Is SDT an error theory?

What is an error theory?

An error theory, for our purposes, is a theory of perception according to which competent human perceivers make global errors and not just local ones. It seems plausible enough to suppose that systems that don't work perfectly might well deliver results which are infelicitous. However, even well-functioning perceptual systems have a certain error rate. Systems that do work perfectly still deliver incorrect results regularly, and in a way that we anticipate. Such a notion will not be strange to anyone familiar with modern notions of inductive inference, in which a trade off must be made between the ampliative nature of induction and the level of guarantee that induction permits for its results. (It is the optimality of the trade off that constitutes working perfectly, not never making a mistake.) Even in non-technical media, it is common to hear of polls estimating quantities of votes, say, with intervals of estimation "using a method reliable nineteen times out of twenty". Readers with some familiarity with standard devices of statistical inference will recognize that those devices not only allow for the possibility of error but must confidently - "using a method unreliable one time in twenty" - insist on the existence of error. But those errors are what I'll term 'local' errors. Any adequate theory of perception must make allowance for that. On the other hand, it would be startling and discomforting to discover that our perceptual systems, when working as reliably as they can, are globally wrong. And it would violate familiar philosophical principles of Charity that insist that our cognitive systems are broadly correct, though with local exceptions.

Is SDT perception systematically misleading?

It has sometimes been said that SDT is 'systematically misleading' . Here's what Michael Tye (Consciousness, Color, and Content. (2000), p. 45-46) has said:

Focus your attention on the scene before your eyes and on how things look to you. You see various objects; and you see these objects by seeing their facing surfaces. Sense-datum theorists claimed that the facing surfaces of the objects are themselves seen by seeing further immaterial surfaces or sense data. The sense-datum theory is unacceptable, however, for a whole host of familiar reasons.Intuitively, the surfaces you see directly are publicly observable physical surfaces. ...

To suppose that the qualities of which perceivers are directly aware in undergoing ordinary, everyday visual experiences are really qualities of the experiences would be to convict such experiences of massive error. That is just not credible. It seems totally implausible to hold that visual experience is systematically misleading in this way. Accordingly, the qualities of which you are directly aware in focusing on the scene before your eyes and how things look are not qualities of your visual experience.

Tye proposes that the sense datum theorist is mistaken in judging that the qualities that we are aware of in experience are internal to us, that is, qualities of the experience (or locally accessible to the experience) rather than qualities of the things in the external world.

The argument that SDT is mistaken in this way is that it implies that SDT is an error theory of perception; it necessarily makes perceivers globally wrong, not just locally or intermittently so. If SDT were true, Tye is saying, then perceivers must take the attributes of which they are aware in their experiences to be qualities of 'publicly observable surfaces'. Such a situation defies belief, and shows that SDT is not to be believed.

Is Tye's target the right SDT?

Tye's description of SDT is rather narrow. He identifies as his target exclusively those versions of SDT, like Frank Jackson's (1977), which appear to be committed to the regressive inner see/outer see conception in which "the facing surfaces of ... objects are themselves seen by seeing further immaterial surfaces or sense data" (emphasis added). However, even if we assume a non-regressive SDT in which the perceiver is aware only of the sense data and not at all of the external, distal stimuli, something like Tye's (2000) complaint would have a prima facie plausibility. If a perceiver were to judge that what is presented in experience is the external world when it is, in fact, an array of sense data in the mind or brain is plausibly a global error rather than a local one.

I propose that Tye's argument is mistaken. It is not true that if we assume SDT, then perceivers must be taken to be in global error. Assuming SDT, it is true that any perceivers who judge that the qualities they experience are qualities of the external world are wrong. But the assumption of SDT does not require that perceivers make any such mistake. Indeed, it would be hard to see how there could be sense datum theorists at all, given that assumption. The fact that there are only a few does not cut it.

Tye's mistake, I believe, is to think that the SDT proponent must concur with his intentionalist assumption that the content of experience is as of the external world.

Grand Stimulus Error

What is at issue, here, is whether the sense datum theorist must concede that a normally functioning human visual system involves awareness of an array of sense data whose representational content includes the proposition that the array is the world outside our bodies.

