Why the singular of “data” is not “anecdote”

This column was first published in the November 2018 issue of Scientific American.

For a documentary on horror movies that seem cursed, I was recently asked to explain the allegedly spooky coincidences associated with some famous films. Months after the release of Poltergeist, for example, its 22-year-old star, Dominique Dunne, was murdered by her abusive ex-boyfriend; Julian Beck, who played the preacher “beast,” succumbed to stomach cancer before Poltergeist II’ s release; and 12-year-old Heather O’Rourke died months before the release of what would be her last starring role in Poltergeist III.

The Exorcist star Linda Blair hurt her back when she was thrown around on her bed when a piece of rigging broke; Ellen Burstyn was injured on set when flung to the ground; and actors Jack MacGowran and Vasiliki Maliaros both died while the film was in postproduction (their characters died in the film).

When Gregory Peck was on his way to London to make The Omen, his plane was struck by lightning, as was producer Mace Neufeld’s plane a few weeks later; Peck avoided aerial disaster again when he canceled another flight at the last moment (that plane crashed, killing everyone onboard); and two weeks after filming, an animal handler who worked on the set was eaten alive by a lion.

During the making of The Crow, star Brandon Lee was accidentally shot to death by a stage gun with blanks; he was the son of Bruce Lee, who also died mysteriously at a young age, possibly from a drug reaction. While filming Twilight Zone: The Movie, star Vic Morrow was killed in a freak helicopter accident.

For some people, such eerie coincidences suggest evil supernatural forces at work. But that conclusion is not warranted. As I explained on camera, picture a 2×2 square with four cells. Cell 1 contains Cursed Horror Movies (Poltergeist, The Exorcist, The Omen, The Crow, Twilight Zone: The Movie). Cell 2 contains Cursed Nonhorror Movies (Superman, The Wizard of Oz, Rebel Without a Cause, Apocalypse Now). Cell 3 contains Noncursed Horror Movies (It, The Ring, The Sixth Sense, The Shining). Cell 4 contains Noncursed, Nonhorror Movies (The Godfather, Star Wars, Casablanca, Citizen Kane). When they are put into this perspective, it is clear that those seeing supernatural intervention are remembering only the horror movies that seemed cursed and forgetting all the other possibilities.

Call it the Fallacy of Excluded Exceptions, or the failure to note instances that do not support the generalization. In cell 1, for example, Halloween is not included, because there are no “curse” stories associated with it; its star, Jamie Lee Curtis, went on to a successful motion picture career, and the film launched a franchise in the horror genre. In cell 2, no one attributes evil forces at work on the California highway where James Dean lost his life after making Rebel Without a Cause. In cell 3, spine-chilling films like The Shining should be loaded with curses, but it isn’t.

The psychology underlying the Fallacy of Excluded Exceptions is confirmation bias, where once one commits to a belief, the tendency is to look for and find only confirming examples while ignoring those that disconfirm. This is very common with paranormal claims. People grasp at predictions by psychics or astrologers when they come true, but what about all the predictions that did not come true or major events that nobody predicted? In the realm of faith, cancers that go into remission after intercessory prayer are often considered religious miracles, but what about the cancers that disappeared without faith-based intervention or the cancer patients who were prayed for but died? Divine providence is often adduced when a few faithful people survive a disaster, but all the religious folks who died and atheists who lived are expediently ignored.

The problem is rampant not just with paranormal and supernatural claims. Claims of medical cures associated with this or that alternative treatment modality typically exclude cases where treated patients were not cured or were cured but possibly by other means. Crime waves are often linked to economic downturns, but this hypothesis is gainsaid by counterexamples, such as the relatively low crime rates during the 1930s depression and the 2008–2010 recession. Excluded exceptions test the rule. Without them, science reverts to subjective speculation.