The Epistemological Event Horizon: Why Dretske’s Model of Perceptual Awareness Fails

Paul J. Gentile

Throughout “Conscious Awareness”, Dretske argues in favor of a thing-awareness of some factive difference attributable to subjects which is sufficient for conscious experience, in short, that if some thing A differs from some other thing B, then, accordingly, some subject's perceptual experience of A must then differ from her experience of B, independently of whether the subject experiences that A and B differ (i.e. whether S has any beliefs concerning differences between A & B), or independently of whether the subject behaves in some fashion consonant with discrimination. I argue that Dretske's model is untenable. The structure of Dretske’s model - prior to any further determinations - presupposes or assumes sight, viz. it relies upon a presupposed and unqualified attribution of sight of some particular x to some S.  The bulk of my attention will be devoted to demonstrating that this assumption - upon which the entirety of Dretske’s model rests - is, at best, inoperable, at worst, nonsensical.  My argument will consist of several interdependent sections: (1) I will highlight problems concerning the attribution of thing-awareness to some subject, (2) I will demonstrate that necessary criteria circumscribing conditions under which such attributions might be effectively performed in ordinary circumstances are absent from Dretske's argument, and (3) I will explore some conceptual confusions concerning Dretske's treatment of seeing, awareness, experience, and difference, and show how such confusions compromise his broader position.

Overview of Dretske's Argument

I will restrict my attention primarily to the centerpiece of Dretske's paper, the “Alpha/Beta” example, wherein he attempts to show that a thing-awareness may obtain independently of any fact-awareness. In this example, the viewer is presented with two diagrams or constellations, each consisting of a collection of black shapes – circles, triangles, rhombi, quadrilaterals, etc – of various size, contained within an implicitly square perimeter, against a white background. The two constellations, we are told, differ in just one respect: Alpha features one black circle (near the 4 o’clock position, call it “Spot”), which is absent in constellation Beta:
The success of this thought experiment requires the observer to imagine that, while glancing first at Alpha, then at Beta, this factive discrepancy, “Spot”, is not explicitly noted or detected. In his own words:

“According to my assumptions, everyone saw Spot. Hence, according to (1), everyone was aware of the thing that constitutes the difference between Alpha and Beta. According to (4), then, everyone consciously experienced (i.e. had a conscious experience of) the thing that distinguishes Alpha from Beta. Everyone, therefore, was thing-aware, but not fact-aware, of the difference between Alpha and Beta.” #

This is a rather compelling example, and at first glance it may appear quite uncontroversial: there is some factive difference between Alpha & Beta, S looked at or saw both Alpha & Beta, thus it follows that S's experience of Alpha must be different than her experience of Beta, independently of whether she explicitly attended to or was fact-aware of that element (“Spot”) which constitutes the factive difference between Alpha & Beta. On this view, there is just a fact that A & B differed, and, if S was aware of thing A, and then aware of thing B, it follows that these two awarenesses must accordingly differ. Thus Dretske wishes to attribute to the individual thing-awareness (without any accompanying fact-awareness) in this and, presumably, all other similar instances.

Dretske advances four assumptions:

(1) “S sees (hears, etc) x (or that P) → S is conscious of x (that P)”. #

Dretske distinguishes perceptual experience (a concept-free mental state) = thing-awareness vs. perceptual belief (a concept-charged mental state) = fact-awareness.  Compare: S sees Bill playing a piano vs. S sees that Bill is playing a piano.

(2) “S is conscious of x ≠ S is conscious that x is F”. #

A conscious awareness of x needn’t be accompanied by any explicit, conceptually-charged assessment that x is F, or an F.  

(3) “S is conscious of x or that P = S is conscious (a conscious being)”.

“That is, transitive (creature) consciousness implies intransitive (creature) consciousness.  You cannot see or hear, taste or smell, a thing without (thereby) being conscious.” #

(4) “S is conscious of x or that P → S is in a conscious mental state of some sort”.

The Stakes:

The stakes involved here are considerable.  If Dretske’s model is correct, we are licensed to infer from observation - with a great degree of specificity - the content of other creature’s visual experiences, and to ascribe to them very particular mental states.  If my reproach of Dretske’s model is correct, the very foundation upon which Dretske’s model rests is in-principle corrupt, and cannot possibly serve as any sound basis for ascribing visual experiences or mental states to others, human or otherwise.  Dretske’s central assumption or foundational premises - concerning (1) what there objectively is to see, and (2) that visually-enabled creatures must have seen “it” - casts an exceedingly broad and philosophically-problematic net, and so requires a correspondingly-broad and comprehensive treatment, viz. demonstrating the depth-of-error embedded within Dretske’s reflexive assumption - that S saw x - will require considerable labor.  

The Problem of Attribution

Of course, we would like to say that persons or other animals quite often see, or are perhaps thing-aware, of many visual features of an environment of which they are not explicitly fact-aware, viz. the visual detection or negotiation of many discrete objects or features is not necessarily accompanied by any articulation of factive propositions by the perceiver, e.g. of the general form “there is a tree immediately to my left” or “there is a blue car”, and so on. And this is a quite trivial point: to imagine that the effective “seeing” of some object or feature should require some explicit accompanying act of propositional articulation, differentiation or identification correspondent to each feature is just absurd. Equally, we would like to say that individuals most certainly discriminate among or see things when no other persons are present to confirm this fact. But, of course, general attributions of “seeing”, without any specificity whatever, tell us precisely nothing: if one maintains “someone in Chicago sometimes sees or is thing-aware of something”, no useful information may derived from such a statement, and there are no conditions or facts of the matter here to empirically verify, no specific state-of-affairs which may or may not obtain.

