Transfer of Learning a Novel Coordinated Rhythmic Movement

My PhD student Daniel Leach has just had his first paper accepted (preregistration, preprint, data & analysis files available here on the OSF) so it's way past time when I should blog this cool work. Danny and I have been developing methods, analyses and a theoretical framework to study learning and transfer of learning, and we have some interesting results (plus MANY more questions :) This post is about the first experiment just published; there's more to come!

We use coordinated rhythmic movement as our task; I've blogged this task in many posts and used this research programme as an example of theoretically driven, mechanistic modelling science. The basic form of the task is described here, the basic pattern of behavioural data is described here, and the model that implements our perception-action approach is described here. The main thing to know is that there are only a couple of rhythmic coordinations that are easy without training (0° and 180°), but other coordinations can be learned with feedback driven training. This gives us a simple model task that can serve as a window on perception-action mechanisms of skilled action and learning. 

Muscle Homology in Coordinated Rhythmic Movements

One of my main experimental tasks is coordinated rhythmic movement. This is a simple lab task in which I ask people to produce rhythmic movements (typically with a joystick) and coordinate those at some mean relative phase. Not all coordinations are equally easy; without training, people can typically only reliably produce 0° (in-phase) and 180° (anti-phase) movements. People can learn other coordinations, however; I typically train the maximally difficult 90° (although my PhD student has just completed a study training people at 60°; more on that awesome data shortly). I use coordination to study the perceptual control of action and learning.

My work is all designed to test and extend Bingham's mechanistic model of coordination dynamics. This model explicitly identifies all the actual components of the perception-action system producing the behaviour, and models them. In particular, it models the perceptual information we use to perceive relative phase; the relative direction of motion. This is an important contributor to coordination stability and this model is a real step up in terms of how we do business in psychology.

There is another factor that affects coordination stability, however, and the model currently only addresses this implicitly. That factor is muscle homology, and it's been repeatedly shown to be an important factor. For a long time, I have avoided worrying about it, because I have had no mechanistic way to talk about it. I think I have the beginnings of a way now, though, and this post is the first of several as I develop my first draft of that analysis.

Ecological Mechanisms and Models of Mechanisms (#MechanismWeek 4)

Mechanistic models are great, but so far cognitive science doesn't have any. We have functional models (of, for example, memory or categorisation) and dynamical models (of, for example, neural networks) but none of these can support the kind of explanations mechanistic models can. Is that it for psychology, or can we do better?

Here we propose that it's possible to do psychology in a way that allows for the development of explanatory, mechanistic models. The trick, as we have discussed, is to identify the correct level of analysis at which to ground those models. These models will definitely end up being multi-level (Craver, 2007), but the form of these final models will be dictated and constrained by the nature of the real parts and operations at the grounding level.

The correct level of analysis, we propose, is the ecological level. Specifically, ecological information is going to be the real component whose nature will place the necessary constraints on both our empirical investigations of psychological mechanisms as well as the mechanistic models we develop.

Let's see how this might work.

A Quick Review and Analysis of Perceptual Control Theory

Perceptual Control Theory (PCT; Powers, 1973) is a theory that proposes behaviour is about the control of perception. We act so as to keep some perceived part of the world at some state, and it's by doing this to sensible variables that we come to exhibit functional behaviour. People have noted the seeming overlap between PCT and the ecological approach, and it's advocates (mainly Richard Marken and Warren Mansell) all talk about it in revolutionary terms that should also feel a bit familiar.

I first encounted it in the context of an interview with Richard Marken on a now defunct blog (pdf of the archived page; link to page and scroll down to "Interview with Richard Marken"). Marken and I got into it a bit in the comments, as you will see! I was not impressed. However, Mansell & Marken (2015) have just published what they pitch as a clear exposition of what PCT actually is and how it works. I took the opportunity to read this and evaluate PCT as a 'grand theory of behaviour'.

My basic opinion has not changed. PCT is not wrong in most of it's basic claims, but it has no theory of information or how that information comes to be made or relate to the dynamics of the world. It's an unconstrained model fitting exercise, and it's central ideas simply don't serve as the kind of guide to discovery as a good theory should. Ecological psychology does a much more effective job of solving the relevant problems. 

Tasks from the First Person Perspective (A Purple Peril)

The great snare of the psychologist is the confusion of his own standpoint with that of the mental fact about which he is making his report. I shall hereafter call this the ‘psychologist’s fallacy’ par excellence.

