 

Reality

. Overview

A Pragmatist Take on ‘Reality’

I began by considering the nature of knowledge in Chapter , articulating
my notion of active knowledge and seeing how propositional knowledge
functions within it. In Chapter  I argued that we should set aside
correspondence realism so that the notion of active knowledge can be
freely developed. Now returning to a more positive vein, there is an urgent
question to be addressed: what is knowledge about? What do we have
knowledge of? When it comes to empirical knowledge, the traditional
intuition is that what we (should try to) know about is reality. We seek
to know facts, which are states of reality. That may be enough of an
answer, when it comes to propositional knowledge. But what are the
objects of active knowledge? Here, too, we come back to the idea that
empirical knowledge is knowledge about reality, since active knowledge is
a matter of our ability to engage productively with reality. These vague
thoughts need to be articulated more precisely, and without falling back
into the notion of reality as completely mind-independent. In order to
achieve clarity on all these issues, I will put forward an operational con-
ception of reality.
In ordinary English, ‘reality’ is just a noun form of ‘real’. What does it
mean in practice when we say that something is real (in the sense of
‘existing’ or ‘actual’)? Philosophical debates about reality should connect
with the concrete methods by which people reach real-life judgements of
what is real. In the discussions concerning realism in metaphysics and the
philosophy of science, it is rare that any concrete criteria for reality are laid
down. In fact, the version of metaphysical realism advocated by Maudlin
and others (see Section .) explicitly removes any inherent connection
between metaphysical truth/reality and the operational procedures of



3 9 084 8:
  Reality
scientific and everyday inquiry. Hacking’s discussion of ‘experimental
realism’ makes a refreshing departure: ‘if you can spray them then they
are real’ (, p. ). In his view ‘reality is parasitic upon representation’:
or better, ‘the concept of reality’ follows from the ‘practice of representing’
(ibid., p. ). To get a sense of the concrete and operational meaning of
reality, we must pay proper attention to our practices, in the spirit of
pragmatism as explained in Section .. I also work, to a point, in the spirit
of ordinary-language philosophy as advocated by Austin ([] ,
pp. –):
our common stock of words embodies all the distinctions men have found
worth drawing, and the connexions they have found worth making, in the
lifetimes of many generations: these surely are likely to be more sound,
since they have stood up to the long test of the survival of the fittest, and
more subtle, at least in all ordinary and reasonably practical matters, than
any that you or I are likely to think up in our arm-chairs of an afternoon –
the most favoured alternative method.
Linguistic and scientific practices are surely not infallible guides to philos-
ophizing, but they are sources for plausible insights worth examining.
They provide as good a starting point as any in epistemology, with clear
implications for metaphysics, too.
In everyday life we do routinely make judgements of what is real.
Treating some things as real and others as not is a very important part of
how we live. Scientists do the same, though they may not often invoke the
terminology of ‘reality’ in expressing their judgements, preferring to talk
about statistical significance and such. Within all kinds of concrete practices
we know very well how to judge what is real and what is not, arriving at
verdicts like ‘Ghosts aren’t real’, ‘The Loch Ness Monster isn’t real’, ‘The
placebo effect is real’, ‘Anti-matter is real, and probably dark matter, too’, or
‘The mesosome, long believed to be a real entity within bacterial cells,
turned out to be an artifact of the chemical fixation process used to prepare
the cells for electron microscopy.’ Scientists sometimes have uncertainties
about specific answers to ‘real or not’ questions in practice (e.g., concerning
mesmerism or the Little Ice Age), but they know how to go about deciding
the answers, while admitting that their judgements are fallible. In all these
situations no one seems to be confused about what being real means, even
when there are disagreements about answers to specific questions.


For an excellent account of the mesosome episode, see Rasmussen ().

A disambiguation of ordinary language is necessary here. The English word ‘real’ can also mean
‘genuine’ (the opposite of ‘fake’ or ‘imitation’), ‘actual’ (as opposed to ‘fictional’), or ‘exemplary’ (as

3 9 084 8:
 Overview 

Operational Coherence and the Meaning of Reality

I propose the following definition as the core of a coherence theory of
reality (real-ness): an entity is real to the extent that there are operationally
coherent activities that can be performed by relying significantly on its existence
and its properties. Recall that ‘operational coherence’ means aim-oriented
coordination in an activity, a matter of doing what makes sense to do (see
Sections . and .). And when I say that an activity ‘relies’ on an entity,
the sense of reliance here is not one of metaphysical necessity, but of actual
use and need. I think this definition of reality (real-ness) is consonant with
many well-established quotidian and scientific uses of the word, and it can
also do very useful philosophical work by helping us spell out what realism
should mean in the context of empiricism and pragmatism. When we say
something like ‘The placebo effect is real’, we are expressing our judge-
ment that the named thing is operative in some processes. Its existence is a
difference that makes a difference. When physicists say that positrons are
real, or when we say that dogs and cats are real, that means one can do
meaningful and effective things with them, like making PET (positron
emission tomography) scans, or taking them to the pet shop.
It is important to note that what I am making here is a semantic move,
in the sense that it concerns the very meaning of the word ‘real’. I am not
proposing operational coherence just as an indication of metaphysical
reality, which serves as evidence that something is real. Rather, it is about
what we mean by something being real, and I suggest that there isn’t
anything else that ‘being real’ means in an operational sense. (I am
proposing a constitutive criterion of real-ness, not an epistemic criterion.)
To draw a rough analogy: if you ask me ‘How do you tell if you have a
headache?’ then the answer is ‘Of course, I check if my head hurts.’ But
that doesn’t mean that a pain in the head is a symptom of a headache; no, a
headache is a pain felt in the head; that is what the word means. There isn’t
some Platonic thing called ‘HEADACHE’ out there, of which the hurt
feeling in my head is merely a symptom or manifestation. Likewise, I am

in ‘a real gentleman’); such meanings are not my immediate concern here (and all these meanings of
‘real’ would not translate into the same word in all languages). People know not to conflate these
different senses of ‘real.’ This ‘Louis Vuitton’ handbag isn’t real, but yes, it is a real handbag. Harry
Potter is a real fictional character, while Parry Hotter isn’t real; however, I am not going to try to take
Harry to lunch like a real person. (I thank James Tartaglia for prompting me to clarify these
distinctions.)

To avoid confusion: this is a very different matter from ‘semantic realism’, which is a matter of
reducing the understanding of a statement to the knowledge of its realist truth-conditions.

3 9 084 8:
  Reality
proposing that we use the term ‘real’ to mean having the capacity to
support coherent activities, in which case it would be misleading to say
that an entity is able to support coherent activities because it is real. That
kind of causal talk would only add an unilluminating and intractable type
of metaphysical layer to our thinking.
Many people will be worried that my proposal distorts or perverts the
meaning of ‘reality’ too greatly. That is a legitimate kind of concern, which
can be debated further. But first of all let me make sure that the nature of
my proposal is conveyed clearly. I am not claiming that the definition of
‘real’ I am proposing here encompasses every existing usage of the word.
Rather, I am proposing that the capacity to support coherent activities is
what we should mean by ‘real’, because I think it will be conducive to
productive discourse, while being reasonably faithful to enough of the
actual usage currently embedded in various practices. So what I am
engaged in can be seen as a project of explication as conceived by
Carnap, or an attempt at ‘conceptual engineering’. The important point
to recognize is that we have some choice in what we mean by a term. So,
we could decide to postulate an unobservable disease-entity called ‘head-
ache’ whose one and only symptom is pain in the head, but we need to ask
how that would be a productive move. Any semantic proposal of this sort
is to be judged by its fruits in a pragmatist manner, because there is no
higher court of appeal.
We should, of course, also address the separate question of how we
know if something is real. According to my meaning of ‘real’, we have first-
hand knowledge that an entity is real if we (personally) know how to
perform some operationally coherent activities that rely on its existence
and its properties. If we are aware of some such activities that other people
can perform, then we have second-hand knowledge of the reality. If there
are coherent activities that someone can perform relying on the entity in
question but we are not aware of this, then the entity is real but we don’t
know that it is real.


This is where I diverge even from Hacking’s ‘entity realism’. As David Resnik () points out,
Hacking’s argument, contrary to his intentions, collapses into a scientific realist argument from
success, taking reality to explain empirical success (see Section .). It makes no great difference here
that Hacking is thinking about the success of interventions rather than predictive success. But
Hacking’s position is less problematic if it is taken as a low-key epistemic thesis. When he says that
‘engineering, not theorizing, is the best proof of scientific realism about entities’ (, p. ), we
should take ‘proof’ simply in the sense of ‘how we know’.

For exemplary work under the banner of conceptual engineering, see Haslanger (); Brun
(); Cappelen (); and Dutilh Novaes ().

3 9 084 8:
 Overview 
Distinguishing this epistemic condition from the definition of reality
should reassure those who fear that my conceptions make the existence of
external reality dependent on our subjective knowledge of it. So, yes, a tree
does fall in the forest even if no one is there to hear the sound – and all
that. Let us not get into an unnecessarily learned discussion of counter-
factuals and modalities here. When people say ‘of course the tree fell even
though there was no one to hear it’, they mean that anyone present would
have heard it, that anyone standing on the unfortunate side of that tree
would have been crushed, that we can go now and see that the tree has
fallen, and so on. You may mean something additional and fancier by the
tree ‘really falling’, but in that case whatever you are asserting there is not
obviously the case. The real-ness of an entity (or an event) is a matter of
whether there are coherent activities it can facilitate, not a matter of
whether our current community of people can and will actually perform
such activities. So something can indeed be real without us knowing
anything about it, and our lack of knowledge does not make anything
non-existent.
Still, I expect many people will be uncomfortable with the idea that real-
ness depends in any way on what we think or do. I think that this worry is
at least exaggerated. Consider, for example, what I will call the inaccessi-
bility argument for metaphysical realism, which encompasses the ‘argu-
ment from the past’ discussed in Section .. The inaccessibility argument
works by pointing to entities that are inaccessible to human inquirers, yet
seem undoubtedly real; if there are such entities, then there are real entities
even if there could be no human cognition or activity involving them.
Surely dinosaurs were real, and so was the asteroid or comet whose impact
 million years ago wiped them out, but there were no humans then, and
therefore no activities performable by anyone involving the dinosaurs or
the asteroid. Then do I not have to deny the reality of these entities?
In response, I would first of all point out that it is not the case that we
(in the present) are completely lacking in the knowledge of these inacces-
sible entities. There are coherent present activities that we perform by
relying on the comet’s past existence and properties, such as identifying
its traces in geological strata. When we engage in a coherent explanatory
activity involving the comet colliding with the earth in the distant past and
causing mass extinctions, or a coherent observational and classificatory
activity involving dinosaur fossils, then we begin to have a knowledge of
the reality of the comet and the dinosaurs. How else does it make sense to
say that scientists know that such things are real? And why else are we
actually so sure that there were dinosaurs?

3 9 084 8:
  Reality
And so it goes with other kinds of situations of inaccessibility, too, such
as entities in a distant corner of the universe that we can never reach, or
black holes that we can never directly see (even as astronomers are now
busy making images of the ‘accretion discs’ around them). Hacking’s
criterion of direct intervention for the knowledge of reality is too restrictive
even for pragmatists, and he need not have conceded serious uncertainty
about the reality of astronomical objects on the basis of that criterion (see
Hacking ; Shapere  for this debate). It is not the case that we
have to be in the same spatio-temporal location with an entity in order to
be engaging with it in coherent activities. Even when I am just looking at
an ordinary object not far away, you could quibble and say that the
interaction is actually indirect, since it involves photons bouncing off the
object and being received by my eyes and causing complicated nerve signals.
You could also point out that actually the information conveyed to my mind
is about something that existed in the past because the photon leaving the
object takes some time to reach my eyes. But nothing of philosophical
significance follows from such points. Sure, we cannot really ‘go to’ the near-
past cat that I see walking down the road, but what of it? And this is just a
very mild version of the inaccessibility of the dinosaurs of the distant past
that I can’t take a time machine to go see. The cat has given me observable
traces, and so have the dinosaurs. I judge them both to be real, as would the
metaphysical realist. No one has scored a point here.

Progressivist Constructivism
A related worry would be that there is a kind of constructivism in my view
of real entities. It may seem strange to tie the metaphysical notion of reality
to operational coherence, which is based on pragmatic understanding.
Doesn’t this deprive reality of its mind-independence? The disambiguation
of mind-independence I made in Section . should be helpful here. I take
reality as mind-framed, but not mind-controlled. In fact, not being subject
to mind-control is an important hallmark of reality; in my previous work
I went as far as to say: ‘I propose to think of external reality as whatever it is
that is not subject to one’s own will’ (Chang a, p. ). This accords
with an ordinary-language sense of ‘reality’ as well; the first definition of
‘reality’ given by my trusty Collins English Dictionary (th edn, ) is
‘the state of things as they are or appear to be, rather than as one might


See Skulberg () for a fascinating account of the history and practice of black-hole imaging.

3 9 084 8:
 Overview 
wish them to be’. Take this as the new realist common sense: all entities are
mind-framed, but only a small portion of them are mind-controlled.
Real entities are, at least in some respects, mind-uncontrolled. They do
not obey our wishes; even when they do what pleases us, they are not
doing as we please. Even something we define in a purely conventional way
cannot be controlled as we wish. Take constellations: after we fancifully
connect up a certain group of stars as ‘Orion’ or ‘the Big Dipper’, we
cannot dictate what shape the group will have a million years later when
the individual stars will have moved around. That is to say, even though
Orion is obviously a mind-framed entity, it is not mind-controlled. The
question of reality is not a question about unqualified mind-independence,
contrary to the metaphysical-realist instinct.
The constructivism inherent in my pragmatist view of reality is not
anything that should worry those who seek empirical knowledge, any more
than empiricists should fear van Fraassen’s constructive empiricism (see
Boon  for a positive take on epistemological constructivism). It is not
that we can create reality just by creating a new concept; however, the
presence of a concept is a prerequisite for there being any specifiable entity
that we can speak or think about. We can make concepts as we like, but
whether the entities they specify turn out to be real is not up to us. If we do
manage to create a new concept that designates an entity with which we
can engage in coherent activities, then that is a successful inventive process.
This process deserves to be called ‘invention’ more than ‘discovery’, but its
success is not in our control, and that is just the same in the technological
processes of invention. Guglielmo Marconi did not simply conjure up
wireless telegraphy in any arbitrary way he fancied; the coherence of his
operations constituted a great achievement precisely because it was
not guaranteed.
I also want to stress that whatever constructivism that is present in my
view of reality does not stand in the way of progress in our knowledge of
reality. On the contrary, as I will discuss further in Chapter , the
pragmatist notion of reality is perfectly suited for encouraging the growth
of knowledge, as it helps us elucidate the various forms that epistemic
progress can take. We may conceive a new entity, and learn of its reality by
successfully crafting some coherent activities on its basis. We may also
improve our knowledge further by working to enhance the coherence of
these activities. We may also seek to learn if the entity is real in additional
domains, by trying to devise coherent activities in those new domains. We
may also try to increase our knowledge by coming up with additional
activities in an already familiar domain. Generally, our knowledge grows as

3 9 084 8:
  Reality
we learn to engage in more coherent activities and more-coherent activ-
ities. There is much more to say about how our learning about reality
actually takes place in practice, as I will discuss further in Sections .
and ..

Can There Be Reality from Mental Activities?

There is one other issue that is worth flagging up briefly before I go on. My
notion of operational coherence would seem to apply in any setting in life,
including the domain of purely mental activities. Does that mean we need
to grant reality to imaginary or fictional entities if they can support
operationally coherent mental activities? A similar question would arise
regarding formal entities postulated in mathematical systems. Are imagi-
nary numbers real because we do carry out many coherent mathematical
activities relying on them? I think the answer is ‘yes’ to all of these
questions, and any initial sense of absurdity should dissipate upon more
careful consideration.
One quick remedy for the sense of absurdity would be to distinguish
different types of reality, going in the direction of a pluralist metaphysical
attitude that John Dupré (, p. ) once advocated under the name of
‘promiscuous realism’: there can be multiple valid taxonomic schemes in
the same domain. This thought can be extended further. We can imme-
diately say that an entity has material or physical reality if it can facilitate
coherent activities that treat it as a material or physical entity. The square
root of !, say, is not a material object and does not do physical work in
any coherent activities, but it has formal reality in the sense that it plays a
crucial role in many very coherent mathematical activities. Likewise, the
Excalibur is a real entity in the fictional realm.
But in all these different domains, there is one thing in common:
whether an activity we devise turns out to be coherent is not up to us.
This is the case even in purely mental activities: the idea of ‘a square circle’
can be put up verbally, but cannot be executed on paper, or even in visual
imagination. Likewise, one can seek the limit of a series, but that will not
be a coherent activity if the series is divergent. In this sense, the lack of
mind-control underlying the intuitive idea of reality is present in mental
activities as well. And so is the lack of pre-determined certainty in the
outcome, which is the fundamental characteristic of the empirical domain.
Consonant with the spirit of pragmatism as I take it (see Section .), we
can recognize that in this sense all activities in life are empirical, and they
deal with real entities.

3 9 084 8:
 Overview 

How Not to Talk about Reality

So far I have discussed what it means for an entity to be real (‘reality’ as
real-ness). Now let me address a different sense of ‘reality’ that is the more
usual subject of metaphysical discourse, namely the thing that exists ‘out
there’, the ‘world’, and so on. I propose to take this concept, too, in the
most humble and concrete way possible. Let us take realities simply to
mean entities that are real. With this meaning, ‘reality’ is a countable
noun, and ‘reality’ without an article in front of it is not grammatical
(unlike ‘reality’ that means real-ness). It is annoying that ‘reality’ as the
noun form of the adjective ‘real’ is ambiguous in its meaning, but the two
meanings as I propose to take them (‘real-ness’ and ‘something real’) are
easily consonant with each other. Some other English nouns formed from
adjectives also exhibit the same duality: e.g. ‘absurdity’ meaning both
absurd-ness and something that is absurd. It is slightly awkward to say
that a sparrow is ‘a reality’, but if you find it difficult to get used to that
usage, you can simply spell it out as ‘a real entity’ each time. The Collins
English Dictionary (th edn, ) gives the following definitions for
‘reality’ (the first of which I quoted earlier):
 the state of things as they are or appear to be, rather than as one might
wish them to be.  something that is real.  the state of being real. 
Philosophy. a that which exists, independent of human awareness. b the
totality of facts as they are independent of human awareness of them.
I am picking up definitions  and  as the chief meanings, and propose to
understand the others on the basis of them.
Now, all this might be quite a let-down to the deep metaphysical mind,
and my notion here certainly does not sit so well with the kind of reality
that many philosophers like to talk about – namely, ‘the world’ (or the
‘external world’), the reality, or ‘Reality’ with a capital R – the totality of
existence. I have difficulty thinking of any coherent activities involving
Reality in this sense. Such a grandiose notion only occurs in the kinds of
metaphysical, religious or mystical discourse that I don’t know how to
engage in sensibly, and will not enter into in print. If I innocently stepped
into questions like ‘Is the external world real?’, I would probably never be
able to come back out of the rabbit-hole. I think we would do well to avoid
thinking in terms of a totalizing kind of Reality, because it is a much-too-


C. I. Lewis (, ch. ) employed this device of distinguishing ‘Reality’ and ‘reality’ for
similar purposes.

3 9 084 8:
  Reality
lofty construction with no concrete operational purchase. Markus Gabriel
([] , p. ) makes a playful yet significant point when he argues:
‘something exists only when it is found in the world . . . the world cannot
in principle exist because it is not found in the world’. Perhaps more
palatable to most philosophers would be Nicholas Rescher’s (, p. )
point that terms such as ‘the world’, ‘the universe’, ‘the true facts’ and so
on are mere placeholders, vacuous unifiers, designating ‘an inherently
empty container into which we can put anything and everything’.
It is very difficult to make sense of the claim that ‘the world’ is real,
according to my notion of real-ness based on operational coherence. There
isn’t anything we can do with ‘the external world’ as a whole. Can we even
talk meaningfully and usefully about the whole of Reality, or the totality of
all realities? Really, what is ‘the world’? I understand what the ‘world’
means in phrases like ‘world champion’ or the ‘World Health
Organization’, as a collection of all the countries or other kinds of human
communities (세계 se-gye, in Korean). Similarly it might mean the Earth
(지구 ji-gu), our planet, as when we talk about the ‘climate map of the
world’. But Anglophone philosophers have a habit of saying ‘world’ to
mean something more like the universe (in Korean the universe is 우주
u-ju, and in everyday contexts hardly anyone would think of translating
English ‘world’ into this word). On reflection, it is really not clear what it
is that we philosophers have in mind when we talk about ‘the world’. If
you think that is a straightforward matter, tell me this: does the world
include God in it? Even when cosmologists theorize about the universe,
they deal with only particular aspects of the universe conceptualized in very
specific ways, not just ‘all that there is’. I find it difficult to see how we can
achieve any sensible aims through the kind of move exemplified by writing
‘Ψ’ for the quantum wavefunction of the whole universe.
It is not easy to avoid talking about the grand universe-scale notion of
reality completely, but I think we should try. When I advanced a doctrine
of ‘active realism’ in a previous publication, I defined it as ‘a commitment
to maximize our learning from reality’, taking ‘reality’ as that mind-
independent something in which we live, which can resist our attempts
to deal with it in some particular way that we might prefer (Chang a,
p. ). That was a mistake. As I discussed in Chapter , reality in this
sense is like the Kantian thing-in-itself, about which we should say
nothing; it doesn’t make sense to think that we can learn anything
expressible about it. As Goodman (, p. ) was at pains to stress, all
that we can ever actually deal with are ‘versions’ of the ‘world’:

3 9 084 8:
 Overview 
We cannot test a version by comparing it with a world undescribed,
undepicted, unperceived . . . While we may speak of determining what
versions are right as ‘learning about the world’, ‘the world’ supposedly
being that which all right versions describe, all we learn about the world
is contained in right versions of it; and while the underlying world, bereft of
these, need not be denied to those who love it, it is perhaps on the whole a
world well lost.
And even Goodman’s way of talking is only a ladder to be kicked away
once we have climbed it. ‘Versions of the world’ is a phrase that inevitably
raises the expectation that there is such a thing as the world, of which we
make versions. That invites the accusation that Goodman is denying the
reality of the evidently existent ‘underlying world’. This is the same
problem that I have mentioned in relation to perspectivism in Section ..

Pragmatist Metaphysics
With my proposed conception of reality I am trying to set the scene for
what Sami Pihlström () calls pragmatist metaphysics. In order to
have any kind of reasoning and discourse, we need to conceive identifiable
and trackable entities with clear properties. This is the business of meta-
physics as I see it. It is commonly thought that pragmatism, like positiv-
ism, should avoid metaphysics altogether or somehow dissolve it into
something non-metaphysical, but that is not the most productive view to
take. I see pragmatist metaphysics as the business of building good ontologies
that will support coherent activities. In science and other empirical realms of
life, the challenging ontological task is to create concepts that specify real
entities (or, realities). As I will discuss further in Chapter , this is what
realists should aspire to do, unless there are particular reasons to take
fictionalist or instrumentalist attitudes in some particular situations.
Identifying the realities that we should be dealing with is a crucial part of
any process of inquiry, and concept-making is not just a matter of thinking
up ideas. For concepts intended to specify real things, their uses have to
involve coherent arrangements of material and social settings. This task of
concept-creation cannot be avoided, and it needs to be done well. Concepts
are not simply handed to us by God or through an intuitive access to
Platonic heaven, and our inborn instincts are not sufficient to give us good
concepts when we need them. Successful concept-building is an ‘engineer-
ing’ process in which we fashion realities, entities that are not subject to
mind-control but amenable to our understanding and engagement.

3 9 084 8:
  Reality
Having a good operational ontology is crucial for any kind of cognitive
activity. Many years ago now I was delighted to meet a computer scientist
whose business card proudly bore the title of ‘Chief Ontologist’ for his
firm; ‘ontology’ is a well-established technical concept in his field, and so it
should be in every field of research. This point has received a new
recognition in the ‘data-centric’ sciences such as genomics or proteomics
with their need to craft the right ‘bio-ontologies’, as discussed by Sabina
Leonelli (, p. , and passim). Without an ontology we cannot say
anything intelligible, make any kind of analysis, or engage with nature in
any specific and directed way. So there is something else fundamentally
right about Hacking’s perspective again: realism concerning entities is
prior to any realism that we can have concerning the truth of the state-
ments that we make about the entities in question. If the entities that we
speak and think about were not real, it would not make any sense to
maintain that the statements we make about them are true. It is necessary
for us to grapple with ontology if we are to talk about truth.
What I am advocating here is a modest and piecemeal practice of
naturalistic metaphysics – not giving a grand view of ‘how the world is’
arising from abstract reflections removed from experience, but allowing
our knowledge of ontology to emerge from well-established practices, in
the spirit of Nancy Cartwright’s work (; ). This is much broader
than ‘naturalism’ as it is often meant, which tries to mould metaphysics in
line with the propositional content of accepted scientific theories. Rather,
we should learn about real things (including the very fact that they are real)
by seeing how we can create various operationally coherent activities
relying on them. Scientists are seriously engaged in the business of crafting
new and better concepts that support ever-multiplying coherent activities.
Again, ‘salvation is through work’, and metaphysicians should pay atten-
tion to scientific work, with full respect for the ingenuity and sustained
effort of scientists. But doing naturalistic metaphysics should not mean a
renunciation of philosophical judgement or responsibility. Scientists do
not always work as well as they could or should. Scientific theoreticians
often make unwarranted pronouncements upon the nature of reality,
which are regarded with suspicion especially by many of their experimen-
talist colleagues. And entire scientific communities may enter into uncrit-
ical groupthink preventing the emergence of more coherent alternatives
(consider, for example, the ‘central dogma’ of molecular genetics that held
back considerations of epigenetic inheritance, or the uncritical acceptance
of Newtonian absolute space and time). Philosophers can and should ask

3 9 084 8:
 How Mind-Framing Works 
critical questions concerning the reality of the items found in current
scientific ontology.
Having articulated pragmatist notions of realness and realities, in the
next chapter I will move on to discuss what it means to make true
statements about realities. (If you are reading at the surface level you
may now want to jump ahead to Chapter  from here.) In the rest of this
chapter I want to consider more carefully several aspects of the practice of
pragmatist metaphysics, outlining some key steps in the practice-based
building of realities. Section . will discuss further how the mind-framing
of entities is done in our activities and how the process of conceptual
development unfolds. Section . will discuss the processes by which the
mind-framed entities may be validated as realities. These discussions will
give some detailed illustrations of how concepts, materials, experience and
aims develop in full mutual entanglement. The kind of pragmatist meta-
physics that I am proposing here will naturally lead to ontological plural-
ism, as shown in Section .. Pluralism is a key aspect of the whole outlook
on knowledge that I am advocating in this book. Finally, Section . will
challenge a common reductionist ontology of physical composition that
provides a strong source of resistance against pragmatist and pluralist
metaphysics.

. How Mind-Framing Works

So far I have rather abstractly advanced the notion that realities are mind-
framed yet mind-uncontrolled entities. Now I want to spell out this idea in
more concrete detail. The first step is to think more carefully about the
actual processes by which the mind-framing of entities works. Realities are
framed to fit what we do, and we need to understand how we actually
create concepts and use them in our epistemic activities. The mind-
framing of entities begins with ontological principles that are adopted
because they are necessary for the performance of certain types of activity;
some of the most fundamental features of the realities in our lives have
their origins as conceptual prerequisites of our actions. Here I am building
on Kant’s fundamental insight that the mind provides certain a priori
principles that frame experience. But the mind-framing of entities also has
a more conscious and deliberate aspect of concept-design. We introduce
new concepts and develop existing ones, in coordination with other
existing and developing concepts, for various specific purposes. As sug-
gested by C. I. Lewis, Michael Friedman, Sami Pihlström and others in a

3 9 084 8:
  Reality
revision of the Kantian insight, we have a choice of which a priori
principles to adopt. This element of freedom becomes obvious when we
consider the processes of concept-development in science, as I will illus-
trate with the cases of the establishment and progressive development of
the concepts of ‘temperature’ and ‘acid’.

Mind-Framing and the A Priori

It is now time for me to say more precisely what mind-framing means. In
relation to ‘mind’ I want to be quite liberal and consider, following Niiniluoto
(, p. ): ‘perspective, point of view, practice, discourse, linguistic or
conceptual framework, scientific paradigm, language-game, form of life, tradi-
tion, and style of thinking’. For my purposes, the important general point is
that we use concepts in order to frame entities. But what, exactly, is ‘framing’?
To ‘frame’ means various things in ordinary English, and the meaning I intend
is something like ‘to formulate; form or articulate’ (definition  in the Google
Dictionary); this is also in line with the meaning of the noun ‘frame’ as ‘a basic
structure that underlies or supports a system, concept, or text’ (definition ).
But doesn’t the image of framing actually support the
correspondence-realist view of reality, if the entity already exists well-formed
and the frame just encloses it, as with a picture-frame? That is not how I intend
the term ‘framing’. Going back to the etymology can help dislodge some
unhelpful intuitions. In Old English, framian meant ‘to be useful’, which then
evolved into the Middle English meaning of ‘to make ready for use’ – a
reassuring word-origin for pragmatists! An important case of ‘making ready’
was to ‘prepare timber for use in building’, which then led to the idea of the
timber frame of a building, hence the basic structure of something. And then
the thing one puts around a painting was metaphorically called a frame! It
might be more useful, for our purposes here, to think in terms of the frame-
work of a house. The frame does not contain or enclose the house; rather, it is
an integral part of the house, without which the house would not stand.
Mind-framing is the specification of an entity in a form that can be
handled by the mind. My view about mind-framing amounts to a denial of
what Sam Page calls ‘individuative mind-independence’ (, p. ): ‘To say
that the natural world is individuatively independent of us is to say that it is
divided up into individual things and kinds of things that are circumscribed by
boundaries that are totally independent of where we draw the lines.’ In Page’s


The etymology is quite intricate. The account I give here is only one strand, extracted from the
discussion in the Google English Dictionary online, provided by Oxford Languages.

3 9 084 8:
 How Mind-Framing Works 
terminology, I want to say that all entities are individuatively mind-dependent.
This may sound like a wild metaphysical claim, but it really is just a productive
tautology that is designed to shift our thinking in a certain direction: nothing
should be called an ‘entity’ if it is not individuated.
But if realities are not controlled by the mind, how exactly is it that
the mind ‘frames’ them? Kant gives us the most productive jumping-off point
with his recognition of the a priori dimension of empirical knowledge, which in
my view remains one of the greatest lessons in all of philosophy. And I think
the full significance of Kant’s insight can only be recognized when we consider
how the a priori operates in the context of action, following the development
of Kantianism in the directions proposed by Grene () and Pihlström (;
). All human perception, thinking and communication take place within
the confines of certain a priori concepts and principles, in terms of which we
perceive, think, talk and act. These principles guide the mind-framing of reality
at the most fundamental level. Where we have to depart from Kant is his
insistence on the apodictic certainty of a priori judgement. Rather, what
happens in mind-framing is a suggestion, a proposal to engage with prospec-
tive realities in a certain way. We start by postulating a certain type of reality, to
see if such a conception can frame coherent experience. The a priori is what
the mind imposes on experience, even though its validity is not guaranteed
and only achieved through operationally coherent activities. But mind-framing
is not a matter of random conjecture. Our starting-point is strongly constrained
by the evolutionary path that humans have taken, in a direction that is
positively adaptive on the whole. We have predispositions to think in certain
ways, and those predispositions are also linked with our bodies, since we have
embodied minds.
Recognizing the lack of absolute and eternal certainty in the a priori
does not mean dismissing the very different roles that a priori and a posteriori
judgements play in any given system of practice at a given stage of its
development. This is one place where Quine’s point about the in-principle
holism of knowledge has often been taken in an unhelpful direction. What we
need to recognize is that in any given situation some propositions are treated
as empirical hypotheses open to testing, and others are taken for granted and
protected from falsification. The latter are operative in the process of mind-
framing; they are the principles that enable us to conceptualize and identify
the entities we want to use and investigate. These principles are neither
unalterably fixed, nor simply dispensable like ordinary empirical hypotheses.
David Stump () has given a very helpful characterization of such principles
as ‘constitutive’ (instead of a priori), and made an informative survey of various
philosophers who have recognized them.

3 9 084 8:
  Reality

Activity-Based Ontology
The first stage of mind-framing stems from the types of mental–physical
activity that we are poised to take in the regular course of life. In previous
work (Chang ; a) I proposed that we should recognize a certain class
of ontological principles (or, metaphysical principles) as necessary conditions
for carrying out certain types of epistemic activity. These are the most funda-
mental principles with which we frame reality, and they are a priori commit-
ments often made without explicit agreement or even articulation, simply
through our deciding to undertake certain activities. If we did articulate these
commitments, they would say: ‘If we want to engage in a certain type of
activity, then we have to presume the truth of some particular metaphysical
principles.’ What we have here is a quasi-Kantian conditional or contingent
transcendental argument – laying out the necessary preconditions for an
activity that we engage in.
For example, if we are to engage in the business of making inductive
predictions, we must take it for granted that the same conditions will results in
the same outcome. Let us call this the ‘principle of uniform consequence’. An
attempt to justify this principle by itself is futile, and that is why the ‘problem of
induction’ is not solvable. It is correct to call induction a ‘custom’ as Hume did,
but that misses the most important aspect of the situation. If we do participate
in the form of life in which we predict what is going to happen next on the
basis of what we have experienced before, then the principle of uniform
consequence becomes an a priori principle. We can deny this principle, but
then it would make no sense to attempt inductive prediction. What underlies
the sense of necessity here is a pragmatic–hermeneutic kind of impossibility of
doing without something. The denial of an ontological principle while we
are engaged in the activity that requires it would generate a sense of
unintelligibility. What is involved here is just the kind of pragmatic sense-
making, doing things designed to lead to the achievement of our aims, that
I discussed in Section . in relation to the hermeneutic dimension of
operational coherence.