The target, then, is to attribute something (thing-awareness) in particular to someone (S) in particular, as an attribution without any specificity is no attribution whatsoever. That is: if we wish to attribute some thing-awareness to some individual – to assert that some peculiar state-of-affairs actually obtains – it follows that we obviously require some constraint or criterion. Thus, moving from generality to specificity, we might wish to say “Jones in Chicago is presently thing-aware (though not fact-aware) of some factive difference between documents A & B”, i.e. to attribute thing-awareness of some factive difference to Jones. But if there exists some factive difference between documents A & B – and a thing-awareness of this difference attributable to our first-person observer Jones (1PO) – how might we justify this attribution in the absence of a third-person observer (3PO)?

For if we maintain that Jones is presently thing-aware of some factive difference between documents A & B, then – of necessity – (1) Jones must be presently looking at such documents, and (2) there must be a factive difference between such documents, and we cannot suppose that Jones is actually looking at such documents nor posit factive differences concerning the documents in the absence of a 3PO who might confirm or establish these specific facts of the matter. Thus, any attribution of thing-awareness of some factive difference to some individual necessarily presupposes some fact of the matter, and such facts of the matter are, quite clearly, not simply given, they are established. And if they are established, they are established by some 3PO. And if such facts of the matter are established exclusively by way of a 3PO, then attributions of specific thing-awareness to some S are necessarily contingent upon some fact-aware assessments performed by a present 3PO. Clearly, these constitute the necessary conditions of any attributions concerning any awareness of differences; exterior to the presence of such a third-person observer, attributions of particular thing-awareness collapse.

Of course, Dretske’s argument - by introducing the assumption “everyone saw Spot” - is in the form of the material conditional, i.e.: if there is some factive difference between A & B, and if 1PO saw both A & B, then 1PO was thing-aware of this difference. This manner of hedging is quite convenient for purposes of securing formal validity - a rather weak criterion - and I hope to present in detail precisely why this assumption is so problematic.  

As such, when Dretske presupposes that “everyone saw Spot”, in order to attribute this fact of the matter to “everyone” – or indeed, anyone – he must concede that any attribution of such thing-awareness is impossible in the absence of some prior fact-awareness possessed by some 3PO, consisting of

(P): a criterion delimiting what gestures, movements, or behaviors effectively count as a surveying of discrete things, ensembles or features.                  

(W): a criterion delimiting what there is or how many discrete features or ensembles there are within some visually-available spectrum.

We will elaborate upon the difficulties of P & W in the sections which follow.  But firstly, we will turn our attention to a few other suppositions built into Dretske’s argument which require highlighting before we proceed.  

The Problem of Explicit Instruction:

Ostensibly, the implicit goal of Dretske's example is to establish some fact of the matter in this circumstance, from which we may justifiably extrapolate that similar facts obtain in other circumstances, viz. that the attribution of thing-awareness in the present example licenses us to attribute thing-awareness to individuals throughout a broader range of real-world visual negotiations of an environment. Moreover, because Dretske’s target is to demonstrate that the non-factive, concept-free experiences of some S result in a conscious awareness, he is clearly not restricting his argument just to concept-capable humans, but is - whether he intended to or otherwise - suggesting that any sighted creature enjoys thing-awareness of factive differences insofar as their sensory organs are appropriately oriented toward the relevant signals or ensembles constitutive of some factive difference. If we could simply ask S whether she indeed saw A & B, we would then have secured a factive guarantee that a discrete seeing of A & B obtained, but Dretske’s target is the attribution of concept-free mental states to some S in instances where the object of awareness (the feature, the difference) is not explicitly denoted.

But things become considerably more difficult when we move away from simplistic, sterile thought-experiments of the present variety wherein concept-charged instructions are issued to the participant - and instead soberly consider the dizzying varieties of complexity which attend real-world examples of visual negotiation.  Perhaps I observed my dog “staring” at my bookshelf.  I cannot instruct my dog (or any other nonhuman animal) to attend to A & B and expect them to obey, nor can I cannot ask my dog whether she saw book A & book B (let alone any differences between them).  Further, I cannot infer from any of her gestures whether she looked at the spines of several books, or all the books, or saw the entire ensemble as one thing, or whether things we call books show up as discrete things for her.  

The visually-available features of my bookshelf contain as many - or as few - possible factive differences as does the entire city, depending upon how one wishes to carve out “difference” here, or what specific schema-of-difference one chooses to apply.  Which schema does Dretske endeavor to impute to my dog, or - in the absence of explicit instruction - to myself?  In the Alpha/Beta example, the answer to questions concerning possible schemas-of-difference is painfully obvious, but the slightest move toward entirely pedestrian circumstances of visual negotiation throws into relief our central difficulty: that Dretske’s Alpha/Beta just gives us pro bono the most difficult to obtain portion of our dilemma - the attribution of discriminate sight of some A & B to some S - absolutely for free.  It just grants a specificity of sight and of what is seen on the basis of assumption, without any meaningful criteria concerning what might count as instances of discrete sight throughout ordinary, non-clinical instances of seeing.  If discerning precisely what animals did or did not see was obvious, or could be so lazily assumed, notice that there would be few if any predicaments whatever concerning perception.

