Two positions have dominated the philosophy of perception: naïve realism and the sense-datum theory. Either we see material objects “directly” or we see only sense-data. I will describe a hybrid theory according to which the objects of perception are indeed material particulars located at some distance away in space but the properties we see these objects as having are not the objective physical properties actually possessed by such particulars. Instead the properties in question are projected by the mind and have no objective counterparts in the physical world. Thus we should be, to use the standard terminology, naïve realists about the objects perceived but sense-datum theorists about the properties these objects are perceived as possessing. We see physical particulars existing in the distal environment, but the attributes we see them as possessing are not their actual physical attributes; they are (in one sense) mental attributes (or possibly topic-neutral attributes). We see objective particulars, but we don’t see them as having attributes objectively possessed (i.e. the attributes they actually objectively exemplify). Perception is both objective and subjective, depending on which facet of it we are discussing. In the matter of perceptual reference to specific objects the senses are objective, but in the matter of perceptual description (general attributes ascribed) the senses are not objective. In yet other terms, perception is exogenous and endogenous, distal and proximate, “external” and “internal”. Thus naïve realism about the objects of perception doesn’t entail naïve realism about how these objects are perceived, and subjectivism about how objects are perceived doesn’t entail subjectivism about the objects perceived. In fact it turns out that this combination of objectivity and subjectivity is precisely what we should expect of perception, given its function and limitations.
I won’t spend much time defending the first part of the hybrid theory—the identification of perceptual objects with objective physical particulars. When an ordinary object is before me in my visual field and acts on my senses in the normal way, we can say truly that it is the object I am seeing. It is the very object that I can also sense in other ways—by touch, taste, etc.—and it is the same object that is described by physics and chemistry, biology and geography. It is a three-dimensional solid object located in physical space consisting of atoms and subject to gravity and other forces; it might have been there long before perceiving organisms ever evolved and it will survive their extinction. Its properties are precisely those ascribed to it in objective physical science, whatever those may be. It is completely mind-independent. Yet it is the (direct!) object of perceptual reference—what the perceptual system singles out as a subject of predication (attribution, characterization). It is what is seen: not some intermediary will-o’-the-wisp entity but a real concrete inhabitant of objectivity reality. Nor is there any other object that is simultaneously seen along with this concrete object; it is the only object referred to by the perceptual system. So yes, we see bunches of atoms, assemblages of electrons, protons, and neutrons, collocations of strings or quarks—things that have nothing of the mind built into them. The objects of perception are just the ordinary physical particulars of the natural world, neither more nor less; and nothing else!
This much will be assented to by most theorists of perception today (not so in the heyday of sense-datum theory); far more controversial is the second part of the theory. It will be supposed that the properties attributed by vision (say) are precisely those objectively possessed by the distal particulars that constitute the objects of sight. For example, I may see an object as round and being round is exactly what that object is—objectively, in reality, as things mind-independently stand. I see the physical particular itself and my eyes ascribe to it the very properties it really has. Maybe I also see it as frightening or delicious-looking, which are mind-involving, but the core of my perceptual representation consists of attributions of objectively instantiated properties—shape, distance, relative position, texture, and so on. I see an object as having edges and by God it has edges! But are things really that simple? Do we see things exactly as they objectively are? Let’s start with color and other secondary qualities: don’t these really have their origin in the mind? Certainly there is a long tradition that supposes so. Again, I won’t go into this question in detail, but it surely is possible that color in objects consists of a disposition to give rise to sensory experiences of a certain sort. And isn’t the way an organism sees color a result of its specific sensory make-up? The same color might elicit fear in one animal but joy in another, depending on what the object means to the animal (is it predator or prey?). Couldn’t there be creatures that by stipulation project colors onto objects with no detriment to the claim that they nevertheless perceive objective physical particulars? This may be admitted but denied that primary qualities are mentally projected: things surely have shape independently of whether they are perceived to have shape! That may well be so, but is it the same shape? We are familiar with the idea that visual geometry is Euclidian while physical geometry is not, so that how we perceive figure differs from how figure objectively is. Also, there is that old point of Plato’s, namely that nothing in nature is ever perfectly circular or rectangular or straight. These are ideals dealt with in pure geometry, but they don’t apply to the empirical world. But don’t we see things in accordance with this natural geometry? The discus looks round, the picture frame rectangular, etc. We see and feel edges as smooth and continuous while in (microscopic) reality they are bumpy and discontinuous. We see things as solid when they are not, being mainly empty space; or empty when they have tiny particles in them, like air. We perceive spatial relations egocentrically, which they are not intrinsically. Then we have that old chestnut: the circular coin that looks elliptical from an angle. An elliptical representation surely enters into such a perception (no matter what we think), but the coin itself lacks this property. All perceptions come with a mode of presentation (visual constancies provide a good example), but modes of presentation are not part of the reality that we perceive—they are not objective features of the physical world. It is hard to think of any perceptual encounter that is free of subjective elements, i.e. mental representations that incorporate the perspective of the subject. We don’t perceive the world sub specie aeternitatus. We never perceive the world just as it objectively is sans any subjective intrusion; no organism does. We always bring a point of view (the Lebenswelt).