I would maintain that even in ‘material’ inductions as John Norton () conceived them, a local
version of the principle of uniform consequence is in action. And even if what we are making is
probabilistic predictions, what we are doing is applying the principle of uniform consequence to
groups of events. Determinism is a commitment to engage in prediction in every individual case.

I want to argue that other types of impossibility are actually grounded in pragmatic impossibility,
being metaphorical extensions of the latter. This is why it would be futile to try to analyse pragmatic
impossibility further. In working out the notions of necessity and possibility sketched here, I wish to
build connections to Roberto Torretti’s ideas on the subject (, ch. ).

3 9 084 8:
 How Mind-Framing Works 
Table .. A partial list of activity–principle pairs

Activity-type Ontological principle

Inductive prediction Uniform consequence
(Contrastive) Explanation Sufficient reason
Narration Subsistence
(Linear) Ordering Transitivity
Voluntary action Agency
Intervention Causality
Empathizing Other minds
Individuation Identity of indiscernibles
Testing-by-overdetermination Single value
Assertion Non-contradiction

For each well-defined type of activity, there is an associated ontolog-

ical principle that makes it performable and intelligible. Table . gives a list of
some important pairs of activity-type and ontological principle, and I will
explain each item very briefly here. If we want to explain why something (as
opposed to something else) happened, we have to assume that when there is
an observed difference, there is a reason behind it; this may be considered a
weak version of the principle of sufficient reason. The activity of narration
requires that the stories we tell have subjects whose identities last through
time, which ‘house’ the changes that are narrated. Paradoxically, without
postulating something that lasts, it is impossible to describe any change. If
we want to put a set of entities in an ordered sequence, we must assume that
the relation that forms the basis of ordering is transitive. Engaging in voluntary
action makes no sense unless we presume that our will directs certain parts of
our bodies to move in certain ways. Intervening in the world with our own
actions requires a presumption that our actions do make other things happen.
In the activity of understanding other people’s intentions and emotions, one
needs to presume that they possess intentions and emotions like oneself. The
identification of an object or a property as a distinct thing depends on an
ontological principle close to Leibniz’s principle of the identity of
indiscernibles. (The principle itself can be divorced from various uncertain
uses that Leibniz and others have made of it.) When it comes to physical
properties of objects, there is what I have called ‘the principle of single value’,
which dictates that it can have no more than one definite value in a given


I thank Roberto Torretti for this suggestion, as well as much inspiration and detailed discussions in
the development of my thinking about ontological principles, even though he advised against using
the term ‘ontological’ in this context.

3 9 084 8:
  Reality
situation (e.g., a stick cannot be  m and  m long). This principle is a
prerequisite for the activity of ‘testing-by-overdetermination’, in which we
determine the value of a quantity in two different ways (e.g., by prediction
and observation); if the values match, that gives credence to the basis on
which we made the two determinations. Even the logical principle of non-
contradiction may be an ontological principle, associated with the epistemic
activity of asserting a proposition. Asserting something makes no sense unless
we refrain from denying what we have just asserted. This one-to-one pairing of
activity-type and ontological principle might seem too neat and contrived, but
it makes sense considering that the ontological principle and the activity-type
partially constitute each other. And the one-to-one correspondence only
applies to the most basic types of activity; concrete and complex activities will
require multiple principles.
It is satisfying to see how our basic metaphysical conceptions arise
from the way we engage with the world. The most fundamental part of
ontology is not abstracted from what we passively observe; rather, it emerges
and becomes established as an essential ingredient of our coherent activities.
Our inclinations to carry out certain types of activity begin to form the basic
ontological shape of the world we live in. It is as Bergson said: ‘The bodies we
perceive are, so to speak, cut out of the stuff of nature by our perception, and
the scissors follow, in some way, the marking of lines along which action might
be taken’ (Bergson [] , p. ; emphases original).

Freedom in Further Framing

The next stage of mind-framing comes when we consciously create or
develop specific concepts in order to use them in various concrete activities.
With the sort of activity-types I have been discussing so far, declining to
engage in them would really make one depart from commonly recognized
human forms of life. But there are also important conceptual–pragmatic
choices to be made at less fundamental levels, and in such cases the freedom
inherent in the process of mind-framing becomes much more visible.
Especially in the long-term development of science, we witness a great deal
of freedom being exercised in concept-creation and concept-development.
I believe that Kant made one major error, namely his commitment to
universalism (cf. Niiniluoto , p. ). Regrettably he fell into the trap of
regarding the trusted systems of knowledge of his age, including Euclidean


For a more careful exposition of the principle of single value, see Chang (, pp. –; ;
a).

3 9 084 8:
 How Mind-Framing Works 
geometry and Newtonian mechanics, as universally and necessarily valid.
Looking at Kant from a little historical distance now makes one thing clear:
his location in Newton-enraptured eighteenth-century Europe must have
exerted a strong hold on his imagination. Take Philipp Frank’s lament about
the poverty of philosophical opposition to new scientific ideas: what parade as
deep metaphysical truths are often simply ‘petrified’ remains of outdated
scientific theories (Frank , pp. –). Metaphysical principles can and
do change with the development of science. This is as acknowledged by a
string of neo-Kantians ranging from William Whewell in the mid-nineteenth
century to Michael Friedman in our time.
My chief inspiration here is Lewis, whose pragmatist notion of the a
priori was explained systematically in his now-forgotten masterpiece of ,
Mind and the World-Order. Lewis once reportedly declared: ‘I am a Kantian who
disagrees with every sentence of the Critique of Pure Reason’ (quoted by Beck
, p. ). The core of Lewis’s disagreement with Kant was his denial of the
existence of synthetic a priori judgements. Lewis stressed the great importance
of a priori elements in knowledge, but argued that they were always analytic:
‘The a priori is not a material truth, delimiting or delineating the content of
experience as such, but is definitive or analytic in its nature’ (Lewis ,
p. , emphasis original). For Lewis, all a priori principles follow from the
nature of the concepts that we choose to craft and use:

The paradigm of the a priori in general is the definition. It has always been clear
that the simplest and most obvious case of truth which can be known in
advance of experience is the explicative proposition and those consequences
of definition which can be derived by purely logical analysis. These are neces-
sarily true, true under all possible circumstances, because definition is legislative.
(ibid., pp. –)

Therefore, ‘the necessity of the a priori is its character as legislative act. It

represents a constraint imposed by the mind, not a constraint imposed upon
mind by something else’ (ibid., p. ). These thoughts form the core of Lewis’s
‘conceptual(istic) pragmatism’: there are a priori statements, which are true by
definition, inherent in ‘conceptual systems’; these systems are constructed by
us, and adopted on ‘instrumental or pragmatic’ grounds (ibid., p. x). We choose
the conceptual system freely, but once we have chosen a conceptual system,
within the system the a priori elements are analytically true. I see Lewis’s legacy
in Anjan Chakravartty’s () recent work on scientific ontology, which
combines realism concerning ontology with voluntarism in epistemology.
As a prime example illustrating his points, Lewis discussed Einstein’s
definition of ‘distant simultaneity’ in special relativity. Einstein defined the

3 9 084 8:
  Reality
simultaneity of two events happening at a distance from each other on the
basis of the principle of the constancy of the speed of light. So, if the two
events happen at locations A and B, and the observer is at the mid-point M
between A and B, then the two events are simultaneous if light signals released
at the time and place of each event’s occurrence reach M at the same time
(local simultaneity being taken as unproblematically meaningful and decid-
able). But what is the status of the assumption that the speed of light is always
the same in all directions? Einstein himself explains: ‘That light requires the
same time to traverse the path A ! M as for the path B ! M is in reality
neither a supposition nor a hypothesis about the physical nature of light, but a
stipulation which I can make of my own free-will in order to arrive at a
definition of simultaneity.’ Such stipulations are necessary in the specification
of the objects of inquiry, as Lewis explains here (, p. ):

we cannot even ask the questions which discovered law would answer until we
have first by a priori stipulation formulated definitive criteria. Such concepts are
not verbal definitions nor classifications merely; they are themselves laws which
prescribe a certain behavior to whatever is thus named. Such definitive laws are
a priori; only so can we enter upon the investigation by which further laws
are sought.

So Einstein framed simultaneity in a fundamentally different way from Newton,

and in a distinctive way that was almost completely unprecedented. It really
does not make sense to claim that Einsteinian simultaneity, or Newtonian
simultaneity, or any other variety, is inherent in nature. Rather, Einstein showed
us how to frame simultaneity, and time itself, and space, too, together with
light (and its constant speed) all in a tight package. Whether there were
realities that could be so framed, which obeyed the a priori rules that
Einstein laid down, was a contingent matter. The answer was to be found by
seeing if the concepts involved supported operationally coherent activities.
(Such pragmatic validation of mind-framing is the subject of Section ..)
Mind-framing is not a once-and-for-all legislation of concepts; this
becomes quite evident if we pay attention to the history of science. A striking
aspect of scientific inquiry, as with any progressive enterprise, is that it con-
tinually introduces new concepts used for the mind-framing of entities, and
updates and elaborates old concepts, too. The concepts change and evolve,
and often there are divergent paths of evolution. This gives rise to competing
systems of mind-framing (or what Goodman  famously called ‘worldmak-
ing’), and scientists make choices between them. As Lewis points out, ‘there


Quoted in Lewis (, p. ), from Einstein (, p. ); emphases original in Einstein.

3 9 084 8:
 How Mind-Framing Works 
will be no assurance that what is a priori will remain fixed and absolute
throughout the history of the [human] race or for the developing individual’.
While Friedman stresses the hidden continuity even in revolutionary change,
and detects continuity and progress at the level of constitutive principles
(Friedman , esp. p. ), Lewis sees the epistemic agent as possessed of
much greater freedom: ‘If the a priori is something made by the mind, mind
may also alter it’; ‘the determination of the a priori is in some sense like free
choice and deliberate action’ (Lewis , pp. –).

The Evolution of Scientific Concepts: Two Cases

We can gain more insights about the mind-framing of entities by examining
some further concrete cases from the history of science in detail. There are
numerous examples to choose from, but I will consider two that I have treated
in some detail in previous works (see Chang a for further examples). The
first example is temperature (Chang ). Temperature is more of a property
than an object, but when I speak of the mind-framing of entities I am taking
‘entity’ in a broad sense. Let’s start with the basic ontological principles
relating to temperature. In building a quantitative concept of temperature
out of the vague notion of hot and cold, scientists first of all presumed that
temperature was a real physical quantity, subject to the principle of single
value. The activity of testing-by-overdetermination (associated with the single-
value principle) manifested itself particularly as the practice of checking ther-
mometers for comparability: does a given thermometer always give the same
temperature reading when placed in the same situation, and do different
thermometers give the same value in the same situation? Similarly, the prin-
ciple of transitivity was also applied to the temperature concept, assuming that
objects could be put into a linear ordering by their temperature. This placed
temperature onto an ordinal scale of measurement. Such stories are very
common in the history of measurement: we propose the existence of a
quantified property in nature, and try to find ways of getting at that quantity.
In the conception of a measurable quantity, it is difficult to imagine the
absence of mind-framing in terms of the single-value and transitivity principles.
Going beyond the conception of temperature as a measurable quan-
tity in itself, an important step was to enrich and refine it in connection with


I will use the term ‘entity’ to designate any kind of thing we may conceive and discuss that may be
designated by a noun term. An entity is not necessarily a material thing; it can be an event or a
process. It may even be abstract, or social or institutional. (I could also take ‘object’ in a similar way,
but there are too many conflicting intuitions about ‘object’, so I try to avoid using that term
altogether.)

3 9 084 8:
  Reality
other concepts. A major step, taken in the late eighteenth century, was to
distinguish temperature from heat, and to clarify the relation between the two.
The concept of heat had separately been developed as a quantity on a ratio
scale (with a physically meaningful zero, and addition and multiplication
operations). For a long time heat was also associated with the notion of the
material substance called caloric, but this association fell away gradually.
Temperature came to be linked to heat chiefly by means of specific heat,
defined as the amount of heat required to raise the temperature of a sub-
stance by a unit amount. This link helped in establishing temperature as a
quantity measurable on an interval scale, on which differences between two
values are physically meaningful: it takes the same amount of heat to raise a
given body by the same temperature interval (e.g., – degrees, or –
degrees), if the specific heat of the body is constant. The next major step was
to link temperature and heat with mechanical concepts such as velocity,
kinetic energy and mechanical work, through the new theories of thermody-
namics and statistical mechanics. Temperature acquired new meanings, on the
one hand as ‘absolute temperature’ in thermodynamics, later linked with
entropy as well (see Chang , ch. ; Chang and Yi ), and on the other
hand as something proportional to the average kinetic energy of molecules.
Each of these developments settled down in the form of a definitional
proposition taken as a priori in its own context, and thereby making a deep
change in the framing of ‘temperature’.
The other example I want to discuss is the concept of ‘acid’ in
chemistry (see Chang a and references therein). As a substance term,
‘acid’ specifies a very different kind of entity from property-terms like temper-
ature. Here we start with framing in terms of the ontological principles that
typically pertain to substances, such as subsistence, causality and the identity
of indiscernibles. In the earliest part of the history, the significant move that
chemists made was to take ‘acid’ as a subsisting thing in its own right, rather
than treating acidity as a transient property exhibited by various substances.
But what kind of thing is it? How do we make the term ‘acid’ refer to some
recognizable set of stuff? The obvious step for the working chemists was to
choose a set of typical properties and behaviours as defining characteristics of
an acid: sour taste, corrosiveness (especially in relation to most metals), and the
ability to alter the colours of various indicators (litmus, juice of violets, turmeric,
etc.) in specific ways. Almost as important was the fact that acids and alkalis
neutralized each other so that they lost their typical properties. These steps
constituted the inevitable intensional route to reference-fixing, which
I discussed in Section ..

3 9 084 8:
 How Mind-Framing Works 
After this property-cluster definition of acid settled down well
enough, many attempts were made to identify the ‘essence’ of acidity, on
the assumption that there was one fundamental characteristic of acids that
was responsible for all their other properties. This was an important additional
layer of mind-framing that acids went through: it was not necessary to
conceive of acidity in such an essentialist way, but that is the choice that
chemists made – not to ask whether there was an essence to acidity, but to ask
what that essence was. Perhaps the best-known attempt was by Lavoisier, who
thought that oxygen was the ‘principle’ of acidity (and coined the term
‘oxygen’ from Greek roots to mean ‘acid-maker’), but there were various other
candidates proposed as well. All these early attempts failed, and success only
came in modern times, when first Svante Arrhenius, and then J. N. Brønsted
and T. M. Lowry developed the notion that it was the ability to give up a
hydrogen ion that defined an acid. And it was only with this advanced
theoretical concept that a convincing quantification of acidity began, via the
pH concept and the glass electrode enabling the construction of pH meters
(Ruthenberg and Chang ). But almost simultaneously the work of Gilbert
Newton Lewis introduced the notion of acid as the acceptor of an electron-
pair, which could be taken as a broader theoretical category encompassing the
Brønsted–Lowry acid concept, but not having a clear link with the pH measure.
Note the variable and unsettled course of development in the
framing of ‘acid’, already evident in the very brief sketch I have given here.
This is a good illustration of the shakiness of the correspondence-realist
notion that there is some well-defined thing out there which our concept
can simply point to or ‘latch on to’. Through the changes mentioned above,
not only the intension but the extension of the concept ‘acid’ changed
significantly. I would say that each notion of acid mentioned above did
frame some real entity; however, the various entities framed by the various
concepts are not at all one and the same thing. And it is difficult to say
which concept is the best one. Perhaps some of the earlier ideas can be
discarded safely enough in the context of modern chemistry, but it is very
difficult to choose between the Brønsted–Lowry concept and the Lewis
concept. The Lewis concept is the most sophisticated one theoretically, but
it is basically not measurable. Most of the experimental activities concern-
ing acidity are carried out on the basis of the Brønsted–Lowry concept, and
there is nothing theoretically deficient about that concept. In practice,
modern chemistry retains both of these concepts. There is no such thing
as mind-unframed ‘acid’, and we can and must choose which mind-framed
‘acid’ we want to engage with.

3 9 084 8:
  Reality

. The Achievement of Reality

We are free to make up concepts as we like, but they need to be validated
pragmatically by supporting coherent activities. Only then are we in
possession of mind-framed realities. Achieving reality requires a dynamic
working-together of experience, action and concepts in close interconnec-
tion and coordination. I continue with the cases of temperature and acid to
illustrate the process of validation. The character of reality achieved in this
way is quite strikingly different from the usual notions of reality. In my
conception, reality (real-ness) is a matter of degrees, and it is also domain-
specific. When we create a new concept that turns out to be able to
support coherent activities, we have a new reality. And when we make
further coherent activities relying on an existing reality, we make it more
real than before. Reality is an achievement made through well-designed
concepts and activities. But as with any other achievements in life, whether
our attempt to achieve reality succeeds or not is ultimately not in
our control.

The Pragmatic Validation of Concepts

Having considered how the mind frames realities, let us turn to the aspects of
realities that the mind has no control over. Experience is the ultimate source
and touchstone of our knowledge about mind-uncontrolled realities. We
could even say that ‘experience’ is the generic name that we give to our
encounters with mind-uncontrolled realities. It is through experience that we
have ‘contact’ (metaphorically and literally) with realities. These are tautologies
that express the empiricist outlook. Even though the objects of experience
clearly embody a priori elements, they are realities as long as they support
operationally coherent activities. So I return to my philosophical idiom that
realities are mind-framed but not mind-controlled.
The question now is how our freely chosen framing of realities can be
validated by experience, so that our proposed entities may be shown to be
realities (or not). Freedom and choice do not mean arbitrariness. A quick
example will illustrate the point. In his attempt to reform the foundations of
classical mechanics, Ernst Mach ([] ) eliminated the traditional
Newtonian concept of force, which he considered too metaphysical. But this
also removed Newton’s second law (F = ma) and, along with it, an obvious
way to define mass, as F/a: the strength of force applied to an object divided
by the acceleration resulting in its motion. To fill this gap, Mach offered a new
definition of mass: if two objects are allowed to interact with each other and

3 9 084 8:
 The Achievement of Reality 
they undergo accelerations as a result, the ratio of their masses is the inverse of
the ratio of the (magnitudes of ) accelerations induced. In a formula: m /m =
a/a. So far, it is just a matter of laying down a definition. But suppose we
have three objects in the system. Then we would have m/m = a/a and m/
m = a/a. Now we have a constraint on the acceleration values, if mass is to
obey the principle of single value (see Section .). To take the simplest
possible case, suppose that objects  and  induce the same magnitudes of
acceleration in each other, and objects  and  do so, too. Then, by Mach’s
definition, objects  and  have the same mass, as do objects  and . Then
objects  and  must have the same mass, too, which implies that they must
induce the same acceleration in each other. But will that be borne out by
experiment? Mach himself notes: ‘No logical necessity exists whatsoever, that
two masses that are equal to a third mass should also be equal to each other’
(ibid., p. ; emphasis original). What should we do if a and a come out
different when we do the experiment? We will not be able to do much
operationally coherent physics in such a situation, and we would be forced
to reject the Machian definition as unworkable. So we see that Mach’s defini-
tion actually contains a hypothesis about how accelerations will go in physical
situations. Definitions and Lewisian a priori principles are not mere tautologies;
whether they are apt or not depends on how things turn out empirically.

Coherence-Building
The pragmatic validation of concepts, when it works out, is an iterative process
of building operational coherence. In order to illustrate this point, I will con-
tinue with the two examples introduced in the previous section. In my
discussion of the temperature concept in the last section, I mentioned that
the presumption of single-valuedness demanded comparability in thermom-
eters. To us moderns it is difficult to imagine what a great challenge it was to
ensure comparability in thermometers, but the fact is that for nearly two and a
half centuries thermometers were made without convincing comparability,
until the monumental and painstaking work of Victor Regnault in the middle
of the nineteenth century. What Regnault’s work revealed was that only
thermometers filled with air (or one of a few other gases) exhibited a sufficient
degree of comparability. The much-loved mercury-in-glass thermometers
failed the test of comparability, and thermometers filled with any other liquids
(including alcohol) were even worse. In other words, temperature-
measurement was a coherent activity only when it used a few particular types
of thermometric fluid.

3 9 084 8:
  Reality
There was also a challenge in identifying the thermometer-based
concept of temperature as a quantified version of people’s sensations of hot
and cold. The indications of thermometers often went against qualitative
perceptions, and such disagreements had to be accommodated by finding
plausible explanations for them. For example, it feels so cold even though the
temperature isn’t so low, because of wind chill. I am having chills even though
the ambient temperature is high according to the thermometer, so I must be
getting a fever (a suspicion that I confirm with the thermometer). And so on.
With such challenges overcome, the thermometer-based temperature con-
cept was able to enhance the operational coherence of a whole host of
activities in many areas of life, including horticulture, brewing, clinical medicine
and experimental chemistry and physics, thanks to the greater precision and
reliability in the judgement of temperature enabled by the thermometer.
Thereby temperature as a quantity became a firm reality.
A different kind of developmental pattern and challenge can be seen
in the development of the temperature concept in relation to the concept of
heat. Specific heat initially arose as a vague material notion of ‘heat capacity’
(or capacity for holding caloric), with the density of caloric representing
temperature. The caloric-based thinking did not survive in the end, but the
more phenomenological concept of specific heat as the ratio between heat
input and temperature increase survived robustly, and supported a set of very
coherent calorimetric activities involving the operations of mixing various
substances at different initial temperatures and predicting and measuring
the temperature of the resulting mixtures. But a major apparent incoherence
loomed: in many situations the addition or subtraction of heat from a body did
not change the temperature at all – is the specific heat infinite in such cases?
Such a consequence was avoided by the invention of the concept of latent
heat: let’s postulate that any heat input that does not serve to raise the
temperature of the receiving body goes into a latent (non-sensible) form; this
made a lot of sense, since latent heat could also be understood as the cause of
observable changes in the state of the receiving body – most notably melting
and boiling. And the reverse changes of state (freezing and condensation)
were duly seen to release the latent heat back into sensible form, again
without a change in temperature. So, latent heat was seen to be real, while
the coherence of a whole range of activities in thermal physics was maintained
and enhanced.
With such an impressive and wide range of coherent activities relying
on it, the reality of temperature as a quantified property became very clear. By
now its reality is exhibited in almost every area of science, industry and
medicine, and it seems that every process in nature is affected by temperature.

3 9 084 8:
 The Achievement of Reality 
It should be stressed again that the concept of temperature has continually
grown and changed over the centuries, within a complex and growing
network of concepts. This also means that the reality designated by the
temperature concept is complex and changeable. An extreme instance of this
is the fact that the modern thermodynamic concept of temperature allows the
reality of negative absolute temperature, if there were a physical system whose
entropy decreases (becoming more orderly) when it absorbs heat. But nega-
tive absolute temperature is clearly not a possibility under the kinetic theory of
heat, according to which absolute temperature is proportional to the average
kinetic energy of molecules, which cannot take on a negative value.
If we look at the case of acid, some different issues emerge.
Concerning material substances there is a patchier track-record in scientists’
postulations settling down as realities, compared to measurable quantities.
Many presumed entities that enjoyed scientific popularity for a time have
come to be considered unreal: the four Galenic humours, caloric and phlogis-
ton, the aether, and many other items that feature in Laudan’s list supporting
the pessimistic induction. In the case of acid, too, it might have been ques-
tioned along various points in its history whether there was really such a thing.
In the property-cluster stage of the concept, it was not clear whether all the
properties were exhibited by all the acids. To put it in Richard Boyd’s termi-
nology (): is the property-cluster in question homeostatic? Do all the key
properties of an acid always go together? If not, a whole range of activities
from classification to prediction based on the concept would lack coherence.
For example, ‘carbonic acid’ (CO in modern terms, or rather, its combination
with water, HCO) has no sour taste but will turn litmus red. Given such gaps,
the reality of ‘acid’ was not convincing.
In the stage of development in which people looked for the essence
of acids, different kinds of activity came into focus. If there really is a ‘principle’
of acidity (such as oxygen), then it should have been possible to apply it to
other substances to turn them into acids. Sometimes this did work out, as
many products of combustion turned out to be acidic, at least when dissolved
in water; for example, carbon dioxide (CO) produced in the combustion of
organic substances formed carbonic acid when it met water. But not all
products of combination were acidic. Another thing that ought to work if
acids contain oxygen is to extract oxygen from known acids, but sometimes
this also turned out to be impossible, as in the famous case of ‘muriatic acid’
(hydrochloric acid, HCl). So the activities of acid-production by oxygen and
oxygen-extraction from acids were seen to lack full coherence. In the end
chemists abandoned not only Lavoisier’s oxygen theory of acids, but also the
general notion that there was an essential substance that conferred acidity on

3 9 084 8:
  Reality
other substances. This is a good illustration of how our concepts may not turn
out to designate realities. In contrast, the rendition of acidity as a disposition
worked out much better. According to the Brønsted–Lowry conception, an
acid is a substance capable of donating a hydrogen ion, and on that basis a
whole range of coherent activities can be performed, ranging from the
definition and measurement of pH to the understanding of neutralization
reactions as the meeting of hydrogen and hydroxide ions (H+ and OH!) to
form water. So we say confidently that Brønsted–Lowry acids are real.

The Character of Reality

Having tried to convey a concrete sense of how entities are framed and their
reality (real-ness) is achieved, let me now consider more carefully the general
character of reality achieved through the establishment of coherent activities.
The first thing to note is that reality, or even our knowledge of it, is not simply
related to observability or any other kind of possibility of immediate access.
There is no fundamental difference between the processes we use to deter-
mine the reality of observable and unobservable entities. To return briefly to
the case of temperature: scientists were able to establish quite convincingly
the reality of something as theoretical and removed from direct observation as
Kelvin’s absolute temperature, which is defined in terms of thermodynamic
theory in the setting of an ideal Carnot engine, which no one has ever been
able to make to any tolerable approximation. The same establishment of reality
can work out for more esoteric, theoretical and unobservable entities like
quarks and dark energy, too.
The next point to note is that being real is a matter of degrees, as
I have tried to make clear in the examples discussed above. At first glance this
will seem absurd, but I hope that the sense of absurdity will dissipate on
careful consideration. We might start by noting that in colloquial speech we
do easily attach degrees to real-ness: ‘the terrorist threat in London is still very
real, even though less media attention has been given to it lately’. An entity
should be considered real to a higher degree if it supports a larger number of
activities that are operationally coherent. In addition, operational coherence
itself is a matter of degrees, so the real-ness of an entity is higher if each activity
it supports has a higher degree of operational coherence. And it is real-ness
itself that is a matter of degrees, not just our knowledge of it. Still, ‘degrees of
reality’ will sound very strange to the ears of those schooled in contemporary
metaphysics (though it was a notion sometimes entertained traditionally): isn’t
real-ness the same thing as existence, and doesn’t existence have to be an

3 9 084 8:
 The Achievement of Reality 
all-or-nothing affair? Can something be just a little bit real, or partly exist?
I think we would do well to find ways of accepting such a notion.
Existence is seen as a black-and-white issue only because we are
accustomed to thinking about extremely clear-cut cases with clear meanings.
There is no reason to maintain that an ill-defined entity either exists, or not.
Consider: ‘Is there any moral turpitude in this man?’ Even in science, existence
may not be a black-and-white matter. Do light rays exist? They do, in the sense
that we can track straight, refracted or reflected paths of light. But they also do
not exist, in the sense that there is no material body in the shape of lines that
can be exhibited along the path of light. Familiar and concrete cases like the
Loch Ness Monster may not be a simple yes/no matter, either. What if there is a
creature in the lake that is a lot like what people have described, but not quite?
Or what if it is a visual effect that surely looks like what people have reported,
but with no material substance behind it? We should say that there is some
reality to the monster, since some activities involving the monster will be quite
coherent if there is a Nessie-like creature or a good visual effect mimicking it.
The reality of entities is not only a matter of degrees, but also
something pertaining to specific domains. ‘Domain’ here may be a spatio-
temporal region, but more generally I intend the term to refer to all kinds of
conditions that affect the coherence of an activity, pointing to a rather general
type of context-dependence. Real entities are only real in their proper
domains, not everywhere. Newtonian point-particles are real enough in situ-
ations where Newtonian activities are coherent (including ‘rocket science’,
solar-system celestial mechanics and pendulum motion, just to take a few
examples). Quantum wavefunctions defined by the Schrödinger equation are
real when it comes to electrons in atoms, and not so much when it comes to
protons and neutrons in the nuclei. Old-fashioned light rays are very real
where geometric optics works, but clearly not when we are doing the
double-slit experiment. Constellations are real within traditional positional
astronomy, but not in modern cosmology. Atomic weight as a fixed number
unique to each chemical element is real in the construction of the periodic
table, but not in nuclear physics.
But you might object: how can a given entity be real in one domain
and not in another? Suppose we refract a ray of light with a prism to direct it
on to a metallic surface, from which it will cause an electron to be ejected by
the photoelectric effect. It seems that my notion of reality will force me into
the absurd view that what goes through the prism is a light ray, which
somehow turns into a photon-bundle as it hits the metal surface. Shouldn’t
we rather learn to make a unified account in terms of what is actually real
everywhere (namely photons, not light rays)? But such a view is based on

3 9 084 8:
  Reality
complacency: you think you have the true picture of Reality, which contains
photons, not light rays. But when you say ‘photon’, what exactly do you have
in mind? You probably think it is a parcel of energy in the amount of hv, where
v is the frequency of light and h is Planck’s constant. But shouldn’t you be
thinking of what quantum field theory or even superstring theory says light is,
or rather, whatever the finished-and-accepted ‘theory of everything’ will say
light is? So, in principle, we can’t pronounce at all about photons until physics
is all finished. But let’s look at the reality of our practices. Whatever our ‘final
theory’ may say, we do already have many coherent practices involving light
rays, photons and also electromagnetic waves. But can these different entities
all be real? I will address this question of ontological pluralism further in the
next section.

. Ontological Pluralism

The message from the discussions given so far in this chapter is clear: if we
want to think seriously about the nature of realities, we should take heed of
what happens when scientists and other investigators try to craft concepts
that facilitate coherent activities. When we do, one significant thing we
learn is the pervasiveness of plurality: diverse types of ontology support
various sets of coherent activities, even within the realm of science, and
even within specific areas of science. This pluralist lesson goes against a
deeply ingrained metaphysical picture, in which the universe has one
correct inventory of things. There is no convincing justification for this
monist ontology, either from the track-record of science or from general
philosophical considerations. Pluralist ontology becomes easily acceptable
when we move away from the fallacy of pre-figuration, the notion that
well-formed real entities simply exist ‘in the world’ (see Chapter ). Real
entities are mind-framed, and there is no absurdity in allowing that many
different sets of real entities are operative in a given field of science.
I advocate ontological pluralism (to be discussed further in Section .):
it is beneficial to encourage multiple ontologies, each of which can facil-
itate coherent epistemic activities.

Ontological Pluralism and Its Traditional Sources

If we accept that operational coherence provides a good criterion of reality,
then it will be difficult to avoid accepting a diverse array of ontologies. If we
add the coherence theory of reality to epistemic pluralism, we are bound to
get a modest, practical sort of ontological pluralism. In my first statement of

3 9 084 8:
 Ontological Pluralism 
pluralism I explicitly limited myself to epistemic pluralism, on the ground that
experience did not teach us sufficiently well about metaphysics (Chang a,
ch. ). However, over the years I have gradually come to see the force of
Cartwright’s view that the successes of certain practices do give us credible
indications about ontology. Here I accommodate her insight in my own way,
by recognizing that what we mean by realities is the entities that facilitate
operationally coherent activities. From the diversity of successful local prac-
tices, Cartwright () takes a single picture of the world that is variegated
and inhomogeneous, and of Nature that works like our own artful practices of
modelling (Cartwright ). What I take is a plurality of ontologies (each of
which may be ‘dappled’, or not), in a way similar to Annemarie Mol’s ()
identification of the overlapping and interacting multiple ontologies emerging
in medical practice, and Chakravartty’s (, p. ) ‘pluralism about packag-
ing’ and ‘pluralism about behavior’.
My conception of reality is stringent, but it is also permissive – or
‘promiscuous’, in Dupré’s provocative phrasing. The criterion of operational
coherence does rule out many things as candidates for reality, but also rules in
a whole variety of other things. In the absence of what else we might
operationally mean by ‘real’, and with the recognition that a concept of reality
is not something we should try to do without, I propose that we muster the
metaphysical courage to admit that there may be different systems of real
entities, even within what we might be compelled to regard as one and the
same domain. My thinking is in the tradition of naturalism, treating ontology as
a field of study best approached through empirical inquiry. Now, if any entity
that plays an indispensable role in coherent activities is taken to be real, then it
opens up the possibility that all sorts of entities may be real, all at the same
time. This is a contingent matter, but what we have seen in the history of
scientific endeavours is that very different ontologies have indeed supported
various successful scientific practices. This issue becomes especially acute if we
dispense with the presumption of ontological reductionism, as I will propose
in Section ..
In embracing and developing ontological pluralism, I want to start by
noting that it is a position that has been advocated by some eminent and
sober modern philosophers (see Stump  for a helpful survey). It is very
well known that Rudolf Carnap argued that we have a choice of conceptual
frameworks, each with its own fundamental ontology:


I thank Helene Scott-Fordsmand and Brooke Holmes for introducing me to Mol’s work.