William James, The Principles of Psychology

This is a video of a baby trying bacon for the first time. The baby gets really really excited, and everyone around him goes 'Ha! Babies love bacon as much as the rest of us, this is great!'. And everyone laughs and cheers.

Except here's the thing. I think what this baby really likes is making his loved ones laugh and cheer. The bacon is fine, but I don't think it's the magical experience his parents assume. They are making the psychologist's fallacy: mistaking what you think is going on for what the person is actually experiencing. (It's our fallacy because our subject matter makes us uniquely susceptible.)

When people come into our labs to take part in experiments, we present them with a situation that we have designed to elicit a specific behaviour from them, and that we manipulate in various ways in order to probe the makeup of that behaviour. We therefore think we know what the person is doing: they are doing the thing we asked them to do. However, this isn't necessarily true, and in order to figure out what our participants did and why, we need to consider how they experienced the experiment. In effect, doing our science right means taking the first person perspective of our participants when we formulate our explanations. 

I take this idea primarily from Louise Barrett's excellent book, Beyond the Brain: How Body & Environment Shape Animal and Human Minds (which I reviewed here). The first couple of chapters spend a lot of time talking about anthropomorphism, and why it's a problem. To be honest, when I read the book I didn't quite know why Louise started with this. But over time, I've realised what an extraordinarily powerful point it is and we now talk about all the time. 

Peril Proposal: The psychologist's fallacy is real, but the ecological approach to understanding task dynamics and the information they create offers a useful framework for avoiding it while we science.

Thinking about representations in relation to mechanisms

As Chemero (2011) observed, there are two ways to think about debates in science. We can either debate about the actual facts of the matter in the world or we can debate about the best way to explain the world. Some debates aren't amenable to the first type of debate, because there is no evidence that can definitively rule out one of the options. The only debate we can really have about representations is in terms of their role in explanations of behaviour. This is different than how I thought about representations a few years back when I wanted to argue that invoking representations was inherently a bad idea. Now I think we need to consider the utility or representations as part of explanations for psychological phenomena. In this post, I will argue that the concept of representations is not helpful in developing a particular class of explanation - ontic mechanistic explanations (described below). This is the first of two posts on this idea. In the next post I will attempt to explicitly compare cognitive and ecological approaches to behaviour in terms of how well they set us up to identify real parts and operations.

Dynamic Mechanistic Explanations in Radical Embodied Cognitive Science

I'm on my way back from an enactivist/embodiment conference in Warsaw. I gave a talk (slides) in which I argued that in order to make theories of distributed/embodied cognition work, you have to have something like a theory of ecological information as the glue to hold it all together. All the talks I saw that discussed any kind of plan for distributing cognition were missing this piece and desperately needed it, so I'm hoping the talk will make people realise this tool exists and can help. Drop me a line if you would like any help!

I argued specifically that information lets us propose mechanistic explanations for distributed cognitive systems. We recently came across the philosophical literature on what mechanisms are and how to make them, and it seemed immediately clear that we should be doing this (and that we already are; see below). 

It turns out, though, that some of the radical camp (specifically Chemero and Silberstein) don't think we can have distributed cognition mechanisms, but that this is ok because we still get explanations out of our dynamical models. 

This post will briefly review what dynamic mechanistic explanation is and why they are so useful (Bechtel & Abrahamsen, 2010). I'll briefly summarise the radical opposition to mechanisms, point out their answer doesn't work, then talk an example that shows we can have radical dynamic mechanistic explanatory models without giving anything up. The trick, as ever, will be to rely heavily on information as the component part that allows cognitive mechanisms to extend out over body and environment. 

The Perturbation Experiment as a Way to Study Perception

When you study perception, your goal is to control the flow of information going into the system so that you can measure the resulting behaviour and evaluate how that information is being used. There are two ways to do this, one (sometimes) used by me, one used by, well, everyone else. In this post I'm going to compare and contrast the methods and describe why the perturbation method is what we should all be doing.

The standard method is to present experimentally isolated cues and test whether people can detect those cues. The perturbation experiment presents a 'full cue' environment but selectively interferes with the link between a single variable and the property it might be information about. These two different methods lead to very different ways of thinking and talking about perceptual abilities. 

Rhythmic constraints on stress timing in English

What kind of embodied constraints affect the production of speech? Can we say anything we like when we like, or are there constraints in play that make some things easier than others? This is the question asked in Cummins & Port (1998) which we recently read in lab meeting (with our PhD student Agnes).

