This must surely be the only book that has ever been devoted to the philosophy of shadows and our perception of them. It might be thought that the subject could at best be one of only minor interest, because it is natural to assume that shadows, being so insubstantial, are things of little importance from either a theoretical or a practical point of view. In fact, however, nothing could be further from the truth. Much of our ability to see ordinary material objects depends crucially on our ability to see their shadows, both those that they cast and those that are cast upon them. For example, top-lit faces are vastly easier for us to recognize than bottom-lit faces. And the metaphysics of shadows is an immensely interesting topic in its own right, alongside the metaphysics of equally insubstantial things, such as holes and cavities.

Here are some of the intriguing questions that Roy Sorensen raises about shadows. What, if anything, are shadows? Are they presences or absences? Can we literally see them? Can we only see them, or can we also feel them? Do they have colours? If so, can they have chromatic colours, or are they only ever black or grey? How do shadows differ from reflections and silhouettes? Can a shadow have a reflection? Is the shadow of a spinning sphere stationary, or does it also spin? Can we see anything when we are in complete darkness? All of these perplexing questions, along with very many more, Sorensen attempts to answer in this remarkable book, which will be of interest not just to philosophers and psychologists but to all readers who retain any sense of the mystery that shadows have for children.

Sorensen begins by distinguishing shadows from silhouettes, focusing our thoughts by the riddle of the ‘double eclipse’. If Earth had two moons, a smaller but nearer one, Near, and a farther but bigger one, Far, then they could, by lining up between Earth and the sun, produce a double eclipse, with Far eclipsing the sun and Near eclipsing Far. In such circumstances, would we be seeing Near or Far when we viewed the double eclipse? Most people are inclined to say Near, since it is in front of Far and they assume that one cannot see an opaque body when it lies behind another. That is true when we see an opaque body in virtue of its reflecting light into our eyes, but in an eclipse we see the eclipsing body in virtue of its blocking light that would otherwise have entered our eyes. Sorensen argues that when we view the double eclipse we must in fact be seeing Far – and that we see it by seeing its silhouette. He contends that Far's silhouette is to be identified with the surface of Far that is nearest to the sun – the surface that blocks the sun's light from us. According to Sorensen, we see a silhouetted object by seeing its surface that is furthermost from us – its back surface – because it is that surface that is causally responsible for blocking light from our eyes. This is exactly the reverse of the case with objects that are seen in virtue of their reflecting light into our eyes, which we see by seeing their surface nearest to us.

Interesting though it is, it is not clear to me that this analysis is correct. Sorensen's claim that an object's silhouette is its back surface is based on two contentions: (i) that ‘I see an object only if I see part of it’ and (ii) that ‘That part must cause my perception in an appropriate way’ (50). For he thinks that the object's back surface is the only part of it that can play that causal role when we see an object in silhouette. However, it the first place, I don't agree that one can see an object only by seeing part of it. I think that one can literally see a distant jet plane by seeing its vapour-trail, or see a charged particle in a bubble-chamber by seeing the track of bubbles that it creates, or indeed see a footballer on television during a game that is televised ‘live’. Secondly, the following hypothetical example seems to undermine Sorensen's claim that an object's silhouette is its back surface. Surely, an infinitely deep object seen against a bright background would still have a silhouette, even though, ex hypothesi, it would have no back surface. Imagine an infinitely long cylindrical solid, with just one circular flat end, viewed end-on against a bright sky: this would surely appear to have a silhouette indistinguishable from that of finitely long cylindrical solid viewed from the same angle – and yet, according to Sorensen, only the second object would actually have a silhouette. Thirdly, when an object is both front-lit and back-lit, its back surface is still blocking light from our eyes and affecting our perception of the object as a consequence – and yet, in such a case, it seems manifestly wrong to say either that we are seeing its back surface or that we are seeing the object ‘in silhouette’.

In fact, it seems to me that if one wants to identify some part or feature of a material object – a house, say – with its silhouette, a much more plausible candidate than its back surface would be its outline in the direction of view (a suggestion which standard dictionary definitions of ‘silhouette’ appear to support). This ‘outline’, it seems, is a linear – that is, one-dimensional – and flat visual object. It is pretty clearly not a material object, unlike the house itself. It cannot even be identified with the perimeter of any cross-section of the house since it is, geometrically speaking, a projection of a set of points on various different edges of the house, lying in different planes, on to a plane orthogonal to the direction of view. (I suppose it might, in principle, be possible to identify the outline with the set of points in question, all of which lie somewhere on the surface of the house: but then we would have to say, counter-intuitively, that the ‘outline’ need not even be a single connected line, but could be a set of disconnected lines in various different planes. If we did take this view, we would have to distinguish between the outline's ‘real’ shape and its ‘projective’ shape, which seems an unnecessary complication.) The shape of the outline (assuming that its real shape is just what it appears to be) is what would be captured by a photograph of the house taken in the given direction of view: it would be the shape of the perimeter of the house's photographic image. Equally, it would be the shape of a piece of card which, if held up in front of the house in a plane orthogonal to the direction of view, would serve to occlude the house exactly. Unsurprisingly, trained artists are very skilful at observing and representing the outlines of objects, but really they are external visual objects with which all of us are very familiar from everyday observation.

