THE NO SELF ALTERNATIVE

THE NO-SELF
ALTERNATIVE
THOMAS METZINGER
WHENEVER popular or academic debates about 'the self' flare up again, we can
often observe an embarrassing fact: Just because-quite obviously, and in many
cultures-there is a folk-metaphysical and a folk-phenomenological concept of'the
self', and just because someone has put this concept back on the agenda, many
participants automatically assume that an entity like 'the self' must actually exist
and that a relevant and well-posed set of scientific and theoretical questions relates
to this entity. However, there seems to be no empirical evidence and no truly
convincing conceptual argument that supports the actual existence of 'a' self.
Nothing forces us to make this assumption. Therefore, many debates of this type
are threatened by a certain shallowness right from the very beginning and, due to
their endorsement of an unwarranted existence assumption, run the risk of triviality.
As it turns out, the 'no-self alternative' may not be an alternative at all-it could
simply be the default assumption for all rational approaches to self-consciousness
and subjectivity.
In the first part of this short chapter, I will differentiate a number of possible
claims regarding the non-existence of 'the self' and will also try to at least sketch
one typical argument for each thesis. In the second part, I will offer some new ideas
on why all such arguments will always remain counterintuitive for many of us.
I am indebted to Jennifer M. Windt and Adrian J. Smith for a number of a number of helpful
critical comments and for their help with the English version of this chapter.
280 THOMAS METZINGER
1. ANTI-REALISM ABOUT 'TI-IE SELF'
Ontological anti-realism about 'the self'
The standard way to be an anti-realist about selves is to deny that selves are
substances:
(ARS0) The self is not a substance.
A substance is an entity that can 'stand in existence' all by itself, even if all other
existing entities were to disappear. It is 'ontologically self-subsistent', because it can
sustain its own existence. It endures over time and is an ontologically fundamental
entity, because it belongs to the basic building blocks of reality. Of course, there is a
highly differentiated universe of metaphysical theories about 'substancehood'. But,
given this simple working definition, the thesis of ARS0 amounts to the claim that
selves are not self-subsistent entities: they do not endure over time and they do not
belong to the basic building blocks of reality.
There are two well-established ways in which one can argue for this sort of
ontological anti-realism about 'the self '. The first strategy is to confine discussion
to the target phenomenon of self-consciousness and related folk-phenomenological
discourse, including its unwarranted metaphysical assumptions. The second
strategy is to deny the existence of all substances or individuals on theoretical
grounds, tout court. Here the obvious traditional example is the anti-substantialist
metaphysics of Buddhist philosophy (see Siderits 2003, 2007, Chapter 12 below; for
examples of recent discussions, cf. Albahari 2007; Westerhoff 2009 ). In addition,
current debates in the theory of science-particularly, in the philosophy of physics-
have long made it obvious that our standard views about individuals, intrinsic
properties, and relations are obsolete. Instead, an ontic form of structural realism
seems to be the most promising candidate in modern metaphysics, proposing that
relations are all that exist. Said relations do not hold between objects with intrinsic
properties, but even the relata themselves can be decomposed into sets of relations,
with structure always being more ontologically fundamental than objects or 'substances'
(for a paradigm example, see Ladyman and Ross 2009). In a certain sense,
this would lead us into an 'unfounded universe', because the ultimate reality would
then be the nomological structure of the world, where even the identity and
individuality of objects would depend upon this (relational) structure.
Just as current physics is fully compatible with the view that quantum-objects
are non-individuals, the corresponding metaphysical underdetermination of cognitive
neuroscience certainly makes it tenable that mental, psychological, or
phenomenological entities like 'the self ' are not proper individuals at all: they simply
possess no clearly specifiable identity criteria. In some contexts, classical objects, as
well as 'selves', are useful heuristic posits, but nonetheless ontologically dispensable
THE NO-SELF ALTERNATIVE 281
entities. If this is correct, we should rather search for an alternative metaphysics that
renders the predictive success of our empirical theories intelligible without committing
us to theoretical terms referring to individual entities in the world (like 'the
self ' ), or to the truth of certain statements in which these terms appear.
Let us now confine discussion to a sketch of the first strategy mentioned above,
considering only the target phenomenon of self-consciousness and the related folkphenomenological
discourse. Why should we assume the existence of selves in the
first place? The main problem, according to ARS0, is that in our widespread and
naively realistic manner of speaking about 'the self ' we introduce an individual,
making an existence assumption which typically is not backed up by independent
argument.
