Proper functions are thus eminently relational in that they make possible the interaction between the biological item and its environment. They are also historical (in the evolutionary sense of the term). In her Language, Thought, and Other Biological Categories (henceforth LTOBC), as well as in her introduction to White Queen Psychology, Millikan further characterizes proper functions F according to two conditions. Given a proper or Normal function F and a biological item B,
- A is a reproduction of some prior item that, because of the possession of certain reproduced properties, actually performed F in the past, and A exists because of this performance; or
- A is the product of a device that had the performance of F as a proper function and normally performs F by way of producing an item like A.
Condition (1) defines what Millikan terms direct proper functions. These are the proper functions which an item has when its very existence is the result of the same evolutionary processes which selected that function. In this way, direct proper functions are always inherited, traits which recur along a whole lineage: e.g. the heart’s functioning as a pump or the spermatozoon’s propelling the sperm cell are both direct Normal functions, as they constitute the proper function of every heart and every spermatozoon.
By contrast, condition (2) defines derived proper functions. These are proper functions which derive from direct Normal functions, and which, unlike the latter, may occur singularly. For instance, a particular “dance” performed by a given been in order to indicate the location of a flower field to the rest of the hive would be a Normal function derived from the direct Normal function of signalling the location of pollen through “dances”.
Millikan claims that intentionality is a proper function of cognitive systems – what cognitive systems have been designed to do, through millennia of evolutionary change and adaptation, is to establish intentional relations between their own inner states and things in the outer world. Intentionality is thus the direct proper function of cognitive systems, and intentional representations appear as functions derived from it.
MAPPING RELATIONS
The second aspect of intentionality is that related to mappings. If intentionality is defined as the property by which a given item a stands in a certain relation to another item b, the relation of intentionality can be described as a mapping from domain A onto domain B.
However, in the case of representations, it seems that we are dealing with a rather peculiar form of mapping. For the number of mathematically isomorphic mappings is potentially infinite – one can map the items in any given domain onto practically anything. However, representation is a mapping between, say, an inner state of mind S and a given item I such that it can be said to be correct or incorrect, true or false. Either there actually is an item I such as S is meant to represent, or there is not. This assessability of mental representations in terms of truth or falsity as often led, following Frege and Davidson, to characterize theories of meaning as theories of truth. However, as Fodor noted in his first “Psychosemantics”, this definition of representational meaning in terms of a correspondence theory of truth begs the question. According to this sort of theories, representations are true or correct if and only if there is some mapping function which maps those representations onto parts of the real world. But the question arises then of what it is that makes a representation true or correct. As Millikan puts it, echoing Fodor (for once),
If any correspondence theory of truth is to avoid vacuousness, it must be a theory that tells what is different or special about the mapping relations that map representations onto representeds. [LTOBC, 86-87]
Millikan’s overall aim is to account for this special quality of representational mappings in naturalistic terms, specifically by grounding it on the Normal character of representations, and thus on the proper functions of cognitive systems. In this way, and in general terms, what would make a representation true (or false) would be its being the result of the proper function of the cognitive system. But how does Millikan specify what counts as a representation?
REPRESENTATIONS AS INTENTIONAL ICONS
Bearing these two aspects of intentionality in mind, we are now in a position to understand Millikan’s rather striking claim that
[i]n the broadest possible sense of “intentionality”, any device with a proper function might be said to display “intentionality”. For the traditional earmark of the intentional is the puzzle that what is intentional apparently stands in relation to something else […] The general solution to the puzzle, I have suggested, is to see that intentionality is at root properness or Normalness. The intentional is “supposed to” stand in a certain relation to something else. [LTOBC, 94]
In this very broad sense, heartbeats would be intentional, insofar as the heart “is meant” by its very nature, or evolutionary function, to beat in the process of pumping blood. However, the “broad intentionality” which relates the heart to its heartbeats obviously lacks the specialness which we referred to when we mentioned the relationship between representations and representeds in the previous section. There is no sense in which it could be said that the heartbeats are “true” or “correct” with respect to anything.
Thus Millikan specifies a narrower (and more usual) sense of intentionality by which the heart indeed would not display intentionality. Entities displaying intentionality in this narrow sense – representations – would be characterized by the fact that they are meant to be interpreted or “consumed” by some “cooperating device”. Indeed, Millikan defines the processes by which representations are as consumer or benefit based (as opposed to such producer or stimuli based accounts of representation as Dretske’s). In this way, heartbeats would not be narrowly intentional, as the evolutionary function of the heart is to pump blood, not for its beats to be heard and interpreted, say, by a cardiologist. Millikan defines representations by means of the notion of the intentional icon. An intentional icon would be any device which is evolutively “supposed” to map onto the world in the special way mentioned above (as Millikan puts it, which maps thusly onto the world) in order to fulfil a direct proper function of the organism possessing it.
Representations are intentional icons the mapping values of the referents of elements of which are supposed to be identified by the cooperating interpreter. [LTOBC, 96. Italics in the original]
In this way, we need to bear three factors in mind when speaking about mental representations:
- the fact that representation is a proper function of the cognitive system;
- the fact that the relation between representations and representeds is a truth-conditionally assessable correspondence;
- the fact that representations are consumer-focused.