We have, to my mind, already sufficiently established the priority of a fact-aware 3PO in order to legitimately attribute thing-awareness to some 1PO. But we have yet to further highlight the impossible bundle of faculties which our hypothetical 3PO must of necessity possess in order to perform the required acts of attribution throughout ordinary circumstances of perception. Let us then transpose these preconditions to alternate scenarios, wherein it is decidedly unclear precisely what constitutes “ensembles”, and equally unclear which pairs of ensembles fall within S's field of vision as items of comparison, i.e. circumstances wherein S is not presented with just two discrete, clearly-defined ensembles, and explicitly instructed to look at or compare them, but rather a situation wherein there are just virtually countless surface features which might potentially serve as discrete items of comparison. To suggest such complicated or unquantifiable circumstances is not at all exotic; it is just to invoke humdrum, everyday instances of seeing.

Let’s begin with a (hopefully) poignant example.  Consider the figure below where constellations Alpha & Beta have been incorporated (Alpha has been placed within Beta):  

Let’s call the above figure Gamma.  On Dretske’s model, the “factive difference” between Alpha and Beta is still just there before some perceiving S: I have not structurally-altered the figures Alpha and Beta or otherwise rearranged their respective constituents; I have merely reduced Alpha to one-twelfth its original size, moved it slightly to the right, and called the broader ensemble “Gamma”.  Dretske’s theorizing on perceptual awareness says absolutely nothing of “implicit borders”, “contiguity” or “edge detection”, nor did it mention that an appreciation of differing size should constitute a difference in kind.  Accordingly, I see no reason why A & B should not still be discrete ensembles available to comparison under the umbrella of Dretske’s model.  

Imagine that some S - human or otherwise - encounters Gamma without any explicit, accompanying instructions - encounters Gamma as any visually-available feature among so many others, or even as dozens or even hundreds of possible features - and (of course) that S possesses no foreknowledge that Alpha & Beta have here been incorporated.  When S has been explicitly instructed to consult two discrete ensembles, and is further pre-alerted to the fact that the two may differ in some respect, there is no question whether S saw A & B, and, moreover, the participant is actively looking to apply some possible schema of difference.  So we already have at least some difference embedded within the instruction: “look left at Alpha, then right at Beta, and report any discrepancies”. That is: the task or instruction - by virtue of being an instruction - is already “concept-charged”, i.e. some regions, things or ensembles have already been pre-perceptually delimited as discrete items of comparison - such that we may assume the seeing of such items - in a way that the common elements of a visually-available field are not so discretely delimited.

Because our hypothetical S (in the case of Gamma) has not received any conceptually-charged instructions concerning what discrete things or ensembles to attend to, we cannot simply assume what S saw.  If we cannot assume what S saw, we cannot know that S saw.  One problem of attribution is that: specificity precedes generality; if our target is an ascription of peculiar thing-awareness of some peculiar x to some S, then (1) an ascription of seeing to some S must necessarily precede any attributions of thing-awareness, and (2) that ascription of seeing cannot be of a general form, e.g. “S sees”,  but rather must be the seeing of something in particular, e.g. S sees some x.  

So in the perceptual laboratory of Alpha/Beta, we have secured a guarantee of seeing: if S understood the already-concept-charged instruction, it is presupposed that S saw A & B, and so saw the factive difference (“Spot”) between A & B.  The understanding of the instruction and of its neatly-established parameters licenses us to assume that seeing occurred in the Alpha/Beta example.  It affords of no skepticism or ambiguity.  That is: the most difficult preliminary challenge of attribution - guaranteeing what S saw and thus that S saw - is simply acquired for free, and is not the product of any painstaking assessment or determination.  

But such licenses and guarantees cannot possibly be extended to our Gamma example: let’s take a second look at Gamma, again imagining that some S has not been explicitly instructed to inspect, survey, or to attend to Gamma in any peculiar fashion, e.g. that Gamma might itself be enclosed within various Deltas and Epsilons, or leaves, or rocks.  That is: the subtraction of any explicit instruction issued to some S to survey Gamma renders Gamma possibly indiscrete; i.e. Gamma itself might blend among a host of other possible features or ensembles, at which point it ceases to be a discrete thing called Gamma:


But more importantly: does the alleged “factive difference” which obtained in the Alpha/Beta example still obtain in Gamma?  We don’t even know that Gamma itself obtains as a perceptually discrete ensemble for some S - i.e. a discrete thing seen - much less the supposed “factive differences” between Alpha & Beta “embedded within” Gamma.  If Gamma was excerpted, enclosed, or perceptually-quarantined as a discrete ensemble, how many possible differences are available or might “obtain”?  That is: what is the total number of “factive differences” present within Gamma?  We might imagine:

- Alpha

different from

- Beta

different from

- the set of all circles.

different from

- the set of all parallelograms.

different from

- the set of all trapezoids.  

different from

- the set of all triangles.  

different from

- the orientation of one triangle

different from

-  the orientation of another triangle.

and so on.

We could - of course - elaborate upon this exercise exhaustively, introducing virtually innumerable possible differences, none of which are necessarily extant or “just there” for any or all sighted creatures.

