Here, I explain why I don’t believe in sense data. I discussed this in Skepticism & the Veil of Perception and my Stanford Encyclopedia entry on “Sense Data” (https://stanford.library.usyd.edu.au/archives/spr2005/entries/sense-data/).
(After I wrote that entry, they kept asking me to edit and update the entry. I had not realized that by writing the entry I was understood to be taking on a lifelong writing project. I did not want such a project. At some point, I told them that I just was not going to update it again. And so, many years later, they got someone else to write an entirely new entry (https://plato.stanford.edu/entries/sense-data/). The new entry is much more historical and less argumentative.)
1. What Are Sense Data?
There are different ways you could understand it, as the people who introduced the term did not seem to be entirely consistent. But here is how I understood it: “sense data” have 3 key characteristics:
They are the things that you are directly aware of in perception;
They are mind-dependent;
They have exactly the characteristics they appear to have.
Ex.: You look at a tomato. The sense-data theorist would say there is a red, round thing of exactly the color and shape that you take yourself to be perceiving, which is dependent on your mind, and that is what you are really directly aware of.
So that’s what I’m denying. I think there is a red, round thing that you’re aware of, but it’s the actual tomato, not an image (or whatever) in your mind. I also think there is a mental state (a sensory experience, a state of seemingly perceiving a tomato), but it is not the object of your awareness, nor is it red or round.
2. Arguments for Sense Data
Why posit sense data? Traditionally, people cite phenomena like the following:
Perspectival Variation
Hume says that when you move farther away from a table, the table you see looks smaller. But the real table does not get smaller. Therefore, what you’re really seeing must be only a mental image, not the real table.
Replies: First, no, the table does not look smaller. It looks farther away.
Second, as Thomas Reid pointed out, the argument is just invalid. The terms “looks smaller” and “is not smaller” are perfectly compatible, so there’s nothing wrong with saying that you see the real table, which looks smaller but isn’t in fact smaller. There’s no need to posit an alternative, ‘mental’ table that really gets smaller.
Illusion
Say you look at a stick half-submerged in water. You see something that is bent. But the physical stick is perfectly straight. Therefore, it must be that you’re seeing a mental image, not the physical stick.
Reply: No, you’re seeing a physical stick, which looks bent but is straight. There’s nothing inconsistent in that and no reason for positing something that is actually bent.
Halllucination
Say you hallucinate a pink elephant. Then you’re directly aware of a pink, elephant-shaped thing. But there is no such thing in the physical world. So it must be a mental object that you’re aware of.
Reply: There is no pink or elephant-shaped thing that you perceive or have awareness of. There is only a state of falsely seeming to perceive a pink, elephant-shaped thing.
Analogy #1: Suppose you have a false belief. What is it that you know by having this belief? Answer: Nothing; false belief isn’t knowledge. It’s pseudo-knowledge, a state that might falsely appear to you to be knowledge. Similarly, hallucinating is not perceiving or being aware of something; it’s only a state that falsely appears to you to be perceiving or being aware of something.
Analogy #2: Suppose someone writes a news story about a murder that took place last week. Only it turns out to be a complete fabrication; no murder such as the story describes in fact occurred on our Earth. Imagine someone concluding that the story must be an accurate description of a “mental murder” that occurred in an alternative, mental realm, since it didn’t occur in the physical world. That’s ridiculous, right? That’s like what the sense data theorist is doing. No murder of any kind occurred; there’s just a story that falsely says that one did. Likewise, in the hallucination scenario, there is no elephant of any kind in the offing; there is just an experience that falsely represents that there is.
3. Arguments Against Sense Data
There are two main arguments against sense data that I like.
Where Are They?
If sense data exist, where exactly are they located? I think there is no satisfying answer to this.
You could try saying that they are located in the same place as the physical object that causes them. But that’s odd; why don’t they get in the way of the physical object? If you look up in the night sky and see a star 1000 light years away, does your experience cause a sense datum to appear 1000 light years away, on the surface of that star? This violates special relativity since it requires superluminal influence.
You could say sense data are wherever they appear to be (just as they have all other characteristics they appear to have). Maybe the star that’s 1000 light years away doesn’t really appear to be 1000 light years away; perhaps it only appears a few hundred feet away, thus avoiding the superluminal influence problem. But then you’d run into problems with sense data that don’t appear to be in any particular actual location. E.g., say you have a hallucination of being in an alternative world – you hallucinate a complete scene in Middle Earth, which doesn’t actually exist. The sense data cannot literally be in Middle Earth, since there is no such place.
You could say they appear in your head, in your brain. But it just sounds super naïve to suggest that when you see a tomato, a tiny tomato-shaped thing appears inside your head. We might wonder why these things don’t get in the way of your brain material. Since sense data have the characteristics they appear to have, if you feel an object that feels solid, that means a solid object appears inside your brain. Wouldn’t that damage your brain?
