Some Thoughts on Millikan's Varieties of Meaning

---

Chapter 1

• Millikan argues that we have many and potentially conflicting purposes, and that there is no clear line dividing them into the special volitional actions of our whole person and some part of the person.
• Note that this also means that Millikan is arguing there is no clear division between, say, the kind of special volitional actions we might call the exercise of free will or at least some kind of autonomy, versus, say, our "biological" purposes like an eye blink or digestion. This is a radical and fascinating claim.
• She considers several examples of behaviors. Some include:
  • Conditioning for eye-blinks
  • Facial expression of (basic) emotions
  • Greeting rituals
  • Preference for sweet foods
  • Becoming aware of your breath
• But each of these presents a conundrum
  • Is the eye-blink conditioning something we "repress"? No. And being aware of it may not change our behavior.
  • Are the facial expressions any less or more meaningful if we don't know we are doing them?
  • Do we really know the purpose of all of our culturally0 transmitted purposes?
  • The preference for sweet foods has a good evolutionary explanation. One can know or not know this and find the motivations unchanged.
  • The role of breath does not change when we become aware of it, does it?
• Human's, we might still argue, have language and complex internal representations. Doesn't this give them special kinds of purposes? Millikan says no.
  • The purposeful devices enabling language must be general (just as the purpose of a camera is not to take any particular picture); so the purposes involved in our best explanations are not going to be primarily the contents we concern ourselves with.
  • Being aware of a purpose does not change the purpose!

Chapter 2

• Here, Millikan endorses a controversial idea, and one that is indepedent of the view we saw in "Biosemantics": she endorses the claim that there are memes.
• Evolution has produced new types of evolution, and this includes memes.
• Memes are patterns of behavior (and this includes ideas). Given that we can learn patters of behavior (and ideas), but that we have limited time and capacity, it is reasonable to conclude that memes can be "in competition" for our attention and support (we support them when we adopt and transmit them).
• In this way, meme's are like genes: they are structures that get transmitted and multiply (more people adopt the idea for memes, more organisms of this kind are born for genes). Competition can mean even that they fall to a kind of natural selection.
• Millikan says the purpose of the meme is whatever causes it to be reproduced (important: note that this can be something we do not know!).
• There is a challenge lurking here: Millikan allows that some memes may have purposes which are not our own. I imagine astology memes might be an example. But then, what purpose could such things have but to be reproduced? And why doesn't the explanation of all memes fall to this single purpose? Presumably we will need to say of some memes that we reproduce them because they are useful; and of others that we reproduce them because they exploit or trick us. How will we draw this line?
• One way to draw this line is to identify devices in humans that transmit memes or support the transmission of memes, for specific kinds of purposes. Cooperation mechanisms might be an example.
• Examples might include devices that in order to facilitate coordination cause some conformity, and this is partly accomplished through sharing common memes. Millikan makes three very interesting observations here:
  • Contrary to much game-theory poisoned philosophy, Millikan asserts that most cooperation is not mysterious and not something that overcomes competition. Rather, cooperation is usually a benefit to all cooperators. (To see an idea of what she means, you can look at some notes on the prisoner's dilemma here.)
  • Millikan claims we use belief-desire psychology after the fact, and not to make predictions as Dennett claims. Rather we predict people mostly on the basis of patterns we see in behavior.
  • Also, we predict behavior based on patterns of social conformity.
• It is a plausible hypothesis then that one mechanism transmitting memes is a mechanism that encourages social conformity (which is really sharing common memes). Meme transmission may often then happen in ways that have nothing to do with utility or reason or sophisticated cognitive systems.
• Memes can thus have a general purpose which covers some but not all instances: they serve to coordinate activity between humans.
• Millikan ends by distinguishing a more fine grained notion for meaning than for representation. The latter refers primarily to "satisfaction condidtions" -- something like a (potential) state of affairs..