In fact, Dretske’s argument does not even require an example as conspicuous as the Alpha/Beta contrast; under his own assumptions, a creature who sees just Alpha (or Beta, or Gamma, or whatever) should be thing-aware of the factive differences resident therein.  In the absence of any explicit instruction to attend to just Alpha, Alpha needn’t necessarily be seen as a discrete or quarantined ensemble; and if Alpha is seen as discrete, it might just as well be seen (or construed, or experienced, or attended to) as a collection of twenty-two different things, or the set of all x’s, y’s, or z’s.  That is: if we assumed (e)A, why not then also assume (e¹)A, (e²)A, and so on, wherein Alpha is not experienced homogeneously - certainly, it’s amply faceted to assume that it may be experienced in an entirely heterogeneous fashion - but rather its very constituents (individual circles, parallelograms, etc) are discretely seen, scrutinized, attended to,               Again: while a concept-charged instruction issued to a concept-capable human to visually attend to A (or to A & B, or to G) might to some degree prohibit the “mistake” of S fixating visually upon the various components of A, this kind of surview cannot be taken for granted in the absence of such instruction.  In short, ordinary instances of visual perception are not at all like Alpha/Beta, and are more like Gamma.

The Necessity of a Two-Way Criterion:

So let us now imagine the necessary presence of our hypothetical fact-aware 3PO in order to positively attribute any certain thing-awarenesses of certain things to certain 1POs: what quantity and what sort of information would our 3PO necessarily require to accurately establish the facts of the matter, in order for us to confidently attribute some peculiar thing-awareness to our 1PO?
The answer is that a two-way criterion - a calibration between what there is and what is seen - must be circumscribed on both sides; one for perceiver (P) and one for world (W).  The problem, then, is that the success of Dretske’s experiment - at the very least - relies upon two wholly unjustified assumptions: (W) what there really is to be seen, and (P) that what there really is really is seen.  In order for his central lesson to be extrapolable - viz. applicable to animals in any meaningful way, or useful for some research project - P & W should be satisfiable:

P: a criterion delimiting what gestures, movements, or behaviors effectively count as a surveying of discrete things, ensembles or features.  

Our 3PO should be possessed of some ability to discern, through observation, that the movements of 1PO's head or her eye saccades effectively scanned the totality of objects or features in question, i.e. that such head and eye movements corresponded firstly to a seeing of just those discrete features. Recall that before we may attempt to attribute thing-awareness of some factive difference to some S, we must firstly establish that S was actually looking at such items of comparison. Dretske's view implicitly presupposes that a “seeing” x & y just occurs when S's eyes are open and x & y are judged to be within S's field of vision. Under this quite broad view, everything present before one is just seen. While any sober consideration of what this view practically entails should be unsettling, I will not dwell much on the point. It should for present purposes suffice to remark that Dretske's short-sided treatment of seeing is not uncontroversial, and, as we shall see, compromises the success of his argument.  

Imagine: we see a man in the forest, pointing a camera toward the two-o’clock position.  In the immediate foreground, there are various shrubs and trees, upon which there are leaves, insects, algae, fungi, and so on.  Just beyond the foreground there are rolling hills, upon which are scattered trees of some different variety.  In the far background, the ridge of a mountain projecting into the distance, itself quite faceted and cluttered with features.  Like the eye, the camera captures or “sees” by means of focus, and one could not in principle detect from observing the mere orientation of a camera what it was focused upon, i.e. whether it was recording objects in the foreground, or minute facets or striations within those foreground objects, or whether the camera was focused upon the mid-ground or the far background.  And like the camera, what is captured by the eye is always a matter of focus, and so a consideration of focus must be included within our anatomy of what is seen or unseen.  

But focus establishes an exclusionary principle concerning sight, or what is seen, which is clearly incompatible with the notion that S is just thing-aware of all factive differences which are immediately before one’s open eyes.  The number of possible factive differences within a visually-available field will be astronomical - if not virtually infinite - and, assuming there just are factive differences within this field, the respective apprehension of each difference will not be a matter of simple availability, but rather of focus.  Because availability neither presupposes nor entails focus, accordingly, “S’s eyes were open before available units x y z” cannot supply a bridge from availablity → focus → sight.  If seeing requires focus, and focus necessarily entails “to the exclusion of x, y, & z features”, then the mere availability of x y z features within some S’s visual field does not guarantee that x y z were in fact registered, captured, experienced or seen.  Again: our goal is firstly to attribute the seeing of some things to some S, and some 3PO observing that S’s head was oriented in this direction, or that S’s camera was pointed toward the two-o’clock position, could not possibly determine - in the absence of explicit instructions or subsequent reports - precisely which available features were registered, captured, or seen.  

The question then is: under what circumstances would it be possible, in principle, for some 3PO to detect that some 1PO's head movements and eye saccades corresponded to just some particular features and not some others? Do 1PO's physical gestures amount to visually attending to the curtain itself, or just the shadow falling upon the curtain? Is 1PO presently attending to the bookshelf, or to the collection of books, or to the spine of one book in particular? If we wish to firstly attribute a seeing of some things or ensembles to some S, an attribution which is logically prior to any further attributions of thing-awareness, then we must clearly establish the parameters or conditions under which this first attribution is to be justified.  That is, we require a criterion: we cannot attribute thing-awareness of x to S before we have attributed a seeing of x to S. But we cannot attribute any seeing of x to S prior to attributing fact-awareness of S's seeing of x to some 3PO. But our 3PO sees in the same fashion as does our 1PO; if there exists any ambiguities concerning the specificities of what 1PO sees or does not see, this ambiguity is then equally extendable to our 3PO. “1PO's eyes were open”, I submit, is a grossly insufficient criterion for the attribution that S saw x.

