakṣaṇa) such as impermanence and so on. However, we ordinarily overlook these aspects of the aggregates, and wrongly conceptualize them as a unitary self that persists through time, performs actions, and then later experiences the result of those actions. This habitual misconception of ourselves, in turn, makes us think that there genuinely is a fundamentally real distinction between ourselves and others. This belief in the self-other distinction gives rise to the peculiar kind of concern – what Vasubandhu calls self-love – towards the aggregates that we falsely take to be our unitary selves. How does such self-love manifest itself? According to Vasubandhu, it motivates us to perform prudentially rational actions, like saving money, that involve making present sacrifices for the sake of our own future well-being. However, the bodhisattvas are not subject to this misconception about the aggregates. Since they habituate themselves to a picture of reality on which there is no strict self-other distinction, they are able to withdraw their self-love from the aggregates and increase their concern for others. This motivates them to perform altruistic actions that require them to subject themselves to suffering for the sake of the well-being of others.

This defence of the possibility of altruism might be taken to support P2. But, on reflection, it is unsatisfactory. Suppose, for the purposes of supporting P2, we were to rely on Vasubandhu's claim that our attitude of self-love is based on a misconception of the aggregates as a unitary self. That would make the argument from self-love flagrantly circular; for one of its premises would now be based on something very much like the conclusion of the argument. Thus, the argument strikes me as dubious.

So, consider the argument from self-ascriptions:

Like the previous argument, this argument would fail to convince the Brahmanical thinkers. Vasubandhu himself anticipates two objections against it.

There are two distinct objections here. The first objection is an objection from figurative use. The claim is that ascriptions of physical properties like ‘I am fair’, and so on, are not literally true; they are figuratively true. Sometimes, we apparently ascribe to an object a property that in fact belongs to something to which the object is related. Consider the following pairs of sentences (Nunberg, Reference Nunberg1995):

1. (1a) I am parked out the back.

2. (1b) My car is parked out the back.

3. (2a) Yeats is still widely read.

4. (2b) Yeats’ poetry is still widely read.

In at least some contexts, (1a) can express the same meaning as (1b), and (2a) can express the same meaning as (2b). One might argue that the sentences (1a) and (2a) aren't literally true. But we can infer (given the context) that they have the same meaning as (1b) and (2b) respectively, and therefore infer that the expressions ‘I’, and ‘Yeats’ in fact are being used non-literally or figuratively to refer to other things that are related to the semantic referents of these expressions: ‘I’ refers to the speaker's car, while ‘Yeats’ refers to Yeats’ poetry. In the same way, as Vasubandhu notes, one can sometimes figuratively speak of one's servant as ‘I’ because one's servant assists oneself. The important point is that, in these cases, the speaker doesn't literally identify Yeats with his poetry, or herself with her car or servant. The same diagnosis can apply to apparent self-ascriptions of physical properties. When we say, ‘I am fair’, or ‘I am dark’, we may be figuratively using ‘I’ to speak of our bodies. So, these self-ascriptions needn't be taken to correctly ascribe physical characteristics to the semantic referent of ‘I’. Therefore, the argument from self-ascriptions needn't be sound.

The second objection is an objection from arbitrariness: if our own body could be the objective basis (ālambana) – the independently existing intentional object – of our ‘I’-awareness, then why could someone else's body not also be picked out by the first-person pronoun? Here is one way of understanding the worry. On Vasubandhu's view about the first-person pronoun, the first-person pronoun can literally refer to one's body, e.g., when one says:

1. (3a) I am fair.

But there are other contexts where the first-person pronoun can literally refer to something that is distinct from a person's body, e.g., when a person says:

1. (3b) My body is fair.

In other words, the first-person pronoun, on this view, is doubly context-sensitive. On the one hand, it can refer to the mental or physical aggregates associated with different persons in different contexts of utterances. On the other hand, depending on the context, it may refer to different sets of aggregates associated with the same person. This, in turn, opens up room for the following question. What are the contextual factors that determine which aggregates will be picked out by ‘I’ in a particular context? If the answer is simply that this is determined by the speaker's intention, then the semantics of the first-person pronoun will be too unconstrained. For, now, if a speaker intends to pick someone else's body by means the first-person pronoun ‘I’, she may indeed succeed in doing so. As a result, the pronoun may indeed end up referring to the physical aggregates associated with another person in that context. But, surely, that will be a misuse of the expression ‘I’. A natural way of blocking this result will be to say that the first-person pronoun has a fixed referent that doesn't vary depending on the context (as Vasubandhu suggests). And the easiest way of implementing this strategy will be to say that ascriptions like (3a) are at best only figuratively or non-literally true; in such cases, the use of ‘I’ in such ascriptions is intended to convey the same meaning that is ordinarily conveyed by the definite description ‘my body’.

It's time to take stock. In this section, we have considered two arguments for the conclusion that the first-person pronoun refers to the aggregates or a stream of aggregates. But neither of these arguments seems to succeed. There is no obvious way for Vasubandhu to defend the argument from self-love without making it flagrantly circular. And the argument from self-ascriptions of physical properties can be resisted either by appealing to the possibility of figurative uses of ‘I’ or by appealing to the danger that, if this argument were sound, someone else's body could also be picked out by ‘I’.

5. ‘I’ and the Causal Theory of Reference

To see how Vasubandhu might resist these objections to his two arguments, it's best to begin with his response to the second objection from arbitrariness to the argument from self-ascriptions. He writes:

The claim is that ‘I’-awareness is directed at the physical or mental aggregates – the body or the mental events – that it is causally connected to (in an appropriate way), but not at anything else. For Vasubandhu, the ‘I’-awareness is one mental event amongst others, embedded through certain relations of causal dependence within a causally connected stream of physical and mental aggregates. So, when we think first-personal thoughts, we can only pick out by ‘I’ other physical or mental aggregates that are part of the relevant stream, but not physical or mental aggregates that fail to overlap with that stream. That is why ‘I’ cannot refer to a body that doesn't overlap with the physical aggregates that are part of the stream of aggregates that the ‘I’-awareness is part of. Why is that? This is because – as Vasubandhu says – we have formed the habit of applying the concept or the conventional designation ‘I’ to the physical or mental aggregates that the relevant ‘I’-awareness is part of.

I wish to show that this theory of what ‘I’ refers to follows from a more general theory of reference that Vasubandhu subscribes to. The aim of this section is to spell out this theory and explore its consequences for the first-person pronoun.