Philosophy of Mind’s New Lease on Life:
Autopoietic Enactivism meets Teleosemiotics
Professor Daniel D. Hutto
Professor of Philosophical Psychology
School of Humanities
University of Hertfordshire
de Havilland Campus
Hatfield,
Hertfordshire AL10 9AB
Telephone: +44 (0)1707 285655
Email: d.d.hutto@herts.ac.uk
“What he saw was a shocking surprise.
Every Who down in Who-ville, the tall and the small,
Was singing! Without any presents at all!
He HADN’T stopped Christmas from coming!
IT CAME!
Somehow or other, it came just the same!
And the Grinch, with his grinch-feet ice-cold in the snow
Stood puzzling and puzzling: ‘How could it be so?’
‘It came without ribbons! It came without tags!
It came without packages, boxes or bags!’
And he puzzled three hours, till his puzzler was sore.
Then the Grinch thought of something he hadn’t before!
‘Maybe Christmas,’ he thought, ‘doesn’t come from a store.
‘Maybe Christmas … perhaps … means a little bit more!’”
-
Dr. Suess, How the Grinch Stole Christmas,
New York: Random House, 1957.
1. Introduction
Mind in Life is a big book – in more ways than one, and in more good ways than one.
In promoting the idea that there is a deep continuity between life and mind, Thompson
defends a boldly anti-representationalist version of enactivism: it challenges, root and branch,
the ‘passive-cognitivist’ view of the mind-brain. It aims for nothing short of the eradication
of misleading Cartesian and Kantian (of the Critique of Pure Reason) dualisms that, despite
challenges, continue to dominate analytic philosophy of mind and some branches of
cognitive science. Thompson sees adoption of the enactive approach as a way to put aside the
mind-body problem, once and for all, and to refocus our investigations on the more fertile –
phenomenologically-inspired body-body problem. We are meant to give up on the traditional
input-output processing model of the mind, one that continues to pay homage, if only tacitly,
to the idea of what is sensorially or informationally ‘given’; of content that is received or
informs, on the one hand, and the idea that such gifts are intellectually categorized,
conceptualized, and schematized – in a downstream and serial manner – by higher forms of
cognitive spontaneity, on the other. By reconceptualizing mentality in essentially active,
2
dynamic and loopy terms, Thompson undermines the traditional boundaries and dichotomies.
The boundaries thought to hold between mind and body and between mind and world are
revealed to be, ultimately, of only heuristic value, having no genuine metaphysical import.
We are asked to trade in the modern, Cartesian, picture of the mind as a sort of
mechanism for a more Aristotelian vision of mentality, which emphasizes its biological
character and the special features it shares with all living systems.1 All of this is in the service
of putting us in a better position to understand aright the place of consciousness in nature,
and thus to deal with the infamous explanatory gap.
In what follows, for reasons of space, I say nothing more about the book’s success in
meeting its principal aim of enlarging and enriching “the philosophical and scientific
resources we have for addressing the gap” (p. x). I think it surely achieves this and I am
wholly sympathetic to Thompson’s ambitions on this front and, by and large, the general sort
of enactivism he promotes. This is not to say that this book closes the gap (it doesn’t seek to)
or that it is the last word on consciousness (it doesn’t promise to be). Still, it is a genuine tour
de force, that adds creatively and convincingly to our ways of understanding basic forms of
mentality. I will take this much for granted here and work instead to ensure that Thompson’s
work gets the kind of reception it should have. For the truth is that many working in analytic
philosophy of mind and cognitive science continue to be utterly mystified about what
enactive approaches have to offer.
To remedy this, this commentary will seek to clarify certain core features of Thompson’s
proposal about the enactive nature of basic mentality, as best it can, and to bring his ideas
into direct conversation with accounts of basic cognition of the sort favoured by analytical
philosophers of mind and more traditional cognitive scientists – i.e. those who tend to be
either suspicious or critical of enactive/embodied approaches (to the extent that they confess
to understanding them at all). My proposed way of opening up this sort of dialogue is to
concentrate on the close similarities between Thompson’s biologically-based proposal about
non-representational forms of basic cognition and what I take to be a reasonable modification
to the ambitions of teleosemantic theories of content. Insofar as today’s theories of mental
representation are less concerned to understand content in properly semantic terms they are
moving ever closer to the sorts of account proposed by enactivists of the Thompsonian stripe
– close enough to have meaningful debates about the nature of basic mentality. It is against
this backdrop that I put a spotlight on the true promise and value of enactivism, providing
some compelling reasons for wanting to go Thompson’s way.
3
This will be achieved, in large part, by showing that there is no wholly agreed way in
which the family of fundamental theoretical notions – content, representation and, especially,
information – are understood by opponents of enactivism. Despite their centrality and critical
importance to traditional cognitivism there are no agreed and well-defined accounts of the
exact nature of contentful, representational or informational properties. As such, it can be
difficult to determine the precise boundaries between representationalist and nonrepresentationalist approaches. Moreover, when we get down to brass tacks, it looks as if the
most scientifically respectable attempts to make sense of these ideas leads straight into the
arms of the sort of enactivism that Thompson proposes (or something near enough). In short,
once we make necessary clarifying adjustments to the best proposals for understanding the
kind of informationally-sensitive responding that constitutes basic mentality, Thompson-style
enactivism may well turn out to be the most promising naturalist game in town, if not the
only one.