The SDT espoused here answers in the negative, although it allows that most of us, most of the time, do indeed adopt the global falsehood that what we experience is the world outside our bodies (with the usual disclaimers about floaters, phosphenes, afterimages, and other comparable phenomena). However, that mistake is not part of the "deliverances of the senses". It arises, not from belief's faithfully taking on the delimited intentional content of experience, but from incorrect interpretation and inductive inference based, in part, on mistaking the signs of properties in the external world for those very properties. Tye is right, that there is a mistake, but it is not part and parcel of the sense datum theory. SDT does not make perception "systematically misleading" as Tye proposes. A perceiver's taking the phenomenal array to be the external world I call "the Grand Stimulus Error" (GSE). (Traditionally, after Titchener, 'stimulus error' is said to occur when reports of subjective experience are biased by subjects' knowledge of the distal stimuli which give rise to the experience. What makes the Grand Stimulus Error "Grand" is that it assumes that what is revealed in the experience is almost entirely the distal world.)

Phenomenology does not "depict" the world: the 'animal' argument

Consider the following plausibility argument for saying that the phenomenology does not intrinsically represent the distal world, and that a mismatch between phenomenology and the world does not necessarily mean that there is a perceptual error. It is not unrealistic to think that the properties exhibited in the phenomenal arrays of many primates are broadly similar in character to those of humans. There is all sort of variation. Further, despite the relative sophistication of primate psychology, we should assume that competent apes, at least in some cases, have no beliefs of the kind "what I am experiencing is the world outside my body".

It may be true that the well-functioning of primate vision incorporates properties in the phenomenal array that stand for properties of the ape body and for properties of remote targets and the spatial relations of the instances of phenomenal properties reflect spatial relations between the ape body and the remote targets. But it is eminently conceivable that an ape with such a well-functioning perceptual system need have absolutely no conceptualization of "what I am experiencing is the world outside my body". And if a human had a suitably similar phenomenology to that of the ape in question, "what I am experiencing is the external world" should not, without further ado, be taken to be an intrinsic part of the representational content of phenomenology, either.

Before elaborating on these claims, I shall turn to consider whether some other notable philosophical accounts of perception are Error Theories. I shall start with Intentional Representationalism, proceed to Projectivism, and then return to Sense Datum Theory.

Intentionalism as an Error Theory

Intentionalist-Representationalism (sometimes, nowadays "Representationalism" or "Intentionalism") is widely endorsed by philosophers of mind and philosophers of perception. Above, I cited Michael Tye as a source of a criticism that SDT was an 'error theory', in that it implied that SDT's account of perceivers made them globally wrong in their perceptual representations. Tye has been a long-standing advocate of Representationalism, explicitly denying sense datum theory among other alternatives.

Fundamentally, Representationalism is the view that what we experience, the 'phenomenal character' of the experience, derives from, or is 'a matter of', what is represented by the experience -- its intentional content. The meat of the position is that what is given in experience is what is represented to be out there, rather than properties of the representing experience itself. There is quite a lot of fine print in the careful formulation of the variants of the position. Some variants allow that certain types of blurriness, say, might be phenomenal character not reflecting intentional content. Some have even allowed that different vehicles for a given intentional content might yield some different experienced properties; a certain content from organic chemistry might have discernible experiential differences when conveyed in a lecture and when enjoyed in a sip of wine.

Apart from those sorts of exceptions, Representationalism is the idea that phenomenology is or is determined by what is represented to be out there.

Now, consider the case of colour. To the extent that this site has an official position, it includes the thesis that we have no good scientific grounds for attributing (phenomenal) colours to what is out there. The evidence is that the formation of colours occurs behind the retina, not in the bodies outside us that are commonly judged to be red, white, and blue, etc.

If that is right, then Representationalists are in trouble. If phenomenology, which includes colours, is not matched in the external world that it attributes properties to, then there is error. If the external world is devoid of colours as experienced by us, then the error is not just local. It's not just that magenta in the phenomenology failed to match yellow in the environment. It's that any colour at all was taken to be out there. Representationalism is an error theory.