“Focus” is but one among a number of facets neglected throughout Dretske’s treatment of vision.  A creature’s body-scale relative to the features of its environment will of course be a determining factor concerning what is seen.  The velocity at which a creature moves throughout an environment will likewise represent a variable concerning what is or is not apprehended.  The discrete ensembles Alpha & Beta (and their respective difference) which were just objectively there in the laboratory are simply not there for the man looking through the window of a speeding train, nor for the sighted insect.

W: a criterion delimiting how many discrete features or ensembles there are within some visually-available spectrum.  

We must confirm that certain, discrete items which fell within the available parameters of S's field of vision - whether such items are physically contiguous or disparate - constituted some factive difference. Is there some factive difference between the edge of the desk and the color of its legs?  “Of course there is a difference”, you will say: because you have heard and understood the very concept-charged question.  But Dretske’s implicit target - to advance a theory of perception wherein we are licensed to attribute a concept-free, conscious thing-awareness of some x to some S in the absence of explicit instruction or corresponding report - cannot rely upon such hearing and understanding on the part of some S; it must be operable in the absence of such qualifications.  What about the thousands upon thousands of pebbles along the path? No two are identical; each rock differs in its various facets from all the others. The top of the glass differs from the bottom, the entire glass from the wooden frame which it encloses it, the color of the frame from the shape, and so on ad infinitum. In any ordinary circumstance of seeing – say, surveying a room – there are billions of possible factive differences among the surface features of items within one's field of vision. Without the imposition of some criterion – not one arbitrarily chosen, but a criterion well-suited to ordinary instances of seeing – then every creature is just thing-aware of all the billions of possible factive differences whenever one's eyes are open. That is, the notion of difference here is not sufficiently discrete; unless it is delimited via some meaningful criterion, nearly anything might count as a factive-difference of which, if S’s eyes were open before it, S must be consciously thing-aware of it.

One very important lesson here is that the concept “awareness” - which Dretske imagines to be synonymous with seeing - requires careful constraint: an awareness of billions of differences is no kind of awareness whatsoever.  Awareness, like the exclusionary concepts “concentration”, “concern”, “focus” or “attention”, is just not the sort of concept whose application may be incautiously stretched without constraint for purposes of accommodating some philosophical theory; if my concentration, concern, or focus is “everywhere”, it is precisely nowhere. Similarly, to maintain that “S is presently enjoying billions of awarenesses of which S is unaware” requires us to forfeit the specificity and value of the term “awareness” altogether. In J. L. Austin’s expression:

“We have here, in fact, a typical case of a word, which already has a very special use, being gradually stretched, without caution or definition or any limit, until it becomes, first perhaps obscurely metaphorical, but ultimately meaningless.  One can’t abuse ordinary language without paying for it.” #

Accordingly, Dretske's treatment of thing-awareness amounts to a complete trivialization of awareness – reducing conscious thing-awareness to little more than seeing – essentially telling us nothing.  It is nonsensical to suggest that S is focused upon every visually-available feature within a given field, and it is in principle impossible for some 3PO to discern in the absence of instruction or report precisely what some S is focusing or not focusing upon.

The Problem of Discrete Experiences:

Following from our discussion concerning the problems of inexactitude which result from this absence of criteria, we might then briefly consult what Dretske means by “experience”, as if it consists of some isolable or solitary state-of-affairs. Recall that under Dretske's model, two experiences must differ if the representational content of those two “experiences” is “different”, e.g. Alpha differs from Beta (Alpha features 'Spot', Beta does not), so, accordingly, S's experience of Alpha must differ from S's experience of Beta. But notice that this model presupposes “experience” as a thing: for each unique event, object or feature, there is imagined the accompaniment of some discrete, quantifiable experience correspondent to the seeing of each thing. In ordinary circumstances of concept-free seeing or visual negotiation, it is exceedingly doubtful that this is the case. That is: there exists no necessary criterion which clearly demarcates the “end” of one experience and the “beginning” of another; the activity of seeing or experiencing is rather a fluid and discontinuous process, and experiences, whatever they are, do not of necessity consist of a series of discrete consultations of unique objects or features.

Here, for purposes of charitability, I will suppose that Dretske is wielding the term “experience” to imply something like “attending”. But even this fails to alleviate our difficulties: it is quite common to attend to complex temporal ensembles which contain a multiplicity of changing features – imagine staring at a busy traffic intersection – e.g. some individual might be said to be attending to the whole, wherein “wholeness” here might imply viewing some constantly changing scene for ten seconds, or for one minute, or for thirty minutes. Now: was the total thirty-minute experience comprised of thousands of discrete micro-experiences, or was it just one experience? Is there a continuity of experience between T¹ and T², such that T¹ and T² might be counted as a single, uninterrupted experience, or is T¹ a different experience than T²? Does there exist some unique experiential event correspondent to each of the cars which entered my field of vision, or, perhaps, should we quantify the seeing of just three cars, or just some features (tires, windows, etc) of these cars, as an experiential event? I will not further belabor this trivial point, but it is worth remarking that if the problem of attribution is reliant upon our ability to say of some individual the she is presently enjoying discrete “experiences” of some things, then the above challenges concerning the precise constituents of an experience deserve some careful consideration.  And again, in the absence of explicit instructions or subsequent reports, Dretske cannot obtain discrete visual experience of some x by some S for free.  