Cummins and Port asked participants to produce sentences over and over and examined when during the cycle a certain stress beat occurred. They set it up so that the beat was timed with a beep to occur throughout the cycle, but showed that people could actually only place the beat in 2 or 3 places in the beat reliably. The big picture result is that speech production is shaped, in part, by the underlying dynamics of production described in terms of the rhythms it is set up to produce.

The nice detail here comes from the theoretical set up and analysis that drives this study. Cummins and Port are directly inspired and guided by work in coordination dynamics. Agnes is interested in this work because she's looking at ways to investigate language and speech using the tools of dynamical systems and embodied cognition - remember, our big pitch is that language is special but not magical and we should be able to study it the way we study, say, rhythmic movement coordination. 

Using coordination to study learning across the lifespan

What happens to our ability to learn new movement skills as we age? There is surprisingly little research on this topic; a relatively recent review (Voelcker-Rehage, 2008) found only 25 articles about learning in old age, and no systematic programme of work. The answer to this question matters a lot; rehabilitation after events such as a stroke pretty much always entail (re)learning movement skills, and if our ability to learn gets worse with age, rehabilitation faces an uphill struggle. 

I have been studying coordinated rhythmic movement for some time now, and now we have a good handle on the task dynamic my colleagues at Indiana and I have begun using it to study the process of learning more generally. We decided to use it to look at learning in old age, to see what we could see.

This project grew out of a grant I had from Remedi when I was a post-doc in Aberdeen. I wanted to use coordination to look at learning post-stroke. One of the problems with studying this is finding useful novel tasks to learn - you need to give the stroke patients something they've never done before so you can be sure that any improvement is about learning, and not simply recovery of function. My thought at the time was that I could use any changes at 180° to assess recovery and changes at 90° to assess learning. We tested a huge number of patients and age matched controls, but the project didn't pan out because neither group (all aged around 65) couldn't learn to move at 90°. The question remained, what was going on? We now have the first of three papers on this question out in press. 

The Small Effect Size Effect - Why Do We Put Up With Small Effects?

Small effects sometimes matter - but psychology can do better

One of the things that bugs me about 'embodied' cognition research is that the effects, while statistically significant, tend to be small. What this means is that the groups were indeed different in the direction the authors claim, but only slightly, and that the authors had enough people showing the effect to make it come out on average. 

The problem with small effect sizes is that they mean all you've done is nudge the system. The embodied nervous system is exquisitely sensitive to variations in the flow of information it is interacting with, and it's not clear to me that merely nudging such a system is all that great an achievement. What's really impressive is when you properly break it - If you can alter the information in a task and simply make it so that the task becomes impossible for an organism, then you have found something that the system considers really important. The reverse is also true, of course - if you find the right way to present the information the system needs, then performance should become trivially easy. 

Psychology has become enthralled by statistical significance (to the point that we're possibly gaming the system in order to cross this magical marker). If your effect comes with a p value of less than .05, it is interesting, regardless of how small the effect is in terms of function. This is a problem, and we don't have to put up with it. If you ask a question about the right thing, you should get an unambiguous answer. If your answer is ambiguous, you may not be asking about the right thing. 

I want to remind readers of a couple of examples of nuisance small effects I've covered here before, then talk a little about some work which either broke or fixed the right thing, to highlight that we don't actually have to suffer from the tyranny of the small effect effect.

Coordination dynamics and relative speed

The Bingham model of coordinated rhythmic movement makes three predictions. First, it predicts that movement stability is a function of perceptual ability, and we confirmed this in two ways (by showing how people can move stably at non-0° with transformed visual feedback (Wilson et al, 2005) and by showing that perceptual learning of 90° led to improved movement stability without practice at the movement task; Wilson et al, 2010). This prediction is also supported by recent work by Kovacs and Shea, who are busy demonstrating that transformed, Lissajous feedback breaks the classic pattern of movement stability in coordination tasks. The second prediction is that relative phase is specified by the relative direction of motion; we confirmed this by selectively perturbing various components of motion and showing selective effects on performance (Wilson & Bingham, 2008). 

The third prediction was that the detection of relative direction was conditioned on the relative speed; the latter was simply a noise term. de Rugy, Oullier & Temprado (2008) tested this prediction by using an amplitude manipulation to alter the relative speeds. Their data did not support the model predictions, and they concluded that the approach taken by the Bingham model was flawed. We recently replicated their experiment (Snapp-Childs, Wilson & Bingham, in press as of Friday; download) and identified numerous critical flaws in their design and analysis which invalidated their criticism.

