# Blindspin 5: History being made?

The picture below shows, for the first time in history, what the visual demand of bicycling probably looks like. We will need a lot more data to make sure that this is actually true. If you live near Turku, and are willing to ride a bike for an hour or two, please contact Jakke.Makela[at]gmail.com.

Roughly speaking, the parameter OT indicates how tightly the cyclist needs to keep his eyes on the road. The smaller the OT, the more dangerous it is to be distracted by something else — a ringing mobile phone, a buzzing insect, a skimpily clad person. An OT of 600 milliseconds means that a glance just 600 ms long could get the cyclist in trouble.

The blue line is the median within a 2-second window, and the red line is the 85th percentile. The red line is probably the more interesting one.

What is significant here is that the OT seems to rise as the speed increases. That is, in some sense it is more dangerous to cycle slowly. The OT rises particularly sharply at about 20 kmh. That is, increasing the speed to that level makes the danger very much smaller.

If phrased as risk, this seems counterintuitive, because of course at smaller speeds the effects of a crash are smaller: risk = probability times impact. The graph above only measures the probability, and even that only indirectly.

Cyclists are constantly being distracted. Usually, any effects from the distraction can be corrected almost unconsciously. However, over time, the risk accumulates. At any speed, there will be many glances of 600 ms duration. These 600 ms glances are more likely to cause problems at slow speeds, and less likely to cause problems at high speeds.

The data are from one experimenter (person P01) who drove back and forth along a straight cycleway track on the outskirts of Turku. Many potential tracks were described in Blindspin 4, but it strongly seems that this particular track is the one we will have to use. There was occasional other traffic, but due to the flat terrain it could be seen from a long distance, and P01 waited until the traffic had passed.

Figure: Track used in this test. The actually usable part of the track is less than the maximal 390 meters.

P01 instructed himself to drive at various speeds: as slow as feasible; at a suitable speed for a crowded cycleway; as fast as seemed sensible.  The technical setup is shown in Blindspin 2. Whenever P01 closed his eyes, he pressed down on the button. When he opened his eyes, he let go. The eyes-closed period is one event. The duration of the eyes-closed period, measured from the data, is the Occlusion Time (OT). The speed at the start of the occlusion period was also measured.

There are many possible sources of error in these measurements, and thus this data set is not highly reliable. The whole experiment will be redone from scratch when we have solved some technical problems. It is certainly possible that these effects will disappear when the test is done properly; hopefully not.

There are other parameters that we could use in these measurements, but for now we will just analyze the OT. We consider OT to be related to the visual demand — that is, the amount of attention that needs to be paid in order to cycle safely.  On a straight track, the visual demand should in principle be constant at all speeds. Even if there are variations, we would expect them to be roughly continuous and linear. However, if bicycling is fundamentally similar to driving a car, we should not be seeing any discontinuities.

In fact, we do see discontinuies. Below speeds of about 10 kmh, P01 was essentially unable to keep his eyes closed for more than 800 ms. Above that speed, much longer occlusions times were possible. An even bigger discontinuity occurs near 20 kmh, when very long occlusion times were easily achieved.

We have a potential explanation. Every cyclist knows that keeping balanced is more difficult when the speed is extremely slow. Staying in balance requires more visual input at slower speeds. This is in direct contrast to driving a car, where slower is always easier. Most bikes have a self-stabilizing speed, which is near or above 20 kmh. At that speed, the bike will be stable even without the cyclist. That is, the cyclist does not need to put any visual effort into keeping balanced.

Is there anything actually new about such a result? Yes, and no. Everyone intuitively knows that this effect exists. What we don’t know is its magnitude, and the speed at which driving starts to become easier again. In terms of pure research, this could also suggest a somewhat controversial idea: when visual demand is studied, great efforts are made to separate the operational demand from the “pure” visual demand. We could make the suggestion that for cycling, such a separation simply cannot be made. Visual input is needed to stay balanced, and this additional visual demand is part and parcel of the task of bicycling. This idea will definitely be difficult to defend, if we ever get to that stage.

Does it have any practical meaning? Yes, and no. We want to be very clear: this study, even if it is 100% successful, will not produce results that could be used to define any safety policies.  This is basic research, which aims to determine how much visual input is needed to cycle adequately. When we know that, we can estimate how badly distraction will degrade driving performance at a given speed.

Given that the impact of slow-speed crashes is very small, the effect might not have any real significance whatsoever. We cannot evaluate the significance; we can only report the facts.

Having said that, if the effect is strong, we could use it to educate cyclists about the need to stay particularly alert at slow speeds. Using a mobile phone while cycling is never exactly safe, but at least in a car, slowing down makes it less dangerous. If slowing down actually makes the situation more precarious in cycling, should we at least tell people about it?

There may be other practical results related to cycleway design; however, we will collect more data before we even venture to say anything about them.

Fundamentally: we really shouldn’t rush to call a phenomenon meaningless before we have actually measured that phenomenon. That is why these experiments should be done, even if most peoples’ gut feeling is that they are probably irrelevant.