"I just didn't see him" is a claim that's repeated over and over in accident reports. Drivers earnestly claim that they simply didn't notice the bicycle/pedestrian/motorcycle they crashed into. The claim is made so frequently that certainly there must be a grain of truth to it. Yet it certainly isn't the case that car drivers can't see such obstacles -- after all, they can see traffic signals that are much smaller than a bike or a motorcycle.
What they mean to say is that their attention was otherwise engaged -- perhaps by a phone conversation, perhaps by other traffic, or perhaps because they were trying to find something -- a street sign, a restaurant, a gas station. Human attention is a fickle thing, and in many cases we don't notice very obvious details changing right before our eyes.
Consider the following movie (QuickTime required): One image will be displayed for a number of seconds, followed by a white screen, and then a second picture -- the same image with one very obvious detail changed. Can you spot the change (don't cheat--just watch it once!)?
Let's make this a poll:
I'll give you another chance at the test later on, so don't repeat it just yet -- the point is that it's often quite difficult for us to notice changes when we don't know what to look for.
But what about when those changes can literally mean the difference between life and death? Perhaps when we're driving a car, we're so attuned to the possibility of an accident that momentary distractions like looking for a street sign don't detract from our ability to notice hazards. Steven Most and Robert Astur developed a simple, elegant experiment to test that notion.
They asked volunteers to participate in a driving simulation task. The simulator had a steering wheel, gas pedal, and brakes, and volunteers were asked to navigate through city streets, turning whenever an arrow flashed in a particular color -- either blue or yellow. The car drove automatically at a constant speed of 30 miles per hour, and the participants were told not to stop unless they were about to hit a pedestrian or another vehicle.
Drivers had to pay close attention to the arrows, because they could change color as they flashed, and they were only supposed to turn when the arrow stopped flashing and was the correct color -- immediately before they reached an intersection. After seven successful turns, a pedestrian walked across the street, forcing a stop. The key manipulation, however, occurred after 9 turns. At this point a motorcycle unexpectedly veered into the driver's path. For half the drivers, the motorcycle was the same color as the arrows they were supposed to follow, and for half the drivers, the motorcycle was the color of the arrows they were supposed to ignore. Was there a difference in stopping time when the motorcycles were different colors? Here are the results:
When the color of the motorcycle matched the color of the arrows the driver was following, the braking time was significantly faster, and almost no drivers collided with the motorcycle, compared to a whopping 36 percent collision rate when the color of the motorcycle didn't match the color of the arrows.
Most and Astur conclude that even in a dangerous situation, the attentional set of a driver has an important impact on driving ability. Just as drug users are quicker to spot changes in drug-related items in a photo, so drivers are faster to react to road hazards if they match other items they're looking for as they drive. So in this case, there's no difference between a dangerous situation and an ordinary situation; what matters is the particular attentional set at of the driver the time of the incident.
One potential application, the authors suggest, might be on road construction sites: if the color of traffic markers such as cones and reflectors matched the color of worker uniforms and construction equipment, then drivers would be quicker to react to those hazards.
Now back to the demo I started with -- can you notice the change in the picture this time? I'm betting a large number of readers will do better the second time they take the test. Go ahead, try it:
Did you see the change this time? Record your response in this poll.
If all goes well, we should see a significant improvement this time. I'll explain why in the comments -- after plenty of readers have had a chance to respond.
[Update: I changed the task to respond to several readers' suggestions that what we may have is a practice effect. See comments 12 and 30 for details]
Most, S.B., & Astur, R.S. (2007). Feature-based attentional set as a cause of traffic accidents. Visual Cognition, 15(2), 125-132.
It took me six tries to spot it. I had to put up a frame from pre- and post- change, and then switch between them several times to figure it out.
It took me six tries to spot it. I had to put up a frame from pre- and post- change, and then switch between them several times to figure it out.
Yeah, it's a tough one. But it looks like we're getting an effect. It's fine to give "spoilers" in the comments if you have a guess as to why people are better the second time around.
