Millikan and Her Critics

Dan Ryder, Justine Kingsbury, Kenneth Williford, Dan Ryder, Justine Kingsbury, Kenneth Williford

• Publisher: Wiley-Blackwell
• Year: 2012
• Print ISBN: 9780470656853, 9780470656846
• eBook ISBN: 9781118328088
• Edition: 1
• Topic: Philosophy
• Subtopic: Philosophy & Ethics in Science

Millikan and Her Critics offers a unique critical discussion of Ruth Millikan's highly regarded, influential, and systematic contributions to philosophy of mind and language, philosophy of biology, epistemology, and metaphysics. These newly written contributions present discussion from some of the most important philosophers in the field today andinclude replies from Millikan herself.

• Comprises 13 new essays that critically examine the highly regarded and influential work of Ruth Millikan
• Covers a wide range of Millikan's most important work, from philosophy of mind and language to philosophy of biology
• Features contributions by some of the most important and influential philosophers working today
• Includes original replies to critics by Millikan

1

TOWARD AN INFORMATIONAL TELEOSEMANTICS

KAREN NEANDER

One might think that a natural place to look for mental content is in the information ­carrying functions of mental representations. One might even think that the proponents of teleosemantics would especially like to look there. Fred Dretske (1986) does look there. However, Ruth Millikan and David Papineau, who have most extensively developed and defended teleosemantics, are critics of this approach. They deny that representations have the function to carry information or that representation producers have the function to produce representations that carry information. In my view, they are wrong, and once we see why they are wrong we can develop an alternative approach, an informational yet teleosemantic theory of content. This chapter makes a start with the contents of sensory representations.

Introduction

Informational theories use a “natural” (as opposed to intentional) notion of information.¹ The starting idea is that sensory representations and perhaps other perceptual, conceptual, and belief-like states, carry information about things in the world insofar as they are caused by them, are correlated with them, or make their presence more probable. On this type of view, the “aboutness” of content is thought to derive from the “aboutness” of natural information. Teleological theories of mental content instead (or possibly, as well) appeal to functions that are given an etiological analysis. A function of something is what it was selected for.² Content is said to be “normative” because representations can ­misrepresent. And these functions are said to be “normative” too because items with such ­functions can malfunction. For teleosemantics, the core idea is that the “normativity” of content derives from the “normativity” of functions.

Each idea is further refined in different ways by different people but, at a first pass, teleosemantic theories, by invoking functions, are said to focus on the effects or results of representations, whereas informational theories, by invoking natural information, are said to focus on the causes of representations or on conditions that co-occur with their causes. The difference between the two approaches can thus seem to be one of output-based versus input-based views of the constitutive conditions for mental content. One might nonetheless think that the two approaches should be combined. A hybrid idea is that representation-producing systems have the function of producing states that carry ­natural information. Then the “aboutness” of content could be explained as originating with the “aboutness” of information and the “normativity” of content could be explained as ­originating with the “normativity” of functions. Here is the Holy Grail! A source of naturalistic normative aboutness! Or so one might think.

Yet Millikan scotches this idea. In her early criticism of informational theories,³ she says that “… [i]f ‘detecting’ is a function of a representational state it must be something that the state effects or produces. For example, it cannot be the function of a state to have been produced in response to something” (Millikan 1989a, 283).⁴ She then argues that a theory of content should ignore the production of representations altogether, in favor of their “consumption.” More recently, this emphasis on consumption versus production has been softened in her theory. Now Millikan speaks of co-adaptation between producing and consuming systems. However, her view is still that “… intentional signs often do carry natural information … [but] … it is not the purpose of an intentional sign to carry natural information” (VOM, 31). And, when discussing Nicholas Shea’s (2007b) argument for an input condition for teleosemantics, she approves his observation that, on her view, “… carrying information is not a function of a representation” and “… it is not a purpose of the producer system to produce items that carry correlational [or I would add [she adds] any other kind of natural] information” (Millikan 2007, 445).

