VI
Mark học AIGorillas in Our Midst · Daniel Simons and Christopher Chabris, Harvard University
When we focus hard on one task, do we see something odd, large and obvious right in front of us? The answer: very often, no. And we hardly ever know what we missed.
Viewers were warned of nothing. They had a single job: count correctly.
Looking isn't seeing. Seeing needs attention to come along.
10 things often said about the gorilla experiment, checked against current research. Tap each line for the reasoning.
Simons and Chabris (1999): 44% noticed the gorilla, 65% the umbrella woman, 54% overall noticed the surprise guest. The effect has been repeated many times, in classrooms and labs, with many video versions.
In the original study, 64% of viewers on the easy task noticed the surprise guest, versus 45% on the hard task. Chabris et al. (2011) found the same in real life: higher attentional load, fewer people noticed the fight.
The black gorilla was seen 58% of the time by viewers tracking the black team and 27% by those tracking the white team (Simons and Chabris, 1999). Most et al. (2001) repeated the pattern with on-screen shapes.
Drew, Võ and Wolfe (2013): a gorilla 48 times the size of an average nodule was inserted in lung scans; 20 of 24 radiologists (83%) did not report it. None of 24 non-experts on the same task saw it.
Strayer and Johnston (2001), 48 students in a simulator: the probability of missing a signal more than doubled during conversation, handheld and hands-free alike; radio or audiobooks made no reliable difference. Strayer's group has repeated this many times since.
Memmert (2006) tracked children's eyes: those who missed it looked at the gorilla for about 1 second, as long as those who saw it. Drew et al. (2013): 12 of 20 radiologists who missed it looked straight at the spot. Same direction, but few eye-tracking studies and small samples so far.
Haynes et al. (2009), 8 hospitals: post-surgical deaths fell from 1.5% to 0.8% and complications from 11.0% to 7.0% after the WHO checklist. But in Ontario, 101 hospitals required checklists and almost nothing changed: 0.71% versus 0.65% (Urbach et al., 2014). Checklists work only when actually used, not ticked for show.
Ophir, Nass and Wagner (2009) found heavy multitaskers filtered distractions worse. A meta-analysis by Parry and le Roux (2021) across 118 assessments found only a small link, and for performance-based tasks it largely disappeared after correction. Cause and effect are unclear.
Simons (2010): people who knew the original video expected and saw the gorilla, but were no better at spotting other surprises (17% versus 29%, not significant). Knowing you have a blind spot doesn't remove it.
Kreitz et al. (2015) measured many cognitive abilities: almost none predicted who would notice. Top experts miss things too (Drew, 2013). It's a general feature of attention, not a flaw of weak people.
One of the sturdiest experiments in psychology. The weak spot is the retelling, not the data. Robustness 9/10.
Many famous experiments fell because of small effects, small samples and flexible analysis. The gorilla is different: the effect is large enough to see with the naked eye, viewers answer yes or no, and anyone can rerun it in a class. Many independent groups have.
The part that needs caution is the exact rate. "50% miss it" is the number for one video and one task. A different video, task or audience gives a different rate. What holds is the pattern: the busier and narrower attention gets, the more we miss the unexpected.
Simons and Chabris won an Ig Nobel Prize in 2004 for this study, a prize for research that "makes people laugh, then think". In 2010 they wrote The Invisible Gorilla, extending it to other "illusions" about memory, confidence and knowledge.
Work skill: do one thing at a time, and know in advance what you'll miss.
Every number in this page, the Facebook carousels and the captions comes from the studies below.