version 1.3
It is very difficult to succinctly state the homunculus problem. Many philosophers have difficulty with it. It is somewhat like the problem with "akrasia" -- the Greek word for the experience of "knowing that what you do is morally wrong, but doing it anyway." It seems obvious that people often experience akrasia, but when you try to explain HOW it can happen, the problem becomes vexing. Did the akratic person really know that an act was wrong, or was the person merely in a state of cognitive and moral confusion? How could truly knowing something and action predicated on that same knowledge come into conflict? It seems impossible. You then wonder if you have stated the nature of what you are trying to describe correctly. Perhaps, you begin to think, "akrasia" does not exist at all.
So it is with the homunculus problem. We DO "see" "pictures" in the mind. And, evidently, the brain is somehow responsible for this ability. The mistake comes in explaining HOW the brain is responsible. Does the brain store "images"? Our first impulse is to answer this in the affirmative. It seems obviously true. How else could you possibly state what the brain seems to be doing? But "image" connotes something that can be SEEN. Is there then an "inner eye" that does the seeing? Subjectively, this also seems true. But how do we explain the workings of the INNER EYE? How does IT see? In order to "see," we have already invoked the idea of the inner eye. Does the INNER EYE itself have an inner eye? The dilemma is sometimes presented this way: we posit some "inner images" that must be seen; in order to explain the seeing, we posit an inner explanatory mechanism -- an inner eye, an inner HOMUNCULUS, that has certain cognitive powers, such as seeing. The inner mechanism posited is like a "little man" -- a homunculus -- that has the power to discern, interpret, and judge events and objects in the inner world -- to "see" these events and objects. But then it becomes obvious that we must explain how the homunculus itself sees. Does the homunculus itself have a mind/brain system, in which resides still another homunculus that provides the power of seeing? One begins to see the problem now!
The "solution" (if it can be called that) is simply to stick with what we know: we have (subjectively) an "inner eye" that "sees" (quotes necessary!) How this is done (precisely) we don't exactly know, other than that it involves processes (and, often, conscious effort) that is similar to those involved in actually seeing something in the external world. We also have to be careful when we say inwardly "see." It's not (generally) the case that we intrinsically accept as REAL the what the "inner eye" delivers to us. Rather, such "inputs" are known to us -- pre-packaged as it were -- as memory, imagination, or thought images. In other words, as one of the TYPES of images I (following the tradition in psychology prior to the invention of cognitive psychology) have been so careful to point out.
Avoiding the homunculus problem is also why the DEFINITION of mental images is so important. If you start by assuming that mental images ARE, for example, "impressions in the brain," "computational states", etc., you are likely to get off track and end in an explanatory framework that invokes an inner homunculus. If you start by defining mental images as inward visual presentations with which consciousness is directly aquainted that are distinct from cases of perception, you limit the range of acceptible forms of explanation, but you adhere to the truth of the phenomena in question.
This homunculus mistake was, in my view, most famously repeated in the contemporary literature by Kosslyn. Pylyshyn was, equally famously, quick to point out the mistake as early, I believe, as 1974. Kosslyn argued that COMPUTATIONAL MECHANISMS explained HOW the mind processes mental images and also how it "inwardly sees" or interprets information contained in mental images. But as Pylyshyn correctly argued, this cannot be the case. This is the homunculus error. Individual subjects, the ego, the individual consciousness, can inwardly see an image. But once we enter the explanatory realm of computational cognitive science, there is no room for the term "image" in the explanatory framework. Look as hard as you may please, you will never find an image -- an image that you can SEE -- anywhere in any computer. You might say: so what! This is clearly wrong, a mere rhetorical twist. We (imagists) do not think there is an image that you can see in a computer! The point is that images are clearly processed all the time in computers around the world! "No," replies Pylyshyn, "you are making the homunculus mistake again." Computers do not process "images." They process electrical states which can then CAUSE events that WE (individual conscious subjects) interpret as being images. Computers have no consciousness of anything at all, much less of anything "being an image." The state of something "being an image" (that is to say, something being an image OF something) is automatically a metaphysical proposition. As both Plato and Aristotole noted, such a proposition states that something is both like and unlike something else, and it states that while the image itself "has being," after a fashion, this being is dependent on human judgement and the relation between the image and the object that image is an image of.
This relates to a fundamental thesis (or as I would prefer, just an "obvious" point) in my work: the being an image is never "given"; the status of "image" is a state of affairs involving the conscious acts of individual subjects. Another way to state the homunculus problem: if we attempt to introduce some third, explanatory mechanism or entity between consciousness and its object, we are likely to commit the homunculus error. It is NOT true that the brain sees images, or that computational mechanisms see images. These things can, however, enable us (conscious agents) to see (or imagine) something and ACCEPT it as an image. It is not true that any mere physical object sees an image. Cameras do not see. Nor does the human eye by itself actually see. While it may be, and probably is, true that some sort of "computational" (loose use of the term) mechanisms CAUSE us to see, this is NOT the same as saying (erroneously) that the mechanisms THEMSELVES "see."
Created on ... November 10, 2005. Revised 12:05 11/16/2005