What are selves? If not 'nuggets of reality', what else could they be? Selves could
be unobservable entities, perhaps conceptual fictions-although in an interesting
sense they are also phenomenological 'everyday objects'. As robust elements of the
phenomenal ontology applied by our brains ( cf. Metzinger and Gallese 2003), they
play a leading role in the everyday phenomenology ofour experiential Lebenswelt
and the pre-scientific folk-psychology it gives rise to. How are we to interpret this
fact on the level of philosophical metaphysics? Let us briefly look at the three main
theoretical options.
If we wanted to establish selves as proper individuals, our first option could be to
take a sober and straightforward approach: we could try to be old-fashioned
scientific realists, and, for example, identify them with their position in space
and time, turning them into single, countable entities by means of the physical
properties they share with no other physical object. With Leibniz (1646-1716) we
would then say that individuality simply amounts to distinguishability, adopting
the Principle ofthe Identity of Indiscernables. However, only bodies can be fully
individuated in this way-but selves are not to be taken as bodies or biological
organisms simpliciter.
If we wanted to find a metaphysical representation of our pre-scientific,
phenomenological intuitions about selfhood, then the second theoretical option
would be to posit some special sort of haecceitas in the sense of Duns Scotus (1266-
1308): a property of primitive 'thisness' or self-identity, a transcendent property of
selves that grounds their intrinsic individuality as well as their numerical identity.
For example, there could be a uniquely instantiable, non-qualitative property of
(say) being identical with Shaun Gallagher, responsible for the irreducible individuality
and the numerical identity of the very person who is the editor of this
Handbook. The haecceity approach has a certain prima facie appeal, because it may
actually capture something phenomenologically important, whereas, metaphysically,
it is less convincing. On the one hand, introducing a 'primitive thisness'
constituting individual substances is, of course, a pure hypostatization. It is a
philosophical move that does not explain anything but just introduces a further,
unobservable property without argument or potential empirical evidence. On the
282 THOMAS METZINGER
other hand, I believe that we may have overlooked an important phenomenological
fact, one that could usefully serve as a constraint on a more comprehensive
theoretical account. There is a distinct phenomenology of singularity, a non-sensory
phenomenology of 'thisness'-for example, in the phenomenology of meditation,
but also in bodily self-consciousness. If we look closely enough, we can discover the
phenomenology of primitive 'thisness' in our own subjective experience. It is
particularly distinct in certain non-conceptual layers of self-awareness (their content
has a non-perceptual but nevertheless 'demonstrative' character), and it certainly is a
feature that requires careful attention in the phenomenology of self-consciousness.
But phenomenological structure per se will never determine metaphysics.c
The third option then would be to develop a theory of'the self ' as a mere collection
of properties. A typical example of anti-substantialist and anti-individualist
approaches to 'the self ' are so-called 'bundle theories'. I have already mentioned
Buddhist philosophy above; the most prominent Western representative, perhaps, is
David Hume (1711-76). Hume would have said that we typically solve the conflict
between experienced sameness across time and the succession of change, between the
phenomenology of identity and that of diversity, by conjuring up a substance: 'the
imagination is apt to feign something unknown and invisible, which it supposes to
continue the same under all these variations; and this unintelligible something it calls
a substance, or original and first matter' (T 220). Conceptually, bundle-theorists
analyze substances as some sort of, possibly complex, relation between properties
(which, however, could in turn either be conceived of as universals or as tropes, i.e.
individual property instantiations). Selves would just be collections of properties.
The open question now becomes what principle exactly is responsible for the
establishment of the complex relation just mentioned-what turns all the features
comprising the self into a coherent whole? Empirically, one could say that selves (or
other entities previously described as substances) are just collections of properties,
which, as a matter of fact, we happen to mentally represent as individual entities. From
the perspective of present -day cognitive neuroscience, this would be a scientifically
plausible strategy: Our brains segment scenes and constitute multimodal, consciously
perceived perceptual objects (e.g. one's own body as a whole) not by attaching
properties to some more basic entity, but by a dynamic, bottom-up process of selforganization
called 'feature-binding' (see e.g. Singer and Gray 1995). All technical
details aside, what is new today is that science offers conceptually clear models of
functional mechanisms which could parsimoniously explain the integration of individual
property-representations into a unified self-representation. This theoretical
model requires no transcendental subject to stand behind the appearance of'a' self as
consciously represented, because it gradually emerges out of the self-organizing
interaction between a large number of simpler components. This possibility-the
appearance of ordered structures without external interaction or a well-defined and
highly specific initial state-simply was not available to thinkers in the past; it is a
novelty in the history of ideas. Therefore, dynamical self-organization is a new
THE NO-SELF ALTERNATIVE 283
theoretical option for the bundle theorist, in metaphysics as well as in phenomenology
(see Metzinger 1995, 2003; Parfit 1982 and Chapter 18; Thompson 2007).