Projectivism and Error

Some sense datum theories must count as what are termed “projectivist” theories of perception. The idea is that perception includes awareness of a phenomenal array – a ‘visual field’ -- whose properties are attributed to the world outside. (I shall use “phenomenal array” rather than “visual field” for reasons given elsewhere.) So the phenomenal array might include a red figure against a green ground and be a vehicle for the intentional content that there is a ripe tomato with leaves in front of the viewer. Hume famously described this sort of situation as the “gilding or staining all natural objects with the colors borrowed from internal sentiments”. (Quoted in Boghossian and Velleman, p. 95.)

This ‘projectivist’ notion has seen more employment in value theory than in perceptual theory, and value projectivists, Hume included, seem to have informally assumed the validity of Hume’s conception of the projection of colour as their template. Boghossian and Velleman have contributed a somewhat developed version of the Projectivist position on colour in the modern literature, and, with some qualifications, they endorse the Humean conception. However, they are willing to accept what they take to be a cost of adopting that conception, which is that the account must be taken to be an Error Theory.

There are two main planks to the Boghossian and Velleman Projectivist position:

1. “visual experience cannot be adequately described without reference to intrinsic sensational qualities of a visual field” (which “may or may not supervene on neural structures; it may or may not be describable by means of adverbs modifying mental verbs rather than by substantives denoting mental items”).

2. “intrinsic colour properties of the visual field are the properties that objects are seen as having when they look coloured”.

Proposition 1 effectively puts the Projectivist position in a form that a sense datum theorist can endorse. Proposition 2, making allowance for rather concise formulation, seems to imply that whatever is instantiated in the phenomenal array will count as the intentional content of the relevant experience and be ‘projected’ onto the external world.

The projection of the full character of the phenomenal array onto the external world is quite compatible with some sense datum theories. I think that it is fair to describe that family as ‘quasi-Aristotelean’. That is because the very same attributes are instantiated in the external world and in the sensory system, recalling Aristotle's notion of perception as the communication of Form without Matter. For example, both the external and the internal world instantiate the same colours. They instantiate the same shapes, although the quantitative details must be not identical but suitably proportionate. Unfortunately, the quasi-Aristotelean account is not empirically plausible, because, at the very least, there are no compelling grounds for attributing phenomenal colours to the distal causes of the phenomenal array and there are plausible grounds for locating the production of colours behind the retina only.

If we assume that the second proposition implies that every instance of a property of the phenomenal array is seen as a property that some distal object has – I’ll call that 'exhaustive projection' --then Projectivism must be counted as an Error Theory. It gets colour globally wrong; there is no colour ‘out there’. As Boghossian and Velleman observe:

External objects do not actually have the colour qualities that projectivism interprets visual experience as attributing to them. The projectivist account thus interprets visual experience as having a content that would be systematically erroneous.

The exhaustive projection of all the attributes of the phenomenal array puts Boghossian and Velleman’s extreme position on all fours with Representationalism. Phenomenal character exactly captures the intentional content attributed to the world by the experience. Equally, if a sense datum theory is understood to insist that the properties of the phenomenal array are exhaustively projected onto the world, and the phenomenal array is taken to include colours, then it must count, too, as an Error Theory. So extreme – exhaustive – versions of projectivism and of sense datum theory fall prey to the same difficulty that Representationalism has: understanding perceptual processes falsely to attribute colours to the external world. The difference between the theories is that less exhaustive versions of projectivism and sense datum theory are feasible options, while Representationalism is essentially constrained to match phenomenal character and intentional content exhaustively.

Although Boghossian and Velleman’s second proposition seems to suggest the exhaustive version of projectivism, they allow that there are exceptions to the rule, while insisting that, as they put it, “visual experience is naive most of the time, or in most respects.” That is, they concede that not all the properties of the phenomenal array need in every case be projected onto the external world. Thus, the properties instantiated in the phenomenal array in the case of an afterimage are not seen as any ‘objects’ looking coloured. They are not projected. Boghossian and Velleman propose similar exceptions for certain kinds of blur and double vision.

