AGAINST DERIVED INTENTIONALITY
David Cole
July 11,2009 – ~6200 words. light rev April 23, 2010; rev Sept 1 2010
Intentionality is a property of an important class of things: things that represent, or are about something. Thus a belief or sentence or story is about something, a painting or photo is of something, a sign is a sign of something, and a desire is a desire for something. These disparate things all display intentionality. They have content; they represent some state of affairs beyond themselves. The represented state of affairs need not be actual, and is not in the cases of false belief, unfulfilled desire, or Salvadore Dali painting.
Many hold a dualism: they hold that we must distinguish betweenderived intentionality and underived (or “original”or “intrinsic”)intentionality. The intentionality of written language is derived; the intentionality of mental states (e.g. beliefs) is intrinsic. The formerintentionality is importantly different from the latter phenomenon, and is secondary because derived intentionality depends upon underived intentionality for its existence.
As Fodor 2009 puts it:
“…the critically important fact [is] that the (derived) intensionality of what happens on the outside [of the head] depends ontologically on the (underived) intensionality of what happens on the inside. Externalism needs internalism; but not vice versa. External representation is a side-show; internal representation is ineliminably the main event.” [Fodor uses “intension” where some use “intention”.]
The distinction between original and derived intentionality is intuitive enough: Thoughts have original intentionality, while pieces of writing have derived intentionality – their intentionality or content exists solely in virtue of the original intentionality of the minds that create or interpret the writing. What’s true of writing is true of artifacts that represent generally – computers, clocks, weather gauges, etc. They have states that represent and hence have intentionality, but only derivatively. John Searle was one who first emphasized the dualism (Searle 1980, 1984). Thus philosophers of mind and language who are at loggerheads about most everything, such as Jerry Fodor and John Searle, agree on this intentionality dualism. Should we be worried? I think so.
If we do worry, we will not be entirely alone. Not all accept intentionality dualism. Daniel Dennett rejects it. Dennett embraces an intentionality monism, holding that all intentionality is derived. The intentionality of books and that of memories inside someone’s head do not differ significantly in this respect – intentionality of any sort requires taking an interpretive stance.
I believe both of these approaches to intentionality, dualism and Dennett’s monism, are mistaken. I will make a case that there is no derived intentionality, intentionality that essentially depends upon interpretation or the intentions of a designer. This leads me to believe there is one kind of intentionality, and it can be in or outside the head. I’ll broach this heresy by way of critical discussion of the arguments for derived intentionality. In the course of that I’ll pick apart a few strands of the tangled web of epistemic and semantic issues that I believe have undermined thinking clearly about intentionality. And I’ll sketch a positive account of intentionality that accords it a full stature as a feature of the world not generally dependent upon acts of interpretation or states of consciousness.
Intentionality Dualism
Let us begin by looking more closely at the distinction between original and derived intentionality as set out by Searle. Searle tells us that mental states (e.g. propositional attitudes) display original intentionality, while language only has derived intentionality.
As a paradigm example of original intentionality Searle 1984 (p. 77) gives us:
Robert believes that Ronald Reagan is President.
By contrast he offers the following as a clear example of derived intentionality:
“Es regnet” means it’s raining.
Of the quoted embedded German sentence he says:
“That very sentence might have meant something else or nothing at all”(p. 78).
To be sure, there is a difference between these examples. But on the face of it, the comparison displays a certain fault: there is a difference in the way the representations are picked out, or “individuated”. This is significant because the apparent difference between the intentionalityin the two cases may be nothing more than an artifact of this difference in description. In the first case, the example of “original intentionality”, the belief is being described by its content or semantics – the belief that it is snowing is essentially (“intrinsically”?) the belief that it is snowing. Change anything about its meaning, what it is about, and it would be a different belief. But this leaves open the possibility that the physical state in the believer’s brain that records and is the physical basis of that belief might, depending on the rest of the brain and the causes of the brain state, “have meant something else or nothing at all”.