The Problem of Difference, and the Necessity of Questions Concerning Difference:

We have yet to thoroughly explore what is implied here by factive differences. For Dretske, some element or feature which constitutes some factive difference between two things needn't be explicitly detected or attended to by S in order for us to attribute thing-awareness of the difference to S. As previously noted, this attribution of thing-awareness to some S is logically contingent upon some 3PO who might establish the facts of the matter, i.e. what the factive differences are. Again, the total number of possible factive differences among possible features of comparison throughout, say, an ordinary bedroom, is virtually without limit. To concede that there are virtually limitless distinctions to be made among detectable visual features throughout any environment – to concede that there are virtually no circumstances wherein the term “difference” might prove inapplicable – does not impel us to forfeit the value of differentiating or positing differences in peculiar circumstances; it simply reminds us that, wherever the concept “difference” is divorced from some circumscribable objective, it might then obtain just anywhere.
That is: any two points of comparison might legitimately be judged as different under some conceptual schema. But if this is the case, then, at least potentially, all there are are differences, which renders any judgment of difference redundant. Where does this leave us? Well, either (a) judgments concerning difference are not reflections of any “real” or constant differences in the world, as just anything might constitute a “real” difference; rather, they are conceptual judgments – that is, the application or imposition of some particular conceptual schema – but conceptual judgments which retain their usefulness or application only insofar as they are appropriately constrained or relevant to some objective. Otherwise, (b) differences are just in the world, and as such precede any perceptions of such differences.

It should go without saying that difference is a relation, the emergent product of some comparison, and that relations are not extant “things”; such relations are not simply present and do not obtain in the absence of perceivers for whom some discontinuity or disjunction – in the form of an implicit or explicit question – is present. That Jack & Jill are of different height does not describe a state-of-affairs which is extant in the world independent of circumstances; “Jack is taller than Jill” is a relation which belongs neither to Jack nor to Jill.  It is a relation which obtains (or does not obtain) where there are present implicit or explicit questions concerning height.  For example, Jack might be gauged as ‘too tall to attack’ by a potential predator, and so Jill may appear to the predator as the preferred prey.  Likewise, this difference will not manifest for an animal whose body-scale is such that it is unable to perceive any difference in height between Jack & Jill, or for an animal whose objectives otherwise do not take height into account.  It is one thing to say “those features which constitute or inform the judgment of difference exist independently of any perceiver”, but quite another to suggest “differences themselves precede perception”. It is quite strange to me – and patently dualistic – to posit the existence of some necessary comparative relation among features that “exists” wholly prior to the application of some conceptual or behavioral schema by a perceiving creature, or a relation which simply “is there” or otherwise obtains where there are no organisms to take note of this comparative relation as a difference.

As such, if difference is a relation which must obtain for some organism, then it makes little sense to posit the necessity of any pre-perceptual factive differences of which some S is thing-aware, as the organism's schema of interpretation when confronted with some possible disjunction or discontinuity among a virtual infinity of possible disjunctions or discontinuities – either explicitly propositional or in the form of a “selective” behavior – will be the deciding factor of whether some proposed relation of difference obtains. Clearly, there is not some constant that is necessarily constitutive of difference; there are all sorts of ways in which one may visually attend to ensembles in accord with peculiar questions.

In the above remarks I invoke the notion of some “selective” behavior when confronted with possible disjunctions or discontinuities for two reasons: firstly, it allows us to rescue the value of perceived differences from redundancy by framing or constraining the perception of differences in terms of relevant tasks. If a difference is a possible relation, then it must be a possible relation for some S. If some relation obtains for some S, then it will not obtain for just any or all possible relations; it will obtain as the product of some task: an instruction (to perform an assessment in accord with some specific criterion of difference, to compare or contrast with some objective in mind) or a utility-based challenge (e.g. does this fruit differ in some respect from the fruit which is safe to eat?). Secondly, it allows us to suggest a possible perception of differences among nonhuman animals by hypothesizing some “general form of question”, in terms of opportunities and consequences for some animal. If our target is a truly general theory of perception, then we naturally endeavor to establish commonalities among the perceptual systems and behaviors of both humans and nonhuman animals; and attributing conscious thing-awareness of the difference between Alpha & Beta to, say, a squirrel or a chicken, is patently absurd, and makes a mockery of both the concepts “difference” and “awareness”. This is akin to asserting, “whenever a creature's eyes are open, it just experiences (is thing-aware of) all factive differences”. Indeed, according to Dretske, the squirrel or the chicken must be thing-aware of all the aforementioned sorts of differences, which is to say, thing-aware of all possible factive differences.

Dretske’s views here are decidedly externalist, and one could derive from his model some partial support for a direct or non-inferential theory of perception, wherein information is just there in the environment, available within the ambient array.  But unlike Gibsonian models - which do not construe environmental information as static signals equally available among all sighted creatures, but instead as creature-specific affordances available in accord with a creature’s body-scale, abilities, objectives and challenges (in crude Wittgensteinian notation, a creature’s ‘form of life’) - Dretske’s model neglects this range of variables in favor of a metaphysical construal of available differences.  And while this is not the occasion to introduce or to defend Gibsonian models - in light of Dretske’s pedestrian treatment of differences - a very brief digression into Gibson’s affordances seems warranted here.  In Gibson’s terms:

“An affordance is neither an objective property nor a subjective property; or it is both

if you like. An affordance cuts across the dichotomy of subjective–objective and helps

us to understand its inadequacy. It is equally a fact of the environment and a fact of

behavior. It is both physical and psychical, yet neither. An affordance points both

ways, to the environment and to the observer.” #

Notice that Gibson’s ambiguous commentary does not offer us any convenient one-size-fits-all solution to questions concerning perception; it is a largely negative exercise, a rebuke of generality which throws into sharp relief the difficulties and complexities which must attend the attribution of specific sight of some x to some S. Such remarks are undoubtedly a thorn in the side of lazy scientists and philosophers who - in realizing the dizzying scope of their task - would prefer to sweep aside the vaulting range of peculiarities which must be represented within one’s model if one endeavors to advance some general theory of sight, or of consciousness.  The unfortunate reflex, “the fewer variables = the better”, simply will not do, and I can imagine no better example of this reflex than Dretske’s Alpha/Beta model.  An adequate science of sight will of necessity be a juggling of innumerable variables, likely involving degrees of specificity, taxonomy and complexity which positively dwarfs the comparatively-tidy notation of physics.  
These remarks don’t suggest that future research programs are doomed; they suggest that such programs - should they endeavor to establish an adequate account of seeing, or of conscious awareness - must first jettison that unfortunate scientific inclination toward generality and adopt something nearer to a creature-specific account concerning what is seen or not seen.

The appreciation of differential relations, or the act of discriminating, must involve discrete circumstances of visual confrontation wherein one scrutinizes meaningful or relevant discrepancies among features, in accord with the specificity of some challenge or objective at hand, and by means of some conceptual or behavioral schemas accommodative of the peculiarity of that challenge, some action consonant with notions of discrimination obtains. And while differences are, in this sense, manufactured, rather than “constants” existing in nature, awaiting notice or discovery, the fact that differences are manufactured is precisely what salvages their utility or meaning: if differences simply existed independently of assessments, patiently awaiting detection, we should have to concede that they existed or obtained just everywhere.

To summarize: if there exists some factive difference among things, and S sees such things, then S is thing-aware of the difference. But without proper constraint, there is nothing to see but differences, which renders Dretske's point entirely redundant. But if we impose some constraint concerning the appearance or detection of differences, then we concede that differences cannot meaningfully obtain exterior to some conceptual schema. But if differences are of conceptual manufacture, then differences are not merely reducible to seeing some identifiable physical feature or element which constitutes some difference. Accordingly, the possibility of attributing any proposed thing-awareness to some S by virtue of some fact-aware third-person observer collapses.

The Problem of Assumption:   

We might now revisit our central complaint.  The success of Dretske’s experiment - at the very least - relies upon two wholly unjustified assumptions: (1) what there really is, (2) that what there really is is really seen.  Taken together, 1 & 2 constitute an assumption concerning what seeing is.  Of course, I see what Dretske is attempting to do here; I’m only suggesting that he can’t get away with it.  The favorite trick of philosophers and logicians - assume P to derive some valid Q - will not work here, for a host of reasons:

Firstly, to ‘assume P’ here would be tantamount to taking for granted the particular content of any creature’s visual experience, viz. advancing a one-size-fits-all de facto solution to the problem concerning precisely what others do or do not perceive. Fortunately for Dretske, logic does not restrict the assumption of any P for purposes of securing validity.  Unfortunately, Dretske’s conclusion - that S enjoys a conscious mental state, the content of which is some seen x - does not differ appreciably from his central assumption, and so his argument closely resembles a tautology.  Recall that this alleged “mental state” - which should be attributable to sighted humans, grasshoppers and chickens alike - is not of any conceptual or articulable variety.  I’ve no clue what Dretske really means here by mental state, but it clearly does not differ meaningfully from what we would ordinarily call seeing.  It boils down to little more than “If S saw x, then S saw x”.  

Notice that any ordinary attribution of sight to other animals is rooted in observations of an animal’s ability to visually discriminate: sighted creatures apparently react or respond to environmental features in deliberate ways, e.g. dodging, leaping, running toward x or away from y, picking up z, etc.  That is: our criterion concerning whether some S saw some x - what we actually mean by “seeing” - involves justifiable inferences from observed behaviors construable as reactions to some peculiar features or events.  Call this the behavior-positive model.  

Yet, it is doubtless that animals see a great deal of textures within a visually-available field to which they do not overtly respond with some explicit gesture, viz. neutrality or neglect are also kinds of behavior.  Silence and motionlessness are behaviors.  Neutrality or neglect are inconvenient for scientific purposes of measuring behavior, but it would be problematic to suggest that creatures only see those items to which their behaviors are apparently correspondent.  Here, we might wield the attribution “S sees x” in an intuitive or pedestrian fashion, because our basis for excluding or denying sight of x to S seems without qualification; it seems the precise inverse of our attribution problem. Call this the behavior-neutral model.

This leaves us with a criterial crisis of sorts: there exists a tension between the measurable vs. the measureless attribution of sight to some S.  Optical organs (open eyes) constitute necessary conditions, but clearly are insufficient for purposes of targeted, specific attributions of sight, in both the former (behavior-positive) and the latter (behavior-neutral) accounts.  We cannot enumerate every possible visually-available detail - along with the emergent differences or contrasts among such details - and subsequently ascribe a concept-free seeing of all such details to some open-eyed S without forfeiting the value of the term “sight” altogether.  Behaviors or reports constitute conditions sufficient for targeted attributions of sight, but are unnecessary according to the behavior-neutral account.  Dretske’s contention that “everyone saw Spot” would fall into the latter, behavior-neutral category, wherein evidence of visual discrimination of any sort is wholly absent from a given sample.  