To sum up, the general point about substances, individuals, and identity criteria
is that none of the currently available scientific data determine our metaphysics in
such a way as to make the assumption of the existence of 'a self' necessary.
Moreover, the principle of parsimony demands that we try to find a simpler
metaphysical representation of our current knowledge about self-consciousness,
of its causal history and its constitutive conditions, than what we have traditionally
assumed to be the self-as-substance.
Before proceeding to a short sketch of epistemological, methodological, and
semantic variations on the no-self alternative, let us briefly pause to look at the
wider context in which such discussions take place. The first strategy mentioned
above often emphasizes simply that no empirical evidence could ever ground a
substantialist metaphysics of selfhood. This more modest approach would refrain
from making any general metaphysical claims about the possibility of substances
per se, but would only demonstrate the -irrationality of positing a special sort of
individual substance, namely, 'the self'. Although our very own naturally evolved
cognitive structures (our inbuilt 'naive physics') almost seem to make it a functional
necessity for us to use substance-concepts like the notions of an enduring
particular or an individual substance, which then work as carriers of properties,
nothing in the brain or the self-conscious biological organism as a whole could
even remotely count as a substance in any philosophically interesting sense. We just
don' t find a substantial self anywhere in the world and nothing on the level of
scientific facts determines our metaphysics in this way.
W hat we do find, however, is the phenomenology of substantiality, on the level of
introspective experience: subjectively, we often experience ourselves exactly as selfsubsistent,
enduring entities forming non-exchangeable and irreducible parts of
reality. Moreover, the deeper core of our theoretical problem might lie in the more
subtle fact that our phe!lomenal experience of selfhood not only expresses an
aspect of 'reality' (i.e. the factual realness of the self), but also an aspect of
'metaphysical necessity' (i.e. the impossibility of non-existence, across all conceivable
scenarios). I will return to these points below, and for two reasons. First, all
deflationary or so-called 'weaker' theories of the self (for examples see Ghin 2005;
Legrand 2005) typically miss the mark, by failing to explain what could ground
the phenomenology of the self-as-substance and what the causal history or the
biological function of this illusion of substantiality could have been. Second, the
phenomenal self is the proto-object as such. If anything grounds our naive-realistic
world-view that reality is composed out of individual substances possessing intrinsic,
context-invariant properties and standing in certain relations to each other, it is
exactly the phenomenology of selfhood. Cognitively, the conscious experience of
selfhood leads directlv to the metaphvsical prototvne of 'obiecthood' and to the
J .�. I L J t /
idea of an individual substance. This observation implies the interesting conclusion
284 THOMAS METZINGER
that many of our irresistible theoretical intuitions about substancehood are ultimately
anchored in the conscious experience of selfhood. With this wider context
in mind, let us now return to our brief sketch of theoretical options.
Epistemic anti-realism about 'the self'
'
For those who cannot resist the intuition that individual substances like selves
exist, the obvious move will be to posit the existence of unknowable individuals
possessing an unknowable intrinsic nature:
(ARSE) The self is part of an unknowable realm of individuals, possessing an unknowable
intrinsic nature.
Just as with the unknowable nature of the Kantian Ding an sich (thing in itself), we
could posit an unknowable self pulling the strings behind the observable behavior
of self-conscious agents and underlying the introspective phenomenology of selfhood.
Consequently, all we could ever know would be the structure of the self--for
example, its form of interaction with other selves and the laws and regularities
guiding its cognitive and bodily behavior. Its nature (the sort of entity it really is)
would remain epistemically inaccessible to us. Epistemological anti-realism treats
the self as an unobservable entity, but derives no specific metaphysical position
from its central claim. We can view it as a form of agnosticism.