3 9 084 8:
  Reality
Let us grant to those who work in any special field of investigation the freedom to
use any form of expression which seems useful to them; the work in the field will
sooner or later lead to the elimination of those forms which have no useful function.
Let us be cautious in making assertions and critical in examining them, but
tolerant in permitting linguistic forms. (Carnap , p. , emphasis original)

Quine’s doctrine of ‘ontological relativity’, briefly mentioned already in Section

., is bound to have pluralist implications. Putnam’s ‘permutation argument’,
fully anticipated by Quine, can also easily be read in a pluralist vein, and Putnam
was explicit in his lasting commitment to ontological plurality even after he
abandoned ‘internal realism’. In response to Maudlin (), whose argument
for metaphysical realism was discussed in Section ., Putnam (b, p. )
admits that he is a metaphysical realist, but proposes a ‘sophisticated meta-
physical realism’ based on an acceptance of ontological plurality: ‘the same state
of affairs can sometimes admit of descriptions that have, taken at face value,
incompatible “ontologies,” in the familiar Quinean sense’ (ibid., p. ).
Among contemporary authors Niiniluoto (, p. ) is notable for
proposing a ‘principle of conceptual pluralism‘, which admits that ‘all inquiry is
relative to some conceptual framework’. With this principle, and an admission
of Peircian fallibilism, he arrives at his ‘critical realism’, which modifies meta-
physical realism so as to make it more operational in practice. Kitcher (,
pp. xxii–xxiii) proposes a modest sort of ontological pluralism as the first key
features of ‘the pragmatist reform of epistemology and metaphysics’: ‘what-
ever is independent of us might be conceptualized in many different ways’.
That sounds innocuous enough for the metaphysical realists, but we should
read on: ‘In one sense there is only one world – the yet-to-be-differentiated
source of our experience; in another, there are many worlds – the diverse
articulated totalities of objects assorted into kinds that reflect all the ways in
which divisions might be drawn.’

Equal-Opportunity Experimental Realism

So there have been plenty of in-principle discussions about alternate ontol-
ogies. I would like to focus more on how multiple ontologies arise in practice.
In looking to practices for ontology I have already drawn inspiration from
Hacking’s experimental realism, and here I will do so again. There is one
obvious objection to Hacking’s position, which will turn out to be a blessing


Niiniluoto’s view is that Kant was driven to scepticism about knowledge of things-in-themselves
because he didn’t recognize pluralism, implying that pluralism allows knowledge of things-in-
themselves. That is where I part company with Niiniluoto.

3 9 084 8:
 Ontological Pluralism 
in disguise for pluralism. The objection is this: might we not misunderstand our
experiments, mistakenly presuming that some non-existent entity is involved
in them? This actually seems to happen with some regularity. The history of
science is full of very successful practical interventions by experimenters who
thought they were using entities that we now regard as unreal. The pessimistic
induction from the history of science would seem to be just as deadly to
Hacking’s experimental realism as it is to standard scientific realism.
Let me illustrate the problem with some concrete examples. William
Herschel discovered infrared radiation in  by realizing that a thermometer
inserted into the dark space beyond the red end of the solar spectrum
detected a good deal of heating effect (see Hentschel , pp. –;
Chang and Leonelli ). Herschel thought that he had successfully used a
prism to separate out the rays of light and the rays of caloric, both contained in
the sunbeam, directing the caloric rays onto the thermometer. If this doesn’t
qualify as Hacking-style ‘spraying’, I don’t know what does. Now, doesn’t such a
case amount to a refutation of Hacking’s experimental realism, and also my
coherence theory of reality, since we now know that caloric isn’t real?
Meanwhile, among Herschel’s contemporaries who opposed the caloric theory
we find Count Rumford, often celebrated as the neglected pioneer of the
kinetic theory of heat. Rumford held that heat consisted in the vibration of
molecules, and even showed by experiment how much heat could be gener-
ated by friction. But most who praise Rumford’s prescience do not realize that
he also postulated the existence of ‘frigorific’ radiation, namely low-frequency
waves emitted by cold objects, which have the effect of cooling down warmer
objects at a distance. Rumford devised successful experiments to reflect and
focus frigorific rays using metallic mirrors and cones (see Chang ).
Similarly, consider the infamous case of phlogiston (Chang a,
ch. , esp. pp. –). The phlogiston theory made sense of a key aspect of age-
old smelting techniques, in which calx (metal oxide, in modern terms) was
transformed into metal by taking up phlogiston from a combustible (i.e.,
phlogiston-rich) substance such as charcoal. Even Kant greatly admired
Georg Ernst Stahl’s laboratory operations transforming one substance to
another, and back to the original one, by giving phlogiston to it and taking
it back out (Kant [] , –). Similarly, Joseph Priestley claimed to be
able to manipulate phlogiston successfully. Matthew Boulton (James Watt’s
business partner) wrote excitedly to Josiah Wedgwood (the famous porcelain-
maker) in :

We have long talked of phlogiston without knowing what we talked about, but
now Dr Priestley hath brought ye matter to light. We can pour that Element out

3 9 084 8:
  Reality
of one Vessell into another, can tell how much of it by accurate measure is
necessary to reduce a Calx to a Metal . . . (Boulton quoted in Musgrave ,
p. )

In the s Priestley made oxygen through his attempt to ‘de-phlogisticate’

air by reducing a calx (rust) back into metallic form in an enclosed space; the
air in that space would give up its phlogiston to the calx, restoring its metallic
nature. This seemed to work out, and he obtained ‘de-phlogisticated air’. This
new gas supported combustion exceptionally well because it was eager to
regain phlogiston, plentiful in combustible substances. Even less ambiguously,
Priestley predicted that a calx could be reduced by heating in inflammable air
(later called hydrogen), which he conceived as pure phlogiston. This experi-
ment succeeded brilliantly. Did Priestley’s successes, including even novel
predictions, mean that he and his contemporaries should have granted reality
to phlogiston?
Such cases are not just relics from bygone eras. Take, for example,
orbitals in modern chemistry. Not only a great deal of theoretical explanation
but numerous experimental interventions in modern chemistry rely on the
general concept of orbitals, and also on a detailed knowledge of the number
and shapes of various types of atomic and molecular orbitals. In addition, the
idea of sequential filling of atomic orbitals explains very nicely, to a limited yet
very significant extent, why the periodic table has the shape it has (see Scerri
). But orbitals inhabited by individual electrons have no reality if we take
quantum mechanics literally, since all electrons are identical and they cannot
be said to occupy different orbitals within a given atom or molecule (see
Ogilvie ). Yet, I think it makes sense to attribute reality to orbitals on the
basis of the successful chemical practices employing them, and that is certainly
how very many chemists think.
All of these examples may appear to show that coherent experimen-
tal activities provide no guarantee of the reality of the entities that the
experimenters themselves presume to be manipulating. You may be able to
spray something without knowing much at all about what it is that you’re
spraying. So, are we back to square one, with Hacking’s attempt to save realism
all in vain? I think the way forward is to admit that caloric, phlogiston and such
entities are real in their proper domains. That is much more defensible than the
selective fault-finding mission, targeting activities involving entities that we
currently think are non-existent. Not only is this strategy unprincipled, but it is a
hostage to fortune as the scientific consensus shifts around, concerning which
cutting-edge theory is the correct one. This is another instance in which the
faith in science that I have critiqued in Section . can lead to unwarranted

3 9 084 8:
 Ontological Pluralism 
complacency. Without presuming that our current most popular theories tell
us the truth about ultimate reality, how do we know that caloric and phlogis-
ton and such things are not real?
What I am proposing is a kind of liberal equal-opportunity realism. Let
us grant reality to all entities that support coherent activities, where they do
and to the extent that they do. It is useful to recall another part of Hacking’s
argument here. Facing those who would doubt that success in practical
intervention can form a secure enough basis for our knowledge of unobser-
vable reality, Hacking hits the ball right back to their court, by asking: why do
you think anything is real? In an argument already alluded to in Section .,
Hacking points out that even the ‘medium-sized dry goods’ are only consid-
ered real because of our ability to handle them. This includes the visual–
muscular coordination that is such an essential part of our normal acts of
seeing. So why not admit that phlogiston within its domain of coherent use
is (nearly) as real as tables-and-chairs are in our daily lives? And here is a
thought for the standard scientific realists who put their trust in the argument
from success: we should be open-minded and generous to all investigators, by
granting reality, provisionally and defeasibly, to the referents of whichever
theoretical conceptions seem to lead to success. This is what we ought to
do if we really take success as our only reliable guide in deciding what to be
realist about.

Living with Plurality

Still, I don’t imagine that standard scientific realists will immediately be happy
with the liberal line that I have just taken. They will point out, quite rightly, that
taking both caloric and phlogiston as real implies accepting two very different
stories in the same domain, for example the science of combustion. So, will
ontological pluralism end in contractions? And would that not destroy the
overall cogency and authority of scientific knowledge? I will argue that this fear
is exaggerated and that the pluralistic situation actually offers tangible bene-
fits. Ontological plurality is a matter of contingent fact about our epistemic life:
it just turns out that we can often engage in coherent activities by employing


The same problem emerges if we try to take the pessimistic induction literally as an induction. If we
know that the terms in Laudan’s infamous list do not refer, that means we do know something about
the shape of reality, and then the realists have won. The pessimistic ‘induction’ is only valid if it is
taken as a reductio, as Psillos (, p. ) points out.

The classic experiments by Richard Held () seemed to show that normal vision failed to
develop when kittens were deprived of muscular activity moving themselves around. See Bermejo,
Hüg and Di Paolo () for a retrospective.

3 9 084 8:
  Reality
many different kinds of presumed entities in various situations. Even within a
given domain, it often happens that there are multiple coherence-conducive
ontologies that cannot easily be conceived in terms of each other – such as
wave and particle, phlogiston and oxygen, caloric fluid and molecular kinetic
energy, or electrons neatly pigeonholed into orbitals and a ‘gas’ of mutually
indistinguishable electrons. If we accept my notion of reality, we should
be open to the reality of sets of entities that are apparently mutually incom-
patible. In fact that would also be the case even for standard scientific realists
who stay true to the spirit of the realist argument from the success of science,
when equal success is achieved on the basis of each of the competing
ontologies.
Above I mentioned some cases of plurality drawn from the history of
the physical sciences in recent centuries. I predict that a more thorough and
extensive look across various sciences and many other areas of life will reveal
the co-existence of very different ontologies in most domains of activity and
thought. For example, our legal thinking is mostly done in terms of individual
persons and their actions, yet at the same time corporations are treated legally
as persons. Most of us think about human actions in terms of free will and
moral responsibility, while at the same time agreeing that the mind is only a
manifestation of the molecular and electrical activities in the brain. As Arthur
Eddington (, pp. –) famously put it, ‘there are duplicates of every
object’. Sitting down to write his words, he was at his ‘two tables’: first, the
familiar substantial thing, and then the ‘scientific table’, which is ‘mostly
emptiness’ with ‘numerous electric charges rushing about with great speed’.
Would you, on reflection, deny reality to either of Eddington’s two tables?
Coming back to physics, currently we have the ontologies of curved space-
time and quantum fields, sitting superimposed on each other as it were, with
dark matter and dark energy tucked away somewhere in the picture.
It is instructive in this connection to hear C. I. Lewis again, this time on
the progress of knowledge and conceptual change:

New ranges of experience such as those due to the invention of the telescope
and microscope have actually led to alteration of our categories in historic time.
The same thing may happen through more penetrating or adequate analysis of
old types of experience – witness Virchow’s redefinition of disease. What was
previously regarded as real – e.g., disease entities – may come to be looked upon
as unreal, and what was previously taken to be unreal – e.g., curved space – may
be admitted to reality. But when this happens the truth remains unaltered and
new truth and old truth do not contradict. Categories and concepts do not literally
change; they are simply given up and replaced by new ones. (Lewis , p. ;
emphasis original)

3 9 084 8:
 Putting Things Together 
Lewis does not draw an explicitly pluralist conclusion in this passage, but he
takes two crucial steps towards it. First, he says that an old truth ‘remains
unaltered’, while he expects conceptual change to continue, and along with it
ontological change. Lewis considers it natural and right that we attribute
reality to the entities that play significant roles in the conceptual schemes
through which we live and learn at each stage of development. There is no
final point or destination of development, which is to say that nothing we
regard as real should be regarded as absolutely and exclusively and eternally
real. That is also to say, to the extent that something is real now, it should not
cease to be real just because it becomes necessary for us to deal with new
experiences, for which we need a different conceptual scheme.
Second, by seeing that ‘new truth’ and ‘old truth’ do not contradict
each other, he allows some logical breathing space for pluralism. It is not
necessary, and in fact not often, that different ontologies directly contradict
each other. Take the case of oxygen and phlogiston again. It is often thought
that Lavoisier’s oxygen-based chemistry proved that phlogiston did not exist,
but that is too hasty. No plausible definition of ‘phlogiston’ and ‘oxygen’ can
give us a logical deduction that ‘phlogiston is real’ implies ‘oxygen is not real’,
or vice versa. So there is no direct logical contradiction in affirming that both
entities are real, and in fact there were chemists who, in the midst of the
Chemical Revolution, made coherent hybrid systems of chemistry which
affirmed the reality of both, using oxygen in the tracking of weights and
phlogiston in the explanation of (what we now recognize as) energy relations
(see Chang a, p. , and references therein). The phlogistonist and
oxygenist systems of chemistry did contain some mutually contradictory
statements (such as ‘Water is an element’ in one and ‘Water is a compound’
in the other). However, if we sufficiently dissect the meanings of ‘element’ and
‘compound’ in those sentences, we find that semantic incommensurability
prevents any direct contradiction (see Chang a, pp. –; cf. Goodman
, p. ). I will comment on this issue further in Section ..

. Putting Things Together

Similarly as correspondence realism can stand in the way of an active view
of knowledge, a certain kind of reductionism can stand in the way of the
kind of pragmatist and pluralist metaphysics advocated in this chapter.
This species of ontological reductionism, which I will call Legoism,
considers any objects to be put together by a simple assembly of unchange-
able units in the manner of building things with Lego bricks. Legoism is
closely linked to the mereological and set-theoretic habits of philosophical

3 9 084 8:
  Reality
thinking discussed in Section .. Legoist metaphysics provides a hostile
climate for pragmatist metaphysics because the assumption of unalterable
basic building-blocks of matter encourages the fallacy of pre-figuration.
And even though in principle there could be multiple sets of immutable
basic building-blocks, that becomes difficult to maintain if the building-
blocks are imagined to be not mind-framed. Legoism is often assumed to
be supported by science; however, a careful look at modern physics shows
that physical combination is not Lego-like assembly. It is also not the case
that Legoist intuitions originate from our everyday experiences. Rather,
they are instilled by the mental habits of making Legoist analysis.

Against Legoism
Before I leave behind the explicit discussion of ontology, I must address a
certain well-entrenched metaphysical doctrine that stands powerfully in the
way of the acceptance of pragmatist metaphysics. In a light-hearted termino-
logical move I am going to call it ‘Legoism’, because it pictures everything in
the world as composed of unchangeable units, like Lego bricks. According to
Legoism, everything is made up by a simple assembly of unchanging parts,
and can be decomposed cleanly into those parts. This view seems to be very
much a part of our philosophical and scientific common sense.
How exactly is it that Legoism stands in the way of pragmatist
metaphysics? The logical connection is not tight, but there is a strong intuitive
push. Legoist metaphysics provides a hostile climate for pragmatist metaphys-
ics, even though it does not directly contradict it. This is because Legoism
often serves as an excellent vehicle for the fallacy of pre-figuration. Legoists
usually assume that the unchanging fundamental units of their analysis are
mind-unframed parts of Reality, even though this is strictly speaking not
required by Legoism itself. In principle it would be possible to combine
Legoist thinking about composition with a pluralist allowance of multiple sets
of fundamental units, but this possibility is usually not entertained. Any other
valid ontology at a higher or more complex level of existence is assumed to be
reducible to the fundamental building-blocks.
Thus, the usual version of Legoism holds that everything can be
ultimately broken down to the one-and-only set of fundamental building-
blocks that cleanly make up everything else. It would seem that everyone
accepts something like the hierarchy of ontological composition as given by
Paul Oppenheim and Hilary Putnam in their classic paper on reductionism
(, p. ): () elementary particles; () atoms; () molecules; () cells; ()
(multicellular) living things; () social groups. Dupré has given a trenchant

3 9 084 8:
 Putting Things Together 
critique of ‘microreductionism’, which he defines as ‘the view that the ultimate
scientific understanding of a range of phenomena is to be gained exclusively
from looking at the constituents of those phenomena and their properties’
(Dupré , p. ). Initially I followed a key line of argument from Dupré: even
if we accept ontological microreduction, epistemic microreduction does not
follow. Now I think that it is also important to subject ontological microreduc-
tionism to full critical scrutiny. We should not concede so readily that compo-
sition is Legoist, that a material whole is just a juxtaposition of its parts. Once it
is agreed that there are elementary building-blocks of nature ‘out there’
independently of all conceptualization, it is easy to argue that all objects are
obviously mereological sums of those building-blocks.

Modern Physics vs. Legoist Metaphysics

Legoism is allegedly backed up by modern physical science, but it is not.
James Ladyman, Don Ross and David Spurrett make this point very strongly as
part of their argument against a priori metaphysics shaped by outdated
scientific common sense (Ladyman and Ross , ch. ). I endorse this aspect
of their critique, and would add that showing the limitations of Legoism does
not even require cutting-edge contemporary physics. Attention to the actual
practices of physical composition and decomposition over the long history of
the physical sciences will show that parts are not all there is to a whole. ‘Parts’
are only the salvaged remnants of a whole that has been shattered, physically
or conceptually. This is easy to see if we think about social situations: if we
destroy society completely, will the individuals left be the same beings that
you recognize in a functioning society? I want to argue that the situation is
actually very similar with physical bodies. If we pay attention to successful
analytic and synthetic practices in chemistry and physics, we will see that they
do not go in Legoist ways.
Legoism is closely related to what I have called compositionism in
chemistry, defined as ‘the notion that chemical substances are made up of
stable units that persist through chemical reactions’ (Chang e, p. ;
also Chang b; a, ch. ). Legoism is generalized compositionism,
extended beyond chemistry. I could have called it ‘atomism’, but I have
avoided that term because what we call ‘atoms’ in modern science are
breakable and changeable – not very ‘atomistic’! Legoist thinking gained
ascendancy in chemistry and physics during the eighteenth and nineteenth
centuries. It had to displace well-entrenched metaphysical alternatives, espe-
cially neo-Aristotelian hylomorphism, which conceived specific materials as the
outcome of the imposition of form on substance. By the early modern period

3 9 084 8:
  Reality
scientists began trying to decompose things into their constituent parts, but
there were worries that the alleged processes of decomposition might be
altering the substances being analysed. For example, the application of strong
heat was commonly thought to break things down, but the cogency of ‘fire-
analysis’ was questioned by the likes of Robert Boyle (see Debus ): how
could one be sure that the application of fire was merely breaking things up
into their constituents, rather than altering their very nature, or at least getting
fire-particles sticking to them? Similar doubts were also raised concerning
other analytical methods, such as the dissolution of substances by the appli-
cation of acids.
That was all in the prehistory of respectable science, you might say.
Let us, then, come into the twentieth century, and take a look at the ‘atom-
smashing’ practices of modern experimental physics. Atom-smashing has
never been Lego-like disassembly: when atomic nuclei are broken up, energy
is almost always added or subtracted; given the interconversion of mass and
energy, this means that the amount of matter is not preserved in decompo-
sition. It won’t do to suggest that the Lego-like picture is approximately true: no
theory that has to dismiss nuclear bombs as unimportant details should be
regarded as ‘approximately true’. As empirical evidence for mass–energy
equivalence (E = mc), what is invoked most often are particle-collision exper-
iments in which the masses of the ingredients do not add up exactly to the
sum of the masses of the products (see Fernflores , esp. sec. ). Especially
famous is the  experiment of John Cockcroft and Ernest Walton, who
bombarded a lithium nucleus with a proton and obtained two helium nuclei
(α-particles). Their measurements showed that the sum of the masses of the
reactants was . + . = . amu, but the masses of the
products only added up to . amu, indicating that . amu had
‘disappeared’, turning into other forms of energy. Thus Lavoisier’s principle of
the conservation of mass was overturned in physical and chemical practice,
after over a century of dominance. An inspection of the periodic table of
elements easily shows that the masses of atoms are slightly different from the
sum of the pre-combination masses of protons and neutrons (and electrons)
that constitute them. These have been regarded as indisputable facts for many
decades now by physicists and chemists, but the basic metaphysical implica-
tions of such facts have not got through to the sensibilities of philosophical
reductionists. If we still regard mass as the primary indicator of the amount of
matter, then it is clear that the amount of matter is not preserved in
elementary-particle collisions or nuclear reactions. That does not present a


 amu (atomic mass unit) is / of the mass of carbon- in its ground state.

3 9 084 8:
 Putting Things Together 
problem for the conservation of energy, of course, but it does destroy the naïve
notion that atoms are simply put together from elementary particles with fixed
masses. Atomic nuclei are not mereological sums of protons and neutrons;
they are not made up of protons and neutrons in a straightforward Legoist
sense.
Experiments in high-energy physics do not support the naïve view of
elementary particles as unchangeable building-blocks of matter. When two
protons collide with each other in a particle accelerator, a whole host of other
particles are created: should we say that a proton (or two protons together,
somehow) already contained these particles? And should the phenomena of
pair-creation and pair-annihilation lead us to conclude that a pair of photons
consists of an electron and a positron, or vice versa? And when a photon is
absorbed by an atom, it ceases its existence but raises the energy level of the
atom; so the photon is not an unchangeable unit, not even a persisting one.
These are merely a handful of illustrative examples. Generally speaking, in the
physics of so-called ‘elementary’ particles, smashed-up pieces do not neces-
sarily pre-exist in the whole. This recognition led Geoffrey Chew to advance his
‘bootstrapping’ view of elementary particles, according to which elementary
particles are made up of one another. This view was sidelined with the
advent of quarks and the Standard Model, but it may be worth revisiting, after
all. That is also to say, the early modern doubts about physical mereology have
returned with a vengeance. Again, I am only invoking very basic experimental
facts here. I am not even entering into the difficult ontological questions raised
by quantum superposition and entanglement, or the indistinguishability of
identical particles, or virtual particles and vacuum fluctuations, or quark con-
finement, all of which are bound to complicate the picture much further and
in all likelihood in anti-Legoist directions.

How Philosophy Should Resist Legoism

At this point some philosophers may say: ‘But what these scientific experi-
ments apparently seem to show can’t really be the true metaphysical picture.
There must be unchangeable basic building blocks, and everything must


See a friendly presentation of the basic facts and ideas in the ‘International Physics Masterclasses’
section on proton collisions (https://atlas.physicsmasterclasses.org/en/zpath_protoncollisions.htm).

Positron–electron pair-annihilation is now even in the realm of familiar technology as the basis of
PET (positron emission tomography) scans, which work by the injection of positron-producing
radioactive atoms into the body; it is not something we can afford to ignore in our thinking about
the world. See Shang () for an illuminating historical and philosophical account of PET.

For more on the metaphysics of bootstrapping and the S-matrix theory, see McKenzie ().

3 9 084 8:
  Reality
ultimately be made of them.’ Of course, no amount of scientific knowledge
can prove that the metaphysical reality is not like Lego. I just think that Legoism
is not a universally productive way of thinking about physical reality. If we look
at the scientific situation closely we will see that Legoism has been most
successful not at the ultimate level of micro-reality, but at an interesting
middle-level that is the realm of molecules and ions (see Chang d). This
certainly does not inspire metaphysical Legoism about the ultimate constitu-
tion of matter.
But where do people get the intuition that reality must be Lego-like?
If Legoism is not supported by successful practices in physical science, then
where do the widespread intuitions in its favour come from? It could be that
the compositionist intuitions are rooted in our everyday life. Robert Northcott,
in a serious joke, observes that we must have Legoist intuitions because we all
grew up playing with Lego. Since Lego itself is a twentieth-century invention it
can’t have been responsible for the advent of compositionism in premodern
science, but could it be that much of our everyday life is like playing with
Lego? We can smash a plate and glue it back together, build a house out of
bricks, and take apart a watch and put it back together. But these quotidian
practices of composition actually do not work in Lego-like ways. ‘Medium-
sized dry goods’ generally do not stick to each other. You can’t build a brick
wall without mortar, or glue things together without glue. How glue works is
not at all like Lego (and actually quite mysterious!), so our experience of gluing
(or stapling, or clamping, or strapping) things together does not explain why
we have Legoist intuitions. In biology, too, a multicellular organism is not just a
bunch of cells put next to each other; the intercellular matrix helps cells hold
together. And a social community cannot be built without individuals
undergoing changes through their mutual associations, changes that enable
the associations in the first place. Lego, first marketed as ‘automatic binding
blocks’, was such a commercial and cultural success precisely because it was a
very clever and unusual arrangement (an ingenious combination of rigidity,
elasticity and friction), in which bricks do stick together without the help of
anything else! Almost nothing else in nature or human life behaves like
Lego, and that is the secret of its success. We do not live in Legoland.
I suggest that that our common Legoist intuitions come not from
practical experience, but from a quasi-Kantian conceptual necessity of the kind


I thank Matt Meizlish for teaching me about the intercellular matrix.

For the history of Lego as told by the company itself, see www.lego.com/en-us/aboutus/lego-group/
the-lego-group-history (last accessed  September ).

Gretchen Siglar points out that Velcro is one rare example.

3 9 084 8:
 Putting Things Together 
that I discussed above in Section .. Legoist intuitions are ontological
principles that are necessitated when we choose to carry out Legoist analysis,
namely the activity of understanding an object as a mereological sum of its
parts, whose identity or essential nature is not affected by any combinations
into which they enter. So, if we do choose to carry out a Legoist analysis,
then of course it makes no sense not to adopt Legoist ontology; that would
render our activity incoherent and unintelligible. And if we do routinely carry
out Legoist analysis, we may understandably form Legoist habits of mind. We
often do engage in such analyses in certain activities of everyday life. Simple-
minded accounting is a very good example; perhaps most fundamentally, the
standard kind of arithmetic that we learn in childhood is firmly founded on
Legoist intuitions. We also engage in Legoist analysis in theoretical science, for
example when we apply various conservation laws. However, this is not to say
that we should always engage in Legoist analysis. It is futile to do so where we
cannot find reliably persistent units out of which the objects of our interest can
be said to be made. Whether or not there are such units is an empirical
question, a contingent matter, which can only be clarified in a pragmatic
way, by devising and attempting to carry out actual operations of physical
assembly and disassembly.
The main source of widespread intuitions in favour of ontological
reductionism is not successful scientific practice, but our predilection for
Legoist analysis. It is unwise to let ourselves be guided by these intuitions in
situations in which attempts at Legoist analysis deliver little empirical success.
Mereology and set theory may apply nicely to some aspects of reality, but that
is a contingent matter. It should not simply be presumed that these schemes
apply to realities. My pragmatist view is that any conceptual scheme should
prove its worth by producing beneficial results of some kind. Judging from the
findings of modern chemistry and physics, it seems that the mereological
part–whole relation is an inappropriate framework for understanding actual
physical combination. Physical fusion and disintegration may violate axioms
that one considers reasonable or even indispensable in mereology, such as
transitivity. If so, the most reasonable conclusion may be that physical com-
position is not a matter of part–whole relation in standard mereology. And if
Legoism doesn’t work in chemistry and physics, then it is not likely to work in
other sciences. This destroys the foundations of the grand microreductionist


Reductionist scientists and philosophers often add two further ideas, which are actually not
necessary for Legoist analysis per se: () that there are fundamental parts that cannot be further
decomposed, and () that there are only a small number of types of fundamental parts or units.

3 9 084 8:
  Reality
strategy. We need to make sure that quasi-Kantian conditional necessity does
not degenerate into pseudo-Kantian metaphysical prejudice.
These thoughts about Legoism also offer an insight into how to do
naturalistic metaphysics. Naturalism should not mean just following any meta-
physical consensus that scientists reach among themselves, or subscribing to
any metaphysical pictures implied by the best scientific theories of the day.
The historical development of science shows plenty of instability and contin-
gency. If we follow the ‘verdict of science’ blindly, we risk mistaking scientists’
presuppositions as warranted conclusions of inquiry. Rather, naturalism should
oblige us to adopt hard-earned insights coming from well-established scien-
tific practices, but only after a thorough philosophical analysis. Attention to the
actual practices of chemistry and physics reveals that there has never been
unequivocal scientific warrant for Legoism. Nineteenth-century structural
chemistry seems to have been a bright blip in the history of Legoism in
science, not the moment of its firm establishment for all future science.

3 9 084 8:
  

Truth

. Overview

Why Worry about the Concept of Truth?

If you are not a professional philosopher, or even if you are one, why
should you concern yourself with the theory of truth? To practical people
it may seem that they know very well how to tell what is true and what is
not true; there is no need for philosophical disquisition. But truth is not
such a straightforward matter, and any complacency you might have felt
about it should have disappeared in this dystopian start to the twenty-first
century, with many societies utterly unable to reach consensus on some
very basic matters of truth. The liberal faith that we can all agree on the
facts and have a polite discussion over opinions and values is being
seriously challenged by people not accepting each other’s facts.
For a telling sign, witness how the New York Times marketed itself in
the crucial election year of  in the United States: ‘Life needs truth.
The truth is essential. The New York Times – Subscribe now.’ But how do
we evaluate the claim that this newspaper will tell you the truth? Easy
enough to trust that the New York Times does not deliberately print lies,
but honesty is not enough, since misinformation is so often conveyed very
earnestly. And it is not good enough to say that it employs stringent
procedures that are designed to filter out misinformation. In order to
know if such procedures work, we already need to know that the specific
pieces of information they filter out are falsities and what they allow in are
truths – which brings us back to square one. We are hit with the same sort
of difficulty when we try to say everyone should trust science to tell us
the truth.


This statement occurred at the end of a video advertisement to potential subscribers: www.nytimes
.com/subscription/truth/truth-is-essential (last accessed  October ).



8: /73 791
  Truth
So there is urgent need to think more carefully about what truth is, and
how we can know it. Unfortunately, some of the most thoughtful cutting-
edge academic work in the humanities originating in the late twentieth
century has had the adverse effect of casting destructive doubt on the
traditional ideas of truth, fact, objectivity and rationality without offering
anything convincing to take their place. There have been attempts to enter
into thoughtful discussions of the ‘post-truth’ phenomenon (e.g.,
Sismondo ; McIntyre ), but philosophical thinking on this issue
tends to be hampered by an attachment to the non-operational idea of
truth as correspondence to the ultimately inaccessible reality (see
Chapter ). We need to rethink the notion of truth itself carefully, in
order to make an effective defence of truth without making indefensible
claims or putting up ideals with no bearings on actual practices.
With that task in mind, it is particularly disappointing to see that
among the most rigorous philosophical thinkers a notable current ten-
dency is deflationism about truth, which tries to avoid vexing questions
about the nature of truth by turning it into a maximally empty concept.
Deflationists take the function of the word ‘true’ as simple affirmation.
Daniel Stoljar and Nic Damnjanovic () open their article on the
subject thus: ‘According to the deflationary theory of truth, to assert that
a statement is true is just to assert the statement itself.’ Stating that
‘Proposition P is true’ only amounts to the assertion of P. A simple-
minded version of Tarski’s disquotation schema is repeated endlessly in
this context: ‘Snow is white’ is true if and only if snow is white. That is all
there is to the meaning of ‘true’, the deflationists argue. Or as Paul
Horwich puts it (a, pp. –): ‘The basic thesis of deflationism, as
I see it, is that the equivalence schema “The proposition that p is true iff p”
is conceptually fundamental.’
Deflationism is not objectionable in itself, but for anyone trying to get a
sense of what it means for certain propositions to be true in concrete
situations, it is not a useful doctrine. Especially in the empirical domain,
it is crucial not to restrict the function of the truth concept to simple


As Greg Ray (, p. ) notes, this is done without mentioning the fact that not all ‘T-sentences’
need to be true, and that Tarski himself wrestled with the liar paradox.

Horwich explains further: ‘By this I mean that we accept its instances in the absence of supporting
argument: more specifically, without deriving them from any reductive premise of the form “For
every x: x is true = x is such-and-such” which characterizes traditional (“inflationary”) accounts
of truth.’

In this judgement I broadly follow Cheryl Misak’s (b) pragmatist critique of various
deflationary accounts of truth. However, I hesitate to follow the Peircian conception of truth
advocated by Misak.

