It was September 2012, and Los Angeles Mayor Antonio Villaraigosa was standing onstage at the crowded Democratic National Convention in the Time-Warner Cable Arena in Charlotte, North Carolina, optioning a resolution to reinsert "God-given talents" and language about the capital of Israel into the party's platform.
Villaraigosa had already strained to hear whether the shouted results of two votes represented a two-thirds majority for "aye" or "no." Now, he stood before the room with a confused grin on his face and glanced desperately to both sides of the room. "I um," he paused. "I guess...." A female organizer walked up to his side, and gave him a nudge. "I'll do that one more time," he continued.
After a third vote, and over the loud boos of audience members, Villaraigosa called it an affirmative decision:
As University of Iowa acoustician Ingo Titze watched the episode unfold on television, he was surprised at the result, as were many other observers at the time. "There were very emotional people in that audience," he recalled. "People were literally jumping up and shouting ‘No! No!’ It got me thinking: What good is a voice vote when there’s a real issue at stake—real emotion involved?'”
Titze teamed up with a researcher from the National Center for Voice and Speech to run a series of experiments designed to gauge the reliability of voice votes, which are used everywhere from the U.S. Senate to local zoning board meetings. In a new paper published this month in the Journal of the Acoustical Society of America, the scientists concluded that this type of vote is extremely unreliable and that even just one boisterous idiot can distort the perceptions of a judge.
After convening a group of 54 speaking subjects and splitting them equally into an experimental group of 27, with 10 participants "successively" speaking at the loudest levels (or "as if shouting to someone across a busy street"), and a test group of 27, which had speakers at moderate loudness levels ("as if teaching a class of about 40 students"), the researchers asked five outside judges to gauge which groups were louder or whether they rated the same. "The judges noticed an increase in voting loudness in the voter group that was increasing in the number of voters speaking louder (linear trend, p 1⁄4 0.008)," the authors wrote. "This difference in voting loudness was noticeable to the judges beginning with one voter speaking louder than the rest (1/27, or 4% of voters in that group) and was clearly obvious to all judges after three voters spoke louder (3/27, or 11% of voters in that group)."
In another experiment, individuals in the experimental group were asked to lower their voices successively to a level that someone would use "if across a table or desk with someone." "They noticed a marginally significant decrease in voting loudness in the voter group that was increasing in the number of voters speaking softer...."
The research suggests that judging a simple majority, or even a two-thirds majority, among a group of voice voters is only possible when everyone uses similar conversational volumes. In other words, a couple of vociferous voices can easily drown out the desires of the group.
A separate acoustic analysis also supported the authors' conclusions. Loud individuals can apparently raise a group's total volume by between three and five decibels, "as if the group size has doubled from, say, 20 to 40." And even just a single very loud speaker "affects the group sound level by more than 2dB up to a group size of 20. Even in a group size of 40, one speaker can raise the group sound level by 1dB, near the just noticeable difference (JND)." What does this mean in practice?
In a small group, simple superposition theory predicts that one voter can bias a vote to the degree that even a two-thirds majority perception is challenged. Also, random fluctuations of sound power of a few voters produce a group loudness fluctuation on the order of the [just noticeable difference], suggesting that a voice vote for a simple majority is highly suspect unless individual loudness is controlled. For example, it requires a group of 40 or more voices before a single loud voice bias is overcome to establish a 60:40 majority.
The authors don't recommend using a voice vote for anything that rises above a decision that's "relatively benign" and done with "cooperative" voters. With larger crowds, there is little hope for determining "a simple majority ... by voice vote unless individuals know how to produce (and agree to produce) a standard sound level, which is unlikely."
At the paper's conclusion, the researchers even worry that vote hijackers will use their findings for political advantage.
"The possibility exists that those who deliberately wish to influence a vote, or discredit it, will realize how easily it can be done and substantiated," they conclude. "It is doubtful, however, that this is not already known and considered by most parliamentarians."