However, there is at least one specific problem, which arises in the context of
'selves' as objects of knowledge. If they really are unknowable individual substances
or have an unknowable intrinsic nature, then self-consciousness can no longer be
seen as a process that provides us with a direct and epistemologically relevant form
of acquaintance with ourselves. On this view, substantive forms of self-knowledge
are no longer possible: in introspection and in phenomenal self-awareness, we J
never grasp our own true nature-it may well be that we have an essence, but this
essence will forever remain inaccessible to us. Therefore, the distinct, characteristic,
and Cartesian phenomenology of certainty which accompanies self-consciousness is
an illusion. Furthermore, as the phenomenology of certainty is not about the
existence of some merely objective, historical person, but about the indubitable
ontological status of oneself as self,1 epistemic anti-realism renders the self not only
epistemologically irrelevant, but also leaves us with no further metaphysical issue
to be resolved.
1 This type of phenomenological first-person certainty is intimately related to the capacity of
thinking Cartesian I* -thoughts of the form [I am certain that I* myself exist]. In other words,
Descartes claimed that he was certain that he* (he himself) existed, not that he was certain that
Descartes existed. Lynne Baker has made this point very clear in Baker 1998, 2007.
THE NO-SELF ALTERNATIVE 285
Methodological anti-realism about (the self'
It is perhaps on this level that we find the most straightforward and convincing
argument for the elimination of the concept of 'a' self:
(ARSM) Nothing in the scientific investigation of self-consciousness commits us to assume
the existence of individual selves.
As already noted in my discussion of ontological anti-realism above, there are no
strong empirical data whatsoever supporting the existence of selves. There are firstperson
reports, and as such they may function as data-points, but of course there
are no first-person data in any more rigorous sense.2 More importantly, the process
of generating and testing new hypotheses in empirical research programs investigating
self-consciousness, agency, social cognition, etc. simply does not require the
assumption of a theoretical entity by the name of 'the self'. Science can achieve its
predictive success, describe and explain the available data, and integrate them into
a larger evolutionary or neuroscientific framework without assuming that there is a
mysterious thing called 'the self ' which is represented in self-consciousness, initiates
actions, or engages in social cognition related to other mysterious individuals
called 'selves'. Prediction, testing, and explanation can take place in a much more
parsimonious conceptual framework, for instance by introducing the concept of a
'transparent self-model' (Metzinger 2003a, 2006, 2008, 2009).
Semantic anti-realism about (the self'
We refer to ourselves using the word 'I'. But what, exactly, is the meaning of the
linguistic expression 'I'? If-as seems obvious-it doesn't refer to a specific part of
reality in an object-mode of reference, what exactly is its logical or semantic
2 Seriously assuming the existence of 'first-person data' rests on an extended usage of a concept that
is well defined in another (namely, scientific) context. First, the whole concept of a 'first-person
perspective' is just a visuo-grammatical metaphor, without a theory to back it up-currently, we
simply don't know what that could be, 'a' first-person perspective (for a first conceptual
differentiation, see Blanke and Metzinger 2009). Second, 'data' are extracted from the physical world
by technical measuring devices, in a public procedure, which is well-defined and well-understood,
replicable, and improvable; and which is necessarily intersubjective. Therefore, speaking of 'firstperson
data' would rest on an extended usage of a concept which is only well-defined in another
context of application. 'Data' are typically (though not always) gathered with the help of technical
measuring devices (and not individual brains) and by groups of people who mutually control and
criticize each other's methods of data gathering (namely, by large scientific communities). In
particular, data are gathered in the context of rational theories aiming at ever better predictions,
theories that-as opposed to phenomenological reports-are capable of falsification.
Autophenomenological reports themselves can be treated as data, the experience itself cannot (for a
dissenting view, cf Thompson 2007: 474 and 338 n. 10 ) . All of this is not to deny that what are
sometirnes called 'first-person methods' could have an enormous impact on our way towards a
rigorous, empirically based theory of self-consciousness.
286 THOMAS METZINGER
function? I n order to be semantic realists about 'I', we would have to assume that it
reliably connects us to an irreducible, elementary aspect of reality. Semantic antirealism
denies this:
(ARS5) The indexical expression 'I' does not refer to any entity that is ontologically
fundamental.
A lot of excellent philosophical work has been done with regard to the semantics <;>f
the indexical term 'I' (see e.g. Boer and Lycan 1980; Castaneda 1966, 1967; Perry
1979, 1993; Recanati 2007). Does 'I' refer to some sort of invisible object, like some
other linguistic expressions do? Does 'I' refer to a Cartesian Ego? Does it perha:IJs
refer to some sort of 'objective self' (see Nagel1986: ch. 4), because sentences like
'I am Thomas Metzinger' always possess a second reading, a second set of truthconditions
superseding the trivial, purely objective self-identification with a particular,
historical person? Is the object of reference for 'I' always a person, that is, an
entity