So far we have seen arguments for the view that experiences have contents, and that those contents are something like Russellian propositions, which have among their constituents certain properties. Given this background, we can ask: which properties can be in the contents of experiences? (As before, we will focus on the case of visual experience.)
Here’s a list of properties which many who accept the above background assumptions would accept as among the properties which can be in the content of an experience:
Call a view on which the above exhaust the properties which can be represented in experience a sparse view of the contents of experience. One important question at this point is: are there any other properties which can be a part of the contents of visual experience? It seems to me that there are at least three reasons why this question is important.
Siegel’s argument against the sparse view takes the form of an argument that there are cases in which a change in the beliefs of a subject about what they’re perceiving plausibly affect the phenomenology of their experience. She gives two examples: (i) learning a language can affect the phenomenology perception of written language, and (ii) acquiring beliefs about a natural kind can affect the phenomenology of perception of members of that kind.
Given some plausible example of this sort, Siegel then runs the argument roughly as follows. Let E1 be an experience before the relevant beliefs are acquired, and E2 an experience after. Further, let E1 and E2 be alike with respect to their ‘sparse’ content.
| 1. | E1 and E2 differ in phenomenology. |
| 2. | Any difference in phenomenology implies a difference in content. |
| 3. | E1 and E2 differ in content. (1,2) |
| 4. | E1 and E2 do not differ in content with respect to their ‘sparse’ content. |
| 5. | E1 and E2 differ in content with respect to some property other than those which are a part of ‘sparse’ contents. (3,4) |
| C. | Properties other than those which can be a part of the sparse contents of experiences can be a part of the contents of experience. (5) |
The argument appears to be valid. We can now consider how a proponent of the sparse view might object to it:
It does not seem plausible to me to claim that there is no phenomenological difference at all between E1 and E2. The more plausible objection is that this difference is to be explained in terms of something other than visual phenomenology. Siegel offers two ways of spelling this out: (i) E1 and E2 differ in the phenomenologies of the thoughts which are accompanying the perceptions, and (ii) E1 and E2 differ in their ‘background phenomenologies.’
Option (i) claims that the phenomenological difference is to be explained by the phenomenology of some non-sensory mental event which accompanies the visual experience. E.g., it could be the judgement that such-and-such is a pine tree; or the judgement that such-and-such is a familiar kind of tree; etc. Against this option, Siegel suggests that there would be a phenomenological difference of the kind suggested even if one knew that one was not looking at pine trees, but at an elaborate hallucination or illusion. This in effect limits the cognitive events the objector can avail himself of: they cannot be events which involve a commitment to there being pine trees in the observed scene.
So suppose instead that the proponent of (i) suggests some non-commitment involving but still non-sensory event, e.g. entertaining the thought that such-and-such is a pine tree. Against this Siegel suggests that there can be such a phenomenological difference in the absence of explicitly entertaining such a thought. This does sound plausible; but is any argument given for this?
Option (ii) invokes some sort of background phenomenology-affecting state akin to drunkenness or depression. Siegel’s main reply is to point out - correctly, it seems to me - that recognitional dispositions do not really seem analogous in any way to moods.
The example of familiarity; impressions of familiarity always represent something as familiar.
This objection also runs afoul of what I was calling the thesis of the transparency of experience last time, in connection with Moore’s remarks about his introspection of an experience of blueness.
A final response open to the theorist inclined toward the sparse view of contents would be to try to make use the the kind of strategy which I suggested that a Russellian should use to handle cases of color constancy: find more properties to account for the relevant phenomenological differences. In this case, Siegel suggests that one might try to account for the differences in terms of perceived shape properties, like being pine-tree shaped. Perhaps one experiences these shape properties in E2, but not in E1. If so, then we can find a representational difference between the two cases without going beyond shape and color properties.
Siegel seems to think that this response is implausible -- or at least less plausible than her alternative. She suggests recognition of expression of emotions on faces as an example. Why this case does not seem to me to be decisive. I suggest that the example of text recognition is a better one, since in this case, as Siegel notes, we can imagine scrutinizing all the relevant shapes before being able to recognize their meaning.
Siegel suggests (in §4 of ‘Misperception’) that some remarks of Robert Stalnaker can be turned into an argument against, in particular, the view that kind properties can be in the contents of perception, and, perhaps, in general, against the view that properties other than color and shape properties can be parts of the contents of perception.
The argument goes like this:
| 1. | An experience of a Eucalyptus tree before familiarity with the kind does not represent the tree as of the kind Eucalyptus. |
| 2. | Being told that the tree of is of the kind Eucalyptus would not change the phenomenology of the experience. |
| 3. | If two experiences have the same phenomenology, they have the same content. |
| C. | An experience of the Eucalyptus tree after being told that it is of the kind Eucalyptus would not represent the tree as of the kind Eucalyptus. (1,2,3) |
First, a note on the limitations of the argument: this would not show that we never represent kind properties in perception, but only that in this case we don’t. But one would wonder when we do, if not in this sort of case. (And one might think that similar arguments could be adapted to other cases.)
Siegel does not focus on this line of response, but it seems to me plausible to respond to this argument by rejecting (3) - which is pretty much our old friend, the Phenomenal Bottleneck Principle.
Now we can, as Siegel suggests, weaken (3) while maintaining the validity of the argument. Consider the following weakening:
| 3*. | For any two successive experiences E and E’, if E and E’ have the same phenomenology, then there is no property which one has and the other does not. |
It seems to me that we should reject this, for much the same reasons as we should reject the original (3).
But Siegel gives us another interesting way around the Learning Argument, which is worth considering as well: deny (1). An argument against (1): one can come to possess the concept of the kind Eucalyptus by demonstrating the tree. But how could this be possible if the experience did not represent the tree as of the kind Eucalyptus?
Siegel’s argument gives us reason to adopt the view that natural kind properties, as well as shape and color properties, can be parts of the contents of experience. One good question at this point is: are there any other properties like this? Siegel’s discussion suggests that there are, and that these include the following:
Are there any other cases like these? What properties, if any, are clearly not like these?