Lissajous feedback and coordination stability

Understanding the perceptual information you provide people in a task is a critical element of the perception-action analysis. Last time I talked about the new form of coordination feedback I developed to allow us to train coordinated rhythmic movements without perturbing the task dynamic. Prior to this, the most common form of augmented feedback was the Lissajous plot - these are the result of plotting the displacements of two harmonic oscillators against one another, and the unique shape associated with each relative phase can be used as a template on the screen. People can then try to move so as to make a dot trace that shape.

Lissajous plots (have a play with them in this Excel file) are transformed feedback, because they take a coordinated movement and represent it on the screen as the motion of a single dot. This type of feedback has been used extensively to train people to perform novel coordinations, but until recently no-one had thought to investigate the consequences of transforming the information about relative phase. Kovacs, Buchanan and Shea have recently begun doing exactly this, and, in line with the perception-action approach developed by Bingham and pushed at every opportunity by myself, these authors have found that Lissajous plots completely alter the nature of the task, with serious consequences for the studies that rely on it.

Visual feedback for training novel coordinations

The key feature of coordinated rhythmic movements is that not all coordinations are stable. Most other rhythms can be learned, however, which is why we can have jazz drumming. People have been training participants to perform novel coordinations (especially 90°, the least stable rhythm without training) for years now, and have been asking all the standard learning questions - how long does learning take? Does it transfer to other coordinations? 

The first real studies on learning were by Kelso and Zanone (Kelso & Zanone, 2002; Zanone & Kelso, 1992a, b, 1997). I briefly reviewed the results of these studies here, which have lead to to the dynamic pattern hypothesis. This account describes stable states as attractors in a state space defined by relative phase as the order parameter, and learning is the creation of a new attractor centred on the target novel phase. This account ran into problems quite quickly but is still alive and kicking in a modified form; stability is the governing principle now, and from this perspective the feedback displays used for training don't matter so long as they support stable action. 

However, from our perception-action standpoint, the feedback displays matter a lot, because these are what's providing the perceptual information about the coordinated movement. Early learning studies all used some kind of transformed feedback, which we could never use because it altered the overall perception-action dynamic. In order to look at action learning directly, we needed a new form of feedback. 

So I invented one.

Rates of learning and the dynamic pattern approach

One of the interesting features of coordinated rhythmic movement is that people start out with a particular pattern to their performance - there is pre-existing structure to our attempts to coordinate these movements. This structure affects our ability to learn new coordinations, and the pattern of the effects reveals a lot about the cause of this pre-existing structure. 

However, the literature is split into two incompatible accounts of learning, and trying to fix this is part of my ongoing interest in this task. The first account is the dynamic pattern approach, which was pioneered by JAS Kelso, and championed by modelling (Gregor Schöner) and behavioural studies (Pier Zanone). I'm more interested in the latter aspect, because it's the motivation for the former. I've already reviewed how this account fails, but it's still alive and well thanks to some creative history, and needs to be tackled again. The second account, which I prefer, is the perception-action account (Bingham) which developed from empirical work on visual and proprioceptive perception as well as action measures, and embodied in a model.

We haven't explicitly tackled the rate of learning issue, although we will and there is already support for our account in the literature (Wenderoth et al, 2002). But it comes up regularly in the dynamic pattern behavioural work, so it's time to work out what's going on in their data.

Perception, Action & Dynamical Systems

Over Easter I visited the Center of Functionally Integrative Neuroscience at Aarhus University in Denmark, courtesy of the Interacting Minds group. I gave a talk, got the tour, and met some of the faculty and students - some interesting opportunities for future collaborations, I hope - thanks for the hospitality!

I wanted to lay out the basics of the talk I gave. I took the opportunity to present some ideas that have been developing as I work on this blog, reading Chemero and working on coordination experiments. There is a core of people in Aarhus interested in things ecological, as well as dynamical systems, so it was a good audience to try these ideas out and they seemed to go over well. This is also the sketch of a paper Sabrina and I are going to work on over the summer.

The take home message of the talk was simple - dynamical systems is the right kind of mindset for cognitive science, but it is not a theory of behaviour. Dynamics merely provides the right kind of modelling tools - the form of the model must be based on hypotheses about the specific kind of dynamical systems we are or else they are merely an exercise in data-fitting. Ecological psychology is the right theory, and the Bingham model of coordinated rhythmic movement is currently the only example of a genuinely perception-action dynamical systems model. My thoughts here are largely from my response to Chapter 4 of Chemero (on 'the dynamical stance') and Chapter 5, his initial attempt to use dynamics to serve as a guide to discovery which I think fails and which Chemero then replaces with ecological psychology. The description of Bingham's model comes from here.