You could say they appear in a separate space, “phenomenal space”, with no spatial relations to physical space. This also violates relativity, if you care about that (nothing can be in physical time without being in physical space, because space and time do not exist as separate things, according to relativity). It’s also a pretty extravagant and bizarre posit. Do we really have to postulate a whole nother space, just to explain our seeing normal physical objects? It would be pretty hard to explain how things in physical space can interact causally with things in this alternate space.
Lastly, you could try saying that sense data have no locations. But in perception, we seem to see things with locations. They also seem to have other spatial properties, like sizes and shapes, and nothing can have spatial properties without being in space. Nor can anything be in space without being in any particular location.
So there seems to be no good answer to where sense data are.
The Argument from Indeterminacy
This is my favorite argument against sense data. Sense data, again, are supposed to have exactly their apparent properties. If you see something that looks bent, your sense datum is bent in just that way.
The problem is that objects can have indeterminate appearances. By this, I mean that an object can appear to fall on some dimension, in some range, but fail to appear to have any specific, exact value within that range. E.g., perhaps an object might appear red but not appear any exact shade of red. Or an object could appear to have a shape, but not appear to have any specific, perfectly precise shape.
Examples: when you look at different shades of colors, there is a limit to how many shades you can distinguish. If a small enough difference is made in an object’s spectral reflectance, you fail to notice the difference. This means that you can’t tell whether the two colors are the same or slightly different. The inability to tell if they are the same has to be explained by saying that you don’t perceive a perfectly precise color. This is also needed to explain why you can have a series of pairwise indistinguishable shades where the first is definitely distinguishable from the last.
Another example: You look at a sign from a distance. You can tell that it has writing on it, but you cannot read it. Now, what exact shape does the writing on the sign appear to you to have? There’s no answer to that; if there were such an exact shape, then you should be able to read the sign (if the shape corresponds to that of some sequence of words in your language); alternately, if the apparent shape is one that doesn’t match any sequence of words in your language, then you should (a) have the experience of seeing that the sign is written in some alien language, and (b) find, as you move closer to the sign, that the letters appear to change their shapes. You’d then conclude that you were hallucinating when looking at the sign from a distance.
Of course none of that happens. You can’t read the sign, nor does it appear to have some alien language, nor does it appear that the shapes change as you move closer, nor are you hallucinating. The only way to explain this is to say that, when you view the sign from a distance, the apparent shapes of the words are indeterminate. That is, your experience fails to represent the exact shapes.
There are two accounts of perceptual appearance. One is the intentionality theory: When something perceptually appears F to you, that’s a matter of your having a mental state that represents that the object is F. The other theory is the sense data theory: When something perceptually appears F to you, that is a matter of your being directly aware of a mental object that is F.
The first theory accommodates indeterminate appearances, because indeterminacy in representations is perfectly possible. Compare the fact that one can easily believe that an object is colored without believing, for any particular color, that the object is that specific color. It’s perfectly possible for an object to be represented as (F or G) without being represented as F and without being represented as G.
But the sense data theory does not accommodate indeterminate appearances. That’s because it’s impossible for an object to actually have indeterminate properties. Nothing can actually be (F or G) without being F and without being G. Hence, on the sense data theory, an object also could not appear (F or G) without appearing F and without appearing G.
So the sense data theory is a false account of what it is for the world to perceptually appear a certain way.
Sense data, and all data, are information. As such, they only exist as an interpretable pattern in some physical medium. If I can have information in my mind or brain without harming myself, I can have sense data in my mind or brain. So arguing as if sense data must consist of distinct physical objects seems odd.
Maybe I am misframing this. Maybe data consist of the electrical charge in a circuit, or the dimple in a DVD, or the electrochemical state of my brain cells, and so in some sense they are physical objects. But if so, sense data are not distinct from my brain. I suppose we can think of my brain as just as dynamic as my thoughts and perceptions. But new perceptions don’t require more space.
Did Hume and the other theorists of sense data distinguish them from other sorts of data that we gain awareness of, interpret, and learn from? Or is all data sense data? Does this argument apply to all data, so we should conclude that datum is an incoherent category?
I clearly don’t really understand the point here. I need to think about this a lot more before I can understand well enough to agree or disagree. Maybe I have not read carefully enough. But the argument from limited space is wrong, unless I missed something big.
sense data approach wrong for 2. and 3. 2 -> because there is no mind-body problem, there's just the body. 3 -> because an object doesn't need to have the exact features it appears to have. our vision (like any other sense) evolved to take physical inputs and probabilistically work (ie, help our chances at surviving without too much cost to other systems/features).
i don't understand the need for the indeterminacy approach, though. not getting it as well as I have with other posts. thanks for sharing!