Imagine a house seen in silhouette against the setting sun. The silhouette surely looks flat and appears to have a shape identical to that of the house's ‘outline’ in the direction of view, as I have just described the latter. So why not identify the two? On Sorensen's theory, by contrast, the house's silhouette, although it looks flat, in fact has all the protuberances and depressions of the back surface of the house. One could never guess this simply by seeing the house in silhouette against the setting sun, only by looking at the house from its sunlit side. It seems strange almost to the point of paradox to say that what one is then seeing – the front surface of the house from that sunlit side – is the very same thing as is seen by the person who sees the house in silhouette from its opposite side. In any case, the outline seems to have at least as good a claim as the back surface to be causally implicated in our seeing of the house: for it is the outline that is replicated in a photographic image, whether the house is photographed in silhouette or in sunlight – and the same holds with regard to the retinal images formed in our eyes when we see the house in either condition. Moreover, if we watch the house continuously from midday to sunset, there will surely be no precise moment at which its silhouette finally becomes visible in addition to, or instead of, its outline: which argues in favour of our simply identifying the silhouette with the outline and saying that a ‘silhouette’ is simply an object's outline as seen when very little or no visible light is reflected from its front surface. Finally, even our hypothetical infinitely deep object has an outline, although not a back surface.

Turning now to shadows themselves, Sorensen urges, first of all, that we should not confuse them with their two-dimensional cross-sections which are cast upon the surfaces of objects lying in their way, since shadows – he maintains – are really three-dimensional in shape. According to Sorensen, shadows are perfectly real enough: but they are not ‘positive’ beings, since they are absences of light – ‘holes in the light’, as he graphically calls them – which are created by opaque material objects when they block light that is incident upon them. Although Sorensen follows most philosophers in adhering to a causal theory of perception, he still maintains that we can literally see shadows, since he rejects the commonly accepted view that absences – or ‘privations’ – cannot be causes. However, whereas he allows that one may see an object by seeing its silhouette, he denies that one may see it by seeing its shadow, on the grounds that the silhouette, but not the shadow, is a part of the object. The shadow, he considers, is only an effect of the object and one cannot – or so he maintains – see an object by seeing its effects. (As I have already implied, he must therefore deny – counterintuitive though this will seem to many – that one can really see an object by means of a television relay.)

However, even if it is accepted that one cannot see an object by seeing its effects, it is both questionable whether one can see an object only by seeing a part of it and also questionable whether shadows are in fact best described as being effects of objects. In the first place, surely one can see an object by seeing a feature of it which is not a part of it – its ‘outline’, as described earlier, plausibly providing an example of this – and, in the second, an object's shadow seems better categorized as being one of its features rather than merely one of its effects. After all, an object's shadow is not ‘detachable’ from it, in the sense that it can go on existing without the object – unlike, for example, a jet plane's vapour-trail or someone's footprint on a sandy beach. In sum, I am inclined to contend, in opposition to Sorensen, that both silhouettes and shadows are features of objects – neither parts nor mere effects of them – by seeing which we can see those objects. Indeed, if Sorensen were right in saying that we can't see objects by seeing their shadows, it would be much harder to understand how it is that – as was mentioned earlier – top-lit objects are typically so much easier for us to see and recognize than bottom-lit ones. Their shadows are much more important than their outlines or colours in this regard. In fact, it is perfectly possible to depict a visually recognizable object by depicting only its shadows – for example, in a charcoal sketch of a human face, or in a grey-tone painting of a white cloth with many folds and creases in it.

Curiously enough, Sorensen denies that we can feel shadows – so that talk of ‘feeling a cool shade’ cannot, according to him, be construed as literally being true – on the grounds that temperature is possessed only by things containing molecules in motion, being ‘defined’ as their average kinetic energy (118). This, as far as I can tell, is one of Sorensen's few slips of a purely scientific nature, since modern physics apparently allows that even the vacuum has a temperature. Sorensen also denies that shadows can have any colour other than black, contending that we need to distinguish shadows themselves from the ‘light pollution’ that may sometimes penetrate and fill them, rather as extraneous matter may sometimes penetrate and fill a hole in a material object. Sorensen coins a snappy new word, ‘filtow’, for the kind of body of coloured light, produced by a filter, that may, according to him, overlap or even coincide with a shadow.

Sorensen's book is certainly fascinating and richly thought-provoking. It discusses many intriguing topics that I have not even been able to touch upon. Much of what he says may well induce doubt and even incredulity, at least on a first reading. But he argues carefully and clearly in favour of his key claims, all of which merit very serious consideration, even if they sometimes provoke one to construct and defend alternative views. That, however, is surely the hallmark of the very best kind of philosophy writing. Seeing Dark Things is a model of this kind.