David Papineau (1998, 3) agrees with Millikan about this and chides me for ignoring her point:

I would argue that Neander has taken insufficient note of Millikan’s point that representational content hinges on how the representation is used, not on what causes it. In her general discussion of teleology, Neander focuses, quite rightly, on the effects of biological traits. But as soon as she turns to representation she shifts to the question of what … it is supposed to detect…. From the teleological point of view this is to start at the wrong end. The teleosemantic strategy requires us first to identify which result the state is supposed to produce and then use this to tell us what it is representing. (Papineau 1998, 7)

I am not about to repent. This chapter explains why, despite being in broad sympathy with Millikan and Papineau, I believe that asking what a system is supposed to detect is to start in the right place.⁵ My plan is first to explain why there are what I call “response functions,” which are functions to respond to something by doing something (in section 2). Sensory systems, I claim, have response functions. Then (in section 3) I argue that ­sensory systems have the function of producing representations that carry natural information, on a certain (singular, causal) understanding of information that I recommend. Section 4 briefly considers the implications of this for what we can say about the informational functions of the sensory representations themselves (the argument in section 4 is not crucial). Then I give a two-part proposal toward an informational teleosemantic (IT) theory, which is for the indicative content of sensory representations. Section 5 explains how, on this version of IT, the problem of error is handled and section 6 explains how the problem of distal content is handled.

Response Functions

A system has a response function if it has the function to respond to something by doing something else. The term ‘response’ is used in a purely causal and non-intentional sense here; a system counts as producing a RED-state in response to red if some red (the instantiating of red) caused the system to produce a RED-state.

The first question I want to consider is whether there can be response functions. Against this, Millikan (2007, 447) claims that “The function of an item, in the teleo­logist’s sense, is always something it effects.” But, before hastening to agree with this apparent truism, bear in mind that the issue is not whether functions must involve effects, which they must, but whether they must be effects to the exclusion of input causes. Might not a sensory system have the function, in part, to produce certain inner states in response to certain features of or occurrences in the environment?

A philosophical problem about biological functions is often posed as the problem of elucidating which effect(s) of an item count as its function(s). Why is it the function of the heart to pump blood and to circulate it through the body? Why are these effects its functions, not its making a whooshing noise? It is assumed at the outset that functions are certain effects of the items that have them, or anyway that functions are certain effects that they have in appropriate circumstances when they are functioning properly and not malfunctioning. But let’s look beyond the initial posing of the problem of functions to consult our best solution to it.

The proponents of teleosemantics invariably agree that the best solution is the etiological theory, and this includes Millikan and Papineau⁶ and me. Leaving aside details that need not concern us now, the etiological theory says that functions are what things were selected for.

The details are variably elucidated, but note that phenotypic traits can be selected for the adaptive things that they do, and doings can involve input causes. Consider an analogous case first.

If I am waiting for my friends to phone to tell me if they’re coming to dinner, there are different things that I can do. I can pick up the phone at random or whenever I remember that my friends will call, or I can wait until it rings and pick it up only then. If I pick it up at random or whenever I remember that my friends will call, I’ll waste my time and possibly frustrate my friends’ efforts to get through. So doing one thing rather than another will serve my purposes better. And the crucial difference in these different things that I could do are differences in the input cause, or in other words in what I am responding to.

Now think about those biological functions that depend on natural selection. One type of mechanism might secrete melatonin in response to the dimming of light,⁷ while another type of mechanism might secrete melatonin in response to light brightening and a third more or less randomly. Since melatonin makes us sleepy, the first mechanism might be more adaptive in creatures like us who have poor night vision. So the first type of mechanism might be selected in preference to the second and third because the first differs in its input cause, in what triggers its production of melatonin.

It is a function of our pineal gland to secrete melatonin in response to the dimming of light. It is also a function of B-lymphocytes to produce antibodies in response to antigens, a function of certain mechanisms to produce shivering in response to the early stages of hypothermia, and a function of the pancreas to secrete insulin in response to elevated levels of glucose in the blood.

A visual system can have the function, in part, to produce REDs in response to red being instantiated. Or, more generally, there can be a range of determinate values, e₁ … e_n, of a determinable (E), and a range of determinate values, r₁ … r_n, of another determinable (R), and a sensory system can have the function, in part, to produce certain values of R in response to certain values of E. It can have the function to produce r₁ in response to e₁, r₂ in response to e₂, r₃ in response to e₃, and so on.⁸

To be sure, functions must also involve effects. For natural selection to occur, there must be downstream replication. But, to repeat, the issue is not whether functions involve effects but whether they can also involve input causes. And they can. Biological mechanisms are selected for their causal roles, which can include dispositions to respond to specific types of causes. Someone could stipulate that by the term ‘function’ she will refer to effects exclusively. But denying that there are response functions is meant to have ­significant consequences for how teleosemantics should be developed and mere stipulation is no motivation.

