Some Thoughts on Millikan's "Biosemantics"

---

• Here Millikan wants to explain what connects a representation up with the thing it represents
• Informational theories are based on the idea that there are some special conditions in which the connection between representation and its object are fixed; mistakes occur when these conditions fail to be met
• What are these "content-fixing" conditions? (That is, the conditions where the connection between the representation and the thing it represents are fixed.)
• It cannot be statistically normal conditions. There are many different cases which can be called statistically normal. And statistically normal causes are too various: is a red face a sign of sunburn or exertion or overheating or embarrasment? Any can be a cause.
• A number of views have focussed upon the idea of something about the way a representation is produced that makes it what it is.
  • Fodor has argued that content is fixed by causal connection, but the role of the representation in a functional system is what determines how its content is fixed. Millikan dismisses this: what is it to "act like a representation" other than to have fixed content?
  • Stampe and Matthen and Dretske say representations are tokens with a detecting or indicating role. Millikan dismisses this as unenlightening: what is it to detect or indicate other than to represent?
  • Dretske uses the notion of a natural sign: something caused by a thing R and thus which we can naturally take as representing R. But she notes Dretske moves from natural sign to the claim that this representation has the function of representing its cause. More important: many representations are not natural signs. Example's include very cautious warning signals, which are made when the danger is not present.
• Millikan's view is that it is the way a representation is used, not how it is made, that fixes its content. She calls this the "consumption" of the representation. The content of a representation thus stands for something not because of some natural information in that thing, or some particular information-bearing connection to that thing, but rather because there is some "device" in the mind that "consumes" that representation as a representation of that thing.
• Millikan needs for this the notion of "proper function." I have and will continue to use the terminology "teleofunction" for this.
• The teleofunction of a device is the role that it was selected to serve. Millikan also sometimes says the role it was designed to serve. The selection or design in organisms is done by evolution.
• Millikan now uses "normal" to mean the role that was selected. This is a bit confusing since "normal" has so many other uses. We might instead say something like, "its purpose" or "its teleofunction."
• The content of a representation is fixed by the role that the representation plays in the device that it serves.
• Natural information, then, even if it exists, can only be used if it is exploited by a device that has as a teleofunction the exploiting of that information.
• Millikan makes the odd claim here that even simple representations like a beaver slap are "articulate" because they happen at a time and place. I think what she means might be that there are several interpretative aspect to the single slap (a time of occurence, a place of occurence, the meaning of danger).
• Contrast with Dretske's information account: the magnetosomes.
  • The interesting issue here is that the oxygen does not cause the direction of the magnetosomes and thus the bacterium, but rather the direction of magnetic north does.
  • On a causal approach, or Dretske's information approach, doesn't this mean that the magnetosomes indicate "north"? But what use is that?
  • When we focus on consumption, we ask, what do these magnetosomes do for the organism? They move it away from oxygen rich water. But then that is their teleofunction.
  • The magnetosomes were selected for this very role: moving away from oxygen.
  • Thus, their content is fixed as something like, this way is away from the poisonous oxygen.
  • Note that the magnetosome represents something "distal" -- that is, not some direct connection but something distant or indirectly connected to the representation.
• Note that Millikan's "consumption" approach is accomplishing two things: first, it is giving us a more fine-grained account of content (so that we can say that the beaver slap means danger and not just "big thing" or something like that), second, it clearly distinguishes useful contents from not useful possible interpretations (magnetosomes direct away from oxygen, not towards north).
• Looking ahead: the magnetosome and the beaver slap are important examples of what Millikan will later call "push-me pull-me" representations, that both direct action and also describe the world.
• How could her account explain something as complex as language? Here Millikan wants to make a general hypothesis: there are devices in humans that have as their teleofunction various language consuming functions. It is, she claims, a very reasonable hypothesis that there were selection pressures benefitting use of language.
• What is the difference then between bacteria with magnetosomes, for example, and humans and their language abilities? Millikan gives six criteria:
  1. Our language includes representations with elements that are not self-representing. By this, Millikan seems to mean that they can refer to things not directly present in the representation itself.
  2. Our language includes elements that can be "stored" -- such as used after they are caused.
  3. Our language can include pure imperatives.
  4. Our language allows and includes inferences.
  5. Our language includes identity information.
  6. Our language allows negation and bivalent truth values.