2. The Information Processing Challenge
The last claim of the previous section is far from obviously true. Originally advanced by
Varela, Thompson and Rosch (1991), the big idea behind enactivism is to treat consciousness
and cognition as emergent phenomena constituted by, and thus to be understood in terms of,
specifiable patterns of organismic activity. Enactivism of this stripe denies that the most
basic forms of genuinely mental activity necessarily involve, or are to be explained by, the
manipulation of contentful representations. Instead, enactivists hold that mentality emerges
from – is ‘brought forth by’ – the self-creating (autopoietic) activities of organisms, and that
the latter are constituted by essentially embodied, diachronic environmental interactions that
are the basis for new possibilities for self-creating activity in a dynamic way. Drawing on
insights from phenomenology and dynamical systems theory, enactivists invert the familiar
explanatory strategies of orthodox cognitive science by supposing that “Abilities are prior to
theories ... Competence is prior to content … [and that] knowing how is the paradigm
cognitive state and it is prior to knowing that” (Fodor 2008, p. 10). The framework has
proved attractive to many. A great variety of enactivist proposals have now been advanced
about many topics, including: consciousness, perception, intentionality, attention, memory,
emotion, intersubjective social cognition and self-consciousness.
Nevertheless this take on basic mentality is viewed by many as at best something that
might supplement existing theories of cognition, and at worst is nothing more than a
4
confused and obscuring gloss which adds nothing positive to the already well-established
cognitivist accounts. Varela-inspired versions of enactivism face a standard objection, given
that their “main explanatory tool … is the theory of self-organizing and autonomous dynamic
systems” (Thompson 2007, p. 26). For even among those who are prepared to accept that the
enactive approach holds promise for understanding how organisms develop and interact over
time, a standard verdict is that it lacks the independent explanatory resources to provide a
genuinely alternative understanding of the basis of mentality (Ramsey 2007, Clark 2008).
Clark (2008) describes the emphasis placed by theories like Thompson’s on the dynamics
of the total state of systems as both a boon and burden. On the positive side, it allows the
theorist to “accurately capture the way two or more systems engage in a continuous realtime, and effectively instantaneous dance of mutual codetermining interaction” (p. 25). On
the downside, it is problematic “insofar as it threatens to obscure the specifically intelligencebased route to evolutionary success” (p. 25). And it does this to the extent that it fails to
recognize “the brain as the principal (though not the only) seat of information-processing
activity” (p. 25, emphasis original). Accordingly, we need not deny the importance of timing,
action and coupled unfolding to cognition so long as we do not forget that these play support
roles in intelligent responses “grounded in processes of information extraction,
transformation and use” (Clark 2008, p. 19). In sum, the complaint is that any version of
enactivism that relies entirely on dynamical systems theory under-appreciates the
fundamental role played by information-processing mechanisms in making mental activity so
much as possible. Call this the Information-Processing Challenge.
The Information-Processing Challenge would present a formidable problem for
enactivists if it could be safely taken for granted that the standard computational/information
processing explanatory strategies of traditional cognitivism are in perfectly good order under
standard renderings. But enactivists question just this. This is precisely what Thompson
(2007) has in his sights when he doubts the truth of the received view in cognitive science.
That view, he maintains, is committed to the idea that “in order to explain cognitive abilities
we need to appeal to information-bearing states inside the system. Such states, by virtue of
the semantic information they carry about the world, qualify as representations” (p. 52,
emphasis added).
This is where the trouble in assessing this debate starts. For there is, in fact, less
consensus about exactly what orthodox cognitive science is committed to than this statement
suggests – in particular it is not clear that the kind of information or content that matters must
be semantic. Still the conservative wing of cognitive science certainly does insist that
5
intelligence depends on information-processing. Textbooks in the field tell us that the core
assumption of traditional cognitive science is “that there are sub-personal contents and subpersonal operations that are truly cognitive in the sense that these operations can be properly
explained only in terms of these contents” (Seager 2000, p. 27). As the Clark quotation
reminds us, information is thought to be the basic currency of cognition; it is received, stored,
manipulated, and transformed by intelligent systems. It is the fuel for cognitive engines. So
conceived “information is a prime commodity, and when it is used in biological theorizing it
is granted a kind of atomistic autonomy as it moves from place to place, is gathered, stored,
imprinted, and translated” (Oyama 2000, p. 1).
This standard metaphor suggests that cognitive operations really involve the manipulation
or processing of information or content of some kind or other – but despite the foundational
importance of this claim it is incredibly difficult to pin down, with any firm grip, the
theoretical commitments of those who propound such stories. In particular, it is hard to get a
clear sense if they are truly committed to this sort of picture and if so exactly what it is that is
supposed to be processed by these intelligent systems and how this is done. I suggest that the
more we work through the possible readings and home in on a credible account that is
naturalistically acceptable, the closer we come to accepting the kind of enactivist proposal
Thompson advocates. What then might possibly fuel cognition?
3. What might Content be?
The sentences of natural language, as expressed in linguistically mediated beliefs and
utterances, are clearly the paradigms of contentful representations, if anything is. It is also
quite clear that when philosophers of mind first developed naturalized theories of content
their aim was to explain how mental representations could have semantic properties of just
the same sort possessed by linguistic representations. A major motivation for this project is
that success in this endeavour would make it possible to explain how language could gain its
semantic properties from underived mental contents, those which could be explained by
appeal to non-semantic properties – such as causation or biological function. Assuming that
an exact parallel exists between the content of thought and language ensures that whatever
can be thought or judged can be said, in principle. Saying something really only requires
finding a public means of expressing oneself. With this in mind psychosemantic theories of
content, including teleosemantics, seek to explain the semantic properties of mental states
6
where these are understood as having the very same sorts of semantic properties possessed by
natural language expressions.
Proponents of Fodor’s ‘language of thought’ hypothesis wear this commitment on their
sleeves. They are interested in mental representations with semantic content of essentially the
same kind as that which natural language sentences possess (this follows given that whatever
content the latter have is wholly derived from the former). Fodor (2008) tells us that “the
content of a mental representation is its referent” (p. 216). If such content is to express truths
then the referents in question would need to be states of affairs (or the equivalent) and the
referents would have to be picked out intensionally, i.e. under some description or mode of
presentation, in order to satisfy the platitudinous disquotational rule (“Snow is white” in L iff
Snow is white).