This retreat from exhaustive projection is highly desirable. If visual experience requires awareness of a ‘visual field’ or ‘phenomenal array’, in all likelihood grounded in the operations of the brain, its proper functioning need not require that everything in its phenomenal character is attributed as such to the represented world. In the case of colour, the claim that Projectivism is an Error Theory rests on the premise that colours in the phenomenal array are falsely attributed to the external world. While Boghossian and Velleman recognize an exception in the case of the colours of afterimages they are committed to the idea that the majority of colours instantiated in the phenomenal array are properly to be understood to be projected and falsely so.

Embracing Projectivism as the theory “that occurs naturally to anyone who learns the rudimentary facts about light and vision”, they write:

External objects do not actually have the colour qualities that projectivism interprets visual experience as attributing to them. The projectivist account thus interprets visual experience as having a content that would be systematically erroneous.

But why should a projectivist or a sense datum approach have to go along with that commitment? Why have colours projected at all, at least in the strong sense that colour experiences are understood to be claiming that there are ‘out there’ instantiations of the same colours as are in the phenomenal array?

Phenomenal character and intentional content in SDT

SDT and Representationalism differ importantly in what they say about the relation of phenomenal character and attributed intentional content.

Representationalism seems to have to say that phenomenal character is dictated by the attributed intentional content, so that the phenomenal character captures what is attributed to the external world. If the phenomenal character includes a certain configuration of coloured regions, then it is because the intentional content of the experience includes (the proposition) that there is that configuration of colours out there. Consequently, the phenomenal character experienced by the viewer is a simulacrum of parts of the external world , at least in the veridical case. Even in the non-veridical case, the phenomenal character is supposed to be like or identical with the possible world that is represented.

Neither SDT nor Projectivism is constrained in the same way. If the phenomenal array contains a certain configuration of colours, it does not follow that the experience has the intentional content that there is the same configuration of colours in the external world. Unlike Representationalism, according to which phenomenal character and attributed content must be congruent and the representation is fully envisioned, SDT can employ sense data to merely stand for things and properties in the outside world which are perhaps qualitatively quite different from them. Sense datum theory is not obliged to have phenomenal character "projected" into the outside world, in the way some Projectivist theories suppose. That is why sense datum theorists are unperturbed by the lack of evidence for the existence of phenomenal colours in distal stimuli. (Indeed, they embrace that lack for as part of the case for accepting sense datum theory!). Just because there are no colours in external things does not mean that colours in sense data are symptoms of the relevant experience being a misrepresentation.

Why should we think, because a sense datum is red, that its role in a given experience is to attribute redness to something? A red sense datum can represent something by simply standing for it, by arbitrary or natural convention. A warning light may be red in order to indicate an excessive temperature in an engine. Its functioning as a representation of that condition does not require that its being red means that the engine condition is red.

This should be a familiar enough point, both from common experience, like the engine light, and from the philosophical tradition. We might recall that John Locke introduced his widely discussed inverted spectrum example as a corollary to his doctrine of a sort of analogue of ‘truth’ for simple ideas of sensation. Simple ideas of sensation have as their 'marks' -- as Locke put it -- the qualities in external bodies which are their causes, because of their immutable link with them, not because they are qualitatively similar to or identical to them. (Locke's use of 'mark' that way is suggestive of its modern use for the target of a fraud or scam; other modern usage would confusingly suggest that a colour could be the 'mark' of some cause.) The employment of red and green in the experiences of primates viewing ripe fruit against leafy backgrounds is not to represent the fruit as being red and the leaves as being green. It is to achieve a portfolio of interests, including being 'labels' -- to use Mohan Matthen's term -- to stand for and distinguish them and also to enhance the boundary between the sensory elements representing stimuli. The latter function helps use figures against grounds to inform us about bodies viewed against backgrounds. That the colours serve those ends doesn't mean they're meant to look the same as fruit and leaves (if indeed fruit and leaves have anything that can count as a 'look'.)

I have offered some clues, above, that hint that SDT and Projectivism are not committed to exhaustive projection of phenomenal character. It remains to spell out ways in which phenomenal character can serve in a broadly representational way to tell us about the external world and ground our actions in it, but without it offering us a perfect simulacrum of the world. Further, we shall see that it is inessential for the operations of the senses that perceivers judge that they are experiencing the outside world. It is for that reason that Tye (2000) was mistaken in claiming that SDT is "systematically misleading".


(to be continued)