Some Observations on Chapters 3 and 4

• Millikan redefines natural signs, natural information, and intentional signs.
• Some distinctions: natural signs cannot be false, but intentional signs can be. Also, some intentional signs depend upon (and when true, are) natural signs, but not all are.
• Dretske's theory of natural information is untenable, since he argues that they depend upon a probability of 1 in a correlation, but then his examples are frequently of statistical relationships (which are not of probability 1).
• Also, clearly there can be information about particulars, but not laws about particulars (e.g., information about George Bush).
• Millikan introduces two notions:
  1. Local natural information
  2. Locally recurrent natural signs
  These are in contrast to Dretske's notion of information, which she renames "context free natural information."
• Context free natural information aims to be a natural correlation in the world (perhaps of probability 1) between As and Bs. Then As are context free signs of Bs.
• Millikan replaces this with the notion that As are signs of Bs if there was some reliable connection in the past that we found we could learn from. (Note: she changes the game from conceptual analysis -- what is a sign? -- to theory construction -- what makes for a reliable sign?)
• These signs are local because they are learned and used only in local contexts. They are recurrent because the relationship that allows them to indicate or represent can occur several/many times.
• Millikan goes on to argue (C4) that these signs can be compositional/productive, and they can be "embedded."
• Natural signs are states of affairs that represent states of affairs. (This is no small move -- there is great controversy in philosophy about "states of affairs" and "facts." They are common-sensical until you start thinking hard about what they might be. Still, the notion seems not particular dangerous for Millikan, since she need only specify enough of a SOA to describe how it reliably is correlated (locally) with some other SOA.)
• Since SOAs can change in terms of various magnitudes, she argues that similar/kinds of SOAs can carry diverse information. Also, when we see the relationship of a natural sign in other contexts, we are seeing different cases of productive use of the sign -- it is being used for different cases (different times, places, objects). (Look at the gas gauge example: we can understand them in different cars, so they productively mean "amount of gas in this car.")
• Embedding is when the information can be taken to refer to more and more SOAs. The geese flying south means the geese are flying south, but also that winter is coming.

Chapter 5

• Teleofunctions will not explain what representations are, not how they connect up with the world, but they will explain the normative nature of representations.
• Millikan attributes the idea that teleofunctions migth explain more to a confusion in Brentano's work. Understand that many philosophers took Brentano's notion of the "intentional inexistent" and made many conundrums of it. Where is the Santa? Is there a possible Santa I represent? A Santa image? What is this inexistent Santa object? Etc.
• On Millikan's account, a false representation is one that fails to represent. They are "representations" in the sense that they are tokens in the device which has the function to represent, but they fail to do so.
• "To represent" is ambiguous: it means (1) there is a thing and it is being represented, but also (2) one is trying to represent something. On (1), there is an object, on (2) there may not be. Some philosophers talk about intentional inexistence and seem to treat (2) like (1) and wonder about the fictional things that are "represented."
• Think of teleosemantics as something we add to some theory of representation to explain its normative aspect.
• There is a potential confusion here. A shorthand gloss of Millikan's view is that the teleofunction of a representation is to represent something. But now suppose that our theory of representation is a causal one (to which we add some teleofunctional account). Teleofunctions describe goals, or effects, of devices. A representation cannot have as its function to have a causal relationship of such and such a kind -- it has to exist first as a representation before it can have any teleofunction, so how could it's teleofunction be to aim for it being a certain kind of thing! It's like saying the goal of every baby is to be born.
• But then how should we clarify this role for teleosemantics? Millikan attempts it by floating three alternatives:
  1. Whatever representations are on your theory (Millikan calls this "basic representations"), it could be that the teleofunction of the devices that produce representations is to make representations of that kind.
  2. There may be a device that makes representations into the kind of thing you take to be basic. Functionalism might be such a view: something acts as a representation because the right kind of relationship gets set up with other internal states.
  3. There are devices which when fulfilling their teleofunction (Millikan says, "acting normally" and here "normal" is used as explained in "Biosemantics") use basic representations. Something is a representation of the right kind when it is produced by or used by (that is, is necessary for the teleofunction of) that device.
• Millikan's approach is the third alternative.
• We observed last time that you can understand Millikan in terms of the Wright Schema. This schema is an analysis in a very general form of how we might reduce teleofunctions. It goes like this.

