The copy has to be exact
Imagine someone taking a sip of lemon tea. They notice the sharpness, recognize the cup, remember yesterday’s drink and decide it needs honey. Now imagine a physical duplicate performing precisely the same sequence without experiencing a taste, a color or anything else.
That last instruction creates the philosophical zombie. It is not a person pretending, emotionally numb or seeing only darkness. Each of those descriptions leaves some experience. The proposed zombie has none.
The Stanford Encyclopedia of Philosophy’s entry stresses that matching behavior is insufficient for the challenge to physicalism. A robot delivering the right sentence would not do. Every physical detail must match, including the processes that produce memory, attention and reports about experience.
A whole zombie world makes the condition especially clear: duplicate the physical universe and its history, then suppose experience is absent. The dispute is whether the final supposition adds a coherent possibility or contradicts what the duplication already includes.
What would the argument establish?
Physicalism, in the form challenged here, says that the complete physical facts determine the facts about consciousness. It need not dismiss feelings as unreal. It can instead hold that the experience is constituted by the relevant physical reality.
In Chalmers’s formulation, the argument moves through three substantive claims: a physical duplicate without experience is conceivable; its conceivability establishes the relevant possibility; that possibility is incompatible with physicalism. The steps should be kept separate.
If a zombie world were genuinely possible, duplicating the physical facts would leave something about consciousness undecided. That is the proposed challenge. Merely finding the story intriguing establishes much less. You could accept that it seems conceivable while rejecting that seeming as evidence of possibility.
A useful way to read the argument is to pause at each inference. Is the starting description coherent? What turns that coherence into a claim about reality? Which version of physicalism would the result contradict? Disagreement at any of those points changes what follows.
Possibility is not a manufacturing forecast
In a Berkeley discussion published in July 2026, Chalmers explicitly separates metaphysical zombie possibilities from his expectations about our world. He says he does not think zombies are actually present and expects an actual physical duplicate to share his consciousness under the laws connecting brains and experience.
The distinction matters. An argument about whether reality could have been otherwise is not a claim that a laboratory can remove experience while leaving a person physically unchanged under current natural laws.
It also prevents the puzzle from turning into a suspicion about other people. A debated description of a possible world is not an observation about a neighbor. Nothing in the thought experiment identifies an actual person whose inner life is missing.
Dennett asks whether we imagined enough
Daniel Dennett’s original objection challenges the ease with which people accept the description. A complete duplicate must also duplicate higher-order reflection: considering its own thoughts, discussing its sensations and reasoning about why consciousness seems puzzling.
Return to the tea. The duplicate must remember how the drink seemed yesterday, compare that with today’s cup, explain the comparison and answer follow-up questions exactly as the conscious original would. We cannot make the copy blank or confused whenever the discussion turns inward. That would change the stipulated capacities.
Dennett doubts that a further removable ingredient remains once the relevant organization is taken seriously. The functionalism discussed in the Stanford entry similarly connects mental states to roles within a system, including internal relationships as well as outputs. On such an account, preserving the constitutive organization while subtracting the state is not an available operation.
Zombie defenders answer that explaining capacities may still leave out why those capacities feel like something. Neither side settles the question simply by treating its preferred account of consciousness as already established.
An easy picture is not ideal conceivability
In Does Conceivability Entail Possibility?, Chalmers distinguishes what seems coherent initially from what would survive ideal rational reflection. He also distinguishes failing to rule something out from forming a coherent positive conception of it.
These distinctions block a careless reading of his argument. It does not say that everything someone can sketch must exist in a possible world. Our first impression may miss a contradiction. The stronger claim concerns a kind of conceivability that would not collapse under better reasoning.
That raises a practical difficulty for the zombie defender: how do we establish that this description passes that demanding test? Attaching a label reading “no experience” to an otherwise complete duplicate is easy. Showing that the label expresses a coherent situation is the work the argument needs done.
A gap in understanding might be a gap in concepts
Another physicalist response accepts the apparent explanatory gap but locates it in how we think. Knowing an experience from the inside and describing a physical process use different concepts; that difference need not show that there are two separate things.
Chalmers examines this approach in Phenomenal Concepts and the Explanatory Gap. Its defenders try to explain why the gap feels persistent while retaining physicalism. Chalmers argues that the strategy cannot do all the explanatory work required. Citing his discussion is not the same as treating his objection as a settled verdict.
For the tea drinker, the question becomes whether an unfamiliar route from a brain description to a taste reveals something missing in reality, or something incomplete in our understanding. Calling both gaps “unexplained” does not make them equivalent.
The AI question needs its own evidence
A language model talking about feelings does not meet the zombie stipulation. Matching a few statements does not establish physical identity with a conscious person. The Berkeley discussion also notes the difficulty of applying human neural correlates directly to systems without neurons.
Our article on whether an AI could be conscious examines that separate evidence problem. For the larger distinction between explaining functions and explaining experience, see the hard problem of consciousness.
A proposed test can easily change the puzzle. If the duplicate fails to remember the tea or gives different answers about it, the copy is no longer exact. If every physical and behavioural detail stays fixed, a test using those details cannot reveal the stipulated difference.
The work is therefore in the argument about what the complete physical account entails. Does it already fix the taste of the tea, or could all of it remain while experience is absent? That is the disputed step, not an overlooked interview question for an unconscious twin.