But the fallacy upon which Dretske’s central assumption is founded should by now be quite obvious: an absence of reasons for denying x is not equivalent to a presence of reasons for affirming x.  “Because we have no evidence of S’s innocence, S must be guilty” does not compute, and represents an argument from ignorance.  Again, it is doubtless that behavior-neutral creatures whose visual apparatuses are in working order practice some manner of visual negotiation, but from this likelihood we may not legitimately infer the content of their visual experiences with any specificity or attribute to such creatures anything like visually-charged “mental states”. Notice that this represents an epistemological event horizon beyond which we are unable to justifiably affirm the presence of visual content or mental states in others not simply because our models lack sufficient detail, but rather because Dretske’s metaphysical presuppositions concerning what there really is are quite untenable.  

Secondly, by assuming a metaphysics of factive arrangements, ensembles, couplings or differences - i.e. of what there is to be represented - it prescribes the visual experiences of creatures in an unjustifiable manner - a manner immediately at-odds with Dretske’s own premises concerning  factive differences - placing the cart squarely before the horse. Dretske’s model concedes that there may exist facts-of-the-matter (factive differences) which are experienced (i.e. looked at and seen), but are not explicitly captured by visual diagnosis such that they may be denoted or reported.  If they cannot be denoted or reported, they cannot be included in the original model of what there is in a given circumstance.  Accordingly, Dretske’s insistence upon thing-awareness compromises his entire thought experiment:

- 3PO supposedly establishes what there really is, i.e. the factive differences of which - assuming 1PO saw such “things” - 1PO must be thing-aware.  This is to say: a great many “factive differences” might be absent from 3PO’s report.  Additionally - in richly-textured natural environments - the virtually-infinite number of possible factive differences is too great for any 3PO to record.  

- 3PO’s visual faculties do not differ appreciably from those of 1PO.  If 3PO sees like anyone else, there may be (many) factive differences of which 3PO is not fact-aware, yet (possibly) thing-aware.  Accordingly, 3PO may not be fact-aware of (some, all) relevant factive differences.  

- If 3PO was not (certainly) fact-aware of the factive differences, then 3PO’s judgment concerning whether S (1PO) in fact saw some A and some B respectively can be questioned: for if there are features or differences in the visual sample of which 3PO might be factually unaware, then we are simply without a reliable sample.  Additionally, as noted in an earlier section, verifying that 1PO saw all the differences in a sample containing billions of possible differences is problematic from a range of perspectives.  

- But then, according to our present operating principle, we could never establish a reliable sample, as surveying some objects or series of ensembles might always and necessarily admit of either errors or alternative schemas-of-interpretation.

This is really a mishandling of the classic ‘seeing vs. seeing-as’ problem, couched within a slightly updated psychological notation.  Dretske’s target is the seeing without the seeing-as, yet a very specific seeing + “conscious mental state”.  Assuming “everyone saw Alpha & Beta (and thus saw Spot)” imposes a kind of seeing-as, albeit a sort of connotative  - as opposed to an explicitly denotative - seeing-as, but one which nevertheless implies a specificity operationally on par with the kind of specificity one would acquire through the imposition of conceptual schemas.  

Whatever a conscious mental state is (or is not), it is difficult to see how one can obtain the concept-free specificity Dretske’s model requires for purposes of meaningful attribution of such states.  When we attribute, we must attribute with specificity; saying “S sees and hears” is saying absolutely nothing.  Dretske would like to have it both ways: he would like to say that some animal is thing-aware of some x, i.e. “is conscious of it”, that somehow the “it” (i.e. the feature, the difference) is uniquely and discretely quarantined as there for the apprehending subject, yet is not there as some “it”.  On Dretske’s model, we - the concept-charged humans - circumscribe what there is to be seen, i.e. a discrete series of features or factive differences, and further suggest that such features or factive differences show up or are discretely there for all animals without the accompanying concepts which excerpted such features as features to begin with.

In closing, the subtext here - supposing the reader has not already intimated it - concerns a problem with philosophical thought-experiments, particularly those which endeavor to extrapolate some broadly-applicable general principle via simplistic, one-size-fits-all models which blatantly disrespect the dynamism and complexity of their subject.  I have here attempted to highlight that
the various assumptions embedded within Dretske’s argument - that S understood the instruction to uniquely survey A, then B, that S saw A & B, that there really are factive differences of x y z variety - those assumptions permitted by formal logic which guarantee mere validity of one’s argument - cannot be assumed exterior to the sterile, concept-charged, instruction-charged climate of Dretske’s hypothetical laboratory, which do not at all resemble the circumstances under which animals ordinarily perceive an environment.  There are no instructions and no tightly-drawn borders in nature.  There are no reports of discriminate seeing of A & B from grammarless animals in a signal-rich coastal forest.  There are few ways in which we might circumscribe what the visually-available factive differences are for a given animal embedded within a feature-cluttered, natural environment.  The neatly-drawn quarantine of the Alpha/Beta example does not secure scientific legitimacy, or even the prospect of further research projects concerning concept-free perception.  From the armchair, it is easy enough to demonstrate that Dretske’s ideas here get us no closer to any useful or applicable models of perception or mind.  But it will require a committed immigration from the armchair to the field - an exceedingly complicated, pluralistic and humbling prostration toward what sighted creatures are actually in the business of doing - if we hope to establish clearer accounts concerning the experiences of sighted creatures.