Chemero (2009), Chapter 5: Guides to Discovery

The dynamical stance laid out by Chemero in the previous chapter has a potential flaw (besides being a bit weak-ass) - it's not clear how it can serve as a guide to discovery. How do you do productive science taking this approach? Chemero is going to make two suggestions, only one of which I think works: first, he's going to suggest dynamical models such as the Haken-Kelso-Bunz (HKB) model can serve to stimulate empirical work even when they are entirely phenomenological. This approach is, I think, entirely incorrect, and this chapter is full of serious problems (only some of which are unique to Chemero). Second, he's going to suggest that Gibsonian ecological psychology can actually solve the problem much more robustly anyway, by serving as an underlying theory of behaviour. This will work better, and I would advocate Bingham's model of coordination as an exemplar of this, more promising route.

But first, the HKB model as guide to discovery (this chapter is largely the material from Chemero, 2000; I intend to turn this post into a paper to rebut that paper and point to the Bingham model as an alternative, so comments are especially welcome on this one). Time to get a little critical, I'm afraid.

Identifying the Visual Information for Relative Phase

Bingham's model predicts that the information for relative phase is the relative direction of movement. The first direct test of this hypothesis was the experiment that followed on from my learning study, in which we systematically perturbed the various candidate information variables to see which affected performance in the perceptual judgement task.

I like this study a lot, if I do say so myself. It's a serious attempt to make a strong test of the model's predictions, and we invested a lot of time in the methodology. This is also that rare paper that benefited from a vigorous review process; the end result is, I think, a clear, careful, and detailed presentation of a critical result for the perception-action approach Geoff and I are developing.

Readers interested in the issue of how you can scientifically study information from an ecological perspective should certainly read the paper (Ken, that's you :); it's my go-to reference for how I believe this has to be done. The main lesson - it's hard to do this properly, but the rewards, in terms of unambiguous data, are clear.

Perceptual Learning Stabilises Action: A Test of the Bingham Model

Bingham's perception-action model was initially inspired by perceptual judgement studies (using vision and proprioception). The HKB phenomena are movement phenomena, however; simply noting that the same qualitative pattern is seen in different judgement and action studies is a good first step but only suggestive, at best. We therefore next took simultaneous judgement & action measures from a movement task where we manipulated the feedback display (Wilson et al, 2005a). For instance, when the display showed 0°, movement was stable, even when the movement was at, for example, 90°. Perception of relative phase was driving the stability of the movements.

On the basis of all this data, the model predicts that the reason 0° and 180° are easy is that the information specifying that you are moving this way is easily perceived. There is provisional evidence to support relative direction of motion as the specifying information (Bogaerts et al, 2003; Wilson et al, 2005b; Wimmers et al, 1992) with relative speed acting as a noise term. This variable certainly predicts the observed pattern, as the relative direction of motion is only stable at 0° and 180°. It is maximally variable at 90°, which would explain why movements here are also maximally unstable. The model is therefore explaining the problem with moving at 90° as a problem detecting the information required to maintain the coordination; as we saw in the case of friction, no information means unstable behaviour.

The model therefore makes a critical prediction. If we could improve people's ability to perceive 90°, they should gain the ability to move at 90° without any practice at the movement itself. All previous learning studies had entailed training people to move by having them move, with the help of various forms of transformed feedback methods (visual metronomes or Lissajous plots; more on this when I discuss feedback). The prediction, that movement stability should improve with improved perceptual ability, is a strong test of both the model and the modelling strategy in general, and the experiment to test it was the first half of my dissertation.

How to Build a Valid Measure of Behaviour

One of the main problems facing psychology as a science is the issue of validity - what is the relationship between what you measured and what you are actually interested in? One of the things I like about studying movement is how straight-forward this issue is - we're interested in the control of action, so I just measure the action! The most common directly measured kinematic variable is displacement, or position over time; you can then derive (via differentiation) the various rates of change of the previous variable (velocity, acceleration, jerk, and, I kid you not, snap, crackle, and pop). In human movement we never tend to go past jerk, and you can do pretty well with just position and it's rate of change, velocity.

This post will discuss how we start from these basic kinematics and derive a measure of coordination that is  entirely valid, covers the entire space of possible states and provides a unique number for every possible state within that space. Psychology doesn't have a lot of these kinds of variables, but you need to be able to characterise your state space to do the kind of modelling I've been describing and advocating.