I am anyway not being revisionist. Let us take a look at some statements of the etiological theory. Larry Wright (1976, 81), for instance, says that “The function of X is Z iff: (i) Z is a consequence (result) of X’s being there, and (ii) X is there because it does (results in) Z.” A result of a sensory system “being there” can be its producing inner states in response to outer states and this can be, in part, why it is there (once we appropriately disentangle talk of types versus tokens, which we would anyway need to do here).

If we change the symbols for the sake of conformity, Millikan (1989b, 288) says that X has the proper function, Z, “… if X originated as a reproduction … of some prior item or items that, due in part to possession of the properties reproduced, have actually performed Z in the past, and X exists because (causally, historically because) of this or these performances.” Presumably, a performance could be a simple effecting. However, without an explicit ruling against this, it could also be an effecting of certain inner states in response to certain conditions obtaining.

In a paper titled “Functions as Selected Effects” I argue that functions are selected effects, thus apparently committing myself to the claim that I here deny. But my point there was that functions are effects of phenotypic traits for which the traits were selected rather than effects of traits that are currently adaptive. That is, the emphasis was on functions being selected effects, not selected effects. Of ordinary physiological functions, I say, “[i]t is the/a proper function of an item (X) of an organism (O) to do that which items of X’s type did to contribute to the inclusive fitness of O’s ancestors, and which caused the genotype, of which X is the phenotypic expression, to be selected by natural selection” (Neander 1991, 174). Again, response functions are not excluded. Something that a sensory system can do and be selected for doing is produce an inner state in response to a certain type of cause.

In short, that functions are selected effects is true enough for most dialectical situations, but don’t mistake the slogan for the theory. The etiological theory, as standardly formulated, and as it should be formulated in my view, does not rule out response functions.

That functions are effects and not causes is often taken to be too obvious to need arguing (Millikan’s and Papineau’s claims are often repeated by others). However, Shea (2007b) offers some motivation for the claim. He says:

Teleosemantics is sometimes mistakenly taken to be a refinement of informational semantics according to which items represent what they were designed by evolution to carry information about. The trouble with this gloss is that evolutionary functions are a matter of effects. Amongst the various effects that an evolved system can produce, those which have contributed systematically to the system’s survival and reproduction in the past are its evolutionary functions. Evolution acts only on effects. It is blind to the mechanism by which those effects are produced. An effect produced by a system at random will be an evolutionary function if its production has contributed systematically to the survival and reproduction of the system. It is not part of any evolutionary function for the effect to have been caused in a particular way. (Shea 2007b, 409; italics added)

Shea goes on to argue that we could nonetheless still supplement teleosemantics with an input condition, and this is an important point. Teleosemantic theories can use other ingredients, besides functions, when they give a recipe for thought. But in defending the possibility of an input-based teleosemantics, Shea concedes more than he needs to.

It is true that natural selection cannot discriminate between two mechanisms that have the same impact on fitness. If one type of pineal gland with associated structures had randomly secreted melatonin and had contributed to survival and reproduction just as well as a system that secreted melatonin in response to the dimming of light, then natural selection could not have discriminated between them. More generally, something that gets something done one way and something that gets exactly the same thing done another way are equal in the eyes of selection (all else being equal).⁹

However, there are cases where different things are done that have different impacts on fitness because they involve responses to different causes. It is safe to assume that a system that secreted melatonin randomly did not contribute to the survival and repro...

Table of contents

• Cover
• Series page
• Title page
• Copyright page
• NOTES ON CONTRIBUTORS
• FOREWORD
• A MILLIKAN BIBLIOGRAPHY
• INTRODUCTION
• 1 TOWARD AN INFORMATIONAL TELEOSEMANTICS
• 2 SIGNALS, ICONS, AND BELIEFS
• 3 MILLIKAN’S ISOMORPHISM REQUIREMENT
• 4 MILLIKAN ON HONEYBEE NAVIGATION AND COMMUNICATION
• 5 CONCEPTS: USEFUL FOR THINKING
• 6 PROPERTIES OVER SUBSTANCE
• 7 MILLIKAN’S HISTORICAL KINDS
• 8 MILLIKAN, REALISM, AND SAMENESS
• 9 CRANING THE ULTIMATE SKYHOOK: MILLIKAN ON THE LAW OF NONCONTRADICTION
• 10 ARE MILLIKAN’S CONCEPTS INSIDE-OUT?
• 11 THE EPISTEMOLOGY OF MEANING
• 12 WEASELS AND THE A PRIORI
• 13 ALL IN THE FAMILY
• AFTERWORD
• REFERENCES
• INDEX