Hallucinations may cause people to believe they have seen a ghost

The argument

This argument holds that ghosts are not real because the sightings offered as evidence are explained by hallucination: the brain, under a range of ordinary conditions, can generate vivid perceptions of people and presences that are not there, and that is what 'seeing a ghost' actually is. The science of hallucination is well established. The human brain does not passively record reality; it constructs perception, and under certain conditions it constructs things that are not present. Grief is a powerful trigger — a large proportion of recently bereaved people report seeing or hearing the dead, a recognised and common experience. Sleep states produce others: hypnagogic and hypnopompic hallucinations at the edges of sleep, and sleep paralysis, routinely generate vivid figures and a felt presence in the room. Fatigue, fever, dehydration, certain drugs and toxins (including carbon-monoxide exposure, which has been tied to classic 'haunting' reports), sensory deprivation in dark and quiet old buildings, and neurological conditions can all conjure convincing apparitions. People in these states genuinely see something — but the something is generated inside the skull, not standing in the room. The argument then applies the basic standard of explanation. Ghost sightings have two candidate causes: that the dead persist and appear, which would overturn the known laws of physics and biology, or that a brain produced a hallucination, a phenomenon already documented, reproducible and common. The second requires nothing new and accounts for the reports — including why sightings cluster around grief, fear, exhaustion and suggestive environments rather than appearing reliably to neutral observers or instruments. From this standpoint, the apparition is a real perception of an unreal thing. Because hallucinations may cause people to believe they have seen a ghost, this argument holds, ghosts are not real.

Premises

Counter-arguments

Critics — including some who share the conclusion — reply that the argument establishes less than it claims. Showing that a phenomenon can be explained without ghosts is not the same as showing there are none: an available alternative explanation makes belief unwarranted rather than makes the belief false, and the step from "hallucination could account for this report" to "ghosts are not real" is a move from a possible explanation to a negative existential that does not follow. Those defending the rival position add that the mechanism fits some reports better than others. Hallucination is a poor account of episodes with multiple simultaneous witnesses reporting the same figure, or of reports from people not bereaved, exhausted, medicated or in a suggestive setting — and the conditions the argument lists are broad enough that almost any report can be assigned to one of them after the fact, which weakens the explanation rather than strengthening it. Critics note that the genuinely load-bearing evidence is different and stronger: that apparitions have not been recorded under controlled conditions despite sustained attempts. That supports the conclusion the argument wants, but by a route the argument has not taken.

Rejecting the premises

[Rejecting P2] That hallucinatory mechanisms are documented and common shows some reports can be explained without ghosts, not that all are; and the list of triggers is broad enough that nearly any report can be assigned to one after the fact, which weakens the explanation rather than confirming it. [Rejecting P3] Preferring the explanation that requires nothing new is a rule about which belief is warranted, not a demonstration that the alternative is false — and hallucination fits poorly the reports involving multiple simultaneous witnesses to the same figure. [Rejecting C] The step from "hallucination could account for this" to "ghosts are not real" moves from a possible explanation to a negative existential; the stronger route to the same conclusion is the failure of apparitions to appear under controlled observation, which the argument does not use.