8: /73 791
 Overview 
affirmation. As Price () argues, a key function of the truth concept is
to prompt people to engage in debates aimed at resolving disagreement. If
I ask you ‘Is this statement really true?’ I am quite likely demanding to
know whether there are good enough grounds for it. If we are asked ‘Is it
true that putting more carbon dioxide in the atmosphere raises the global
mean temperature?’, I think the questioner would probably be expecting
an answer like ‘It must be the case because carbon dioxide absorbs infrared
radiation very well, trapping the heat escaping from the earth into space in
that form’, or ‘At least there seems to be a clear correlation between higher
carbon dioxide levels and average global temperature.’ If instead we
responded to this demand in a deflationist way by simply reasserting the
statement, we would be missing the point of the question. In the empirical
domain we seek truth that is learned and tested by experience, rather than
truth that is a matter of mere assertion, self-consistency and logical
inference.
Deflationists may be quite correct that ‘philosophers looking for the
nature of truth are bound to be frustrated . . . because they are looking for
something that isn’t there’ (Stoljar and Damnjanovic ). But I think
deflationists are wrong in their diagnosis of where the trouble for ‘infla-
tionism’ lies. It’s not that there isn’t anything substantive to truth, but only
that there are many things meant by truth, so that attempts to find a
universal notion more substantive than the deflationary one fail. In my
view, more productive than deflationism is ‘truth pluralism’ as advocated
by Michael P. Lynch and Nikolaj Pedersen: ‘There is thus a range of
properties (correspondence, superassertibility, coherence, etc.) that consti-
tute truth for different domains of discourse.’ In Section . I will give a
more considered view of truth pluralism, also building on the homespun
version of it that I have given before (Chang a, sec. ..) without
realizing that there was already considerable literature on it.

Primary and Secondary Truth

In thinking about the different meanings and functions of truth, it is
helpful to start with a distinction between what I will call primary truth
and secondary truth. A true proposition is true in the secondary sense if
its truth derives from the truth of other propositions (or we might say that


Kevin Scharp () argues that even deflationary truth needs to be separated into two notions, in
order to avoid contradictions.

Pedersen and Lynch (, p. ); see Lynch () for further discussion.

8: /73 791
  Truth
it possesses secondary truth, or that it is a secondary truth). A true
proposition whose truth does not derive from the truth of other proposi-
tions is true in the primary sense (or, it possesses primary truth, or it is a
primary truth). The expression ‘derives from’ is a loose way of speaking,
and gets at the sense of grounding. Rather than going deeply into the
nature of grounding, I would like to think more practically about specific
ways in which the truth of one proposition depends on the truth of other
propositions. Perhaps the most straightforward way in which secondary
truth can be constituted is by enumerative induction in a finite set. ‘All of
my cats are black’ is true in virtue of it being true that each and every cat of
mine is black. Similarly straightforward would be grounding by deductive
consequence from other propositions that are true. Various other ways of
grounding secondary truth in primary truth will enter the discussion
later on.
One distinction that needs to be made carefully at the outset is between
the constitution of truth and the justification of belief. I want to focus on
the question of what makes something true, not the epistemic conditions
concerning how we can know what is true and what is not. If justification
means finding good reasons for believing something, then the line of
justification may or may not follow the line of constitution of truth. You
may justifiably believe that all of my cats are black because Stuart told you
so and Stuart is generally a reliable witness, but the truth of the proposition
‘All of my cats are black’ does not actually depend on Stuart being a
reliable witness (or on your wifi connection being good enough to allow
you to hear clearly what he said). Now, sometimes the line of justification
does follow the constitution of truth – if, for instance, you justified your
belief by seeking out each of my cats and confirming their blackness in
each case.
For our present purposes, it is helpful to set aside the question of
justification. The constitution of truth does not suffer from vicious
circularity or infinite regress. The primary–secondary distinction concern-
ing truth is clearly a hierarchical one, and we need to recognize that the
constitution of primary truth is a fundamentally different kind of thing
from the constitution of secondary truth. But this is a tricky process, as
Wittgenstein memorably indicated in On Certainty: ‘It is so difficult to
find the beginning. Or, better: it is difficult to begin at the beginning. And


My terminology harbours a deliberate ambiguity that I think is convenient, and harmless enough:
‘truth’ means both the quality of being true, and a proposition (or statement) that is true. This is
similar to ‘reality’ meaning both the quality of being real, and an entity that is real.

8: /73 791
 Overview 
not try to go further back.’ I will make some further clarifications on the
interactions between primary and secondary truth in Section ., but
I think I have now said enough to allow me to get on to the main idea
I want to advance in this chapter.

Truth-by-Operational-Coherence
I want to craft a non-deflationary theory of truth suitable for empirical
domains, including science and much of daily life as well. In an empirical
domain we find realities (as defined in Chapter ), entities that do not do
as we wish, which are mind-framed but not mind-controlled. In an
empirical domain we learn facts through experience; it is a different kind
of setting from the a priori domains of logic and mathematics as they are
commonly understood, where truth just follows from postulates that are
adopted by the mind. It is also different from fictional domains, where
truth can be imagined as we wish. Whether moral or religious truths are in
empirical domains is a controversial question, which I will not try to
answer here. But I do want to offer a conception of truth that can be
usefully applied in whichever domains that one may treat as empirical.
The key task is to understand what constitutes primary truth in empir-
ical domains, and this is my proposal: a statement is true to the extent that
there are operationally coherent activities that can be performed by relying on
its content. In parallel to that, taking a proposition as the content of a
statement, we can say: a proposition is true to the extent that there are
operationally coherent activities that can be performed by relying on it. Let
me call this truth-by-operational-coherence. The notion of ‘operational
coherence‘ here is what I developed initially in Sections . and ., and
already used in my characterization of reality in Chapter . I hasten to add
that I am by no means presenting truth-by-operational-coherence as the
only kind of truth there is. As I will explain fully in Section ., I subscribe
to pluralism concerning truth: there are many different notions of truth,
which have different uses in various domains and contexts. But I do want
to propose that truth-by-operational-coherence is what constitutes primary
truth in empirical domains, therefore something that we should centrally
concern ourselves with in the philosophy of science.


‘Es ist so schwer den Anfang zu finden. Oder besser: Es ist schwer am Anfang anzufangen. Und
nicht versuchen weiter zurück zu gehen’ (Wittgenstein , p. , §, emphasis original).

I reject the idea that statements do not bear truth-value, as it does excessive violence to
normal language.

8: /73 791
  Truth
Now, let’s consider more carefully how operational coherence, which is
a property of an activity, relates to truth, which is a property of a
proposition or a statement. In order to build up some initial intuitions,
let us take a fresh look at the kind of empirical statements that we do take
to be so secure that they can serve as the ground for other true statements,
and ask why we regard them to be true:
‘The ground is firm.’
‘Here is a hand.’ (G. E. Moore)
‘The people that I see when I walk into a room are really there.’
‘When I wake up in the morning, the earth will still be here.’
Why do we take such propositions as unquestioned truths? It is not
because they are justified by some other propositions that are more
fundamental or more secure. This is where we need to remember
Wittgenstein’s warning against trying to go back further than the begin-
ning. In my own idiom: we should not mistake the above propositions as
secondary truths. They are primary truths, which means that the grounds
for the truth of these propositions need to be found in themselves (or
rather, in their functions). So my proposal is that they are true-by-opera-
tional-coherence, true in the sense that numerous activities that we rou-
tinely carry out are reliant on them.
Take the proposition that the ground is firm. We carry out most
activities of our earthly lives on the basis of this assumption, though freak
events do sometimes happen to disturb the assumption, for example when
sinkholes suddenly open up. I remember experiencing a major earthquake
(and various aftershocks following the big one) in October  in Palo
Alto, near San Francisco. (I was just starting my graduate work in philos-
ophy, in fact sitting in a colloquium when the tremors started, so my
philosophizing will never be entirely free from that experience.) For a short
while afterwards all aspects of life were different, not even being able to
walk around assuming that the ground was fixed. This strange existence
did stop after a while, but that would have been different if major
earthquakes had kept happening every few days. But as long as the reliance
on the statement in question supports an effective way of life, then we go
on regarding it as true. And why shouldn’t we say that it is true that the
ground is fixed (even though it is also true that the earth is moving through
space incredibly fast), when and where our activities premised on that idea
are operationally coherent?
In science, too, the kind of truth possessed by foundational propositions
is truth-by-operational-coherence. Why should we take direct empirical

8: /73 791
 Overview 
observations as true in general, even though philosophical sceptics have
shown plenty of reasons for doubting them? Because it turns out that we
can carry out a great number of operationally coherent activities on the
basis of taking our own and others’ sincere observational reports at face
value (reserving doubt for particular and unusual circumstances). The
same also goes for basic propositions of a more theoretical nature – for
example, that the speed of light is constant in all directions regardless of
the motions of its source or receiver, or that all genetic information is
contained in the DNA molecule. Even though these basic theoretical
propositions are often taken as axioms, that is, held true by decree,
continued adherence to the axioms would be pointless unless the activities
that they support were operationally coherent. This is the spirit of the
arguments by Dewey and C. I. Lewis that a priori elements, even
the axioms of logic, can in the end only be justified pragmatically by the
operational coherence of the reasoning activities that they support (see
Sections . and .). Certainly in the realm of empirical science, theo-
retical propositions cannot be upheld as true simply by decree. Even
though some central empirical propositions are treated dogmatically (like
those in the ‘hard core’ of a research programme in Imre Lakatos’s view of
science), that should only be a temporary and provisional situation. We try
to devise epistemic activities on the basis of the postulates, to see how well
we can come up with coherent activities.
As I will explain further in Section ., truth-by-operational-coherence
is a robustly pragmatist notion – not only in paying attention to practical
consequences of our beliefs, but in having our thoughts fully rooted in
experience. In relation to the standard critique of the classical pragmatist
theory of truth, there is one point worth stressing immediately, and briefly
for now: the definition I offer here is precisely not in the spirit of
‘something is true if it is convenient for me to believe’. A coherent activity
is a difficult thing to devise, and it will only work out if our assumptions
entering into it are suitable; that is how empirical statements must be put
through the test of experience. Operational coherence carries within it the
constraint by nature, which gives truth-by-operational-coherence the mark
of mind-independence in the sense of the absence of mind-control, which
is something that many realists rightly value in their favoured correspon-
dence notion of truth. The basic pragmatist intuition is that primary truth
is something that we can live by. Unlike religious truth in its ordinary
conception, which one also lives by, empirical primary truth is tested by
experience; it is revisable in response to the expansion of our experience or
changes in our situation. When we conceive truth in terms of coherent

8: /73 791
  Truth
activities, it is squarely placed in the realm of active knowledge as
I presented it in Chapter .
If we understand empirical primary truth as truth-by-operational-
coherence, what about empirical secondary truth? Here we have a natural
way of rehabilitating the correspondence notion of truth, in a more down-
to-earth form, along the lines that I suggested in Section .. I think it
would be uncontroversial that the substantive theory of truth most suitable
for empirical secondary truth is the correspondence theory. One would
have to work out exactly how secondary truth is constituted from primary
truth in empirical domains, going beyond deduction and enumerative
induction as discussed above, but that is a tractable problem that can be
tackled plausibly. When we see the nature of truth-by-correspondence as
secondary truth, there is no need to imagine a ‘transcendent’ type of reality
to which our theories need to correspond in order to possess truth-by-
correspondence.
Again, it is important not to confuse the constitution of truth itself and
the justification of our knowledge of it. The operational coherence of
activities relying on a true proposition are not consequences or indications of
its truth, from which we may infer the truth. Rather, operational coherence
is constitutive of truth. As Dewey put it ([] , pp. –): ‘the
effective working of an idea and its truth are one and the same thing – this
working being neither the cause nor the evidence of truth but its nature’. As
with the notion of reality in Chapter , what I am making here concerning
truth is a semantic move, in the sense that I am proposing what we ought to
mean by ‘truth’, in order to render it as a useful concept.

Against Truth-Absolutism
Truth-by-operational-coherence is a robust enough notion to serve all the
main functions that we would want a concept of truth to serve in empirical
domains. And it can serve those functions without being an absolutist
notion. For example, we should say that Newtonian mechanics remains
quite true (true-by-operational-coherence) in its proper domain. It would
be not only annoying and pedantic, but actually unwise, to insist that we
now know that Newtonian mechanics is false and that it just gives
approximately correct empirical predictions in certain situations. If we
do that, by the same lights we would be obliged to say the same about
general relativity, quantum mechanics, or any other theory we now have.
Then ‘true’ would become a designation that can never actually be used.

8: /73 791
 Overview 
Should we really want to discard or incapacitate such a crucial concept for
our intelligent life? That would be like insisting that we should never say
things like ‘She is a good person’ unless someone is unfailingly and
perfectly good. Should we be saying that she is an ‘approximately good’
person, or a ‘benisimilar’ person? No, better to stick with the common
sense of saying that she is a good person, but like all of us she could
become still better in various ways. Many physicists have now become
content to say that every theory is an ‘effective theory’, true with limited
scope (see Cao and Schweber ); their main focus is on the pertinent
energy level at which each theory functions, but the lesson can be gener-
alized to cover other parameters, too.
Generally speaking, truth-by-operational-coherence is a qualitative attri-
bute. A proposition is true to a higher degree if it supports a larger number
and variety of operationally coherent activities, and if such activities rely on
the proposition in question more strongly; moreover, operational
coherence itself is a matter of degrees. And it is the quality of truth itself
that is a matter of degrees, not just our knowledge of it or our belief in it.
This is actually consonant with everyday usage: as Austin noted ([]
, pp. , –), ‘very true’, ‘true enough’, etc. are perfectly sensible
locutions, and it is unreasonable to try to reduce ordinary judgements of
truth to yes/no. The spirit of Austin’s observation has now become current
again, thanks to its revival by Catherine Elgin in her True Enough ().
And the degree of truth I am speaking of will not be quantifiable in a
simple numerical way, nor reducible to probabilities. In practice, truth-in-
degrees is already a widely accepted notion in the philosophy of science.
Many philosophers, most of all in the course of defending standard
scientific realism, have already fallen into the habit of speaking about
‘approximate truth’, and Richard Boyd () has argued convincingly
that it is not possible to maintain scientific realism without relying on
some notion of approximate truth. Perhaps one could say that ‘approxi-
mate truth’ is an imprecise way of speaking, and what we are really talking
about is an approximation to the truth, while truth itself remains a yes-or-
no matter. But I do not see what would be gained in preserving binarity for
truth in that way. On the contrary, as hinted in Section ., the current
flourishing of many-valued logic with its use in computing and artificial
intelligence would tend to suggest that departing from a binary notion of
truth is a legitimate and potentially very useful move.
Another factor that makes truth non-absolute is its finite scope, or
domain-specificity. According to my definition of truth-by-operational

8: /73 791
  Truth
coherence, a statement that is true in a certain domain can easily fail to be
true in other domains (i.e., it may not support coherent activities there).
And who would deny, on reflection, that a given statement can be true in
some domains and not in others? We are all familiar with situations where
a law of nature is only true in some cases, such as the classical laws of
motion being quite wildly invalid in various quantum-mechanical
domains. It would be useful for us to get into the habit of always asking
‘where/when is this statement true?’ as an antidote to absolutist and
universalist tendencies. Now, many people will be uneasy about attaching
limited scopes to truth. This leads to what I might call the ‘witchcraft
objection’: a patently untrue theory (e.g., of witchcraft) may work well
enough (support coherent activities) within a small domain, but it seems
wrong to say that it is true even within that domain. This issue will be dealt
with fully in Section ., but for now let me just express my view that what
we are dealing with here is an unnecessary fear. There are reasons for which
we should not want to grant truth to the theory of witchcraft even in a
narrow domain (and similarly for climate change denial, vaccine refusal,
‘young-earth’ creationism, flat-earth cosmology, and so on): it does not in
fact support operationally coherent activities very well in any domain; the
claim that it worked well in certain situations is wildly exaggerated. On the
other hand, if we do have a theory that actually works well and does not
contradict other established truths within a given domain, there is nothing
wrong with regarding such a theory as an empirical truth in that domain.
I think it is correct to say things like: ‘Newtonian mechanics remains true
in its domain of application.’ And it has not shown to be less true in most
of its old domain than quantum mechanics and general relativity are,
because no one has yet tried applying the latter theories (unmixed with
classical mechanics) to most classical situations.
Yet another non-absolute aspect of truth-by-operational-coherence is
plurality. If truth-by-operational-coherence only makes sense in the con-
text of an activity (or within a whole system of practice), then there can be
different sets of truths belonging to mutually incommensurable activities
or systems. This points to epistemic and ontological pluralism, as a
separate issue from ‘truth pluralism’, which is about the meaning of ‘truth’
itself. The plurality I have in mind now is all exhibited under one concept
of truth, namely truth-by-operational-coherence. It is true that light is an


One could avoid the domain-specificity of truth by writing the domain-restriction explicitly into the
statement in question and treating the restricted statement as strictly true or not. That would be a
losing game because not all possible domain-restrictions can be anticipated and specified.

8: /73 791
 Different Kinds of Truth 
electromagnetic wave, and it is also true that light is composed of photons.
This feature of truth-by-operational-coherence is very consonant with the
pluralism concerning science articulated in my earlier work, which is ‘the
doctrine advocating the cultivation of multiple systems of practice in any
given field of science’ (Chang a, p. ). I will discuss pluralism
further in Section ., and again in relation to realism in Chapter .
Recognizing the simultaneous truth of statements made in different sys-
tems of practice does not easily result in a logical contradiction, even when
they seem to say conflicting things. This is for two reasons. In many cases
the apparently contradictory statements are not employed in the same
situations, so they do not actually clash. And when they do concern the
same situations, there is usually a sufficient degree of semantic incommen-
surability so that they do not directly contradict each other, talking past
each other instead. It is important to note that the multiplicity and non-
absoluteness of truth do not amount to a crude or extreme form of
relativism, or ‘idiot relativism’ as Arthur Fine (, p. ) calls it.
Within each system of practice, truth-by-operational-coherence is fixed
in a way that is not controlled by our wishes or expectations. There is no
reason to fear the multiplicity of such mind-framed yet mind-
uncontrolled truth.
In the remainder of this chapter I will give further elaborations on my
perspective on truth. As a preliminary step, in Section . I will present
various other valid conceptions of ‘truth’ differing from truth-by-opera-
tional-coherence, and discuss what kinds of situations render these notions
suitable. In Section . I will give a fuller discussion of truth-by-opera-
tional-coherence, and elaborate further on how this notion of truth is not
an absolutist one, and try to anticipate worries that people may have
concerning this lack of absoluteness. Section . will argue that we should
have a concept of empirical truth according to which truth comes attached
with a specific scope, and Section . will come to terms with the fact that
multiple truths-by-operational-coherence can co-exist in a given subject
area. Finally in Section . I will link this notion of truth back to the
traditional pragmatist ideas about truth, and clear up some persistent
misunderstandings about the classical pragmatist ideas.

. Different Kinds of Truth

Before I go on to discuss my notion of truth-by-operational-coherence in
detail, it will be useful to see it in relation to various other meanings of
truth. Following the general spirit of pragmatism, and in sympathy with

8: /73 791
  Truth
‘truth functionalism’, I identify and distinguish various meanings of truth
according to the different types of activity within which the terms ‘true’
and ‘truth’ acquire their meanings, focusing on empirical science and the
most common practices of daily life. This analysis will produce partially
overlapping meanings of truth, rather than a disjunctive set of mutually
exclusive truth-concepts. Truth-by-assertion is based on a minimalist idea
that asserting a proposition is equivalent to saying that it is true. We have
truth-by-honesty when someone tells you without distortion what e thinks
or feels. Truth-by-decree obtains when we postulate something to be true.
None of these senses of truth captures what we mean when we say that
something is empirically true, which points to truth-by-operational-coher-
ence. And as discussed in Section ., there is also secondary truth (which
I will also call truth-by-comparison), which is a matter of agreement with
other propositions that are already established as true. It is important to
note that various types of truth often operate in conjunction with one
another, which creates interesting complications in using the notion of
truth in real practices.

Functionalism and Pluralism Concerning Truth

What should a philosophical theory of truth do? I think it should give a
synoptic view elucidating how the concept of truth is used in various domains.
This is according to my late-Wittgensteinian inclination to take meaning as
arising from use, treating truth in this regard like any other term. My debt to
Austin’s perspective, in general and specifically on truth, has already been
noted. Within the ‘neo-pragmatist’ tradition Robert Brandom () and Huw
Price (; ) have stressed the need to consider functions of the concept
of truth. More recently Lynch (; ) has added to the advocacy of ‘truth
functionalism’. There are subtle differences among these various positions and
perspectives, especially concerning whether we are inquiring into the meaning
or function of truth. Instead of entering into a debate about the differences,
I will build on what I see as a common core: attention to how the concept of
truth is actually used in various activities. So the task here is not to review
various theories of truth in order to say which is the correct one, but to see
how different conceptions of truth are operative in different contexts. And
this approach is also consonant with a cross-cultural comparative approach to
the notion of truth, as advocated and practised for example by Alexus McLeod
().


For general surveys of various theories of truth, see Kirkham () and Glanzberg (b).

8: /73 791
 Different Kinds of Truth 
Table .. Various concepts of truth

Concept Core intuition Associations

Truth-by-assertion ‘P is true’ means ‘P’. Deflationism, disquotation

Truth-by-honesty P is true if it is asserted sincerely. Truth-telling, witnessing
Truth-by-decree P is true if we decide to affirm it. Postulation, axiomatic
systems
Truth-by-operational- P is true if it facilitates coherent Pragmatism, embedding in
coherence activities. active knowledge
Truth-by-comparison P is true if it agrees with other Correspondence,
(secondary truth) truths already established. confirmation

It is easy to see that the term ‘truth’ or ‘true’ is used in many different
ways in practice. In Lynch’s wry expression (, p. ): ‘The history of
attempts to identify the property that all and only true propositions have in
common has not been a happy one.’ This has led some philosophers to adopt
a pluralism concerning the concept of truth. Crispin Wright notes that in
certain domains truth is based on a correspondence-like relation, while in
others it is a matter of assertibility. Pedersen and Lynch (, p. ) draw a
very instructive analogy with the concept of ‘winning’: there is ‘variation in
terms of what winning amounts to across different games’, with ‘a range of
properties (scoring more goals than the opponent, checkmating the oppo-
nent’s king, etc.) that constitute winning for different games’. Likewise for truth:
the properties that ‘constitute truth’ may be very different in ‘different domains
of discourse’ – sometimes correspondence, sometimes superassertibility,
sometimes coherence. Kevin Scharp (; ) does not think that there
can be any concept of truth that satisfies all of the ‘platitudes’ normally
associated with truth, opting instead for a pluralist project of conceptual
engineering in which we try to craft a set of distinct alethetic concepts that
can replace truth.
In the remainder of this section I will give my own view on the
different meanings and functions of ‘truth’ in different domains, drawing freely
from the works of the authors mentioned above. Table . gives a quick
overview of the upcoming discussion. Going beyond identifying whole
domains of discourse, my analysis will dig down into the level of particular


See Lynch (); Wright (; ); and overviews in Pedersen and Wright (), Pedersen
and Lynch (), and Glanzberg (a, sec. .).

I will also be building on my own previous work on ‘truth and its multiple meanings’ (Chang
a, pp. –), which I am modifying significantly here.

8: /73 791
  Truth
types of activity in which each notion of truth is put to work. What I am
offering is not a disjunctive taxonomy but several overlapping categories.

Truth-by-Assertion and Deflationism

The simplest activity relevant to the notion of truth is that of assertion-and-
denial: to assert the proposition P is to say that P is true, and to deny P is to say
that P is false. What ‘truth’ means in this activity is that a proposition is being
asserted, regardless of whether one has good reasons for asserting it. Let’s call
it truth-by-assertion. These thoughts lead to the deflationary theory of truth.
Since any statement is liable to being asserted, it would seem that the
deflationary theory applies everywhere and it is the most basic idea of truth
underlying any and all others. But it would be unwise to presume that all other
notions of truth can and should be eliminated or reduced to the deflationary
one.
What exactly is this game of assertion, and when do we actually play
it? Assertion in the wild is not the dry thing that it may seem in a logic class.
There are other, more significant functions of saying something is true than a
simple repetition. If I follow someone’s utterance (including my own) by ‘It’s
true’, that is an act of emphasis, more effective and forceful than repeating the
statement in full. If I say ‘That’s true . . .’ in that intonation typically followed by
‘. . . but’, that is an act of approving the bare content of what someone said
while disputing some connotations or implications of it. Whether it is played
bluntly or subtly, assertion is a game of personal commitment. When I assert
something, I lend it whatever credibility I have, imply that I will act in
accordance with it, and open myself up to others’ judgement. That is to say,
there are different illocutionary functions served by truth-by-assertion, while
the narrow meaning of ‘is true’ is a simple affirmation of the proposition in
question. So even the context of assertion may not be completely suited for
the deflationary theory of truth.

Truth-by-Honesty and Truth-Telling

Another function of the word ‘true’ or ‘truth’ is to indicate the speaker’s
honesty or sincerity. The ‘truth’ involved here is a different kind of thing from
the more objectivist conceptions of truth, as indicated by Price (; ) in
his critique of the conflation between norms of sincerity and norms of truth.


I also think it is too hasty to rule out non-assertoric functions of truth that may be served by
unarticulated content that is only implicitly assumed.

8: /73 791
 Different Kinds of Truth 
The relevant activity for this notion of truth, which I will call truth-by-honesty,
is that of truth-telling, or giving witness. Truth-by-honesty could be considered
a kind of correspondence, between what I assert and what is in my mind, a
match between what I say and what I perceive or think. But ‘match’ is not
always quite right here, because the exact mental process going on in truth-
telling is often not such a straightforward correspondence. In any case, truth-
telling is operationally meaningful. ‘I am telling you the truth when I say I saw a
snow leopard on the hill.’ This meaning is cogent whether or not my thoughts
are true in some other sense (‘A snow leopard in London? – surely not’). Truth-
telling is not mere assertion, and a lie is also an assertion. Without the activity
of truth-telling no epistemic community would be able to sustain itself, and no
collective pursuit of knowledge would be possible. Truth-telling is not only a
matter of personal character and disposition, but also an important part of the
foundation of knowledge in any empirical system of practice.

Truth-by-Decree: Axioms and Conventions

Some statements are true because we decide that they shall be true. But such
truth is only valid within our commitment to uphold it, and only interesting
when we go on to do something on the basis of that commitment. I will call
this truth-by-decree, the archetypes of which are definitions and tautologies.
Of course it is true that the standard meter is one meter long (if it still is the
standard of length, which it actually isn’t any more in real metrology). And one
cannot doubt the truth of the statement ‘All bachelors are unmarried’ since it is
a tautology. These are truths that we construct, judge and maintain by making,
using and enforcing specific meanings. They are also rendered true by pre-
sumption when we take them as given and engage in activities on their basis.
In mathematics we build systems of inference starting from axioms
that are true by decree. Scientific theories can also be constructed and applied
in an axiomatic manner. For example, when Einstein in  postulated that
the speed of light was the same regardless of the motion of either the
observer or the source, this was a new axiom, which by no means followed
from the previous definitions of ‘light’ or ‘speed’. Another kind of truth by
decree is conventions, both in the colloquial sense and in Poincaré’s sense.
Colloquially, a convention is something arbitrary that is socially agreed upon,
like driving on one side of the road or the other, or when new technical terms
or measurement units are defined. According to Poincaré, scientists decide to
‘elevate’ the best-confirmed empirical laws to the status of conventions, to put
them beyond questioning. But even in that scenario truth-by-decree does not
in itself amount to empirical truth.

8: /73 791
  Truth

Truth-by-Operational-Coherence
Now I come back to the kind of truth that I am most concerned with, which
I named truth-by-operational-coherence in the last section. (I will be brief
here, since this concept has already been introduced, and will be elaborated
further in the rest of the chapter.) This is the notion of truth at the heart of
empiricism, embodying the idea that empirical truth is determined by what
mind-uncontrolled realities are like, not by what we believe, say, do or wish.
None of the other notions of truth discussed above captures this essential
empiricist requirement. In connection with Price’s theory (), it may seem
that truth-by-operational-coherence only answers to the norm of personal
warranted assertibility; however, to the extent that epistemic activities are
conducted by communities of people, the Pricean norm of truth, which
demands mutual agreement, is also in play.

Truth-by-Comparison
In Section . I drew a distinction between primary and secondary truth.
Truth-by-decree and truth-by-operational-coherence are matters of primary
truth, and truth-by-honesty could also be considered a matter of primary truth.
Secondary truth is truth-by-comparison, taking ‘comparison’ in a broad
sense. The grounding of secondary truth in primary truth can happen in
various ways. Earlier I mentioned deduction and enumerative induction. If
we want to be most liberal about secondary truth, we might allow it to be
established by mere lack of contradiction with primary truths, or by inference
to the best explanation. Interestingly, the traditional correspondence and
coherence theories of truth both belong cogently in the realm of truth-by-
comparison.

The Complex Functioning of the Notion of Truth in Real Practices

I hope that the distinctions made so far will be helpful in clarifying our thinking
about truth in various domains of life. In typical situations in everyday life or
scientific practice, various notions of truth are simultaneously in play, often in
conjunction with each other. If we try to think about the whole conglomerate
situation with just one undifferentiated notion of ‘truth’, we are bound to
encounter difficulties.


It is closely related to what I have called ‘truth’ previously (Chang a, p. ), but that notion
also encompasses truth-by-comparison.

8: /73 791
 Empirical Truth and Operational Coherence 
When different notions of truth are applied simultaneously to a given
statement, the verdict of truth may vary. It often happens that a statement that
is unwaveringly asserted or upheld honestly cannot support coherent activi-
ties. And it may be the case that Donald Trump often tells truth-by-honesty
when he expresses his opinions (which is something that his supporters value),
but the things he says really lack truth-by-operational-coherence in empirical
domains. It can also happen that a proposition has a high degree of truth-by-
operational-coherence, but is contradicted by a theoretical principle that is a
truth-by-decree. In empirical science we routinely examine whether a postu-
late laid down as a truth-by-decree also has truth-by-operational-coherence;
ultimately that is the process of empirical confirmation of theories. It is a
wonderful scientific moment when truth-by-decree and truth-by-operational-
coherence line up with each other, but that is by no means to be taken for
granted. Poincaré’s conventionalism plays it safer, by recommending that we
ascertain the truth-by-operational-coherence of a statement to a high degree
first, and then assign truth-by-decree to it.
It is also important to note that one and the same proposition may
work both as a primary truth and as a secondary truth. Take the proposition
that the atomic weight of chlorine is .. This may be a primary truth
determined by how well certain activities relying on this statement work out
in chemistry and physics. For example, on this basis we can determine the
amount of chlorine that will enter into chemical combination with certain
amounts of other substances, and that works out well. But we may also
establish the atomic weight of chlorine by measurement. In that case it is
established as a secondary truth, grounded in the truth of certain physical
assumptions underlying the operation of the measuring instrument. For
another example, take the proposition that the surface area of a sphere is
proportional to the square of its radius. This is a secondary truth in Euclidean
geometry, provable from the axioms. But we can also take it as an empirical
hypothesis, in which case it is established as a primary truth if we can rely on it
for a set of coherent activities, ranging from figuring out the amounts of paint
needed to paint balls of different sizes, to the Kantian deduction of the inverse
square law of gravitation (Kant [] , pp. –; – in the original).
Generally speaking, knowing how to manage different senses of truth is part of
the art of inquiry.

. Empirical Truth and Operational Coherence

Turning now to truth-by-operational-coherence, in this section I will focus
on explaining and justifying the basic idea that truth consists in the

8: /73 791
  Truth
facilitation of coherent activities, which will seem far-fetched to many
people. Again I emphasize that my proposal of the notion of truth-by-
operational-coherence is a semantic move concerning what primary truth
should mean in empirical domains. Being relied on in coherent activities is
constitutive of the truth-by-operational-coherence of a proposition, rather
than being an effect or evidence of truth. I will also try to say more about
what exactly is meant by the ‘reliance’ that coherent activities have on true
propositions. The key point is that reliance is a practical kind of necessity:
is the use of the proposition in question needed for the coherence of the
activity in question? I also address the problem of coherent activities that
rely on propositions that are not even intended as true, and argue that such
propositions should be regarded to be true in relevant respects, or to an
appropriate extent. Unlike the idea of correspondence to an inaccessible
kind of ultimate reality, truth-by-operational-coherence is a concept that is
in operation in all sorts of practices, including the actual procedures for
theory-testing.

Coherence as Constitutive of Truth

Having set out various notions of truth and their uses in Section ., I can now
give a more focused discussion of truth-by-operational-coherence. Recall the
brief definition of the notion that I gave in Section .: a proposition is true to
the extent that there are operationally coherent activities that can be per-
formed by relying on it. I believe that this notion is the chief sense of primary
truth operative in empirical domains, and so it provides crucial foundation for
any empiricist epistemology or philosophy of science.
The first issue to consider more carefully is the semantic nature of my
pragmatist proposal, to take operational coherence as constitutive of primary
truth in empirical domains. Against this view you might want to maintain a
more traditional stance, according to which truth is whatever it is (e.g.,
correspondence to ultimate Reality), and things like operational coherence
are only relevant to the epistemic questions of how we know and learn truths.
This approach amounts to separating out the metaphysics and the epistemol-
ogy of truth; the epistemological side can be rightfully empiricist, while the
metaphysical side remains with correspondence realism (as defined in Section
.). The common intuition is that an empirical statement possesses mind-
independent truth, and that our epistemic activities such as hypothesis-testing
are designed to find out this pre-existing truth. That is the intuition that I am
trying to steer us away from.