Unlike Robert’s mental state, Searle characterizes the German sentence on the basis of its syntax. ThoseGerman words – the marks “es regnet” - might have meant something else or nothing at all. But if instead we use the same content-based individuated that Searle used for the belief, e.g. “Otto just said that it is raining” then we have also individuated language in terms of its semantics, parallel to the way Searle individuated the mental state. And Otto’s saying that it is raining cannot have meant something else, or else it would not be an instance of saying that it is raining.
Note that the same distinction between two methods of individuation can be made even with instances of signs that are not directly produced by a human or something with consciousness. My microwave dings. The ding means something, but the ding might have meant something else, or nothing at all. But if we say that my microwave signaled that its cooking time is complete, then that signaling can’t mean something else. As in Searle’s examples, the difference is not in the phenomena, but in the method of individuation.
Given their physical description, brain states might have had other, or no, content, as might the German words. Given their semantic descriptions, the examples of language and of mental states have their respective meaning properties essentially. Thus we have not at the outset been given clear evidence of a fundamental difference between the two cases, and hence motivation for intentionality dualism.
Intentionality monism
Dennett agrees with Fodor and Searle that language and other artifacts are paradigms of derived intentionality. But Dennett then goes on to argue that the requirement of interpretation that makes the intentionality of language derived applies also to inner states that Fodor and Searle have taken to have a different and more fundamental “intrinsic” form of intentionality. Regarding language Dennett agrees completely with the dualists. He says:
“The point about the dependent status of artifactual representations is undeniable. Manifestly the pencil marks in themselves don’t mean a thing. This is particularly clear in the cases of ambiguous sentences. The philosopher W. V. O. Quine gives us the nice example:
Our mothers bore us.
What is this thing about? … You have to ask the person who created the sentence. Nothing about the marks in themselves could possibly determine the answer. They certainly don’t have intrinsic intentionality, whatever that might be. If they mean anything at all, it is because of the role they play in a system of representation that is anchored to the minds of the representers. (Dennett 1997 “Intentionality” in Kind of Minds p. 51)
Where Dennett disagrees with intentionality dualists is in holding that all intentionality is of this derived type. And Dennett holds that the failure by intentionality dualists to appreciate this is a mistake of the utmost importance. He writes (Dennett and Haugeland 1987):
“I have argued in … The Intentional Stance (Dennett, 1987) that clinging to the doctrine of original intentionality is the primary source of perplexity in contemporary Anglo-American philosophy of mind…..”
However I believe that in Dennett’s treatment of intentionality we can discern a problem that afflicts many discussions of intentionality and representation. Intentionality dualists, as well as those monists who see all intentionality as derived,show a lamentable tendency to conflate epistemology and semantics. Consider Dennett’s argument that it is obvious that the quoted sentence has derived intentionality. The argument turns on the point that to know what an ambiguous sentence means, you have to ask the author. You can’t tell from the marks themselves what they mean.
These are epistemic complaints. How one can tell what something means, and what it means, are two importantly different things. One is semantic, the other epistemic. In fact, perhaps conflating these is in fact “the primary source of perplexity in contemporary philosophy of mind” - and philosophy of language and semantic theory. Consider natural signs. One can often figure out what something means by knowing what caused it. The world is such that some causes have unique effects – in the wild, smoke generally means fire. Other phenomena - spots on skin, say, to take a Grice example - can mean more than one thing. You can’t tell what the spots mean, what they are signs of, what disease is causing them in a particular case, without some more diagnostic efforts. But these epistemic problems have nothing to do with meaning, the significance of the symptoms and signs.
Even when a sign (e.g. a tokening of “Fire!”) is directly caused by human agency (still the typical case with language, even as machines become increasingly talkative), the meaning arguably comes from the conventional link with a state of the world (the presence of a fire). With a tokening of something we have not encountered before – “Gavagai”, perchance – it is not our possibly mistaken interpretation that determines the meaning, but rather, if it is meaningful, the meaning is determined by a conventional connection with the world. “Nothing about the marks in themselves” determines the meaning, to be sure, but that is not because meaning is derived, but rather because meaning always depends on factors external to “the marks in themselves”.