It is quite clear that others working in this area are also primarily interested in mental
representations with content of this sort. Millikan (2005), for example, thinks that
representational content is essentially truth conditional. Hence, “intentionality has to do with
truth conditions” (p. 93). For her this requirement goes all the way down, thus “the
intentionality of language is exactly parallel to the intentionality of bee dances” (p. 98). The
same goes for Papineau who offers a “naturalistically acceptable explanation of
representation: namely that the biological purpose of beliefs was to occur in the presence of
certain states of affairs, which states of affairs counted as their truth conditions” (1987, p.
xvi). And, again, McGinn tells us that the aim of teleosemantic theories of content was to
show how “teleology turns into truth conditions” (McGinn 1989, p. 148).
All of this connects with what Fodor claims is a truism – that “the mind’s main concern is
not acting but thinking, and that paradigmatic thinking is directed to ascertaining truths”
(Fodor 2008, p. 8). It follows that if the mind’s main business is to ascertain truths, and if
mental representations are the tools for conducting such business, then they must aim at truth.
Thus contents must be at least truth-apt. Ascertaining truth is typically a risky affair. A
formal requirement on genuinely representational contents of this sort is that they might be
false. Representation always admits of the possibility of error or misrepresentation. Some
hold that certain types of mental states, such as perception and memory, are factive (Hopkins,
forthcoming). Such states necessarily reflect the facts. They come with a cast-iron epistemic
guarantee: if a token mental state of that kind represents it as the case that p, then it is the
case that p. Here’s the rule: If it should turn out that the content of the state in question does
not represent the facts then we are not dealing with a mental state of that kind. Note that
7
factive here qualifies the mental states in question, not their contents. To be a
representational content that expresses p always allows for the possibility that not-p.
Now if this were the received view in orthodox cognitive science about the kind of
content possessed by mental representations, which play a role in the information-processing
challenge, then it would be directly at odds with the kind of enactivist approach that
Thompson defends. But things are not so simple.
The first thing to note is that linguistically-mediated beliefs and utterances are conceptual
representations. They represent instensionally (with an ‘s’), under guises; they represent
things as this-or-that. As Fodor (2008) insists “if a symbol represents a such and such, it must
represent it as something or other” (p. 178). However, many theorists believe in the
existence of non-conceptual content. This is traditionally understood as a kind of
representational content which presents the world as being a certain way, despite the fact that
the creature or system doing the representing lacks the concepts that would canonically
express the content in question. Now it’s a nice question whether we can really make sense
of the idea of a kind of truth-conditional content that is non-conceptual. We might worry that
we can’t so long as we think that intensionality (with an ‘s’) and concepts necessarily go
together.
However, this worry can be avoided if there is a kind of content that is subject to norms
other than those to do with truth and falsity. The idea that this might be so is gaining in
popularity. Thus, Crane (2009) rejects what he calls the ‘propositional attitude thesis’ about
perception without surrendering the idea that perceptual states possess representational
content. Instead, he claims such states have accuracy and correctness conditions that are not
any kind of truth conditions. A main motivation for this is the observation that accuracy and
correctness come in degrees whereas truth or falsity do not. Crane compares experiences to
pictures in this respect. Like pictures, he holds, experiences can be more or less accurate, but
they are not intrinsically true or false. Nor can pictures stand in logical relations. Thus
although Crane thinks they have a kind of representational content he denies it is of the truthconditional sort.2 They have a kind of content that is more primitive, more basic than that had
by propositional attitudes such as beliefs.
Fodor (2008) has advanced a similar line. He observes: “pictures don’t have truth
conditions. In the root case, for a symbol to be true it has to pick out an individual and
property and predicate the latter of the former; but iconic representations have no way to do
either. So, the camera doesn’t lie, but nor does it tell the truth” (p. 175-176). This is his way
of making room for the “possibility that some mental representation is nonconceptual” (p.
8
179, emphasis added). He promotes the idea that there can be representing that isn’t
representing as. Thus “X represents Y insofar as X carries information about Y, where
‘carries information about …’ is read as transparent … [this allows] for representation that’s
not ‘under a description’” (p. 179).
The idea that being an information-carrying state suffices for being a mental
representation with nonconceptual content is a radical departure from standard thinking about
what is minimally required to qualify as a representational state. As noted, many now
question the idea that content is necessarily truth-conditional. We are told that “it isn’t
apparent that an intentional state, event, or object about something other than a state of
affairs should be evaluated in terms of truth/falsity” (Gunther 2003, p. 5). But it is generally
held that “what is true of any state (event, experience, and so forth) with content, is that it is
governed by semantic normativity. For whether its content is conceptual or nonconceptual,
propositional or not, an intentional state presents the world as being a certain way; and
intrinsic to this presentation, to its content, is a set of (semantic) conditions under which it
does this correctly, truthfully, satisfactorily, appropriately, skillfully, and so on” (Gunther
2003, p. 5-6). On this view, “semantic normativity is the mark of intentionality” (Gunther
2003, p. 6). We can assume, along with Millikan (2005) that the normativity at play here is of
the non-evaluative (e.g. possibly of the merely biological) sort. But even if symbols and
concepts are not involved or crunched, and even if this isn’t any kind of intensional (with an
‘s’), truth-conditional representing, there is more going on than being in states that merely
carry or contain information in the sense Fodor describes. What appears to matter to those
who invoke the normative requirement is the character of the organism’s response,
understood as involving some kind of norm other than truth. I will pick up this thread in a
more positive vein in the next section.
