The idea for this post came from an article I came across recently while looking for information on treatment of speech disorders caused by Parkinson’s Disease. The article is “New Conceptualizations of Practice: Common Principles in Three Paradigms Suggest New Concepts for Training,” by Richard Schmidt and Robert Bjork. It appears in the book Disorders of Motor Speech: Assessment, Treatment, and Clinical Characterization, edited by Donald Robin, Kathryn Yorkston, and Davbid Beukelman. (Brookes Publishing, 1996). Naturally, the article is concerned with speech and language skills, but much of it is relevant to learning sculling. In this post I will show how I think the results discussed in this article relate to coaching and to practicing technique. As the authors say,
“ {The findings we discuss} are obtained from diverse research paradigms that employ several different verbal and motor paradigms… Taken together, however, these finding suggest that certain conceptualizations about how and when to practice are at best incomplete, and at worst incorrect. These findings also have some theoretical implications with respect to the processes involved in practice, particularly as they relate to the acquisition of real-world skills.” (p. 4)
The general issue that the authors discuss is one which all you scullers will like, if you feel the sound of the coach’s launch makes your technique go to pot. That is the issue of whether the conditions that result in the best performance during practice are the same that result in the best performance later on. They are not, the authors say. In fact, acquisition performance is an imperfect indicator of learning.
There are certain criteria that a teacher of any sort uses to decide how effective the lesson was. Among these are long-term retention, generalizability, and resistance to altered contexts. And what research shows is that there are at least three kinds of situations where something about the learning stage actually decreases performance toward the end of practice, but results in better post-training performance (as measured by long-term retention, generalizability, and resistance to altered contexts).
This is especially interesting to me, because I very much want my scullers to perform well during their sessions with me. I don’t want to praise them for poor performance, but I also don’t want them to feel as though they don’t get what I’m trying to teach them or that they are failing to acquire the skill. In fact, I consider it the mark of a good coach to be able to adjust her teaching style to match the learning style of the sculler. With the information from Schmidt and Bjork, I now want to be careful that, in doing so, I am not inadvertently reducing how well my scullers eventually learn the skills they want to acquire. The information will also be interesting to those who self-coach. It will tell all of us how we might best practice.
First, how should you order the tasks in your practice session to maximize skill retention? Two possibilities are “blocked practice” and “random practice”. In blocked practice, you complete all trials for a given task before going on to the next one. For example, if you were learning to type, blocked practice might involve working through all of the exercises that involve the left index finger’s movements on the top rowof letters before going on to the exercises that involve the left index finger’s movements on the second row of letters. But in random practice, you do the same number of trials overall, just in an order such that the same task is never practiced on two trials in a row. So you might work on top-row, left index finge;, then middle row, right index finger; the bottom row, left index finger, then top row, right – and so on. Research that Schmidt and Bjork cite shows that random practice always resulted in better retention, even though it was also detrimental to performance during skill practice sessions.
What seems to be the active effect is that random practice provides “forgetting time” for the learner during acquisition (thus the poorer performance during practice). The need to continually work at retrieving the information on how to perform the action correctly seems to be the factor that actually enhances long-term retention. In the words of the authors, “Information-processing activities that cause forgetting of the to-be-remembered information, and thus require practice at retrieving it again on a subsequent trial, are beneficial for retention.” (p. 13).
Clearly, there are types of interruptions that cause forgetting and do not aid in long-term retention (like that annoying noise your rigger makes every time you feather). But if the practice schedule creates the need for kinds of information processing that are also needed for retention (like remembering the skill), then it will be a good schedule. I think this practice in remembering the skill is what I have found so useful for myself and the scullers I work with. If you look back at the drill sequences I list, though they all concern a certain general topic, the motor tasks each one demands differ. That is, taking your blades out of the water at arms-and-back only differs in some important respects from taking them out at full-slide. So introducing the perturbation of adding in more of the stroke, while directing the scullers to focus back on the release (or the catch, or what-have-you) after each change, may be very helpful. I wonder, however, whether mixing up the drills a little more may be more effective. That is, instead of practicing the release at arms-and-body, then half-slide, then full slide, might it be better over the long term to practice it at half-slide, then arms-and-body, then full slide? Or in some other order?
