Verified by 
Participants in a 2012 No Pants T-Ride in Boston (from weeklydig's photostream)
On a cold day in early January of 2002, a man entered a New York subway car wearing a hat, scarf, gloves, and winter coat, plus a brightly-colored pair of ladybug boxer shorts. That's right, no pants.
At the next stop, he was joined by a second pantless rider, who appeared not to know him. This happened five more times.
While one pantless New Yorker hardly bears notice, seven demand an explanation. What was going on?
Innocent observers were quick to generate their own hypotheses. "Is there a man-without-pants convention today or something?" one man asked.
In fact, the event was a stunt by Improv Everywhere, a New York City-Based "prank collective" that stages semi-improvised events in public places, leaving laughter and confusion in their wake. The 2002 No Pants Subway Ride has inspired an annual event, with thousands of New Yorkers participating in the tenth anniversary held just last week.
The original, 2002 prank was compelling because it cleverly exploited the human drive for explanations. (Though admittedly, the near-nudity and ladybug boxer shorts did help.) Whether they realized it or not, the creative minds behind the stunt were capitalizing on how explanations change as evidence accumulates. Somewhere between pantless rider two and seven, explanations that appealed to chance ("what a bizarre coincidence!") morphed into an appreciation that larger forces were at work.
In research published in the journal Cognition, researchers Tom Griffiths* and Joshua Tenenbaum develop a mathematical model of the psychology of coincidences, a calculus for how and why we dismiss some events as mere coincidences and treat others as compelling evidence for an underlying explanation, whether it's an improvisational stunt or a man-without-pants convention.
According to their theory, coincidences aren't simply unlikely sets of co-occurrences, but rather constellations of events that provide evidence for a causal explanation. What differentiates a coincidence from a more mundane occurrence is the strength of the evidence and whether the explanation is plausible.
To illustrate, consider four hypothetical cases, the result of varying whether evidence is weak or strong and whether the explanation is plausible or implausible.
Case #1 (weak evidence + plausible explanation): Two men enter the subway with umbrellas. This provides weak evidence for a plausible cause (e.g., that it's raining or forecast to rain). This isn't a coincidence, it's a typical January morning in New York.
Case #2 (strong evidence + plausible explanation): Seven men enter the subway with umbrellas. Now we have plenty of evidence for a plausible common cause: actual or expected rain. Umbrella-carrying is no coincidence, it's the result of that morning's weather forecast.
Case #3 (weak evidence + implausible explanation): One pantless man boards the subway. This provides some evidence for an improvisational stunt or a man-without-pants convention, but both possibilities are sufficiently implausible that we're unlikely to entertain them without further prompting. Instead, we might appeal to something more plausible, like run-of-the-mill, New York eccentricity. (As one woman explained to her startled husband when a man removed his pants in preparation to board the pantless subway car, "Honey, it's NEW YORK.") This isn't a coincidence, it's just a little weird.
Finally, consider Case #4 (strong evidence + implausible explanation): Seven pantless men board the same subway car. Now we have strong evidence that something is going on, but it isn't immediately apparent what it is – there aren't any obvious and plausible explanations. This is what Griffiths and Tenenbaum call a "suspicious" coincidence, and it's the kind of event that can presage scientific discovery (when, for example, Halley noticed the suspiciously similar orbits that led him to posit a recurring comet) or laughter (when, for example, you observe yet another pantless stranger board your subway car and decide you've been had).
In the end, observers in the 2002 No Pants Subway Ride concluded that their observations were no coincidence, but conclusive evidence for a prank, performance art, a man-without-pants convention, or - as speculated by one Arizona man in a 2012 reenactment - "a coming out party or something."
The event just wouldn't have been funny with only one man (weak evidence), or if the men had all carried umbrellas instead of shedding their pants (too many plausible explanations). Improv Everywhere hit the sweet spot predicted by Griffiths and Tenenbaum's model: strong evidence plus an implausible explanation that ultimately had to be true.
Griffiths and Tenenbaum's model also explains why coincidences are associated both with wacky beliefs (the paranormal, conspiracy theories) and with scientific breakthroughs. In the former case, we're dealing with what most people judge to be implausible causes: ESP, alien interventions, supernatural forces, or secret societies that somehow explain why Thomas Jefferson and John Adams both died on the 4th of July. In the latter case, we're dealing with strong evidence for an alternative to accepted scientific theory, and are therefore in the territory of unanticipated explanations. Unfortunately, it isn't always obvious whether a novel explanation is fringe lunacy or the key to scientific progress until the evidence (that seventh pantless man) is overwhelming.
* Full disclosure: Tom Griffiths is my husband.