Thank you SOOO much for including the entire article in the RSS feed. I don't know when you started doing that, but I really enjoyed going through the entire article in my feed. I wasn't one of your vocal proponents for the change when you asked about it weeks ago, but I am really enjoying the effects. Thank you!
Likewise. If that's a "very obvious detail," then I'm next to brain-dead ...
We're just doing the full RSS experimentally for now. Consider this a trial run, but I'm glad you like it!
I think I saw it on the eighth or ninth try, and then had to run it one more time to be sure.
NOT "very obvious": I had to drag the 'slider' back and forth quickly (DIY 'blink comparitor') many times while systematically studying each part of the picture in turn before I could spot it. The change is a very low contrast change, easily missed unless you know precisely where to look.
The first time I was completely clueless. The second time I thought I had it, but had to run it again to be certain.
I got it the second time. The first time I was focused on the colorful more contrasty central parts of the image assuming "very obvious" would have something to do with those, and didn't see anything, so I loosened my focus and looked at the "big picture" the second time and caught it. I'm not sure I'd call it "very obvious" either, though; bright color change or motion would have been more obvious. If I'd been told to look for a moderate change I might have apporached the first run differently.
Same here.
Hmmm... Maybe "very obvious" isn't the best way to put it. The point is that once you see the difference, it's easy to spot. It's not just a shadow or nuance of color.
So here's the manipulation, let's see if you buy it. As of now about 11 percent of respondents got it on the first try, and 32 percent got it on the second try -- a pretty big improvement!
In the instructions for the first try I say it's a "bookshop," priming viewers to look for changes there. In the second set of instructions I say "a large number of readers," and I also say the numerical term "second," subtly priming people to think about numbers. Indeed, the numbers are what change, and that may be what led to the larger response the second time around.
So, do you buy that? Do you think that's why we got better results the second time? Or is it just a practice effect?
I saw the change both times - but I think that the intro gave an unfair advantage as it encouraged me to scan the whole picture and identify all possible features quickly. Interestingly, the first time around I actually thought I saw TWO changes - the real one plus a number I wrongly believed to change between the two images. Don't know what that's about.
I saw the change both times - but I think that the intro gave an unfair advantage as it encouraged me to scan the whole picture and identify all possible features quickly. Interestingly, the first time around I actually thought I saw TWO changes - the real one plus a number I wrongly believed to change between the two images. Don't know what that's about.
I got it the second time, but I think it's purely a practice effect. In fact, the first time I was looking mostly at the bookstore and the two signs, while the change is not in the center of the field.
The second time, I was spending time looking more closely at areas I hadn't been looking at before, and oddly enough, I keyed in on the right location before the whitescreen in a sort of "oh, I didn't notice that there last time" way. And then it disappeared.
Dave - I went back and read everything again, and I'm thinking that the title with its reference to color was what I was priming off of on the first run. The dropped references to numbers might have made a difference, because when I first looked at it I didn't even take in the numbers and I definitely did the second time. But the "very obvious" term certainly dominated what I was looking for overall both times, and it looks like that was true for several other commenters as well.
Ok first of all this page crashed Firefox three out of the four times I tried to load it (I finally had to resort to posting this comment in...gulp...IE), so there's no way I'm double checking my results here, but I *think* I got the change on the second go-round of the second try. I have no idea how "color" or "bookstore" could have anything to do with priming the change I noticed, though, so maybe I'm seeing completely the wrong thing.
I still have no clue.
But I'd like to say that I think one reason drivers don't "see" bikers is because we're driving more or less automatically and bikers aren't cars. I'd be willing to bet that where there are lots of bikers they get "seen" more.
I thought I saw a change the first time round, so I clicked "Yes". On the second try, I saw that I had been mistaken, but I was sure of another change, in the sign on the right this time.
So I clicked "Yes" again. And I was wrong again.
It wasn't until the 4th go-round that I saw the real change.
Our eyes are not made to work with a white screen blanking out our visual field. This gimmick defeats what normally works very well with our vision -- noticing something pop up or vanish while looking very near the thing. (Think of a chipmunk poking its head out of a pipe.)