Fodor’s proposal, however, is apparently much more radical than this. It allows that there
can be mental representation with content but without ‘semantic normativity’ in the reduced
sense just described. Simply being in a state that registers information suffices for
representing (but not representing as). But there’s a tension in his calling nonconceptual
states with this sort of content mental representations. For Fodor (2009) tells us that “The
mark of the mental is its intensionality (with an ‘s’) that’s to say that mental states have
content; they are typically about things … only what’s literally and unmetaphorically mental
has content”. If this is taken to mean that intensionality (with an ‘s’) is a minimal, necessary
requirement for genuine mentality then it turns out that states bearing only informational
content not only lack semantic properties, they are not truly mental. This fits with Fodor’s
9
earlier verdicts about the sorts of creatures that belong to the class of cognizers or
mentalizers; “wherever precisely the line is to be drawn, and however thick it may be, it is
vastly plausible that we fall on one side and the paramecium fall on the other” (Fodor 1986,
p. 12). I shall pick up this thread more positively in the next section too.
Frankly, all of this makes my puzzler sore. Content, it seems, is a bit like Christmas: it can
come without truth conditions, without concepts, without intensionality (with an ‘s’), without
semantics, without mentality. Content, I guess, means something more. It would be nice –
very nice, indeed – if there were some agreed, unequivocal, non-slippery and fully stable
understanding of just what ‘content’ and ‘representation’ are; given that mainstream
cognitive science apparently depends on these notions so heavily. It is perhaps too much to
say that the history of the use of these terms is shrouded in equivocation, equivocation,
equivocation. But it is no exaggeration that despite their utterly foundational importance to
orthodox cognitive science they are extremely elastic, far from unambiguous and not yet very
well-understood.
One thing is certain. There is no point in looking to our pre-scientific folk intuitions to
decide the matter. As Jackson and Pettit (1993) point out, “‘Content’ is a recently prominent
term of art and may well mean different things to different practitioners of the art” (p. 269).
Nonetheless, Crane (2009) assures us that his usage of the term, which deviates from the
propositional attitude rendering, corresponds to the way that many professional philosophers
use it. To back that up he cites the pedigree of this usage, reminding us that the notion of
content belongs with the theory of intentionality that Brentano offered us. But, be this as it
may, it does not by itself provide us with a robust and perspicuously clear understanding of
content or its properties, nor even any ready way to demarcate it from other phenomena.
Indeed, as Crane (2008) has recently admitted in response to a challenge by Nes (2008),
making sense of intentionality itself seems to require calling on the notion of representation
to do foundational work. He says: “It is the notion of representation, I think, that will
distinguish intentionality from … other phenomena” (p. 216). If so then we must call on our
notion of representations to do important work in demarcating intentionality. But presumably
we need a notion of content in order to distinguish representational phenomena from nonrepresentational phenomena. And, as we just noted, a notion of intentionality is needed to
help us understand the notion of content. So it seems that trying to make sense of these
notions in this way is to move in a rather tight, and seemingly incestuous, circle.
And we must be on our guard here for another reason. Brentano’s understanding of
intentionality is complex; it embeds more than one notion – he speaks not only of being
10
directed at objects but also of such objects having intentional inexistence (see Menary 2009
for an excellent exegesis). There are different ways of understanding these ideas in today’s
context. Yet those who look to Brentano for a lead on the nature of intentionality typically
end up making appeal to a notion of representation or content that is modeled directly on the
kind of semantic content (whether truth conditional or referential) associated with
linguistically mediated states of mind such as propositional attitudes. It is by this route that
many of today’s philosophers come to endorse what I will call the thesis of semantic
intentionality. Flanagan (1991) supplies us with a neat reminder of how the standard thinking
goes on this issue in his discussion of the central tenets of James’ philosophy of mind.
The concept of intentionality is a medieval notion with philosophical roots in Aristotle
and etymological roots in the Latin verb intendo, meaning ‘to aim at’ or ‘point toward’.
The concept of intentionality was resurrected by and clarified by ... Franz Brentano …
Brentano distinguished between mental acts and mental contents. My belief that today is
Monday has two components. There is my act of believing and there is the content of my
belief, namely, that today is Monday … Beliefs are not alone in having meaningful
intentional content … Language wears this fact on its sleeve. We say that people desire
that [ ---- ], hope that [ ---- ], expect that [ ---- ], perceive that [ ---- ], and so on, where
whatever fills the blank is the intentional content of the mental act. Intentionality refers to
the widespread fact that mental acts have meaningful content …The fact that we are
capable of having beliefs, desires, or opinions about non-existing things secures the thesis
that the contents of mental states are mental representations, not the things themselves –
since in the case of unicorns, ghosts, devils and our plans for the future there simply are
not real things to be the contents of our mental states! On this interpretation, James, is an
advocate of what Jerry Fodor calls the representational theory of mind (p. 28, second and
third emphases added).
Going this way only takes us back to square one. Just as the notion of content is a term of
art so too is representation as Millikan (1993) reminds us, “the name ‘representation’ does
not come from scripture” (p. 103). The short exercise of this section is meant to remind us
that, after all, as Matthen (2006) helpfully notes, representation is “a new and controversial
concept … The natural home of this concept is in the study of communication between
agents who possess intentions and goals. It is not immediately clear how it can be extended
to states issued by automatic sub-personal systems” (p. 147). Thus relying on our intuitive
11
grasp of what these notions mean renders us none the wiser about the core nature of
intentional, representational or contentful properties.
4. A Fresh Start
It seems we need a different approach to these issues. I suggest putting aside our antecedent
philosophical commitments and intuitions about the nature of ‘representation’ and ‘content’
and working forward from an agreed and non-controversial understanding of the nature of
information, as the notion called upon in a variety of sciences. This should be the lowest
common denominator in driving our thinking about what is needed to understand basic
mentality and any advance beyond it requires justification. The question is – does basic
cognition or mentality require information processing?