A second issue is the nature and frequency of feedback during acquisition. I had thought that the more immediate feedback was, the more frequent it was, and the more useful it was in changing the sculler’s behavior in the moment, the better. But Schmidt and Bjork cite research involving a complex task, where feedback was given after each trial, after each 5 trials, or after each 15. You may already have guessed that during practice, the longer the time the learners had to wait for feedback, the more errors they made. But when delayed retention was measured, it was the “after-every-15” learners who made the fewest errors. Similarly, learners who got feedback that was faded over the course of the practice session made more errors during those sessions than learners who got feedback on every trial. But they performed better on post-training retention test, even though they only got half as much feedback.
Again, what seems to matter here is that the expanded spacing of feedback allows time for forgetting – and it is really the need to retrieve the skill after forgetting that builds the long-term representation of the skill in the mind.
What this result tells me is that immediate corrective feedback, which I had felt was vital to getting a sculler to perform a certain movement accurately, may only be useful in the beginning of a session. Typically, I structure one-on-one sessions with explanation and drills for the first half, followed by “alone time” – time for the sculler to work out on his or her own how to perform the movement without me saying much. I will sometimes provide reminders for the sculler, either when I see their technique reverting, or after some kind of distracting event like having to steer around an obstacle. In fact, I often recommend to scullers to tie their concentrating on a certain skill to something like bridges. The lower Charles, where I most often coach, is blessed with many bridges and turns. When I was training, I used the bridges as reminders to bring my mind back to the technical issue I was working on that day. I had to start looking over my shoulder more frequently and steering more as I approached a bridge, but as soon as I was done steering for the time being and was in a bit of a straightaway, I returned to the issue at hand. Now, I recommend the same to the scullers I work with.
The final issue that Schmidt and Bjork discuss is “induced variability” – the idea that practicing (say) just throwing a beanbag at a target 3 feet away isn’t necessarily the best way to get good at throwing beanbags at targets 3 feet away. Instead, throwing them at targets 2 feet away and 4 feet away give better results. This was true in the experiments Schmidt and Bjork discuss even though the 2-feet/4-feet group never practiced on targets 3 feet away.
As the authors say, “These experiments suggest that variable practice alters the practice context, forcing a change in behavior from trial to trial and encouraging additional information-processing activities about the lawful relationships among the task variants.” (p. 18) In other words, if you must perform essentially the same task in a variety of contexts, you will be better at learning what the important aspects of the task are. But you probably won’t do as well at the task while you're still figuring that out.
I don’t know whether the drills I like to use in teaching are enough of a variation in context to induce this kind of learning. But perhaps it's relevant that I do feel that it is important to be able to do drills well. A sculler once asked me, “Who cares whether I can do square blades? You don’t race square blades.” My feeling was that it’s true that you don’t race square blades. But if rowing square blades helps you learn more about what releasing the blade from the water involves, then it helps you. It makes your mental model of the release that much more detailed, and it gives you that much more ability to adjust your stroke to the immediate demands of the conditions you are rowing in.
Here’s another metaphor: we’ve all been in boats that just aren’t rigged right for us. While I wouldn’t advocate rowing in poorly-rigged boats as a way of improving your sculling in general, it is also true that the best scullers can make even non-optimally rigged boats move well. Switching boats means that you have to take some time to figure out how that boat wants to be rowed, and until you do it can feel very annoying. But maybe having to figure it out isn’t a bad thing, in moderation.
I’ll close with a quote from Schmidt and Bjork:
“If certain acquisition conditions force the learner to engage in processes that are also critical for test performance, then those conditions will be judged as effective for learning (because they facilitate test performance), even though they may exhibit different superficial conditions. Also, those conditions that maximize learning may not be very effective for performance during the acquisition phase, as they provide various “difficulties” for the learners… In other words, these conditions can be considered as effective for learning because they prepare the learner for the processing that will be required at testing.” (p. 20)
So here’s what I’m going to be doing on Sunday morning. I’m going to try a variety of different drills centering on the same basic issue. I’m going to give relatively frequent feedback at first, and then I’m going to taper it off. And I’m going to make my scullers practice their new skills in lots of different situations, like at different ratings that I call out randomly. I’m going to keep perturbing my scullers.