8: /73 791
 Empirical Truth and Operational Coherence 
We must not overplay the distinction between truth itself and the
evidence for it, when it comes to truth-by-operational-coherence. I, too, want
to preserve a distinction between truth and our personal knowledge of it, but
taking truth to be separate from actual epistemic activities is not the way to do
it. I think that is what James had in mind when he said ([] , p. ):
‘The reasons why we call things true is the reason why they are true.’ What it
means for a statement to be true-by-operational-coherence is the same sort of
thing as how we tell that it is true-by-operational-coherence. I propose to
make the distinction between truth and knowledge as follows, in parallel with
how I dealt with reality vs. our knowledge of it in Chapter :

• A proposition is true to the extent that there are operationally coherent

activities that can be performed by relying on it.
• I know that a proposition is true to the extent that I (personally) actually know
how to carry out some operationally coherent activities by relying on it.

So we can retain the role of operational coherence in the constitution of truth,

while removing matters of personal circumstance from it. I fully expect that
there are coherent activities that I have not yet learned to perform, or
even conceived.
Truth-by-operational-coherence is the same thing as empirical
confirmation, if we take ‘confirmation’ in a broad sense. Let’s get back to some
basic intuitions here. What should confirmation mean, other than surviving
‘the test of experience’? And why should the test of experience be confined
only to explicit testing, let alone testing that follows a particular formal scheme
(e.g., the hypothetico–deductive model, Bayesian probability-updating, or null-
hypothesis testing)? Explicit theory-testing might be the nearest activity-
version of the idea of a statement corresponding to the world, but that is
not a compelling reason to privilege it. If a theory can facilitate coherent
activities, and not as a result of any strange accident or coincidence as far as
we can see, then we can and should take (the relevant statements in) this
theory to be ‘true’ – in the same down-to-earth sense as we say that it is true
that rabbits have whiskers and live in underground burrows.
Theory-testing is not a distinct kind of activity from theory-use, but a
subset of theory-use. Every attempt to use a theory may in principle function
as an empirical test of it, and the truth or falsity of a proposition often emerges
in activities that are not explicitly conceived as tests. ‘The distance from Madrid
to Barcelona is  kilometres’ is untrue, and would be exposed as such, in that
most activities we might try to carry out on its basis (such as walking from one
place to the other in an afternoon, or hitting one from the other with artillery)
would be operationally incoherent. Scientists in various scientific fields ranging

8: /73 791
  Truth
from epidemiology to cosmology are very familiar with the situation in which
no explicit tests can be devised, and hypotheses just have to be evaluated by
how well they account for observed phenomena ‘in the wild’. (The process of
validating new vaccines goes slowly because we do not have human guinea
pigs to whom we can deliberately give the pathogen after giving them the
vaccine, but that is just a constraint we have to live with, unless we want to
change the fundamental ethical basis of social life.) And even the classic
hypothetico–deductive method of theory confirmation involves using the
hypothesis to derive predictions, which we then evaluate by comparison with
observations. Theory-confirmation is a measure of how coherently this activity
works out, of making empirically accurate predictions relying on the content of
the theory under test.
Any positive test-outcomes partially constitute the theory’s truth; it
would be a category-mistake to say that positive test-outcomes are caused by
the theory’s truth. James warned against the reification of truth as with other
‘words in th’ (wealth, health, strength and truth), quoting an epigram from
Lessing: ‘How come it’s exactly the richest people in the world who have the
most money?’ There is, of course, nothing to explain here, because having
lots of money is exactly what being wealthy means (James [] ,
pp. –). So it is, too, with truth-as-operational-coherence. Here again I am
rejecting the relation between empirical success and truth envisaged in
standard scientific realism, according to which truth explains success. It may
be instructive to contrast my view with Kitcher’s, as he tries to preserve an
explanatory role for correspondence truth. Although he conceives the realist
success-to-truth inference in a pragmatist or activity-based way, Kitcher still
preserves the inference from ‘S plays a crucial role in a systematic practice of
fine-grained prediction and intervention’ to ‘S is approximately true’ (Kitcher
, p. ). In more recent work, he takes this success–truth link as a
‘plausible empirical conjecture’; if the conjecture holds, it forms the basis for
treating truth as the explanation for success (Kitcher forthcoming, ch. ). In
contrast, I do not take the success–truth relation as a matter of inference or
explanation; instead I take success (or rather, operational coherence) as the
core of the very meaning of ‘approximately true’, or rather, ‘true (to a degree)’.
In James’s evocative expression, ‘truth happens to an idea. It
becomes true, is made true by events.’ This is a pointer to what happens as


A finite number of positive test-outcomes will not add up to the complete truth of a general theory;
this is like the relation between instances of empirical success and the full empirical adequacy of a
theory, in terms of van Fraassen’s epistemology.

‘Wie kommt es, Vetter Fritzen / Dass grad’ die Reichsten in der Welt / Das meiste Geld besitzen?’

James [] (, p. ) emphases original; quoted in Kitcher (, p. xxiii).

8: /73 791
 Empirical Truth and Operational Coherence 
the idea goes on to facilitate further coherent activities. Truth-by-operational-
coherence consists in the significant roles that propositions play in our various
activities, and more such roles emerge as we continue to learn to do more
things. In an imperfect yet suggestive metaphor: think of a proposition attain-
ing truth like a child growing up. Even though we would say that a particular
child has the potential to grow up to be a certain kind of adult, the adult is not
fully there in the child. So with truth. We can allow that there are coherent
activities as yet unknown to us, but when we have to make actual concrete
judgements of truth we have nothing else to go on except the truth that is
already known in activities that we can actually perform.

The Meaning of Reliance

Return to the definition of truth-by-operational-coherence: a proposition is
true to the extent that there are operationally coherent activities that can be
performed by relying on it. What exactly does it mean for an activity to ‘rely on’
a proposition? In evaluating the truth of a proposition, we should ask: can the
coherence of the activity in question be maintained, if we do not let the
proposition in question inform what we do? The sense of ‘reliance’ incorpo-
rates a practical and empirical need, a sense that the activity in question
cannot, as a matter of fact, be performed without making use of the propo-
sition in question. For example, most problem-solving in classical electrody-
namics cannot be done without relying on Maxwell’s equations. The notion of
‘reliance’ is meant to capture the sense in which the employment of a
proposition is productive and meaningful. For what I have in mind by ‘employ-
ment’ here, it is not sufficient that a proposition is asserted in the course of an
activity. So we can perfectly well use Maxwell’s equations while denying that
the aether exists, even though many pre-Einsteinian physicists did frequently
assert that the aether existed when they were using Maxwell’s equations. The
activity of solving Maxwell’s equations for various problems in electrodynamics
does not rely on the proposition that the aether exists. Insisting on actual
employment also guards against the possibility that the proposition in ques-
tion might be involved in the activity in a superfluous way reminiscent of the
tacking paradox or the Gettier problem, in which case we would not want to
make an attribution of truth.


In a previous formulation I expressed the idea as follows: ‘a statement is true in a given circumstance
if (belief in) it is needed in a coherent activity’ (Chang b, p. ). What I meant to express by
the term ‘need’ there was not logical or metaphysical necessity, but a pragmatic necessity – what we
actually need, which can only be learned empirically. Now I propose to avoid invoking a notion of
necessity, since it is liable to cause confusion.

8: /73 791
  Truth
Reliance was the issue, for example, at the core of my debate with
Psillos regarding the successes of the caloric theory (Chang ), mentioned
briefly in Section .. I think I have shown to most people’s satisfaction that
some very coherent activities of thermal physicists around  did rely on key
tenets of the caloric theory. For example, Pierre-Simon Laplace’s derivation of
the ideal gas law relied on assumptions about the material reality of caloric,
about its corpuscular constitution, and the basic nature of the forces operating
between particles of caloric. Psillos had stated that no substantive assumptions
about the nature of caloric were relied upon in the successes of the caloric
theorists, and I disagreed. Which one of us was right about the historical facts is
not the main issue for my present purposes. I think we both agreed that this
kind of reliance, which we can judge by a concrete examination of the
practices of scientists, is what underpins the truth of the propositions in
question. To the problem of suspected superfluous propositions there is no
magic solution, only the hard work of empirical investigations. This sort of
checking may not live up to some overblown image of a philosophical test,
but it is how we get on in science, and in the rest of life, too.
It may be objected that reliance is too loose a notion, especially as it
would seem to incorporate psychological factors. Doesn’t it sometimes
happen that people rely on certain propositions in some psychological way
that should not be considered to have any bearing on truth? What if our
mental make-up is such that belief in a certain fantastical proposition is
psychologically necessary for us to carry out some activities? For example,
what if I can only swim by believing that I am a dolphin? Is it then true that
I am a dolphin, because my coherent activity of swimming relies on that
belief? On the face of it, this seems like a straightforward and devastating
objection to my notion of truth-by-operational-coherence, or any pragmatist
notion of truth for that matter. Actually the problem arises from positing an
inherent connection between truth and belief. Thinking in terms of reliance
on content instead of belief allows us to avoid unnecessary tangles with
psychologistic considerations. Psychological reliance on a belief is a different
issue from the employment of a proposition in an activity. Imagine a
nineteenth-century scientist who can only get enough motivation to do
his work in optics because he believes in the reality of the noble luminiferous
aether, although he does not rely on any particular assumptions about the
nature of the aether in his reasoning.


I thank Mike Martin for raising a version of this worry at the Aristotelian Society in January .
His example was about delusional self-confidence that allows scientists to tackle very challenging
problems (successfully).

8: /73 791
 Empirical Truth and Operational Coherence 

Coherent Use without Truth?

Now, there may be times when people put to good use some propositions
which they expressly do not regard as true. This is a common situation in
scientific research, for example when admittedly false models are used in
various coherent epistemic activities including prediction and explanation
(see Suárez ; Toon ; Rowbottom ). Doesn’t this make trouble
for my notion of truth-by-operational-coherence, according to which propo-
sitions relied upon in coherent activities are true? It would appear that my
account faces a threatening conceptual tangle in the idea of a coherent use of
an untrue proposition, while more standard accounts of truth can freely allow
that useful and even correct consequences can sometimes follow from false-
hoods. For several different reasons, this is not a problem.
A relatively simple type of case is where we may declare a theory or
model false on the whole, but actually trust or affirm the aspects of it that are
being actively relied upon in coherent usage. And then there are cases in
which we regard the theory or model in question not to be strictly true but
‘true enough’, and also freely allow that its uses are only coherent enough, so
there is no glaring mismatch between coherence and truth. Or we may make a
working commitment to hypotheses whose probability would have to be
assessed as not comfortably close to  per cent. Hypotheses are very often
used in the absence of conclusive confirmation, as we should know from the
existence of the phrase ‘working hypothesis’. In fact, life is very much a game of
relying on less-than-certain hypotheses, which are open to doubting and
testing. For example, we may count on the victory of a leading candidate in
an election, assume that a currently dominant scientific theory will not be
rejected for some time, or even plan the next year on the basis that we will not
be involved in car crashes or contract serious illnesses. Such working hypoth-
eses are relied upon provisionally because we do not have a better basis for
our activities. But if these activities turn out to be coherent, then the working
hypothesis does earn a measure of truth-by-operational-coherence. If not, our
commitment is withdrawn. So the mismatch between truth and coherent use
would be quite temporary.
It may also happen that the proposition in question is true in one
sense of truth, but not true in another sense of truth (recall the various senses
of ‘truth’ discussed in Section .). For instance, it can easily happen that a
theory or model lacks truth-by-comparison because it conflicts with another
that is more firmly established as true, but it possesses truth-by-operational-
coherence through its usage. For example, as mentioned in Section ., the
theory of atomic and molecular orbitals supports many extremely coherent

8: /73 791
  Truth
activities in chemistry, but is declared to be false by later versions of quantum
mechanics that treat electrons as identical (indistinguishable) particles. In such
cases, it would be wrong to say that the successfully used model is simply false.
It is true-by-operational-coherence at least in a limited domain.
Sometimes we take a theory or model as only apparently proposi-
tional. We may even subscribe to a generalized instrumentalism, which treats
theories or models as tools of reasoning rather than sets of propositions. In
that case the question of truth does not arise. That is a tenable position,
though I prefer a realist–pluralist outlook that attributes some truth-by-oper-
ational-coherence to each reliably successful model, as I will explain further in
Chapter . Or we may take a fictionalist attitude. Think back to the Newtonian
physicists with their supreme confidence in the truth of Newton’s laws of
motion and gravitation. Now we look back and say that Newton’s laws are not
entirely true, but we can make fictionalist sense of Newtonian practice, or in a
surrealist way, as Tim Lyons () would have it: the world, in many domains,
behaves as if Newton’s laws were true. So here is a kind of pessimistic
induction: who is to say that our current most-trusted theories will not turn
out to be only fictionally true? But I would prefer to say that they are true
within the recognized limits. This amounts to an optimist–realist answer to the
pessimistic induction: within the successful activities that we carry out, there is
truth that should not be negated by a later fictionalist account of the situation.

. Truth as a Quality

One key feature of the concept of truth-by-operational-coherence is that it
is not an absolutist notion. In my opinion this is both important and
beneficial, so some elaboration is called for. My definition of truth-by-
operational-coherence states that a proposition is true ‘to the extent’ that it
facilitates coherent activities. ‘To the extent’ there was a deliberately vague
designation for the sake of brevity, and now I must say more about what it
means. There are two main aspects, which I will address in turn. First, the
truth-as-operational-coherence of a proposition comes attached with a
specific and finite scope, given by the set of coherent activities that relies
on it. Secondly, the extent of truth depends on the number and variety of
the coherent activities, the degree to which each of those activities relies on
the proposition in question, and the degree of coherence of each activity. It


This is different from constructive empiricism (van Fraassen ), which regards statements
concerning unobservables as truth-apt while recommending against trying to find out their truth-
values.

8: /73 791
 Truth as a Quality 
will be futile to try to summarize all of those considerations into a single
measure. So truth-by-operational-coherence is not a matter of binary true/
false, not even a simple matter of degree, but a multi-faceted and contex-
tual quality.

The Finite Scope of Truth

In unpacking the sense in which truth-by-operational-coherence obtains ‘to
the extent’ that it facilitates operationally coherent activities, the first thing to
note is that truth has a finite scope. This is a very useful insight when we are
trying to make sense of the history of science (or of any other evolving field of
knowledge). We do often say that certain theories are only true within a certain
domain. The special theory of relativity is true only where there is no
gravitation. Ordinary quantum mechanics is true enough in the world of atoms
and molecules, but it is not designed to handle the interaction between
matter and light, or the collision of elementary particles at high energies. As
noted in Section ., there is a temptation to say that all theories are false
except the most cutting-edge one that we now believe, but that would be to
walk right into a trap, since we have to grant the possibility or even likelihood
that our current best theories will later be shown to be false, too. This makes
‘truth’ a basically unusable designation. The only sensible way to shield
ourselves from the pessimistic induction from the history of science is to grant
that all good theories are true, and remain true, within their own scope. And
with my notion of truth-by-operational-coherence, such liberal granting of
truth is not only possible but very natural. (This has a clear implication of
pluralism, which will be discussed further in Sections . and ..)
Now, the usual tendency among scientific realists is to assume that
later and better theories will be true in larger domains, and that this is a
cumulative process, so that the domain of a previous theory is completely
included in the domain of a later and better theory (recall Einstein and Infeld’s
image of the ever-expanding vista as one climbs a mountain, discussed in
Section .). If that is the correct picture of the history, then the limited scope
of the truth of a current theory becomes bearable in the optimistic hope that
we are approaching, step by step, ‘a final theory, one that would be of
unlimited validity and entirely satisfying in its completeness and consistency’
(Weinberg , p. ). But there are good reasons to be sceptical about this
picture (even if we set aside the problem of incommensurability, which I will
discuss in Section .).
We tend to have an overblown sense of how well later theories
account for the successes of earlier theories in science. For example, so often

8: /73 791
  Truth
we hear that quantum mechanics is true everywhere, while classical mechan-
ics only approximates it well enough in macroscopic domains. But we actually
have no direct empirical evidence that quantum mechanics is true in the
standard macroscopic domains where classical mechanics does its best. It is
considered a stunning achievement to demonstrate unequivocally quantum-
mechanical effects even in bodies just large enough to be above the threshold
of what we can see with a microscope, for example when two Bose–Einstein
condensates come together and make an interference pattern. Even in such
amazing cases, what we have is only a qualitative demonstration of a quantum
effect, not the confirmation of a precise prediction made from a quantum-
mechanical calculation. When it comes to typically macroscopic situations
concerning, say, the trajectory of a rocket ship, or a spinning top, it would be a
mind-boggling technical problem to figure out how to adapt the Schrödinger
equation to such objects. Now, if someone figured out how to do that, can we
presume that the quantum-mechanical calculation would generate empirically
correct results, and more correct than the result of the classical-mechanical
calculation? When I raise this question, the usual response I get is simple
incredulity: how can you question such a thing? It is a common article of faith
that quantum predictions will be truer than classical predictions everywhere
because quantum mechanics is a generally superior theory to classical
mechanics. But how is the assumption of such a uniform superiority justified?
(If you are a better person than I am, are you better at everything than I am?) As
for the elaborate accounts given on how quantum descriptions approach
classical ones in the macroscopic realm, ranging from Ehrenfest’s theorem to
decoherence, they only demonstrate certain sorts of convergence between
the two theories, but cannot in themselves say which theory has the empirical
edge where they differ. In my opinion, it is better to recognize that classical
mechanics is still true where it always has been, and that its truth does not
consist in a successful reduction to quantum mechanics. The same point could
be made even more strikingly if we considered the relation between classical
mechanics and quantum field theory or the standard model of elementary
particle physics.
Let us try to discern the scope of truth-by-operational-coherence
where it actually obtains, without getting distracted by the ‘dreams of a final
theory’. For another example, take what was perhaps the single most impor-
tant proposition in the history of organic structural chemistry in the nineteenth
century: ‘Carbon has valency ’, meaning that it is capable of bonding with


See Tomczyk () for an insightful historical and philosophical discussion of Bose–
Einstein condensates.

8: /73 791
 Truth as a Quality 
four other chemical units (atoms or radicals) at the same time. At least to begin
with, this proposition was a truth-by-operational-coherence: there was no way
to compare it directly with observations, and the theories of the internal
structure of atoms from which it could have been deduced were still many
decades away. But the highly productive activity of working out the molecular
structures of countless organic substances relied crucially on this tetravalency
of carbon. It was also relied upon in the understanding and planning of
countless chemical reactions, especially ‘substitutions’. For example, a body
of methane gas could be made to absorb a volume of chlorine gas and emit
an equal volume of hydrogen gas, turning the methane into chloromethane;
such a substitution could be made four successive times, in the end yielding
what we call carbon tetrachloride. All this makes perfect sense when we use
the tetravalency of carbon to understand methane as CH, chloromethane as
CHCl, and carbon tetrachloride as CCl (see Figure . in Section . again).
Such highly coherent activities in analytic and synthetic chemistry were what
constituted the truth of ‘Carbon has valency ’, and those activities still
maintain their coherence to this day. But this truth is a limited one. We know,
for example, that the structure of carbon monoxide remained a mystery for a
long time. Even carbon dioxide (CO) was not trivial to understand, but it could
be accommodated by saying that the carbon atom in it formed a double bond
with each of the two atoms of oxygen (which has valency ), thereby using up
all of its  bonding-potentials, as indicated by the graphic formula O=C=O. But
it was not clear at all how carbon monoxide (CO) could be understood until
quantum chemistry arrived with its sophisticated and entirely different
account of chemical bonds. And in the new theoretical regime the old
concept of ‘valency’ as such had no place (and the sticks that represent the
bonds in the old ball-and-stick models have no straightforward quantum-
mechanical equivalents). However, the nineteenth-century valency values are
still true, to the extent that they continue to enable a myriad of highly
coherent analytic and synthetic activities of chemistry.
The recognition that any given truth has a limited scope is also crucial
for dealing with what I will call the ‘objection from effective false beliefs‘
(essentially the same as what I called, less soberly, the ‘witchcraft objection‘
in Section .). The objection is quite simple: we know that false beliefs can
sometimes lead to success, so we should not equate truth with success (or
rather, with what leads to success). The same sort of objection can easily be
directed to my notion of truth: in a whole range of cases, including the reality
of witches and various attributes of God, it would seem that people can
engage in some operationally coherent activities by relying on false beliefs.
For example, on the basis of the assumption that the earth is flat, one can carry

8: /73 791
  Truth
out a lot of quite coherent activities; surely that doesn’t mean it is true that the
earth is flat? But we need to be careful about this kind of case. Coherent flat-
earth activities rely on the assumption of local, not global flatness, which is true
enough; there is no harm in allowing truth to that (see Teller , p. S,
footnote ). Now, global flatness can also be ‘held true come what may’, but
this becomes a classic degenerating research programme in the Lakatosian
sense. After all sorts of excuses are made and risky tests are declined, the
coherence of global flatness that remains is in a pitifully tiny domain excluding
images from satellites, the global positioning system, the existence of
Antarctica, etc. It is very instructive to read Lee McIntyre’s account () of
his efforts to engage flat-earth advocates in an empirical discourse.
When we say that mere convenience should not be mistaken for
truth, that is normally because ‘the truth will out’ (recall James’s image of
experience ‘boiling over’, cited in Section .). That is to say, we should not
assert that a statement is ‘true’ without qualification, if we expect that it might
be shown not to be true in some other circumstances. When we say ‘It may
seem as if P were true in these circumstances, but P is actually not true’, what
else can we be meaningfully asserting, other than that P will fail to be true in
some other circumstances? Those who invoke the objection from effective
false belief tend to be overly fearful of easy and undeserved coherence, and it
might be reassuring to be reminded of just how difficult it is to devise and
maintain coherent activities in real life. They also tend to be overly dogmatic in
denying truth to propositions at the basis of various coherent activities. I have
already argued that it is not wise to deny truth altogether to the best scientific
theories from the past. I see a disturbing pattern of thinking: one makes up
one’s mind, a priori, about the falsity of some doctrine; when there are
coherent activities relying upon it, instead of allowing that the doctrine might
be true in those circumstances, one concludes that facilitating operational
coherence cannot be what truth is about, retreating to the empty notion of
truth by correspondence to ultimate Reality instead.

Truth as a Quality
By saying that a proposition is true to the extent that there are coherent
activities relying on it, I also mean that its truth is a matter of degrees,
depending on how well it plays this role. Actually it is not quite right to say
‘matter of degrees’, because ‘degree’ has a connotation of a numerical quan-
tity, and ‘how well’ here is not easily quantifiable, at least not into a single
number. There are many factors that feed into truth-as-operational-coherence.
What is the number and variety of the coherent activities that rely on the

8: /73 791
 Truth as a Quality 
proposition in question? How strong is the reliance that each of those activities
has on the proposition? How coherent is each activity? And if we were to try to
get an overall measure of the degree of truth, we would have to also consider
how valuable each of those activities is, so they can be given proper weights,
and there would also be a need to balance out coherent and incoherent
activities attempted within the same domain. Truth-by-operational-coherence
would at best be a multi-dimensional quantity, or an index composed of
several quantities, but even such a move could be worked out convincingly
only if each dimension is properly quantifiable.
Instead of trying to construct an elaborate and unworkable quantita-
tive measure, I think it would be better to accept that truth-by-operational-
coherence is a quality, of which we can discern a sense of ‘more’ or ‘less’ only in
an imprecise way. Truth-by-operational-coherence should be seen as yet
another case in the class of qualities that allow comparisons but not in
unequivocal and fully quantitative ways. There are a number of such qualities.
With such qualities, sometimes there are clear more-and-less judgements, as in
the hardness of diamond versus talc. Some cases are much more equivocal:
glass is harder than iron according to the what-can-scratch-what criterion of
the Mohs conception of hardness, but it is clearly less hard than iron in the
sense of being more fragile. Some other qualities only allow comparisons in
extreme cases or in very specific contexts. We can confidently say Mother
Theresa had a higher degree of benevolence than Adolf Hitler, but it will be
difficult to be definitive about John F. Kennedy versus Franklin D. Roosevelt.
Truth-by-operational-coherence is akin to these latter cases. And perhaps it is
only appropriate that truth comes out looking like other major virtues in life,
such as beauty, justice and benevolence.
It may be useful here to consider some other attempts that have
been made to acknowledge that empirical truth is not a black-and-white
binary judgement. The idea of approximate truth is often invoked, especially
by scientific realists, but hardly ever precisely defined. In a similar direction
serious work has been done on the quantification of the truth-likeness
(verisimilitude) of theories, starting with Popper’s attempts and evolving into
versions like the one given by Niiniluoto (, sec. .), ultimately based on
the idea of counting up the number of true and false consequences of a
theory. But such quantification is workable only because it rests on the
assumption that the truth and falsity of the consequences of a theory are
themselves unequivocally defined. Another popular move in recent episte-
mology has been a resort to probability. But it is difficult to reduce truth-by-
operational-coherence to probability, because it is not a matter of well-defined
frequency or degree-of-belief. And even if we could somehow squash the

8: /73 791
  Truth
notion of truth-by-operational-coherence into one dimension, it does not sit
well with the kind of quantification as probability, plotted on a scale of  to .
Admitting truth to be a quality raises a very significant logical issue, to
which I cannot do anything like full justice here. With truth as a quality, what
happens to all the rules of logic based on the strict mutual exclusion of true/
false? As indicated in Section ., I think we should take this as a serious warning
that bivalent logic is not the ultimate guide to real life. It is an interesting
quandary to think what ‘false’ should mean when ‘true’ is a quality. My very
amateurish view, for the purpose of the philosophy of science, is that ‘true’ and
‘not true’ should still be taken to be mutually exclusive just to preserve the
meaning of ‘not’, though that isn’t all that useful when ‘true’ itself is a qualitative
notion. And ‘false’ could be given a more interesting meaning than ‘not true’.
I think there is much use for a notion of falsity which can be given to a
proposition that would, if relied upon, actively destroy the coherence of existing
activities. Under some circumstances we can work conveniently treating truth
and falsity as if they were simply mutually exclusive categories, but we should
not mistake that as the ultimate and universal rule of thinking.
As indicated by the logical questions just mentioned, I realize that the
idea of applying the term ‘truth’ to what is less than perfect and universal will
grate against many philosophers’ intuitions. To soothe the irritation, it will be
helpful to consider more carefully some of the numerous instances in which
we freely and confidently assign truth to statements that we know to be quite
likely to be imperfect and limited. In fact, I would argue that most cases of
truth-assignment in any real practices outside classical formal logic are of this
type. For the purpose of having a philosophy that is equipped to make sense
of real life, it is actually a good thing that our concept of truth does not
designate something perfect. ‘True enough’ may be the operative criterion in
many practical and scientific decisions. As James once put it: ‘Any idea upon
which we can ride, so to speak . . . is true for just so much, true in so far forth’
(James [] , p. ).

. Plurality and Incommensurability

If truth-as-operational-coherence is qualitative in the sense explained
in Section ., it becomes natural that there may be multiple true


I thank Dorothy Edgington for making me realize that I could not avoid this issue.

In the remainder of this passage James states that an idea is ‘true instrumentally’; this is a
fundamentally different attitude from what usually goes by ‘instrumentalism’ these days, which
amounts to denying truth-value to propositions regarded as instruments.

8: /73 791
 Plurality and Incommensurability 
theories in a given domain. This plurality is a separate issue from the
recognition that there are multiple concepts of truth. But doesn’t the
plurality of truth-as-operational-coherence lead to contradictions? I argue
that this is not a worry at least in most cases, because successful competing
theories generally employ sets of concepts that are sufficiently different
from each other, so that propositions in the competing theories do not
directly contradict each other. This is an expression of Kuhn’s doctrine of
semantic incommensurability. The recognition of incommensurability
leads to epistemic pluralism, to go with the metaphysical pluralism
expressed in Section .: it makes sense that scientists tend to maintain a
plurality of theories by conserving successful theories even after the emer-
gence of alternative theories that are superior in some respects.

The Plurality of Truth-by-Operational-Coherence

I expect that my views expressed so far will have raised a worry about
relativism in many readers’ minds. I do not think that relativism, in the sense
of recognizing the absence of absolutes, is to be feared so much, but at any
rate the implications of my views on truth are more pluralist than relativist
(Chang b). My perspective on truth is pluralist at two levels. First of all,
I think there are many valid and useful concepts of truth, as discussed in Section
.. At another level, under the concept of truth-by-operational-coherence,
which is my main focus here, we see that multiple truths can hold concerning
the same subject-area. It is this latter kind of plurality that I want to discuss
further now.
The first point I want to stress is that the plurality of truth-by-opera-
tional-coherence does not necessarily imply contradiction, and that it very
rarely does in practice. The plurality of theories is threatening only if we
assume that each theory is complete and perfect. As discussed in Section
., the scope of the actual use of theories is often narrower than imagined,
and they do not overlap as much as we might fear (or hope). Take, for
example, the business of wave–particle duality in quantum mechanics. Niels
Bohr recognized the truth and necessity of both wave-descriptions and


Instead of theories it might be better to think about ‘accounts’ as defined by Catherine Elgin (,
p. ): an account ‘consists of contentions about a topic and the reasons adduced to support them,
the ways they can be used to support other contentions, and higher-order commitments that specify
why and how evidence supports them. It also contains normative and methodological commitments
specifying the suitability of categories, the criteria of justificatory adequacy, and the ways to
establish that the criteria have been met.’

8: /73 791
  Truth
particle-descriptions of quanta, and generalized this insight into his doctrine of
complementarity (see Murdoch , chs. –). Even if we do not accept the
deeply metaphysical aspects of Bohr’s view, it must be agreed as a matter of
fact that electrons and other quanta exhibit particle-like behaviour under
certain experimental conditions (as in the photoelectric effect or the
Compton–Simon experiment), and wave-like behaviour under certain other
experimental conditions (as in the double-slit experiment). A vivid image of
this wave–particle duality was given by Bruce R. Wheaton (, epigraph),
quoting J. J. Thomson from : ‘It is like a struggle between a tiger and a
shark, each is supreme in his own element, but helpless in that of the other.’
Even though we are dealing with the same objects in one sense, we are
looking at very separate sets of phenomena that are exhibited in very different
experimental activities that we engage in. We could quite plausibly say that
the wave-like description is true in one domain of phenomena, and the
particle-like description in another, even though we can also discern the same
entities that are involved in both sets of phenomena.
But wouldn’t such breezy pluralism raise a difficulty with my own
notion of truth? The definition of truth-by-operational-coherence requires
that a true proposition should be relied upon in coherent activities. But if the
same job may be done in two different ways, neither is necessary. So how can
either be said to be relied upon? For example, if the same experiment can be
interpreted in two wholly different ways, wouldn’t that make all the state-
ments in either of the interpretations unnecessary? This is where the subtle
difference between reliance and a stronger notion of necessity comes in.
I take reliance as a matter of actual need within a given system of practice: so
we might say, for example, that having adopted the caloric theory of heat
and the explanatory practices that come with it, I have to rely on the
assumption of the self-repulsion of the caloric fluid in order to be able to
explain the phenomena relating to the pressure of gases (for example that it
increases with temperature). But if I do not work in the calorist system at all,
then of course my explanation of pressure does not rely on any assumptions
about caloric; if I am working with the kinetic theory of gases instead, then
the same phenomena can be explained by relying on the assumption that
gas molecules are bouncing around at random and that temperature is
proportional to the molecules’ kinetic energy. So both the assumptions of
the caloric theory and the assumptions of the kinetic theory have some truth-
by-operational-coherence. To take another example briefly: there are coher-
ent activities that I can perform using the phlogiston theory (which Priestley
and others did perform), and there are also coherent activities that you can
perform using Lavoisier’s oxygen theory. I am relying on the phlogiston

8: /73 791
 Plurality and Incommensurability 
theory, and you are relying on the oxygen theory; each theory is true to
that extent.

