Reaction Times in HEMA

Over the last year or so, a Ukrainian HEMA Youtube channel called Bent Blades has uploaded several videos showing various experiments and tests they have done to determine the time it takes to execute various attacks. I found this very interesting, because I also have done tests to attempt to measure the speed of attacks in longsword, particularly to compare the ballistic passing step ideas that we were talking about in the HEMA Discord in 2020-2021 to the conventional dominant-foot-forward lunging attack style. It makes me happy to see other people and groups trying similar experiments. 

Their latest video was on the topic of applying this to reaction time and decision making, which can be found here. With regards to reaction time, there have been many studies showing that reaction time is a highly skill context-dependent attribute, and top level athletes do not possess superhuman rote reaction times, they are comparable to the general population. This video could easily have fallen into the trap of thinking that you need to improve your general reaction time in order to improve at fencing, but thankfully it did not. I generally agree with the conclusions that they came to, however the conclusions as well as the basis of the experiment themselves are thoroughly steeped in information processing approach theory. Here I will explain why I think that the conclusions are much better explained and more easily reached using the Ecological Approach model. 

In this video, there were two different experiments, a pure reaction time experiment, and a reaction time + decision making experiment. For both, they used a device called a Blazepod, which is marketed as a reaction time training device. The device lights up at a random time, the fencer strikes when they see the light, and the time between the light and the impact is recorded. For the first experiment, a single light and a single attack was done in order to isolate reaction time (this is the opposite of my attack timing experiments, in which I subtracted my average reaction time in order to isolate the time of the attacking motion). In the second experiment, one of two lights appeared at random, if a red light appeared then the fencer would have to cut, if a blue light appeared the fencer would have to thrust, in order to add a decision element. Here the attack time was a known quantity taken from previous experiments (~200-300ms), so the result was 200-300ms extra for reaction, plus 200-300ms additional for a decision. 

I’m now going to paraphrase the conclusions, which I will then talk about. Please watch the video instead of taking my word for it, my purpose is not to misrepresent anyone or put words in their mouth. The conclusion of the first experiment was that when you are defending, if you wait until the attack starts and only then begin to react, it will be too late because the time of the attack is faster than the time it will take you to react. For the second one, the conclusion was that fencers must shift the decision and reaction portion of the exchange to before the action happens in order to be fast enough, and if their plan fails and they need to process a new plan in the middle then that leaves them open to getting hit. In this way, fencing becomes the positional chess game that we see when two fencers are at distance preparing their actions. 

In both of these cases, there is a necessity to reconcile what we see in actual fencing with the results of the experiments. For the first one, reaction time is slow, but fencers can still react to fast attacks and defend in time. Of course, as the video states, if they are waiting for the initial attack they will be too late, but this is easy enough to reconcile with IPA theory, just draw them in using fake cues so you can execute a pre-planned defense, or simply be the person who is attacking instead. In the Ecological Approach model, you are not reacting to any individual cue warning you of your opponent’s attack, but you are perceiving and acting upon all information in your environment as a whole, therefore you should not be waiting for the terminal phase of your opponent’s attack to start your defensive action. There is not one discrete decision point, the decision is a process that occurs throughout the action – see my previous article on decision points. The idea of drawing your opponent in by giving them false information applies to both approaches.

The decision making experiment is where this really gets interesting. The crux of the information processing approach is that the entire decision making loop (react, decide, execute action) must exist whenever an action is taken. However, we can readily observe that fencing happens on much smaller timescales than the ~700-900 ms at which the loop was measured in the experiment. This is explained by the idea of shifting the decision making process to before you engage, basically pre-deciding what you are going to do so you can execute it as soon as you perceive that cue. The issue with this is that you start seeing edge cases that need to be accounted for, like how do you explain when someone reacts instantly in an unplanned way deep into the exchange, or when someone changes their plan halfway through and does something different. It is possible to do this, but it gets messy and complicated.

In the decision making case, the ecological approach simplifies the situation significantly. The downfall of the IPA in this case comes from the way the experiment was run, the cue. If you try to distill a fencing situation into one specific cue (especially one that has nothing to do with fencing like a light turning on), you are losing tons of vital environmental information. According to the ecological approach, you are constantly perceiving and acting on information available in the environment in real time, so the entire basis of the reaction + decision + execution time evaporates. The decision making process is happening constantly up until the moment of execution, and continues as the exchange unfolds. This easily accounts for both the approach phases and lightning fast reactions when something unexpected happens. 

I want to make it clear that I really like this experiment and I wish more people would do things like this. I have done it in the past and would like to do it more in the future. It would be interesting to see what we can come up with if approaching these types of situations through an ecological lens, maybe by asking people to execute specific plans or actions of varying levels of detail in a live fencing situation and seeing how it affects their time and accuracy of execution. Ultimately I think this shows that both models can be used to describe reality, but as things get closer to a performance scenario, the information processing model requires more exceptions for edge cases, and the ecological model holds firm and does not need to be modified. It’s like using Newtonian mechanics when you start to get into relativistic situations, or trying to draw a circle with cartesian coordinates as opposed to polar. It can be done, it just loses coherence, and therefore usefulness.