Before we get started – just what is the most basic kind of cognition or mentality? One
suggestion is that “cognitive interactions are those in which sensory responses guide action
and actions have consequences for subsequent sensory stimulation, subject to the constraint
that the system maintain its viability. ‘Sensory response’ and ‘action’ are taken broadly to
include, for example, a bacterium’s ability to sense the concentration of sucrose in its
immediate environment and to move itself accordingly” (Thompson 2007, p. 125). There
seems no good reason to rule this out as an instance of cognition or mentality, albeit basic,
other than attachment to the idea that true cognition or mentality must involve symbols and
concepts. But, as we saw in the previous section, even champions of a more restrictive
understanding of cognition have apparently begun to waver on this point. If so, then as
Thompson (2007) argues, the thesis of deep continuity between life and mind is secure, so
long as we adopt a liberal understanding of autopoiesis as “internal self-production sufficient
for constructive and interactive processes in relation to the environment” (p. 127).
Any living creature capable of this will need to be informationally sensitive (see Hutto
1999, ch 2 &3, 2008, ch. 3). But it doesn’t follow that they need to process information – if
this means that information is some sort of commodity that is in some way contentful, as
such talk appears to suggest. Godfrey Smith (2007) distinguishes two senses of information,
a weaker and stronger one. The weak notion is the familiar one that derives from the work of
Shannon and which has played a pivotal role in the development of communication
technology. It assumes that informational relations are nothing more than covariance
relations; they exist wherever correlations between facts, events or properties obtain. Let us
12
call this the information-as-covariance notion. As Godfrey Smith notes this conception of
information is ‘unproblematic’ and does not require much philosophical attention.3
It has a richer cousin that is much more controversial. It is referred to as ‘semantic’ or
‘intentional’ information, the kind of contentful information – the message – that some
communications convey. That notion significantly adds to the basic Shannon notion. Let us
call it the information-as-content notion.
There is a real danger of conflating these two notions. Jacob (1997) tells us that “the
relevant notion of information at stake in informational semantics is the notion involved in
many areas of scientific investigation as when it is said that a footprint or a fingerprint carries
information about the individual whose footprint or fingerprint it is. In this sense, it may also
be said that a fossil carries information about a past organism. The number of tree rings in a
tree trunk carries information about the age of the tree … In all of these cases, it is not
unreasonable to assume that the informational relation holds between an indicator and what it
indicates (or a source) independently of the presence of an agent with propositional attitudes”
(Jacob 1997, p. 45, emphasis added). To stress this last point, he adds that “the information
or indication relation is going to be a relation between states or facts … It is an ‘objective’
relation” (p. 49-50, emphasis added).
There is no doubt that information-as-covariance is an objective relation. And as the
quotation suggests it has wide currency in a number of sciences. But to talk of informational
semantics and to speak of indication when describing it, as Dretske (1981, 1988) does, courts
confusion with its richer, sister notion of information-as-content. As Cummins et al (2006)
point out, “‘indication’ is just a semantic-sounding word for detection” (p. 200). This being
so equating information and indication relations is doubly problematic – not only is there a
risk of smuggling in a notion of semantic content where it does not belong, there is also the
fact that the idea of detection undermines the idea that the information in question is a purely
objective relation; it makes no sense to talk of detection in the absence of an agent (or
equivalent) that does the detecting.
This highlights something important. It suggests that if we stick only with the weak
notion of information then there are no grounds for thinking that the world, standing apart
from agentive systems, contains anything that could be called informational content. To see
what’s at stake it helps to consider Cummins et al’s (2006) attempt to expose a special
problem for teleosemantics, by noting that we really have no choice but to believe in the
existence of unexploited content, content of a sort that must exist independently and logically
prior to the capacity of systems to make use of it. Apparently, this is a problem for
13
teleosemantic theories because they insist that natural signs or signals lack representational,
semantic content until a consuming response is selected for, one that governs a system’s
reactions. The worry raised is that cognitive systems are surely able to come to use
previously unexploited content – either by individual learning or evolution in the species. But
this apparently presents teleosemantics with a conundrum for it “requires content to pre-date
selection and teleosemantics requires selection to pre-date content” (p. 199).
Although interesting in its own right, it is not my purpose here to review the argument
advanced by Cummins and co. against teleosemantics. Rather I want to highlight an
observation they make at a crucial juncture when considering possible replies. For it is
important to their argument that the content in question is that allegedly contained in
representations used by cognitive systems. Thus they write: “A very natural response is to
say that unexploited content isn’t really content. After all, there is a lot of unexploited
information in the environment, information that cognitive systems must acquire abilities to
exploit. We do not call that information content” (p. 204, emphases original).
But in cases of basic mentality – that of paramecia – this is exactly the kind of situation
we have to deal with. Agents are interacting in reliable, informationally sensitive but noncontentful ways with their environments. Why non-contentful? Well, as we have just seen, if
we stick to the information-as-covariance notion there is no content out there for them to
interact with (let alone register, pick up, and so on).
If there’s no objective content in the world, then perhaps content comes into being along
with the activity of agents or consuming systems. This idea lies at the heart of
teleosemantics. As Millikan (2006) says “the content of a representation is determined, in a
very important part, by the systems that interpret it” (p. 100). But, once again, the metaphors
can mislead. We should not think of agents as content-consuming systems – for this suggests
that there is already pre-existing content to be consumed; and we have just ruled that idea
out. Perhaps then we should speak of content-creating systems instead.
That’s a step in the right direction, but note – now we are getting very close to the
enactivist story. After all, Thompson (2007) maintains that “Cognition is behaviour in
relation to meaning and norms that the system itself enacts or brings forth on the basis of its
autonomy” (p. 126, emphases added). There is doubtless much to be learned and perhaps
salvaged from teleosemantic theories – which are widely regarded as the best, if still
imperfect, attempts to naturalize representational content. Perhaps, with modifications, such
accounts might help to augment autopoietic enactivism. Some contemporary proponents of
enactivism believe that the basic idea requires supplement by appeal to additional notions.