Incommensurability Revisited
The type of situation just discussed, in which we have very different theories
being used in the same domain, raises a different type of worry for my view on
truth as well. Won’t we end up with contradictions in such cases? The
phlogiston and the oxygen theories do say things that apparently conflict
with each other, for example whether water and metals are elements (the
phlogiston theory says that water is an element and metals are not, and the
oxygen theory says the opposite). Or consider the case of light again. Light is
taken as a collection of rays that obeys certain rules of reflection and refraction
(geometric optics), or as waves in the aether that can exhibit diffraction and
interference (the original wave optics), or as patterns of propagation of inter-
twined electric and magnetic fields (modern Maxwellian electrodynamics), or
as a bunch of photons each carrying a little bundle of energy (Einstein’s
quantum theory of light), or as photons that mediate electromagnetic forces
between charged particles (Feynman’s quantum electrodynamics), or excited
states of a quantized field (quantum field theory). All of these conceptions are
at least in principle applicable to every situation where light is present, and
each conception embodies a set of truths. While some combinations of these
conceptions are quite harmonious, some other combinations are not. So how
can we avoid contradictions, if we grant truth freely and pluralistically as I am
proposing?
I would argue that in most such cases in actual science there is still no
contradiction. In order to unpack and justify that claim, I need to enter into a
discussion of incommensurability. There is clearly a kind of dissonance, a lack
of agreement, between saying that light is a continuous wave-like distribution
of electric and magnetic fields, and saying that light is a collection of discrete
photons each carrying a definite amount of energy. But no physicist would
deny that discrete photons pass through the slits in the double-slit experiment,
or that there is an electromagnetic wave hitting the metallic surface in the
photoelectric effect. In situations like this, we have to admit that we have two
parallel descriptions, which are related to each other (for example, the energy
of a photon is proportional to its frequency, a notion rooted in the wave-
description) yet not fully translatable to each other. This is a clear case of
semantic incommensurability as discussed by Kuhn.
Kuhnian semantic incommensurability does not imply contradiction.
This issue is complex and deserves further consideration (see Hoyningen-Huene

8: /73 791
  Truth
and Sankey  for a thorough discussion). To continue with the example of
light: at the general level of description, the idea of photons as discrete
packets of energy does seem to contradict the idea of an electromagnetic
wave which carries energy in an entirely continuous manner. But when it
comes to the level of specific statements, the contradiction often evaporates
into a mist of untranslatability. For example, how do we render into wave-
theoretic language the statement ‘One of five photons, each of energy E,
travelling along the x-axis, was absorbed by an atom at location L and thereby
put the atom into an excited state’? We can try to depict a plane wave
propagating in a definite direction, but in that case the wave theory has no
terminology for expressing the discreteness of photons, or localizing the wave
to the location L. The statement in question cannot be faithfully rendered into
wave-theoretic terms, which means that we cannot even ask what the wave
theory would say about that situation, to see if it would contradict what the
particle theory says.
Therefore, if we take the early twentieth-century wave- and particle-
descriptions of light and ask which is true, the only sensible answer is that they
are each true, to an extent. Which one is truer would be difficult to say, as they
do better in different situations (recall ‘a tiger and a shark’). How should we
choose between the two, then? That was the very controversial question of
theory-choice (or paradigm-choice) that Kuhn’s work raised. The pluralist
answer is that we do not need to choose between the alternatives in a way
that preserves only one and kills off the other. We keep both descriptions, and
use one or the other or both, wherever each can facilitate coherent activities.
By facilitating coherent activities each description acquires truth. The actual
history of physics bears out this possibility; quantum physicists learned to live
with wave–particle duality, even if they did not all follow Bohr on the general
doctrine of complementarity. In earlier optics, too, the particle theory and the
wave theory co-existed and each demonstrated some truth. Many scientists
saw this situation as one of mere ignorance and indecision, and expected the
debate to be resolved in favour of one theory or the other, but the subsequent
history did not bear out that monist expectation. Think back to the inspiration
that Kuhn took from gestalt psychology. Is the duck–rabbit really a duck, or a
rabbit? Does the Necker cube drawing really depict the cube one way, or the
other? Both interpretations are correct, depending on how we are looking at it.
There may be contextual factors which make one more apt than the other (for
example if the duck–rabbit figure is part of a depiction of a duck farm), but
there is no sense in which one can be absolutely true and the other one false.
Less metaphorically, when there are mutually incommensurable descriptions,
the truth of each depends on how coherent its uses are, but not on how
coherent the uses of other descriptions are.

8: /73 791
 Rehabilitating the Pragmatists 

. Rehabilitating the Pragmatists

Having laid out my own ideas about the nature and functions of the
concept(s) of truth, I now want to make some reflections on the so-called
pragmatic theory of truth. It is important to get this aspect of pragmatism
right, because truth is the issue on which classical pragmatism was most
seriously misunderstood and attacked. The classical pragmatists’ views on
truth should not be caricatured as a notion that whatever pleases the
believer is true. In particular, I argue that William James’s controversial
notion of truth can be rehabilitated through my notion of truth-by-
operational-coherence. I will also try to dispel some common misunder-
standings of the pragmatic theory of truth. To start with, it is important to
note that pragmatism does not equate truth and utility. More subtly,
according to my interpretation what the pragmatic theory offers is a shift
in how we conceive the very meaning of truth, not merely a new set of
criteria by which truth can be judged.

What Is the Pragmatist Theory of Truth?

In this section I want to make a defence and productive reinterpretation of the
classical pragmatists’ views on truth. The account of truth was a central part of
pragmatism, especially as James presented it, and it was also where pragmatism
drew the fiercest criticism. The classical pragmatist notions of truth are not far
from my notion of truth-by-operational-coherence, which is no surprise since my
thinking has been strongly inspired by the pragmatists. I cannot enter into an in-
depth discussion of the classical pragmatists’ views here, but given the impor-
tance of pragmatism in the general orientation of this book, it would be remiss of
me not to comment on how I understand the chief pragmatists’ views on truth.
I also believe that looking back on the classical pragmatists through the lens of
my own ideas can suggest a productive reinterpretation of their thoughts.
There are two main streams of thought on truth within the pragma-
tist tradition. One is due to Peirce, who led the common pragmatist aversion to
the typical correspondence theory or any other ‘transcendental’ accounts of
truth referring to inaccessible metaphysical realms, which made truth ‘a useless
word’ (Peirce , §.). But what exactly was the pragmatic ‘upshot’
(Peirce , §.) of the concept of truth? Wanting to preserve a sense of
objectivity in the idea of truth while not falling back into correspondence
realism, Peirce famously opted for the following formulation: ‘The opinion
which is fated to be ultimately agreed to by all who investigate, is what we
mean by the true, and the object represented in that opinion is the real’ (Peirce
, W.). Similarly, in a later formulation: ‘Truth is that concordance of an

8: /73 791
  Truth
abstract statement with the ideal limit towards which endless investigation
would tend to bring scientific belief’ (Peirce , §.). So, when all is said
and done, at the end of all inquiry, a community of sincere and able inquirers
will be left with a set of truths.
There is an obvious futility to the idea of ‘the end of inquiry’ for
anyone wanting to stay true to the basic spirit of pragmatism. The ultimate
end of inquiry that Peirce envisages here is quite as inaccessible as the
metaphysical realist’s ‘external world’ or ‘Reality’. So Peirce’s notion of truth is
just as useless as the correspondence notion of truth, and detracts from
Peirce’s own spirit of pragmatism. I am inclined to agree with Cheryl Misak’s
argument (b) that the real core of Peirce’s notion of truth was not the
idea of convergence or the fate of knowledge in the long run, but the idea of
‘indefeasible belief’. Indefeasibility means that the idea under consideration
‘would not be improved upon; or would never lead to disappointment; or
would forever meet the challenges of reasons, argument, and evidence’ (Misak
, p. ; see also Misak b, p. ). This is surely correct as a reading of
Peirce’s later works, but it still does not solve the problem that worries me.
Indefeasibility, in the sense of being able to meet certain challenges ‘forever’, is
not something we can evaluate in the present at all, any more than we can
foresee the outcomes of inquiry at the ultimate end of the process.
Thus setting aside Peirce’s view for the moment, I now come to the
other main stream of the pragmatist theory of truth, which is the more contro-
versial view usually attributed to James. This view requires a careful exposition
and defence. James himself ([] , p. ) called it the ‘Schiller–Dewey
view of truth’, but I will focus on James’s presentation here, not only because it is
the best-known version but because he (unlike Dewey) considered the elucida-
tion of the truth concept such a central issue, viewing it as one of the two key
components of pragmatism, the other being ‘the pragmatic method’ (ibid.,
p. ). He gave significant attention to the concept of truth in his flagship text
Pragmatism published in , and followed that up two years later with The
Meaning of Truth, billed as a sequel to Pragmatism. James starts by seemingly
endorsing a common-sense correspondence view (ibid., pp. –):
Truth, as any dictionary will tell you, is a property of certain of our ideas. It means
their ‘agreement,’ as falsity means their disagreement, with ‘reality.’ Pragmatists
and intellectualists both accept this definition as a matter of course.

But, James continues, they ‘begin to quarrel’ when ‘the question is raised as to
what may precisely be meant by the term “agreement,” and what by the term
“reality,” when reality is taken as something for our ideas to agree with’. James
rejects the idea that our ideas should ‘copy’ reality. Instead, ‘agreement with
reality’ is spelled out in terms of ‘the truth’s cash-value in experiential terms’.

8: /73 791
 Rehabilitating the Pragmatists 
This is just pragmatism asking ‘its usual question’: ‘Grant an idea or belief to be
true’ – ‘what concrete difference will its being true make in anyone’s actual
life?’ James concludes:

True ideas are those that we can assimilate, validate, corroborate and verify. False
ideas are those that we cannot. That is the practical difference it makes to us to
have true ideas; that, therefore, is the meaning of truth, for it is all that truth is
known-as. (ibid., p. ; emphasis original)

James makes various attempts to express what is involved in the

pragmatic validation of ideas, and there are three important points. First, truth
is a matter of connection between experiences. James stresses that the appar-
ent sense of ‘agreement’ exhibited in a true idea is to be found in productive
connections between different experiences that it enables us to form: ‘ideas
(which themselves are but parts of our experience) become true just in so far as they
help us to get into satisfactory relation with other parts of our experience, to
summarize them and get about among them by conceptual short-cuts’ (ibid.,
p. ; emphasis original). James says that the main function of true thoughts is
their ‘go-between function’ (ibid., p. ). Second, a truth needs to incorporate
previously known truths: ‘our theory must mediate between all previous truths
and certain new experiences’ (ibid., p. ). And ‘a new opinion counts as “true”
just in proportion as it gratifies the individual’s desire to assimilate the novel in
his experience to his beliefs in stock. It must both lean on old truth and grasp
new fact’ (ibid., p. ). Third, truth has a ‘leading’ or ‘guiding’ function:

To ‘agree’ in the widest sense with a reality, can only mean to be guided either
straight up to it or into its surroundings, or to be put into such working touch with it
as to handle either it or something connected with it better than if we disagreed.
Better either intellectually or practically! (ibid., p. ; emphasis original)

The sense of being ‘guided to’ a reality is obscure. Being put into ‘working
touch with it’ is clearer, as James explains further:

Any idea that helps us to deal, whether practically or intellectually, with either the
reality or its belongings, that doesn’t entangle our progress in frustrations, that
fits, in fact, and adapts our life to the reality’s whole setting, will agree sufficiently
to meet the requirement. It will hold true of that reality. (Ibid.; emphases original)


Recognizing this point, we can also make better sense of some mysterious statement such as this:
‘Truths emerge from facts; but they dip forward into facts again and add to them; which facts again
create or reveal new truth (the word is indifferent) and so on indefinitely’ (James [] ,
p. ).

8: /73 791
  Truth
Pragmatist Truth as Truth-by-Operational-Coherence
Truth-by-operational-coherence is something achievable and verifiable in
practice to various degrees, and its pursuit is clearly useful. So it is certainly a
notion of truth fit for pragmatist philosophy. And I believe that reinterpretation
in terms of my ideas can give added clarity and plausibility to the classic
pragmatist views, especially to James’s view. Let me note some instructive
points of comparison and contrast.
James’s expressions can be blurry, sometimes verging on the poetic.
This is the case, too, when he tries to explain what the practical bearings of
truth are. I developed the notion of operational coherence precisely for the
purpose of making James’s kind of intuition more precise and systematic.
James holds that pragmatism ‘converts the absolutely empty notion of a static
relation of “correspondence” . . . between our minds and reality, into that of a
rich and active commerce (that anyone may follow in detail and understand)
between particular thoughts of ours, and the great universe of other experi-
ences in which they play their parts and have their uses’ (ibid., p. ). James’s
idea of ‘expediency’ embodies the coercion of reality on thought, as explained
in Section ., and operational coherence does the same job. Both notions
should be taken in a strong empiricist vein. (I think Dewey’s notion of ‘war-
ranted assertibility’ can also plausibly be incorporated into my notion of truth-
by-operational-coherence.)
Recognizing truth as a quality rather than a black-and-white binary
judgement, as I have explained in Section ., can help shut down the ridicule
that critics have heaped on the pragmatist attribution of truth to imperfect
ideas. James’s irritated response is understandable (‘expedient in the long run,
of course’), and so is the temptation felt by Peirce to reserve the designation of
truth to the ‘indefeasible’. But with my notion of truth we do not need to wait
for the long run, and we can certainly refrain from Peirce’s self-defeating move
of appealing to the ‘end of inquiry’. James easily recognized the qualitative and
multi-dimensional nature of truth. Defining truth in terms of operational
coherence also helps the pragmatist by taking truth one step away from direct
verification in terms of success. This may be unsatisfying at first glance, but it
does make it possible not to grant truth to accidental successes and failures
due to variations of fringe circumstances.
I also think that an explicit adoption of pluralism, at both levels, can
enrich and ease the pragmatist view on truth. First, adopting pluralism con-
cerning the concept of truth (Section .) liberates the pragmatist theory of
truth from having to account for all uses of ‘truth’ by means of one idea.
Identifying pragmatist truth with truth-by-operational coherence does not

8: /73 791
 Rehabilitating the Pragmatists 
mean giving the same treatment to other kinds of truth. I think James ([]
, p. ) went too far when he stated that the pragmatist account of truth
was about ‘what truth everywhere signifies’ (if that meant unsuspecting
participation in truth monism). Secondly, I think that losing a monist presump-
tion concerning truth-by-operational-coherence itself (Section .) can help
the pragmatist view reach its full potential. I find Christopher Hookway’s
interpretation of Peirce instructive. Hookway (, p. ) argues that
Peirce’s early convergentist view of truth was formulated so as to prop up a
monist view of mind-independent reality (defining reality as the object of a
true proposition), and that this view of truth was no longer necessary when
Peirce’s notion of reality moved on to an idea of direct perception. Hookway
(ibid., p. ) goes on to give a broader reading of what Peirce meant by any
investigator being ‘fated’ to arrive at the truth: ‘If a proposition is true, then
anyone who investigates some question to which that proposition provides
the answer is fated to believe it.’ Hookway notes that this is ‘compatible with
rejection of the absolute conception of reality, for it is compatible with the
view that our different perspectives are reflected in the varying ranges of
questions that we can understand or take seriously’. Whether or not Hookway’s
view is the best interpretation of Peirce, I think it is a very plausible view on
truth in its own right, consonant with the general spirit of pragmatism and
very compatible with the plurality of truth-by-operational-coherence.

Beyond Misunderstanding
Having gone through the above presentation and interpretation, I think we
can fully appreciate James’s complaint about how the pragmatist view of truth
‘suffered a hailstorm of contempt and ridicule’ (James [] , p. ), ‘so
ferociously attacked by rationalistic philosophers, and so abominably misun-
derstood’ (ibid., p. ). He issued this complaint already as part of his best-
known exposition of pragmatism in  in defence of Dewey and Schiller as
well as himself, but his attempts at clarification apparently had no effect in
preventing the same kind of misunderstanding from recurring.
One factor must be noted, and set aside. Much vitriol directed against
James’s view on truth was excited by his willingness to entertain the possibility
that religious belief could be justified as a pragmatic truth, if it leads one to live
a good life. Russell (, p. ) made no secret of the fact that this was his
main objection to the pragmatist view of truth: ‘It is chiefly in regard to religion
that the pragmatist use of “truth” seems to me misleading.’ For some philos-
ophers this must have been the ultimate sign that there was something rotten
in the pragmatist notion of truth, because they knew that God did not exist.

8: /73 791
  Truth
But as James was keen to emphasize, atheist dogmatism is just as unwarranted
as theist dogmatism. In terms of my own notion of truth-by-operational-
coherence, I admit that many coherent activities are carried out by many
people on the basis of their belief in God, so I would be happy to accept that
‘God exists’ is true, to that extent. But such qualified truth would probably be
rejected by the theists themselves: no, if God exists, he does so absolutely.
Then we are asking if every activity in life can be carried out coherently by
relying of the existence of God, and I think that is not the case – certainly not if
we are talking about a bearded white man who lives in a place up there called
heaven and listens to everyone’s prayers, and even if we are talking more
sensibly about an ultimate being who is omnipotent and supremely benevo-
lent yet somehow allows all sorts of evil to take place in human life. But if it
were the case (and it is a big ‘if’) that some societies could work out a truly all-
encompassing and coherent way of life based on a well-crafted notion of a
deity, then who are we to presume that such a God doesn’t exist? Why would
belief in God in that case be any less credible than the modern atheists’
widespread belief in a rather extreme sort of materialism?
John Capps (, sec. ) gives a convenient list of standard objec-
tions that have been raised against pragmatist theories of truth. I will examine
each of objection and point out how I think it is based on a misunderstanding.
First, Capps points out: ‘if the pragmatic theory of truth equates truth with
utility, this definition is (obviously!) refuted by the existence of useful but false
beliefs, on the one hand, and by the existence of true but useless beliefs on the
other’. Paul Horwich (b, p. ) also takes the pragmatist to offer a straight-
forward definition of truth in terms of utility: ‘Here truth is utility; true assump-
tions are those that work best – those which provoke actions with desirable
results.’ And some have even taken utility as happiness or psychological
satisfaction, which is evidently not the main thing that was meant by the
pragmatists. On this point I think it is enough to hear the lament from James
([] , p. ): ‘A favorite formula for describing Mr. Schiller’s doctrines
and mine is that we are persons who think that by saying whatever you find it
pleasant to say and calling it truth you fulfil every pragmatistic [sic] require-
ment.’ Even worse is the interpretation of utility as material benefit. I am
perhaps not alone in having been confronted by otherwise serious and well-
informed scholars denouncing pragmatism as an American obsession with
money-making, by taking James’s talk of ‘cash-value’ (and what ‘pays’) quite
literally!
If the first objection chides pragmatism for giving a wrong definition
of truth, the second one charges it with confusion, of mistaking epistemic
criteria for truth as the definition of truth. Here is how Capps expresses the

8: /73 791
 Rehabilitating the Pragmatists 
objection: ‘utility, long-term durability, and assertibility (etc.) should be viewed
not as definitions but rather as criteria of truth.’ This objection arises from not
understanding the advantages of the pragmatists’ semantic proposal concern-
ing the meaning of ‘truth’, which I have tried to explain in Sections . and ..
It was for a reason that James gave the title of The Meaning of Truth to his
sequel to Pragmatism, with an emphasis on ‘meaning’. Russell did understand
the pragmatists’ semantic move, but gave a distorted view of it (, p. ;
emphases original): ‘The arguments of pragmatists are almost wholly directed
to proving that utility is a criterion; that utility is the meaning of truth is then
supposed to follow.’ In my reading, at least James tackled the meaning of
truth directly, rather than confusing criteria and definitions.
Thirdly on Capps’s list, it is alleged that ‘assessing the usefulness (etc.)
of a belief is no more clear-cut than assessing its truth’, which means that
pragmatism fails to turn truth into a concept any more accessible and useful
than the correspondence notion. Russell says: ‘it is so often harder to deter-
mine whether a belief is useful than whether it is true’ (Russell , p. ; see
also p. ). This objection draws a false equivalence between the in-principle
inaccessiblity of the correspondence between idea and world, and the
detailed vagueness and multi-dimensionality of very tangible successes, which
I discussed in Section .. The messiness of determining what is useful is only a
problem if one has a monist notion of usefulness.
And finally, there is what Capps calls ‘the fundamental objection’:
‘pragmatic theories of truth are anti-realist and, as such, violate basic intuitions
about the nature and meaning of truth’. It is difficult to know where to start in
response to this. First of all, the objection is question-begging, as it presumes
to know what the right intuitions about truth are. But there is also a misun-
derstanding at a deeper level, as it ignores the realist dimension of pragma-
tism, in terms of how pragmatism demands that our ideas answer to
experience, and to realities in the sense I defined them in Chapter . But in
order to explain that point fully, I need to give a full exposition of pragmatist
realism, which will be done in Chapter .


And Russell resorts to ridicule instead of analysis: ‘According to the pragmatists, to say “it is true
that other people exist” means “it is useful to believe that other people exist”’ (Russell , p. ;
emphasis original).

8: /73 791
  

Realism

. Overview

Why Do I Claim to Be a Realist?

I promised to offer a ‘realism for realistic people’ in this book, and I am
now ready to spell out what this involves, having re-examined the nature of
knowledge, reality and truth. What is realism, and what do I mean by
making it realistic? Let us start by taking a step back: why should we care at
all about something that might usefully be called ‘realism’, concerning
science and various other practices in life? It is because we want to respect
facts, not fantasies. Because we want to be open to learning from experi-
ence, instead of hiding behind the comforts of fixed opinions. Because we
appreciate that empirical knowledge can help us live better in the material
world. And because we want to live with the optimism that we can learn
some truths about realities, rather than resign ourselves to living without
improving our current state of ignorance.
Many people who have seen previous versions of my thoughts expressed
in this book have cast doubt on the wisdom of calling my position ‘realism’
at all. There are different reasons for this. Some say that my position does
not deserve to be called ‘realism’, because what I am calling ‘reality’ and
‘truth’ are too bound up with our human activities and concepts to be
objects of anything called realism. But I think that is an unnecessary worry.
As I emphasized in Section ., entities being mind-framed does not imply
that they are mind-controlled; there is no reason to be afraid to call them
realities. Even though real entities are concept-bound, they do not obey
our wishes. We can design a concept as we wish, but whether our concept
can facilitate coherent activities is a matter that is quite outside our control.
So I think my position does retain something very important in what
many people value in realism. And my position is also clearly different
from positivist or instrumentalist anti-realism, since I do think that



8: /73 791
 Overview 
theoretical propositions can be truth-apt. It is also different from van
Fraassen’s constructive empiricism, since I think that science can often
learn truths about unobservable entities.
Identifying my position as realism involves a reconceptualization of the
relation between realism and empiricism. In the philosophy of science the
two are often perceived as opposing doctrines, and I think that this is a
significant mistake. Realism should not mean anything contradictory to
learning from experience, which is the essence of empiricism. I suspect that
the perceived opposition between realism and empiricism is a historical
accident. It is a legacy of the philosophical landscape of the early twentieth
century, in which logical positivism was the standard-bearer of empiricism.
The positivists on the whole shunned realism (with Neurath as a notable
exception), as they saw it as part of the metaphysical tradition that they
were trying to purge from philosophy. But as I have tried to show, realism
concerning science and other practices does not need to rest on metaphys-
ical realism. And empiricism is very much in line with the spirit of being
realistic: if we stick with learning from experience, then we will not be
getting into overblown claims or unattainable dreams of knowledge.
Some friendlier critics have advised against calling my position ‘realism’
for a different reason: the term has already been spoiled, as it were, with
common meanings attached to it that I should not want. Better to leave
this word behind, they say, and focus on articulating my ideas under some
other name. However, there are strong motivations for keeping the ‘real-
ism’ label. The same goes for the terms ‘reality’ and ‘truth’, which I have
also proposed reconceptualizing in a pragmatist manner rather than setting
aside altogether. Such powerful words should not be left to people who
will use them in unhelpful ways. To make an extreme parallel: if those in
power commit atrocities in the name of ‘law and order’, it is not the right
response for the rest of us to retreat and stop insisting on the rule of law
and an orderly society. The right response is to reclaim the terms, and use
them to designate actions and situations that deserve to be described by
those honorific words. We say this is order, and that is not the rule of law.
Thankfully the current situation in philosophy is not so politically and
morally charged, yet the same methodological point applies here, too. It
would be a grave mistake to say that we don’t care about truth, knowledge,
reality and realism, just because these terms have been misused by others.
So a certain amount of wrangling about what the important terms should
mean is necessary. Recall how Peirce achieved nothing by trying to
distinguish himself from other pragmatists by calling his own view ‘prag-
maticism’, a label that no one took up (and he did not even expect people

8: /73 791
  Realism
to take up). Dewey’s move to talk about ‘warranted assertibility’ instead of
truth was similarly unwise.

Being Realistic about Realism

It is an important part of my pragmatist stance that I think philosophy
should play a role in guiding action. So I suggest that realism should be re-
conceived as a doctrine that we can actually put into practice, as an outlook
on life that can guide the development of empirical knowledge in all walks
of life. Being a realist or not should make some difference to what people
do. Here we should heed Arthur Fine’s (, p. ) observation that
everyone lives and works on the basis of what he calls the ‘natural
ontological attitude (NOA)’, and philosophical realism layered on top of
that only amounts to adding a ‘desk-thumping, foot-stamping shout of
“really!”’ about what is already accepted. This is why we should set aside
metaphysical realism (see Section .): not because metaphysics is entirely
meaningless as the positivists maintained, but because it doesn’t change
anything in practice, while disputes about it take attention away from
more important issues. To the extent that ‘scientific realism’ is based on
metaphysical realism, it also runs the same risk of futility. As I will discuss
further in Section ., I follow in the footsteps of various philosophers who
have attempted to articulate productive versions of realism in a pragmatist
mode, including Torretti, Vihalemm, Pihlström and Kitcher.
Now, if realism is to provide a guide to action, then it should be realistic,
in the normal everyday sense of that word: ‘having or showing a sensible
and practical idea of what can be achieved or expected’ (Google
Dictionary, first definition). This also expresses a down-to-earth sense of
rationality: to pursue aims that we can have some hope of achieving, at
least aims that we can meaningfully work towards. Realism should be
about the kind of knowledge that we can actually try to attain. As discussed
in Chapter , standard scientific realism concerns itself with a kind of truth
that we cannot meaningfully pursue, since we do not even know how to
tell whether we are getting closer to truth as conceived in the traditional
form of the correspondence theory. In contrast, realism as I intend it is not
the search for an ultimate kind of truth. Rather, it is the pursuit of
operational kinds of truth – namely, truth-by-operational-coherence, and
secondary truth based on that (as explained in Chapter ). These are the
kinds of truth we can attain in the here-and-now; we can work on
improving the truths we have, and we can clearly tell when we are making
progress in that task. The same spirit holds for my notion of reality

8: /73 791
 Overview 
explained in Chapter . The term ‘reality’ should be reserved for things
that we can meaningfully interact with, not for some inaccessible realm of
Being that we only entertain in our abstract thought. My concepts make
truth about real entities a very realistic aim to achieve, dissolving a central
difficulty concerning standard scientific realism.
Several philosophers of science have noted the irony that many who call
themselves ‘realists’ (subscribing to standard scientific realism) are not very
realistic in their realism. Either they are wildly optimistic about what
science has achieved or can achieve, or they talk at an incredible level of
abstraction without any sense of what happens in real scientific research. In
the Introduction I have already mentioned Peter Kosso’s call for rendering
realism more realistic, but he is by no means alone. The  workshop
on practical realism in Tartu was subtitled ‘Towards a Realistic Account of
Science’ (Lõhkivi and Vihalemm , p. ). Kerry McKenzie ()
advocates ‘being realistic’ in the realism debate, seeking a kind of realist
attitude that we can apply to science here and now. Kitcher’s (, ch. )
‘real realism’ also seeks to ground realist reasoning in the knowledge of
actually accessible objects. Teller’s () argument against the ‘perfect
model model’ of science is also very much in the spirit of philosophically
engaging with scientific knowledge in a form that humans can realistically
pursue and attain. Putnam concludes his ‘Defense of Internal Realism’
thus: ‘It is my view that reviving and revitalizing the realistic spirit is the
important task for a philosopher at this time’ (Putnam c, p. ). This
realistic spirit has been traced back to Wittgenstein by Cora Diamond
() and to Ramsey by Cheryl Misak (, pp. xxvi, ).
There are two important ways in which the realistic spirit manifests
itself in the context of debates on realism. First, it is often expressed as a
kind of ‘internalist’ stance, which amounts to a commitment to deal with
knowledge in terms of what is accessible to the knower. I think internalism
is productive, though I would want to avoid the exclusive focus on
conscious elements of an individual mind that is usually found in intern-
alism in epistemology and the philosophy of mind. The internalist point
I want to stress is that realism should offer operational guides to actual
practice, rather than putting up inaccessible ‘external’ criteria. In this
context Putnam’s ‘internal realism’ is a very important cognate position


Uskali Mäki () advocates ‘realistic realism’, but with a rather different intention, namely that of
curbing an excessive demand for literal truth in models in economics.

It is not entirely clear exactly what Putnam meant by this, but David Macarthur () takes it as a
matter of respecting common sense, which remained an important part of Putnam’s
philosophical outlook.

8: /73 791
  Realism
to mine, as I will discuss further in Section .. There I will also relate
Putnam’s position to a long tradition of seeing knowledge as situated in the
knower, reaching ultimately back to Kant and currently very much alive in
perspectivism as advocated by Ronald Giere, Michela Massimi and others.
The other manifestation of the realistic spirit is progress in an iterative
form. This is because any inquiry, to be realistic, needs to start from a
situation that we inherit from our personal and collective pasts. And our
starting point is hardly ever going to be anything like a firm foundation –
recall Peirce and Dewey pointing out that inquiry begins with a state of
disorder. But if we don’t have a firm foundation, where can knowledge
rest? (One answer, of course, is that knowledge does not rest – or certainly
inquiry does not.) As we try to forge ahead in inquiry, all we can do is to
accept something in our inheritance as given, and learn what we can learn
on that basis. And if we are fortunate enough, we will be able to use what
we learn in order to come back to our starting point and improve it. I have
used the term epistemic iteration to designate such a process of ‘getting
on’ in the absence of indubitable foundations. In epistemic iteration
successive stages of knowledge are created, each building on the preceding
one, in order to enhance the achievement of certain epistemic goals. In
each step, the later stage is based on the earlier stage but cannot be deduced
from it; later developments often correct and refine the presuppositions
made in the earlier stage. Initially I crafted this notion in the context of
trying to solve the puzzle of how measurement standards can be estab-
lished in the absence of pre-existing standards against which they might be
validated (Chang , ch. , and also pp. –). By now I am quite
confident that epistemic iteration is a very general feature of empirical
inquiry in many different types of situation (see Chang , a,
a). I will give a more in-depth consideration of the iterative nature
of realistic scientific progress in Section ..

Activist Realism
Realism as I take it is not a descriptive thesis, but a commitment to an
ideal. In ordinary usage an ‘ism’ might be a political ideology (as in
‘nationalism’, ‘conservatism’ or ‘environmentalism’) or an artistic move-
ment (as in ‘cubism’ or ‘impressionism’). It ought to be similar in
philosophy: for example, positivism as practised by the Vienna Circle
was certainly a committed stance, as is reductionism in the hands of

More broadly it can designate a discernible stance or tendency of any kind: ‘nihilism’, ‘nudism’,
‘egoism’, ‘narcissism’ or even ‘rheumatism’.

8: /73 791
 Overview 
practising scientists. The core of realism as I see it is an activist ideal of
inquiry: a commitment to seek more and better knowledge about realities,
along with a commitment to improve our epistemic practices to that end.
The overall objective of realism ought to be the improvement of knowl-
edge in its extent and quality, not only propositional knowledge but all
aspects of active knowledge as characterized in Chapter . The realism
I defend involves not only a recognition of realities, but a strenuous
commitment to promote inquiry and facilitate learning. Realism concern-
ing science should be about how we should pursue scientific knowledge.
In previous work (Chang a, p. ) I called this ideal of inquiry
‘active (scientific) realism’: science should strive to maximize our contact
with realities and our learning about them. To avoid conflation with the
terminology of ‘active’ knowledge adopted in this book, I now propose to
rename my brand of realism slightly, as activist realism. The ‘activist’
label accentuates a normative outlook, focused on an imperative of
progress. In this regard, my view on realism has not changed much from
what was expressed a decade ago (ibid., ch. , esp. pp. –), though
I hope it has become better founded now, with a more detailed articulation
of key concepts.
Activist realism is a commitment to do whatever we can in order to
extend and enhance our knowledge concerning realities, as much as
possible in the context of other aims and values. It is not an attitude of
sitting content in appreciation of the knowledge that we have already
attained, or condemning people who do not accept and use our knowl-
edge. Activist realism means not only accepting the verdict of empirical
tests of hypotheses, but also devising more and better tests, and producing
more hypotheses to be tested. Activist realism also dictates that we should
ask entirely new questions, make new theories, and even create more real
entities and learn about them.
Like all good ideals, the activist-realist ideal I am advocating here may
acquire an obvious (‘duh!’) quality to it, once it is seriously entertained:
who would not want more knowledge about more realities? But it is not a
trivial assumption that knowledge is a good thing; I would certainly not
want to argue that knowledge is an absolute good that trumps all others,
which is why the statement of activist realism only says that we should


‘Scientific realism’ would have been a good phrase for designating my position, of course, except that
the phrase was taken long ago by people who mean something quite different. See Section ..

It is possible to think about active knowledge (knowledge-as-ability) without much interest in
increasing it, in which case one would be non-activist about active knowledge.