I spotted it the first time, but unfortunately I didn't remember what changed, only where it changed and what colour it was. I could extrapolate based on the similar object on the other side, but I couldn't be sure until I watched the second time.
I did see it the second or third time but the change didn't register with me until I viewed the picture several more times. I guess I was looking for something more obvious.
Check out the choice blindness experiment for perhaps even more surprising effect:
http://en.wikipedia.org/wiki/Choice_blindness
Oh, BFD, so that faint 44 disappeared. You made me read through all these comments and then I still had to find it myself? When is a car accident going to happen from not looking up that much, from the airplane part falling from the sky, from being like they talked about in Star Trek II, deficient in paying attention to the Z axis? Yes, we have strategies for guessing where the most relevant change is going to come, and your numeric clue was better than nothing. Yes, such strategies might kill us if we're wrong. Hey, stand in line with everything else that will kill me. Whenever I die it will be human nature that kills me one way or another. No experiments are necessary to know that.
I feel so much better having given my authentic reaction. It's too bad emotions don't get the same degree of analysis as perception and cognition.
I got it the second time! YaY! LOL.
I am soooo going to blog about this! :)
I have ridden street motorcycles for more than thirty years -- carefully, as you may surmise, since I am still alive and healthy. Drivers of cars and trucks fail to acknowledge my presence so often that I simply assume that I am never seen and ride accordingly. For several years, though, I rode a side car rig, both on cross-country tours and in local traffic: I never, even once, had a car or truck pull in front of, crowd, or otherwise fail to acknowledge the presence of that rig. I suspect that the novelty of the vehicle raised its awareness factor significantly. The motorcycle and side car rig were both black, by the way.
I saw it the second time, and I have to agree with the other commenter who thought it was a practice effect. The interesting thing is that I just happened to be looking at the right spot on the picture when the screen changed the second time around. Was that just a coincidence, or did I subconsciously know that something was going on with that portion of the picture?
So what colour should I wear when on my Bike ? Should I wear black as its the colour of the road, white as its the colour of road signs, or flourescent yellow so people think I'm a copper ?
Okay, let's see if we can suss out whether this was a practice effect. I just changed the instructions for the first picture to include primes for "number" and "second."
As of now, 85 said they could see the change and 394 said they could not, for a success rate of 17.7 percent. If we get a hundred or so more responses, we may be able to see if people are better at noticing the difference when they are primed for numbers.
It seems to me that just putting on your lights when you are driving anything (car, motorcycle, bike) would cut down on the number of accidents. Lights are pretty easy to see. So are unusual colors, such as bright yellow or fluorescent orange. There is good reason why those those colors are used at construction sites and for cautionary traffic signs. There is method to the madness.
By the way, looking at something on a screen is simply NOT the same as being out there in real life and seeing it. Peripheral vision is not needed for a screen, whereas in real life it is not only present, but highly useful. I just don't believe one can compare with the other or discover much that is useful while driving.
It seems to me that just putting on your lights when you are driving anything (car, motorcycle, bike) would cut down on the number of accidents. Lights are pretty easy to see. So are unusual colors, such as bright yellow or fluorescent orange. There is good reason why those those colors are used at construction sites and for cautionary traffic signs. There is method to the madness.
By the way, looking at something on a screen is simply NOT the same as being out there in real life and seeing it. Peripheral vision is not needed for a screen, whereas in real life it is not only present, but highly useful. I just don't believe one can compare with the other or discover much that is useful for real-time driving.
I am wondering if the Most and Asur (2007) experiment actually implicates an attentional phenomenon? It could potentially be explained as a perceptual-motor learning phenomenon.
Consider the hypothetical case where participants are asked to pay attention to the blue arrow and ignore the yellow one. Upon encountering a blue motorcycle, they stop more quickly (presumably) because they have learned to attend to the color blue. It could also be the case, though, that a perceptual-motor learning association occurs at turn 7--namely, blue = put foot on brake, which is then applied at the incident at 9. Alternatively, another perceptual-motor learning association could be, yellow = don't put foot on brake (or, keep foot on accelerator). Thus participants might be slower at stopping upon seeing a yellow motorcycle because they have not learned to associate yellow with putting one's foot on the brake.