This is not remotely surprising given how perceptual systems evolved. Perception evolved in conjunction with (inseparably from) the organism’s motor system, and the motor system is a practical capacity. Like all biological systems the sensorimotor system obeys principles of economy and has inherent limitations; it doesn’t build in more than is necessary to achieve its biological purpose (survival, reproduction). Perceptual representational primitives are geared to practical ends not to ideal science, so they suffice to carry out their function if they approximate to objective reality; complete veridicality is not in their job description. Getting the exact geometry of the physical world right is not their aim, so long as there is an adequate correlation between how they represent things and how things objectively are. We don’t find systematic radical discrepancies between perceptual content and objective reality, because that would thwart the purposes of the sensorimotor system; but it can tolerate small degrees of inaccuracy or coarseness or opacity. Animals don’t need “microspical eyes” (to use Locke’s phrase). As long as the perceived external object is represented in practically useful ways, the sensorimotor system is doing its job; it can get by with non-veridical representations as long as they make no difference to practical outcomes. Indeed too much accuracy and precision might interfere with the smooth functioning of the organism (compare vagueness). Of course the human visual system evolved from much simpler visual systems and inherits many of their design features, so we should expect it to have the pragmatic character of earlier systems. Thus some degree of objectivity is desirable, but not complete veridicality, absolute precision. The mind generates perceptual primitives that suit its biological purposes, and these map onto objective reality without necessarily coinciding with it. Thus the properties attributed to external objects by the sensorimotor system are not possessed by those objects as they are independently. Those smooth surfaces that we see and touch are not possessed by physical objects as such; their surfaces are granular and discontinuous—so they are not parts of physical objects objectively considered. The data of sense consist of attributes cobbled together by the mind (ultimately the genes) over the long course of evolution, so sense-datum theory is not wrong on this score (however mangled its expression tended to be). We perceive objects as we represent them for our specific purposes not as they independently are (precisely, objectively, absolutely). Of course objects are disposed to cause in us perceptions that only approximate to their intrinsic objective properties, so they have these dispositional properties; but the properties we attribute are not identical with any properties of the object in itself. Yet we are seeing that object: it is the particular that our perception singles out for comment (so to speak). So our perceptual states have two sides: the objective physical particular that is singled out and the mind-contributed property that is attributed to that particular. We are both locked in our head (as in the traditional sense-datum theory) and oriented to outer reality (as naïve realism supposes). But we are not completely locked in our head, because of the correlations and approximations I have mentioned. We are not a mirror but we are not a black box either.
There is a question about how mental the perceptual attributes are, and hence how subjective perception is. Earlier sense-datum theorists tended to take sense-data as completely mental—assuming that perceived shape and color were themselves mental phenomena (though not all did this: some took them to be neutral between mental and physical). I would say they are not mental at all, not strictly speaking anyway. I don’t think color properties are really mental properties, despite being projected by the mind. I also believe that geometrical attributes are not themselves mental: being circular, say, is not a property of the mind. They are “abstract” for want of a better word. My point has just been that the attributes ascribed by the perceptual system to physical particulars are not in general identical to attributes objectively possessed by those particulars, so how we see things is not (precisely) how they are objectively. I don’t say that these attributes are mental or subjective, though it is true that they are projected by the mind (they are probably lurking somewhere in the genes). Certainly they are not derived from external objects in the manner of classical empiricism. They are “subjective” only in the sense that they have their origin in the mind not in the sense that they are attributes of the mind (like pain or belief). What I have chiefly wanted to urge is that it is perfectly consistent to maintain a hybrid view of perception, and that both sides of this composite picture are plausible. Perception is of (de re) physical particulars outside the perceiving subject and yet it is as of (de dicto) properties that fail to coincide with the properties objectively possessed by such particulars. Perceptual reference is to physical particulars, but perceptual predication is not of physical properties (i.e. those objectively possessed by physical particulars). So both naïve realism and the sense-datum theory are partly wrong and partly not wrong.
 I am adopting some of the terminology developed by Tyler Burge in his Perception: First Form of Mind (2022).
 See J.L. Austin’s Sense and Sensibilia (1962) on the mangling.
 Burge doesn’t discuss the hybrid theory, though it seems to me consistent with his overall position (except that he tends towards attribute veridicality). I should note that a projective view of perceptual properties is consistent with the existence of perceptual constancies with respect to such properties.