8: /73 791
  Realism
pursue knowledge as much as possible in the context of other aims and
values. With that qualification, however, I do think that activist realism is
an ideal that should have a very broad appeal. Even those who identify as
anti-realists in relation to standard scientific realism should be able to join
in with activist realism; there are hardly any anti-realists who would argue
against learning what we can actually learn. Rather, their objection to
standard scientific realism is that it attributes to science the kind of
knowledge that we cannot reasonably claim to be able to attain. In
Section . I will show how activist realism relates to various positions
in the debate on scientific realism, and hopefully prompt a productive
realignment of positions in the realism debate.
Still, one might doubt that the articulation of activist realism would
make any difference to scientific practice – aren’t scientists already
doing their utmost to increase knowledge? Shouldn’t we simply leave
it up to the scientists, to do what they do so well? What is the role of
philosophical thinking in relation to this realism concerning science?
But scientists do not always behave in a way that is maximally condu-
cive to learning. Other thoughtful people can suggest, respectfully yet
critically, how science might be done even better, while acknowledging
that it is already done very well indeed. If philosophers of science do not
perform this function, who will? I find great inspiration from a neigh-
bouring discipline: Paul Forman (, p. ) urges historians of
science to embrace ‘the obligation to decide for ourselves what is the
good of science, and by our historical research and writing to advance
that good’. The same goes for the function of the philosophy of science
even more urgently.
Scientists often refuse to accept empirical evidence, or they explain it
away by means of ad hoc hypotheses, while shutting down questions and
conceptual schemes that offer alternatives to the currently dominant one.
Such practices are anti-realist in my sense, since they constitute attempts to
avoid full inquisitive engagement with realities. Any number of examples
of such anti-realist practices can be found from the history of science,
ranging from the early modern suppression of Copernicanism to the initial
hostility to the ideas of prions and epigenetic inheritance in the late
twentieth century. Sometimes scientists display a lack of concern that
science should deal in concepts that facilitate empirical learning. We have
seen theories shift in the direction of speculativeness, with fewer opportu-
nities for empirical tests. Often the turn towards untestable hypotheses is
motivated by metaphysical desiderata, such as unity and beauty; a lasting
tradition of such work runs through the history of physics, from the

8: /73 791
 Overview 
ancient theory of four elements through the eighteenth-century system of
imponderable fluids to the superstring theories of today.
Activist realism can make a difference in practice, as one should expect
from a pragmatist doctrine. The sense of futility about the realism debate
(as expressed by Fine, quoted above) is only apt in connection with
metaphysical realism or standard scientific realism. In relation to those
realist doctrines, it can indeed be the case that a clash between philosoph-
ical realism and anti-realism makes very little difference in practice. For
example, in much of nineteenth-century organic chemistry those who
professed their belief in the reality of atoms seem to have done basically
the same kind of work as those who didn’t. Add an appropriate suspension
of belief or disbelief, and the practices come out looking much the same on
both sides. (A shouting match between scientists about realism is no more
productive than one between philosophers.)
One particular debate that would not be productive to engage in is the
one between realism and constructivism. Activist realism recognizes con-
cepts and theories as human constructions. And according to my view of
reality (see Chapter ), the referents of conceptual constructions that
facilitate operationally coherent activities are real. This blending of realism
and constructivism should not worry anyone. We may propose the initial
construction of a concept freely, but whether coherent activities can be
carried out on the basis of it is not at all up to our will. In a well-executed
process of inquiry there will be an iterative process of concept-building that
responds to empirical successes and failures, akin to the process of resis-
tance-and-accommodation that Pickering () identifies in scientific
practice. We do construct realities, but that process of construction is
not arbitrary. Seeing realism and constructivism thus going together helps
us make sense of an apparently constructivist statement from the staunch
realist Ludwig Boltzmann: ‘According to my feeling, the task of theory lies
in the construction of an image of the exterior world, which exists only in
our mind and should serve to guide us in all our thoughts and all our
experiments’ (quoted in Nye , p. ).

Activist Realism and Scientific Progress

Activist realism is an inherently progressivist doctrine, as it is a com-
mitment to attaining more and better knowledge. So, in order to make

See Hossenfelder () for a critique of recent theoretical physics in this regard, and Dawid ()
for a nuanced defence of string theory.

8: /73 791
  Realism
a full articulation of activist realism, I must make a careful consider-
ation of the idea of progress in science. Progressivism, in the basic sense
of ‘wanting to make things better’, has been a powerful modern
ideal ever since the Age of Enlightenment. The advancement of science
has been a central part of the Enlightenment vision, and progress has
been an unquestioned credo within science itself. Although the desir-
ability of progress is nearly tautological (how can making something
better not be a good thing?), we should recognize that there are also
some worthy non-progressivist ideals, such as the conservative yearning
for order and stability, or the acceptance and resignation counselled by
Buddhism. Even when it comes to knowledge, one may feel that we
already know enough to support a decent kind of life, and that learning
more is only likely to cause trouble. Whether this may be the correct
attitude would need to be judged on a case-by-case basis. But on the
whole I don’t think that humanity is at a point at which further
knowledge should be generally refused. Even though there are specific
fronts on which a headlong dash to learn more is unwise and needs to
be regulated through a careful consideration of likely consequences,
such local restraint is compatible with an overall imperative of progress.
Progress was a central issue in the philosophy of science in the s
and the s, but it has taken a back seat nowadays. Perhaps it is
considered an outmoded modernist ideal by some. But scientific progress
has been abandoned as a topic of discussion even by many who value it
greatly, perhaps due to the lack of philosophical progress in giving a good
characterization of scientific progress. But a lack of clear progress is not a
good reason to abandon a philosophical question, and the neglect of
progress in the philosophy of science is a grave mistake. It is difficult to
deny that science does continue to progress, and we need to understand
how it manages to do so, and in what sense exactly. It is ironic that many
of the sophisticated thinkers who shun the talk of scientific progress
nowadays are at the same time fervent and outspoken advocates of social
and political progress. Although there is no logical contradiction in
holding both of these attitudes, there is certainly discomfort. If the idea
of ‘progress’ is so lacking in cogency as to be not worth considering even
in the realm of science, how do we presume to have such a secure notion
of it in the realm of ethics or politics? Moreover, science does and must
continue to play a crucial role in social progress, both intellectually and
materially. I endorse wholeheartedly Kitcher’s sentiment, inherited from
Dewey: progress has been a neglected concept lately, and we need to

8: /73 791
 Overview 
‘rehabilitate’ this ‘endangered concept’, and restore its importance to
inquiry.
The considerations I have made in earlier chapters about knowledge,
truth and inquiry strongly suggest that we should break away from the
view of scientific progress that is found at the foundation of standard
scientific realism, which sees progress as an approach to the Truth. Kitcher
(forthcoming, ch. ) presents ‘pragmatic progress’, which works as progress
from a problematic situation in Peirce’s and Dewey’s spirit of inquiry, as a
direct contrast to teleological progress. Activist realism adds another layer
of motivation to move away from the teleological view of progress, and
instead recommends a view of progress as an abundant form of develop-
ment. The abundant shape of progress follows from the nature of inquiry
as explained in Section .. Inquiry is ultimately unrestricted, in the sense
that any part of the problematic situation may be modified to resolve it.
Different types of resolution point to different ways in which knowledge
can be improved and science can progress.
Here I will only present a very brief (and not exhaustive) taxonomy of
different ways of scientific progress, to give a sense of the abundance.
Inspired by Kitcher’s (, ch. ) analysis of the ‘varieties of progress’,
I give a non-exhaustive list here. () Given an existing question, we can
answer it, or improve the quality of an existing answer. Improvements may
involve correcting known isolated errors, or finding better methods of
observation, measurement or computation. This will be a routine occur-
rence in the fact-gathering type of inquiry. () Within the realm of fact-
gathering inquiry, we can also make progress by asking and answering new
questions (Shan ). Asking new questions without discarding old ones
allows an accumulation of propositional knowledge. () Not only can we
make progress by asking new questions, but we can also improve the
questions that we have been asking. Better questions seek information
that is more relevant and useful to the achievement of the broader aims
that we have. () Active knowledge can be improved by increasing the
coherence of our activities. It happens routinely that we learn to execute a
task more effectively, as anyone who has learned any skill can testify. ()
Propositional knowledge can also be improved through the increase of
coherence. Starting with a proposition that we accept, we can enhance its


Kitcher (forthcoming, ch. ); he has kept the issue of progress alive at least since Kitcher ().

The first sequencing of the human genome was a painstaking enterprise that took thirteen years on a
budget of about  billion dollars; nowadays it is done in a day or two for a few thousand dollars.
Similarly for the manufacturing of an atomic bomb.

8: /73 791
  Realism
truth-by-operational-coherence by improving the coherence of the activi-
ties relying on it, or by devising more activities relying on it. () Active
knowledge can be increased if we can find new (and preferably better)
methods for achieving the same aim; multiple methods for achieving the
same end may be maintained, and they may enhance each other. () When
we create new material objects and phenomena, we can of course gain new
knowledge about them, as well as the knowledge of how to make them. ()
We can create new concepts, and devise epistemic activities involving
them. If these activities turn out to be coherent, then we have new realities,
about which we can ask further questions. () We can set ourselves new
aims, which we then learn how to achieve by devising new activities. If we
are successful, we will certainly have acquired new active knowledge, and
most likely some new propositional knowledge as well. () We can
modify the aims of existing activities, so that they are more compatible
with the aims of other activities we are engaged in, and they serve their
external functions better. Such modifications increase the operational
coherence of the systems of practice in which the activities in
question occur.
So, if we ask ‘what is scientific progress?’, there will be no single answer
that is informative enough. There are many different senses of scientific
progress, ranging from the narrow and well prescribed to the completely
open-ended. These different types of progress will be difficult to aggregate
into one measure, and that is just fine. What is important is that we
recognize all significant types of progress, so that our general notion of
progress does not become impoverished or imbalanced. All of the modes of
progress identified above should be, and are, pursued in science. It would
be foolish to restrict our effort to only some of them, not to mention just
one. When we consider all these multifarious ways of progress, Peirce’s
injunction comes truly alive: ‘Do not block the way of inquiry.’ Rejecting
the traditional view of scientific progress towards perfect truth, it would be
more sensible to adopt the ‘progress from’ perspective advocated by Kuhn
instead, inspired by an evolutionary model of scientific development. But
if we are to think on the evolutionary template, it would perhaps be better
to go with Rouse’s () picture of niche-construction, which fully
acknowledges the dynamic and constructive interaction between the
organism and its environment.


This theme is central to Amy McLaughlin’s interpretation of Peirce, articulated in McLaughlin
() and McLaughlin (); likewise for Susan Haack ().

8: /73 791
 Overview 
Activist realism dictates a pluralistic shape of progress on a larger scale,
too, as I will discuss in more detail in Section .. In trying to gain
knowledge in any way we can, we will most likely end up with multiple
systems of practice in a given field of study. Each system will yield valuable
knowledge in a distinctive way. In fact, my activist perspective has strong
affinity to the very old-fashioned cumulative view of scientific progress,
according to which we keep producing and storing up more and more
knowledge (within each system of practice, and by the accumulation of
systems). And we may create more knowledge yet, if we can establish
meaningful links between the truths and realities found in different
systems of practice. I am not making a logical deduction from activist
realism to pluralism, but a cheerful and exuberant acceptance of the actual
shape of the history of science and other types of inquiry. I am also
recognizing that there is no compelling reason to expect that the future
of inquiry will be radically different in this regard. Realism is generally
taken as a monist position both in metaphysics and in the philosophy of
science, but in my view monism must be rejected if it stands in the way of
progress. Various successes of science will easily result in the establishment
of various real entities and reveal various truths about each.
Such abundance of truth and reality should not embarrass us, as I have
already argued in Sections . and .. True realism will pursue knowledge
unafraid in the spirit of Feyerabend, freed up from the unnecessary
constraints of monism. All theories and all systems of practice that facilitate
successful activities provide ways of learning, and they should all be
maintained and developed actively. The picture of inquiry that emerges
is that of the humble cultivation of abundance. Activist realism is a
commitment to keep learning about the ever-multiplying realities in life.
In order to maximize our learning about realities, we should continue to
invent new concepts, while keeping all the old ones that can support
coherent activities in variously separate and overlapping domains. This is
how we acquire more knowledge about more realities.
These considerations on the nature of scientific progress also give us a
fresh perspective on the state of the debate concerning scientific realism.
Realism should be taken as a business of ampliative inquiry, not ampliative
inference. Activist realism is not ampliative in the sense of drawing
conclusions that are stronger than what the available evidence warrants,
but in the sense of encouraging the actual creation of more knowledge. In
science, all modes of inquiry should be put to work in this relentless drive
to increase and improve knowledge. Bridgman (, p. ) put it
bluntly: ‘The scientific method, as far as it is a method, is nothing more

8: /73 791
  Realism
than doing one’s damnedest with one’s mind, no holds barred.’ And rather
than engaging in defensive manoeuvres attempting to show that science
can approach an unattainable type of truth in some way, realism in
philosophy should focus on articulating all the productive ways in which
knowledge can grow.

. Pragmatism and Realism

As already indicated, I see pragmatism as the most suitable philosophical
tradition in which to base my conception of realism. Various other
thinkers with cognate views to mine have also framed their ideas explicitly
in pragmatist terms. I have been inspired by their attempts to show that
realism can be conceived as a pragmatist position, and in this section
I want to pay homage to these predecessors and draw more attention to
their work. Roberto Torretti has explicitly advocated a ‘pragmatic real-
ism’, and very clearly pointed out what I have called the fallacy of pre-
figuration. Rein Vihalemm and Endla Lõhkivi’s ‘practical realism’ is a
practice-based approach that has its roots in Marxist philosophy. Sami
Pihlström’s ‘pragmatic scientific realism’ is a notable synthesis of
Kantianism and pragmatism, with affinities to C. I. Lewis’s work. Philip
Kitcher presents a revival of classical pragmatism in a way that is compat-
ible with correspondence realism, which I think makes perfect sense if
correspondence is taken as a relation that obtains in actual practice.

Roberto Torretti’s Pragmatic Realism

Among my predecessors who have advocated placing realism within the
pragmatist tradition, I cannot think of anyone more evocative than Roberto
Torretti, whom I have quoted a few times already. Torretti (, p. )
advocates ‘pragmatic realism’, which I have already acknowledged (Chang
b) as a strong inspiration for my thinking. At a very basic level, he sees
realism and pragmatism as motivated by the same drive: realists are concerned
about having correct knowledge of things; so are pragmatists. This is similar to
my perception of affinity between realism and empiricism, and between
empiricism and pragmatism. Torretti (, pp. –) states that science is
‘the continuation of common sense by other means’, reflecting the insistence


The phrase ‘pragmatic realism’, or something very close to it, is used by many authors, including
Pihlström and Kitcher, whose works are also discussed in this section. I should also note Pickering
(, p. ) and Timothy Lenoir (, p. ).

8: /73 791
 Pragmatism and Realism 
by Dewey and other pragmatists that the process of inquiry is continuous from
everyday life to the most esoteric branches of science. In all cases, knowledge
is about getting on in the world, dealing with our situation in the most
effective way possible.
Yet, what philosophers commonly mean by (scientific) realism
diverges quite strongly from pragmatism. Torretti suspects, correctly I think,
that the kind of ‘realism’ stemming from what I have called the fallacy of pre-
figuration (see Section .) is a hangover from a monotheistic perspective on
knowledge and scholarship: ‘The existence of a well-defined or . . . ready-made
reality is no doubt implied by the standard monotheistic conception of God,
but I have not the slightest ground for thinking that God’s worldview can be
articulated in human discourse. To entertain the notion that we could convey
that view in words is a symptom of acute provincialism’ – even though this
provincialism is so often dressed up as universalism! This is something that a
mature philosophy of science ought to be able to transcend: ‘it is pragmatic
realism, not the nostalgic kryptotheology of “scientific realism”, that best
expresses the real facts of human knowledge and the working scientist’s
understanding of reality’ (ibid., p. ). Torretti maintains that ‘science as it is
actually practiced’ is not concerned with looking for an ‘absolute structure’ of
reality (ibid., p. ).

The Tartu School of Practical Realism

My discussion in Chapter  benefited greatly from Rein Vihalemm’s character-
ization of standard scientific realism. Now it is time to pay attention to his own
positive view on realism. Vihalemm, Endla Lõhkivi and their colleagues in Tartu
have developed a position that they call ‘practical realism’ (see Lõhkivi and
Vihalemm  for a recent statement of their position, and a collection of
papers by some of their most prominent fellow-travellers). Vihalemm does not
quite identify his ‘Marxist practice-based realism’ with pragmatism, but I think
the affinity with pragmatism is quite clear. The main tenets of practical
realism as given by Lõhkivi and Vihalemm (, p. ) can be summarized as
follows:

() Science does not represent the world ‘as it really is’ from a god’s-eye
point of view. It is not an ideal of science to pursue a ‘one-to-one
representation of reality’, for which we do not have criteria of judgement.


See Vihalemm (, p. ), for phrase quoted here, and p.  for the distancing from pragmatism.

8: /73 791
  Realism
() The inaccessiblity of the world independently of scientific theories,
paradigms or practices does not argue for internal realism or ‘radical’
social constructivism.
() Scientific research is a ‘practical activity’, whose main form is ‘the scientific
experiment that takes place in the real world, being a purposeful and
critical theory-guided . . . material interference with nature’.
() ‘Science as practice is also a social-historical activity, which means . . . that
scientific practice includes a normative aspect, too.’
() This position qualifies as realism, ‘as it claims that . . . science as practice is
a way in which we are engaged with the world’.

I must say that I agree entirely with all of these tenets, at least if we understand
the terms used in them (such as ‘the world’) in my pragmatist sense. As I see it,
Vihalemm’s fundamental idea behind practical realism is that all knowledge, or
even all discourse, is rooted in practice:

To speak about the world outside practice means to speak about something
indefinable or illusory. It is only through practice that the objective world can
really exist for humans. Therefore, knowledge must be regarded as the process of
understanding how the world becomes defined in practice. (Vihalemm ,
p. )

And what exactly is practice? It is ‘human activity as a social-historical, critically

purposeful-normative, constructive, material interference with nature and soci-
ety producing and reproducing the human world – culture – in nature’ (ibid.).
Not only discourse, but objects themselves are grounded in practice.
Vihalemm (ibid., p. ) quotes Rouse (, p. ) approvingly in this con-
nection: ‘Belonging to the realm of possible determinations open within our
practices is constitutive of a thing’s being a thing at all.’ Vihalemm’s metaphys-
ics is a radical one:

The practice-based approach implies that practical activity has a more funda-
mental status than the status of individual objects-things. Concrete determina-
tion of the existence of individual objects in this case is determined by
specifically defined activities in the context of which these objects-things appear
as specific invariants. (Vihalemm , p. )

As his foil, Vihalemm identifies ‘standard scientific realism’, as discussed in

Chapter . As for anti-realism, he defines it as any position that rejects one
or more tenets of standard realism. He says that practical realism is opposed to
both standard realism and instrumentalist/constructivist anti-realism. The main

8: /73 791
 Pragmatism and Realism 
thing he finds wrong with both sides of the debate is that each position is
‘isolated from practice’ and ‘does not proceed from the practice of real science’
(ibid., p. ).
Tenet  of practical realism is notable in its explicit disavowal of
Putnam’s internal realism, which I will discuss further in Section ..
Vihalemm regards internal realism as not worthy of the title ‘realism’, which
is somewhat puzzling since his own position seems to share a great deal with
internal realism. He adds that an ‘essential difference between internal
realism and practical realism’ is that internal realism ‘belongs to the tradition
of Kantianism and cannot actually be qualified as realism at all’ (Vihalemm
, p. ). Why Vihalemm objected so much to Kantianism is not entirely
clear to me. Perhaps an important clue is given in his comment on the notion
of truth (ibid., p. ): ‘I cannot speak for pragmatists, but in practical realism,
“truth” can be interpreted in a deflationary way and this interpretation is
compatible with semantic realism.’ And Vihalemm traces semantic realism
back to Niiniluoto’s view that ‘truth is a semantical relation between language
and reality. Its meaning is given by a modern (Tarskian) version of the corre-
spondence theory, and its best indicator is given by systematic enquiry using
the methods of science’ (Niiniluoto , p. , quoted in Vihalemm ,
p. ). Like Putnam in his later phase, Vihalemm is pulled into disquotation and
correspondence. But how is that compatible with the practice-based view of
everything, which seemed to me very much like internalism? Vihalemm (ibid.,
p. ) quotes Niiniluoto (, p. ) who, in contrast to standard scientific
realists, says that ‘THE WORLD contains unidentified objects which are identi-
fiable, but not “self-identifying objects” in the bad metaphysical sense’. What
Vihalemm and Niiniluoto mean by ‘THE WORLD’ may seem like the Kantian
noumenal world, but Vihalemm says that ‘for practical realists or [Marxist]
materialists it [the world] is not ungraspable, but identifiable in its concrete
forms of existence through practice’. As Vihalemm (, p. ) notes with
approval, Niiniluoto (p. ) quotes Friedrich Engels from his  text on
Feuerbach in this connection: ‘If we are able to prove the correctness of our
conception of a natural process by making it ourselves . . . then there is an end
to the Kantian ungraspable “thing-in-itself”.’ But I am not sure whether being
graspable through practice is really different from only being cognizable as
phenomena inextricably wrapped in the Jamesian ‘trail of the human serpent’.


Lõhkivi thinks that this might have been due to a current of externalism in Vihalemm’s thought
(conversation in Tartu, March ).

8: /73 791
  Realism

Sami Pihlström’s ‘Pragmatic Scientific Realism’

Much as I find practical realism congenial, I do not find Vihalemm’s treatment
of Kant convincing. In my view, the best way to build on Kantian insights is
shown by Sami Pihlström’s exciting blend of pragmatism and Kantianism (see
Pihlström  for the latest synthesis). He is an avowed pragmatist (unlike
Putnam and Vihalemm) and he regards realism as a crucially important
problem to be tackled by philosophy (Pihlström , p. ), so it will be
very important for me to see how he positions himself. According to my
understanding, Pihlström’s main insight on the realism debate is twofold: we
need Kant, and what we need is a pragmatist version of Kant. To be more
precise: Pihlström (, p. ) wants to hold on to the fundamental Kantian
take on the realism debate, which is a combination of transcendental idealism
(cognition requires a conceptual scheme, which is indeed mind-dependent)
and empirical realism (the empirical properties of objects, cognized within the
conceptual scheme, are not controlled by the cognizing subject). This is
exactly how I see the situation: reality is mind-framed but not mind-controlled;
we cannot claim any absolute validity for our human framing of cognition, but
within that framing the objects we deal with are real. (Pihlström also speaks of
‘scheme-internal’ realism, which seems to be a nod to Carnap, and in good
accord with Putnam’s internal realism.)
In making a pragmatist rendition of the fundamental Kantian insight,
Pihlström recognizes conceptual schemes as rooted in practice. Pihlström
(, p. ; also , p. ) characterizes this move as a sort of ‘pragmatic
“naturalization” of Kantian transcendental idealism’. Conceptual schemes are
not universal and static as Kant had assumed; rather, as epistemic subjects ‘we
are fully naturally situated within context-dependent and context-changing
practices’. It is these practices that ‘contain “relative a priori” conditions that
structure our ways of experiencing reality’. Pihlström’s stance on this is quite
consonant with Friedman’s neo-Kantianism, and I think Pihlström is also
correct in reading Kuhn along these lines, and in tracing the idea of the
‘relative a priori’ to C. I. Lewis (Pihlström , pp. –). What is shared by
all these quasi-Kantian positions is the recognition that cognition is only
possible on the presumption of certain a priori principles. They also share a
denial of the Kantian insistence that the a priori principles to be presupposed
are universal, fixed and inevitable. To put the point paradoxically: the a priori is
necessary, but only contingently so (see Section .).


For non-experts, the exposition by Nicholas Stang () is quite helpful on these basic Kantian
doctrines involved here.

8: /73 791
 Pragmatism and Realism 
For Pihlström, ontology is rooted in practice: ‘It is in our goal-directed
activities and practices themselves that our ontological ways of taking the
world to be in some particular manner are to be located.’ This he identifies as
consonant with Dewey’s view that ‘scientific objects are not “ready made” prior
to inquiry but rather arise out of, or are constructed and/or identified in the
course of, inquiry’ (ibid., p. , paraphrasing Dewey). Furthermore, ‘there are,
and can be, no beliefs at all apart from such activities and practices’ (ibid.,
p. ). As Pihlström himself recognizes, these thoughts are strongly resonant
with Vihalemm’s view that truth and reality are only meaningful within prac-
tice, and similarly with Putnam’s internal realist view. Here is one cogent
summary of Pihlström’s position (, p. ): ‘the Kantian or quasi-Kantian
“transcendental” element – whatever it is that must be presupposed for
inquiry, representation, or cognition to be possible – may lie in the local
practices themselves’. And he points out that ‘a very basic transcendental issue
concerning the practice-laden representability and experienceability of reality
must be taken up from the perspective of [Vihalemm’s] practical realism, too’
(ibid., p. ).
I am mostly in enthusiastic agreement with Pihlström, and it seems to
me that there is an important further step to take, after accepting his basic
outlook: we must go into the details of specific practices and demonstrate, by
some sort of transcendental argument, which specific assumptions are needed
for which specific practices. We should also think again about the nature of
transcendental arguments (see Chang ): what exactly is the method of
sussing out the necessary enabling conditions of something? It is usually seen
as a deduction, but it is not a straightforward kind of deduction – if it were, it
wouldn’t be such hard work following Kant’s arguments! These are the lines of
work that I have tried to carry out in my discussion of ‘mind-framing’ in Chapter .

Philip Kitcher’s ‘Hybrid Pragmatism’ and ‘Real Realism’

I cannot possibly close this discussion of philosophers of science currently
pursuing the integration of pragmatism and realism without attention to the
recent and ongoing work of Philip Kitcher. I have alluded to various aspects of
his thought in the course of the earlier chapters, but this is an appropriate
place to present his work in a more systematic way. Kitcher’s brand of
pragmatism has been laid out in various articles collected in the  volume
Preludes to Pragmatism, and in more recent works including the
 Descartes Lectures at Tilburg University, which will be synthesized into
a forthcoming book entitled Homo Quaerens: Progress, Truth, and Values.

8: /73 791
  Realism
Building on the ideas advanced in the paper on ‘Pragmatism and
Realism: A Modest Proposal’ (Kitcher , ch. ), his recent work explicitly
advocates a ‘hybrid pragmatism’, at the core of which is a synthesis of the
pragmatist and the correspondence theories of truth. Kitcher makes a new
Peircian definition of truth: ‘A sentence S of language L is true just in case, as
used in L, that sentence would be stably retained as inquiry progresses
indefinitely.’ But he adds: ‘Truth for descriptive sentences may be an amalgam,
to which both the idea of correspondence and the Peircean exposition make
necessary contributions’ (Kitcher forthcoming, ch. ). It is important for Kitcher
to hold on to a notion of truth as correspondence, most of all because it serves
the function of explaining pragmatic success (Kitcher , ch. ). Preserving
the explanation of success by reference to truth is central to the version of
scientific realism that Kitcher wants to preserve and defend. But ‘real realism’,
as he puts it, is a down-to-earth doctrine, according to which the ‘truth causes
success’ thesis is an empirical hypothesis, directly verifiable in some cases and
to be trusted in other situations. What he calls the ‘Galilean strategy’ of real
realism (realism in practice) is a piecemeal extension of that successful-
because-true hypothesis from a well-confirmed domain to a new uncharted
domain. Galileo’s work with the telescope provides a persuasive illustration: the
successful use of the telescope in terrestrial situations can be explained by the
verifiable correspondence between how things look close-up to the naked
eye, and how they look through a telescope from afar. In using the telescope
to learn about astronomical objects, Galileo was conjecturing that its prag-
matic success was also due to the correspondence of the telescopic images
and the real shape of the objects, in this case not accessible to direct
observation (Kitcher , ch. ; Kitcher forthcoming, ch. ).
What I find most appealing in Kitcher’s synthesis is the fact that he
treats correspondence as something that takes place within the ‘world of
experience’; in that regard Kitcher’s pragmatist realism is quite consonant with
internal realism and perspectival realism (see Section .). While he is with
many metaphysical realists and standard scientific realists in invoking Tarski for
the understanding of truth, he is also clear about a Tarskian point not usually
emphasized by them: ‘If (as I prefer) we want to apply truth to sentences, we
have to make, or presuppose, reference to a language’ (Kitcher forthcoming,
ch. ). And being language-bound means already having the affordances of
thinking in terms of the entities that exist in the ‘world of experience’ of the
speakers of a language, which are realities bound up with various primary
truths expressed in that language. This is quite consonant with the view of
mind-framed reality that I have proposed in Chapter .

8: /73 791
 Internal and Perspectival Realism 
Yet my view differs from Kitcher’s in a few important respects. While
he recognizes that correspondence is not correspondence to inaccessible
noumenal reality, he does retain the traditional thought that empirical truth
is always a matter of correspondence. Therefore, truth concerning objects that
are not directly accessible to us can only be modelled on the notion of
correspondence, and has to be treated as a hypothesis. In my view, it is more
productive to allow truth-by-operational-coherence to play a role, for two
main reasons. Correspondence truth is secondary truth, which only makes
sense if there are previously established facts to which the propositions under
consideration can correspond, and ultimately in the chain of correspondence
there must be some primary truths. For the pragmatist the only primary truth
one can rely on in empirical domains is truth-by-operational-coherence. Also,
we should allow the possibility that a statement that is in principle able to
attain correspondence truth may be functioning as truth-by-operational-
coherence.

. Internal and Perspectival Realism

In this section I examine two further positions on realism that are cognate
to my own, especially in their embodiment of the realistic spirit. One is
Putnam’s internal realism, based on the insight that we need to make sense
of ontology, truth and correspondence within given conceptual frame-
works. I will also consider why Putnam renounced his internal realism, and
suggest that a more strongly pragmatist interpretation of internal realism
with the help of my notion of truth-as-operational-coherence would have
made it more defensible. Closely related to internal realism is perspectivism
(or, perspectival realism), which is also a species of internalism guided by
the realistic spirit. I believe that the lasting lesson of perspectivism is the
recognition that perspectival truth is something we can have and value, in
line with truth-as-operational-coherence as I conceive it.

What Is Internal Realism?

Many philosophers over the ages have advanced positions that embody the
realistic spirit. One notable tradition may be identified as epistemological
internalism: the view that knowledge only exists as situated in actual knowers,
recognizing the crucial importance of what I call the mind-framing of realities.
Under various notions such as conceptual frameworks, paradigms and world-
versions, internalists have recognized that knowledge can only exist within a

8: /73 791
  Realism
specific system of practice. In the late twentieth century the most prominent
expression of this view was Putnam’s internal realism.
What exactly did Putnam mean by ‘internal realism’? This in itself is a
contentious issue, on which there is considerable commentary. One of the
reasons why it is difficult to give a straightforward characterization of Putnam’s
internal realism is that he defined it primarily in a negative way, as the opposite
of ‘external’ or ‘metaphysical’ realism. In Reason, Truth and History, he pre-
sents ‘the perspective of metaphysical realism’ as follows:

On this perspective, the world consists of some fixed totality of mind-

independent objects. There is exactly one true and complete description of
‘the way the world is’. Truth involves some sort of correspondence relation
between words or thought-signs and external things and sets of things. I shall
call this perspective the externalist perspective, because its favorite point of view
is a God’s Eye point of view. (Putnam , p. ; emphasis original)

In contrast, Putnam’s own perspective is that both ontology and truth are
‘internal’ matters, only meaningful within a given conceptual framework. So,
‘what objects does the world consist of? is a question that it only makes sense to
ask within a theory or description’ (ibid.; emphases original). He adds: ‘“Objects”
do not exist independently of conceptual schemes’ (ibid., p. ). And if
ontology is internal, then it is inevitable that truth will be, too:

If objects are . . . theory-dependent, then the whole idea of truth’s being defined
or explained in terms of a ‘correspondence’ between items in a language and
items in a fixed theory-independent reality has to be given up. (Putnam c,
p. )

So what is truth for an internalist?

‘Truth’, in an internalist view, is some sort of (idealized) rational acceptability –

some sort of ideal coherence of our beliefs with each other and with our


See especially Hacking ; Steinhoff ; Sosa ; Clark and Hale ; Niiniluoto ;
Baghramian ; and Button . Putnam himself (a, p. ) locates the first statement of
‘internal realism’ in his APA Eastern Division Presidential Address given in December ,
published in Meaning and the Moral Sciences (), and says that the position was expounded
further in Reason, Truth and History (). Putnam () and Putnam (a) are also key texts.

The exposition in this text is as clear and explicit a statement of internal realism as one can find in
Putnam’s own words, and it is taken as definitive by Niiniluoto (, p. ) and Hacking (,
pp. –).

Putnam (c, pp. –) later notes that Field has distinguished the three theses as separate
versions of metaphysical realism, and notes that ‘the natural way of understanding’ the second of
these theses involves accepting the first one.