To make it more of an attentional (or at least, cognitive) phenomenon, a third condition would seem critical. Specifically, what would happen if participants were asked to ignore the blue arrow, attend to the yellow one, and the color of the motorcycle encountered is also blue? If participants in this condition are responding to the motorcycle as slowly as participants trained to attend to the yellow arrow (but then encounter a blue motorcycle), it would suggest the slowed reaction is more a result of not paying attention to a particular color, and not so much the possibility of learning a different motor response to a different color.
The significance here, is if what is actually happening is some kind of associative learning involving colors, it might not be effective to place cones and persons in the same color--especially if one the attention system is trained to ignore particular colors.
I'm *still* not sure what the change is, and I've gone back and forth a dozen times. The only difference I can find is that there's a very faint "44" sign in one view that's not in the other - it's at the upper right. I can't imagine any circumstance under which this could be described as an obvious or large change, particularly given the lack of contrast between the sign and the background wall.
Whenever Cognitive Daily encourages us to participate in some cool experiment, I can't help but think that some other aspect of our behavior is being observed.
The unknowingly chumped rubes, if you will.
Hmmmm.... Looks like my efforts to prime for number changes weren't successful: After 171 people responded with the new instructions, we're at a 21 percent success rate. Not much better than the 17 percent we got with the old instructions.
So it looks like this demo didn't work. Sigh. Such is the life of a cognitive psychology blogger. Still, it's interesting how few people noticed the change.
Also, if you think the change was too subtle, try this one, with a more obvious change. I bet you still don't get it on the first try!
As for the change-blindness effect, I find that this waries significantly by people. Most people have difficulty seeing changes, which has always had me stumped. I see them almost always, and definitely after 3-5 seconds. I have, however, seen cases when room-fulls of people couldn't see a change for 30-60 seconds... Why some people are inherently better at seeing these things is completely unclear as far as I know.
As a daily motorcycle rider, I can appreciate this article, as I have been a victim of "I just didn't see him" a few times, once resulting in a broken leg, most others near misses. I noticed the change immediately on the first viewing as well as the second of course.
I got to thinking about why it was so obvious to me, and the only reason I came up with is the fact that if you are a motorcyle rider and you have had sense enough to take a Motorcycle Safety Foundation course, you are taught to constantly scan everything around you as you ride, both near and far (where you will be in the next 12 seconds of travel) simply to avoid the "can't seem to pay attention" car drivers. It really becomes a habit to scan many things at once and deduce which are a potential danger and which can be ignored (somewhat). I only WISH that car drivers (cagers as we call them) could do the same. We do it because OUR LIVES DEPEND ON IT.
To make the test truly representative of the idiots out there in their tin boxes driving around running into each other you need to tell them:
Pick up your cell phone and start yacking, lean over and change a CD or fiddle with the radio, yell at the kids in the back seat... *now* watch this video while doing all of the above and try not kill anyone.
SMIDSY should be an automatic forfeit of drivers license
Very good test. I did not read any of the article, but went straight to the test and caught the change right away. I don't know why, maybe to many years of "Where's Waldo" with my kids. I am usually oblivious to changes like that one. But I find that has little to do with the original post about cars. I have driven many colors of cars over the years, and the times that I had red cars, I would alway seem to get hit in the rear while sitting at a stop light.
Hate to say it but I don't know how I ended up watching this on a Friday night but the rest of you ought to to know that the 6% of us caucasians that have color vision deficiencies can't see crap in this video however many times we watch it. Can somebody tell me what it is that changes?
Another theory on SMIDSY accidents is that cage drivers do not see motorcycles as a threat, and subconsciously dismiss them.
Much in the same way that the difference between a wasp and a fly will prompt different instinctive behaviour.