Let us return to consider Searle’s intentionality dualism for a moment. It seems likely that a conflation of epistemic and semantic issues may be at work in Searle’s later treatments of intentionality as well. Searle is struck by the fact that mental states such as one’s belief don’t require an interpretation step. They are experienced as having content; the conscious experience itself lets us know what the satisfaction conditions of our belief are. A written sentence may be ambiguous; a thought is not.
Three points: first, yet again this is an epistemic point, not a semantic point. Knowing what a state of my brain means is distinct from what it means. The intentionality – a semantic property of aboutness – is not determined by how I come to know about it. Even when a state is experienced as wearing its significance on its sleeve, that epistemic fact about the state, while undoubtedly interesting, is distinct from its intentionality.
Second, it is not clear, as Dennett notes, that this state of consciousness is not the result of a process of unconscious interpretation. Perhaps I am only conscious of states that I can interpret and indeed have unconsciously interpreted, but I have other representational stateswith content that I cannot interpret and so they necessarily remain unconscious. Perhaps I have more beliefs than I can express in language or that I can be conscious of. Perhaps, e.g., there are emotional states that are intentional but localized in e.g. the amygdala and not fully available to cortex and interpretation.
Third, it is not at all clear that this epistemic privilege is more than trivial. I knows what I knows – I can disquote my thoughts and have the satisfaction conditions manifest. My thought “it is raining” means that it is raining. This will always be true, and may cause wonder at the transparency of consciousness, but it is not clear what it buys. In particular, as again Dennett notes, there appear to be externalist considerations that count against any robust knowledge of satisfaction conditions of mental states. I think “water, water everywhere!” – now is this satisfied by a superabundance of H2O, or of XYZ? Turning inward does not afford an answer.
Back to intentionality monism. Dennett 1987 offers us two additional arguments that all intentionality is derived. Dennett and the dualists take it as given that derived intentionality is a feature of artifacts. In many cases, the source of the artifact is the source of the intentionality; intentionality is not intrinsic to states of an artifact. But, Dennett argues, WE are artifacts, designed by evolution. Evolution is (presumably) not a conscious agent, but it is the source of our design, including the design of our cognitive system. We are built by the Blind Watchmaker, and so if all artifacts have derived intentionality, then we do.
Dennett goes on to argue, against others who take evolution and design seriously (especially Dretske) that "If we are such artifacts, not only have we no guaranteed privileged access to the deeper facts that fix the meanings of our thoughts, but there are no such deeper facts.” [Dennett’s emphasis]
Dretske and Millikan have tried to find such content-determining “deeper facts” in a principled way by identifying what Millikan (1984) calls the “proper function” of representational states. The problem, Dennett claims, is that there is no fact of the matter about proper function. There is no answer from evolution as to whether frog visual systems are fly or moving dark spot detectors, or whether our visual bilateral symmetry detector is a predator detector, prey detector, mate detector, etc. Function is indeterminate in nature, and so assigning one function of the many possible is a matter of interpretation. This is true of evolved systemsgenerally, not just in the special case of representation in cognitive science. Hence the intentionality of cognitive states is necessarily based on an interpretation of their function – and this dependence of content on interpretation again makes the intentionality of mental states as derived as that of a text external to the head.
So it seems that on Dennett’s view, the intentionality of mental states is doubly derived. First, it depends on “Nature's purposes”, and intentionality that depends on a designer's purposes is derived. Second, Nature’s purposes (evolutionary factors) are too ambiguous and indeterminate to support specific attribution of function and content. And so interpretation is required to assign determinate content – and requiring interpretation is surely a sufficient condition for making the content derived intentionality. Let us call the first of these arguments, the argument that states that depend upon a designer can only have derived intentionality, the Argument from Design. Let us call the argument that artifacts have their intentionality only through an intelligent agent’s acts of interpretation, the Indeterminacy Argument.
