In undertaking this wide-ranging, compassionate and ultimately revelatory survey of the strange terrain of humans' delusional capability, Oliver Sacks says he has been inspired by William James's example – in The Varieties of Religious Experience and elsewhere – to create a sort of "natural history or anthology" of the hallucinatory. "My own favourite definition" of an hallucination, Sacks writes in one of his characteristically compendious footnotes, "is that given by William James in his 1890 Principles of Psychology: 'An hallucination is a strictly sensational form of consciousness, as good and true a sensation as if there were a real object there. The object happens not to be there, that is all' (his italics)."

So, following this plan – which appeals to common sense by eliding tricky questions about the reality or otherwise of the external world with a strictly phenomenological approach – Sacks introduces us to a realm of compelling oddities. He begins with Charles Bonnet Syndrome, which affects a significant proportion of blind and visually impaired people. It can give rise to hallucinations that range from simple geometrical patterns and phosphenes (rings or blots of colour), through galleries of faces – both realistic, cartoon-like and even horribly grotesque – to entire moving scenes, often featuring figures in "exotic dress". CBS hallucinations differ markedly from our conception of normal mental imagery in having far greater complexity; and when those experiencing them are observed, their sightless eyes do indeed saccade across and then focus on these objects that happen not to be there.

From his earliest writings Sacks has managed to convey in relatively accessible prose (I say "relatively", because of necessity he employs a wide range of specialist neurological and other medical terminology) the forensic nature of his discipline. Following the founding father of modern neurology AR Luria – whose Mind of a Mnemonist and The Man with a Shattered World he took as his templates – Sacks has presented individual case histories, linked them to diagnostic criteria, and then shown how particular psychic phenomena may well have a discoverable organic basis. Luria's patients were Soviet second world war casualties, often with pinpoint wounds that had produced localised brain damage, and, like him, Sacks proceeds from the palpable effect of a given impairment, such as blindness, to hypotheses concerning the functional organisation of the brain. So, with CBS, the tendency of these hallucinations to consist of strongly defined images, although not determinate things (nothing, for example, that the hallucinator remembers from their sighted experience), suggests to Sacks that "at some lower level, in the early visual system, there is a categorical dictionary of images or part images – of generic 'noses', for example, or 'headwear'." It follows that in the absence of expected visual stimuli, the brain gets to work generating its own phantasmal world, using this lexicon of innate types. Sacks delves deeper into this phenomenon, quoting William Burke, a neurophysiologist who has himself experienced the geometric hallucinations associated with CBS and who hypothesises that the actual imagery may well reflect the very anatomy of the visual cortex itself.

CBS hallucinations, while they may be extremely fanciful and fantastical, produce remarkably little emotional impact on those who experience them, and Sacks is at pains to distinguish these from other phenomena – such as the hectoring, persecutory voices endured by schizophrenics or the classic "bad trip" occasioned by LSD – that may lead to lasting psychological disturbance. With CBS, Sacks was surprised, given his own clinical experience, by how little prevalence the literature indicated historically. Contemporary studies suggest that complex hallucinations may be experienced by up to 15% of blind and partially sighted people, while 80% may be subject to the basic geometric type. Sacks thinks that the failure to report these hallucinations – and many others that have no implications for mental pathology – may be a function of people not wishing to appear crazy.

Certainly, once the whole panoply of the hallucinatory has been brought into plain view – from the sharp "tears" in the visual field and multicoloured swirls associated with migraine, to rare Lilliputian hallucinations in which people see tiny humanoid figures – it seems hard to resist Sacks's contention that these phenomena have always "had an important place in our mental economy", influencing traditional art, folklore and even – for example, in the case of the distinctive transcendent "aura" that precedes an epileptic fit – "generating our sense of the divine".

Divine some hallucinations may be, but most of those recounted by Sacks seem on a sliding scale between the mundane and the terrifying: a sufferer from the herpes simplex virus, which can attack the olfactory nerve, found himself assailed first by the smell of wet cardboard and then – for an entire year! – by that of rotting fish. When this stench finally blew through his mind it took with it his ability to discriminate smell (and by extension, taste), leaving behind only the cruel taunt of an olfactory hallucination as a precursor to his next attack of neuritis. Another sufferer – this time from Parkinson's – began to notice tactile hallucinations soon after being diagnosed: "the surfaces of various objects were covered by a film of fuzz, like peach fuzz, or the down in a pillow." Another Parkinsonian, almost certainly suffering from "Lewy bodies" (abnormal aggregates of protein inside the nerve cells), saw myriad little figures – "Chuckys", she called them – running around her apartment, as well as horror scenes enacted in front of her: "I saw my son murdered right in front of my eyes."

For those unacquainted with Sacks's earlier work, Hallucinations is a perfectly respectable place to start, but for those already familiar with such classic works as The Man Who Mistook his Wife for a Hat, Awakenings and the more recent Musicophilia, there is pleasure to be gained from the way he cross-references, with one case history – say, in the case of the latter, of musical hallucination (some people have both auditory hallucinations of music they've never heard, and also "see" equally bizarre musical notation plastered all over their visual field) – reappearing, and being re-examined from a different angle. This general inspissation of the Sacks worldview can seem both stimulating and disturbing, and about of a third of the way into Hallucinations, I began to feel that the entire world must be full of people furiously hallucinating; while my own phenomenological oddities – migraine, sleep paralysis, quite a lot of major psychotropic-induced visions – seemed curiously diminished.

There was also – for me – a paradoxical sense of distancing from Sacks himself. He's never afraid of putting himself in his own narratives as a character. He has also written a rather wonderful "scientific memoir" of his childhood, Uncle Tungsten, and he always – and with exemplary humanity – makes it clear when he's writing about his own patients. But for all this, the sheer strangeness of his material has, over the years, tended to make his undoubtedly compassionate involvement seem nonetheless a little bit creepy. I was getting this quite familiar sensation from Hallucinations, when, in the section on the visions engendered by intoxicants, the whole book caught fire. Writing without a scintilla of sensationalism, Sacks relates his experiences in the late 60s with LSD, mescaline, cannabis, amphetamine, chloral hydrate and even injectable morphine, culminating in the astonishingly complex and extreme hallucinations occasioned by his taking 20 Artane pills (a synthetic drug related to belladonna).

Suggested to him by his pals on Muscle Beach (a casual aside in the text, but I happen to know that Sacks was a champion weightlifter in his youth), the Artane provoked completely veridical hallucinations of friends coming to visit him in his LA home, his parents arriving by helicopter, and even a conversation with a philosophic spider who's opening comment was: "did I think that Bertrand Russell had exploded Frege's paradox?" Normally such wacky reminiscences are a turn-off for the enquiring reader, but in the context of Sacks's natural history of hallucinations they are revelatory, making of the author not simply an interpreter of others' dreams, but one who experiences them as well. The last two-thirds of the book gain immeasurably in force and depth from the centrality he himself has taken in the narrative; indeed, for seasoned Sacks-watchers, I would say that Hallucinations is really the keystone of the amazing edifice that is this remarkable thinker's oeuvre; a body of work that sets out to do nothing more or less than examine the totality of human being from the perspective of neurology.

• Will Self's Umbrella is published by Bloomsbury.