Italian-American neuroscientist, born 1960

Giulio Tononi

A digital camera can tell apart millions of images, yet nobody thinks it sees anything. Tononi's answer is that its pixels never form a whole: each works alone. Consciousness, he proposes, is information that is integrated, and the amount can in principle be measured.

Parisians relax on a sunny riverside lawn, among them a couple in dark clothes with a parasol and a small monkey on a lead. Up close, the whole scene is made of tiny separate dots of colour.
Georges Seurat, A Sunday on La Grande Jatte, 1884–86. Thousands of separate dots, and yet we see one scene. A camera sensor has the dots without the scene. For Tononi, the difference between the two is the difference consciousness makes.Art Institute of Chicago. Public domain, via Wikimedia Commons.

The view

Tononi trained in medicine and psychiatry in Pisa, worked with the Nobel laureate Gerald Edelman in the United States, and leads a laboratory at the University of Wisconsin–Madison that studies sleep and consciousness. He first set out integrated information theory, or IIT, in 2004, and has revised it several times since. The neuroscientist Christof Koch, who spent years looking for the neural correlates of consciousness with Francis Crick, became its best-known supporter.

IIT runs in the opposite direction from most theories. Instead of starting with the brain and asking how it produces experience, it starts with experience and asks what a physical system would have to be like to have it. Every experience, Tononi says, exists for itself, is structured, is the specific experience it is, is unified and is definite. From these features he derives conditions on physical systems, and arrives at a quantity he calls phi (Φ): roughly, how much a system as a whole does over and above what its parts do separately. A system is conscious to the degree that it has phi, and what its experience is like is fixed by the shape of its cause-and-effect structure.

The argument

Being conscious is not a matter of what a system does, but of what it is: how far its parts make a difference to one another as a single whole. After Tononi, "An Information Integration Theory of Consciousness" (2004)

The camera is the starting point. Its million photodiodes between them can distinguish a vast number of images, so in one sense it holds a lot of information. But each diode responds only to its own patch of light and has no effect on any other. Cut the sensor in half and each half works exactly as before: nothing is lost, because nothing was shared. Its phi is zero. A brain is different. Its parts constantly constrain one another, so cutting it would destroy information that exists only at the level of the whole. Try cutting each system below.

Choose a system, then cut it in two.

The theory explains some puzzling facts about the brain. The cerebellum has more neurons than the rest of the brain put together, yet people born without one are fully conscious; IIT says its parallel, modular wiring integrates little. Consciousness fades in dreamless sleep and under anaesthesia even though neurons keep firing; IIT says integration breaks down. Tononi and Marcello Massimini turned this into a practical test, nicknamed "zap and zip": stimulate the brain with a magnetic pulse, record the echo with EEG, and measure how complex it is. The index has helped to detect hidden consciousness in patients who cannot respond.

Other consequences are stranger. A network whose signals run only one way, however clever its behaviour, has no phi at all. A perfect simulation of your brain running on an ordinary computer would behave exactly as you do and feel nothing, because the computer's own physical wiring integrates almost nothing. And some very simple structures, such as a large regular grid of logic gates, would have more phi than a human brain. The computer scientist Scott Aaronson pointed this out in 2014; Tononi accepted the result. In 2023 more than a hundred researchers signed an open letter calling IIT pseudoscience, which in turn drew sharp replies. It remains among the most debated theories in the field.

Where he meets Dennett

Not often, but there is some common ground. Both think consciousness is a natural phenomenon that science can study, with no need for a soul or a separate mental substance. Both think it comes in degrees rather than being simply on or off, and both expect to find it, in some form, in many animals. And Dennett, who always asked what a theory could do in practice, would have recognised the value of a measure that helps doctors tell whether an unresponsive patient is aware, whatever the theory behind it.

Where he parts

At the root. IIT says that consciousness is an intrinsic property, there for the system itself, and fixed by physical structure rather than by what the system does. That is precisely the kind of property Dennett argued does not exist, and that illusionists think the brain wrongly represents itself as having. On IIT, two systems could behave identically in every way, one conscious and one not; for Dennett that would make consciousness a difference that makes no difference. IIT also starts from introspection, treating the apparent features of experience as self-evident truths, which is the step Dennett's heterophenomenology refuses to take. And where Dennett thought a machine that does what we do would be conscious, IIT says that no program on a conventional computer ever could be, while a grid of simple gates might be more conscious than you. See Could a machine be conscious?

Tononi and Dennett side by side
TononiDennett
What is the mind?Consciousness is integrated information: a physical system's cause-and-effect structure.What the brain does: a pattern of functions.
Is phenomenal consciousness real?Yes: it is the starting point, and the one thing we cannot doubt.No, not as philosophers define it. It is an illusion produced by the brain.
Could a machine have a mind?Not by running software on ordinary chips. Hardware with the right physical structure might.Yes, in principle, if it does what we do.
Is there a hard problem?Answered by starting from experience and deriving its physical basis.No. There is the Hard Question: and then what happens?

See every view compared → · Could a machine be conscious? →

Further reading