As it stands, I don’t believe the examples we are given support the conclusion. I will focus on the Interpretation Argument. Dennett’s first argument, the Argument from Design, it seems to me, is not compelling. For one thing, Unintelligent Design is arguably not design at all. Hence it is not clear that we are artifacts in the usual sense, as a product of real agents and their intentions. To call everything that has survived a process of change an “artifact” is to stretch the concept beyond its ability to do the useful work we need it for in everyday life.
Second, artifact or not, it seems implausible that the intentionality of my and your current mental states turns on some ancient etiological fact about the species. Whether the first humans were designed by evolution, by God, by gods, or were the products of a lightning strike in SwampEden has, it seems to me, no bearing on whether our mental states now have original or derived intentionality. It is reasonable to hold that the first mutant creature with conscious perceptual states had states that carried information about and were used as representations ofits world (I’ll discuss this further discussion below). And no matter what our ultimate origins, all of us here now have a long pedigree of beings endowed with intentional states recursively begetting. Thus whether something is ultimatelydescended from an artifact is not in itself a determinate of whether its representing states have derived or original intentionality. Dennett sometimes seems to agree with this denial of a crucial difference between organisms and artifacts, but decides to go with “all intentionality is derived”. The path not taken, the neglected monism, comports better with the externalist considerations that Dennett sometimes raises.
I find Dennett’s second, Indeterminacy, argument to be more interesting. Dennett holds that all attribution of function or content requires interpretation. But at the outset there is something paradoxical about this position. What, after all, is interpretation? An interpretation is an interpretation of something, it must have an intentional object, and hence any interpretation is an intentional state. If all intentional states require interpretation, and interpretations are themselves intentional states, then we appear to have a pernicious regress.
Second, just how ambiguous really is the biological function of hearts? How indeterminate is the function of a frog’s visual system? How free are we to interpret and assign functions to these systems? There is a requirement of interpretation only if biological function is deeply indeterminate. Frogs notoriously will eat seductively dangled BB’s until they sink with lead-bellies through their lily pads to the bottom of the pond. So is the frog visual system a BB detector? A fly detector? A moving-black-spot detector? Millikan and Dretske would like to appeal to evolutionary history for a principled way of rejecting these alternate attributions of function, and to permit saying the frogs are making a mistake when they down a pound of BBs. Thefrog visual system evolved as a fly detector, not a BB detector. But, presses Dennett, what entitles them to say this, rather than say that the system is a moving-dark-spot detector?
It is not clear to me that these are rival characterizations. The frog visual system does incorporate a moving-dark-spot detector. In environments that prevailed during frog evolution, flies were the predominate generators of moving-dark-spots on frog retinas, and so the frog visual system alsofunctioned as a fly detector. But this second, it seems to me, is a privileged description. The key to its privilege lies in counterfactuals. Had the main source of moving-black-spots not been edible flies (say they had beenrobotic BBs), the frog would not have evolved the moving spot detector system, no matter how that system is characterized. So it is the fact that moving spots meant flies that led to this feature of the frog visual system. In general, animals evolve to have detectors of salient features of their environments. The detectors are perforce largely “indirect” – predator detection is via scent, sounds, and optical properties. But “predator” and “scent” are not rival characterizations of the detection system. Organisms detect prey or predator by their scent – one is more mediate than the other. This is at best a distinction between narrow and wide function, and is not a case of indeterminacy. And, crucial to the issue at hand, this entire functional chain exists apart from interpretation – frogs were detecting flies via the moving spots flies make on frog retinas long before any interpreter engaged in the practice of attributing content to frog brain states. The biological function of the states – what they did that promoted froggy gene pool enlargement - is not a product of interpretation. Clearly there is more to be said, but the moral I draw is that this particular worry about evolutionary function has not been shown to be a sufficient basis for holding that all intentionality is derived.