14
We are told, “It is a mistake to take the theory of autopoiesis as originally formulated as a
finished theory … autopoiesis leaves many questions unanswered. In particular, several
essential issues that could serve as a bridge between mind and life (like a proper grounding of
teleology and agency) are given scant or null treatment in the primary literature” (Di Paolo
2009, p. 12).
Even if we accept this, it is unwise and unnecessary, for the reasons gestured at above, to
buy into the teleosemanticists’ semantic ambitions. Indeed, despite initial optimism, many
now doubt that attempts to naturalize semantic content have any chance of success. GodfreySmith (2006) provides an astute assessment “there is a growing suspicion that we have been
looking for the wrong kind of theory, in some big sense. Naturalistic treatments of semantic
properties have somehow lost proper contact with the phenomena” (p. 42). Nevertheless, he
also acknowledges that the driving idea behind teleosemantics – that evolved structures can
have a kind of ‘specificity’ or ‘directedness’ – is essentially correct: “there is an important
kind of natural involvement relation that is picked out by selection-based concepts of
function. But this relation is found in many cases that do not involve representation or
anything close to it” (p. 60).
What should we make of this? To quote a famous Rolling Stones lyric, “You can’t always
get what you want, but if you try sometimes, you just might find, you get what you need”.
Teleosemantic accounts fail to provide an adequate basis for naturalizing semantic or
intensional (with an ‘s’) content but they are proceeding along the right lines. Crucially, with
adjustment they provide serviceable tools for making sense of something more modest – i.e.
organismic responses involving intentionality (with a ‘t’).
What if in the place of teleosemantics we put teleosemiotics? Teleosemiotics is an order
of ‘teleosemantics – hold the semantics’. Teleosemiotics borrows what is best from
teleosemantics and covariance accounts of information to provide a content-free naturalistic
account of the determinate intentional directedness that organisms exhibit towards aspects of
their environments (Hutto 2008, ch. 3). Yet unlike teleosemantics, it does not seek to
understand the most basic forms of directedness, such as registering, in semantic, contentful
or representational terms. Such modes of responding are not to be understood as contentinvolving or even content-creating to the extent that these notions are understood in terms of
reference or truth conditions.
Compare this with Thompson’s discussion of virtual milieus, vital norms and meaning as
essential features of cognition, as inspired by Merleau-Ponty’s The Structure of Behaviour.
As he stresses, the even bacterial cells, the simplest life forms on earth (where life is
15
understood autopoietically) have needs that are fulfilled by deriving nutrition from the
environment. They achieve their ends by ingesting sucrose – an environmental feature.
Following Merleau-Ponty, Thompson (2007) holds that the property of being a nutrient is a
virtual property, not something found ‘objectively’ in the environment. Rather “it is enacted
or brought forth by the way the organism, given its autonomy and the norms its autonomy
brings about, couples with the environment” (p. 74). Thus, “sucrose has meaning and value
as food but only in the milieu that ‘the system itself brings into existence’ or ‘constitutes for
itself’” (p. 74). The whole point is that “Behaviour … expresses meaning-constitution rather
than information processing” (p. 71).
This way of talking will seem alien, or misplaced, to many analytic philosophers but
Thompson is quite clear that he is concerned with norms in “the biological sense of norms”
(p. 75). Moreover, the ‘meaning’ and ‘value’ that are brought forth neither constitute nor
depend on semantic content. Nor does it imply the existence of representations in anything
like the standard sense. We are told “if we wish to continue using the term representation,
then we need to be aware of what sense this term can have for the enactive approach …
autonomous systems do not operate on the basis of internal representations … they enact an
environment” (p. 58-59). Substantively I agree, but it will only breed confusion to use terms
like ‘meaning’ and ‘representation’ to describe the cognitive antics of bacteria – hence I
prefer the more austere teleosemiotic talk of informationally-sensitive responses to natural
signs. But this comes to much the same thing.
Can a modified teleosemantics – i.e. teleosemiotics – serve as a secure point of contact
between what phenomenologically-inspired enactivists have to offer and some of the best
work in the analytic tradition on theories of content? Making allowance for differences in
language and connation I think the answer is clearly ‘Yes’. However, there are more
important twists in this tale.
Thompson (2007) rejects the idea of evolution as driven by external forces, such as
natural selection. This is not to say he denies the existence or importance of natural selection,
only that he objects to the standard interpretations of it. So, this is something that
philosophers of biology need to debate – I say no more about it here. Rather I want to focus
on another aspect of his view of ‘enactive evolution’, one that has a direct bearing on the
issue at hand. It is his rejection of the idea that organisms are “systems that have atomistic
traits as their proper parts” (p. 203).
To accept this requires surrendering the idea that we should understand the intentionality
of biologically basic cognitive systems as being a property of their individual mental states.
16
While this will no doubt shock some – and goes against the grain of the standard way of
talking in analytic philosophy of mind – there is much to be said for it. For on reflection there
is every reason to think that the intentionality exhibited by basic cognizers differs
significantly from that of beings whose thinking or perceiving takes the form of, or involves,
propositional attitudes. To think otherwise would be to imagine intentionality as a property of
individual mental states – i.e. states of mind that bear special kinds of mental content. This
would be to model such states of mind on isolated words or sentences in the heads of
thinkers, however weakly. To be wholly free of this idea we ought to think of the intentional
directedness as simultaneously focused on both virtual and actual worldly targets and as
involving the goal-directed activity of the whole organism. Consider that, in the style of
David Attenborough, we say of baby, Sheba or Rover that they are trying to do this or
achieve that. Moreover, we say that they succeed or fail because of what they know, think,
notice or sense. Notice – it is the activity itself, and not some sub-part of it, that we can
coherently regard as being successful or not. Deciding if it is, or not, requires appeal to some
set of norms that specifies the goal in question. For this we must make appeals to the
creature’s evolutionary history, individual learning or the norms of an established practice,
and so on.
