An exercise in exorcism

For 300 years, the mind-matter dualism of René Descartes dominated science and philosophy of mind and kept them separate. Then, in the middle of the 20th century, Gilbert Ryle undertook his self-styled "hatchet job" on Descartes's dogma of the Ghost in the Machine. Suddenly, the scientific study of consciousness was philosophically respectable and among Ryle's students one in particular rose to the challenge: Daniel Dennett.

Dennett's views are laid out with great clarity in Consciousness Explained . He claims that most materialist philosophers, having publicly shown Descartes's "ghost" the door, were letting it slip back. He coined the term "Cartesian materialism" for the erroneous belief that at some specific point in the brain, the hitherto unconscious processing of information from the sense organs is transformed into conscious experience. He dubs this imaginary boundary the "Cartesian theatre" and insists that it was a wrong-headed concept. It was wrong because it implied an imaginary audience - a subject, a conscious self - who has the experience. Dress it up how you will, that is Descartes's ghost back in the machine.

Dennett's opponents, from veteran John Searle (a fellow student of Ryle's) to new-generation David Chalmers, accused him of denying consciousness altogether, of explaining it away rather than explaining it. In Sweet Dreams , Dennett revisits battles that he has fought over two decades. He denies denying consciousness altogether, but insists it is not what people think it is.

The litmus test for whether you are with him or against him is whether you find zombies conceivable. A zombie in this context is a fictional creature that is physically and functionally identical to a human being but without consciousness. Only if you think of consciousness as a non-physical added extra can you possibly conceive of such a being, and in that case you have not fully exorcised your Cartesian ghost. Nobody believes that zombies actually exist, but Searle and Chalmers say they are conceivable. Dennett finds the idea "preposterous" and cannot understand why so many neuroscientists are attracted by it.

Next, he sets out what he claims is the only method for studying consciousness scientifically: heterophenomenology, so-called because it is a third-person investigation of someone else's first-person phenomenology or conscious experience. Dennett claims that a science of consciousness should study and seek to explain a person's beliefs about his or her experiences. His critics say a satisfactory theory considers experiences themselves, but he argues against this on grounds of their inaccessibility (in theory and practice).

Another old friend to emerge in a new guise in Sweet Dreams is philosopher Frank Jackson's "Mary", the fictitious neuroscientist whose transformation from a monochrome world to a coloured one has long been a test-bed for theories of qualia (phenomenal qualities such as the colour red). Dennett tries to rehabilitate the battered fable by presenting RoboMary, a more tractable creature than her human counterpart. Dennett derides qualia because philosophers have endowed the ill-defined term with "a variety of ill-considered associations and special powers", but it will take more than RoboMary to kill them off.

When Dennett followed in Ryle's steps and exiled the conscious self, he propounded the multiple-drafts model of consciousness in its place. Though widely discussed, this theory was rarely adopted, so he developed a "more imagination-friendly antidote to the Cartesian imagery", called the fame-in-the-brain model. This stated that what distinguishes conscious from unconscious contents of the brain is ubiquity. There is no change in kind, only in quantity. Crucially, this change does not cause something else to happen that brings about consciousness, it simply is consciousness. Now fame-in-the-brain has itself given way to a new theory: the Fantasy Echo.

The nearest we shall ever get to a defining feature of consciousness is, Dennett says, our capacity to relive or rekindle contentful events. His best guess is that this echoic capacity is due in large part to habits of self-stimulation picked up from human culture.

Dennett is always stimulating, always provocative and often more convincing than is comfortable. All that is true of this volume, but it is marred by being something of a ragbag of talks and articles produced for other occasions. This results in too much repetition - sometimes of whole paragraphs - to be entirely satisfying.

Anthony Freeman is managing editor, Journal of Consciousness Studies .