Experiment One: Explicit knowledge of the elevator rule
In our first experiment, we used Amazon Mechanical Turk (2) to ask 1,039 participants the following question: “Imagine that you’re in a building, waiting by yourself for an empty elevator while blindfolded. Without removing the blindfold or getting on the elevator, would you be able to tell whether an arriving elevator was going up or down?” For those responding “yes”, we asked how they would do so, and then, after they responded, gave them a list of six possible answers and asked them to select the one closest to theirs. Those participants who indicated that they did not know how to solve this problem were given the same list of possibilities and asked to select which answer seemed most likely to work
Overall, 370 participants (36%) responded that they knew how to determine the direction of the elevator, but only 3 of these (<1%) chose the correct answer from the list—the elevator “would ding twice if it were going down” (Fig. 1). Upon review of their written responses, it was clear that two of the three had not known the correct answer, while the third provided an ambiguous, but most likely incorrect, answer ("I recently learned that elevators make one tone for up and another for down"; Table 1). Among the 669 participants who reported that they did not have a solution for the dilemma, 46 (7%) selected the correct option from the list of six possibilities provided, less than half of what would have been expected by chance. Thus, none of the 1,039 subjects provided convincing evidence that they knew the solution to the Elevator Dilemma, a rate significantly less than one in 200 (binomial test, p = 0.04).
Table 1. Responses of participants who answered that they knew the answer to the Elevator Dilemma and then selected that "the elevator would ding twice if it were going down” (n = 3).
| |
Please describe in detail how you would explain to a friend how they could know which way an elevator is going.
|
How do you know this works?
|
|
X
|
"The closer the sound the elevator appears to be, the more likely it's going down"
|
“I’m just guessing”
|
|
X
|
(no answer given)
|
(no answer given)
|
|
?
|
"I recently learned that elevators make one tone for up and another for down."
|
"Observation & experience riding elevators."
|
Experiment 2: Implicit knowledge of the elevator rule
Despite their total lack of explicit awareness of the elevator rule, it remained possible that individuals had implicit knowledge of the meaning of the different sounds made by elevators. We investigated this in a second experiment involving 150 participants, of whom 114 completed the protocol. Participants were presented with recorded elevator sounds in two blocks of 16 trials each. For the first block, participants heard one ding on eight trials and two dings on the other eight. After each presentation, they were asked to indicate the elevator's direction of travel and their confidence in their answer. In the second block, participants completed the same task but heard either a higher or lower pitch sound on each trial—a distinction that is not, in fact, ever used to indicate the direction of an elevators (1, p 135). At the end of the experiment, participants were asked to rate the accuracy of a series of statements about elevator sound-direction combinations (e.g., 'one ding means up' or 'a higher pitch means down') on a 1–8 scale ranging from completely false to completely true.
Overall, participants were at chance for correctly assigning direction in the 16 one- and two-ding trials (47.6% correct; t (114) = 0.80, p = 0.43). But a chi-square test comparing the symmetry of the distribution with the random binomial curve (Fig. 2A, blue curve) was significant (χ ² = 50.4, p < 0.0001). Indeed, despite this nearly 50–50 overall split on correct responses, 26 participants (23%) responded correctly on at least 14 (88%) of the 16 trials (binomial test, p ≤ 0. 002 for each participant) while an even larger number (n = 28, 25%) responded incorrectly on the same number of trials. Those who answered correctly on 3–13 trials (n = 60) appeared to have guessed randomly, approximating a binomial distribution (χ2 = 6.4, p = 0.78; Fig. 2A).
These differences were also reflected in participant’s judgements about sound-direction combinations at the end of the experiment. Participants who answered 14–16 trials correctly rated the statement “one ding means up” as significantly truer than those who answered approximately half (7–9 out of 16) correct (6.5 ± 0.3 vs. 4.0 ± 0.2, df = 50, t = 7.7, p < 0.0001). But the reverse was also true; participants who answered 14–16 trials incorrectly rated the statement “one ding means down” as significantly truer than participants who answered 7–9 incorrectly (5.5 ± 0.3 vs. 4.0 ± 0.2l, df = 52, t = 4.0, p = 0.0001).
In contrast, for the high vs. low pitch judgement, which, in fact, is not used to indicate elevator directions, 97 participants (85.1%) assigned an upward direction for the higher pitch sound on most (9–16) trials, while only 10 participants (8.8%) preferentially assigned an upward direction to the lower pitch tone (binomial test, p < 0.0001). In this case, only 2 participants (1.4%) scored 14–16 trials as “high = down” or “low = up”, whereas 54 participants (47%) scored a similar number of trials as “high = up" or "low = down” (Fig. 2B). Thus, the vast majority of participants (n = 97, 85%) drew the erroneous conclusion that higher pitched tones indicate an elevator going up and lower pitched tones indicate one going down. In subsequent evaluations, participants who scored at least 14 trials as high = up or low = down rated the truth of the statement “a high-pitched sound indicates up” higher than those who only scored about half (7–9) the trials in this manner (6.7 vs. 3.6, df = 75, t = 9.9, p < 0.0001).
Taken together, these results offer no evidence of subjects having implicit knowledge of the relationship between ding-number and elevator direction, suggesting that over the course of these participants’ lives, information about the second ding was not effectively processed. This is even though half the participants in Experiment 2 reported that they rode elevators at least twice a month and a quarter reported riding one at least 7 times a month; indeed, individual ridership did not correlate with accuracy predicting elevator direction (Fig. 3).
Experiment 3: Experiencing blindness
A third study clearly demonstrated that participants’ lack of knowledge of this elevator rule was not due to any inherent difficulty in determining this relationship=. After failing to correctly provide an answer to the Elevator Dilemma question, participants (n = 20) were taken to a bank of elevators and blindfolded. The experimenter called for elevators in both directions. Upon arrival of an elevator, the participant was asked which direction they thought the elevator was going and, after responding, was given feedback on whether they were right or wrong. This continued until the participant reported that they had learned that one ding indicated that an elevator was going up and two indicated that it was going down. Most participants solved the problem in fewer than 6 trials (median = 4, mean = 5.29, SD = 4.72, range = 2–20). Thus, when this auditory information was presented without the usual visual cues, participants rapidly deduced its significance, with most participants discovering the rule in four trials or fewer, suggesting that information such as that provided by the second ding (for descending elevators), though normally filtered out, is effectively attended to when needed.