Whether a bit of goal-directed organismic activity succeeds, or not, depends on whether
certain facts obtain. Well-designed organisms have many (and often quite complex) means of
responding to the natural signs of environmental correspondences that are important to them.
Responding to such signs is meant to guide their behaviour with respect to the state of the
world so they succeed in their activities. And, if they are well-built and conditions are
normal, their activities non-accidently succeed often enough to fulfill their needs.
All of this can be true without it being the case that some sub-part of the organismic
system – e.g. an internal mental state – contentfully represents some part of the external
world correctly or incorrectly by saying that it stands thus or so. Indeed, in very basic cases
there is no principled basis for picking out one segment, or part, of a much larger organismic
response to some external natural sign as a discrete, contentful state of mind that represents
some more distal state of affairs. In normal conditions it is the totality of an organism’s
response that ensures the non-accidental success of its activities. As such, it is the attitude of
the whole organism engaged in such activities that exhibits intentional directedness. It is the
response as a whole that targets certain aspects of the world and not some sub-part of the
response. If so, it must be possible to be intentionally directed without having discrete mental
states that possess any kind of mental content at all.
17
I call such non-contentful but world-directed attitudes – intentional attitudes. They are to
be contrasted with properly contentful, sententially-mediated propositional attitudes, such as
truth-conditional beliefs and desires. The attitudes of the latter sort do possess semantic
content and linguistic structure. Indeed, I have long held that our “ordinary concept of belief
ranges over cases which, from the philosophical point of view, we should distinguish as
instances of beliefs-as-propositional-attitudes and beliefs-as-intentional attitudes’ (Hutto
1999, pp. 109–110). To have a content-involving thought, it is not enough for an organism to
be merely intentionally directed at a situation or state of affairs, even in the sorts of complex
and systematic ways intimated above. A creature could engage in many highly sophisticated
activities while only having attitudes of an intentionally directed sort that are to be
understood in purely non-intensional (with an ‘s’) terms.
5. Conclusion
Adequately responding to the Information-Processing Challenge requires acknowledging the
special importance of the informational sensitivities of sentient and sapient systems and
understanding these correctly. This requires resisting the temptation to think of information
as a kind of contentful, object-like, commodity, or to otherwise assume that basic mentality
depends on the manipulation of content-bearing mental states.
Where does this leave us? On the positive side, this conclusion is consistent with
accepting that cognitive systems exploit the relations of covariance that hold between
environmental states of affairs in various ways. This idea is consistent with – and, indeed,
inspired by – the idea that well-fashioned organisms are responsive to natural signs and that
these can guide actions successfully (in historically normal conditions), even if such signs do
not supply the creatures’ cognitive mechanisms with contentful information and even if the
signs are left semantically uninterpreted.
That said successful action requires informational sensitivity and a kind of responsiveness
to natural signs that introduces asymmetries (e.g. an organism’s sensitivity to one state of
affairs enables it to respond appropriately to a more distal state of affairs – in historically
normal circumstances). To be sure, organismic actions do not always succeed. But the mere
possibility of worldly misalignment does not imply (and need not be explained in terms of)
the existence of semantic relations of truth and reference.
18
How does this help the prospects of enactivism? The appeal to teleosemiotics shows that
the Information-Processing Challenge can be defused by enactivists without abandoning their
core motivating insight.
Now, it might be thought that even if contentful representations are not necessary for
understanding basic mentality, surely this cannot be the whole story about cognition. True.
But we can go a long way (even if not the whole way) in making sense of very complex,
elaborate and sophisticated worldly engagements without assuming they are either
constitutively contentfully or contentfully mediated.
The great bulk of living, thinking organisms act successfully by making appropriate
responses to objects or states of affairs in ways mediated by their sensitivity to natural signs.
But this does not involve contentfully representing those objects or states of affairs as such or
even representing them non-conceptually. Basic forms of mentality depend on informational
sensitivity and response, sometimes of a quite sophisticated variety, not processing
informational content.
Undoubtedly, some states of mind exhibit semantic intentionality – propositional attitudes,
for instance. They are properly contentful. Nevertheless, a great deal of sophisticated, worlddirected cognition exhibits intentional directedness that is not contentful in the sense just
discriminated. Teleosemiotics understands on-line perceptual responding as informationally
sensitive but it rejects the idea that this equates to a purely informational kind of
nonconceptual representing. It denies that such responding constitutes “a way of representing
X without representing it as anything” (Fodor 2008, p. 182).
Radical Enactivism – of the sort that both Thompson and I promote – explicitly rejects the
idea that content, whether informational or representational, is an inevitable ingredient in the
process that enables basic mentality. In this, not only is this brand of enactivism wholly in
line with the spirit of the original and most philosophically challenging conception of
enactivism, it is independently well-motivated.
Surely, it will be objected, some behaviour is too off-line, ‘plastic and flexible’ to be
explained without appeal to the manipulation of contentful propositional attitudes and
symbolic representations. Just how far can we go? Much further than is commonly thought, I
think, before we need to introduce anything like contentful states of mind into the picture
(see Hutto 2008, ch. 4 and 5). Still, there are obvious limit cases. Certain kinds of
deliberative planning – such as acting on the basis of considerations that are explicitly
represented – must be content-involving. We should accept that “the ability to think the kind
of thoughts that have truth-values is, in the nature of the case, prior to the ability to plan a
19
course of action. The reason is perfectly transparent: Acting on plans (as opposed to, say,
merely behaving reflexively or just thrashing about) requires being able to think about the
world” (Fodor 2008, p. 13).