  The teleofunction of structure S is to do f if and only if:

  1. S does f
  2. S exists because it does f

  Here we generally should read S and f as types. Also, I described S as a structure; I consider this more conservative but Millikan will also allow that memes have teleofunctions. The heavy lifting is done by the second clause. We can interpret Millikan as saying that the structure is there because it was selected to do f.
• I mentioned the systematist view. My work on this includes the following paper: Ontology and Teleofunctions, which I include only for those who may be interested.

Chapters 6 and 7
Our interest in these chapters concerns primarily just a few facts.

• Millikan proposes three ways in which representations can be used by a system in terms of how they relate to the world.
  1. Direct: the representation is local natural information that is taken to describe something about the world.
  2. Directive: the representation is local natural information that is used in part to motivate an action which in turn may cause some change in the world; and it is this change which is actually represented.
  3. Pushmi-Pullyu (hereafter "PP"): these are a mix of descriptive and directive.
• Millikan argues that we can easily explain the strange features of intensionality with the claim that representations be in a context where "grammar alone does not tell what likeness is intended."
• Thus, an intensional context may include that "Jones believes Malcolm X is tall" but we cannot conclude that "Jones believes that Malcolm Little was tall" even though MalcolmX = MalcolmLittle. She appears to be saying that the problem is that in the first sentence grammar does not tell us whether "Malcolm X" means "the person referred to by me as 'Malcolm x'" or "That man I'm pointing at now" or "The person Jones believes is called Malcolm X and which she believes has no other name" and so on. This is a rather simplistic and bold move to do away with a great many philosophical problems.

Some thoughts on Part IV
It is fruitful to think of Millikan as in these chapters primarily aiming to speculate about how representational devices, and so representations, evolved. Since she sees PP representations as common to all representing organisms, she in part sees this as the task of explaining what distinguishes some of the abilities of humans from nonhuman organisms, partly in terms of the evolution of PPs.

• One striking feature of human representations, according to Millikan, is that we have pure descriptive, and pure directive, representations.
• Recall that PPs both describe and direct at the same time. The beaver tail slap describes both danger and directs flight. She speculates that it is possible that all nonhuman representations are PPs.
• Many (most?) internal representations are PPs. Examples include representations used by our bodies for various homeostatic balancing acts. Reflexes and some emotional behaviors are directed by PPs.
• Millikan argues that PPs can
  • be abstract, contrary to the traditional idea that primitive signs are all concrete. But it all really turns on how the device that "consumes" the relevant PP is using it; if it needs and can find abstract information in the PP, then that PP is abstract.
  • refer to distal (distant) affairs
  • refer to future or past affairs
• Presumably PPs evolved to be articulate and combinatorial so that they could direct more and more complex and useful behaviors.
• Some of these combinatorial parts may come to play a role that is largely descriptive -- one way perhaps that we evolved descriptive representations.
• There are two aspects of the PP: identification of affordances (opportunities for action) and identification of the relation of self to that affordance. There is evidence in the separate motor and descriptive visual tracks that vertebrate vision is hardwired for two such aspects of vision.
• We can see highly repititious and ineffectual behavior in the "pure PP" animal because it does not represent the success conditions (or, really, the purpose) of its own PP-directed actions. Examples include the Sphex wasp, etc.
• Note that there is a difference between representing things in order to complete some task, and representing what it is to have completed that task. The latter is not present in most animals and so they can exhibit very repititiuos behavior even when it is unsuccessful.
• Millikan argues that the first step in separating description from directive is to represent the goal state of a PP. Millikan hypotheses that a common representational system evolved in animals to represent by current (for pushmi) and future (pullyu) information
• Millikan has a suggestion for what separates us from other animals: we will collect (remember) information for which we have no specific (present) use.