These three facts and their interconnection now enable us to characterize Millikan’s theory of inner representations. Cognitive systems have evolved so as to perceive certain features of their environments. This is the result of natural selection – those frogs endowed with cognitive systems which, by random mutation, were slightly attuned to the perception of flies would have had higher survival rates, and thus higher reproduction rates, than the frogs with less accurate cognitive systems. In this way, because of this evolutionary history the perception of flies became one of the proper functions of the frog’s cognitive system. The correspondence between the state of the frog’s cognitive state which means “fly” (or something like it) and the actual fly is thus guaranteed by the evolutionary origins of representation: the “fly” representation is a function of the cognitive system derived from its general proper function of accurate representation. The cognitive system may occasionally misrepresent under certain conditions – the frog may mistake a speck of dust for a fly – but, as a rule, the frog will represent a fly when there is actually a fly out there because its cognitive system is in itself the result of a selective process which ensured the survival of those frogs who accurately represented flies (among other things). The relation between representation and represented is thus assessable in terms of truth or correctness because that very truth or correctness is vital for the performance of the frog’s proper function, and thus, ultimately, for its survival and reproduction.
In the last term, it is this double evolutionary aim of survival and reproduction which grounds the whole of Millikan’s theory of representation, constituting the telos in teleosemantics. In this way, Millikan’s definition of representation – unlike Fodor’s – manages to steer clear from self-referential loops, as its ultimate basis is biological. Representations are consumer-interpreted intentional icons: that is, (a) they stand in a truth-conditionally assessable correspondence to their representeds because such is the proper function of the cognitive system which produces representations (and proper functions developed because they afforded their possessors higher survival and reproduction rates); and they are interpreted by their consumers for identical reasons.
Insofar as it claims that representation is a truth-conditionally assessable correspondence relation, Millikan’s theory is similar to Wittgenstein’s picture theory in the Tractatus. Representations must picture their representeds in some way in order to be true – that is, they must be isomorphic to the reality they represent. However, the trouble with isomorphisms, as mentioned above, is that they are potentially infinite and ubiquitous. Millikan solves this problem for (truth-functional) representational isomorphisms by specifying the constraints that (1) they must be the result of the cognitive system’s proper functions and (2) they must be interpreted by their consumers – two constraints which ultimately are grounded to the evolutionary summum bonum of surviving and reproducing.
In this sense, it might be said that Millikan’s move consists in the reduction of semantics to biology through a Darwinization of Wittgenstein’s picture theory of meaning. Even at the risk of reading too much into what might be no more than mere homonymy, I think that Millikan’s insight has been to link, through evolutionary teleology, both senses of “meaning” in English: its representational sense (by which a word means something) and its evolutionary-normative sense (by which an organ is meant to do something). Representations mean what they are mean because they are (evolutively) meant to.
DESIRE AND BELIEF
How does Millikan’s theory of representation apply to the desire/belief model of commonsense psychology? One of the shared features of desires and beliefs is the fact that they both have satisfaction conditions which are roughly equivalent to the truth conditions of sentences in Fregean semantics. Thus my desire to see Walter is satisfied if I actually get to see Walter, and my belief that the Earth is round is validated if the Earth is actually round, just in the same way as the expression “snow is white” is true if snow is actually white. In this way, the constraints which guaranteed and explained the “specialness” of the representational correspondence also apply here.
Millikan distinguishes between indicative and imperative intentional icons. Imperative intentional icons, or goal icons, claims Millikan, are those which
map in accordance with historically Normal mapping rules onto the configurations or world affairs that they produce when obeyed. [LTOBC, 99]
On the other hand, indicative intentional icons, or fact icons, map onto the world
in accordance with historically Normal mapping rules onto configurations or world affairs whenever they cause true beliefs in hearers in accordance with Normal explanations. [LTOBC, 99. Italics in the original]
This is interesting, because it would seem that desires are imperative intentional icons, as their satisfaction conditions obtain when the desire is “obeyed”. The proper function of desires is “to produce a state of affairs onto which it maps in accordance with certain mapping rules.” (LTOBC, 140). Beliefs, however, appear to be more complex: Millikan describes them as indicative intentional icons – but, by her definition, indicative intentional icons cause beliefs in turn. So it would seem that beliefs, as Millikan puts it, “are supposed to participate in inference processes”.
CRITICISMS
One criticism to Millikan’s theory might be her (to my mind) lack of clarity concerning the increasing complexity of representations – e.g. desires and beliefs – in the human species. Even though she does accept the particular character of human mental representations as compared to other species (the fact that human thought often seems to take a propositional form, for instance), I find no convincing account of this particularity in her theory (although, of course, it is highly probable that I may have missed it altogether. LTOBC is quite a book to digest).
Another possible criticism to Millikan’s teleosemantics would be the place (or the lack thereof) it ascribes to consciousness. According to the now famous Swampman argument first formulated by Davidson, a randomly materialized creature with exactly a physical configuration type-identical to mine would have intentional, meaningful mental states.
Against this “Swampman intuitionism” Millikan and others have claimed that intentionality and meaning supervene on the proper functions of the organism in question, and such proper functions are determined by the evolutionary history of the organism. The poor Swampman, lacking such evolutionary history (in Davidson’s account, it would have been created by the molecular transfiguration of a lightning-struck tree trunk), would have no proper functions, and therefore no mental representations, no intentionality, no meaningful mental content. However, grants Millikan, the Swampman’s mental states would have a purely phenomenal content, insofar as it would be able of perception.
It has been suggested – by David Chalmers among others – that it may not be really possible to give an account of intentionality which does not refer to consciousness, and vice versa. Certainly, it does seem rather counterintuitive to see intentionality as (exclusively) ethologically motivated. Moreover, it is not very clear what the teleological purpose of consciousness might be – what proper function of our cognitive systems is fulfilled by our being conscious?
However, as we shall see, Millikan conceives of intentionality in a broad and a narrow sense, and in it will be the narrow, more intuitive sense of the word that we will be concerned with in dealing with representations.
Which of course does not mean that desires are always fulfilled – as Millikan strikingly puts it, “many desires, like sperm, emerge in a world that does not permit their proper functions to be performed”.