There are two things to note about this remark. First, if the above arguments hold then we
should resist the idea that all worldly engagements that do not involve sophisticated,
contentful deliberation and symbol-crunching are nothing but bits of non-cognitive reflex or
‘thrashing about’. Secondly, we must ask exactly who – i.e. which cognitive beings – are
capable of reflective planning and deliberation? If that class contains only us – i.e. adult,
linguistically competent and typically developing human beings – beings who have benefited
from a wealth of specialized social scaffolding through engaging in communal practices –
then we will want a story about how we get in a position to do so. Thompson-style
enactivism promises to tell that story. It may be possible that we can account for the
contentful or meaningful basis of such activities without gaps and without having to believe
in the existence of underived mental contents of the sort that have resisted naturalistic
explanation for so long. Perhaps, after all, we have reason to believe the enactivist credo, that
practice logically precedes theory, and not – pace Fodor – the other way around. If so,
philosophy of mind gets a new lease on life.
References
Clark A. 2008. Supersizing the Mind: Embodiment, Action, and Cognitive Extension.
Oxford: Oxford University Press
Crane T. 2008. Reply to Nes. Analysis 68:215-18
Crane T. 2009. Is Perception a Propositional Attitude? Philosophical Quarterly 59:45269.
Cummins R, Blackmon J, Byrd D, Lee A, Roth M. 2006. Representation and
Unexploited Content. In Teleosemantics, ed. G MacDonald, D Papineau, pp. 195207. Oxford: Oxford University Press
Di Paolo EA. 2009. Extended Life Topoi: 9-21.
Dretske F. 1981. Knowledge and the Flow of Information. Cambridge, MA: MIT Press
Dretske F. 1988. Explaining Behaviour: Reasons in a World of Causes. Cambridge, MA:
MIT Press
20
Flanagan O. 1991. The Science of the Mind. Cambridge, MA: MIT Press
Fodor JA. 1986. Why Paramecia Don’t Have Mental Representations. In Midwest
Studies in Philosophy, ed. PA French, pp. 3-23. Minneapolis: University of
Minnesota Press
Fodor JA. 2008. LOT 2: The Language of Thought Revisited. Oxford: Oxford University
Press
Fodor JA. 2009. Where is My Mind? London Review of Books, 12 February.
Godfrey-Smith P. 2006. Mental Representation and Naturalism. In Teleosemantics, ed.
G Macdonald, D Papineau, pp. 42-68. Oxford: Oxford University Press
Godfrey Smith P. 2007. Information in Biology. In The Cambridge Companion to the
Philosophy of Biology, ed. D Hull, M Ruse, pp. 103-19. Cambridge: Cambridge
University Press
Gunther YH. 2003. General Introduction. In Essays on Nonconceptual Content, ed. YH
Gunther, pp. 1-19. Cambridge, MA: MIT Press
Hutto DD. 1999. The Presence of Mind. Amsterdam: John Benjamins
Hutto DD. 2008. Folk Psychological Narratives: The Socio-Cultural Basis of Understanding
Reasons. Cambridge, MA: MIT Press
Jackson F, Pettit P. 1993. Some Content is Narrow. In Mental Causation, ed. J Heil, A
Mele, pp. 259-82. Oxford: Oxford University Press
Matthen M. 2006. Teleosemantics and the Consumer. In Teleosemantics, ed. G
MacDonald, D Papineau, pp. 146-66. Oxford: Oxford University Press
Menary R. 2009. Intentionality, Cognitive Integration and the Continuity Thesis.
Topoi 28:31-43
McGinn C. 1989. Mental Content. Oxford: Basil Blackwell
Millikan RG. 1993. White Queen Psychology and Other Essays for Alice. Cambridge,
MA: MIT Press
Millikan RG. 2005. Language: A Biological Model. Oxford: Oxford University Press
Millikan RG. 2006. Useless Content. In Teleosemantics, ed. G MacDonald, D
Papineau, pp. 100-14. Oxford: Oxford University Press
Nes A. 2008. Are Only Mental Phenomena Intentional? Analysis 68: 205-15
21
Oyama S. 2000. The Ontogeny of Information: Developmental Systems and Evolution.
Durham: Duke University Press
Papineau D. 1987. Reality and Representation. Oxford: Oxford University Press
Ramsey WM. 2007. Representation Reconsidered. Cambridge: Cambridge University
Press
Seager W. 2000. Theories of Consciousness. London: Routledge.
Thompson E. 2007. Mind in Life: Biology, Phenomenology, and the Sciences of Mind.
Cambridge, MA: Harvard University Press
Varela FJ, Thompson E, Rosch E. 1991. The Embodied Mind: Cognitive Science and
Human Experience. Cambridge, M.A: MIT Press
Wheeler M. 2005. Reconstructing the Cognitive World. Cambridge, MA: MIT Press
1
For a discussion of this contrast see Thompson (2007) p. 80.
2
I am reporting Crane’s view, not suggesting that it is unproblematic. One immediate worry
about this proposal is that is hard to understand what it would be for an experience or picture
to be accurate simpliciter. It looks as if to be accurate is to be accurate in this or that respect.
And that would seem to imply that that a condition on being an experience or picture is that
the ways in which the picture or experience might be accurate or not would have to be
independently specificable. If so, that makes it look as for an experience or picture to be
accurate or not (to whatever degree) depends its being used for a particular purpose that
determines in what respect it might be so.
3
This notion of information has two features that make it of great value to the naturalist. It
doesn’t presuppose what it sets out to explain – i.e. the existence of semantic content and it is
a resoundingly non-mysterious, having put to good work in a number of hard sciences.
Following Wheeler (2005) we can say that it is entirely muggle – there’s nothing magical
about it.
22