8: /73 791
 Internal and Perspectival Realism 
experiences as those experiences are themselves represented in our belief system –
and not correspondence with mind-independent or discourse-independent
‘states of affairs’. (Putnam , pp. –; emphasis original)

A fundamental affinity I find with internalism is its focus on what is accessible

to us. If internalism is reconceived as a commitment to understanding truth
and reality in terms of what we can experience, then it becomes seamlessly
connected with empiricism and pragmatism. Niiniluoto (, p. ) explains
that Putnam’s internal realism ‘belongs to the tradition of Kantianism in its
denial that the world has a “ready-made” structure, and to pragmatism in its
linkage between truth and the epistemic concepts of verification and accep-
tance’. This blend of Kantianism and pragmatism, fully and explicitly developed
by Pihlström (see Section .), is also at the heart of my own take on realism.
Although Putnam did not clearly identify himself as a pragmatist and called
pragmatism an ‘open question’ (Putnam , subtitle), I believe it is produc-
tive to make a pragmatist reading of Putnam. Interestingly, Torretti (,
p. ) cites Putnam’s internal realism as a direct inspiration for his own
‘pragmatic realism’ (see Section .), and notes that Putnam (, p. )
stated later that he should have called his position ‘pragmatic realism’ instead
of ‘internal realism’.
In motivating and appreciating a pragmatist interpretation of
Putnam, there are two key items to consider further: truth and correspon-
dence. A pragmatist reading of Putnam’s internal realism rests on taking
‘experience‘ in a broad sense linked to actions and practices (see Section
.), not in the sense of information-input through perception. Read Putnam
again with that in mind, when he says that truth consists in the ‘coherence of
our beliefs . . . with our experiences’. With a fuller notion of experience,
Putnam’s notion of truth becomes quite close to my notion of truth-by-
operational-coherence (Chapter ). I believe that this reading is faithful to
Putnam’s own spirit, in wanting an operational notion of truth, rendering it
an internal notion: ‘All I ask is that what is supposed to be “true” be warrantable
on the basis of experience and intelligence for creatures with a “rational and
sensible nature”’ (Putnam c, p. ). This impulse squares well with that of
the pragmatists: ‘What I believe is that there is a notion of truth, or, more
humbly, of being “right,” which we use constantly and which is not at all the
metaphysical realist’s notion of a description which “corresponds” to the
noumenal facts’ (ibid., p. ). Even after he moved away from internal realism,
Putnam maintained this orientation towards the concept of truth. Here is a
pithy statement from his commentary on James: ‘truth, James believes, must
be such that we can say how it is possible for us to grasp what it is’ (Putnam
, p. ).

8: /73 791
  Realism
Putnam’s internal-realist view on correspondence is also strikingly
pragmatist, and very much like my own (see Section .), so much so that
I wonder if I had long ago absorbed it from him without realizing its impor-
tance. The main point is that correspondence is a perfectly operable notion
within a system:

[A] sign that is actually employed in a particular way by a particular community

of users can correspond to particular objects within the conceptual scheme of
those users . . . We cut up the world into objects when we introduce one or
another scheme of description. Since the objects and the signs are alike internal
to the scheme of description, it is possible to say what matches what. (Putnam
, p. ; emphases original)

Such passages clearly indicate Putnam’s view that reference is an internal

matter. And from this it is only a short step to recognize that there is a
perfectly sensible internalist correspondence notion of truth, which I have
called truth-by-comparison in Section .. I prefer to distinguish truth-by-
comparison, which is a matter of secondary truth, from truth-by-operational-
coherence, which is primary. Putnam does not make that sort of distinction,
but both of these types of truth are internalist in Putnam’s sense.

Putnam’s Arguments for Internal Realism

So much for what internal realism is and why I find it congenial. How did
Putnam himself actually argue for this position? One line of argument begins
with a commitment to common sense:

If there is any appeal of Realism which is wholly legitimate it is the appeal to the
commonsense feeling that of course there are tables and chairs, and any philos-
ophy that tells us that there really aren’t – that there are really only sense data, or
only ‘texts’, or whatever, is more than slightly crazy. (Putnam , pp. –;
emphasis original)

Putnam says that what is standardly called Realism is a betrayal of this insight:
this Realism defeats Anti-realism by appealing to common sense, and then
turns itself into ‘Scientific Realism‘ that says everyday objects actually aren’t real
and only god-knows-what sanctioned in ‘finished science’ are real (ibid., p. ).
Rejecting that betrayal, Putnam opts for ‘realism (with a small “r”) . . . that takes
our familiar commonsense scheme, as well as our scientific and artistic and
other schemes, at face value, without helping itself to the notion of the thing
“in itself”.’ He declares: ‘Realism with a capital “R” is, sad to say, the foe, not the
defender, of realism with a small “r”.’ So we arrive at another brief statement on

8: /73 791
 Internal and Perspectival Realism 
internal realism: it is ‘the key to working out the programme of preserving
commonsense realism while avoiding the absurdities and antinomies of meta-
physical realism’ (ibid., p. ). These intuitions were in fact preserved even as
Putnam renounced internal realism, as we shall see below.
But the argument for internal realism that got most attention in the
literature is the ‘model-theoretic argument’, which is not so much an argu-
ment for internal realism but an argument against external/metaphysical
realism. I will not try to retrace the technical intricacies of Putnam’s argu-
ment(s), on which I would defer most of all to Tim Button’s () excellent
and refined critical exposition. Here I only wish to offer an intuitive perspective
on what useful lessons we can take from Putnam. Of the two classes of
Putnam’s model-theoretic argument that Button distinguishes, the more intu-
itively forceful is the class of ‘indeterminacy arguments’, which seek to dem-
onstrate: ‘If there is any way to make a theory true, then there are many ways to
do so’ (Button , p. ; emphasis original). Putnam (, pp. –) wants
to demonstrate that the ‘received view of interpretation’, according to which
the extensions and intensions of terms are fixed by fixing the truth-conditions
of whole sentences, does not work. Extending Quine’s insights, Putnam argues
that ‘it is possible to interpret the entire language in violently different ways,
each of them compatible with the requirement that the truth-value of each
sentence in each possible world would be the one specified’. In Hacking’s
homespun rendition, the thought goes like this: ‘Every time you speak of
cherries, you could be referring to what I call cats, and vice versa. Were
I seriously to say that a cat is on a mat, you could assent, because you took
me to be saying that a cherry is on a tree. We can reach total agreement on the
facts of the world’ (Hacking , p. ). Somewhat more abstractly, Button
(, pp. –) explains the ‘permutation argument’ (as the easiest of the
model-theoretic arguments): ‘Imagine that we were to lay out all the objects in
the world, together with various labels (names) for them . . . Suppose we now
shuffle the objects around. So long as we do not disturb the labels, exactly the
same sentences will come out as true after the shuffling as were true before
the shuffling.’ This destroys the definite correspondence between words and
objects that metaphysical realists presume as a fundamental tenet of their
view.


Putnam’s critique of metaphysical realism can appear self-defeating, because he sets up the model-
theoretic arguments by first adopting the basic terms of metaphysical realism. Inherent in the very
set-up of model theory, we have language L, domain X of individuals, and interpretation-function I
from L to X (see Niiniluoto , pp. ff.). This presupposes at the outset a world composed of
well-defined individuals (without mind-framing). Putnam’s model-theoretic arguments only make
sense as a reductio ad absurdum of metaphysical realism, as Button suggests: ‘Putnam does not

8: /73 791
  Realism

Why Did Putnam Move on from Internal Realism?

As I feel such a strong affinity to Putnam’s internal realism, I must deal with one
familiar issue concerning Putnam’s philosophy: he often changed his mind. His
view on realism was no exception, and from the early s he explicitly
disavowed internal realism. There is something uncomfortable, of course,
about building on a philosophy that its author himself later renounced. Hear
Putnam’s own retrospective accounts of how he gave up internal realism (or,
‘antirealism’!):

I publicly renounced the thesis that true statements are those that we would
accept were conditions to become sufficiently ‘ideal,’ which was the form of
antirealism I defended in the late s and ’s, as a mistaken ‘concession to
verificationism’ . . . (Putnam b, p. , note )

Here Putnam points to the pragmatist-leaning account of truth as the main

thing that was wrong with his internal realism.
Interestingly, Putnam actually retained some crucial aspects of his
internal realism after his explicit rejection of it. Even after renouncing internal
realism, Putnam said that he remained committed to ‘conceptual relativity’,
and I think he was right in maintaining that commitment. Putnam did not
quite renounce pragmatism in general, either. On the contrary, he followed his
renunciation of internal realism with a very sympathetic study of pragmatism
(Putnam ). His  retrospective says that what he did in  was to
return to what James had called ‘natural realism’, or ‘direct realism‘ (Putnam
a, pp. –), which was indeed how he framed his new attitude in The
Threefold Cord, published in . I think he also retained the realistic spirit (as
discussed in Section .), and this was part of his humanism: ‘Our ideas of
interpretation, explanation, and the rest flow as much from deep and complex
human needs as our ethical values do’ (Putnam c, p. ). All in all, I do not
find that Putnam moved on to another convincing and coherent view after
renouncing internal realism. Rather, I wish he would have retained internal
realism and developed it further, and I think having a defensible pragmatist
theory of truth would have helped him not lose his nerve about internal
realism. I like to imagine that Putnam would have approved of my work.
I believe that the notion of truth-by-operational-coherence allows a

embrace meaning scepticism; instead, he uses it as a reductio of opposing positions, such as external
realism’ (Button , p. ; emphasis original). Putnam (, p. ) does say that he is showing
how metaphysical realism ‘collapses into incoherence’.

See also Putnam (a, pp. –, ; b, p. ).

8: /73 791
 Internal and Perspectival Realism 
productive reinterpretation of both internal realism and pragmatism, and a
synthesis of the two.

Perspectival Realism
What I see as the core of internal realism was present long before Putnam. As
mentioned in Section ., it was very much the spirit of Carnap’s late work (see
Carnap , p. ). It was strong in the works of pragmatist-leaning neo-
Kantians before Putnam, including C. I. Lewis, Goodman and Kuhn. In current
philosophy of science, this tradition continues under the name of perspecti-
vism. Ronald Giere’s conception of perspectivism (, pp. –) starts
metaphorically with ‘the idea of viewing objects or scenes from different
places’. But what he calls his ‘prototype for a scientific perspectivism’ is colour
vision: ‘colors are real enough, but . . . their reality is perspectival’. He proposes
to understand perception in general this way, and also instrument-aided
observation (ibid., ch. ). And then comes the most ‘controversial’ and inter-
esting extension of this idea, to scientific theorizing (ibid., ch. ): ‘the grand
principles objectivists cite as universal laws of nature are better understood as
defining highly generalized models that characterize a theoretical perspective’
(ibid., p. ). Building on Giere’s work, Michela Massimi understands perspecti-
vism as ‘a family of positions that in different ways place emphasis on our
scientific knowledge being situated’ historically and culturally (Massimi a,
p. ; emphasis original). For Teller (, p. ) the key source of perspec-
tivality is ‘the different, even incompatible, modelling idealizations needed in
practice for treatment of different aspects of a subject matter’.
Having heard that much, one might fairly ask a series of questions.
What exactly is a perspective? What does situatedness consist in? And what
exactly is it that gets situated? According to Massimi (b, p. , footnote
), a perspective is ‘the actual – historically and intellectually situated – scien-
tific practice of any real scientific community at any given historical time’. What
she means by ‘practice’ includes knowledge claims, methods, and norms of
justification. In another place Massimi (a, p. ) lists ‘scientific represen-
tations, modelling practices, data gathering, and scientific theories’ as the
elements of knowledge that are situated. Put that way, a perspective is not
so different from a Kuhnian paradigm in the sense of ‘disciplinary matrix’; Giere
(, p. ) states that perspective is a narrower notion than paradigm, but
does not elaborate on that point.
I think it is useful to distinguish three separate layers of perspectivality,
which I laid out in a recent work comparing perspectivism and pragmatism
(Chang a; cf. Chakravartty ). () The same content can be expressed

8: /73 791
  Realism
in different ways, in different languages, or using different expressions, that are
not incommensurable with each other. The different expressions will typically
have different connotations embodying divergent expectations and prompt-
ing divergent courses of action. For example, the Newtonian, Lagrangian and
Hamiltonian formulations of classical mechanics are equivalent to each other
in content, but with significantly differing affordances in problem-solving and
further theorizing. () Different perspectives can highlight different aspects of a
given object, and also conceal other aspects. This sense of perspectivism is
consonant with quite a literal reading of ‘perspective’. If we look at a three-
dimensional object in the normal way, we will only see a two-dimensional
picture whose content depends on the direction of gaze. A cylinder may look
like a circle or a square, depending on the perspective one takes on it; the view
of the circle conceals the square-ness of the other view, and vice versa.
Following Giere, we can generalize and extend this thought to both observa-
tional and theoretical perspectives. () Going deeper, one can argue that the
relation between our knowledge and the world cannot be spelled out in an
objectivist way. Any phenomena that we can discuss are already expressed in
terms of concepts (mind-framed), and we can only choose from different
conceptual frameworks that are liable to be incommensurable with each
other. Even to say that two representations are different perspectives on the
same object is to take too much for granted. Each perspective offers knowl-
edge about realities, but not the same realities.
Both Giere and Massimi present their perspectivism as a realist posi-
tion, a happy medium that recognizes both the situatedness of knowledge
and the expression of mind-uncontrolled truth in situated knowledge.
Perspectival realism allows us to transcend the opposition between
‘objectivism (or objectivist realism)’ and ‘constructivism’ (Giere , p. ).
Massimi sees perspectival realism as ‘the latest attempt at bypassing dichoto-
mous divisions’ in the realism debate: it is possible our ‘scientific knowledge
claims’ are ‘perspectival, while also being claims about the world as it is’
(Massimi b, p. , emphases original). But how is that possible? Giere
(, p. ) stresses that perspectival facts are clearly facts; there is ‘truth
within a perspective’ that is as robust as anything. To adapt one of his
examples: looking north from Taiwan, Japan lies to the right of Korea; this is
easily verifiable by experience, and quite indisputable, and entirely compatible
with the fact that within the perspective from Kamchatka, Japan is to the left of
Korea. And since any observing and theorizing must take place perspectivally,
all truth claims are ‘relative to a perspective’. But one can and should take a
maximally realist attitude about such perspectival truth, in any reasonable
sense of ‘realist’: in Giere’s words again (, p. , emphasis original): ‘claims

8: /73 791
 Pluralism and Realism 
made from within a perspective are nevertheless intended to be genuinely
about the world, and thus “realistic,” even though not fully precise or com-
plete’. This notion of perspectival truth is very consonant with my notion of
truth-by-operational-coherence, especially given Massimi’s view that perspec-
tives consist in sets of practices. And embracing the pragmatist notion of truth
would push perspectivism into adopting the deepest sense of perspectivality
(option  above).
But one may worry that we are now leaning too far in the direction of
constructivism. Massimi pushes beyond Giere’s argument, in a bid to show that
perspectival truth can actually tell us something non-perspectival. Different
perspectives may come to an agreement on specific points (‘agreeing-whilst-
perpsectivally-diagreeing’) such as the charge of the electron as a minimal
unit of electric charge (Massimi ). More generally, a productive interac-
tion of different perspectives can point us to non-perspectival knowledge.
But if the kind of scenario envisaged by Massimi worked out too well, it would
provide fodder for the critique of perspectivism by Chakravartty, who argues
that perspectivist arguments merely point to incomplete or idealized ver-
sions of fully objective and non-perspectival truth: ‘one may speak the
truth . . . without thereby speaking the whole truth . . . But this is not
tantamount to perspectivism’ (Chakravartty , p. ). Concerning the
fact that scientists work with mutually inconsistent models of the same sets
of phenomena or objects, Chakravartty contends (ibid., p. ): ‘even though
there are thoroughly reasonable senses in which scientific models – and in
particular, inconsistent models . . . are perspectival, this does not entail that
we do not or cannot learn non-perspectival facts relating to the things these
models model’. So if Massimi’s triangulation from perspectival to non-
perspectival knowledge is successful, that will only pave the way to
Chakravartty’s dismissal of perspectivism, which brings us right back to
standard scientific realism.

. Pluralism and Realism

The activist and realistic spirit of realism that I advocate recommends that
we do whatever we plausibly can in order to enhance our knowledge of
realities. Adopting realism as I take it enhances the arguments for pluralism
advanced in Sections . and .. Concerning pluralism more generally,
I mostly stand by the view that I have expressed in previous publications,
including the answers given there to worries about the plausibility and
consequences of it. Here I will further develop some aspects of my
thinking about pluralism that are enhanced by the articulation of realism

8: /73 791
  Realism
made in this book. First, the realistic spirit of internalism (see Section
.) naturally allows the plurality of conceptual schemes. Second, the
realistic and the activist inclinations together provide strong support for
what I have called ‘conservationist pluralism’, which takes care not to
discard practices with good track records of success. Third, activist realism
is best served by interactive pluralism, which seeks to reap benefits from
productive interactions between different systems of practice, going
beyond the non-interactive co-existence of different systems found in
‘tolerant pluralism’ or ‘foliated pluralism’.

Internalism and Pluralism

Internal realism and other related positions discussed in Section . are
strongly conducive to pluralism. The link between perspectivism and pluralism
is obvious: a perspective on an object wouldn’t be a perspective if it were
impossible to take other perspectives on the same object. Similarly the
internal–external distinction almost implies pluralism: if internally valid propo-
sitions are formulated within a conceptual scheme, then whole other sets of
valid propositions will be formulated within other conceptual schemes. I say
‘almost’ because it is possible to insist, as Kant did, that there is only one
fundamental conceptual scheme for cognition. But the whole histories of neo-
Kantian philosophy, modern mathematics and modern physics are testaments
to the fact that Kant’s monism in that regard was not convincing, even to
those most sympathetic to his views in general. Carnap, Lewis and many
others freely admitted that there were alternative languages or conceptual
schemes. As Rescher put it (, p. ): ‘different languages afford us
different ways of talking – of saying different sorts of things, rather than saying
“the same things” differently or making different claims about “the same
thing”’. And in the internal sense, if we are making valid statements about
something, then that something must exist, so a straightforward kind of
ontological pluralism follows, as discussed in Section ..
Putnam’s internal realism was based on the understanding that
different groups of humans do routinely develop divergent conceptual
schemes. Early on he noted that for the internalist ‘there is no God’s Eye point
of view that we can know or usefully imagine; there are only various points of
view of actual persons reflecting various interests and purposes that their
descriptions and theories subserve’ (Putnam , p. ). Putnam went as


Here I build on the ideas expressed in Chang (). Israel Scheffler () has long argued for the
compatibility of realism and pluralism. So has Feyerabend, as I argue in Chang ().

8: /73 791
 Pluralism and Realism 
far as to say: ‘internal realism is, at bottom, just the insistence that realism is not
incompatible with conceptual relativity’ (Putnam , p. ; emphasis orig-
inal). He gives some instructive examples: ‘from the point of view of life and
intellectual practice, a theory which treats points as individuals and a theory
which treats points as limits may (in their proper contexts) both be right’. Or in
physical science, ‘a theory which represents the physical interactions between
bodies in terms of action at a distance and a physical theory which represents
the same situation in terms of fields may both be right’. Generally, ‘theories
with incompatible ontologies can both be right’ (Putnam c, p. ). Here
Putnam begins to sound a bit like the Goodman of Ways of Worldmaking: ‘That
we do not, in practice, actually construct a unique version of the world, but
only a vast number of versions . . . is something that “realism” hides from us’
(ibid., p. ).
Putnam the internal realist emphasizes that the stability of reference
is not a matter of objective truth, but interpretation: ‘Why do we regard it as
reasonable of Bohr to keep the same word “electron” (Elektron) in  and
, and thereby treat his two different theories . . . as theories which
describe the same objects, and regard it as unreasonable to say that
phlogiston referred to valence electrons?’ (Putnam c, p. ). And I do
think it is eminently reasonable to say that phlogiston refers to valence
electrons (or conduction electrons)! (Chang a, pp. –). Contrary to
the notion of ‘rigid designators’, Putnam argues: ‘reference, like causality, is a
flexible, interest-relative notion: what we count as referring to something
depends on background knowledge and our willingness to be charitable in
interpretation. To read a relation so deeply human and so pervasively inten-
tional into the world and to call the resulting metaphysical picture
satisfactory . . . is absurd’ (Putnam , p. ). Putnam pre-empts the
attempt to ground the metaphysical objectivity of reference in the meta-
physical objectivity of causation, by noting that causation itself is ‘radically
perspectival’. The latter is the familiar point about context-dependence: ‘For
Earthians it may be a discarded cigarette that causes a forest fire, while for
Martians it is the presence of oxygen’ (Sosa , p. ). Putnam concludes
(c, p. ): ‘the claim that we have a notion of reference which is
independent of the procedures and practices by which we decide that
people in different situations . . . do, in fact, refer to the same things . . .
seems unintelligible’.
Even after renouncing his internal realism, Putnam maintained a
commitment to ontological pluralism. In response to Maudlin’s push for an
‘unsophisticated metaphysical realism’, Putnam (b, p. ) stated that he
advocated ‘sophisticated realism’, ‘a realism that accepts the idea that the

8: /73 791
  Realism
same state of affairs can sometimes admit of descriptions that have, taken at
face value, incompatible “ontologies,” in the familiar Quinean sense of “ontol-
ogy”.’ Similarly, while Niiniluoto (, p. ) does not accept internal realism,
his own position of ‘critical scientific realism’ subscribes to ‘the principle of
conceptual pluralism’, according to which ‘all inquiry is relative to some
conceptual framework’. In Niiniluoto’s view (, p. ), ‘the true ingredient
of internal realism and the cookie cutter metaphor is conceptual pluralism: the
world can be described or conceptualized with several different linguistic
frameworks’.

Conservationist Pluralism
If we take a fresh look at the history of science with activist realism in mind, we
can see a particularly important way in which epistemic pluralism manifests
itself in the practices of working scientists. There is a long-standing unspoken
policy among scientists, to which I have given the name of conservationist
pluralism (Chang a, pp. , ): retain previously successful systems of
practice for what they are still good at, and add new systems that will give us
knowledge about other realities. This practice is quite widespread in actual
science, contrary to what is often imagined by standard scientific realists. It is a
realistic attitude that practising scientists take. They tend to take care not to
discard useful theories from the past – even while declaring them to be false,
even while paying lip-service to reductionism and partaking in dreams of the
grand unified theory. A whole host of examples of such protected old theories
that are put to effective use can be given, ranging from geometric optics to
orbital theory in chemistry. And contrary to what one might expect, physics
shows this conservationist pattern of development more starkly than any other
science. Physicists and others who use physics have retained various successful
systems that are good in particular domains: geocentrism (for navigation),
Newtonian mechanics (for other terrestrial activities and for space travel within
the solar system), ordinary quantum mechanics (for much microphysics and
almost all quantum chemistry), as well as special and general relativity, quan-
tum field theory, and more recent theories. For those who would feel upset by
the suggestion that the applicability of something like general relativity is
‘local’, I can only point to all the situations in which we would not dream of
using general relativity, as well as the fact that there have been very few
specific empirical tests of general relativity. Goodman (, p. ) was correct
to observe: ‘The pluralist, far from being anti-scientific, accepts the sciences at
full value.’

8: /73 791
 Pluralism and Realism 
The standard scientific realist inference from success to truth is
typically made in a monist framework. According to that reasoning, the
most successful of all theories in a given domain is the true theory; when a
more successful new theory emerges, the attribution of truth should be
withdrawn from the formerly most successful theory. In contrast, realistic
realism can make perfect sense of scientists’ conservationist behaviour: the
success of a theory only tells us that it has a degree of truth-by-operational-
coherence; this does not amount to proof, and moreover it does not negate
the truth-by-operational-coherence of competing theories. But success does
provide a credible promise of continued and further success, a promise to be
accepted with our eyes wide open to the problem of induction. That promise
of further success can be shared by many competing theories within a given
domain. When a system of practice has produced success time and again, it
makes sense to keep it for future use. Preserving a successful old system is a
modest and reasonable inductivist policy, firmly rooted in the kind of basic
inductive reasoning and action that Hume taught us we cannot live without.
The successfulness of a tried-and-tested system should be robust in the face of
another system that does something else well (or even the same thing well in
a different way), which deserves its own credence. To put the point most
generally and vaguely: whatever we regard as responsible for success should
be preserved so that it may continue giving us that success. With conserva-
tionist pluralism we can, once again, understand the progress of science as
cumulative: not an accumulation of simple unalterable facts (from which more
and more general theories would be formed), but of various locally effective
systems of practice which somehow continue to be successful.
To illustrate that point, return to Einstein for a moment, this time to
his work on special relativity. Einstein’s renunciation of the aether and absolute
space and time (see Sections . and .) should not be taken as a moment of
metaphysical enlightenment. Instead, I propose that we take Einstein’s work
as a pluralist move that demonstrated how such a fundamentally different way
of doing physics could be coherent. Most people who have successfully
mastered special relativity would remember what a struggle it was to learn
to think relativistically, yet how much sense it made once the learning was
done. But Einstein did not invalidate the whole range of activities based on the
presumption of absolute space and time, which continue to take place in a


I have pushed against this monism by recognizing that success is a multi-dimensional thing (Chang
a, pp. –, and references therein).

Nor should it be seen as the dawning of a general operationalist conscience, a reading which
Einstein himself rejected, against Bridgman, Heisenberg and Dingle (see Chang b).

8: /73 791
  Realism
whole range of domains from Newtonian mechanics to molecular biology, not
to mention everyday life. Rather, relativity was an amazing move in physics in
that it was a piece of maximally unrestricted inquiry, which solved a difficult
problem by means of unprecedented methods that led to unforeseen results.
It resulted in new pragmatic understanding based in new conceptual activities.
The setting-up of the relativistic reference-frame gave new meaning to the
very concepts of space and time, and to the concept of simultaneity.
This pluralist perspective has a strong implication for the problem of
theory-choice in the philosophy of science. I suggest that the kind of theory-
choice (or paradigm-choice) conceived in the traditional way is not necessary.
If we stop worrying about choosing the winner and eliminating all other
competitors, we can have a much more relaxed and open-minded view
about the nature and assessment of scientific progress. Pluralists can allow
that any system of practice with sufficient promise of progress should be
permitted and encouraged. Now, ‘sufficient’ is a vague notion, of course, and
the judgement of sufficient promise depends on how many systems we can
afford to maintain at once, within the constraints of the material and
cognitive resources available to us. But the vagueness of judgement here is
fine, because it suits the vagueness of the actions that need to be taken.
Allowing or supporting a system of practice is not a binary decision of life
and death; different degrees of support are available, ranging from barely
tolerating something to going ‘all in’ for it. So, if enough people have enough
faith that a certain system of practice shows enough promise of progress,
then that is good enough prima facie reason for activist realists to consider
giving the system some measure of support.
Renouncing monistic theory-choice also means giving up on the
notion of progress towards a fixed final destination. Unrestricted inquiry
does not have a pre-determined destination because the realities we want
to learn about are not pre-figured independently of our inquiry. They
become realized in different ways as we push inquiry in different directions.
I resist the alluring vision of the development of knowledge steadily con-
verging on a final point, which has captivated a range of thinkers from Peirce
to Friedman who had insights that would otherwise have led them to
pluralism. Instead I have come to embrace a vision of knowledge advanced
by Feyerabend, which is a picture of abundance. When we let inquiry go
unrestricted, its results seem to diverge in interesting ways, even as all
inquirers strive to learn in the most successful way possible. We may all build
upward on earth, but ‘up’ is not all the same direction! Going ‘up’ in Uruguay
and going ‘up’ in Korea are both progressive in the sense of going higher,
but they go in precisely opposite directions if we are looking from outer

8: /73 791
 Pluralism and Realism 
space. The point is not merely that we do not know which direction of
development is right, but that there may not even be such a thing as the
correct or even the best direction of development.

How Different Types of Pluralism Serve Activist Realism

Now I want to consider more carefully how pluralism can serve the cause of
activist realism. Here it will be useful to distinguish different types of pluralism,
which have different takes on the relation between the different knowledge-
systems. (‘Knowledge-system’ here is a deliberately vague term, to encompass
theories, models, research programmes, paradigms and systems of practice.)
One important modification I need to introduce is that in addition to the
‘tolerant’ and ‘interactive’ forms of pluralism that I previously distinguished,
there is another major form: foliated pluralism.
According to the weakest type of pluralism, different knowledge-
systems that are all valid in the same domain are quite compatible with each
other. I adopt Stéphanie Ruphy’s () term ‘foliated pluralism’ to designate
this class of views, which can also be understood as a mild kind of
perspectivism (see the ‘first layer’ of perspectivism that I identified in Section
.). According to Ruphy, ontology is enriched by the coherent addition of
new perspectives on to existing ones. Foliation provides a contrast to the
image of patchwork given by Cartwright: the different layers of knowledge
cover the same area, and they are intimately connected with each other, while
each adds something different. Does foliated pluralism (or first-layer perspec-
tivism) contribute to activist realism? This very much depends on one’s view of
the nature of knowledge. If one takes an entirely propositional view of
knowledge, one might deny that having two sets of statements that are
entirely translatable into each other adds anything to knowledge, compared
to the situation in which we have just one of those sets of statements.
However, according to my own view, a perspective overlaid upon another
one can indeed constitute or create new knowledge, if it supports a new
system of practice. But in that case it is quite likely that the different ‘leaves’
(folios) will start to develop in different directions, eventually creating diver-
gent bodies of active and propositional knowledge. When that happens the
neat foliation will cease.
In the next type of pluralism, which I have called tolerant pluralism,
the different knowledge-systems are not presumed to be fully compatible with
each other, but they co-exist peacefully without interfering with each other. An
excellent example of tolerant pluralism is given in Werner Heisenberg’s notion
of ‘closed theories’, as articulated in an insightful presentation by Alisa Bokulich

8: /73 791
  Realism
(, ch. ). According to Heisenberg, classical mechanics and quantum
mechanics are both closed theories, meaning that they are each perfect in
themselves and cannot be improved in minor ways; they have to be accepted
as complete packages, or rejected altogether. In that way closed theories are
similar to Kuhnian paradigms, but in Heisenberg’s view tolerant pluralism was
the correct attitude, rather than the revolutionary abolition of the previous
paradigm. Tolerant pluralism can certainly contribute to activist realism.
Recall Feyerabend’s vision of proliferation. Less flamboyantly, conservationist
pluralism points in the same direction: each system of practice can produce its
own knowledge, and a tolerant pluralist society can reap the knowledge from
all of the systems. However, questions do arise as to how the strands of
knowledge emerging from the different systems can be put together.
Depending on the situation, the answer may be that they are not brought
together – different sub-communities may exist in different spheres of activity
and not interact with one another in any epistemically meaningful way. Even
one and the same person or community may engage in activities in different
systems of practice at different times, with no strong connection between
them. Teller points out various situations in which physicists use multiple
models of the same object and do not try to bring them together: sometimes
water needs to be modelled quantum-mechanically, and sometimes as a
classical fluid (Teller , pp. –); having both representations involving
gravitational force and representations involving space-time curvature gives us
‘much richer access to the way things are’ (Teller , p. ).
It is interactive pluralism that can serve the needs of activist realism
most fully, with productive interactions between the different knowledge-
systems generating new avenues of inquiry in addition to those offered
separately by each system. I have previously discussed competition, co-
optation and integration as three main modes of inter-system interaction
(Chang a, sec. ..). A deeper and richer set of insights about inter-
system interactions comes from the close examination of the quantum–
classical relation by Bokulich (). She sees these interactions as going
beyond pluralism, but in my terms her vision is actually in the spirit of the
interactive variety of pluralism. Particularly striking is her discussion of ‘semi-
classical mechanics’, which ‘uses classical quantities to investigate, calculate,
and even explain quantum phenomena’, whose methods employ ‘an unor-
thodox blending of quantum and classical ideas, such as a classical trajectory
with an associated quantum phase’ (Bokulich , p. ). She stresses that


Bokulich (, p. ) notes a similarity to Cartwright’s ‘metaphysical nomological pluralism’.

8: /73 791
 Epistemic Iteration Revisited 
there are various types of benefit that have accrued from this line of work,
including not only computational convenience but new physical insights and
explanations, and even the discovery of new phenomena. Similar kinds of
interaction can be seen to be taking place in the GPS example discussed in
Section .. Massimi (b, p. ) has also emphasized the importance of
productive inter-perspectival interactions: ‘Each scientific perspective . . . func-
tions then both as a context of use (for its own knowledge claims) and as a
context of assessments (for evaluating the ongoing performance-adequacy of
knowledge claims of other scientific perspectives).’ I am not convinced that
such interactions can give us non-perspectival knowledge, but they surely
create more and better perspectival knowledge. These thoughts extend
Sandra Mitchell’s (; ) picture of the ad hoc integration of different
systems to meet particular situations. Interactive pluralism is an essential
feature of unrestricted inquiry; without it the full potential of our knowledge-
seeking activities cannot be realized.

. Epistemic Iteration Revisited

In this section I will consider the iterative character of scientific progress
more carefully. The realistic spirit accepts that progress must start from
some inherited starting point, and that no starting point given to us will be
fully justified. In the process of epistemic iteration, we knowingly start
inquiry on the basis of an imperfect starting point, and use the outcome of
that inquiry in order to improve its own starting point. The combination
of conservatism and optimism in epistemic iteration exemplifies realistic
realism. I will start by reviewing and restating my previously articulated
ideas about epistemic iteration. And then I will show how epistemic
iteration can lead to progressive changes in our concepts and ontologies,
our methods and principles, and even our aims. We can recognize all of
these patterns of iterative progress with no need for the notion that we get
closer to an absolute kind of truth. Accepting a given situation and starting
inquiry on that basis is a rational method of achieving progress, which
enables deeper kinds of progress than a mere accumulation of facts or a
simple increase of precision and scope in our knowledge.

Conservatism and Optimism in Epistemic Iteration

I first articulated the notion of ‘epistemic iteration’ in relation to the justification
of measurement methods (Chang , ch. ), and have extended the notion
in various directions since then (Chang ; a; a). I will start by

8: /73 791